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 *, struct socket *,
 130         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                                 /* FALLTHROUGH */
 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 the 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         /*
 496          * If there is a cached route,
 497          * check that it is to the same destination
 498          * and is still up. If not, free it and try again.
 499          */
 500         if (ro->ro_rt != NULL &&
 501             (!(ro->ro_rt->rt_flags & RTF_UP) || dst->sin6_family != AF_INET6 ||
 502              !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
 503                 RTFREE(ro->ro_rt);
 504                 ro->ro_rt = NULL;
 505         }
 506         if (ro->ro_rt == NULL) {
 507                 bzero(dst, sizeof(*dst));
 508                 dst->sin6_family = AF_INET6;
 509                 dst->sin6_len = sizeof(struct sockaddr_in6);
 510                 dst->sin6_addr = ip6->ip6_dst;
 511         }
 512 #if defined(IPSEC) || defined(FAST_IPSEC)
 513         if (needipsec && needipsectun) {
 514                 struct ipsec_output_state state;
 515
 516                 /*
 517                  * All the extension headers will become inaccessible
 518                  * (since they can be encrypted).
 519                  * Don't panic, we need no more updates to extension headers
 520                  * on inner IPv6 packet (since they are now encapsulated).
 521                  *
 522                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 523                  */
 524                 bzero(&exthdrs, sizeof(exthdrs));
 525                 exthdrs.ip6e_ip6 = m;
 526
 527                 bzero(&state, sizeof(state));
 528                 state.m = m;
 529                 state.ro = (struct route *)ro;
 530                 state.dst = (struct sockaddr *)dst;
 531
 532                 error = ipsec6_output_tunnel(&state, sp, flags);
 533
 534                 m = state.m;
 535                 ro = (struct route_in6 *)state.ro;
 536                 dst = (struct sockaddr_in6 *)state.dst;
 537                 if (error) {
 538                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 539                         m0 = m = NULL;
 540                         m = NULL;
 541                         switch (error) {
 542                         case EHOSTUNREACH:
 543                         case ENETUNREACH:
 544                         case EMSGSIZE:
 545                         case ENOBUFS:
 546                         case ENOMEM:
 547                                 break;
 548                         default:
 549                                 kprintf("ip6_output (ipsec): error code %d\n", error);
 550                                 /* FALLTHROUGH */
 551                         case ENOENT:
 552                                 /* don't show these error codes to the user */
 553                                 error = 0;
 554                                 break;
 555                         }
 556                         goto bad;
 557                 }
 558
 559                 exthdrs.ip6e_ip6 = m;
 560         }
 561 #endif
 562
 563         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 564                 /* Unicast */
 565
 566 #define ifatoia6(ifa)   ((struct in6_ifaddr *)(ifa))
 567 #define sin6tosa(sin6)  ((struct sockaddr *)(sin6))
 568                 /* xxx
 569                  * interface selection comes here
 570                  * if an interface is specified from an upper layer,
 571                  * ifp must point it.
 572                  */
 573                 if (ro->ro_rt == NULL) {
 574                         /*
 575                          * non-bsdi always clone routes, if parent is
 576                          * PRF_CLONING.
 577                          */
 578                         rtalloc((struct route *)ro);
 579                 }
 580                 if (ro->ro_rt == NULL) {
 581                         ip6stat.ip6s_noroute++;
 582                         error = EHOSTUNREACH;
 583                         /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
 584                         goto bad;
 585                 }
 586                 ia = ifatoia6(ro->ro_rt->rt_ifa);
 587                 ifp = ro->ro_rt->rt_ifp;
 588                 ro->ro_rt->rt_use++;
 589                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
 590                         dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
 591                 m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
 592
 593                 in6_ifstat_inc(ifp, ifs6_out_request);
 594
 595                 /*
 596                  * Check if the outgoing interface conflicts with
 597                  * the interface specified by ifi6_ifindex (if specified).
 598                  * Note that loopback interface is always okay.
 599                  * (this may happen when we are sending a packet to one of
 600                  *  our own addresses.)
 601                  */
 602                 if (opt && opt->ip6po_pktinfo
 603                  && opt->ip6po_pktinfo->ipi6_ifindex) {
 604                         if (!(ifp->if_flags & IFF_LOOPBACK)
 605                          && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
 606                                 ip6stat.ip6s_noroute++;
 607                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 608                                 error = EHOSTUNREACH;
 609                                 goto bad;
 610                         }
 611                 }
 612
 613                 if (opt && opt->ip6po_hlim != -1)
 614                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 615         } else {
 616                 /* Multicast */
 617                 struct  in6_multi *in6m;
 618
 619                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 620
 621                 /*
 622                  * See if the caller provided any multicast options
 623                  */
 624                 ifp = NULL;
 625                 if (im6o != NULL) {
 626                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 627                         if (im6o->im6o_multicast_ifp != NULL)
 628                                 ifp = im6o->im6o_multicast_ifp;
 629                 } else
 630                         ip6->ip6_hlim = ip6_defmcasthlim;
 631
 632                 /*
 633                  * See if the caller provided the outgoing interface
 634                  * as an ancillary data.
 635                  * Boundary check for ifindex is assumed to be already done.
 636                  */
 637                 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
 638                         ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
 639
 640                 /*
 641                  * If the destination is a node-local scope multicast,
 642                  * the packet should be loop-backed only.
 643                  */
 644                 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
 645                         /*
 646                          * If the outgoing interface is already specified,
 647                          * it should be a loopback interface.
 648                          */
 649                         if (ifp && !(ifp->if_flags & IFF_LOOPBACK)) {
 650                                 ip6stat.ip6s_badscope++;
 651                                 error = ENETUNREACH; /* XXX: better error? */
 652                                 /* XXX correct ifp? */
 653                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 654                                 goto bad;
 655                         } else {
 656                                 ifp = &loif[0];
 657                         }
 658                 }
 659
 660                 if (opt && opt->ip6po_hlim != -1)
 661                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 662
 663                 /*
 664                  * If caller did not provide an interface lookup a
 665                  * default in the routing table.  This is either a
 666                  * default for the speicfied group (i.e. a host
 667                  * route), or a multicast default (a route for the
 668                  * ``net'' ff00::/8).
 669                  */
 670                 if (ifp == NULL) {
 671                         if (ro->ro_rt == NULL) {
 672                                 ro->ro_rt =
 673                                   rtpurelookup((struct sockaddr *)&ro->ro_dst);
 674                         }
 675                         if (ro->ro_rt == NULL) {
 676                                 ip6stat.ip6s_noroute++;
 677                                 error = EHOSTUNREACH;
 678                                 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
 679                                 goto bad;
 680                         }
 681                         ia = ifatoia6(ro->ro_rt->rt_ifa);
 682                         ifp = ro->ro_rt->rt_ifp;
 683                         ro->ro_rt->rt_use++;
 684                 }
 685
 686                 if (!(flags & IPV6_FORWARDING))
 687                         in6_ifstat_inc(ifp, ifs6_out_request);
 688                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 689
 690                 /*
 691                  * Confirm that the outgoing interface supports multicast.
 692                  */
 693                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 694                         ip6stat.ip6s_noroute++;
 695                         in6_ifstat_inc(ifp, ifs6_out_discard);
 696                         error = ENETUNREACH;
 697                         goto bad;
 698                 }
 699                 in6m = IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp);
 700                 if (in6m != NULL &&
 701                    (im6o == NULL || im6o->im6o_multicast_loop)) {
 702                         /*
 703                          * If we belong to the destination multicast group
 704                          * on the outgoing interface, and the caller did not
 705                          * forbid loopback, loop back a copy.
 706                          */
 707                         ip6_mloopback(ifp, m, dst);
 708                 } else {
 709                         /*
 710                          * If we are acting as a multicast router, perform
 711                          * multicast forwarding as if the packet had just
 712                          * arrived on the interface to which we are about
 713                          * to send.  The multicast forwarding function
 714                          * recursively calls this function, using the
 715                          * IPV6_FORWARDING flag to prevent infinite recursion.
 716                          *
 717                          * Multicasts that are looped back by ip6_mloopback(),
 718                          * above, will be forwarded by the ip6_input() routine,
 719                          * if necessary.
 720                          */
 721                         if (ip6_mrouter && !(flags & IPV6_FORWARDING)) {
 722                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 723                                         m_freem(m);
 724                                         goto done;
 725                                 }
 726                         }
 727                 }
 728                 /*
 729                  * Multicasts with a hoplimit of zero may be looped back,
 730                  * above, but must not be transmitted on a network.
 731                  * Also, multicasts addressed to the loopback interface
 732                  * are not sent -- the above call to ip6_mloopback() will
 733                  * loop back a copy if this host actually belongs to the
 734                  * destination group on the loopback interface.
 735                  */
 736                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
 737                         m_freem(m);
 738                         goto done;
 739                 }
 740         }
 741
 742         /*
 743          * Fill the outgoing inteface to tell the upper layer
 744          * to increment per-interface statistics.
 745          */
 746         if (ifpp)
 747                 *ifpp = ifp;
 748
 749         /* Determine path MTU. */
 750         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 751             &alwaysfrag)) != 0)
 752                 goto bad;
 753
 754         /*
 755          * The caller of this function may specify to use the minimum MTU
 756          * in some cases.
 757          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 758          * setting.  The logic is a bit complicated; by default, unicast
 759          * packets will follow path MTU while multicast packets will be sent at
 760          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 761          * including unicast ones will be sent at the minimum MTU.  Multicast
 762          * packets will always be sent at the minimum MTU unless
 763          * IP6PO_MINMTU_DISABLE is explicitly specified.
 764          * See RFC 3542 for more details.
 765          */
 766         if (mtu > IPV6_MMTU) {
 767                 if ((flags & IPV6_MINMTU))
 768                         mtu = IPV6_MMTU;
 769                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 770                         mtu = IPV6_MMTU;
 771                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 772                          (opt == NULL ||
 773                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 774                         mtu = IPV6_MMTU;
 775                 }
 776         }
 777
 778         /* Fake scoped addresses */
 779         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
 780                 /*
 781                  * If source or destination address is a scoped address, and
 782                  * the packet is going to be sent to a loopback interface,
 783                  * we should keep the original interface.
 784                  */
 785
 786                 /*
 787                  * XXX: this is a very experimental and temporary solution.
 788                  * We eventually have sockaddr_in6 and use the sin6_scope_id
 789                  * field of the structure here.
 790                  * We rely on the consistency between two scope zone ids
 791                  * of source and destination, which should already be assured.
 792                  * Larger scopes than link will be supported in the future.
 793                  */
 794                 origifp = NULL;
 795                 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
 796                         origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
 797                 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
 798                         origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
 799                 /*
 800                  * XXX: origifp can be NULL even in those two cases above.
 801                  * For example, if we remove the (only) link-local address
 802                  * from the loopback interface, and try to send a link-local
 803                  * address without link-id information.  Then the source
 804                  * address is ::1, and the destination address is the
 805                  * link-local address with its s6_addr16[1] being zero.
 806                  * What is worse, if the packet goes to the loopback interface
 807                  * by a default rejected route, the null pointer would be
 808                  * passed to looutput, and the kernel would hang.
 809                  * The following last resort would prevent such disaster.
 810                  */
 811                 if (origifp == NULL)
 812                         origifp = ifp;
 813         }
 814         else
 815                 origifp = ifp;
 816         /*
 817          * clear embedded scope identifiers if necessary.
 818          * in6_clearscope will touch the addresses only when necessary.
 819          */
 820         in6_clearscope(&ip6->ip6_src);
 821         in6_clearscope(&ip6->ip6_dst);
 822
 823         /*
 824          * Check with the firewall...
 825          */
 826         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 827                 u_short port = 0;
 828
 829                 m->m_pkthdr.rcvif = NULL;       /* XXX */
 830                 /* If ipfw says divert, we have to just drop packet */
 831                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
 832                         m_freem(m);
 833                         goto done;
 834                 }
 835                 if (!m) {
 836                         error = EACCES;
 837                         goto done;
 838                 }
 839         }
 840
 841         /*
 842          * If the outgoing packet contains a hop-by-hop options header,
 843          * it must be examined and processed even by the source node.
 844          * (RFC 2460, section 4.)
 845          */
 846         if (exthdrs.ip6e_hbh) {
 847                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 848                 u_int32_t dummy1; /* XXX unused */
 849                 u_int32_t dummy2; /* XXX unused */
 850
 851 #ifdef DIAGNOSTIC
 852                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 853                         panic("ip6e_hbh is not continuous");
 854 #endif
 855                 /*
 856                  *  XXX: if we have to send an ICMPv6 error to the sender,
 857                  *       we need the M_LOOP flag since icmp6_error() expects
 858                  *       the IPv6 and the hop-by-hop options header are
 859                  *       continuous unless the flag is set.
 860                  */
 861                 m->m_flags |= M_LOOP;
 862                 m->m_pkthdr.rcvif = ifp;
 863                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 864                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 865                     &dummy1, &dummy2) < 0) {
 866                         /* m was already freed at this point */
 867                         error = EINVAL;/* better error? */
 868                         goto done;
 869                 }
 870                 m->m_flags &= ~M_LOOP; /* XXX */
 871                 m->m_pkthdr.rcvif = NULL;
 872         }
 873
 874         /*
 875          * Run through list of hooks for output packets.
 876          */
 877         if (pfil_has_hooks(&inet6_pfil_hook)) {
 878                 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
 879                 if (error != 0 || m == NULL)
 880                         goto done;
 881                 ip6 = mtod(m, struct ip6_hdr *);
 882         }
 883
 884         /*
 885          * Send the packet to the outgoing interface.
 886          * If necessary, do IPv6 fragmentation before sending.
 887          *
 888          * the logic here is rather complex:
 889          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 890          * 1-a: send as is if tlen <= path mtu
 891          * 1-b: fragment if tlen > path mtu
 892          *
 893          * 2: if user asks us not to fragment (dontfrag == 1)
 894          * 2-a: send as is if tlen <= interface mtu
 895          * 2-b: error if tlen > interface mtu
 896          *
 897          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 898          *      always fragment
 899          *
 900          * 4: if dontfrag == 1 && alwaysfrag == 1
 901          *      error, as we cannot handle this conflicting request
 902          */
 903         tlen = m->m_pkthdr.len;
 904
 905         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
 906                 dontfrag = 1;
 907         else
 908                 dontfrag = 0;
 909         if (dontfrag && alwaysfrag) {   /* case 4 */
 910                 /* conflicting request - can't transmit */
 911                 error = EMSGSIZE;
 912                 goto bad;
 913         }
 914         if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
 915                 /*
 916                  * Even if the DONTFRAG option is specified, we cannot send the
 917                  * packet when the data length is larger than the MTU of the
 918                  * outgoing interface.
 919                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 920                  * well as returning an error code (the latter is not described
 921                  * in the API spec.)
 922                  */
 923                 u_int32_t mtu32;
 924                 struct ip6ctlparam ip6cp;
 925
 926                 mtu32 = (u_int32_t)mtu;
 927                 bzero(&ip6cp, sizeof(ip6cp));
 928                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
 929                 kpfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
 930                     (void *)&ip6cp);
 931
 932                 error = EMSGSIZE;
 933                 goto bad;
 934         }
 935
 936         /*
 937          * transmit packet without fragmentation
 938          */
 939         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 940                 struct in6_ifaddr *ia6;
 941
 942                 ip6 = mtod(m, struct ip6_hdr *);
 943                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 944                 if (ia6) {
 945                         /* Record statistics for this interface address. */
 946                         IFA_STAT_INC(&ia6->ia_ifa, opackets, 1);
 947                         IFA_STAT_INC(&ia6->ia_ifa, obytes, m->m_pkthdr.len);
 948                 }
 949 #ifdef IPSEC
 950                 /* clean ipsec history once it goes out of the node */
 951                 ipsec_delaux(m);
 952 #endif
 953                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 954                 goto done;
 955         }
 956
 957         /*
 958          * try to fragment the packet.  case 1-b and 3
 959          */
 960         if (mtu < IPV6_MMTU) {
 961                 /*
 962                  * note that path MTU is never less than IPV6_MMTU
 963                  * (see icmp6_input).
 964                  */
 965                 error = EMSGSIZE;
 966                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 967                 goto bad;
 968         } else if (ip6->ip6_plen == 0) {
 969                 /* 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, M_NOWAIT, 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) - sizeof(struct ip6_hdr)));
1048                         if ((m_frgpart = m_copy(m0, off, len)) == NULL) {
1049                                 error = ENOBUFS;
1050                                 ip6stat.ip6s_odropped++;
1051                                 goto sendorfree;
1052                         }
1053                         m_cat(m, m_frgpart);
1054                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1055                         m->m_pkthdr.rcvif = NULL;
1056                         ip6f->ip6f_reserved = 0;
1057                         ip6f->ip6f_ident = id;
1058                         ip6f->ip6f_nxt = nextproto;
1059                         ip6stat.ip6s_ofragments++;
1060                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1061                 }
1062
1063                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1064         }
1065
1066         /*
1067          * Remove leading garbages.
1068          */
1069 sendorfree:
1070         m = m0->m_nextpkt;
1071         m0->m_nextpkt = NULL;
1072         m_freem(m0);
1073         for (m0 = m; m; m = m0) {
1074                 m0 = m->m_nextpkt;
1075                 m->m_nextpkt = NULL;
1076                 if (error == 0) {
1077                         /* Record statistics for this interface address. */
1078                         if (ia) {
1079                                 IFA_STAT_INC(&ia->ia_ifa, opackets, 1);
1080                                 IFA_STAT_INC(&ia->ia_ifa, obytes,
1081                                     m->m_pkthdr.len);
1082                         }
1083 #ifdef IPSEC
1084                         /* clean ipsec history once it goes out of the node */
1085                         ipsec_delaux(m);
1086 #endif
1087                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1088                 } else
1089                         m_freem(m);
1090         }
1091
1092         if (error == 0)
1093                 ip6stat.ip6s_fragmented++;
1094
1095 done:
1096         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1097                 RTFREE(ro->ro_rt);
1098         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1099                 RTFREE(ro_pmtu->ro_rt);
1100         }
1101
1102 #ifdef IPSEC
1103         if (sp != NULL)
1104                 key_freesp(sp);
1105 #endif
1106 #ifdef FAST_IPSEC
1107         if (sp != NULL)
1108                 KEY_FREESP(&sp);
1109 #endif
1110
1111         return (error);
1112
1113 freehdrs:
1114         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1115         m_freem(exthdrs.ip6e_dest1);
1116         m_freem(exthdrs.ip6e_rthdr);
1117         m_freem(exthdrs.ip6e_dest2);
1118         /* FALLTHROUGH */
1119 bad:
1120         m_freem(m);
1121         goto done;
1122 }
1123
1124 static int
1125 copyexthdr(void *h, struct mbuf **mp)
1126 {
1127         struct ip6_ext *hdr = h;
1128         int hlen;
1129         struct mbuf *m;
1130
1131         if (hdr == NULL)
1132                 return 0;
1133
1134         hlen = (hdr->ip6e_len + 1) * 8;
1135         if (hlen > MCLBYTES)
1136                 return ENOBUFS; /* XXX */
1137
1138         m = m_getb(hlen, M_NOWAIT, MT_DATA, 0);
1139         if (!m)
1140                 return ENOBUFS;
1141         m->m_len = hlen;
1142
1143         bcopy(hdr, mtod(m, caddr_t), hlen);
1144
1145         *mp = m;
1146         return 0;
1147 }
1148
1149 /*
1150  * Insert jumbo payload option.
1151  */
1152 static int
1153 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1154 {
1155         struct mbuf *mopt;
1156         u_char *optbuf;
1157         u_int32_t v;
1158
1159 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1160
1161         /*
1162          * If there is no hop-by-hop options header, allocate new one.
1163          * If there is one but it doesn't have enough space to store the
1164          * jumbo payload option, allocate a cluster to store the whole options.
1165          * Otherwise, use it to store the options.
1166          */
1167         if (exthdrs->ip6e_hbh == NULL) {
1168                 MGET(mopt, M_NOWAIT, MT_DATA);
1169                 if (mopt == NULL)
1170                         return (ENOBUFS);
1171                 mopt->m_len = JUMBOOPTLEN;
1172                 optbuf = mtod(mopt, u_char *);
1173                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1174                 exthdrs->ip6e_hbh = mopt;
1175         } else {
1176                 struct ip6_hbh *hbh;
1177
1178                 mopt = exthdrs->ip6e_hbh;
1179                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1180                         /*
1181                          * XXX assumption:
1182                          * - exthdrs->ip6e_hbh is not referenced from places
1183                          *   other than exthdrs.
1184                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1185                          */
1186                         int oldoptlen = mopt->m_len;
1187                         struct mbuf *n;
1188
1189                         /*
1190                          * XXX: give up if the whole (new) hbh header does
1191                          * not fit even in an mbuf cluster.
1192                          */
1193                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1194                                 return (ENOBUFS);
1195
1196                         /*
1197                          * As a consequence, we must always prepare a cluster
1198                          * at this point.
1199                          */
1200                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1201                         if (!n)
1202                                 return (ENOBUFS);
1203                         n->m_len = oldoptlen + JUMBOOPTLEN;
1204                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen);
1205                         optbuf = mtod(n, caddr_t) + oldoptlen;
1206                         m_freem(mopt);
1207                         mopt = exthdrs->ip6e_hbh = n;
1208                 } else {
1209                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1210                         mopt->m_len += JUMBOOPTLEN;
1211                 }
1212                 optbuf[0] = IP6OPT_PADN;
1213                 optbuf[1] = 1;
1214
1215                 /*
1216                  * Adjust the header length according to the pad and
1217                  * the jumbo payload option.
1218                  */
1219                 hbh = mtod(mopt, struct ip6_hbh *);
1220                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1221         }
1222
1223         /* fill in the option. */
1224         optbuf[2] = IP6OPT_JUMBO;
1225         optbuf[3] = 4;
1226         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1227         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1228
1229         /* finally, adjust the packet header length */
1230         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1231
1232         return (0);
1233 #undef JUMBOOPTLEN
1234 }
1235
1236 /*
1237  * Insert fragment header and copy unfragmentable header portions.
1238  */
1239 static int
1240 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1241                   struct ip6_frag **frghdrp)
1242 {
1243         struct mbuf *n, *mlast;
1244
1245         if (hlen > sizeof(struct ip6_hdr)) {
1246                 n = m_copym(m0, sizeof(struct ip6_hdr),
1247                             hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1248                 if (n == NULL)
1249                         return (ENOBUFS);
1250                 m->m_next = n;
1251         } else
1252                 n = m;
1253
1254         /* Search for the last mbuf of unfragmentable part. */
1255         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1256                 ;
1257
1258         if (!(mlast->m_flags & M_EXT) &&
1259             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1260                 /* use the trailing space of the last mbuf for the fragment hdr */
1261                 *frghdrp = (struct ip6_frag *)
1262                     (mtod(mlast, caddr_t) + mlast->m_len);
1263                 mlast->m_len += sizeof(struct ip6_frag);
1264                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1265         } else {
1266                 /* allocate a new mbuf for the fragment header */
1267                 struct mbuf *mfrg;
1268
1269                 MGET(mfrg, M_NOWAIT, MT_DATA);
1270                 if (mfrg == NULL)
1271                         return (ENOBUFS);
1272                 mfrg->m_len = sizeof(struct ip6_frag);
1273                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1274                 mlast->m_next = mfrg;
1275         }
1276
1277         return (0);
1278 }
1279
1280 static int
1281 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1282     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1283     int *alwaysfragp)
1284 {
1285         u_int32_t mtu = 0;
1286         int alwaysfrag = 0;
1287         int error = 0;
1288
1289         if (ro_pmtu != ro) {
1290                 /* The first hop and the final destination may differ. */
1291                 struct sockaddr_in6 *sa6_dst =
1292                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1293                 if (ro_pmtu->ro_rt &&
1294                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1295                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1296                         RTFREE(ro_pmtu->ro_rt);
1297                         ro_pmtu->ro_rt = NULL;
1298                 }
1299                 if (ro_pmtu->ro_rt == NULL) {
1300                         bzero(sa6_dst, sizeof(*sa6_dst));
1301                         sa6_dst->sin6_family = AF_INET6;
1302                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1303                         sa6_dst->sin6_addr = *dst;
1304
1305                         rtalloc((struct route *)ro_pmtu);
1306                 }
1307         }
1308         if (ro_pmtu->ro_rt) {
1309                 u_int32_t ifmtu;
1310                 struct in_conninfo inc;
1311
1312                 bzero(&inc, sizeof(inc));
1313                 inc.inc_flags = 1; /* IPv6 */
1314                 inc.inc6_faddr = *dst;
1315
1316                 if (ifp == NULL)
1317                         ifp = ro_pmtu->ro_rt->rt_ifp;
1318                 ifmtu = IN6_LINKMTU(ifp);
1319                 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1320                 if (mtu == 0)
1321                         mtu = ifmtu;
1322                 else if (mtu < IPV6_MMTU) {
1323                         /*
1324                          * RFC2460 section 5, last paragraph:
1325                          * if we record ICMPv6 too big message with
1326                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1327                          * or smaller, with framgent header attached.
1328                          * (fragment header is needed regardless from the
1329                          * packet size, for translators to identify packets)
1330                          */
1331                         alwaysfrag = 1;
1332                         mtu = IPV6_MMTU;
1333                 } else if (mtu > ifmtu) {
1334                         /*
1335                          * The MTU on the route is larger than the MTU on
1336                          * the interface!  This shouldn't happen, unless the
1337                          * MTU of the interface has been changed after the
1338                          * interface was brought up.  Change the MTU in the
1339                          * route to match the interface MTU (as long as the
1340                          * field isn't locked).
1341                          */
1342                         mtu = ifmtu;
1343                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1344                 }
1345         } else if (ifp) {
1346                 mtu = IN6_LINKMTU(ifp);
1347         } else
1348                 error = EHOSTUNREACH; /* XXX */
1349
1350         *mtup = mtu;
1351         if (alwaysfragp)
1352                 *alwaysfragp = alwaysfrag;
1353         return (error);
1354 }
1355
1356 /*
1357  * IP6 socket option processing.
1358  */
1359 void
1360 ip6_ctloutput_dispatch(netmsg_t msg)
1361 {
1362         int error;
1363
1364         error = ip6_ctloutput(msg->ctloutput.base.nm_so,
1365                               msg->ctloutput.nm_sopt);
1366         lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
1367 }
1368
1369 int
1370 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1371 {
1372         int optdatalen,uproto;
1373         int privileged;
1374         struct inpcb *in6p = so->so_pcb;
1375         void *optdata;
1376         int error, optval;
1377         int level, op, optname;
1378         int optlen;
1379         struct thread *td;
1380
1381         if (sopt) {
1382                 level = sopt->sopt_level;
1383                 op = sopt->sopt_dir;
1384                 optname = sopt->sopt_name;
1385                 optlen = sopt->sopt_valsize;
1386                 td = sopt->sopt_td;
1387         } else {
1388                 panic("ip6_ctloutput: arg soopt is NULL");
1389                 /* NOT REACHED */
1390                 td = NULL;
1391         }
1392         error = optval = 0;
1393
1394         uproto = (int)so->so_proto->pr_protocol;
1395         privileged = (td == NULL || priv_check(td, PRIV_ROOT)) ? 0 : 1;
1396
1397         if (level == IPPROTO_IPV6) {
1398                 switch (op) {
1399
1400                 case SOPT_SET:
1401                         switch (optname) {
1402                         case IPV6_2292PKTOPTIONS:
1403 #ifdef IPV6_PKTOPTIONS
1404                         case IPV6_PKTOPTIONS:
1405 #endif
1406                         {
1407                                 struct mbuf *m;
1408
1409                                 error = soopt_getm(sopt, &m); /* XXX */
1410                                 if (error != 0)
1411                                         break;
1412                                 soopt_to_mbuf(sopt, m); /* XXX */
1413                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1414                                                     m, so, sopt);
1415                                 m_freem(m); /* XXX */
1416                                 break;
1417                         }
1418
1419                         /*
1420                          * Use of some Hop-by-Hop options or some
1421                          * Destination options, might require special
1422                          * privilege.  That is, normal applications
1423                          * (without special privilege) might be forbidden
1424                          * from setting certain options in outgoing packets,
1425                          * and might never see certain options in received
1426                          * packets. [RFC 2292 Section 6]
1427                          * KAME specific note:
1428                          *  KAME prevents non-privileged users from sending or
1429                          *  receiving ANY hbh/dst options in order to avoid
1430                          *  overhead of parsing options in the kernel.
1431                          */
1432                         case IPV6_RECVHOPOPTS:
1433                         case IPV6_RECVDSTOPTS:
1434                         case IPV6_RECVRTHDRDSTOPTS:
1435                                 if (!privileged)
1436                                         return (EPERM);
1437                         case IPV6_RECVPKTINFO:
1438                         case IPV6_RECVHOPLIMIT:
1439                         case IPV6_RECVRTHDR:
1440                         case IPV6_RECVPATHMTU:
1441                         case IPV6_RECVTCLASS:
1442                         case IPV6_AUTOFLOWLABEL:
1443                         case IPV6_HOPLIMIT:
1444                         /* FALLTHROUGH */
1445                         case IPV6_UNICAST_HOPS:
1446                         case IPV6_FAITH:
1447
1448                         case IPV6_V6ONLY:
1449                                 if (optlen != sizeof(int)) {
1450                                         error = EINVAL;
1451                                         break;
1452                                 }
1453                                 error = soopt_to_kbuf(sopt, &optval,
1454                                         sizeof optval, sizeof optval);
1455                                 if (error)
1456                                         break;
1457                                 switch (optname) {
1458
1459                                 case IPV6_UNICAST_HOPS:
1460                                         if (optval < -1 || optval >= 256)
1461                                                 error = EINVAL;
1462                                         else {
1463                                                 /* -1 = kernel default */
1464                                                 in6p->in6p_hops = optval;
1465                                         }
1466                                         break;
1467 #define OPTSET(bit) \
1468 do { \
1469         if (optval) \
1470                 in6p->in6p_flags |= (bit); \
1471         else \
1472                 in6p->in6p_flags &= ~(bit); \
1473 } while (0)
1474 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1475 /*
1476  * Although changed to RFC3542, It's better to also support RFC2292 API
1477  */
1478 #define OPTSET2292(bit) \
1479 do { \
1480         in6p->in6p_flags |= IN6P_RFC2292; \
1481         if (optval) \
1482                 in6p->in6p_flags |= (bit); \
1483         else \
1484                 in6p->in6p_flags &= ~(bit); \
1485 } while (/*CONSTCOND*/ 0)
1486
1487                                 case IPV6_RECVPKTINFO:
1488                                         /* cannot mix with RFC2292 */
1489                                         if (OPTBIT(IN6P_RFC2292)) {
1490                                                 error = EINVAL;
1491                                                 break;
1492                                         }
1493                                         OPTSET(IN6P_PKTINFO);
1494                                         break;
1495
1496                                 case IPV6_HOPLIMIT:
1497                                 {
1498                                         struct ip6_pktopts **optp;
1499
1500                                         /* cannot mix with RFC2292 */
1501                                         if (OPTBIT(IN6P_RFC2292)) {
1502                                                 error = EINVAL;
1503                                                 break;
1504                                         }
1505                                         optp = &in6p->in6p_outputopts;
1506                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1507                                             (u_char *)&optval, sizeof(optval),
1508                                             optp, uproto);
1509                                         break;
1510                                 }
1511
1512                                 case IPV6_RECVHOPLIMIT:
1513                                         /* cannot mix with RFC2292 */
1514                                         if (OPTBIT(IN6P_RFC2292)) {
1515                                                 error = EINVAL;
1516                                                 break;
1517                                         }
1518                                         OPTSET(IN6P_HOPLIMIT);
1519                                         break;
1520
1521                                 case IPV6_RECVHOPOPTS:
1522                                         /* cannot mix with RFC2292 */
1523                                         if (OPTBIT(IN6P_RFC2292)) {
1524                                                 error = EINVAL;
1525                                                 break;
1526                                         }
1527                                         OPTSET(IN6P_HOPOPTS);
1528                                         break;
1529
1530                                 case IPV6_RECVDSTOPTS:
1531                                         /* cannot mix with RFC2292 */
1532                                         if (OPTBIT(IN6P_RFC2292)) {
1533                                                 error = EINVAL;
1534                                                 break;
1535                                         }
1536                                         OPTSET(IN6P_DSTOPTS);
1537                                         break;
1538
1539                                 case IPV6_RECVRTHDRDSTOPTS:
1540                                         /* cannot mix with RFC2292 */
1541                                         if (OPTBIT(IN6P_RFC2292)) {
1542                                                 error = EINVAL;
1543                                                 break;
1544                                         }
1545                                         OPTSET(IN6P_RTHDRDSTOPTS);
1546                                         break;
1547
1548                                 case IPV6_RECVRTHDR:
1549                                         /* cannot mix with RFC2292 */
1550                                         if (OPTBIT(IN6P_RFC2292)) {
1551                                                 error = EINVAL;
1552                                                 break;
1553                                         }
1554                                         OPTSET(IN6P_RTHDR);
1555                                         break;
1556
1557                                 case IPV6_RECVPATHMTU:
1558                                         /*
1559                                          * We ignore this option for TCP
1560                                          * sockets.
1561                                          * (RFC3542 leaves this case
1562                                          * unspecified.)
1563                                          */
1564                                         if (uproto != IPPROTO_TCP)
1565                                                 OPTSET(IN6P_MTU);
1566                                         break;
1567
1568                                 case IPV6_RECVTCLASS:
1569                                         /* cannot mix with RFC2292 XXX */
1570                                         if (OPTBIT(IN6P_RFC2292)) {
1571                                                 error = EINVAL;
1572                                                 break;
1573                                         }
1574                                         OPTSET(IN6P_TCLASS);
1575                                         break;
1576
1577                                 case IPV6_AUTOFLOWLABEL:
1578                                         OPTSET(IN6P_AUTOFLOWLABEL);
1579                                         break;
1580
1581                                 case IPV6_FAITH:
1582                                         OPTSET(IN6P_FAITH);
1583                                         break;
1584
1585                                 case IPV6_V6ONLY:
1586                                         /*
1587                                          * make setsockopt(IPV6_V6ONLY)
1588                                          * available only prior to bind(2).
1589                                          */
1590                                         if (in6p->in6p_lport ||
1591                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1592                                         {
1593                                                 error = EINVAL;
1594                                                 break;
1595                                         }
1596                                         if (!optval) {
1597                                                 /* Don't allow v4-mapped */
1598                                                 error = EOPNOTSUPP;
1599                                         }
1600                                         break;
1601                                 }
1602                                 break;
1603
1604                         case IPV6_TCLASS:
1605                         case IPV6_DONTFRAG:
1606                         case IPV6_USE_MIN_MTU:
1607                         case IPV6_PREFER_TEMPADDR:
1608                                 if (optlen != sizeof(optval)) {
1609                                         error = EINVAL;
1610                                         break;
1611                                 }
1612                                 error = soopt_to_kbuf(sopt, &optval,
1613                                         sizeof optval, sizeof optval);
1614                                 if (error)
1615                                         break;
1616                                 {
1617                                         struct ip6_pktopts **optp;
1618                                         optp = &in6p->in6p_outputopts;
1619                                         error = ip6_pcbopt(optname,
1620                                             (u_char *)&optval, sizeof(optval),
1621                                             optp, uproto);
1622                                         break;
1623                                 }
1624
1625                         case IPV6_2292PKTINFO:
1626                         case IPV6_2292HOPLIMIT:
1627                         case IPV6_2292HOPOPTS:
1628                         case IPV6_2292DSTOPTS:
1629                         case IPV6_2292RTHDR:
1630                                 /* RFC 2292 */
1631                                 if (optlen != sizeof(int)) {
1632                                         error = EINVAL;
1633                                         break;
1634                                 }
1635                                 error = soopt_to_kbuf(sopt, &optval,
1636                                         sizeof optval, sizeof optval);
1637                                 if (error)
1638                                         break;
1639                                 switch (optname) {
1640                                 case IPV6_2292PKTINFO:
1641                                         OPTSET2292(IN6P_PKTINFO);
1642                                         break;
1643                                 case IPV6_2292HOPLIMIT:
1644                                         OPTSET2292(IN6P_HOPLIMIT);
1645                                         break;
1646                                 case IPV6_2292HOPOPTS:
1647                                         /*
1648                                          * Check super-user privilege.
1649                                          * See comments for IPV6_RECVHOPOPTS.
1650                                          */
1651                                         if (!privileged)
1652                                                 return (EPERM);
1653                                         OPTSET2292(IN6P_HOPOPTS);
1654                                         break;
1655                                 case IPV6_2292DSTOPTS:
1656                                         if (!privileged)
1657                                                 return (EPERM);
1658                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1659                                         break;
1660                                 case IPV6_2292RTHDR:
1661                                         OPTSET2292(IN6P_RTHDR);
1662                                         break;
1663                                 }
1664                                 break;
1665
1666                         case IPV6_PKTINFO:
1667                         case IPV6_HOPOPTS:
1668                         case IPV6_RTHDR:
1669                         case IPV6_DSTOPTS:
1670                         case IPV6_RTHDRDSTOPTS:
1671                         case IPV6_NEXTHOP:
1672                         {
1673                                 /*
1674                                  * New advanced API (RFC3542)
1675                                  */
1676                                 u_char *optbuf;
1677                                 u_char optbuf_storage[MCLBYTES];
1678                                 int optlen;
1679                                 struct ip6_pktopts **optp;
1680
1681                                 /* cannot mix with RFC2292 */
1682                                 if (OPTBIT(IN6P_RFC2292)) {
1683                                         error = EINVAL;
1684                                         break;
1685                                 }
1686
1687                                 /*
1688                                  * We only ensure valsize is not too large
1689                                  * here.  Further validation will be done
1690                                  * later.
1691                                  */
1692                                 error = soopt_to_kbuf(sopt, optbuf_storage,
1693                                     sizeof(optbuf_storage), 0);
1694                                 if (error)
1695                                         break;
1696                                 optlen = sopt->sopt_valsize;
1697                                 optbuf = optbuf_storage;
1698                                 optp = &in6p->in6p_outputopts;
1699                                 error = ip6_pcbopt(optname, optbuf, optlen,
1700                                     optp, uproto);
1701                                 break;
1702                         }
1703 #undef OPTSET
1704
1705                         case IPV6_MULTICAST_IF:
1706                         case IPV6_MULTICAST_HOPS:
1707                         case IPV6_MULTICAST_LOOP:
1708                         case IPV6_JOIN_GROUP:
1709                         case IPV6_LEAVE_GROUP:
1710                             {
1711                                 struct mbuf *m;
1712
1713                                 if (sopt->sopt_valsize > MLEN) {
1714                                         error = EMSGSIZE;
1715                                         break;
1716                                 }
1717                                 /* XXX */
1718                                 MGET(m, sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_HEADER);
1719                                 if (m == NULL) {
1720                                         error = ENOBUFS;
1721                                         break;
1722                                 }
1723                                 m->m_len = sopt->sopt_valsize;
1724                                 error = soopt_to_kbuf(sopt, mtod(m, char *),
1725                                                     m->m_len, m->m_len);
1726                                 error = ip6_setmoptions(sopt->sopt_name,
1727                                                         &in6p->in6p_moptions,
1728                                                         m);
1729                                 m_free(m);
1730                             }
1731                                 break;
1732
1733                         case IPV6_PORTRANGE:
1734                                 error = soopt_to_kbuf(sopt, &optval,
1735                                     sizeof optval, sizeof optval);
1736                                 if (error)
1737                                         break;
1738
1739                                 switch (optval) {
1740                                 case IPV6_PORTRANGE_DEFAULT:
1741                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1742                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1743                                         break;
1744
1745                                 case IPV6_PORTRANGE_HIGH:
1746                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1747                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1748                                         break;
1749
1750                                 case IPV6_PORTRANGE_LOW:
1751                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1752                                         in6p->in6p_flags |= IN6P_LOWPORT;
1753                                         break;
1754
1755                                 default:
1756                                         error = EINVAL;
1757                                         break;
1758                                 }
1759                                 break;
1760
1761 #if defined(IPSEC) || defined(FAST_IPSEC)
1762                         case IPV6_IPSEC_POLICY:
1763                             {
1764                                 caddr_t req = NULL;
1765                                 size_t len = 0;
1766                                 struct mbuf *m;
1767
1768                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1769                                         break;
1770                                 soopt_to_mbuf(sopt, m);         /* XXX */
1771                                 if (m) {
1772                                         req = mtod(m, caddr_t);
1773                                         len = m->m_len;
1774                                 }
1775                                 error = ipsec6_set_policy(in6p, optname, req,
1776                                                           len, privileged);
1777                                 m_freem(m);
1778                             }
1779                                 break;
1780 #endif /* KAME IPSEC */
1781
1782                         case IPV6_FW_ADD:
1783                         case IPV6_FW_DEL:
1784                         case IPV6_FW_FLUSH:
1785                         case IPV6_FW_ZERO:
1786                             {
1787                                 struct mbuf *m;
1788                                 struct mbuf **mp = &m;
1789
1790                                 if (ip6_fw_ctl_ptr == NULL)
1791                                         return EINVAL;
1792                                 /* XXX */
1793                                 if ((error = soopt_getm(sopt, &m)) != 0)
1794                                         break;
1795                                 /* XXX */
1796                                 soopt_to_mbuf(sopt, m);
1797                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1798                                 m = *mp;
1799                             }
1800                                 break;
1801
1802                         default:
1803                                 error = ENOPROTOOPT;
1804                                 break;
1805                         }
1806                         break;
1807
1808                 case SOPT_GET:
1809                         switch (optname) {
1810                         case IPV6_2292PKTOPTIONS:
1811 #ifdef IPV6_PKTOPTIONS
1812                         case IPV6_PKTOPTIONS:
1813 #endif
1814                                 /*
1815                                  * RFC3542 (effectively) deprecated the
1816                                  * semantics of the 2292-style pktoptions.
1817                                  * Since it was not reliable in nature (i.e.,
1818                                  * applications had to expect the lack of some
1819                                  * information after all), it would make sense
1820                                  * to simplify this part by always returning
1821                                  * empty data.
1822                                  */
1823                                 if (in6p->in6p_options) {
1824                                         struct mbuf *m;
1825                                         m = m_copym(in6p->in6p_options,
1826                                             0, M_COPYALL, M_WAITOK);
1827                                         error = soopt_from_mbuf(sopt, m);
1828                                         if (error == 0)
1829                                                 m_freem(m);
1830                                 } else
1831                                         sopt->sopt_valsize = 0;
1832                                 break;
1833
1834                         case IPV6_RECVHOPOPTS:
1835                         case IPV6_RECVDSTOPTS:
1836                         case IPV6_RECVRTHDRDSTOPTS:
1837                         case IPV6_UNICAST_HOPS:
1838                         case IPV6_RECVPKTINFO:
1839                         case IPV6_RECVHOPLIMIT:
1840                         case IPV6_RECVRTHDR:
1841                         case IPV6_RECVPATHMTU:
1842                         case IPV6_RECVTCLASS:
1843                         case IPV6_AUTOFLOWLABEL:
1844                         case IPV6_FAITH:
1845                         case IPV6_V6ONLY:
1846                         case IPV6_PORTRANGE:
1847                                 switch (optname) {
1848
1849                                 case IPV6_RECVHOPOPTS:
1850                                         optval = OPTBIT(IN6P_HOPOPTS);
1851                                         break;
1852
1853                                 case IPV6_RECVDSTOPTS:
1854                                         optval = OPTBIT(IN6P_DSTOPTS);
1855                                         break;
1856
1857                                 case IPV6_RECVRTHDRDSTOPTS:
1858                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1859                                         break;
1860
1861                                 case IPV6_RECVPKTINFO:
1862                                         optval = OPTBIT(IN6P_PKTINFO);
1863                                         break;
1864
1865                                 case IPV6_RECVHOPLIMIT:
1866                                         optval = OPTBIT(IN6P_HOPLIMIT);
1867                                         break;
1868
1869                                 case IPV6_RECVRTHDR:
1870                                         optval = OPTBIT(IN6P_RTHDR);
1871                                         break;
1872
1873                                 case IPV6_RECVPATHMTU:
1874                                         optval = OPTBIT(IN6P_MTU);
1875                                         break;
1876
1877                                 case IPV6_RECVTCLASS:
1878                                         optval = OPTBIT(IN6P_TCLASS);
1879                                         break;
1880
1881                                 case IPV6_AUTOFLOWLABEL:
1882                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1883                                         break;
1884
1885
1886                                 case IPV6_UNICAST_HOPS:
1887                                         optval = in6p->in6p_hops;
1888                                         break;
1889
1890                                 case IPV6_FAITH:
1891                                         optval = OPTBIT(IN6P_FAITH);
1892                                         break;
1893
1894                                 case IPV6_V6ONLY:
1895                                         optval = 1;
1896                                         break;
1897
1898                                 case IPV6_PORTRANGE:
1899                                     {
1900                                         int flags;
1901                                         flags = in6p->in6p_flags;
1902                                         if (flags & IN6P_HIGHPORT)
1903                                                 optval = IPV6_PORTRANGE_HIGH;
1904                                         else if (flags & IN6P_LOWPORT)
1905                                                 optval = IPV6_PORTRANGE_LOW;
1906                                         else
1907                                                 optval = 0;
1908                                         break;
1909                                     }
1910                                 }
1911                                 soopt_from_kbuf(sopt, &optval,
1912                                         sizeof optval);
1913                                 break;
1914
1915                         case IPV6_PATHMTU:
1916                         {
1917                                 u_long pmtu = 0;
1918                                 struct ip6_mtuinfo mtuinfo;
1919                                 struct route_in6 sro;
1920
1921                                 bzero(&sro, sizeof(sro));
1922
1923                                 if (!(so->so_state & SS_ISCONNECTED))
1924                                         return (ENOTCONN);
1925                                 /*
1926                                  * XXX: we dot not consider the case of source
1927                                  * routing, or optional information to specify
1928                                  * the outgoing interface.
1929                                  */
1930                                 error = ip6_getpmtu(&sro, NULL, NULL,
1931                                     &in6p->in6p_faddr, &pmtu, NULL);
1932                                 if (sro.ro_rt)
1933                                         RTFREE(sro.ro_rt);
1934                                 if (error)
1935                                         break;
1936                                 if (pmtu > IPV6_MAXPACKET)
1937                                         pmtu = IPV6_MAXPACKET;
1938
1939                                 bzero(&mtuinfo, sizeof(mtuinfo));
1940                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1941                                 optdata = (void *)&mtuinfo;
1942                                 optdatalen = sizeof(mtuinfo);
1943                                 soopt_from_kbuf(sopt, optdata,
1944                                     optdatalen);
1945                                 break;
1946                         }
1947
1948                         case IPV6_2292PKTINFO:
1949                         case IPV6_2292HOPLIMIT:
1950                         case IPV6_2292HOPOPTS:
1951                         case IPV6_2292RTHDR:
1952                         case IPV6_2292DSTOPTS:
1953                                 if (optname == IPV6_2292HOPOPTS ||
1954                                     optname == IPV6_2292DSTOPTS ||
1955                                     !privileged)
1956                                         return (EPERM);
1957                                 switch (optname) {
1958                                 case IPV6_2292PKTINFO:
1959                                         optval = OPTBIT(IN6P_PKTINFO);
1960                                         break;
1961                                 case IPV6_2292HOPLIMIT:
1962                                         optval = OPTBIT(IN6P_HOPLIMIT);
1963                                         break;
1964                                 case IPV6_2292HOPOPTS:
1965                                         if (!privileged)
1966                                                 return (EPERM);
1967                                         optval = OPTBIT(IN6P_HOPOPTS);
1968                                         break;
1969                                 case IPV6_2292RTHDR:
1970                                         optval = OPTBIT(IN6P_RTHDR);
1971                                         break;
1972                                 case IPV6_2292DSTOPTS:
1973                                         if (!privileged)
1974                                                 return (EPERM);
1975                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1976                                         break;
1977                                 }
1978                                 soopt_from_kbuf(sopt, &optval,
1979                                         sizeof optval);
1980                                 break;
1981
1982                         case IPV6_PKTINFO:
1983                         case IPV6_HOPOPTS:
1984                         case IPV6_RTHDR:
1985                         case IPV6_DSTOPTS:
1986                         case IPV6_RTHDRDSTOPTS:
1987                         case IPV6_NEXTHOP:
1988                         case IPV6_TCLASS:
1989                         case IPV6_DONTFRAG:
1990                         case IPV6_USE_MIN_MTU:
1991                         case IPV6_PREFER_TEMPADDR:
1992                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
1993                                     optname, sopt);
1994                                 break;
1995
1996                         case IPV6_MULTICAST_IF:
1997                         case IPV6_MULTICAST_HOPS:
1998                         case IPV6_MULTICAST_LOOP:
1999                         case IPV6_JOIN_GROUP:
2000                         case IPV6_LEAVE_GROUP:
2001                             {
2002                                 struct mbuf *m;
2003                                 error = ip6_getmoptions(sopt->sopt_name,
2004                                     in6p->in6p_moptions, &m);
2005                                 if (error == 0) {
2006                                         soopt_from_kbuf(sopt,
2007                                             mtod(m, char *), m->m_len);
2008                                 }
2009                                 m_freem(m);
2010                             }
2011                                 break;
2012
2013 #if defined(IPSEC) || defined(FAST_IPSEC)
2014                         case IPV6_IPSEC_POLICY:
2015                           {
2016                                 caddr_t req = NULL;
2017                                 size_t len = 0;
2018                                 struct mbuf *m = NULL;
2019                                 struct mbuf **mp = &m;
2020
2021                                 error = soopt_getm(sopt, &m); /* XXX */
2022                                 if (error != 0)
2023                                         break;
2024                                 soopt_to_mbuf(sopt, m); /* XXX */
2025                                 if (m) {
2026                                         req = mtod(m, caddr_t);
2027                                         len = m->m_len;
2028                                 }
2029                                 error = ipsec6_get_policy(in6p, req, len, mp);
2030                                 if (error == 0)
2031                                         error = soopt_from_mbuf(sopt, m);/*XXX*/
2032                                 if (error == 0 && m != NULL)
2033                                         m_freem(m);
2034                                 break;
2035                           }
2036 #endif /* KAME IPSEC */
2037
2038                         case IPV6_FW_GET:
2039                           {
2040                                 struct mbuf *m;
2041                                 struct mbuf **mp = &m;
2042
2043                                 if (ip6_fw_ctl_ptr == NULL)
2044                                 {
2045                                         return EINVAL;
2046                                 }
2047                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
2048                                 if (error == 0)
2049                                         error = soopt_from_mbuf(sopt, m); /* XXX */
2050                                 if (error == 0 && m != NULL)
2051                                         m_freem(m);
2052                           }
2053                                 break;
2054
2055                         default:
2056                                 error = ENOPROTOOPT;
2057                                 break;
2058                         }
2059                         break;
2060                 }
2061         } else {
2062                 error = EINVAL;
2063         }
2064         return (error);
2065 }
2066
2067 int
2068 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2069 {
2070         int error = 0, optval, optlen;
2071         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2072         struct in6pcb *in6p = sotoin6pcb(so);
2073         int level, op, optname;
2074
2075         if (sopt) {
2076                 level = sopt->sopt_level;
2077                 op = sopt->sopt_dir;
2078                 optname = sopt->sopt_name;
2079                 optlen = sopt->sopt_valsize;
2080         } else
2081                 panic("ip6_raw_ctloutput: arg soopt is NULL");
2082
2083         if (level != IPPROTO_IPV6) {
2084                 return (EINVAL);
2085         }
2086
2087         switch (optname) {
2088         case IPV6_CHECKSUM:
2089                 /*
2090                  * For ICMPv6 sockets, no modification allowed for checksum
2091                  * offset, permit "no change" values to help existing apps.
2092                  *
2093                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2094                  * for an ICMPv6 socket will fail."
2095                  * The current behavior does not meet RFC3542.
2096                  */
2097                 switch (op) {
2098                 case SOPT_SET:
2099                         if (optlen != sizeof(int)) {
2100                                 error = EINVAL;
2101                                 break;
2102                         }
2103                         error = soopt_to_kbuf(sopt, &optval,
2104                                     sizeof optval, sizeof optval);
2105                         if (error)
2106                                 break;
2107                         if ((optval % 2) != 0) {
2108                                 /* the API assumes even offset values */
2109                                 error = EINVAL;
2110                         } else if (so->so_proto->pr_protocol ==
2111                             IPPROTO_ICMPV6) {
2112                                 if (optval != icmp6off)
2113                                         error = EINVAL;
2114                         } else
2115                                 in6p->in6p_cksum = optval;
2116                         break;
2117
2118                 case SOPT_GET:
2119                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2120                                 optval = icmp6off;
2121                         else
2122                                 optval = in6p->in6p_cksum;
2123
2124                         soopt_from_kbuf(sopt, &optval, sizeof(optval));
2125                         break;
2126
2127                 default:
2128                         error = EINVAL;
2129                         break;
2130                 }
2131                 break;
2132
2133         default:
2134                 error = ENOPROTOOPT;
2135                 break;
2136         }
2137
2138         return (error);
2139 }
2140
2141 /*
2142  * Set up IP6 options in pcb for insertion in output packets or
2143  * specifying behavior of outgoing packets.
2144  */
2145 static int
2146 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2147     struct socket *so, struct sockopt *sopt)
2148 {
2149         int priv = 0;
2150         struct ip6_pktopts *opt = *pktopt;
2151         int error = 0;
2152
2153         /* turn off any old options. */
2154         if (opt) {
2155 #ifdef DIAGNOSTIC
2156                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2157                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2158                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2159                         kprintf("ip6_pcbopts: all specified options are cleared.\n");
2160 #endif
2161                 ip6_clearpktopts(opt, -1);
2162         } else
2163                 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2164         *pktopt = NULL;
2165
2166         if (!m || m->m_len == 0) {
2167                 /*
2168                  * Only turning off any previous options, regardless of
2169                  * whether the opt is just created or given.
2170                  */
2171                 kfree(opt, M_IP6OPT);
2172                 return (0);
2173         }
2174
2175         /*  set options specified by user. */
2176         if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) {
2177                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2178                 kfree(opt, M_IP6OPT);
2179                 return (error);
2180         }
2181         *pktopt = opt;
2182         return (0);
2183 }
2184
2185
2186 /*
2187  * Below three functions are introduced by merge to RFC3542
2188  */
2189
2190 static int
2191 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2192 {
2193         void *optdata = NULL;
2194         int optdatalen = 0;
2195         struct ip6_ext *ip6e;
2196         int error = 0;
2197         struct in6_pktinfo null_pktinfo;
2198         int deftclass = 0, on;
2199         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2200         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2201
2202         switch (optname) {
2203         case IPV6_PKTINFO:
2204                 if (pktopt && pktopt->ip6po_pktinfo)
2205                         optdata = (void *)pktopt->ip6po_pktinfo;
2206                 else {
2207                         /* XXX: we don't have to do this every time... */
2208                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2209                         optdata = (void *)&null_pktinfo;
2210                 }
2211                 optdatalen = sizeof(struct in6_pktinfo);
2212                 break;
2213         case IPV6_TCLASS:
2214                 if (pktopt && pktopt->ip6po_tclass >= 0)
2215                         optdata = (void *)&pktopt->ip6po_tclass;
2216                 else
2217                         optdata = (void *)&deftclass;
2218                 optdatalen = sizeof(int);
2219                 break;
2220         case IPV6_HOPOPTS:
2221                 if (pktopt && pktopt->ip6po_hbh) {
2222                         optdata = (void *)pktopt->ip6po_hbh;
2223                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2224                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2225                 }
2226                 break;
2227         case IPV6_RTHDR:
2228                 if (pktopt && pktopt->ip6po_rthdr) {
2229                         optdata = (void *)pktopt->ip6po_rthdr;
2230                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2231                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2232                 }
2233                 break;
2234         case IPV6_RTHDRDSTOPTS:
2235                 if (pktopt && pktopt->ip6po_dest1) {
2236                         optdata = (void *)pktopt->ip6po_dest1;
2237                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2238                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2239                 }
2240                 break;
2241         case IPV6_DSTOPTS:
2242                 if (pktopt && pktopt->ip6po_dest2) {
2243                         optdata = (void *)pktopt->ip6po_dest2;
2244                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2245                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2246                 }
2247                 break;
2248         case IPV6_NEXTHOP:
2249                 if (pktopt && pktopt->ip6po_nexthop) {
2250                         optdata = (void *)pktopt->ip6po_nexthop;
2251                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2252                 }
2253                 break;
2254         case IPV6_USE_MIN_MTU:
2255                 if (pktopt)
2256                         optdata = (void *)&pktopt->ip6po_minmtu;
2257                 else
2258                         optdata = (void *)&defminmtu;
2259                 optdatalen = sizeof(int);
2260                 break;
2261         case IPV6_DONTFRAG:
2262                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2263                         on = 1;
2264                 else
2265                         on = 0;
2266                 optdata = (void *)&on;
2267                 optdatalen = sizeof(on);
2268                 break;
2269         case IPV6_PREFER_TEMPADDR:
2270                 if (pktopt)
2271                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2272                 else
2273                         optdata = (void *)&defpreftemp;
2274                 optdatalen = sizeof(int);
2275                 break;
2276         default:                /* should not happen */
2277 #ifdef DIAGNOSTIC
2278                 panic("ip6_getpcbopt: unexpected option");
2279 #endif
2280                 return (ENOPROTOOPT);
2281         }
2282
2283         soopt_from_kbuf(sopt, optdata, optdatalen);
2284
2285         return (error);
2286 }
2287
2288 /*
2289  * initialize ip6_pktopts.  beware that there are non-zero default values in
2290  * the struct.
2291  */
2292
2293 static int
2294 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto)
2295 {
2296         struct ip6_pktopts *opt;
2297         int priv =0;
2298         if (*pktopt == NULL) {
2299                 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2300                 init_ip6pktopts(*pktopt);
2301         }
2302         opt = *pktopt;
2303
2304         return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv));
2305 }
2306
2307 /*
2308  * initialize ip6_pktopts.  beware that there are non-zero default values in
2309  * the struct.
2310  */
2311 void
2312 init_ip6pktopts(struct ip6_pktopts *opt)
2313 {
2314
2315         bzero(opt, sizeof(*opt));
2316         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2317         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2318         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2319         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2320 }
2321
2322 void
2323 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2324 {
2325         if (pktopt == NULL)
2326                 return;
2327
2328         if (optname == -1 || optname == IPV6_PKTINFO) {
2329                 if (pktopt->ip6po_pktinfo)
2330                         kfree(pktopt->ip6po_pktinfo, M_IP6OPT);
2331                 pktopt->ip6po_pktinfo = NULL;
2332         }
2333         if (optname == -1 || optname == IPV6_HOPLIMIT)
2334                 pktopt->ip6po_hlim = -1;
2335         if (optname == -1 || optname == IPV6_TCLASS)
2336                 pktopt->ip6po_tclass = -1;
2337         if (optname == -1 || optname == IPV6_NEXTHOP) {
2338                 if (pktopt->ip6po_nextroute.ro_rt) {
2339                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2340                         pktopt->ip6po_nextroute.ro_rt = NULL;
2341                 }
2342                 if (pktopt->ip6po_nexthop)
2343                         kfree(pktopt->ip6po_nexthop, M_IP6OPT);
2344                 pktopt->ip6po_nexthop = NULL;
2345         }
2346         if (optname == -1 || optname == IPV6_HOPOPTS) {
2347                 if (pktopt->ip6po_hbh)
2348                         kfree(pktopt->ip6po_hbh, M_IP6OPT);
2349                 pktopt->ip6po_hbh = NULL;
2350         }
2351         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2352                 if (pktopt->ip6po_dest1)
2353                         kfree(pktopt->ip6po_dest1, M_IP6OPT);
2354                 pktopt->ip6po_dest1 = NULL;
2355         }
2356         if (optname == -1 || optname == IPV6_RTHDR) {
2357                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2358                         kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2359                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2360                 if (pktopt->ip6po_route.ro_rt) {
2361                         RTFREE(pktopt->ip6po_route.ro_rt);
2362                         pktopt->ip6po_route.ro_rt = NULL;
2363                 }
2364         }
2365         if (optname == -1 || optname == IPV6_DSTOPTS) {
2366                 if (pktopt->ip6po_dest2)
2367                         kfree(pktopt->ip6po_dest2, M_IP6OPT);
2368                 pktopt->ip6po_dest2 = NULL;
2369         }
2370 }
2371
2372 #define PKTOPT_EXTHDRCPY(type) \
2373 do {\
2374         if (src->type) {\
2375                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2376                 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\
2377                 if (dst->type == NULL)\
2378                         goto bad;\
2379                 bcopy(src->type, dst->type, hlen);\
2380         }\
2381 } while (0)
2382
2383 struct ip6_pktopts *
2384 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2385 {
2386         struct ip6_pktopts *dst;
2387
2388         if (src == NULL) {
2389                 kprintf("ip6_clearpktopts: invalid argument\n");
2390                 return (NULL);
2391         }
2392
2393         dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO);
2394         if (dst == NULL)
2395                 return (NULL);
2396
2397         dst->ip6po_hlim = src->ip6po_hlim;
2398         if (src->ip6po_pktinfo) {
2399                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2400                     M_IP6OPT, canwait);
2401                 if (dst->ip6po_pktinfo == NULL)
2402                         goto bad;
2403                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2404         }
2405         if (src->ip6po_nexthop) {
2406                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2407                     M_IP6OPT, canwait);
2408                 if (dst->ip6po_nexthop == NULL)
2409                         goto bad;
2410                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2411                     src->ip6po_nexthop->sa_len);
2412         }
2413         PKTOPT_EXTHDRCPY(ip6po_hbh);
2414         PKTOPT_EXTHDRCPY(ip6po_dest1);
2415         PKTOPT_EXTHDRCPY(ip6po_dest2);
2416         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2417         return (dst);
2418
2419 bad:
2420         if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT);
2421         if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT);
2422         if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT);
2423         if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT);
2424         if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT);
2425         if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT);
2426         kfree(dst, M_IP6OPT);
2427         return (NULL);
2428 }
2429
2430 static int
2431 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2432 {
2433         if (dst == NULL || src == NULL)  {
2434 #ifdef DIAGNOSTIC
2435                 kprintf("ip6_clearpktopts: invalid argument\n");
2436 #endif
2437                 return (EINVAL);
2438         }
2439
2440         dst->ip6po_hlim = src->ip6po_hlim;
2441         dst->ip6po_tclass = src->ip6po_tclass;
2442         dst->ip6po_flags = src->ip6po_flags;
2443         if (src->ip6po_pktinfo) {
2444                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2445                     M_IP6OPT, canwait);
2446                 if (dst->ip6po_pktinfo == NULL)
2447                         goto bad;
2448                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2449         }
2450         if (src->ip6po_nexthop) {
2451                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2452                     M_IP6OPT, canwait);
2453                 if (dst->ip6po_nexthop == NULL)
2454                         goto bad;
2455                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2456                     src->ip6po_nexthop->sa_len);
2457         }
2458         PKTOPT_EXTHDRCPY(ip6po_hbh);
2459         PKTOPT_EXTHDRCPY(ip6po_dest1);
2460         PKTOPT_EXTHDRCPY(ip6po_dest2);
2461         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2462         return (0);
2463
2464   bad:
2465         ip6_clearpktopts(dst, -1);
2466         return (ENOBUFS);
2467 }
2468 #undef PKTOPT_EXTHDRCPY
2469
2470 void
2471 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2472 {
2473         if (pktopt == NULL)
2474                 return;
2475
2476         ip6_clearpktopts(pktopt, -1);
2477
2478         kfree(pktopt, M_IP6OPT);
2479 }
2480
2481 /*
2482  * Set the IP6 multicast options in response to user setsockopt().
2483  */
2484 static int
2485 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2486 {
2487         int error = 0;
2488         u_int loop, ifindex;
2489         struct ipv6_mreq *mreq;
2490         struct ifnet *ifp;
2491         struct ip6_moptions *im6o = *im6op;
2492         struct route_in6 ro;
2493         struct sockaddr_in6 *dst;
2494         struct in6_multi_mship *imm;
2495         struct thread *td = curthread;
2496
2497         if (im6o == NULL) {
2498                 /*
2499                  * No multicast option buffer attached to the pcb;
2500                  * allocate one and initialize to default values.
2501                  */
2502                 im6o = (struct ip6_moptions *)
2503                         kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2504
2505                 *im6op = im6o;
2506                 im6o->im6o_multicast_ifp = NULL;
2507                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2508                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2509                 LIST_INIT(&im6o->im6o_memberships);
2510         }
2511
2512         switch (optname) {
2513
2514         case IPV6_MULTICAST_IF:
2515                 /*
2516                  * Select the interface for outgoing multicast packets.
2517                  */
2518                 if (m == NULL || m->m_len != sizeof(u_int)) {
2519                         error = EINVAL;
2520                         break;
2521                 }
2522                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2523                 if (ifindex < 0 || if_index < ifindex) {
2524                         error = ENXIO;  /* XXX EINVAL? */
2525                         break;
2526                 }
2527                 ifp = ifindex2ifnet[ifindex];
2528                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2529                         error = EADDRNOTAVAIL;
2530                         break;
2531                 }
2532                 im6o->im6o_multicast_ifp = ifp;
2533                 break;
2534
2535         case IPV6_MULTICAST_HOPS:
2536             {
2537                 /*
2538                  * Set the IP6 hoplimit for outgoing multicast packets.
2539                  */
2540                 int optval;
2541                 if (m == NULL || m->m_len != sizeof(int)) {
2542                         error = EINVAL;
2543                         break;
2544                 }
2545                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2546                 if (optval < -1 || optval >= 256)
2547                         error = EINVAL;
2548                 else if (optval == -1)
2549                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2550                 else
2551                         im6o->im6o_multicast_hlim = optval;
2552                 break;
2553             }
2554
2555         case IPV6_MULTICAST_LOOP:
2556                 /*
2557                  * Set the loopback flag for outgoing multicast packets.
2558                  * Must be zero or one.
2559                  */
2560                 if (m == NULL || m->m_len != sizeof(u_int)) {
2561                         error = EINVAL;
2562                         break;
2563                 }
2564                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2565                 if (loop > 1) {
2566                         error = EINVAL;
2567                         break;
2568                 }
2569                 im6o->im6o_multicast_loop = loop;
2570                 break;
2571
2572         case IPV6_JOIN_GROUP:
2573                 /*
2574                  * Add a multicast group membership.
2575                  * Group must be a valid IP6 multicast address.
2576                  */
2577                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2578                         error = EINVAL;
2579                         break;
2580                 }
2581                 mreq = mtod(m, struct ipv6_mreq *);
2582                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2583                         /*
2584                          * We use the unspecified address to specify to accept
2585                          * all multicast addresses. Only super user is allowed
2586                          * to do this.
2587                          */
2588                         if (priv_check(td, PRIV_ROOT)) {
2589                                 error = EACCES;
2590                                 break;
2591                         }
2592                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2593                         error = EINVAL;
2594                         break;
2595                 }
2596
2597                 /*
2598                  * If the interface is specified, validate it.
2599                  */
2600                 if (mreq->ipv6mr_interface < 0
2601                  || if_index < mreq->ipv6mr_interface) {
2602                         error = ENXIO;  /* XXX EINVAL? */
2603                         break;
2604                 }
2605                 /*
2606                  * If no interface was explicitly specified, choose an
2607                  * appropriate one according to the given multicast address.
2608                  */
2609                 if (mreq->ipv6mr_interface == 0) {
2610                         /*
2611                          * If the multicast address is in node-local scope,
2612                          * the interface should be a loopback interface.
2613                          * Otherwise, look up the routing table for the
2614                          * address, and choose the outgoing interface.
2615                          *   XXX: is it a good approach?
2616                          */
2617                         if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2618                                 ifp = &loif[0];
2619                         } else {
2620                                 ro.ro_rt = NULL;
2621                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2622                                 bzero(dst, sizeof(*dst));
2623                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2624                                 dst->sin6_family = AF_INET6;
2625                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2626                                 rtalloc((struct route *)&ro);
2627                                 if (ro.ro_rt == NULL) {
2628                                         error = EADDRNOTAVAIL;
2629                                         break;
2630                                 }
2631                                 ifp = ro.ro_rt->rt_ifp;
2632                                 rtfree(ro.ro_rt);
2633                         }
2634                 } else
2635                         ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2636
2637                 /*
2638                  * See if we found an interface, and confirm that it
2639                  * supports multicast
2640                  */
2641                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2642                         error = EADDRNOTAVAIL;
2643                         break;
2644                 }
2645                 /*
2646                  * Put interface index into the multicast address,
2647                  * if the address has link-local scope.
2648                  */
2649                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2650                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2651                                 = htons(mreq->ipv6mr_interface);
2652                 }
2653                 /*
2654                  * See if the membership already exists.
2655                  */
2656                 for (imm = im6o->im6o_memberships.lh_first;
2657                      imm != NULL; imm = imm->i6mm_chain.le_next)
2658                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2659                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2660                                                &mreq->ipv6mr_multiaddr))
2661                                 break;
2662                 if (imm != NULL) {
2663                         error = EADDRINUSE;
2664                         break;
2665                 }
2666                 /*
2667                  * Everything looks good; add a new record to the multicast
2668                  * address list for the given interface.
2669                  */
2670                 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2671                 if ((imm->i6mm_maddr =
2672                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2673                         kfree(imm, M_IPMADDR);
2674                         break;
2675                 }
2676                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2677                 break;
2678
2679         case IPV6_LEAVE_GROUP:
2680                 /*
2681                  * Drop a multicast group membership.
2682                  * Group must be a valid IP6 multicast address.
2683                  */
2684                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2685                         error = EINVAL;
2686                         break;
2687                 }
2688                 mreq = mtod(m, struct ipv6_mreq *);
2689                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2690                         if (priv_check(td, PRIV_ROOT)) {
2691                                 error = EACCES;
2692                                 break;
2693                         }
2694                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2695                         error = EINVAL;
2696                         break;
2697                 }
2698                 /*
2699                  * If an interface address was specified, get a pointer
2700                  * to its ifnet structure.
2701                  */
2702                 if (mreq->ipv6mr_interface < 0
2703                  || if_index < mreq->ipv6mr_interface) {
2704                         error = ENXIO;  /* XXX EINVAL? */
2705                         break;
2706                 }
2707                 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2708                 /*
2709                  * Put interface index into the multicast address,
2710                  * if the address has link-local scope.
2711                  */
2712                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2713                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2714                                 = htons(mreq->ipv6mr_interface);
2715                 }
2716
2717                 /*
2718                  * Find the membership in the membership list.
2719                  */
2720                 for (imm = im6o->im6o_memberships.lh_first;
2721                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2722                         if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2723                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2724                             &mreq->ipv6mr_multiaddr))
2725                                 break;
2726                 }
2727                 if (imm == NULL) {
2728                         /* Unable to resolve interface */
2729                         error = EADDRNOTAVAIL;
2730                         break;
2731                 }
2732                 /*
2733                  * Give up the multicast address record to which the
2734                  * membership points.
2735                  */
2736                 LIST_REMOVE(imm, i6mm_chain);
2737                 in6_delmulti(imm->i6mm_maddr);
2738                 kfree(imm, M_IPMADDR);
2739                 break;
2740
2741         default:
2742                 error = EOPNOTSUPP;
2743                 break;
2744         }
2745
2746         /*
2747          * If all options have default values, no need to keep the mbuf.
2748          */
2749         if (im6o->im6o_multicast_ifp == NULL &&
2750             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2751             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2752             im6o->im6o_memberships.lh_first == NULL) {
2753                 kfree(*im6op, M_IPMOPTS);
2754                 *im6op = NULL;
2755         }
2756
2757         return (error);
2758 }
2759
2760 /*
2761  * Return the IP6 multicast options in response to user getsockopt().
2762  */
2763 static int
2764 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2765 {
2766         u_int *hlim, *loop, *ifindex;
2767
2768         *mp = m_get(M_WAITOK, MT_HEADER);               /* XXX */
2769
2770         switch (optname) {
2771
2772         case IPV6_MULTICAST_IF:
2773                 ifindex = mtod(*mp, u_int *);
2774                 (*mp)->m_len = sizeof(u_int);
2775                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2776                         *ifindex = 0;
2777                 else
2778                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2779                 return (0);
2780
2781         case IPV6_MULTICAST_HOPS:
2782                 hlim = mtod(*mp, u_int *);
2783                 (*mp)->m_len = sizeof(u_int);
2784                 if (im6o == NULL)
2785                         *hlim = ip6_defmcasthlim;
2786                 else
2787                         *hlim = im6o->im6o_multicast_hlim;
2788                 return (0);
2789
2790         case IPV6_MULTICAST_LOOP:
2791                 loop = mtod(*mp, u_int *);
2792                 (*mp)->m_len = sizeof(u_int);
2793                 if (im6o == NULL)
2794                         *loop = ip6_defmcasthlim;
2795                 else
2796                         *loop = im6o->im6o_multicast_loop;
2797                 return (0);
2798
2799         default:
2800                 return (EOPNOTSUPP);
2801         }
2802 }
2803
2804 /*
2805  * Discard the IP6 multicast options.
2806  */
2807 void
2808 ip6_freemoptions(struct ip6_moptions *im6o)
2809 {
2810         struct in6_multi_mship *imm;
2811
2812         if (im6o == NULL)
2813                 return;
2814
2815         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2816                 LIST_REMOVE(imm, i6mm_chain);
2817                 if (imm->i6mm_maddr)
2818                         in6_delmulti(imm->i6mm_maddr);
2819                 kfree(imm, M_IPMADDR);
2820         }
2821         kfree(im6o, M_IPMOPTS);
2822 }
2823
2824 /*
2825  * Set a particular packet option, as a sticky option or an ancillary data
2826  * item.  "len" can be 0 only when it's a sticky option.
2827  * We have 4 cases of combination of "sticky" and "cmsg":
2828  * "sticky=0, cmsg=0": impossible
2829  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2830  * "sticky=1, cmsg=0": RFC3542 socket option
2831  * "sticky=1, cmsg=1": RFC2292 socket option
2832  */
2833 static int
2834 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2835      int sticky, int cmsg, int uproto, int priv)
2836 {
2837         int minmtupolicy, preftemp;
2838         //int error;
2839
2840         if (!sticky && !cmsg) {
2841                 kprintf("ip6_setpktoption: impossible case\n");
2842                 return (EINVAL);
2843         }
2844
2845         /*
2846          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2847          * not be specified in the context of RFC3542.  Conversely,
2848          * RFC3542 types should not be specified in the context of RFC2292.
2849          */
2850         if (!cmsg) {
2851                 switch (optname) {
2852                 case IPV6_2292PKTINFO:
2853                 case IPV6_2292HOPLIMIT:
2854                 case IPV6_2292NEXTHOP:
2855                 case IPV6_2292HOPOPTS:
2856                 case IPV6_2292DSTOPTS:
2857                 case IPV6_2292RTHDR:
2858                 case IPV6_2292PKTOPTIONS:
2859                         return (ENOPROTOOPT);
2860                 }
2861         }
2862         if (sticky && cmsg) {
2863                 switch (optname) {
2864                 case IPV6_PKTINFO:
2865                 case IPV6_HOPLIMIT:
2866                 case IPV6_NEXTHOP:
2867                 case IPV6_HOPOPTS:
2868                 case IPV6_DSTOPTS:
2869                 case IPV6_RTHDRDSTOPTS:
2870                 case IPV6_RTHDR:
2871                 case IPV6_USE_MIN_MTU:
2872                 case IPV6_DONTFRAG:
2873                 case IPV6_TCLASS:
2874                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2875                         return (ENOPROTOOPT);
2876                 }
2877         }
2878
2879         switch (optname) {
2880         case IPV6_2292PKTINFO:
2881         case IPV6_PKTINFO:
2882         {
2883                 struct in6_pktinfo *pktinfo;
2884                 if (len != sizeof(struct in6_pktinfo))
2885                         return (EINVAL);
2886                 pktinfo = (struct in6_pktinfo *)buf;
2887
2888                 /*
2889                  * An application can clear any sticky IPV6_PKTINFO option by
2890                  * doing a "regular" setsockopt with ipi6_addr being
2891                  * in6addr_any and ipi6_ifindex being zero.
2892                  * [RFC 3542, Section 6]
2893                  */
2894                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2895                     pktinfo->ipi6_ifindex == 0 &&
2896                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2897                         ip6_clearpktopts(opt, optname);
2898                         break;
2899                 }
2900
2901                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2902                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2903                         return (EINVAL);
2904                 }
2905
2906                 /* validate the interface index if specified. */
2907                 if (pktinfo->ipi6_ifindex > if_index ||
2908                     pktinfo->ipi6_ifindex < 0) {
2909                          return (ENXIO);
2910                 }
2911                 /*
2912                  * Check if the requested source address is indeed a
2913                  * unicast address assigned to the node, and can be
2914                  * used as the packet's source address.
2915                  */
2916                 if (opt->ip6po_pktinfo != NULL &&
2917                     !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2918                         struct in6_ifaddr *ia6;
2919                         struct sockaddr_in6 sin6;
2920
2921                         bzero(&sin6, sizeof(sin6));
2922                         sin6.sin6_len = sizeof(sin6);
2923                         sin6.sin6_family = AF_INET6;
2924                         sin6.sin6_addr =
2925                         opt->ip6po_pktinfo->ipi6_addr;
2926                         ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2927                         if (ia6 == NULL ||
2928                                 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2929                                         IN6_IFF_NOTREADY)) != 0)
2930                         return (EADDRNOTAVAIL);
2931                 }
2932
2933                 /*
2934                  * We store the address anyway, and let in6_selectsrc()
2935                  * validate the specified address.  This is because ipi6_addr
2936                  * may not have enough information about its scope zone, and
2937                  * we may need additional information (such as outgoing
2938                  * interface or the scope zone of a destination address) to
2939                  * disambiguate the scope.
2940                  * XXX: the delay of the validation may confuse the
2941                  * application when it is used as a sticky option.
2942                  */
2943                 if (opt->ip6po_pktinfo == NULL) {
2944                         opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo),
2945                             M_IP6OPT, M_NOWAIT);
2946                         if (opt->ip6po_pktinfo == NULL)
2947                                 return (ENOBUFS);
2948                 }
2949                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2950                 break;
2951         }
2952
2953         case IPV6_2292HOPLIMIT:
2954         case IPV6_HOPLIMIT:
2955         {
2956                 int *hlimp;
2957
2958                 /*
2959                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2960                  * to simplify the ordering among hoplimit options.
2961                  */
2962                 if (optname == IPV6_HOPLIMIT && sticky)
2963                         return (ENOPROTOOPT);
2964
2965                 if (len != sizeof(int))
2966                         return (EINVAL);
2967                 hlimp = (int *)buf;
2968                 if (*hlimp < -1 || *hlimp > 255)
2969                         return (EINVAL);
2970
2971                 opt->ip6po_hlim = *hlimp;
2972                 break;
2973         }
2974
2975         case IPV6_TCLASS:
2976         {
2977                 int tclass;
2978
2979                 if (len != sizeof(int))
2980                         return (EINVAL);
2981                 tclass = *(int *)buf;
2982                 if (tclass < -1 || tclass > 255)
2983                         return (EINVAL);
2984
2985                 opt->ip6po_tclass = tclass;
2986                 break;
2987         }
2988
2989         case IPV6_2292NEXTHOP:
2990         case IPV6_NEXTHOP:
2991                 if (!priv)
2992                         return (EPERM);
2993
2994                 if (len == 0) { /* just remove the option */
2995                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2996                         break;
2997                 }
2998
2999                 /* check if cmsg_len is large enough for sa_len */
3000                 if (len < sizeof(struct sockaddr) || len < *buf)
3001                         return (EINVAL);
3002
3003                 switch (((struct sockaddr *)buf)->sa_family) {
3004                 case AF_INET6:
3005                 {
3006                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3007                         //int error;
3008
3009                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3010                                 return (EINVAL);
3011
3012                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3013                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3014                                 return (EINVAL);
3015                         }
3016                         break;
3017                 }
3018                 case AF_LINK:   /* should eventually be supported */
3019                 default:
3020                         return (EAFNOSUPPORT);
3021                 }
3022
3023                 /* turn off the previous option, then set the new option. */
3024                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3025                 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT);
3026                 if (opt->ip6po_nexthop == NULL)
3027                         return (ENOBUFS);
3028                 bcopy(buf, opt->ip6po_nexthop, *buf);
3029                 break;
3030
3031         case IPV6_2292HOPOPTS:
3032         case IPV6_HOPOPTS:
3033         {
3034                 struct ip6_hbh *hbh;
3035                 int hbhlen;
3036
3037                 /*
3038                  * XXX: We don't allow a non-privileged user to set ANY HbH
3039                  * options, since per-option restriction has too much
3040                  * overhead.
3041                  */
3042                 if (!priv)
3043                         return (EPERM);
3044                 if (len == 0) {
3045                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3046                         break;  /* just remove the option */
3047                 }
3048
3049                 /* message length validation */
3050                 if (len < sizeof(struct ip6_hbh))
3051                         return (EINVAL);
3052                 hbh = (struct ip6_hbh *)buf;
3053                 hbhlen = (hbh->ip6h_len + 1) << 3;
3054                 if (len != hbhlen)
3055                         return (EINVAL);
3056
3057                 /* turn off the previous option, then set the new option. */
3058                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3059                 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT);
3060                 if (opt->ip6po_hbh == NULL)
3061                         return (ENOBUFS);
3062                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3063
3064                 break;
3065         }
3066
3067         case IPV6_2292DSTOPTS:
3068         case IPV6_DSTOPTS:
3069         case IPV6_RTHDRDSTOPTS:
3070         {
3071                 struct ip6_dest *dest, **newdest = NULL;
3072                 int destlen;
3073                 if (!priv)
3074                         return (EPERM);
3075
3076                 if (len == 0) {
3077                         ip6_clearpktopts(opt, optname);
3078                         break;  /* just remove the option */
3079                 }
3080
3081                 /* message length validation */
3082                 if (len < sizeof(struct ip6_dest))
3083                         return (EINVAL);
3084                 dest = (struct ip6_dest *)buf;
3085                 destlen = (dest->ip6d_len + 1) << 3;
3086                 if (len != destlen)
3087                         return (EINVAL);
3088
3089                 /*
3090                  * Determine the position that the destination options header
3091                  * should be inserted; before or after the routing header.
3092                  */
3093                 switch (optname) {
3094                 case IPV6_2292DSTOPTS:
3095                         /*
3096                          * The old advacned API is ambiguous on this point.
3097                          * Our approach is to determine the position based
3098                          * according to the existence of a routing header.
3099                          * Note, however, that this depends on the order of the
3100                          * extension headers in the ancillary data; the 1st
3101                          * part of the destination options header must appear
3102                          * before the routing header in the ancillary data,
3103                          * too.
3104                          * RFC3542 solved the ambiguity by introducing
3105                          * separate ancillary data or option types.
3106                          */
3107                         if (opt->ip6po_rthdr == NULL)
3108                                 newdest = &opt->ip6po_dest1;
3109                         else
3110                                 newdest = &opt->ip6po_dest2;
3111                         break;
3112                 case IPV6_RTHDRDSTOPTS:
3113                         newdest = &opt->ip6po_dest1;
3114                         break;
3115                 case IPV6_DSTOPTS:
3116                         newdest = &opt->ip6po_dest2;
3117                         break;
3118                 }
3119
3120                 /* turn off the previous option, then set the new option. */
3121                 ip6_clearpktopts(opt, optname);
3122                 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT);
3123                 if (*newdest == NULL)
3124                         return (ENOBUFS);
3125                 bcopy(dest, *newdest, destlen);
3126
3127                 break;
3128         }
3129
3130         case IPV6_2292RTHDR:
3131         case IPV6_RTHDR:
3132         {
3133                 struct ip6_rthdr *rth;
3134                 int rthlen;
3135
3136                 if (len == 0) {
3137                         ip6_clearpktopts(opt, IPV6_RTHDR);
3138                         break;  /* just remove the option */
3139                 }
3140
3141                 /* message length validation */
3142                 if (len < sizeof(struct ip6_rthdr))
3143                         return (EINVAL);
3144                 rth = (struct ip6_rthdr *)buf;
3145                 rthlen = (rth->ip6r_len + 1) << 3;
3146                 if (len != rthlen)
3147                         return (EINVAL);
3148
3149                 switch (rth->ip6r_type) {
3150                 default:
3151                         return (EINVAL);        /* not supported */
3152                 }
3153
3154                 /* turn off the previous option */
3155                 ip6_clearpktopts(opt, IPV6_RTHDR);
3156                 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT);
3157                 if (opt->ip6po_rthdr == NULL)
3158                         return (ENOBUFS);
3159                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3160
3161                 break;
3162         }
3163
3164         case IPV6_USE_MIN_MTU:
3165                 if (len != sizeof(int))
3166                         return (EINVAL);
3167                 minmtupolicy = *(int *)buf;
3168                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3169                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3170                     minmtupolicy != IP6PO_MINMTU_ALL) {
3171                         return (EINVAL);
3172                 }
3173                 opt->ip6po_minmtu = minmtupolicy;
3174                 break;
3175
3176         case IPV6_DONTFRAG:
3177                 if (len != sizeof(int))
3178                         return (EINVAL);
3179
3180                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3181                         /*
3182                          * we ignore this option for TCP sockets.
3183                          * (RFC3542 leaves this case unspecified.)
3184                          */
3185                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3186                 } else
3187                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3188                 break;
3189
3190         case IPV6_PREFER_TEMPADDR:
3191                 if (len != sizeof(int))
3192                         return (EINVAL);
3193                 preftemp = *(int *)buf;
3194                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3195                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3196                     preftemp != IP6PO_TEMPADDR_PREFER) {
3197                         return (EINVAL);
3198                 }
3199                 opt->ip6po_prefer_tempaddr = preftemp;
3200                 break;
3201
3202         default:
3203                 return (ENOPROTOOPT);
3204         } /* end of switch */
3205
3206         return (0);
3207 }
3208
3209
3210 /*
3211  * Set IPv6 outgoing packet options based on advanced API.
3212  */
3213 int
3214 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt,
3215     struct ip6_pktopts *stickyopt, int uproto, int priv)
3216 {
3217         struct cmsghdr *cm = NULL;
3218
3219         if (control == NULL || opt == NULL)
3220                 return (EINVAL);
3221
3222         init_ip6pktopts(opt);
3223
3224         /*
3225          * XXX: Currently, we assume all the optional information is stored
3226          * in a single mbuf.
3227          */
3228         if (stickyopt) {
3229                 int error;
3230
3231                 /*
3232                  * If stickyopt is provided, make a local copy of the options
3233                  * for this particular packet, then override them by ancillary
3234                  * objects.
3235                  * XXX: copypktopts() does not copy the cached route to a next
3236                  * hop (if any).  This is not very good in terms of efficiency,
3237                  * but we can allow this since this option should be rarely
3238                  * used.
3239                  */
3240                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
3241                         return (error);
3242         }
3243
3244         /*
3245          * XXX: Currently, we assume all the optional information is stored
3246          * in a single mbuf.
3247          */
3248         if (control->m_next)
3249                 return (EINVAL);
3250
3251         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3252             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3253                 int error;
3254
3255                 if (control->m_len < CMSG_LEN(0))
3256                         return (EINVAL);
3257
3258                 cm = mtod(control, struct cmsghdr *);
3259                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3260                         return (EINVAL);
3261                 if (cm->cmsg_level != IPPROTO_IPV6)
3262                         continue;
3263
3264                 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3265                     cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto, priv);
3266                 if (error)
3267                         return (error);
3268         }
3269
3270         return (0);
3271 }
3272
3273 /*
3274  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3275  * packet to the input queue of a specified interface.  Note that this
3276  * calls the output routine of the loopback "driver", but with an interface
3277  * pointer that might NOT be &loif -- easier than replicating that code here.
3278  */
3279 void
3280 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3281 {
3282         struct mbuf *copym;
3283         struct ip6_hdr *ip6;
3284
3285         copym = m_copy(m, 0, M_COPYALL);
3286         if (copym == NULL)
3287                 return;
3288
3289         /*
3290          * Make sure to deep-copy IPv6 header portion in case the data
3291          * is in an mbuf cluster, so that we can safely override the IPv6
3292          * header portion later.
3293          */
3294         if ((copym->m_flags & M_EXT) != 0 ||
3295             copym->m_len < sizeof(struct ip6_hdr)) {
3296                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3297                 if (copym == NULL)
3298                         return;
3299         }
3300
3301 #ifdef DIAGNOSTIC
3302         if (copym->m_len < sizeof(*ip6)) {
3303                 m_freem(copym);
3304                 return;
3305         }
3306 #endif
3307
3308         ip6 = mtod(copym, struct ip6_hdr *);
3309         /*
3310          * clear embedded scope identifiers if necessary.
3311          * in6_clearscope will touch the addresses only when necessary.
3312          */
3313         in6_clearscope(&ip6->ip6_src);
3314         in6_clearscope(&ip6->ip6_dst);
3315
3316         if_simloop(ifp, copym, dst->sin6_family, 0);
3317 }
3318
3319 /*
3320  * Separate the IPv6 header from the payload into its own mbuf.
3321  *
3322  * Returns the new mbuf chain or the original mbuf if no payload.
3323  * Returns NULL if can't allocate new mbuf for header.
3324  */
3325 static struct mbuf *
3326 ip6_splithdr(struct mbuf *m)
3327 {
3328         struct mbuf *mh;
3329
3330         if (m->m_len <= sizeof(struct ip6_hdr))         /* no payload */
3331                 return (m);
3332
3333         MGETHDR(mh, M_NOWAIT, MT_HEADER);
3334         if (mh == NULL)
3335                 return (NULL);
3336         mh->m_len = sizeof(struct ip6_hdr);
3337         M_MOVE_PKTHDR(mh, m);
3338         MH_ALIGN(mh, sizeof(struct ip6_hdr));
3339         bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr));
3340         m->m_data += sizeof(struct ip6_hdr);
3341         m->m_len -= sizeof(struct ip6_hdr);
3342         mh->m_next = m;
3343         return (mh);
3344 }
3345
3346 /*
3347  * Compute IPv6 extension header length.
3348  */
3349 int
3350 ip6_optlen(struct in6pcb *in6p)
3351 {
3352         int len;
3353
3354         if (!in6p->in6p_outputopts)
3355                 return 0;
3356
3357         len = 0;
3358 #define elen(x) \
3359     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3360
3361         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3362         if (in6p->in6p_outputopts->ip6po_rthdr)
3363                 /* dest1 is valid with rthdr only */
3364                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3365         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3366         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3367         return len;
3368 #undef elen
3369 }