sys/netinet6/ip6_output.c

   1 /*      $FreeBSD: src/sys/netinet6/ip6_output.c,v 1.13.2.18 2003/01/24 05:11:35 sam Exp $       */
   2 /*      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $    */
   3
   4 /*
   5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   6  * All rights reserved.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the name of the project nor the names of its contributors
  17  *    may be used to endorse or promote products derived from this software
  18  *    without specific prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  30  * SUCH DAMAGE.
  31  */
  32
  33 /*
  34  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  35  *      The Regents of the University of California.  All rights reserved.
  36  *
  37  * Redistribution and use in source and binary forms, with or without
  38  * modification, are permitted provided that the following conditions
  39  * are met:
  40  * 1. Redistributions of source code must retain the above copyright
  41  *    notice, this list of conditions and the following disclaimer.
  42  * 2. Redistributions in binary form must reproduce the above copyright
  43  *    notice, this list of conditions and the following disclaimer in the
  44  *    documentation and/or other materials provided with the distribution.
  45  * 3. All advertising materials mentioning features or use of this software
  46  *    must display the following acknowledgement:
  47  *      This product includes software developed by the University of
  48  *      California, Berkeley and its contributors.
  49  * 4. Neither the name of the University nor the names of its contributors
  50  *    may be used to endorse or promote products derived from this software
  51  *    without specific prior written permission.
  52  *
  53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  63  * SUCH DAMAGE.
  64  *
  65  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  66  */
  67
  68 #include "opt_ip6fw.h"
  69 #include "opt_inet.h"
  70 #include "opt_inet6.h"
  71 #include "opt_ipsec.h"
  72
  73 #include <sys/param.h>
  74 #include <sys/malloc.h>
  75 #include <sys/mbuf.h>
  76 #include <sys/errno.h>
  77 #include <sys/protosw.h>
  78 #include <sys/socket.h>
  79 #include <sys/socketvar.h>
  80 #include <sys/systm.h>
  81 #include <sys/kernel.h>
  82 #include <sys/proc.h>
  83 #include <sys/priv.h>
  84
  85 #include <sys/thread2.h>
  86 #include <sys/msgport2.h>
  87
  88 #include <net/if.h>
  89 #include <net/route.h>
  90 #include <net/pfil.h>
  91
  92 #include <netinet/in.h>
  93 #include <netinet/in_var.h>
  94 #include <netinet6/in6_var.h>
  95 #include <netinet/ip6.h>
  96 #include <netinet/icmp6.h>
  97 #include <netinet6/ip6_var.h>
  98 #include <netinet/in_pcb.h>
  99 #include <netinet6/nd6.h>
 100 #include <netinet6/ip6protosw.h>
 101
 102 #ifdef IPSEC
 103 #include <netinet6/ipsec.h>
 104 #ifdef INET6
 105 #include <netinet6/ipsec6.h>
 106 #endif
 107 #include <netproto/key/key.h>
 108 #endif /* IPSEC */
 109
 110 #ifdef FAST_IPSEC
 111 #include <netproto/ipsec/ipsec.h>
 112 #include <netproto/ipsec/ipsec6.h>
 113 #include <netproto/ipsec/key.h>
 114 #endif
 115
 116 #include <net/ip6fw/ip6_fw.h>
 117
 118 #include <net/net_osdep.h>
 119
 120 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
 121
 122 struct ip6_exthdrs {
 123         struct mbuf *ip6e_ip6;
 124         struct mbuf *ip6e_hbh;
 125         struct mbuf *ip6e_dest1;
 126         struct mbuf *ip6e_rthdr;
 127         struct mbuf *ip6e_dest2;
 128 };
 129
 130 static int ip6_pcbopt (int, u_char *, int, struct ip6_pktopts **, int);
 131 static int ip6_setpktoption (int, u_char *, int, struct ip6_pktopts *,
 132          int, int, int, int);
 133 static int ip6_pcbopts (struct ip6_pktopts **, struct mbuf *,
 134                             struct socket *, struct sockopt *);
 135 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 136 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *);
 137 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **);
 138 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
 139         struct ifnet *, struct in6_addr *, u_long *, int *);
 140 static int copyexthdr (void *, struct mbuf **);
 141 static int ip6_insertfraghdr (struct mbuf *, struct mbuf *, int,
 142                                   struct ip6_frag **);
 143 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t);
 144 static struct mbuf *ip6_splithdr (struct mbuf *);
 145 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 146
 147 /*
 148  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 149  * header (with pri, len, nxt, hlim, src, dst).
 150  * This function may modify ver and hlim only.
 151  * The mbuf chain containing the packet will be freed.
 152  * The mbuf opt, if present, will not be freed.
 153  *
 154  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 155  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 156  * which is rt_rmx.rmx_mtu.
 157  */
 158 int
 159 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
 160            int flags, struct ip6_moptions *im6o,
 161            struct ifnet **ifpp,         /* XXX: just for statistics */
 162            struct inpcb *inp)
 163 {
 164         struct ip6_hdr *ip6, *mhip6;
 165         struct ifnet *ifp, *origifp;
 166         struct mbuf *m = m0;
 167         struct mbuf *mprev;
 168         u_char *nexthdrp;
 169         int hlen, tlen, len, off;
 170         struct route_in6 ip6route;
 171         struct sockaddr_in6 *dst;
 172         int error = 0;
 173         struct in6_ifaddr *ia = NULL;
 174         u_long mtu;
 175         int alwaysfrag, dontfrag;
 176         u_int32_t optlen, plen = 0, unfragpartlen;
 177         struct ip6_exthdrs exthdrs;
 178         struct in6_addr finaldst;
 179         struct route_in6 *ro_pmtu = NULL;
 180         boolean_t hdrsplit = FALSE;
 181         boolean_t needipsec = FALSE;
 182 #ifdef IPSEC
 183         boolean_t needipsectun = FALSE;
 184         struct secpolicy *sp = NULL;
 185         struct socket *so = inp ? inp->inp_socket : NULL;
 186
 187         ip6 = mtod(m, struct ip6_hdr *);
 188 #endif
 189 #ifdef FAST_IPSEC
 190         boolean_t needipsectun = FALSE;
 191         struct secpolicy *sp = NULL;
 192
 193         ip6 = mtod(m, struct ip6_hdr *);
 194 #endif
 195
 196         bzero(&exthdrs, sizeof exthdrs);
 197
 198         if (opt) {
 199                 if ((error = copyexthdr(opt->ip6po_hbh, &exthdrs.ip6e_hbh)))
 200                         goto freehdrs;
 201                 if ((error = copyexthdr(opt->ip6po_dest1, &exthdrs.ip6e_dest1)))
 202                         goto freehdrs;
 203                 if ((error = copyexthdr(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr)))
 204                         goto freehdrs;
 205                 if ((error = copyexthdr(opt->ip6po_dest2, &exthdrs.ip6e_dest2)))
 206                         goto freehdrs;
 207         }
 208
 209 #ifdef IPSEC
 210         /* get a security policy for this packet */
 211         if (so == NULL)
 212                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 213         else
 214                 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
 215
 216         if (sp == NULL) {
 217                 ipsec6stat.out_inval++;
 218                 goto freehdrs;
 219         }
 220
 221         error = 0;
 222
 223         /* check policy */
 224         switch (sp->policy) {
 225         case IPSEC_POLICY_DISCARD:
 226                 /*
 227                  * This packet is just discarded.
 228                  */
 229                 ipsec6stat.out_polvio++;
 230                 goto freehdrs;
 231
 232         case IPSEC_POLICY_BYPASS:
 233         case IPSEC_POLICY_NONE:
 234                 /* no need to do IPsec. */
 235                 needipsec = FALSE;
 236                 break;
 237
 238         case IPSEC_POLICY_IPSEC:
 239                 if (sp->req == NULL) {
 240                         error = key_spdacquire(sp);     /* acquire a policy */
 241                         goto freehdrs;
 242                 }
 243                 needipsec = TRUE;
 244                 break;
 245
 246         case IPSEC_POLICY_ENTRUST:
 247         default:
 248                 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
 249         }
 250 #endif /* IPSEC */
 251 #ifdef FAST_IPSEC
 252         /* get a security policy for this packet */
 253         if (inp == NULL)
 254                 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 255         else
 256                 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
 257
 258         if (sp == NULL) {
 259                 newipsecstat.ips_out_inval++;
 260                 goto freehdrs;
 261         }
 262
 263         error = 0;
 264
 265         /* check policy */
 266         switch (sp->policy) {
 267         case IPSEC_POLICY_DISCARD:
 268                 /*
 269                  * This packet is just discarded.
 270                  */
 271                 newipsecstat.ips_out_polvio++;
 272                 goto freehdrs;
 273
 274         case IPSEC_POLICY_BYPASS:
 275         case IPSEC_POLICY_NONE:
 276                 /* no need to do IPsec. */
 277                 needipsec = FALSE;
 278                 break;
 279
 280         case IPSEC_POLICY_IPSEC:
 281                 if (sp->req == NULL) {
 282                         error = key_spdacquire(sp);     /* acquire a policy */
 283                         goto freehdrs;
 284                 }
 285                 needipsec = TRUE;
 286                 break;
 287
 288         case IPSEC_POLICY_ENTRUST:
 289         default:
 290                 kprintf("ip6_output: Invalid policy found. %d\n", sp->policy);
 291         }
 292 #endif /* FAST_IPSEC */
 293
 294         /*
 295          * Calculate the total length of the extension header chain.
 296          * Keep the length of the unfragmentable part for fragmentation.
 297          */
 298         optlen = m_lengthm(exthdrs.ip6e_hbh, NULL) +
 299             m_lengthm(exthdrs.ip6e_dest1, NULL) +
 300             m_lengthm(exthdrs.ip6e_rthdr, NULL);
 301
 302         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 303
 304         /* NOTE: we don't add AH/ESP length here. do that later. */
 305         optlen += m_lengthm(exthdrs.ip6e_dest2, NULL);
 306
 307         /*
 308          * If we need IPsec, or there is at least one extension header,
 309          * separate IP6 header from the payload.
 310          */
 311         if ((needipsec || optlen) && !hdrsplit) {
 312                 exthdrs.ip6e_ip6 = ip6_splithdr(m);
 313                 if (exthdrs.ip6e_ip6 == NULL) {
 314                         error = ENOBUFS;
 315                         goto freehdrs;
 316                 }
 317                 m = exthdrs.ip6e_ip6;
 318                 hdrsplit = TRUE;
 319         }
 320
 321         /* adjust pointer */
 322         ip6 = mtod(m, struct ip6_hdr *);
 323
 324         /* adjust mbuf packet header length */
 325         m->m_pkthdr.len += optlen;
 326         plen = m->m_pkthdr.len - sizeof(*ip6);
 327
 328         /* If this is a jumbo payload, insert a jumbo payload option. */
 329         if (plen > IPV6_MAXPACKET) {
 330                 if (!hdrsplit) {
 331                         exthdrs.ip6e_ip6 = ip6_splithdr(m);
 332                         if (exthdrs.ip6e_ip6 == NULL) {
 333                                 error = ENOBUFS;
 334                                 goto freehdrs;
 335                         }
 336                         m = exthdrs.ip6e_ip6;
 337                         hdrsplit = TRUE;
 338                 }
 339                 /* adjust pointer */
 340                 ip6 = mtod(m, struct ip6_hdr *);
 341                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 342                         goto freehdrs;
 343                 ip6->ip6_plen = 0;
 344         } else
 345                 ip6->ip6_plen = htons(plen);
 346
 347         /*
 348          * Concatenate headers and fill in next header fields.
 349          * Here we have, on "m"
 350          *      IPv6 payload
 351          * and we insert headers accordingly.  Finally, we should be getting:
 352          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 353          *
 354          * during the header composing process, "m" points to IPv6 header.
 355          * "mprev" points to an extension header prior to esp.
 356          */
 357
 358         nexthdrp = &ip6->ip6_nxt;
 359         mprev = m;
 360
 361         /*
 362          * we treat dest2 specially.  this makes IPsec processing
 363          * much easier.  the goal here is to make mprev point the
 364          * mbuf prior to dest2.
 365          *
 366          * result: IPv6 dest2 payload
 367          * m and mprev will point to IPv6 header.
 368          */
 369         if (exthdrs.ip6e_dest2) {
 370                 if (!hdrsplit)
 371                         panic("assumption failed: hdr not split");
 372                 exthdrs.ip6e_dest2->m_next = m->m_next;
 373                 m->m_next = exthdrs.ip6e_dest2;
 374                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 375                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 376         }
 377
 378 /*
 379  * Place m1 after mprev.
 380  */
 381 #define MAKE_CHAIN(m1, mprev, nexthdrp, i)\
 382     do {\
 383         if (m1) {\
 384                 if (!hdrsplit)\
 385                         panic("assumption failed: hdr not split");\
 386                 *mtod(m1, u_char *) = *nexthdrp;\
 387                 *nexthdrp = (i);\
 388                 nexthdrp = mtod(m1, u_char *);\
 389                 m1->m_next = mprev->m_next;\
 390                 mprev->m_next = m1;\
 391                 mprev = m1;\
 392         }\
 393     } while (0)
 394
 395         /*
 396          * result: IPv6 hbh dest1 rthdr dest2 payload
 397          * m will point to IPv6 header.  mprev will point to the
 398          * extension header prior to dest2 (rthdr in the above case).
 399          */
 400         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 401         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS);
 402         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING);
 403
 404 #if defined(IPSEC) || defined(FAST_IPSEC)
 405         if (needipsec) {
 406                 struct ipsec_output_state state;
 407                 int segleft_org = 0;
 408                 struct ip6_rthdr *rh = NULL;
 409
 410                 /*
 411                  * pointers after IPsec headers are not valid any more.
 412                  * other pointers need a great care too.
 413                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
 414                  */
 415                 exthdrs.ip6e_dest2 = NULL;
 416
 417                 if (exthdrs.ip6e_rthdr) {
 418                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 419                         segleft_org = rh->ip6r_segleft;
 420                         rh->ip6r_segleft = 0;
 421                 }
 422
 423                 bzero(&state, sizeof state);
 424                 state.m = m;
 425                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 426                                             &needipsectun);
 427                 m = state.m;
 428                 if (error) {
 429                         /* mbuf is already reclaimed in ipsec6_output_trans. */
 430                         m = NULL;
 431                         switch (error) {
 432                         case EHOSTUNREACH:
 433                         case ENETUNREACH:
 434                         case EMSGSIZE:
 435                         case ENOBUFS:
 436                         case ENOMEM:
 437                                 break;
 438                         default:
 439                                 kprintf("ip6_output (ipsec): error code %d\n",
 440                                        error);
 441                                 /* fall through */
 442                         case ENOENT:
 443                                 /* don't show these error codes to the user */
 444                                 error = 0;
 445                                 break;
 446                         }
 447                         goto bad;
 448                 }
 449                 if (exthdrs.ip6e_rthdr) {
 450                         /* ah6_output doesn't modify mbuf chain */
 451                         rh->ip6r_segleft = segleft_org;
 452                 }
 453         }
 454 #endif
 455
 456         /*
 457          * If there is a routing header, replace destination address field
 458          * with the first hop of the routing header.
 459          */
 460         if (exthdrs.ip6e_rthdr) {
 461                 struct ip6_rthdr *rh;
 462
 463                 finaldst = ip6->ip6_dst;
 464                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 465                 switch (rh->ip6r_type) {
 466                 default:        /* is it possible? */
 467                          error = EINVAL;
 468                          goto bad;
 469                 }
 470         }
 471
 472         /* Source address validation */
 473         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 474             !(flags & IPV6_DADOUTPUT)) {
 475                 error = EOPNOTSUPP;
 476                 ip6stat.ip6s_badscope++;
 477                 goto bad;
 478         }
 479         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 480                 error = EOPNOTSUPP;
 481                 ip6stat.ip6s_badscope++;
 482                 goto bad;
 483         }
 484
 485         ip6stat.ip6s_localout++;
 486
 487         /*
 488          * Route packet.
 489          */
 490         if (ro == NULL) {
 491                 ro = &ip6route;
 492                 bzero(ro, sizeof(*ro));
 493         }
 494         ro_pmtu = ro;
 495         if (opt && opt->ip6po_rthdr)
 496                 ro = &opt->ip6po_route;
 497         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 498         /*
 499          * If there is a cached route,
 500          * check that it is to the same destination
 501          * and is still up. If not, free it and try again.
 502          */
 503         if (ro->ro_rt != NULL &&
 504             (!(ro->ro_rt->rt_flags & RTF_UP) || dst->sin6_family != AF_INET6 ||
 505              !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
 506                 RTFREE(ro->ro_rt);
 507                 ro->ro_rt = NULL;
 508         }
 509         if (ro->ro_rt == NULL) {
 510                 bzero(dst, sizeof(*dst));
 511                 dst->sin6_family = AF_INET6;
 512                 dst->sin6_len = sizeof(struct sockaddr_in6);
 513                 dst->sin6_addr = ip6->ip6_dst;
 514         }
 515 #if defined(IPSEC) || defined(FAST_IPSEC)
 516         if (needipsec && needipsectun) {
 517                 struct ipsec_output_state state;
 518
 519                 /*
 520                  * All the extension headers will become inaccessible
 521                  * (since they can be encrypted).
 522                  * Don't panic, we need no more updates to extension headers
 523                  * on inner IPv6 packet (since they are now encapsulated).
 524                  *
 525                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 526                  */
 527                 bzero(&exthdrs, sizeof(exthdrs));
 528                 exthdrs.ip6e_ip6 = m;
 529
 530                 bzero(&state, sizeof(state));
 531                 state.m = m;
 532                 state.ro = (struct route *)ro;
 533                 state.dst = (struct sockaddr *)dst;
 534
 535                 error = ipsec6_output_tunnel(&state, sp, flags);
 536
 537                 m = state.m;
 538                 ro = (struct route_in6 *)state.ro;
 539                 dst = (struct sockaddr_in6 *)state.dst;
 540                 if (error) {
 541                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 542                         m0 = m = NULL;
 543                         m = NULL;
 544                         switch (error) {
 545                         case EHOSTUNREACH:
 546                         case ENETUNREACH:
 547                         case EMSGSIZE:
 548                         case ENOBUFS:
 549                         case ENOMEM:
 550                                 break;
 551                         default:
 552                                 kprintf("ip6_output (ipsec): error code %d\n", error);
 553                                 /* fall through */
 554                         case ENOENT:
 555                                 /* don't show these error codes to the user */
 556                                 error = 0;
 557                                 break;
 558                         }
 559                         goto bad;
 560                 }
 561
 562                 exthdrs.ip6e_ip6 = m;
 563         }
 564 #endif
 565
 566         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 567                 /* Unicast */
 568
 569 #define ifatoia6(ifa)   ((struct in6_ifaddr *)(ifa))
 570 #define sin6tosa(sin6)  ((struct sockaddr *)(sin6))
 571                 /* xxx
 572                  * interface selection comes here
 573                  * if an interface is specified from an upper layer,
 574                  * ifp must point it.
 575                  */
 576                 if (ro->ro_rt == NULL) {
 577                         /*
 578                          * non-bsdi always clone routes, if parent is
 579                          * PRF_CLONING.
 580                          */
 581                         rtalloc((struct route *)ro);
 582                 }
 583                 if (ro->ro_rt == NULL) {
 584                         ip6stat.ip6s_noroute++;
 585                         error = EHOSTUNREACH;
 586                         /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
 587                         goto bad;
 588                 }
 589                 ia = ifatoia6(ro->ro_rt->rt_ifa);
 590                 ifp = ro->ro_rt->rt_ifp;
 591                 ro->ro_rt->rt_use++;
 592                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
 593                         dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
 594                 m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
 595
 596                 in6_ifstat_inc(ifp, ifs6_out_request);
 597
 598                 /*
 599                  * Check if the outgoing interface conflicts with
 600                  * the interface specified by ifi6_ifindex (if specified).
 601                  * Note that loopback interface is always okay.
 602                  * (this may happen when we are sending a packet to one of
 603                  *  our own addresses.)
 604                  */
 605                 if (opt && opt->ip6po_pktinfo
 606                  && opt->ip6po_pktinfo->ipi6_ifindex) {
 607                         if (!(ifp->if_flags & IFF_LOOPBACK)
 608                          && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
 609                                 ip6stat.ip6s_noroute++;
 610                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 611                                 error = EHOSTUNREACH;
 612                                 goto bad;
 613                         }
 614                 }
 615
 616                 if (opt && opt->ip6po_hlim != -1)
 617                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 618         } else {
 619                 /* Multicast */
 620                 struct  in6_multi *in6m;
 621
 622                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 623
 624                 /*
 625                  * See if the caller provided any multicast options
 626                  */
 627                 ifp = NULL;
 628                 if (im6o != NULL) {
 629                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 630                         if (im6o->im6o_multicast_ifp != NULL)
 631                                 ifp = im6o->im6o_multicast_ifp;
 632                 } else
 633                         ip6->ip6_hlim = ip6_defmcasthlim;
 634
 635                 /*
 636                  * See if the caller provided the outgoing interface
 637                  * as an ancillary data.
 638                  * Boundary check for ifindex is assumed to be already done.
 639                  */
 640                 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
 641                         ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
 642
 643                 /*
 644                  * If the destination is a node-local scope multicast,
 645                  * the packet should be loop-backed only.
 646                  */
 647                 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
 648                         /*
 649                          * If the outgoing interface is already specified,
 650                          * it should be a loopback interface.
 651                          */
 652                         if (ifp && !(ifp->if_flags & IFF_LOOPBACK)) {
 653                                 ip6stat.ip6s_badscope++;
 654                                 error = ENETUNREACH; /* XXX: better error? */
 655                                 /* XXX correct ifp? */
 656                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 657                                 goto bad;
 658                         } else {
 659                                 ifp = &loif[0];
 660                         }
 661                 }
 662
 663                 if (opt && opt->ip6po_hlim != -1)
 664                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 665
 666                 /*
 667                  * If caller did not provide an interface lookup a
 668                  * default in the routing table.  This is either a
 669                  * default for the speicfied group (i.e. a host
 670                  * route), or a multicast default (a route for the
 671                  * ``net'' ff00::/8).
 672                  */
 673                 if (ifp == NULL) {
 674                         if (ro->ro_rt == NULL) {
 675                                 ro->ro_rt =
 676                                   rtpurelookup((struct sockaddr *)&ro->ro_dst);
 677                         }
 678                         if (ro->ro_rt == NULL) {
 679                                 ip6stat.ip6s_noroute++;
 680                                 error = EHOSTUNREACH;
 681                                 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
 682                                 goto bad;
 683                         }
 684                         ia = ifatoia6(ro->ro_rt->rt_ifa);
 685                         ifp = ro->ro_rt->rt_ifp;
 686                         ro->ro_rt->rt_use++;
 687                 }
 688
 689                 if (!(flags & IPV6_FORWARDING))
 690                         in6_ifstat_inc(ifp, ifs6_out_request);
 691                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 692
 693                 /*
 694                  * Confirm that the outgoing interface supports multicast.
 695                  */
 696                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 697                         ip6stat.ip6s_noroute++;
 698                         in6_ifstat_inc(ifp, ifs6_out_discard);
 699                         error = ENETUNREACH;
 700                         goto bad;
 701                 }
 702                 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
 703                 if (in6m != NULL &&
 704                    (im6o == NULL || im6o->im6o_multicast_loop)) {
 705                         /*
 706                          * If we belong to the destination multicast group
 707                          * on the outgoing interface, and the caller did not
 708                          * forbid loopback, loop back a copy.
 709                          */
 710                         ip6_mloopback(ifp, m, dst);
 711                 } else {
 712                         /*
 713                          * If we are acting as a multicast router, perform
 714                          * multicast forwarding as if the packet had just
 715                          * arrived on the interface to which we are about
 716                          * to send.  The multicast forwarding function
 717                          * recursively calls this function, using the
 718                          * IPV6_FORWARDING flag to prevent infinite recursion.
 719                          *
 720                          * Multicasts that are looped back by ip6_mloopback(),
 721                          * above, will be forwarded by the ip6_input() routine,
 722                          * if necessary.
 723                          */
 724                         if (ip6_mrouter && !(flags & IPV6_FORWARDING)) {
 725                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 726                                         m_freem(m);
 727                                         goto done;
 728                                 }
 729                         }
 730                 }
 731                 /*
 732                  * Multicasts with a hoplimit of zero may be looped back,
 733                  * above, but must not be transmitted on a network.
 734                  * Also, multicasts addressed to the loopback interface
 735                  * are not sent -- the above call to ip6_mloopback() will
 736                  * loop back a copy if this host actually belongs to the
 737                  * destination group on the loopback interface.
 738                  */
 739                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
 740                         m_freem(m);
 741                         goto done;
 742                 }
 743         }
 744
 745         /*
 746          * Fill the outgoing inteface to tell the upper layer
 747          * to increment per-interface statistics.
 748          */
 749         if (ifpp)
 750                 *ifpp = ifp;
 751
 752         /* Determine path MTU. */
 753         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 754             &alwaysfrag)) != 0)
 755                 goto bad;
 756
 757         /*
 758          * The caller of this function may specify to use the minimum MTU
 759          * in some cases.
 760          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 761          * setting.  The logic is a bit complicated; by default, unicast
 762          * packets will follow path MTU while multicast packets will be sent at
 763          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 764          * including unicast ones will be sent at the minimum MTU.  Multicast
 765          * packets will always be sent at the minimum MTU unless
 766          * IP6PO_MINMTU_DISABLE is explicitly specified.
 767          * See RFC 3542 for more details.
 768          */
 769         if (mtu > IPV6_MMTU) {
 770                 if ((flags & IPV6_MINMTU))
 771                         mtu = IPV6_MMTU;
 772                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 773                         mtu = IPV6_MMTU;
 774                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 775                          (opt == NULL ||
 776                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 777                         mtu = IPV6_MMTU;
 778                 }
 779         }
 780
 781         /* Fake scoped addresses */
 782         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
 783                 /*
 784                  * If source or destination address is a scoped address, and
 785                  * the packet is going to be sent to a loopback interface,
 786                  * we should keep the original interface.
 787                  */
 788
 789                 /*
 790                  * XXX: this is a very experimental and temporary solution.
 791                  * We eventually have sockaddr_in6 and use the sin6_scope_id
 792                  * field of the structure here.
 793                  * We rely on the consistency between two scope zone ids
 794                  * of source and destination, which should already be assured.
 795                  * Larger scopes than link will be supported in the future.
 796                  */
 797                 origifp = NULL;
 798                 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
 799                         origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
 800                 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
 801                         origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
 802                 /*
 803                  * XXX: origifp can be NULL even in those two cases above.
 804                  * For example, if we remove the (only) link-local address
 805                  * from the loopback interface, and try to send a link-local
 806                  * address without link-id information.  Then the source
 807                  * address is ::1, and the destination address is the
 808                  * link-local address with its s6_addr16[1] being zero.
 809                  * What is worse, if the packet goes to the loopback interface
 810                  * by a default rejected route, the null pointer would be
 811                  * passed to looutput, and the kernel would hang.
 812                  * The following last resort would prevent such disaster.
 813                  */
 814                 if (origifp == NULL)
 815                         origifp = ifp;
 816         }
 817         else
 818                 origifp = ifp;
 819         /*
 820          * clear embedded scope identifiers if necessary.
 821          * in6_clearscope will touch the addresses only when necessary.
 822          */
 823         in6_clearscope(&ip6->ip6_src);
 824         in6_clearscope(&ip6->ip6_dst);
 825
 826         /*
 827          * Check with the firewall...
 828          */
 829         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 830                 u_short port = 0;
 831
 832                 m->m_pkthdr.rcvif = NULL;       /* XXX */
 833                 /* If ipfw says divert, we have to just drop packet */
 834                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
 835                         m_freem(m);
 836                         goto done;
 837                 }
 838                 if (!m) {
 839                         error = EACCES;
 840                         goto done;
 841                 }
 842         }
 843
 844         /*
 845          * If the outgoing packet contains a hop-by-hop options header,
 846          * it must be examined and processed even by the source node.
 847          * (RFC 2460, section 4.)
 848          */
 849         if (exthdrs.ip6e_hbh) {
 850                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 851                 u_int32_t dummy1; /* XXX unused */
 852                 u_int32_t dummy2; /* XXX unused */
 853
 854 #ifdef DIAGNOSTIC
 855                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 856                         panic("ip6e_hbh is not continuous");
 857 #endif
 858                 /*
 859                  *  XXX: if we have to send an ICMPv6 error to the sender,
 860                  *       we need the M_LOOP flag since icmp6_error() expects
 861                  *       the IPv6 and the hop-by-hop options header are
 862                  *       continuous unless the flag is set.
 863                  */
 864                 m->m_flags |= M_LOOP;
 865                 m->m_pkthdr.rcvif = ifp;
 866                 if (ip6_process_hopopts(m,
 867                                         (u_int8_t *)(hbh + 1),
 868                                         ((hbh->ip6h_len + 1) << 3) -
 869                                         sizeof(struct ip6_hbh),
 870                                         &dummy1, &dummy2) < 0) {
 871                         /* m was already freed at this point */
 872                         error = EINVAL;/* better error? */
 873                         goto done;
 874                 }
 875                 m->m_flags &= ~M_LOOP; /* XXX */
 876                 m->m_pkthdr.rcvif = NULL;
 877         }
 878
 879         /*
 880          * Run through list of hooks for output packets.
 881          */
 882         if (pfil_has_hooks(&inet6_pfil_hook)) {
 883                 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
 884                 if (error != 0 || m == NULL)
 885                         goto done;
 886                 ip6 = mtod(m, struct ip6_hdr *);
 887         }
 888
 889         /*
 890          * Send the packet to the outgoing interface.
 891          * If necessary, do IPv6 fragmentation before sending.
 892          *
 893          * the logic here is rather complex:
 894          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 895          * 1-a: send as is if tlen <= path mtu
 896          * 1-b: fragment if tlen > path mtu
 897          *
 898          * 2: if user asks us not to fragment (dontfrag == 1)
 899          * 2-a: send as is if tlen <= interface mtu
 900          * 2-b: error if tlen > interface mtu
 901          *
 902          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 903          *      always fragment
 904          *
 905          * 4: if dontfrag == 1 && alwaysfrag == 1
 906          *      error, as we cannot handle this conflicting request
 907          */
 908         tlen = m->m_pkthdr.len;
 909
 910         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
 911                 dontfrag = 1;
 912         else
 913                 dontfrag = 0;
 914         if (dontfrag && alwaysfrag) {   /* case 4 */
 915                 /* conflicting request - can't transmit */
 916                 error = EMSGSIZE;
 917                 goto bad;
 918         }
 919         if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
 920                 /*
 921                  * Even if the DONTFRAG option is specified, we cannot send the
 922                  * packet when the data length is larger than the MTU of the
 923                  * outgoing interface.
 924                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 925                  * well as returning an error code (the latter is not described
 926                  * in the API spec.)
 927                  */
 928                 u_int32_t mtu32;
 929                 struct ip6ctlparam ip6cp;
 930
 931                 mtu32 = (u_int32_t)mtu;
 932                 bzero(&ip6cp, sizeof(ip6cp));
 933                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
 934                 kpfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
 935                     (void *)&ip6cp);
 936
 937                 error = EMSGSIZE;
 938                 goto bad;
 939         }
 940
 941         /*
 942          * transmit packet without fragmentation
 943          */
 944         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 945                 struct in6_ifaddr *ia6;
 946
 947                 ip6 = mtod(m, struct ip6_hdr *);
 948                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 949                 if (ia6) {
 950                         /* Record statistics for this interface address. */
 951                         ia6->ia_ifa.if_opackets++;
 952                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
 953                 }
 954 #ifdef IPSEC
 955                 /* clean ipsec history once it goes out of the node */
 956                 ipsec_delaux(m);
 957 #endif
 958                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 959                 goto done;
 960         }
 961
 962         /*
 963          * try to fragment the packet.  case 1-b and 3
 964          */
 965         if (mtu < IPV6_MMTU) {
 966                 /*
 967                  * note that path MTU is never less than IPV6_MMTU
 968                  * (see icmp6_input).
 969                  */
 970                 error = EMSGSIZE;
 971                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 972                 goto bad;
 973         } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
 974                 error = EMSGSIZE;
 975                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 976                 goto bad;
 977         } else {
 978                 struct mbuf **mnext, *m_frgpart;
 979                 struct ip6_frag *ip6f;
 980                 u_int32_t id = htonl(ip6_id++);
 981                 u_char nextproto;
 982
 983                 /*
 984                  * Too large for the destination or interface;
 985                  * fragment if possible.
 986                  * Must be able to put at least 8 bytes per fragment.
 987                  */
 988                 hlen = unfragpartlen;
 989                 if (mtu > IPV6_MAXPACKET)
 990                         mtu = IPV6_MAXPACKET;
 991
 992                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 993                 if (len < 8) {
 994                         error = EMSGSIZE;
 995                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 996                         goto bad;
 997                 }
 998
 999                 mnext = &m->m_nextpkt;
1000
1001                 /*
1002                  * Change the next header field of the last header in the
1003                  * unfragmentable part.
1004                  */
1005                 if (exthdrs.ip6e_rthdr) {
1006                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1007                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1008                 } else if (exthdrs.ip6e_dest1) {
1009                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1010                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1011                 } else if (exthdrs.ip6e_hbh) {
1012                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1013                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1014                 } else {
1015                         nextproto = ip6->ip6_nxt;
1016                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1017                 }
1018
1019                 /*
1020                  * Loop through length of segment after first fragment,
1021                  * make new header and copy data of each part and link onto
1022                  * chain.
1023                  */
1024                 m0 = m;
1025                 for (off = hlen; off < tlen; off += len) {
1026                         MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1027                         if (!m) {
1028                                 error = ENOBUFS;
1029                                 ip6stat.ip6s_odropped++;
1030                                 goto sendorfree;
1031                         }
1032                         m->m_pkthdr.rcvif = NULL;
1033                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1034                         *mnext = m;
1035                         mnext = &m->m_nextpkt;
1036                         m->m_data += max_linkhdr;
1037                         mhip6 = mtod(m, struct ip6_hdr *);
1038                         *mhip6 = *ip6;
1039                         m->m_len = sizeof(*mhip6);
1040                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1041                         if (error) {
1042                                 ip6stat.ip6s_odropped++;
1043                                 goto sendorfree;
1044                         }
1045                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1046                         if (off + len >= tlen)
1047                                 len = tlen - off;
1048                         else
1049                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1050                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1051                                                           sizeof(*ip6f) -
1052                                                           sizeof(struct ip6_hdr)));
1053                         if ((m_frgpart = m_copy(m0, off, len)) == NULL) {
1054                                 error = ENOBUFS;
1055                                 ip6stat.ip6s_odropped++;
1056                                 goto sendorfree;
1057                         }
1058                         m_cat(m, m_frgpart);
1059                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1060                         m->m_pkthdr.rcvif = NULL;
1061                         ip6f->ip6f_reserved = 0;
1062                         ip6f->ip6f_ident = id;
1063                         ip6f->ip6f_nxt = nextproto;
1064                         ip6stat.ip6s_ofragments++;
1065                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1066                 }
1067
1068                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1069         }
1070
1071         /*
1072          * Remove leading garbages.
1073          */
1074 sendorfree:
1075         m = m0->m_nextpkt;
1076         m0->m_nextpkt = NULL;
1077         m_freem(m0);
1078         for (m0 = m; m; m = m0) {
1079                 m0 = m->m_nextpkt;
1080                 m->m_nextpkt = NULL;
1081                 if (error == 0) {
1082                         /* Record statistics for this interface address. */
1083                         if (ia) {
1084                                 ia->ia_ifa.if_opackets++;
1085                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1086                         }
1087 #ifdef IPSEC
1088                         /* clean ipsec history once it goes out of the node */
1089                         ipsec_delaux(m);
1090 #endif
1091                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1092                 } else
1093                         m_freem(m);
1094         }
1095
1096         if (error == 0)
1097                 ip6stat.ip6s_fragmented++;
1098
1099 done:
1100         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1101                 RTFREE(ro->ro_rt);
1102         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1103                 RTFREE(ro_pmtu->ro_rt);
1104         }
1105
1106 #ifdef IPSEC
1107         if (sp != NULL)
1108                 key_freesp(sp);
1109 #endif
1110 #ifdef FAST_IPSEC
1111         if (sp != NULL)
1112                 KEY_FREESP(&sp);
1113 #endif
1114
1115         return (error);
1116
1117 freehdrs:
1118         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1119         m_freem(exthdrs.ip6e_dest1);
1120         m_freem(exthdrs.ip6e_rthdr);
1121         m_freem(exthdrs.ip6e_dest2);
1122         /* fall through */
1123 bad:
1124         m_freem(m);
1125         goto done;
1126 }
1127
1128 static int
1129 copyexthdr(void *h, struct mbuf **mp)
1130 {
1131         struct ip6_ext *hdr = h;
1132         int hlen;
1133         struct mbuf *m;
1134
1135         if (hdr == NULL)
1136                 return 0;
1137
1138         hlen = (hdr->ip6e_len + 1) * 8;
1139         if (hlen > MCLBYTES)
1140                 return ENOBUFS; /* XXX */
1141
1142         m = m_getb(hlen, MB_DONTWAIT, MT_DATA, 0);
1143         if (!m)
1144                 return ENOBUFS;
1145         m->m_len = hlen;
1146
1147         bcopy(hdr, mtod(m, caddr_t), hlen);
1148
1149         *mp = m;
1150         return 0;
1151 }
1152
1153 /*
1154  * Insert jumbo payload option.
1155  */
1156 static int
1157 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1158 {
1159         struct mbuf *mopt;
1160         u_char *optbuf;
1161         u_int32_t v;
1162
1163 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1164
1165         /*
1166          * If there is no hop-by-hop options header, allocate new one.
1167          * If there is one but it doesn't have enough space to store the
1168          * jumbo payload option, allocate a cluster to store the whole options.
1169          * Otherwise, use it to store the options.
1170          */
1171         if (exthdrs->ip6e_hbh == NULL) {
1172                 MGET(mopt, MB_DONTWAIT, MT_DATA);
1173                 if (mopt == NULL)
1174                         return (ENOBUFS);
1175                 mopt->m_len = JUMBOOPTLEN;
1176                 optbuf = mtod(mopt, u_char *);
1177                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1178                 exthdrs->ip6e_hbh = mopt;
1179         } else {
1180                 struct ip6_hbh *hbh;
1181
1182                 mopt = exthdrs->ip6e_hbh;
1183                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1184                         /*
1185                          * XXX assumption:
1186                          * - exthdrs->ip6e_hbh is not referenced from places
1187                          *   other than exthdrs.
1188                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1189                          */
1190                         int oldoptlen = mopt->m_len;
1191                         struct mbuf *n;
1192
1193                         /*
1194                          * XXX: give up if the whole (new) hbh header does
1195                          * not fit even in an mbuf cluster.
1196                          */
1197                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1198                                 return (ENOBUFS);
1199
1200                         /*
1201                          * As a consequence, we must always prepare a cluster
1202                          * at this point.
1203                          */
1204                         n = m_getcl(MB_DONTWAIT, MT_DATA, 0);
1205                         if (!n)
1206                                 return (ENOBUFS);
1207                         n->m_len = oldoptlen + JUMBOOPTLEN;
1208                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen);
1209                         optbuf = mtod(n, caddr_t) + oldoptlen;
1210                         m_freem(mopt);
1211                         mopt = exthdrs->ip6e_hbh = n;
1212                 } else {
1213                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1214                         mopt->m_len += JUMBOOPTLEN;
1215                 }
1216                 optbuf[0] = IP6OPT_PADN;
1217                 optbuf[1] = 1;
1218
1219                 /*
1220                  * Adjust the header length according to the pad and
1221                  * the jumbo payload option.
1222                  */
1223                 hbh = mtod(mopt, struct ip6_hbh *);
1224                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1225         }
1226
1227         /* fill in the option. */
1228         optbuf[2] = IP6OPT_JUMBO;
1229         optbuf[3] = 4;
1230         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1231         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1232
1233         /* finally, adjust the packet header length */
1234         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1235
1236         return (0);
1237 #undef JUMBOOPTLEN
1238 }
1239
1240 /*
1241  * Insert fragment header and copy unfragmentable header portions.
1242  */
1243 static int
1244 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1245                   struct ip6_frag **frghdrp)
1246 {
1247         struct mbuf *n, *mlast;
1248
1249         if (hlen > sizeof(struct ip6_hdr)) {
1250                 n = m_copym(m0, sizeof(struct ip6_hdr),
1251                             hlen - sizeof(struct ip6_hdr), MB_DONTWAIT);
1252                 if (n == NULL)
1253                         return (ENOBUFS);
1254                 m->m_next = n;
1255         } else
1256                 n = m;
1257
1258         /* Search for the last mbuf of unfragmentable part. */
1259         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1260                 ;
1261
1262         if (!(mlast->m_flags & M_EXT) &&
1263             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1264                 /* use the trailing space of the last mbuf for the fragment hdr */
1265                 *frghdrp =
1266                         (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1267                 mlast->m_len += sizeof(struct ip6_frag);
1268                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1269         } else {
1270                 /* allocate a new mbuf for the fragment header */
1271                 struct mbuf *mfrg;
1272
1273                 MGET(mfrg, MB_DONTWAIT, MT_DATA);
1274                 if (mfrg == NULL)
1275                         return (ENOBUFS);
1276                 mfrg->m_len = sizeof(struct ip6_frag);
1277                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1278                 mlast->m_next = mfrg;
1279         }
1280
1281         return (0);
1282 }
1283
1284 static int
1285 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1286     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1287     int *alwaysfragp)
1288 {
1289         u_int32_t mtu = 0;
1290         int alwaysfrag = 0;
1291         int error = 0;
1292
1293         if (ro_pmtu != ro) {
1294                 /* The first hop and the final destination may differ. */
1295                 struct sockaddr_in6 *sa6_dst =
1296                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1297                 if (ro_pmtu->ro_rt &&
1298                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1299                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1300                         RTFREE(ro_pmtu->ro_rt);
1301                         ro_pmtu->ro_rt = NULL;
1302                 }
1303                 if (ro_pmtu->ro_rt == NULL) {
1304                         bzero(sa6_dst, sizeof(*sa6_dst));
1305                         sa6_dst->sin6_family = AF_INET6;
1306                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1307                         sa6_dst->sin6_addr = *dst;
1308
1309                         rtalloc((struct route *)ro_pmtu);
1310                 }
1311         }
1312         if (ro_pmtu->ro_rt) {
1313                 u_int32_t ifmtu;
1314                 struct in_conninfo inc;
1315
1316                 bzero(&inc, sizeof(inc));
1317                 inc.inc_flags = 1; /* IPv6 */
1318                 inc.inc6_faddr = *dst;
1319
1320                 if (ifp == NULL)
1321                         ifp = ro_pmtu->ro_rt->rt_ifp;
1322                 ifmtu = IN6_LINKMTU(ifp);
1323                 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1324                 if (mtu == 0)
1325                         mtu = ifmtu;
1326                 else if (mtu < IPV6_MMTU) {
1327                         /*
1328                          * RFC2460 section 5, last paragraph:
1329                          * if we record ICMPv6 too big message with
1330                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1331                          * or smaller, with framgent header attached.
1332                          * (fragment header is needed regardless from the
1333                          * packet size, for translators to identify packets)
1334                          */
1335                         alwaysfrag = 1;
1336                         mtu = IPV6_MMTU;
1337                 } else if (mtu > ifmtu) {
1338                         /*
1339                          * The MTU on the route is larger than the MTU on
1340                          * the interface!  This shouldn't happen, unless the
1341                          * MTU of the interface has been changed after the
1342                          * interface was brought up.  Change the MTU in the
1343                          * route to match the interface MTU (as long as the
1344                          * field isn't locked).
1345                          */
1346                         mtu = ifmtu;
1347                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1348                 }
1349         } else if (ifp) {
1350                 mtu = IN6_LINKMTU(ifp);
1351         } else
1352                 error = EHOSTUNREACH; /* XXX */
1353
1354         *mtup = mtu;
1355         if (alwaysfragp)
1356                 *alwaysfragp = alwaysfrag;
1357         return (error);
1358 }
1359
1360 /*
1361  * IP6 socket option processing.
1362  */
1363 void
1364 ip6_ctloutput_dispatch(netmsg_t msg)
1365 {
1366         int error;
1367
1368         error = ip6_ctloutput(msg->ctloutput.base.nm_so,
1369                               msg->ctloutput.nm_sopt);
1370         lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
1371 }
1372
1373 int
1374 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1375 {
1376         int optdatalen,uproto;
1377         int privileged;
1378         struct inpcb *in6p = so->so_pcb;
1379         void *optdata;
1380         int error, optval;
1381         int level, op, optname;
1382         int optlen;
1383         struct thread *td;
1384
1385         if (sopt) {
1386                 level = sopt->sopt_level;
1387                 op = sopt->sopt_dir;
1388                 optname = sopt->sopt_name;
1389                 optlen = sopt->sopt_valsize;
1390                 td = sopt->sopt_td;
1391         } else {
1392                 panic("ip6_ctloutput: arg soopt is NULL");
1393                 /* NOT REACHED */
1394                 td = NULL;
1395         }
1396         error = optval = 0;
1397
1398         uproto = (int)so->so_proto->pr_protocol;
1399         privileged = (td == NULL || priv_check(td, PRIV_ROOT)) ? 0 : 1;
1400
1401         if (level == IPPROTO_IPV6) {
1402                 switch (op) {
1403
1404                 case SOPT_SET:
1405                         switch (optname) {
1406                         case IPV6_2292PKTOPTIONS:
1407 #ifdef IPV6_PKTOPTIONS
1408                         case IPV6_PKTOPTIONS:
1409 #endif
1410                         {
1411                                 struct mbuf *m;
1412
1413                                 error = soopt_getm(sopt, &m); /* XXX */
1414                                 if (error != 0)
1415                                         break;
1416                                 soopt_to_mbuf(sopt, m); /* XXX */
1417                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1418                                                     m, so, sopt);
1419                                 m_freem(m); /* XXX */
1420                                 break;
1421                         }
1422
1423                         /*
1424                          * Use of some Hop-by-Hop options or some
1425                          * Destination options, might require special
1426                          * privilege.  That is, normal applications
1427                          * (without special privilege) might be forbidden
1428                          * from setting certain options in outgoing packets,
1429                          * and might never see certain options in received
1430                          * packets. [RFC 2292 Section 6]
1431                          * KAME specific note:
1432                          *  KAME prevents non-privileged users from sending or
1433                          *  receiving ANY hbh/dst options in order to avoid
1434                          *  overhead of parsing options in the kernel.
1435                          */
1436                         case IPV6_RECVHOPOPTS:
1437                         case IPV6_RECVDSTOPTS:
1438                         case IPV6_RECVRTHDRDSTOPTS:
1439                                 if (!privileged)
1440                                         return (EPERM);
1441                         case IPV6_RECVPKTINFO:
1442                         case IPV6_RECVHOPLIMIT:
1443                         case IPV6_RECVRTHDR:
1444                         case IPV6_RECVPATHMTU:
1445                         case IPV6_RECVTCLASS:
1446                         case IPV6_AUTOFLOWLABEL:
1447                         case IPV6_HOPLIMIT:
1448                         /* FALLTHROUGH */
1449                         case IPV6_UNICAST_HOPS:
1450                         case IPV6_FAITH:
1451
1452                         case IPV6_V6ONLY:
1453                                 if (optlen != sizeof(int)) {
1454                                         error = EINVAL;
1455                                         break;
1456                                 }
1457                                 error = soopt_to_kbuf(sopt, &optval,
1458                                         sizeof optval, sizeof optval);
1459                                 if (error)
1460                                         break;
1461                                 switch (optname) {
1462
1463                                 case IPV6_UNICAST_HOPS:
1464                                         if (optval < -1 || optval >= 256)
1465                                                 error = EINVAL;
1466                                         else {
1467                                                 /* -1 = kernel default */
1468                                                 in6p->in6p_hops = optval;
1469
1470                                                 if ((in6p->in6p_vflag &
1471                                                      INP_IPV4) != 0)
1472                                                         in6p->inp_ip_ttl = optval;
1473                                         }
1474                                         break;
1475 #define OPTSET(bit) \
1476 do { \
1477         if (optval) \
1478                 in6p->in6p_flags |= (bit); \
1479         else \
1480                 in6p->in6p_flags &= ~(bit); \
1481 } while (0)
1482 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1483 /*
1484  * Although changed to RFC3542, It's better to also support RFC2292 API
1485  */
1486 #define OPTSET2292(bit) \
1487 do { \
1488         in6p->in6p_flags |= IN6P_RFC2292; \
1489         if (optval) \
1490                 in6p->in6p_flags |= (bit); \
1491         else \
1492                 in6p->in6p_flags &= ~(bit); \
1493 } while (/*CONSTCOND*/ 0)
1494
1495                                 case IPV6_RECVPKTINFO:
1496                                         /* cannot mix with RFC2292 */
1497                                         if (OPTBIT(IN6P_RFC2292)) {
1498                                                 error = EINVAL;
1499                                                 break;
1500                                         }
1501                                         OPTSET(IN6P_PKTINFO);
1502                                         break;
1503
1504                                 case IPV6_HOPLIMIT:
1505                                 {
1506                                         struct ip6_pktopts **optp;
1507
1508                                         /* cannot mix with RFC2292 */
1509                                         if (OPTBIT(IN6P_RFC2292)) {
1510                                                 error = EINVAL;
1511                                                 break;
1512                                         }
1513                                         optp = &in6p->in6p_outputopts;
1514                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1515                                             (u_char *)&optval, sizeof(optval),
1516                                             optp, uproto);
1517                                         break;
1518                                 }
1519
1520                                 case IPV6_RECVHOPLIMIT:
1521                                         /* cannot mix with RFC2292 */
1522                                         if (OPTBIT(IN6P_RFC2292)) {
1523                                                 error = EINVAL;
1524                                                 break;
1525                                         }
1526                                         OPTSET(IN6P_HOPLIMIT);
1527                                         break;
1528
1529                                 case IPV6_RECVHOPOPTS:
1530                                         /* cannot mix with RFC2292 */
1531                                         if (OPTBIT(IN6P_RFC2292)) {
1532                                                 error = EINVAL;
1533                                                 break;
1534                                         }
1535                                         OPTSET(IN6P_HOPOPTS);
1536                                         break;
1537
1538                                 case IPV6_RECVDSTOPTS:
1539                                         /* cannot mix with RFC2292 */
1540                                         if (OPTBIT(IN6P_RFC2292)) {
1541                                                 error = EINVAL;
1542                                                 break;
1543                                         }
1544                                         OPTSET(IN6P_DSTOPTS);
1545                                         break;
1546
1547                                 case IPV6_RECVRTHDRDSTOPTS:
1548                                         /* cannot mix with RFC2292 */
1549                                         if (OPTBIT(IN6P_RFC2292)) {
1550                                                 error = EINVAL;
1551                                                 break;
1552                                         }
1553                                         OPTSET(IN6P_RTHDRDSTOPTS);
1554                                         break;
1555
1556                                 case IPV6_RECVRTHDR:
1557                                         /* cannot mix with RFC2292 */
1558                                         if (OPTBIT(IN6P_RFC2292)) {
1559                                                 error = EINVAL;
1560                                                 break;
1561                                         }
1562                                         OPTSET(IN6P_RTHDR);
1563                                         break;
1564
1565                                 case IPV6_RECVPATHMTU:
1566                                         /*
1567                                          * We ignore this option for TCP
1568                                          * sockets.
1569                                          * (RFC3542 leaves this case
1570                                          * unspecified.)
1571                                          */
1572                                         if (uproto != IPPROTO_TCP)
1573                                                 OPTSET(IN6P_MTU);
1574                                         break;
1575
1576                                 case IPV6_RECVTCLASS:
1577                                         /* cannot mix with RFC2292 XXX */
1578                                         if (OPTBIT(IN6P_RFC2292)) {
1579                                                 error = EINVAL;
1580                                                 break;
1581                                         }
1582                                         OPTSET(IN6P_TCLASS);
1583                                         break;
1584
1585                                 case IPV6_AUTOFLOWLABEL:
1586                                         OPTSET(IN6P_AUTOFLOWLABEL);
1587                                         break;
1588
1589                                 case IPV6_FAITH:
1590                                         OPTSET(IN6P_FAITH);
1591                                         break;
1592
1593                                 case IPV6_V6ONLY:
1594                                         /*
1595                                          * make setsockopt(IPV6_V6ONLY)
1596                                          * available only prior to bind(2).
1597                                          */
1598                                         if (in6p->in6p_lport ||
1599                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1600                                         {
1601                                                 error = EINVAL;
1602                                                 break;
1603                                         }
1604                                         OPTSET(IN6P_IPV6_V6ONLY);
1605                                         if (optval)
1606                                                 in6p->in6p_vflag &= ~INP_IPV4;
1607                                         else
1608                                                 in6p->in6p_vflag |= INP_IPV4;
1609                                         break;
1610                                 }
1611                                 break;
1612
1613                         case IPV6_TCLASS:
1614                         case IPV6_DONTFRAG:
1615                         case IPV6_USE_MIN_MTU:
1616                         case IPV6_PREFER_TEMPADDR:
1617                                 if (optlen != sizeof(optval)) {
1618                                         error = EINVAL;
1619                                         break;
1620                                 }
1621                                 error = soopt_to_kbuf(sopt, &optval,
1622                                         sizeof optval, sizeof optval);
1623                                 if (error)
1624                                         break;
1625                                 {
1626                                         struct ip6_pktopts **optp;
1627                                         optp = &in6p->in6p_outputopts;
1628                                         error = ip6_pcbopt(optname,
1629                                             (u_char *)&optval, sizeof(optval),
1630                                             optp, uproto);
1631                                         break;
1632                                 }
1633
1634                         case IPV6_2292PKTINFO:
1635                         case IPV6_2292HOPLIMIT:
1636                         case IPV6_2292HOPOPTS:
1637                         case IPV6_2292DSTOPTS:
1638                         case IPV6_2292RTHDR:
1639                                 /* RFC 2292 */
1640                                 if (optlen != sizeof(int)) {
1641                                         error = EINVAL;
1642                                         break;
1643                                 }
1644                                 error = soopt_to_kbuf(sopt, &optval,
1645                                         sizeof optval, sizeof optval);
1646                                 if (error)
1647                                         break;
1648                                 switch (optname) {
1649                                 case IPV6_2292PKTINFO:
1650                                         OPTSET2292(IN6P_PKTINFO);
1651                                         break;
1652                                 case IPV6_2292HOPLIMIT:
1653                                         OPTSET2292(IN6P_HOPLIMIT);
1654                                         break;
1655                                 case IPV6_2292HOPOPTS:
1656                                         /*
1657                                          * Check super-user privilege.
1658                                          * See comments for IPV6_RECVHOPOPTS.
1659                                          */
1660                                         if (!privileged)
1661                                                 return (EPERM);
1662                                         OPTSET2292(IN6P_HOPOPTS);
1663                                         break;
1664                                 case IPV6_2292DSTOPTS:
1665                                         if (!privileged)
1666                                                 return (EPERM);
1667                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1668                                         break;
1669                                 case IPV6_2292RTHDR:
1670                                         OPTSET2292(IN6P_RTHDR);
1671                                         break;
1672                                 }
1673                                 break;
1674
1675                         case IPV6_PKTINFO:
1676                         case IPV6_HOPOPTS:
1677                         case IPV6_RTHDR:
1678                         case IPV6_DSTOPTS:
1679                         case IPV6_RTHDRDSTOPTS:
1680                         case IPV6_NEXTHOP:
1681                         {
1682                                 /*
1683                                  * New advanced API (RFC3542)
1684                                  */
1685                                 u_char *optbuf;
1686                                 u_char optbuf_storage[MCLBYTES];
1687                                 int optlen;
1688                                 struct ip6_pktopts **optp;
1689
1690                                 /* cannot mix with RFC2292 */
1691                                 if (OPTBIT(IN6P_RFC2292)) {
1692                                         error = EINVAL;
1693                                         break;
1694                                 }
1695
1696                                 /*
1697                                  * We only ensure valsize is not too large
1698                                  * here.  Further validation will be done
1699                                  * later.
1700                                  */
1701                                 error = soopt_to_kbuf(sopt, optbuf_storage,
1702                                     sizeof(optbuf_storage), 0);
1703                                 if (error)
1704                                         break;
1705                                 optlen = sopt->sopt_valsize;
1706                                 optbuf = optbuf_storage;
1707                                 optp = &in6p->in6p_outputopts;
1708                                 error = ip6_pcbopt(optname, optbuf, optlen,
1709                                     optp, uproto);
1710                                 break;
1711                         }
1712 #undef OPTSET
1713
1714                         case IPV6_MULTICAST_IF:
1715                         case IPV6_MULTICAST_HOPS:
1716                         case IPV6_MULTICAST_LOOP:
1717                         case IPV6_JOIN_GROUP:
1718                         case IPV6_LEAVE_GROUP:
1719                             {
1720                                 struct mbuf *m;
1721                                 if (sopt->sopt_valsize > MLEN) {
1722                                         error = EMSGSIZE;
1723                                         break;
1724                                 }
1725                                 /* XXX */
1726                                 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1727                                 if (m == NULL) {
1728                                         error = ENOBUFS;
1729                                         break;
1730                                 }
1731                                 m->m_len = sopt->sopt_valsize;
1732                                 error = soopt_to_kbuf(sopt, mtod(m, char *),
1733                                                     m->m_len, m->m_len);
1734                                 error = ip6_setmoptions(sopt->sopt_name,
1735                                                         &in6p->in6p_moptions,
1736                                                         m);
1737                                 m_free(m);
1738                             }
1739                                 break;
1740
1741                         case IPV6_PORTRANGE:
1742                                 error = soopt_to_kbuf(sopt, &optval,
1743                                     sizeof optval, sizeof optval);
1744                                 if (error)
1745                                         break;
1746
1747                                 switch (optval) {
1748                                 case IPV6_PORTRANGE_DEFAULT:
1749                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1750                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1751                                         break;
1752
1753                                 case IPV6_PORTRANGE_HIGH:
1754                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1755                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1756                                         break;
1757
1758                                 case IPV6_PORTRANGE_LOW:
1759                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1760                                         in6p->in6p_flags |= IN6P_LOWPORT;
1761                                         break;
1762
1763                                 default:
1764                                         error = EINVAL;
1765                                         break;
1766                                 }
1767                                 break;
1768
1769 #if defined(IPSEC) || defined(FAST_IPSEC)
1770                         case IPV6_IPSEC_POLICY:
1771                             {
1772                                 caddr_t req = NULL;
1773                                 size_t len = 0;
1774                                 struct mbuf *m;
1775
1776                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1777                                         break;
1778                                 soopt_to_mbuf(sopt, m);         /* XXX */
1779                                 if (m) {
1780                                         req = mtod(m, caddr_t);
1781                                         len = m->m_len;
1782                                 }
1783                                 error = ipsec6_set_policy(in6p, optname, req,
1784                                                           len, privileged);
1785                                 m_freem(m);
1786                             }
1787                                 break;
1788 #endif /* KAME IPSEC */
1789
1790                         case IPV6_FW_ADD:
1791                         case IPV6_FW_DEL:
1792                         case IPV6_FW_FLUSH:
1793                         case IPV6_FW_ZERO:
1794                             {
1795                                 struct mbuf *m;
1796                                 struct mbuf **mp = &m;
1797
1798                                 if (ip6_fw_ctl_ptr == NULL)
1799                                         return EINVAL;
1800                                 /* XXX */
1801                                 if ((error = soopt_getm(sopt, &m)) != 0)
1802                                         break;
1803                                 /* XXX */
1804                                 soopt_to_mbuf(sopt, m);
1805                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1806                                 m = *mp;
1807                             }
1808                                 break;
1809
1810                         default:
1811                                 error = ENOPROTOOPT;
1812                                 break;
1813                         }
1814                         break;
1815
1816                 case SOPT_GET:
1817                         switch (optname) {
1818                         case IPV6_2292PKTOPTIONS:
1819 #ifdef IPV6_PKTOPTIONS
1820                         case IPV6_PKTOPTIONS:
1821 #endif
1822                                 /*
1823                                  * RFC3542 (effectively) deprecated the
1824                                  * semantics of the 2292-style pktoptions.
1825                                  * Since it was not reliable in nature (i.e.,
1826                                  * applications had to expect the lack of some
1827                                  * information after all), it would make sense
1828                                  * to simplify this part by always returning
1829                                  * empty data.
1830                                  */
1831                                 if (in6p->in6p_options) {
1832                                         struct mbuf *m;
1833                                         m = m_copym(in6p->in6p_options,
1834                                             0, M_COPYALL, MB_WAIT);
1835                                         error = soopt_from_mbuf(sopt, m);
1836                                         if (error == 0)
1837                                                 m_freem(m);
1838                                 } else
1839                                         sopt->sopt_valsize = 0;
1840                                 break;
1841
1842                         case IPV6_RECVHOPOPTS:
1843                         case IPV6_RECVDSTOPTS:
1844                         case IPV6_RECVRTHDRDSTOPTS:
1845                         case IPV6_UNICAST_HOPS:
1846                         case IPV6_RECVPKTINFO:
1847                         case IPV6_RECVHOPLIMIT:
1848                         case IPV6_RECVRTHDR:
1849                         case IPV6_RECVPATHMTU:
1850                         case IPV6_RECVTCLASS:
1851                         case IPV6_AUTOFLOWLABEL:
1852                         case IPV6_FAITH:
1853                         case IPV6_V6ONLY:
1854                         case IPV6_PORTRANGE:
1855                                 switch (optname) {
1856
1857                                 case IPV6_RECVHOPOPTS:
1858                                         optval = OPTBIT(IN6P_HOPOPTS);
1859                                         break;
1860
1861                                 case IPV6_RECVDSTOPTS:
1862                                         optval = OPTBIT(IN6P_DSTOPTS);
1863                                         break;
1864
1865                                 case IPV6_RECVRTHDRDSTOPTS:
1866                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1867                                         break;
1868
1869                                 case IPV6_RECVPKTINFO:
1870                                         optval = OPTBIT(IN6P_PKTINFO);
1871                                         break;
1872
1873                                 case IPV6_RECVHOPLIMIT:
1874                                         optval = OPTBIT(IN6P_HOPLIMIT);
1875                                         break;
1876
1877                                 case IPV6_RECVRTHDR:
1878                                         optval = OPTBIT(IN6P_RTHDR);
1879                                         break;
1880
1881                                 case IPV6_RECVPATHMTU:
1882                                         optval = OPTBIT(IN6P_MTU);
1883                                         break;
1884
1885                                 case IPV6_RECVTCLASS:
1886                                         optval = OPTBIT(IN6P_TCLASS);
1887                                         break;
1888
1889                                 case IPV6_AUTOFLOWLABEL:
1890                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1891                                         break;
1892
1893
1894                                 case IPV6_UNICAST_HOPS:
1895                                         optval = in6p->in6p_hops;
1896                                         break;
1897
1898                                 case IPV6_FAITH:
1899                                         optval = OPTBIT(IN6P_FAITH);
1900                                         break;
1901
1902                                 case IPV6_V6ONLY:
1903                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1904                                         break;
1905
1906                                 case IPV6_PORTRANGE:
1907                                     {
1908                                         int flags;
1909                                         flags = in6p->in6p_flags;
1910                                         if (flags & IN6P_HIGHPORT)
1911                                                 optval = IPV6_PORTRANGE_HIGH;
1912                                         else if (flags & IN6P_LOWPORT)
1913                                                 optval = IPV6_PORTRANGE_LOW;
1914                                         else
1915                                                 optval = 0;
1916                                         break;
1917                                     }
1918                                 }
1919                                 soopt_from_kbuf(sopt, &optval,
1920                                         sizeof optval);
1921                                 break;
1922
1923                         case IPV6_PATHMTU:
1924                         {
1925                                 u_long pmtu = 0;
1926                                 struct ip6_mtuinfo mtuinfo;
1927                                 struct route_in6 sro;
1928
1929                                 bzero(&sro, sizeof(sro));
1930
1931                                 if (!(so->so_state & SS_ISCONNECTED))
1932                                         return (ENOTCONN);
1933                                 /*
1934                                  * XXX: we dot not consider the case of source
1935                                  * routing, or optional information to specify
1936                                  * the outgoing interface.
1937                                  */
1938                                 error = ip6_getpmtu(&sro, NULL, NULL,
1939                                     &in6p->in6p_faddr, &pmtu, NULL);
1940                                 if (sro.ro_rt)
1941                                         RTFREE(sro.ro_rt);
1942                                 if (error)
1943                                         break;
1944                                 if (pmtu > IPV6_MAXPACKET)
1945                                         pmtu = IPV6_MAXPACKET;
1946
1947                                 bzero(&mtuinfo, sizeof(mtuinfo));
1948                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1949                                 optdata = (void *)&mtuinfo;
1950                                 optdatalen = sizeof(mtuinfo);
1951                                 soopt_from_kbuf(sopt, optdata,
1952                                     optdatalen);
1953                                 break;
1954                         }
1955
1956                         case IPV6_2292PKTINFO:
1957                         case IPV6_2292HOPLIMIT:
1958                         case IPV6_2292HOPOPTS:
1959                         case IPV6_2292RTHDR:
1960                         case IPV6_2292DSTOPTS:
1961                                 if (optname == IPV6_2292HOPOPTS ||
1962                                     optname == IPV6_2292DSTOPTS ||
1963                                     !privileged)
1964                                         return (EPERM);
1965                                 switch (optname) {
1966                                 case IPV6_2292PKTINFO:
1967                                         optval = OPTBIT(IN6P_PKTINFO);
1968                                         break;
1969                                 case IPV6_2292HOPLIMIT:
1970                                         optval = OPTBIT(IN6P_HOPLIMIT);
1971                                         break;
1972                                 case IPV6_2292HOPOPTS:
1973                                         if (!privileged)
1974                                                 return (EPERM);
1975                                         optval = OPTBIT(IN6P_HOPOPTS);
1976                                         break;
1977                                 case IPV6_2292RTHDR:
1978                                         optval = OPTBIT(IN6P_RTHDR);
1979                                         break;
1980                                 case IPV6_2292DSTOPTS:
1981                                         if (!privileged)
1982                                                 return (EPERM);
1983                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1984                                         break;
1985                                 }
1986                                 soopt_from_kbuf(sopt, &optval,
1987                                         sizeof optval);
1988                                 break;
1989
1990                         case IPV6_PKTINFO:
1991                         case IPV6_HOPOPTS:
1992                         case IPV6_RTHDR:
1993                         case IPV6_DSTOPTS:
1994                         case IPV6_RTHDRDSTOPTS:
1995                         case IPV6_NEXTHOP:
1996                         case IPV6_TCLASS:
1997                         case IPV6_DONTFRAG:
1998                         case IPV6_USE_MIN_MTU:
1999                         case IPV6_PREFER_TEMPADDR:
2000                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2001                                     optname, sopt);
2002                                 break;
2003
2004                         case IPV6_MULTICAST_IF:
2005                         case IPV6_MULTICAST_HOPS:
2006                         case IPV6_MULTICAST_LOOP:
2007                         case IPV6_JOIN_GROUP:
2008                         case IPV6_LEAVE_GROUP:
2009                             {
2010                                 struct mbuf *m;
2011                                 error = ip6_getmoptions(sopt->sopt_name,
2012                                                 in6p->in6p_moptions, &m);
2013                                 if (error == 0)
2014                                         soopt_from_kbuf(sopt,
2015                                                 mtod(m, char *), m->m_len);
2016                                 m_freem(m);
2017                             }
2018                                 break;
2019
2020 #if defined(IPSEC) || defined(FAST_IPSEC)
2021                         case IPV6_IPSEC_POLICY:
2022                           {
2023                                 caddr_t req = NULL;
2024                                 size_t len = 0;
2025                                 struct mbuf *m = NULL;
2026                                 struct mbuf **mp = &m;
2027
2028                                 error = soopt_getm(sopt, &m); /* XXX */
2029                                 if (error != 0)
2030                                         break;
2031                                 soopt_to_mbuf(sopt, m); /* XXX */
2032                                 if (m) {
2033                                         req = mtod(m, caddr_t);
2034                                         len = m->m_len;
2035                                 }
2036                                 error = ipsec6_get_policy(in6p, req, len, mp);
2037                                 if (error == 0)
2038                                         error = soopt_from_mbuf(sopt, m); /*XXX*/
2039                                 if (error == 0 && m != NULL)
2040                                         m_freem(m);
2041                                 break;
2042                           }
2043 #endif /* KAME IPSEC */
2044
2045                         case IPV6_FW_GET:
2046                           {
2047                                 struct mbuf *m;
2048                                 struct mbuf **mp = &m;
2049
2050                                 if (ip6_fw_ctl_ptr == NULL)
2051                                 {
2052                                         return EINVAL;
2053                                 }
2054                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
2055                                 if (error == 0)
2056                                         error = soopt_from_mbuf(sopt, m); /* XXX */
2057                                 if (error == 0 && m != NULL)
2058                                         m_freem(m);
2059                           }
2060                                 break;
2061
2062                         default:
2063                                 error = ENOPROTOOPT;
2064                                 break;
2065                         }
2066                         break;
2067                 }
2068         } else {
2069                 error = EINVAL;
2070         }
2071         return (error);
2072 }
2073
2074 int
2075 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2076 {
2077         int error = 0, optval, optlen;
2078         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2079         struct in6pcb *in6p = sotoin6pcb(so);
2080         int level, op, optname;
2081
2082         if (sopt) {
2083                 level = sopt->sopt_level;
2084                 op = sopt->sopt_dir;
2085                 optname = sopt->sopt_name;
2086                 optlen = sopt->sopt_valsize;
2087         } else
2088                 panic("ip6_raw_ctloutput: arg soopt is NULL");
2089
2090         if (level != IPPROTO_IPV6) {
2091                 return (EINVAL);
2092         }
2093
2094         switch (optname) {
2095         case IPV6_CHECKSUM:
2096                 /*
2097                  * For ICMPv6 sockets, no modification allowed for checksum
2098                  * offset, permit "no change" values to help existing apps.
2099                  *
2100                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2101                  * for an ICMPv6 socket will fail."
2102                  * The current behavior does not meet RFC3542.
2103                  */
2104                 switch (op) {
2105                 case SOPT_SET:
2106                         if (optlen != sizeof(int)) {
2107                                 error = EINVAL;
2108                                 break;
2109                         }
2110                         error = soopt_to_kbuf(sopt, &optval,
2111                                     sizeof optval, sizeof optval);
2112                         if (error)
2113                                 break;
2114                         if ((optval % 2) != 0) {
2115                                 /* the API assumes even offset values */
2116                                 error = EINVAL;
2117                         } else if (so->so_proto->pr_protocol ==
2118                             IPPROTO_ICMPV6) {
2119                                 if (optval != icmp6off)
2120                                         error = EINVAL;
2121                         } else
2122                                 in6p->in6p_cksum = optval;
2123                         break;
2124
2125                 case SOPT_GET:
2126                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2127                                 optval = icmp6off;
2128                         else
2129                                 optval = in6p->in6p_cksum;
2130
2131                         soopt_from_kbuf(sopt, &optval, sizeof(optval));
2132                         break;
2133
2134                 default:
2135                         error = EINVAL;
2136                         break;
2137                 }
2138                 break;
2139
2140         default:
2141                 error = ENOPROTOOPT;
2142                 break;
2143         }
2144
2145         return (error);
2146 }
2147
2148 /*
2149  * Set up IP6 options in pcb for insertion in output packets or
2150  * specifying behavior of outgoing packets.
2151  */
2152 static int
2153 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2154     struct socket *so, struct sockopt *sopt)
2155 {
2156         int priv = 0;
2157         struct ip6_pktopts *opt = *pktopt;
2158         int error = 0;
2159
2160         /* turn off any old options. */
2161         if (opt) {
2162 #ifdef DIAGNOSTIC
2163                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2164                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2165                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2166                         kprintf("ip6_pcbopts: all specified options are cleared.\n");
2167 #endif
2168                 ip6_clearpktopts(opt, -1);
2169         } else
2170                 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2171         *pktopt = NULL;
2172
2173         if (!m || m->m_len == 0) {
2174                 /*
2175                  * Only turning off any previous options, regardless of
2176                  * whether the opt is just created or given.
2177                  */
2178                 kfree(opt, M_IP6OPT);
2179                 return (0);
2180         }
2181
2182         /*  set options specified by user. */
2183         if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) {
2184                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2185                 kfree(opt, M_IP6OPT);
2186                 return (error);
2187         }
2188         *pktopt = opt;
2189         return (0);
2190 }
2191
2192
2193 /*
2194  * Below three functions are introduced by merge to RFC3542
2195  */
2196
2197 static int
2198 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2199 {
2200         void *optdata = NULL;
2201         int optdatalen = 0;
2202         struct ip6_ext *ip6e;
2203         int error = 0;
2204         struct in6_pktinfo null_pktinfo;
2205         int deftclass = 0, on;
2206         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2207         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2208
2209         switch (optname) {
2210         case IPV6_PKTINFO:
2211                 if (pktopt && pktopt->ip6po_pktinfo)
2212                         optdata = (void *)pktopt->ip6po_pktinfo;
2213                 else {
2214                         /* XXX: we don't have to do this every time... */
2215                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2216                         optdata = (void *)&null_pktinfo;
2217                 }
2218                 optdatalen = sizeof(struct in6_pktinfo);
2219                 break;
2220         case IPV6_TCLASS:
2221                 if (pktopt && pktopt->ip6po_tclass >= 0)
2222                         optdata = (void *)&pktopt->ip6po_tclass;
2223                 else
2224                         optdata = (void *)&deftclass;
2225                 optdatalen = sizeof(int);
2226                 break;
2227         case IPV6_HOPOPTS:
2228                 if (pktopt && pktopt->ip6po_hbh) {
2229                         optdata = (void *)pktopt->ip6po_hbh;
2230                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2231                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2232                 }
2233                 break;
2234         case IPV6_RTHDR:
2235                 if (pktopt && pktopt->ip6po_rthdr) {
2236                         optdata = (void *)pktopt->ip6po_rthdr;
2237                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2238                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2239                 }
2240                 break;
2241         case IPV6_RTHDRDSTOPTS:
2242                 if (pktopt && pktopt->ip6po_dest1) {
2243                         optdata = (void *)pktopt->ip6po_dest1;
2244                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2245                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2246                 }
2247                 break;
2248         case IPV6_DSTOPTS:
2249                 if (pktopt && pktopt->ip6po_dest2) {
2250                         optdata = (void *)pktopt->ip6po_dest2;
2251                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2252                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2253                 }
2254                 break;
2255         case IPV6_NEXTHOP:
2256                 if (pktopt && pktopt->ip6po_nexthop) {
2257                         optdata = (void *)pktopt->ip6po_nexthop;
2258                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2259                 }
2260                 break;
2261         case IPV6_USE_MIN_MTU:
2262                 if (pktopt)
2263                         optdata = (void *)&pktopt->ip6po_minmtu;
2264                 else
2265                         optdata = (void *)&defminmtu;
2266                 optdatalen = sizeof(int);
2267                 break;
2268         case IPV6_DONTFRAG:
2269                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2270                         on = 1;
2271                 else
2272                         on = 0;
2273                 optdata = (void *)&on;
2274                 optdatalen = sizeof(on);
2275                 break;
2276         case IPV6_PREFER_TEMPADDR:
2277                 if (pktopt)
2278                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2279                 else
2280                         optdata = (void *)&defpreftemp;
2281                 optdatalen = sizeof(int);
2282                 break;
2283         default:                /* should not happen */
2284 #ifdef DIAGNOSTIC
2285                 panic("ip6_getpcbopt: unexpected option");
2286 #endif
2287                 return (ENOPROTOOPT);
2288         }
2289
2290         soopt_from_kbuf(sopt, optdata, optdatalen);
2291
2292         return (error);
2293 }
2294
2295 /*
2296  * initialize ip6_pktopts.  beware that there are non-zero default values in
2297  * the struct.
2298  */
2299
2300 static int
2301 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto)
2302 {
2303         struct ip6_pktopts *opt;
2304         int priv =0;
2305         if (*pktopt == NULL) {
2306                 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2307                 init_ip6pktopts(*pktopt);
2308         }
2309         opt = *pktopt;
2310
2311         return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv));
2312 }
2313
2314 /*
2315  * initialize ip6_pktopts.  beware that there are non-zero default values in
2316  * the struct.
2317  */
2318 void
2319 init_ip6pktopts(struct ip6_pktopts *opt)
2320 {
2321
2322         bzero(opt, sizeof(*opt));
2323         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2324         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2325         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2326         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2327 }
2328
2329 void
2330 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2331 {
2332         if (pktopt == NULL)
2333                 return;
2334
2335         if (optname == -1 || optname == IPV6_PKTINFO) {
2336                 if (pktopt->ip6po_pktinfo)
2337                         kfree(pktopt->ip6po_pktinfo, M_IP6OPT);
2338                 pktopt->ip6po_pktinfo = NULL;
2339         }
2340         if (optname == -1 || optname == IPV6_HOPLIMIT)
2341                 pktopt->ip6po_hlim = -1;
2342         if (optname == -1 || optname == IPV6_TCLASS)
2343                 pktopt->ip6po_tclass = -1;
2344         if (optname == -1 || optname == IPV6_NEXTHOP) {
2345                 if (pktopt->ip6po_nextroute.ro_rt) {
2346                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2347                         pktopt->ip6po_nextroute.ro_rt = NULL;
2348                 }
2349                 if (pktopt->ip6po_nexthop)
2350                         kfree(pktopt->ip6po_nexthop, M_IP6OPT);
2351                 pktopt->ip6po_nexthop = NULL;
2352         }
2353         if (optname == -1 || optname == IPV6_HOPOPTS) {
2354                 if (pktopt->ip6po_hbh)
2355                         kfree(pktopt->ip6po_hbh, M_IP6OPT);
2356                 pktopt->ip6po_hbh = NULL;
2357         }
2358         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2359                 if (pktopt->ip6po_dest1)
2360                         kfree(pktopt->ip6po_dest1, M_IP6OPT);
2361                 pktopt->ip6po_dest1 = NULL;
2362         }
2363         if (optname == -1 || optname == IPV6_RTHDR) {
2364                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2365                         kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2366                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2367                 if (pktopt->ip6po_route.ro_rt) {
2368                         RTFREE(pktopt->ip6po_route.ro_rt);
2369                         pktopt->ip6po_route.ro_rt = NULL;
2370                 }
2371         }
2372         if (optname == -1 || optname == IPV6_DSTOPTS) {
2373                 if (pktopt->ip6po_dest2)
2374                         kfree(pktopt->ip6po_dest2, M_IP6OPT);
2375                 pktopt->ip6po_dest2 = NULL;
2376         }
2377 }
2378
2379 #define PKTOPT_EXTHDRCPY(type) \
2380 do {\
2381         if (src->type) {\
2382                 int hlen =\
2383                         (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2384                 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\
2385                 if (dst->type == NULL)\
2386                         goto bad;\
2387                 bcopy(src->type, dst->type, hlen);\
2388         }\
2389 } while (0)
2390
2391 struct ip6_pktopts *
2392 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2393 {
2394         struct ip6_pktopts *dst;
2395
2396         if (src == NULL) {
2397                 kprintf("ip6_clearpktopts: invalid argument\n");
2398                 return (NULL);
2399         }
2400
2401         dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO);
2402         if (dst == NULL)
2403                 return (NULL);
2404
2405         dst->ip6po_hlim = src->ip6po_hlim;
2406         if (src->ip6po_pktinfo) {
2407                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2408                                             M_IP6OPT, canwait);
2409                 if (dst->ip6po_pktinfo == NULL)
2410                         goto bad;
2411                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2412         }
2413         if (src->ip6po_nexthop) {
2414                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2415                                             M_IP6OPT, canwait);
2416                 if (dst->ip6po_nexthop == NULL)
2417                         goto bad;
2418                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2419                       src->ip6po_nexthop->sa_len);
2420         }
2421         PKTOPT_EXTHDRCPY(ip6po_hbh);
2422         PKTOPT_EXTHDRCPY(ip6po_dest1);
2423         PKTOPT_EXTHDRCPY(ip6po_dest2);
2424         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2425         return (dst);
2426
2427 bad:
2428         if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT);
2429         if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT);
2430         if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT);
2431         if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT);
2432         if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT);
2433         if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT);
2434         kfree(dst, M_IP6OPT);
2435         return (NULL);
2436 }
2437
2438 static int
2439 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2440 {
2441         if (dst == NULL || src == NULL)  {
2442 #ifdef DIAGNOSTIC
2443                 kprintf("ip6_clearpktopts: invalid argument\n");
2444 #endif
2445                 return (EINVAL);
2446         }
2447
2448         dst->ip6po_hlim = src->ip6po_hlim;
2449         dst->ip6po_tclass = src->ip6po_tclass;
2450         dst->ip6po_flags = src->ip6po_flags;
2451         if (src->ip6po_pktinfo) {
2452                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2453                     M_IP6OPT, canwait);
2454                 if (dst->ip6po_pktinfo == NULL)
2455                         goto bad;
2456                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2457         }
2458         if (src->ip6po_nexthop) {
2459                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2460                     M_IP6OPT, canwait);
2461                 if (dst->ip6po_nexthop == NULL)
2462                         goto bad;
2463                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2464                     src->ip6po_nexthop->sa_len);
2465         }
2466         PKTOPT_EXTHDRCPY(ip6po_hbh);
2467         PKTOPT_EXTHDRCPY(ip6po_dest1);
2468         PKTOPT_EXTHDRCPY(ip6po_dest2);
2469         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2470         return (0);
2471
2472   bad:
2473         ip6_clearpktopts(dst, -1);
2474         return (ENOBUFS);
2475 }
2476 #undef PKTOPT_EXTHDRCPY
2477
2478 void
2479 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2480 {
2481         if (pktopt == NULL)
2482                 return;
2483
2484         ip6_clearpktopts(pktopt, -1);
2485
2486         kfree(pktopt, M_IP6OPT);
2487 }
2488
2489 /*
2490  * Set the IP6 multicast options in response to user setsockopt().
2491  */
2492 static int
2493 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2494 {
2495         int error = 0;
2496         u_int loop, ifindex;
2497         struct ipv6_mreq *mreq;
2498         struct ifnet *ifp;
2499         struct ip6_moptions *im6o = *im6op;
2500         struct route_in6 ro;
2501         struct sockaddr_in6 *dst;
2502         struct in6_multi_mship *imm;
2503         struct thread *td = curthread;  /* XXX */
2504
2505         if (im6o == NULL) {
2506                 /*
2507                  * No multicast option buffer attached to the pcb;
2508                  * allocate one and initialize to default values.
2509                  */
2510                 im6o = (struct ip6_moptions *)
2511                         kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2512
2513                 *im6op = im6o;
2514                 im6o->im6o_multicast_ifp = NULL;
2515                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2516                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2517                 LIST_INIT(&im6o->im6o_memberships);
2518         }
2519
2520         switch (optname) {
2521
2522         case IPV6_MULTICAST_IF:
2523                 /*
2524                  * Select the interface for outgoing multicast packets.
2525                  */
2526                 if (m == NULL || m->m_len != sizeof(u_int)) {
2527                         error = EINVAL;
2528                         break;
2529                 }
2530                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2531                 if (ifindex < 0 || if_index < ifindex) {
2532                         error = ENXIO;  /* XXX EINVAL? */
2533                         break;
2534                 }
2535                 ifp = ifindex2ifnet[ifindex];
2536                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2537                         error = EADDRNOTAVAIL;
2538                         break;
2539                 }
2540                 im6o->im6o_multicast_ifp = ifp;
2541                 break;
2542
2543         case IPV6_MULTICAST_HOPS:
2544             {
2545                 /*
2546                  * Set the IP6 hoplimit for outgoing multicast packets.
2547                  */
2548                 int optval;
2549                 if (m == NULL || m->m_len != sizeof(int)) {
2550                         error = EINVAL;
2551                         break;
2552                 }
2553                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2554                 if (optval < -1 || optval >= 256)
2555                         error = EINVAL;
2556                 else if (optval == -1)
2557                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2558                 else
2559                         im6o->im6o_multicast_hlim = optval;
2560                 break;
2561             }
2562
2563         case IPV6_MULTICAST_LOOP:
2564                 /*
2565                  * Set the loopback flag for outgoing multicast packets.
2566                  * Must be zero or one.
2567                  */
2568                 if (m == NULL || m->m_len != sizeof(u_int)) {
2569                         error = EINVAL;
2570                         break;
2571                 }
2572                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2573                 if (loop > 1) {
2574                         error = EINVAL;
2575                         break;
2576                 }
2577                 im6o->im6o_multicast_loop = loop;
2578                 break;
2579
2580         case IPV6_JOIN_GROUP:
2581                 /*
2582                  * Add a multicast group membership.
2583                  * Group must be a valid IP6 multicast address.
2584                  */
2585                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2586                         error = EINVAL;
2587                         break;
2588                 }
2589                 mreq = mtod(m, struct ipv6_mreq *);
2590                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2591                         /*
2592                          * We use the unspecified address to specify to accept
2593                          * all multicast addresses. Only super user is allowed
2594                          * to do this.
2595                          */
2596                         if (priv_check(td, PRIV_ROOT))
2597                         {
2598                                 error = EACCES;
2599                                 break;
2600                         }
2601                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2602                         error = EINVAL;
2603                         break;
2604                 }
2605
2606                 /*
2607                  * If the interface is specified, validate it.
2608                  */
2609                 if (mreq->ipv6mr_interface < 0
2610                  || if_index < mreq->ipv6mr_interface) {
2611                         error = ENXIO;  /* XXX EINVAL? */
2612                         break;
2613                 }
2614                 /*
2615                  * If no interface was explicitly specified, choose an
2616                  * appropriate one according to the given multicast address.
2617                  */
2618                 if (mreq->ipv6mr_interface == 0) {
2619                         /*
2620                          * If the multicast address is in node-local scope,
2621                          * the interface should be a loopback interface.
2622                          * Otherwise, look up the routing table for the
2623                          * address, and choose the outgoing interface.
2624                          *   XXX: is it a good approach?
2625                          */
2626                         if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2627                                 ifp = &loif[0];
2628                         } else {
2629                                 ro.ro_rt = NULL;
2630                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2631                                 bzero(dst, sizeof(*dst));
2632                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2633                                 dst->sin6_family = AF_INET6;
2634                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2635                                 rtalloc((struct route *)&ro);
2636                                 if (ro.ro_rt == NULL) {
2637                                         error = EADDRNOTAVAIL;
2638                                         break;
2639                                 }
2640                                 ifp = ro.ro_rt->rt_ifp;
2641                                 rtfree(ro.ro_rt);
2642                         }
2643                 } else
2644                         ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2645
2646                 /*
2647                  * See if we found an interface, and confirm that it
2648                  * supports multicast
2649                  */
2650                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2651                         error = EADDRNOTAVAIL;
2652                         break;
2653                 }
2654                 /*
2655                  * Put interface index into the multicast address,
2656                  * if the address has link-local scope.
2657                  */
2658                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2659                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2660                                 = htons(mreq->ipv6mr_interface);
2661                 }
2662                 /*
2663                  * See if the membership already exists.
2664                  */
2665                 for (imm = im6o->im6o_memberships.lh_first;
2666                      imm != NULL; imm = imm->i6mm_chain.le_next)
2667                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2668                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2669                                                &mreq->ipv6mr_multiaddr))
2670                                 break;
2671                 if (imm != NULL) {
2672                         error = EADDRINUSE;
2673                         break;
2674                 }
2675                 /*
2676                  * Everything looks good; add a new record to the multicast
2677                  * address list for the given interface.
2678                  */
2679                 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2680                 if ((imm->i6mm_maddr =
2681                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2682                         kfree(imm, M_IPMADDR);
2683                         break;
2684                 }
2685                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2686                 break;
2687
2688         case IPV6_LEAVE_GROUP:
2689                 /*
2690                  * Drop a multicast group membership.
2691                  * Group must be a valid IP6 multicast address.
2692                  */
2693                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2694                         error = EINVAL;
2695                         break;
2696                 }
2697                 mreq = mtod(m, struct ipv6_mreq *);
2698                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2699                         if (priv_check(td, PRIV_ROOT)) {
2700                                 error = EACCES;
2701                                 break;
2702                         }
2703                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2704                         error = EINVAL;
2705                         break;
2706                 }
2707                 /*
2708                  * If an interface address was specified, get a pointer
2709                  * to its ifnet structure.
2710                  */
2711                 if (mreq->ipv6mr_interface < 0
2712                  || if_index < mreq->ipv6mr_interface) {
2713                         error = ENXIO;  /* XXX EINVAL? */
2714                         break;
2715                 }
2716                 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2717                 /*
2718                  * Put interface index into the multicast address,
2719                  * if the address has link-local scope.
2720                  */
2721                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2722                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2723                                 = htons(mreq->ipv6mr_interface);
2724                 }
2725                 /*
2726                  * Find the membership in the membership list.
2727                  */
2728                 for (imm = im6o->im6o_memberships.lh_first;
2729                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2730                         if ((ifp == NULL ||
2731                              imm->i6mm_maddr->in6m_ifp == ifp) &&
2732                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2733                                                &mreq->ipv6mr_multiaddr))
2734                                 break;
2735                 }
2736                 if (imm == NULL) {
2737                         /* Unable to resolve interface */
2738                         error = EADDRNOTAVAIL;
2739                         break;
2740                 }
2741                 /*
2742                  * Give up the multicast address record to which the
2743                  * membership points.
2744                  */
2745                 LIST_REMOVE(imm, i6mm_chain);
2746                 in6_delmulti(imm->i6mm_maddr);
2747                 kfree(imm, M_IPMADDR);
2748                 break;
2749
2750         default:
2751                 error = EOPNOTSUPP;
2752                 break;
2753         }
2754
2755         /*
2756          * If all options have default values, no need to keep the mbuf.
2757          */
2758         if (im6o->im6o_multicast_ifp == NULL &&
2759             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2760             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2761             im6o->im6o_memberships.lh_first == NULL) {
2762                 kfree(*im6op, M_IPMOPTS);
2763                 *im6op = NULL;
2764         }
2765
2766         return (error);
2767 }
2768
2769 /*
2770  * Return the IP6 multicast options in response to user getsockopt().
2771  */
2772 static int
2773 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2774 {
2775         u_int *hlim, *loop, *ifindex;
2776
2777         *mp = m_get(MB_WAIT, MT_HEADER);                /* XXX */
2778
2779         switch (optname) {
2780
2781         case IPV6_MULTICAST_IF:
2782                 ifindex = mtod(*mp, u_int *);
2783                 (*mp)->m_len = sizeof(u_int);
2784                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2785                         *ifindex = 0;
2786                 else
2787                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2788                 return (0);
2789
2790         case IPV6_MULTICAST_HOPS:
2791                 hlim = mtod(*mp, u_int *);
2792                 (*mp)->m_len = sizeof(u_int);
2793                 if (im6o == NULL)
2794                         *hlim = ip6_defmcasthlim;
2795                 else
2796                         *hlim = im6o->im6o_multicast_hlim;
2797                 return (0);
2798
2799         case IPV6_MULTICAST_LOOP:
2800                 loop = mtod(*mp, u_int *);
2801                 (*mp)->m_len = sizeof(u_int);
2802                 if (im6o == NULL)
2803                         *loop = ip6_defmcasthlim;
2804                 else
2805                         *loop = im6o->im6o_multicast_loop;
2806                 return (0);
2807
2808         default:
2809                 return (EOPNOTSUPP);
2810         }
2811 }
2812
2813 /*
2814  * Discard the IP6 multicast options.
2815  */
2816 void
2817 ip6_freemoptions(struct ip6_moptions *im6o)
2818 {
2819         struct in6_multi_mship *imm;
2820
2821         if (im6o == NULL)
2822                 return;
2823
2824         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2825                 LIST_REMOVE(imm, i6mm_chain);
2826                 if (imm->i6mm_maddr)
2827                         in6_delmulti(imm->i6mm_maddr);
2828                 kfree(imm, M_IPMADDR);
2829         }
2830         kfree(im6o, M_IPMOPTS);
2831 }
2832
2833 /*
2834  * Set a particular packet option, as a sticky option or an ancillary data
2835  * item.  "len" can be 0 only when it's a sticky option.
2836  * We have 4 cases of combination of "sticky" and "cmsg":
2837  * "sticky=0, cmsg=0": impossible
2838  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2839  * "sticky=1, cmsg=0": RFC3542 socket option
2840  * "sticky=1, cmsg=1": RFC2292 socket option
2841  */
2842 static int
2843 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2844      int sticky, int cmsg, int uproto, int priv)
2845 {
2846         int minmtupolicy, preftemp;
2847         //int error;
2848
2849         if (!sticky && !cmsg) {
2850                 kprintf("ip6_setpktoption: impossible case\n");
2851                 return (EINVAL);
2852         }
2853
2854         /*
2855          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2856          * not be specified in the context of RFC3542.  Conversely,
2857          * RFC3542 types should not be specified in the context of RFC2292.
2858          */
2859         if (!cmsg) {
2860                 switch (optname) {
2861                 case IPV6_2292PKTINFO:
2862                 case IPV6_2292HOPLIMIT:
2863                 case IPV6_2292NEXTHOP:
2864                 case IPV6_2292HOPOPTS:
2865                 case IPV6_2292DSTOPTS:
2866                 case IPV6_2292RTHDR:
2867                 case IPV6_2292PKTOPTIONS:
2868                         return (ENOPROTOOPT);
2869                 }
2870         }
2871         if (sticky && cmsg) {
2872                 switch (optname) {
2873                 case IPV6_PKTINFO:
2874                 case IPV6_HOPLIMIT:
2875                 case IPV6_NEXTHOP:
2876                 case IPV6_HOPOPTS:
2877                 case IPV6_DSTOPTS:
2878                 case IPV6_RTHDRDSTOPTS:
2879                 case IPV6_RTHDR:
2880                 case IPV6_USE_MIN_MTU:
2881                 case IPV6_DONTFRAG:
2882                 case IPV6_TCLASS:
2883                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2884                         return (ENOPROTOOPT);
2885                 }
2886         }
2887
2888         switch (optname) {
2889         case IPV6_2292PKTINFO:
2890         case IPV6_PKTINFO:
2891         {
2892                 struct in6_pktinfo *pktinfo;
2893                 if (len != sizeof(struct in6_pktinfo))
2894                         return (EINVAL);
2895                 pktinfo = (struct in6_pktinfo *)buf;
2896
2897                 /*
2898                  * An application can clear any sticky IPV6_PKTINFO option by
2899                  * doing a "regular" setsockopt with ipi6_addr being
2900                  * in6addr_any and ipi6_ifindex being zero.
2901                  * [RFC 3542, Section 6]
2902                  */
2903                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2904                     pktinfo->ipi6_ifindex == 0 &&
2905                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2906                         ip6_clearpktopts(opt, optname);
2907                         break;
2908                 }
2909
2910                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2911                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2912                         return (EINVAL);
2913                 }
2914
2915                 /* validate the interface index if specified. */
2916                 if (pktinfo->ipi6_ifindex > if_index ||
2917                     pktinfo->ipi6_ifindex < 0) {
2918                          return (ENXIO);
2919                 }
2920                 /*
2921                  * Check if the requested source address is indeed a
2922                  * unicast address assigned to the node, and can be
2923                  * used as the packet's source address.
2924                  */
2925                 if (opt->ip6po_pktinfo != NULL &&
2926                     !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2927                         struct in6_ifaddr *ia6;
2928                         struct sockaddr_in6 sin6;
2929
2930                         bzero(&sin6, sizeof(sin6));
2931                         sin6.sin6_len = sizeof(sin6);
2932                         sin6.sin6_family = AF_INET6;
2933                         sin6.sin6_addr =
2934                         opt->ip6po_pktinfo->ipi6_addr;
2935                         ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2936                         if (ia6 == NULL ||
2937                                 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2938                                         IN6_IFF_NOTREADY)) != 0)
2939                         return (EADDRNOTAVAIL);
2940                 }
2941
2942                 /*
2943                  * We store the address anyway, and let in6_selectsrc()
2944                  * validate the specified address.  This is because ipi6_addr
2945                  * may not have enough information about its scope zone, and
2946                  * we may need additional information (such as outgoing
2947                  * interface or the scope zone of a destination address) to
2948                  * disambiguate the scope.
2949                  * XXX: the delay of the validation may confuse the
2950                  * application when it is used as a sticky option.
2951                  */
2952                 if (opt->ip6po_pktinfo == NULL) {
2953                         opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo),
2954                             M_IP6OPT, M_NOWAIT);
2955                         if (opt->ip6po_pktinfo == NULL)
2956                                 return (ENOBUFS);
2957                 }
2958                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2959                 break;
2960         }
2961
2962         case IPV6_2292HOPLIMIT:
2963         case IPV6_HOPLIMIT:
2964         {
2965                 int *hlimp;
2966
2967                 /*
2968                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2969                  * to simplify the ordering among hoplimit options.
2970                  */
2971                 if (optname == IPV6_HOPLIMIT && sticky)
2972                         return (ENOPROTOOPT);
2973
2974                 if (len != sizeof(int))
2975                         return (EINVAL);
2976                 hlimp = (int *)buf;
2977                 if (*hlimp < -1 || *hlimp > 255)
2978                         return (EINVAL);
2979
2980                 opt->ip6po_hlim = *hlimp;
2981                 break;
2982         }
2983
2984         case IPV6_TCLASS:
2985         {
2986                 int tclass;
2987
2988                 if (len != sizeof(int))
2989                         return (EINVAL);
2990                 tclass = *(int *)buf;
2991                 if (tclass < -1 || tclass > 255)
2992                         return (EINVAL);
2993
2994                 opt->ip6po_tclass = tclass;
2995                 break;
2996         }
2997
2998         case IPV6_2292NEXTHOP:
2999         case IPV6_NEXTHOP:
3000                 if (!priv)
3001                         return (EPERM);
3002
3003                 if (len == 0) { /* just remove the option */
3004                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3005                         break;
3006                 }
3007
3008                 /* check if cmsg_len is large enough for sa_len */
3009                 if (len < sizeof(struct sockaddr) || len < *buf)
3010                         return (EINVAL);
3011
3012                 switch (((struct sockaddr *)buf)->sa_family) {
3013                 case AF_INET6:
3014                 {
3015                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3016                         //int error;
3017
3018                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3019                                 return (EINVAL);
3020
3021                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3022                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3023                                 return (EINVAL);
3024                         }
3025                         break;
3026                 }
3027                 case AF_LINK:   /* should eventually be supported */
3028                 default:
3029                         return (EAFNOSUPPORT);
3030                 }
3031
3032                 /* turn off the previous option, then set the new option. */
3033                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3034                 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT);
3035                 if (opt->ip6po_nexthop == NULL)
3036                         return (ENOBUFS);
3037                 bcopy(buf, opt->ip6po_nexthop, *buf);
3038                 break;
3039
3040         case IPV6_2292HOPOPTS:
3041         case IPV6_HOPOPTS:
3042         {
3043                 struct ip6_hbh *hbh;
3044                 int hbhlen;
3045
3046                 /*
3047                  * XXX: We don't allow a non-privileged user to set ANY HbH
3048                  * options, since per-option restriction has too much
3049                  * overhead.
3050                  */
3051                 if (!priv)
3052                         return (EPERM);
3053                 if (len == 0) {
3054                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3055                         break;  /* just remove the option */
3056                 }
3057
3058                 /* message length validation */
3059                 if (len < sizeof(struct ip6_hbh))
3060                         return (EINVAL);
3061                 hbh = (struct ip6_hbh *)buf;
3062                 hbhlen = (hbh->ip6h_len + 1) << 3;
3063                 if (len != hbhlen)
3064                         return (EINVAL);
3065
3066                 /* turn off the previous option, then set the new option. */
3067                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3068                 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT);
3069                 if (opt->ip6po_hbh == NULL)
3070                         return (ENOBUFS);
3071                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3072
3073                 break;
3074         }
3075
3076         case IPV6_2292DSTOPTS:
3077         case IPV6_DSTOPTS:
3078         case IPV6_RTHDRDSTOPTS:
3079         {
3080                 struct ip6_dest *dest, **newdest = NULL;
3081                 int destlen;
3082                 if (!priv)
3083                         return (EPERM);
3084
3085                 if (len == 0) {
3086                         ip6_clearpktopts(opt, optname);
3087                         break;  /* just remove the option */
3088                 }
3089
3090                 /* message length validation */
3091                 if (len < sizeof(struct ip6_dest))
3092                         return (EINVAL);
3093                 dest = (struct ip6_dest *)buf;
3094                 destlen = (dest->ip6d_len + 1) << 3;
3095                 if (len != destlen)
3096                         return (EINVAL);
3097
3098                 /*
3099                  * Determine the position that the destination options header
3100                  * should be inserted; before or after the routing header.
3101                  */
3102                 switch (optname) {
3103                 case IPV6_2292DSTOPTS:
3104                         /*
3105                          * The old advacned API is ambiguous on this point.
3106                          * Our approach is to determine the position based
3107                          * according to the existence of a routing header.
3108                          * Note, however, that this depends on the order of the
3109                          * extension headers in the ancillary data; the 1st
3110                          * part of the destination options header must appear
3111                          * before the routing header in the ancillary data,
3112                          * too.
3113                          * RFC3542 solved the ambiguity by introducing
3114                          * separate ancillary data or option types.
3115                          */
3116                         if (opt->ip6po_rthdr == NULL)
3117                                 newdest = &opt->ip6po_dest1;
3118                         else
3119                                 newdest = &opt->ip6po_dest2;
3120                         break;
3121                 case IPV6_RTHDRDSTOPTS:
3122                         newdest = &opt->ip6po_dest1;
3123                         break;
3124                 case IPV6_DSTOPTS:
3125                         newdest = &opt->ip6po_dest2;
3126                         break;
3127                 }
3128
3129                 /* turn off the previous option, then set the new option. */
3130                 ip6_clearpktopts(opt, optname);
3131                 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT);
3132                 if (*newdest == NULL)
3133                         return (ENOBUFS);
3134                 bcopy(dest, *newdest, destlen);
3135
3136                 break;
3137         }
3138
3139         case IPV6_2292RTHDR:
3140         case IPV6_RTHDR:
3141         {
3142                 struct ip6_rthdr *rth;
3143                 int rthlen;
3144
3145                 if (len == 0) {
3146                         ip6_clearpktopts(opt, IPV6_RTHDR);
3147                         break;  /* just remove the option */
3148                 }
3149
3150                 /* message length validation */
3151                 if (len < sizeof(struct ip6_rthdr))
3152                         return (EINVAL);
3153                 rth = (struct ip6_rthdr *)buf;
3154                 rthlen = (rth->ip6r_len + 1) << 3;
3155                 if (len != rthlen)
3156                         return (EINVAL);
3157
3158                 switch (rth->ip6r_type) {
3159                 default:
3160                         return (EINVAL);        /* not supported */
3161                 }
3162
3163                 /* turn off the previous option */
3164                 ip6_clearpktopts(opt, IPV6_RTHDR);
3165                 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT);
3166                 if (opt->ip6po_rthdr == NULL)
3167                         return (ENOBUFS);
3168                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3169
3170                 break;
3171         }
3172
3173         case IPV6_USE_MIN_MTU:
3174                 if (len != sizeof(int))
3175                         return (EINVAL);
3176                 minmtupolicy = *(int *)buf;
3177                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3178                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3179                     minmtupolicy != IP6PO_MINMTU_ALL) {
3180                         return (EINVAL);
3181                 }
3182                 opt->ip6po_minmtu = minmtupolicy;
3183                 break;
3184
3185         case IPV6_DONTFRAG:
3186                 if (len != sizeof(int))
3187                         return (EINVAL);
3188
3189                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3190                         /*
3191                          * we ignore this option for TCP sockets.
3192                          * (RFC3542 leaves this case unspecified.)
3193                          */
3194                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3195                 } else
3196                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3197                 break;
3198
3199         case IPV6_PREFER_TEMPADDR:
3200                 if (len != sizeof(int))
3201                         return (EINVAL);
3202                 preftemp = *(int *)buf;
3203                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3204                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3205                     preftemp != IP6PO_TEMPADDR_PREFER) {
3206                         return (EINVAL);
3207                 }
3208                 opt->ip6po_prefer_tempaddr = preftemp;
3209                 break;
3210
3211         default:
3212                 return (ENOPROTOOPT);
3213         } /* end of switch */
3214
3215         return (0);
3216 }
3217
3218
3219 /*
3220  * Set IPv6 outgoing packet options based on advanced API.
3221  */
3222 int
3223 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt,
3224     struct ip6_pktopts *stickyopt, int uproto, int priv)
3225 {
3226         struct cmsghdr *cm = NULL;
3227
3228         if (control == NULL || opt == NULL)
3229                 return (EINVAL);
3230
3231         init_ip6pktopts(opt);
3232
3233         /*
3234          * XXX: Currently, we assume all the optional information is stored
3235          * in a single mbuf.
3236          */
3237         if (stickyopt) {
3238                 int error;
3239
3240                 /*
3241                  * If stickyopt is provided, make a local copy of the options
3242                  * for this particular packet, then override them by ancillary
3243                  * objects.
3244                  * XXX: copypktopts() does not copy the cached route to a next
3245                  * hop (if any).  This is not very good in terms of efficiency,
3246                  * but we can allow this since this option should be rarely
3247                  * used.
3248                  */
3249                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
3250                         return (error);
3251         }
3252
3253         /*
3254          * XXX: Currently, we assume all the optional information is stored
3255          * in a single mbuf.
3256          */
3257         if (control->m_next)
3258                 return (EINVAL);
3259
3260         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3261             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3262                 int error;
3263
3264                 if (control->m_len < CMSG_LEN(0))
3265                         return (EINVAL);
3266
3267                 cm = mtod(control, struct cmsghdr *);
3268                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3269                         return (EINVAL);
3270                 if (cm->cmsg_level != IPPROTO_IPV6)
3271                         continue;
3272
3273                 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3274                     cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto, priv);
3275                 if (error)
3276                         return (error);
3277         }
3278
3279         return (0);
3280 }
3281
3282 /*
3283  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3284  * packet to the input queue of a specified interface.  Note that this
3285  * calls the output routine of the loopback "driver", but with an interface
3286  * pointer that might NOT be &loif -- easier than replicating that code here.
3287  */
3288 void
3289 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3290 {
3291         struct mbuf *copym;
3292         struct ip6_hdr *ip6;
3293
3294         copym = m_copy(m, 0, M_COPYALL);
3295         if (copym == NULL)
3296                 return;
3297
3298         /*
3299          * Make sure to deep-copy IPv6 header portion in case the data
3300          * is in an mbuf cluster, so that we can safely override the IPv6
3301          * header portion later.
3302          */
3303         if ((copym->m_flags & M_EXT) != 0 ||
3304             copym->m_len < sizeof(struct ip6_hdr)) {
3305                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3306                 if (copym == NULL)
3307                         return;
3308         }
3309
3310 #ifdef DIAGNOSTIC
3311         if (copym->m_len < sizeof(*ip6)) {
3312                 m_freem(copym);
3313                 return;
3314         }
3315 #endif
3316
3317         ip6 = mtod(copym, struct ip6_hdr *);
3318         /*
3319          * clear embedded scope identifiers if necessary.
3320          * in6_clearscope will touch the addresses only when necessary.
3321          */
3322         in6_clearscope(&ip6->ip6_src);
3323         in6_clearscope(&ip6->ip6_dst);
3324
3325         if_simloop(ifp, copym, dst->sin6_family, 0);
3326 }
3327
3328 /*
3329  * Separate the IPv6 header from the payload into its own mbuf.
3330  *
3331  * Returns the new mbuf chain or the original mbuf if no payload.
3332  * Returns NULL if can't allocate new mbuf for header.
3333  */
3334 static struct mbuf *
3335 ip6_splithdr(struct mbuf *m)
3336 {
3337         struct mbuf *mh;
3338
3339         if (m->m_len <= sizeof(struct ip6_hdr))         /* no payload */
3340                 return (m);
3341
3342         MGETHDR(mh, MB_DONTWAIT, MT_HEADER);
3343         if (mh == NULL)
3344                 return (NULL);
3345         mh->m_len = sizeof(struct ip6_hdr);
3346         M_MOVE_PKTHDR(mh, m);
3347         MH_ALIGN(mh, sizeof(struct ip6_hdr));
3348         bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr));
3349         m->m_data += sizeof(struct ip6_hdr);
3350         m->m_len -= sizeof(struct ip6_hdr);
3351         mh->m_next = m;
3352         return (mh);
3353 }
3354
3355 /*
3356  * Compute IPv6 extension header length.
3357  */
3358 int
3359 ip6_optlen(struct in6pcb *in6p)
3360 {
3361         int len;
3362
3363         if (!in6p->in6p_outputopts)
3364                 return 0;
3365
3366         len = 0;
3367 #define elen(x) \
3368     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3369
3370         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3371         if (in6p->in6p_outputopts->ip6po_rthdr)
3372                 /* dest1 is valid with rthdr only */
3373                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3374         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3375         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3376         return len;
3377 #undef elen
3378 }