sys/netinet6/ip6_output.c

   1 /*-
   2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. Neither the name of the project nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
  30  */
  31
  32 /*-
  33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  34  *      The Regents of the University of California.  All rights reserved.
  35  *
  36  * Redistribution and use in source and binary forms, with or without
  37  * modification, are permitted provided that the following conditions
  38  * are met:
  39  * 1. Redistributions of source code must retain the above copyright
  40  *    notice, this list of conditions and the following disclaimer.
  41  * 2. Redistributions in binary form must reproduce the above copyright
  42  *    notice, this list of conditions and the following disclaimer in the
  43  *    documentation and/or other materials provided with the distribution.
  44  * 4. Neither the name of the University nor the names of its contributors
  45  *    may be used to endorse or promote products derived from this software
  46  *    without specific prior written permission.
  47  *
  48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  58  * SUCH DAMAGE.
  59  *
  60  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  61  */
  62
  63 #include <sys/cdefs.h>
  64 __FBSDID("$FreeBSD$");
  65
  66 #include "opt_inet.h"
  67 #include "opt_inet6.h"
  68 #include "opt_ipfw.h"
  69 #include "opt_ipsec.h"
  70 #include "opt_sctp.h"
  71 #include "opt_route.h"
  72
  73 #include <sys/param.h>
  74 #include <sys/kernel.h>
  75 #include <sys/malloc.h>
  76 #include <sys/mbuf.h>
  77 #include <sys/errno.h>
  78 #include <sys/priv.h>
  79 #include <sys/proc.h>
  80 #include <sys/protosw.h>
  81 #include <sys/socket.h>
  82 #include <sys/socketvar.h>
  83 #include <sys/syslog.h>
  84 #include <sys/ucred.h>
  85
  86 #include <machine/in_cksum.h>
  87
  88 #include <net/if.h>
  89 #include <net/netisr.h>
  90 #include <net/route.h>
  91 #include <net/pfil.h>
  92 #include <net/vnet.h>
  93
  94 #include <netinet/in.h>
  95 #include <netinet/in_var.h>
  96 #include <netinet/ip_var.h>
  97 #include <netinet6/in6_var.h>
  98 #include <netinet/ip6.h>
  99 #include <netinet/icmp6.h>
 100 #include <netinet6/ip6_var.h>
 101 #include <netinet/in_pcb.h>
 102 #include <netinet/tcp_var.h>
 103 #include <netinet6/nd6.h>
 104
 105 #ifdef IPSEC
 106 #include <netipsec/ipsec.h>
 107 #include <netipsec/ipsec6.h>
 108 #include <netipsec/key.h>
 109 #include <netinet6/ip6_ipsec.h>
 110 #endif /* IPSEC */
 111 #ifdef SCTP
 112 #include <netinet/sctp.h>
 113 #include <netinet/sctp_crc32.h>
 114 #endif
 115
 116 #include <netinet6/ip6protosw.h>
 117 #include <netinet6/scope6_var.h>
 118
 119 #ifdef FLOWTABLE
 120 #include <net/flowtable.h>
 121 #endif
 122
 123 extern int in6_mcast_loop;
 124
 125 struct ip6_exthdrs {
 126         struct mbuf *ip6e_ip6;
 127         struct mbuf *ip6e_hbh;
 128         struct mbuf *ip6e_dest1;
 129         struct mbuf *ip6e_rthdr;
 130         struct mbuf *ip6e_dest2;
 131 };
 132
 133 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
 134                            struct ucred *, int);
 135 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
 136         struct socket *, struct sockopt *);
 137 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 138 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
 139         struct ucred *, int, int, int);
 140
 141 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 142 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 143         struct ip6_frag **);
 144 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 145 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 146 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
 147         struct ifnet *, struct in6_addr *, u_long *, int *, u_int);
 148 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 149
 150
 151 /*
 152  * Make an extension header from option data.  hp is the source, and
 153  * mp is the destination.
 154  */
 155 #define MAKE_EXTHDR(hp, mp)                                             \
 156     do {                                                                \
 157         if (hp) {                                                       \
 158                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 159                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 160                     ((eh)->ip6e_len + 1) << 3);                         \
 161                 if (error)                                              \
 162                         goto freehdrs;                                  \
 163         }                                                               \
 164     } while (/*CONSTCOND*/ 0)
 165
 166 /*
 167  * Form a chain of extension headers.
 168  * m is the extension header mbuf
 169  * mp is the previous mbuf in the chain
 170  * p is the next header
 171  * i is the type of option.
 172  */
 173 #define MAKE_CHAIN(m, mp, p, i)\
 174     do {\
 175         if (m) {\
 176                 if (!hdrsplit) \
 177                         panic("assumption failed: hdr not split"); \
 178                 *mtod((m), u_char *) = *(p);\
 179                 *(p) = (i);\
 180                 p = mtod((m), u_char *);\
 181                 (m)->m_next = (mp)->m_next;\
 182                 (mp)->m_next = (m);\
 183                 (mp) = (m);\
 184         }\
 185     } while (/*CONSTCOND*/ 0)
 186
 187 void
 188 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 189 {
 190         u_short csum;
 191
 192         csum = in_cksum_skip(m, offset + plen, offset);
 193         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 194                 csum = 0xffff;
 195         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 196
 197         if (offset + sizeof(u_short) > m->m_len) {
 198                 printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
 199                     "csum_flags=0x%04x\n", __func__, m->m_len, plen, offset,
 200                     m->m_pkthdr.csum_flags);
 201                 /*
 202                  * XXX this should not happen, but if it does, the correct
 203                  * behavior may be to insert the checksum in the appropriate
 204                  * next mbuf in the chain.
 205                  */
 206                 return;
 207         }
 208         *(u_short *)(m->m_data + offset) = csum;
 209 }
 210
 211 /*
 212  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 213  * header (with pri, len, nxt, hlim, src, dst).
 214  * This function may modify ver and hlim only.
 215  * The mbuf chain containing the packet will be freed.
 216  * The mbuf opt, if present, will not be freed.
 217  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
 218  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 219  * then result of route lookup is stored in ro->ro_rt.
 220  *
 221  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 222  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 223  * which is rt_rmx.rmx_mtu.
 224  *
 225  * ifpp - XXX: just for statistics
 226  */
 227 int
 228 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 229     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 230     struct ifnet **ifpp, struct inpcb *inp)
 231 {
 232         struct ip6_hdr *ip6, *mhip6;
 233         struct ifnet *ifp, *origifp;
 234         struct mbuf *m = m0;
 235         struct mbuf *mprev = NULL;
 236         int hlen, tlen, len, off;
 237         struct route_in6 ip6route;
 238         struct rtentry *rt = NULL;
 239         struct sockaddr_in6 *dst, src_sa, dst_sa;
 240         struct in6_addr odst;
 241         int error = 0;
 242         struct in6_ifaddr *ia = NULL;
 243         u_long mtu;
 244         int alwaysfrag, dontfrag;
 245         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 246         struct ip6_exthdrs exthdrs;
 247         struct in6_addr finaldst, src0, dst0;
 248         u_int32_t zone;
 249         struct route_in6 *ro_pmtu = NULL;
 250         int hdrsplit = 0;
 251         int needipsec = 0;
 252         int sw_csum, tso;
 253 #ifdef IPSEC
 254         struct ipsec_output_state state;
 255         struct ip6_rthdr *rh = NULL;
 256         int needipsectun = 0;
 257         int segleft_org = 0;
 258         struct secpolicy *sp = NULL;
 259 #endif /* IPSEC */
 260         struct m_tag *fwd_tag = NULL;
 261
 262         ip6 = mtod(m, struct ip6_hdr *);
 263         if (ip6 == NULL) {
 264                 printf ("ip6 is NULL");
 265                 goto bad;
 266         }
 267
 268         if (inp != NULL)
 269                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 270
 271         finaldst = ip6->ip6_dst;
 272         bzero(&exthdrs, sizeof(exthdrs));
 273         if (opt) {
 274                 /* Hop-by-Hop options header */
 275                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 276                 /* Destination options header(1st part) */
 277                 if (opt->ip6po_rthdr) {
 278                         /*
 279                          * Destination options header(1st part)
 280                          * This only makes sense with a routing header.
 281                          * See Section 9.2 of RFC 3542.
 282                          * Disabling this part just for MIP6 convenience is
 283                          * a bad idea.  We need to think carefully about a
 284                          * way to make the advanced API coexist with MIP6
 285                          * options, which might automatically be inserted in
 286                          * the kernel.
 287                          */
 288                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 289                 }
 290                 /* Routing header */
 291                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 292                 /* Destination options header(2nd part) */
 293                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 294         }
 295
 296 #ifdef IPSEC
 297         /*
 298          * IPSec checking which handles several cases.
 299          * FAST IPSEC: We re-injected the packet.
 300          */
 301         switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
 302         {
 303         case 1:                 /* Bad packet */
 304                 goto freehdrs;
 305         case -1:                /* Do IPSec */
 306                 needipsec = 1;
 307                 /*
 308                  * Do delayed checksums now, as we may send before returning.
 309                  */
 310                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 311                         plen = m->m_pkthdr.len - sizeof(*ip6);
 312                         in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 313                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 314                 }
 315 #ifdef SCTP
 316                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
 317                         sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 318                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 319                 }
 320 #endif
 321         case 0:                 /* No IPSec */
 322         default:
 323                 break;
 324         }
 325 #endif /* IPSEC */
 326
 327         /*
 328          * Calculate the total length of the extension header chain.
 329          * Keep the length of the unfragmentable part for fragmentation.
 330          */
 331         optlen = 0;
 332         if (exthdrs.ip6e_hbh)
 333                 optlen += exthdrs.ip6e_hbh->m_len;
 334         if (exthdrs.ip6e_dest1)
 335                 optlen += exthdrs.ip6e_dest1->m_len;
 336         if (exthdrs.ip6e_rthdr)
 337                 optlen += exthdrs.ip6e_rthdr->m_len;
 338         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 339
 340         /* NOTE: we don't add AH/ESP length here. do that later. */
 341         if (exthdrs.ip6e_dest2)
 342                 optlen += exthdrs.ip6e_dest2->m_len;
 343
 344         /*
 345          * If we need IPsec, or there is at least one extension header,
 346          * separate IP6 header from the payload.
 347          */
 348         if ((needipsec || optlen) && !hdrsplit) {
 349                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 350                         m = NULL;
 351                         goto freehdrs;
 352                 }
 353                 m = exthdrs.ip6e_ip6;
 354                 hdrsplit++;
 355         }
 356
 357         /* adjust pointer */
 358         ip6 = mtod(m, struct ip6_hdr *);
 359
 360         /* adjust mbuf packet header length */
 361         m->m_pkthdr.len += optlen;
 362         plen = m->m_pkthdr.len - sizeof(*ip6);
 363
 364         /* If this is a jumbo payload, insert a jumbo payload option. */
 365         if (plen > IPV6_MAXPACKET) {
 366                 if (!hdrsplit) {
 367                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 368                                 m = NULL;
 369                                 goto freehdrs;
 370                         }
 371                         m = exthdrs.ip6e_ip6;
 372                         hdrsplit++;
 373                 }
 374                 /* adjust pointer */
 375                 ip6 = mtod(m, struct ip6_hdr *);
 376                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 377                         goto freehdrs;
 378                 ip6->ip6_plen = 0;
 379         } else
 380                 ip6->ip6_plen = htons(plen);
 381
 382         /*
 383          * Concatenate headers and fill in next header fields.
 384          * Here we have, on "m"
 385          *      IPv6 payload
 386          * and we insert headers accordingly.  Finally, we should be getting:
 387          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 388          *
 389          * during the header composing process, "m" points to IPv6 header.
 390          * "mprev" points to an extension header prior to esp.
 391          */
 392         u_char *nexthdrp = &ip6->ip6_nxt;
 393         mprev = m;
 394
 395         /*
 396          * we treat dest2 specially.  this makes IPsec processing
 397          * much easier.  the goal here is to make mprev point the
 398          * mbuf prior to dest2.
 399          *
 400          * result: IPv6 dest2 payload
 401          * m and mprev will point to IPv6 header.
 402          */
 403         if (exthdrs.ip6e_dest2) {
 404                 if (!hdrsplit)
 405                         panic("assumption failed: hdr not split");
 406                 exthdrs.ip6e_dest2->m_next = m->m_next;
 407                 m->m_next = exthdrs.ip6e_dest2;
 408                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 409                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 410         }
 411
 412         /*
 413          * result: IPv6 hbh dest1 rthdr dest2 payload
 414          * m will point to IPv6 header.  mprev will point to the
 415          * extension header prior to dest2 (rthdr in the above case).
 416          */
 417         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 418         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 419                    IPPROTO_DSTOPTS);
 420         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 421                    IPPROTO_ROUTING);
 422
 423 #ifdef IPSEC
 424         if (!needipsec)
 425                 goto skip_ipsec2;
 426
 427         /*
 428          * pointers after IPsec headers are not valid any more.
 429          * other pointers need a great care too.
 430          * (IPsec routines should not mangle mbufs prior to AH/ESP)
 431          */
 432         exthdrs.ip6e_dest2 = NULL;
 433
 434         if (exthdrs.ip6e_rthdr) {
 435                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 436                 segleft_org = rh->ip6r_segleft;
 437                 rh->ip6r_segleft = 0;
 438         }
 439
 440         bzero(&state, sizeof(state));
 441         state.m = m;
 442         error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 443                                     &needipsectun);
 444         m = state.m;
 445         if (error == EJUSTRETURN) {
 446                 /*
 447                  * We had a SP with a level of 'use' and no SA. We
 448                  * will just continue to process the packet without
 449                  * IPsec processing.
 450                  */
 451                 ;
 452         } else if (error) {
 453                 /* mbuf is already reclaimed in ipsec6_output_trans. */
 454                 m = NULL;
 455                 switch (error) {
 456                 case EHOSTUNREACH:
 457                 case ENETUNREACH:
 458                 case EMSGSIZE:
 459                 case ENOBUFS:
 460                 case ENOMEM:
 461                         break;
 462                 default:
 463                         printf("[%s:%d] (ipsec): error code %d\n",
 464                             __func__, __LINE__, error);
 465                         /* FALLTHROUGH */
 466                 case ENOENT:
 467                         /* don't show these error codes to the user */
 468                         error = 0;
 469                         break;
 470                 }
 471                 goto bad;
 472         } else if (!needipsectun) {
 473                 /*
 474                  * In the FAST IPSec case we have already
 475                  * re-injected the packet and it has been freed
 476                  * by the ipsec_done() function.  So, just clean
 477                  * up after ourselves.
 478                  */
 479                 m = NULL;
 480                 goto done;
 481         }
 482         if (exthdrs.ip6e_rthdr) {
 483                 /* ah6_output doesn't modify mbuf chain */
 484                 rh->ip6r_segleft = segleft_org;
 485         }
 486 skip_ipsec2:;
 487 #endif /* IPSEC */
 488
 489         /*
 490          * If there is a routing header, discard the packet.
 491          */
 492         if (exthdrs.ip6e_rthdr) {
 493                  error = EINVAL;
 494                  goto bad;
 495         }
 496
 497         /* Source address validation */
 498         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 499             (flags & IPV6_UNSPECSRC) == 0) {
 500                 error = EOPNOTSUPP;
 501                 IP6STAT_INC(ip6s_badscope);
 502                 goto bad;
 503         }
 504         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 505                 error = EOPNOTSUPP;
 506                 IP6STAT_INC(ip6s_badscope);
 507                 goto bad;
 508         }
 509
 510         IP6STAT_INC(ip6s_localout);
 511
 512         /*
 513          * Route packet.
 514          */
 515         if (ro == 0) {
 516                 ro = &ip6route;
 517                 bzero((caddr_t)ro, sizeof(*ro));
 518         }
 519         ro_pmtu = ro;
 520         if (opt && opt->ip6po_rthdr)
 521                 ro = &opt->ip6po_route;
 522         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 523 #ifdef FLOWTABLE
 524         if (ro->ro_rt == NULL) {
 525                 struct flentry *fle;
 526
 527                 /*
 528                  * The flow table returns route entries valid for up to 30
 529                  * seconds; we rely on the remainder of ip_output() taking no
 530                  * longer than that long for the stability of ro_rt.  The
 531                  * flow ID assignment must have happened before this point.
 532                  */
 533                 fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6);
 534                 if (fle != NULL)
 535                         flow_to_route_in6(fle, ro);
 536         }
 537 #endif
 538 again:
 539         /*
 540          * if specified, try to fill in the traffic class field.
 541          * do not override if a non-zero value is already set.
 542          * we check the diffserv field and the ecn field separately.
 543          */
 544         if (opt && opt->ip6po_tclass >= 0) {
 545                 int mask = 0;
 546
 547                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 548                         mask |= 0xfc;
 549                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 550                         mask |= 0x03;
 551                 if (mask != 0)
 552                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 553         }
 554
 555         /* fill in or override the hop limit field, if necessary. */
 556         if (opt && opt->ip6po_hlim != -1)
 557                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 558         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 559                 if (im6o != NULL)
 560                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 561                 else
 562                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 563         }
 564
 565 #ifdef IPSEC
 566         /*
 567          * We may re-inject packets into the stack here.
 568          */
 569         if (needipsec && needipsectun) {
 570                 struct ipsec_output_state state;
 571
 572                 /*
 573                  * All the extension headers will become inaccessible
 574                  * (since they can be encrypted).
 575                  * Don't panic, we need no more updates to extension headers
 576                  * on inner IPv6 packet (since they are now encapsulated).
 577                  *
 578                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 579                  */
 580                 bzero(&exthdrs, sizeof(exthdrs));
 581                 exthdrs.ip6e_ip6 = m;
 582
 583                 bzero(&state, sizeof(state));
 584                 state.m = m;
 585                 state.ro = (struct route *)ro;
 586                 state.dst = (struct sockaddr *)dst;
 587
 588                 error = ipsec6_output_tunnel(&state, sp, flags);
 589
 590                 m = state.m;
 591                 ro = (struct route_in6 *)state.ro;
 592                 dst = (struct sockaddr_in6 *)state.dst;
 593                 if (error == EJUSTRETURN) {
 594                         /*
 595                          * We had a SP with a level of 'use' and no SA. We
 596                          * will just continue to process the packet without
 597                          * IPsec processing.
 598                          */
 599                         ;
 600                 } else if (error) {
 601                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 602                         m0 = m = NULL;
 603                         m = NULL;
 604                         switch (error) {
 605                         case EHOSTUNREACH:
 606                         case ENETUNREACH:
 607                         case EMSGSIZE:
 608                         case ENOBUFS:
 609                         case ENOMEM:
 610                                 break;
 611                         default:
 612                                 printf("[%s:%d] (ipsec): error code %d\n",
 613                                     __func__, __LINE__, error);
 614                                 /* FALLTHROUGH */
 615                         case ENOENT:
 616                                 /* don't show these error codes to the user */
 617                                 error = 0;
 618                                 break;
 619                         }
 620                         goto bad;
 621                 } else {
 622                         /*
 623                          * In the FAST IPSec case we have already
 624                          * re-injected the packet and it has been freed
 625                          * by the ipsec_done() function.  So, just clean
 626                          * up after ourselves.
 627                          */
 628                         m = NULL;
 629                         goto done;
 630                 }
 631
 632                 exthdrs.ip6e_ip6 = m;
 633         }
 634 #endif /* IPSEC */
 635
 636         /* adjust pointer */
 637         ip6 = mtod(m, struct ip6_hdr *);
 638
 639         if (ro->ro_rt && fwd_tag == NULL) {
 640                 rt = ro->ro_rt;
 641                 ifp = ro->ro_rt->rt_ifp;
 642         } else {
 643                 if (fwd_tag == NULL) {
 644                         bzero(&dst_sa, sizeof(dst_sa));
 645                         dst_sa.sin6_family = AF_INET6;
 646                         dst_sa.sin6_len = sizeof(dst_sa);
 647                         dst_sa.sin6_addr = ip6->ip6_dst;
 648                 }
 649                 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
 650                     &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
 651                 if (error != 0) {
 652                         if (ifp != NULL)
 653                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 654                         goto bad;
 655                 }
 656         }
 657         if (rt == NULL) {
 658                 /*
 659                  * If in6_selectroute() does not return a route entry,
 660                  * dst may not have been updated.
 661                  */
 662                 *dst = dst_sa;  /* XXX */
 663         }
 664
 665         /*
 666          * then rt (for unicast) and ifp must be non-NULL valid values.
 667          */
 668         if ((flags & IPV6_FORWARDING) == 0) {
 669                 /* XXX: the FORWARDING flag can be set for mrouting. */
 670                 in6_ifstat_inc(ifp, ifs6_out_request);
 671         }
 672         if (rt != NULL) {
 673                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 674                 rt->rt_use++;
 675         }
 676
 677
 678         /*
 679          * The outgoing interface must be in the zone of source and
 680          * destination addresses.
 681          */
 682         origifp = ifp;
 683
 684         src0 = ip6->ip6_src;
 685         if (in6_setscope(&src0, origifp, &zone))
 686                 goto badscope;
 687         bzero(&src_sa, sizeof(src_sa));
 688         src_sa.sin6_family = AF_INET6;
 689         src_sa.sin6_len = sizeof(src_sa);
 690         src_sa.sin6_addr = ip6->ip6_src;
 691         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 692                 goto badscope;
 693
 694         dst0 = ip6->ip6_dst;
 695         if (in6_setscope(&dst0, origifp, &zone))
 696                 goto badscope;
 697         /* re-initialize to be sure */
 698         bzero(&dst_sa, sizeof(dst_sa));
 699         dst_sa.sin6_family = AF_INET6;
 700         dst_sa.sin6_len = sizeof(dst_sa);
 701         dst_sa.sin6_addr = ip6->ip6_dst;
 702         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 703                 goto badscope;
 704         }
 705
 706         /* We should use ia_ifp to support the case of
 707          * sending packets to an address of our own.
 708          */
 709         if (ia != NULL && ia->ia_ifp)
 710                 ifp = ia->ia_ifp;
 711
 712         /* scope check is done. */
 713         goto routefound;
 714
 715   badscope:
 716         IP6STAT_INC(ip6s_badscope);
 717         in6_ifstat_inc(origifp, ifs6_out_discard);
 718         if (error == 0)
 719                 error = EHOSTUNREACH; /* XXX */
 720         goto bad;
 721
 722   routefound:
 723         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 724                 if (opt && opt->ip6po_nextroute.ro_rt) {
 725                         /*
 726                          * The nexthop is explicitly specified by the
 727                          * application.  We assume the next hop is an IPv6
 728                          * address.
 729                          */
 730                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 731                 }
 732                 else if ((rt->rt_flags & RTF_GATEWAY))
 733                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 734         }
 735
 736         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 737                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 738         } else {
 739                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 740                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 741                 /*
 742                  * Confirm that the outgoing interface supports multicast.
 743                  */
 744                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 745                         IP6STAT_INC(ip6s_noroute);
 746                         in6_ifstat_inc(ifp, ifs6_out_discard);
 747                         error = ENETUNREACH;
 748                         goto bad;
 749                 }
 750                 if ((im6o == NULL && in6_mcast_loop) ||
 751                     (im6o && im6o->im6o_multicast_loop)) {
 752                         /*
 753                          * Loop back multicast datagram if not expressly
 754                          * forbidden to do so, even if we have not joined
 755                          * the address; protocols will filter it later,
 756                          * thus deferring a hash lookup and lock acquisition
 757                          * at the expense of an m_copym().
 758                          */
 759                         ip6_mloopback(ifp, m, dst);
 760                 } else {
 761                         /*
 762                          * If we are acting as a multicast router, perform
 763                          * multicast forwarding as if the packet had just
 764                          * arrived on the interface to which we are about
 765                          * to send.  The multicast forwarding function
 766                          * recursively calls this function, using the
 767                          * IPV6_FORWARDING flag to prevent infinite recursion.
 768                          *
 769                          * Multicasts that are looped back by ip6_mloopback(),
 770                          * above, will be forwarded by the ip6_input() routine,
 771                          * if necessary.
 772                          */
 773                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 774                                 /*
 775                                  * XXX: ip6_mforward expects that rcvif is NULL
 776                                  * when it is called from the originating path.
 777                                  * However, it is not always the case, since
 778                                  * some versions of MGETHDR() does not
 779                                  * initialize the field.
 780                                  */
 781                                 m->m_pkthdr.rcvif = NULL;
 782                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 783                                         m_freem(m);
 784                                         goto done;
 785                                 }
 786                         }
 787                 }
 788                 /*
 789                  * Multicasts with a hoplimit of zero may be looped back,
 790                  * above, but must not be transmitted on a network.
 791                  * Also, multicasts addressed to the loopback interface
 792                  * are not sent -- the above call to ip6_mloopback() will
 793                  * loop back a copy if this host actually belongs to the
 794                  * destination group on the loopback interface.
 795                  */
 796                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 797                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 798                         m_freem(m);
 799                         goto done;
 800                 }
 801         }
 802
 803         /*
 804          * Fill the outgoing inteface to tell the upper layer
 805          * to increment per-interface statistics.
 806          */
 807         if (ifpp)
 808                 *ifpp = ifp;
 809
 810         /* Determine path MTU. */
 811         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 812             &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
 813                 goto bad;
 814
 815         /*
 816          * The caller of this function may specify to use the minimum MTU
 817          * in some cases.
 818          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 819          * setting.  The logic is a bit complicated; by default, unicast
 820          * packets will follow path MTU while multicast packets will be sent at
 821          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 822          * including unicast ones will be sent at the minimum MTU.  Multicast
 823          * packets will always be sent at the minimum MTU unless
 824          * IP6PO_MINMTU_DISABLE is explicitly specified.
 825          * See RFC 3542 for more details.
 826          */
 827         if (mtu > IPV6_MMTU) {
 828                 if ((flags & IPV6_MINMTU))
 829                         mtu = IPV6_MMTU;
 830                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 831                         mtu = IPV6_MMTU;
 832                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 833                          (opt == NULL ||
 834                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 835                         mtu = IPV6_MMTU;
 836                 }
 837         }
 838
 839         /*
 840          * clear embedded scope identifiers if necessary.
 841          * in6_clearscope will touch the addresses only when necessary.
 842          */
 843         in6_clearscope(&ip6->ip6_src);
 844         in6_clearscope(&ip6->ip6_dst);
 845
 846         /*
 847          * If the outgoing packet contains a hop-by-hop options header,
 848          * it must be examined and processed even by the source node.
 849          * (RFC 2460, section 4.)
 850          */
 851         if (exthdrs.ip6e_hbh) {
 852                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 853                 u_int32_t dummy; /* XXX unused */
 854                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 855
 856 #ifdef DIAGNOSTIC
 857                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 858                         panic("ip6e_hbh is not contiguous");
 859 #endif
 860                 /*
 861                  *  XXX: if we have to send an ICMPv6 error to the sender,
 862                  *       we need the M_LOOP flag since icmp6_error() expects
 863                  *       the IPv6 and the hop-by-hop options header are
 864                  *       contiguous unless the flag is set.
 865                  */
 866                 m->m_flags |= M_LOOP;
 867                 m->m_pkthdr.rcvif = ifp;
 868                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 869                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 870                     &dummy, &plen) < 0) {
 871                         /* m was already freed at this point */
 872                         error = EINVAL;/* better error? */
 873                         goto done;
 874                 }
 875                 m->m_flags &= ~M_LOOP; /* XXX */
 876                 m->m_pkthdr.rcvif = NULL;
 877         }
 878
 879         /* Jump over all PFIL processing if hooks are not active. */
 880         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 881                 goto passout;
 882
 883         odst = ip6->ip6_dst;
 884         /* Run through list of hooks for output packets. */
 885         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 886         if (error != 0 || m == NULL)
 887                 goto done;
 888         ip6 = mtod(m, struct ip6_hdr *);
 889
 890         /* See if destination IP address was changed by packet filter. */
 891         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 892                 m->m_flags |= M_SKIP_FIREWALL;
 893                 /* If destination is now ourself drop to ip6_input(). */
 894                 if (in6_localip(&ip6->ip6_dst)) {
 895                         m->m_flags |= M_FASTFWD_OURS;
 896                         if (m->m_pkthdr.rcvif == NULL)
 897                                 m->m_pkthdr.rcvif = V_loif;
 898                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 899                                 m->m_pkthdr.csum_flags |=
 900                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 901                                 m->m_pkthdr.csum_data = 0xffff;
 902                         }
 903 #ifdef SCTP
 904                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 905                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 906 #endif
 907                         error = netisr_queue(NETISR_IPV6, m);
 908                         goto done;
 909                 } else
 910                         goto again;     /* Redo the routing table lookup. */
 911         }
 912
 913         /* See if local, if yes, send it to netisr. */
 914         if (m->m_flags & M_FASTFWD_OURS) {
 915                 if (m->m_pkthdr.rcvif == NULL)
 916                         m->m_pkthdr.rcvif = V_loif;
 917                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 918                         m->m_pkthdr.csum_flags |=
 919                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 920                         m->m_pkthdr.csum_data = 0xffff;
 921                 }
 922 #ifdef SCTP
 923                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 924                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 925 #endif
 926                 error = netisr_queue(NETISR_IPV6, m);
 927                 goto done;
 928         }
 929         /* Or forward to some other address? */
 930         if ((m->m_flags & M_IP6_NEXTHOP) &&
 931             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 932                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 933                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
 934                 m->m_flags |= M_SKIP_FIREWALL;
 935                 m->m_flags &= ~M_IP6_NEXTHOP;
 936                 m_tag_delete(m, fwd_tag);
 937                 goto again;
 938         }
 939
 940 passout:
 941         /*
 942          * Send the packet to the outgoing interface.
 943          * If necessary, do IPv6 fragmentation before sending.
 944          *
 945          * the logic here is rather complex:
 946          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 947          * 1-a: send as is if tlen <= path mtu
 948          * 1-b: fragment if tlen > path mtu
 949          *
 950          * 2: if user asks us not to fragment (dontfrag == 1)
 951          * 2-a: send as is if tlen <= interface mtu
 952          * 2-b: error if tlen > interface mtu
 953          *
 954          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 955          *      always fragment
 956          *
 957          * 4: if dontfrag == 1 && alwaysfrag == 1
 958          *      error, as we cannot handle this conflicting request
 959          */
 960         sw_csum = m->m_pkthdr.csum_flags;
 961         if (!hdrsplit) {
 962                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 963                 sw_csum &= ~ifp->if_hwassist;
 964         } else
 965                 tso = 0;
 966         /*
 967          * If we added extension headers, we will not do TSO and calculate the
 968          * checksums ourselves for now.
 969          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 970          * with ext. hdrs.
 971          */
 972         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 973                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 974                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 975         }
 976 #ifdef SCTP
 977         if (sw_csum & CSUM_SCTP_IPV6) {
 978                 sw_csum &= ~CSUM_SCTP_IPV6;
 979                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 980         }
 981 #endif
 982         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 983         tlen = m->m_pkthdr.len;
 984
 985         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 986                 dontfrag = 1;
 987         else
 988                 dontfrag = 0;
 989         if (dontfrag && alwaysfrag) {   /* case 4 */
 990                 /* conflicting request - can't transmit */
 991                 error = EMSGSIZE;
 992                 goto bad;
 993         }
 994         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 995                 /*
 996                  * Even if the DONTFRAG option is specified, we cannot send the
 997                  * packet when the data length is larger than the MTU of the
 998                  * outgoing interface.
 999                  * Notify the error by sending IPV6_PATHMTU ancillary data as
1000                  * well as returning an error code (the latter is not described
1001                  * in the API spec.)
1002                  */
1003                 u_int32_t mtu32;
1004                 struct ip6ctlparam ip6cp;
1005
1006                 mtu32 = (u_int32_t)mtu;
1007                 bzero(&ip6cp, sizeof(ip6cp));
1008                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1009                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1010                     (void *)&ip6cp);
1011
1012                 error = EMSGSIZE;
1013                 goto bad;
1014         }
1015
1016         /*
1017          * transmit packet without fragmentation
1018          */
1019         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1020                 struct in6_ifaddr *ia6;
1021
1022                 ip6 = mtod(m, struct ip6_hdr *);
1023                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1024                 if (ia6) {
1025                         /* Record statistics for this interface address. */
1026                         ia6->ia_ifa.if_opackets++;
1027                         ia6->ia_ifa.if_obytes += 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                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1126                         if (!m) {
1127                                 error = ENOBUFS;
1128                                 IP6STAT_INC(ip6s_odropped);
1129                                 goto sendorfree;
1130                         }
1131                         m->m_pkthdr.rcvif = NULL;
1132                         m->m_flags = m0->m_flags & M_COPYFLAGS; /* incl. FIB */
1133                         *mnext = m;
1134                         mnext = &m->m_nextpkt;
1135                         m->m_data += max_linkhdr;
1136                         mhip6 = mtod(m, struct ip6_hdr *);
1137                         *mhip6 = *ip6;
1138                         m->m_len = sizeof(*mhip6);
1139                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1140                         if (error) {
1141                                 IP6STAT_INC(ip6s_odropped);
1142                                 goto sendorfree;
1143                         }
1144                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1145                         if (off + len >= tlen)
1146                                 len = tlen - off;
1147                         else
1148                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1149                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1150                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1151                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1152                                 error = ENOBUFS;
1153                                 IP6STAT_INC(ip6s_odropped);
1154                                 goto sendorfree;
1155                         }
1156                         m_cat(m, m_frgpart);
1157                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
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                                 ia->ia_ifa.if_opackets++;
1183                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1184                         }
1185                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1186                 } else
1187                         m_freem(m);
1188         }
1189
1190         if (error == 0)
1191                 IP6STAT_INC(ip6s_fragmented);
1192
1193 done:
1194         if (ro == &ip6route)
1195                 RO_RTFREE(ro);
1196         if (ro_pmtu == &ip6route)
1197                 RO_RTFREE(ro_pmtu);
1198 #ifdef IPSEC
1199         if (sp != NULL)
1200                 KEY_FREESP(&sp);
1201 #endif
1202
1203         return (error);
1204
1205 freehdrs:
1206         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1207         m_freem(exthdrs.ip6e_dest1);
1208         m_freem(exthdrs.ip6e_rthdr);
1209         m_freem(exthdrs.ip6e_dest2);
1210         /* FALLTHROUGH */
1211 bad:
1212         if (m)
1213                 m_freem(m);
1214         goto done;
1215 }
1216
1217 static int
1218 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1219 {
1220         struct mbuf *m;
1221
1222         if (hlen > MCLBYTES)
1223                 return (ENOBUFS); /* XXX */
1224
1225         MGET(m, M_DONTWAIT, MT_DATA);
1226         if (!m)
1227                 return (ENOBUFS);
1228
1229         if (hlen > MLEN) {
1230                 MCLGET(m, M_DONTWAIT);
1231                 if ((m->m_flags & M_EXT) == 0) {
1232                         m_free(m);
1233                         return (ENOBUFS);
1234                 }
1235         }
1236         m->m_len = hlen;
1237         if (hdr)
1238                 bcopy(hdr, mtod(m, caddr_t), hlen);
1239
1240         *mp = m;
1241         return (0);
1242 }
1243
1244 /*
1245  * Insert jumbo payload option.
1246  */
1247 static int
1248 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1249 {
1250         struct mbuf *mopt;
1251         u_char *optbuf;
1252         u_int32_t v;
1253
1254 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1255
1256         /*
1257          * If there is no hop-by-hop options header, allocate new one.
1258          * If there is one but it doesn't have enough space to store the
1259          * jumbo payload option, allocate a cluster to store the whole options.
1260          * Otherwise, use it to store the options.
1261          */
1262         if (exthdrs->ip6e_hbh == 0) {
1263                 MGET(mopt, M_DONTWAIT, MT_DATA);
1264                 if (mopt == 0)
1265                         return (ENOBUFS);
1266                 mopt->m_len = JUMBOOPTLEN;
1267                 optbuf = mtod(mopt, u_char *);
1268                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1269                 exthdrs->ip6e_hbh = mopt;
1270         } else {
1271                 struct ip6_hbh *hbh;
1272
1273                 mopt = exthdrs->ip6e_hbh;
1274                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1275                         /*
1276                          * XXX assumption:
1277                          * - exthdrs->ip6e_hbh is not referenced from places
1278                          *   other than exthdrs.
1279                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1280                          */
1281                         int oldoptlen = mopt->m_len;
1282                         struct mbuf *n;
1283
1284                         /*
1285                          * XXX: give up if the whole (new) hbh header does
1286                          * not fit even in an mbuf cluster.
1287                          */
1288                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1289                                 return (ENOBUFS);
1290
1291                         /*
1292                          * As a consequence, we must always prepare a cluster
1293                          * at this point.
1294                          */
1295                         MGET(n, M_DONTWAIT, MT_DATA);
1296                         if (n) {
1297                                 MCLGET(n, M_DONTWAIT);
1298                                 if ((n->m_flags & M_EXT) == 0) {
1299                                         m_freem(n);
1300                                         n = NULL;
1301                                 }
1302                         }
1303                         if (!n)
1304                                 return (ENOBUFS);
1305                         n->m_len = oldoptlen + JUMBOOPTLEN;
1306                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1307                             oldoptlen);
1308                         optbuf = mtod(n, caddr_t) + oldoptlen;
1309                         m_freem(mopt);
1310                         mopt = exthdrs->ip6e_hbh = n;
1311                 } else {
1312                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1313                         mopt->m_len += JUMBOOPTLEN;
1314                 }
1315                 optbuf[0] = IP6OPT_PADN;
1316                 optbuf[1] = 1;
1317
1318                 /*
1319                  * Adjust the header length according to the pad and
1320                  * the jumbo payload option.
1321                  */
1322                 hbh = mtod(mopt, struct ip6_hbh *);
1323                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1324         }
1325
1326         /* fill in the option. */
1327         optbuf[2] = IP6OPT_JUMBO;
1328         optbuf[3] = 4;
1329         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1330         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1331
1332         /* finally, adjust the packet header length */
1333         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1334
1335         return (0);
1336 #undef JUMBOOPTLEN
1337 }
1338
1339 /*
1340  * Insert fragment header and copy unfragmentable header portions.
1341  */
1342 static int
1343 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1344     struct ip6_frag **frghdrp)
1345 {
1346         struct mbuf *n, *mlast;
1347
1348         if (hlen > sizeof(struct ip6_hdr)) {
1349                 n = m_copym(m0, sizeof(struct ip6_hdr),
1350                     hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1351                 if (n == 0)
1352                         return (ENOBUFS);
1353                 m->m_next = n;
1354         } else
1355                 n = m;
1356
1357         /* Search for the last mbuf of unfragmentable part. */
1358         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1359                 ;
1360
1361         if ((mlast->m_flags & M_EXT) == 0 &&
1362             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1363                 /* use the trailing space of the last mbuf for the fragment hdr */
1364                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1365                     mlast->m_len);
1366                 mlast->m_len += sizeof(struct ip6_frag);
1367                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1368         } else {
1369                 /* allocate a new mbuf for the fragment header */
1370                 struct mbuf *mfrg;
1371
1372                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1373                 if (mfrg == 0)
1374                         return (ENOBUFS);
1375                 mfrg->m_len = sizeof(struct ip6_frag);
1376                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1377                 mlast->m_next = mfrg;
1378         }
1379
1380         return (0);
1381 }
1382
1383 static int
1384 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1385     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1386     int *alwaysfragp, u_int fibnum)
1387 {
1388         u_int32_t mtu = 0;
1389         int alwaysfrag = 0;
1390         int error = 0;
1391
1392         if (ro_pmtu != ro) {
1393                 /* The first hop and the final destination may differ. */
1394                 struct sockaddr_in6 *sa6_dst =
1395                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1396                 if (ro_pmtu->ro_rt &&
1397                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1398                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1399                         RTFREE(ro_pmtu->ro_rt);
1400                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1401                 }
1402                 if (ro_pmtu->ro_rt == NULL) {
1403                         bzero(sa6_dst, sizeof(*sa6_dst));
1404                         sa6_dst->sin6_family = AF_INET6;
1405                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1406                         sa6_dst->sin6_addr = *dst;
1407
1408                         in6_rtalloc(ro_pmtu, fibnum);
1409                 }
1410         }
1411         if (ro_pmtu->ro_rt) {
1412                 u_int32_t ifmtu;
1413                 struct in_conninfo inc;
1414
1415                 bzero(&inc, sizeof(inc));
1416                 inc.inc_flags |= INC_ISIPV6;
1417                 inc.inc6_faddr = *dst;
1418
1419                 if (ifp == NULL)
1420                         ifp = ro_pmtu->ro_rt->rt_ifp;
1421                 ifmtu = IN6_LINKMTU(ifp);
1422                 mtu = tcp_hc_getmtu(&inc);
1423                 if (mtu)
1424                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1425                 else
1426                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1427                 if (mtu == 0)
1428                         mtu = ifmtu;
1429                 else if (mtu < IPV6_MMTU) {
1430                         /*
1431                          * RFC2460 section 5, last paragraph:
1432                          * if we record ICMPv6 too big message with
1433                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1434                          * or smaller, with framgent header attached.
1435                          * (fragment header is needed regardless from the
1436                          * packet size, for translators to identify packets)
1437                          */
1438                         alwaysfrag = 1;
1439                         mtu = IPV6_MMTU;
1440                 } else if (mtu > ifmtu) {
1441                         /*
1442                          * The MTU on the route is larger than the MTU on
1443                          * the interface!  This shouldn't happen, unless the
1444                          * MTU of the interface has been changed after the
1445                          * interface was brought up.  Change the MTU in the
1446                          * route to match the interface MTU (as long as the
1447                          * field isn't locked).
1448                          */
1449                         mtu = ifmtu;
1450                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1451                 }
1452         } else if (ifp) {
1453                 mtu = IN6_LINKMTU(ifp);
1454         } else
1455                 error = EHOSTUNREACH; /* XXX */
1456
1457         *mtup = mtu;
1458         if (alwaysfragp)
1459                 *alwaysfragp = alwaysfrag;
1460         return (error);
1461 }
1462
1463 /*
1464  * IP6 socket option processing.
1465  */
1466 int
1467 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1468 {
1469         int optdatalen, uproto;
1470         void *optdata;
1471         struct inpcb *in6p = sotoinpcb(so);
1472         int error, optval;
1473         int level, op, optname;
1474         int optlen;
1475         struct thread *td;
1476
1477         level = sopt->sopt_level;
1478         op = sopt->sopt_dir;
1479         optname = sopt->sopt_name;
1480         optlen = sopt->sopt_valsize;
1481         td = sopt->sopt_td;
1482         error = 0;
1483         optval = 0;
1484         uproto = (int)so->so_proto->pr_protocol;
1485
1486         if (level != IPPROTO_IPV6) {
1487                 error = EINVAL;
1488
1489                 if (sopt->sopt_level == SOL_SOCKET &&
1490                     sopt->sopt_dir == SOPT_SET) {
1491                         switch (sopt->sopt_name) {
1492                         case SO_REUSEADDR:
1493                                 INP_WLOCK(in6p);
1494                                 if ((so->so_options & SO_REUSEADDR) != 0)
1495                                         in6p->inp_flags2 |= INP_REUSEADDR;
1496                                 else
1497                                         in6p->inp_flags2 &= ~INP_REUSEADDR;
1498                                 INP_WUNLOCK(in6p);
1499                                 error = 0;
1500                                 break;
1501                         case SO_REUSEPORT:
1502                                 INP_WLOCK(in6p);
1503                                 if ((so->so_options & SO_REUSEPORT) != 0)
1504                                         in6p->inp_flags2 |= INP_REUSEPORT;
1505                                 else
1506                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1507                                 INP_WUNLOCK(in6p);
1508                                 error = 0;
1509                                 break;
1510                         case SO_SETFIB:
1511                                 INP_WLOCK(in6p);
1512                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1513                                 INP_WUNLOCK(in6p);
1514                                 error = 0;
1515                                 break;
1516                         default:
1517                                 break;
1518                         }
1519                 }
1520         } else {                /* level == IPPROTO_IPV6 */
1521                 switch (op) {
1522
1523                 case SOPT_SET:
1524                         switch (optname) {
1525                         case IPV6_2292PKTOPTIONS:
1526 #ifdef IPV6_PKTOPTIONS
1527                         case IPV6_PKTOPTIONS:
1528 #endif
1529                         {
1530                                 struct mbuf *m;
1531
1532                                 error = soopt_getm(sopt, &m); /* XXX */
1533                                 if (error != 0)
1534                                         break;
1535                                 error = soopt_mcopyin(sopt, m); /* XXX */
1536                                 if (error != 0)
1537                                         break;
1538                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1539                                                     m, so, sopt);
1540                                 m_freem(m); /* XXX */
1541                                 break;
1542                         }
1543
1544                         /*
1545                          * Use of some Hop-by-Hop options or some
1546                          * Destination options, might require special
1547                          * privilege.  That is, normal applications
1548                          * (without special privilege) might be forbidden
1549                          * from setting certain options in outgoing packets,
1550                          * and might never see certain options in received
1551                          * packets. [RFC 2292 Section 6]
1552                          * KAME specific note:
1553                          *  KAME prevents non-privileged users from sending or
1554                          *  receiving ANY hbh/dst options in order to avoid
1555                          *  overhead of parsing options in the kernel.
1556                          */
1557                         case IPV6_RECVHOPOPTS:
1558                         case IPV6_RECVDSTOPTS:
1559                         case IPV6_RECVRTHDRDSTOPTS:
1560                                 if (td != NULL) {
1561                                         error = priv_check(td,
1562                                             PRIV_NETINET_SETHDROPTS);
1563                                         if (error)
1564                                                 break;
1565                                 }
1566                                 /* FALLTHROUGH */
1567                         case IPV6_UNICAST_HOPS:
1568                         case IPV6_HOPLIMIT:
1569                         case IPV6_FAITH:
1570
1571                         case IPV6_RECVPKTINFO:
1572                         case IPV6_RECVHOPLIMIT:
1573                         case IPV6_RECVRTHDR:
1574                         case IPV6_RECVPATHMTU:
1575                         case IPV6_RECVTCLASS:
1576                         case IPV6_V6ONLY:
1577                         case IPV6_AUTOFLOWLABEL:
1578                         case IPV6_BINDANY:
1579                                 if (optname == IPV6_BINDANY && td != NULL) {
1580                                         error = priv_check(td,
1581                                             PRIV_NETINET_BINDANY);
1582                                         if (error)
1583                                                 break;
1584                                 }
1585
1586                                 if (optlen != sizeof(int)) {
1587                                         error = EINVAL;
1588                                         break;
1589                                 }
1590                                 error = sooptcopyin(sopt, &optval,
1591                                         sizeof optval, sizeof optval);
1592                                 if (error)
1593                                         break;
1594                                 switch (optname) {
1595
1596                                 case IPV6_UNICAST_HOPS:
1597                                         if (optval < -1 || optval >= 256)
1598                                                 error = EINVAL;
1599                                         else {
1600                                                 /* -1 = kernel default */
1601                                                 in6p->in6p_hops = optval;
1602                                                 if ((in6p->inp_vflag &
1603                                                      INP_IPV4) != 0)
1604                                                         in6p->inp_ip_ttl = optval;
1605                                         }
1606                                         break;
1607 #define OPTSET(bit) \
1608 do { \
1609         INP_WLOCK(in6p); \
1610         if (optval) \
1611                 in6p->inp_flags |= (bit); \
1612         else \
1613                 in6p->inp_flags &= ~(bit); \
1614         INP_WUNLOCK(in6p); \
1615 } while (/*CONSTCOND*/ 0)
1616 #define OPTSET2292(bit) \
1617 do { \
1618         INP_WLOCK(in6p); \
1619         in6p->inp_flags |= IN6P_RFC2292; \
1620         if (optval) \
1621                 in6p->inp_flags |= (bit); \
1622         else \
1623                 in6p->inp_flags &= ~(bit); \
1624         INP_WUNLOCK(in6p); \
1625 } while (/*CONSTCOND*/ 0)
1626 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1627
1628                                 case IPV6_RECVPKTINFO:
1629                                         /* cannot mix with RFC2292 */
1630                                         if (OPTBIT(IN6P_RFC2292)) {
1631                                                 error = EINVAL;
1632                                                 break;
1633                                         }
1634                                         OPTSET(IN6P_PKTINFO);
1635                                         break;
1636
1637                                 case IPV6_HOPLIMIT:
1638                                 {
1639                                         struct ip6_pktopts **optp;
1640
1641                                         /* cannot mix with RFC2292 */
1642                                         if (OPTBIT(IN6P_RFC2292)) {
1643                                                 error = EINVAL;
1644                                                 break;
1645                                         }
1646                                         optp = &in6p->in6p_outputopts;
1647                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1648                                             (u_char *)&optval, sizeof(optval),
1649                                             optp, (td != NULL) ? td->td_ucred :
1650                                             NULL, uproto);
1651                                         break;
1652                                 }
1653
1654                                 case IPV6_RECVHOPLIMIT:
1655                                         /* cannot mix with RFC2292 */
1656                                         if (OPTBIT(IN6P_RFC2292)) {
1657                                                 error = EINVAL;
1658                                                 break;
1659                                         }
1660                                         OPTSET(IN6P_HOPLIMIT);
1661                                         break;
1662
1663                                 case IPV6_RECVHOPOPTS:
1664                                         /* cannot mix with RFC2292 */
1665                                         if (OPTBIT(IN6P_RFC2292)) {
1666                                                 error = EINVAL;
1667                                                 break;
1668                                         }
1669                                         OPTSET(IN6P_HOPOPTS);
1670                                         break;
1671
1672                                 case IPV6_RECVDSTOPTS:
1673                                         /* cannot mix with RFC2292 */
1674                                         if (OPTBIT(IN6P_RFC2292)) {
1675                                                 error = EINVAL;
1676                                                 break;
1677                                         }
1678                                         OPTSET(IN6P_DSTOPTS);
1679                                         break;
1680
1681                                 case IPV6_RECVRTHDRDSTOPTS:
1682                                         /* cannot mix with RFC2292 */
1683                                         if (OPTBIT(IN6P_RFC2292)) {
1684                                                 error = EINVAL;
1685                                                 break;
1686                                         }
1687                                         OPTSET(IN6P_RTHDRDSTOPTS);
1688                                         break;
1689
1690                                 case IPV6_RECVRTHDR:
1691                                         /* cannot mix with RFC2292 */
1692                                         if (OPTBIT(IN6P_RFC2292)) {
1693                                                 error = EINVAL;
1694                                                 break;
1695                                         }
1696                                         OPTSET(IN6P_RTHDR);
1697                                         break;
1698
1699                                 case IPV6_FAITH:
1700                                         OPTSET(INP_FAITH);
1701                                         break;
1702
1703                                 case IPV6_RECVPATHMTU:
1704                                         /*
1705                                          * We ignore this option for TCP
1706                                          * sockets.
1707                                          * (RFC3542 leaves this case
1708                                          * unspecified.)
1709                                          */
1710                                         if (uproto != IPPROTO_TCP)
1711                                                 OPTSET(IN6P_MTU);
1712                                         break;
1713
1714                                 case IPV6_V6ONLY:
1715                                         /*
1716                                          * make setsockopt(IPV6_V6ONLY)
1717                                          * available only prior to bind(2).
1718                                          * see ipng mailing list, Jun 22 2001.
1719                                          */
1720                                         if (in6p->inp_lport ||
1721                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1722                                                 error = EINVAL;
1723                                                 break;
1724                                         }
1725                                         OPTSET(IN6P_IPV6_V6ONLY);
1726                                         if (optval)
1727                                                 in6p->inp_vflag &= ~INP_IPV4;
1728                                         else
1729                                                 in6p->inp_vflag |= INP_IPV4;
1730                                         break;
1731                                 case IPV6_RECVTCLASS:
1732                                         /* cannot mix with RFC2292 XXX */
1733                                         if (OPTBIT(IN6P_RFC2292)) {
1734                                                 error = EINVAL;
1735                                                 break;
1736                                         }
1737                                         OPTSET(IN6P_TCLASS);
1738                                         break;
1739                                 case IPV6_AUTOFLOWLABEL:
1740                                         OPTSET(IN6P_AUTOFLOWLABEL);
1741                                         break;
1742
1743                                 case IPV6_BINDANY:
1744                                         OPTSET(INP_BINDANY);
1745                                         break;
1746                                 }
1747                                 break;
1748
1749                         case IPV6_TCLASS:
1750                         case IPV6_DONTFRAG:
1751                         case IPV6_USE_MIN_MTU:
1752                         case IPV6_PREFER_TEMPADDR:
1753                                 if (optlen != sizeof(optval)) {
1754                                         error = EINVAL;
1755                                         break;
1756                                 }
1757                                 error = sooptcopyin(sopt, &optval,
1758                                         sizeof optval, sizeof optval);
1759                                 if (error)
1760                                         break;
1761                                 {
1762                                         struct ip6_pktopts **optp;
1763                                         optp = &in6p->in6p_outputopts;
1764                                         error = ip6_pcbopt(optname,
1765                                             (u_char *)&optval, sizeof(optval),
1766                                             optp, (td != NULL) ? td->td_ucred :
1767                                             NULL, uproto);
1768                                         break;
1769                                 }
1770
1771                         case IPV6_2292PKTINFO:
1772                         case IPV6_2292HOPLIMIT:
1773                         case IPV6_2292HOPOPTS:
1774                         case IPV6_2292DSTOPTS:
1775                         case IPV6_2292RTHDR:
1776                                 /* RFC 2292 */
1777                                 if (optlen != sizeof(int)) {
1778                                         error = EINVAL;
1779                                         break;
1780                                 }
1781                                 error = sooptcopyin(sopt, &optval,
1782                                         sizeof optval, sizeof optval);
1783                                 if (error)
1784                                         break;
1785                                 switch (optname) {
1786                                 case IPV6_2292PKTINFO:
1787                                         OPTSET2292(IN6P_PKTINFO);
1788                                         break;
1789                                 case IPV6_2292HOPLIMIT:
1790                                         OPTSET2292(IN6P_HOPLIMIT);
1791                                         break;
1792                                 case IPV6_2292HOPOPTS:
1793                                         /*
1794                                          * Check super-user privilege.
1795                                          * See comments for IPV6_RECVHOPOPTS.
1796                                          */
1797                                         if (td != NULL) {
1798                                                 error = priv_check(td,
1799                                                     PRIV_NETINET_SETHDROPTS);
1800                                                 if (error)
1801                                                         return (error);
1802                                         }
1803                                         OPTSET2292(IN6P_HOPOPTS);
1804                                         break;
1805                                 case IPV6_2292DSTOPTS:
1806                                         if (td != NULL) {
1807                                                 error = priv_check(td,
1808                                                     PRIV_NETINET_SETHDROPTS);
1809                                                 if (error)
1810                                                         return (error);
1811                                         }
1812                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1813                                         break;
1814                                 case IPV6_2292RTHDR:
1815                                         OPTSET2292(IN6P_RTHDR);
1816                                         break;
1817                                 }
1818                                 break;
1819                         case IPV6_PKTINFO:
1820                         case IPV6_HOPOPTS:
1821                         case IPV6_RTHDR:
1822                         case IPV6_DSTOPTS:
1823                         case IPV6_RTHDRDSTOPTS:
1824                         case IPV6_NEXTHOP:
1825                         {
1826                                 /* new advanced API (RFC3542) */
1827                                 u_char *optbuf;
1828                                 u_char optbuf_storage[MCLBYTES];
1829                                 int optlen;
1830                                 struct ip6_pktopts **optp;
1831
1832                                 /* cannot mix with RFC2292 */
1833                                 if (OPTBIT(IN6P_RFC2292)) {
1834                                         error = EINVAL;
1835                                         break;
1836                                 }
1837
1838                                 /*
1839                                  * We only ensure valsize is not too large
1840                                  * here.  Further validation will be done
1841                                  * later.
1842                                  */
1843                                 error = sooptcopyin(sopt, optbuf_storage,
1844                                     sizeof(optbuf_storage), 0);
1845                                 if (error)
1846                                         break;
1847                                 optlen = sopt->sopt_valsize;
1848                                 optbuf = optbuf_storage;
1849                                 optp = &in6p->in6p_outputopts;
1850                                 error = ip6_pcbopt(optname, optbuf, optlen,
1851                                     optp, (td != NULL) ? td->td_ucred : NULL,
1852                                     uproto);
1853                                 break;
1854                         }
1855 #undef OPTSET
1856
1857                         case IPV6_MULTICAST_IF:
1858                         case IPV6_MULTICAST_HOPS:
1859                         case IPV6_MULTICAST_LOOP:
1860                         case IPV6_JOIN_GROUP:
1861                         case IPV6_LEAVE_GROUP:
1862                         case IPV6_MSFILTER:
1863                         case MCAST_BLOCK_SOURCE:
1864                         case MCAST_UNBLOCK_SOURCE:
1865                         case MCAST_JOIN_GROUP:
1866                         case MCAST_LEAVE_GROUP:
1867                         case MCAST_JOIN_SOURCE_GROUP:
1868                         case MCAST_LEAVE_SOURCE_GROUP:
1869                                 error = ip6_setmoptions(in6p, sopt);
1870                                 break;
1871
1872                         case IPV6_PORTRANGE:
1873                                 error = sooptcopyin(sopt, &optval,
1874                                     sizeof optval, sizeof optval);
1875                                 if (error)
1876                                         break;
1877
1878                                 INP_WLOCK(in6p);
1879                                 switch (optval) {
1880                                 case IPV6_PORTRANGE_DEFAULT:
1881                                         in6p->inp_flags &= ~(INP_LOWPORT);
1882                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1883                                         break;
1884
1885                                 case IPV6_PORTRANGE_HIGH:
1886                                         in6p->inp_flags &= ~(INP_LOWPORT);
1887                                         in6p->inp_flags |= INP_HIGHPORT;
1888                                         break;
1889
1890                                 case IPV6_PORTRANGE_LOW:
1891                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1892                                         in6p->inp_flags |= INP_LOWPORT;
1893                                         break;
1894
1895                                 default:
1896                                         error = EINVAL;
1897                                         break;
1898                                 }
1899                                 INP_WUNLOCK(in6p);
1900                                 break;
1901
1902 #ifdef IPSEC
1903                         case IPV6_IPSEC_POLICY:
1904                         {
1905                                 caddr_t req;
1906                                 struct mbuf *m;
1907
1908                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1909                                         break;
1910                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1911                                         break;
1912                                 req = mtod(m, caddr_t);
1913                                 error = ipsec_set_policy(in6p, optname, req,
1914                                     m->m_len, (sopt->sopt_td != NULL) ?
1915                                     sopt->sopt_td->td_ucred : NULL);
1916                                 m_freem(m);
1917                                 break;
1918                         }
1919 #endif /* IPSEC */
1920
1921                         default:
1922                                 error = ENOPROTOOPT;
1923                                 break;
1924                         }
1925                         break;
1926
1927                 case SOPT_GET:
1928                         switch (optname) {
1929
1930                         case IPV6_2292PKTOPTIONS:
1931 #ifdef IPV6_PKTOPTIONS
1932                         case IPV6_PKTOPTIONS:
1933 #endif
1934                                 /*
1935                                  * RFC3542 (effectively) deprecated the
1936                                  * semantics of the 2292-style pktoptions.
1937                                  * Since it was not reliable in nature (i.e.,
1938                                  * applications had to expect the lack of some
1939                                  * information after all), it would make sense
1940                                  * to simplify this part by always returning
1941                                  * empty data.
1942                                  */
1943                                 sopt->sopt_valsize = 0;
1944                                 break;
1945
1946                         case IPV6_RECVHOPOPTS:
1947                         case IPV6_RECVDSTOPTS:
1948                         case IPV6_RECVRTHDRDSTOPTS:
1949                         case IPV6_UNICAST_HOPS:
1950                         case IPV6_RECVPKTINFO:
1951                         case IPV6_RECVHOPLIMIT:
1952                         case IPV6_RECVRTHDR:
1953                         case IPV6_RECVPATHMTU:
1954
1955                         case IPV6_FAITH:
1956                         case IPV6_V6ONLY:
1957                         case IPV6_PORTRANGE:
1958                         case IPV6_RECVTCLASS:
1959                         case IPV6_AUTOFLOWLABEL:
1960                         case IPV6_BINDANY:
1961                                 switch (optname) {
1962
1963                                 case IPV6_RECVHOPOPTS:
1964                                         optval = OPTBIT(IN6P_HOPOPTS);
1965                                         break;
1966
1967                                 case IPV6_RECVDSTOPTS:
1968                                         optval = OPTBIT(IN6P_DSTOPTS);
1969                                         break;
1970
1971                                 case IPV6_RECVRTHDRDSTOPTS:
1972                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1973                                         break;
1974
1975                                 case IPV6_UNICAST_HOPS:
1976                                         optval = in6p->in6p_hops;
1977                                         break;
1978
1979                                 case IPV6_RECVPKTINFO:
1980                                         optval = OPTBIT(IN6P_PKTINFO);
1981                                         break;
1982
1983                                 case IPV6_RECVHOPLIMIT:
1984                                         optval = OPTBIT(IN6P_HOPLIMIT);
1985                                         break;
1986
1987                                 case IPV6_RECVRTHDR:
1988                                         optval = OPTBIT(IN6P_RTHDR);
1989                                         break;
1990
1991                                 case IPV6_RECVPATHMTU:
1992                                         optval = OPTBIT(IN6P_MTU);
1993                                         break;
1994
1995                                 case IPV6_FAITH:
1996                                         optval = OPTBIT(INP_FAITH);
1997                                         break;
1998
1999                                 case IPV6_V6ONLY:
2000                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2001                                         break;
2002
2003                                 case IPV6_PORTRANGE:
2004                                     {
2005                                         int flags;
2006                                         flags = in6p->inp_flags;
2007                                         if (flags & INP_HIGHPORT)
2008                                                 optval = IPV6_PORTRANGE_HIGH;
2009                                         else if (flags & INP_LOWPORT)
2010                                                 optval = IPV6_PORTRANGE_LOW;
2011                                         else
2012                                                 optval = 0;
2013                                         break;
2014                                     }
2015                                 case IPV6_RECVTCLASS:
2016                                         optval = OPTBIT(IN6P_TCLASS);
2017                                         break;
2018
2019                                 case IPV6_AUTOFLOWLABEL:
2020                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2021                                         break;
2022
2023                                 case IPV6_BINDANY:
2024                                         optval = OPTBIT(INP_BINDANY);
2025                                         break;
2026                                 }
2027                                 if (error)
2028                                         break;
2029                                 error = sooptcopyout(sopt, &optval,
2030                                         sizeof optval);
2031                                 break;
2032
2033                         case IPV6_PATHMTU:
2034                         {
2035                                 u_long pmtu = 0;
2036                                 struct ip6_mtuinfo mtuinfo;
2037                                 struct route_in6 sro;
2038
2039                                 bzero(&sro, sizeof(sro));
2040
2041                                 if (!(so->so_state & SS_ISCONNECTED))
2042                                         return (ENOTCONN);
2043                                 /*
2044                                  * XXX: we dot not consider the case of source
2045                                  * routing, or optional information to specify
2046                                  * the outgoing interface.
2047                                  */
2048                                 error = ip6_getpmtu(&sro, NULL, NULL,
2049                                     &in6p->in6p_faddr, &pmtu, NULL,
2050                                     so->so_fibnum);
2051                                 if (sro.ro_rt)
2052                                         RTFREE(sro.ro_rt);
2053                                 if (error)
2054                                         break;
2055                                 if (pmtu > IPV6_MAXPACKET)
2056                                         pmtu = IPV6_MAXPACKET;
2057
2058                                 bzero(&mtuinfo, sizeof(mtuinfo));
2059                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2060                                 optdata = (void *)&mtuinfo;
2061                                 optdatalen = sizeof(mtuinfo);
2062                                 error = sooptcopyout(sopt, optdata,
2063                                     optdatalen);
2064                                 break;
2065                         }
2066
2067                         case IPV6_2292PKTINFO:
2068                         case IPV6_2292HOPLIMIT:
2069                         case IPV6_2292HOPOPTS:
2070                         case IPV6_2292RTHDR:
2071                         case IPV6_2292DSTOPTS:
2072                                 switch (optname) {
2073                                 case IPV6_2292PKTINFO:
2074                                         optval = OPTBIT(IN6P_PKTINFO);
2075                                         break;
2076                                 case IPV6_2292HOPLIMIT:
2077                                         optval = OPTBIT(IN6P_HOPLIMIT);
2078                                         break;
2079                                 case IPV6_2292HOPOPTS:
2080                                         optval = OPTBIT(IN6P_HOPOPTS);
2081                                         break;
2082                                 case IPV6_2292RTHDR:
2083                                         optval = OPTBIT(IN6P_RTHDR);
2084                                         break;
2085                                 case IPV6_2292DSTOPTS:
2086                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2087                                         break;
2088                                 }
2089                                 error = sooptcopyout(sopt, &optval,
2090                                     sizeof optval);
2091                                 break;
2092                         case IPV6_PKTINFO:
2093                         case IPV6_HOPOPTS:
2094                         case IPV6_RTHDR:
2095                         case IPV6_DSTOPTS:
2096                         case IPV6_RTHDRDSTOPTS:
2097                         case IPV6_NEXTHOP:
2098                         case IPV6_TCLASS:
2099                         case IPV6_DONTFRAG:
2100                         case IPV6_USE_MIN_MTU:
2101                         case IPV6_PREFER_TEMPADDR:
2102                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2103                                     optname, sopt);
2104                                 break;
2105
2106                         case IPV6_MULTICAST_IF:
2107                         case IPV6_MULTICAST_HOPS:
2108                         case IPV6_MULTICAST_LOOP:
2109                         case IPV6_MSFILTER:
2110                                 error = ip6_getmoptions(in6p, sopt);
2111                                 break;
2112
2113 #ifdef IPSEC
2114                         case IPV6_IPSEC_POLICY:
2115                           {
2116                                 caddr_t req = NULL;
2117                                 size_t len = 0;
2118                                 struct mbuf *m = NULL;
2119                                 struct mbuf **mp = &m;
2120                                 size_t ovalsize = sopt->sopt_valsize;
2121                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2122
2123                                 error = soopt_getm(sopt, &m); /* XXX */
2124                                 if (error != 0)
2125                                         break;
2126                                 error = soopt_mcopyin(sopt, m); /* XXX */
2127                                 if (error != 0)
2128                                         break;
2129                                 sopt->sopt_valsize = ovalsize;
2130                                 sopt->sopt_val = oval;
2131                                 if (m) {
2132                                         req = mtod(m, caddr_t);
2133                                         len = m->m_len;
2134                                 }
2135                                 error = ipsec_get_policy(in6p, req, len, mp);
2136                                 if (error == 0)
2137                                         error = soopt_mcopyout(sopt, m); /* XXX */
2138                                 if (error == 0 && m)
2139                                         m_freem(m);
2140                                 break;
2141                           }
2142 #endif /* IPSEC */
2143
2144                         default:
2145                                 error = ENOPROTOOPT;
2146                                 break;
2147                         }
2148                         break;
2149                 }
2150         }
2151         return (error);
2152 }
2153
2154 int
2155 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2156 {
2157         int error = 0, optval, optlen;
2158         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2159         struct inpcb *in6p = sotoinpcb(so);
2160         int level, op, optname;
2161
2162         level = sopt->sopt_level;
2163         op = sopt->sopt_dir;
2164         optname = sopt->sopt_name;
2165         optlen = sopt->sopt_valsize;
2166
2167         if (level != IPPROTO_IPV6) {
2168                 return (EINVAL);
2169         }
2170
2171         switch (optname) {
2172         case IPV6_CHECKSUM:
2173                 /*
2174                  * For ICMPv6 sockets, no modification allowed for checksum
2175                  * offset, permit "no change" values to help existing apps.
2176                  *
2177                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2178                  * for an ICMPv6 socket will fail."
2179                  * The current behavior does not meet RFC3542.
2180                  */
2181                 switch (op) {
2182                 case SOPT_SET:
2183                         if (optlen != sizeof(int)) {
2184                                 error = EINVAL;
2185                                 break;
2186                         }
2187                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2188                                             sizeof(optval));
2189                         if (error)
2190                                 break;
2191                         if ((optval % 2) != 0) {
2192                                 /* the API assumes even offset values */
2193                                 error = EINVAL;
2194                         } else if (so->so_proto->pr_protocol ==
2195                             IPPROTO_ICMPV6) {
2196                                 if (optval != icmp6off)
2197                                         error = EINVAL;
2198                         } else
2199                                 in6p->in6p_cksum = optval;
2200                         break;
2201
2202                 case SOPT_GET:
2203                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2204                                 optval = icmp6off;
2205                         else
2206                                 optval = in6p->in6p_cksum;
2207
2208                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2209                         break;
2210
2211                 default:
2212                         error = EINVAL;
2213                         break;
2214                 }
2215                 break;
2216
2217         default:
2218                 error = ENOPROTOOPT;
2219                 break;
2220         }
2221
2222         return (error);
2223 }
2224
2225 /*
2226  * Set up IP6 options in pcb for insertion in output packets or
2227  * specifying behavior of outgoing packets.
2228  */
2229 static int
2230 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2231     struct socket *so, struct sockopt *sopt)
2232 {
2233         struct ip6_pktopts *opt = *pktopt;
2234         int error = 0;
2235         struct thread *td = sopt->sopt_td;
2236
2237         /* turn off any old options. */
2238         if (opt) {
2239 #ifdef DIAGNOSTIC
2240                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2241                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2242                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2243                         printf("ip6_pcbopts: all specified options are cleared.\n");
2244 #endif
2245                 ip6_clearpktopts(opt, -1);
2246         } else
2247                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2248         *pktopt = NULL;
2249
2250         if (!m || m->m_len == 0) {
2251                 /*
2252                  * Only turning off any previous options, regardless of
2253                  * whether the opt is just created or given.
2254                  */
2255                 free(opt, M_IP6OPT);
2256                 return (0);
2257         }
2258
2259         /*  set options specified by user. */
2260         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2261             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2262                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2263                 free(opt, M_IP6OPT);
2264                 return (error);
2265         }
2266         *pktopt = opt;
2267         return (0);
2268 }
2269
2270 /*
2271  * initialize ip6_pktopts.  beware that there are non-zero default values in
2272  * the struct.
2273  */
2274 void
2275 ip6_initpktopts(struct ip6_pktopts *opt)
2276 {
2277
2278         bzero(opt, sizeof(*opt));
2279         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2280         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2281         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2282         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2283 }
2284
2285 static int
2286 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2287     struct ucred *cred, int uproto)
2288 {
2289         struct ip6_pktopts *opt;
2290
2291         if (*pktopt == NULL) {
2292                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2293                     M_WAITOK);
2294                 ip6_initpktopts(*pktopt);
2295         }
2296         opt = *pktopt;
2297
2298         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2299 }
2300
2301 static int
2302 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2303 {
2304         void *optdata = NULL;
2305         int optdatalen = 0;
2306         struct ip6_ext *ip6e;
2307         int error = 0;
2308         struct in6_pktinfo null_pktinfo;
2309         int deftclass = 0, on;
2310         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2311         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2312
2313         switch (optname) {
2314         case IPV6_PKTINFO:
2315                 if (pktopt && pktopt->ip6po_pktinfo)
2316                         optdata = (void *)pktopt->ip6po_pktinfo;
2317                 else {
2318                         /* XXX: we don't have to do this every time... */
2319                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2320                         optdata = (void *)&null_pktinfo;
2321                 }
2322                 optdatalen = sizeof(struct in6_pktinfo);
2323                 break;
2324         case IPV6_TCLASS:
2325                 if (pktopt && pktopt->ip6po_tclass >= 0)
2326                         optdata = (void *)&pktopt->ip6po_tclass;
2327                 else
2328                         optdata = (void *)&deftclass;
2329                 optdatalen = sizeof(int);
2330                 break;
2331         case IPV6_HOPOPTS:
2332                 if (pktopt && pktopt->ip6po_hbh) {
2333                         optdata = (void *)pktopt->ip6po_hbh;
2334                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2335                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2336                 }
2337                 break;
2338         case IPV6_RTHDR:
2339                 if (pktopt && pktopt->ip6po_rthdr) {
2340                         optdata = (void *)pktopt->ip6po_rthdr;
2341                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2342                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2343                 }
2344                 break;
2345         case IPV6_RTHDRDSTOPTS:
2346                 if (pktopt && pktopt->ip6po_dest1) {
2347                         optdata = (void *)pktopt->ip6po_dest1;
2348                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2349                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2350                 }
2351                 break;
2352         case IPV6_DSTOPTS:
2353                 if (pktopt && pktopt->ip6po_dest2) {
2354                         optdata = (void *)pktopt->ip6po_dest2;
2355                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2356                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2357                 }
2358                 break;
2359         case IPV6_NEXTHOP:
2360                 if (pktopt && pktopt->ip6po_nexthop) {
2361                         optdata = (void *)pktopt->ip6po_nexthop;
2362                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2363                 }
2364                 break;
2365         case IPV6_USE_MIN_MTU:
2366                 if (pktopt)
2367                         optdata = (void *)&pktopt->ip6po_minmtu;
2368                 else
2369                         optdata = (void *)&defminmtu;
2370                 optdatalen = sizeof(int);
2371                 break;
2372         case IPV6_DONTFRAG:
2373                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2374                         on = 1;
2375                 else
2376                         on = 0;
2377                 optdata = (void *)&on;
2378                 optdatalen = sizeof(on);
2379                 break;
2380         case IPV6_PREFER_TEMPADDR:
2381                 if (pktopt)
2382                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2383                 else
2384                         optdata = (void *)&defpreftemp;
2385                 optdatalen = sizeof(int);
2386                 break;
2387         default:                /* should not happen */
2388 #ifdef DIAGNOSTIC
2389                 panic("ip6_getpcbopt: unexpected option\n");
2390 #endif
2391                 return (ENOPROTOOPT);
2392         }
2393
2394         error = sooptcopyout(sopt, optdata, optdatalen);
2395
2396         return (error);
2397 }
2398
2399 void
2400 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2401 {
2402         if (pktopt == NULL)
2403                 return;
2404
2405         if (optname == -1 || optname == IPV6_PKTINFO) {
2406                 if (pktopt->ip6po_pktinfo)
2407                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2408                 pktopt->ip6po_pktinfo = NULL;
2409         }
2410         if (optname == -1 || optname == IPV6_HOPLIMIT)
2411                 pktopt->ip6po_hlim = -1;
2412         if (optname == -1 || optname == IPV6_TCLASS)
2413                 pktopt->ip6po_tclass = -1;
2414         if (optname == -1 || optname == IPV6_NEXTHOP) {
2415                 if (pktopt->ip6po_nextroute.ro_rt) {
2416                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2417                         pktopt->ip6po_nextroute.ro_rt = NULL;
2418                 }
2419                 if (pktopt->ip6po_nexthop)
2420                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2421                 pktopt->ip6po_nexthop = NULL;
2422         }
2423         if (optname == -1 || optname == IPV6_HOPOPTS) {
2424                 if (pktopt->ip6po_hbh)
2425                         free(pktopt->ip6po_hbh, M_IP6OPT);
2426                 pktopt->ip6po_hbh = NULL;
2427         }
2428         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2429                 if (pktopt->ip6po_dest1)
2430                         free(pktopt->ip6po_dest1, M_IP6OPT);
2431                 pktopt->ip6po_dest1 = NULL;
2432         }
2433         if (optname == -1 || optname == IPV6_RTHDR) {
2434                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2435                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2436                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2437                 if (pktopt->ip6po_route.ro_rt) {
2438                         RTFREE(pktopt->ip6po_route.ro_rt);
2439                         pktopt->ip6po_route.ro_rt = NULL;
2440                 }
2441         }
2442         if (optname == -1 || optname == IPV6_DSTOPTS) {
2443                 if (pktopt->ip6po_dest2)
2444                         free(pktopt->ip6po_dest2, M_IP6OPT);
2445                 pktopt->ip6po_dest2 = NULL;
2446         }
2447 }
2448
2449 #define PKTOPT_EXTHDRCPY(type) \
2450 do {\
2451         if (src->type) {\
2452                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2453                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2454                 if (dst->type == NULL && canwait == M_NOWAIT)\
2455                         goto bad;\
2456                 bcopy(src->type, dst->type, hlen);\
2457         }\
2458 } while (/*CONSTCOND*/ 0)
2459
2460 static int
2461 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2462 {
2463         if (dst == NULL || src == NULL)  {
2464                 printf("ip6_clearpktopts: invalid argument\n");
2465                 return (EINVAL);
2466         }
2467
2468         dst->ip6po_hlim = src->ip6po_hlim;
2469         dst->ip6po_tclass = src->ip6po_tclass;
2470         dst->ip6po_flags = src->ip6po_flags;
2471         dst->ip6po_minmtu = src->ip6po_minmtu;
2472         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2473         if (src->ip6po_pktinfo) {
2474                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2475                     M_IP6OPT, canwait);
2476                 if (dst->ip6po_pktinfo == NULL)
2477                         goto bad;
2478                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2479         }
2480         if (src->ip6po_nexthop) {
2481                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2482                     M_IP6OPT, canwait);
2483                 if (dst->ip6po_nexthop == NULL)
2484                         goto bad;
2485                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2486                     src->ip6po_nexthop->sa_len);
2487         }
2488         PKTOPT_EXTHDRCPY(ip6po_hbh);
2489         PKTOPT_EXTHDRCPY(ip6po_dest1);
2490         PKTOPT_EXTHDRCPY(ip6po_dest2);
2491         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2492         return (0);
2493
2494   bad:
2495         ip6_clearpktopts(dst, -1);
2496         return (ENOBUFS);
2497 }
2498 #undef PKTOPT_EXTHDRCPY
2499
2500 struct ip6_pktopts *
2501 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2502 {
2503         int error;
2504         struct ip6_pktopts *dst;
2505
2506         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2507         if (dst == NULL)
2508                 return (NULL);
2509         ip6_initpktopts(dst);
2510
2511         if ((error = copypktopts(dst, src, canwait)) != 0) {
2512                 free(dst, M_IP6OPT);
2513                 return (NULL);
2514         }
2515
2516         return (dst);
2517 }
2518
2519 void
2520 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2521 {
2522         if (pktopt == NULL)
2523                 return;
2524
2525         ip6_clearpktopts(pktopt, -1);
2526
2527         free(pktopt, M_IP6OPT);
2528 }
2529
2530 /*
2531  * Set IPv6 outgoing packet options based on advanced API.
2532  */
2533 int
2534 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2535     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2536 {
2537         struct cmsghdr *cm = 0;
2538
2539         if (control == NULL || opt == NULL)
2540                 return (EINVAL);
2541
2542         ip6_initpktopts(opt);
2543         if (stickyopt) {
2544                 int error;
2545
2546                 /*
2547                  * If stickyopt is provided, make a local copy of the options
2548                  * for this particular packet, then override them by ancillary
2549                  * objects.
2550                  * XXX: copypktopts() does not copy the cached route to a next
2551                  * hop (if any).  This is not very good in terms of efficiency,
2552                  * but we can allow this since this option should be rarely
2553                  * used.
2554                  */
2555                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2556                         return (error);
2557         }
2558
2559         /*
2560          * XXX: Currently, we assume all the optional information is stored
2561          * in a single mbuf.
2562          */
2563         if (control->m_next)
2564                 return (EINVAL);
2565
2566         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2567             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2568                 int error;
2569
2570                 if (control->m_len < CMSG_LEN(0))
2571                         return (EINVAL);
2572
2573                 cm = mtod(control, struct cmsghdr *);
2574                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2575                         return (EINVAL);
2576                 if (cm->cmsg_level != IPPROTO_IPV6)
2577                         continue;
2578
2579                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2580                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2581                 if (error)
2582                         return (error);
2583         }
2584
2585         return (0);
2586 }
2587
2588 /*
2589  * Set a particular packet option, as a sticky option or an ancillary data
2590  * item.  "len" can be 0 only when it's a sticky option.
2591  * We have 4 cases of combination of "sticky" and "cmsg":
2592  * "sticky=0, cmsg=0": impossible
2593  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2594  * "sticky=1, cmsg=0": RFC3542 socket option
2595  * "sticky=1, cmsg=1": RFC2292 socket option
2596  */
2597 static int
2598 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2599     struct ucred *cred, int sticky, int cmsg, int uproto)
2600 {
2601         int minmtupolicy, preftemp;
2602         int error;
2603
2604         if (!sticky && !cmsg) {
2605 #ifdef DIAGNOSTIC
2606                 printf("ip6_setpktopt: impossible case\n");
2607 #endif
2608                 return (EINVAL);
2609         }
2610
2611         /*
2612          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2613          * not be specified in the context of RFC3542.  Conversely,
2614          * RFC3542 types should not be specified in the context of RFC2292.
2615          */
2616         if (!cmsg) {
2617                 switch (optname) {
2618                 case IPV6_2292PKTINFO:
2619                 case IPV6_2292HOPLIMIT:
2620                 case IPV6_2292NEXTHOP:
2621                 case IPV6_2292HOPOPTS:
2622                 case IPV6_2292DSTOPTS:
2623                 case IPV6_2292RTHDR:
2624                 case IPV6_2292PKTOPTIONS:
2625                         return (ENOPROTOOPT);
2626                 }
2627         }
2628         if (sticky && cmsg) {
2629                 switch (optname) {
2630                 case IPV6_PKTINFO:
2631                 case IPV6_HOPLIMIT:
2632                 case IPV6_NEXTHOP:
2633                 case IPV6_HOPOPTS:
2634                 case IPV6_DSTOPTS:
2635                 case IPV6_RTHDRDSTOPTS:
2636                 case IPV6_RTHDR:
2637                 case IPV6_USE_MIN_MTU:
2638                 case IPV6_DONTFRAG:
2639                 case IPV6_TCLASS:
2640                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2641                         return (ENOPROTOOPT);
2642                 }
2643         }
2644
2645         switch (optname) {
2646         case IPV6_2292PKTINFO:
2647         case IPV6_PKTINFO:
2648         {
2649                 struct ifnet *ifp = NULL;
2650                 struct in6_pktinfo *pktinfo;
2651
2652                 if (len != sizeof(struct in6_pktinfo))
2653                         return (EINVAL);
2654
2655                 pktinfo = (struct in6_pktinfo *)buf;
2656
2657                 /*
2658                  * An application can clear any sticky IPV6_PKTINFO option by
2659                  * doing a "regular" setsockopt with ipi6_addr being
2660                  * in6addr_any and ipi6_ifindex being zero.
2661                  * [RFC 3542, Section 6]
2662                  */
2663                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2664                     pktinfo->ipi6_ifindex == 0 &&
2665                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2666                         ip6_clearpktopts(opt, optname);
2667                         break;
2668                 }
2669
2670                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2671                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2672                         return (EINVAL);
2673                 }
2674
2675                 /* validate the interface index if specified. */
2676                 if (pktinfo->ipi6_ifindex > V_if_index ||
2677                     pktinfo->ipi6_ifindex < 0) {
2678                          return (ENXIO);
2679                 }
2680                 if (pktinfo->ipi6_ifindex) {
2681                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2682                         if (ifp == NULL)
2683                                 return (ENXIO);
2684                 }
2685
2686                 /*
2687                  * We store the address anyway, and let in6_selectsrc()
2688                  * validate the specified address.  This is because ipi6_addr
2689                  * may not have enough information about its scope zone, and
2690                  * we may need additional information (such as outgoing
2691                  * interface or the scope zone of a destination address) to
2692                  * disambiguate the scope.
2693                  * XXX: the delay of the validation may confuse the
2694                  * application when it is used as a sticky option.
2695                  */
2696                 if (opt->ip6po_pktinfo == NULL) {
2697                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2698                             M_IP6OPT, M_NOWAIT);
2699                         if (opt->ip6po_pktinfo == NULL)
2700                                 return (ENOBUFS);
2701                 }
2702                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2703                 break;
2704         }
2705
2706         case IPV6_2292HOPLIMIT:
2707         case IPV6_HOPLIMIT:
2708         {
2709                 int *hlimp;
2710
2711                 /*
2712                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2713                  * to simplify the ordering among hoplimit options.
2714                  */
2715                 if (optname == IPV6_HOPLIMIT && sticky)
2716                         return (ENOPROTOOPT);
2717
2718                 if (len != sizeof(int))
2719                         return (EINVAL);
2720                 hlimp = (int *)buf;
2721                 if (*hlimp < -1 || *hlimp > 255)
2722                         return (EINVAL);
2723
2724                 opt->ip6po_hlim = *hlimp;
2725                 break;
2726         }
2727
2728         case IPV6_TCLASS:
2729         {
2730                 int tclass;
2731
2732                 if (len != sizeof(int))
2733                         return (EINVAL);
2734                 tclass = *(int *)buf;
2735                 if (tclass < -1 || tclass > 255)
2736                         return (EINVAL);
2737
2738                 opt->ip6po_tclass = tclass;
2739                 break;
2740         }
2741
2742         case IPV6_2292NEXTHOP:
2743         case IPV6_NEXTHOP:
2744                 if (cred != NULL) {
2745                         error = priv_check_cred(cred,
2746                             PRIV_NETINET_SETHDROPTS, 0);
2747                         if (error)
2748                                 return (error);
2749                 }
2750
2751                 if (len == 0) { /* just remove the option */
2752                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2753                         break;
2754                 }
2755
2756                 /* check if cmsg_len is large enough for sa_len */
2757                 if (len < sizeof(struct sockaddr) || len < *buf)
2758                         return (EINVAL);
2759
2760                 switch (((struct sockaddr *)buf)->sa_family) {
2761                 case AF_INET6:
2762                 {
2763                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2764                         int error;
2765
2766                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2767                                 return (EINVAL);
2768
2769                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2770                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2771                                 return (EINVAL);
2772                         }
2773                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2774                             != 0) {
2775                                 return (error);
2776                         }
2777                         break;
2778                 }
2779                 case AF_LINK:   /* should eventually be supported */
2780                 default:
2781                         return (EAFNOSUPPORT);
2782                 }
2783
2784                 /* turn off the previous option, then set the new option. */
2785                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2786                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2787                 if (opt->ip6po_nexthop == NULL)
2788                         return (ENOBUFS);
2789                 bcopy(buf, opt->ip6po_nexthop, *buf);
2790                 break;
2791
2792         case IPV6_2292HOPOPTS:
2793         case IPV6_HOPOPTS:
2794         {
2795                 struct ip6_hbh *hbh;
2796                 int hbhlen;
2797
2798                 /*
2799                  * XXX: We don't allow a non-privileged user to set ANY HbH
2800                  * options, since per-option restriction has too much
2801                  * overhead.
2802                  */
2803                 if (cred != NULL) {
2804                         error = priv_check_cred(cred,
2805                             PRIV_NETINET_SETHDROPTS, 0);
2806                         if (error)
2807                                 return (error);
2808                 }
2809
2810                 if (len == 0) {
2811                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2812                         break;  /* just remove the option */
2813                 }
2814
2815                 /* message length validation */
2816                 if (len < sizeof(struct ip6_hbh))
2817                         return (EINVAL);
2818                 hbh = (struct ip6_hbh *)buf;
2819                 hbhlen = (hbh->ip6h_len + 1) << 3;
2820                 if (len != hbhlen)
2821                         return (EINVAL);
2822
2823                 /* turn off the previous option, then set the new option. */
2824                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2825                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2826                 if (opt->ip6po_hbh == NULL)
2827                         return (ENOBUFS);
2828                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2829
2830                 break;
2831         }
2832
2833         case IPV6_2292DSTOPTS:
2834         case IPV6_DSTOPTS:
2835         case IPV6_RTHDRDSTOPTS:
2836         {
2837                 struct ip6_dest *dest, **newdest = NULL;
2838                 int destlen;
2839
2840                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2841                         error = priv_check_cred(cred,
2842                             PRIV_NETINET_SETHDROPTS, 0);
2843                         if (error)
2844                                 return (error);
2845                 }
2846
2847                 if (len == 0) {
2848                         ip6_clearpktopts(opt, optname);
2849                         break;  /* just remove the option */
2850                 }
2851
2852                 /* message length validation */
2853                 if (len < sizeof(struct ip6_dest))
2854                         return (EINVAL);
2855                 dest = (struct ip6_dest *)buf;
2856                 destlen = (dest->ip6d_len + 1) << 3;
2857                 if (len != destlen)
2858                         return (EINVAL);
2859
2860                 /*
2861                  * Determine the position that the destination options header
2862                  * should be inserted; before or after the routing header.
2863                  */
2864                 switch (optname) {
2865                 case IPV6_2292DSTOPTS:
2866                         /*
2867                          * The old advacned API is ambiguous on this point.
2868                          * Our approach is to determine the position based
2869                          * according to the existence of a routing header.
2870                          * Note, however, that this depends on the order of the
2871                          * extension headers in the ancillary data; the 1st
2872                          * part of the destination options header must appear
2873                          * before the routing header in the ancillary data,
2874                          * too.
2875                          * RFC3542 solved the ambiguity by introducing
2876                          * separate ancillary data or option types.
2877                          */
2878                         if (opt->ip6po_rthdr == NULL)
2879                                 newdest = &opt->ip6po_dest1;
2880                         else
2881                                 newdest = &opt->ip6po_dest2;
2882                         break;
2883                 case IPV6_RTHDRDSTOPTS:
2884                         newdest = &opt->ip6po_dest1;
2885                         break;
2886                 case IPV6_DSTOPTS:
2887                         newdest = &opt->ip6po_dest2;
2888                         break;
2889                 }
2890
2891                 /* turn off the previous option, then set the new option. */
2892                 ip6_clearpktopts(opt, optname);
2893                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2894                 if (*newdest == NULL)
2895                         return (ENOBUFS);
2896                 bcopy(dest, *newdest, destlen);
2897
2898                 break;
2899         }
2900
2901         case IPV6_2292RTHDR:
2902         case IPV6_RTHDR:
2903         {
2904                 struct ip6_rthdr *rth;
2905                 int rthlen;
2906
2907                 if (len == 0) {
2908                         ip6_clearpktopts(opt, IPV6_RTHDR);
2909                         break;  /* just remove the option */
2910                 }
2911
2912                 /* message length validation */
2913                 if (len < sizeof(struct ip6_rthdr))
2914                         return (EINVAL);
2915                 rth = (struct ip6_rthdr *)buf;
2916                 rthlen = (rth->ip6r_len + 1) << 3;
2917                 if (len != rthlen)
2918                         return (EINVAL);
2919
2920                 switch (rth->ip6r_type) {
2921                 case IPV6_RTHDR_TYPE_0:
2922                         if (rth->ip6r_len == 0) /* must contain one addr */
2923                                 return (EINVAL);
2924                         if (rth->ip6r_len % 2) /* length must be even */
2925                                 return (EINVAL);
2926                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2927                                 return (EINVAL);
2928                         break;
2929                 default:
2930                         return (EINVAL);        /* not supported */
2931                 }
2932
2933                 /* turn off the previous option */
2934                 ip6_clearpktopts(opt, IPV6_RTHDR);
2935                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2936                 if (opt->ip6po_rthdr == NULL)
2937                         return (ENOBUFS);
2938                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2939
2940                 break;
2941         }
2942
2943         case IPV6_USE_MIN_MTU:
2944                 if (len != sizeof(int))
2945                         return (EINVAL);
2946                 minmtupolicy = *(int *)buf;
2947                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2948                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2949                     minmtupolicy != IP6PO_MINMTU_ALL) {
2950                         return (EINVAL);
2951                 }
2952                 opt->ip6po_minmtu = minmtupolicy;
2953                 break;
2954
2955         case IPV6_DONTFRAG:
2956                 if (len != sizeof(int))
2957                         return (EINVAL);
2958
2959                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2960                         /*
2961                          * we ignore this option for TCP sockets.
2962                          * (RFC3542 leaves this case unspecified.)
2963                          */
2964                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2965                 } else
2966                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2967                 break;
2968
2969         case IPV6_PREFER_TEMPADDR:
2970                 if (len != sizeof(int))
2971                         return (EINVAL);
2972                 preftemp = *(int *)buf;
2973                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2974                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2975                     preftemp != IP6PO_TEMPADDR_PREFER) {
2976                         return (EINVAL);
2977                 }
2978                 opt->ip6po_prefer_tempaddr = preftemp;
2979                 break;
2980
2981         default:
2982                 return (ENOPROTOOPT);
2983         } /* end of switch */
2984
2985         return (0);
2986 }
2987
2988 /*
2989  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2990  * packet to the input queue of a specified interface.  Note that this
2991  * calls the output routine of the loopback "driver", but with an interface
2992  * pointer that might NOT be &loif -- easier than replicating that code here.
2993  */
2994 void
2995 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2996 {
2997         struct mbuf *copym;
2998         struct ip6_hdr *ip6;
2999
3000         copym = m_copy(m, 0, M_COPYALL);
3001         if (copym == NULL)
3002                 return;
3003
3004         /*
3005          * Make sure to deep-copy IPv6 header portion in case the data
3006          * is in an mbuf cluster, so that we can safely override the IPv6
3007          * header portion later.
3008          */
3009         if ((copym->m_flags & M_EXT) != 0 ||
3010             copym->m_len < sizeof(struct ip6_hdr)) {
3011                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3012                 if (copym == NULL)
3013                         return;
3014         }
3015
3016 #ifdef DIAGNOSTIC
3017         if (copym->m_len < sizeof(*ip6)) {
3018                 m_freem(copym);
3019                 return;
3020         }
3021 #endif
3022
3023         ip6 = mtod(copym, struct ip6_hdr *);
3024         /*
3025          * clear embedded scope identifiers if necessary.
3026          * in6_clearscope will touch the addresses only when necessary.
3027          */
3028         in6_clearscope(&ip6->ip6_src);
3029         in6_clearscope(&ip6->ip6_dst);
3030
3031         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3032 }
3033
3034 /*
3035  * Chop IPv6 header off from the payload.
3036  */
3037 static int
3038 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3039 {
3040         struct mbuf *mh;
3041         struct ip6_hdr *ip6;
3042
3043         ip6 = mtod(m, struct ip6_hdr *);
3044         if (m->m_len > sizeof(*ip6)) {
3045                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3046                 if (mh == 0) {
3047                         m_freem(m);
3048                         return ENOBUFS;
3049                 }
3050                 M_MOVE_PKTHDR(mh, m);
3051                 MH_ALIGN(mh, sizeof(*ip6));
3052                 m->m_len -= sizeof(*ip6);
3053                 m->m_data += sizeof(*ip6);
3054                 mh->m_next = m;
3055                 m = mh;
3056                 m->m_len = sizeof(*ip6);
3057                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3058         }
3059         exthdrs->ip6e_ip6 = m;
3060         return 0;
3061 }
3062
3063 /*
3064  * Compute IPv6 extension header length.
3065  */
3066 int
3067 ip6_optlen(struct inpcb *in6p)
3068 {
3069         int len;
3070
3071         if (!in6p->in6p_outputopts)
3072                 return 0;
3073
3074         len = 0;
3075 #define elen(x) \
3076     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3077
3078         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3079         if (in6p->in6p_outputopts->ip6po_rthdr)
3080                 /* dest1 is valid with rthdr only */
3081                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3082         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3083         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3084         return len;
3085 #undef elen
3086 }