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/if_var.h>
  90 #include <net/netisr.h>
  91 #include <net/route.h>
  92 #include <net/pfil.h>
  93 #include <net/vnet.h>
  94
  95 #include <netinet/in.h>
  96 #include <netinet/in_var.h>
  97 #include <netinet/ip_var.h>
  98 #include <netinet6/in6_var.h>
  99 #include <netinet/ip6.h>
 100 #include <netinet/icmp6.h>
 101 #include <netinet6/ip6_var.h>
 102 #include <netinet/in_pcb.h>
 103 #include <netinet/tcp_var.h>
 104 #include <netinet6/nd6.h>
 105
 106 #ifdef IPSEC
 107 #include <netipsec/ipsec.h>
 108 #include <netipsec/ipsec6.h>
 109 #include <netipsec/key.h>
 110 #include <netinet6/ip6_ipsec.h>
 111 #endif /* IPSEC */
 112 #ifdef SCTP
 113 #include <netinet/sctp.h>
 114 #include <netinet/sctp_crc32.h>
 115 #endif
 116
 117 #include <netinet6/ip6protosw.h>
 118 #include <netinet6/scope6_var.h>
 119
 120 #ifdef FLOWTABLE
 121 #include <net/flowtable.h>
 122 #endif
 123
 124 extern int in6_mcast_loop;
 125
 126 struct ip6_exthdrs {
 127         struct mbuf *ip6e_ip6;
 128         struct mbuf *ip6e_hbh;
 129         struct mbuf *ip6e_dest1;
 130         struct mbuf *ip6e_rthdr;
 131         struct mbuf *ip6e_dest2;
 132 };
 133
 134 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
 135                            struct ucred *, int);
 136 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
 137         struct socket *, struct sockopt *);
 138 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 139 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
 140         struct ucred *, int, int, int);
 141
 142 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 143 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 144         struct ip6_frag **);
 145 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 146 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 147 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
 148         struct ifnet *, struct in6_addr *, u_long *, int *, u_int);
 149 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 150
 151
 152 /*
 153  * Make an extension header from option data.  hp is the source, and
 154  * mp is the destination.
 155  */
 156 #define MAKE_EXTHDR(hp, mp)                                             \
 157     do {                                                                \
 158         if (hp) {                                                       \
 159                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 160                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 161                     ((eh)->ip6e_len + 1) << 3);                         \
 162                 if (error)                                              \
 163                         goto freehdrs;                                  \
 164         }                                                               \
 165     } while (/*CONSTCOND*/ 0)
 166
 167 /*
 168  * Form a chain of extension headers.
 169  * m is the extension header mbuf
 170  * mp is the previous mbuf in the chain
 171  * p is the next header
 172  * i is the type of option.
 173  */
 174 #define MAKE_CHAIN(m, mp, p, i)\
 175     do {\
 176         if (m) {\
 177                 if (!hdrsplit) \
 178                         panic("assumption failed: hdr not split"); \
 179                 *mtod((m), u_char *) = *(p);\
 180                 *(p) = (i);\
 181                 p = mtod((m), u_char *);\
 182                 (m)->m_next = (mp)->m_next;\
 183                 (mp)->m_next = (m);\
 184                 (mp) = (m);\
 185         }\
 186     } while (/*CONSTCOND*/ 0)
 187
 188 static void
 189 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 190 {
 191         u_short csum;
 192
 193         csum = in_cksum_skip(m, offset + plen, offset);
 194         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 195                 csum = 0xffff;
 196         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 197
 198         if (offset + sizeof(u_short) > m->m_len) {
 199                 printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
 200                     "csum_flags=%b\n", __func__, m->m_len, plen, offset,
 201                     (int)m->m_pkthdr.csum_flags, CSUM_BITS);
 202                 /*
 203                  * XXX this should not happen, but if it does, the correct
 204                  * behavior may be to insert the checksum in the appropriate
 205                  * next mbuf in the chain.
 206                  */
 207                 return;
 208         }
 209         *(u_short *)(m->m_data + offset) = csum;
 210 }
 211
 212 /*
 213  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 214  * header (with pri, len, nxt, hlim, src, dst).
 215  * This function may modify ver and hlim only.
 216  * The mbuf chain containing the packet will be freed.
 217  * The mbuf opt, if present, will not be freed.
 218  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
 219  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 220  * then result of route lookup is stored in ro->ro_rt.
 221  *
 222  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 223  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 224  * which is rt_rmx.rmx_mtu.
 225  *
 226  * ifpp - XXX: just for statistics
 227  */
 228 int
 229 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 230     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 231     struct ifnet **ifpp, struct inpcb *inp)
 232 {
 233         struct ip6_hdr *ip6, *mhip6;
 234         struct ifnet *ifp, *origifp;
 235         struct mbuf *m = m0;
 236         struct mbuf *mprev = NULL;
 237         int hlen, tlen, len, off;
 238         struct route_in6 ip6route;
 239         struct rtentry *rt = NULL;
 240         struct sockaddr_in6 *dst, src_sa, dst_sa;
 241         struct in6_addr odst;
 242         int error = 0;
 243         struct in6_ifaddr *ia = NULL;
 244         u_long mtu;
 245         int alwaysfrag, dontfrag;
 246         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 247         struct ip6_exthdrs exthdrs;
 248         struct in6_addr finaldst, src0, dst0;
 249         u_int32_t zone;
 250         struct route_in6 *ro_pmtu = NULL;
 251         int hdrsplit = 0;
 252         int needipsec = 0;
 253         int sw_csum, tso;
 254 #ifdef IPSEC
 255         struct ipsec_output_state state;
 256         struct ip6_rthdr *rh = NULL;
 257         int needipsectun = 0;
 258         int segleft_org = 0;
 259         struct secpolicy *sp = NULL;
 260 #endif /* IPSEC */
 261         struct m_tag *fwd_tag = NULL;
 262
 263         ip6 = mtod(m, struct ip6_hdr *);
 264         if (ip6 == NULL) {
 265                 printf ("ip6 is NULL");
 266                 goto bad;
 267         }
 268
 269         if (inp != NULL)
 270                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 271
 272         finaldst = ip6->ip6_dst;
 273         bzero(&exthdrs, sizeof(exthdrs));
 274         if (opt) {
 275                 /* Hop-by-Hop options header */
 276                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 277                 /* Destination options header(1st part) */
 278                 if (opt->ip6po_rthdr) {
 279                         /*
 280                          * Destination options header(1st part)
 281                          * This only makes sense with a routing header.
 282                          * See Section 9.2 of RFC 3542.
 283                          * Disabling this part just for MIP6 convenience is
 284                          * a bad idea.  We need to think carefully about a
 285                          * way to make the advanced API coexist with MIP6
 286                          * options, which might automatically be inserted in
 287                          * the kernel.
 288                          */
 289                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 290                 }
 291                 /* Routing header */
 292                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 293                 /* Destination options header(2nd part) */
 294                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 295         }
 296
 297 #ifdef IPSEC
 298         /*
 299          * IPSec checking which handles several cases.
 300          * FAST IPSEC: We re-injected the packet.
 301          */
 302         switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
 303         {
 304         case 1:                 /* Bad packet */
 305                 goto freehdrs;
 306         case -1:                /* Do IPSec */
 307                 needipsec = 1;
 308                 /*
 309                  * Do delayed checksums now, as we may send before returning.
 310                  */
 311                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 312                         plen = m->m_pkthdr.len - sizeof(*ip6);
 313                         in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 314                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 315                 }
 316 #ifdef SCTP
 317                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
 318                         sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 319                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 320                 }
 321 #endif
 322         case 0:                 /* No IPSec */
 323         default:
 324                 break;
 325         }
 326 #endif /* IPSEC */
 327
 328         /*
 329          * Calculate the total length of the extension header chain.
 330          * Keep the length of the unfragmentable part for fragmentation.
 331          */
 332         optlen = 0;
 333         if (exthdrs.ip6e_hbh)
 334                 optlen += exthdrs.ip6e_hbh->m_len;
 335         if (exthdrs.ip6e_dest1)
 336                 optlen += exthdrs.ip6e_dest1->m_len;
 337         if (exthdrs.ip6e_rthdr)
 338                 optlen += exthdrs.ip6e_rthdr->m_len;
 339         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 340
 341         /* NOTE: we don't add AH/ESP length here. do that later. */
 342         if (exthdrs.ip6e_dest2)
 343                 optlen += exthdrs.ip6e_dest2->m_len;
 344
 345         /*
 346          * If we need IPsec, or there is at least one extension header,
 347          * separate IP6 header from the payload.
 348          */
 349         if ((needipsec || optlen) && !hdrsplit) {
 350                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 351                         m = NULL;
 352                         goto freehdrs;
 353                 }
 354                 m = exthdrs.ip6e_ip6;
 355                 hdrsplit++;
 356         }
 357
 358         /* adjust pointer */
 359         ip6 = mtod(m, struct ip6_hdr *);
 360
 361         /* adjust mbuf packet header length */
 362         m->m_pkthdr.len += optlen;
 363         plen = m->m_pkthdr.len - sizeof(*ip6);
 364
 365         /* If this is a jumbo payload, insert a jumbo payload option. */
 366         if (plen > IPV6_MAXPACKET) {
 367                 if (!hdrsplit) {
 368                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 369                                 m = NULL;
 370                                 goto freehdrs;
 371                         }
 372                         m = exthdrs.ip6e_ip6;
 373                         hdrsplit++;
 374                 }
 375                 /* adjust pointer */
 376                 ip6 = mtod(m, struct ip6_hdr *);
 377                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 378                         goto freehdrs;
 379                 ip6->ip6_plen = 0;
 380         } else
 381                 ip6->ip6_plen = htons(plen);
 382
 383         /*
 384          * Concatenate headers and fill in next header fields.
 385          * Here we have, on "m"
 386          *      IPv6 payload
 387          * and we insert headers accordingly.  Finally, we should be getting:
 388          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 389          *
 390          * during the header composing process, "m" points to IPv6 header.
 391          * "mprev" points to an extension header prior to esp.
 392          */
 393         u_char *nexthdrp = &ip6->ip6_nxt;
 394         mprev = m;
 395
 396         /*
 397          * we treat dest2 specially.  this makes IPsec processing
 398          * much easier.  the goal here is to make mprev point the
 399          * mbuf prior to dest2.
 400          *
 401          * result: IPv6 dest2 payload
 402          * m and mprev will point to IPv6 header.
 403          */
 404         if (exthdrs.ip6e_dest2) {
 405                 if (!hdrsplit)
 406                         panic("assumption failed: hdr not split");
 407                 exthdrs.ip6e_dest2->m_next = m->m_next;
 408                 m->m_next = exthdrs.ip6e_dest2;
 409                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 410                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 411         }
 412
 413         /*
 414          * result: IPv6 hbh dest1 rthdr dest2 payload
 415          * m will point to IPv6 header.  mprev will point to the
 416          * extension header prior to dest2 (rthdr in the above case).
 417          */
 418         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 419         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 420                    IPPROTO_DSTOPTS);
 421         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 422                    IPPROTO_ROUTING);
 423
 424 #ifdef IPSEC
 425         if (!needipsec)
 426                 goto skip_ipsec2;
 427
 428         /*
 429          * pointers after IPsec headers are not valid any more.
 430          * other pointers need a great care too.
 431          * (IPsec routines should not mangle mbufs prior to AH/ESP)
 432          */
 433         exthdrs.ip6e_dest2 = NULL;
 434
 435         if (exthdrs.ip6e_rthdr) {
 436                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 437                 segleft_org = rh->ip6r_segleft;
 438                 rh->ip6r_segleft = 0;
 439         }
 440
 441         bzero(&state, sizeof(state));
 442         state.m = m;
 443         error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 444                                     &needipsectun);
 445         m = state.m;
 446         if (error == EJUSTRETURN) {
 447                 /*
 448                  * We had a SP with a level of 'use' and no SA. We
 449                  * will just continue to process the packet without
 450                  * IPsec processing.
 451                  */
 452                 ;
 453         } else if (error) {
 454                 /* mbuf is already reclaimed in ipsec6_output_trans. */
 455                 m = NULL;
 456                 switch (error) {
 457                 case EHOSTUNREACH:
 458                 case ENETUNREACH:
 459                 case EMSGSIZE:
 460                 case ENOBUFS:
 461                 case ENOMEM:
 462                         break;
 463                 default:
 464                         printf("[%s:%d] (ipsec): error code %d\n",
 465                             __func__, __LINE__, error);
 466                         /* FALLTHROUGH */
 467                 case ENOENT:
 468                         /* don't show these error codes to the user */
 469                         error = 0;
 470                         break;
 471                 }
 472                 goto bad;
 473         } else if (!needipsectun) {
 474                 /*
 475                  * In the FAST IPSec case we have already
 476                  * re-injected the packet and it has been freed
 477                  * by the ipsec_done() function.  So, just clean
 478                  * up after ourselves.
 479                  */
 480                 m = NULL;
 481                 goto done;
 482         }
 483         if (exthdrs.ip6e_rthdr) {
 484                 /* ah6_output doesn't modify mbuf chain */
 485                 rh->ip6r_segleft = segleft_org;
 486         }
 487 skip_ipsec2:;
 488 #endif /* IPSEC */
 489
 490         /*
 491          * If there is a routing header, discard the packet.
 492          */
 493         if (exthdrs.ip6e_rthdr) {
 494                  error = EINVAL;
 495                  goto bad;
 496         }
 497
 498         /* Source address validation */
 499         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 500             (flags & IPV6_UNSPECSRC) == 0) {
 501                 error = EOPNOTSUPP;
 502                 IP6STAT_INC(ip6s_badscope);
 503                 goto bad;
 504         }
 505         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 506                 error = EOPNOTSUPP;
 507                 IP6STAT_INC(ip6s_badscope);
 508                 goto bad;
 509         }
 510
 511         IP6STAT_INC(ip6s_localout);
 512
 513         /*
 514          * Route packet.
 515          */
 516         if (ro == 0) {
 517                 ro = &ip6route;
 518                 bzero((caddr_t)ro, sizeof(*ro));
 519         }
 520         ro_pmtu = ro;
 521         if (opt && opt->ip6po_rthdr)
 522                 ro = &opt->ip6po_route;
 523         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 524 #ifdef FLOWTABLE
 525         if (ro->ro_rt == NULL) {
 526                 struct flentry *fle;
 527
 528                 /*
 529                  * The flow table returns route entries valid for up to 30
 530                  * seconds; we rely on the remainder of ip_output() taking no
 531                  * longer than that long for the stability of ro_rt.  The
 532                  * flow ID assignment must have happened before this point.
 533                  */
 534                 fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6);
 535                 if (fle != NULL)
 536                         flow_to_route_in6(fle, ro);
 537         }
 538 #endif
 539 again:
 540         /*
 541          * if specified, try to fill in the traffic class field.
 542          * do not override if a non-zero value is already set.
 543          * we check the diffserv field and the ecn field separately.
 544          */
 545         if (opt && opt->ip6po_tclass >= 0) {
 546                 int mask = 0;
 547
 548                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 549                         mask |= 0xfc;
 550                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 551                         mask |= 0x03;
 552                 if (mask != 0)
 553                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 554         }
 555
 556         /* fill in or override the hop limit field, if necessary. */
 557         if (opt && opt->ip6po_hlim != -1)
 558                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 559         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 560                 if (im6o != NULL)
 561                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 562                 else
 563                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 564         }
 565
 566 #ifdef IPSEC
 567         /*
 568          * We may re-inject packets into the stack here.
 569          */
 570         if (needipsec && needipsectun) {
 571                 struct ipsec_output_state state;
 572
 573                 /*
 574                  * All the extension headers will become inaccessible
 575                  * (since they can be encrypted).
 576                  * Don't panic, we need no more updates to extension headers
 577                  * on inner IPv6 packet (since they are now encapsulated).
 578                  *
 579                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 580                  */
 581                 bzero(&exthdrs, sizeof(exthdrs));
 582                 exthdrs.ip6e_ip6 = m;
 583
 584                 bzero(&state, sizeof(state));
 585                 state.m = m;
 586                 state.ro = (struct route *)ro;
 587                 state.dst = (struct sockaddr *)dst;
 588
 589                 error = ipsec6_output_tunnel(&state, sp, flags);
 590
 591                 m = state.m;
 592                 ro = (struct route_in6 *)state.ro;
 593                 dst = (struct sockaddr_in6 *)state.dst;
 594                 if (error == EJUSTRETURN) {
 595                         /*
 596                          * We had a SP with a level of 'use' and no SA. We
 597                          * will just continue to process the packet without
 598                          * IPsec processing.
 599                          */
 600                         ;
 601                 } else if (error) {
 602                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 603                         m0 = m = NULL;
 604                         m = NULL;
 605                         switch (error) {
 606                         case EHOSTUNREACH:
 607                         case ENETUNREACH:
 608                         case EMSGSIZE:
 609                         case ENOBUFS:
 610                         case ENOMEM:
 611                                 break;
 612                         default:
 613                                 printf("[%s:%d] (ipsec): error code %d\n",
 614                                     __func__, __LINE__, error);
 615                                 /* FALLTHROUGH */
 616                         case ENOENT:
 617                                 /* don't show these error codes to the user */
 618                                 error = 0;
 619                                 break;
 620                         }
 621                         goto bad;
 622                 } else {
 623                         /*
 624                          * In the FAST IPSec case we have already
 625                          * re-injected the packet and it has been freed
 626                          * by the ipsec_done() function.  So, just clean
 627                          * up after ourselves.
 628                          */
 629                         m = NULL;
 630                         goto done;
 631                 }
 632
 633                 exthdrs.ip6e_ip6 = m;
 634         }
 635 #endif /* IPSEC */
 636
 637         /* adjust pointer */
 638         ip6 = mtod(m, struct ip6_hdr *);
 639
 640         if (ro->ro_rt && fwd_tag == NULL) {
 641                 rt = ro->ro_rt;
 642                 ifp = ro->ro_rt->rt_ifp;
 643         } else {
 644                 if (fwd_tag == NULL) {
 645                         bzero(&dst_sa, sizeof(dst_sa));
 646                         dst_sa.sin6_family = AF_INET6;
 647                         dst_sa.sin6_len = sizeof(dst_sa);
 648                         dst_sa.sin6_addr = ip6->ip6_dst;
 649                 }
 650                 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
 651                     &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
 652                 if (error != 0) {
 653                         if (ifp != NULL)
 654                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 655                         goto bad;
 656                 }
 657         }
 658         if (rt == NULL) {
 659                 /*
 660                  * If in6_selectroute() does not return a route entry,
 661                  * dst may not have been updated.
 662                  */
 663                 *dst = dst_sa;  /* XXX */
 664         }
 665
 666         /*
 667          * then rt (for unicast) and ifp must be non-NULL valid values.
 668          */
 669         if ((flags & IPV6_FORWARDING) == 0) {
 670                 /* XXX: the FORWARDING flag can be set for mrouting. */
 671                 in6_ifstat_inc(ifp, ifs6_out_request);
 672         }
 673         if (rt != NULL) {
 674                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 675                 rt->rt_use++;
 676         }
 677
 678
 679         /*
 680          * The outgoing interface must be in the zone of source and
 681          * destination addresses.
 682          */
 683         origifp = ifp;
 684
 685         src0 = ip6->ip6_src;
 686         if (in6_setscope(&src0, origifp, &zone))
 687                 goto badscope;
 688         bzero(&src_sa, sizeof(src_sa));
 689         src_sa.sin6_family = AF_INET6;
 690         src_sa.sin6_len = sizeof(src_sa);
 691         src_sa.sin6_addr = ip6->ip6_src;
 692         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 693                 goto badscope;
 694
 695         dst0 = ip6->ip6_dst;
 696         if (in6_setscope(&dst0, origifp, &zone))
 697                 goto badscope;
 698         /* re-initialize to be sure */
 699         bzero(&dst_sa, sizeof(dst_sa));
 700         dst_sa.sin6_family = AF_INET6;
 701         dst_sa.sin6_len = sizeof(dst_sa);
 702         dst_sa.sin6_addr = ip6->ip6_dst;
 703         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 704                 goto badscope;
 705         }
 706
 707         /* We should use ia_ifp to support the case of
 708          * sending packets to an address of our own.
 709          */
 710         if (ia != NULL && ia->ia_ifp)
 711                 ifp = ia->ia_ifp;
 712
 713         /* scope check is done. */
 714         goto routefound;
 715
 716   badscope:
 717         IP6STAT_INC(ip6s_badscope);
 718         in6_ifstat_inc(origifp, ifs6_out_discard);
 719         if (error == 0)
 720                 error = EHOSTUNREACH; /* XXX */
 721         goto bad;
 722
 723   routefound:
 724         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 725                 if (opt && opt->ip6po_nextroute.ro_rt) {
 726                         /*
 727                          * The nexthop is explicitly specified by the
 728                          * application.  We assume the next hop is an IPv6
 729                          * address.
 730                          */
 731                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 732                 }
 733                 else if ((rt->rt_flags & RTF_GATEWAY))
 734                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 735         }
 736
 737         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 738                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 739         } else {
 740                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 741                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 742                 /*
 743                  * Confirm that the outgoing interface supports multicast.
 744                  */
 745                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 746                         IP6STAT_INC(ip6s_noroute);
 747                         in6_ifstat_inc(ifp, ifs6_out_discard);
 748                         error = ENETUNREACH;
 749                         goto bad;
 750                 }
 751                 if ((im6o == NULL && in6_mcast_loop) ||
 752                     (im6o && im6o->im6o_multicast_loop)) {
 753                         /*
 754                          * Loop back multicast datagram if not expressly
 755                          * forbidden to do so, even if we have not joined
 756                          * the address; protocols will filter it later,
 757                          * thus deferring a hash lookup and lock acquisition
 758                          * at the expense of an m_copym().
 759                          */
 760                         ip6_mloopback(ifp, m, dst);
 761                 } else {
 762                         /*
 763                          * If we are acting as a multicast router, perform
 764                          * multicast forwarding as if the packet had just
 765                          * arrived on the interface to which we are about
 766                          * to send.  The multicast forwarding function
 767                          * recursively calls this function, using the
 768                          * IPV6_FORWARDING flag to prevent infinite recursion.
 769                          *
 770                          * Multicasts that are looped back by ip6_mloopback(),
 771                          * above, will be forwarded by the ip6_input() routine,
 772                          * if necessary.
 773                          */
 774                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 775                                 /*
 776                                  * XXX: ip6_mforward expects that rcvif is NULL
 777                                  * when it is called from the originating path.
 778                                  * However, it may not always be the case.
 779                                  */
 780                                 m->m_pkthdr.rcvif = NULL;
 781                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 782                                         m_freem(m);
 783                                         goto done;
 784                                 }
 785                         }
 786                 }
 787                 /*
 788                  * Multicasts with a hoplimit of zero may be looped back,
 789                  * above, but must not be transmitted on a network.
 790                  * Also, multicasts addressed to the loopback interface
 791                  * are not sent -- the above call to ip6_mloopback() will
 792                  * loop back a copy if this host actually belongs to the
 793                  * destination group on the loopback interface.
 794                  */
 795                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 796                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 797                         m_freem(m);
 798                         goto done;
 799                 }
 800         }
 801
 802         /*
 803          * Fill the outgoing inteface to tell the upper layer
 804          * to increment per-interface statistics.
 805          */
 806         if (ifpp)
 807                 *ifpp = ifp;
 808
 809         /* Determine path MTU. */
 810         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 811             &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
 812                 goto bad;
 813
 814         /*
 815          * The caller of this function may specify to use the minimum MTU
 816          * in some cases.
 817          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 818          * setting.  The logic is a bit complicated; by default, unicast
 819          * packets will follow path MTU while multicast packets will be sent at
 820          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 821          * including unicast ones will be sent at the minimum MTU.  Multicast
 822          * packets will always be sent at the minimum MTU unless
 823          * IP6PO_MINMTU_DISABLE is explicitly specified.
 824          * See RFC 3542 for more details.
 825          */
 826         if (mtu > IPV6_MMTU) {
 827                 if ((flags & IPV6_MINMTU))
 828                         mtu = IPV6_MMTU;
 829                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 830                         mtu = IPV6_MMTU;
 831                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 832                          (opt == NULL ||
 833                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 834                         mtu = IPV6_MMTU;
 835                 }
 836         }
 837
 838         /*
 839          * clear embedded scope identifiers if necessary.
 840          * in6_clearscope will touch the addresses only when necessary.
 841          */
 842         in6_clearscope(&ip6->ip6_src);
 843         in6_clearscope(&ip6->ip6_dst);
 844
 845         /*
 846          * If the outgoing packet contains a hop-by-hop options header,
 847          * it must be examined and processed even by the source node.
 848          * (RFC 2460, section 4.)
 849          */
 850         if (exthdrs.ip6e_hbh) {
 851                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 852                 u_int32_t dummy; /* XXX unused */
 853                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 854
 855 #ifdef DIAGNOSTIC
 856                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 857                         panic("ip6e_hbh is not contiguous");
 858 #endif
 859                 /*
 860                  *  XXX: if we have to send an ICMPv6 error to the sender,
 861                  *       we need the M_LOOP flag since icmp6_error() expects
 862                  *       the IPv6 and the hop-by-hop options header are
 863                  *       contiguous unless the flag is set.
 864                  */
 865                 m->m_flags |= M_LOOP;
 866                 m->m_pkthdr.rcvif = ifp;
 867                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 868                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 869                     &dummy, &plen) < 0) {
 870                         /* m was already freed at this point */
 871                         error = EINVAL;/* better error? */
 872                         goto done;
 873                 }
 874                 m->m_flags &= ~M_LOOP; /* XXX */
 875                 m->m_pkthdr.rcvif = NULL;
 876         }
 877
 878         /* Jump over all PFIL processing if hooks are not active. */
 879         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 880                 goto passout;
 881
 882         odst = ip6->ip6_dst;
 883         /* Run through list of hooks for output packets. */
 884         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 885         if (error != 0 || m == NULL)
 886                 goto done;
 887         ip6 = mtod(m, struct ip6_hdr *);
 888
 889         /* See if destination IP address was changed by packet filter. */
 890         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 891                 m->m_flags |= M_SKIP_FIREWALL;
 892                 /* If destination is now ourself drop to ip6_input(). */
 893                 if (in6_localip(&ip6->ip6_dst)) {
 894                         m->m_flags |= M_FASTFWD_OURS;
 895                         if (m->m_pkthdr.rcvif == NULL)
 896                                 m->m_pkthdr.rcvif = V_loif;
 897                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 898                                 m->m_pkthdr.csum_flags |=
 899                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 900                                 m->m_pkthdr.csum_data = 0xffff;
 901                         }
 902 #ifdef SCTP
 903                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 904                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 905 #endif
 906                         error = netisr_queue(NETISR_IPV6, m);
 907                         goto done;
 908                 } else
 909                         goto again;     /* Redo the routing table lookup. */
 910         }
 911
 912         /* See if local, if yes, send it to netisr. */
 913         if (m->m_flags & M_FASTFWD_OURS) {
 914                 if (m->m_pkthdr.rcvif == NULL)
 915                         m->m_pkthdr.rcvif = V_loif;
 916                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 917                         m->m_pkthdr.csum_flags |=
 918                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 919                         m->m_pkthdr.csum_data = 0xffff;
 920                 }
 921 #ifdef SCTP
 922                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 923                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 924 #endif
 925                 error = netisr_queue(NETISR_IPV6, m);
 926                 goto done;
 927         }
 928         /* Or forward to some other address? */
 929         if ((m->m_flags & M_IP6_NEXTHOP) &&
 930             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 931                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 932                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
 933                 m->m_flags |= M_SKIP_FIREWALL;
 934                 m->m_flags &= ~M_IP6_NEXTHOP;
 935                 m_tag_delete(m, fwd_tag);
 936                 goto again;
 937         }
 938
 939 passout:
 940         /*
 941          * Send the packet to the outgoing interface.
 942          * If necessary, do IPv6 fragmentation before sending.
 943          *
 944          * the logic here is rather complex:
 945          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 946          * 1-a: send as is if tlen <= path mtu
 947          * 1-b: fragment if tlen > path mtu
 948          *
 949          * 2: if user asks us not to fragment (dontfrag == 1)
 950          * 2-a: send as is if tlen <= interface mtu
 951          * 2-b: error if tlen > interface mtu
 952          *
 953          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 954          *      always fragment
 955          *
 956          * 4: if dontfrag == 1 && alwaysfrag == 1
 957          *      error, as we cannot handle this conflicting request
 958          */
 959         sw_csum = m->m_pkthdr.csum_flags;
 960         if (!hdrsplit) {
 961                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 962                 sw_csum &= ~ifp->if_hwassist;
 963         } else
 964                 tso = 0;
 965         /*
 966          * If we added extension headers, we will not do TSO and calculate the
 967          * checksums ourselves for now.
 968          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 969          * with ext. hdrs.
 970          */
 971         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 972                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 973                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 974         }
 975 #ifdef SCTP
 976         if (sw_csum & CSUM_SCTP_IPV6) {
 977                 sw_csum &= ~CSUM_SCTP_IPV6;
 978                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 979         }
 980 #endif
 981         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 982         tlen = m->m_pkthdr.len;
 983
 984         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 985                 dontfrag = 1;
 986         else
 987                 dontfrag = 0;
 988         if (dontfrag && alwaysfrag) {   /* case 4 */
 989                 /* conflicting request - can't transmit */
 990                 error = EMSGSIZE;
 991                 goto bad;
 992         }
 993         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 994                 /*
 995                  * Even if the DONTFRAG option is specified, we cannot send the
 996                  * packet when the data length is larger than the MTU of the
 997                  * outgoing interface.
 998                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 999                  * well as returning an error code (the latter is not described
1000                  * in the API spec.)
1001                  */
1002                 u_int32_t mtu32;
1003                 struct ip6ctlparam ip6cp;
1004
1005                 mtu32 = (u_int32_t)mtu;
1006                 bzero(&ip6cp, sizeof(ip6cp));
1007                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1008                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1009                     (void *)&ip6cp);
1010
1011                 error = EMSGSIZE;
1012                 goto bad;
1013         }
1014
1015         /*
1016          * transmit packet without fragmentation
1017          */
1018         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1019                 struct in6_ifaddr *ia6;
1020
1021                 ip6 = mtod(m, struct ip6_hdr *);
1022                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1023                 if (ia6) {
1024                         /* Record statistics for this interface address. */
1025                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1026                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
1027                             m->m_pkthdr.len);
1028                         ifa_free(&ia6->ia_ifa);
1029                 }
1030                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1031                 goto done;
1032         }
1033
1034         /*
1035          * try to fragment the packet.  case 1-b and 3
1036          */
1037         if (mtu < IPV6_MMTU) {
1038                 /* path MTU cannot be less than IPV6_MMTU */
1039                 error = EMSGSIZE;
1040                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1041                 goto bad;
1042         } else if (ip6->ip6_plen == 0) {
1043                 /* jumbo payload cannot be fragmented */
1044                 error = EMSGSIZE;
1045                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1046                 goto bad;
1047         } else {
1048                 struct mbuf **mnext, *m_frgpart;
1049                 struct ip6_frag *ip6f;
1050                 u_int32_t id = htonl(ip6_randomid());
1051                 u_char nextproto;
1052
1053                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1054
1055                 /*
1056                  * Too large for the destination or interface;
1057                  * fragment if possible.
1058                  * Must be able to put at least 8 bytes per fragment.
1059                  */
1060                 hlen = unfragpartlen;
1061                 if (mtu > IPV6_MAXPACKET)
1062                         mtu = IPV6_MAXPACKET;
1063
1064                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1065                 if (len < 8) {
1066                         error = EMSGSIZE;
1067                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1068                         goto bad;
1069                 }
1070
1071                 /*
1072                  * Verify that we have any chance at all of being able to queue
1073                  *      the packet or packet fragments
1074                  */
1075                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1076                     < tlen  /* - hlen */)) {
1077                         error = ENOBUFS;
1078                         IP6STAT_INC(ip6s_odropped);
1079                         goto bad;
1080                 }
1081
1082
1083                 /*
1084                  * If the interface will not calculate checksums on
1085                  * fragmented packets, then do it here.
1086                  * XXX-BZ handle the hw offloading case.  Need flags.
1087                  */
1088                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1089                         in6_delayed_cksum(m, plen, hlen);
1090                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1091                 }
1092 #ifdef SCTP
1093                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1094                         sctp_delayed_cksum(m, hlen);
1095                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1096                 }
1097 #endif
1098                 mnext = &m->m_nextpkt;
1099
1100                 /*
1101                  * Change the next header field of the last header in the
1102                  * unfragmentable part.
1103                  */
1104                 if (exthdrs.ip6e_rthdr) {
1105                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1106                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1107                 } else if (exthdrs.ip6e_dest1) {
1108                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1109                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1110                 } else if (exthdrs.ip6e_hbh) {
1111                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1112                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1113                 } else {
1114                         nextproto = ip6->ip6_nxt;
1115                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1116                 }
1117
1118                 /*
1119                  * Loop through length of segment after first fragment,
1120                  * make new header and copy data of each part and link onto
1121                  * chain.
1122                  */
1123                 m0 = m;
1124                 for (off = hlen; off < tlen; off += len) {
1125                         m = m_gethdr(M_NOWAIT, MT_DATA);
1126                         if (!m) {
1127                                 error = ENOBUFS;
1128                                 IP6STAT_INC(ip6s_odropped);
1129                                 goto sendorfree;
1130                         }
1131                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1132                         *mnext = m;
1133                         mnext = &m->m_nextpkt;
1134                         m->m_data += max_linkhdr;
1135                         mhip6 = mtod(m, struct ip6_hdr *);
1136                         *mhip6 = *ip6;
1137                         m->m_len = sizeof(*mhip6);
1138                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1139                         if (error) {
1140                                 IP6STAT_INC(ip6s_odropped);
1141                                 goto sendorfree;
1142                         }
1143                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1144                         if (off + len >= tlen)
1145                                 len = tlen - off;
1146                         else
1147                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1148                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1149                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1150                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1151                                 error = ENOBUFS;
1152                                 IP6STAT_INC(ip6s_odropped);
1153                                 goto sendorfree;
1154                         }
1155                         m_cat(m, m_frgpart);
1156                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1157                         m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
1158                         m->m_pkthdr.rcvif = NULL;
1159                         ip6f->ip6f_reserved = 0;
1160                         ip6f->ip6f_ident = id;
1161                         ip6f->ip6f_nxt = nextproto;
1162                         IP6STAT_INC(ip6s_ofragments);
1163                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1164                 }
1165
1166                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1167         }
1168
1169         /*
1170          * Remove leading garbages.
1171          */
1172 sendorfree:
1173         m = m0->m_nextpkt;
1174         m0->m_nextpkt = 0;
1175         m_freem(m0);
1176         for (m0 = m; m; m = m0) {
1177                 m0 = m->m_nextpkt;
1178                 m->m_nextpkt = 0;
1179                 if (error == 0) {
1180                         /* Record statistics for this interface address. */
1181                         if (ia) {
1182                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1183                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1184                                     m->m_pkthdr.len);
1185                         }
1186                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1187                 } else
1188                         m_freem(m);
1189         }
1190
1191         if (error == 0)
1192                 IP6STAT_INC(ip6s_fragmented);
1193
1194 done:
1195         if (ro == &ip6route)
1196                 RO_RTFREE(ro);
1197         if (ro_pmtu == &ip6route)
1198                 RO_RTFREE(ro_pmtu);
1199 #ifdef IPSEC
1200         if (sp != NULL)
1201                 KEY_FREESP(&sp);
1202 #endif
1203
1204         return (error);
1205
1206 freehdrs:
1207         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1208         m_freem(exthdrs.ip6e_dest1);
1209         m_freem(exthdrs.ip6e_rthdr);
1210         m_freem(exthdrs.ip6e_dest2);
1211         /* FALLTHROUGH */
1212 bad:
1213         if (m)
1214                 m_freem(m);
1215         goto done;
1216 }
1217
1218 static int
1219 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1220 {
1221         struct mbuf *m;
1222
1223         if (hlen > MCLBYTES)
1224                 return (ENOBUFS); /* XXX */
1225
1226         if (hlen > MLEN)
1227                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1228         else
1229                 m = m_get(M_NOWAIT, MT_DATA);
1230         if (m == NULL)
1231                 return (ENOBUFS);
1232         m->m_len = hlen;
1233         if (hdr)
1234                 bcopy(hdr, mtod(m, caddr_t), hlen);
1235
1236         *mp = m;
1237         return (0);
1238 }
1239
1240 /*
1241  * Insert jumbo payload option.
1242  */
1243 static int
1244 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1245 {
1246         struct mbuf *mopt;
1247         u_char *optbuf;
1248         u_int32_t v;
1249
1250 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1251
1252         /*
1253          * If there is no hop-by-hop options header, allocate new one.
1254          * If there is one but it doesn't have enough space to store the
1255          * jumbo payload option, allocate a cluster to store the whole options.
1256          * Otherwise, use it to store the options.
1257          */
1258         if (exthdrs->ip6e_hbh == 0) {
1259                 mopt = m_get(M_NOWAIT, MT_DATA);
1260                 if (mopt == NULL)
1261                         return (ENOBUFS);
1262                 mopt->m_len = JUMBOOPTLEN;
1263                 optbuf = mtod(mopt, u_char *);
1264                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1265                 exthdrs->ip6e_hbh = mopt;
1266         } else {
1267                 struct ip6_hbh *hbh;
1268
1269                 mopt = exthdrs->ip6e_hbh;
1270                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1271                         /*
1272                          * XXX assumption:
1273                          * - exthdrs->ip6e_hbh is not referenced from places
1274                          *   other than exthdrs.
1275                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1276                          */
1277                         int oldoptlen = mopt->m_len;
1278                         struct mbuf *n;
1279
1280                         /*
1281                          * XXX: give up if the whole (new) hbh header does
1282                          * not fit even in an mbuf cluster.
1283                          */
1284                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1285                                 return (ENOBUFS);
1286
1287                         /*
1288                          * As a consequence, we must always prepare a cluster
1289                          * at this point.
1290                          */
1291                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1292                         if (n == NULL)
1293                                 return (ENOBUFS);
1294                         n->m_len = oldoptlen + JUMBOOPTLEN;
1295                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1296                             oldoptlen);
1297                         optbuf = mtod(n, caddr_t) + oldoptlen;
1298                         m_freem(mopt);
1299                         mopt = exthdrs->ip6e_hbh = n;
1300                 } else {
1301                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1302                         mopt->m_len += JUMBOOPTLEN;
1303                 }
1304                 optbuf[0] = IP6OPT_PADN;
1305                 optbuf[1] = 1;
1306
1307                 /*
1308                  * Adjust the header length according to the pad and
1309                  * the jumbo payload option.
1310                  */
1311                 hbh = mtod(mopt, struct ip6_hbh *);
1312                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1313         }
1314
1315         /* fill in the option. */
1316         optbuf[2] = IP6OPT_JUMBO;
1317         optbuf[3] = 4;
1318         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1319         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1320
1321         /* finally, adjust the packet header length */
1322         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1323
1324         return (0);
1325 #undef JUMBOOPTLEN
1326 }
1327
1328 /*
1329  * Insert fragment header and copy unfragmentable header portions.
1330  */
1331 static int
1332 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1333     struct ip6_frag **frghdrp)
1334 {
1335         struct mbuf *n, *mlast;
1336
1337         if (hlen > sizeof(struct ip6_hdr)) {
1338                 n = m_copym(m0, sizeof(struct ip6_hdr),
1339                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1340                 if (n == 0)
1341                         return (ENOBUFS);
1342                 m->m_next = n;
1343         } else
1344                 n = m;
1345
1346         /* Search for the last mbuf of unfragmentable part. */
1347         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1348                 ;
1349
1350         if ((mlast->m_flags & M_EXT) == 0 &&
1351             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1352                 /* use the trailing space of the last mbuf for the fragment hdr */
1353                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1354                     mlast->m_len);
1355                 mlast->m_len += sizeof(struct ip6_frag);
1356                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1357         } else {
1358                 /* allocate a new mbuf for the fragment header */
1359                 struct mbuf *mfrg;
1360
1361                 mfrg = m_get(M_NOWAIT, MT_DATA);
1362                 if (mfrg == NULL)
1363                         return (ENOBUFS);
1364                 mfrg->m_len = sizeof(struct ip6_frag);
1365                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1366                 mlast->m_next = mfrg;
1367         }
1368
1369         return (0);
1370 }
1371
1372 static int
1373 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1374     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1375     int *alwaysfragp, u_int fibnum)
1376 {
1377         u_int32_t mtu = 0;
1378         int alwaysfrag = 0;
1379         int error = 0;
1380
1381         if (ro_pmtu != ro) {
1382                 /* The first hop and the final destination may differ. */
1383                 struct sockaddr_in6 *sa6_dst =
1384                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1385                 if (ro_pmtu->ro_rt &&
1386                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1387                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1388                         RTFREE(ro_pmtu->ro_rt);
1389                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1390                 }
1391                 if (ro_pmtu->ro_rt == NULL) {
1392                         bzero(sa6_dst, sizeof(*sa6_dst));
1393                         sa6_dst->sin6_family = AF_INET6;
1394                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1395                         sa6_dst->sin6_addr = *dst;
1396
1397                         in6_rtalloc(ro_pmtu, fibnum);
1398                 }
1399         }
1400         if (ro_pmtu->ro_rt) {
1401                 u_int32_t ifmtu;
1402                 struct in_conninfo inc;
1403
1404                 bzero(&inc, sizeof(inc));
1405                 inc.inc_flags |= INC_ISIPV6;
1406                 inc.inc6_faddr = *dst;
1407
1408                 if (ifp == NULL)
1409                         ifp = ro_pmtu->ro_rt->rt_ifp;
1410                 ifmtu = IN6_LINKMTU(ifp);
1411                 mtu = tcp_hc_getmtu(&inc);
1412                 if (mtu)
1413                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1414                 else
1415                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1416                 if (mtu == 0)
1417                         mtu = ifmtu;
1418                 else if (mtu < IPV6_MMTU) {
1419                         /*
1420                          * RFC2460 section 5, last paragraph:
1421                          * if we record ICMPv6 too big message with
1422                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1423                          * or smaller, with framgent header attached.
1424                          * (fragment header is needed regardless from the
1425                          * packet size, for translators to identify packets)
1426                          */
1427                         alwaysfrag = 1;
1428                         mtu = IPV6_MMTU;
1429                 } else if (mtu > ifmtu) {
1430                         /*
1431                          * The MTU on the route is larger than the MTU on
1432                          * the interface!  This shouldn't happen, unless the
1433                          * MTU of the interface has been changed after the
1434                          * interface was brought up.  Change the MTU in the
1435                          * route to match the interface MTU (as long as the
1436                          * field isn't locked).
1437                          */
1438                         mtu = ifmtu;
1439                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1440                 }
1441         } else if (ifp) {
1442                 mtu = IN6_LINKMTU(ifp);
1443         } else
1444                 error = EHOSTUNREACH; /* XXX */
1445
1446         *mtup = mtu;
1447         if (alwaysfragp)
1448                 *alwaysfragp = alwaysfrag;
1449         return (error);
1450 }
1451
1452 /*
1453  * IP6 socket option processing.
1454  */
1455 int
1456 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1457 {
1458         int optdatalen, uproto;
1459         void *optdata;
1460         struct inpcb *in6p = sotoinpcb(so);
1461         int error, optval;
1462         int level, op, optname;
1463         int optlen;
1464         struct thread *td;
1465
1466         level = sopt->sopt_level;
1467         op = sopt->sopt_dir;
1468         optname = sopt->sopt_name;
1469         optlen = sopt->sopt_valsize;
1470         td = sopt->sopt_td;
1471         error = 0;
1472         optval = 0;
1473         uproto = (int)so->so_proto->pr_protocol;
1474
1475         if (level != IPPROTO_IPV6) {
1476                 error = EINVAL;
1477
1478                 if (sopt->sopt_level == SOL_SOCKET &&
1479                     sopt->sopt_dir == SOPT_SET) {
1480                         switch (sopt->sopt_name) {
1481                         case SO_REUSEADDR:
1482                                 INP_WLOCK(in6p);
1483                                 if ((so->so_options & SO_REUSEADDR) != 0)
1484                                         in6p->inp_flags2 |= INP_REUSEADDR;
1485                                 else
1486                                         in6p->inp_flags2 &= ~INP_REUSEADDR;
1487                                 INP_WUNLOCK(in6p);
1488                                 error = 0;
1489                                 break;
1490                         case SO_REUSEPORT:
1491                                 INP_WLOCK(in6p);
1492                                 if ((so->so_options & SO_REUSEPORT) != 0)
1493                                         in6p->inp_flags2 |= INP_REUSEPORT;
1494                                 else
1495                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1496                                 INP_WUNLOCK(in6p);
1497                                 error = 0;
1498                                 break;
1499                         case SO_SETFIB:
1500                                 INP_WLOCK(in6p);
1501                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1502                                 INP_WUNLOCK(in6p);
1503                                 error = 0;
1504                                 break;
1505                         default:
1506                                 break;
1507                         }
1508                 }
1509         } else {                /* level == IPPROTO_IPV6 */
1510                 switch (op) {
1511
1512                 case SOPT_SET:
1513                         switch (optname) {
1514                         case IPV6_2292PKTOPTIONS:
1515 #ifdef IPV6_PKTOPTIONS
1516                         case IPV6_PKTOPTIONS:
1517 #endif
1518                         {
1519                                 struct mbuf *m;
1520
1521                                 error = soopt_getm(sopt, &m); /* XXX */
1522                                 if (error != 0)
1523                                         break;
1524                                 error = soopt_mcopyin(sopt, m); /* XXX */
1525                                 if (error != 0)
1526                                         break;
1527                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1528                                                     m, so, sopt);
1529                                 m_freem(m); /* XXX */
1530                                 break;
1531                         }
1532
1533                         /*
1534                          * Use of some Hop-by-Hop options or some
1535                          * Destination options, might require special
1536                          * privilege.  That is, normal applications
1537                          * (without special privilege) might be forbidden
1538                          * from setting certain options in outgoing packets,
1539                          * and might never see certain options in received
1540                          * packets. [RFC 2292 Section 6]
1541                          * KAME specific note:
1542                          *  KAME prevents non-privileged users from sending or
1543                          *  receiving ANY hbh/dst options in order to avoid
1544                          *  overhead of parsing options in the kernel.
1545                          */
1546                         case IPV6_RECVHOPOPTS:
1547                         case IPV6_RECVDSTOPTS:
1548                         case IPV6_RECVRTHDRDSTOPTS:
1549                                 if (td != NULL) {
1550                                         error = priv_check(td,
1551                                             PRIV_NETINET_SETHDROPTS);
1552                                         if (error)
1553                                                 break;
1554                                 }
1555                                 /* FALLTHROUGH */
1556                         case IPV6_UNICAST_HOPS:
1557                         case IPV6_HOPLIMIT:
1558                         case IPV6_FAITH:
1559
1560                         case IPV6_RECVPKTINFO:
1561                         case IPV6_RECVHOPLIMIT:
1562                         case IPV6_RECVRTHDR:
1563                         case IPV6_RECVPATHMTU:
1564                         case IPV6_RECVTCLASS:
1565                         case IPV6_V6ONLY:
1566                         case IPV6_AUTOFLOWLABEL:
1567                         case IPV6_BINDANY:
1568                                 if (optname == IPV6_BINDANY && td != NULL) {
1569                                         error = priv_check(td,
1570                                             PRIV_NETINET_BINDANY);
1571                                         if (error)
1572                                                 break;
1573                                 }
1574
1575                                 if (optlen != sizeof(int)) {
1576                                         error = EINVAL;
1577                                         break;
1578                                 }
1579                                 error = sooptcopyin(sopt, &optval,
1580                                         sizeof optval, sizeof optval);
1581                                 if (error)
1582                                         break;
1583                                 switch (optname) {
1584
1585                                 case IPV6_UNICAST_HOPS:
1586                                         if (optval < -1 || optval >= 256)
1587                                                 error = EINVAL;
1588                                         else {
1589                                                 /* -1 = kernel default */
1590                                                 in6p->in6p_hops = optval;
1591                                                 if ((in6p->inp_vflag &
1592                                                      INP_IPV4) != 0)
1593                                                         in6p->inp_ip_ttl = optval;
1594                                         }
1595                                         break;
1596 #define OPTSET(bit) \
1597 do { \
1598         INP_WLOCK(in6p); \
1599         if (optval) \
1600                 in6p->inp_flags |= (bit); \
1601         else \
1602                 in6p->inp_flags &= ~(bit); \
1603         INP_WUNLOCK(in6p); \
1604 } while (/*CONSTCOND*/ 0)
1605 #define OPTSET2292(bit) \
1606 do { \
1607         INP_WLOCK(in6p); \
1608         in6p->inp_flags |= IN6P_RFC2292; \
1609         if (optval) \
1610                 in6p->inp_flags |= (bit); \
1611         else \
1612                 in6p->inp_flags &= ~(bit); \
1613         INP_WUNLOCK(in6p); \
1614 } while (/*CONSTCOND*/ 0)
1615 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1616
1617                                 case IPV6_RECVPKTINFO:
1618                                         /* cannot mix with RFC2292 */
1619                                         if (OPTBIT(IN6P_RFC2292)) {
1620                                                 error = EINVAL;
1621                                                 break;
1622                                         }
1623                                         OPTSET(IN6P_PKTINFO);
1624                                         break;
1625
1626                                 case IPV6_HOPLIMIT:
1627                                 {
1628                                         struct ip6_pktopts **optp;
1629
1630                                         /* cannot mix with RFC2292 */
1631                                         if (OPTBIT(IN6P_RFC2292)) {
1632                                                 error = EINVAL;
1633                                                 break;
1634                                         }
1635                                         optp = &in6p->in6p_outputopts;
1636                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1637                                             (u_char *)&optval, sizeof(optval),
1638                                             optp, (td != NULL) ? td->td_ucred :
1639                                             NULL, uproto);
1640                                         break;
1641                                 }
1642
1643                                 case IPV6_RECVHOPLIMIT:
1644                                         /* cannot mix with RFC2292 */
1645                                         if (OPTBIT(IN6P_RFC2292)) {
1646                                                 error = EINVAL;
1647                                                 break;
1648                                         }
1649                                         OPTSET(IN6P_HOPLIMIT);
1650                                         break;
1651
1652                                 case IPV6_RECVHOPOPTS:
1653                                         /* cannot mix with RFC2292 */
1654                                         if (OPTBIT(IN6P_RFC2292)) {
1655                                                 error = EINVAL;
1656                                                 break;
1657                                         }
1658                                         OPTSET(IN6P_HOPOPTS);
1659                                         break;
1660
1661                                 case IPV6_RECVDSTOPTS:
1662                                         /* cannot mix with RFC2292 */
1663                                         if (OPTBIT(IN6P_RFC2292)) {
1664                                                 error = EINVAL;
1665                                                 break;
1666                                         }
1667                                         OPTSET(IN6P_DSTOPTS);
1668                                         break;
1669
1670                                 case IPV6_RECVRTHDRDSTOPTS:
1671                                         /* cannot mix with RFC2292 */
1672                                         if (OPTBIT(IN6P_RFC2292)) {
1673                                                 error = EINVAL;
1674                                                 break;
1675                                         }
1676                                         OPTSET(IN6P_RTHDRDSTOPTS);
1677                                         break;
1678
1679                                 case IPV6_RECVRTHDR:
1680                                         /* cannot mix with RFC2292 */
1681                                         if (OPTBIT(IN6P_RFC2292)) {
1682                                                 error = EINVAL;
1683                                                 break;
1684                                         }
1685                                         OPTSET(IN6P_RTHDR);
1686                                         break;
1687
1688                                 case IPV6_FAITH:
1689                                         OPTSET(INP_FAITH);
1690                                         break;
1691
1692                                 case IPV6_RECVPATHMTU:
1693                                         /*
1694                                          * We ignore this option for TCP
1695                                          * sockets.
1696                                          * (RFC3542 leaves this case
1697                                          * unspecified.)
1698                                          */
1699                                         if (uproto != IPPROTO_TCP)
1700                                                 OPTSET(IN6P_MTU);
1701                                         break;
1702
1703                                 case IPV6_V6ONLY:
1704                                         /*
1705                                          * make setsockopt(IPV6_V6ONLY)
1706                                          * available only prior to bind(2).
1707                                          * see ipng mailing list, Jun 22 2001.
1708                                          */
1709                                         if (in6p->inp_lport ||
1710                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1711                                                 error = EINVAL;
1712                                                 break;
1713                                         }
1714                                         OPTSET(IN6P_IPV6_V6ONLY);
1715                                         if (optval)
1716                                                 in6p->inp_vflag &= ~INP_IPV4;
1717                                         else
1718                                                 in6p->inp_vflag |= INP_IPV4;
1719                                         break;
1720                                 case IPV6_RECVTCLASS:
1721                                         /* cannot mix with RFC2292 XXX */
1722                                         if (OPTBIT(IN6P_RFC2292)) {
1723                                                 error = EINVAL;
1724                                                 break;
1725                                         }
1726                                         OPTSET(IN6P_TCLASS);
1727                                         break;
1728                                 case IPV6_AUTOFLOWLABEL:
1729                                         OPTSET(IN6P_AUTOFLOWLABEL);
1730                                         break;
1731
1732                                 case IPV6_BINDANY:
1733                                         OPTSET(INP_BINDANY);
1734                                         break;
1735                                 }
1736                                 break;
1737
1738                         case IPV6_TCLASS:
1739                         case IPV6_DONTFRAG:
1740                         case IPV6_USE_MIN_MTU:
1741                         case IPV6_PREFER_TEMPADDR:
1742                                 if (optlen != sizeof(optval)) {
1743                                         error = EINVAL;
1744                                         break;
1745                                 }
1746                                 error = sooptcopyin(sopt, &optval,
1747                                         sizeof optval, sizeof optval);
1748                                 if (error)
1749                                         break;
1750                                 {
1751                                         struct ip6_pktopts **optp;
1752                                         optp = &in6p->in6p_outputopts;
1753                                         error = ip6_pcbopt(optname,
1754                                             (u_char *)&optval, sizeof(optval),
1755                                             optp, (td != NULL) ? td->td_ucred :
1756                                             NULL, uproto);
1757                                         break;
1758                                 }
1759
1760                         case IPV6_2292PKTINFO:
1761                         case IPV6_2292HOPLIMIT:
1762                         case IPV6_2292HOPOPTS:
1763                         case IPV6_2292DSTOPTS:
1764                         case IPV6_2292RTHDR:
1765                                 /* RFC 2292 */
1766                                 if (optlen != sizeof(int)) {
1767                                         error = EINVAL;
1768                                         break;
1769                                 }
1770                                 error = sooptcopyin(sopt, &optval,
1771                                         sizeof optval, sizeof optval);
1772                                 if (error)
1773                                         break;
1774                                 switch (optname) {
1775                                 case IPV6_2292PKTINFO:
1776                                         OPTSET2292(IN6P_PKTINFO);
1777                                         break;
1778                                 case IPV6_2292HOPLIMIT:
1779                                         OPTSET2292(IN6P_HOPLIMIT);
1780                                         break;
1781                                 case IPV6_2292HOPOPTS:
1782                                         /*
1783                                          * Check super-user privilege.
1784                                          * See comments for IPV6_RECVHOPOPTS.
1785                                          */
1786                                         if (td != NULL) {
1787                                                 error = priv_check(td,
1788                                                     PRIV_NETINET_SETHDROPTS);
1789                                                 if (error)
1790                                                         return (error);
1791                                         }
1792                                         OPTSET2292(IN6P_HOPOPTS);
1793                                         break;
1794                                 case IPV6_2292DSTOPTS:
1795                                         if (td != NULL) {
1796                                                 error = priv_check(td,
1797                                                     PRIV_NETINET_SETHDROPTS);
1798                                                 if (error)
1799                                                         return (error);
1800                                         }
1801                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1802                                         break;
1803                                 case IPV6_2292RTHDR:
1804                                         OPTSET2292(IN6P_RTHDR);
1805                                         break;
1806                                 }
1807                                 break;
1808                         case IPV6_PKTINFO:
1809                         case IPV6_HOPOPTS:
1810                         case IPV6_RTHDR:
1811                         case IPV6_DSTOPTS:
1812                         case IPV6_RTHDRDSTOPTS:
1813                         case IPV6_NEXTHOP:
1814                         {
1815                                 /* new advanced API (RFC3542) */
1816                                 u_char *optbuf;
1817                                 u_char optbuf_storage[MCLBYTES];
1818                                 int optlen;
1819                                 struct ip6_pktopts **optp;
1820
1821                                 /* cannot mix with RFC2292 */
1822                                 if (OPTBIT(IN6P_RFC2292)) {
1823                                         error = EINVAL;
1824                                         break;
1825                                 }
1826
1827                                 /*
1828                                  * We only ensure valsize is not too large
1829                                  * here.  Further validation will be done
1830                                  * later.
1831                                  */
1832                                 error = sooptcopyin(sopt, optbuf_storage,
1833                                     sizeof(optbuf_storage), 0);
1834                                 if (error)
1835                                         break;
1836                                 optlen = sopt->sopt_valsize;
1837                                 optbuf = optbuf_storage;
1838                                 optp = &in6p->in6p_outputopts;
1839                                 error = ip6_pcbopt(optname, optbuf, optlen,
1840                                     optp, (td != NULL) ? td->td_ucred : NULL,
1841                                     uproto);
1842                                 break;
1843                         }
1844 #undef OPTSET
1845
1846                         case IPV6_MULTICAST_IF:
1847                         case IPV6_MULTICAST_HOPS:
1848                         case IPV6_MULTICAST_LOOP:
1849                         case IPV6_JOIN_GROUP:
1850                         case IPV6_LEAVE_GROUP:
1851                         case IPV6_MSFILTER:
1852                         case MCAST_BLOCK_SOURCE:
1853                         case MCAST_UNBLOCK_SOURCE:
1854                         case MCAST_JOIN_GROUP:
1855                         case MCAST_LEAVE_GROUP:
1856                         case MCAST_JOIN_SOURCE_GROUP:
1857                         case MCAST_LEAVE_SOURCE_GROUP:
1858                                 error = ip6_setmoptions(in6p, sopt);
1859                                 break;
1860
1861                         case IPV6_PORTRANGE:
1862                                 error = sooptcopyin(sopt, &optval,
1863                                     sizeof optval, sizeof optval);
1864                                 if (error)
1865                                         break;
1866
1867                                 INP_WLOCK(in6p);
1868                                 switch (optval) {
1869                                 case IPV6_PORTRANGE_DEFAULT:
1870                                         in6p->inp_flags &= ~(INP_LOWPORT);
1871                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1872                                         break;
1873
1874                                 case IPV6_PORTRANGE_HIGH:
1875                                         in6p->inp_flags &= ~(INP_LOWPORT);
1876                                         in6p->inp_flags |= INP_HIGHPORT;
1877                                         break;
1878
1879                                 case IPV6_PORTRANGE_LOW:
1880                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1881                                         in6p->inp_flags |= INP_LOWPORT;
1882                                         break;
1883
1884                                 default:
1885                                         error = EINVAL;
1886                                         break;
1887                                 }
1888                                 INP_WUNLOCK(in6p);
1889                                 break;
1890
1891 #ifdef IPSEC
1892                         case IPV6_IPSEC_POLICY:
1893                         {
1894                                 caddr_t req;
1895                                 struct mbuf *m;
1896
1897                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1898                                         break;
1899                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1900                                         break;
1901                                 req = mtod(m, caddr_t);
1902                                 error = ipsec_set_policy(in6p, optname, req,
1903                                     m->m_len, (sopt->sopt_td != NULL) ?
1904                                     sopt->sopt_td->td_ucred : NULL);
1905                                 m_freem(m);
1906                                 break;
1907                         }
1908 #endif /* IPSEC */
1909
1910                         default:
1911                                 error = ENOPROTOOPT;
1912                                 break;
1913                         }
1914                         break;
1915
1916                 case SOPT_GET:
1917                         switch (optname) {
1918
1919                         case IPV6_2292PKTOPTIONS:
1920 #ifdef IPV6_PKTOPTIONS
1921                         case IPV6_PKTOPTIONS:
1922 #endif
1923                                 /*
1924                                  * RFC3542 (effectively) deprecated the
1925                                  * semantics of the 2292-style pktoptions.
1926                                  * Since it was not reliable in nature (i.e.,
1927                                  * applications had to expect the lack of some
1928                                  * information after all), it would make sense
1929                                  * to simplify this part by always returning
1930                                  * empty data.
1931                                  */
1932                                 sopt->sopt_valsize = 0;
1933                                 break;
1934
1935                         case IPV6_RECVHOPOPTS:
1936                         case IPV6_RECVDSTOPTS:
1937                         case IPV6_RECVRTHDRDSTOPTS:
1938                         case IPV6_UNICAST_HOPS:
1939                         case IPV6_RECVPKTINFO:
1940                         case IPV6_RECVHOPLIMIT:
1941                         case IPV6_RECVRTHDR:
1942                         case IPV6_RECVPATHMTU:
1943
1944                         case IPV6_FAITH:
1945                         case IPV6_V6ONLY:
1946                         case IPV6_PORTRANGE:
1947                         case IPV6_RECVTCLASS:
1948                         case IPV6_AUTOFLOWLABEL:
1949                         case IPV6_BINDANY:
1950                                 switch (optname) {
1951
1952                                 case IPV6_RECVHOPOPTS:
1953                                         optval = OPTBIT(IN6P_HOPOPTS);
1954                                         break;
1955
1956                                 case IPV6_RECVDSTOPTS:
1957                                         optval = OPTBIT(IN6P_DSTOPTS);
1958                                         break;
1959
1960                                 case IPV6_RECVRTHDRDSTOPTS:
1961                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1962                                         break;
1963
1964                                 case IPV6_UNICAST_HOPS:
1965                                         optval = in6p->in6p_hops;
1966                                         break;
1967
1968                                 case IPV6_RECVPKTINFO:
1969                                         optval = OPTBIT(IN6P_PKTINFO);
1970                                         break;
1971
1972                                 case IPV6_RECVHOPLIMIT:
1973                                         optval = OPTBIT(IN6P_HOPLIMIT);
1974                                         break;
1975
1976                                 case IPV6_RECVRTHDR:
1977                                         optval = OPTBIT(IN6P_RTHDR);
1978                                         break;
1979
1980                                 case IPV6_RECVPATHMTU:
1981                                         optval = OPTBIT(IN6P_MTU);
1982                                         break;
1983
1984                                 case IPV6_FAITH:
1985                                         optval = OPTBIT(INP_FAITH);
1986                                         break;
1987
1988                                 case IPV6_V6ONLY:
1989                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1990                                         break;
1991
1992                                 case IPV6_PORTRANGE:
1993                                     {
1994                                         int flags;
1995                                         flags = in6p->inp_flags;
1996                                         if (flags & INP_HIGHPORT)
1997                                                 optval = IPV6_PORTRANGE_HIGH;
1998                                         else if (flags & INP_LOWPORT)
1999                                                 optval = IPV6_PORTRANGE_LOW;
2000                                         else
2001                                                 optval = 0;
2002                                         break;
2003                                     }
2004                                 case IPV6_RECVTCLASS:
2005                                         optval = OPTBIT(IN6P_TCLASS);
2006                                         break;
2007
2008                                 case IPV6_AUTOFLOWLABEL:
2009                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2010                                         break;
2011
2012                                 case IPV6_BINDANY:
2013                                         optval = OPTBIT(INP_BINDANY);
2014                                         break;
2015                                 }
2016                                 if (error)
2017                                         break;
2018                                 error = sooptcopyout(sopt, &optval,
2019                                         sizeof optval);
2020                                 break;
2021
2022                         case IPV6_PATHMTU:
2023                         {
2024                                 u_long pmtu = 0;
2025                                 struct ip6_mtuinfo mtuinfo;
2026                                 struct route_in6 sro;
2027
2028                                 bzero(&sro, sizeof(sro));
2029
2030                                 if (!(so->so_state & SS_ISCONNECTED))
2031                                         return (ENOTCONN);
2032                                 /*
2033                                  * XXX: we dot not consider the case of source
2034                                  * routing, or optional information to specify
2035                                  * the outgoing interface.
2036                                  */
2037                                 error = ip6_getpmtu(&sro, NULL, NULL,
2038                                     &in6p->in6p_faddr, &pmtu, NULL,
2039                                     so->so_fibnum);
2040                                 if (sro.ro_rt)
2041                                         RTFREE(sro.ro_rt);
2042                                 if (error)
2043                                         break;
2044                                 if (pmtu > IPV6_MAXPACKET)
2045                                         pmtu = IPV6_MAXPACKET;
2046
2047                                 bzero(&mtuinfo, sizeof(mtuinfo));
2048                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2049                                 optdata = (void *)&mtuinfo;
2050                                 optdatalen = sizeof(mtuinfo);
2051                                 error = sooptcopyout(sopt, optdata,
2052                                     optdatalen);
2053                                 break;
2054                         }
2055
2056                         case IPV6_2292PKTINFO:
2057                         case IPV6_2292HOPLIMIT:
2058                         case IPV6_2292HOPOPTS:
2059                         case IPV6_2292RTHDR:
2060                         case IPV6_2292DSTOPTS:
2061                                 switch (optname) {
2062                                 case IPV6_2292PKTINFO:
2063                                         optval = OPTBIT(IN6P_PKTINFO);
2064                                         break;
2065                                 case IPV6_2292HOPLIMIT:
2066                                         optval = OPTBIT(IN6P_HOPLIMIT);
2067                                         break;
2068                                 case IPV6_2292HOPOPTS:
2069                                         optval = OPTBIT(IN6P_HOPOPTS);
2070                                         break;
2071                                 case IPV6_2292RTHDR:
2072                                         optval = OPTBIT(IN6P_RTHDR);
2073                                         break;
2074                                 case IPV6_2292DSTOPTS:
2075                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2076                                         break;
2077                                 }
2078                                 error = sooptcopyout(sopt, &optval,
2079                                     sizeof optval);
2080                                 break;
2081                         case IPV6_PKTINFO:
2082                         case IPV6_HOPOPTS:
2083                         case IPV6_RTHDR:
2084                         case IPV6_DSTOPTS:
2085                         case IPV6_RTHDRDSTOPTS:
2086                         case IPV6_NEXTHOP:
2087                         case IPV6_TCLASS:
2088                         case IPV6_DONTFRAG:
2089                         case IPV6_USE_MIN_MTU:
2090                         case IPV6_PREFER_TEMPADDR:
2091                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2092                                     optname, sopt);
2093                                 break;
2094
2095                         case IPV6_MULTICAST_IF:
2096                         case IPV6_MULTICAST_HOPS:
2097                         case IPV6_MULTICAST_LOOP:
2098                         case IPV6_MSFILTER:
2099                                 error = ip6_getmoptions(in6p, sopt);
2100                                 break;
2101
2102 #ifdef IPSEC
2103                         case IPV6_IPSEC_POLICY:
2104                           {
2105                                 caddr_t req = NULL;
2106                                 size_t len = 0;
2107                                 struct mbuf *m = NULL;
2108                                 struct mbuf **mp = &m;
2109                                 size_t ovalsize = sopt->sopt_valsize;
2110                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2111
2112                                 error = soopt_getm(sopt, &m); /* XXX */
2113                                 if (error != 0)
2114                                         break;
2115                                 error = soopt_mcopyin(sopt, m); /* XXX */
2116                                 if (error != 0)
2117                                         break;
2118                                 sopt->sopt_valsize = ovalsize;
2119                                 sopt->sopt_val = oval;
2120                                 if (m) {
2121                                         req = mtod(m, caddr_t);
2122                                         len = m->m_len;
2123                                 }
2124                                 error = ipsec_get_policy(in6p, req, len, mp);
2125                                 if (error == 0)
2126                                         error = soopt_mcopyout(sopt, m); /* XXX */
2127                                 if (error == 0 && m)
2128                                         m_freem(m);
2129                                 break;
2130                           }
2131 #endif /* IPSEC */
2132
2133                         default:
2134                                 error = ENOPROTOOPT;
2135                                 break;
2136                         }
2137                         break;
2138                 }
2139         }
2140         return (error);
2141 }
2142
2143 int
2144 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2145 {
2146         int error = 0, optval, optlen;
2147         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2148         struct inpcb *in6p = sotoinpcb(so);
2149         int level, op, optname;
2150
2151         level = sopt->sopt_level;
2152         op = sopt->sopt_dir;
2153         optname = sopt->sopt_name;
2154         optlen = sopt->sopt_valsize;
2155
2156         if (level != IPPROTO_IPV6) {
2157                 return (EINVAL);
2158         }
2159
2160         switch (optname) {
2161         case IPV6_CHECKSUM:
2162                 /*
2163                  * For ICMPv6 sockets, no modification allowed for checksum
2164                  * offset, permit "no change" values to help existing apps.
2165                  *
2166                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2167                  * for an ICMPv6 socket will fail."
2168                  * The current behavior does not meet RFC3542.
2169                  */
2170                 switch (op) {
2171                 case SOPT_SET:
2172                         if (optlen != sizeof(int)) {
2173                                 error = EINVAL;
2174                                 break;
2175                         }
2176                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2177                                             sizeof(optval));
2178                         if (error)
2179                                 break;
2180                         if ((optval % 2) != 0) {
2181                                 /* the API assumes even offset values */
2182                                 error = EINVAL;
2183                         } else if (so->so_proto->pr_protocol ==
2184                             IPPROTO_ICMPV6) {
2185                                 if (optval != icmp6off)
2186                                         error = EINVAL;
2187                         } else
2188                                 in6p->in6p_cksum = optval;
2189                         break;
2190
2191                 case SOPT_GET:
2192                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2193                                 optval = icmp6off;
2194                         else
2195                                 optval = in6p->in6p_cksum;
2196
2197                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2198                         break;
2199
2200                 default:
2201                         error = EINVAL;
2202                         break;
2203                 }
2204                 break;
2205
2206         default:
2207                 error = ENOPROTOOPT;
2208                 break;
2209         }
2210
2211         return (error);
2212 }
2213
2214 /*
2215  * Set up IP6 options in pcb for insertion in output packets or
2216  * specifying behavior of outgoing packets.
2217  */
2218 static int
2219 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2220     struct socket *so, struct sockopt *sopt)
2221 {
2222         struct ip6_pktopts *opt = *pktopt;
2223         int error = 0;
2224         struct thread *td = sopt->sopt_td;
2225
2226         /* turn off any old options. */
2227         if (opt) {
2228 #ifdef DIAGNOSTIC
2229                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2230                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2231                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2232                         printf("ip6_pcbopts: all specified options are cleared.\n");
2233 #endif
2234                 ip6_clearpktopts(opt, -1);
2235         } else
2236                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2237         *pktopt = NULL;
2238
2239         if (!m || m->m_len == 0) {
2240                 /*
2241                  * Only turning off any previous options, regardless of
2242                  * whether the opt is just created or given.
2243                  */
2244                 free(opt, M_IP6OPT);
2245                 return (0);
2246         }
2247
2248         /*  set options specified by user. */
2249         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2250             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2251                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2252                 free(opt, M_IP6OPT);
2253                 return (error);
2254         }
2255         *pktopt = opt;
2256         return (0);
2257 }
2258
2259 /*
2260  * initialize ip6_pktopts.  beware that there are non-zero default values in
2261  * the struct.
2262  */
2263 void
2264 ip6_initpktopts(struct ip6_pktopts *opt)
2265 {
2266
2267         bzero(opt, sizeof(*opt));
2268         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2269         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2270         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2271         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2272 }
2273
2274 static int
2275 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2276     struct ucred *cred, int uproto)
2277 {
2278         struct ip6_pktopts *opt;
2279
2280         if (*pktopt == NULL) {
2281                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2282                     M_WAITOK);
2283                 ip6_initpktopts(*pktopt);
2284         }
2285         opt = *pktopt;
2286
2287         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2288 }
2289
2290 static int
2291 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2292 {
2293         void *optdata = NULL;
2294         int optdatalen = 0;
2295         struct ip6_ext *ip6e;
2296         int error = 0;
2297         struct in6_pktinfo null_pktinfo;
2298         int deftclass = 0, on;
2299         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2300         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2301
2302         switch (optname) {
2303         case IPV6_PKTINFO:
2304                 if (pktopt && pktopt->ip6po_pktinfo)
2305                         optdata = (void *)pktopt->ip6po_pktinfo;
2306                 else {
2307                         /* XXX: we don't have to do this every time... */
2308                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2309                         optdata = (void *)&null_pktinfo;
2310                 }
2311                 optdatalen = sizeof(struct in6_pktinfo);
2312                 break;
2313         case IPV6_TCLASS:
2314                 if (pktopt && pktopt->ip6po_tclass >= 0)
2315                         optdata = (void *)&pktopt->ip6po_tclass;
2316                 else
2317                         optdata = (void *)&deftclass;
2318                 optdatalen = sizeof(int);
2319                 break;
2320         case IPV6_HOPOPTS:
2321                 if (pktopt && pktopt->ip6po_hbh) {
2322                         optdata = (void *)pktopt->ip6po_hbh;
2323                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2324                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2325                 }
2326                 break;
2327         case IPV6_RTHDR:
2328                 if (pktopt && pktopt->ip6po_rthdr) {
2329                         optdata = (void *)pktopt->ip6po_rthdr;
2330                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2331                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2332                 }
2333                 break;
2334         case IPV6_RTHDRDSTOPTS:
2335                 if (pktopt && pktopt->ip6po_dest1) {
2336                         optdata = (void *)pktopt->ip6po_dest1;
2337                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2338                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2339                 }
2340                 break;
2341         case IPV6_DSTOPTS:
2342                 if (pktopt && pktopt->ip6po_dest2) {
2343                         optdata = (void *)pktopt->ip6po_dest2;
2344                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2345                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2346                 }
2347                 break;
2348         case IPV6_NEXTHOP:
2349                 if (pktopt && pktopt->ip6po_nexthop) {
2350                         optdata = (void *)pktopt->ip6po_nexthop;
2351                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2352                 }
2353                 break;
2354         case IPV6_USE_MIN_MTU:
2355                 if (pktopt)
2356                         optdata = (void *)&pktopt->ip6po_minmtu;
2357                 else
2358                         optdata = (void *)&defminmtu;
2359                 optdatalen = sizeof(int);
2360                 break;
2361         case IPV6_DONTFRAG:
2362                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2363                         on = 1;
2364                 else
2365                         on = 0;
2366                 optdata = (void *)&on;
2367                 optdatalen = sizeof(on);
2368                 break;
2369         case IPV6_PREFER_TEMPADDR:
2370                 if (pktopt)
2371                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2372                 else
2373                         optdata = (void *)&defpreftemp;
2374                 optdatalen = sizeof(int);
2375                 break;
2376         default:                /* should not happen */
2377 #ifdef DIAGNOSTIC
2378                 panic("ip6_getpcbopt: unexpected option\n");
2379 #endif
2380                 return (ENOPROTOOPT);
2381         }
2382
2383         error = sooptcopyout(sopt, optdata, optdatalen);
2384
2385         return (error);
2386 }
2387
2388 void
2389 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2390 {
2391         if (pktopt == NULL)
2392                 return;
2393
2394         if (optname == -1 || optname == IPV6_PKTINFO) {
2395                 if (pktopt->ip6po_pktinfo)
2396                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2397                 pktopt->ip6po_pktinfo = NULL;
2398         }
2399         if (optname == -1 || optname == IPV6_HOPLIMIT)
2400                 pktopt->ip6po_hlim = -1;
2401         if (optname == -1 || optname == IPV6_TCLASS)
2402                 pktopt->ip6po_tclass = -1;
2403         if (optname == -1 || optname == IPV6_NEXTHOP) {
2404                 if (pktopt->ip6po_nextroute.ro_rt) {
2405                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2406                         pktopt->ip6po_nextroute.ro_rt = NULL;
2407                 }
2408                 if (pktopt->ip6po_nexthop)
2409                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2410                 pktopt->ip6po_nexthop = NULL;
2411         }
2412         if (optname == -1 || optname == IPV6_HOPOPTS) {
2413                 if (pktopt->ip6po_hbh)
2414                         free(pktopt->ip6po_hbh, M_IP6OPT);
2415                 pktopt->ip6po_hbh = NULL;
2416         }
2417         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2418                 if (pktopt->ip6po_dest1)
2419                         free(pktopt->ip6po_dest1, M_IP6OPT);
2420                 pktopt->ip6po_dest1 = NULL;
2421         }
2422         if (optname == -1 || optname == IPV6_RTHDR) {
2423                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2424                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2425                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2426                 if (pktopt->ip6po_route.ro_rt) {
2427                         RTFREE(pktopt->ip6po_route.ro_rt);
2428                         pktopt->ip6po_route.ro_rt = NULL;
2429                 }
2430         }
2431         if (optname == -1 || optname == IPV6_DSTOPTS) {
2432                 if (pktopt->ip6po_dest2)
2433                         free(pktopt->ip6po_dest2, M_IP6OPT);
2434                 pktopt->ip6po_dest2 = NULL;
2435         }
2436 }
2437
2438 #define PKTOPT_EXTHDRCPY(type) \
2439 do {\
2440         if (src->type) {\
2441                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2442                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2443                 if (dst->type == NULL && canwait == M_NOWAIT)\
2444                         goto bad;\
2445                 bcopy(src->type, dst->type, hlen);\
2446         }\
2447 } while (/*CONSTCOND*/ 0)
2448
2449 static int
2450 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2451 {
2452         if (dst == NULL || src == NULL)  {
2453                 printf("ip6_clearpktopts: invalid argument\n");
2454                 return (EINVAL);
2455         }
2456
2457         dst->ip6po_hlim = src->ip6po_hlim;
2458         dst->ip6po_tclass = src->ip6po_tclass;
2459         dst->ip6po_flags = src->ip6po_flags;
2460         dst->ip6po_minmtu = src->ip6po_minmtu;
2461         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2462         if (src->ip6po_pktinfo) {
2463                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2464                     M_IP6OPT, canwait);
2465                 if (dst->ip6po_pktinfo == NULL)
2466                         goto bad;
2467                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2468         }
2469         if (src->ip6po_nexthop) {
2470                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2471                     M_IP6OPT, canwait);
2472                 if (dst->ip6po_nexthop == NULL)
2473                         goto bad;
2474                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2475                     src->ip6po_nexthop->sa_len);
2476         }
2477         PKTOPT_EXTHDRCPY(ip6po_hbh);
2478         PKTOPT_EXTHDRCPY(ip6po_dest1);
2479         PKTOPT_EXTHDRCPY(ip6po_dest2);
2480         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2481         return (0);
2482
2483   bad:
2484         ip6_clearpktopts(dst, -1);
2485         return (ENOBUFS);
2486 }
2487 #undef PKTOPT_EXTHDRCPY
2488
2489 struct ip6_pktopts *
2490 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2491 {
2492         int error;
2493         struct ip6_pktopts *dst;
2494
2495         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2496         if (dst == NULL)
2497                 return (NULL);
2498         ip6_initpktopts(dst);
2499
2500         if ((error = copypktopts(dst, src, canwait)) != 0) {
2501                 free(dst, M_IP6OPT);
2502                 return (NULL);
2503         }
2504
2505         return (dst);
2506 }
2507
2508 void
2509 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2510 {
2511         if (pktopt == NULL)
2512                 return;
2513
2514         ip6_clearpktopts(pktopt, -1);
2515
2516         free(pktopt, M_IP6OPT);
2517 }
2518
2519 /*
2520  * Set IPv6 outgoing packet options based on advanced API.
2521  */
2522 int
2523 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2524     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2525 {
2526         struct cmsghdr *cm = 0;
2527
2528         if (control == NULL || opt == NULL)
2529                 return (EINVAL);
2530
2531         ip6_initpktopts(opt);
2532         if (stickyopt) {
2533                 int error;
2534
2535                 /*
2536                  * If stickyopt is provided, make a local copy of the options
2537                  * for this particular packet, then override them by ancillary
2538                  * objects.
2539                  * XXX: copypktopts() does not copy the cached route to a next
2540                  * hop (if any).  This is not very good in terms of efficiency,
2541                  * but we can allow this since this option should be rarely
2542                  * used.
2543                  */
2544                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2545                         return (error);
2546         }
2547
2548         /*
2549          * XXX: Currently, we assume all the optional information is stored
2550          * in a single mbuf.
2551          */
2552         if (control->m_next)
2553                 return (EINVAL);
2554
2555         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2556             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2557                 int error;
2558
2559                 if (control->m_len < CMSG_LEN(0))
2560                         return (EINVAL);
2561
2562                 cm = mtod(control, struct cmsghdr *);
2563                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2564                         return (EINVAL);
2565                 if (cm->cmsg_level != IPPROTO_IPV6)
2566                         continue;
2567
2568                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2569                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2570                 if (error)
2571                         return (error);
2572         }
2573
2574         return (0);
2575 }
2576
2577 /*
2578  * Set a particular packet option, as a sticky option or an ancillary data
2579  * item.  "len" can be 0 only when it's a sticky option.
2580  * We have 4 cases of combination of "sticky" and "cmsg":
2581  * "sticky=0, cmsg=0": impossible
2582  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2583  * "sticky=1, cmsg=0": RFC3542 socket option
2584  * "sticky=1, cmsg=1": RFC2292 socket option
2585  */
2586 static int
2587 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2588     struct ucred *cred, int sticky, int cmsg, int uproto)
2589 {
2590         int minmtupolicy, preftemp;
2591         int error;
2592
2593         if (!sticky && !cmsg) {
2594 #ifdef DIAGNOSTIC
2595                 printf("ip6_setpktopt: impossible case\n");
2596 #endif
2597                 return (EINVAL);
2598         }
2599
2600         /*
2601          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2602          * not be specified in the context of RFC3542.  Conversely,
2603          * RFC3542 types should not be specified in the context of RFC2292.
2604          */
2605         if (!cmsg) {
2606                 switch (optname) {
2607                 case IPV6_2292PKTINFO:
2608                 case IPV6_2292HOPLIMIT:
2609                 case IPV6_2292NEXTHOP:
2610                 case IPV6_2292HOPOPTS:
2611                 case IPV6_2292DSTOPTS:
2612                 case IPV6_2292RTHDR:
2613                 case IPV6_2292PKTOPTIONS:
2614                         return (ENOPROTOOPT);
2615                 }
2616         }
2617         if (sticky && cmsg) {
2618                 switch (optname) {
2619                 case IPV6_PKTINFO:
2620                 case IPV6_HOPLIMIT:
2621                 case IPV6_NEXTHOP:
2622                 case IPV6_HOPOPTS:
2623                 case IPV6_DSTOPTS:
2624                 case IPV6_RTHDRDSTOPTS:
2625                 case IPV6_RTHDR:
2626                 case IPV6_USE_MIN_MTU:
2627                 case IPV6_DONTFRAG:
2628                 case IPV6_TCLASS:
2629                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2630                         return (ENOPROTOOPT);
2631                 }
2632         }
2633
2634         switch (optname) {
2635         case IPV6_2292PKTINFO:
2636         case IPV6_PKTINFO:
2637         {
2638                 struct ifnet *ifp = NULL;
2639                 struct in6_pktinfo *pktinfo;
2640
2641                 if (len != sizeof(struct in6_pktinfo))
2642                         return (EINVAL);
2643
2644                 pktinfo = (struct in6_pktinfo *)buf;
2645
2646                 /*
2647                  * An application can clear any sticky IPV6_PKTINFO option by
2648                  * doing a "regular" setsockopt with ipi6_addr being
2649                  * in6addr_any and ipi6_ifindex being zero.
2650                  * [RFC 3542, Section 6]
2651                  */
2652                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2653                     pktinfo->ipi6_ifindex == 0 &&
2654                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2655                         ip6_clearpktopts(opt, optname);
2656                         break;
2657                 }
2658
2659                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2660                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2661                         return (EINVAL);
2662                 }
2663
2664                 /* validate the interface index if specified. */
2665                 if (pktinfo->ipi6_ifindex > V_if_index)
2666                          return (ENXIO);
2667                 if (pktinfo->ipi6_ifindex) {
2668                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2669                         if (ifp == NULL)
2670                                 return (ENXIO);
2671                 }
2672
2673                 /*
2674                  * We store the address anyway, and let in6_selectsrc()
2675                  * validate the specified address.  This is because ipi6_addr
2676                  * may not have enough information about its scope zone, and
2677                  * we may need additional information (such as outgoing
2678                  * interface or the scope zone of a destination address) to
2679                  * disambiguate the scope.
2680                  * XXX: the delay of the validation may confuse the
2681                  * application when it is used as a sticky option.
2682                  */
2683                 if (opt->ip6po_pktinfo == NULL) {
2684                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2685                             M_IP6OPT, M_NOWAIT);
2686                         if (opt->ip6po_pktinfo == NULL)
2687                                 return (ENOBUFS);
2688                 }
2689                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2690                 break;
2691         }
2692
2693         case IPV6_2292HOPLIMIT:
2694         case IPV6_HOPLIMIT:
2695         {
2696                 int *hlimp;
2697
2698                 /*
2699                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2700                  * to simplify the ordering among hoplimit options.
2701                  */
2702                 if (optname == IPV6_HOPLIMIT && sticky)
2703                         return (ENOPROTOOPT);
2704
2705                 if (len != sizeof(int))
2706                         return (EINVAL);
2707                 hlimp = (int *)buf;
2708                 if (*hlimp < -1 || *hlimp > 255)
2709                         return (EINVAL);
2710
2711                 opt->ip6po_hlim = *hlimp;
2712                 break;
2713         }
2714
2715         case IPV6_TCLASS:
2716         {
2717                 int tclass;
2718
2719                 if (len != sizeof(int))
2720                         return (EINVAL);
2721                 tclass = *(int *)buf;
2722                 if (tclass < -1 || tclass > 255)
2723                         return (EINVAL);
2724
2725                 opt->ip6po_tclass = tclass;
2726                 break;
2727         }
2728
2729         case IPV6_2292NEXTHOP:
2730         case IPV6_NEXTHOP:
2731                 if (cred != NULL) {
2732                         error = priv_check_cred(cred,
2733                             PRIV_NETINET_SETHDROPTS, 0);
2734                         if (error)
2735                                 return (error);
2736                 }
2737
2738                 if (len == 0) { /* just remove the option */
2739                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2740                         break;
2741                 }
2742
2743                 /* check if cmsg_len is large enough for sa_len */
2744                 if (len < sizeof(struct sockaddr) || len < *buf)
2745                         return (EINVAL);
2746
2747                 switch (((struct sockaddr *)buf)->sa_family) {
2748                 case AF_INET6:
2749                 {
2750                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2751                         int error;
2752
2753                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2754                                 return (EINVAL);
2755
2756                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2757                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2758                                 return (EINVAL);
2759                         }
2760                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2761                             != 0) {
2762                                 return (error);
2763                         }
2764                         break;
2765                 }
2766                 case AF_LINK:   /* should eventually be supported */
2767                 default:
2768                         return (EAFNOSUPPORT);
2769                 }
2770
2771                 /* turn off the previous option, then set the new option. */
2772                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2773                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2774                 if (opt->ip6po_nexthop == NULL)
2775                         return (ENOBUFS);
2776                 bcopy(buf, opt->ip6po_nexthop, *buf);
2777                 break;
2778
2779         case IPV6_2292HOPOPTS:
2780         case IPV6_HOPOPTS:
2781         {
2782                 struct ip6_hbh *hbh;
2783                 int hbhlen;
2784
2785                 /*
2786                  * XXX: We don't allow a non-privileged user to set ANY HbH
2787                  * options, since per-option restriction has too much
2788                  * overhead.
2789                  */
2790                 if (cred != NULL) {
2791                         error = priv_check_cred(cred,
2792                             PRIV_NETINET_SETHDROPTS, 0);
2793                         if (error)
2794                                 return (error);
2795                 }
2796
2797                 if (len == 0) {
2798                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2799                         break;  /* just remove the option */
2800                 }
2801
2802                 /* message length validation */
2803                 if (len < sizeof(struct ip6_hbh))
2804                         return (EINVAL);
2805                 hbh = (struct ip6_hbh *)buf;
2806                 hbhlen = (hbh->ip6h_len + 1) << 3;
2807                 if (len != hbhlen)
2808                         return (EINVAL);
2809
2810                 /* turn off the previous option, then set the new option. */
2811                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2812                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2813                 if (opt->ip6po_hbh == NULL)
2814                         return (ENOBUFS);
2815                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2816
2817                 break;
2818         }
2819
2820         case IPV6_2292DSTOPTS:
2821         case IPV6_DSTOPTS:
2822         case IPV6_RTHDRDSTOPTS:
2823         {
2824                 struct ip6_dest *dest, **newdest = NULL;
2825                 int destlen;
2826
2827                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2828                         error = priv_check_cred(cred,
2829                             PRIV_NETINET_SETHDROPTS, 0);
2830                         if (error)
2831                                 return (error);
2832                 }
2833
2834                 if (len == 0) {
2835                         ip6_clearpktopts(opt, optname);
2836                         break;  /* just remove the option */
2837                 }
2838
2839                 /* message length validation */
2840                 if (len < sizeof(struct ip6_dest))
2841                         return (EINVAL);
2842                 dest = (struct ip6_dest *)buf;
2843                 destlen = (dest->ip6d_len + 1) << 3;
2844                 if (len != destlen)
2845                         return (EINVAL);
2846
2847                 /*
2848                  * Determine the position that the destination options header
2849                  * should be inserted; before or after the routing header.
2850                  */
2851                 switch (optname) {
2852                 case IPV6_2292DSTOPTS:
2853                         /*
2854                          * The old advacned API is ambiguous on this point.
2855                          * Our approach is to determine the position based
2856                          * according to the existence of a routing header.
2857                          * Note, however, that this depends on the order of the
2858                          * extension headers in the ancillary data; the 1st
2859                          * part of the destination options header must appear
2860                          * before the routing header in the ancillary data,
2861                          * too.
2862                          * RFC3542 solved the ambiguity by introducing
2863                          * separate ancillary data or option types.
2864                          */
2865                         if (opt->ip6po_rthdr == NULL)
2866                                 newdest = &opt->ip6po_dest1;
2867                         else
2868                                 newdest = &opt->ip6po_dest2;
2869                         break;
2870                 case IPV6_RTHDRDSTOPTS:
2871                         newdest = &opt->ip6po_dest1;
2872                         break;
2873                 case IPV6_DSTOPTS:
2874                         newdest = &opt->ip6po_dest2;
2875                         break;
2876                 }
2877
2878                 /* turn off the previous option, then set the new option. */
2879                 ip6_clearpktopts(opt, optname);
2880                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2881                 if (*newdest == NULL)
2882                         return (ENOBUFS);
2883                 bcopy(dest, *newdest, destlen);
2884
2885                 break;
2886         }
2887
2888         case IPV6_2292RTHDR:
2889         case IPV6_RTHDR:
2890         {
2891                 struct ip6_rthdr *rth;
2892                 int rthlen;
2893
2894                 if (len == 0) {
2895                         ip6_clearpktopts(opt, IPV6_RTHDR);
2896                         break;  /* just remove the option */
2897                 }
2898
2899                 /* message length validation */
2900                 if (len < sizeof(struct ip6_rthdr))
2901                         return (EINVAL);
2902                 rth = (struct ip6_rthdr *)buf;
2903                 rthlen = (rth->ip6r_len + 1) << 3;
2904                 if (len != rthlen)
2905                         return (EINVAL);
2906
2907                 switch (rth->ip6r_type) {
2908                 case IPV6_RTHDR_TYPE_0:
2909                         if (rth->ip6r_len == 0) /* must contain one addr */
2910                                 return (EINVAL);
2911                         if (rth->ip6r_len % 2) /* length must be even */
2912                                 return (EINVAL);
2913                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2914                                 return (EINVAL);
2915                         break;
2916                 default:
2917                         return (EINVAL);        /* not supported */
2918                 }
2919
2920                 /* turn off the previous option */
2921                 ip6_clearpktopts(opt, IPV6_RTHDR);
2922                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2923                 if (opt->ip6po_rthdr == NULL)
2924                         return (ENOBUFS);
2925                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2926
2927                 break;
2928         }
2929
2930         case IPV6_USE_MIN_MTU:
2931                 if (len != sizeof(int))
2932                         return (EINVAL);
2933                 minmtupolicy = *(int *)buf;
2934                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2935                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2936                     minmtupolicy != IP6PO_MINMTU_ALL) {
2937                         return (EINVAL);
2938                 }
2939                 opt->ip6po_minmtu = minmtupolicy;
2940                 break;
2941
2942         case IPV6_DONTFRAG:
2943                 if (len != sizeof(int))
2944                         return (EINVAL);
2945
2946                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2947                         /*
2948                          * we ignore this option for TCP sockets.
2949                          * (RFC3542 leaves this case unspecified.)
2950                          */
2951                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2952                 } else
2953                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2954                 break;
2955
2956         case IPV6_PREFER_TEMPADDR:
2957                 if (len != sizeof(int))
2958                         return (EINVAL);
2959                 preftemp = *(int *)buf;
2960                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2961                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2962                     preftemp != IP6PO_TEMPADDR_PREFER) {
2963                         return (EINVAL);
2964                 }
2965                 opt->ip6po_prefer_tempaddr = preftemp;
2966                 break;
2967
2968         default:
2969                 return (ENOPROTOOPT);
2970         } /* end of switch */
2971
2972         return (0);
2973 }
2974
2975 /*
2976  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2977  * packet to the input queue of a specified interface.  Note that this
2978  * calls the output routine of the loopback "driver", but with an interface
2979  * pointer that might NOT be &loif -- easier than replicating that code here.
2980  */
2981 void
2982 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2983 {
2984         struct mbuf *copym;
2985         struct ip6_hdr *ip6;
2986
2987         copym = m_copy(m, 0, M_COPYALL);
2988         if (copym == NULL)
2989                 return;
2990
2991         /*
2992          * Make sure to deep-copy IPv6 header portion in case the data
2993          * is in an mbuf cluster, so that we can safely override the IPv6
2994          * header portion later.
2995          */
2996         if ((copym->m_flags & M_EXT) != 0 ||
2997             copym->m_len < sizeof(struct ip6_hdr)) {
2998                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2999                 if (copym == NULL)
3000                         return;
3001         }
3002
3003 #ifdef DIAGNOSTIC
3004         if (copym->m_len < sizeof(*ip6)) {
3005                 m_freem(copym);
3006                 return;
3007         }
3008 #endif
3009
3010         ip6 = mtod(copym, struct ip6_hdr *);
3011         /*
3012          * clear embedded scope identifiers if necessary.
3013          * in6_clearscope will touch the addresses only when necessary.
3014          */
3015         in6_clearscope(&ip6->ip6_src);
3016         in6_clearscope(&ip6->ip6_dst);
3017
3018         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3019 }
3020
3021 /*
3022  * Chop IPv6 header off from the payload.
3023  */
3024 static int
3025 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3026 {
3027         struct mbuf *mh;
3028         struct ip6_hdr *ip6;
3029
3030         ip6 = mtod(m, struct ip6_hdr *);
3031         if (m->m_len > sizeof(*ip6)) {
3032                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3033                 if (mh == NULL) {
3034                         m_freem(m);
3035                         return ENOBUFS;
3036                 }
3037                 m_move_pkthdr(mh, m);
3038                 MH_ALIGN(mh, sizeof(*ip6));
3039                 m->m_len -= sizeof(*ip6);
3040                 m->m_data += sizeof(*ip6);
3041                 mh->m_next = m;
3042                 m = mh;
3043                 m->m_len = sizeof(*ip6);
3044                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3045         }
3046         exthdrs->ip6e_ip6 = m;
3047         return 0;
3048 }
3049
3050 /*
3051  * Compute IPv6 extension header length.
3052  */
3053 int
3054 ip6_optlen(struct inpcb *in6p)
3055 {
3056         int len;
3057
3058         if (!in6p->in6p_outputopts)
3059                 return 0;
3060
3061         len = 0;
3062 #define elen(x) \
3063     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3064
3065         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3066         if (in6p->in6p_outputopts->ip6po_rthdr)
3067                 /* dest1 is valid with rthdr only */
3068                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3069         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3070         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3071         return len;
3072 #undef elen
3073 }