sys/netinet6/ip6_output.c

   1 /*-
   2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. Neither the name of the project nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
  30  */
  31
  32 /*-
  33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  34  *      The Regents of the University of California.  All rights reserved.
  35  *
  36  * Redistribution and use in source and binary forms, with or without
  37  * modification, are permitted provided that the following conditions
  38  * are met:
  39  * 1. Redistributions of source code must retain the above copyright
  40  *    notice, this list of conditions and the following disclaimer.
  41  * 2. Redistributions in binary form must reproduce the above copyright
  42  *    notice, this list of conditions and the following disclaimer in the
  43  *    documentation and/or other materials provided with the distribution.
  44  * 4. Neither the name of the University nor the names of its contributors
  45  *    may be used to endorse or promote products derived from this software
  46  *    without specific prior written permission.
  47  *
  48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  58  * SUCH DAMAGE.
  59  *
  60  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  61  */
  62
  63 #include <sys/cdefs.h>
  64 __FBSDID("$FreeBSD$");
  65
  66 #include "opt_inet.h"
  67 #include "opt_inet6.h"
  68 #include "opt_ipfw.h"
  69 #include "opt_ipsec.h"
  70 #include "opt_sctp.h"
  71 #include "opt_route.h"
  72
  73 #include <sys/param.h>
  74 #include <sys/kernel.h>
  75 #include <sys/malloc.h>
  76 #include <sys/mbuf.h>
  77 #include <sys/errno.h>
  78 #include <sys/priv.h>
  79 #include <sys/proc.h>
  80 #include <sys/protosw.h>
  81 #include <sys/socket.h>
  82 #include <sys/socketvar.h>
  83 #include <sys/syslog.h>
  84 #include <sys/ucred.h>
  85
  86 #include <net/if.h>
  87 #include <net/netisr.h>
  88 #include <net/route.h>
  89 #include <net/pfil.h>
  90 #include <net/vnet.h>
  91
  92 #include <netinet/in.h>
  93 #include <netinet/in_var.h>
  94 #include <netinet/ip_var.h>
  95 #include <netinet6/in6_var.h>
  96 #include <netinet/ip6.h>
  97 #include <netinet/icmp6.h>
  98 #include <netinet6/ip6_var.h>
  99 #include <netinet/in_pcb.h>
 100 #include <netinet/tcp_var.h>
 101 #include <netinet6/nd6.h>
 102
 103 #ifdef IPSEC
 104 #include <netipsec/ipsec.h>
 105 #include <netipsec/ipsec6.h>
 106 #include <netipsec/key.h>
 107 #include <netinet6/ip6_ipsec.h>
 108 #endif /* IPSEC */
 109 #ifdef SCTP
 110 #include <netinet/sctp.h>
 111 #include <netinet/sctp_crc32.h>
 112 #endif
 113
 114 #include <netinet6/ip6protosw.h>
 115 #include <netinet6/scope6_var.h>
 116
 117 #ifdef FLOWTABLE
 118 #include <net/flowtable.h>
 119 #endif
 120
 121 extern int in6_mcast_loop;
 122
 123 struct ip6_exthdrs {
 124         struct mbuf *ip6e_ip6;
 125         struct mbuf *ip6e_hbh;
 126         struct mbuf *ip6e_dest1;
 127         struct mbuf *ip6e_rthdr;
 128         struct mbuf *ip6e_dest2;
 129 };
 130
 131 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
 132                            struct ucred *, int));
 133 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
 134         struct socket *, struct sockopt *));
 135 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 136 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *,
 137         struct ucred *, int, int, int));
 138
 139 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 140 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
 141         struct ip6_frag **));
 142 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 143 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 144 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
 145         struct ifnet *, struct in6_addr *, u_long *, int *));
 146 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 147
 148
 149 /*
 150  * Make an extension header from option data.  hp is the source, and
 151  * mp is the destination.
 152  */
 153 #define MAKE_EXTHDR(hp, mp)                                             \
 154     do {                                                                \
 155         if (hp) {                                                       \
 156                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 157                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 158                     ((eh)->ip6e_len + 1) << 3);                         \
 159                 if (error)                                              \
 160                         goto freehdrs;                                  \
 161         }                                                               \
 162     } while (/*CONSTCOND*/ 0)
 163
 164 /*
 165  * Form a chain of extension headers.
 166  * m is the extension header mbuf
 167  * mp is the previous mbuf in the chain
 168  * p is the next header
 169  * i is the type of option.
 170  */
 171 #define MAKE_CHAIN(m, mp, p, i)\
 172     do {\
 173         if (m) {\
 174                 if (!hdrsplit) \
 175                         panic("assumption failed: hdr not split"); \
 176                 *mtod((m), u_char *) = *(p);\
 177                 *(p) = (i);\
 178                 p = mtod((m), u_char *);\
 179                 (m)->m_next = (mp)->m_next;\
 180                 (mp)->m_next = (m);\
 181                 (mp) = (m);\
 182         }\
 183     } while (/*CONSTCOND*/ 0)
 184
 185 /*
 186  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 187  * header (with pri, len, nxt, hlim, src, dst).
 188  * This function may modify ver and hlim only.
 189  * The mbuf chain containing the packet will be freed.
 190  * The mbuf opt, if present, will not be freed.
 191  *
 192  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 193  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 194  * which is rt_rmx.rmx_mtu.
 195  *
 196  * ifpp - XXX: just for statistics
 197  */
 198 int
 199 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 200     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 201     struct ifnet **ifpp, struct inpcb *inp)
 202 {
 203         struct ip6_hdr *ip6, *mhip6;
 204         struct ifnet *ifp, *origifp;
 205         struct mbuf *m = m0;
 206         struct mbuf *mprev = NULL;
 207         int hlen, tlen, len, off;
 208         struct route_in6 ip6route;
 209         struct rtentry *rt = NULL;
 210         struct sockaddr_in6 *dst, src_sa, dst_sa;
 211         struct in6_addr odst;
 212         int error = 0;
 213         struct in6_ifaddr *ia = NULL;
 214         u_long mtu;
 215         int alwaysfrag, dontfrag;
 216         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 217         struct ip6_exthdrs exthdrs;
 218         struct in6_addr finaldst, src0, dst0;
 219         u_int32_t zone;
 220         struct route_in6 *ro_pmtu = NULL;
 221         int flevalid = 0;
 222         int hdrsplit = 0;
 223         int needipsec = 0;
 224 #ifdef SCTP
 225         int sw_csum;
 226 #endif
 227 #ifdef IPSEC
 228         struct ipsec_output_state state;
 229         struct ip6_rthdr *rh = NULL;
 230         int needipsectun = 0;
 231         int segleft_org = 0;
 232         struct secpolicy *sp = NULL;
 233 #endif /* IPSEC */
 234 #ifdef IPFIREWALL_FORWARD
 235         struct m_tag *fwd_tag;
 236 #endif
 237
 238         ip6 = mtod(m, struct ip6_hdr *);
 239         if (ip6 == NULL) {
 240                 printf ("ip6 is NULL");
 241                 goto bad;
 242         }
 243
 244         finaldst = ip6->ip6_dst;
 245         bzero(&exthdrs, sizeof(exthdrs));
 246         if (opt) {
 247                 /* Hop-by-Hop options header */
 248                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 249                 /* Destination options header(1st part) */
 250                 if (opt->ip6po_rthdr) {
 251                         /*
 252                          * Destination options header(1st part)
 253                          * This only makes sense with a routing header.
 254                          * See Section 9.2 of RFC 3542.
 255                          * Disabling this part just for MIP6 convenience is
 256                          * a bad idea.  We need to think carefully about a
 257                          * way to make the advanced API coexist with MIP6
 258                          * options, which might automatically be inserted in
 259                          * the kernel.
 260                          */
 261                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 262                 }
 263                 /* Routing header */
 264                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 265                 /* Destination options header(2nd part) */
 266                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 267         }
 268
 269         /*
 270          * IPSec checking which handles several cases.
 271          * FAST IPSEC: We re-injected the packet.
 272          */
 273 #ifdef IPSEC
 274         switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
 275         {
 276         case 1:                 /* Bad packet */
 277                 goto freehdrs;
 278         case -1:                /* Do IPSec */
 279                 needipsec = 1;
 280         case 0:                 /* No IPSec */
 281         default:
 282                 break;
 283         }
 284 #endif /* IPSEC */
 285
 286         /*
 287          * Calculate the total length of the extension header chain.
 288          * Keep the length of the unfragmentable part for fragmentation.
 289          */
 290         optlen = 0;
 291         if (exthdrs.ip6e_hbh)
 292                 optlen += exthdrs.ip6e_hbh->m_len;
 293         if (exthdrs.ip6e_dest1)
 294                 optlen += exthdrs.ip6e_dest1->m_len;
 295         if (exthdrs.ip6e_rthdr)
 296                 optlen += exthdrs.ip6e_rthdr->m_len;
 297         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 298
 299         /* NOTE: we don't add AH/ESP length here. do that later. */
 300         if (exthdrs.ip6e_dest2)
 301                 optlen += exthdrs.ip6e_dest2->m_len;
 302
 303         /*
 304          * If we need IPsec, or there is at least one extension header,
 305          * separate IP6 header from the payload.
 306          */
 307         if ((needipsec || optlen) && !hdrsplit) {
 308                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 309                         m = NULL;
 310                         goto freehdrs;
 311                 }
 312                 m = exthdrs.ip6e_ip6;
 313                 hdrsplit++;
 314         }
 315
 316         /* adjust pointer */
 317         ip6 = mtod(m, struct ip6_hdr *);
 318
 319         /* adjust mbuf packet header length */
 320         m->m_pkthdr.len += optlen;
 321         plen = m->m_pkthdr.len - sizeof(*ip6);
 322
 323         /* If this is a jumbo payload, insert a jumbo payload option. */
 324         if (plen > IPV6_MAXPACKET) {
 325                 if (!hdrsplit) {
 326                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 327                                 m = NULL;
 328                                 goto freehdrs;
 329                         }
 330                         m = exthdrs.ip6e_ip6;
 331                         hdrsplit++;
 332                 }
 333                 /* adjust pointer */
 334                 ip6 = mtod(m, struct ip6_hdr *);
 335                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 336                         goto freehdrs;
 337                 ip6->ip6_plen = 0;
 338         } else
 339                 ip6->ip6_plen = htons(plen);
 340
 341         /*
 342          * Concatenate headers and fill in next header fields.
 343          * Here we have, on "m"
 344          *      IPv6 payload
 345          * and we insert headers accordingly.  Finally, we should be getting:
 346          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 347          *
 348          * during the header composing process, "m" points to IPv6 header.
 349          * "mprev" points to an extension header prior to esp.
 350          */
 351         u_char *nexthdrp = &ip6->ip6_nxt;
 352         mprev = m;
 353
 354         /*
 355          * we treat dest2 specially.  this makes IPsec processing
 356          * much easier.  the goal here is to make mprev point the
 357          * mbuf prior to dest2.
 358          *
 359          * result: IPv6 dest2 payload
 360          * m and mprev will point to IPv6 header.
 361          */
 362         if (exthdrs.ip6e_dest2) {
 363                 if (!hdrsplit)
 364                         panic("assumption failed: hdr not split");
 365                 exthdrs.ip6e_dest2->m_next = m->m_next;
 366                 m->m_next = exthdrs.ip6e_dest2;
 367                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 368                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 369         }
 370
 371         /*
 372          * result: IPv6 hbh dest1 rthdr dest2 payload
 373          * m will point to IPv6 header.  mprev will point to the
 374          * extension header prior to dest2 (rthdr in the above case).
 375          */
 376         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 377         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 378                    IPPROTO_DSTOPTS);
 379         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 380                    IPPROTO_ROUTING);
 381
 382 #ifdef IPSEC
 383         if (!needipsec)
 384                 goto skip_ipsec2;
 385
 386         /*
 387          * pointers after IPsec headers are not valid any more.
 388          * other pointers need a great care too.
 389          * (IPsec routines should not mangle mbufs prior to AH/ESP)
 390          */
 391         exthdrs.ip6e_dest2 = NULL;
 392
 393         if (exthdrs.ip6e_rthdr) {
 394                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 395                 segleft_org = rh->ip6r_segleft;
 396                 rh->ip6r_segleft = 0;
 397         }
 398
 399         bzero(&state, sizeof(state));
 400         state.m = m;
 401         error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 402                                     &needipsectun);
 403         m = state.m;
 404         if (error == EJUSTRETURN) {
 405                 /*
 406                  * We had a SP with a level of 'use' and no SA. We
 407                  * will just continue to process the packet without
 408                  * IPsec processing.
 409                  */
 410                 ;
 411         } else if (error) {
 412                 /* mbuf is already reclaimed in ipsec6_output_trans. */
 413                 m = NULL;
 414                 switch (error) {
 415                 case EHOSTUNREACH:
 416                 case ENETUNREACH:
 417                 case EMSGSIZE:
 418                 case ENOBUFS:
 419                 case ENOMEM:
 420                         break;
 421                 default:
 422                         printf("[%s:%d] (ipsec): error code %d\n",
 423                             __func__, __LINE__, error);
 424                         /* FALLTHROUGH */
 425                 case ENOENT:
 426                         /* don't show these error codes to the user */
 427                         error = 0;
 428                         break;
 429                 }
 430                 goto bad;
 431         } else if (!needipsectun) {
 432                 /*
 433                  * In the FAST IPSec case we have already
 434                  * re-injected the packet and it has been freed
 435                  * by the ipsec_done() function.  So, just clean
 436                  * up after ourselves.
 437                  */
 438                 m = NULL;
 439                 goto done;
 440         }
 441         if (exthdrs.ip6e_rthdr) {
 442                 /* ah6_output doesn't modify mbuf chain */
 443                 rh->ip6r_segleft = segleft_org;
 444         }
 445 skip_ipsec2:;
 446 #endif /* IPSEC */
 447
 448         /*
 449          * If there is a routing header, discard the packet.
 450          */
 451         if (exthdrs.ip6e_rthdr) {
 452                  error = EINVAL;
 453                  goto bad;
 454         }
 455
 456         /* Source address validation */
 457         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 458             (flags & IPV6_UNSPECSRC) == 0) {
 459                 error = EOPNOTSUPP;
 460                 V_ip6stat.ip6s_badscope++;
 461                 goto bad;
 462         }
 463         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 464                 error = EOPNOTSUPP;
 465                 V_ip6stat.ip6s_badscope++;
 466                 goto bad;
 467         }
 468
 469         V_ip6stat.ip6s_localout++;
 470
 471         /*
 472          * Route packet.
 473          */
 474         if (ro == 0) {
 475                 ro = &ip6route;
 476                 bzero((caddr_t)ro, sizeof(*ro));
 477         }
 478         ro_pmtu = ro;
 479         if (opt && opt->ip6po_rthdr)
 480                 ro = &opt->ip6po_route;
 481         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 482 #ifdef FLOWTABLE
 483         if (ro == &ip6route) {
 484                 struct flentry *fle;
 485
 486                 /*
 487                  * The flow table returns route entries valid for up to 30
 488                  * seconds; we rely on the remainder of ip_output() taking no
 489                  * longer than that long for the stability of ro_rt.  The
 490                  * flow ID assignment must have happened before this point.
 491                  */
 492                 if ((fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6)) != NULL) {
 493                         flow_to_route_in6(fle, ro);
 494                         if (ro->ro_rt != NULL && ro->ro_lle != NULL)
 495                                 flevalid = 1;
 496                 }
 497         }
 498 #endif
 499 again:
 500         /*
 501          * if specified, try to fill in the traffic class field.
 502          * do not override if a non-zero value is already set.
 503          * we check the diffserv field and the ecn field separately.
 504          */
 505         if (opt && opt->ip6po_tclass >= 0) {
 506                 int mask = 0;
 507
 508                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 509                         mask |= 0xfc;
 510                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 511                         mask |= 0x03;
 512                 if (mask != 0)
 513                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 514         }
 515
 516         /* fill in or override the hop limit field, if necessary. */
 517         if (opt && opt->ip6po_hlim != -1)
 518                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 519         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 520                 if (im6o != NULL)
 521                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 522                 else
 523                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 524         }
 525
 526 #ifdef IPSEC
 527         /*
 528          * We may re-inject packets into the stack here.
 529          */
 530         if (needipsec && needipsectun) {
 531                 struct ipsec_output_state state;
 532
 533                 /*
 534                  * All the extension headers will become inaccessible
 535                  * (since they can be encrypted).
 536                  * Don't panic, we need no more updates to extension headers
 537                  * on inner IPv6 packet (since they are now encapsulated).
 538                  *
 539                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 540                  */
 541                 bzero(&exthdrs, sizeof(exthdrs));
 542                 exthdrs.ip6e_ip6 = m;
 543
 544                 bzero(&state, sizeof(state));
 545                 state.m = m;
 546                 state.ro = (struct route *)ro;
 547                 state.dst = (struct sockaddr *)dst;
 548
 549                 error = ipsec6_output_tunnel(&state, sp, flags);
 550
 551                 m = state.m;
 552                 ro = (struct route_in6 *)state.ro;
 553                 dst = (struct sockaddr_in6 *)state.dst;
 554                 if (error == EJUSTRETURN) {
 555                         /*
 556                          * We had a SP with a level of 'use' and no SA. We
 557                          * will just continue to process the packet without
 558                          * IPsec processing.
 559                          */
 560                         ;
 561                 } else if (error) {
 562                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 563                         m0 = m = NULL;
 564                         m = NULL;
 565                         switch (error) {
 566                         case EHOSTUNREACH:
 567                         case ENETUNREACH:
 568                         case EMSGSIZE:
 569                         case ENOBUFS:
 570                         case ENOMEM:
 571                                 break;
 572                         default:
 573                                 printf("[%s:%d] (ipsec): error code %d\n",
 574                                     __func__, __LINE__, error);
 575                                 /* FALLTHROUGH */
 576                         case ENOENT:
 577                                 /* don't show these error codes to the user */
 578                                 error = 0;
 579                                 break;
 580                         }
 581                         goto bad;
 582                 } else {
 583                         /*
 584                          * In the FAST IPSec case we have already
 585                          * re-injected the packet and it has been freed
 586                          * by the ipsec_done() function.  So, just clean
 587                          * up after ourselves.
 588                          */
 589                         m = NULL;
 590                         goto done;
 591                 }
 592
 593                 exthdrs.ip6e_ip6 = m;
 594         }
 595 #endif /* IPSEC */
 596
 597         /* adjust pointer */
 598         ip6 = mtod(m, struct ip6_hdr *);
 599
 600         bzero(&dst_sa, sizeof(dst_sa));
 601         dst_sa.sin6_family = AF_INET6;
 602         dst_sa.sin6_len = sizeof(dst_sa);
 603         dst_sa.sin6_addr = ip6->ip6_dst;
 604         if (flevalid) {
 605                 rt = ro->ro_rt;
 606                 ifp = ro->ro_rt->rt_ifp;
 607         } else if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
 608             &ifp, &rt)) != 0) {
 609                 switch (error) {
 610                 case EHOSTUNREACH:
 611                         V_ip6stat.ip6s_noroute++;
 612                         break;
 613                 case EADDRNOTAVAIL:
 614                 default:
 615                         break; /* XXX statistics? */
 616                 }
 617                 if (ifp != NULL)
 618                         in6_ifstat_inc(ifp, ifs6_out_discard);
 619                 goto bad;
 620         }
 621         if (rt == NULL) {
 622                 /*
 623                  * If in6_selectroute() does not return a route entry,
 624                  * dst may not have been updated.
 625                  */
 626                 *dst = dst_sa;  /* XXX */
 627         }
 628
 629         /*
 630          * then rt (for unicast) and ifp must be non-NULL valid values.
 631          */
 632         if ((flags & IPV6_FORWARDING) == 0) {
 633                 /* XXX: the FORWARDING flag can be set for mrouting. */
 634                 in6_ifstat_inc(ifp, ifs6_out_request);
 635         }
 636         if (rt != NULL) {
 637                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 638                 rt->rt_use++;
 639         }
 640
 641
 642         /*
 643          * The outgoing interface must be in the zone of source and
 644          * destination addresses.
 645          */
 646         origifp = ifp;
 647
 648         src0 = ip6->ip6_src;
 649         if (in6_setscope(&src0, origifp, &zone))
 650                 goto badscope;
 651         bzero(&src_sa, sizeof(src_sa));
 652         src_sa.sin6_family = AF_INET6;
 653         src_sa.sin6_len = sizeof(src_sa);
 654         src_sa.sin6_addr = ip6->ip6_src;
 655         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 656                 goto badscope;
 657
 658         dst0 = ip6->ip6_dst;
 659         if (in6_setscope(&dst0, origifp, &zone))
 660                 goto badscope;
 661         /* re-initialize to be sure */
 662         bzero(&dst_sa, sizeof(dst_sa));
 663         dst_sa.sin6_family = AF_INET6;
 664         dst_sa.sin6_len = sizeof(dst_sa);
 665         dst_sa.sin6_addr = ip6->ip6_dst;
 666         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 667                 goto badscope;
 668         }
 669
 670         /* We should use ia_ifp to support the case of
 671          * sending packets to an address of our own.
 672          */
 673         if (ia != NULL && ia->ia_ifp)
 674                 ifp = ia->ia_ifp;
 675
 676         /* scope check is done. */
 677         goto routefound;
 678
 679   badscope:
 680         V_ip6stat.ip6s_badscope++;
 681         in6_ifstat_inc(origifp, ifs6_out_discard);
 682         if (error == 0)
 683                 error = EHOSTUNREACH; /* XXX */
 684         goto bad;
 685
 686   routefound:
 687         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 688                 if (opt && opt->ip6po_nextroute.ro_rt) {
 689                         /*
 690                          * The nexthop is explicitly specified by the
 691                          * application.  We assume the next hop is an IPv6
 692                          * address.
 693                          */
 694                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 695                 }
 696                 else if ((rt->rt_flags & RTF_GATEWAY))
 697                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 698         }
 699
 700         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 701                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 702         } else {
 703                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 704                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 705                 /*
 706                  * Confirm that the outgoing interface supports multicast.
 707                  */
 708                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 709                         V_ip6stat.ip6s_noroute++;
 710                         in6_ifstat_inc(ifp, ifs6_out_discard);
 711                         error = ENETUNREACH;
 712                         goto bad;
 713                 }
 714                 if ((im6o == NULL && in6_mcast_loop) ||
 715                     (im6o && im6o->im6o_multicast_loop)) {
 716                         /*
 717                          * Loop back multicast datagram if not expressly
 718                          * forbidden to do so, even if we have not joined
 719                          * the address; protocols will filter it later,
 720                          * thus deferring a hash lookup and lock acquisition
 721                          * at the expense of an m_copym().
 722                          */
 723                         ip6_mloopback(ifp, m, dst);
 724                 } else {
 725                         /*
 726                          * If we are acting as a multicast router, perform
 727                          * multicast forwarding as if the packet had just
 728                          * arrived on the interface to which we are about
 729                          * to send.  The multicast forwarding function
 730                          * recursively calls this function, using the
 731                          * IPV6_FORWARDING flag to prevent infinite recursion.
 732                          *
 733                          * Multicasts that are looped back by ip6_mloopback(),
 734                          * above, will be forwarded by the ip6_input() routine,
 735                          * if necessary.
 736                          */
 737                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 738                                 /*
 739                                  * XXX: ip6_mforward expects that rcvif is NULL
 740                                  * when it is called from the originating path.
 741                                  * However, it is not always the case, since
 742                                  * some versions of MGETHDR() does not
 743                                  * initialize the field.
 744                                  */
 745                                 m->m_pkthdr.rcvif = NULL;
 746                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 747                                         m_freem(m);
 748                                         goto done;
 749                                 }
 750                         }
 751                 }
 752                 /*
 753                  * Multicasts with a hoplimit of zero may be looped back,
 754                  * above, but must not be transmitted on a network.
 755                  * Also, multicasts addressed to the loopback interface
 756                  * are not sent -- the above call to ip6_mloopback() will
 757                  * loop back a copy if this host actually belongs to the
 758                  * destination group on the loopback interface.
 759                  */
 760                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 761                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 762                         m_freem(m);
 763                         goto done;
 764                 }
 765         }
 766
 767         /*
 768          * Fill the outgoing inteface to tell the upper layer
 769          * to increment per-interface statistics.
 770          */
 771         if (ifpp)
 772                 *ifpp = ifp;
 773
 774         /* Determine path MTU. */
 775         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 776             &alwaysfrag)) != 0)
 777                 goto bad;
 778
 779         /*
 780          * The caller of this function may specify to use the minimum MTU
 781          * in some cases.
 782          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 783          * setting.  The logic is a bit complicated; by default, unicast
 784          * packets will follow path MTU while multicast packets will be sent at
 785          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 786          * including unicast ones will be sent at the minimum MTU.  Multicast
 787          * packets will always be sent at the minimum MTU unless
 788          * IP6PO_MINMTU_DISABLE is explicitly specified.
 789          * See RFC 3542 for more details.
 790          */
 791         if (mtu > IPV6_MMTU) {
 792                 if ((flags & IPV6_MINMTU))
 793                         mtu = IPV6_MMTU;
 794                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 795                         mtu = IPV6_MMTU;
 796                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 797                          (opt == NULL ||
 798                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 799                         mtu = IPV6_MMTU;
 800                 }
 801         }
 802
 803         /*
 804          * clear embedded scope identifiers if necessary.
 805          * in6_clearscope will touch the addresses only when necessary.
 806          */
 807         in6_clearscope(&ip6->ip6_src);
 808         in6_clearscope(&ip6->ip6_dst);
 809
 810         /*
 811          * If the outgoing packet contains a hop-by-hop options header,
 812          * it must be examined and processed even by the source node.
 813          * (RFC 2460, section 4.)
 814          */
 815         if (exthdrs.ip6e_hbh) {
 816                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 817                 u_int32_t dummy; /* XXX unused */
 818                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 819
 820 #ifdef DIAGNOSTIC
 821                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 822                         panic("ip6e_hbh is not contiguous");
 823 #endif
 824                 /*
 825                  *  XXX: if we have to send an ICMPv6 error to the sender,
 826                  *       we need the M_LOOP flag since icmp6_error() expects
 827                  *       the IPv6 and the hop-by-hop options header are
 828                  *       contiguous unless the flag is set.
 829                  */
 830                 m->m_flags |= M_LOOP;
 831                 m->m_pkthdr.rcvif = ifp;
 832                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 833                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 834                     &dummy, &plen) < 0) {
 835                         /* m was already freed at this point */
 836                         error = EINVAL;/* better error? */
 837                         goto done;
 838                 }
 839                 m->m_flags &= ~M_LOOP; /* XXX */
 840                 m->m_pkthdr.rcvif = NULL;
 841         }
 842
 843         /* Jump over all PFIL processing if hooks are not active. */
 844         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 845                 goto passout;
 846
 847         odst = ip6->ip6_dst;
 848         /* Run through list of hooks for output packets. */
 849         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 850         if (error != 0 || m == NULL)
 851                 goto done;
 852         ip6 = mtod(m, struct ip6_hdr *);
 853
 854         /* See if destination IP address was changed by packet filter. */
 855         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 856                 m->m_flags |= M_SKIP_FIREWALL;
 857                 /* If destination is now ourself drop to ip6_input(). */
 858                 if (in6_localip(&ip6->ip6_dst)) {
 859                         m->m_flags |= M_FASTFWD_OURS;
 860                         if (m->m_pkthdr.rcvif == NULL)
 861                                 m->m_pkthdr.rcvif = V_loif;
 862                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 863                                 m->m_pkthdr.csum_flags |=
 864                                     CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 865                                 m->m_pkthdr.csum_data = 0xffff;
 866                         }
 867                         m->m_pkthdr.csum_flags |=
 868                             CSUM_IP_CHECKED | CSUM_IP_VALID;
 869 #ifdef SCTP
 870                         if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 871                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 872 #endif
 873                         error = netisr_queue(NETISR_IPV6, m);
 874                         goto done;
 875                 } else
 876                         goto again;     /* Redo the routing table lookup. */
 877         }
 878
 879 #ifdef IPFIREWALL_FORWARD
 880         /* See if local, if yes, send it to netisr. */
 881         if (m->m_flags & M_FASTFWD_OURS) {
 882                 if (m->m_pkthdr.rcvif == NULL)
 883                         m->m_pkthdr.rcvif = V_loif;
 884                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 885                         m->m_pkthdr.csum_flags |=
 886                             CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 887                         m->m_pkthdr.csum_data = 0xffff;
 888                 }
 889 #ifdef SCTP
 890                 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 891                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 892 #endif
 893                 error = netisr_queue(NETISR_IPV6, m);
 894                 goto done;
 895         }
 896         /* Or forward to some other address? */
 897         fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
 898         if (fwd_tag) {
 899                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 900                 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6));
 901                 m->m_flags |= M_SKIP_FIREWALL;
 902                 m_tag_delete(m, fwd_tag);
 903                 goto again;
 904         }
 905 #endif /* IPFIREWALL_FORWARD */
 906
 907 passout:
 908         /*
 909          * Send the packet to the outgoing interface.
 910          * If necessary, do IPv6 fragmentation before sending.
 911          *
 912          * the logic here is rather complex:
 913          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 914          * 1-a: send as is if tlen <= path mtu
 915          * 1-b: fragment if tlen > path mtu
 916          *
 917          * 2: if user asks us not to fragment (dontfrag == 1)
 918          * 2-a: send as is if tlen <= interface mtu
 919          * 2-b: error if tlen > interface mtu
 920          *
 921          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 922          *      always fragment
 923          *
 924          * 4: if dontfrag == 1 && alwaysfrag == 1
 925          *      error, as we cannot handle this conflicting request
 926          */
 927 #ifdef SCTP
 928         sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
 929         if (sw_csum & CSUM_SCTP) {
 930                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 931                 sw_csum &= ~CSUM_SCTP;
 932         }
 933 #endif
 934         tlen = m->m_pkthdr.len;
 935
 936         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
 937                 dontfrag = 1;
 938         else
 939                 dontfrag = 0;
 940         if (dontfrag && alwaysfrag) {   /* case 4 */
 941                 /* conflicting request - can't transmit */
 942                 error = EMSGSIZE;
 943                 goto bad;
 944         }
 945         if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
 946                 /*
 947                  * Even if the DONTFRAG option is specified, we cannot send the
 948                  * packet when the data length is larger than the MTU of the
 949                  * outgoing interface.
 950                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 951                  * well as returning an error code (the latter is not described
 952                  * in the API spec.)
 953                  */
 954                 u_int32_t mtu32;
 955                 struct ip6ctlparam ip6cp;
 956
 957                 mtu32 = (u_int32_t)mtu;
 958                 bzero(&ip6cp, sizeof(ip6cp));
 959                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
 960                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
 961                     (void *)&ip6cp);
 962
 963                 error = EMSGSIZE;
 964                 goto bad;
 965         }
 966
 967         /*
 968          * transmit packet without fragmentation
 969          */
 970         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 971                 struct in6_ifaddr *ia6;
 972
 973                 ip6 = mtod(m, struct ip6_hdr *);
 974                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 975                 if (ia6) {
 976                         /* Record statistics for this interface address. */
 977                         ia6->ia_ifa.if_opackets++;
 978                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
 979                         ifa_free(&ia6->ia_ifa);
 980                 }
 981                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 982                 goto done;
 983         }
 984
 985         /*
 986          * try to fragment the packet.  case 1-b and 3
 987          */
 988         if (mtu < IPV6_MMTU) {
 989                 /* path MTU cannot be less than IPV6_MMTU */
 990                 error = EMSGSIZE;
 991                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 992                 goto bad;
 993         } else if (ip6->ip6_plen == 0) {
 994                 /* jumbo payload cannot be fragmented */
 995                 error = EMSGSIZE;
 996                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 997                 goto bad;
 998         } else {
 999                 struct mbuf **mnext, *m_frgpart;
1000                 struct ip6_frag *ip6f;
1001                 u_int32_t id = htonl(ip6_randomid());
1002                 u_char nextproto;
1003
1004                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1005
1006                 /*
1007                  * Too large for the destination or interface;
1008                  * fragment if possible.
1009                  * Must be able to put at least 8 bytes per fragment.
1010                  */
1011                 hlen = unfragpartlen;
1012                 if (mtu > IPV6_MAXPACKET)
1013                         mtu = IPV6_MAXPACKET;
1014
1015                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1016                 if (len < 8) {
1017                         error = EMSGSIZE;
1018                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1019                         goto bad;
1020                 }
1021
1022                 /*
1023                  * Verify that we have any chance at all of being able to queue
1024                  *      the packet or packet fragments
1025                  */
1026                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1027                     < tlen  /* - hlen */)) {
1028                         error = ENOBUFS;
1029                         V_ip6stat.ip6s_odropped++;
1030                         goto bad;
1031                 }
1032
1033                 mnext = &m->m_nextpkt;
1034
1035                 /*
1036                  * Change the next header field of the last header in the
1037                  * unfragmentable part.
1038                  */
1039                 if (exthdrs.ip6e_rthdr) {
1040                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1041                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1042                 } else if (exthdrs.ip6e_dest1) {
1043                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1044                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1045                 } else if (exthdrs.ip6e_hbh) {
1046                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1047                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1048                 } else {
1049                         nextproto = ip6->ip6_nxt;
1050                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1051                 }
1052
1053                 /*
1054                  * Loop through length of segment after first fragment,
1055                  * make new header and copy data of each part and link onto
1056                  * chain.
1057                  */
1058                 m0 = m;
1059                 for (off = hlen; off < tlen; off += len) {
1060                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1061                         if (!m) {
1062                                 error = ENOBUFS;
1063                                 V_ip6stat.ip6s_odropped++;
1064                                 goto sendorfree;
1065                         }
1066                         m->m_pkthdr.rcvif = NULL;
1067                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1068                         *mnext = m;
1069                         mnext = &m->m_nextpkt;
1070                         m->m_data += max_linkhdr;
1071                         mhip6 = mtod(m, struct ip6_hdr *);
1072                         *mhip6 = *ip6;
1073                         m->m_len = sizeof(*mhip6);
1074                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1075                         if (error) {
1076                                 V_ip6stat.ip6s_odropped++;
1077                                 goto sendorfree;
1078                         }
1079                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1080                         if (off + len >= tlen)
1081                                 len = tlen - off;
1082                         else
1083                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1084                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1085                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1086                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1087                                 error = ENOBUFS;
1088                                 V_ip6stat.ip6s_odropped++;
1089                                 goto sendorfree;
1090                         }
1091                         m_cat(m, m_frgpart);
1092                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1093                         m->m_pkthdr.rcvif = NULL;
1094                         ip6f->ip6f_reserved = 0;
1095                         ip6f->ip6f_ident = id;
1096                         ip6f->ip6f_nxt = nextproto;
1097                         V_ip6stat.ip6s_ofragments++;
1098                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1099                 }
1100
1101                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1102         }
1103
1104         /*
1105          * Remove leading garbages.
1106          */
1107 sendorfree:
1108         m = m0->m_nextpkt;
1109         m0->m_nextpkt = 0;
1110         m_freem(m0);
1111         for (m0 = m; m; m = m0) {
1112                 m0 = m->m_nextpkt;
1113                 m->m_nextpkt = 0;
1114                 if (error == 0) {
1115                         /* Record statistics for this interface address. */
1116                         if (ia) {
1117                                 ia->ia_ifa.if_opackets++;
1118                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1119                         }
1120                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1121                 } else
1122                         m_freem(m);
1123         }
1124
1125         if (error == 0)
1126                 V_ip6stat.ip6s_fragmented++;
1127
1128 done:
1129         if (ro == &ip6route && ro->ro_rt && flevalid == 0) {
1130                 /* brace necessary for RTFREE */
1131                 RTFREE(ro->ro_rt);
1132         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt &&
1133             ((flevalid == 0) || (ro_pmtu != ro))) {
1134                 RTFREE(ro_pmtu->ro_rt);
1135         }
1136 #ifdef IPSEC
1137         if (sp != NULL)
1138                 KEY_FREESP(&sp);
1139 #endif
1140
1141         return (error);
1142
1143 freehdrs:
1144         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1145         m_freem(exthdrs.ip6e_dest1);
1146         m_freem(exthdrs.ip6e_rthdr);
1147         m_freem(exthdrs.ip6e_dest2);
1148         /* FALLTHROUGH */
1149 bad:
1150         if (m)
1151                 m_freem(m);
1152         goto done;
1153 }
1154
1155 static int
1156 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1157 {
1158         struct mbuf *m;
1159
1160         if (hlen > MCLBYTES)
1161                 return (ENOBUFS); /* XXX */
1162
1163         MGET(m, M_DONTWAIT, MT_DATA);
1164         if (!m)
1165                 return (ENOBUFS);
1166
1167         if (hlen > MLEN) {
1168                 MCLGET(m, M_DONTWAIT);
1169                 if ((m->m_flags & M_EXT) == 0) {
1170                         m_free(m);
1171                         return (ENOBUFS);
1172                 }
1173         }
1174         m->m_len = hlen;
1175         if (hdr)
1176                 bcopy(hdr, mtod(m, caddr_t), hlen);
1177
1178         *mp = m;
1179         return (0);
1180 }
1181
1182 /*
1183  * Insert jumbo payload option.
1184  */
1185 static int
1186 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1187 {
1188         struct mbuf *mopt;
1189         u_char *optbuf;
1190         u_int32_t v;
1191
1192 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1193
1194         /*
1195          * If there is no hop-by-hop options header, allocate new one.
1196          * If there is one but it doesn't have enough space to store the
1197          * jumbo payload option, allocate a cluster to store the whole options.
1198          * Otherwise, use it to store the options.
1199          */
1200         if (exthdrs->ip6e_hbh == 0) {
1201                 MGET(mopt, M_DONTWAIT, MT_DATA);
1202                 if (mopt == 0)
1203                         return (ENOBUFS);
1204                 mopt->m_len = JUMBOOPTLEN;
1205                 optbuf = mtod(mopt, u_char *);
1206                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1207                 exthdrs->ip6e_hbh = mopt;
1208         } else {
1209                 struct ip6_hbh *hbh;
1210
1211                 mopt = exthdrs->ip6e_hbh;
1212                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1213                         /*
1214                          * XXX assumption:
1215                          * - exthdrs->ip6e_hbh is not referenced from places
1216                          *   other than exthdrs.
1217                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1218                          */
1219                         int oldoptlen = mopt->m_len;
1220                         struct mbuf *n;
1221
1222                         /*
1223                          * XXX: give up if the whole (new) hbh header does
1224                          * not fit even in an mbuf cluster.
1225                          */
1226                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1227                                 return (ENOBUFS);
1228
1229                         /*
1230                          * As a consequence, we must always prepare a cluster
1231                          * at this point.
1232                          */
1233                         MGET(n, M_DONTWAIT, MT_DATA);
1234                         if (n) {
1235                                 MCLGET(n, M_DONTWAIT);
1236                                 if ((n->m_flags & M_EXT) == 0) {
1237                                         m_freem(n);
1238                                         n = NULL;
1239                                 }
1240                         }
1241                         if (!n)
1242                                 return (ENOBUFS);
1243                         n->m_len = oldoptlen + JUMBOOPTLEN;
1244                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1245                             oldoptlen);
1246                         optbuf = mtod(n, caddr_t) + oldoptlen;
1247                         m_freem(mopt);
1248                         mopt = exthdrs->ip6e_hbh = n;
1249                 } else {
1250                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1251                         mopt->m_len += JUMBOOPTLEN;
1252                 }
1253                 optbuf[0] = IP6OPT_PADN;
1254                 optbuf[1] = 1;
1255
1256                 /*
1257                  * Adjust the header length according to the pad and
1258                  * the jumbo payload option.
1259                  */
1260                 hbh = mtod(mopt, struct ip6_hbh *);
1261                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1262         }
1263
1264         /* fill in the option. */
1265         optbuf[2] = IP6OPT_JUMBO;
1266         optbuf[3] = 4;
1267         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1268         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1269
1270         /* finally, adjust the packet header length */
1271         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1272
1273         return (0);
1274 #undef JUMBOOPTLEN
1275 }
1276
1277 /*
1278  * Insert fragment header and copy unfragmentable header portions.
1279  */
1280 static int
1281 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1282     struct ip6_frag **frghdrp)
1283 {
1284         struct mbuf *n, *mlast;
1285
1286         if (hlen > sizeof(struct ip6_hdr)) {
1287                 n = m_copym(m0, sizeof(struct ip6_hdr),
1288                     hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1289                 if (n == 0)
1290                         return (ENOBUFS);
1291                 m->m_next = n;
1292         } else
1293                 n = m;
1294
1295         /* Search for the last mbuf of unfragmentable part. */
1296         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1297                 ;
1298
1299         if ((mlast->m_flags & M_EXT) == 0 &&
1300             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1301                 /* use the trailing space of the last mbuf for the fragment hdr */
1302                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1303                     mlast->m_len);
1304                 mlast->m_len += sizeof(struct ip6_frag);
1305                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1306         } else {
1307                 /* allocate a new mbuf for the fragment header */
1308                 struct mbuf *mfrg;
1309
1310                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1311                 if (mfrg == 0)
1312                         return (ENOBUFS);
1313                 mfrg->m_len = sizeof(struct ip6_frag);
1314                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1315                 mlast->m_next = mfrg;
1316         }
1317
1318         return (0);
1319 }
1320
1321 static int
1322 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1323     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1324     int *alwaysfragp)
1325 {
1326         u_int32_t mtu = 0;
1327         int alwaysfrag = 0;
1328         int error = 0;
1329
1330         if (ro_pmtu != ro) {
1331                 /* The first hop and the final destination may differ. */
1332                 struct sockaddr_in6 *sa6_dst =
1333                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1334                 if (ro_pmtu->ro_rt &&
1335                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1336                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1337                         RTFREE(ro_pmtu->ro_rt);
1338                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1339                 }
1340                 if (ro_pmtu->ro_rt == NULL) {
1341                         bzero(sa6_dst, sizeof(*sa6_dst));
1342                         sa6_dst->sin6_family = AF_INET6;
1343                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1344                         sa6_dst->sin6_addr = *dst;
1345
1346                         rtalloc((struct route *)ro_pmtu);
1347                 }
1348         }
1349         if (ro_pmtu->ro_rt) {
1350                 u_int32_t ifmtu;
1351                 struct in_conninfo inc;
1352
1353                 bzero(&inc, sizeof(inc));
1354                 inc.inc_flags |= INC_ISIPV6;
1355                 inc.inc6_faddr = *dst;
1356
1357                 if (ifp == NULL)
1358                         ifp = ro_pmtu->ro_rt->rt_ifp;
1359                 ifmtu = IN6_LINKMTU(ifp);
1360                 mtu = tcp_hc_getmtu(&inc);
1361                 if (mtu)
1362                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1363                 else
1364                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1365                 if (mtu == 0)
1366                         mtu = ifmtu;
1367                 else if (mtu < IPV6_MMTU) {
1368                         /*
1369                          * RFC2460 section 5, last paragraph:
1370                          * if we record ICMPv6 too big message with
1371                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1372                          * or smaller, with framgent header attached.
1373                          * (fragment header is needed regardless from the
1374                          * packet size, for translators to identify packets)
1375                          */
1376                         alwaysfrag = 1;
1377                         mtu = IPV6_MMTU;
1378                 } else if (mtu > ifmtu) {
1379                         /*
1380                          * The MTU on the route is larger than the MTU on
1381                          * the interface!  This shouldn't happen, unless the
1382                          * MTU of the interface has been changed after the
1383                          * interface was brought up.  Change the MTU in the
1384                          * route to match the interface MTU (as long as the
1385                          * field isn't locked).
1386                          */
1387                         mtu = ifmtu;
1388                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1389                 }
1390         } else if (ifp) {
1391                 mtu = IN6_LINKMTU(ifp);
1392         } else
1393                 error = EHOSTUNREACH; /* XXX */
1394
1395         *mtup = mtu;
1396         if (alwaysfragp)
1397                 *alwaysfragp = alwaysfrag;
1398         return (error);
1399 }
1400
1401 /*
1402  * IP6 socket option processing.
1403  */
1404 int
1405 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1406 {
1407         int optdatalen, uproto;
1408         void *optdata;
1409         struct inpcb *in6p = sotoinpcb(so);
1410         int error, optval;
1411         int level, op, optname;
1412         int optlen;
1413         struct thread *td;
1414
1415         level = sopt->sopt_level;
1416         op = sopt->sopt_dir;
1417         optname = sopt->sopt_name;
1418         optlen = sopt->sopt_valsize;
1419         td = sopt->sopt_td;
1420         error = 0;
1421         optval = 0;
1422         uproto = (int)so->so_proto->pr_protocol;
1423
1424         if (level == IPPROTO_IPV6) {
1425                 switch (op) {
1426
1427                 case SOPT_SET:
1428                         switch (optname) {
1429                         case IPV6_2292PKTOPTIONS:
1430 #ifdef IPV6_PKTOPTIONS
1431                         case IPV6_PKTOPTIONS:
1432 #endif
1433                         {
1434                                 struct mbuf *m;
1435
1436                                 error = soopt_getm(sopt, &m); /* XXX */
1437                                 if (error != 0)
1438                                         break;
1439                                 error = soopt_mcopyin(sopt, m); /* XXX */
1440                                 if (error != 0)
1441                                         break;
1442                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1443                                                     m, so, sopt);
1444                                 m_freem(m); /* XXX */
1445                                 break;
1446                         }
1447
1448                         /*
1449                          * Use of some Hop-by-Hop options or some
1450                          * Destination options, might require special
1451                          * privilege.  That is, normal applications
1452                          * (without special privilege) might be forbidden
1453                          * from setting certain options in outgoing packets,
1454                          * and might never see certain options in received
1455                          * packets. [RFC 2292 Section 6]
1456                          * KAME specific note:
1457                          *  KAME prevents non-privileged users from sending or
1458                          *  receiving ANY hbh/dst options in order to avoid
1459                          *  overhead of parsing options in the kernel.
1460                          */
1461                         case IPV6_RECVHOPOPTS:
1462                         case IPV6_RECVDSTOPTS:
1463                         case IPV6_RECVRTHDRDSTOPTS:
1464                                 if (td != NULL) {
1465                                         error = priv_check(td,
1466                                             PRIV_NETINET_SETHDROPTS);
1467                                         if (error)
1468                                                 break;
1469                                 }
1470                                 /* FALLTHROUGH */
1471                         case IPV6_UNICAST_HOPS:
1472                         case IPV6_HOPLIMIT:
1473                         case IPV6_FAITH:
1474
1475                         case IPV6_RECVPKTINFO:
1476                         case IPV6_RECVHOPLIMIT:
1477                         case IPV6_RECVRTHDR:
1478                         case IPV6_RECVPATHMTU:
1479                         case IPV6_RECVTCLASS:
1480                         case IPV6_V6ONLY:
1481                         case IPV6_AUTOFLOWLABEL:
1482                         case IPV6_BINDANY:
1483                                 if (optname == IPV6_BINDANY && td != NULL) {
1484                                         error = priv_check(td,
1485                                             PRIV_NETINET_BINDANY);
1486                                         if (error)
1487                                                 break;
1488                                 }
1489
1490                                 if (optlen != sizeof(int)) {
1491                                         error = EINVAL;
1492                                         break;
1493                                 }
1494                                 error = sooptcopyin(sopt, &optval,
1495                                         sizeof optval, sizeof optval);
1496                                 if (error)
1497                                         break;
1498                                 switch (optname) {
1499
1500                                 case IPV6_UNICAST_HOPS:
1501                                         if (optval < -1 || optval >= 256)
1502                                                 error = EINVAL;
1503                                         else {
1504                                                 /* -1 = kernel default */
1505                                                 in6p->in6p_hops = optval;
1506                                                 if ((in6p->inp_vflag &
1507                                                      INP_IPV4) != 0)
1508                                                         in6p->inp_ip_ttl = optval;
1509                                         }
1510                                         break;
1511 #define OPTSET(bit) \
1512 do { \
1513         if (optval) \
1514                 in6p->inp_flags |= (bit); \
1515         else \
1516                 in6p->inp_flags &= ~(bit); \
1517 } while (/*CONSTCOND*/ 0)
1518 #define OPTSET2292(bit) \
1519 do { \
1520         in6p->inp_flags |= IN6P_RFC2292; \
1521         if (optval) \
1522                 in6p->inp_flags |= (bit); \
1523         else \
1524                 in6p->inp_flags &= ~(bit); \
1525 } while (/*CONSTCOND*/ 0)
1526 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1527
1528                                 case IPV6_RECVPKTINFO:
1529                                         /* cannot mix with RFC2292 */
1530                                         if (OPTBIT(IN6P_RFC2292)) {
1531                                                 error = EINVAL;
1532                                                 break;
1533                                         }
1534                                         OPTSET(IN6P_PKTINFO);
1535                                         break;
1536
1537                                 case IPV6_HOPLIMIT:
1538                                 {
1539                                         struct ip6_pktopts **optp;
1540
1541                                         /* cannot mix with RFC2292 */
1542                                         if (OPTBIT(IN6P_RFC2292)) {
1543                                                 error = EINVAL;
1544                                                 break;
1545                                         }
1546                                         optp = &in6p->in6p_outputopts;
1547                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1548                                             (u_char *)&optval, sizeof(optval),
1549                                             optp, (td != NULL) ? td->td_ucred :
1550                                             NULL, uproto);
1551                                         break;
1552                                 }
1553
1554                                 case IPV6_RECVHOPLIMIT:
1555                                         /* cannot mix with RFC2292 */
1556                                         if (OPTBIT(IN6P_RFC2292)) {
1557                                                 error = EINVAL;
1558                                                 break;
1559                                         }
1560                                         OPTSET(IN6P_HOPLIMIT);
1561                                         break;
1562
1563                                 case IPV6_RECVHOPOPTS:
1564                                         /* cannot mix with RFC2292 */
1565                                         if (OPTBIT(IN6P_RFC2292)) {
1566                                                 error = EINVAL;
1567                                                 break;
1568                                         }
1569                                         OPTSET(IN6P_HOPOPTS);
1570                                         break;
1571
1572                                 case IPV6_RECVDSTOPTS:
1573                                         /* cannot mix with RFC2292 */
1574                                         if (OPTBIT(IN6P_RFC2292)) {
1575                                                 error = EINVAL;
1576                                                 break;
1577                                         }
1578                                         OPTSET(IN6P_DSTOPTS);
1579                                         break;
1580
1581                                 case IPV6_RECVRTHDRDSTOPTS:
1582                                         /* cannot mix with RFC2292 */
1583                                         if (OPTBIT(IN6P_RFC2292)) {
1584                                                 error = EINVAL;
1585                                                 break;
1586                                         }
1587                                         OPTSET(IN6P_RTHDRDSTOPTS);
1588                                         break;
1589
1590                                 case IPV6_RECVRTHDR:
1591                                         /* cannot mix with RFC2292 */
1592                                         if (OPTBIT(IN6P_RFC2292)) {
1593                                                 error = EINVAL;
1594                                                 break;
1595                                         }
1596                                         OPTSET(IN6P_RTHDR);
1597                                         break;
1598
1599                                 case IPV6_FAITH:
1600                                         OPTSET(INP_FAITH);
1601                                         break;
1602
1603                                 case IPV6_RECVPATHMTU:
1604                                         /*
1605                                          * We ignore this option for TCP
1606                                          * sockets.
1607                                          * (RFC3542 leaves this case
1608                                          * unspecified.)
1609                                          */
1610                                         if (uproto != IPPROTO_TCP)
1611                                                 OPTSET(IN6P_MTU);
1612                                         break;
1613
1614                                 case IPV6_V6ONLY:
1615                                         /*
1616                                          * make setsockopt(IPV6_V6ONLY)
1617                                          * available only prior to bind(2).
1618                                          * see ipng mailing list, Jun 22 2001.
1619                                          */
1620                                         if (in6p->inp_lport ||
1621                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1622                                                 error = EINVAL;
1623                                                 break;
1624                                         }
1625                                         OPTSET(IN6P_IPV6_V6ONLY);
1626                                         if (optval)
1627                                                 in6p->inp_vflag &= ~INP_IPV4;
1628                                         else
1629                                                 in6p->inp_vflag |= INP_IPV4;
1630                                         break;
1631                                 case IPV6_RECVTCLASS:
1632                                         /* cannot mix with RFC2292 XXX */
1633                                         if (OPTBIT(IN6P_RFC2292)) {
1634                                                 error = EINVAL;
1635                                                 break;
1636                                         }
1637                                         OPTSET(IN6P_TCLASS);
1638                                         break;
1639                                 case IPV6_AUTOFLOWLABEL:
1640                                         OPTSET(IN6P_AUTOFLOWLABEL);
1641                                         break;
1642
1643                                 case IPV6_BINDANY:
1644                                         OPTSET(INP_BINDANY);
1645                                         break;
1646                                 }
1647                                 break;
1648
1649                         case IPV6_TCLASS:
1650                         case IPV6_DONTFRAG:
1651                         case IPV6_USE_MIN_MTU:
1652                         case IPV6_PREFER_TEMPADDR:
1653                                 if (optlen != sizeof(optval)) {
1654                                         error = EINVAL;
1655                                         break;
1656                                 }
1657                                 error = sooptcopyin(sopt, &optval,
1658                                         sizeof optval, sizeof optval);
1659                                 if (error)
1660                                         break;
1661                                 {
1662                                         struct ip6_pktopts **optp;
1663                                         optp = &in6p->in6p_outputopts;
1664                                         error = ip6_pcbopt(optname,
1665                                             (u_char *)&optval, sizeof(optval),
1666                                             optp, (td != NULL) ? td->td_ucred :
1667                                             NULL, uproto);
1668                                         break;
1669                                 }
1670
1671                         case IPV6_2292PKTINFO:
1672                         case IPV6_2292HOPLIMIT:
1673                         case IPV6_2292HOPOPTS:
1674                         case IPV6_2292DSTOPTS:
1675                         case IPV6_2292RTHDR:
1676                                 /* RFC 2292 */
1677                                 if (optlen != sizeof(int)) {
1678                                         error = EINVAL;
1679                                         break;
1680                                 }
1681                                 error = sooptcopyin(sopt, &optval,
1682                                         sizeof optval, sizeof optval);
1683                                 if (error)
1684                                         break;
1685                                 switch (optname) {
1686                                 case IPV6_2292PKTINFO:
1687                                         OPTSET2292(IN6P_PKTINFO);
1688                                         break;
1689                                 case IPV6_2292HOPLIMIT:
1690                                         OPTSET2292(IN6P_HOPLIMIT);
1691                                         break;
1692                                 case IPV6_2292HOPOPTS:
1693                                         /*
1694                                          * Check super-user privilege.
1695                                          * See comments for IPV6_RECVHOPOPTS.
1696                                          */
1697                                         if (td != NULL) {
1698                                                 error = priv_check(td,
1699                                                     PRIV_NETINET_SETHDROPTS);
1700                                                 if (error)
1701                                                         return (error);
1702                                         }
1703                                         OPTSET2292(IN6P_HOPOPTS);
1704                                         break;
1705                                 case IPV6_2292DSTOPTS:
1706                                         if (td != NULL) {
1707                                                 error = priv_check(td,
1708                                                     PRIV_NETINET_SETHDROPTS);
1709                                                 if (error)
1710                                                         return (error);
1711                                         }
1712                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1713                                         break;
1714                                 case IPV6_2292RTHDR:
1715                                         OPTSET2292(IN6P_RTHDR);
1716                                         break;
1717                                 }
1718                                 break;
1719                         case IPV6_PKTINFO:
1720                         case IPV6_HOPOPTS:
1721                         case IPV6_RTHDR:
1722                         case IPV6_DSTOPTS:
1723                         case IPV6_RTHDRDSTOPTS:
1724                         case IPV6_NEXTHOP:
1725                         {
1726                                 /* new advanced API (RFC3542) */
1727                                 u_char *optbuf;
1728                                 u_char optbuf_storage[MCLBYTES];
1729                                 int optlen;
1730                                 struct ip6_pktopts **optp;
1731
1732                                 /* cannot mix with RFC2292 */
1733                                 if (OPTBIT(IN6P_RFC2292)) {
1734                                         error = EINVAL;
1735                                         break;
1736                                 }
1737
1738                                 /*
1739                                  * We only ensure valsize is not too large
1740                                  * here.  Further validation will be done
1741                                  * later.
1742                                  */
1743                                 error = sooptcopyin(sopt, optbuf_storage,
1744                                     sizeof(optbuf_storage), 0);
1745                                 if (error)
1746                                         break;
1747                                 optlen = sopt->sopt_valsize;
1748                                 optbuf = optbuf_storage;
1749                                 optp = &in6p->in6p_outputopts;
1750                                 error = ip6_pcbopt(optname, optbuf, optlen,
1751                                     optp, (td != NULL) ? td->td_ucred : NULL,
1752                                     uproto);
1753                                 break;
1754                         }
1755 #undef OPTSET
1756
1757                         case IPV6_MULTICAST_IF:
1758                         case IPV6_MULTICAST_HOPS:
1759                         case IPV6_MULTICAST_LOOP:
1760                         case IPV6_JOIN_GROUP:
1761                         case IPV6_LEAVE_GROUP:
1762                         case IPV6_MSFILTER:
1763                         case MCAST_BLOCK_SOURCE:
1764                         case MCAST_UNBLOCK_SOURCE:
1765                         case MCAST_JOIN_GROUP:
1766                         case MCAST_LEAVE_GROUP:
1767                         case MCAST_JOIN_SOURCE_GROUP:
1768                         case MCAST_LEAVE_SOURCE_GROUP:
1769                                 error = ip6_setmoptions(in6p, sopt);
1770                                 break;
1771
1772                         case IPV6_PORTRANGE:
1773                                 error = sooptcopyin(sopt, &optval,
1774                                     sizeof optval, sizeof optval);
1775                                 if (error)
1776                                         break;
1777
1778                                 switch (optval) {
1779                                 case IPV6_PORTRANGE_DEFAULT:
1780                                         in6p->inp_flags &= ~(INP_LOWPORT);
1781                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1782                                         break;
1783
1784                                 case IPV6_PORTRANGE_HIGH:
1785                                         in6p->inp_flags &= ~(INP_LOWPORT);
1786                                         in6p->inp_flags |= INP_HIGHPORT;
1787                                         break;
1788
1789                                 case IPV6_PORTRANGE_LOW:
1790                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1791                                         in6p->inp_flags |= INP_LOWPORT;
1792                                         break;
1793
1794                                 default:
1795                                         error = EINVAL;
1796                                         break;
1797                                 }
1798                                 break;
1799
1800 #ifdef IPSEC
1801                         case IPV6_IPSEC_POLICY:
1802                         {
1803                                 caddr_t req;
1804                                 struct mbuf *m;
1805
1806                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1807                                         break;
1808                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1809                                         break;
1810                                 req = mtod(m, caddr_t);
1811                                 error = ipsec_set_policy(in6p, optname, req,
1812                                     m->m_len, (sopt->sopt_td != NULL) ?
1813                                     sopt->sopt_td->td_ucred : NULL);
1814                                 m_freem(m);
1815                                 break;
1816                         }
1817 #endif /* IPSEC */
1818
1819                         default:
1820                                 error = ENOPROTOOPT;
1821                                 break;
1822                         }
1823                         break;
1824
1825                 case SOPT_GET:
1826                         switch (optname) {
1827
1828                         case IPV6_2292PKTOPTIONS:
1829 #ifdef IPV6_PKTOPTIONS
1830                         case IPV6_PKTOPTIONS:
1831 #endif
1832                                 /*
1833                                  * RFC3542 (effectively) deprecated the
1834                                  * semantics of the 2292-style pktoptions.
1835                                  * Since it was not reliable in nature (i.e.,
1836                                  * applications had to expect the lack of some
1837                                  * information after all), it would make sense
1838                                  * to simplify this part by always returning
1839                                  * empty data.
1840                                  */
1841                                 sopt->sopt_valsize = 0;
1842                                 break;
1843
1844                         case IPV6_RECVHOPOPTS:
1845                         case IPV6_RECVDSTOPTS:
1846                         case IPV6_RECVRTHDRDSTOPTS:
1847                         case IPV6_UNICAST_HOPS:
1848                         case IPV6_RECVPKTINFO:
1849                         case IPV6_RECVHOPLIMIT:
1850                         case IPV6_RECVRTHDR:
1851                         case IPV6_RECVPATHMTU:
1852
1853                         case IPV6_FAITH:
1854                         case IPV6_V6ONLY:
1855                         case IPV6_PORTRANGE:
1856                         case IPV6_RECVTCLASS:
1857                         case IPV6_AUTOFLOWLABEL:
1858                         case IPV6_BINDANY:
1859                                 switch (optname) {
1860
1861                                 case IPV6_RECVHOPOPTS:
1862                                         optval = OPTBIT(IN6P_HOPOPTS);
1863                                         break;
1864
1865                                 case IPV6_RECVDSTOPTS:
1866                                         optval = OPTBIT(IN6P_DSTOPTS);
1867                                         break;
1868
1869                                 case IPV6_RECVRTHDRDSTOPTS:
1870                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1871                                         break;
1872
1873                                 case IPV6_UNICAST_HOPS:
1874                                         optval = in6p->in6p_hops;
1875                                         break;
1876
1877                                 case IPV6_RECVPKTINFO:
1878                                         optval = OPTBIT(IN6P_PKTINFO);
1879                                         break;
1880
1881                                 case IPV6_RECVHOPLIMIT:
1882                                         optval = OPTBIT(IN6P_HOPLIMIT);
1883                                         break;
1884
1885                                 case IPV6_RECVRTHDR:
1886                                         optval = OPTBIT(IN6P_RTHDR);
1887                                         break;
1888
1889                                 case IPV6_RECVPATHMTU:
1890                                         optval = OPTBIT(IN6P_MTU);
1891                                         break;
1892
1893                                 case IPV6_FAITH:
1894                                         optval = OPTBIT(INP_FAITH);
1895                                         break;
1896
1897                                 case IPV6_V6ONLY:
1898                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1899                                         break;
1900
1901                                 case IPV6_PORTRANGE:
1902                                     {
1903                                         int flags;
1904                                         flags = in6p->inp_flags;
1905                                         if (flags & INP_HIGHPORT)
1906                                                 optval = IPV6_PORTRANGE_HIGH;
1907                                         else if (flags & INP_LOWPORT)
1908                                                 optval = IPV6_PORTRANGE_LOW;
1909                                         else
1910                                                 optval = 0;
1911                                         break;
1912                                     }
1913                                 case IPV6_RECVTCLASS:
1914                                         optval = OPTBIT(IN6P_TCLASS);
1915                                         break;
1916
1917                                 case IPV6_AUTOFLOWLABEL:
1918                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1919                                         break;
1920
1921                                 case IPV6_BINDANY:
1922                                         optval = OPTBIT(INP_BINDANY);
1923                                         break;
1924                                 }
1925                                 if (error)
1926                                         break;
1927                                 error = sooptcopyout(sopt, &optval,
1928                                         sizeof optval);
1929                                 break;
1930
1931                         case IPV6_PATHMTU:
1932                         {
1933                                 u_long pmtu = 0;
1934                                 struct ip6_mtuinfo mtuinfo;
1935                                 struct route_in6 sro;
1936
1937                                 bzero(&sro, sizeof(sro));
1938
1939                                 if (!(so->so_state & SS_ISCONNECTED))
1940                                         return (ENOTCONN);
1941                                 /*
1942                                  * XXX: we dot not consider the case of source
1943                                  * routing, or optional information to specify
1944                                  * the outgoing interface.
1945                                  */
1946                                 error = ip6_getpmtu(&sro, NULL, NULL,
1947                                     &in6p->in6p_faddr, &pmtu, NULL);
1948                                 if (sro.ro_rt)
1949                                         RTFREE(sro.ro_rt);
1950                                 if (error)
1951                                         break;
1952                                 if (pmtu > IPV6_MAXPACKET)
1953                                         pmtu = IPV6_MAXPACKET;
1954
1955                                 bzero(&mtuinfo, sizeof(mtuinfo));
1956                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1957                                 optdata = (void *)&mtuinfo;
1958                                 optdatalen = sizeof(mtuinfo);
1959                                 error = sooptcopyout(sopt, optdata,
1960                                     optdatalen);
1961                                 break;
1962                         }
1963
1964                         case IPV6_2292PKTINFO:
1965                         case IPV6_2292HOPLIMIT:
1966                         case IPV6_2292HOPOPTS:
1967                         case IPV6_2292RTHDR:
1968                         case IPV6_2292DSTOPTS:
1969                                 switch (optname) {
1970                                 case IPV6_2292PKTINFO:
1971                                         optval = OPTBIT(IN6P_PKTINFO);
1972                                         break;
1973                                 case IPV6_2292HOPLIMIT:
1974                                         optval = OPTBIT(IN6P_HOPLIMIT);
1975                                         break;
1976                                 case IPV6_2292HOPOPTS:
1977                                         optval = OPTBIT(IN6P_HOPOPTS);
1978                                         break;
1979                                 case IPV6_2292RTHDR:
1980                                         optval = OPTBIT(IN6P_RTHDR);
1981                                         break;
1982                                 case IPV6_2292DSTOPTS:
1983                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1984                                         break;
1985                                 }
1986                                 error = sooptcopyout(sopt, &optval,
1987                                     sizeof optval);
1988                                 break;
1989                         case IPV6_PKTINFO:
1990                         case IPV6_HOPOPTS:
1991                         case IPV6_RTHDR:
1992                         case IPV6_DSTOPTS:
1993                         case IPV6_RTHDRDSTOPTS:
1994                         case IPV6_NEXTHOP:
1995                         case IPV6_TCLASS:
1996                         case IPV6_DONTFRAG:
1997                         case IPV6_USE_MIN_MTU:
1998                         case IPV6_PREFER_TEMPADDR:
1999                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2000                                     optname, sopt);
2001                                 break;
2002
2003                         case IPV6_MULTICAST_IF:
2004                         case IPV6_MULTICAST_HOPS:
2005                         case IPV6_MULTICAST_LOOP:
2006                         case IPV6_MSFILTER:
2007                                 error = ip6_getmoptions(in6p, sopt);
2008                                 break;
2009
2010 #ifdef IPSEC
2011                         case IPV6_IPSEC_POLICY:
2012                           {
2013                                 caddr_t req = NULL;
2014                                 size_t len = 0;
2015                                 struct mbuf *m = NULL;
2016                                 struct mbuf **mp = &m;
2017                                 size_t ovalsize = sopt->sopt_valsize;
2018                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2019
2020                                 error = soopt_getm(sopt, &m); /* XXX */
2021                                 if (error != 0)
2022                                         break;
2023                                 error = soopt_mcopyin(sopt, m); /* XXX */
2024                                 if (error != 0)
2025                                         break;
2026                                 sopt->sopt_valsize = ovalsize;
2027                                 sopt->sopt_val = oval;
2028                                 if (m) {
2029                                         req = mtod(m, caddr_t);
2030                                         len = m->m_len;
2031                                 }
2032                                 error = ipsec_get_policy(in6p, req, len, mp);
2033                                 if (error == 0)
2034                                         error = soopt_mcopyout(sopt, m); /* XXX */
2035                                 if (error == 0 && m)
2036                                         m_freem(m);
2037                                 break;
2038                           }
2039 #endif /* IPSEC */
2040
2041                         default:
2042                                 error = ENOPROTOOPT;
2043                                 break;
2044                         }
2045                         break;
2046                 }
2047         } else {                /* level != IPPROTO_IPV6 */
2048                 error = EINVAL;
2049         }
2050         return (error);
2051 }
2052
2053 int
2054 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2055 {
2056         int error = 0, optval, optlen;
2057         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2058         struct inpcb *in6p = sotoinpcb(so);
2059         int level, op, optname;
2060
2061         level = sopt->sopt_level;
2062         op = sopt->sopt_dir;
2063         optname = sopt->sopt_name;
2064         optlen = sopt->sopt_valsize;
2065
2066         if (level != IPPROTO_IPV6) {
2067                 return (EINVAL);
2068         }
2069
2070         switch (optname) {
2071         case IPV6_CHECKSUM:
2072                 /*
2073                  * For ICMPv6 sockets, no modification allowed for checksum
2074                  * offset, permit "no change" values to help existing apps.
2075                  *
2076                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2077                  * for an ICMPv6 socket will fail."
2078                  * The current behavior does not meet RFC3542.
2079                  */
2080                 switch (op) {
2081                 case SOPT_SET:
2082                         if (optlen != sizeof(int)) {
2083                                 error = EINVAL;
2084                                 break;
2085                         }
2086                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2087                                             sizeof(optval));
2088                         if (error)
2089                                 break;
2090                         if ((optval % 2) != 0) {
2091                                 /* the API assumes even offset values */
2092                                 error = EINVAL;
2093                         } else if (so->so_proto->pr_protocol ==
2094                             IPPROTO_ICMPV6) {
2095                                 if (optval != icmp6off)
2096                                         error = EINVAL;
2097                         } else
2098                                 in6p->in6p_cksum = optval;
2099                         break;
2100
2101                 case SOPT_GET:
2102                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2103                                 optval = icmp6off;
2104                         else
2105                                 optval = in6p->in6p_cksum;
2106
2107                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2108                         break;
2109
2110                 default:
2111                         error = EINVAL;
2112                         break;
2113                 }
2114                 break;
2115
2116         default:
2117                 error = ENOPROTOOPT;
2118                 break;
2119         }
2120
2121         return (error);
2122 }
2123
2124 /*
2125  * Set up IP6 options in pcb for insertion in output packets or
2126  * specifying behavior of outgoing packets.
2127  */
2128 static int
2129 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2130     struct socket *so, struct sockopt *sopt)
2131 {
2132         struct ip6_pktopts *opt = *pktopt;
2133         int error = 0;
2134         struct thread *td = sopt->sopt_td;
2135
2136         /* turn off any old options. */
2137         if (opt) {
2138 #ifdef DIAGNOSTIC
2139                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2140                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2141                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2142                         printf("ip6_pcbopts: all specified options are cleared.\n");
2143 #endif
2144                 ip6_clearpktopts(opt, -1);
2145         } else
2146                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2147         *pktopt = NULL;
2148
2149         if (!m || m->m_len == 0) {
2150                 /*
2151                  * Only turning off any previous options, regardless of
2152                  * whether the opt is just created or given.
2153                  */
2154                 free(opt, M_IP6OPT);
2155                 return (0);
2156         }
2157
2158         /*  set options specified by user. */
2159         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2160             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2161                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2162                 free(opt, M_IP6OPT);
2163                 return (error);
2164         }
2165         *pktopt = opt;
2166         return (0);
2167 }
2168
2169 /*
2170  * initialize ip6_pktopts.  beware that there are non-zero default values in
2171  * the struct.
2172  */
2173 void
2174 ip6_initpktopts(struct ip6_pktopts *opt)
2175 {
2176
2177         bzero(opt, sizeof(*opt));
2178         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2179         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2180         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2181         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2182 }
2183
2184 static int
2185 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2186     struct ucred *cred, int uproto)
2187 {
2188         struct ip6_pktopts *opt;
2189
2190         if (*pktopt == NULL) {
2191                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2192                     M_WAITOK);
2193                 ip6_initpktopts(*pktopt);
2194         }
2195         opt = *pktopt;
2196
2197         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2198 }
2199
2200 static int
2201 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2202 {
2203         void *optdata = NULL;
2204         int optdatalen = 0;
2205         struct ip6_ext *ip6e;
2206         int error = 0;
2207         struct in6_pktinfo null_pktinfo;
2208         int deftclass = 0, on;
2209         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2210         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2211
2212         switch (optname) {
2213         case IPV6_PKTINFO:
2214                 if (pktopt && pktopt->ip6po_pktinfo)
2215                         optdata = (void *)pktopt->ip6po_pktinfo;
2216                 else {
2217                         /* XXX: we don't have to do this every time... */
2218                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2219                         optdata = (void *)&null_pktinfo;
2220                 }
2221                 optdatalen = sizeof(struct in6_pktinfo);
2222                 break;
2223         case IPV6_TCLASS:
2224                 if (pktopt && pktopt->ip6po_tclass >= 0)
2225                         optdata = (void *)&pktopt->ip6po_tclass;
2226                 else
2227                         optdata = (void *)&deftclass;
2228                 optdatalen = sizeof(int);
2229                 break;
2230         case IPV6_HOPOPTS:
2231                 if (pktopt && pktopt->ip6po_hbh) {
2232                         optdata = (void *)pktopt->ip6po_hbh;
2233                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2234                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2235                 }
2236                 break;
2237         case IPV6_RTHDR:
2238                 if (pktopt && pktopt->ip6po_rthdr) {
2239                         optdata = (void *)pktopt->ip6po_rthdr;
2240                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2241                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2242                 }
2243                 break;
2244         case IPV6_RTHDRDSTOPTS:
2245                 if (pktopt && pktopt->ip6po_dest1) {
2246                         optdata = (void *)pktopt->ip6po_dest1;
2247                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2248                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2249                 }
2250                 break;
2251         case IPV6_DSTOPTS:
2252                 if (pktopt && pktopt->ip6po_dest2) {
2253                         optdata = (void *)pktopt->ip6po_dest2;
2254                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2255                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2256                 }
2257                 break;
2258         case IPV6_NEXTHOP:
2259                 if (pktopt && pktopt->ip6po_nexthop) {
2260                         optdata = (void *)pktopt->ip6po_nexthop;
2261                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2262                 }
2263                 break;
2264         case IPV6_USE_MIN_MTU:
2265                 if (pktopt)
2266                         optdata = (void *)&pktopt->ip6po_minmtu;
2267                 else
2268                         optdata = (void *)&defminmtu;
2269                 optdatalen = sizeof(int);
2270                 break;
2271         case IPV6_DONTFRAG:
2272                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2273                         on = 1;
2274                 else
2275                         on = 0;
2276                 optdata = (void *)&on;
2277                 optdatalen = sizeof(on);
2278                 break;
2279         case IPV6_PREFER_TEMPADDR:
2280                 if (pktopt)
2281                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2282                 else
2283                         optdata = (void *)&defpreftemp;
2284                 optdatalen = sizeof(int);
2285                 break;
2286         default:                /* should not happen */
2287 #ifdef DIAGNOSTIC
2288                 panic("ip6_getpcbopt: unexpected option\n");
2289 #endif
2290                 return (ENOPROTOOPT);
2291         }
2292
2293         error = sooptcopyout(sopt, optdata, optdatalen);
2294
2295         return (error);
2296 }
2297
2298 void
2299 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2300 {
2301         if (pktopt == NULL)
2302                 return;
2303
2304         if (optname == -1 || optname == IPV6_PKTINFO) {
2305                 if (pktopt->ip6po_pktinfo)
2306                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2307                 pktopt->ip6po_pktinfo = NULL;
2308         }
2309         if (optname == -1 || optname == IPV6_HOPLIMIT)
2310                 pktopt->ip6po_hlim = -1;
2311         if (optname == -1 || optname == IPV6_TCLASS)
2312                 pktopt->ip6po_tclass = -1;
2313         if (optname == -1 || optname == IPV6_NEXTHOP) {
2314                 if (pktopt->ip6po_nextroute.ro_rt) {
2315                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2316                         pktopt->ip6po_nextroute.ro_rt = NULL;
2317                 }
2318                 if (pktopt->ip6po_nexthop)
2319                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2320                 pktopt->ip6po_nexthop = NULL;
2321         }
2322         if (optname == -1 || optname == IPV6_HOPOPTS) {
2323                 if (pktopt->ip6po_hbh)
2324                         free(pktopt->ip6po_hbh, M_IP6OPT);
2325                 pktopt->ip6po_hbh = NULL;
2326         }
2327         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2328                 if (pktopt->ip6po_dest1)
2329                         free(pktopt->ip6po_dest1, M_IP6OPT);
2330                 pktopt->ip6po_dest1 = NULL;
2331         }
2332         if (optname == -1 || optname == IPV6_RTHDR) {
2333                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2334                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2335                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2336                 if (pktopt->ip6po_route.ro_rt) {
2337                         RTFREE(pktopt->ip6po_route.ro_rt);
2338                         pktopt->ip6po_route.ro_rt = NULL;
2339                 }
2340         }
2341         if (optname == -1 || optname == IPV6_DSTOPTS) {
2342                 if (pktopt->ip6po_dest2)
2343                         free(pktopt->ip6po_dest2, M_IP6OPT);
2344                 pktopt->ip6po_dest2 = NULL;
2345         }
2346 }
2347
2348 #define PKTOPT_EXTHDRCPY(type) \
2349 do {\
2350         if (src->type) {\
2351                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2352                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2353                 if (dst->type == NULL && canwait == M_NOWAIT)\
2354                         goto bad;\
2355                 bcopy(src->type, dst->type, hlen);\
2356         }\
2357 } while (/*CONSTCOND*/ 0)
2358
2359 static int
2360 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2361 {
2362         if (dst == NULL || src == NULL)  {
2363                 printf("ip6_clearpktopts: invalid argument\n");
2364                 return (EINVAL);
2365         }
2366
2367         dst->ip6po_hlim = src->ip6po_hlim;
2368         dst->ip6po_tclass = src->ip6po_tclass;
2369         dst->ip6po_flags = src->ip6po_flags;
2370         dst->ip6po_minmtu = src->ip6po_minmtu;
2371         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2372         if (src->ip6po_pktinfo) {
2373                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2374                     M_IP6OPT, canwait);
2375                 if (dst->ip6po_pktinfo == NULL)
2376                         goto bad;
2377                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2378         }
2379         if (src->ip6po_nexthop) {
2380                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2381                     M_IP6OPT, canwait);
2382                 if (dst->ip6po_nexthop == NULL)
2383                         goto bad;
2384                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2385                     src->ip6po_nexthop->sa_len);
2386         }
2387         PKTOPT_EXTHDRCPY(ip6po_hbh);
2388         PKTOPT_EXTHDRCPY(ip6po_dest1);
2389         PKTOPT_EXTHDRCPY(ip6po_dest2);
2390         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2391         return (0);
2392
2393   bad:
2394         ip6_clearpktopts(dst, -1);
2395         return (ENOBUFS);
2396 }
2397 #undef PKTOPT_EXTHDRCPY
2398
2399 struct ip6_pktopts *
2400 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2401 {
2402         int error;
2403         struct ip6_pktopts *dst;
2404
2405         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2406         if (dst == NULL)
2407                 return (NULL);
2408         ip6_initpktopts(dst);
2409
2410         if ((error = copypktopts(dst, src, canwait)) != 0) {
2411                 free(dst, M_IP6OPT);
2412                 return (NULL);
2413         }
2414
2415         return (dst);
2416 }
2417
2418 void
2419 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2420 {
2421         if (pktopt == NULL)
2422                 return;
2423
2424         ip6_clearpktopts(pktopt, -1);
2425
2426         free(pktopt, M_IP6OPT);
2427 }
2428
2429 /*
2430  * Set IPv6 outgoing packet options based on advanced API.
2431  */
2432 int
2433 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2434     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2435 {
2436         struct cmsghdr *cm = 0;
2437
2438         if (control == NULL || opt == NULL)
2439                 return (EINVAL);
2440
2441         ip6_initpktopts(opt);
2442         if (stickyopt) {
2443                 int error;
2444
2445                 /*
2446                  * If stickyopt is provided, make a local copy of the options
2447                  * for this particular packet, then override them by ancillary
2448                  * objects.
2449                  * XXX: copypktopts() does not copy the cached route to a next
2450                  * hop (if any).  This is not very good in terms of efficiency,
2451                  * but we can allow this since this option should be rarely
2452                  * used.
2453                  */
2454                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2455                         return (error);
2456         }
2457
2458         /*
2459          * XXX: Currently, we assume all the optional information is stored
2460          * in a single mbuf.
2461          */
2462         if (control->m_next)
2463                 return (EINVAL);
2464
2465         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2466             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2467                 int error;
2468
2469                 if (control->m_len < CMSG_LEN(0))
2470                         return (EINVAL);
2471
2472                 cm = mtod(control, struct cmsghdr *);
2473                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2474                         return (EINVAL);
2475                 if (cm->cmsg_level != IPPROTO_IPV6)
2476                         continue;
2477
2478                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2479                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2480                 if (error)
2481                         return (error);
2482         }
2483
2484         return (0);
2485 }
2486
2487 /*
2488  * Set a particular packet option, as a sticky option or an ancillary data
2489  * item.  "len" can be 0 only when it's a sticky option.
2490  * We have 4 cases of combination of "sticky" and "cmsg":
2491  * "sticky=0, cmsg=0": impossible
2492  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2493  * "sticky=1, cmsg=0": RFC3542 socket option
2494  * "sticky=1, cmsg=1": RFC2292 socket option
2495  */
2496 static int
2497 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2498     struct ucred *cred, int sticky, int cmsg, int uproto)
2499 {
2500         int minmtupolicy, preftemp;
2501         int error;
2502
2503         if (!sticky && !cmsg) {
2504 #ifdef DIAGNOSTIC
2505                 printf("ip6_setpktopt: impossible case\n");
2506 #endif
2507                 return (EINVAL);
2508         }
2509
2510         /*
2511          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2512          * not be specified in the context of RFC3542.  Conversely,
2513          * RFC3542 types should not be specified in the context of RFC2292.
2514          */
2515         if (!cmsg) {
2516                 switch (optname) {
2517                 case IPV6_2292PKTINFO:
2518                 case IPV6_2292HOPLIMIT:
2519                 case IPV6_2292NEXTHOP:
2520                 case IPV6_2292HOPOPTS:
2521                 case IPV6_2292DSTOPTS:
2522                 case IPV6_2292RTHDR:
2523                 case IPV6_2292PKTOPTIONS:
2524                         return (ENOPROTOOPT);
2525                 }
2526         }
2527         if (sticky && cmsg) {
2528                 switch (optname) {
2529                 case IPV6_PKTINFO:
2530                 case IPV6_HOPLIMIT:
2531                 case IPV6_NEXTHOP:
2532                 case IPV6_HOPOPTS:
2533                 case IPV6_DSTOPTS:
2534                 case IPV6_RTHDRDSTOPTS:
2535                 case IPV6_RTHDR:
2536                 case IPV6_USE_MIN_MTU:
2537                 case IPV6_DONTFRAG:
2538                 case IPV6_TCLASS:
2539                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2540                         return (ENOPROTOOPT);
2541                 }
2542         }
2543
2544         switch (optname) {
2545         case IPV6_2292PKTINFO:
2546         case IPV6_PKTINFO:
2547         {
2548                 struct ifnet *ifp = NULL;
2549                 struct in6_pktinfo *pktinfo;
2550
2551                 if (len != sizeof(struct in6_pktinfo))
2552                         return (EINVAL);
2553
2554                 pktinfo = (struct in6_pktinfo *)buf;
2555
2556                 /*
2557                  * An application can clear any sticky IPV6_PKTINFO option by
2558                  * doing a "regular" setsockopt with ipi6_addr being
2559                  * in6addr_any and ipi6_ifindex being zero.
2560                  * [RFC 3542, Section 6]
2561                  */
2562                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2563                     pktinfo->ipi6_ifindex == 0 &&
2564                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2565                         ip6_clearpktopts(opt, optname);
2566                         break;
2567                 }
2568
2569                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2570                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2571                         return (EINVAL);
2572                 }
2573
2574                 /* validate the interface index if specified. */
2575                 if (pktinfo->ipi6_ifindex > V_if_index ||
2576                     pktinfo->ipi6_ifindex < 0) {
2577                          return (ENXIO);
2578                 }
2579                 if (pktinfo->ipi6_ifindex) {
2580                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2581                         if (ifp == NULL)
2582                                 return (ENXIO);
2583                 }
2584
2585                 /*
2586                  * We store the address anyway, and let in6_selectsrc()
2587                  * validate the specified address.  This is because ipi6_addr
2588                  * may not have enough information about its scope zone, and
2589                  * we may need additional information (such as outgoing
2590                  * interface or the scope zone of a destination address) to
2591                  * disambiguate the scope.
2592                  * XXX: the delay of the validation may confuse the
2593                  * application when it is used as a sticky option.
2594                  */
2595                 if (opt->ip6po_pktinfo == NULL) {
2596                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2597                             M_IP6OPT, M_NOWAIT);
2598                         if (opt->ip6po_pktinfo == NULL)
2599                                 return (ENOBUFS);
2600                 }
2601                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2602                 break;
2603         }
2604
2605         case IPV6_2292HOPLIMIT:
2606         case IPV6_HOPLIMIT:
2607         {
2608                 int *hlimp;
2609
2610                 /*
2611                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2612                  * to simplify the ordering among hoplimit options.
2613                  */
2614                 if (optname == IPV6_HOPLIMIT && sticky)
2615                         return (ENOPROTOOPT);
2616
2617                 if (len != sizeof(int))
2618                         return (EINVAL);
2619                 hlimp = (int *)buf;
2620                 if (*hlimp < -1 || *hlimp > 255)
2621                         return (EINVAL);
2622
2623                 opt->ip6po_hlim = *hlimp;
2624                 break;
2625         }
2626
2627         case IPV6_TCLASS:
2628         {
2629                 int tclass;
2630
2631                 if (len != sizeof(int))
2632                         return (EINVAL);
2633                 tclass = *(int *)buf;
2634                 if (tclass < -1 || tclass > 255)
2635                         return (EINVAL);
2636
2637                 opt->ip6po_tclass = tclass;
2638                 break;
2639         }
2640
2641         case IPV6_2292NEXTHOP:
2642         case IPV6_NEXTHOP:
2643                 if (cred != NULL) {
2644                         error = priv_check_cred(cred,
2645                             PRIV_NETINET_SETHDROPTS, 0);
2646                         if (error)
2647                                 return (error);
2648                 }
2649
2650                 if (len == 0) { /* just remove the option */
2651                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2652                         break;
2653                 }
2654
2655                 /* check if cmsg_len is large enough for sa_len */
2656                 if (len < sizeof(struct sockaddr) || len < *buf)
2657                         return (EINVAL);
2658
2659                 switch (((struct sockaddr *)buf)->sa_family) {
2660                 case AF_INET6:
2661                 {
2662                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2663                         int error;
2664
2665                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2666                                 return (EINVAL);
2667
2668                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2669                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2670                                 return (EINVAL);
2671                         }
2672                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2673                             != 0) {
2674                                 return (error);
2675                         }
2676                         break;
2677                 }
2678                 case AF_LINK:   /* should eventually be supported */
2679                 default:
2680                         return (EAFNOSUPPORT);
2681                 }
2682
2683                 /* turn off the previous option, then set the new option. */
2684                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2685                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2686                 if (opt->ip6po_nexthop == NULL)
2687                         return (ENOBUFS);
2688                 bcopy(buf, opt->ip6po_nexthop, *buf);
2689                 break;
2690
2691         case IPV6_2292HOPOPTS:
2692         case IPV6_HOPOPTS:
2693         {
2694                 struct ip6_hbh *hbh;
2695                 int hbhlen;
2696
2697                 /*
2698                  * XXX: We don't allow a non-privileged user to set ANY HbH
2699                  * options, since per-option restriction has too much
2700                  * overhead.
2701                  */
2702                 if (cred != NULL) {
2703                         error = priv_check_cred(cred,
2704                             PRIV_NETINET_SETHDROPTS, 0);
2705                         if (error)
2706                                 return (error);
2707                 }
2708
2709                 if (len == 0) {
2710                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2711                         break;  /* just remove the option */
2712                 }
2713
2714                 /* message length validation */
2715                 if (len < sizeof(struct ip6_hbh))
2716                         return (EINVAL);
2717                 hbh = (struct ip6_hbh *)buf;
2718                 hbhlen = (hbh->ip6h_len + 1) << 3;
2719                 if (len != hbhlen)
2720                         return (EINVAL);
2721
2722                 /* turn off the previous option, then set the new option. */
2723                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2724                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2725                 if (opt->ip6po_hbh == NULL)
2726                         return (ENOBUFS);
2727                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2728
2729                 break;
2730         }
2731
2732         case IPV6_2292DSTOPTS:
2733         case IPV6_DSTOPTS:
2734         case IPV6_RTHDRDSTOPTS:
2735         {
2736                 struct ip6_dest *dest, **newdest = NULL;
2737                 int destlen;
2738
2739                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2740                         error = priv_check_cred(cred,
2741                             PRIV_NETINET_SETHDROPTS, 0);
2742                         if (error)
2743                                 return (error);
2744                 }
2745
2746                 if (len == 0) {
2747                         ip6_clearpktopts(opt, optname);
2748                         break;  /* just remove the option */
2749                 }
2750
2751                 /* message length validation */
2752                 if (len < sizeof(struct ip6_dest))
2753                         return (EINVAL);
2754                 dest = (struct ip6_dest *)buf;
2755                 destlen = (dest->ip6d_len + 1) << 3;
2756                 if (len != destlen)
2757                         return (EINVAL);
2758
2759                 /*
2760                  * Determine the position that the destination options header
2761                  * should be inserted; before or after the routing header.
2762                  */
2763                 switch (optname) {
2764                 case IPV6_2292DSTOPTS:
2765                         /*
2766                          * The old advacned API is ambiguous on this point.
2767                          * Our approach is to determine the position based
2768                          * according to the existence of a routing header.
2769                          * Note, however, that this depends on the order of the
2770                          * extension headers in the ancillary data; the 1st
2771                          * part of the destination options header must appear
2772                          * before the routing header in the ancillary data,
2773                          * too.
2774                          * RFC3542 solved the ambiguity by introducing
2775                          * separate ancillary data or option types.
2776                          */
2777                         if (opt->ip6po_rthdr == NULL)
2778                                 newdest = &opt->ip6po_dest1;
2779                         else
2780                                 newdest = &opt->ip6po_dest2;
2781                         break;
2782                 case IPV6_RTHDRDSTOPTS:
2783                         newdest = &opt->ip6po_dest1;
2784                         break;
2785                 case IPV6_DSTOPTS:
2786                         newdest = &opt->ip6po_dest2;
2787                         break;
2788                 }
2789
2790                 /* turn off the previous option, then set the new option. */
2791                 ip6_clearpktopts(opt, optname);
2792                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2793                 if (*newdest == NULL)
2794                         return (ENOBUFS);
2795                 bcopy(dest, *newdest, destlen);
2796
2797                 break;
2798         }
2799
2800         case IPV6_2292RTHDR:
2801         case IPV6_RTHDR:
2802         {
2803                 struct ip6_rthdr *rth;
2804                 int rthlen;
2805
2806                 if (len == 0) {
2807                         ip6_clearpktopts(opt, IPV6_RTHDR);
2808                         break;  /* just remove the option */
2809                 }
2810
2811                 /* message length validation */
2812                 if (len < sizeof(struct ip6_rthdr))
2813                         return (EINVAL);
2814                 rth = (struct ip6_rthdr *)buf;
2815                 rthlen = (rth->ip6r_len + 1) << 3;
2816                 if (len != rthlen)
2817                         return (EINVAL);
2818
2819                 switch (rth->ip6r_type) {
2820                 case IPV6_RTHDR_TYPE_0:
2821                         if (rth->ip6r_len == 0) /* must contain one addr */
2822                                 return (EINVAL);
2823                         if (rth->ip6r_len % 2) /* length must be even */
2824                                 return (EINVAL);
2825                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2826                                 return (EINVAL);
2827                         break;
2828                 default:
2829                         return (EINVAL);        /* not supported */
2830                 }
2831
2832                 /* turn off the previous option */
2833                 ip6_clearpktopts(opt, IPV6_RTHDR);
2834                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2835                 if (opt->ip6po_rthdr == NULL)
2836                         return (ENOBUFS);
2837                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2838
2839                 break;
2840         }
2841
2842         case IPV6_USE_MIN_MTU:
2843                 if (len != sizeof(int))
2844                         return (EINVAL);
2845                 minmtupolicy = *(int *)buf;
2846                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2847                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2848                     minmtupolicy != IP6PO_MINMTU_ALL) {
2849                         return (EINVAL);
2850                 }
2851                 opt->ip6po_minmtu = minmtupolicy;
2852                 break;
2853
2854         case IPV6_DONTFRAG:
2855                 if (len != sizeof(int))
2856                         return (EINVAL);
2857
2858                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2859                         /*
2860                          * we ignore this option for TCP sockets.
2861                          * (RFC3542 leaves this case unspecified.)
2862                          */
2863                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2864                 } else
2865                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2866                 break;
2867
2868         case IPV6_PREFER_TEMPADDR:
2869                 if (len != sizeof(int))
2870                         return (EINVAL);
2871                 preftemp = *(int *)buf;
2872                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2873                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2874                     preftemp != IP6PO_TEMPADDR_PREFER) {
2875                         return (EINVAL);
2876                 }
2877                 opt->ip6po_prefer_tempaddr = preftemp;
2878                 break;
2879
2880         default:
2881                 return (ENOPROTOOPT);
2882         } /* end of switch */
2883
2884         return (0);
2885 }
2886
2887 /*
2888  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2889  * packet to the input queue of a specified interface.  Note that this
2890  * calls the output routine of the loopback "driver", but with an interface
2891  * pointer that might NOT be &loif -- easier than replicating that code here.
2892  */
2893 void
2894 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2895 {
2896         struct mbuf *copym;
2897         struct ip6_hdr *ip6;
2898
2899         copym = m_copy(m, 0, M_COPYALL);
2900         if (copym == NULL)
2901                 return;
2902
2903         /*
2904          * Make sure to deep-copy IPv6 header portion in case the data
2905          * is in an mbuf cluster, so that we can safely override the IPv6
2906          * header portion later.
2907          */
2908         if ((copym->m_flags & M_EXT) != 0 ||
2909             copym->m_len < sizeof(struct ip6_hdr)) {
2910                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2911                 if (copym == NULL)
2912                         return;
2913         }
2914
2915 #ifdef DIAGNOSTIC
2916         if (copym->m_len < sizeof(*ip6)) {
2917                 m_freem(copym);
2918                 return;
2919         }
2920 #endif
2921
2922         ip6 = mtod(copym, struct ip6_hdr *);
2923         /*
2924          * clear embedded scope identifiers if necessary.
2925          * in6_clearscope will touch the addresses only when necessary.
2926          */
2927         in6_clearscope(&ip6->ip6_src);
2928         in6_clearscope(&ip6->ip6_dst);
2929
2930         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
2931 }
2932
2933 /*
2934  * Chop IPv6 header off from the payload.
2935  */
2936 static int
2937 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2938 {
2939         struct mbuf *mh;
2940         struct ip6_hdr *ip6;
2941
2942         ip6 = mtod(m, struct ip6_hdr *);
2943         if (m->m_len > sizeof(*ip6)) {
2944                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2945                 if (mh == 0) {
2946                         m_freem(m);
2947                         return ENOBUFS;
2948                 }
2949                 M_MOVE_PKTHDR(mh, m);
2950                 MH_ALIGN(mh, sizeof(*ip6));
2951                 m->m_len -= sizeof(*ip6);
2952                 m->m_data += sizeof(*ip6);
2953                 mh->m_next = m;
2954                 m = mh;
2955                 m->m_len = sizeof(*ip6);
2956                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2957         }
2958         exthdrs->ip6e_ip6 = m;
2959         return 0;
2960 }
2961
2962 /*
2963  * Compute IPv6 extension header length.
2964  */
2965 int
2966 ip6_optlen(struct inpcb *in6p)
2967 {
2968         int len;
2969
2970         if (!in6p->in6p_outputopts)
2971                 return 0;
2972
2973         len = 0;
2974 #define elen(x) \
2975     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2976
2977         len += elen(in6p->in6p_outputopts->ip6po_hbh);
2978         if (in6p->in6p_outputopts->ip6po_rthdr)
2979                 /* dest1 is valid with rthdr only */
2980                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
2981         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2982         len += elen(in6p->in6p_outputopts->ip6po_dest2);
2983         return len;
2984 #undef elen
2985 }