sys/netinet6/ip6_output.c

   1 /*      $FreeBSD: src/sys/netinet6/ip6_output.c,v 1.13.2.18 2003/01/24 05:11:35 sam Exp $       */
   2 /*      $DragonFly: src/sys/netinet6/ip6_output.c,v 1.37 2008/09/04 09:08:22 hasso Exp $        */
   3 /*      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $    */
   4
   5 /*
   6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   7  * All rights reserved.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. Neither the name of the project nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  */
  33
  34 /*
  35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. All advertising materials mentioning features or use of this software
  47  *    must display the following acknowledgement:
  48  *      This product includes software developed by the University of
  49  *      California, Berkeley and its contributors.
  50  * 4. Neither the name of the University nor the names of its contributors
  51  *    may be used to endorse or promote products derived from this software
  52  *    without specific prior written permission.
  53  *
  54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  64  * SUCH DAMAGE.
  65  *
  66  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  67  */
  68
  69 #include "opt_ip6fw.h"
  70 #include "opt_inet.h"
  71 #include "opt_inet6.h"
  72 #include "opt_ipsec.h"
  73
  74 #include <sys/param.h>
  75 #include <sys/malloc.h>
  76 #include <sys/mbuf.h>
  77 #include <sys/errno.h>
  78 #include <sys/protosw.h>
  79 #include <sys/socket.h>
  80 #include <sys/socketvar.h>
  81 #include <sys/systm.h>
  82 #include <sys/kernel.h>
  83 #include <sys/proc.h>
  84
  85 #include <net/if.h>
  86 #include <net/route.h>
  87 #include <net/pfil.h>
  88
  89 #include <netinet/in.h>
  90 #include <netinet/in_var.h>
  91 #include <netinet6/in6_var.h>
  92 #include <netinet/ip6.h>
  93 #include <netinet/icmp6.h>
  94 #include <netinet6/ip6_var.h>
  95 #include <netinet/in_pcb.h>
  96 #include <netinet6/nd6.h>
  97 #include <netinet6/ip6protosw.h>
  98
  99 #ifdef IPSEC
 100 #include <netinet6/ipsec.h>
 101 #ifdef INET6
 102 #include <netinet6/ipsec6.h>
 103 #endif
 104 #include <netproto/key/key.h>
 105 #endif /* IPSEC */
 106
 107 #ifdef FAST_IPSEC
 108 #include <netproto/ipsec/ipsec.h>
 109 #include <netproto/ipsec/ipsec6.h>
 110 #include <netproto/ipsec/key.h>
 111 #endif
 112
 113 #include <net/ip6fw/ip6_fw.h>
 114
 115 #include <net/net_osdep.h>
 116
 117 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
 118
 119 struct ip6_exthdrs {
 120         struct mbuf *ip6e_ip6;
 121         struct mbuf *ip6e_hbh;
 122         struct mbuf *ip6e_dest1;
 123         struct mbuf *ip6e_rthdr;
 124         struct mbuf *ip6e_dest2;
 125 };
 126
 127 static int ip6_pcbopt (int, u_char *, int, struct ip6_pktopts **, int);
 128 static int ip6_setpktoption (int, u_char *, int, struct ip6_pktopts *,
 129          int, int, int, int);
 130 static int ip6_pcbopts (struct ip6_pktopts **, struct mbuf *,
 131                             struct socket *, struct sockopt *);
 132 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 133 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *);
 134 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **);
 135 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
 136         struct ifnet *, struct in6_addr *, u_long *, int *);
 137 static int copyexthdr (void *, struct mbuf **);
 138 static int ip6_insertfraghdr (struct mbuf *, struct mbuf *, int,
 139                                   struct ip6_frag **);
 140 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t);
 141 static struct mbuf *ip6_splithdr (struct mbuf *);
 142 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 143
 144 /*
 145  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 146  * header (with pri, len, nxt, hlim, src, dst).
 147  * This function may modify ver and hlim only.
 148  * The mbuf chain containing the packet will be freed.
 149  * The mbuf opt, if present, will not be freed.
 150  *
 151  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 152  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 153  * which is rt_rmx.rmx_mtu.
 154  */
 155 int
 156 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
 157            int flags, struct ip6_moptions *im6o,
 158            struct ifnet **ifpp,         /* XXX: just for statistics */
 159            struct inpcb *inp)
 160 {
 161         struct ip6_hdr *ip6, *mhip6;
 162         struct ifnet *ifp, *origifp;
 163         struct mbuf *m = m0;
 164         struct mbuf *mprev;
 165         u_char *nexthdrp;
 166         int hlen, tlen, len, off;
 167         struct route_in6 ip6route;
 168         struct sockaddr_in6 *dst;
 169         int error = 0;
 170         struct in6_ifaddr *ia = NULL;
 171         u_long mtu;
 172         int alwaysfrag, dontfrag;
 173         u_int32_t optlen, plen = 0, unfragpartlen;
 174         struct ip6_exthdrs exthdrs;
 175         struct in6_addr finaldst;
 176         struct route_in6 *ro_pmtu = NULL;
 177         boolean_t hdrsplit = FALSE;
 178         boolean_t needipsec = FALSE;
 179 #ifdef IPSEC
 180         boolean_t needipsectun = FALSE;
 181         struct secpolicy *sp = NULL;
 182         struct socket *so = inp ? inp->inp_socket : NULL;
 183
 184         ip6 = mtod(m, struct ip6_hdr *);
 185 #endif
 186 #ifdef FAST_IPSEC
 187         boolean_t needipsectun = FALSE;
 188         struct secpolicy *sp = NULL;
 189
 190         ip6 = mtod(m, struct ip6_hdr *);
 191 #endif
 192
 193         bzero(&exthdrs, sizeof exthdrs);
 194
 195         if (opt) {
 196                 if ((error = copyexthdr(opt->ip6po_hbh, &exthdrs.ip6e_hbh)))
 197                         goto freehdrs;
 198                 if ((error = copyexthdr(opt->ip6po_dest1, &exthdrs.ip6e_dest1)))
 199                         goto freehdrs;
 200                 if ((error = copyexthdr(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr)))
 201                         goto freehdrs;
 202                 if ((error = copyexthdr(opt->ip6po_dest2, &exthdrs.ip6e_dest2)))
 203                         goto freehdrs;
 204         }
 205
 206 #ifdef IPSEC
 207         /* get a security policy for this packet */
 208         if (so == NULL)
 209                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 210         else
 211                 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
 212
 213         if (sp == NULL) {
 214                 ipsec6stat.out_inval++;
 215                 goto freehdrs;
 216         }
 217
 218         error = 0;
 219
 220         /* check policy */
 221         switch (sp->policy) {
 222         case IPSEC_POLICY_DISCARD:
 223                 /*
 224                  * This packet is just discarded.
 225                  */
 226                 ipsec6stat.out_polvio++;
 227                 goto freehdrs;
 228
 229         case IPSEC_POLICY_BYPASS:
 230         case IPSEC_POLICY_NONE:
 231                 /* no need to do IPsec. */
 232                 needipsec = FALSE;
 233                 break;
 234
 235         case IPSEC_POLICY_IPSEC:
 236                 if (sp->req == NULL) {
 237                         error = key_spdacquire(sp);     /* acquire a policy */
 238                         goto freehdrs;
 239                 }
 240                 needipsec = TRUE;
 241                 break;
 242
 243         case IPSEC_POLICY_ENTRUST:
 244         default:
 245                 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
 246         }
 247 #endif /* IPSEC */
 248 #ifdef FAST_IPSEC
 249         /* get a security policy for this packet */
 250         if (inp == NULL)
 251                 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 252         else
 253                 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
 254
 255         if (sp == NULL) {
 256                 newipsecstat.ips_out_inval++;
 257                 goto freehdrs;
 258         }
 259
 260         error = 0;
 261
 262         /* check policy */
 263         switch (sp->policy) {
 264         case IPSEC_POLICY_DISCARD:
 265                 /*
 266                  * This packet is just discarded.
 267                  */
 268                 newipsecstat.ips_out_polvio++;
 269                 goto freehdrs;
 270
 271         case IPSEC_POLICY_BYPASS:
 272         case IPSEC_POLICY_NONE:
 273                 /* no need to do IPsec. */
 274                 needipsec = FALSE;
 275                 break;
 276
 277         case IPSEC_POLICY_IPSEC:
 278                 if (sp->req == NULL) {
 279                         error = key_spdacquire(sp);     /* acquire a policy */
 280                         goto freehdrs;
 281                 }
 282                 needipsec = TRUE;
 283                 break;
 284
 285         case IPSEC_POLICY_ENTRUST:
 286         default:
 287                 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
 288         }
 289 #endif /* FAST_IPSEC */
 290
 291         /*
 292          * Calculate the total length of the extension header chain.
 293          * Keep the length of the unfragmentable part for fragmentation.
 294          */
 295         optlen = m_lengthm(exthdrs.ip6e_hbh, NULL) +
 296             m_lengthm(exthdrs.ip6e_dest1, NULL) +
 297             m_lengthm(exthdrs.ip6e_rthdr, NULL);
 298
 299         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 300
 301         /* NOTE: we don't add AH/ESP length here. do that later. */
 302         optlen += m_lengthm(exthdrs.ip6e_dest2, NULL);
 303
 304         /*
 305          * If we need IPsec, or there is at least one extension header,
 306          * separate IP6 header from the payload.
 307          */
 308         if ((needipsec || optlen) && !hdrsplit) {
 309                 exthdrs.ip6e_ip6 = ip6_splithdr(m);
 310                 if (exthdrs.ip6e_ip6 == NULL) {
 311                         error = ENOBUFS;
 312                         goto freehdrs;
 313                 }
 314                 m = exthdrs.ip6e_ip6;
 315                 hdrsplit = TRUE;
 316         }
 317
 318         /* adjust pointer */
 319         ip6 = mtod(m, struct ip6_hdr *);
 320
 321         /* adjust mbuf packet header length */
 322         m->m_pkthdr.len += optlen;
 323         plen = m->m_pkthdr.len - sizeof(*ip6);
 324
 325         /* If this is a jumbo payload, insert a jumbo payload option. */
 326         if (plen > IPV6_MAXPACKET) {
 327                 if (!hdrsplit) {
 328                         exthdrs.ip6e_ip6 = ip6_splithdr(m);
 329                         if (exthdrs.ip6e_ip6 == NULL) {
 330                                 error = ENOBUFS;
 331                                 goto freehdrs;
 332                         }
 333                         m = exthdrs.ip6e_ip6;
 334                         hdrsplit = TRUE;
 335                 }
 336                 /* adjust pointer */
 337                 ip6 = mtod(m, struct ip6_hdr *);
 338                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 339                         goto freehdrs;
 340                 ip6->ip6_plen = 0;
 341         } else
 342                 ip6->ip6_plen = htons(plen);
 343
 344         /*
 345          * Concatenate headers and fill in next header fields.
 346          * Here we have, on "m"
 347          *      IPv6 payload
 348          * and we insert headers accordingly.  Finally, we should be getting:
 349          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 350          *
 351          * during the header composing process, "m" points to IPv6 header.
 352          * "mprev" points to an extension header prior to esp.
 353          */
 354
 355         nexthdrp = &ip6->ip6_nxt;
 356         mprev = m;
 357
 358         /*
 359          * we treat dest2 specially.  this makes IPsec processing
 360          * much easier.  the goal here is to make mprev point the
 361          * mbuf prior to dest2.
 362          *
 363          * result: IPv6 dest2 payload
 364          * m and mprev will point to IPv6 header.
 365          */
 366         if (exthdrs.ip6e_dest2) {
 367                 if (!hdrsplit)
 368                         panic("assumption failed: hdr not split");
 369                 exthdrs.ip6e_dest2->m_next = m->m_next;
 370                 m->m_next = exthdrs.ip6e_dest2;
 371                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 372                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 373         }
 374
 375 /*
 376  * Place m1 after mprev.
 377  */
 378 #define MAKE_CHAIN(m1, mprev, nexthdrp, i)\
 379     do {\
 380         if (m1) {\
 381                 if (!hdrsplit)\
 382                         panic("assumption failed: hdr not split");\
 383                 *mtod(m1, u_char *) = *nexthdrp;\
 384                 *nexthdrp = (i);\
 385                 nexthdrp = mtod(m1, u_char *);\
 386                 m1->m_next = mprev->m_next;\
 387                 mprev->m_next = m1;\
 388                 mprev = m1;\
 389         }\
 390     } while (0)
 391
 392         /*
 393          * result: IPv6 hbh dest1 rthdr dest2 payload
 394          * m will point to IPv6 header.  mprev will point to the
 395          * extension header prior to dest2 (rthdr in the above case).
 396          */
 397         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 398         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS);
 399         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING);
 400
 401 #if defined(IPSEC) || defined(FAST_IPSEC)
 402         if (needipsec) {
 403                 struct ipsec_output_state state;
 404                 int segleft_org = 0;
 405                 struct ip6_rthdr *rh = NULL;
 406
 407                 /*
 408                  * pointers after IPsec headers are not valid any more.
 409                  * other pointers need a great care too.
 410                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
 411                  */
 412                 exthdrs.ip6e_dest2 = NULL;
 413
 414                 if (exthdrs.ip6e_rthdr) {
 415                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 416                         segleft_org = rh->ip6r_segleft;
 417                         rh->ip6r_segleft = 0;
 418                 }
 419
 420                 bzero(&state, sizeof state);
 421                 state.m = m;
 422                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 423                                             &needipsectun);
 424                 m = state.m;
 425                 if (error) {
 426                         /* mbuf is already reclaimed in ipsec6_output_trans. */
 427                         m = NULL;
 428                         switch (error) {
 429                         case EHOSTUNREACH:
 430                         case ENETUNREACH:
 431                         case EMSGSIZE:
 432                         case ENOBUFS:
 433                         case ENOMEM:
 434                                 break;
 435                         default:
 436                                 kprintf("ip6_output (ipsec): error code %d\n",
 437                                        error);
 438                                 /* fall through */
 439                         case ENOENT:
 440                                 /* don't show these error codes to the user */
 441                                 error = 0;
 442                                 break;
 443                         }
 444                         goto bad;
 445                 }
 446                 if (exthdrs.ip6e_rthdr) {
 447                         /* ah6_output doesn't modify mbuf chain */
 448                         rh->ip6r_segleft = segleft_org;
 449                 }
 450         }
 451 #endif
 452
 453         /*
 454          * If there is a routing header, replace destination address field
 455          * with the first hop of the routing header.
 456          */
 457         if (exthdrs.ip6e_rthdr) {
 458                 struct ip6_rthdr *rh;
 459
 460                 finaldst = ip6->ip6_dst;
 461                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 462                 switch (rh->ip6r_type) {
 463                 default:        /* is it possible? */
 464                          error = EINVAL;
 465                          goto bad;
 466                 }
 467         }
 468
 469         /* Source address validation */
 470         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 471             !(flags & IPV6_DADOUTPUT)) {
 472                 error = EOPNOTSUPP;
 473                 ip6stat.ip6s_badscope++;
 474                 goto bad;
 475         }
 476         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 477                 error = EOPNOTSUPP;
 478                 ip6stat.ip6s_badscope++;
 479                 goto bad;
 480         }
 481
 482         ip6stat.ip6s_localout++;
 483
 484         /*
 485          * Route packet.
 486          */
 487         if (ro == NULL) {
 488                 ro = &ip6route;
 489                 bzero(ro, sizeof(*ro));
 490         }
 491         ro_pmtu = ro;
 492         if (opt && opt->ip6po_rthdr)
 493                 ro = &opt->ip6po_route;
 494         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 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                                 /* fall through */
 551                         case ENOENT:
 552                                 /* don't show these error codes to the user */
 553                                 error = 0;
 554                                 break;
 555                         }
 556                         goto bad;
 557                 }
 558
 559                 exthdrs.ip6e_ip6 = m;
 560         }
 561 #endif
 562
 563         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 564                 /* Unicast */
 565
 566 #define ifatoia6(ifa)   ((struct in6_ifaddr *)(ifa))
 567 #define sin6tosa(sin6)  ((struct sockaddr *)(sin6))
 568                 /* xxx
 569                  * interface selection comes here
 570                  * if an interface is specified from an upper layer,
 571                  * ifp must point it.
 572                  */
 573                 if (ro->ro_rt == NULL) {
 574                         /*
 575                          * non-bsdi always clone routes, if parent is
 576                          * PRF_CLONING.
 577                          */
 578                         rtalloc((struct route *)ro);
 579                 }
 580                 if (ro->ro_rt == NULL) {
 581                         ip6stat.ip6s_noroute++;
 582                         error = EHOSTUNREACH;
 583                         /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
 584                         goto bad;
 585                 }
 586                 ia = ifatoia6(ro->ro_rt->rt_ifa);
 587                 ifp = ro->ro_rt->rt_ifp;
 588                 ro->ro_rt->rt_use++;
 589                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
 590                         dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
 591                 m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
 592
 593                 in6_ifstat_inc(ifp, ifs6_out_request);
 594
 595                 /*
 596                  * Check if the outgoing interface conflicts with
 597                  * the interface specified by ifi6_ifindex (if specified).
 598                  * Note that loopback interface is always okay.
 599                  * (this may happen when we are sending a packet to one of
 600                  *  our own addresses.)
 601                  */
 602                 if (opt && opt->ip6po_pktinfo
 603                  && opt->ip6po_pktinfo->ipi6_ifindex) {
 604                         if (!(ifp->if_flags & IFF_LOOPBACK)
 605                          && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
 606                                 ip6stat.ip6s_noroute++;
 607                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 608                                 error = EHOSTUNREACH;
 609                                 goto bad;
 610                         }
 611                 }
 612
 613                 if (opt && opt->ip6po_hlim != -1)
 614                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 615         } else {
 616                 /* Multicast */
 617                 struct  in6_multi *in6m;
 618
 619                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 620
 621                 /*
 622                  * See if the caller provided any multicast options
 623                  */
 624                 ifp = NULL;
 625                 if (im6o != NULL) {
 626                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 627                         if (im6o->im6o_multicast_ifp != NULL)
 628                                 ifp = im6o->im6o_multicast_ifp;
 629                 } else
 630                         ip6->ip6_hlim = ip6_defmcasthlim;
 631
 632                 /*
 633                  * See if the caller provided the outgoing interface
 634                  * as an ancillary data.
 635                  * Boundary check for ifindex is assumed to be already done.
 636                  */
 637                 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
 638                         ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
 639
 640                 /*
 641                  * If the destination is a node-local scope multicast,
 642                  * the packet should be loop-backed only.
 643                  */
 644                 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
 645                         /*
 646                          * If the outgoing interface is already specified,
 647                          * it should be a loopback interface.
 648                          */
 649                         if (ifp && !(ifp->if_flags & IFF_LOOPBACK)) {
 650                                 ip6stat.ip6s_badscope++;
 651                                 error = ENETUNREACH; /* XXX: better error? */
 652                                 /* XXX correct ifp? */
 653                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 654                                 goto bad;
 655                         } else {
 656                                 ifp = &loif[0];
 657                         }
 658                 }
 659
 660                 if (opt && opt->ip6po_hlim != -1)
 661                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 662
 663                 /*
 664                  * If caller did not provide an interface lookup a
 665                  * default in the routing table.  This is either a
 666                  * default for the speicfied group (i.e. a host
 667                  * route), or a multicast default (a route for the
 668                  * ``net'' ff00::/8).
 669                  */
 670                 if (ifp == NULL) {
 671                         if (ro->ro_rt == NULL) {
 672                                 ro->ro_rt =
 673                                   rtpurelookup((struct sockaddr *)&ro->ro_dst);
 674                         }
 675                         if (ro->ro_rt == NULL) {
 676                                 ip6stat.ip6s_noroute++;
 677                                 error = EHOSTUNREACH;
 678                                 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
 679                                 goto bad;
 680                         }
 681                         ia = ifatoia6(ro->ro_rt->rt_ifa);
 682                         ifp = ro->ro_rt->rt_ifp;
 683                         ro->ro_rt->rt_use++;
 684                 }
 685
 686                 if (!(flags & IPV6_FORWARDING))
 687                         in6_ifstat_inc(ifp, ifs6_out_request);
 688                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 689
 690                 /*
 691                  * Confirm that the outgoing interface supports multicast.
 692                  */
 693                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 694                         ip6stat.ip6s_noroute++;
 695                         in6_ifstat_inc(ifp, ifs6_out_discard);
 696                         error = ENETUNREACH;
 697                         goto bad;
 698                 }
 699                 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
 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,
 864                                         (u_int8_t *)(hbh + 1),
 865                                         ((hbh->ip6h_len + 1) << 3) -
 866                                         sizeof(struct ip6_hbh),
 867                                         &dummy1, &dummy2) < 0) {
 868                         /* m was already freed at this point */
 869                         error = EINVAL;/* better error? */
 870                         goto done;
 871                 }
 872                 m->m_flags &= ~M_LOOP; /* XXX */
 873                 m->m_pkthdr.rcvif = NULL;
 874         }
 875
 876         /*
 877          * Run through list of hooks for output packets.
 878          */
 879         if (pfil_has_hooks(&inet6_pfil_hook)) {
 880                 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
 881                 if (error != 0 || m == NULL)
 882                         goto done;
 883                 ip6 = mtod(m, struct ip6_hdr *);
 884         }
 885
 886         /*
 887          * Send the packet to the outgoing interface.
 888          * If necessary, do IPv6 fragmentation before sending.
 889          *
 890          * the logic here is rather complex:
 891          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 892          * 1-a: send as is if tlen <= path mtu
 893          * 1-b: fragment if tlen > path mtu
 894          *
 895          * 2: if user asks us not to fragment (dontfrag == 1)
 896          * 2-a: send as is if tlen <= interface mtu
 897          * 2-b: error if tlen > interface mtu
 898          *
 899          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 900          *      always fragment
 901          *
 902          * 4: if dontfrag == 1 && alwaysfrag == 1
 903          *      error, as we cannot handle this conflicting request
 904          */
 905         tlen = m->m_pkthdr.len;
 906
 907         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
 908                 dontfrag = 1;
 909         else
 910                 dontfrag = 0;
 911         if (dontfrag && alwaysfrag) {   /* case 4 */
 912                 /* conflicting request - can't transmit */
 913                 error = EMSGSIZE;
 914                 goto bad;
 915         }
 916         if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
 917                 /*
 918                  * Even if the DONTFRAG option is specified, we cannot send the
 919                  * packet when the data length is larger than the MTU of the
 920                  * outgoing interface.
 921                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 922                  * well as returning an error code (the latter is not described
 923                  * in the API spec.)
 924                  */
 925                 u_int32_t mtu32;
 926                 struct ip6ctlparam ip6cp;
 927
 928                 mtu32 = (u_int32_t)mtu;
 929                 bzero(&ip6cp, sizeof(ip6cp));
 930                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
 931                 kpfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
 932                     (void *)&ip6cp);
 933
 934                 error = EMSGSIZE;
 935                 goto bad;
 936         }
 937
 938         /*
 939          * transmit packet without fragmentation
 940          */
 941         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 942                 struct in6_ifaddr *ia6;
 943
 944                 ip6 = mtod(m, struct ip6_hdr *);
 945                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 946                 if (ia6) {
 947                         /* Record statistics for this interface address. */
 948                         ia6->ia_ifa.if_opackets++;
 949                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
 950                 }
 951 #ifdef IPSEC
 952                 /* clean ipsec history once it goes out of the node */
 953                 ipsec_delaux(m);
 954 #endif
 955                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 956                 goto done;
 957         }
 958
 959         /*
 960          * try to fragment the packet.  case 1-b and 3
 961          */
 962         if (mtu < IPV6_MMTU) {
 963                 /*
 964                  * note that path MTU is never less than IPV6_MMTU
 965                  * (see icmp6_input).
 966                  */
 967                 error = EMSGSIZE;
 968                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 969                 goto bad;
 970         } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
 971                 error = EMSGSIZE;
 972                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 973                 goto bad;
 974         } else {
 975                 struct mbuf **mnext, *m_frgpart;
 976                 struct ip6_frag *ip6f;
 977                 u_int32_t id = htonl(ip6_id++);
 978                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
 979                 u_char nextproto;
 980
 981                 /*
 982                  * Too large for the destination or interface;
 983                  * fragment if possible.
 984                  * Must be able to put at least 8 bytes per fragment.
 985                  */
 986                 hlen = unfragpartlen;
 987                 if (mtu > IPV6_MAXPACKET)
 988                         mtu = IPV6_MAXPACKET;
 989
 990                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 991                 if (len < 8) {
 992                         error = EMSGSIZE;
 993                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 994                         goto bad;
 995                 }
 996
 997                 /*
 998                  * Verify that we have any chance at all of being able to queue
 999                  *      the packet or packet fragments
1000                  */
1001                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1002                     < tlen  /* - hlen */)) {
1003                         error = ENOBUFS;
1004                         ip6stat.ip6s_odropped++;
1005                         goto bad;
1006                 }
1007
1008                 mnext = &m->m_nextpkt;
1009
1010                 /*
1011                  * Change the next header field of the last header in the
1012                  * unfragmentable part.
1013                  */
1014                 if (exthdrs.ip6e_rthdr) {
1015                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1016                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1017                 } else if (exthdrs.ip6e_dest1) {
1018                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1019                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1020                 } else if (exthdrs.ip6e_hbh) {
1021                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1022                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1023                 } else {
1024                         nextproto = ip6->ip6_nxt;
1025                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1026                 }
1027
1028                 /*
1029                  * Loop through length of segment after first fragment,
1030                  * make new header and copy data of each part and link onto
1031                  * chain.
1032                  */
1033                 m0 = m;
1034                 for (off = hlen; off < tlen; off += len) {
1035                         MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1036                         if (!m) {
1037                                 error = ENOBUFS;
1038                                 ip6stat.ip6s_odropped++;
1039                                 goto sendorfree;
1040                         }
1041                         m->m_pkthdr.rcvif = NULL;
1042                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1043                         *mnext = m;
1044                         mnext = &m->m_nextpkt;
1045                         m->m_data += max_linkhdr;
1046                         mhip6 = mtod(m, struct ip6_hdr *);
1047                         *mhip6 = *ip6;
1048                         m->m_len = sizeof(*mhip6);
1049                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1050                         if (error) {
1051                                 ip6stat.ip6s_odropped++;
1052                                 goto sendorfree;
1053                         }
1054                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1055                         if (off + len >= tlen)
1056                                 len = tlen - off;
1057                         else
1058                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1059                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1060                                                           sizeof(*ip6f) -
1061                                                           sizeof(struct ip6_hdr)));
1062                         if ((m_frgpart = m_copy(m0, off, len)) == NULL) {
1063                                 error = ENOBUFS;
1064                                 ip6stat.ip6s_odropped++;
1065                                 goto sendorfree;
1066                         }
1067                         m_cat(m, m_frgpart);
1068                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1069                         m->m_pkthdr.rcvif = (struct ifnet *)0;
1070                         ip6f->ip6f_reserved = 0;
1071                         ip6f->ip6f_ident = id;
1072                         ip6f->ip6f_nxt = nextproto;
1073                         ip6stat.ip6s_ofragments++;
1074                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1075                 }
1076
1077                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1078         }
1079
1080         /*
1081          * Remove leading garbages.
1082          */
1083 sendorfree:
1084         m = m0->m_nextpkt;
1085         m0->m_nextpkt = NULL;
1086         m_freem(m0);
1087         for (m0 = m; m; m = m0) {
1088                 m0 = m->m_nextpkt;
1089                 m->m_nextpkt = NULL;
1090                 if (error == 0) {
1091                         /* Record statistics for this interface address. */
1092                         if (ia) {
1093                                 ia->ia_ifa.if_opackets++;
1094                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1095                         }
1096 #ifdef IPSEC
1097                         /* clean ipsec history once it goes out of the node */
1098                         ipsec_delaux(m);
1099 #endif
1100                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1101                 } else
1102                         m_freem(m);
1103         }
1104
1105         if (error == 0)
1106                 ip6stat.ip6s_fragmented++;
1107
1108 done:
1109         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1110                 RTFREE(ro->ro_rt);
1111         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1112                 RTFREE(ro_pmtu->ro_rt);
1113         }
1114
1115 #ifdef IPSEC
1116         if (sp != NULL)
1117                 key_freesp(sp);
1118 #endif
1119 #ifdef FAST_IPSEC
1120         if (sp != NULL)
1121                 KEY_FREESP(&sp);
1122 #endif
1123
1124         return (error);
1125
1126 freehdrs:
1127         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1128         m_freem(exthdrs.ip6e_dest1);
1129         m_freem(exthdrs.ip6e_rthdr);
1130         m_freem(exthdrs.ip6e_dest2);
1131         /* fall through */
1132 bad:
1133         m_freem(m);
1134         goto done;
1135 }
1136
1137 static int
1138 copyexthdr(void *h, struct mbuf **mp)
1139 {
1140         struct ip6_ext *hdr = h;
1141         int hlen;
1142         struct mbuf *m;
1143
1144         if (hdr == NULL)
1145                 return 0;
1146
1147         hlen = (hdr->ip6e_len + 1) * 8;
1148         if (hlen > MCLBYTES)
1149                 return ENOBUFS; /* XXX */
1150
1151         m = m_getb(hlen, MB_DONTWAIT, MT_DATA, 0);
1152         if (!m)
1153                 return ENOBUFS;
1154         m->m_len = hlen;
1155
1156         bcopy(hdr, mtod(m, caddr_t), hlen);
1157
1158         *mp = m;
1159         return 0;
1160 }
1161
1162 /*
1163  * Insert jumbo payload option.
1164  */
1165 static int
1166 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1167 {
1168         struct mbuf *mopt;
1169         u_char *optbuf;
1170         u_int32_t v;
1171
1172 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1173
1174         /*
1175          * If there is no hop-by-hop options header, allocate new one.
1176          * If there is one but it doesn't have enough space to store the
1177          * jumbo payload option, allocate a cluster to store the whole options.
1178          * Otherwise, use it to store the options.
1179          */
1180         if (exthdrs->ip6e_hbh == NULL) {
1181                 MGET(mopt, MB_DONTWAIT, MT_DATA);
1182                 if (mopt == NULL)
1183                         return (ENOBUFS);
1184                 mopt->m_len = JUMBOOPTLEN;
1185                 optbuf = mtod(mopt, u_char *);
1186                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1187                 exthdrs->ip6e_hbh = mopt;
1188         } else {
1189                 struct ip6_hbh *hbh;
1190
1191                 mopt = exthdrs->ip6e_hbh;
1192                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1193                         /*
1194                          * XXX assumption:
1195                          * - exthdrs->ip6e_hbh is not referenced from places
1196                          *   other than exthdrs.
1197                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1198                          */
1199                         int oldoptlen = mopt->m_len;
1200                         struct mbuf *n;
1201
1202                         /*
1203                          * XXX: give up if the whole (new) hbh header does
1204                          * not fit even in an mbuf cluster.
1205                          */
1206                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1207                                 return (ENOBUFS);
1208
1209                         /*
1210                          * As a consequence, we must always prepare a cluster
1211                          * at this point.
1212                          */
1213                         n = m_getcl(MB_DONTWAIT, MT_DATA, 0);
1214                         if (!n)
1215                                 return (ENOBUFS);
1216                         n->m_len = oldoptlen + JUMBOOPTLEN;
1217                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen);
1218                         optbuf = mtod(n, caddr_t) + oldoptlen;
1219                         m_freem(mopt);
1220                         mopt = exthdrs->ip6e_hbh = n;
1221                 } else {
1222                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1223                         mopt->m_len += JUMBOOPTLEN;
1224                 }
1225                 optbuf[0] = IP6OPT_PADN;
1226                 optbuf[1] = 1;
1227
1228                 /*
1229                  * Adjust the header length according to the pad and
1230                  * the jumbo payload option.
1231                  */
1232                 hbh = mtod(mopt, struct ip6_hbh *);
1233                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1234         }
1235
1236         /* fill in the option. */
1237         optbuf[2] = IP6OPT_JUMBO;
1238         optbuf[3] = 4;
1239         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1240         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1241
1242         /* finally, adjust the packet header length */
1243         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1244
1245         return (0);
1246 #undef JUMBOOPTLEN
1247 }
1248
1249 /*
1250  * Insert fragment header and copy unfragmentable header portions.
1251  */
1252 static int
1253 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1254                   struct ip6_frag **frghdrp)
1255 {
1256         struct mbuf *n, *mlast;
1257
1258         if (hlen > sizeof(struct ip6_hdr)) {
1259                 n = m_copym(m0, sizeof(struct ip6_hdr),
1260                             hlen - sizeof(struct ip6_hdr), MB_DONTWAIT);
1261                 if (n == NULL)
1262                         return (ENOBUFS);
1263                 m->m_next = n;
1264         } else
1265                 n = m;
1266
1267         /* Search for the last mbuf of unfragmentable part. */
1268         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1269                 ;
1270
1271         if (!(mlast->m_flags & M_EXT) &&
1272             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1273                 /* use the trailing space of the last mbuf for the fragment hdr */
1274                 *frghdrp =
1275                         (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1276                 mlast->m_len += sizeof(struct ip6_frag);
1277                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1278         } else {
1279                 /* allocate a new mbuf for the fragment header */
1280                 struct mbuf *mfrg;
1281
1282                 MGET(mfrg, MB_DONTWAIT, MT_DATA);
1283                 if (mfrg == NULL)
1284                         return (ENOBUFS);
1285                 mfrg->m_len = sizeof(struct ip6_frag);
1286                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1287                 mlast->m_next = mfrg;
1288         }
1289
1290         return (0);
1291 }
1292
1293 static int
1294 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1295     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1296     int *alwaysfragp)
1297 {
1298         u_int32_t mtu = 0;
1299         int alwaysfrag = 0;
1300         int error = 0;
1301
1302         if (ro_pmtu != ro) {
1303                 /* The first hop and the final destination may differ. */
1304                 struct sockaddr_in6 *sa6_dst =
1305                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1306                 if (ro_pmtu->ro_rt &&
1307                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1308                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1309                         RTFREE(ro_pmtu->ro_rt);
1310                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1311                 }
1312                 if (ro_pmtu->ro_rt == NULL) {
1313                         bzero(sa6_dst, sizeof(*sa6_dst));
1314                         sa6_dst->sin6_family = AF_INET6;
1315                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1316                         sa6_dst->sin6_addr = *dst;
1317
1318                         rtalloc((struct route *)ro_pmtu);
1319                 }
1320         }
1321         if (ro_pmtu->ro_rt) {
1322                 u_int32_t ifmtu;
1323                 struct in_conninfo inc;
1324
1325                 bzero(&inc, sizeof(inc));
1326                 inc.inc_flags = 1; /* IPv6 */
1327                 inc.inc6_faddr = *dst;
1328
1329                 if (ifp == NULL)
1330                         ifp = ro_pmtu->ro_rt->rt_ifp;
1331                 ifmtu = IN6_LINKMTU(ifp);
1332                 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1333                 if (mtu == 0)
1334                         mtu = ifmtu;
1335                 else if (mtu < IPV6_MMTU) {
1336                         /*
1337                          * RFC2460 section 5, last paragraph:
1338                          * if we record ICMPv6 too big message with
1339                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1340                          * or smaller, with framgent header attached.
1341                          * (fragment header is needed regardless from the
1342                          * packet size, for translators to identify packets)
1343                          */
1344                         alwaysfrag = 1;
1345                         mtu = IPV6_MMTU;
1346                 } else if (mtu > ifmtu) {
1347                         /*
1348                          * The MTU on the route is larger than the MTU on
1349                          * the interface!  This shouldn't happen, unless the
1350                          * MTU of the interface has been changed after the
1351                          * interface was brought up.  Change the MTU in the
1352                          * route to match the interface MTU (as long as the
1353                          * field isn't locked).
1354                          */
1355                         mtu = ifmtu;
1356                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1357                 }
1358         } else if (ifp) {
1359                 mtu = IN6_LINKMTU(ifp);
1360         } else
1361                 error = EHOSTUNREACH; /* XXX */
1362
1363         *mtup = mtu;
1364         if (alwaysfragp)
1365                 *alwaysfragp = alwaysfrag;
1366         return (error);
1367 }
1368
1369 /*
1370  * IP6 socket option processing.
1371  */
1372 int
1373 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1374 {
1375         int optdatalen,uproto;
1376         int privileged;
1377         struct inpcb *in6p = so->so_pcb;
1378         void *optdata;
1379         int error, optval;
1380         int level, op, optname;
1381         int optlen;
1382         struct thread *td;
1383
1384         if (sopt) {
1385                 level = sopt->sopt_level;
1386                 op = sopt->sopt_dir;
1387                 optname = sopt->sopt_name;
1388                 optlen = sopt->sopt_valsize;
1389                 td = sopt->sopt_td;
1390         } else {
1391                 panic("ip6_ctloutput: arg soopt is NULL");
1392                 /* NOT REACHED */
1393                 td = NULL;
1394         }
1395         error = optval = 0;
1396
1397         uproto = (int)so->so_proto->pr_protocol;
1398         privileged = (td == NULL || suser(td)) ? 0 : 1;
1399
1400         if (level == IPPROTO_IPV6) {
1401                 switch (op) {
1402
1403                 case SOPT_SET:
1404                         switch (optname) {
1405                         case IPV6_2292PKTOPTIONS:
1406 #ifdef IPV6_PKTOPTIONS
1407                         case IPV6_PKTOPTIONS:
1408 #endif
1409                         {
1410                                 struct mbuf *m;
1411
1412                                 error = soopt_getm(sopt, &m); /* XXX */
1413                                 if (error != 0)
1414                                         break;
1415                                 soopt_to_mbuf(sopt, m); /* XXX */
1416                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1417                                                     m, so, sopt);
1418                                 m_freem(m); /* XXX */
1419                                 break;
1420                         }
1421
1422                         /*
1423                          * Use of some Hop-by-Hop options or some
1424                          * Destination options, might require special
1425                          * privilege.  That is, normal applications
1426                          * (without special privilege) might be forbidden
1427                          * from setting certain options in outgoing packets,
1428                          * and might never see certain options in received
1429                          * packets. [RFC 2292 Section 6]
1430                          * KAME specific note:
1431                          *  KAME prevents non-privileged users from sending or
1432                          *  receiving ANY hbh/dst options in order to avoid
1433                          *  overhead of parsing options in the kernel.
1434                          */
1435                         case IPV6_RECVHOPOPTS:
1436                         case IPV6_RECVDSTOPTS:
1437                         case IPV6_RECVRTHDRDSTOPTS:
1438                                 if (!privileged)
1439                                         return (EPERM);
1440                         case IPV6_RECVPKTINFO:
1441                         case IPV6_RECVHOPLIMIT:
1442                         case IPV6_RECVRTHDR:
1443                         case IPV6_RECVPATHMTU:
1444                         case IPV6_RECVTCLASS:
1445                         case IPV6_AUTOFLOWLABEL:
1446                         case IPV6_HOPLIMIT:
1447                         /* FALLTHROUGH */
1448                         case IPV6_UNICAST_HOPS:
1449                         case IPV6_FAITH:
1450
1451                         case IPV6_V6ONLY:
1452                                 if (optlen != sizeof(int)) {
1453                                         error = EINVAL;
1454                                         break;
1455                                 }
1456                                 error = soopt_to_kbuf(sopt, &optval,
1457                                         sizeof optval, sizeof optval);
1458                                 if (error)
1459                                         break;
1460                                 switch (optname) {
1461
1462                                 case IPV6_UNICAST_HOPS:
1463                                         if (optval < -1 || optval >= 256)
1464                                                 error = EINVAL;
1465                                         else {
1466                                                 /* -1 = kernel default */
1467                                                 in6p->in6p_hops = optval;
1468
1469                                                 if ((in6p->in6p_vflag &
1470                                                      INP_IPV4) != 0)
1471                                                         in6p->inp_ip_ttl = optval;
1472                                         }
1473                                         break;
1474 #define OPTSET(bit) \
1475 do { \
1476         if (optval) \
1477                 in6p->in6p_flags |= (bit); \
1478         else \
1479                 in6p->in6p_flags &= ~(bit); \
1480 } while (0)
1481 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1482 /*
1483  * Although changed to RFC3542, It's better to also support RFC2292 API
1484  */
1485 #define OPTSET2292(bit) \
1486 do { \
1487         in6p->in6p_flags |= IN6P_RFC2292; \
1488         if (optval) \
1489                 in6p->in6p_flags |= (bit); \
1490         else \
1491                 in6p->in6p_flags &= ~(bit); \
1492 } while (/*CONSTCOND*/ 0)
1493
1494                                 case IPV6_RECVPKTINFO:
1495                                         /* cannot mix with RFC2292 */
1496                                         if (OPTBIT(IN6P_RFC2292)) {
1497                                                 error = EINVAL;
1498                                                 break;
1499                                         }
1500                                         OPTSET(IN6P_PKTINFO);
1501                                         break;
1502
1503                                 case IPV6_HOPLIMIT:
1504                                 {
1505                                         struct ip6_pktopts **optp;
1506
1507                                         /* cannot mix with RFC2292 */
1508                                         if (OPTBIT(IN6P_RFC2292)) {
1509                                                 error = EINVAL;
1510                                                 break;
1511                                         }
1512                                         optp = &in6p->in6p_outputopts;
1513                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1514                                             (u_char *)&optval, sizeof(optval),
1515                                             optp, uproto);
1516                                         break;
1517                                 }
1518
1519                                 case IPV6_RECVHOPLIMIT:
1520                                         /* cannot mix with RFC2292 */
1521                                         if (OPTBIT(IN6P_RFC2292)) {
1522                                                 error = EINVAL;
1523                                                 break;
1524                                         }
1525                                         OPTSET(IN6P_HOPLIMIT);
1526                                         break;
1527
1528                                 case IPV6_RECVHOPOPTS:
1529                                         /* cannot mix with RFC2292 */
1530                                         if (OPTBIT(IN6P_RFC2292)) {
1531                                                 error = EINVAL;
1532                                                 break;
1533                                         }
1534                                         OPTSET(IN6P_HOPOPTS);
1535                                         break;
1536
1537                                 case IPV6_RECVDSTOPTS:
1538                                         /* cannot mix with RFC2292 */
1539                                         if (OPTBIT(IN6P_RFC2292)) {
1540                                                 error = EINVAL;
1541                                                 break;
1542                                         }
1543                                         OPTSET(IN6P_DSTOPTS);
1544                                         break;
1545
1546                                 case IPV6_RECVRTHDRDSTOPTS:
1547                                         /* cannot mix with RFC2292 */
1548                                         if (OPTBIT(IN6P_RFC2292)) {
1549                                                 error = EINVAL;
1550                                                 break;
1551                                         }
1552                                         OPTSET(IN6P_RTHDRDSTOPTS);
1553                                         break;
1554
1555                                 case IPV6_RECVRTHDR:
1556                                         /* cannot mix with RFC2292 */
1557                                         if (OPTBIT(IN6P_RFC2292)) {
1558                                                 error = EINVAL;
1559                                                 break;
1560                                         }
1561                                         OPTSET(IN6P_RTHDR);
1562                                         break;
1563
1564                                 case IPV6_RECVPATHMTU:
1565                                         /*
1566                                          * We ignore this option for TCP
1567                                          * sockets.
1568                                          * (RFC3542 leaves this case
1569                                          * unspecified.)
1570                                          */
1571                                         if (uproto != IPPROTO_TCP)
1572                                                 OPTSET(IN6P_MTU);
1573                                         break;
1574
1575                                 case IPV6_RECVTCLASS:
1576                                         /* cannot mix with RFC2292 XXX */
1577                                         if (OPTBIT(IN6P_RFC2292)) {
1578                                                 error = EINVAL;
1579                                                 break;
1580                                         }
1581                                         OPTSET(IN6P_TCLASS);
1582                                         break;
1583
1584                                 case IPV6_AUTOFLOWLABEL:
1585                                         OPTSET(IN6P_AUTOFLOWLABEL);
1586                                         break;
1587
1588                                 case IPV6_FAITH:
1589                                         OPTSET(IN6P_FAITH);
1590                                         break;
1591
1592                                 case IPV6_V6ONLY:
1593                                         /*
1594                                          * make setsockopt(IPV6_V6ONLY)
1595                                          * available only prior to bind(2).
1596                                          */
1597                                         if (in6p->in6p_lport ||
1598                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1599                                         {
1600                                                 error = EINVAL;
1601                                                 break;
1602                                         }
1603                                         OPTSET(IN6P_IPV6_V6ONLY);
1604                                         if (optval)
1605                                                 in6p->in6p_vflag &= ~INP_IPV4;
1606                                         else
1607                                                 in6p->in6p_vflag |= INP_IPV4;
1608                                         break;
1609                                 }
1610                                 break;
1611
1612                         case IPV6_TCLASS:
1613                         case IPV6_DONTFRAG:
1614                         case IPV6_USE_MIN_MTU:
1615                         case IPV6_PREFER_TEMPADDR:
1616                                 if (optlen != sizeof(optval)) {
1617                                         error = EINVAL;
1618                                         break;
1619                                 }
1620                                 error = soopt_to_kbuf(sopt, &optval,
1621                                         sizeof optval, sizeof optval);
1622                                 if (error)
1623                                         break;
1624                                 {
1625                                         struct ip6_pktopts **optp;
1626                                         optp = &in6p->in6p_outputopts;
1627                                         error = ip6_pcbopt(optname,
1628                                             (u_char *)&optval, sizeof(optval),
1629                                             optp, uproto);
1630                                         break;
1631                                 }
1632
1633                         case IPV6_2292PKTINFO:
1634                         case IPV6_2292HOPLIMIT:
1635                         case IPV6_2292HOPOPTS:
1636                         case IPV6_2292DSTOPTS:
1637                         case IPV6_2292RTHDR:
1638                                 /* RFC 2292 */
1639                                 if (optlen != sizeof(int)) {
1640                                         error = EINVAL;
1641                                         break;
1642                                 }
1643                                 error = soopt_to_kbuf(sopt, &optval,
1644                                         sizeof optval, sizeof optval);
1645                                 if (error)
1646                                         break;
1647                                 switch (optname) {
1648                                 case IPV6_2292PKTINFO:
1649                                         OPTSET2292(IN6P_PKTINFO);
1650                                         break;
1651                                 case IPV6_2292HOPLIMIT:
1652                                         OPTSET2292(IN6P_HOPLIMIT);
1653                                         break;
1654                                 case IPV6_2292HOPOPTS:
1655                                         /*
1656                                          * Check super-user privilege.
1657                                          * See comments for IPV6_RECVHOPOPTS.
1658                                          */
1659                                         if (!privileged)
1660                                                 return (EPERM);
1661                                         OPTSET2292(IN6P_HOPOPTS);
1662                                         break;
1663                                 case IPV6_2292DSTOPTS:
1664                                         if (!privileged)
1665                                                 return (EPERM);
1666                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1667                                         break;
1668                                 case IPV6_2292RTHDR:
1669                                         OPTSET2292(IN6P_RTHDR);
1670                                         break;
1671                                 }
1672                                 break;
1673
1674                         case IPV6_PKTINFO:
1675                         case IPV6_HOPOPTS:
1676                         case IPV6_RTHDR:
1677                         case IPV6_DSTOPTS:
1678                         case IPV6_RTHDRDSTOPTS:
1679                         case IPV6_NEXTHOP:
1680                         {
1681                                 /*
1682                                  * New advanced API (RFC3542)
1683                                  */
1684                                 u_char *optbuf;
1685                                 u_char optbuf_storage[MCLBYTES];
1686                                 int optlen;
1687                                 struct ip6_pktopts **optp;
1688
1689                                 /* cannot mix with RFC2292 */
1690                                 if (OPTBIT(IN6P_RFC2292)) {
1691                                         error = EINVAL;
1692                                         break;
1693                                 }
1694
1695                                 /*
1696                                  * We only ensure valsize is not too large
1697                                  * here.  Further validation will be done
1698                                  * later.
1699                                  */
1700                                 error = soopt_to_kbuf(sopt, optbuf_storage,
1701                                     sizeof(optbuf_storage), 0);
1702                                 if (error)
1703                                         break;
1704                                 optlen = sopt->sopt_valsize;
1705                                 optbuf = optbuf_storage;
1706                                 optp = &in6p->in6p_outputopts;
1707                                 error = ip6_pcbopt(optname, optbuf, optlen,
1708                                     optp, uproto);
1709                                 break;
1710                         }
1711 #undef OPTSET
1712
1713                         case IPV6_MULTICAST_IF:
1714                         case IPV6_MULTICAST_HOPS:
1715                         case IPV6_MULTICAST_LOOP:
1716                         case IPV6_JOIN_GROUP:
1717                         case IPV6_LEAVE_GROUP:
1718                             {
1719                                 struct mbuf *m;
1720                                 if (sopt->sopt_valsize > MLEN) {
1721                                         error = EMSGSIZE;
1722                                         break;
1723                                 }
1724                                 /* XXX */
1725                                 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1726                                 if (m == NULL) {
1727                                         error = ENOBUFS;
1728                                         break;
1729                                 }
1730                                 m->m_len = sopt->sopt_valsize;
1731                                 error = soopt_to_kbuf(sopt, mtod(m, char *),
1732                                                     m->m_len, m->m_len);
1733                                 error = ip6_setmoptions(sopt->sopt_name,
1734                                                         &in6p->in6p_moptions,
1735                                                         m);
1736                                 m_free(m);
1737                             }
1738                                 break;
1739
1740                         case IPV6_PORTRANGE:
1741                                 error = soopt_to_kbuf(sopt, &optval,
1742                                     sizeof optval, sizeof optval);
1743                                 if (error)
1744                                         break;
1745
1746                                 switch (optval) {
1747                                 case IPV6_PORTRANGE_DEFAULT:
1748                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1749                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1750                                         break;
1751
1752                                 case IPV6_PORTRANGE_HIGH:
1753                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1754                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1755                                         break;
1756
1757                                 case IPV6_PORTRANGE_LOW:
1758                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1759                                         in6p->in6p_flags |= IN6P_LOWPORT;
1760                                         break;
1761
1762                                 default:
1763                                         error = EINVAL;
1764                                         break;
1765                                 }
1766                                 break;
1767
1768 #if defined(IPSEC) || defined(FAST_IPSEC)
1769                         case IPV6_IPSEC_POLICY:
1770                             {
1771                                 caddr_t req = NULL;
1772                                 size_t len = 0;
1773                                 struct mbuf *m;
1774
1775                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1776                                         break;
1777                                 soopt_to_mbuf(sopt, m);         /* XXX */
1778                                 if (m) {
1779                                         req = mtod(m, caddr_t);
1780                                         len = m->m_len;
1781                                 }
1782                                 error = ipsec6_set_policy(in6p, optname, req,
1783                                                           len, privileged);
1784                                 m_freem(m);
1785                             }
1786                                 break;
1787 #endif /* KAME IPSEC */
1788
1789                         case IPV6_FW_ADD:
1790                         case IPV6_FW_DEL:
1791                         case IPV6_FW_FLUSH:
1792                         case IPV6_FW_ZERO:
1793                             {
1794                                 struct mbuf *m;
1795                                 struct mbuf **mp = &m;
1796
1797                                 if (ip6_fw_ctl_ptr == NULL)
1798                                         return EINVAL;
1799                                 /* XXX */
1800                                 if ((error = soopt_getm(sopt, &m)) != 0)
1801                                         break;
1802                                 /* XXX */
1803                                 soopt_to_mbuf(sopt, m);
1804                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1805                                 m = *mp;
1806                             }
1807                                 break;
1808
1809                         default:
1810                                 error = ENOPROTOOPT;
1811                                 break;
1812                         }
1813                         break;
1814
1815                 case SOPT_GET:
1816                         switch (optname) {
1817                         case IPV6_2292PKTOPTIONS:
1818 #ifdef IPV6_PKTOPTIONS
1819                         case IPV6_PKTOPTIONS:
1820 #endif
1821                                 /*
1822                                  * RFC3542 (effectively) deprecated the
1823                                  * semantics of the 2292-style pktoptions.
1824                                  * Since it was not reliable in nature (i.e.,
1825                                  * applications had to expect the lack of some
1826                                  * information after all), it would make sense
1827                                  * to simplify this part by always returning
1828                                  * empty data.
1829                                  */
1830                                 if (in6p->in6p_options) {
1831                                         struct mbuf *m;
1832                                         m = m_copym(in6p->in6p_options,
1833                                             0, M_COPYALL, MB_WAIT);
1834                                         error = soopt_from_mbuf(sopt, m);
1835                                         if (error == 0)
1836                                                 m_freem(m);
1837                                 } else
1838                                         sopt->sopt_valsize = 0;
1839                                 break;
1840
1841                         case IPV6_RECVHOPOPTS:
1842                         case IPV6_RECVDSTOPTS:
1843                         case IPV6_RECVRTHDRDSTOPTS:
1844                         case IPV6_UNICAST_HOPS:
1845                         case IPV6_RECVPKTINFO:
1846                         case IPV6_RECVHOPLIMIT:
1847                         case IPV6_RECVRTHDR:
1848                         case IPV6_RECVPATHMTU:
1849                         case IPV6_RECVTCLASS:
1850                         case IPV6_AUTOFLOWLABEL:
1851                         case IPV6_FAITH:
1852                         case IPV6_V6ONLY:
1853                         case IPV6_PORTRANGE:
1854                                 switch (optname) {
1855
1856                                 case IPV6_RECVHOPOPTS:
1857                                         optval = OPTBIT(IN6P_HOPOPTS);
1858                                         break;
1859
1860                                 case IPV6_RECVDSTOPTS:
1861                                         optval = OPTBIT(IN6P_DSTOPTS);
1862                                         break;
1863
1864                                 case IPV6_RECVRTHDRDSTOPTS:
1865                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1866                                         break;
1867
1868                                 case IPV6_RECVPKTINFO:
1869                                         optval = OPTBIT(IN6P_PKTINFO);
1870                                         break;
1871
1872                                 case IPV6_RECVHOPLIMIT:
1873                                         optval = OPTBIT(IN6P_HOPLIMIT);
1874                                         break;
1875
1876                                 case IPV6_RECVRTHDR:
1877                                         optval = OPTBIT(IN6P_RTHDR);
1878                                         break;
1879
1880                                 case IPV6_RECVPATHMTU:
1881                                         optval = OPTBIT(IN6P_MTU);
1882                                         break;
1883
1884                                 case IPV6_RECVTCLASS:
1885                                         optval = OPTBIT(IN6P_TCLASS);
1886                                         break;
1887
1888                                 case IPV6_AUTOFLOWLABEL:
1889                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1890                                         break;
1891
1892
1893                                 case IPV6_UNICAST_HOPS:
1894                                         optval = in6p->in6p_hops;
1895                                         break;
1896
1897                                 case IPV6_FAITH:
1898                                         optval = OPTBIT(IN6P_FAITH);
1899                                         break;
1900
1901                                 case IPV6_V6ONLY:
1902                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1903                                         break;
1904
1905                                 case IPV6_PORTRANGE:
1906                                     {
1907                                         int flags;
1908                                         flags = in6p->in6p_flags;
1909                                         if (flags & IN6P_HIGHPORT)
1910                                                 optval = IPV6_PORTRANGE_HIGH;
1911                                         else if (flags & IN6P_LOWPORT)
1912                                                 optval = IPV6_PORTRANGE_LOW;
1913                                         else
1914                                                 optval = 0;
1915                                         break;
1916                                     }
1917                                 }
1918                                 soopt_from_kbuf(sopt, &optval,
1919                                         sizeof optval);
1920                                 break;
1921
1922                         case IPV6_PATHMTU:
1923                         {
1924                                 u_long pmtu = 0;
1925                                 struct ip6_mtuinfo mtuinfo;
1926                                 struct route_in6 sro;
1927
1928                                 bzero(&sro, sizeof(sro));
1929
1930                                 if (!(so->so_state & SS_ISCONNECTED))
1931                                         return (ENOTCONN);
1932                                 /*
1933                                  * XXX: we dot not consider the case of source
1934                                  * routing, or optional information to specify
1935                                  * the outgoing interface.
1936                                  */
1937                                 error = ip6_getpmtu(&sro, NULL, NULL,
1938                                     &in6p->in6p_faddr, &pmtu, NULL);
1939                                 if (sro.ro_rt)
1940                                         RTFREE(sro.ro_rt);
1941                                 if (error)
1942                                         break;
1943                                 if (pmtu > IPV6_MAXPACKET)
1944                                         pmtu = IPV6_MAXPACKET;
1945
1946                                 bzero(&mtuinfo, sizeof(mtuinfo));
1947                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1948                                 optdata = (void *)&mtuinfo;
1949                                 optdatalen = sizeof(mtuinfo);
1950                                 soopt_from_kbuf(sopt, optdata,
1951                                     optdatalen);
1952                                 break;
1953                         }
1954
1955                         case IPV6_2292PKTINFO:
1956                         case IPV6_2292HOPLIMIT:
1957                         case IPV6_2292HOPOPTS:
1958                         case IPV6_2292RTHDR:
1959                         case IPV6_2292DSTOPTS:
1960                                 if (optname == IPV6_2292HOPOPTS ||
1961                                     optname == IPV6_2292DSTOPTS ||
1962                                     !privileged)
1963                                         return (EPERM);
1964                                 switch (optname) {
1965                                 case IPV6_2292PKTINFO:
1966                                         optval = OPTBIT(IN6P_PKTINFO);
1967                                         break;
1968                                 case IPV6_2292HOPLIMIT:
1969                                         optval = OPTBIT(IN6P_HOPLIMIT);
1970                                         break;
1971                                 case IPV6_2292HOPOPTS:
1972                                         if (!privileged)
1973                                                 return (EPERM);
1974                                         optval = OPTBIT(IN6P_HOPOPTS);
1975                                         break;
1976                                 case IPV6_2292RTHDR:
1977                                         optval = OPTBIT(IN6P_RTHDR);
1978                                         break;
1979                                 case IPV6_2292DSTOPTS:
1980                                         if (!privileged)
1981                                                 return (EPERM);
1982                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1983                                         break;
1984                                 }
1985                                 soopt_from_kbuf(sopt, &optval,
1986                                         sizeof optval);
1987                                 break;
1988
1989                         case IPV6_PKTINFO:
1990                         case IPV6_HOPOPTS:
1991                         case IPV6_RTHDR:
1992                         case IPV6_DSTOPTS:
1993                         case IPV6_RTHDRDSTOPTS:
1994                         case IPV6_NEXTHOP:
1995                         case IPV6_TCLASS:
1996                         case IPV6_DONTFRAG:
1997                         case IPV6_USE_MIN_MTU:
1998                         case IPV6_PREFER_TEMPADDR:
1999                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2000                                     optname, sopt);
2001                                 break;
2002
2003                         case IPV6_MULTICAST_IF:
2004                         case IPV6_MULTICAST_HOPS:
2005                         case IPV6_MULTICAST_LOOP:
2006                         case IPV6_JOIN_GROUP:
2007                         case IPV6_LEAVE_GROUP:
2008                             {
2009                                 struct mbuf *m;
2010                                 error = ip6_getmoptions(sopt->sopt_name,
2011                                                 in6p->in6p_moptions, &m);
2012                                 if (error == 0)
2013                                         soopt_from_kbuf(sopt,
2014                                                 mtod(m, char *), m->m_len);
2015                                 m_freem(m);
2016                             }
2017                                 break;
2018
2019 #if defined(IPSEC) || defined(FAST_IPSEC)
2020                         case IPV6_IPSEC_POLICY:
2021                           {
2022                                 caddr_t req = NULL;
2023                                 size_t len = 0;
2024                                 struct mbuf *m = NULL;
2025                                 struct mbuf **mp = &m;
2026
2027                                 error = soopt_getm(sopt, &m); /* XXX */
2028                                 if (error != 0)
2029                                         break;
2030                                 soopt_to_mbuf(sopt, m); /* XXX */
2031                                 if (m) {
2032                                         req = mtod(m, caddr_t);
2033                                         len = m->m_len;
2034                                 }
2035                                 error = ipsec6_get_policy(in6p, req, len, mp);
2036                                 if (error == 0)
2037                                         error = soopt_from_mbuf(sopt, m); /*XXX*/
2038                                 if (error == 0 && m != NULL)
2039                                         m_freem(m);
2040                                 break;
2041                           }
2042 #endif /* KAME IPSEC */
2043
2044                         case IPV6_FW_GET:
2045                           {
2046                                 struct mbuf *m;
2047                                 struct mbuf **mp = &m;
2048
2049                                 if (ip6_fw_ctl_ptr == NULL)
2050                                 {
2051                                         return EINVAL;
2052                                 }
2053                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
2054                                 if (error == 0)
2055                                         error = soopt_from_mbuf(sopt, m); /* XXX */
2056                                 if (error == 0 && m != NULL)
2057                                         m_freem(m);
2058                           }
2059                                 break;
2060
2061                         default:
2062                                 error = ENOPROTOOPT;
2063                                 break;
2064                         }
2065                         break;
2066                 }
2067         } else {
2068                 error = EINVAL;
2069         }
2070         return (error);
2071 }
2072
2073 int
2074 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2075 {
2076         int error = 0, optval, optlen;
2077         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2078         struct in6pcb *in6p = sotoin6pcb(so);
2079         int level, op, optname;
2080
2081         if (sopt) {
2082                 level = sopt->sopt_level;
2083                 op = sopt->sopt_dir;
2084                 optname = sopt->sopt_name;
2085                 optlen = sopt->sopt_valsize;
2086         } else
2087                 panic("ip6_raw_ctloutput: arg soopt is NULL");
2088
2089         if (level != IPPROTO_IPV6) {
2090                 return (EINVAL);
2091         }
2092
2093         switch (optname) {
2094         case IPV6_CHECKSUM:
2095                 /*
2096                  * For ICMPv6 sockets, no modification allowed for checksum
2097                  * offset, permit "no change" values to help existing apps.
2098                  *
2099                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2100                  * for an ICMPv6 socket will fail."
2101                  * The current behavior does not meet RFC3542.
2102                  */
2103                 switch (op) {
2104                 case SOPT_SET:
2105                         if (optlen != sizeof(int)) {
2106                                 error = EINVAL;
2107                                 break;
2108                         }
2109                         error = soopt_to_kbuf(sopt, &optval,
2110                                     sizeof optval, sizeof optval);
2111                         if (error)
2112                                 break;
2113                         if ((optval % 2) != 0) {
2114                                 /* the API assumes even offset values */
2115                                 error = EINVAL;
2116                         } else if (so->so_proto->pr_protocol ==
2117                             IPPROTO_ICMPV6) {
2118                                 if (optval != icmp6off)
2119                                         error = EINVAL;
2120                         } else
2121                                 in6p->in6p_cksum = optval;
2122                         break;
2123
2124                 case SOPT_GET:
2125                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2126                                 optval = icmp6off;
2127                         else
2128                                 optval = in6p->in6p_cksum;
2129
2130                         soopt_from_kbuf(sopt, &optval, sizeof(optval));
2131                         break;
2132
2133                 default:
2134                         error = EINVAL;
2135                         break;
2136                 }
2137                 break;
2138
2139         default:
2140                 error = ENOPROTOOPT;
2141                 break;
2142         }
2143
2144         return (error);
2145 }
2146
2147 /*
2148  * Set up IP6 options in pcb for insertion in output packets or
2149  * specifying behavior of outgoing packets.
2150  */
2151 static int
2152 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2153     struct socket *so, struct sockopt *sopt)
2154 {
2155         int priv = 0;
2156         struct ip6_pktopts *opt = *pktopt;
2157         int error = 0;
2158
2159         /* turn off any old options. */
2160         if (opt) {
2161 #ifdef DIAGNOSTIC
2162                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2163                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2164                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2165                         kprintf("ip6_pcbopts: all specified options are cleared.\n");
2166 #endif
2167                 ip6_clearpktopts(opt, -1);
2168         } else
2169                 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2170         *pktopt = NULL;
2171
2172         if (!m || m->m_len == 0) {
2173                 /*
2174                  * Only turning off any previous options, regardless of
2175                  * whether the opt is just created or given.
2176                  */
2177                 kfree(opt, M_IP6OPT);
2178                 return (0);
2179         }
2180
2181         /*  set options specified by user. */
2182         if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) {
2183                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2184                 kfree(opt, M_IP6OPT);
2185                 return (error);
2186         }
2187         *pktopt = opt;
2188         return (0);
2189 }
2190
2191
2192 /*
2193  * Below three functions are introduced by merge to RFC3542
2194  */
2195
2196 static int
2197 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2198 {
2199         void *optdata = NULL;
2200         int optdatalen = 0;
2201         struct ip6_ext *ip6e;
2202         int error = 0;
2203         struct in6_pktinfo null_pktinfo;
2204         int deftclass = 0, on;
2205         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2206         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2207
2208         switch (optname) {
2209         case IPV6_PKTINFO:
2210                 if (pktopt && pktopt->ip6po_pktinfo)
2211                         optdata = (void *)pktopt->ip6po_pktinfo;
2212                 else {
2213                         /* XXX: we don't have to do this every time... */
2214                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2215                         optdata = (void *)&null_pktinfo;
2216                 }
2217                 optdatalen = sizeof(struct in6_pktinfo);
2218                 break;
2219         case IPV6_TCLASS:
2220                 if (pktopt && pktopt->ip6po_tclass >= 0)
2221                         optdata = (void *)&pktopt->ip6po_tclass;
2222                 else
2223                         optdata = (void *)&deftclass;
2224                 optdatalen = sizeof(int);
2225                 break;
2226         case IPV6_HOPOPTS:
2227                 if (pktopt && pktopt->ip6po_hbh) {
2228                         optdata = (void *)pktopt->ip6po_hbh;
2229                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2230                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2231                 }
2232                 break;
2233         case IPV6_RTHDR:
2234                 if (pktopt && pktopt->ip6po_rthdr) {
2235                         optdata = (void *)pktopt->ip6po_rthdr;
2236                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2237                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2238                 }
2239                 break;
2240         case IPV6_RTHDRDSTOPTS:
2241                 if (pktopt && pktopt->ip6po_dest1) {
2242                         optdata = (void *)pktopt->ip6po_dest1;
2243                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2244                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2245                 }
2246                 break;
2247         case IPV6_DSTOPTS:
2248                 if (pktopt && pktopt->ip6po_dest2) {
2249                         optdata = (void *)pktopt->ip6po_dest2;
2250                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2251                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2252                 }
2253                 break;
2254         case IPV6_NEXTHOP:
2255                 if (pktopt && pktopt->ip6po_nexthop) {
2256                         optdata = (void *)pktopt->ip6po_nexthop;
2257                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2258                 }
2259                 break;
2260         case IPV6_USE_MIN_MTU:
2261                 if (pktopt)
2262                         optdata = (void *)&pktopt->ip6po_minmtu;
2263                 else
2264                         optdata = (void *)&defminmtu;
2265                 optdatalen = sizeof(int);
2266                 break;
2267         case IPV6_DONTFRAG:
2268                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2269                         on = 1;
2270                 else
2271                         on = 0;
2272                 optdata = (void *)&on;
2273                 optdatalen = sizeof(on);
2274                 break;
2275         case IPV6_PREFER_TEMPADDR:
2276                 if (pktopt)
2277                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2278                 else
2279                         optdata = (void *)&defpreftemp;
2280                 optdatalen = sizeof(int);
2281                 break;
2282         default:                /* should not happen */
2283 #ifdef DIAGNOSTIC
2284                 panic("ip6_getpcbopt: unexpected option\n");
2285 #endif
2286                 return (ENOPROTOOPT);
2287         }
2288
2289         soopt_from_kbuf(sopt, optdata, optdatalen);
2290
2291         return (error);
2292 }
2293
2294 /*
2295  * initialize ip6_pktopts.  beware that there are non-zero default values in
2296  * the struct.
2297  */
2298
2299 static int
2300 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto)
2301 {
2302         struct ip6_pktopts *opt;
2303         int priv =0;
2304         if (*pktopt == NULL) {
2305                 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2306                 init_ip6pktopts(*pktopt);
2307         }
2308         opt = *pktopt;
2309
2310         return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv));
2311 }
2312
2313 /*
2314  * initialize ip6_pktopts.  beware that there are non-zero default values in
2315  * the struct.
2316  */
2317 void
2318 init_ip6pktopts(struct ip6_pktopts *opt)
2319 {
2320
2321         bzero(opt, sizeof(*opt));
2322         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2323         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2324         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2325         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2326 }
2327
2328 void
2329 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2330 {
2331         if (pktopt == NULL)
2332                 return;
2333
2334         if (optname == -1 || optname == IPV6_PKTINFO) {
2335                 if (pktopt->ip6po_pktinfo)
2336                         kfree(pktopt->ip6po_pktinfo, M_IP6OPT);
2337                 pktopt->ip6po_pktinfo = NULL;
2338         }
2339         if (optname == -1 || optname == IPV6_HOPLIMIT)
2340                 pktopt->ip6po_hlim = -1;
2341         if (optname == -1 || optname == IPV6_TCLASS)
2342                 pktopt->ip6po_tclass = -1;
2343         if (optname == -1 || optname == IPV6_NEXTHOP) {
2344                 if (pktopt->ip6po_nextroute.ro_rt) {
2345                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2346                         pktopt->ip6po_nextroute.ro_rt = NULL;
2347                 }
2348                 if (pktopt->ip6po_nexthop)
2349                         kfree(pktopt->ip6po_nexthop, M_IP6OPT);
2350                 pktopt->ip6po_nexthop = NULL;
2351         }
2352         if (optname == -1 || optname == IPV6_HOPOPTS) {
2353                 if (pktopt->ip6po_hbh)
2354                         kfree(pktopt->ip6po_hbh, M_IP6OPT);
2355                 pktopt->ip6po_hbh = NULL;
2356         }
2357         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2358                 if (pktopt->ip6po_dest1)
2359                         kfree(pktopt->ip6po_dest1, M_IP6OPT);
2360                 pktopt->ip6po_dest1 = NULL;
2361         }
2362         if (optname == -1 || optname == IPV6_RTHDR) {
2363                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2364                         kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2365                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2366                 if (pktopt->ip6po_route.ro_rt) {
2367                         RTFREE(pktopt->ip6po_route.ro_rt);
2368                         pktopt->ip6po_route.ro_rt = NULL;
2369                 }
2370         }
2371         if (optname == -1 || optname == IPV6_DSTOPTS) {
2372                 if (pktopt->ip6po_dest2)
2373                         kfree(pktopt->ip6po_dest2, M_IP6OPT);
2374                 pktopt->ip6po_dest2 = NULL;
2375         }
2376 }
2377
2378 #define PKTOPT_EXTHDRCPY(type) \
2379 do {\
2380         if (src->type) {\
2381                 int hlen =\
2382                         (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2383                 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\
2384                 if (dst->type == NULL)\
2385                         goto bad;\
2386                 bcopy(src->type, dst->type, hlen);\
2387         }\
2388 } while (0)
2389
2390 struct ip6_pktopts *
2391 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2392 {
2393         struct ip6_pktopts *dst;
2394
2395         if (src == NULL) {
2396                 kprintf("ip6_clearpktopts: invalid argument\n");
2397                 return (NULL);
2398         }
2399
2400         dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO);
2401         if (dst == NULL)
2402                 return (NULL);
2403
2404         dst->ip6po_hlim = src->ip6po_hlim;
2405         if (src->ip6po_pktinfo) {
2406                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2407                                             M_IP6OPT, canwait);
2408                 if (dst->ip6po_pktinfo == NULL)
2409                         goto bad;
2410                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2411         }
2412         if (src->ip6po_nexthop) {
2413                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2414                                             M_IP6OPT, canwait);
2415                 if (dst->ip6po_nexthop == NULL)
2416                         goto bad;
2417                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2418                       src->ip6po_nexthop->sa_len);
2419         }
2420         PKTOPT_EXTHDRCPY(ip6po_hbh);
2421         PKTOPT_EXTHDRCPY(ip6po_dest1);
2422         PKTOPT_EXTHDRCPY(ip6po_dest2);
2423         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2424         return (dst);
2425
2426 bad:
2427         if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT);
2428         if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT);
2429         if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT);
2430         if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT);
2431         if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT);
2432         if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT);
2433         kfree(dst, M_IP6OPT);
2434         return (NULL);
2435 }
2436
2437 static int
2438 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2439 {
2440         if (dst == NULL || src == NULL)  {
2441 #ifdef DIAGNOSTIC
2442                 kprintf("ip6_clearpktopts: invalid argument\n");
2443 #endif
2444                 return (EINVAL);
2445         }
2446
2447         dst->ip6po_hlim = src->ip6po_hlim;
2448         dst->ip6po_tclass = src->ip6po_tclass;
2449         dst->ip6po_flags = src->ip6po_flags;
2450         if (src->ip6po_pktinfo) {
2451                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2452                     M_IP6OPT, canwait);
2453                 if (dst->ip6po_pktinfo == NULL)
2454                         goto bad;
2455                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2456         }
2457         if (src->ip6po_nexthop) {
2458                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2459                     M_IP6OPT, canwait);
2460                 if (dst->ip6po_nexthop == NULL)
2461                         goto bad;
2462                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2463                     src->ip6po_nexthop->sa_len);
2464         }
2465         PKTOPT_EXTHDRCPY(ip6po_hbh);
2466         PKTOPT_EXTHDRCPY(ip6po_dest1);
2467         PKTOPT_EXTHDRCPY(ip6po_dest2);
2468         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2469         return (0);
2470
2471   bad:
2472         ip6_clearpktopts(dst, -1);
2473         return (ENOBUFS);
2474 }
2475 #undef PKTOPT_EXTHDRCPY
2476
2477 void
2478 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2479 {
2480         if (pktopt == NULL)
2481                 return;
2482
2483         ip6_clearpktopts(pktopt, -1);
2484
2485         kfree(pktopt, M_IP6OPT);
2486 }
2487
2488 /*
2489  * Set the IP6 multicast options in response to user setsockopt().
2490  */
2491 static int
2492 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2493 {
2494         int error = 0;
2495         u_int loop, ifindex;
2496         struct ipv6_mreq *mreq;
2497         struct ifnet *ifp;
2498         struct ip6_moptions *im6o = *im6op;
2499         struct route_in6 ro;
2500         struct sockaddr_in6 *dst;
2501         struct in6_multi_mship *imm;
2502         struct thread *td = curthread;  /* XXX */
2503
2504         if (im6o == NULL) {
2505                 /*
2506                  * No multicast option buffer attached to the pcb;
2507                  * allocate one and initialize to default values.
2508                  */
2509                 im6o = (struct ip6_moptions *)
2510                         kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2511
2512                 *im6op = im6o;
2513                 im6o->im6o_multicast_ifp = NULL;
2514                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2515                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2516                 LIST_INIT(&im6o->im6o_memberships);
2517         }
2518
2519         switch (optname) {
2520
2521         case IPV6_MULTICAST_IF:
2522                 /*
2523                  * Select the interface for outgoing multicast packets.
2524                  */
2525                 if (m == NULL || m->m_len != sizeof(u_int)) {
2526                         error = EINVAL;
2527                         break;
2528                 }
2529                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2530                 if (ifindex < 0 || if_index < ifindex) {
2531                         error = ENXIO;  /* XXX EINVAL? */
2532                         break;
2533                 }
2534                 ifp = ifindex2ifnet[ifindex];
2535                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2536                         error = EADDRNOTAVAIL;
2537                         break;
2538                 }
2539                 im6o->im6o_multicast_ifp = ifp;
2540                 break;
2541
2542         case IPV6_MULTICAST_HOPS:
2543             {
2544                 /*
2545                  * Set the IP6 hoplimit for outgoing multicast packets.
2546                  */
2547                 int optval;
2548                 if (m == NULL || m->m_len != sizeof(int)) {
2549                         error = EINVAL;
2550                         break;
2551                 }
2552                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2553                 if (optval < -1 || optval >= 256)
2554                         error = EINVAL;
2555                 else if (optval == -1)
2556                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2557                 else
2558                         im6o->im6o_multicast_hlim = optval;
2559                 break;
2560             }
2561
2562         case IPV6_MULTICAST_LOOP:
2563                 /*
2564                  * Set the loopback flag for outgoing multicast packets.
2565                  * Must be zero or one.
2566                  */
2567                 if (m == NULL || m->m_len != sizeof(u_int)) {
2568                         error = EINVAL;
2569                         break;
2570                 }
2571                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2572                 if (loop > 1) {
2573                         error = EINVAL;
2574                         break;
2575                 }
2576                 im6o->im6o_multicast_loop = loop;
2577                 break;
2578
2579         case IPV6_JOIN_GROUP:
2580                 /*
2581                  * Add a multicast group membership.
2582                  * Group must be a valid IP6 multicast address.
2583                  */
2584                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2585                         error = EINVAL;
2586                         break;
2587                 }
2588                 mreq = mtod(m, struct ipv6_mreq *);
2589                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2590                         /*
2591                          * We use the unspecified address to specify to accept
2592                          * all multicast addresses. Only super user is allowed
2593                          * to do this.
2594                          */
2595                         if (suser(td))
2596                         {
2597                                 error = EACCES;
2598                                 break;
2599                         }
2600                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2601                         error = EINVAL;
2602                         break;
2603                 }
2604
2605                 /*
2606                  * If the interface is specified, validate it.
2607                  */
2608                 if (mreq->ipv6mr_interface < 0
2609                  || if_index < mreq->ipv6mr_interface) {
2610                         error = ENXIO;  /* XXX EINVAL? */
2611                         break;
2612                 }
2613                 /*
2614                  * If no interface was explicitly specified, choose an
2615                  * appropriate one according to the given multicast address.
2616                  */
2617                 if (mreq->ipv6mr_interface == 0) {
2618                         /*
2619                          * If the multicast address is in node-local scope,
2620                          * the interface should be a loopback interface.
2621                          * Otherwise, look up the routing table for the
2622                          * address, and choose the outgoing interface.
2623                          *   XXX: is it a good approach?
2624                          */
2625                         if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2626                                 ifp = &loif[0];
2627                         } else {
2628                                 ro.ro_rt = NULL;
2629                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2630                                 bzero(dst, sizeof(*dst));
2631                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2632                                 dst->sin6_family = AF_INET6;
2633                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2634                                 rtalloc((struct route *)&ro);
2635                                 if (ro.ro_rt == NULL) {
2636                                         error = EADDRNOTAVAIL;
2637                                         break;
2638                                 }
2639                                 ifp = ro.ro_rt->rt_ifp;
2640                                 rtfree(ro.ro_rt);
2641                         }
2642                 } else
2643                         ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2644
2645                 /*
2646                  * See if we found an interface, and confirm that it
2647                  * supports multicast
2648                  */
2649                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2650                         error = EADDRNOTAVAIL;
2651                         break;
2652                 }
2653                 /*
2654                  * Put interface index into the multicast address,
2655                  * if the address has link-local scope.
2656                  */
2657                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2658                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2659                                 = htons(mreq->ipv6mr_interface);
2660                 }
2661                 /*
2662                  * See if the membership already exists.
2663                  */
2664                 for (imm = im6o->im6o_memberships.lh_first;
2665                      imm != NULL; imm = imm->i6mm_chain.le_next)
2666                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2667                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2668                                                &mreq->ipv6mr_multiaddr))
2669                                 break;
2670                 if (imm != NULL) {
2671                         error = EADDRINUSE;
2672                         break;
2673                 }
2674                 /*
2675                  * Everything looks good; add a new record to the multicast
2676                  * address list for the given interface.
2677                  */
2678                 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2679                 if ((imm->i6mm_maddr =
2680                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2681                         kfree(imm, M_IPMADDR);
2682                         break;
2683                 }
2684                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2685                 break;
2686
2687         case IPV6_LEAVE_GROUP:
2688                 /*
2689                  * Drop a multicast group membership.
2690                  * Group must be a valid IP6 multicast address.
2691                  */
2692                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2693                         error = EINVAL;
2694                         break;
2695                 }
2696                 mreq = mtod(m, struct ipv6_mreq *);
2697                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2698                         if (suser(td)) {
2699                                 error = EACCES;
2700                                 break;
2701                         }
2702                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2703                         error = EINVAL;
2704                         break;
2705                 }
2706                 /*
2707                  * If an interface address was specified, get a pointer
2708                  * to its ifnet structure.
2709                  */
2710                 if (mreq->ipv6mr_interface < 0
2711                  || if_index < mreq->ipv6mr_interface) {
2712                         error = ENXIO;  /* XXX EINVAL? */
2713                         break;
2714                 }
2715                 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2716                 /*
2717                  * Put interface index into the multicast address,
2718                  * if the address has link-local scope.
2719                  */
2720                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2721                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2722                                 = htons(mreq->ipv6mr_interface);
2723                 }
2724                 /*
2725                  * Find the membership in the membership list.
2726                  */
2727                 for (imm = im6o->im6o_memberships.lh_first;
2728                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2729                         if ((ifp == NULL ||
2730                              imm->i6mm_maddr->in6m_ifp == ifp) &&
2731                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2732                                                &mreq->ipv6mr_multiaddr))
2733                                 break;
2734                 }
2735                 if (imm == NULL) {
2736                         /* Unable to resolve interface */
2737                         error = EADDRNOTAVAIL;
2738                         break;
2739                 }
2740                 /*
2741                  * Give up the multicast address record to which the
2742                  * membership points.
2743                  */
2744                 LIST_REMOVE(imm, i6mm_chain);
2745                 in6_delmulti(imm->i6mm_maddr);
2746                 kfree(imm, M_IPMADDR);
2747                 break;
2748
2749         default:
2750                 error = EOPNOTSUPP;
2751                 break;
2752         }
2753
2754         /*
2755          * If all options have default values, no need to keep the mbuf.
2756          */
2757         if (im6o->im6o_multicast_ifp == NULL &&
2758             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2759             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2760             im6o->im6o_memberships.lh_first == NULL) {
2761                 kfree(*im6op, M_IPMOPTS);
2762                 *im6op = NULL;
2763         }
2764
2765         return (error);
2766 }
2767
2768 /*
2769  * Return the IP6 multicast options in response to user getsockopt().
2770  */
2771 static int
2772 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2773 {
2774         u_int *hlim, *loop, *ifindex;
2775
2776         *mp = m_get(MB_WAIT, MT_HEADER);                /* XXX */
2777
2778         switch (optname) {
2779
2780         case IPV6_MULTICAST_IF:
2781                 ifindex = mtod(*mp, u_int *);
2782                 (*mp)->m_len = sizeof(u_int);
2783                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2784                         *ifindex = 0;
2785                 else
2786                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2787                 return (0);
2788
2789         case IPV6_MULTICAST_HOPS:
2790                 hlim = mtod(*mp, u_int *);
2791                 (*mp)->m_len = sizeof(u_int);
2792                 if (im6o == NULL)
2793                         *hlim = ip6_defmcasthlim;
2794                 else
2795                         *hlim = im6o->im6o_multicast_hlim;
2796                 return (0);
2797
2798         case IPV6_MULTICAST_LOOP:
2799                 loop = mtod(*mp, u_int *);
2800                 (*mp)->m_len = sizeof(u_int);
2801                 if (im6o == NULL)
2802                         *loop = ip6_defmcasthlim;
2803                 else
2804                         *loop = im6o->im6o_multicast_loop;
2805                 return (0);
2806
2807         default:
2808                 return (EOPNOTSUPP);
2809         }
2810 }
2811
2812 /*
2813  * Discard the IP6 multicast options.
2814  */
2815 void
2816 ip6_freemoptions(struct ip6_moptions *im6o)
2817 {
2818         struct in6_multi_mship *imm;
2819
2820         if (im6o == NULL)
2821                 return;
2822
2823         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2824                 LIST_REMOVE(imm, i6mm_chain);
2825                 if (imm->i6mm_maddr)
2826                         in6_delmulti(imm->i6mm_maddr);
2827                 kfree(imm, M_IPMADDR);
2828         }
2829         kfree(im6o, M_IPMOPTS);
2830 }
2831
2832 /*
2833  * Set a particular packet option, as a sticky option or an ancillary data
2834  * item.  "len" can be 0 only when it's a sticky option.
2835  * We have 4 cases of combination of "sticky" and "cmsg":
2836  * "sticky=0, cmsg=0": impossible
2837  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2838  * "sticky=1, cmsg=0": RFC3542 socket option
2839  * "sticky=1, cmsg=1": RFC2292 socket option
2840  */
2841 static int
2842 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2843      int sticky, int cmsg, int uproto, int priv)
2844 {
2845         int minmtupolicy, preftemp;
2846         //int error;
2847
2848         if (!sticky && !cmsg) {
2849                 kprintf("ip6_setpktoption: impossible case\n");
2850                 return (EINVAL);
2851         }
2852
2853         /*
2854          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2855          * not be specified in the context of RFC3542.  Conversely,
2856          * RFC3542 types should not be specified in the context of RFC2292.
2857          */
2858         if (!cmsg) {
2859                 switch (optname) {
2860                 case IPV6_2292PKTINFO:
2861                 case IPV6_2292HOPLIMIT:
2862                 case IPV6_2292NEXTHOP:
2863                 case IPV6_2292HOPOPTS:
2864                 case IPV6_2292DSTOPTS:
2865                 case IPV6_2292RTHDR:
2866                 case IPV6_2292PKTOPTIONS:
2867                         return (ENOPROTOOPT);
2868                 }
2869         }
2870         if (sticky && cmsg) {
2871                 switch (optname) {
2872                 case IPV6_PKTINFO:
2873                 case IPV6_HOPLIMIT:
2874                 case IPV6_NEXTHOP:
2875                 case IPV6_HOPOPTS:
2876                 case IPV6_DSTOPTS:
2877                 case IPV6_RTHDRDSTOPTS:
2878                 case IPV6_RTHDR:
2879                 case IPV6_USE_MIN_MTU:
2880                 case IPV6_DONTFRAG:
2881                 case IPV6_TCLASS:
2882                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2883                         return (ENOPROTOOPT);
2884                 }
2885         }
2886
2887         switch (optname) {
2888         case IPV6_2292PKTINFO:
2889         case IPV6_PKTINFO:
2890         {
2891                 struct in6_pktinfo *pktinfo;
2892                 if (len != sizeof(struct in6_pktinfo))
2893                         return (EINVAL);
2894                 pktinfo = (struct in6_pktinfo *)buf;
2895
2896                 /*
2897                  * An application can clear any sticky IPV6_PKTINFO option by
2898                  * doing a "regular" setsockopt with ipi6_addr being
2899                  * in6addr_any and ipi6_ifindex being zero.
2900                  * [RFC 3542, Section 6]
2901                  */
2902                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2903                     pktinfo->ipi6_ifindex == 0 &&
2904                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2905                         ip6_clearpktopts(opt, optname);
2906                         break;
2907                 }
2908
2909                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2910                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2911                         return (EINVAL);
2912                 }
2913
2914                 /* validate the interface index if specified. */
2915                 if (pktinfo->ipi6_ifindex > if_index ||
2916                     pktinfo->ipi6_ifindex < 0) {
2917                          return (ENXIO);
2918                 }
2919                 /*
2920                  * Check if the requested source address is indeed a
2921                  * unicast address assigned to the node, and can be
2922                  * used as the packet's source address.
2923                  */
2924                 if (opt->ip6po_pktinfo != NULL &&
2925                     !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2926                         struct in6_ifaddr *ia6;
2927                         struct sockaddr_in6 sin6;
2928
2929                         bzero(&sin6, sizeof(sin6));
2930                         sin6.sin6_len = sizeof(sin6);
2931                         sin6.sin6_family = AF_INET6;
2932                         sin6.sin6_addr =
2933                         opt->ip6po_pktinfo->ipi6_addr;
2934                         ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2935                         if (ia6 == NULL ||
2936                                 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2937                                         IN6_IFF_NOTREADY)) != 0)
2938                         return (EADDRNOTAVAIL);
2939                 }
2940
2941                 /*
2942                  * We store the address anyway, and let in6_selectsrc()
2943                  * validate the specified address.  This is because ipi6_addr
2944                  * may not have enough information about its scope zone, and
2945                  * we may need additional information (such as outgoing
2946                  * interface or the scope zone of a destination address) to
2947                  * disambiguate the scope.
2948                  * XXX: the delay of the validation may confuse the
2949                  * application when it is used as a sticky option.
2950                  */
2951                 if (opt->ip6po_pktinfo == NULL) {
2952                         opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo),
2953                             M_IP6OPT, M_NOWAIT);
2954                         if (opt->ip6po_pktinfo == NULL)
2955                                 return (ENOBUFS);
2956                 }
2957                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2958                 break;
2959         }
2960
2961         case IPV6_2292HOPLIMIT:
2962         case IPV6_HOPLIMIT:
2963         {
2964                 int *hlimp;
2965
2966                 /*
2967                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2968                  * to simplify the ordering among hoplimit options.
2969                  */
2970                 if (optname == IPV6_HOPLIMIT && sticky)
2971                         return (ENOPROTOOPT);
2972
2973                 if (len != sizeof(int))
2974                         return (EINVAL);
2975                 hlimp = (int *)buf;
2976                 if (*hlimp < -1 || *hlimp > 255)
2977                         return (EINVAL);
2978
2979                 opt->ip6po_hlim = *hlimp;
2980                 break;
2981         }
2982
2983         case IPV6_TCLASS:
2984         {
2985                 int tclass;
2986
2987                 if (len != sizeof(int))
2988                         return (EINVAL);
2989                 tclass = *(int *)buf;
2990                 if (tclass < -1 || tclass > 255)
2991                         return (EINVAL);
2992
2993                 opt->ip6po_tclass = tclass;
2994                 break;
2995         }
2996
2997         case IPV6_2292NEXTHOP:
2998         case IPV6_NEXTHOP:
2999                 if (!priv)
3000                         return (EPERM);
3001
3002                 if (len == 0) { /* just remove the option */
3003                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3004                         break;
3005                 }
3006
3007                 /* check if cmsg_len is large enough for sa_len */
3008                 if (len < sizeof(struct sockaddr) || len < *buf)
3009                         return (EINVAL);
3010
3011                 switch (((struct sockaddr *)buf)->sa_family) {
3012                 case AF_INET6:
3013                 {
3014                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3015                         //int error;
3016
3017                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3018                                 return (EINVAL);
3019
3020                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3021                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3022                                 return (EINVAL);
3023                         }
3024                         break;
3025                 }
3026                 case AF_LINK:   /* should eventually be supported */
3027                 default:
3028                         return (EAFNOSUPPORT);
3029                 }
3030
3031                 /* turn off the previous option, then set the new option. */
3032                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3033                 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT);
3034                 if (opt->ip6po_nexthop == NULL)
3035                         return (ENOBUFS);
3036                 bcopy(buf, opt->ip6po_nexthop, *buf);
3037                 break;
3038
3039         case IPV6_2292HOPOPTS:
3040         case IPV6_HOPOPTS:
3041         {
3042                 struct ip6_hbh *hbh;
3043                 int hbhlen;
3044
3045                 /*
3046                  * XXX: We don't allow a non-privileged user to set ANY HbH
3047                  * options, since per-option restriction has too much
3048                  * overhead.
3049                  */
3050                 if (!priv)
3051                         return (EPERM);
3052                 if (len == 0) {
3053                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3054                         break;  /* just remove the option */
3055                 }
3056
3057                 /* message length validation */
3058                 if (len < sizeof(struct ip6_hbh))
3059                         return (EINVAL);
3060                 hbh = (struct ip6_hbh *)buf;
3061                 hbhlen = (hbh->ip6h_len + 1) << 3;
3062                 if (len != hbhlen)
3063                         return (EINVAL);
3064
3065                 /* turn off the previous option, then set the new option. */
3066                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3067                 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT);
3068                 if (opt->ip6po_hbh == NULL)
3069                         return (ENOBUFS);
3070                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3071
3072                 break;
3073         }
3074
3075         case IPV6_2292DSTOPTS:
3076         case IPV6_DSTOPTS:
3077         case IPV6_RTHDRDSTOPTS:
3078         {
3079                 struct ip6_dest *dest, **newdest = NULL;
3080                 int destlen;
3081                 if (!priv)
3082                         return (EPERM);
3083
3084                 if (len == 0) {
3085                         ip6_clearpktopts(opt, optname);
3086                         break;  /* just remove the option */
3087                 }
3088
3089                 /* message length validation */
3090                 if (len < sizeof(struct ip6_dest))
3091                         return (EINVAL);
3092                 dest = (struct ip6_dest *)buf;
3093                 destlen = (dest->ip6d_len + 1) << 3;
3094                 if (len != destlen)
3095                         return (EINVAL);
3096
3097                 /*
3098                  * Determine the position that the destination options header
3099                  * should be inserted; before or after the routing header.
3100                  */
3101                 switch (optname) {
3102                 case IPV6_2292DSTOPTS:
3103                         /*
3104                          * The old advacned API is ambiguous on this point.
3105                          * Our approach is to determine the position based
3106                          * according to the existence of a routing header.
3107                          * Note, however, that this depends on the order of the
3108                          * extension headers in the ancillary data; the 1st
3109                          * part of the destination options header must appear
3110                          * before the routing header in the ancillary data,
3111                          * too.
3112                          * RFC3542 solved the ambiguity by introducing
3113                          * separate ancillary data or option types.
3114                          */
3115                         if (opt->ip6po_rthdr == NULL)
3116                                 newdest = &opt->ip6po_dest1;
3117                         else
3118                                 newdest = &opt->ip6po_dest2;
3119                         break;
3120                 case IPV6_RTHDRDSTOPTS:
3121                         newdest = &opt->ip6po_dest1;
3122                         break;
3123                 case IPV6_DSTOPTS:
3124                         newdest = &opt->ip6po_dest2;
3125                         break;
3126                 }
3127
3128                 /* turn off the previous option, then set the new option. */
3129                 ip6_clearpktopts(opt, optname);
3130                 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT);
3131                 if (*newdest == NULL)
3132                         return (ENOBUFS);
3133                 bcopy(dest, *newdest, destlen);
3134
3135                 break;
3136         }
3137
3138         case IPV6_2292RTHDR:
3139         case IPV6_RTHDR:
3140         {
3141                 struct ip6_rthdr *rth;
3142                 int rthlen;
3143
3144                 if (len == 0) {
3145                         ip6_clearpktopts(opt, IPV6_RTHDR);
3146                         break;  /* just remove the option */
3147                 }
3148
3149                 /* message length validation */
3150                 if (len < sizeof(struct ip6_rthdr))
3151                         return (EINVAL);
3152                 rth = (struct ip6_rthdr *)buf;
3153                 rthlen = (rth->ip6r_len + 1) << 3;
3154                 if (len != rthlen)
3155                         return (EINVAL);
3156
3157                 switch (rth->ip6r_type) {
3158                 default:
3159                         return (EINVAL);        /* not supported */
3160                 }
3161
3162                 /* turn off the previous option */
3163                 ip6_clearpktopts(opt, IPV6_RTHDR);
3164                 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT);
3165                 if (opt->ip6po_rthdr == NULL)
3166                         return (ENOBUFS);
3167                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3168
3169                 break;
3170         }
3171
3172         case IPV6_USE_MIN_MTU:
3173                 if (len != sizeof(int))
3174                         return (EINVAL);
3175                 minmtupolicy = *(int *)buf;
3176                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3177                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3178                     minmtupolicy != IP6PO_MINMTU_ALL) {
3179                         return (EINVAL);
3180                 }
3181                 opt->ip6po_minmtu = minmtupolicy;
3182                 break;
3183
3184         case IPV6_DONTFRAG:
3185                 if (len != sizeof(int))
3186                         return (EINVAL);
3187
3188                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3189                         /*
3190                          * we ignore this option for TCP sockets.
3191                          * (RFC3542 leaves this case unspecified.)
3192                          */
3193                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3194                 } else
3195                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3196                 break;
3197
3198         case IPV6_PREFER_TEMPADDR:
3199                 if (len != sizeof(int))
3200                         return (EINVAL);
3201                 preftemp = *(int *)buf;
3202                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3203                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3204                     preftemp != IP6PO_TEMPADDR_PREFER) {
3205                         return (EINVAL);
3206                 }
3207                 opt->ip6po_prefer_tempaddr = preftemp;
3208                 break;
3209
3210         default:
3211                 return (ENOPROTOOPT);
3212         } /* end of switch */
3213
3214         return (0);
3215 }
3216
3217
3218 /*
3219  * Set IPv6 outgoing packet options based on advanced API.
3220  */
3221 int
3222 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt,
3223     struct ip6_pktopts *stickyopt, int uproto, int priv)
3224 {
3225         struct cmsghdr *cm = NULL;
3226
3227         if (control == NULL || opt == NULL)
3228                 return (EINVAL);
3229
3230         init_ip6pktopts(opt);
3231
3232         /*
3233          * XXX: Currently, we assume all the optional information is stored
3234          * in a single mbuf.
3235          */
3236         if (stickyopt) {
3237                 int error;
3238
3239                 /*
3240                  * If stickyopt is provided, make a local copy of the options
3241                  * for this particular packet, then override them by ancillary
3242                  * objects.
3243                  * XXX: copypktopts() does not copy the cached route to a next
3244                  * hop (if any).  This is not very good in terms of efficiency,
3245                  * but we can allow this since this option should be rarely
3246                  * used.
3247                  */
3248                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
3249                         return (error);
3250         }
3251
3252         /*
3253          * XXX: Currently, we assume all the optional information is stored
3254          * in a single mbuf.
3255          */
3256         if (control->m_next)
3257                 return (EINVAL);
3258
3259         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3260             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3261                 int error;
3262
3263                 if (control->m_len < CMSG_LEN(0))
3264                         return (EINVAL);
3265
3266                 cm = mtod(control, struct cmsghdr *);
3267                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3268                         return (EINVAL);
3269                 if (cm->cmsg_level != IPPROTO_IPV6)
3270                         continue;
3271
3272                 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3273                     cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto, priv);
3274                 if (error)
3275                         return (error);
3276         }
3277
3278         return (0);
3279 }
3280
3281 /*
3282  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3283  * packet to the input queue of a specified interface.  Note that this
3284  * calls the output routine of the loopback "driver", but with an interface
3285  * pointer that might NOT be &loif -- easier than replicating that code here.
3286  */
3287 void
3288 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3289 {
3290         struct mbuf *copym;
3291         struct ip6_hdr *ip6;
3292
3293         copym = m_copy(m, 0, M_COPYALL);
3294         if (copym == NULL)
3295                 return;
3296
3297         /*
3298          * Make sure to deep-copy IPv6 header portion in case the data
3299          * is in an mbuf cluster, so that we can safely override the IPv6
3300          * header portion later.
3301          */
3302         if ((copym->m_flags & M_EXT) != 0 ||
3303             copym->m_len < sizeof(struct ip6_hdr)) {
3304                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3305                 if (copym == NULL)
3306                         return;
3307         }
3308
3309 #ifdef DIAGNOSTIC
3310         if (copym->m_len < sizeof(*ip6)) {
3311                 m_freem(copym);
3312                 return;
3313         }
3314 #endif
3315
3316         ip6 = mtod(copym, struct ip6_hdr *);
3317         /*
3318          * clear embedded scope identifiers if necessary.
3319          * in6_clearscope will touch the addresses only when necessary.
3320          */
3321         in6_clearscope(&ip6->ip6_src);
3322         in6_clearscope(&ip6->ip6_dst);
3323
3324         if_simloop(ifp, copym, dst->sin6_family, 0);
3325 }
3326
3327 /*
3328  * Separate the IPv6 header from the payload into its own mbuf.
3329  *
3330  * Returns the new mbuf chain or the original mbuf if no payload.
3331  * Returns NULL if can't allocate new mbuf for header.
3332  */
3333 static struct mbuf *
3334 ip6_splithdr(struct mbuf *m)
3335 {
3336         struct mbuf *mh;
3337
3338         if (m->m_len <= sizeof(struct ip6_hdr))         /* no payload */
3339                 return (m);
3340
3341         MGETHDR(mh, MB_DONTWAIT, MT_HEADER);
3342         if (mh == NULL)
3343                 return (NULL);
3344         mh->m_len = sizeof(struct ip6_hdr);
3345         M_MOVE_PKTHDR(mh, m);
3346         MH_ALIGN(mh, sizeof(struct ip6_hdr));
3347         bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr));
3348         m->m_data += sizeof(struct ip6_hdr);
3349         m->m_len -= sizeof(struct ip6_hdr);
3350         mh->m_next = m;
3351         return (mh);
3352 }
3353
3354 /*
3355  * Compute IPv6 extension header length.
3356  */
3357 int
3358 ip6_optlen(struct in6pcb *in6p)
3359 {
3360         int len;
3361
3362         if (!in6p->in6p_outputopts)
3363                 return 0;
3364
3365         len = 0;
3366 #define elen(x) \
3367     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3368
3369         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3370         if (in6p->in6p_outputopts->ip6po_rthdr)
3371                 /* dest1 is valid with rthdr only */
3372                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3373         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3374         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3375         return len;
3376 #undef elen
3377 }