sys/netinet6/ip6_output.c

   1 /*      $FreeBSD$       */
   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  * 4. Neither the name of the University nor the names of its contributors
  46  *    may be used to endorse or promote products derived from this software
  47  *    without specific prior written permission.
  48  *
  49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  59  * SUCH DAMAGE.
  60  *
  61  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  62  */
  63
  64 #include "opt_ip6fw.h"
  65 #include "opt_inet.h"
  66 #include "opt_inet6.h"
  67 #include "opt_ipsec.h"
  68
  69 #include <sys/param.h>
  70 #include <sys/malloc.h>
  71 #include <sys/mbuf.h>
  72 #include <sys/proc.h>
  73 #include <sys/errno.h>
  74 #include <sys/protosw.h>
  75 #include <sys/socket.h>
  76 #include <sys/socketvar.h>
  77 #include <sys/systm.h>
  78 #include <sys/kernel.h>
  79
  80 #include <net/if.h>
  81 #include <net/netisr.h>
  82 #include <net/route.h>
  83 #include <net/pfil.h>
  84
  85 #include <netinet/in.h>
  86 #include <netinet/in_var.h>
  87 #include <netinet6/in6_var.h>
  88 #include <netinet/ip6.h>
  89 #include <netinet/icmp6.h>
  90 #include <netinet6/ip6_var.h>
  91 #include <netinet/in_pcb.h>
  92 #include <netinet/tcp_var.h>
  93 #include <netinet6/nd6.h>
  94
  95 #ifdef IPSEC
  96 #include <netinet6/ipsec.h>
  97 #ifdef INET6
  98 #include <netinet6/ipsec6.h>
  99 #endif
 100 #include <netkey/key.h>
 101 #endif /* IPSEC */
 102
 103 #ifdef FAST_IPSEC
 104 #include <netipsec/ipsec.h>
 105 #include <netipsec/ipsec6.h>
 106 #include <netipsec/key.h>
 107 #endif /* FAST_IPSEC */
 108
 109 #include <netinet6/ip6_fw.h>
 110
 111 #include <net/net_osdep.h>
 112
 113 #include <netinet6/ip6protosw.h>
 114
 115 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
 116
 117 struct ip6_exthdrs {
 118         struct mbuf *ip6e_ip6;
 119         struct mbuf *ip6e_hbh;
 120         struct mbuf *ip6e_dest1;
 121         struct mbuf *ip6e_rthdr;
 122         struct mbuf *ip6e_dest2;
 123 };
 124
 125 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
 126                            int, int));
 127 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
 128         struct socket *, struct sockopt *));
 129 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct sockopt *));
 130 static int ip6_setpktoption __P((int, u_char *, int, struct ip6_pktopts *, int,
 131         int, int, int));
 132
 133 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
 134 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
 135 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
 136 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
 137         struct ip6_frag **));
 138 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
 139 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
 140 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
 141         struct ifnet *, struct in6_addr *, u_long *, int *));
 142
 143
 144 /*
 145  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 146  * header (with pri, len, nxt, hlim, src, dst).
 147  * This function may modify ver and hlim only.
 148  * The mbuf chain containing the packet will be freed.
 149  * The mbuf opt, if present, will not be freed.
 150  *
 151  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 152  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 153  * which is rt_rmx.rmx_mtu.
 154  */
 155 int
 156 ip6_output(m0, opt, ro, flags, im6o, ifpp, inp)
 157         struct mbuf *m0;
 158         struct ip6_pktopts *opt;
 159         struct route_in6 *ro;
 160         int flags;
 161         struct ip6_moptions *im6o;
 162         struct ifnet **ifpp;            /* XXX: just for statistics */
 163         struct inpcb *inp;
 164 {
 165         struct ip6_hdr *ip6, *mhip6;
 166         struct ifnet *ifp, *origifp;
 167         struct mbuf *m = m0;
 168         int hlen, tlen, len, off;
 169         struct route_in6 ip6route;
 170         struct sockaddr_in6 *dst;
 171         struct in6_addr odst;
 172         int error = 0;
 173         struct in6_ifaddr *ia = NULL;
 174         u_long mtu;
 175         int alwaysfrag, dontfrag;
 176         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 177         struct ip6_exthdrs exthdrs;
 178         struct in6_addr finaldst;
 179         struct route_in6 *ro_pmtu = NULL;
 180         int hdrsplit = 0;
 181         int needipsec = 0;
 182 #if defined(IPSEC) || defined(FAST_IPSEC)
 183         int needipsectun = 0;
 184         struct secpolicy *sp = NULL;
 185 #endif /*IPSEC || FAST_IPSEC*/
 186
 187         ip6 = mtod(m, struct ip6_hdr *);
 188         finaldst = ip6->ip6_dst;
 189
 190 #define MAKE_EXTHDR(hp, mp)                                             \
 191     do {                                                                \
 192         if (hp) {                                                       \
 193                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 194                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 195                     ((eh)->ip6e_len + 1) << 3);                         \
 196                 if (error)                                              \
 197                         goto freehdrs;                                  \
 198         }                                                               \
 199     } while (/*CONSTCOND*/ 0)
 200
 201         bzero(&exthdrs, sizeof(exthdrs));
 202
 203         if (opt) {
 204                 /* Hop-by-Hop options header */
 205                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 206                 /* Destination options header(1st part) */
 207                 if (opt->ip6po_rthdr) {
 208                         /*
 209                          * Destination options header(1st part)
 210                          * This only makes sence with a routing header.
 211                          * See Section 9.2 of RFC 3542.
 212                          * Disabling this part just for MIP6 convenience is
 213                          * a bad idea.  We need to think carefully about a
 214                          * way to make the advanced API coexist with MIP6
 215                          * options, which might automatically be inserted in
 216                          * the kernel.
 217                          */
 218                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 219                 }
 220                 /* Routing header */
 221                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 222                 /* Destination options header(2nd part) */
 223                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 224         }
 225
 226 #ifdef IPSEC
 227         /* get a security policy for this packet */
 228         if (inp == NULL)
 229                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 230         else
 231                 sp = ipsec6_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error);
 232
 233         if (sp == NULL) {
 234                 ipsec6stat.out_inval++;
 235                 goto freehdrs;
 236         }
 237
 238         error = 0;
 239
 240         /* check policy */
 241         switch (sp->policy) {
 242         case IPSEC_POLICY_DISCARD:
 243                 /*
 244                  * This packet is just discarded.
 245                  */
 246                 ipsec6stat.out_polvio++;
 247                 goto freehdrs;
 248
 249         case IPSEC_POLICY_BYPASS:
 250         case IPSEC_POLICY_NONE:
 251                 /* no need to do IPsec. */
 252                 needipsec = 0;
 253                 break;
 254
 255         case IPSEC_POLICY_IPSEC:
 256                 if (sp->req == NULL) {
 257                         /* acquire a policy */
 258                         error = key_spdacquire(sp);
 259                         goto freehdrs;
 260                 }
 261                 needipsec = 1;
 262                 break;
 263
 264         case IPSEC_POLICY_ENTRUST:
 265         default:
 266                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 267         }
 268 #endif /* IPSEC */
 269 #ifdef FAST_IPSEC
 270         /* get a security policy for this packet */
 271         if (inp == NULL)
 272                 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 273         else
 274                 sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
 275
 276         if (sp == NULL) {
 277                 newipsecstat.ips_out_inval++;
 278                 goto freehdrs;
 279         }
 280
 281         error = 0;
 282
 283         /* check policy */
 284         switch (sp->policy) {
 285         case IPSEC_POLICY_DISCARD:
 286                 /*
 287                  * This packet is just discarded.
 288                  */
 289                 newipsecstat.ips_out_polvio++;
 290                 goto freehdrs;
 291
 292         case IPSEC_POLICY_BYPASS:
 293         case IPSEC_POLICY_NONE:
 294                 /* no need to do IPsec. */
 295                 needipsec = 0;
 296                 break;
 297
 298         case IPSEC_POLICY_IPSEC:
 299                 if (sp->req == NULL) {
 300                         /* acquire a policy */
 301                         error = key_spdacquire(sp);
 302                         goto freehdrs;
 303                 }
 304                 needipsec = 1;
 305                 break;
 306
 307         case IPSEC_POLICY_ENTRUST:
 308         default:
 309                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 310         }
 311 #endif /* FAST_IPSEC */
 312
 313         /*
 314          * Calculate the total length of the extension header chain.
 315          * Keep the length of the unfragmentable part for fragmentation.
 316          */
 317         optlen = 0;
 318         if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
 319         if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
 320         if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
 321         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 322         /* NOTE: we don't add AH/ESP length here. do that later. */
 323         if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
 324
 325         /*
 326          * If we need IPsec, or there is at least one extension header,
 327          * separate IP6 header from the payload.
 328          */
 329         if ((needipsec || optlen) && !hdrsplit) {
 330                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 331                         m = NULL;
 332                         goto freehdrs;
 333                 }
 334                 m = exthdrs.ip6e_ip6;
 335                 hdrsplit++;
 336         }
 337
 338         /* adjust pointer */
 339         ip6 = mtod(m, struct ip6_hdr *);
 340
 341         /* adjust mbuf packet header length */
 342         m->m_pkthdr.len += optlen;
 343         plen = m->m_pkthdr.len - sizeof(*ip6);
 344
 345         /* If this is a jumbo payload, insert a jumbo payload option. */
 346         if (plen > IPV6_MAXPACKET) {
 347                 if (!hdrsplit) {
 348                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 349                                 m = NULL;
 350                                 goto freehdrs;
 351                         }
 352                         m = exthdrs.ip6e_ip6;
 353                         hdrsplit++;
 354                 }
 355                 /* adjust pointer */
 356                 ip6 = mtod(m, struct ip6_hdr *);
 357                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 358                         goto freehdrs;
 359                 ip6->ip6_plen = 0;
 360         } else
 361                 ip6->ip6_plen = htons(plen);
 362
 363         /*
 364          * Concatenate headers and fill in next header fields.
 365          * Here we have, on "m"
 366          *      IPv6 payload
 367          * and we insert headers accordingly.  Finally, we should be getting:
 368          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 369          *
 370          * during the header composing process, "m" points to IPv6 header.
 371          * "mprev" points to an extension header prior to esp.
 372          */
 373         {
 374                 u_char *nexthdrp = &ip6->ip6_nxt;
 375                 struct mbuf *mprev = m;
 376
 377                 /*
 378                  * we treat dest2 specially.  this makes IPsec processing
 379                  * much easier.  the goal here is to make mprev point the
 380                  * mbuf prior to dest2.
 381                  *
 382                  * result: IPv6 dest2 payload
 383                  * m and mprev will point to IPv6 header.
 384                  */
 385                 if (exthdrs.ip6e_dest2) {
 386                         if (!hdrsplit)
 387                                 panic("assumption failed: hdr not split");
 388                         exthdrs.ip6e_dest2->m_next = m->m_next;
 389                         m->m_next = exthdrs.ip6e_dest2;
 390                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 391                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 392                 }
 393
 394 #define MAKE_CHAIN(m, mp, p, i)\
 395     do {\
 396         if (m) {\
 397                 if (!hdrsplit) \
 398                         panic("assumption failed: hdr not split"); \
 399                 *mtod((m), u_char *) = *(p);\
 400                 *(p) = (i);\
 401                 p = mtod((m), u_char *);\
 402                 (m)->m_next = (mp)->m_next;\
 403                 (mp)->m_next = (m);\
 404                 (mp) = (m);\
 405         }\
 406     } while (/*CONSTCOND*/ 0)
 407                 /*
 408                  * result: IPv6 hbh dest1 rthdr dest2 payload
 409                  * m will point to IPv6 header.  mprev will point to the
 410                  * extension header prior to dest2 (rthdr in the above case).
 411                  */
 412                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 413                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 414                     IPPROTO_DSTOPTS);
 415                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 416                     IPPROTO_ROUTING);
 417
 418 #if defined(IPSEC) || defined(FAST_IPSEC)
 419                 if (!needipsec)
 420                         goto skip_ipsec2;
 421
 422                 /*
 423                  * pointers after IPsec headers are not valid any more.
 424                  * other pointers need a great care too.
 425                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
 426                  */
 427                 exthdrs.ip6e_dest2 = NULL;
 428
 429             {
 430                 struct ip6_rthdr *rh = NULL;
 431                 int segleft_org = 0;
 432                 struct ipsec_output_state state;
 433
 434                 if (exthdrs.ip6e_rthdr) {
 435                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 436                         segleft_org = rh->ip6r_segleft;
 437                         rh->ip6r_segleft = 0;
 438                 }
 439
 440                 bzero(&state, sizeof(state));
 441                 state.m = m;
 442                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 443                     &needipsectun);
 444                 m = state.m;
 445                 if (error) {
 446                         /* mbuf is already reclaimed in ipsec6_output_trans. */
 447                         m = NULL;
 448                         switch (error) {
 449                         case EHOSTUNREACH:
 450                         case ENETUNREACH:
 451                         case EMSGSIZE:
 452                         case ENOBUFS:
 453                         case ENOMEM:
 454                                 break;
 455                         default:
 456                                 printf("ip6_output (ipsec): error code %d\n", error);
 457                                 /* FALLTHROUGH */
 458                         case ENOENT:
 459                                 /* don't show these error codes to the user */
 460                                 error = 0;
 461                                 break;
 462                         }
 463                         goto bad;
 464                 }
 465                 if (exthdrs.ip6e_rthdr) {
 466                         /* ah6_output doesn't modify mbuf chain */
 467                         rh->ip6r_segleft = segleft_org;
 468                 }
 469             }
 470 skip_ipsec2:;
 471 #endif
 472         }
 473
 474         /*
 475          * If there is a routing header, replace the destination address field
 476          * with the first hop of the routing header.
 477          */
 478         if (exthdrs.ip6e_rthdr) {
 479                 struct ip6_rthdr *rh =
 480                         (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
 481                                                   struct ip6_rthdr *));
 482                 struct ip6_rthdr0 *rh0;
 483                 struct in6_addr *addrs;
 484
 485                 switch (rh->ip6r_type) {
 486                 case IPV6_RTHDR_TYPE_0:
 487                          rh0 = (struct ip6_rthdr0 *)rh;
 488                          addrs = (struct in6_addr *)(rh0 + 1);
 489
 490                          ip6->ip6_dst = *addrs;
 491                          bcopy((caddr_t)(addrs + 1), (caddr_t)addrs,
 492                                sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
 493                                  );
 494                          *(addrs + rh0->ip6r0_segleft - 1) = finaldst;
 495                          break;
 496                 default:        /* is it possible? */
 497                          error = EINVAL;
 498                          goto bad;
 499                 }
 500         }
 501
 502         /* Source address validation */
 503         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 504             (flags & IPV6_DADOUTPUT) == 0) {
 505                 error = EOPNOTSUPP;
 506                 ip6stat.ip6s_badscope++;
 507                 goto bad;
 508         }
 509         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 510                 error = EOPNOTSUPP;
 511                 ip6stat.ip6s_badscope++;
 512                 goto bad;
 513         }
 514
 515         ip6stat.ip6s_localout++;
 516
 517         /*
 518          * Route packet.
 519          */
 520         if (ro == 0) {
 521                 ro = &ip6route;
 522                 bzero((caddr_t)ro, sizeof(*ro));
 523         }
 524         ro_pmtu = ro;
 525         if (opt && opt->ip6po_rthdr)
 526                 ro = &opt->ip6po_route;
 527         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 528
 529 again:
 530         /*
 531          * If there is a cached route,
 532          * check that it is to the same destination
 533          * and is still up. If not, free it and try again.
 534          */
 535         if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
 536                          dst->sin6_family != AF_INET6 ||
 537                          !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
 538                 RTFREE(ro->ro_rt);
 539                 ro->ro_rt = (struct rtentry *)0;
 540         }
 541         if (ro->ro_rt == 0) {
 542                 bzero(dst, sizeof(*dst));
 543                 dst->sin6_family = AF_INET6;
 544                 dst->sin6_len = sizeof(struct sockaddr_in6);
 545                 dst->sin6_addr = ip6->ip6_dst;
 546         }
 547
 548         /*
 549          * if specified, try to fill in the traffic class field.
 550          * do not override if a non-zero value is already set.
 551          * we check the diffserv field and the ecn field separately.
 552          */
 553         if (opt && opt->ip6po_tclass >= 0) {
 554                 int mask = 0;
 555
 556                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 557                         mask |= 0xfc;
 558                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 559                         mask |= 0x03;
 560                 if (mask != 0)
 561                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 562         }
 563
 564         /* fill in or override the hop limit field, if necessary. */
 565         if (opt && opt->ip6po_hlim != -1)
 566                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 567         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 568                 if (im6o != NULL)
 569                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 570                 else
 571                         ip6->ip6_hlim = ip6_defmcasthlim;
 572         }
 573
 574 #if defined(IPSEC) || defined(FAST_IPSEC)
 575         if (needipsec && needipsectun) {
 576                 struct ipsec_output_state state;
 577
 578                 /*
 579                  * All the extension headers will become inaccessible
 580                  * (since they can be encrypted).
 581                  * Don't panic, we need no more updates to extension headers
 582                  * on inner IPv6 packet (since they are now encapsulated).
 583                  *
 584                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 585                  */
 586                 bzero(&exthdrs, sizeof(exthdrs));
 587                 exthdrs.ip6e_ip6 = m;
 588
 589                 bzero(&state, sizeof(state));
 590                 state.m = m;
 591                 state.ro = (struct route *)ro;
 592                 state.dst = (struct sockaddr *)dst;
 593
 594                 error = ipsec6_output_tunnel(&state, sp, flags);
 595
 596                 m = state.m;
 597                 ro = (struct route_in6 *)state.ro;
 598                 dst = (struct sockaddr_in6 *)state.dst;
 599                 if (error) {
 600                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 601                         m0 = m = NULL;
 602                         m = NULL;
 603                         switch (error) {
 604                         case EHOSTUNREACH:
 605                         case ENETUNREACH:
 606                         case EMSGSIZE:
 607                         case ENOBUFS:
 608                         case ENOMEM:
 609                                 break;
 610                         default:
 611                                 printf("ip6_output (ipsec): error code %d\n", error);
 612                                 /* FALLTHROUGH */
 613                         case ENOENT:
 614                                 /* don't show these error codes to the user */
 615                                 error = 0;
 616                                 break;
 617                         }
 618                         goto bad;
 619                 }
 620
 621                 exthdrs.ip6e_ip6 = m;
 622         }
 623 #endif /* IPSEC */
 624
 625         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 626                 /* Unicast */
 627
 628 #define ifatoia6(ifa)   ((struct in6_ifaddr *)(ifa))
 629 #define sin6tosa(sin6)  ((struct sockaddr *)(sin6))
 630                 /* xxx
 631                  * interface selection comes here
 632                  * if an interface is specified from an upper layer,
 633                  * ifp must point it.
 634                  */
 635                 if (ro->ro_rt == 0) {
 636                         /*
 637                          * non-bsdi always clone routes, if parent is
 638                          * PRF_CLONING.
 639                          */
 640                         rtalloc((struct route *)ro);
 641                 }
 642                 if (ro->ro_rt == 0) {
 643                         ip6stat.ip6s_noroute++;
 644                         error = EHOSTUNREACH;
 645                         /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
 646                         goto bad;
 647                 }
 648                 /* XXX rt not locked */
 649                 ia = ifatoia6(ro->ro_rt->rt_ifa);
 650                 ifp = ro->ro_rt->rt_ifp;
 651                 ro->ro_rt->rt_rmx.rmx_pksent++;
 652                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
 653                         dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
 654                 m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
 655
 656                 in6_ifstat_inc(ifp, ifs6_out_request);
 657
 658                 /*
 659                  * Check if the outgoing interface conflicts with
 660                  * the interface specified by ifi6_ifindex (if specified).
 661                  * Note that loopback interface is always okay.
 662                  * (this may happen when we are sending a packet to one of
 663                  *  our own addresses.)
 664                  */
 665                 if (opt && opt->ip6po_pktinfo
 666                  && opt->ip6po_pktinfo->ipi6_ifindex) {
 667                         if (!(ifp->if_flags & IFF_LOOPBACK)
 668                          && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
 669                                 ip6stat.ip6s_noroute++;
 670                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 671                                 error = EHOSTUNREACH;
 672                                 goto bad;
 673                         }
 674                 }
 675
 676                 if (opt && opt->ip6po_hlim != -1)
 677                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 678         } else {
 679                 /* Multicast */
 680                 struct  in6_multi *in6m;
 681
 682                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 683
 684                 /*
 685                  * See if the caller provided any multicast options
 686                  */
 687                 ifp = NULL;
 688                 if (im6o != NULL) {
 689                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 690                         if (im6o->im6o_multicast_ifp != NULL)
 691                                 ifp = im6o->im6o_multicast_ifp;
 692                 } else
 693                         ip6->ip6_hlim = ip6_defmcasthlim;
 694
 695                 /*
 696                  * See if the caller provided the outgoing interface
 697                  * as an ancillary data.
 698                  * Boundary check for ifindex is assumed to be already done.
 699                  */
 700                 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
 701                         ifp = ifnet_byindex(opt->ip6po_pktinfo->ipi6_ifindex);
 702
 703                 /*
 704                  * If the destination is a node-local scope multicast,
 705                  * the packet should be loop-backed only.
 706                  */
 707                 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 708                         /*
 709                          * If the outgoing interface is already specified,
 710                          * it should be a loopback interface.
 711                          */
 712                         if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
 713                                 ip6stat.ip6s_badscope++;
 714                                 error = ENETUNREACH; /* XXX: better error? */
 715                                 /* XXX correct ifp? */
 716                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 717                                 goto bad;
 718                         } else {
 719                                 ifp = &loif[0];
 720                         }
 721                 }
 722
 723                 if (opt && opt->ip6po_hlim != -1)
 724                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 725
 726                 /*
 727                  * If caller did not provide an interface lookup a
 728                  * default in the routing table.  This is either a
 729                  * default for the speicfied group (i.e. a host
 730                  * route), or a multicast default (a route for the
 731                  * ``net'' ff00::/8).
 732                  */
 733                 if (ifp == NULL) {
 734                         if (ro->ro_rt == 0)
 735                                 ro->ro_rt = rtalloc1((struct sockaddr *)
 736                                                 &ro->ro_dst, 0, 0UL);
 737                         else
 738                                 RT_LOCK(ro->ro_rt);
 739                         if (ro->ro_rt == 0) {
 740                                 ip6stat.ip6s_noroute++;
 741                                 error = EHOSTUNREACH;
 742                                 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
 743                                 goto bad;
 744                         }
 745                         ia = ifatoia6(ro->ro_rt->rt_ifa);
 746                         ifp = ro->ro_rt->rt_ifp;
 747                         ro->ro_rt->rt_rmx.rmx_pksent++;
 748                         RT_UNLOCK(ro->ro_rt);
 749                 }
 750
 751                 if ((flags & IPV6_FORWARDING) == 0)
 752                         in6_ifstat_inc(ifp, ifs6_out_request);
 753                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 754
 755                 /*
 756                  * Confirm that the outgoing interface supports multicast.
 757                  */
 758                 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
 759                         ip6stat.ip6s_noroute++;
 760                         in6_ifstat_inc(ifp, ifs6_out_discard);
 761                         error = ENETUNREACH;
 762                         goto bad;
 763                 }
 764                 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
 765                 if (in6m != NULL &&
 766                    (im6o == NULL || im6o->im6o_multicast_loop)) {
 767                         /*
 768                          * If we belong to the destination multicast group
 769                          * on the outgoing interface, and the caller did not
 770                          * forbid loopback, loop back a copy.
 771                          */
 772                         ip6_mloopback(ifp, m, dst);
 773                 } else {
 774                         /*
 775                          * If we are acting as a multicast router, perform
 776                          * multicast forwarding as if the packet had just
 777                          * arrived on the interface to which we are about
 778                          * to send.  The multicast forwarding function
 779                          * recursively calls this function, using the
 780                          * IPV6_FORWARDING flag to prevent infinite recursion.
 781                          *
 782                          * Multicasts that are looped back by ip6_mloopback(),
 783                          * above, will be forwarded by the ip6_input() routine,
 784                          * if necessary.
 785                          */
 786                         if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 787                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 788                                         m_freem(m);
 789                                         goto done;
 790                                 }
 791                         }
 792                 }
 793                 /*
 794                  * Multicasts with a hoplimit of zero may be looped back,
 795                  * above, but must not be transmitted on a network.
 796                  * Also, multicasts addressed to the loopback interface
 797                  * are not sent -- the above call to ip6_mloopback() will
 798                  * loop back a copy if this host actually belongs to the
 799                  * destination group on the loopback interface.
 800                  */
 801                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 802                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 803                         m_freem(m);
 804                         goto done;
 805                 }
 806         }
 807
 808         /*
 809          * Fill the outgoing inteface to tell the upper layer
 810          * to increment per-interface statistics.
 811          */
 812         if (ifpp)
 813                 *ifpp = ifp;
 814
 815         /* Determine path MTU. */
 816         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 817             &alwaysfrag)) != 0)
 818                 goto bad;
 819
 820         /*
 821          * The caller of this function may specify to use the minimum MTU
 822          * in some cases.
 823          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 824          * setting.  The logic is a bit complicated; by default, unicast
 825          * packets will follow path MTU while multicast packets will be sent at
 826          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 827          * including unicast ones will be sent at the minimum MTU.  Multicast
 828          * packets will always be sent at the minimum MTU unless
 829          * IP6PO_MINMTU_DISABLE is explicitly specified.
 830          * See RFC 3542 for more details.
 831          */
 832         if (mtu > IPV6_MMTU) {
 833                 if ((flags & IPV6_MINMTU))
 834                         mtu = IPV6_MMTU;
 835                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 836                         mtu = IPV6_MMTU;
 837                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 838                          (opt == NULL ||
 839                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 840                         mtu = IPV6_MMTU;
 841                 }
 842         }
 843
 844         /* Fake scoped addresses */
 845         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
 846                 /*
 847                  * If source or destination address is a scoped address, and
 848                  * the packet is going to be sent to a loopback interface,
 849                  * we should keep the original interface.
 850                  */
 851
 852                 /*
 853                  * XXX: this is a very experimental and temporary solution.
 854                  * We eventually have sockaddr_in6 and use the sin6_scope_id
 855                  * field of the structure here.
 856                  * We rely on the consistency between two scope zone ids
 857                  * of source and destination, which should already be assured.
 858                  * Larger scopes than link will be supported in the future.
 859                  */
 860                 origifp = NULL;
 861                 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
 862                         origifp = ifnet_byindex(ntohs(ip6->ip6_src.s6_addr16[1]));
 863                 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
 864                         origifp = ifnet_byindex(ntohs(ip6->ip6_dst.s6_addr16[1]));
 865                 /*
 866                  * XXX: origifp can be NULL even in those two cases above.
 867                  * For example, if we remove the (only) link-local address
 868                  * from the loopback interface, and try to send a link-local
 869                  * address without link-id information.  Then the source
 870                  * address is ::1, and the destination address is the
 871                  * link-local address with its s6_addr16[1] being zero.
 872                  * What is worse, if the packet goes to the loopback interface
 873                  * by a default rejected route, the null pointer would be
 874                  * passed to looutput, and the kernel would hang.
 875                  * The following last resort would prevent such disaster.
 876                  */
 877                 if (origifp == NULL)
 878                         origifp = ifp;
 879         }
 880         else
 881                 origifp = ifp;
 882         /*
 883          * clear embedded scope identifiers if necessary.
 884          * in6_clearscope will touch the addresses only when necessary.
 885          */
 886         in6_clearscope(&ip6->ip6_src);
 887         in6_clearscope(&ip6->ip6_dst);
 888
 889         /*
 890          * Check with the firewall...
 891          */
 892         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 893                 u_short port = 0;
 894                 m->m_pkthdr.rcvif = NULL;       /* XXX */
 895                 /* If ipfw says divert, we have to just drop packet */
 896                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
 897                         m_freem(m);
 898                         goto done;
 899                 }
 900                 if (!m) {
 901                         error = EACCES;
 902                         goto done;
 903                 }
 904         }
 905
 906         /*
 907          * If the outgoing packet contains a hop-by-hop options header,
 908          * it must be examined and processed even by the source node.
 909          * (RFC 2460, section 4.)
 910          */
 911         if (exthdrs.ip6e_hbh) {
 912                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 913                 u_int32_t dummy; /* XXX unused */
 914                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 915
 916 #ifdef DIAGNOSTIC
 917                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 918                         panic("ip6e_hbh is not continuous");
 919 #endif
 920                 /*
 921                  *  XXX: if we have to send an ICMPv6 error to the sender,
 922                  *       we need the M_LOOP flag since icmp6_error() expects
 923                  *       the IPv6 and the hop-by-hop options header are
 924                  *       continuous unless the flag is set.
 925                  */
 926                 m->m_flags |= M_LOOP;
 927                 m->m_pkthdr.rcvif = ifp;
 928                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 929                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 930                     &dummy, &plen) < 0) {
 931                         /* m was already freed at this point */
 932                         error = EINVAL;/* better error? */
 933                         goto done;
 934                 }
 935                 m->m_flags &= ~M_LOOP; /* XXX */
 936                 m->m_pkthdr.rcvif = NULL;
 937         }
 938
 939         /* Jump over all PFIL processing if hooks are not active. */
 940         if (inet6_pfil_hook.ph_busy_count == -1)
 941                 goto passout;
 942
 943         odst = ip6->ip6_dst;
 944         /* Run through list of hooks for output packets. */
 945         error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 946         if (error != 0 || m == NULL)
 947                 goto done;
 948         ip6 = mtod(m, struct ip6_hdr *);
 949
 950         /* See if destination IP address was changed by packet filter. */
 951         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 952                 m->m_flags |= M_SKIP_FIREWALL;
 953                 /* If destination is now ourself drop to ip6_input(). */
 954                 if (in6_localaddr(&ip6->ip6_dst)) {
 955                         if (m->m_pkthdr.rcvif == NULL)
 956                                 m->m_pkthdr.rcvif = loif;
 957                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 958                                 m->m_pkthdr.csum_flags |=
 959                                     CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 960                                 m->m_pkthdr.csum_data = 0xffff;
 961                         }
 962                         m->m_pkthdr.csum_flags |=
 963                             CSUM_IP_CHECKED | CSUM_IP_VALID;
 964                         error = netisr_queue(NETISR_IPV6, m);
 965                         goto done;
 966                 } else
 967                         goto again;     /* Redo the routing table lookup. */
 968         }
 969
 970         /* XXX: IPFIREWALL_FORWARD */
 971
 972 passout:
 973         /*
 974          * Send the packet to the outgoing interface.
 975          * If necessary, do IPv6 fragmentation before sending.
 976          *
 977          * the logic here is rather complex:
 978          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 979          * 1-a: send as is if tlen <= path mtu
 980          * 1-b: fragment if tlen > path mtu
 981          *
 982          * 2: if user asks us not to fragment (dontfrag == 1)
 983          * 2-a: send as is if tlen <= interface mtu
 984          * 2-b: error if tlen > interface mtu
 985          *
 986          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 987          *      always fragment
 988          *
 989          * 4: if dontfrag == 1 && alwaysfrag == 1
 990          *      error, as we cannot handle this conflicting request
 991          */
 992         tlen = m->m_pkthdr.len;
 993
 994         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
 995                 dontfrag = 1;
 996         else
 997                 dontfrag = 0;
 998         if (dontfrag && alwaysfrag) {   /* case 4 */
 999                 /* conflicting request - can't transmit */
1000                 error = EMSGSIZE;
1001                 goto bad;
1002         }
1003         if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
1004                 /*
1005                  * Even if the DONTFRAG option is specified, we cannot send the
1006                  * packet when the data length is larger than the MTU of the
1007                  * outgoing interface.
1008                  * Notify the error by sending IPV6_PATHMTU ancillary data as
1009                  * well as returning an error code (the latter is not described
1010                  * in the API spec.)
1011                  */
1012                 u_int32_t mtu32;
1013                 struct ip6ctlparam ip6cp;
1014
1015                 mtu32 = (u_int32_t)mtu;
1016                 bzero(&ip6cp, sizeof(ip6cp));
1017                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1018                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1019                     (void *)&ip6cp);
1020
1021                 error = EMSGSIZE;
1022                 goto bad;
1023         }
1024
1025         /*
1026          * transmit packet without fragmentation
1027          */
1028         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1029                 struct in6_ifaddr *ia6;
1030
1031                 ip6 = mtod(m, struct ip6_hdr *);
1032                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1033                 if (ia6) {
1034                         /* Record statistics for this interface address. */
1035                         ia6->ia_ifa.if_opackets++;
1036                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1037                 }
1038 #ifdef IPSEC
1039                 /* clean ipsec history once it goes out of the node */
1040                 ipsec_delaux(m);
1041 #endif
1042                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1043                 goto done;
1044         }
1045
1046         /*
1047          * try to fragment the packet.  case 1-b and 3
1048          */
1049         if (mtu < IPV6_MMTU) {
1050                 /* path MTU cannot be less than IPV6_MMTU */
1051                 error = EMSGSIZE;
1052                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1053                 goto bad;
1054         } else if (ip6->ip6_plen == 0) {
1055                 /* jumbo payload cannot be fragmented */
1056                 error = EMSGSIZE;
1057                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1058                 goto bad;
1059         } else {
1060                 struct mbuf **mnext, *m_frgpart;
1061                 struct ip6_frag *ip6f;
1062                 u_int32_t id = htonl(ip6_randomid());
1063                 u_char nextproto;
1064 #if 0
1065                 struct ip6ctlparam ip6cp;
1066                 u_int32_t mtu32;
1067 #endif
1068                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1069
1070                 /*
1071                  * Too large for the destination or interface;
1072                  * fragment if possible.
1073                  * Must be able to put at least 8 bytes per fragment.
1074                  */
1075                 hlen = unfragpartlen;
1076                 if (mtu > IPV6_MAXPACKET)
1077                         mtu = IPV6_MAXPACKET;
1078
1079 #if 0
1080                 /*
1081                  * It is believed this code is a leftover from the
1082                  * development of the IPV6_RECVPATHMTU sockopt and
1083                  * associated work to implement RFC3542.
1084                  * It's not entirely clear what the intent of the API
1085                  * is at this point, so disable this code for now.
1086                  * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
1087                  * will send notifications if the application requests.
1088                  */
1089
1090                 /* Notify a proper path MTU to applications. */
1091                 mtu32 = (u_int32_t)mtu;
1092                 bzero(&ip6cp, sizeof(ip6cp));
1093                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1094                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1095                     (void *)&ip6cp);
1096 #endif
1097
1098                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1099                 if (len < 8) {
1100                         error = EMSGSIZE;
1101                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1102                         goto bad;
1103                 }
1104
1105                 /*
1106                  * Verify that we have any chance at all of being able to queue
1107                  *      the packet or packet fragments
1108                  */
1109                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1110                     < tlen  /* - hlen */)) {
1111                         error = ENOBUFS;
1112                         ip6stat.ip6s_odropped++;
1113                         goto bad;
1114                 }
1115
1116                 mnext = &m->m_nextpkt;
1117
1118                 /*
1119                  * Change the next header field of the last header in the
1120                  * unfragmentable part.
1121                  */
1122                 if (exthdrs.ip6e_rthdr) {
1123                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1124                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1125                 } else if (exthdrs.ip6e_dest1) {
1126                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1127                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1128                 } else if (exthdrs.ip6e_hbh) {
1129                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1130                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1131                 } else {
1132                         nextproto = ip6->ip6_nxt;
1133                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1134                 }
1135
1136                 /*
1137                  * Loop through length of segment after first fragment,
1138                  * make new header and copy data of each part and link onto
1139                  * chain.
1140                  */
1141                 m0 = m;
1142                 for (off = hlen; off < tlen; off += len) {
1143                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1144                         if (!m) {
1145                                 error = ENOBUFS;
1146                                 ip6stat.ip6s_odropped++;
1147                                 goto sendorfree;
1148                         }
1149                         m->m_pkthdr.rcvif = NULL;
1150                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1151                         *mnext = m;
1152                         mnext = &m->m_nextpkt;
1153                         m->m_data += max_linkhdr;
1154                         mhip6 = mtod(m, struct ip6_hdr *);
1155                         *mhip6 = *ip6;
1156                         m->m_len = sizeof(*mhip6);
1157                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1158                         if (error) {
1159                                 ip6stat.ip6s_odropped++;
1160                                 goto sendorfree;
1161                         }
1162                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1163                         if (off + len >= tlen)
1164                                 len = tlen - off;
1165                         else
1166                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1167                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1168                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1169                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1170                                 error = ENOBUFS;
1171                                 ip6stat.ip6s_odropped++;
1172                                 goto sendorfree;
1173                         }
1174                         m_cat(m, m_frgpart);
1175                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1176                         m->m_pkthdr.rcvif = NULL;
1177                         ip6f->ip6f_reserved = 0;
1178                         ip6f->ip6f_ident = id;
1179                         ip6f->ip6f_nxt = nextproto;
1180                         ip6stat.ip6s_ofragments++;
1181                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1182                 }
1183
1184                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1185         }
1186
1187         /*
1188          * Remove leading garbages.
1189          */
1190 sendorfree:
1191         m = m0->m_nextpkt;
1192         m0->m_nextpkt = 0;
1193         m_freem(m0);
1194         for (m0 = m; m; m = m0) {
1195                 m0 = m->m_nextpkt;
1196                 m->m_nextpkt = 0;
1197                 if (error == 0) {
1198                         /* Record statistics for this interface address. */
1199                         if (ia) {
1200                                 ia->ia_ifa.if_opackets++;
1201                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1202                         }
1203 #ifdef IPSEC
1204                         /* clean ipsec history once it goes out of the node */
1205                         ipsec_delaux(m);
1206 #endif
1207                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1208                 } else
1209                         m_freem(m);
1210         }
1211
1212         if (error == 0)
1213                 ip6stat.ip6s_fragmented++;
1214
1215 done:
1216         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1217                 RTFREE(ro->ro_rt);
1218         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1219                 RTFREE(ro_pmtu->ro_rt);
1220         }
1221
1222 #ifdef IPSEC
1223         if (sp != NULL)
1224                 key_freesp(sp);
1225 #endif /* IPSEC */
1226 #ifdef FAST_IPSEC
1227         if (sp != NULL)
1228                 KEY_FREESP(&sp);
1229 #endif /* FAST_IPSEC */
1230
1231         return (error);
1232
1233 freehdrs:
1234         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1235         m_freem(exthdrs.ip6e_dest1);
1236         m_freem(exthdrs.ip6e_rthdr);
1237         m_freem(exthdrs.ip6e_dest2);
1238         /* FALLTHROUGH */
1239 bad:
1240         m_freem(m);
1241         goto done;
1242 }
1243
1244 static int
1245 ip6_copyexthdr(mp, hdr, hlen)
1246         struct mbuf **mp;
1247         caddr_t hdr;
1248         int hlen;
1249 {
1250         struct mbuf *m;
1251
1252         if (hlen > MCLBYTES)
1253                 return (ENOBUFS); /* XXX */
1254
1255         MGET(m, M_DONTWAIT, MT_DATA);
1256         if (!m)
1257                 return (ENOBUFS);
1258
1259         if (hlen > MLEN) {
1260                 MCLGET(m, M_DONTWAIT);
1261                 if ((m->m_flags & M_EXT) == 0) {
1262                         m_free(m);
1263                         return (ENOBUFS);
1264                 }
1265         }
1266         m->m_len = hlen;
1267         if (hdr)
1268                 bcopy(hdr, mtod(m, caddr_t), hlen);
1269
1270         *mp = m;
1271         return (0);
1272 }
1273
1274 /*
1275  * Insert jumbo payload option.
1276  */
1277 static int
1278 ip6_insert_jumboopt(exthdrs, plen)
1279         struct ip6_exthdrs *exthdrs;
1280         u_int32_t plen;
1281 {
1282         struct mbuf *mopt;
1283         u_char *optbuf;
1284         u_int32_t v;
1285
1286 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1287
1288         /*
1289          * If there is no hop-by-hop options header, allocate new one.
1290          * If there is one but it doesn't have enough space to store the
1291          * jumbo payload option, allocate a cluster to store the whole options.
1292          * Otherwise, use it to store the options.
1293          */
1294         if (exthdrs->ip6e_hbh == 0) {
1295                 MGET(mopt, M_DONTWAIT, MT_DATA);
1296                 if (mopt == 0)
1297                         return (ENOBUFS);
1298                 mopt->m_len = JUMBOOPTLEN;
1299                 optbuf = mtod(mopt, u_char *);
1300                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1301                 exthdrs->ip6e_hbh = mopt;
1302         } else {
1303                 struct ip6_hbh *hbh;
1304
1305                 mopt = exthdrs->ip6e_hbh;
1306                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1307                         /*
1308                          * XXX assumption:
1309                          * - exthdrs->ip6e_hbh is not referenced from places
1310                          *   other than exthdrs.
1311                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1312                          */
1313                         int oldoptlen = mopt->m_len;
1314                         struct mbuf *n;
1315
1316                         /*
1317                          * XXX: give up if the whole (new) hbh header does
1318                          * not fit even in an mbuf cluster.
1319                          */
1320                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1321                                 return (ENOBUFS);
1322
1323                         /*
1324                          * As a consequence, we must always prepare a cluster
1325                          * at this point.
1326                          */
1327                         MGET(n, M_DONTWAIT, MT_DATA);
1328                         if (n) {
1329                                 MCLGET(n, M_DONTWAIT);
1330                                 if ((n->m_flags & M_EXT) == 0) {
1331                                         m_freem(n);
1332                                         n = NULL;
1333                                 }
1334                         }
1335                         if (!n)
1336                                 return (ENOBUFS);
1337                         n->m_len = oldoptlen + JUMBOOPTLEN;
1338                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1339                             oldoptlen);
1340                         optbuf = mtod(n, caddr_t) + oldoptlen;
1341                         m_freem(mopt);
1342                         mopt = exthdrs->ip6e_hbh = n;
1343                 } else {
1344                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1345                         mopt->m_len += JUMBOOPTLEN;
1346                 }
1347                 optbuf[0] = IP6OPT_PADN;
1348                 optbuf[1] = 1;
1349
1350                 /*
1351                  * Adjust the header length according to the pad and
1352                  * the jumbo payload option.
1353                  */
1354                 hbh = mtod(mopt, struct ip6_hbh *);
1355                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1356         }
1357
1358         /* fill in the option. */
1359         optbuf[2] = IP6OPT_JUMBO;
1360         optbuf[3] = 4;
1361         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1362         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1363
1364         /* finally, adjust the packet header length */
1365         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1366
1367         return (0);
1368 #undef JUMBOOPTLEN
1369 }
1370
1371 /*
1372  * Insert fragment header and copy unfragmentable header portions.
1373  */
1374 static int
1375 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1376         struct mbuf *m0, *m;
1377         int hlen;
1378         struct ip6_frag **frghdrp;
1379 {
1380         struct mbuf *n, *mlast;
1381
1382         if (hlen > sizeof(struct ip6_hdr)) {
1383                 n = m_copym(m0, sizeof(struct ip6_hdr),
1384                     hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1385                 if (n == 0)
1386                         return (ENOBUFS);
1387                 m->m_next = n;
1388         } else
1389                 n = m;
1390
1391         /* Search for the last mbuf of unfragmentable part. */
1392         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1393                 ;
1394
1395         if ((mlast->m_flags & M_EXT) == 0 &&
1396             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1397                 /* use the trailing space of the last mbuf for the fragment hdr */
1398                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1399                     mlast->m_len);
1400                 mlast->m_len += sizeof(struct ip6_frag);
1401                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1402         } else {
1403                 /* allocate a new mbuf for the fragment header */
1404                 struct mbuf *mfrg;
1405
1406                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1407                 if (mfrg == 0)
1408                         return (ENOBUFS);
1409                 mfrg->m_len = sizeof(struct ip6_frag);
1410                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1411                 mlast->m_next = mfrg;
1412         }
1413
1414         return (0);
1415 }
1416
1417 static int
1418 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp)
1419         struct route_in6 *ro_pmtu, *ro;
1420         struct ifnet *ifp;
1421         struct in6_addr *dst;
1422         u_long *mtup;
1423         int *alwaysfragp;
1424 {
1425         u_int32_t mtu = 0;
1426         int alwaysfrag = 0;
1427         int error = 0;
1428
1429         if (ro_pmtu != ro) {
1430                 /* The first hop and the final destination may differ. */
1431                 struct sockaddr_in6 *sa6_dst =
1432                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1433                 if (ro_pmtu->ro_rt &&
1434                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1435                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1436                         RTFREE(ro_pmtu->ro_rt);
1437                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1438                 }
1439                 if (ro_pmtu->ro_rt == NULL) {
1440                         bzero(sa6_dst, sizeof(*sa6_dst));
1441                         sa6_dst->sin6_family = AF_INET6;
1442                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1443                         sa6_dst->sin6_addr = *dst;
1444
1445                         rtalloc((struct route *)ro_pmtu);
1446                 }
1447         }
1448         if (ro_pmtu->ro_rt) {
1449                 u_int32_t ifmtu;
1450                 struct in_conninfo inc;
1451
1452                 bzero(&inc, sizeof(inc));
1453                 inc.inc_flags = 1; /* IPv6 */
1454                 inc.inc6_faddr = *dst;
1455
1456                 if (ifp == NULL)
1457                         ifp = ro_pmtu->ro_rt->rt_ifp;
1458                 ifmtu = IN6_LINKMTU(ifp);
1459                 mtu = tcp_hc_getmtu(&inc);
1460                 if (mtu)
1461                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1462                 else
1463                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1464                 if (mtu == 0)
1465                         mtu = ifmtu;
1466                 else if (mtu < IPV6_MMTU) {
1467                         /*
1468                          * RFC2460 section 5, last paragraph:
1469                          * if we record ICMPv6 too big message with
1470                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1471                          * or smaller, with framgent header attached.
1472                          * (fragment header is needed regardless from the
1473                          * packet size, for translators to identify packets)
1474                          */
1475                         alwaysfrag = 1;
1476                         mtu = IPV6_MMTU;
1477                 } else if (mtu > ifmtu) {
1478                         /*
1479                          * The MTU on the route is larger than the MTU on
1480                          * the interface!  This shouldn't happen, unless the
1481                          * MTU of the interface has been changed after the
1482                          * interface was brought up.  Change the MTU in the
1483                          * route to match the interface MTU (as long as the
1484                          * field isn't locked).
1485                          */
1486                         mtu = ifmtu;
1487                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1488                 }
1489         } else if (ifp) {
1490                 mtu = IN6_LINKMTU(ifp);
1491         } else
1492                 error = EHOSTUNREACH; /* XXX */
1493
1494         *mtup = mtu;
1495         if (alwaysfragp)
1496                 *alwaysfragp = alwaysfrag;
1497         return (error);
1498 }
1499
1500 /*
1501  * IP6 socket option processing.
1502  */
1503 int
1504 ip6_ctloutput(so, sopt)
1505         struct socket *so;
1506         struct sockopt *sopt;
1507 {
1508         int privileged, optdatalen, uproto;
1509         void *optdata;
1510         struct inpcb *in6p = sotoinpcb(so);
1511         int error, optval;
1512         int level, op, optname;
1513         int optlen;
1514         struct thread *td;
1515
1516         if (sopt) {
1517                 level = sopt->sopt_level;
1518                 op = sopt->sopt_dir;
1519                 optname = sopt->sopt_name;
1520                 optlen = sopt->sopt_valsize;
1521                 td = sopt->sopt_td;
1522         } else {
1523                 panic("ip6_ctloutput: arg soopt is NULL");
1524         }
1525         error = optval = 0;
1526
1527         privileged = (td == 0 || suser(td)) ? 0 : 1;
1528         uproto = (int)so->so_proto->pr_protocol;
1529
1530         if (level == IPPROTO_IPV6) {
1531                 switch (op) {
1532
1533                 case SOPT_SET:
1534                         switch (optname) {
1535                         case IPV6_2292PKTOPTIONS:
1536 #ifdef IPV6_PKTOPTIONS
1537                         case IPV6_PKTOPTIONS:
1538 #endif
1539                         {
1540                                 struct mbuf *m;
1541
1542                                 error = soopt_getm(sopt, &m); /* XXX */
1543                                 if (error != 0)
1544                                         break;
1545                                 error = soopt_mcopyin(sopt, m); /* XXX */
1546                                 if (error != 0)
1547                                         break;
1548                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1549                                                     m, so, sopt);
1550                                 m_freem(m); /* XXX */
1551                                 break;
1552                         }
1553
1554                         /*
1555                          * Use of some Hop-by-Hop options or some
1556                          * Destination options, might require special
1557                          * privilege.  That is, normal applications
1558                          * (without special privilege) might be forbidden
1559                          * from setting certain options in outgoing packets,
1560                          * and might never see certain options in received
1561                          * packets. [RFC 2292 Section 6]
1562                          * KAME specific note:
1563                          *  KAME prevents non-privileged users from sending or
1564                          *  receiving ANY hbh/dst options in order to avoid
1565                          *  overhead of parsing options in the kernel.
1566                          */
1567                         case IPV6_RECVHOPOPTS:
1568                         case IPV6_RECVDSTOPTS:
1569                         case IPV6_RECVRTHDRDSTOPTS:
1570                                 if (!privileged) {
1571                                         error = EPERM;
1572                                         break;
1573                                 }
1574                                 /* FALLTHROUGH */
1575                         case IPV6_UNICAST_HOPS:
1576                         case IPV6_HOPLIMIT:
1577                         case IPV6_FAITH:
1578
1579                         case IPV6_RECVPKTINFO:
1580                         case IPV6_RECVHOPLIMIT:
1581                         case IPV6_RECVRTHDR:
1582                         case IPV6_RECVPATHMTU:
1583                         case IPV6_RECVTCLASS:
1584                         case IPV6_V6ONLY:
1585                         case IPV6_AUTOFLOWLABEL:
1586                                 if (optlen != sizeof(int)) {
1587                                         error = EINVAL;
1588                                         break;
1589                                 }
1590                                 error = sooptcopyin(sopt, &optval,
1591                                         sizeof optval, sizeof optval);
1592                                 if (error)
1593                                         break;
1594                                 switch (optname) {
1595
1596                                 case IPV6_UNICAST_HOPS:
1597                                         if (optval < -1 || optval >= 256)
1598                                                 error = EINVAL;
1599                                         else {
1600                                                 /* -1 = kernel default */
1601                                                 in6p->in6p_hops = optval;
1602                                                 if ((in6p->in6p_vflag &
1603                                                      INP_IPV4) != 0)
1604                                                         in6p->inp_ip_ttl = optval;
1605                                         }
1606                                         break;
1607 #define OPTSET(bit) \
1608 do { \
1609         if (optval) \
1610                 in6p->in6p_flags |= (bit); \
1611         else \
1612                 in6p->in6p_flags &= ~(bit); \
1613 } while (/*CONSTCOND*/ 0)
1614 #define OPTSET2292(bit) \
1615 do { \
1616         in6p->in6p_flags |= IN6P_RFC2292; \
1617         if (optval) \
1618                 in6p->in6p_flags |= (bit); \
1619         else \
1620                 in6p->in6p_flags &= ~(bit); \
1621 } while (/*CONSTCOND*/ 0)
1622 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1623
1624                                 case IPV6_RECVPKTINFO:
1625                                         /* cannot mix with RFC2292 */
1626                                         if (OPTBIT(IN6P_RFC2292)) {
1627                                                 error = EINVAL;
1628                                                 break;
1629                                         }
1630                                         OPTSET(IN6P_PKTINFO);
1631                                         break;
1632
1633                                 case IPV6_HOPLIMIT:
1634                                 {
1635                                         struct ip6_pktopts **optp;
1636
1637                                         /* cannot mix with RFC2292 */
1638                                         if (OPTBIT(IN6P_RFC2292)) {
1639                                                 error = EINVAL;
1640                                                 break;
1641                                         }
1642                                         optp = &in6p->in6p_outputopts;
1643                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1644                                                            (u_char *)&optval,
1645                                                            sizeof(optval),
1646                                                            optp,
1647                                                            privileged, uproto);
1648                                         break;
1649                                 }
1650
1651                                 case IPV6_RECVHOPLIMIT:
1652                                         /* cannot mix with RFC2292 */
1653                                         if (OPTBIT(IN6P_RFC2292)) {
1654                                                 error = EINVAL;
1655                                                 break;
1656                                         }
1657                                         OPTSET(IN6P_HOPLIMIT);
1658                                         break;
1659
1660                                 case IPV6_RECVHOPOPTS:
1661                                         /* cannot mix with RFC2292 */
1662                                         if (OPTBIT(IN6P_RFC2292)) {
1663                                                 error = EINVAL;
1664                                                 break;
1665                                         }
1666                                         OPTSET(IN6P_HOPOPTS);
1667                                         break;
1668
1669                                 case IPV6_RECVDSTOPTS:
1670                                         /* cannot mix with RFC2292 */
1671                                         if (OPTBIT(IN6P_RFC2292)) {
1672                                                 error = EINVAL;
1673                                                 break;
1674                                         }
1675                                         OPTSET(IN6P_DSTOPTS);
1676                                         break;
1677
1678                                 case IPV6_RECVRTHDRDSTOPTS:
1679                                         /* cannot mix with RFC2292 */
1680                                         if (OPTBIT(IN6P_RFC2292)) {
1681                                                 error = EINVAL;
1682                                                 break;
1683                                         }
1684                                         OPTSET(IN6P_RTHDRDSTOPTS);
1685                                         break;
1686
1687                                 case IPV6_RECVRTHDR:
1688                                         /* cannot mix with RFC2292 */
1689                                         if (OPTBIT(IN6P_RFC2292)) {
1690                                                 error = EINVAL;
1691                                                 break;
1692                                         }
1693                                         OPTSET(IN6P_RTHDR);
1694                                         break;
1695
1696                                 case IPV6_FAITH:
1697                                         OPTSET(IN6P_FAITH);
1698                                         break;
1699
1700                                 case IPV6_RECVPATHMTU:
1701                                         /*
1702                                          * We ignore this option for TCP
1703                                          * sockets.
1704                                          * (rfc2292bis leaves this case
1705                                          * unspecified.)
1706                                          */
1707                                         if (uproto != IPPROTO_TCP)
1708                                                 OPTSET(IN6P_MTU);
1709                                         break;
1710
1711                                 case IPV6_V6ONLY:
1712                                         /*
1713                                          * make setsockopt(IPV6_V6ONLY)
1714                                          * available only prior to bind(2).
1715                                          * see ipng mailing list, Jun 22 2001.
1716                                          */
1717                                         if (in6p->in6p_lport ||
1718                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1719                                                 error = EINVAL;
1720                                                 break;
1721                                         }
1722                                         OPTSET(IN6P_IPV6_V6ONLY);
1723                                         if (optval)
1724                                                 in6p->in6p_vflag &= ~INP_IPV4;
1725                                         else
1726                                                 in6p->in6p_vflag |= INP_IPV4;
1727                                         break;
1728                                 case IPV6_RECVTCLASS:
1729                                         /* cannot mix with RFC2292 XXX */
1730                                         if (OPTBIT(IN6P_RFC2292)) {
1731                                                 error = EINVAL;
1732                                                 break;
1733                                         }
1734                                         OPTSET(IN6P_TCLASS);
1735                                         break;
1736                                 case IPV6_AUTOFLOWLABEL:
1737                                         OPTSET(IN6P_AUTOFLOWLABEL);
1738                                         break;
1739
1740                                 }
1741                                 break;
1742
1743                         case IPV6_TCLASS:
1744                         case IPV6_DONTFRAG:
1745                         case IPV6_USE_MIN_MTU:
1746                         case IPV6_PREFER_TEMPADDR:
1747                                 if (optlen != sizeof(optval)) {
1748                                         error = EINVAL;
1749                                         break;
1750                                 }
1751                                 error = sooptcopyin(sopt, &optval,
1752                                         sizeof optval, sizeof optval);
1753                                 if (error)
1754                                         break;
1755                                 {
1756                                         struct ip6_pktopts **optp;
1757                                         optp = &in6p->in6p_outputopts;
1758                                         error = ip6_pcbopt(optname,
1759                                                            (u_char *)&optval,
1760                                                            sizeof(optval),
1761                                                            optp,
1762                                                            privileged, uproto);
1763                                         break;
1764                                 }
1765
1766                         case IPV6_2292PKTINFO:
1767                         case IPV6_2292HOPLIMIT:
1768                         case IPV6_2292HOPOPTS:
1769                         case IPV6_2292DSTOPTS:
1770                         case IPV6_2292RTHDR:
1771                                 /* RFC 2292 */
1772                                 if (optlen != sizeof(int)) {
1773                                         error = EINVAL;
1774                                         break;
1775                                 }
1776                                 error = sooptcopyin(sopt, &optval,
1777                                         sizeof optval, sizeof optval);
1778                                 if (error)
1779                                         break;
1780                                 switch (optname) {
1781                                 case IPV6_2292PKTINFO:
1782                                         OPTSET2292(IN6P_PKTINFO);
1783                                         break;
1784                                 case IPV6_2292HOPLIMIT:
1785                                         OPTSET2292(IN6P_HOPLIMIT);
1786                                         break;
1787                                 case IPV6_2292HOPOPTS:
1788                                         /*
1789                                          * Check super-user privilege.
1790                                          * See comments for IPV6_RECVHOPOPTS.
1791                                          */
1792                                         if (!privileged)
1793                                                 return (EPERM);
1794                                         OPTSET2292(IN6P_HOPOPTS);
1795                                         break;
1796                                 case IPV6_2292DSTOPTS:
1797                                         if (!privileged)
1798                                                 return (EPERM);
1799                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1800                                         break;
1801                                 case IPV6_2292RTHDR:
1802                                         OPTSET2292(IN6P_RTHDR);
1803                                         break;
1804                                 }
1805                                 break;
1806                         case IPV6_PKTINFO:
1807                         case IPV6_HOPOPTS:
1808                         case IPV6_RTHDR:
1809                         case IPV6_DSTOPTS:
1810                         case IPV6_RTHDRDSTOPTS:
1811                         case IPV6_NEXTHOP:
1812                         {
1813                                 /* new advanced API (2292bis) */
1814                                 u_char *optbuf;
1815                                 int optlen;
1816                                 struct ip6_pktopts **optp;
1817
1818                                 /* cannot mix with RFC2292 */
1819                                 if (OPTBIT(IN6P_RFC2292)) {
1820                                         error = EINVAL;
1821                                         break;
1822                                 }
1823
1824                                 switch (optname) {
1825                                 case IPV6_HOPOPTS:
1826                                 case IPV6_DSTOPTS:
1827                                 case IPV6_RTHDRDSTOPTS:
1828                                 case IPV6_NEXTHOP:
1829                                         if (!privileged)
1830                                                 error = EPERM;
1831                                         break;
1832                                 }
1833                                 if (error)
1834                                         break;
1835
1836                                 switch (optname) {
1837                                 case IPV6_PKTINFO:
1838                                         optlen = sizeof(struct in6_pktinfo);
1839                                         break;
1840                                 case IPV6_NEXTHOP:
1841                                         optlen = SOCK_MAXADDRLEN;
1842                                         break;
1843                                 default:
1844                                         optlen = IPV6_MAXOPTHDR;
1845                                         break;
1846                                 }
1847                                 if (sopt->sopt_valsize > optlen) {
1848                                         error = EINVAL;
1849                                         break;
1850                                 }
1851
1852                                 optlen = sopt->sopt_valsize;
1853                                 optbuf = malloc(optlen, M_TEMP, M_WAITOK);
1854                                 error = sooptcopyin(sopt, optbuf, optlen,
1855                                     optlen);
1856                                 if (error) {
1857                                         free(optbuf, M_TEMP);
1858                                         break;
1859                                 }
1860
1861                                 optp = &in6p->in6p_outputopts;
1862                                 error = ip6_pcbopt(optname,
1863                                                    optbuf, optlen,
1864                                                    optp, privileged, uproto);
1865                                 free(optbuf, M_TEMP);
1866                                 break;
1867                         }
1868 #undef OPTSET
1869
1870                         case IPV6_MULTICAST_IF:
1871                         case IPV6_MULTICAST_HOPS:
1872                         case IPV6_MULTICAST_LOOP:
1873                         case IPV6_JOIN_GROUP:
1874                         case IPV6_LEAVE_GROUP:
1875                             {
1876                                 if (sopt->sopt_valsize > MLEN) {
1877                                         error = EMSGSIZE;
1878                                         break;
1879                                 }
1880                                 /* XXX */
1881                             }
1882                             /* FALLTHROUGH */
1883                             {
1884                                 struct mbuf *m;
1885
1886                                 if (sopt->sopt_valsize > MCLBYTES) {
1887                                         error = EMSGSIZE;
1888                                         break;
1889                                 }
1890                                 /* XXX */
1891                                 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_HEADER);
1892                                 if (m == 0) {
1893                                         error = ENOBUFS;
1894                                         break;
1895                                 }
1896                                 if (sopt->sopt_valsize > MLEN) {
1897                                         MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1898                                         if ((m->m_flags & M_EXT) == 0) {
1899                                                 m_free(m);
1900                                                 error = ENOBUFS;
1901                                                 break;
1902                                         }
1903                                 }
1904                                 m->m_len = sopt->sopt_valsize;
1905                                 error = sooptcopyin(sopt, mtod(m, char *),
1906                                                     m->m_len, m->m_len);
1907                                 if (error) {
1908                                         (void)m_free(m);
1909                                         break;
1910                                 }
1911                                 error = ip6_setmoptions(sopt->sopt_name,
1912                                                         &in6p->in6p_moptions,
1913                                                         m);
1914                                 (void)m_free(m);
1915                             }
1916                                 break;
1917
1918                         case IPV6_PORTRANGE:
1919                                 error = sooptcopyin(sopt, &optval,
1920                                     sizeof optval, sizeof optval);
1921                                 if (error)
1922                                         break;
1923
1924                                 switch (optval) {
1925                                 case IPV6_PORTRANGE_DEFAULT:
1926                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1927                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1928                                         break;
1929
1930                                 case IPV6_PORTRANGE_HIGH:
1931                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1932                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1933                                         break;
1934
1935                                 case IPV6_PORTRANGE_LOW:
1936                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1937                                         in6p->in6p_flags |= IN6P_LOWPORT;
1938                                         break;
1939
1940                                 default:
1941                                         error = EINVAL;
1942                                         break;
1943                                 }
1944                                 break;
1945
1946 #if defined(IPSEC) || defined(FAST_IPSEC)
1947                         case IPV6_IPSEC_POLICY:
1948                             {
1949                                 caddr_t req = NULL;
1950                                 size_t len = 0;
1951                                 struct mbuf *m;
1952
1953                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1954                                         break;
1955                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1956                                         break;
1957                                 if (m) {
1958                                         req = mtod(m, caddr_t);
1959                                         len = m->m_len;
1960                                 }
1961                                 error = ipsec6_set_policy(in6p, optname, req,
1962                                                           len, privileged);
1963                                 m_freem(m);
1964                             }
1965                                 break;
1966 #endif /* KAME IPSEC */
1967
1968                         case IPV6_FW_ADD:
1969                         case IPV6_FW_DEL:
1970                         case IPV6_FW_FLUSH:
1971                         case IPV6_FW_ZERO:
1972                             {
1973                                 struct mbuf *m;
1974                                 struct mbuf **mp = &m;
1975
1976                                 if (ip6_fw_ctl_ptr == NULL)
1977                                         return EINVAL;
1978                                 /* XXX */
1979                                 if ((error = soopt_getm(sopt, &m)) != 0)
1980                                         break;
1981                                 /* XXX */
1982                                 if ((error = soopt_mcopyin(sopt, m)) != 0)
1983                                         break;
1984                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1985                                 m = *mp;
1986                             }
1987                                 break;
1988
1989                         default:
1990                                 error = ENOPROTOOPT;
1991                                 break;
1992                         }
1993                         break;
1994
1995                 case SOPT_GET:
1996                         switch (optname) {
1997
1998                         case IPV6_2292PKTOPTIONS:
1999 #ifdef IPV6_PKTOPTIONS
2000                         case IPV6_PKTOPTIONS:
2001 #endif
2002                                 /*
2003                                  * RFC3542 (effectively) deprecated the
2004                                  * semantics of the 2292-style pktoptions.
2005                                  * Since it was not reliable in nature (i.e.,
2006                                  * applications had to expect the lack of some
2007                                  * information after all), it would make sense
2008                                  * to simplify this part by always returning
2009                                  * empty data.
2010                                  */
2011                                 sopt->sopt_valsize = 0;
2012                                 break;
2013
2014                         case IPV6_RECVHOPOPTS:
2015                         case IPV6_RECVDSTOPTS:
2016                         case IPV6_RECVRTHDRDSTOPTS:
2017                         case IPV6_UNICAST_HOPS:
2018                         case IPV6_RECVPKTINFO:
2019                         case IPV6_RECVHOPLIMIT:
2020                         case IPV6_RECVRTHDR:
2021                         case IPV6_RECVPATHMTU:
2022
2023                         case IPV6_FAITH:
2024                         case IPV6_V6ONLY:
2025                         case IPV6_PORTRANGE:
2026                         case IPV6_RECVTCLASS:
2027                         case IPV6_AUTOFLOWLABEL:
2028                                 switch (optname) {
2029
2030                                 case IPV6_RECVHOPOPTS:
2031                                         optval = OPTBIT(IN6P_HOPOPTS);
2032                                         break;
2033
2034                                 case IPV6_RECVDSTOPTS:
2035                                         optval = OPTBIT(IN6P_DSTOPTS);
2036                                         break;
2037
2038                                 case IPV6_RECVRTHDRDSTOPTS:
2039                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2040                                         break;
2041
2042                                 case IPV6_UNICAST_HOPS:
2043                                         optval = in6p->in6p_hops;
2044                                         break;
2045
2046                                 case IPV6_RECVPKTINFO:
2047                                         optval = OPTBIT(IN6P_PKTINFO);
2048                                         break;
2049
2050                                 case IPV6_RECVHOPLIMIT:
2051                                         optval = OPTBIT(IN6P_HOPLIMIT);
2052                                         break;
2053
2054                                 case IPV6_RECVRTHDR:
2055                                         optval = OPTBIT(IN6P_RTHDR);
2056                                         break;
2057
2058                                 case IPV6_RECVPATHMTU:
2059                                         optval = OPTBIT(IN6P_MTU);
2060                                         break;
2061
2062                                 case IPV6_FAITH:
2063                                         optval = OPTBIT(IN6P_FAITH);
2064                                         break;
2065
2066                                 case IPV6_V6ONLY:
2067                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2068                                         break;
2069
2070                                 case IPV6_PORTRANGE:
2071                                     {
2072                                         int flags;
2073                                         flags = in6p->in6p_flags;
2074                                         if (flags & IN6P_HIGHPORT)
2075                                                 optval = IPV6_PORTRANGE_HIGH;
2076                                         else if (flags & IN6P_LOWPORT)
2077                                                 optval = IPV6_PORTRANGE_LOW;
2078                                         else
2079                                                 optval = 0;
2080                                         break;
2081                                     }
2082                                 case IPV6_RECVTCLASS:
2083                                         optval = OPTBIT(IN6P_TCLASS);
2084                                         break;
2085
2086                                 case IPV6_AUTOFLOWLABEL:
2087                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2088                                         break;
2089                                 }
2090                                 if (error)
2091                                         break;
2092                                 error = sooptcopyout(sopt, &optval,
2093                                         sizeof optval);
2094                                 break;
2095
2096                         case IPV6_PATHMTU:
2097                         {
2098                                 u_long pmtu = 0;
2099                                 struct ip6_mtuinfo mtuinfo;
2100                                 struct route_in6 sro;
2101
2102                                 bzero(&sro, sizeof(sro));
2103
2104                                 if (!(so->so_state & SS_ISCONNECTED))
2105                                         return (ENOTCONN);
2106                                 /*
2107                                  * XXX: we dot not consider the case of source
2108                                  * routing, or optional information to specify
2109                                  * the outgoing interface.
2110                                  */
2111                                 error = ip6_getpmtu(&sro, NULL, NULL,
2112                                     &in6p->in6p_faddr, &pmtu, NULL);
2113                                 if (sro.ro_rt)
2114                                         RTFREE(sro.ro_rt);
2115                                 if (error)
2116                                         break;
2117                                 if (pmtu > IPV6_MAXPACKET)
2118                                         pmtu = IPV6_MAXPACKET;
2119
2120                                 bzero(&mtuinfo, sizeof(mtuinfo));
2121                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2122                                 optdata = (void *)&mtuinfo;
2123                                 optdatalen = sizeof(mtuinfo);
2124                                 error = sooptcopyout(sopt, optdata,
2125                                     optdatalen);
2126                                 break;
2127                         }
2128
2129                         case IPV6_2292PKTINFO:
2130                         case IPV6_2292HOPLIMIT:
2131                         case IPV6_2292HOPOPTS:
2132                         case IPV6_2292RTHDR:
2133                         case IPV6_2292DSTOPTS:
2134                                 switch (optname) {
2135                                 case IPV6_2292PKTINFO:
2136                                         optval = OPTBIT(IN6P_PKTINFO);
2137                                         break;
2138                                 case IPV6_2292HOPLIMIT:
2139                                         optval = OPTBIT(IN6P_HOPLIMIT);
2140                                         break;
2141                                 case IPV6_2292HOPOPTS:
2142                                         optval = OPTBIT(IN6P_HOPOPTS);
2143                                         break;
2144                                 case IPV6_2292RTHDR:
2145                                         optval = OPTBIT(IN6P_RTHDR);
2146                                         break;
2147                                 case IPV6_2292DSTOPTS:
2148                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2149                                         break;
2150                                 }
2151                                 error = sooptcopyout(sopt, &optval,
2152                                     sizeof optval);
2153                                 break;
2154                         case IPV6_PKTINFO:
2155                         case IPV6_HOPOPTS:
2156                         case IPV6_RTHDR:
2157                         case IPV6_DSTOPTS:
2158                         case IPV6_RTHDRDSTOPTS:
2159                         case IPV6_NEXTHOP:
2160                         case IPV6_TCLASS:
2161                         case IPV6_DONTFRAG:
2162                         case IPV6_USE_MIN_MTU:
2163                         case IPV6_PREFER_TEMPADDR:
2164                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2165                                     optname, sopt);
2166                                 break;
2167
2168                         case IPV6_MULTICAST_IF:
2169                         case IPV6_MULTICAST_HOPS:
2170                         case IPV6_MULTICAST_LOOP:
2171                         case IPV6_JOIN_GROUP:
2172                         case IPV6_LEAVE_GROUP:
2173                             {
2174                                 struct mbuf *m;
2175                                 error = ip6_getmoptions(sopt->sopt_name,
2176                                     in6p->in6p_moptions, &m);
2177                                 if (error == 0)
2178                                         error = sooptcopyout(sopt,
2179                                             mtod(m, char *), m->m_len);
2180                                 m_freem(m);
2181                             }
2182                                 break;
2183
2184 #if defined(IPSEC) || defined(FAST_IPSEC)
2185                         case IPV6_IPSEC_POLICY:
2186                           {
2187                                 caddr_t req = NULL;
2188                                 size_t len = 0;
2189                                 struct mbuf *m = NULL;
2190                                 struct mbuf **mp = &m;
2191                                 size_t ovalsize = sopt->sopt_valsize;
2192                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2193
2194                                 error = soopt_getm(sopt, &m); /* XXX */
2195                                 if (error != 0)
2196                                         break;
2197                                 error = soopt_mcopyin(sopt, m); /* XXX */
2198                                 if (error != 0)
2199                                         break;
2200                                 sopt->sopt_valsize = ovalsize;
2201                                 sopt->sopt_val = oval;
2202                                 if (m) {
2203                                         req = mtod(m, caddr_t);
2204                                         len = m->m_len;
2205                                 }
2206                                 error = ipsec6_get_policy(in6p, req, len, mp);
2207                                 if (error == 0)
2208                                         error = soopt_mcopyout(sopt, m); /* XXX */
2209                                 if (error == 0 && m)
2210                                         m_freem(m);
2211                                 break;
2212                           }
2213 #endif /* KAME IPSEC */
2214
2215                         case IPV6_FW_GET:
2216                           {
2217                                 struct mbuf *m;
2218                                 struct mbuf **mp = &m;
2219
2220                                 if (ip6_fw_ctl_ptr == NULL)
2221                                 {
2222                                         return EINVAL;
2223                                 }
2224                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
2225                                 if (error == 0)
2226                                         error = soopt_mcopyout(sopt, m); /* XXX */
2227                                 if (error == 0 && m)
2228                                         m_freem(m);
2229                           }
2230                                 break;
2231
2232                         default:
2233                                 error = ENOPROTOOPT;
2234                                 break;
2235                         }
2236                         break;
2237                 }
2238         } else {                /* level != IPPROTO_IPV6 */
2239                 error = EINVAL;
2240         }
2241         return (error);
2242 }
2243
2244 int
2245 ip6_raw_ctloutput(so, sopt)
2246         struct socket *so;
2247         struct sockopt *sopt;
2248 {
2249         int error = 0, optval, optlen;
2250         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2251         struct in6pcb *in6p = sotoin6pcb(so);
2252         int level, op, optname;
2253
2254         if (sopt) {
2255                 level = sopt->sopt_level;
2256                 op = sopt->sopt_dir;
2257                 optname = sopt->sopt_name;
2258                 optlen = sopt->sopt_valsize;
2259         } else
2260                 panic("ip6_raw_ctloutput: arg soopt is NULL");
2261
2262         if (level != IPPROTO_IPV6) {
2263                 return (EINVAL);
2264         }
2265
2266         switch (optname) {
2267         case IPV6_CHECKSUM:
2268                 /*
2269                  * For ICMPv6 sockets, no modification allowed for checksum
2270                  * offset, permit "no change" values to help existing apps.
2271                  *
2272                  * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM
2273                  * for an ICMPv6 socket will fail."
2274                  * The current behavior does not meet 2292bis.
2275                  */
2276                 switch (op) {
2277                 case SOPT_SET:
2278                         if (optlen != sizeof(int)) {
2279                                 error = EINVAL;
2280                                 break;
2281                         }
2282                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2283                                             sizeof(optval));
2284                         if (error)
2285                                 break;
2286                         if ((optval % 2) != 0) {
2287                                 /* the API assumes even offset values */
2288                                 error = EINVAL;
2289                         } else if (so->so_proto->pr_protocol ==
2290                             IPPROTO_ICMPV6) {
2291                                 if (optval != icmp6off)
2292                                         error = EINVAL;
2293                         } else
2294                                 in6p->in6p_cksum = optval;
2295                         break;
2296
2297                 case SOPT_GET:
2298                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2299                                 optval = icmp6off;
2300                         else
2301                                 optval = in6p->in6p_cksum;
2302
2303                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2304                         break;
2305
2306                 default:
2307                         error = EINVAL;
2308                         break;
2309                 }
2310                 break;
2311
2312         default:
2313                 error = ENOPROTOOPT;
2314                 break;
2315         }
2316
2317         return (error);
2318 }
2319
2320 /*
2321  * Set up IP6 options in pcb for insertion in output packets or
2322  * specifying behavior of outgoing packets.
2323  */
2324 static int
2325 ip6_pcbopts(pktopt, m, so, sopt)
2326         struct ip6_pktopts **pktopt;
2327         struct mbuf *m;
2328         struct socket *so;
2329         struct sockopt *sopt;
2330 {
2331         struct ip6_pktopts *opt = *pktopt;
2332         int error = 0;
2333         struct thread *td = sopt->sopt_td;
2334         int priv = 0;
2335
2336         /* turn off any old options. */
2337         if (opt) {
2338 #ifdef DIAGNOSTIC
2339                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2340                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2341                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2342                         printf("ip6_pcbopts: all specified options are cleared.\n");
2343 #endif
2344                 ip6_clearpktopts(opt, -1);
2345         } else
2346                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2347         *pktopt = NULL;
2348
2349         if (!m || m->m_len == 0) {
2350                 /*
2351                  * Only turning off any previous options, regardless of
2352                  * whether the opt is just created or given.
2353                  */
2354                 free(opt, M_IP6OPT);
2355                 return (0);
2356         }
2357
2358         /*  set options specified by user. */
2359         if (td && !suser(td))
2360                 priv = 1;
2361         if ((error = ip6_setpktoptions(m, opt, NULL, priv, 1,
2362             so->so_proto->pr_protocol)) != 0) {
2363                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2364                 free(opt, M_IP6OPT);
2365                 return (error);
2366         }
2367         *pktopt = opt;
2368         return (0);
2369 }
2370
2371 /*
2372  * initialize ip6_pktopts.  beware that there are non-zero default values in
2373  * the struct.
2374  */
2375 void
2376 init_ip6pktopts(opt)
2377         struct ip6_pktopts *opt;
2378 {
2379
2380         bzero(opt, sizeof(*opt));
2381         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2382         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2383         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2384         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2385 }
2386
2387 static int
2388 ip6_pcbopt(optname, buf, len, pktopt, priv, uproto)
2389         int optname, len, priv;
2390         u_char *buf;
2391         struct ip6_pktopts **pktopt;
2392         int uproto;
2393 {
2394         struct ip6_pktopts *opt;
2395
2396         if (*pktopt == NULL) {
2397                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2398                     M_WAITOK);
2399                 init_ip6pktopts(*pktopt);
2400                 (*pktopt)->needfree = 1;
2401         }
2402         opt = *pktopt;
2403
2404         return (ip6_setpktoption(optname, buf, len, opt, priv, 1, 0, uproto));
2405 }
2406
2407 static int
2408 ip6_getpcbopt(pktopt, optname, sopt)
2409         struct ip6_pktopts *pktopt;
2410         struct sockopt *sopt;
2411         int optname;
2412 {
2413         void *optdata = NULL;
2414         int optdatalen = 0;
2415         struct ip6_ext *ip6e;
2416         int error = 0;
2417         struct in6_pktinfo null_pktinfo;
2418         int deftclass = 0, on;
2419         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2420         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2421
2422         switch (optname) {
2423         case IPV6_PKTINFO:
2424                 if (pktopt && pktopt->ip6po_pktinfo)
2425                         optdata = (void *)pktopt->ip6po_pktinfo;
2426                 else {
2427                         /* XXX: we don't have to do this every time... */
2428                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2429                         optdata = (void *)&null_pktinfo;
2430                 }
2431                 optdatalen = sizeof(struct in6_pktinfo);
2432                 break;
2433         case IPV6_TCLASS:
2434                 if (pktopt && pktopt->ip6po_tclass >= 0)
2435                         optdata = (void *)&pktopt->ip6po_tclass;
2436                 else
2437                         optdata = (void *)&deftclass;
2438                 optdatalen = sizeof(int);
2439                 break;
2440         case IPV6_HOPOPTS:
2441                 if (pktopt && pktopt->ip6po_hbh) {
2442                         optdata = (void *)pktopt->ip6po_hbh;
2443                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2444                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2445                 }
2446                 break;
2447         case IPV6_RTHDR:
2448                 if (pktopt && pktopt->ip6po_rthdr) {
2449                         optdata = (void *)pktopt->ip6po_rthdr;
2450                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2451                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2452                 }
2453                 break;
2454         case IPV6_RTHDRDSTOPTS:
2455                 if (pktopt && pktopt->ip6po_dest1) {
2456                         optdata = (void *)pktopt->ip6po_dest1;
2457                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2458                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2459                 }
2460                 break;
2461         case IPV6_DSTOPTS:
2462                 if (pktopt && pktopt->ip6po_dest2) {
2463                         optdata = (void *)pktopt->ip6po_dest2;
2464                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2465                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2466                 }
2467                 break;
2468         case IPV6_NEXTHOP:
2469                 if (pktopt && pktopt->ip6po_nexthop) {
2470                         optdata = (void *)pktopt->ip6po_nexthop;
2471                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2472                 }
2473                 break;
2474         case IPV6_USE_MIN_MTU:
2475                 if (pktopt)
2476                         optdata = (void *)&pktopt->ip6po_minmtu;
2477                 else
2478                         optdata = (void *)&defminmtu;
2479                 optdatalen = sizeof(int);
2480                 break;
2481         case IPV6_DONTFRAG:
2482                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2483                         on = 1;
2484                 else
2485                         on = 0;
2486                 optdata = (void *)&on;
2487                 optdatalen = sizeof(on);
2488                 break;
2489         case IPV6_PREFER_TEMPADDR:
2490                 if (pktopt)
2491                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2492                 else
2493                         optdata = (void *)&defpreftemp;
2494                 optdatalen = sizeof(int);
2495                 break;
2496         default:                /* should not happen */
2497 #ifdef DIAGNOSTIC
2498                 panic("ip6_getpcbopt: unexpected option\n");
2499 #endif
2500                 return (ENOPROTOOPT);
2501         }
2502
2503         error = sooptcopyout(sopt, optdata, optdatalen);
2504
2505         return (error);
2506 }
2507
2508 void
2509 ip6_clearpktopts(pktopt, optname)
2510         struct ip6_pktopts *pktopt;
2511         int optname;
2512 {
2513         int needfree;
2514
2515         if (pktopt == NULL)
2516                 return;
2517
2518         needfree = pktopt->needfree;
2519
2520         if (optname == -1 || optname == IPV6_PKTINFO) {
2521                 if (needfree && pktopt->ip6po_pktinfo)
2522                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2523                 pktopt->ip6po_pktinfo = NULL;
2524         }
2525         if (optname == -1 || optname == IPV6_HOPLIMIT)
2526                 pktopt->ip6po_hlim = -1;
2527         if (optname == -1 || optname == IPV6_TCLASS)
2528                 pktopt->ip6po_tclass = -1;
2529         if (optname == -1 || optname == IPV6_NEXTHOP) {
2530                 if (pktopt->ip6po_nextroute.ro_rt) {
2531                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2532                         pktopt->ip6po_nextroute.ro_rt = NULL;
2533                 }
2534                 if (needfree && pktopt->ip6po_nexthop)
2535                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2536                 pktopt->ip6po_nexthop = NULL;
2537         }
2538         if (optname == -1 || optname == IPV6_HOPOPTS) {
2539                 if (needfree && pktopt->ip6po_hbh)
2540                         free(pktopt->ip6po_hbh, M_IP6OPT);
2541                 pktopt->ip6po_hbh = NULL;
2542         }
2543         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2544                 if (needfree && pktopt->ip6po_dest1)
2545                         free(pktopt->ip6po_dest1, M_IP6OPT);
2546                 pktopt->ip6po_dest1 = NULL;
2547         }
2548         if (optname == -1 || optname == IPV6_RTHDR) {
2549                 if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2550                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2551                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2552                 if (pktopt->ip6po_route.ro_rt) {
2553                         RTFREE(pktopt->ip6po_route.ro_rt);
2554                         pktopt->ip6po_route.ro_rt = NULL;
2555                 }
2556         }
2557         if (optname == -1 || optname == IPV6_DSTOPTS) {
2558                 if (needfree && pktopt->ip6po_dest2)
2559                         free(pktopt->ip6po_dest2, M_IP6OPT);
2560                 pktopt->ip6po_dest2 = NULL;
2561         }
2562 }
2563
2564 #define PKTOPT_EXTHDRCPY(type) \
2565 do {\
2566         if (src->type) {\
2567                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2568                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2569                 if (dst->type == NULL && canwait == M_NOWAIT)\
2570                         goto bad;\
2571                 bcopy(src->type, dst->type, hlen);\
2572         }\
2573 } while (/*CONSTCOND*/ 0)
2574
2575 struct ip6_pktopts *
2576 ip6_copypktopts(src, canwait)
2577         struct ip6_pktopts *src;
2578         int canwait;
2579 {
2580         struct ip6_pktopts *dst;
2581
2582         if (src == NULL) {
2583                 printf("ip6_clearpktopts: invalid argument\n");
2584                 return (NULL);
2585         }
2586
2587         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2588         if (dst == NULL && canwait == M_NOWAIT)
2589                 return (NULL);
2590         bzero(dst, sizeof(*dst));
2591         dst->needfree = 1;
2592
2593         dst->ip6po_hlim = src->ip6po_hlim;
2594         dst->ip6po_tclass = src->ip6po_tclass;
2595         dst->ip6po_flags = src->ip6po_flags;
2596         if (src->ip6po_pktinfo) {
2597                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2598                     M_IP6OPT, canwait);
2599                 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
2600                         goto bad;
2601                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2602         }
2603         if (src->ip6po_nexthop) {
2604                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2605                     M_IP6OPT, canwait);
2606                 if (dst->ip6po_nexthop == NULL)
2607                         goto bad;
2608                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2609                     src->ip6po_nexthop->sa_len);
2610         }
2611         PKTOPT_EXTHDRCPY(ip6po_hbh);
2612         PKTOPT_EXTHDRCPY(ip6po_dest1);
2613         PKTOPT_EXTHDRCPY(ip6po_dest2);
2614         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2615         return (dst);
2616
2617   bad:
2618         if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2619         if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2620         if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2621         if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2622         if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2623         if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2624         free(dst, M_IP6OPT);
2625         return (NULL);
2626 }
2627 #undef PKTOPT_EXTHDRCPY
2628
2629 void
2630 ip6_freepcbopts(pktopt)
2631         struct ip6_pktopts *pktopt;
2632 {
2633         if (pktopt == NULL)
2634                 return;
2635
2636         ip6_clearpktopts(pktopt, -1);
2637
2638         free(pktopt, M_IP6OPT);
2639 }
2640
2641 /*
2642  * Set the IP6 multicast options in response to user setsockopt().
2643  */
2644 static int
2645 ip6_setmoptions(optname, im6op, m)
2646         int optname;
2647         struct ip6_moptions **im6op;
2648         struct mbuf *m;
2649 {
2650         int error = 0;
2651         u_int loop, ifindex;
2652         struct ipv6_mreq *mreq;
2653         struct ifnet *ifp;
2654         struct ip6_moptions *im6o = *im6op;
2655         struct route_in6 ro;
2656         struct sockaddr_in6 *dst;
2657         struct in6_multi_mship *imm;
2658         struct thread *td = curthread;
2659
2660         if (im6o == NULL) {
2661                 /*
2662                  * No multicast option buffer attached to the pcb;
2663                  * allocate one and initialize to default values.
2664                  */
2665                 im6o = (struct ip6_moptions *)
2666                         malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2667
2668                 if (im6o == NULL)
2669                         return (ENOBUFS);
2670                 *im6op = im6o;
2671                 im6o->im6o_multicast_ifp = NULL;
2672                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2673                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2674                 LIST_INIT(&im6o->im6o_memberships);
2675         }
2676
2677         switch (optname) {
2678
2679         case IPV6_MULTICAST_IF:
2680                 /*
2681                  * Select the interface for outgoing multicast packets.
2682                  */
2683                 if (m == NULL || m->m_len != sizeof(u_int)) {
2684                         error = EINVAL;
2685                         break;
2686                 }
2687                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2688                 if (ifindex < 0 || if_index < ifindex) {
2689                         error = ENXIO;  /* XXX EINVAL? */
2690                         break;
2691                 }
2692                 ifp = ifnet_byindex(ifindex);
2693                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2694                         error = EADDRNOTAVAIL;
2695                         break;
2696                 }
2697                 im6o->im6o_multicast_ifp = ifp;
2698                 break;
2699
2700         case IPV6_MULTICAST_HOPS:
2701             {
2702                 /*
2703                  * Set the IP6 hoplimit for outgoing multicast packets.
2704                  */
2705                 int optval;
2706                 if (m == NULL || m->m_len != sizeof(int)) {
2707                         error = EINVAL;
2708                         break;
2709                 }
2710                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2711                 if (optval < -1 || optval >= 256)
2712                         error = EINVAL;
2713                 else if (optval == -1)
2714                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2715                 else
2716                         im6o->im6o_multicast_hlim = optval;
2717                 break;
2718             }
2719
2720         case IPV6_MULTICAST_LOOP:
2721                 /*
2722                  * Set the loopback flag for outgoing multicast packets.
2723                  * Must be zero or one.
2724                  */
2725                 if (m == NULL || m->m_len != sizeof(u_int)) {
2726                         error = EINVAL;
2727                         break;
2728                 }
2729                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2730                 if (loop > 1) {
2731                         error = EINVAL;
2732                         break;
2733                 }
2734                 im6o->im6o_multicast_loop = loop;
2735                 break;
2736
2737         case IPV6_JOIN_GROUP:
2738                 /*
2739                  * Add a multicast group membership.
2740                  * Group must be a valid IP6 multicast address.
2741                  */
2742                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2743                         error = EINVAL;
2744                         break;
2745                 }
2746                 mreq = mtod(m, struct ipv6_mreq *);
2747                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2748                         /*
2749                          * We use the unspecified address to specify to accept
2750                          * all multicast addresses. Only super user is allowed
2751                          * to do this.
2752                          */
2753                         if (suser(td)) {
2754                                 error = EACCES;
2755                                 break;
2756                         }
2757                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2758                         error = EINVAL;
2759                         break;
2760                 }
2761
2762                 /*
2763                  * If the interface is specified, validate it.
2764                  */
2765                 if (mreq->ipv6mr_interface < 0 ||
2766                     if_index < mreq->ipv6mr_interface) {
2767                         error = ENXIO;  /* XXX EINVAL? */
2768                         break;
2769                 }
2770                 /*
2771                  * If no interface was explicitly specified, choose an
2772                  * appropriate one according to the given multicast address.
2773                  */
2774                 if (mreq->ipv6mr_interface == 0) {
2775                         /*
2776                          * If the multicast address is in node-local scope,
2777                          * the interface should be a loopback interface.
2778                          * Otherwise, look up the routing table for the
2779                          * address, and choose the outgoing interface.
2780                          *   XXX: is it a good approach?
2781                          */
2782                         if (IN6_IS_ADDR_MC_INTFACELOCAL(&mreq->ipv6mr_multiaddr)) {
2783                                 ifp = &loif[0];
2784                         } else {
2785                                 ro.ro_rt = NULL;
2786                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2787                                 bzero(dst, sizeof(*dst));
2788                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2789                                 dst->sin6_family = AF_INET6;
2790                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2791                                 rtalloc((struct route *)&ro);
2792                                 if (ro.ro_rt == NULL) {
2793                                         error = EADDRNOTAVAIL;
2794                                         break;
2795                                 }
2796                                 ifp = ro.ro_rt->rt_ifp;
2797                                 RTFREE(ro.ro_rt);
2798                         }
2799                 } else
2800                         ifp = ifnet_byindex(mreq->ipv6mr_interface);
2801
2802                 /*
2803                  * See if we found an interface, and confirm that it
2804                  * supports multicast
2805                  */
2806                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2807                         error = EADDRNOTAVAIL;
2808                         break;
2809                 }
2810                 /*
2811                  * Put interface index into the multicast address,
2812                  * if the address has link-local scope.
2813                  */
2814                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2815                         mreq->ipv6mr_multiaddr.s6_addr16[1] =
2816                             htons(ifp->if_index);
2817                 }
2818                 /*
2819                  * See if the membership already exists.
2820                  */
2821                 for (imm = im6o->im6o_memberships.lh_first;
2822                      imm != NULL; imm = imm->i6mm_chain.le_next)
2823                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2824                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2825                                                &mreq->ipv6mr_multiaddr))
2826                                 break;
2827                 if (imm != NULL) {
2828                         error = EADDRINUSE;
2829                         break;
2830                 }
2831                 /*
2832                  * Everything looks good; add a new record to the multicast
2833                  * address list for the given interface.
2834                  */
2835                 imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2836                 if (imm == NULL) {
2837                         error = ENOBUFS;
2838                         break;
2839                 }
2840                 if ((imm->i6mm_maddr =
2841                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2842                         free(imm, M_IPMADDR);
2843                         break;
2844                 }
2845                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2846                 break;
2847
2848         case IPV6_LEAVE_GROUP:
2849                 /*
2850                  * Drop a multicast group membership.
2851                  * Group must be a valid IP6 multicast address.
2852                  */
2853                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2854                         error = EINVAL;
2855                         break;
2856                 }
2857                 mreq = mtod(m, struct ipv6_mreq *);
2858                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2859                         if (suser(td)) {
2860                                 error = EACCES;
2861                                 break;
2862                         }
2863                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2864                         error = EINVAL;
2865                         break;
2866                 }
2867                 /*
2868                  * If an interface address was specified, get a pointer
2869                  * to its ifnet structure.
2870                  */
2871                 if (mreq->ipv6mr_interface < 0
2872                  || if_index < mreq->ipv6mr_interface) {
2873                         error = ENXIO;  /* XXX EINVAL? */
2874                         break;
2875                 }
2876                 ifp = ifnet_byindex(mreq->ipv6mr_interface);
2877                 /*
2878                  * Put interface index into the multicast address,
2879                  * if the address has link-local scope.
2880                  */
2881                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2882                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2883                                 = htons(mreq->ipv6mr_interface);
2884                 }
2885
2886                 /*
2887                  * Find the membership in the membership list.
2888                  */
2889                 for (imm = im6o->im6o_memberships.lh_first;
2890                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2891                         if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2892                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2893                             &mreq->ipv6mr_multiaddr))
2894                                 break;
2895                 }
2896                 if (imm == NULL) {
2897                         /* Unable to resolve interface */
2898                         error = EADDRNOTAVAIL;
2899                         break;
2900                 }
2901                 /*
2902                  * Give up the multicast address record to which the
2903                  * membership points.
2904                  */
2905                 LIST_REMOVE(imm, i6mm_chain);
2906                 in6_delmulti(imm->i6mm_maddr);
2907                 free(imm, M_IPMADDR);
2908                 break;
2909
2910         default:
2911                 error = EOPNOTSUPP;
2912                 break;
2913         }
2914
2915         /*
2916          * If all options have default values, no need to keep the mbuf.
2917          */
2918         if (im6o->im6o_multicast_ifp == NULL &&
2919             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2920             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2921             im6o->im6o_memberships.lh_first == NULL) {
2922                 free(*im6op, M_IPMOPTS);
2923                 *im6op = NULL;
2924         }
2925
2926         return (error);
2927 }
2928
2929 /*
2930  * Return the IP6 multicast options in response to user getsockopt().
2931  */
2932 static int
2933 ip6_getmoptions(optname, im6o, mp)
2934         int optname;
2935         struct ip6_moptions *im6o;
2936         struct mbuf **mp;
2937 {
2938         u_int *hlim, *loop, *ifindex;
2939
2940         *mp = m_get(M_TRYWAIT, MT_HEADER);              /* XXX */
2941
2942         switch (optname) {
2943
2944         case IPV6_MULTICAST_IF:
2945                 ifindex = mtod(*mp, u_int *);
2946                 (*mp)->m_len = sizeof(u_int);
2947                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2948                         *ifindex = 0;
2949                 else
2950                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2951                 return (0);
2952
2953         case IPV6_MULTICAST_HOPS:
2954                 hlim = mtod(*mp, u_int *);
2955                 (*mp)->m_len = sizeof(u_int);
2956                 if (im6o == NULL)
2957                         *hlim = ip6_defmcasthlim;
2958                 else
2959                         *hlim = im6o->im6o_multicast_hlim;
2960                 return (0);
2961
2962         case IPV6_MULTICAST_LOOP:
2963                 loop = mtod(*mp, u_int *);
2964                 (*mp)->m_len = sizeof(u_int);
2965                 if (im6o == NULL)
2966                         *loop = ip6_defmcasthlim;
2967                 else
2968                         *loop = im6o->im6o_multicast_loop;
2969                 return (0);
2970
2971         default:
2972                 return (EOPNOTSUPP);
2973         }
2974 }
2975
2976 /*
2977  * Discard the IP6 multicast options.
2978  */
2979 void
2980 ip6_freemoptions(im6o)
2981         struct ip6_moptions *im6o;
2982 {
2983         struct in6_multi_mship *imm;
2984
2985         if (im6o == NULL)
2986                 return;
2987
2988         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2989                 LIST_REMOVE(imm, i6mm_chain);
2990                 if (imm->i6mm_maddr)
2991                         in6_delmulti(imm->i6mm_maddr);
2992                 free(imm, M_IPMADDR);
2993         }
2994         free(im6o, M_IPMOPTS);
2995 }
2996
2997 /*
2998  * Set IPv6 outgoing packet options based on advanced API.
2999  */
3000 int
3001 ip6_setpktoptions(control, opt, stickyopt, priv, needcopy, uproto)
3002         struct mbuf *control;
3003         struct ip6_pktopts *opt, *stickyopt;
3004         int priv, needcopy, uproto;
3005 {
3006         struct cmsghdr *cm = 0;
3007
3008         if (control == 0 || opt == 0)
3009                 return (EINVAL);
3010
3011         if (stickyopt) {
3012                 /*
3013                  * If stickyopt is provided, make a local copy of the options
3014                  * for this particular packet, then override them by ancillary
3015                  * objects.
3016                  * XXX: need to gain a reference for the cached route of the
3017                  * next hop in case of the overriding.
3018                  */
3019                 *opt = *stickyopt;
3020                 if (opt->ip6po_nextroute.ro_rt) {
3021                         RT_LOCK(opt->ip6po_nextroute.ro_rt);
3022                         RT_ADDREF(opt->ip6po_nextroute.ro_rt);
3023                         RT_UNLOCK(opt->ip6po_nextroute.ro_rt);
3024                 }
3025         } else
3026                 init_ip6pktopts(opt);
3027         opt->needfree = needcopy;
3028
3029         /*
3030          * XXX: Currently, we assume all the optional information is stored
3031          * in a single mbuf.
3032          */
3033         if (control->m_next)
3034                 return (EINVAL);
3035
3036         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3037             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3038                 int error;
3039
3040                 if (control->m_len < CMSG_LEN(0))
3041                         return (EINVAL);
3042
3043                 cm = mtod(control, struct cmsghdr *);
3044                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3045                         return (EINVAL);
3046                 if (cm->cmsg_level != IPPROTO_IPV6)
3047                         continue;
3048
3049                 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3050                     cm->cmsg_len - CMSG_LEN(0), opt, priv, needcopy, 1, uproto);
3051                 if (error)
3052                         return (error);
3053         }
3054
3055         return (0);
3056 }
3057
3058 /*
3059  * Set a particular packet option, as a sticky option or an ancillary data
3060  * item.  "len" can be 0 only when it's a sticky option.
3061  * We have 4 cases of combination of "sticky" and "cmsg":
3062  * "sticky=0, cmsg=0": impossible
3063  * "sticky=0, cmsg=1": RFC2292 or rfc2292bis ancillary data
3064  * "sticky=1, cmsg=0": rfc2292bis socket option
3065  * "sticky=1, cmsg=1": RFC2292 socket option
3066  */
3067 static int
3068 ip6_setpktoption(optname, buf, len, opt, priv, sticky, cmsg, uproto)
3069         int optname, len, priv, sticky, cmsg, uproto;
3070         u_char *buf;
3071         struct ip6_pktopts *opt;
3072 {
3073         int minmtupolicy, preftemp;
3074
3075         if (!sticky && !cmsg) {
3076 #ifdef DIAGNOSTIC
3077                 printf("ip6_setpktoption: impossible case\n");
3078 #endif
3079                 return (EINVAL);
3080         }
3081
3082         /*
3083          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
3084          * not be specified in the context of rfc2292bis.  Conversely,
3085          * rfc2292bis types should not be specified in the context of RFC2292.
3086          */
3087         if (!cmsg) {
3088                 switch (optname) {
3089                 case IPV6_2292PKTINFO:
3090                 case IPV6_2292HOPLIMIT:
3091                 case IPV6_2292NEXTHOP:
3092                 case IPV6_2292HOPOPTS:
3093                 case IPV6_2292DSTOPTS:
3094                 case IPV6_2292RTHDR:
3095                 case IPV6_2292PKTOPTIONS:
3096                         return (ENOPROTOOPT);
3097                 }
3098         }
3099         if (sticky && cmsg) {
3100                 switch (optname) {
3101                 case IPV6_PKTINFO:
3102                 case IPV6_HOPLIMIT:
3103                 case IPV6_NEXTHOP:
3104                 case IPV6_HOPOPTS:
3105                 case IPV6_DSTOPTS:
3106                 case IPV6_RTHDRDSTOPTS:
3107                 case IPV6_RTHDR:
3108                 case IPV6_USE_MIN_MTU:
3109                 case IPV6_DONTFRAG:
3110                 case IPV6_TCLASS:
3111                 case IPV6_PREFER_TEMPADDR: /* XXX: not an rfc2292bis option */
3112                         return (ENOPROTOOPT);
3113                 }
3114         }
3115
3116         switch (optname) {
3117         case IPV6_2292PKTINFO:
3118         case IPV6_PKTINFO:
3119         {
3120                 struct ifnet *ifp = NULL;
3121                 struct in6_pktinfo *pktinfo;
3122
3123                 if (len != sizeof(struct in6_pktinfo))
3124                         return (EINVAL);
3125
3126                 pktinfo = (struct in6_pktinfo *)buf;
3127
3128                 /*
3129                  * An application can clear any sticky IPV6_PKTINFO option by
3130                  * doing a "regular" setsockopt with ipi6_addr being
3131                  * in6addr_any and ipi6_ifindex being zero.
3132                  * [RFC 3542, Section 6]
3133                  */
3134                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
3135                     pktinfo->ipi6_ifindex == 0 &&
3136                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3137                         ip6_clearpktopts(opt, optname);
3138                         break;
3139                 }
3140
3141                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
3142                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3143                         return (EINVAL);
3144                 }
3145
3146                 /* validate the interface index if specified. */
3147                 if (pktinfo->ipi6_ifindex > if_index ||
3148                     pktinfo->ipi6_ifindex < 0) {
3149                          return (ENXIO);
3150                 }
3151                 if (pktinfo->ipi6_ifindex) {
3152                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
3153                         if (ifp == NULL)
3154                                 return (ENXIO);
3155                 }
3156
3157                 /*
3158                  * We store the address anyway, and let in6_selectsrc()
3159                  * validate the specified address.  This is because ipi6_addr
3160                  * may not have enough information about its scope zone, and
3161                  * we may need additional information (such as outgoing
3162                  * interface or the scope zone of a destination address) to
3163                  * disambiguate the scope.
3164                  * XXX: the delay of the validation may confuse the
3165                  * application when it is used as a sticky option.
3166                  */
3167                 if (sticky) {
3168                         if (opt->ip6po_pktinfo == NULL) {
3169                                 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3170                                     M_IP6OPT, M_WAITOK);
3171                         }
3172                         bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3173                 } else
3174                         opt->ip6po_pktinfo = pktinfo;
3175                 break;
3176         }
3177
3178         case IPV6_2292HOPLIMIT:
3179         case IPV6_HOPLIMIT:
3180         {
3181                 int *hlimp;
3182
3183                 /*
3184                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3185                  * to simplify the ordering among hoplimit options.
3186                  */
3187                 if (optname == IPV6_HOPLIMIT && sticky)
3188                         return (ENOPROTOOPT);
3189
3190                 if (len != sizeof(int))
3191                         return (EINVAL);
3192                 hlimp = (int *)buf;
3193                 if (*hlimp < -1 || *hlimp > 255)
3194                         return (EINVAL);
3195
3196                 opt->ip6po_hlim = *hlimp;
3197                 break;
3198         }
3199
3200         case IPV6_TCLASS:
3201         {
3202                 int tclass;
3203
3204                 if (len != sizeof(int))
3205                         return (EINVAL);
3206                 tclass = *(int *)buf;
3207                 if (tclass < -1 || tclass > 255)
3208                         return (EINVAL);
3209
3210                 opt->ip6po_tclass = tclass;
3211                 break;
3212         }
3213
3214         case IPV6_2292NEXTHOP:
3215         case IPV6_NEXTHOP:
3216                 if (!priv)
3217                         return (EPERM);
3218
3219                 if (len == 0) { /* just remove the option */
3220                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3221                         break;
3222                 }
3223
3224                 /* check if cmsg_len is large enough for sa_len */
3225                 if (len < sizeof(struct sockaddr) || len < *buf)
3226                         return (EINVAL);
3227
3228                 switch (((struct sockaddr *)buf)->sa_family) {
3229                 case AF_INET6:
3230                 {
3231                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3232 #if 0
3233                         int error;
3234 #endif
3235
3236                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3237                                 return (EINVAL);
3238
3239                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3240                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3241                                 return (EINVAL);
3242                         }
3243 #if 0
3244                         if ((error = scope6_check_id(sa6, ip6_use_defzone))
3245                             != 0) {
3246                                 return (error);
3247                         }
3248 #endif
3249                         sa6->sin6_scope_id = 0; /* XXX */
3250                         break;
3251                 }
3252                 case AF_LINK:   /* should eventually be supported */
3253                 default:
3254                         return (EAFNOSUPPORT);
3255                 }
3256
3257                 /* turn off the previous option, then set the new option. */
3258                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3259                 if (sticky) {
3260                         opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_WAITOK);
3261                         bcopy(buf, opt->ip6po_nexthop, *buf);
3262                 } else
3263                         opt->ip6po_nexthop = (struct sockaddr *)buf;
3264                 break;
3265
3266         case IPV6_2292HOPOPTS:
3267         case IPV6_HOPOPTS:
3268         {
3269                 struct ip6_hbh *hbh;
3270                 int hbhlen;
3271
3272                 /*
3273                  * XXX: We don't allow a non-privileged user to set ANY HbH
3274                  * options, since per-option restriction has too much
3275                  * overhead.
3276                  */
3277                 if (!priv)
3278                         return (EPERM);
3279
3280                 if (len == 0) {
3281                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3282                         break;  /* just remove the option */
3283                 }
3284
3285                 /* message length validation */
3286                 if (len < sizeof(struct ip6_hbh))
3287                         return (EINVAL);
3288                 hbh = (struct ip6_hbh *)buf;
3289                 hbhlen = (hbh->ip6h_len + 1) << 3;
3290                 if (len != hbhlen)
3291                         return (EINVAL);
3292
3293                 /* turn off the previous option, then set the new option. */
3294                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3295                 if (sticky) {
3296                         opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_WAITOK);
3297                         bcopy(hbh, opt->ip6po_hbh, hbhlen);
3298                 } else
3299                         opt->ip6po_hbh = hbh;
3300
3301                 break;
3302         }
3303
3304         case IPV6_2292DSTOPTS:
3305         case IPV6_DSTOPTS:
3306         case IPV6_RTHDRDSTOPTS:
3307         {
3308                 struct ip6_dest *dest, **newdest = NULL;
3309                 int destlen;
3310
3311                 if (!priv)      /* XXX: see the comment for IPV6_HOPOPTS */
3312                         return (EPERM);
3313
3314                 if (len == 0) {
3315                         ip6_clearpktopts(opt, optname);
3316                         break;  /* just remove the option */
3317                 }
3318
3319                 /* message length validation */
3320                 if (len < sizeof(struct ip6_dest))
3321                         return (EINVAL);
3322                 dest = (struct ip6_dest *)buf;
3323                 destlen = (dest->ip6d_len + 1) << 3;
3324                 if (len != destlen)
3325                         return (EINVAL);
3326
3327                 /*
3328                  * Determine the position that the destination options header
3329                  * should be inserted; before or after the routing header.
3330                  */
3331                 switch (optname) {
3332                 case IPV6_2292DSTOPTS:
3333                         /*
3334                          * The old advacned API is ambiguous on this point.
3335                          * Our approach is to determine the position based
3336                          * according to the existence of a routing header.
3337                          * Note, however, that this depends on the order of the
3338                          * extension headers in the ancillary data; the 1st
3339                          * part of the destination options header must appear
3340                          * before the routing header in the ancillary data,
3341                          * too.
3342                          * RFC2292bis solved the ambiguity by introducing
3343                          * separate ancillary data or option types.
3344                          */
3345                         if (opt->ip6po_rthdr == NULL)
3346                                 newdest = &opt->ip6po_dest1;
3347                         else
3348                                 newdest = &opt->ip6po_dest2;
3349                         break;
3350                 case IPV6_RTHDRDSTOPTS:
3351                         newdest = &opt->ip6po_dest1;
3352                         break;
3353                 case IPV6_DSTOPTS:
3354                         newdest = &opt->ip6po_dest2;
3355                         break;
3356                 }
3357
3358                 /* turn off the previous option, then set the new option. */
3359                 ip6_clearpktopts(opt, optname);
3360                 if (sticky) {
3361                         *newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
3362                         bcopy(dest, *newdest, destlen);
3363                 } else
3364                         *newdest = dest;
3365
3366                 break;
3367         }
3368
3369         case IPV6_2292RTHDR:
3370         case IPV6_RTHDR:
3371         {
3372                 struct ip6_rthdr *rth;
3373                 int rthlen;
3374
3375                 if (len == 0) {
3376                         ip6_clearpktopts(opt, IPV6_RTHDR);
3377                         break;  /* just remove the option */
3378                 }
3379
3380                 /* message length validation */
3381                 if (len < sizeof(struct ip6_rthdr))
3382                         return (EINVAL);
3383                 rth = (struct ip6_rthdr *)buf;
3384                 rthlen = (rth->ip6r_len + 1) << 3;
3385                 if (len != rthlen)
3386                         return (EINVAL);
3387
3388                 switch (rth->ip6r_type) {
3389                 case IPV6_RTHDR_TYPE_0:
3390                         if (rth->ip6r_len == 0) /* must contain one addr */
3391                                 return (EINVAL);
3392                         if (rth->ip6r_len % 2) /* length must be even */
3393                                 return (EINVAL);
3394                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3395                                 return (EINVAL);
3396                         break;
3397                 default:
3398                         return (EINVAL);        /* not supported */
3399                 }
3400
3401                 /* turn off the previous option */
3402                 ip6_clearpktopts(opt, IPV6_RTHDR);
3403                 if (sticky) {
3404                         opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_WAITOK);
3405                         bcopy(rth, opt->ip6po_rthdr, rthlen);
3406                 } else
3407                         opt->ip6po_rthdr = rth;
3408
3409                 break;
3410         }
3411
3412         case IPV6_USE_MIN_MTU:
3413                 if (len != sizeof(int))
3414                         return (EINVAL);
3415                 minmtupolicy = *(int *)buf;
3416                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3417                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3418                     minmtupolicy != IP6PO_MINMTU_ALL) {
3419                         return (EINVAL);
3420                 }
3421                 opt->ip6po_minmtu = minmtupolicy;
3422                 break;
3423
3424         case IPV6_DONTFRAG:
3425                 if (len != sizeof(int))
3426                         return (EINVAL);
3427
3428                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3429                         /*
3430                          * we ignore this option for TCP sockets.
3431                          * (rfc2292bis leaves this case unspecified.)
3432                          */
3433                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3434                 } else
3435                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3436                 break;
3437
3438         case IPV6_PREFER_TEMPADDR:
3439                 if (len != sizeof(int))
3440                         return (EINVAL);
3441                 preftemp = *(int *)buf;
3442                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3443                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3444                     preftemp != IP6PO_TEMPADDR_PREFER) {
3445                         return (EINVAL);
3446                 }
3447                 opt->ip6po_prefer_tempaddr = preftemp;
3448                 break;
3449
3450         default:
3451                 return (ENOPROTOOPT);
3452         } /* end of switch */
3453
3454         return (0);
3455 }
3456
3457 /*
3458  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3459  * packet to the input queue of a specified interface.  Note that this
3460  * calls the output routine of the loopback "driver", but with an interface
3461  * pointer that might NOT be &loif -- easier than replicating that code here.
3462  */
3463 void
3464 ip6_mloopback(ifp, m, dst)
3465         struct ifnet *ifp;
3466         struct mbuf *m;
3467         struct sockaddr_in6 *dst;
3468 {
3469         struct mbuf *copym;
3470         struct ip6_hdr *ip6;
3471
3472         copym = m_copy(m, 0, M_COPYALL);
3473         if (copym == NULL)
3474                 return;
3475
3476         /*
3477          * Make sure to deep-copy IPv6 header portion in case the data
3478          * is in an mbuf cluster, so that we can safely override the IPv6
3479          * header portion later.
3480          */
3481         if ((copym->m_flags & M_EXT) != 0 ||
3482             copym->m_len < sizeof(struct ip6_hdr)) {
3483                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3484                 if (copym == NULL)
3485                         return;
3486         }
3487
3488 #ifdef DIAGNOSTIC
3489         if (copym->m_len < sizeof(*ip6)) {
3490                 m_freem(copym);
3491                 return;
3492         }
3493 #endif
3494
3495         ip6 = mtod(copym, struct ip6_hdr *);
3496         /*
3497          * clear embedded scope identifiers if necessary.
3498          * in6_clearscope will touch the addresses only when necessary.
3499          */
3500         in6_clearscope(&ip6->ip6_src);
3501         in6_clearscope(&ip6->ip6_dst);
3502
3503         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3504 }
3505
3506 /*
3507  * Chop IPv6 header off from the payload.
3508  */
3509 static int
3510 ip6_splithdr(m, exthdrs)
3511         struct mbuf *m;
3512         struct ip6_exthdrs *exthdrs;
3513 {
3514         struct mbuf *mh;
3515         struct ip6_hdr *ip6;
3516
3517         ip6 = mtod(m, struct ip6_hdr *);
3518         if (m->m_len > sizeof(*ip6)) {
3519                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3520                 if (mh == 0) {
3521                         m_freem(m);
3522                         return ENOBUFS;
3523                 }
3524                 M_MOVE_PKTHDR(mh, m);
3525                 MH_ALIGN(mh, sizeof(*ip6));
3526                 m->m_len -= sizeof(*ip6);
3527                 m->m_data += sizeof(*ip6);
3528                 mh->m_next = m;
3529                 m = mh;
3530                 m->m_len = sizeof(*ip6);
3531                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3532         }
3533         exthdrs->ip6e_ip6 = m;
3534         return 0;
3535 }
3536
3537 /*
3538  * Compute IPv6 extension header length.
3539  */
3540 int
3541 ip6_optlen(in6p)
3542         struct in6pcb *in6p;
3543 {
3544         int len;
3545
3546         if (!in6p->in6p_outputopts)
3547                 return 0;
3548
3549         len = 0;
3550 #define elen(x) \
3551     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3552
3553         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3554         if (in6p->in6p_outputopts->ip6po_rthdr)
3555                 /* dest1 is valid with rthdr only */
3556                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3557         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3558         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3559         return len;
3560 #undef elen
3561 }