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_INTFACELOCAL(&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;
 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
1311                 if (ifp == NULL)
1312                         ifp = ro_pmtu->ro_rt->rt_ifp;
1313                 ifmtu = IN6_LINKMTU(ifp);
1314                 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1315                 if (mtu == 0) {
1316                         mtu = ifmtu;
1317                 } else if (mtu < IPV6_MMTU) {
1318                         /*
1319                          * RFC2460 section 5, last paragraph:
1320                          * if we record ICMPv6 too big message with
1321                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1322                          * or smaller, with framgent header attached.
1323                          * (fragment header is needed regardless from the
1324                          * packet size, for translators to identify packets)
1325                          */
1326                         alwaysfrag = 1;
1327                         mtu = IPV6_MMTU;
1328                 } else if (mtu > ifmtu) {
1329                         /*
1330                          * The MTU on the route is larger than the MTU on
1331                          * the interface!  This shouldn't happen, unless the
1332                          * MTU of the interface has been changed after the
1333                          * interface was brought up.  Change the MTU in the
1334                          * route to match the interface MTU (as long as the
1335                          * field isn't locked).
1336                          */
1337                         mtu = ifmtu;
1338                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1339                 }
1340         } else if (ifp) {
1341                 mtu = IN6_LINKMTU(ifp);
1342         } else {
1343                 error = EHOSTUNREACH; /* XXX */
1344         }
1345
1346         *mtup = mtu;
1347         if (alwaysfragp)
1348                 *alwaysfragp = alwaysfrag;
1349         return (error);
1350 }
1351
1352 /*
1353  * IP6 socket option processing.
1354  */
1355 void
1356 ip6_ctloutput_dispatch(netmsg_t msg)
1357 {
1358         int error;
1359
1360         error = ip6_ctloutput(msg->ctloutput.base.nm_so,
1361                               msg->ctloutput.nm_sopt);
1362         lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
1363 }
1364
1365 int
1366 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1367 {
1368         int optdatalen,uproto;
1369         int privileged;
1370         struct inpcb *in6p = so->so_pcb;
1371         void *optdata;
1372         int error, optval;
1373         int level, op, optname;
1374         int optlen;
1375         struct thread *td;
1376
1377         if (sopt) {
1378                 level = sopt->sopt_level;
1379                 op = sopt->sopt_dir;
1380                 optname = sopt->sopt_name;
1381                 optlen = sopt->sopt_valsize;
1382                 td = sopt->sopt_td;
1383         } else {
1384                 panic("ip6_ctloutput: arg soopt is NULL");
1385                 /* NOT REACHED */
1386                 td = NULL;
1387         }
1388         error = optval = 0;
1389
1390         uproto = (int)so->so_proto->pr_protocol;
1391         privileged = (td == NULL || priv_check(td, PRIV_ROOT)) ? 0 : 1;
1392
1393         if (level == IPPROTO_IPV6) {
1394                 switch (op) {
1395
1396                 case SOPT_SET:
1397                         switch (optname) {
1398                         case IPV6_2292PKTOPTIONS:
1399 #ifdef IPV6_PKTOPTIONS
1400                         case IPV6_PKTOPTIONS:
1401 #endif
1402                         {
1403                                 struct mbuf *m;
1404
1405                                 error = soopt_getm(sopt, &m); /* XXX */
1406                                 if (error != 0)
1407                                         break;
1408                                 soopt_to_mbuf(sopt, m); /* XXX */
1409                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1410                                                     m, so, sopt);
1411                                 m_freem(m); /* XXX */
1412                                 break;
1413                         }
1414
1415                         /*
1416                          * Use of some Hop-by-Hop options or some
1417                          * Destination options, might require special
1418                          * privilege.  That is, normal applications
1419                          * (without special privilege) might be forbidden
1420                          * from setting certain options in outgoing packets,
1421                          * and might never see certain options in received
1422                          * packets. [RFC 2292 Section 6]
1423                          * KAME specific note:
1424                          *  KAME prevents non-privileged users from sending or
1425                          *  receiving ANY hbh/dst options in order to avoid
1426                          *  overhead of parsing options in the kernel.
1427                          */
1428                         case IPV6_RECVHOPOPTS:
1429                         case IPV6_RECVDSTOPTS:
1430                         case IPV6_RECVRTHDRDSTOPTS:
1431                                 if (!privileged)
1432                                         return (EPERM);
1433                         case IPV6_RECVPKTINFO:
1434                         case IPV6_RECVHOPLIMIT:
1435                         case IPV6_RECVRTHDR:
1436                         case IPV6_RECVPATHMTU:
1437                         case IPV6_RECVTCLASS:
1438                         case IPV6_AUTOFLOWLABEL:
1439                         case IPV6_HOPLIMIT:
1440                         /* FALLTHROUGH */
1441                         case IPV6_UNICAST_HOPS:
1442                         case IPV6_FAITH:
1443
1444                         case IPV6_V6ONLY:
1445                                 if (optlen != sizeof(int)) {
1446                                         error = EINVAL;
1447                                         break;
1448                                 }
1449                                 error = soopt_to_kbuf(sopt, &optval,
1450                                         sizeof optval, sizeof optval);
1451                                 if (error)
1452                                         break;
1453                                 switch (optname) {
1454
1455                                 case IPV6_UNICAST_HOPS:
1456                                         if (optval < -1 || optval >= 256)
1457                                                 error = EINVAL;
1458                                         else {
1459                                                 /* -1 = kernel default */
1460                                                 in6p->in6p_hops = optval;
1461                                         }
1462                                         break;
1463 #define OPTSET(bit) \
1464 do { \
1465         if (optval) \
1466                 in6p->in6p_flags |= (bit); \
1467         else \
1468                 in6p->in6p_flags &= ~(bit); \
1469 } while (0)
1470 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1471 /*
1472  * Although changed to RFC3542, It's better to also support RFC2292 API
1473  */
1474 #define OPTSET2292(bit) \
1475 do { \
1476         in6p->in6p_flags |= IN6P_RFC2292; \
1477         if (optval) \
1478                 in6p->in6p_flags |= (bit); \
1479         else \
1480                 in6p->in6p_flags &= ~(bit); \
1481 } while (/*CONSTCOND*/ 0)
1482
1483                                 case IPV6_RECVPKTINFO:
1484                                         /* cannot mix with RFC2292 */
1485                                         if (OPTBIT(IN6P_RFC2292)) {
1486                                                 error = EINVAL;
1487                                                 break;
1488                                         }
1489                                         OPTSET(IN6P_PKTINFO);
1490                                         break;
1491
1492                                 case IPV6_HOPLIMIT:
1493                                 {
1494                                         struct ip6_pktopts **optp;
1495
1496                                         /* cannot mix with RFC2292 */
1497                                         if (OPTBIT(IN6P_RFC2292)) {
1498                                                 error = EINVAL;
1499                                                 break;
1500                                         }
1501                                         optp = &in6p->in6p_outputopts;
1502                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1503                                             (u_char *)&optval, sizeof(optval),
1504                                             optp, uproto);
1505                                         break;
1506                                 }
1507
1508                                 case IPV6_RECVHOPLIMIT:
1509                                         /* cannot mix with RFC2292 */
1510                                         if (OPTBIT(IN6P_RFC2292)) {
1511                                                 error = EINVAL;
1512                                                 break;
1513                                         }
1514                                         OPTSET(IN6P_HOPLIMIT);
1515                                         break;
1516
1517                                 case IPV6_RECVHOPOPTS:
1518                                         /* cannot mix with RFC2292 */
1519                                         if (OPTBIT(IN6P_RFC2292)) {
1520                                                 error = EINVAL;
1521                                                 break;
1522                                         }
1523                                         OPTSET(IN6P_HOPOPTS);
1524                                         break;
1525
1526                                 case IPV6_RECVDSTOPTS:
1527                                         /* cannot mix with RFC2292 */
1528                                         if (OPTBIT(IN6P_RFC2292)) {
1529                                                 error = EINVAL;
1530                                                 break;
1531                                         }
1532                                         OPTSET(IN6P_DSTOPTS);
1533                                         break;
1534
1535                                 case IPV6_RECVRTHDRDSTOPTS:
1536                                         /* cannot mix with RFC2292 */
1537                                         if (OPTBIT(IN6P_RFC2292)) {
1538                                                 error = EINVAL;
1539                                                 break;
1540                                         }
1541                                         OPTSET(IN6P_RTHDRDSTOPTS);
1542                                         break;
1543
1544                                 case IPV6_RECVRTHDR:
1545                                         /* cannot mix with RFC2292 */
1546                                         if (OPTBIT(IN6P_RFC2292)) {
1547                                                 error = EINVAL;
1548                                                 break;
1549                                         }
1550                                         OPTSET(IN6P_RTHDR);
1551                                         break;
1552
1553                                 case IPV6_RECVPATHMTU:
1554                                         /*
1555                                          * We ignore this option for TCP
1556                                          * sockets.
1557                                          * (RFC3542 leaves this case
1558                                          * unspecified.)
1559                                          */
1560                                         if (uproto != IPPROTO_TCP)
1561                                                 OPTSET(IN6P_MTU);
1562                                         break;
1563
1564                                 case IPV6_RECVTCLASS:
1565                                         /* cannot mix with RFC2292 XXX */
1566                                         if (OPTBIT(IN6P_RFC2292)) {
1567                                                 error = EINVAL;
1568                                                 break;
1569                                         }
1570                                         OPTSET(IN6P_TCLASS);
1571                                         break;
1572
1573                                 case IPV6_AUTOFLOWLABEL:
1574                                         OPTSET(IN6P_AUTOFLOWLABEL);
1575                                         break;
1576
1577                                 case IPV6_FAITH:
1578                                         OPTSET(IN6P_FAITH);
1579                                         break;
1580
1581                                 case IPV6_V6ONLY:
1582                                         /*
1583                                          * make setsockopt(IPV6_V6ONLY)
1584                                          * available only prior to bind(2).
1585                                          */
1586                                         if (in6p->in6p_lport ||
1587                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1588                                         {
1589                                                 error = EINVAL;
1590                                                 break;
1591                                         }
1592                                         if (!optval) {
1593                                                 /* Don't allow v4-mapped */
1594                                                 error = EOPNOTSUPP;
1595                                         }
1596                                         break;
1597                                 }
1598                                 break;
1599
1600                         case IPV6_TCLASS:
1601                         case IPV6_DONTFRAG:
1602                         case IPV6_USE_MIN_MTU:
1603                         case IPV6_PREFER_TEMPADDR:
1604                                 if (optlen != sizeof(optval)) {
1605                                         error = EINVAL;
1606                                         break;
1607                                 }
1608                                 error = soopt_to_kbuf(sopt, &optval,
1609                                         sizeof optval, sizeof optval);
1610                                 if (error)
1611                                         break;
1612                                 {
1613                                         struct ip6_pktopts **optp;
1614                                         optp = &in6p->in6p_outputopts;
1615                                         error = ip6_pcbopt(optname,
1616                                             (u_char *)&optval, sizeof(optval),
1617                                             optp, uproto);
1618                                         break;
1619                                 }
1620
1621                         case IPV6_2292PKTINFO:
1622                         case IPV6_2292HOPLIMIT:
1623                         case IPV6_2292HOPOPTS:
1624                         case IPV6_2292DSTOPTS:
1625                         case IPV6_2292RTHDR:
1626                                 /* RFC 2292 */
1627                                 if (optlen != sizeof(int)) {
1628                                         error = EINVAL;
1629                                         break;
1630                                 }
1631                                 error = soopt_to_kbuf(sopt, &optval,
1632                                         sizeof optval, sizeof optval);
1633                                 if (error)
1634                                         break;
1635                                 switch (optname) {
1636                                 case IPV6_2292PKTINFO:
1637                                         OPTSET2292(IN6P_PKTINFO);
1638                                         break;
1639                                 case IPV6_2292HOPLIMIT:
1640                                         OPTSET2292(IN6P_HOPLIMIT);
1641                                         break;
1642                                 case IPV6_2292HOPOPTS:
1643                                         /*
1644                                          * Check super-user privilege.
1645                                          * See comments for IPV6_RECVHOPOPTS.
1646                                          */
1647                                         if (!privileged)
1648                                                 return (EPERM);
1649                                         OPTSET2292(IN6P_HOPOPTS);
1650                                         break;
1651                                 case IPV6_2292DSTOPTS:
1652                                         if (!privileged)
1653                                                 return (EPERM);
1654                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1655                                         break;
1656                                 case IPV6_2292RTHDR:
1657                                         OPTSET2292(IN6P_RTHDR);
1658                                         break;
1659                                 }
1660                                 break;
1661
1662                         case IPV6_PKTINFO:
1663                         case IPV6_HOPOPTS:
1664                         case IPV6_RTHDR:
1665                         case IPV6_DSTOPTS:
1666                         case IPV6_RTHDRDSTOPTS:
1667                         case IPV6_NEXTHOP:
1668                         {
1669                                 /*
1670                                  * New advanced API (RFC3542)
1671                                  */
1672                                 u_char *optbuf;
1673                                 u_char optbuf_storage[MCLBYTES];
1674                                 int optlen;
1675                                 struct ip6_pktopts **optp;
1676
1677                                 /* cannot mix with RFC2292 */
1678                                 if (OPTBIT(IN6P_RFC2292)) {
1679                                         error = EINVAL;
1680                                         break;
1681                                 }
1682
1683                                 /*
1684                                  * We only ensure valsize is not too large
1685                                  * here.  Further validation will be done
1686                                  * later.
1687                                  */
1688                                 error = soopt_to_kbuf(sopt, optbuf_storage,
1689                                     sizeof(optbuf_storage), 0);
1690                                 if (error)
1691                                         break;
1692                                 optlen = sopt->sopt_valsize;
1693                                 optbuf = optbuf_storage;
1694                                 optp = &in6p->in6p_outputopts;
1695                                 error = ip6_pcbopt(optname, optbuf, optlen,
1696                                     optp, uproto);
1697                                 break;
1698                         }
1699 #undef OPTSET
1700
1701                         case IPV6_MULTICAST_IF:
1702                         case IPV6_MULTICAST_HOPS:
1703                         case IPV6_MULTICAST_LOOP:
1704                         case IPV6_JOIN_GROUP:
1705                         case IPV6_LEAVE_GROUP:
1706                             {
1707                                 struct mbuf *m;
1708
1709                                 if (sopt->sopt_valsize > MLEN) {
1710                                         error = EMSGSIZE;
1711                                         break;
1712                                 }
1713                                 /* XXX */
1714                                 MGET(m, sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_HEADER);
1715                                 if (m == NULL) {
1716                                         error = ENOBUFS;
1717                                         break;
1718                                 }
1719                                 m->m_len = sopt->sopt_valsize;
1720                                 error = soopt_to_kbuf(sopt, mtod(m, char *),
1721                                                     m->m_len, m->m_len);
1722                                 error = ip6_setmoptions(sopt->sopt_name,
1723                                                         &in6p->in6p_moptions,
1724                                                         m);
1725                                 m_free(m);
1726                             }
1727                                 break;
1728
1729                         case IPV6_PORTRANGE:
1730                                 error = soopt_to_kbuf(sopt, &optval,
1731                                     sizeof optval, sizeof optval);
1732                                 if (error)
1733                                         break;
1734
1735                                 switch (optval) {
1736                                 case IPV6_PORTRANGE_DEFAULT:
1737                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1738                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1739                                         break;
1740
1741                                 case IPV6_PORTRANGE_HIGH:
1742                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1743                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1744                                         break;
1745
1746                                 case IPV6_PORTRANGE_LOW:
1747                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1748                                         in6p->in6p_flags |= IN6P_LOWPORT;
1749                                         break;
1750
1751                                 default:
1752                                         error = EINVAL;
1753                                         break;
1754                                 }
1755                                 break;
1756
1757 #if defined(IPSEC) || defined(FAST_IPSEC)
1758                         case IPV6_IPSEC_POLICY:
1759                             {
1760                                 caddr_t req = NULL;
1761                                 size_t len = 0;
1762                                 struct mbuf *m;
1763
1764                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1765                                         break;
1766                                 soopt_to_mbuf(sopt, m);         /* XXX */
1767                                 if (m) {
1768                                         req = mtod(m, caddr_t);
1769                                         len = m->m_len;
1770                                 }
1771                                 error = ipsec6_set_policy(in6p, optname, req,
1772                                                           len, privileged);
1773                                 m_freem(m);
1774                             }
1775                                 break;
1776 #endif /* KAME IPSEC */
1777
1778                         case IPV6_FW_ADD:
1779                         case IPV6_FW_DEL:
1780                         case IPV6_FW_FLUSH:
1781                         case IPV6_FW_ZERO:
1782                             {
1783                                 struct mbuf *m;
1784                                 struct mbuf **mp = &m;
1785
1786                                 if (ip6_fw_ctl_ptr == NULL)
1787                                         return EINVAL;
1788                                 /* XXX */
1789                                 if ((error = soopt_getm(sopt, &m)) != 0)
1790                                         break;
1791                                 /* XXX */
1792                                 soopt_to_mbuf(sopt, m);
1793                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1794                                 m = *mp;
1795                             }
1796                                 break;
1797
1798                         default:
1799                                 error = ENOPROTOOPT;
1800                                 break;
1801                         }
1802                         break;
1803
1804                 case SOPT_GET:
1805                         switch (optname) {
1806                         case IPV6_2292PKTOPTIONS:
1807 #ifdef IPV6_PKTOPTIONS
1808                         case IPV6_PKTOPTIONS:
1809 #endif
1810                                 /*
1811                                  * RFC3542 (effectively) deprecated the
1812                                  * semantics of the 2292-style pktoptions.
1813                                  * Since it was not reliable in nature (i.e.,
1814                                  * applications had to expect the lack of some
1815                                  * information after all), it would make sense
1816                                  * to simplify this part by always returning
1817                                  * empty data.
1818                                  */
1819                                 if (in6p->in6p_options) {
1820                                         struct mbuf *m;
1821                                         m = m_copym(in6p->in6p_options,
1822                                             0, M_COPYALL, M_WAITOK);
1823                                         error = soopt_from_mbuf(sopt, m);
1824                                         if (error == 0)
1825                                                 m_freem(m);
1826                                 } else
1827                                         sopt->sopt_valsize = 0;
1828                                 break;
1829
1830                         case IPV6_RECVHOPOPTS:
1831                         case IPV6_RECVDSTOPTS:
1832                         case IPV6_RECVRTHDRDSTOPTS:
1833                         case IPV6_UNICAST_HOPS:
1834                         case IPV6_RECVPKTINFO:
1835                         case IPV6_RECVHOPLIMIT:
1836                         case IPV6_RECVRTHDR:
1837                         case IPV6_RECVPATHMTU:
1838                         case IPV6_RECVTCLASS:
1839                         case IPV6_AUTOFLOWLABEL:
1840                         case IPV6_FAITH:
1841                         case IPV6_V6ONLY:
1842                         case IPV6_PORTRANGE:
1843                                 switch (optname) {
1844
1845                                 case IPV6_RECVHOPOPTS:
1846                                         optval = OPTBIT(IN6P_HOPOPTS);
1847                                         break;
1848
1849                                 case IPV6_RECVDSTOPTS:
1850                                         optval = OPTBIT(IN6P_DSTOPTS);
1851                                         break;
1852
1853                                 case IPV6_RECVRTHDRDSTOPTS:
1854                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1855                                         break;
1856
1857                                 case IPV6_RECVPKTINFO:
1858                                         optval = OPTBIT(IN6P_PKTINFO);
1859                                         break;
1860
1861                                 case IPV6_RECVHOPLIMIT:
1862                                         optval = OPTBIT(IN6P_HOPLIMIT);
1863                                         break;
1864
1865                                 case IPV6_RECVRTHDR:
1866                                         optval = OPTBIT(IN6P_RTHDR);
1867                                         break;
1868
1869                                 case IPV6_RECVPATHMTU:
1870                                         optval = OPTBIT(IN6P_MTU);
1871                                         break;
1872
1873                                 case IPV6_RECVTCLASS:
1874                                         optval = OPTBIT(IN6P_TCLASS);
1875                                         break;
1876
1877                                 case IPV6_AUTOFLOWLABEL:
1878                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1879                                         break;
1880
1881
1882                                 case IPV6_UNICAST_HOPS:
1883                                         optval = in6p->in6p_hops;
1884                                         break;
1885
1886                                 case IPV6_FAITH:
1887                                         optval = OPTBIT(IN6P_FAITH);
1888                                         break;
1889
1890                                 case IPV6_V6ONLY:
1891                                         optval = 1;
1892                                         break;
1893
1894                                 case IPV6_PORTRANGE:
1895                                     {
1896                                         int flags;
1897                                         flags = in6p->in6p_flags;
1898                                         if (flags & IN6P_HIGHPORT)
1899                                                 optval = IPV6_PORTRANGE_HIGH;
1900                                         else if (flags & IN6P_LOWPORT)
1901                                                 optval = IPV6_PORTRANGE_LOW;
1902                                         else
1903                                                 optval = 0;
1904                                         break;
1905                                     }
1906                                 }
1907                                 soopt_from_kbuf(sopt, &optval,
1908                                         sizeof optval);
1909                                 break;
1910
1911                         case IPV6_PATHMTU:
1912                         {
1913                                 u_long pmtu = 0;
1914                                 struct ip6_mtuinfo mtuinfo;
1915                                 struct route_in6 sro;
1916
1917                                 bzero(&sro, sizeof(sro));
1918
1919                                 if (!(so->so_state & SS_ISCONNECTED))
1920                                         return (ENOTCONN);
1921                                 /*
1922                                  * XXX: we dot not consider the case of source
1923                                  * routing, or optional information to specify
1924                                  * the outgoing interface.
1925                                  */
1926                                 error = ip6_getpmtu(&sro, NULL, NULL,
1927                                     &in6p->in6p_faddr, &pmtu, NULL);
1928                                 if (sro.ro_rt)
1929                                         RTFREE(sro.ro_rt);
1930                                 if (error)
1931                                         break;
1932                                 if (pmtu > IPV6_MAXPACKET)
1933                                         pmtu = IPV6_MAXPACKET;
1934
1935                                 bzero(&mtuinfo, sizeof(mtuinfo));
1936                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1937                                 optdata = (void *)&mtuinfo;
1938                                 optdatalen = sizeof(mtuinfo);
1939                                 soopt_from_kbuf(sopt, optdata,
1940                                     optdatalen);
1941                                 break;
1942                         }
1943
1944                         case IPV6_2292PKTINFO:
1945                         case IPV6_2292HOPLIMIT:
1946                         case IPV6_2292HOPOPTS:
1947                         case IPV6_2292RTHDR:
1948                         case IPV6_2292DSTOPTS:
1949                                 if (optname == IPV6_2292HOPOPTS ||
1950                                     optname == IPV6_2292DSTOPTS ||
1951                                     !privileged)
1952                                         return (EPERM);
1953                                 switch (optname) {
1954                                 case IPV6_2292PKTINFO:
1955                                         optval = OPTBIT(IN6P_PKTINFO);
1956                                         break;
1957                                 case IPV6_2292HOPLIMIT:
1958                                         optval = OPTBIT(IN6P_HOPLIMIT);
1959                                         break;
1960                                 case IPV6_2292HOPOPTS:
1961                                         if (!privileged)
1962                                                 return (EPERM);
1963                                         optval = OPTBIT(IN6P_HOPOPTS);
1964                                         break;
1965                                 case IPV6_2292RTHDR:
1966                                         optval = OPTBIT(IN6P_RTHDR);
1967                                         break;
1968                                 case IPV6_2292DSTOPTS:
1969                                         if (!privileged)
1970                                                 return (EPERM);
1971                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1972                                         break;
1973                                 }
1974                                 soopt_from_kbuf(sopt, &optval,
1975                                         sizeof optval);
1976                                 break;
1977
1978                         case IPV6_PKTINFO:
1979                         case IPV6_HOPOPTS:
1980                         case IPV6_RTHDR:
1981                         case IPV6_DSTOPTS:
1982                         case IPV6_RTHDRDSTOPTS:
1983                         case IPV6_NEXTHOP:
1984                         case IPV6_TCLASS:
1985                         case IPV6_DONTFRAG:
1986                         case IPV6_USE_MIN_MTU:
1987                         case IPV6_PREFER_TEMPADDR:
1988                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
1989                                     optname, sopt);
1990                                 break;
1991
1992                         case IPV6_MULTICAST_IF:
1993                         case IPV6_MULTICAST_HOPS:
1994                         case IPV6_MULTICAST_LOOP:
1995                         case IPV6_JOIN_GROUP:
1996                         case IPV6_LEAVE_GROUP:
1997                             {
1998                                 struct mbuf *m;
1999                                 error = ip6_getmoptions(sopt->sopt_name,
2000                                     in6p->in6p_moptions, &m);
2001                                 if (error == 0) {
2002                                         soopt_from_kbuf(sopt,
2003                                             mtod(m, char *), m->m_len);
2004                                 }
2005                                 m_freem(m);
2006                             }
2007                                 break;
2008
2009 #if defined(IPSEC) || defined(FAST_IPSEC)
2010                         case IPV6_IPSEC_POLICY:
2011                           {
2012                                 caddr_t req = NULL;
2013                                 size_t len = 0;
2014                                 struct mbuf *m = NULL;
2015                                 struct mbuf **mp = &m;
2016
2017                                 error = soopt_getm(sopt, &m); /* XXX */
2018                                 if (error != 0)
2019                                         break;
2020                                 soopt_to_mbuf(sopt, m); /* XXX */
2021                                 if (m) {
2022                                         req = mtod(m, caddr_t);
2023                                         len = m->m_len;
2024                                 }
2025                                 error = ipsec6_get_policy(in6p, req, len, mp);
2026                                 if (error == 0)
2027                                         error = soopt_from_mbuf(sopt, m);/*XXX*/
2028                                 if (error == 0 && m != NULL)
2029                                         m_freem(m);
2030                                 break;
2031                           }
2032 #endif /* KAME IPSEC */
2033
2034                         case IPV6_FW_GET:
2035                           {
2036                                 struct mbuf *m;
2037                                 struct mbuf **mp = &m;
2038
2039                                 if (ip6_fw_ctl_ptr == NULL)
2040                                 {
2041                                         return EINVAL;
2042                                 }
2043                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
2044                                 if (error == 0)
2045                                         error = soopt_from_mbuf(sopt, m); /* XXX */
2046                                 if (error == 0 && m != NULL)
2047                                         m_freem(m);
2048                           }
2049                                 break;
2050
2051                         default:
2052                                 error = ENOPROTOOPT;
2053                                 break;
2054                         }
2055                         break;
2056                 }
2057         } else {
2058                 error = EINVAL;
2059         }
2060         return (error);
2061 }
2062
2063 int
2064 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2065 {
2066         int error = 0, optval, optlen;
2067         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2068         struct in6pcb *in6p = sotoin6pcb(so);
2069         int level, op, optname;
2070
2071         if (sopt) {
2072                 level = sopt->sopt_level;
2073                 op = sopt->sopt_dir;
2074                 optname = sopt->sopt_name;
2075                 optlen = sopt->sopt_valsize;
2076         } else
2077                 panic("ip6_raw_ctloutput: arg soopt is NULL");
2078
2079         if (level != IPPROTO_IPV6) {
2080                 return (EINVAL);
2081         }
2082
2083         switch (optname) {
2084         case IPV6_CHECKSUM:
2085                 /*
2086                  * For ICMPv6 sockets, no modification allowed for checksum
2087                  * offset, permit "no change" values to help existing apps.
2088                  *
2089                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2090                  * for an ICMPv6 socket will fail."
2091                  * The current behavior does not meet RFC3542.
2092                  */
2093                 switch (op) {
2094                 case SOPT_SET:
2095                         if (optlen != sizeof(int)) {
2096                                 error = EINVAL;
2097                                 break;
2098                         }
2099                         error = soopt_to_kbuf(sopt, &optval,
2100                                     sizeof optval, sizeof optval);
2101                         if (error)
2102                                 break;
2103                         if ((optval % 2) != 0) {
2104                                 /* the API assumes even offset values */
2105                                 error = EINVAL;
2106                         } else if (so->so_proto->pr_protocol ==
2107                             IPPROTO_ICMPV6) {
2108                                 if (optval != icmp6off)
2109                                         error = EINVAL;
2110                         } else
2111                                 in6p->in6p_cksum = optval;
2112                         break;
2113
2114                 case SOPT_GET:
2115                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2116                                 optval = icmp6off;
2117                         else
2118                                 optval = in6p->in6p_cksum;
2119
2120                         soopt_from_kbuf(sopt, &optval, sizeof(optval));
2121                         break;
2122
2123                 default:
2124                         error = EINVAL;
2125                         break;
2126                 }
2127                 break;
2128
2129         default:
2130                 error = ENOPROTOOPT;
2131                 break;
2132         }
2133
2134         return (error);
2135 }
2136
2137 /*
2138  * Set up IP6 options in pcb for insertion in output packets or
2139  * specifying behavior of outgoing packets.
2140  */
2141 static int
2142 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2143     struct socket *so, struct sockopt *sopt)
2144 {
2145         int priv = 0;
2146         struct ip6_pktopts *opt = *pktopt;
2147         int error = 0;
2148
2149         /* turn off any old options. */
2150         if (opt) {
2151 #ifdef DIAGNOSTIC
2152                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2153                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2154                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2155                         kprintf("ip6_pcbopts: all specified options are cleared.\n");
2156 #endif
2157                 ip6_clearpktopts(opt, -1);
2158         } else
2159                 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2160         *pktopt = NULL;
2161
2162         if (!m || m->m_len == 0) {
2163                 /*
2164                  * Only turning off any previous options, regardless of
2165                  * whether the opt is just created or given.
2166                  */
2167                 kfree(opt, M_IP6OPT);
2168                 return (0);
2169         }
2170
2171         /*  set options specified by user. */
2172         if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) {
2173                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2174                 kfree(opt, M_IP6OPT);
2175                 return (error);
2176         }
2177         *pktopt = opt;
2178         return (0);
2179 }
2180
2181
2182 /*
2183  * Below three functions are introduced by merge to RFC3542
2184  */
2185
2186 static int
2187 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2188 {
2189         void *optdata = NULL;
2190         int optdatalen = 0;
2191         struct ip6_ext *ip6e;
2192         int error = 0;
2193         struct in6_pktinfo null_pktinfo;
2194         int deftclass = 0, on;
2195         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2196         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2197
2198         switch (optname) {
2199         case IPV6_PKTINFO:
2200                 if (pktopt && pktopt->ip6po_pktinfo)
2201                         optdata = (void *)pktopt->ip6po_pktinfo;
2202                 else {
2203                         /* XXX: we don't have to do this every time... */
2204                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2205                         optdata = (void *)&null_pktinfo;
2206                 }
2207                 optdatalen = sizeof(struct in6_pktinfo);
2208                 break;
2209         case IPV6_TCLASS:
2210                 if (pktopt && pktopt->ip6po_tclass >= 0)
2211                         optdata = (void *)&pktopt->ip6po_tclass;
2212                 else
2213                         optdata = (void *)&deftclass;
2214                 optdatalen = sizeof(int);
2215                 break;
2216         case IPV6_HOPOPTS:
2217                 if (pktopt && pktopt->ip6po_hbh) {
2218                         optdata = (void *)pktopt->ip6po_hbh;
2219                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2220                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2221                 }
2222                 break;
2223         case IPV6_RTHDR:
2224                 if (pktopt && pktopt->ip6po_rthdr) {
2225                         optdata = (void *)pktopt->ip6po_rthdr;
2226                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2227                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2228                 }
2229                 break;
2230         case IPV6_RTHDRDSTOPTS:
2231                 if (pktopt && pktopt->ip6po_dest1) {
2232                         optdata = (void *)pktopt->ip6po_dest1;
2233                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2234                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2235                 }
2236                 break;
2237         case IPV6_DSTOPTS:
2238                 if (pktopt && pktopt->ip6po_dest2) {
2239                         optdata = (void *)pktopt->ip6po_dest2;
2240                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2241                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2242                 }
2243                 break;
2244         case IPV6_NEXTHOP:
2245                 if (pktopt && pktopt->ip6po_nexthop) {
2246                         optdata = (void *)pktopt->ip6po_nexthop;
2247                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2248                 }
2249                 break;
2250         case IPV6_USE_MIN_MTU:
2251                 if (pktopt)
2252                         optdata = (void *)&pktopt->ip6po_minmtu;
2253                 else
2254                         optdata = (void *)&defminmtu;
2255                 optdatalen = sizeof(int);
2256                 break;
2257         case IPV6_DONTFRAG:
2258                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2259                         on = 1;
2260                 else
2261                         on = 0;
2262                 optdata = (void *)&on;
2263                 optdatalen = sizeof(on);
2264                 break;
2265         case IPV6_PREFER_TEMPADDR:
2266                 if (pktopt)
2267                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2268                 else
2269                         optdata = (void *)&defpreftemp;
2270                 optdatalen = sizeof(int);
2271                 break;
2272         default:                /* should not happen */
2273 #ifdef DIAGNOSTIC
2274                 panic("ip6_getpcbopt: unexpected option");
2275 #endif
2276                 return (ENOPROTOOPT);
2277         }
2278
2279         soopt_from_kbuf(sopt, optdata, optdatalen);
2280
2281         return (error);
2282 }
2283
2284 /*
2285  * initialize ip6_pktopts.  beware that there are non-zero default values in
2286  * the struct.
2287  */
2288
2289 static int
2290 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto)
2291 {
2292         struct ip6_pktopts *opt;
2293         int priv =0;
2294         if (*pktopt == NULL) {
2295                 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2296                 init_ip6pktopts(*pktopt);
2297         }
2298         opt = *pktopt;
2299
2300         return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv));
2301 }
2302
2303 /*
2304  * initialize ip6_pktopts.  beware that there are non-zero default values in
2305  * the struct.
2306  */
2307 void
2308 init_ip6pktopts(struct ip6_pktopts *opt)
2309 {
2310
2311         bzero(opt, sizeof(*opt));
2312         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2313         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2314         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2315         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2316 }
2317
2318 void
2319 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2320 {
2321         if (pktopt == NULL)
2322                 return;
2323
2324         if (optname == -1 || optname == IPV6_PKTINFO) {
2325                 if (pktopt->ip6po_pktinfo)
2326                         kfree(pktopt->ip6po_pktinfo, M_IP6OPT);
2327                 pktopt->ip6po_pktinfo = NULL;
2328         }
2329         if (optname == -1 || optname == IPV6_HOPLIMIT)
2330                 pktopt->ip6po_hlim = -1;
2331         if (optname == -1 || optname == IPV6_TCLASS)
2332                 pktopt->ip6po_tclass = -1;
2333         if (optname == -1 || optname == IPV6_NEXTHOP) {
2334                 if (pktopt->ip6po_nextroute.ro_rt) {
2335                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2336                         pktopt->ip6po_nextroute.ro_rt = NULL;
2337                 }
2338                 if (pktopt->ip6po_nexthop)
2339                         kfree(pktopt->ip6po_nexthop, M_IP6OPT);
2340                 pktopt->ip6po_nexthop = NULL;
2341         }
2342         if (optname == -1 || optname == IPV6_HOPOPTS) {
2343                 if (pktopt->ip6po_hbh)
2344                         kfree(pktopt->ip6po_hbh, M_IP6OPT);
2345                 pktopt->ip6po_hbh = NULL;
2346         }
2347         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2348                 if (pktopt->ip6po_dest1)
2349                         kfree(pktopt->ip6po_dest1, M_IP6OPT);
2350                 pktopt->ip6po_dest1 = NULL;
2351         }
2352         if (optname == -1 || optname == IPV6_RTHDR) {
2353                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2354                         kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2355                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2356                 if (pktopt->ip6po_route.ro_rt) {
2357                         RTFREE(pktopt->ip6po_route.ro_rt);
2358                         pktopt->ip6po_route.ro_rt = NULL;
2359                 }
2360         }
2361         if (optname == -1 || optname == IPV6_DSTOPTS) {
2362                 if (pktopt->ip6po_dest2)
2363                         kfree(pktopt->ip6po_dest2, M_IP6OPT);
2364                 pktopt->ip6po_dest2 = NULL;
2365         }
2366 }
2367
2368 #define PKTOPT_EXTHDRCPY(type) \
2369 do {\
2370         if (src->type) {\
2371                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2372                 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\
2373                 if (dst->type == NULL)\
2374                         goto bad;\
2375                 bcopy(src->type, dst->type, hlen);\
2376         }\
2377 } while (0)
2378
2379 struct ip6_pktopts *
2380 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2381 {
2382         struct ip6_pktopts *dst;
2383
2384         if (src == NULL) {
2385                 kprintf("ip6_clearpktopts: invalid argument\n");
2386                 return (NULL);
2387         }
2388
2389         dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO);
2390         if (dst == NULL)
2391                 return (NULL);
2392
2393         dst->ip6po_hlim = src->ip6po_hlim;
2394         if (src->ip6po_pktinfo) {
2395                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2396                     M_IP6OPT, canwait);
2397                 if (dst->ip6po_pktinfo == NULL)
2398                         goto bad;
2399                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2400         }
2401         if (src->ip6po_nexthop) {
2402                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2403                     M_IP6OPT, canwait);
2404                 if (dst->ip6po_nexthop == NULL)
2405                         goto bad;
2406                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2407                     src->ip6po_nexthop->sa_len);
2408         }
2409         PKTOPT_EXTHDRCPY(ip6po_hbh);
2410         PKTOPT_EXTHDRCPY(ip6po_dest1);
2411         PKTOPT_EXTHDRCPY(ip6po_dest2);
2412         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2413         return (dst);
2414
2415 bad:
2416         if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT);
2417         if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT);
2418         if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT);
2419         if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT);
2420         if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT);
2421         if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT);
2422         kfree(dst, M_IP6OPT);
2423         return (NULL);
2424 }
2425
2426 static int
2427 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2428 {
2429         if (dst == NULL || src == NULL)  {
2430 #ifdef DIAGNOSTIC
2431                 kprintf("ip6_clearpktopts: invalid argument\n");
2432 #endif
2433                 return (EINVAL);
2434         }
2435
2436         dst->ip6po_hlim = src->ip6po_hlim;
2437         dst->ip6po_tclass = src->ip6po_tclass;
2438         dst->ip6po_flags = src->ip6po_flags;
2439         if (src->ip6po_pktinfo) {
2440                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2441                     M_IP6OPT, canwait);
2442                 if (dst->ip6po_pktinfo == NULL)
2443                         goto bad;
2444                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2445         }
2446         if (src->ip6po_nexthop) {
2447                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2448                     M_IP6OPT, canwait);
2449                 if (dst->ip6po_nexthop == NULL)
2450                         goto bad;
2451                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2452                     src->ip6po_nexthop->sa_len);
2453         }
2454         PKTOPT_EXTHDRCPY(ip6po_hbh);
2455         PKTOPT_EXTHDRCPY(ip6po_dest1);
2456         PKTOPT_EXTHDRCPY(ip6po_dest2);
2457         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2458         return (0);
2459
2460   bad:
2461         ip6_clearpktopts(dst, -1);
2462         return (ENOBUFS);
2463 }
2464 #undef PKTOPT_EXTHDRCPY
2465
2466 void
2467 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2468 {
2469         if (pktopt == NULL)
2470                 return;
2471
2472         ip6_clearpktopts(pktopt, -1);
2473
2474         kfree(pktopt, M_IP6OPT);
2475 }
2476
2477 /*
2478  * Set the IP6 multicast options in response to user setsockopt().
2479  */
2480 static int
2481 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2482 {
2483         int error = 0;
2484         u_int loop, ifindex;
2485         struct ipv6_mreq *mreq;
2486         struct ifnet *ifp;
2487         struct ip6_moptions *im6o = *im6op;
2488         struct route_in6 ro;
2489         struct sockaddr_in6 *dst;
2490         struct in6_multi_mship *imm;
2491         struct thread *td = curthread;
2492
2493         if (im6o == NULL) {
2494                 /*
2495                  * No multicast option buffer attached to the pcb;
2496                  * allocate one and initialize to default values.
2497                  */
2498                 im6o = (struct ip6_moptions *)
2499                         kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2500
2501                 *im6op = im6o;
2502                 im6o->im6o_multicast_ifp = NULL;
2503                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2504                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2505                 LIST_INIT(&im6o->im6o_memberships);
2506         }
2507
2508         switch (optname) {
2509
2510         case IPV6_MULTICAST_IF:
2511                 /*
2512                  * Select the interface for outgoing multicast packets.
2513                  */
2514                 if (m == NULL || m->m_len != sizeof(u_int)) {
2515                         error = EINVAL;
2516                         break;
2517                 }
2518                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2519                 if (ifindex < 0 || if_index < ifindex) {
2520                         error = ENXIO;  /* XXX EINVAL? */
2521                         break;
2522                 }
2523                 ifp = ifindex2ifnet[ifindex];
2524                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2525                         error = EADDRNOTAVAIL;
2526                         break;
2527                 }
2528                 im6o->im6o_multicast_ifp = ifp;
2529                 break;
2530
2531         case IPV6_MULTICAST_HOPS:
2532             {
2533                 /*
2534                  * Set the IP6 hoplimit for outgoing multicast packets.
2535                  */
2536                 int optval;
2537                 if (m == NULL || m->m_len != sizeof(int)) {
2538                         error = EINVAL;
2539                         break;
2540                 }
2541                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2542                 if (optval < -1 || optval >= 256)
2543                         error = EINVAL;
2544                 else if (optval == -1)
2545                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2546                 else
2547                         im6o->im6o_multicast_hlim = optval;
2548                 break;
2549             }
2550
2551         case IPV6_MULTICAST_LOOP:
2552                 /*
2553                  * Set the loopback flag for outgoing multicast packets.
2554                  * Must be zero or one.
2555                  */
2556                 if (m == NULL || m->m_len != sizeof(u_int)) {
2557                         error = EINVAL;
2558                         break;
2559                 }
2560                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2561                 if (loop > 1) {
2562                         error = EINVAL;
2563                         break;
2564                 }
2565                 im6o->im6o_multicast_loop = loop;
2566                 break;
2567
2568         case IPV6_JOIN_GROUP:
2569                 /*
2570                  * Add a multicast group membership.
2571                  * Group must be a valid IP6 multicast address.
2572                  */
2573                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2574                         error = EINVAL;
2575                         break;
2576                 }
2577                 mreq = mtod(m, struct ipv6_mreq *);
2578                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2579                         /*
2580                          * We use the unspecified address to specify to accept
2581                          * all multicast addresses. Only super user is allowed
2582                          * to do this.
2583                          */
2584                         if (priv_check(td, PRIV_ROOT)) {
2585                                 error = EACCES;
2586                                 break;
2587                         }
2588                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2589                         error = EINVAL;
2590                         break;
2591                 }
2592
2593                 /*
2594                  * If the interface is specified, validate it.
2595                  */
2596                 if (mreq->ipv6mr_interface < 0
2597                  || if_index < mreq->ipv6mr_interface) {
2598                         error = ENXIO;  /* XXX EINVAL? */
2599                         break;
2600                 }
2601                 /*
2602                  * If no interface was explicitly specified, choose an
2603                  * appropriate one according to the given multicast address.
2604                  */
2605                 if (mreq->ipv6mr_interface == 0) {
2606                         /*
2607                          * If the multicast address is in node-local scope,
2608                          * the interface should be a loopback interface.
2609                          * Otherwise, look up the routing table for the
2610                          * address, and choose the outgoing interface.
2611                          *   XXX: is it a good approach?
2612                          */
2613                         if (IN6_IS_ADDR_MC_INTFACELOCAL(&mreq->ipv6mr_multiaddr)) {
2614                                 ifp = loif;
2615                         } else {
2616                                 ro.ro_rt = NULL;
2617                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2618                                 bzero(dst, sizeof(*dst));
2619                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2620                                 dst->sin6_family = AF_INET6;
2621                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2622                                 rtalloc((struct route *)&ro);
2623                                 if (ro.ro_rt == NULL) {
2624                                         error = EADDRNOTAVAIL;
2625                                         break;
2626                                 }
2627                                 ifp = ro.ro_rt->rt_ifp;
2628                                 rtfree(ro.ro_rt);
2629                         }
2630                 } else
2631                         ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2632
2633                 /*
2634                  * See if we found an interface, and confirm that it
2635                  * supports multicast
2636                  */
2637                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2638                         error = EADDRNOTAVAIL;
2639                         break;
2640                 }
2641                 /*
2642                  * Put interface index into the multicast address,
2643                  * if the address has link-local scope.
2644                  */
2645                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2646                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2647                                 = htons(mreq->ipv6mr_interface);
2648                 }
2649                 /*
2650                  * See if the membership already exists.
2651                  */
2652                 for (imm = im6o->im6o_memberships.lh_first;
2653                      imm != NULL; imm = imm->i6mm_chain.le_next)
2654                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2655                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2656                                                &mreq->ipv6mr_multiaddr))
2657                                 break;
2658                 if (imm != NULL) {
2659                         error = EADDRINUSE;
2660                         break;
2661                 }
2662                 /*
2663                  * Everything looks good; add a new record to the multicast
2664                  * address list for the given interface.
2665                  */
2666                 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2667                 if ((imm->i6mm_maddr =
2668                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2669                         kfree(imm, M_IPMADDR);
2670                         break;
2671                 }
2672                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2673                 break;
2674
2675         case IPV6_LEAVE_GROUP:
2676                 /*
2677                  * Drop a multicast group membership.
2678                  * Group must be a valid IP6 multicast address.
2679                  */
2680                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2681                         error = EINVAL;
2682                         break;
2683                 }
2684                 mreq = mtod(m, struct ipv6_mreq *);
2685                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2686                         if (priv_check(td, PRIV_ROOT)) {
2687                                 error = EACCES;
2688                                 break;
2689                         }
2690                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2691                         error = EINVAL;
2692                         break;
2693                 }
2694                 /*
2695                  * If an interface address was specified, get a pointer
2696                  * to its ifnet structure.
2697                  */
2698                 if (mreq->ipv6mr_interface < 0
2699                  || if_index < mreq->ipv6mr_interface) {
2700                         error = ENXIO;  /* XXX EINVAL? */
2701                         break;
2702                 }
2703                 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2704                 /*
2705                  * Put interface index into the multicast address,
2706                  * if the address has link-local scope.
2707                  */
2708                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2709                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2710                                 = htons(mreq->ipv6mr_interface);
2711                 }
2712
2713                 /*
2714                  * Find the membership in the membership list.
2715                  */
2716                 for (imm = im6o->im6o_memberships.lh_first;
2717                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2718                         if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2719                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2720                             &mreq->ipv6mr_multiaddr))
2721                                 break;
2722                 }
2723                 if (imm == NULL) {
2724                         /* Unable to resolve interface */
2725                         error = EADDRNOTAVAIL;
2726                         break;
2727                 }
2728                 /*
2729                  * Give up the multicast address record to which the
2730                  * membership points.
2731                  */
2732                 LIST_REMOVE(imm, i6mm_chain);
2733                 in6_delmulti(imm->i6mm_maddr);
2734                 kfree(imm, M_IPMADDR);
2735                 break;
2736
2737         default:
2738                 error = EOPNOTSUPP;
2739                 break;
2740         }
2741
2742         /*
2743          * If all options have default values, no need to keep the mbuf.
2744          */
2745         if (im6o->im6o_multicast_ifp == NULL &&
2746             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2747             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2748             im6o->im6o_memberships.lh_first == NULL) {
2749                 kfree(*im6op, M_IPMOPTS);
2750                 *im6op = NULL;
2751         }
2752
2753         return (error);
2754 }
2755
2756 /*
2757  * Return the IP6 multicast options in response to user getsockopt().
2758  */
2759 static int
2760 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2761 {
2762         u_int *hlim, *loop, *ifindex;
2763
2764         *mp = m_get(M_WAITOK, MT_HEADER);               /* XXX */
2765
2766         switch (optname) {
2767
2768         case IPV6_MULTICAST_IF:
2769                 ifindex = mtod(*mp, u_int *);
2770                 (*mp)->m_len = sizeof(u_int);
2771                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2772                         *ifindex = 0;
2773                 else
2774                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2775                 return (0);
2776
2777         case IPV6_MULTICAST_HOPS:
2778                 hlim = mtod(*mp, u_int *);
2779                 (*mp)->m_len = sizeof(u_int);
2780                 if (im6o == NULL)
2781                         *hlim = ip6_defmcasthlim;
2782                 else
2783                         *hlim = im6o->im6o_multicast_hlim;
2784                 return (0);
2785
2786         case IPV6_MULTICAST_LOOP:
2787                 loop = mtod(*mp, u_int *);
2788                 (*mp)->m_len = sizeof(u_int);
2789                 if (im6o == NULL)
2790                         *loop = ip6_defmcasthlim;
2791                 else
2792                         *loop = im6o->im6o_multicast_loop;
2793                 return (0);
2794
2795         default:
2796                 return (EOPNOTSUPP);
2797         }
2798 }
2799
2800 /*
2801  * Discard the IP6 multicast options.
2802  */
2803 void
2804 ip6_freemoptions(struct ip6_moptions *im6o)
2805 {
2806         struct in6_multi_mship *imm;
2807
2808         if (im6o == NULL)
2809                 return;
2810
2811         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2812                 LIST_REMOVE(imm, i6mm_chain);
2813                 if (imm->i6mm_maddr)
2814                         in6_delmulti(imm->i6mm_maddr);
2815                 kfree(imm, M_IPMADDR);
2816         }
2817         kfree(im6o, M_IPMOPTS);
2818 }
2819
2820 /*
2821  * Set a particular packet option, as a sticky option or an ancillary data
2822  * item.  "len" can be 0 only when it's a sticky option.
2823  * We have 4 cases of combination of "sticky" and "cmsg":
2824  * "sticky=0, cmsg=0": impossible
2825  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2826  * "sticky=1, cmsg=0": RFC3542 socket option
2827  * "sticky=1, cmsg=1": RFC2292 socket option
2828  */
2829 static int
2830 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2831      int sticky, int cmsg, int uproto, int priv)
2832 {
2833         int minmtupolicy, preftemp;
2834         //int error;
2835
2836         if (!sticky && !cmsg) {
2837                 kprintf("ip6_setpktoption: impossible case\n");
2838                 return (EINVAL);
2839         }
2840
2841         /*
2842          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2843          * not be specified in the context of RFC3542.  Conversely,
2844          * RFC3542 types should not be specified in the context of RFC2292.
2845          */
2846         if (!cmsg) {
2847                 switch (optname) {
2848                 case IPV6_2292PKTINFO:
2849                 case IPV6_2292HOPLIMIT:
2850                 case IPV6_2292NEXTHOP:
2851                 case IPV6_2292HOPOPTS:
2852                 case IPV6_2292DSTOPTS:
2853                 case IPV6_2292RTHDR:
2854                 case IPV6_2292PKTOPTIONS:
2855                         return (ENOPROTOOPT);
2856                 }
2857         }
2858         if (sticky && cmsg) {
2859                 switch (optname) {
2860                 case IPV6_PKTINFO:
2861                 case IPV6_HOPLIMIT:
2862                 case IPV6_NEXTHOP:
2863                 case IPV6_HOPOPTS:
2864                 case IPV6_DSTOPTS:
2865                 case IPV6_RTHDRDSTOPTS:
2866                 case IPV6_RTHDR:
2867                 case IPV6_USE_MIN_MTU:
2868                 case IPV6_DONTFRAG:
2869                 case IPV6_TCLASS:
2870                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2871                         return (ENOPROTOOPT);
2872                 }
2873         }
2874
2875         switch (optname) {
2876         case IPV6_2292PKTINFO:
2877         case IPV6_PKTINFO:
2878         {
2879                 struct in6_pktinfo *pktinfo;
2880                 if (len != sizeof(struct in6_pktinfo))
2881                         return (EINVAL);
2882                 pktinfo = (struct in6_pktinfo *)buf;
2883
2884                 /*
2885                  * An application can clear any sticky IPV6_PKTINFO option by
2886                  * doing a "regular" setsockopt with ipi6_addr being
2887                  * in6addr_any and ipi6_ifindex being zero.
2888                  * [RFC 3542, Section 6]
2889                  */
2890                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2891                     pktinfo->ipi6_ifindex == 0 &&
2892                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2893                         ip6_clearpktopts(opt, optname);
2894                         break;
2895                 }
2896
2897                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2898                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2899                         return (EINVAL);
2900                 }
2901
2902                 /* validate the interface index if specified. */
2903                 if (pktinfo->ipi6_ifindex > if_index ||
2904                     pktinfo->ipi6_ifindex < 0) {
2905                          return (ENXIO);
2906                 }
2907                 /*
2908                  * Check if the requested source address is indeed a
2909                  * unicast address assigned to the node, and can be
2910                  * used as the packet's source address.
2911                  */
2912                 if (opt->ip6po_pktinfo != NULL &&
2913                     !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2914                         struct in6_ifaddr *ia6;
2915                         struct sockaddr_in6 sin6;
2916
2917                         bzero(&sin6, sizeof(sin6));
2918                         sin6.sin6_len = sizeof(sin6);
2919                         sin6.sin6_family = AF_INET6;
2920                         sin6.sin6_addr =
2921                         opt->ip6po_pktinfo->ipi6_addr;
2922                         ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2923                         if (ia6 == NULL ||
2924                                 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2925                                         IN6_IFF_NOTREADY)) != 0)
2926                         return (EADDRNOTAVAIL);
2927                 }
2928
2929                 /*
2930                  * We store the address anyway, and let in6_selectsrc()
2931                  * validate the specified address.  This is because ipi6_addr
2932                  * may not have enough information about its scope zone, and
2933                  * we may need additional information (such as outgoing
2934                  * interface or the scope zone of a destination address) to
2935                  * disambiguate the scope.
2936                  * XXX: the delay of the validation may confuse the
2937                  * application when it is used as a sticky option.
2938                  */
2939                 if (opt->ip6po_pktinfo == NULL) {
2940                         opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo),
2941                             M_IP6OPT, M_NOWAIT);
2942                         if (opt->ip6po_pktinfo == NULL)
2943                                 return (ENOBUFS);
2944                 }
2945                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2946                 break;
2947         }
2948
2949         case IPV6_2292HOPLIMIT:
2950         case IPV6_HOPLIMIT:
2951         {
2952                 int *hlimp;
2953
2954                 /*
2955                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2956                  * to simplify the ordering among hoplimit options.
2957                  */
2958                 if (optname == IPV6_HOPLIMIT && sticky)
2959                         return (ENOPROTOOPT);
2960
2961                 if (len != sizeof(int))
2962                         return (EINVAL);
2963                 hlimp = (int *)buf;
2964                 if (*hlimp < -1 || *hlimp > 255)
2965                         return (EINVAL);
2966
2967                 opt->ip6po_hlim = *hlimp;
2968                 break;
2969         }
2970
2971         case IPV6_TCLASS:
2972         {
2973                 int tclass;
2974
2975                 if (len != sizeof(int))
2976                         return (EINVAL);
2977                 tclass = *(int *)buf;
2978                 if (tclass < -1 || tclass > 255)
2979                         return (EINVAL);
2980
2981                 opt->ip6po_tclass = tclass;
2982                 break;
2983         }
2984
2985         case IPV6_2292NEXTHOP:
2986         case IPV6_NEXTHOP:
2987                 if (!priv)
2988                         return (EPERM);
2989
2990                 if (len == 0) { /* just remove the option */
2991                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2992                         break;
2993                 }
2994
2995                 /* check if cmsg_len is large enough for sa_len */
2996                 if (len < sizeof(struct sockaddr) || len < *buf)
2997                         return (EINVAL);
2998
2999                 switch (((struct sockaddr *)buf)->sa_family) {
3000                 case AF_INET6:
3001                 {
3002                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3003                         //int error;
3004
3005                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3006                                 return (EINVAL);
3007
3008                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3009                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3010                                 return (EINVAL);
3011                         }
3012                         break;
3013                 }
3014                 case AF_LINK:   /* should eventually be supported */
3015                 default:
3016                         return (EAFNOSUPPORT);
3017                 }
3018
3019                 /* turn off the previous option, then set the new option. */
3020                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3021                 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT);
3022                 if (opt->ip6po_nexthop == NULL)
3023                         return (ENOBUFS);
3024                 bcopy(buf, opt->ip6po_nexthop, *buf);
3025                 break;
3026
3027         case IPV6_2292HOPOPTS:
3028         case IPV6_HOPOPTS:
3029         {
3030                 struct ip6_hbh *hbh;
3031                 int hbhlen;
3032
3033                 /*
3034                  * XXX: We don't allow a non-privileged user to set ANY HbH
3035                  * options, since per-option restriction has too much
3036                  * overhead.
3037                  */
3038                 if (!priv)
3039                         return (EPERM);
3040                 if (len == 0) {
3041                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3042                         break;  /* just remove the option */
3043                 }
3044
3045                 /* message length validation */
3046                 if (len < sizeof(struct ip6_hbh))
3047                         return (EINVAL);
3048                 hbh = (struct ip6_hbh *)buf;
3049                 hbhlen = (hbh->ip6h_len + 1) << 3;
3050                 if (len != hbhlen)
3051                         return (EINVAL);
3052
3053                 /* turn off the previous option, then set the new option. */
3054                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3055                 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT);
3056                 if (opt->ip6po_hbh == NULL)
3057                         return (ENOBUFS);
3058                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3059
3060                 break;
3061         }
3062
3063         case IPV6_2292DSTOPTS:
3064         case IPV6_DSTOPTS:
3065         case IPV6_RTHDRDSTOPTS:
3066         {
3067                 struct ip6_dest *dest, **newdest = NULL;
3068                 int destlen;
3069                 if (!priv)
3070                         return (EPERM);
3071
3072                 if (len == 0) {
3073                         ip6_clearpktopts(opt, optname);
3074                         break;  /* just remove the option */
3075                 }
3076
3077                 /* message length validation */
3078                 if (len < sizeof(struct ip6_dest))
3079                         return (EINVAL);
3080                 dest = (struct ip6_dest *)buf;
3081                 destlen = (dest->ip6d_len + 1) << 3;
3082                 if (len != destlen)
3083                         return (EINVAL);
3084
3085                 /*
3086                  * Determine the position that the destination options header
3087                  * should be inserted; before or after the routing header.
3088                  */
3089                 switch (optname) {
3090                 case IPV6_2292DSTOPTS:
3091                         /*
3092                          * The old advacned API is ambiguous on this point.
3093                          * Our approach is to determine the position based
3094                          * according to the existence of a routing header.
3095                          * Note, however, that this depends on the order of the
3096                          * extension headers in the ancillary data; the 1st
3097                          * part of the destination options header must appear
3098                          * before the routing header in the ancillary data,
3099                          * too.
3100                          * RFC3542 solved the ambiguity by introducing
3101                          * separate ancillary data or option types.
3102                          */
3103                         if (opt->ip6po_rthdr == NULL)
3104                                 newdest = &opt->ip6po_dest1;
3105                         else
3106                                 newdest = &opt->ip6po_dest2;
3107                         break;
3108                 case IPV6_RTHDRDSTOPTS:
3109                         newdest = &opt->ip6po_dest1;
3110                         break;
3111                 case IPV6_DSTOPTS:
3112                         newdest = &opt->ip6po_dest2;
3113                         break;
3114                 }
3115
3116                 /* turn off the previous option, then set the new option. */
3117                 ip6_clearpktopts(opt, optname);
3118                 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT);
3119                 if (*newdest == NULL)
3120                         return (ENOBUFS);
3121                 bcopy(dest, *newdest, destlen);
3122
3123                 break;
3124         }
3125
3126         case IPV6_2292RTHDR:
3127         case IPV6_RTHDR:
3128         {
3129                 struct ip6_rthdr *rth;
3130                 int rthlen;
3131
3132                 if (len == 0) {
3133                         ip6_clearpktopts(opt, IPV6_RTHDR);
3134                         break;  /* just remove the option */
3135                 }
3136
3137                 /* message length validation */
3138                 if (len < sizeof(struct ip6_rthdr))
3139                         return (EINVAL);
3140                 rth = (struct ip6_rthdr *)buf;
3141                 rthlen = (rth->ip6r_len + 1) << 3;
3142                 if (len != rthlen)
3143                         return (EINVAL);
3144
3145                 switch (rth->ip6r_type) {
3146                 default:
3147                         return (EINVAL);        /* not supported */
3148                 }
3149
3150                 /* turn off the previous option */
3151                 ip6_clearpktopts(opt, IPV6_RTHDR);
3152                 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT);
3153                 if (opt->ip6po_rthdr == NULL)
3154                         return (ENOBUFS);
3155                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3156
3157                 break;
3158         }
3159
3160         case IPV6_USE_MIN_MTU:
3161                 if (len != sizeof(int))
3162                         return (EINVAL);
3163                 minmtupolicy = *(int *)buf;
3164                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3165                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3166                     minmtupolicy != IP6PO_MINMTU_ALL) {
3167                         return (EINVAL);
3168                 }
3169                 opt->ip6po_minmtu = minmtupolicy;
3170                 break;
3171
3172         case IPV6_DONTFRAG:
3173                 if (len != sizeof(int))
3174                         return (EINVAL);
3175
3176                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3177                         /*
3178                          * we ignore this option for TCP sockets.
3179                          * (RFC3542 leaves this case unspecified.)
3180                          */
3181                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3182                 } else
3183                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3184                 break;
3185
3186         case IPV6_PREFER_TEMPADDR:
3187                 if (len != sizeof(int))
3188                         return (EINVAL);
3189                 preftemp = *(int *)buf;
3190                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3191                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3192                     preftemp != IP6PO_TEMPADDR_PREFER) {
3193                         return (EINVAL);
3194                 }
3195                 opt->ip6po_prefer_tempaddr = preftemp;
3196                 break;
3197
3198         default:
3199                 return (ENOPROTOOPT);
3200         } /* end of switch */
3201
3202         return (0);
3203 }
3204
3205
3206 /*
3207  * Set IPv6 outgoing packet options based on advanced API.
3208  */
3209 int
3210 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt,
3211     struct ip6_pktopts *stickyopt, int uproto, int priv)
3212 {
3213         struct cmsghdr *cm = NULL;
3214
3215         if (control == NULL || opt == NULL)
3216                 return (EINVAL);
3217
3218         init_ip6pktopts(opt);
3219
3220         /*
3221          * XXX: Currently, we assume all the optional information is stored
3222          * in a single mbuf.
3223          */
3224         if (stickyopt) {
3225                 int error;
3226
3227                 /*
3228                  * If stickyopt is provided, make a local copy of the options
3229                  * for this particular packet, then override them by ancillary
3230                  * objects.
3231                  * XXX: copypktopts() does not copy the cached route to a next
3232                  * hop (if any).  This is not very good in terms of efficiency,
3233                  * but we can allow this since this option should be rarely
3234                  * used.
3235                  */
3236                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
3237                         return (error);
3238         }
3239
3240         /*
3241          * XXX: Currently, we assume all the optional information is stored
3242          * in a single mbuf.
3243          */
3244         if (control->m_next)
3245                 return (EINVAL);
3246
3247         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3248             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3249                 int error;
3250
3251                 if (control->m_len < CMSG_LEN(0))
3252                         return (EINVAL);
3253
3254                 cm = mtod(control, struct cmsghdr *);
3255                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3256                         return (EINVAL);
3257                 if (cm->cmsg_level != IPPROTO_IPV6)
3258                         continue;
3259
3260                 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3261                     cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto, priv);
3262                 if (error)
3263                         return (error);
3264         }
3265
3266         return (0);
3267 }
3268
3269 /*
3270  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3271  * packet to the input queue of a specified interface.  Note that this
3272  * calls the output routine of the loopback "driver", but with an interface
3273  * pointer that might NOT be loif -- easier than replicating that code here.
3274  */
3275 void
3276 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3277 {
3278         struct mbuf *copym;
3279         struct ip6_hdr *ip6;
3280
3281         copym = m_copy(m, 0, M_COPYALL);
3282         if (copym == NULL)
3283                 return;
3284
3285         /*
3286          * Make sure to deep-copy IPv6 header portion in case the data
3287          * is in an mbuf cluster, so that we can safely override the IPv6
3288          * header portion later.
3289          */
3290         if ((copym->m_flags & M_EXT) != 0 ||
3291             copym->m_len < sizeof(struct ip6_hdr)) {
3292                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3293                 if (copym == NULL)
3294                         return;
3295         }
3296
3297 #ifdef DIAGNOSTIC
3298         if (copym->m_len < sizeof(*ip6)) {
3299                 m_freem(copym);
3300                 return;
3301         }
3302 #endif
3303
3304         ip6 = mtod(copym, struct ip6_hdr *);
3305         /*
3306          * clear embedded scope identifiers if necessary.
3307          * in6_clearscope will touch the addresses only when necessary.
3308          */
3309         in6_clearscope(&ip6->ip6_src);
3310         in6_clearscope(&ip6->ip6_dst);
3311
3312         if_simloop(ifp, copym, dst->sin6_family, 0);
3313 }
3314
3315 /*
3316  * Separate the IPv6 header from the payload into its own mbuf.
3317  *
3318  * Returns the new mbuf chain or the original mbuf if no payload.
3319  * Returns NULL if can't allocate new mbuf for header.
3320  */
3321 static struct mbuf *
3322 ip6_splithdr(struct mbuf *m)
3323 {
3324         struct mbuf *mh;
3325
3326         if (m->m_len <= sizeof(struct ip6_hdr))         /* no payload */
3327                 return (m);
3328
3329         MGETHDR(mh, M_NOWAIT, MT_HEADER);
3330         if (mh == NULL)
3331                 return (NULL);
3332         mh->m_len = sizeof(struct ip6_hdr);
3333         M_MOVE_PKTHDR(mh, m);
3334         MH_ALIGN(mh, sizeof(struct ip6_hdr));
3335         bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr));
3336         m->m_data += sizeof(struct ip6_hdr);
3337         m->m_len -= sizeof(struct ip6_hdr);
3338         mh->m_next = m;
3339         return (mh);
3340 }
3341
3342 /*
3343  * Compute IPv6 extension header length.
3344  */
3345 int
3346 ip6_optlen(struct in6pcb *in6p)
3347 {
3348         int len;
3349
3350         if (!in6p->in6p_outputopts)
3351                 return 0;
3352
3353         len = 0;
3354 #define elen(x) \
3355     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3356
3357         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3358         if (in6p->in6p_outputopts->ip6po_rthdr)
3359                 /* dest1 is valid with rthdr only */
3360                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3361         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3362         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3363         return len;
3364 #undef elen
3365 }