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 <machine/in_cksum.h>
  87
  88 #include <net/if.h>
  89 #include <net/netisr.h>
  90 #include <net/route.h>
  91 #include <net/pfil.h>
  92 #include <net/vnet.h>
  93
  94 #include <netinet/in.h>
  95 #include <netinet/in_var.h>
  96 #include <netinet/ip_var.h>
  97 #include <netinet6/in6_var.h>
  98 #include <netinet/ip6.h>
  99 #include <netinet/icmp6.h>
 100 #include <netinet6/ip6_var.h>
 101 #include <netinet/in_pcb.h>
 102 #include <netinet/tcp_var.h>
 103 #include <netinet6/nd6.h>
 104
 105 #ifdef IPSEC
 106 #include <netipsec/ipsec.h>
 107 #include <netipsec/ipsec6.h>
 108 #include <netipsec/key.h>
 109 #include <netinet6/ip6_ipsec.h>
 110 #endif /* IPSEC */
 111 #ifdef SCTP
 112 #include <netinet/sctp.h>
 113 #include <netinet/sctp_crc32.h>
 114 #endif
 115
 116 #include <netinet6/ip6protosw.h>
 117 #include <netinet6/scope6_var.h>
 118
 119 #ifdef FLOWTABLE
 120 #include <net/flowtable.h>
 121 #endif
 122
 123 extern int in6_mcast_loop;
 124
 125 struct ip6_exthdrs {
 126         struct mbuf *ip6e_ip6;
 127         struct mbuf *ip6e_hbh;
 128         struct mbuf *ip6e_dest1;
 129         struct mbuf *ip6e_rthdr;
 130         struct mbuf *ip6e_dest2;
 131 };
 132
 133 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
 134                            struct ucred *, int));
 135 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
 136         struct socket *, struct sockopt *));
 137 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 138 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *,
 139         struct ucred *, int, int, int));
 140
 141 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 142 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
 143         struct ip6_frag **));
 144 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 145 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 146 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
 147         struct ifnet *, struct in6_addr *, u_long *, int *, u_int));
 148 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 149
 150
 151 /*
 152  * Make an extension header from option data.  hp is the source, and
 153  * mp is the destination.
 154  */
 155 #define MAKE_EXTHDR(hp, mp)                                             \
 156     do {                                                                \
 157         if (hp) {                                                       \
 158                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 159                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 160                     ((eh)->ip6e_len + 1) << 3);                         \
 161                 if (error)                                              \
 162                         goto freehdrs;                                  \
 163         }                                                               \
 164     } while (/*CONSTCOND*/ 0)
 165
 166 /*
 167  * Form a chain of extension headers.
 168  * m is the extension header mbuf
 169  * mp is the previous mbuf in the chain
 170  * p is the next header
 171  * i is the type of option.
 172  */
 173 #define MAKE_CHAIN(m, mp, p, i)\
 174     do {\
 175         if (m) {\
 176                 if (!hdrsplit) \
 177                         panic("assumption failed: hdr not split"); \
 178                 *mtod((m), u_char *) = *(p);\
 179                 *(p) = (i);\
 180                 p = mtod((m), u_char *);\
 181                 (m)->m_next = (mp)->m_next;\
 182                 (mp)->m_next = (m);\
 183                 (mp) = (m);\
 184         }\
 185     } while (/*CONSTCOND*/ 0)
 186
 187 static void
 188 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 189 {
 190         u_short csum;
 191
 192         csum = in_cksum_skip(m, offset + plen, offset);
 193         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 194                 csum = 0xffff;
 195         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 196
 197         if (offset + sizeof(u_short) > m->m_len) {
 198                 printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
 199                     "csum_flags=0x%04x\n", __func__, m->m_len, plen, offset,
 200                     m->m_pkthdr.csum_flags);
 201                 /*
 202                  * XXX this should not happen, but if it does, the correct
 203                  * behavior may be to insert the checksum in the appropriate
 204                  * next mbuf in the chain.
 205                  */
 206                 return;
 207         }
 208         *(u_short *)(m->m_data + offset) = csum;
 209 }
 210
 211 /*
 212  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 213  * header (with pri, len, nxt, hlim, src, dst).
 214  * This function may modify ver and hlim only.
 215  * The mbuf chain containing the packet will be freed.
 216  * The mbuf opt, if present, will not be freed.
 217  *
 218  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 219  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 220  * which is rt_rmx.rmx_mtu.
 221  *
 222  * ifpp - XXX: just for statistics
 223  */
 224 int
 225 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 226     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 227     struct ifnet **ifpp, struct inpcb *inp)
 228 {
 229         struct ip6_hdr *ip6, *mhip6;
 230         struct ifnet *ifp, *origifp;
 231         struct mbuf *m = m0;
 232         struct mbuf *mprev = NULL;
 233         int hlen, tlen, len, off;
 234         struct route_in6 ip6route;
 235         struct rtentry *rt = NULL;
 236         struct sockaddr_in6 *dst, src_sa, dst_sa;
 237         struct in6_addr odst;
 238         int error = 0;
 239         struct in6_ifaddr *ia = NULL;
 240         u_long mtu;
 241         int alwaysfrag, dontfrag;
 242         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 243         struct ip6_exthdrs exthdrs;
 244         struct in6_addr finaldst, src0, dst0;
 245         u_int32_t zone;
 246         struct route_in6 *ro_pmtu = NULL;
 247         int flevalid = 0;
 248         int hdrsplit = 0;
 249         int needipsec = 0;
 250         int sw_csum, tso;
 251 #ifdef IPSEC
 252         struct ipsec_output_state state;
 253         struct ip6_rthdr *rh = NULL;
 254         int needipsectun = 0;
 255         int segleft_org = 0;
 256         struct secpolicy *sp = NULL;
 257 #endif /* IPSEC */
 258 #ifdef IPFIREWALL_FORWARD
 259         struct m_tag *fwd_tag;
 260 #endif
 261
 262         ip6 = mtod(m, struct ip6_hdr *);
 263         if (ip6 == NULL) {
 264                 printf ("ip6 is NULL");
 265                 goto bad;
 266         }
 267
 268         if (inp != NULL)
 269                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 270
 271         finaldst = ip6->ip6_dst;
 272         bzero(&exthdrs, sizeof(exthdrs));
 273         if (opt) {
 274                 /* Hop-by-Hop options header */
 275                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 276                 /* Destination options header(1st part) */
 277                 if (opt->ip6po_rthdr) {
 278                         /*
 279                          * Destination options header(1st part)
 280                          * This only makes sense with a routing header.
 281                          * See Section 9.2 of RFC 3542.
 282                          * Disabling this part just for MIP6 convenience is
 283                          * a bad idea.  We need to think carefully about a
 284                          * way to make the advanced API coexist with MIP6
 285                          * options, which might automatically be inserted in
 286                          * the kernel.
 287                          */
 288                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 289                 }
 290                 /* Routing header */
 291                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 292                 /* Destination options header(2nd part) */
 293                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 294         }
 295
 296 #ifdef IPSEC
 297         /*
 298          * IPSec checking which handles several cases.
 299          * FAST IPSEC: We re-injected the packet.
 300          */
 301         switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
 302         {
 303         case 1:                 /* Bad packet */
 304                 goto freehdrs;
 305         case -1:                /* Do IPSec */
 306                 needipsec = 1;
 307                 /*
 308                  * Do delayed checksums now, as we may send before returning.
 309                  */
 310                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 311                         plen = m->m_pkthdr.len - sizeof(*ip6);
 312                         in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 313                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 314                 }
 315 #ifdef SCTP
 316                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
 317                         sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 318                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 319                 }
 320 #endif
 321         case 0:                 /* No IPSec */
 322         default:
 323                 break;
 324         }
 325 #endif /* IPSEC */
 326
 327         /*
 328          * Calculate the total length of the extension header chain.
 329          * Keep the length of the unfragmentable part for fragmentation.
 330          */
 331         optlen = 0;
 332         if (exthdrs.ip6e_hbh)
 333                 optlen += exthdrs.ip6e_hbh->m_len;
 334         if (exthdrs.ip6e_dest1)
 335                 optlen += exthdrs.ip6e_dest1->m_len;
 336         if (exthdrs.ip6e_rthdr)
 337                 optlen += exthdrs.ip6e_rthdr->m_len;
 338         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 339
 340         /* NOTE: we don't add AH/ESP length here. do that later. */
 341         if (exthdrs.ip6e_dest2)
 342                 optlen += exthdrs.ip6e_dest2->m_len;
 343
 344         /*
 345          * If we need IPsec, or there is at least one extension header,
 346          * separate IP6 header from the payload.
 347          */
 348         if ((needipsec || optlen) && !hdrsplit) {
 349                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 350                         m = NULL;
 351                         goto freehdrs;
 352                 }
 353                 m = exthdrs.ip6e_ip6;
 354                 hdrsplit++;
 355         }
 356
 357         /* adjust pointer */
 358         ip6 = mtod(m, struct ip6_hdr *);
 359
 360         /* adjust mbuf packet header length */
 361         m->m_pkthdr.len += optlen;
 362         plen = m->m_pkthdr.len - sizeof(*ip6);
 363
 364         /* If this is a jumbo payload, insert a jumbo payload option. */
 365         if (plen > IPV6_MAXPACKET) {
 366                 if (!hdrsplit) {
 367                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 368                                 m = NULL;
 369                                 goto freehdrs;
 370                         }
 371                         m = exthdrs.ip6e_ip6;
 372                         hdrsplit++;
 373                 }
 374                 /* adjust pointer */
 375                 ip6 = mtod(m, struct ip6_hdr *);
 376                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 377                         goto freehdrs;
 378                 ip6->ip6_plen = 0;
 379         } else
 380                 ip6->ip6_plen = htons(plen);
 381
 382         /*
 383          * Concatenate headers and fill in next header fields.
 384          * Here we have, on "m"
 385          *      IPv6 payload
 386          * and we insert headers accordingly.  Finally, we should be getting:
 387          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 388          *
 389          * during the header composing process, "m" points to IPv6 header.
 390          * "mprev" points to an extension header prior to esp.
 391          */
 392         u_char *nexthdrp = &ip6->ip6_nxt;
 393         mprev = m;
 394
 395         /*
 396          * we treat dest2 specially.  this makes IPsec processing
 397          * much easier.  the goal here is to make mprev point the
 398          * mbuf prior to dest2.
 399          *
 400          * result: IPv6 dest2 payload
 401          * m and mprev will point to IPv6 header.
 402          */
 403         if (exthdrs.ip6e_dest2) {
 404                 if (!hdrsplit)
 405                         panic("assumption failed: hdr not split");
 406                 exthdrs.ip6e_dest2->m_next = m->m_next;
 407                 m->m_next = exthdrs.ip6e_dest2;
 408                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 409                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 410         }
 411
 412         /*
 413          * result: IPv6 hbh dest1 rthdr dest2 payload
 414          * m will point to IPv6 header.  mprev will point to the
 415          * extension header prior to dest2 (rthdr in the above case).
 416          */
 417         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 418         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 419                    IPPROTO_DSTOPTS);
 420         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 421                    IPPROTO_ROUTING);
 422
 423 #ifdef IPSEC
 424         if (!needipsec)
 425                 goto skip_ipsec2;
 426
 427         /*
 428          * pointers after IPsec headers are not valid any more.
 429          * other pointers need a great care too.
 430          * (IPsec routines should not mangle mbufs prior to AH/ESP)
 431          */
 432         exthdrs.ip6e_dest2 = NULL;
 433
 434         if (exthdrs.ip6e_rthdr) {
 435                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 436                 segleft_org = rh->ip6r_segleft;
 437                 rh->ip6r_segleft = 0;
 438         }
 439
 440         bzero(&state, sizeof(state));
 441         state.m = m;
 442         error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 443                                     &needipsectun);
 444         m = state.m;
 445         if (error == EJUSTRETURN) {
 446                 /*
 447                  * We had a SP with a level of 'use' and no SA. We
 448                  * will just continue to process the packet without
 449                  * IPsec processing.
 450                  */
 451                 ;
 452         } else if (error) {
 453                 /* mbuf is already reclaimed in ipsec6_output_trans. */
 454                 m = NULL;
 455                 switch (error) {
 456                 case EHOSTUNREACH:
 457                 case ENETUNREACH:
 458                 case EMSGSIZE:
 459                 case ENOBUFS:
 460                 case ENOMEM:
 461                         break;
 462                 default:
 463                         printf("[%s:%d] (ipsec): error code %d\n",
 464                             __func__, __LINE__, error);
 465                         /* FALLTHROUGH */
 466                 case ENOENT:
 467                         /* don't show these error codes to the user */
 468                         error = 0;
 469                         break;
 470                 }
 471                 goto bad;
 472         } else if (!needipsectun) {
 473                 /*
 474                  * In the FAST IPSec case we have already
 475                  * re-injected the packet and it has been freed
 476                  * by the ipsec_done() function.  So, just clean
 477                  * up after ourselves.
 478                  */
 479                 m = NULL;
 480                 goto done;
 481         }
 482         if (exthdrs.ip6e_rthdr) {
 483                 /* ah6_output doesn't modify mbuf chain */
 484                 rh->ip6r_segleft = segleft_org;
 485         }
 486 skip_ipsec2:;
 487 #endif /* IPSEC */
 488
 489         /*
 490          * If there is a routing header, discard the packet.
 491          */
 492         if (exthdrs.ip6e_rthdr) {
 493                  error = EINVAL;
 494                  goto bad;
 495         }
 496
 497         /* Source address validation */
 498         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 499             (flags & IPV6_UNSPECSRC) == 0) {
 500                 error = EOPNOTSUPP;
 501                 V_ip6stat.ip6s_badscope++;
 502                 goto bad;
 503         }
 504         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 505                 error = EOPNOTSUPP;
 506                 V_ip6stat.ip6s_badscope++;
 507                 goto bad;
 508         }
 509
 510         V_ip6stat.ip6s_localout++;
 511
 512         /*
 513          * Route packet.
 514          */
 515         if (ro == 0) {
 516                 ro = &ip6route;
 517                 bzero((caddr_t)ro, sizeof(*ro));
 518         }
 519         ro_pmtu = ro;
 520         if (opt && opt->ip6po_rthdr)
 521                 ro = &opt->ip6po_route;
 522         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 523 #ifdef FLOWTABLE
 524         if (ro == &ip6route) {
 525                 struct flentry *fle;
 526
 527                 /*
 528                  * The flow table returns route entries valid for up to 30
 529                  * seconds; we rely on the remainder of ip_output() taking no
 530                  * longer than that long for the stability of ro_rt.  The
 531                  * flow ID assignment must have happened before this point.
 532                  */
 533                 if ((fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6)) != NULL) {
 534                         flow_to_route_in6(fle, ro);
 535                         if (ro->ro_rt != NULL && ro->ro_lle != NULL)
 536                                 flevalid = 1;
 537                 }
 538         }
 539 #endif
 540 again:
 541         /*
 542          * if specified, try to fill in the traffic class field.
 543          * do not override if a non-zero value is already set.
 544          * we check the diffserv field and the ecn field separately.
 545          */
 546         if (opt && opt->ip6po_tclass >= 0) {
 547                 int mask = 0;
 548
 549                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 550                         mask |= 0xfc;
 551                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 552                         mask |= 0x03;
 553                 if (mask != 0)
 554                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 555         }
 556
 557         /* fill in or override the hop limit field, if necessary. */
 558         if (opt && opt->ip6po_hlim != -1)
 559                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 560         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 561                 if (im6o != NULL)
 562                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 563                 else
 564                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 565         }
 566
 567 #ifdef IPSEC
 568         /*
 569          * We may re-inject packets into the stack here.
 570          */
 571         if (needipsec && needipsectun) {
 572                 struct ipsec_output_state state;
 573
 574                 /*
 575                  * All the extension headers will become inaccessible
 576                  * (since they can be encrypted).
 577                  * Don't panic, we need no more updates to extension headers
 578                  * on inner IPv6 packet (since they are now encapsulated).
 579                  *
 580                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 581                  */
 582                 bzero(&exthdrs, sizeof(exthdrs));
 583                 exthdrs.ip6e_ip6 = m;
 584
 585                 bzero(&state, sizeof(state));
 586                 state.m = m;
 587                 state.ro = (struct route *)ro;
 588                 state.dst = (struct sockaddr *)dst;
 589
 590                 error = ipsec6_output_tunnel(&state, sp, flags);
 591
 592                 m = state.m;
 593                 ro = (struct route_in6 *)state.ro;
 594                 dst = (struct sockaddr_in6 *)state.dst;
 595                 if (error == EJUSTRETURN) {
 596                         /*
 597                          * We had a SP with a level of 'use' and no SA. We
 598                          * will just continue to process the packet without
 599                          * IPsec processing.
 600                          */
 601                         ;
 602                 } else if (error) {
 603                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 604                         m0 = m = NULL;
 605                         m = NULL;
 606                         switch (error) {
 607                         case EHOSTUNREACH:
 608                         case ENETUNREACH:
 609                         case EMSGSIZE:
 610                         case ENOBUFS:
 611                         case ENOMEM:
 612                                 break;
 613                         default:
 614                                 printf("[%s:%d] (ipsec): error code %d\n",
 615                                     __func__, __LINE__, error);
 616                                 /* FALLTHROUGH */
 617                         case ENOENT:
 618                                 /* don't show these error codes to the user */
 619                                 error = 0;
 620                                 break;
 621                         }
 622                         goto bad;
 623                 } else {
 624                         /*
 625                          * In the FAST IPSec case we have already
 626                          * re-injected the packet and it has been freed
 627                          * by the ipsec_done() function.  So, just clean
 628                          * up after ourselves.
 629                          */
 630                         m = NULL;
 631                         goto done;
 632                 }
 633
 634                 exthdrs.ip6e_ip6 = m;
 635         }
 636 #endif /* IPSEC */
 637
 638         /* adjust pointer */
 639         ip6 = mtod(m, struct ip6_hdr *);
 640
 641         bzero(&dst_sa, sizeof(dst_sa));
 642         dst_sa.sin6_family = AF_INET6;
 643         dst_sa.sin6_len = sizeof(dst_sa);
 644         dst_sa.sin6_addr = ip6->ip6_dst;
 645         if (flevalid) {
 646                 rt = ro->ro_rt;
 647                 ifp = ro->ro_rt->rt_ifp;
 648         } else if ((error = in6_selectroute_fib(&dst_sa, opt, im6o, ro,
 649             &ifp, &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0) {
 650                 switch (error) {
 651                 case EHOSTUNREACH:
 652                         V_ip6stat.ip6s_noroute++;
 653                         break;
 654                 case EADDRNOTAVAIL:
 655                 default:
 656                         break; /* XXX statistics? */
 657                 }
 658                 if (ifp != NULL)
 659                         in6_ifstat_inc(ifp, ifs6_out_discard);
 660                 goto bad;
 661         }
 662         if (rt == NULL) {
 663                 /*
 664                  * If in6_selectroute() does not return a route entry,
 665                  * dst may not have been updated.
 666                  */
 667                 *dst = dst_sa;  /* XXX */
 668         }
 669
 670         /*
 671          * then rt (for unicast) and ifp must be non-NULL valid values.
 672          */
 673         if ((flags & IPV6_FORWARDING) == 0) {
 674                 /* XXX: the FORWARDING flag can be set for mrouting. */
 675                 in6_ifstat_inc(ifp, ifs6_out_request);
 676         }
 677         if (rt != NULL) {
 678                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 679                 rt->rt_use++;
 680         }
 681
 682
 683         /*
 684          * The outgoing interface must be in the zone of source and
 685          * destination addresses.
 686          */
 687         origifp = ifp;
 688
 689         src0 = ip6->ip6_src;
 690         if (in6_setscope(&src0, origifp, &zone))
 691                 goto badscope;
 692         bzero(&src_sa, sizeof(src_sa));
 693         src_sa.sin6_family = AF_INET6;
 694         src_sa.sin6_len = sizeof(src_sa);
 695         src_sa.sin6_addr = ip6->ip6_src;
 696         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 697                 goto badscope;
 698
 699         dst0 = ip6->ip6_dst;
 700         if (in6_setscope(&dst0, origifp, &zone))
 701                 goto badscope;
 702         /* re-initialize to be sure */
 703         bzero(&dst_sa, sizeof(dst_sa));
 704         dst_sa.sin6_family = AF_INET6;
 705         dst_sa.sin6_len = sizeof(dst_sa);
 706         dst_sa.sin6_addr = ip6->ip6_dst;
 707         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 708                 goto badscope;
 709         }
 710
 711         /* We should use ia_ifp to support the case of
 712          * sending packets to an address of our own.
 713          */
 714         if (ia != NULL && ia->ia_ifp)
 715                 ifp = ia->ia_ifp;
 716
 717         /* scope check is done. */
 718         goto routefound;
 719
 720   badscope:
 721         V_ip6stat.ip6s_badscope++;
 722         in6_ifstat_inc(origifp, ifs6_out_discard);
 723         if (error == 0)
 724                 error = EHOSTUNREACH; /* XXX */
 725         goto bad;
 726
 727   routefound:
 728         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 729                 if (opt && opt->ip6po_nextroute.ro_rt) {
 730                         /*
 731                          * The nexthop is explicitly specified by the
 732                          * application.  We assume the next hop is an IPv6
 733                          * address.
 734                          */
 735                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 736                 }
 737                 else if ((rt->rt_flags & RTF_GATEWAY))
 738                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 739         }
 740
 741         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 742                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 743         } else {
 744                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 745                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 746                 /*
 747                  * Confirm that the outgoing interface supports multicast.
 748                  */
 749                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 750                         V_ip6stat.ip6s_noroute++;
 751                         in6_ifstat_inc(ifp, ifs6_out_discard);
 752                         error = ENETUNREACH;
 753                         goto bad;
 754                 }
 755                 if ((im6o == NULL && in6_mcast_loop) ||
 756                     (im6o && im6o->im6o_multicast_loop)) {
 757                         /*
 758                          * Loop back multicast datagram if not expressly
 759                          * forbidden to do so, even if we have not joined
 760                          * the address; protocols will filter it later,
 761                          * thus deferring a hash lookup and lock acquisition
 762                          * at the expense of an m_copym().
 763                          */
 764                         ip6_mloopback(ifp, m, dst);
 765                 } else {
 766                         /*
 767                          * If we are acting as a multicast router, perform
 768                          * multicast forwarding as if the packet had just
 769                          * arrived on the interface to which we are about
 770                          * to send.  The multicast forwarding function
 771                          * recursively calls this function, using the
 772                          * IPV6_FORWARDING flag to prevent infinite recursion.
 773                          *
 774                          * Multicasts that are looped back by ip6_mloopback(),
 775                          * above, will be forwarded by the ip6_input() routine,
 776                          * if necessary.
 777                          */
 778                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 779                                 /*
 780                                  * XXX: ip6_mforward expects that rcvif is NULL
 781                                  * when it is called from the originating path.
 782                                  * However, it is not always the case, since
 783                                  * some versions of MGETHDR() does not
 784                                  * initialize the field.
 785                                  */
 786                                 m->m_pkthdr.rcvif = NULL;
 787                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 788                                         m_freem(m);
 789                                         goto done;
 790                                 }
 791                         }
 792                 }
 793                 /*
 794                  * Multicasts with a hoplimit of zero may be looped back,
 795                  * above, but must not be transmitted on a network.
 796                  * Also, multicasts addressed to the loopback interface
 797                  * are not sent -- the above call to ip6_mloopback() will
 798                  * loop back a copy if this host actually belongs to the
 799                  * destination group on the loopback interface.
 800                  */
 801                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 802                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 803                         m_freem(m);
 804                         goto done;
 805                 }
 806         }
 807
 808         /*
 809          * Fill the outgoing inteface to tell the upper layer
 810          * to increment per-interface statistics.
 811          */
 812         if (ifpp)
 813                 *ifpp = ifp;
 814
 815         /* Determine path MTU. */
 816         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 817             &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
 818                 goto bad;
 819
 820         /*
 821          * The caller of this function may specify to use the minimum MTU
 822          * in some cases.
 823          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 824          * setting.  The logic is a bit complicated; by default, unicast
 825          * packets will follow path MTU while multicast packets will be sent at
 826          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 827          * including unicast ones will be sent at the minimum MTU.  Multicast
 828          * packets will always be sent at the minimum MTU unless
 829          * IP6PO_MINMTU_DISABLE is explicitly specified.
 830          * See RFC 3542 for more details.
 831          */
 832         if (mtu > IPV6_MMTU) {
 833                 if ((flags & IPV6_MINMTU))
 834                         mtu = IPV6_MMTU;
 835                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 836                         mtu = IPV6_MMTU;
 837                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 838                          (opt == NULL ||
 839                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 840                         mtu = IPV6_MMTU;
 841                 }
 842         }
 843
 844         /*
 845          * clear embedded scope identifiers if necessary.
 846          * in6_clearscope will touch the addresses only when necessary.
 847          */
 848         in6_clearscope(&ip6->ip6_src);
 849         in6_clearscope(&ip6->ip6_dst);
 850
 851         /*
 852          * If the outgoing packet contains a hop-by-hop options header,
 853          * it must be examined and processed even by the source node.
 854          * (RFC 2460, section 4.)
 855          */
 856         if (exthdrs.ip6e_hbh) {
 857                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 858                 u_int32_t dummy; /* XXX unused */
 859                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 860
 861 #ifdef DIAGNOSTIC
 862                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 863                         panic("ip6e_hbh is not contiguous");
 864 #endif
 865                 /*
 866                  *  XXX: if we have to send an ICMPv6 error to the sender,
 867                  *       we need the M_LOOP flag since icmp6_error() expects
 868                  *       the IPv6 and the hop-by-hop options header are
 869                  *       contiguous unless the flag is set.
 870                  */
 871                 m->m_flags |= M_LOOP;
 872                 m->m_pkthdr.rcvif = ifp;
 873                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 874                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 875                     &dummy, &plen) < 0) {
 876                         /* m was already freed at this point */
 877                         error = EINVAL;/* better error? */
 878                         goto done;
 879                 }
 880                 m->m_flags &= ~M_LOOP; /* XXX */
 881                 m->m_pkthdr.rcvif = NULL;
 882         }
 883
 884         /* Jump over all PFIL processing if hooks are not active. */
 885         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 886                 goto passout;
 887
 888         odst = ip6->ip6_dst;
 889         /* Run through list of hooks for output packets. */
 890         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 891         if (error != 0 || m == NULL)
 892                 goto done;
 893         ip6 = mtod(m, struct ip6_hdr *);
 894
 895         /* See if destination IP address was changed by packet filter. */
 896         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 897                 m->m_flags |= M_SKIP_FIREWALL;
 898                 /* If destination is now ourself drop to ip6_input(). */
 899                 if (in6_localip(&ip6->ip6_dst)) {
 900                         m->m_flags |= M_FASTFWD_OURS;
 901                         if (m->m_pkthdr.rcvif == NULL)
 902                                 m->m_pkthdr.rcvif = V_loif;
 903                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 904                                 m->m_pkthdr.csum_flags |=
 905                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 906                                 m->m_pkthdr.csum_data = 0xffff;
 907                         }
 908 #ifdef SCTP
 909                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 910                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 911 #endif
 912                         error = netisr_queue(NETISR_IPV6, m);
 913                         goto done;
 914                 } else
 915                         goto again;     /* Redo the routing table lookup. */
 916         }
 917
 918 #ifdef IPFIREWALL_FORWARD
 919         /* See if local, if yes, send it to netisr. */
 920         if (m->m_flags & M_FASTFWD_OURS) {
 921                 if (m->m_pkthdr.rcvif == NULL)
 922                         m->m_pkthdr.rcvif = V_loif;
 923                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 924                         m->m_pkthdr.csum_flags |=
 925                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 926                         m->m_pkthdr.csum_data = 0xffff;
 927                 }
 928 #ifdef SCTP
 929                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 930                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 931 #endif
 932                 error = netisr_queue(NETISR_IPV6, m);
 933                 goto done;
 934         }
 935         /* Or forward to some other address? */
 936         fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
 937         if (fwd_tag) {
 938                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 939                 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6));
 940                 m->m_flags |= M_SKIP_FIREWALL;
 941                 m_tag_delete(m, fwd_tag);
 942                 goto again;
 943         }
 944 #endif /* IPFIREWALL_FORWARD */
 945
 946 passout:
 947         /*
 948          * Send the packet to the outgoing interface.
 949          * If necessary, do IPv6 fragmentation before sending.
 950          *
 951          * the logic here is rather complex:
 952          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 953          * 1-a: send as is if tlen <= path mtu
 954          * 1-b: fragment if tlen > path mtu
 955          *
 956          * 2: if user asks us not to fragment (dontfrag == 1)
 957          * 2-a: send as is if tlen <= interface mtu
 958          * 2-b: error if tlen > interface mtu
 959          *
 960          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 961          *      always fragment
 962          *
 963          * 4: if dontfrag == 1 && alwaysfrag == 1
 964          *      error, as we cannot handle this conflicting request
 965          */
 966         sw_csum = m->m_pkthdr.csum_flags;
 967         if (!hdrsplit) {
 968                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 969                 sw_csum &= ~ifp->if_hwassist;
 970         } else
 971                 tso = 0;
 972         /*
 973          * If we added extension headers, we will not do TSO and calculate the
 974          * checksums ourselves for now.
 975          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 976          * with ext. hdrs.
 977          */
 978         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 979                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 980                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 981         }
 982 #ifdef SCTP
 983         if (sw_csum & CSUM_SCTP_IPV6) {
 984                 sw_csum &= ~CSUM_SCTP_IPV6;
 985                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 986         }
 987 #endif
 988         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 989         tlen = m->m_pkthdr.len;
 990
 991         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 992                 dontfrag = 1;
 993         else
 994                 dontfrag = 0;
 995         if (dontfrag && alwaysfrag) {   /* case 4 */
 996                 /* conflicting request - can't transmit */
 997                 error = EMSGSIZE;
 998                 goto bad;
 999         }
1000         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
1001                 /*
1002                  * Even if the DONTFRAG option is specified, we cannot send the
1003                  * packet when the data length is larger than the MTU of the
1004                  * outgoing interface.
1005                  * Notify the error by sending IPV6_PATHMTU ancillary data as
1006                  * well as returning an error code (the latter is not described
1007                  * in the API spec.)
1008                  */
1009                 u_int32_t mtu32;
1010                 struct ip6ctlparam ip6cp;
1011
1012                 mtu32 = (u_int32_t)mtu;
1013                 bzero(&ip6cp, sizeof(ip6cp));
1014                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1015                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1016                     (void *)&ip6cp);
1017
1018                 error = EMSGSIZE;
1019                 goto bad;
1020         }
1021
1022         /*
1023          * transmit packet without fragmentation
1024          */
1025         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1026                 struct in6_ifaddr *ia6;
1027
1028                 ip6 = mtod(m, struct ip6_hdr *);
1029                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1030                 if (ia6) {
1031                         /* Record statistics for this interface address. */
1032                         ia6->ia_ifa.if_opackets++;
1033                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1034                         ifa_free(&ia6->ia_ifa);
1035                 }
1036                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1037                 goto done;
1038         }
1039
1040         /*
1041          * try to fragment the packet.  case 1-b and 3
1042          */
1043         if (mtu < IPV6_MMTU) {
1044                 /* path MTU cannot be less than IPV6_MMTU */
1045                 error = EMSGSIZE;
1046                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1047                 goto bad;
1048         } else if (ip6->ip6_plen == 0) {
1049                 /* jumbo payload cannot be fragmented */
1050                 error = EMSGSIZE;
1051                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1052                 goto bad;
1053         } else {
1054                 struct mbuf **mnext, *m_frgpart;
1055                 struct ip6_frag *ip6f;
1056                 u_int32_t id = htonl(ip6_randomid());
1057                 u_char nextproto;
1058
1059                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1060
1061                 /*
1062                  * Too large for the destination or interface;
1063                  * fragment if possible.
1064                  * Must be able to put at least 8 bytes per fragment.
1065                  */
1066                 hlen = unfragpartlen;
1067                 if (mtu > IPV6_MAXPACKET)
1068                         mtu = IPV6_MAXPACKET;
1069
1070                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1071                 if (len < 8) {
1072                         error = EMSGSIZE;
1073                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1074                         goto bad;
1075                 }
1076
1077                 /*
1078                  * Verify that we have any chance at all of being able to queue
1079                  *      the packet or packet fragments
1080                  */
1081                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1082                     < tlen  /* - hlen */)) {
1083                         error = ENOBUFS;
1084                         V_ip6stat.ip6s_odropped++;
1085                         goto bad;
1086                 }
1087
1088
1089                 /*
1090                  * If the interface will not calculate checksums on
1091                  * fragmented packets, then do it here.
1092                  * XXX-BZ handle the hw offloading case.  Need flags.
1093                  */
1094                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1095                         in6_delayed_cksum(m, plen, hlen);
1096                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1097                 }
1098 #ifdef SCTP
1099                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1100                         sctp_delayed_cksum(m, hlen);
1101                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1102                 }
1103 #endif
1104                 mnext = &m->m_nextpkt;
1105
1106                 /*
1107                  * Change the next header field of the last header in the
1108                  * unfragmentable part.
1109                  */
1110                 if (exthdrs.ip6e_rthdr) {
1111                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1112                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1113                 } else if (exthdrs.ip6e_dest1) {
1114                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1115                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1116                 } else if (exthdrs.ip6e_hbh) {
1117                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1118                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1119                 } else {
1120                         nextproto = ip6->ip6_nxt;
1121                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1122                 }
1123
1124                 /*
1125                  * Loop through length of segment after first fragment,
1126                  * make new header and copy data of each part and link onto
1127                  * chain.
1128                  */
1129                 m0 = m;
1130                 for (off = hlen; off < tlen; off += len) {
1131                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1132                         if (!m) {
1133                                 error = ENOBUFS;
1134                                 V_ip6stat.ip6s_odropped++;
1135                                 goto sendorfree;
1136                         }
1137                         m->m_pkthdr.rcvif = NULL;
1138                         m->m_flags = m0->m_flags & M_COPYFLAGS; /* incl. FIB */
1139                         *mnext = m;
1140                         mnext = &m->m_nextpkt;
1141                         m->m_data += max_linkhdr;
1142                         mhip6 = mtod(m, struct ip6_hdr *);
1143                         *mhip6 = *ip6;
1144                         m->m_len = sizeof(*mhip6);
1145                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1146                         if (error) {
1147                                 V_ip6stat.ip6s_odropped++;
1148                                 goto sendorfree;
1149                         }
1150                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1151                         if (off + len >= tlen)
1152                                 len = tlen - off;
1153                         else
1154                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1155                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1156                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1157                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1158                                 error = ENOBUFS;
1159                                 V_ip6stat.ip6s_odropped++;
1160                                 goto sendorfree;
1161                         }
1162                         m_cat(m, m_frgpart);
1163                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1164                         m->m_pkthdr.rcvif = NULL;
1165                         ip6f->ip6f_reserved = 0;
1166                         ip6f->ip6f_ident = id;
1167                         ip6f->ip6f_nxt = nextproto;
1168                         V_ip6stat.ip6s_ofragments++;
1169                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1170                 }
1171
1172                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1173         }
1174
1175         /*
1176          * Remove leading garbages.
1177          */
1178 sendorfree:
1179         m = m0->m_nextpkt;
1180         m0->m_nextpkt = 0;
1181         m_freem(m0);
1182         for (m0 = m; m; m = m0) {
1183                 m0 = m->m_nextpkt;
1184                 m->m_nextpkt = 0;
1185                 if (error == 0) {
1186                         /* Record statistics for this interface address. */
1187                         if (ia) {
1188                                 ia->ia_ifa.if_opackets++;
1189                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1190                         }
1191                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1192                 } else
1193                         m_freem(m);
1194         }
1195
1196         if (error == 0)
1197                 V_ip6stat.ip6s_fragmented++;
1198
1199 done:
1200         if (ro == &ip6route && ro->ro_rt && flevalid == 0) {
1201                 /* brace necessary for RTFREE */
1202                 RTFREE(ro->ro_rt);
1203         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt &&
1204             ((flevalid == 0) || (ro_pmtu != ro))) {
1205                 RTFREE(ro_pmtu->ro_rt);
1206         }
1207 #ifdef IPSEC
1208         if (sp != NULL)
1209                 KEY_FREESP(&sp);
1210 #endif
1211
1212         return (error);
1213
1214 freehdrs:
1215         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1216         m_freem(exthdrs.ip6e_dest1);
1217         m_freem(exthdrs.ip6e_rthdr);
1218         m_freem(exthdrs.ip6e_dest2);
1219         /* FALLTHROUGH */
1220 bad:
1221         if (m)
1222                 m_freem(m);
1223         goto done;
1224 }
1225
1226 static int
1227 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1228 {
1229         struct mbuf *m;
1230
1231         if (hlen > MCLBYTES)
1232                 return (ENOBUFS); /* XXX */
1233
1234         MGET(m, M_DONTWAIT, MT_DATA);
1235         if (!m)
1236                 return (ENOBUFS);
1237
1238         if (hlen > MLEN) {
1239                 MCLGET(m, M_DONTWAIT);
1240                 if ((m->m_flags & M_EXT) == 0) {
1241                         m_free(m);
1242                         return (ENOBUFS);
1243                 }
1244         }
1245         m->m_len = hlen;
1246         if (hdr)
1247                 bcopy(hdr, mtod(m, caddr_t), hlen);
1248
1249         *mp = m;
1250         return (0);
1251 }
1252
1253 /*
1254  * Insert jumbo payload option.
1255  */
1256 static int
1257 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1258 {
1259         struct mbuf *mopt;
1260         u_char *optbuf;
1261         u_int32_t v;
1262
1263 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1264
1265         /*
1266          * If there is no hop-by-hop options header, allocate new one.
1267          * If there is one but it doesn't have enough space to store the
1268          * jumbo payload option, allocate a cluster to store the whole options.
1269          * Otherwise, use it to store the options.
1270          */
1271         if (exthdrs->ip6e_hbh == 0) {
1272                 MGET(mopt, M_DONTWAIT, MT_DATA);
1273                 if (mopt == 0)
1274                         return (ENOBUFS);
1275                 mopt->m_len = JUMBOOPTLEN;
1276                 optbuf = mtod(mopt, u_char *);
1277                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1278                 exthdrs->ip6e_hbh = mopt;
1279         } else {
1280                 struct ip6_hbh *hbh;
1281
1282                 mopt = exthdrs->ip6e_hbh;
1283                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1284                         /*
1285                          * XXX assumption:
1286                          * - exthdrs->ip6e_hbh is not referenced from places
1287                          *   other than exthdrs.
1288                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1289                          */
1290                         int oldoptlen = mopt->m_len;
1291                         struct mbuf *n;
1292
1293                         /*
1294                          * XXX: give up if the whole (new) hbh header does
1295                          * not fit even in an mbuf cluster.
1296                          */
1297                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1298                                 return (ENOBUFS);
1299
1300                         /*
1301                          * As a consequence, we must always prepare a cluster
1302                          * at this point.
1303                          */
1304                         MGET(n, M_DONTWAIT, MT_DATA);
1305                         if (n) {
1306                                 MCLGET(n, M_DONTWAIT);
1307                                 if ((n->m_flags & M_EXT) == 0) {
1308                                         m_freem(n);
1309                                         n = NULL;
1310                                 }
1311                         }
1312                         if (!n)
1313                                 return (ENOBUFS);
1314                         n->m_len = oldoptlen + JUMBOOPTLEN;
1315                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1316                             oldoptlen);
1317                         optbuf = mtod(n, caddr_t) + oldoptlen;
1318                         m_freem(mopt);
1319                         mopt = exthdrs->ip6e_hbh = n;
1320                 } else {
1321                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1322                         mopt->m_len += JUMBOOPTLEN;
1323                 }
1324                 optbuf[0] = IP6OPT_PADN;
1325                 optbuf[1] = 1;
1326
1327                 /*
1328                  * Adjust the header length according to the pad and
1329                  * the jumbo payload option.
1330                  */
1331                 hbh = mtod(mopt, struct ip6_hbh *);
1332                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1333         }
1334
1335         /* fill in the option. */
1336         optbuf[2] = IP6OPT_JUMBO;
1337         optbuf[3] = 4;
1338         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1339         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1340
1341         /* finally, adjust the packet header length */
1342         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1343
1344         return (0);
1345 #undef JUMBOOPTLEN
1346 }
1347
1348 /*
1349  * Insert fragment header and copy unfragmentable header portions.
1350  */
1351 static int
1352 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1353     struct ip6_frag **frghdrp)
1354 {
1355         struct mbuf *n, *mlast;
1356
1357         if (hlen > sizeof(struct ip6_hdr)) {
1358                 n = m_copym(m0, sizeof(struct ip6_hdr),
1359                     hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1360                 if (n == 0)
1361                         return (ENOBUFS);
1362                 m->m_next = n;
1363         } else
1364                 n = m;
1365
1366         /* Search for the last mbuf of unfragmentable part. */
1367         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1368                 ;
1369
1370         if ((mlast->m_flags & M_EXT) == 0 &&
1371             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1372                 /* use the trailing space of the last mbuf for the fragment hdr */
1373                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1374                     mlast->m_len);
1375                 mlast->m_len += sizeof(struct ip6_frag);
1376                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1377         } else {
1378                 /* allocate a new mbuf for the fragment header */
1379                 struct mbuf *mfrg;
1380
1381                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1382                 if (mfrg == 0)
1383                         return (ENOBUFS);
1384                 mfrg->m_len = sizeof(struct ip6_frag);
1385                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1386                 mlast->m_next = mfrg;
1387         }
1388
1389         return (0);
1390 }
1391
1392 static int
1393 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1394     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1395     int *alwaysfragp, u_int fibnum)
1396 {
1397         u_int32_t mtu = 0;
1398         int alwaysfrag = 0;
1399         int error = 0;
1400
1401         if (ro_pmtu != ro) {
1402                 /* The first hop and the final destination may differ. */
1403                 struct sockaddr_in6 *sa6_dst =
1404                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1405                 if (ro_pmtu->ro_rt &&
1406                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1407                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1408                         RTFREE(ro_pmtu->ro_rt);
1409                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1410                 }
1411                 if (ro_pmtu->ro_rt == NULL) {
1412                         bzero(sa6_dst, sizeof(*sa6_dst));
1413                         sa6_dst->sin6_family = AF_INET6;
1414                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1415                         sa6_dst->sin6_addr = *dst;
1416
1417                         in6_rtalloc(ro_pmtu, fibnum);
1418                 }
1419         }
1420         if (ro_pmtu->ro_rt) {
1421                 u_int32_t ifmtu;
1422                 struct in_conninfo inc;
1423
1424                 bzero(&inc, sizeof(inc));
1425                 inc.inc_flags |= INC_ISIPV6;
1426                 inc.inc6_faddr = *dst;
1427
1428                 if (ifp == NULL)
1429                         ifp = ro_pmtu->ro_rt->rt_ifp;
1430                 ifmtu = IN6_LINKMTU(ifp);
1431                 mtu = tcp_hc_getmtu(&inc);
1432                 if (mtu)
1433                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1434                 else
1435                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1436                 if (mtu == 0)
1437                         mtu = ifmtu;
1438                 else if (mtu < IPV6_MMTU) {
1439                         /*
1440                          * RFC2460 section 5, last paragraph:
1441                          * if we record ICMPv6 too big message with
1442                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1443                          * or smaller, with framgent header attached.
1444                          * (fragment header is needed regardless from the
1445                          * packet size, for translators to identify packets)
1446                          */
1447                         alwaysfrag = 1;
1448                         mtu = IPV6_MMTU;
1449                 } else if (mtu > ifmtu) {
1450                         /*
1451                          * The MTU on the route is larger than the MTU on
1452                          * the interface!  This shouldn't happen, unless the
1453                          * MTU of the interface has been changed after the
1454                          * interface was brought up.  Change the MTU in the
1455                          * route to match the interface MTU (as long as the
1456                          * field isn't locked).
1457                          */
1458                         mtu = ifmtu;
1459                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1460                 }
1461         } else if (ifp) {
1462                 mtu = IN6_LINKMTU(ifp);
1463         } else
1464                 error = EHOSTUNREACH; /* XXX */
1465
1466         *mtup = mtu;
1467         if (alwaysfragp)
1468                 *alwaysfragp = alwaysfrag;
1469         return (error);
1470 }
1471
1472 /*
1473  * IP6 socket option processing.
1474  */
1475 int
1476 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1477 {
1478         int optdatalen, uproto;
1479         void *optdata;
1480         struct inpcb *in6p = sotoinpcb(so);
1481         int error, optval;
1482         int level, op, optname;
1483         int optlen;
1484         struct thread *td;
1485
1486         level = sopt->sopt_level;
1487         op = sopt->sopt_dir;
1488         optname = sopt->sopt_name;
1489         optlen = sopt->sopt_valsize;
1490         td = sopt->sopt_td;
1491         error = 0;
1492         optval = 0;
1493         uproto = (int)so->so_proto->pr_protocol;
1494
1495         if (level != IPPROTO_IPV6) {
1496                 error = EINVAL;
1497
1498                 if (sopt->sopt_level == SOL_SOCKET &&
1499                     sopt->sopt_dir == SOPT_SET) {
1500                         switch (sopt->sopt_name) {
1501                         case SO_REUSEADDR:
1502                                 INP_WLOCK(in6p);
1503                                 if (IN_MULTICAST(ntohl(in6p->inp_laddr.s_addr))) {
1504                                         if ((so->so_options &
1505                                             (SO_REUSEADDR | SO_REUSEPORT)) != 0)
1506                                                 in6p->inp_flags2 |= INP_REUSEPORT;
1507                                         else
1508                                                 in6p->inp_flags2 &= ~INP_REUSEPORT;
1509                                 }
1510                                 INP_WUNLOCK(in6p);
1511                                 error = 0;
1512                                 break;
1513                         case SO_REUSEPORT:
1514                                 INP_WLOCK(in6p);
1515                                 if ((so->so_options & SO_REUSEPORT) != 0)
1516                                         in6p->inp_flags2 |= INP_REUSEPORT;
1517                                 else
1518                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1519                                 INP_WUNLOCK(in6p);
1520                                 error = 0;
1521                                 break;
1522                         case SO_SETFIB:
1523                                 INP_WLOCK(in6p);
1524                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1525                                 INP_WUNLOCK(in6p);
1526                                 error = 0;
1527                                 break;
1528                         default:
1529                                 break;
1530                         }
1531                 }
1532         } else {                /* level == IPPROTO_IPV6 */
1533                 switch (op) {
1534
1535                 case SOPT_SET:
1536                         switch (optname) {
1537                         case IPV6_2292PKTOPTIONS:
1538 #ifdef IPV6_PKTOPTIONS
1539                         case IPV6_PKTOPTIONS:
1540 #endif
1541                         {
1542                                 struct mbuf *m;
1543
1544                                 error = soopt_getm(sopt, &m); /* XXX */
1545                                 if (error != 0)
1546                                         break;
1547                                 error = soopt_mcopyin(sopt, m); /* XXX */
1548                                 if (error != 0)
1549                                         break;
1550                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1551                                                     m, so, sopt);
1552                                 m_freem(m); /* XXX */
1553                                 break;
1554                         }
1555
1556                         /*
1557                          * Use of some Hop-by-Hop options or some
1558                          * Destination options, might require special
1559                          * privilege.  That is, normal applications
1560                          * (without special privilege) might be forbidden
1561                          * from setting certain options in outgoing packets,
1562                          * and might never see certain options in received
1563                          * packets. [RFC 2292 Section 6]
1564                          * KAME specific note:
1565                          *  KAME prevents non-privileged users from sending or
1566                          *  receiving ANY hbh/dst options in order to avoid
1567                          *  overhead of parsing options in the kernel.
1568                          */
1569                         case IPV6_RECVHOPOPTS:
1570                         case IPV6_RECVDSTOPTS:
1571                         case IPV6_RECVRTHDRDSTOPTS:
1572                                 if (td != NULL) {
1573                                         error = priv_check(td,
1574                                             PRIV_NETINET_SETHDROPTS);
1575                                         if (error)
1576                                                 break;
1577                                 }
1578                                 /* FALLTHROUGH */
1579                         case IPV6_UNICAST_HOPS:
1580                         case IPV6_HOPLIMIT:
1581                         case IPV6_FAITH:
1582
1583                         case IPV6_RECVPKTINFO:
1584                         case IPV6_RECVHOPLIMIT:
1585                         case IPV6_RECVRTHDR:
1586                         case IPV6_RECVPATHMTU:
1587                         case IPV6_RECVTCLASS:
1588                         case IPV6_V6ONLY:
1589                         case IPV6_AUTOFLOWLABEL:
1590                         case IPV6_BINDANY:
1591                                 if (optname == IPV6_BINDANY && td != NULL) {
1592                                         error = priv_check(td,
1593                                             PRIV_NETINET_BINDANY);
1594                                         if (error)
1595                                                 break;
1596                                 }
1597
1598                                 if (optlen != sizeof(int)) {
1599                                         error = EINVAL;
1600                                         break;
1601                                 }
1602                                 error = sooptcopyin(sopt, &optval,
1603                                         sizeof optval, sizeof optval);
1604                                 if (error)
1605                                         break;
1606                                 switch (optname) {
1607
1608                                 case IPV6_UNICAST_HOPS:
1609                                         if (optval < -1 || optval >= 256)
1610                                                 error = EINVAL;
1611                                         else {
1612                                                 /* -1 = kernel default */
1613                                                 in6p->in6p_hops = optval;
1614                                                 if ((in6p->inp_vflag &
1615                                                      INP_IPV4) != 0)
1616                                                         in6p->inp_ip_ttl = optval;
1617                                         }
1618                                         break;
1619 #define OPTSET(bit) \
1620 do { \
1621         if (optval) \
1622                 in6p->inp_flags |= (bit); \
1623         else \
1624                 in6p->inp_flags &= ~(bit); \
1625 } while (/*CONSTCOND*/ 0)
1626 #define OPTSET2292(bit) \
1627 do { \
1628         in6p->inp_flags |= IN6P_RFC2292; \
1629         if (optval) \
1630                 in6p->inp_flags |= (bit); \
1631         else \
1632                 in6p->inp_flags &= ~(bit); \
1633 } while (/*CONSTCOND*/ 0)
1634 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1635
1636                                 case IPV6_RECVPKTINFO:
1637                                         /* cannot mix with RFC2292 */
1638                                         if (OPTBIT(IN6P_RFC2292)) {
1639                                                 error = EINVAL;
1640                                                 break;
1641                                         }
1642                                         OPTSET(IN6P_PKTINFO);
1643                                         break;
1644
1645                                 case IPV6_HOPLIMIT:
1646                                 {
1647                                         struct ip6_pktopts **optp;
1648
1649                                         /* cannot mix with RFC2292 */
1650                                         if (OPTBIT(IN6P_RFC2292)) {
1651                                                 error = EINVAL;
1652                                                 break;
1653                                         }
1654                                         optp = &in6p->in6p_outputopts;
1655                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1656                                             (u_char *)&optval, sizeof(optval),
1657                                             optp, (td != NULL) ? td->td_ucred :
1658                                             NULL, uproto);
1659                                         break;
1660                                 }
1661
1662                                 case IPV6_RECVHOPLIMIT:
1663                                         /* cannot mix with RFC2292 */
1664                                         if (OPTBIT(IN6P_RFC2292)) {
1665                                                 error = EINVAL;
1666                                                 break;
1667                                         }
1668                                         OPTSET(IN6P_HOPLIMIT);
1669                                         break;
1670
1671                                 case IPV6_RECVHOPOPTS:
1672                                         /* cannot mix with RFC2292 */
1673                                         if (OPTBIT(IN6P_RFC2292)) {
1674                                                 error = EINVAL;
1675                                                 break;
1676                                         }
1677                                         OPTSET(IN6P_HOPOPTS);
1678                                         break;
1679
1680                                 case IPV6_RECVDSTOPTS:
1681                                         /* cannot mix with RFC2292 */
1682                                         if (OPTBIT(IN6P_RFC2292)) {
1683                                                 error = EINVAL;
1684                                                 break;
1685                                         }
1686                                         OPTSET(IN6P_DSTOPTS);
1687                                         break;
1688
1689                                 case IPV6_RECVRTHDRDSTOPTS:
1690                                         /* cannot mix with RFC2292 */
1691                                         if (OPTBIT(IN6P_RFC2292)) {
1692                                                 error = EINVAL;
1693                                                 break;
1694                                         }
1695                                         OPTSET(IN6P_RTHDRDSTOPTS);
1696                                         break;
1697
1698                                 case IPV6_RECVRTHDR:
1699                                         /* cannot mix with RFC2292 */
1700                                         if (OPTBIT(IN6P_RFC2292)) {
1701                                                 error = EINVAL;
1702                                                 break;
1703                                         }
1704                                         OPTSET(IN6P_RTHDR);
1705                                         break;
1706
1707                                 case IPV6_FAITH:
1708                                         OPTSET(INP_FAITH);
1709                                         break;
1710
1711                                 case IPV6_RECVPATHMTU:
1712                                         /*
1713                                          * We ignore this option for TCP
1714                                          * sockets.
1715                                          * (RFC3542 leaves this case
1716                                          * unspecified.)
1717                                          */
1718                                         if (uproto != IPPROTO_TCP)
1719                                                 OPTSET(IN6P_MTU);
1720                                         break;
1721
1722                                 case IPV6_V6ONLY:
1723                                         /*
1724                                          * make setsockopt(IPV6_V6ONLY)
1725                                          * available only prior to bind(2).
1726                                          * see ipng mailing list, Jun 22 2001.
1727                                          */
1728                                         if (in6p->inp_lport ||
1729                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1730                                                 error = EINVAL;
1731                                                 break;
1732                                         }
1733                                         OPTSET(IN6P_IPV6_V6ONLY);
1734                                         if (optval)
1735                                                 in6p->inp_vflag &= ~INP_IPV4;
1736                                         else
1737                                                 in6p->inp_vflag |= INP_IPV4;
1738                                         break;
1739                                 case IPV6_RECVTCLASS:
1740                                         /* cannot mix with RFC2292 XXX */
1741                                         if (OPTBIT(IN6P_RFC2292)) {
1742                                                 error = EINVAL;
1743                                                 break;
1744                                         }
1745                                         OPTSET(IN6P_TCLASS);
1746                                         break;
1747                                 case IPV6_AUTOFLOWLABEL:
1748                                         OPTSET(IN6P_AUTOFLOWLABEL);
1749                                         break;
1750
1751                                 case IPV6_BINDANY:
1752                                         OPTSET(INP_BINDANY);
1753                                         break;
1754                                 }
1755                                 break;
1756
1757                         case IPV6_TCLASS:
1758                         case IPV6_DONTFRAG:
1759                         case IPV6_USE_MIN_MTU:
1760                         case IPV6_PREFER_TEMPADDR:
1761                                 if (optlen != sizeof(optval)) {
1762                                         error = EINVAL;
1763                                         break;
1764                                 }
1765                                 error = sooptcopyin(sopt, &optval,
1766                                         sizeof optval, sizeof optval);
1767                                 if (error)
1768                                         break;
1769                                 {
1770                                         struct ip6_pktopts **optp;
1771                                         optp = &in6p->in6p_outputopts;
1772                                         error = ip6_pcbopt(optname,
1773                                             (u_char *)&optval, sizeof(optval),
1774                                             optp, (td != NULL) ? td->td_ucred :
1775                                             NULL, uproto);
1776                                         break;
1777                                 }
1778
1779                         case IPV6_2292PKTINFO:
1780                         case IPV6_2292HOPLIMIT:
1781                         case IPV6_2292HOPOPTS:
1782                         case IPV6_2292DSTOPTS:
1783                         case IPV6_2292RTHDR:
1784                                 /* RFC 2292 */
1785                                 if (optlen != sizeof(int)) {
1786                                         error = EINVAL;
1787                                         break;
1788                                 }
1789                                 error = sooptcopyin(sopt, &optval,
1790                                         sizeof optval, sizeof optval);
1791                                 if (error)
1792                                         break;
1793                                 switch (optname) {
1794                                 case IPV6_2292PKTINFO:
1795                                         OPTSET2292(IN6P_PKTINFO);
1796                                         break;
1797                                 case IPV6_2292HOPLIMIT:
1798                                         OPTSET2292(IN6P_HOPLIMIT);
1799                                         break;
1800                                 case IPV6_2292HOPOPTS:
1801                                         /*
1802                                          * Check super-user privilege.
1803                                          * See comments for IPV6_RECVHOPOPTS.
1804                                          */
1805                                         if (td != NULL) {
1806                                                 error = priv_check(td,
1807                                                     PRIV_NETINET_SETHDROPTS);
1808                                                 if (error)
1809                                                         return (error);
1810                                         }
1811                                         OPTSET2292(IN6P_HOPOPTS);
1812                                         break;
1813                                 case IPV6_2292DSTOPTS:
1814                                         if (td != NULL) {
1815                                                 error = priv_check(td,
1816                                                     PRIV_NETINET_SETHDROPTS);
1817                                                 if (error)
1818                                                         return (error);
1819                                         }
1820                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1821                                         break;
1822                                 case IPV6_2292RTHDR:
1823                                         OPTSET2292(IN6P_RTHDR);
1824                                         break;
1825                                 }
1826                                 break;
1827                         case IPV6_PKTINFO:
1828                         case IPV6_HOPOPTS:
1829                         case IPV6_RTHDR:
1830                         case IPV6_DSTOPTS:
1831                         case IPV6_RTHDRDSTOPTS:
1832                         case IPV6_NEXTHOP:
1833                         {
1834                                 /* new advanced API (RFC3542) */
1835                                 u_char *optbuf;
1836                                 u_char optbuf_storage[MCLBYTES];
1837                                 int optlen;
1838                                 struct ip6_pktopts **optp;
1839
1840                                 /* cannot mix with RFC2292 */
1841                                 if (OPTBIT(IN6P_RFC2292)) {
1842                                         error = EINVAL;
1843                                         break;
1844                                 }
1845
1846                                 /*
1847                                  * We only ensure valsize is not too large
1848                                  * here.  Further validation will be done
1849                                  * later.
1850                                  */
1851                                 error = sooptcopyin(sopt, optbuf_storage,
1852                                     sizeof(optbuf_storage), 0);
1853                                 if (error)
1854                                         break;
1855                                 optlen = sopt->sopt_valsize;
1856                                 optbuf = optbuf_storage;
1857                                 optp = &in6p->in6p_outputopts;
1858                                 error = ip6_pcbopt(optname, optbuf, optlen,
1859                                     optp, (td != NULL) ? td->td_ucred : NULL,
1860                                     uproto);
1861                                 break;
1862                         }
1863 #undef OPTSET
1864
1865                         case IPV6_MULTICAST_IF:
1866                         case IPV6_MULTICAST_HOPS:
1867                         case IPV6_MULTICAST_LOOP:
1868                         case IPV6_JOIN_GROUP:
1869                         case IPV6_LEAVE_GROUP:
1870                         case IPV6_MSFILTER:
1871                         case MCAST_BLOCK_SOURCE:
1872                         case MCAST_UNBLOCK_SOURCE:
1873                         case MCAST_JOIN_GROUP:
1874                         case MCAST_LEAVE_GROUP:
1875                         case MCAST_JOIN_SOURCE_GROUP:
1876                         case MCAST_LEAVE_SOURCE_GROUP:
1877                                 error = ip6_setmoptions(in6p, sopt);
1878                                 break;
1879
1880                         case IPV6_PORTRANGE:
1881                                 error = sooptcopyin(sopt, &optval,
1882                                     sizeof optval, sizeof optval);
1883                                 if (error)
1884                                         break;
1885
1886                                 switch (optval) {
1887                                 case IPV6_PORTRANGE_DEFAULT:
1888                                         in6p->inp_flags &= ~(INP_LOWPORT);
1889                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1890                                         break;
1891
1892                                 case IPV6_PORTRANGE_HIGH:
1893                                         in6p->inp_flags &= ~(INP_LOWPORT);
1894                                         in6p->inp_flags |= INP_HIGHPORT;
1895                                         break;
1896
1897                                 case IPV6_PORTRANGE_LOW:
1898                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1899                                         in6p->inp_flags |= INP_LOWPORT;
1900                                         break;
1901
1902                                 default:
1903                                         error = EINVAL;
1904                                         break;
1905                                 }
1906                                 break;
1907
1908 #ifdef IPSEC
1909                         case IPV6_IPSEC_POLICY:
1910                         {
1911                                 caddr_t req;
1912                                 struct mbuf *m;
1913
1914                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1915                                         break;
1916                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1917                                         break;
1918                                 req = mtod(m, caddr_t);
1919                                 error = ipsec_set_policy(in6p, optname, req,
1920                                     m->m_len, (sopt->sopt_td != NULL) ?
1921                                     sopt->sopt_td->td_ucred : NULL);
1922                                 m_freem(m);
1923                                 break;
1924                         }
1925 #endif /* IPSEC */
1926
1927                         default:
1928                                 error = ENOPROTOOPT;
1929                                 break;
1930                         }
1931                         break;
1932
1933                 case SOPT_GET:
1934                         switch (optname) {
1935
1936                         case IPV6_2292PKTOPTIONS:
1937 #ifdef IPV6_PKTOPTIONS
1938                         case IPV6_PKTOPTIONS:
1939 #endif
1940                                 /*
1941                                  * RFC3542 (effectively) deprecated the
1942                                  * semantics of the 2292-style pktoptions.
1943                                  * Since it was not reliable in nature (i.e.,
1944                                  * applications had to expect the lack of some
1945                                  * information after all), it would make sense
1946                                  * to simplify this part by always returning
1947                                  * empty data.
1948                                  */
1949                                 sopt->sopt_valsize = 0;
1950                                 break;
1951
1952                         case IPV6_RECVHOPOPTS:
1953                         case IPV6_RECVDSTOPTS:
1954                         case IPV6_RECVRTHDRDSTOPTS:
1955                         case IPV6_UNICAST_HOPS:
1956                         case IPV6_RECVPKTINFO:
1957                         case IPV6_RECVHOPLIMIT:
1958                         case IPV6_RECVRTHDR:
1959                         case IPV6_RECVPATHMTU:
1960
1961                         case IPV6_FAITH:
1962                         case IPV6_V6ONLY:
1963                         case IPV6_PORTRANGE:
1964                         case IPV6_RECVTCLASS:
1965                         case IPV6_AUTOFLOWLABEL:
1966                         case IPV6_BINDANY:
1967                                 switch (optname) {
1968
1969                                 case IPV6_RECVHOPOPTS:
1970                                         optval = OPTBIT(IN6P_HOPOPTS);
1971                                         break;
1972
1973                                 case IPV6_RECVDSTOPTS:
1974                                         optval = OPTBIT(IN6P_DSTOPTS);
1975                                         break;
1976
1977                                 case IPV6_RECVRTHDRDSTOPTS:
1978                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1979                                         break;
1980
1981                                 case IPV6_UNICAST_HOPS:
1982                                         optval = in6p->in6p_hops;
1983                                         break;
1984
1985                                 case IPV6_RECVPKTINFO:
1986                                         optval = OPTBIT(IN6P_PKTINFO);
1987                                         break;
1988
1989                                 case IPV6_RECVHOPLIMIT:
1990                                         optval = OPTBIT(IN6P_HOPLIMIT);
1991                                         break;
1992
1993                                 case IPV6_RECVRTHDR:
1994                                         optval = OPTBIT(IN6P_RTHDR);
1995                                         break;
1996
1997                                 case IPV6_RECVPATHMTU:
1998                                         optval = OPTBIT(IN6P_MTU);
1999                                         break;
2000
2001                                 case IPV6_FAITH:
2002                                         optval = OPTBIT(INP_FAITH);
2003                                         break;
2004
2005                                 case IPV6_V6ONLY:
2006                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2007                                         break;
2008
2009                                 case IPV6_PORTRANGE:
2010                                     {
2011                                         int flags;
2012                                         flags = in6p->inp_flags;
2013                                         if (flags & INP_HIGHPORT)
2014                                                 optval = IPV6_PORTRANGE_HIGH;
2015                                         else if (flags & INP_LOWPORT)
2016                                                 optval = IPV6_PORTRANGE_LOW;
2017                                         else
2018                                                 optval = 0;
2019                                         break;
2020                                     }
2021                                 case IPV6_RECVTCLASS:
2022                                         optval = OPTBIT(IN6P_TCLASS);
2023                                         break;
2024
2025                                 case IPV6_AUTOFLOWLABEL:
2026                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2027                                         break;
2028
2029                                 case IPV6_BINDANY:
2030                                         optval = OPTBIT(INP_BINDANY);
2031                                         break;
2032                                 }
2033                                 if (error)
2034                                         break;
2035                                 error = sooptcopyout(sopt, &optval,
2036                                         sizeof optval);
2037                                 break;
2038
2039                         case IPV6_PATHMTU:
2040                         {
2041                                 u_long pmtu = 0;
2042                                 struct ip6_mtuinfo mtuinfo;
2043                                 struct route_in6 sro;
2044
2045                                 bzero(&sro, sizeof(sro));
2046
2047                                 if (!(so->so_state & SS_ISCONNECTED))
2048                                         return (ENOTCONN);
2049                                 /*
2050                                  * XXX: we dot not consider the case of source
2051                                  * routing, or optional information to specify
2052                                  * the outgoing interface.
2053                                  */
2054                                 error = ip6_getpmtu(&sro, NULL, NULL,
2055                                     &in6p->in6p_faddr, &pmtu, NULL,
2056                                     so->so_fibnum);
2057                                 if (sro.ro_rt)
2058                                         RTFREE(sro.ro_rt);
2059                                 if (error)
2060                                         break;
2061                                 if (pmtu > IPV6_MAXPACKET)
2062                                         pmtu = IPV6_MAXPACKET;
2063
2064                                 bzero(&mtuinfo, sizeof(mtuinfo));
2065                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2066                                 optdata = (void *)&mtuinfo;
2067                                 optdatalen = sizeof(mtuinfo);
2068                                 error = sooptcopyout(sopt, optdata,
2069                                     optdatalen);
2070                                 break;
2071                         }
2072
2073                         case IPV6_2292PKTINFO:
2074                         case IPV6_2292HOPLIMIT:
2075                         case IPV6_2292HOPOPTS:
2076                         case IPV6_2292RTHDR:
2077                         case IPV6_2292DSTOPTS:
2078                                 switch (optname) {
2079                                 case IPV6_2292PKTINFO:
2080                                         optval = OPTBIT(IN6P_PKTINFO);
2081                                         break;
2082                                 case IPV6_2292HOPLIMIT:
2083                                         optval = OPTBIT(IN6P_HOPLIMIT);
2084                                         break;
2085                                 case IPV6_2292HOPOPTS:
2086                                         optval = OPTBIT(IN6P_HOPOPTS);
2087                                         break;
2088                                 case IPV6_2292RTHDR:
2089                                         optval = OPTBIT(IN6P_RTHDR);
2090                                         break;
2091                                 case IPV6_2292DSTOPTS:
2092                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2093                                         break;
2094                                 }
2095                                 error = sooptcopyout(sopt, &optval,
2096                                     sizeof optval);
2097                                 break;
2098                         case IPV6_PKTINFO:
2099                         case IPV6_HOPOPTS:
2100                         case IPV6_RTHDR:
2101                         case IPV6_DSTOPTS:
2102                         case IPV6_RTHDRDSTOPTS:
2103                         case IPV6_NEXTHOP:
2104                         case IPV6_TCLASS:
2105                         case IPV6_DONTFRAG:
2106                         case IPV6_USE_MIN_MTU:
2107                         case IPV6_PREFER_TEMPADDR:
2108                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2109                                     optname, sopt);
2110                                 break;
2111
2112                         case IPV6_MULTICAST_IF:
2113                         case IPV6_MULTICAST_HOPS:
2114                         case IPV6_MULTICAST_LOOP:
2115                         case IPV6_MSFILTER:
2116                                 error = ip6_getmoptions(in6p, sopt);
2117                                 break;
2118
2119 #ifdef IPSEC
2120                         case IPV6_IPSEC_POLICY:
2121                           {
2122                                 caddr_t req = NULL;
2123                                 size_t len = 0;
2124                                 struct mbuf *m = NULL;
2125                                 struct mbuf **mp = &m;
2126                                 size_t ovalsize = sopt->sopt_valsize;
2127                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2128
2129                                 error = soopt_getm(sopt, &m); /* XXX */
2130                                 if (error != 0)
2131                                         break;
2132                                 error = soopt_mcopyin(sopt, m); /* XXX */
2133                                 if (error != 0)
2134                                         break;
2135                                 sopt->sopt_valsize = ovalsize;
2136                                 sopt->sopt_val = oval;
2137                                 if (m) {
2138                                         req = mtod(m, caddr_t);
2139                                         len = m->m_len;
2140                                 }
2141                                 error = ipsec_get_policy(in6p, req, len, mp);
2142                                 if (error == 0)
2143                                         error = soopt_mcopyout(sopt, m); /* XXX */
2144                                 if (error == 0 && m)
2145                                         m_freem(m);
2146                                 break;
2147                           }
2148 #endif /* IPSEC */
2149
2150                         default:
2151                                 error = ENOPROTOOPT;
2152                                 break;
2153                         }
2154                         break;
2155                 }
2156         }
2157         return (error);
2158 }
2159
2160 int
2161 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2162 {
2163         int error = 0, optval, optlen;
2164         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2165         struct inpcb *in6p = sotoinpcb(so);
2166         int level, op, optname;
2167
2168         level = sopt->sopt_level;
2169         op = sopt->sopt_dir;
2170         optname = sopt->sopt_name;
2171         optlen = sopt->sopt_valsize;
2172
2173         if (level != IPPROTO_IPV6) {
2174                 return (EINVAL);
2175         }
2176
2177         switch (optname) {
2178         case IPV6_CHECKSUM:
2179                 /*
2180                  * For ICMPv6 sockets, no modification allowed for checksum
2181                  * offset, permit "no change" values to help existing apps.
2182                  *
2183                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2184                  * for an ICMPv6 socket will fail."
2185                  * The current behavior does not meet RFC3542.
2186                  */
2187                 switch (op) {
2188                 case SOPT_SET:
2189                         if (optlen != sizeof(int)) {
2190                                 error = EINVAL;
2191                                 break;
2192                         }
2193                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2194                                             sizeof(optval));
2195                         if (error)
2196                                 break;
2197                         if ((optval % 2) != 0) {
2198                                 /* the API assumes even offset values */
2199                                 error = EINVAL;
2200                         } else if (so->so_proto->pr_protocol ==
2201                             IPPROTO_ICMPV6) {
2202                                 if (optval != icmp6off)
2203                                         error = EINVAL;
2204                         } else
2205                                 in6p->in6p_cksum = optval;
2206                         break;
2207
2208                 case SOPT_GET:
2209                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2210                                 optval = icmp6off;
2211                         else
2212                                 optval = in6p->in6p_cksum;
2213
2214                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2215                         break;
2216
2217                 default:
2218                         error = EINVAL;
2219                         break;
2220                 }
2221                 break;
2222
2223         default:
2224                 error = ENOPROTOOPT;
2225                 break;
2226         }
2227
2228         return (error);
2229 }
2230
2231 /*
2232  * Set up IP6 options in pcb for insertion in output packets or
2233  * specifying behavior of outgoing packets.
2234  */
2235 static int
2236 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2237     struct socket *so, struct sockopt *sopt)
2238 {
2239         struct ip6_pktopts *opt = *pktopt;
2240         int error = 0;
2241         struct thread *td = sopt->sopt_td;
2242
2243         /* turn off any old options. */
2244         if (opt) {
2245 #ifdef DIAGNOSTIC
2246                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2247                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2248                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2249                         printf("ip6_pcbopts: all specified options are cleared.\n");
2250 #endif
2251                 ip6_clearpktopts(opt, -1);
2252         } else
2253                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2254         *pktopt = NULL;
2255
2256         if (!m || m->m_len == 0) {
2257                 /*
2258                  * Only turning off any previous options, regardless of
2259                  * whether the opt is just created or given.
2260                  */
2261                 free(opt, M_IP6OPT);
2262                 return (0);
2263         }
2264
2265         /*  set options specified by user. */
2266         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2267             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2268                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2269                 free(opt, M_IP6OPT);
2270                 return (error);
2271         }
2272         *pktopt = opt;
2273         return (0);
2274 }
2275
2276 /*
2277  * initialize ip6_pktopts.  beware that there are non-zero default values in
2278  * the struct.
2279  */
2280 void
2281 ip6_initpktopts(struct ip6_pktopts *opt)
2282 {
2283
2284         bzero(opt, sizeof(*opt));
2285         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2286         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2287         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2288         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2289 }
2290
2291 static int
2292 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2293     struct ucred *cred, int uproto)
2294 {
2295         struct ip6_pktopts *opt;
2296
2297         if (*pktopt == NULL) {
2298                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2299                     M_WAITOK);
2300                 ip6_initpktopts(*pktopt);
2301         }
2302         opt = *pktopt;
2303
2304         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2305 }
2306
2307 static int
2308 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2309 {
2310         void *optdata = NULL;
2311         int optdatalen = 0;
2312         struct ip6_ext *ip6e;
2313         int error = 0;
2314         struct in6_pktinfo null_pktinfo;
2315         int deftclass = 0, on;
2316         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2317         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2318
2319         switch (optname) {
2320         case IPV6_PKTINFO:
2321                 if (pktopt && pktopt->ip6po_pktinfo)
2322                         optdata = (void *)pktopt->ip6po_pktinfo;
2323                 else {
2324                         /* XXX: we don't have to do this every time... */
2325                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2326                         optdata = (void *)&null_pktinfo;
2327                 }
2328                 optdatalen = sizeof(struct in6_pktinfo);
2329                 break;
2330         case IPV6_TCLASS:
2331                 if (pktopt && pktopt->ip6po_tclass >= 0)
2332                         optdata = (void *)&pktopt->ip6po_tclass;
2333                 else
2334                         optdata = (void *)&deftclass;
2335                 optdatalen = sizeof(int);
2336                 break;
2337         case IPV6_HOPOPTS:
2338                 if (pktopt && pktopt->ip6po_hbh) {
2339                         optdata = (void *)pktopt->ip6po_hbh;
2340                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2341                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2342                 }
2343                 break;
2344         case IPV6_RTHDR:
2345                 if (pktopt && pktopt->ip6po_rthdr) {
2346                         optdata = (void *)pktopt->ip6po_rthdr;
2347                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2348                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2349                 }
2350                 break;
2351         case IPV6_RTHDRDSTOPTS:
2352                 if (pktopt && pktopt->ip6po_dest1) {
2353                         optdata = (void *)pktopt->ip6po_dest1;
2354                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2355                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2356                 }
2357                 break;
2358         case IPV6_DSTOPTS:
2359                 if (pktopt && pktopt->ip6po_dest2) {
2360                         optdata = (void *)pktopt->ip6po_dest2;
2361                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2362                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2363                 }
2364                 break;
2365         case IPV6_NEXTHOP:
2366                 if (pktopt && pktopt->ip6po_nexthop) {
2367                         optdata = (void *)pktopt->ip6po_nexthop;
2368                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2369                 }
2370                 break;
2371         case IPV6_USE_MIN_MTU:
2372                 if (pktopt)
2373                         optdata = (void *)&pktopt->ip6po_minmtu;
2374                 else
2375                         optdata = (void *)&defminmtu;
2376                 optdatalen = sizeof(int);
2377                 break;
2378         case IPV6_DONTFRAG:
2379                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2380                         on = 1;
2381                 else
2382                         on = 0;
2383                 optdata = (void *)&on;
2384                 optdatalen = sizeof(on);
2385                 break;
2386         case IPV6_PREFER_TEMPADDR:
2387                 if (pktopt)
2388                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2389                 else
2390                         optdata = (void *)&defpreftemp;
2391                 optdatalen = sizeof(int);
2392                 break;
2393         default:                /* should not happen */
2394 #ifdef DIAGNOSTIC
2395                 panic("ip6_getpcbopt: unexpected option\n");
2396 #endif
2397                 return (ENOPROTOOPT);
2398         }
2399
2400         error = sooptcopyout(sopt, optdata, optdatalen);
2401
2402         return (error);
2403 }
2404
2405 void
2406 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2407 {
2408         if (pktopt == NULL)
2409                 return;
2410
2411         if (optname == -1 || optname == IPV6_PKTINFO) {
2412                 if (pktopt->ip6po_pktinfo)
2413                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2414                 pktopt->ip6po_pktinfo = NULL;
2415         }
2416         if (optname == -1 || optname == IPV6_HOPLIMIT)
2417                 pktopt->ip6po_hlim = -1;
2418         if (optname == -1 || optname == IPV6_TCLASS)
2419                 pktopt->ip6po_tclass = -1;
2420         if (optname == -1 || optname == IPV6_NEXTHOP) {
2421                 if (pktopt->ip6po_nextroute.ro_rt) {
2422                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2423                         pktopt->ip6po_nextroute.ro_rt = NULL;
2424                 }
2425                 if (pktopt->ip6po_nexthop)
2426                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2427                 pktopt->ip6po_nexthop = NULL;
2428         }
2429         if (optname == -1 || optname == IPV6_HOPOPTS) {
2430                 if (pktopt->ip6po_hbh)
2431                         free(pktopt->ip6po_hbh, M_IP6OPT);
2432                 pktopt->ip6po_hbh = NULL;
2433         }
2434         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2435                 if (pktopt->ip6po_dest1)
2436                         free(pktopt->ip6po_dest1, M_IP6OPT);
2437                 pktopt->ip6po_dest1 = NULL;
2438         }
2439         if (optname == -1 || optname == IPV6_RTHDR) {
2440                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2441                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2442                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2443                 if (pktopt->ip6po_route.ro_rt) {
2444                         RTFREE(pktopt->ip6po_route.ro_rt);
2445                         pktopt->ip6po_route.ro_rt = NULL;
2446                 }
2447         }
2448         if (optname == -1 || optname == IPV6_DSTOPTS) {
2449                 if (pktopt->ip6po_dest2)
2450                         free(pktopt->ip6po_dest2, M_IP6OPT);
2451                 pktopt->ip6po_dest2 = NULL;
2452         }
2453 }
2454
2455 #define PKTOPT_EXTHDRCPY(type) \
2456 do {\
2457         if (src->type) {\
2458                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2459                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2460                 if (dst->type == NULL && canwait == M_NOWAIT)\
2461                         goto bad;\
2462                 bcopy(src->type, dst->type, hlen);\
2463         }\
2464 } while (/*CONSTCOND*/ 0)
2465
2466 static int
2467 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2468 {
2469         if (dst == NULL || src == NULL)  {
2470                 printf("ip6_clearpktopts: invalid argument\n");
2471                 return (EINVAL);
2472         }
2473
2474         dst->ip6po_hlim = src->ip6po_hlim;
2475         dst->ip6po_tclass = src->ip6po_tclass;
2476         dst->ip6po_flags = src->ip6po_flags;
2477         dst->ip6po_minmtu = src->ip6po_minmtu;
2478         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2479         if (src->ip6po_pktinfo) {
2480                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2481                     M_IP6OPT, canwait);
2482                 if (dst->ip6po_pktinfo == NULL)
2483                         goto bad;
2484                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2485         }
2486         if (src->ip6po_nexthop) {
2487                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2488                     M_IP6OPT, canwait);
2489                 if (dst->ip6po_nexthop == NULL)
2490                         goto bad;
2491                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2492                     src->ip6po_nexthop->sa_len);
2493         }
2494         PKTOPT_EXTHDRCPY(ip6po_hbh);
2495         PKTOPT_EXTHDRCPY(ip6po_dest1);
2496         PKTOPT_EXTHDRCPY(ip6po_dest2);
2497         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2498         return (0);
2499
2500   bad:
2501         ip6_clearpktopts(dst, -1);
2502         return (ENOBUFS);
2503 }
2504 #undef PKTOPT_EXTHDRCPY
2505
2506 struct ip6_pktopts *
2507 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2508 {
2509         int error;
2510         struct ip6_pktopts *dst;
2511
2512         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2513         if (dst == NULL)
2514                 return (NULL);
2515         ip6_initpktopts(dst);
2516
2517         if ((error = copypktopts(dst, src, canwait)) != 0) {
2518                 free(dst, M_IP6OPT);
2519                 return (NULL);
2520         }
2521
2522         return (dst);
2523 }
2524
2525 void
2526 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2527 {
2528         if (pktopt == NULL)
2529                 return;
2530
2531         ip6_clearpktopts(pktopt, -1);
2532
2533         free(pktopt, M_IP6OPT);
2534 }
2535
2536 /*
2537  * Set IPv6 outgoing packet options based on advanced API.
2538  */
2539 int
2540 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2541     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2542 {
2543         struct cmsghdr *cm = 0;
2544
2545         if (control == NULL || opt == NULL)
2546                 return (EINVAL);
2547
2548         ip6_initpktopts(opt);
2549         if (stickyopt) {
2550                 int error;
2551
2552                 /*
2553                  * If stickyopt is provided, make a local copy of the options
2554                  * for this particular packet, then override them by ancillary
2555                  * objects.
2556                  * XXX: copypktopts() does not copy the cached route to a next
2557                  * hop (if any).  This is not very good in terms of efficiency,
2558                  * but we can allow this since this option should be rarely
2559                  * used.
2560                  */
2561                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2562                         return (error);
2563         }
2564
2565         /*
2566          * XXX: Currently, we assume all the optional information is stored
2567          * in a single mbuf.
2568          */
2569         if (control->m_next)
2570                 return (EINVAL);
2571
2572         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2573             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2574                 int error;
2575
2576                 if (control->m_len < CMSG_LEN(0))
2577                         return (EINVAL);
2578
2579                 cm = mtod(control, struct cmsghdr *);
2580                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2581                         return (EINVAL);
2582                 if (cm->cmsg_level != IPPROTO_IPV6)
2583                         continue;
2584
2585                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2586                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2587                 if (error)
2588                         return (error);
2589         }
2590
2591         return (0);
2592 }
2593
2594 /*
2595  * Set a particular packet option, as a sticky option or an ancillary data
2596  * item.  "len" can be 0 only when it's a sticky option.
2597  * We have 4 cases of combination of "sticky" and "cmsg":
2598  * "sticky=0, cmsg=0": impossible
2599  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2600  * "sticky=1, cmsg=0": RFC3542 socket option
2601  * "sticky=1, cmsg=1": RFC2292 socket option
2602  */
2603 static int
2604 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2605     struct ucred *cred, int sticky, int cmsg, int uproto)
2606 {
2607         int minmtupolicy, preftemp;
2608         int error;
2609
2610         if (!sticky && !cmsg) {
2611 #ifdef DIAGNOSTIC
2612                 printf("ip6_setpktopt: impossible case\n");
2613 #endif
2614                 return (EINVAL);
2615         }
2616
2617         /*
2618          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2619          * not be specified in the context of RFC3542.  Conversely,
2620          * RFC3542 types should not be specified in the context of RFC2292.
2621          */
2622         if (!cmsg) {
2623                 switch (optname) {
2624                 case IPV6_2292PKTINFO:
2625                 case IPV6_2292HOPLIMIT:
2626                 case IPV6_2292NEXTHOP:
2627                 case IPV6_2292HOPOPTS:
2628                 case IPV6_2292DSTOPTS:
2629                 case IPV6_2292RTHDR:
2630                 case IPV6_2292PKTOPTIONS:
2631                         return (ENOPROTOOPT);
2632                 }
2633         }
2634         if (sticky && cmsg) {
2635                 switch (optname) {
2636                 case IPV6_PKTINFO:
2637                 case IPV6_HOPLIMIT:
2638                 case IPV6_NEXTHOP:
2639                 case IPV6_HOPOPTS:
2640                 case IPV6_DSTOPTS:
2641                 case IPV6_RTHDRDSTOPTS:
2642                 case IPV6_RTHDR:
2643                 case IPV6_USE_MIN_MTU:
2644                 case IPV6_DONTFRAG:
2645                 case IPV6_TCLASS:
2646                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2647                         return (ENOPROTOOPT);
2648                 }
2649         }
2650
2651         switch (optname) {
2652         case IPV6_2292PKTINFO:
2653         case IPV6_PKTINFO:
2654         {
2655                 struct ifnet *ifp = NULL;
2656                 struct in6_pktinfo *pktinfo;
2657
2658                 if (len != sizeof(struct in6_pktinfo))
2659                         return (EINVAL);
2660
2661                 pktinfo = (struct in6_pktinfo *)buf;
2662
2663                 /*
2664                  * An application can clear any sticky IPV6_PKTINFO option by
2665                  * doing a "regular" setsockopt with ipi6_addr being
2666                  * in6addr_any and ipi6_ifindex being zero.
2667                  * [RFC 3542, Section 6]
2668                  */
2669                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2670                     pktinfo->ipi6_ifindex == 0 &&
2671                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2672                         ip6_clearpktopts(opt, optname);
2673                         break;
2674                 }
2675
2676                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2677                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2678                         return (EINVAL);
2679                 }
2680
2681                 /* validate the interface index if specified. */
2682                 if (pktinfo->ipi6_ifindex > V_if_index ||
2683                     pktinfo->ipi6_ifindex < 0) {
2684                          return (ENXIO);
2685                 }
2686                 if (pktinfo->ipi6_ifindex) {
2687                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2688                         if (ifp == NULL)
2689                                 return (ENXIO);
2690                 }
2691
2692                 /*
2693                  * We store the address anyway, and let in6_selectsrc()
2694                  * validate the specified address.  This is because ipi6_addr
2695                  * may not have enough information about its scope zone, and
2696                  * we may need additional information (such as outgoing
2697                  * interface or the scope zone of a destination address) to
2698                  * disambiguate the scope.
2699                  * XXX: the delay of the validation may confuse the
2700                  * application when it is used as a sticky option.
2701                  */
2702                 if (opt->ip6po_pktinfo == NULL) {
2703                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2704                             M_IP6OPT, M_NOWAIT);
2705                         if (opt->ip6po_pktinfo == NULL)
2706                                 return (ENOBUFS);
2707                 }
2708                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2709                 break;
2710         }
2711
2712         case IPV6_2292HOPLIMIT:
2713         case IPV6_HOPLIMIT:
2714         {
2715                 int *hlimp;
2716
2717                 /*
2718                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2719                  * to simplify the ordering among hoplimit options.
2720                  */
2721                 if (optname == IPV6_HOPLIMIT && sticky)
2722                         return (ENOPROTOOPT);
2723
2724                 if (len != sizeof(int))
2725                         return (EINVAL);
2726                 hlimp = (int *)buf;
2727                 if (*hlimp < -1 || *hlimp > 255)
2728                         return (EINVAL);
2729
2730                 opt->ip6po_hlim = *hlimp;
2731                 break;
2732         }
2733
2734         case IPV6_TCLASS:
2735         {
2736                 int tclass;
2737
2738                 if (len != sizeof(int))
2739                         return (EINVAL);
2740                 tclass = *(int *)buf;
2741                 if (tclass < -1 || tclass > 255)
2742                         return (EINVAL);
2743
2744                 opt->ip6po_tclass = tclass;
2745                 break;
2746         }
2747
2748         case IPV6_2292NEXTHOP:
2749         case IPV6_NEXTHOP:
2750                 if (cred != NULL) {
2751                         error = priv_check_cred(cred,
2752                             PRIV_NETINET_SETHDROPTS, 0);
2753                         if (error)
2754                                 return (error);
2755                 }
2756
2757                 if (len == 0) { /* just remove the option */
2758                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2759                         break;
2760                 }
2761
2762                 /* check if cmsg_len is large enough for sa_len */
2763                 if (len < sizeof(struct sockaddr) || len < *buf)
2764                         return (EINVAL);
2765
2766                 switch (((struct sockaddr *)buf)->sa_family) {
2767                 case AF_INET6:
2768                 {
2769                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2770                         int error;
2771
2772                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2773                                 return (EINVAL);
2774
2775                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2776                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2777                                 return (EINVAL);
2778                         }
2779                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2780                             != 0) {
2781                                 return (error);
2782                         }
2783                         break;
2784                 }
2785                 case AF_LINK:   /* should eventually be supported */
2786                 default:
2787                         return (EAFNOSUPPORT);
2788                 }
2789
2790                 /* turn off the previous option, then set the new option. */
2791                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2792                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2793                 if (opt->ip6po_nexthop == NULL)
2794                         return (ENOBUFS);
2795                 bcopy(buf, opt->ip6po_nexthop, *buf);
2796                 break;
2797
2798         case IPV6_2292HOPOPTS:
2799         case IPV6_HOPOPTS:
2800         {
2801                 struct ip6_hbh *hbh;
2802                 int hbhlen;
2803
2804                 /*
2805                  * XXX: We don't allow a non-privileged user to set ANY HbH
2806                  * options, since per-option restriction has too much
2807                  * overhead.
2808                  */
2809                 if (cred != NULL) {
2810                         error = priv_check_cred(cred,
2811                             PRIV_NETINET_SETHDROPTS, 0);
2812                         if (error)
2813                                 return (error);
2814                 }
2815
2816                 if (len == 0) {
2817                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2818                         break;  /* just remove the option */
2819                 }
2820
2821                 /* message length validation */
2822                 if (len < sizeof(struct ip6_hbh))
2823                         return (EINVAL);
2824                 hbh = (struct ip6_hbh *)buf;
2825                 hbhlen = (hbh->ip6h_len + 1) << 3;
2826                 if (len != hbhlen)
2827                         return (EINVAL);
2828
2829                 /* turn off the previous option, then set the new option. */
2830                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2831                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2832                 if (opt->ip6po_hbh == NULL)
2833                         return (ENOBUFS);
2834                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2835
2836                 break;
2837         }
2838
2839         case IPV6_2292DSTOPTS:
2840         case IPV6_DSTOPTS:
2841         case IPV6_RTHDRDSTOPTS:
2842         {
2843                 struct ip6_dest *dest, **newdest = NULL;
2844                 int destlen;
2845
2846                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2847                         error = priv_check_cred(cred,
2848                             PRIV_NETINET_SETHDROPTS, 0);
2849                         if (error)
2850                                 return (error);
2851                 }
2852
2853                 if (len == 0) {
2854                         ip6_clearpktopts(opt, optname);
2855                         break;  /* just remove the option */
2856                 }
2857
2858                 /* message length validation */
2859                 if (len < sizeof(struct ip6_dest))
2860                         return (EINVAL);
2861                 dest = (struct ip6_dest *)buf;
2862                 destlen = (dest->ip6d_len + 1) << 3;
2863                 if (len != destlen)
2864                         return (EINVAL);
2865
2866                 /*
2867                  * Determine the position that the destination options header
2868                  * should be inserted; before or after the routing header.
2869                  */
2870                 switch (optname) {
2871                 case IPV6_2292DSTOPTS:
2872                         /*
2873                          * The old advacned API is ambiguous on this point.
2874                          * Our approach is to determine the position based
2875                          * according to the existence of a routing header.
2876                          * Note, however, that this depends on the order of the
2877                          * extension headers in the ancillary data; the 1st
2878                          * part of the destination options header must appear
2879                          * before the routing header in the ancillary data,
2880                          * too.
2881                          * RFC3542 solved the ambiguity by introducing
2882                          * separate ancillary data or option types.
2883                          */
2884                         if (opt->ip6po_rthdr == NULL)
2885                                 newdest = &opt->ip6po_dest1;
2886                         else
2887                                 newdest = &opt->ip6po_dest2;
2888                         break;
2889                 case IPV6_RTHDRDSTOPTS:
2890                         newdest = &opt->ip6po_dest1;
2891                         break;
2892                 case IPV6_DSTOPTS:
2893                         newdest = &opt->ip6po_dest2;
2894                         break;
2895                 }
2896
2897                 /* turn off the previous option, then set the new option. */
2898                 ip6_clearpktopts(opt, optname);
2899                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2900                 if (*newdest == NULL)
2901                         return (ENOBUFS);
2902                 bcopy(dest, *newdest, destlen);
2903
2904                 break;
2905         }
2906
2907         case IPV6_2292RTHDR:
2908         case IPV6_RTHDR:
2909         {
2910                 struct ip6_rthdr *rth;
2911                 int rthlen;
2912
2913                 if (len == 0) {
2914                         ip6_clearpktopts(opt, IPV6_RTHDR);
2915                         break;  /* just remove the option */
2916                 }
2917
2918                 /* message length validation */
2919                 if (len < sizeof(struct ip6_rthdr))
2920                         return (EINVAL);
2921                 rth = (struct ip6_rthdr *)buf;
2922                 rthlen = (rth->ip6r_len + 1) << 3;
2923                 if (len != rthlen)
2924                         return (EINVAL);
2925
2926                 switch (rth->ip6r_type) {
2927                 case IPV6_RTHDR_TYPE_0:
2928                         if (rth->ip6r_len == 0) /* must contain one addr */
2929                                 return (EINVAL);
2930                         if (rth->ip6r_len % 2) /* length must be even */
2931                                 return (EINVAL);
2932                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2933                                 return (EINVAL);
2934                         break;
2935                 default:
2936                         return (EINVAL);        /* not supported */
2937                 }
2938
2939                 /* turn off the previous option */
2940                 ip6_clearpktopts(opt, IPV6_RTHDR);
2941                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2942                 if (opt->ip6po_rthdr == NULL)
2943                         return (ENOBUFS);
2944                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2945
2946                 break;
2947         }
2948
2949         case IPV6_USE_MIN_MTU:
2950                 if (len != sizeof(int))
2951                         return (EINVAL);
2952                 minmtupolicy = *(int *)buf;
2953                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2954                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2955                     minmtupolicy != IP6PO_MINMTU_ALL) {
2956                         return (EINVAL);
2957                 }
2958                 opt->ip6po_minmtu = minmtupolicy;
2959                 break;
2960
2961         case IPV6_DONTFRAG:
2962                 if (len != sizeof(int))
2963                         return (EINVAL);
2964
2965                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2966                         /*
2967                          * we ignore this option for TCP sockets.
2968                          * (RFC3542 leaves this case unspecified.)
2969                          */
2970                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2971                 } else
2972                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2973                 break;
2974
2975         case IPV6_PREFER_TEMPADDR:
2976                 if (len != sizeof(int))
2977                         return (EINVAL);
2978                 preftemp = *(int *)buf;
2979                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2980                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2981                     preftemp != IP6PO_TEMPADDR_PREFER) {
2982                         return (EINVAL);
2983                 }
2984                 opt->ip6po_prefer_tempaddr = preftemp;
2985                 break;
2986
2987         default:
2988                 return (ENOPROTOOPT);
2989         } /* end of switch */
2990
2991         return (0);
2992 }
2993
2994 /*
2995  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2996  * packet to the input queue of a specified interface.  Note that this
2997  * calls the output routine of the loopback "driver", but with an interface
2998  * pointer that might NOT be &loif -- easier than replicating that code here.
2999  */
3000 void
3001 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3002 {
3003         struct mbuf *copym;
3004         struct ip6_hdr *ip6;
3005
3006         copym = m_copy(m, 0, M_COPYALL);
3007         if (copym == NULL)
3008                 return;
3009
3010         /*
3011          * Make sure to deep-copy IPv6 header portion in case the data
3012          * is in an mbuf cluster, so that we can safely override the IPv6
3013          * header portion later.
3014          */
3015         if ((copym->m_flags & M_EXT) != 0 ||
3016             copym->m_len < sizeof(struct ip6_hdr)) {
3017                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3018                 if (copym == NULL)
3019                         return;
3020         }
3021
3022 #ifdef DIAGNOSTIC
3023         if (copym->m_len < sizeof(*ip6)) {
3024                 m_freem(copym);
3025                 return;
3026         }
3027 #endif
3028
3029         ip6 = mtod(copym, struct ip6_hdr *);
3030         /*
3031          * clear embedded scope identifiers if necessary.
3032          * in6_clearscope will touch the addresses only when necessary.
3033          */
3034         in6_clearscope(&ip6->ip6_src);
3035         in6_clearscope(&ip6->ip6_dst);
3036
3037         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3038 }
3039
3040 /*
3041  * Chop IPv6 header off from the payload.
3042  */
3043 static int
3044 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3045 {
3046         struct mbuf *mh;
3047         struct ip6_hdr *ip6;
3048
3049         ip6 = mtod(m, struct ip6_hdr *);
3050         if (m->m_len > sizeof(*ip6)) {
3051                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3052                 if (mh == 0) {
3053                         m_freem(m);
3054                         return ENOBUFS;
3055                 }
3056                 M_MOVE_PKTHDR(mh, m);
3057                 MH_ALIGN(mh, sizeof(*ip6));
3058                 m->m_len -= sizeof(*ip6);
3059                 m->m_data += sizeof(*ip6);
3060                 mh->m_next = m;
3061                 m = mh;
3062                 m->m_len = sizeof(*ip6);
3063                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3064         }
3065         exthdrs->ip6e_ip6 = m;
3066         return 0;
3067 }
3068
3069 /*
3070  * Compute IPv6 extension header length.
3071  */
3072 int
3073 ip6_optlen(struct inpcb *in6p)
3074 {
3075         int len;
3076
3077         if (!in6p->in6p_outputopts)
3078                 return 0;
3079
3080         len = 0;
3081 #define elen(x) \
3082     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3083
3084         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3085         if (in6p->in6p_outputopts->ip6po_rthdr)
3086                 /* dest1 is valid with rthdr only */
3087                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3088         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3089         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3090         return len;
3091 #undef elen
3092 }