sys/netinet6/ip6_output.c

   1 /*-
   2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. Neither the name of the project nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
  30  */
  31
  32 /*-
  33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  34  *      The Regents of the University of California.  All rights reserved.
  35  *
  36  * Redistribution and use in source and binary forms, with or without
  37  * modification, are permitted provided that the following conditions
  38  * are met:
  39  * 1. Redistributions of source code must retain the above copyright
  40  *    notice, this list of conditions and the following disclaimer.
  41  * 2. Redistributions in binary form must reproduce the above copyright
  42  *    notice, this list of conditions and the following disclaimer in the
  43  *    documentation and/or other materials provided with the distribution.
  44  * 4. Neither the name of the University nor the names of its contributors
  45  *    may be used to endorse or promote products derived from this software
  46  *    without specific prior written permission.
  47  *
  48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  58  * SUCH DAMAGE.
  59  *
  60  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  61  */
  62
  63 #include <sys/cdefs.h>
  64 __FBSDID("$FreeBSD$");
  65
  66 #include "opt_inet.h"
  67 #include "opt_inet6.h"
  68 #include "opt_ipfw.h"
  69 #include "opt_ipsec.h"
  70 #include "opt_sctp.h"
  71 #include "opt_route.h"
  72
  73 #include <sys/param.h>
  74 #include <sys/kernel.h>
  75 #include <sys/malloc.h>
  76 #include <sys/mbuf.h>
  77 #include <sys/errno.h>
  78 #include <sys/priv.h>
  79 #include <sys/proc.h>
  80 #include <sys/protosw.h>
  81 #include <sys/socket.h>
  82 #include <sys/socketvar.h>
  83 #include <sys/syslog.h>
  84 #include <sys/ucred.h>
  85
  86 #include <machine/in_cksum.h>
  87
  88 #include <net/if.h>
  89 #include <net/netisr.h>
  90 #include <net/route.h>
  91 #include <net/pfil.h>
  92 #include <net/vnet.h>
  93
  94 #include <netinet/in.h>
  95 #include <netinet/in_var.h>
  96 #include <netinet/ip_var.h>
  97 #include <netinet6/in6_var.h>
  98 #include <netinet/ip6.h>
  99 #include <netinet/icmp6.h>
 100 #include <netinet6/ip6_var.h>
 101 #include <netinet/in_pcb.h>
 102 #include <netinet/tcp_var.h>
 103 #include <netinet6/nd6.h>
 104
 105 #ifdef IPSEC
 106 #include <netipsec/ipsec.h>
 107 #include <netipsec/ipsec6.h>
 108 #include <netipsec/key.h>
 109 #include <netinet6/ip6_ipsec.h>
 110 #endif /* IPSEC */
 111 #ifdef SCTP
 112 #include <netinet/sctp.h>
 113 #include <netinet/sctp_crc32.h>
 114 #endif
 115
 116 #include <netinet6/ip6protosw.h>
 117 #include <netinet6/scope6_var.h>
 118
 119 #ifdef FLOWTABLE
 120 #include <net/flowtable.h>
 121 #endif
 122
 123 extern int in6_mcast_loop;
 124
 125 struct ip6_exthdrs {
 126         struct mbuf *ip6e_ip6;
 127         struct mbuf *ip6e_hbh;
 128         struct mbuf *ip6e_dest1;
 129         struct mbuf *ip6e_rthdr;
 130         struct mbuf *ip6e_dest2;
 131 };
 132
 133 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
 134                            struct ucred *, int));
 135 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
 136         struct socket *, struct sockopt *));
 137 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 138 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *,
 139         struct ucred *, int, int, int));
 140
 141 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 142 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
 143         struct ip6_frag **));
 144 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 145 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 146 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
 147         struct ifnet *, struct in6_addr *, u_long *, int *, u_int));
 148 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 149
 150
 151 /*
 152  * Make an extension header from option data.  hp is the source, and
 153  * mp is the destination.
 154  */
 155 #define MAKE_EXTHDR(hp, mp)                                             \
 156     do {                                                                \
 157         if (hp) {                                                       \
 158                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 159                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 160                     ((eh)->ip6e_len + 1) << 3);                         \
 161                 if (error)                                              \
 162                         goto freehdrs;                                  \
 163         }                                                               \
 164     } while (/*CONSTCOND*/ 0)
 165
 166 /*
 167  * Form a chain of extension headers.
 168  * m is the extension header mbuf
 169  * mp is the previous mbuf in the chain
 170  * p is the next header
 171  * i is the type of option.
 172  */
 173 #define MAKE_CHAIN(m, mp, p, i)\
 174     do {\
 175         if (m) {\
 176                 if (!hdrsplit) \
 177                         panic("assumption failed: hdr not split"); \
 178                 *mtod((m), u_char *) = *(p);\
 179                 *(p) = (i);\
 180                 p = mtod((m), u_char *);\
 181                 (m)->m_next = (mp)->m_next;\
 182                 (mp)->m_next = (m);\
 183                 (mp) = (m);\
 184         }\
 185     } while (/*CONSTCOND*/ 0)
 186
 187 static void
 188 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 189 {
 190         u_short csum;
 191
 192         csum = in_cksum_skip(m, offset + plen, offset);
 193         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 194                 csum = 0xffff;
 195         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 196
 197         if (offset + sizeof(u_short) > m->m_len) {
 198                 printf("%s: delayed m_pullup, m->len: %d off: %d\n",
 199                     __func__, m->m_len, offset);
 200                 /*
 201                  * XXX this should not happen, but if it does, the correct
 202                  * behavior may be to insert the checksum in the appropriate
 203                  * next mbuf in the chain.
 204                  */
 205                 return;
 206         }
 207         *(u_short *)(m->m_data + offset) = csum;
 208 }
 209
 210 /*
 211  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 212  * header (with pri, len, nxt, hlim, src, dst).
 213  * This function may modify ver and hlim only.
 214  * The mbuf chain containing the packet will be freed.
 215  * The mbuf opt, if present, will not be freed.
 216  *
 217  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 218  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 219  * which is rt_rmx.rmx_mtu.
 220  *
 221  * ifpp - XXX: just for statistics
 222  */
 223 int
 224 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 225     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 226     struct ifnet **ifpp, struct inpcb *inp)
 227 {
 228         struct ip6_hdr *ip6, *mhip6;
 229         struct ifnet *ifp, *origifp;
 230         struct mbuf *m = m0;
 231         struct mbuf *mprev = NULL;
 232         int hlen, tlen, len, off;
 233         struct route_in6 ip6route;
 234         struct rtentry *rt = NULL;
 235         struct sockaddr_in6 *dst, src_sa, dst_sa;
 236         struct in6_addr odst;
 237         int error = 0;
 238         struct in6_ifaddr *ia = NULL;
 239         u_long mtu;
 240         int alwaysfrag, dontfrag;
 241         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 242         struct ip6_exthdrs exthdrs;
 243         struct in6_addr finaldst, src0, dst0;
 244         u_int32_t zone;
 245         struct route_in6 *ro_pmtu = NULL;
 246         int flevalid = 0;
 247         int hdrsplit = 0;
 248         int needipsec = 0;
 249         int sw_csum, tso;
 250 #ifdef IPSEC
 251         struct ipsec_output_state state;
 252         struct ip6_rthdr *rh = NULL;
 253         int needipsectun = 0;
 254         int segleft_org = 0;
 255         struct secpolicy *sp = NULL;
 256 #endif /* IPSEC */
 257 #ifdef IPFIREWALL_FORWARD
 258         struct m_tag *fwd_tag;
 259 #endif
 260
 261         ip6 = mtod(m, struct ip6_hdr *);
 262         if (ip6 == NULL) {
 263                 printf ("ip6 is NULL");
 264                 goto bad;
 265         }
 266
 267         if (inp != NULL)
 268                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 269
 270         finaldst = ip6->ip6_dst;
 271         bzero(&exthdrs, sizeof(exthdrs));
 272         if (opt) {
 273                 /* Hop-by-Hop options header */
 274                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 275                 /* Destination options header(1st part) */
 276                 if (opt->ip6po_rthdr) {
 277                         /*
 278                          * Destination options header(1st part)
 279                          * This only makes sense with a routing header.
 280                          * See Section 9.2 of RFC 3542.
 281                          * Disabling this part just for MIP6 convenience is
 282                          * a bad idea.  We need to think carefully about a
 283                          * way to make the advanced API coexist with MIP6
 284                          * options, which might automatically be inserted in
 285                          * the kernel.
 286                          */
 287                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 288                 }
 289                 /* Routing header */
 290                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 291                 /* Destination options header(2nd part) */
 292                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 293         }
 294
 295         /*
 296          * IPSec checking which handles several cases.
 297          * FAST IPSEC: We re-injected the packet.
 298          */
 299 #ifdef IPSEC
 300         switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
 301         {
 302         case 1:                 /* Bad packet */
 303                 goto freehdrs;
 304         case -1:                /* Do IPSec */
 305                 needipsec = 1;
 306         case 0:                 /* No IPSec */
 307         default:
 308                 break;
 309         }
 310 #endif /* IPSEC */
 311
 312         /*
 313          * Calculate the total length of the extension header chain.
 314          * Keep the length of the unfragmentable part for fragmentation.
 315          */
 316         optlen = 0;
 317         if (exthdrs.ip6e_hbh)
 318                 optlen += exthdrs.ip6e_hbh->m_len;
 319         if (exthdrs.ip6e_dest1)
 320                 optlen += exthdrs.ip6e_dest1->m_len;
 321         if (exthdrs.ip6e_rthdr)
 322                 optlen += exthdrs.ip6e_rthdr->m_len;
 323         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 324
 325         /* NOTE: we don't add AH/ESP length here. do that later. */
 326         if (exthdrs.ip6e_dest2)
 327                 optlen += exthdrs.ip6e_dest2->m_len;
 328
 329         /*
 330          * If we need IPsec, or there is at least one extension header,
 331          * separate IP6 header from the payload.
 332          */
 333         if ((needipsec || optlen) && !hdrsplit) {
 334                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 335                         m = NULL;
 336                         goto freehdrs;
 337                 }
 338                 m = exthdrs.ip6e_ip6;
 339                 hdrsplit++;
 340         }
 341
 342         /* adjust pointer */
 343         ip6 = mtod(m, struct ip6_hdr *);
 344
 345         /* adjust mbuf packet header length */
 346         m->m_pkthdr.len += optlen;
 347         plen = m->m_pkthdr.len - sizeof(*ip6);
 348
 349         /* If this is a jumbo payload, insert a jumbo payload option. */
 350         if (plen > IPV6_MAXPACKET) {
 351                 if (!hdrsplit) {
 352                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 353                                 m = NULL;
 354                                 goto freehdrs;
 355                         }
 356                         m = exthdrs.ip6e_ip6;
 357                         hdrsplit++;
 358                 }
 359                 /* adjust pointer */
 360                 ip6 = mtod(m, struct ip6_hdr *);
 361                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 362                         goto freehdrs;
 363                 ip6->ip6_plen = 0;
 364         } else
 365                 ip6->ip6_plen = htons(plen);
 366
 367         /*
 368          * Concatenate headers and fill in next header fields.
 369          * Here we have, on "m"
 370          *      IPv6 payload
 371          * and we insert headers accordingly.  Finally, we should be getting:
 372          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 373          *
 374          * during the header composing process, "m" points to IPv6 header.
 375          * "mprev" points to an extension header prior to esp.
 376          */
 377         u_char *nexthdrp = &ip6->ip6_nxt;
 378         mprev = m;
 379
 380         /*
 381          * we treat dest2 specially.  this makes IPsec processing
 382          * much easier.  the goal here is to make mprev point the
 383          * mbuf prior to dest2.
 384          *
 385          * result: IPv6 dest2 payload
 386          * m and mprev will point to IPv6 header.
 387          */
 388         if (exthdrs.ip6e_dest2) {
 389                 if (!hdrsplit)
 390                         panic("assumption failed: hdr not split");
 391                 exthdrs.ip6e_dest2->m_next = m->m_next;
 392                 m->m_next = exthdrs.ip6e_dest2;
 393                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 394                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 395         }
 396
 397         /*
 398          * result: IPv6 hbh dest1 rthdr dest2 payload
 399          * m will point to IPv6 header.  mprev will point to the
 400          * extension header prior to dest2 (rthdr in the above case).
 401          */
 402         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 403         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 404                    IPPROTO_DSTOPTS);
 405         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 406                    IPPROTO_ROUTING);
 407
 408 #ifdef IPSEC
 409         if (!needipsec)
 410                 goto skip_ipsec2;
 411
 412         /*
 413          * pointers after IPsec headers are not valid any more.
 414          * other pointers need a great care too.
 415          * (IPsec routines should not mangle mbufs prior to AH/ESP)
 416          */
 417         exthdrs.ip6e_dest2 = NULL;
 418
 419         if (exthdrs.ip6e_rthdr) {
 420                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 421                 segleft_org = rh->ip6r_segleft;
 422                 rh->ip6r_segleft = 0;
 423         }
 424
 425         bzero(&state, sizeof(state));
 426         state.m = m;
 427         error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 428                                     &needipsectun);
 429         m = state.m;
 430         if (error == EJUSTRETURN) {
 431                 /*
 432                  * We had a SP with a level of 'use' and no SA. We
 433                  * will just continue to process the packet without
 434                  * IPsec processing.
 435                  */
 436                 ;
 437         } else if (error) {
 438                 /* mbuf is already reclaimed in ipsec6_output_trans. */
 439                 m = NULL;
 440                 switch (error) {
 441                 case EHOSTUNREACH:
 442                 case ENETUNREACH:
 443                 case EMSGSIZE:
 444                 case ENOBUFS:
 445                 case ENOMEM:
 446                         break;
 447                 default:
 448                         printf("[%s:%d] (ipsec): error code %d\n",
 449                             __func__, __LINE__, error);
 450                         /* FALLTHROUGH */
 451                 case ENOENT:
 452                         /* don't show these error codes to the user */
 453                         error = 0;
 454                         break;
 455                 }
 456                 goto bad;
 457         } else if (!needipsectun) {
 458                 /*
 459                  * In the FAST IPSec case we have already
 460                  * re-injected the packet and it has been freed
 461                  * by the ipsec_done() function.  So, just clean
 462                  * up after ourselves.
 463                  */
 464                 m = NULL;
 465                 goto done;
 466         }
 467         if (exthdrs.ip6e_rthdr) {
 468                 /* ah6_output doesn't modify mbuf chain */
 469                 rh->ip6r_segleft = segleft_org;
 470         }
 471 skip_ipsec2:;
 472 #endif /* IPSEC */
 473
 474         /*
 475          * If there is a routing header, discard the packet.
 476          */
 477         if (exthdrs.ip6e_rthdr) {
 478                  error = EINVAL;
 479                  goto bad;
 480         }
 481
 482         /* Source address validation */
 483         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 484             (flags & IPV6_UNSPECSRC) == 0) {
 485                 error = EOPNOTSUPP;
 486                 V_ip6stat.ip6s_badscope++;
 487                 goto bad;
 488         }
 489         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 490                 error = EOPNOTSUPP;
 491                 V_ip6stat.ip6s_badscope++;
 492                 goto bad;
 493         }
 494
 495         V_ip6stat.ip6s_localout++;
 496
 497         /*
 498          * Route packet.
 499          */
 500         if (ro == 0) {
 501                 ro = &ip6route;
 502                 bzero((caddr_t)ro, sizeof(*ro));
 503         }
 504         ro_pmtu = ro;
 505         if (opt && opt->ip6po_rthdr)
 506                 ro = &opt->ip6po_route;
 507         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 508 #ifdef FLOWTABLE
 509         if (ro == &ip6route) {
 510                 struct flentry *fle;
 511
 512                 /*
 513                  * The flow table returns route entries valid for up to 30
 514                  * seconds; we rely on the remainder of ip_output() taking no
 515                  * longer than that long for the stability of ro_rt.  The
 516                  * flow ID assignment must have happened before this point.
 517                  */
 518                 if ((fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6)) != NULL) {
 519                         flow_to_route_in6(fle, ro);
 520                         if (ro->ro_rt != NULL && ro->ro_lle != NULL)
 521                                 flevalid = 1;
 522                 }
 523         }
 524 #endif
 525 again:
 526         /*
 527          * if specified, try to fill in the traffic class field.
 528          * do not override if a non-zero value is already set.
 529          * we check the diffserv field and the ecn field separately.
 530          */
 531         if (opt && opt->ip6po_tclass >= 0) {
 532                 int mask = 0;
 533
 534                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 535                         mask |= 0xfc;
 536                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 537                         mask |= 0x03;
 538                 if (mask != 0)
 539                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 540         }
 541
 542         /* fill in or override the hop limit field, if necessary. */
 543         if (opt && opt->ip6po_hlim != -1)
 544                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 545         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 546                 if (im6o != NULL)
 547                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 548                 else
 549                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 550         }
 551
 552 #ifdef IPSEC
 553         /*
 554          * We may re-inject packets into the stack here.
 555          */
 556         if (needipsec && needipsectun) {
 557                 struct ipsec_output_state state;
 558
 559                 /*
 560                  * All the extension headers will become inaccessible
 561                  * (since they can be encrypted).
 562                  * Don't panic, we need no more updates to extension headers
 563                  * on inner IPv6 packet (since they are now encapsulated).
 564                  *
 565                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 566                  */
 567                 bzero(&exthdrs, sizeof(exthdrs));
 568                 exthdrs.ip6e_ip6 = m;
 569
 570                 bzero(&state, sizeof(state));
 571                 state.m = m;
 572                 state.ro = (struct route *)ro;
 573                 state.dst = (struct sockaddr *)dst;
 574
 575                 error = ipsec6_output_tunnel(&state, sp, flags);
 576
 577                 m = state.m;
 578                 ro = (struct route_in6 *)state.ro;
 579                 dst = (struct sockaddr_in6 *)state.dst;
 580                 if (error == EJUSTRETURN) {
 581                         /*
 582                          * We had a SP with a level of 'use' and no SA. We
 583                          * will just continue to process the packet without
 584                          * IPsec processing.
 585                          */
 586                         ;
 587                 } else if (error) {
 588                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 589                         m0 = m = NULL;
 590                         m = NULL;
 591                         switch (error) {
 592                         case EHOSTUNREACH:
 593                         case ENETUNREACH:
 594                         case EMSGSIZE:
 595                         case ENOBUFS:
 596                         case ENOMEM:
 597                                 break;
 598                         default:
 599                                 printf("[%s:%d] (ipsec): error code %d\n",
 600                                     __func__, __LINE__, error);
 601                                 /* FALLTHROUGH */
 602                         case ENOENT:
 603                                 /* don't show these error codes to the user */
 604                                 error = 0;
 605                                 break;
 606                         }
 607                         goto bad;
 608                 } else {
 609                         /*
 610                          * In the FAST IPSec case we have already
 611                          * re-injected the packet and it has been freed
 612                          * by the ipsec_done() function.  So, just clean
 613                          * up after ourselves.
 614                          */
 615                         m = NULL;
 616                         goto done;
 617                 }
 618
 619                 exthdrs.ip6e_ip6 = m;
 620         }
 621 #endif /* IPSEC */
 622
 623         /* adjust pointer */
 624         ip6 = mtod(m, struct ip6_hdr *);
 625
 626         bzero(&dst_sa, sizeof(dst_sa));
 627         dst_sa.sin6_family = AF_INET6;
 628         dst_sa.sin6_len = sizeof(dst_sa);
 629         dst_sa.sin6_addr = ip6->ip6_dst;
 630         if (flevalid) {
 631                 rt = ro->ro_rt;
 632                 ifp = ro->ro_rt->rt_ifp;
 633         } else if ((error = in6_selectroute_fib(&dst_sa, opt, im6o, ro,
 634             &ifp, &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0) {
 635                 switch (error) {
 636                 case EHOSTUNREACH:
 637                         V_ip6stat.ip6s_noroute++;
 638                         break;
 639                 case EADDRNOTAVAIL:
 640                 default:
 641                         break; /* XXX statistics? */
 642                 }
 643                 if (ifp != NULL)
 644                         in6_ifstat_inc(ifp, ifs6_out_discard);
 645                 goto bad;
 646         }
 647         if (rt == NULL) {
 648                 /*
 649                  * If in6_selectroute() does not return a route entry,
 650                  * dst may not have been updated.
 651                  */
 652                 *dst = dst_sa;  /* XXX */
 653         }
 654
 655         /*
 656          * then rt (for unicast) and ifp must be non-NULL valid values.
 657          */
 658         if ((flags & IPV6_FORWARDING) == 0) {
 659                 /* XXX: the FORWARDING flag can be set for mrouting. */
 660                 in6_ifstat_inc(ifp, ifs6_out_request);
 661         }
 662         if (rt != NULL) {
 663                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 664                 rt->rt_use++;
 665         }
 666
 667
 668         /*
 669          * The outgoing interface must be in the zone of source and
 670          * destination addresses.
 671          */
 672         origifp = ifp;
 673
 674         src0 = ip6->ip6_src;
 675         if (in6_setscope(&src0, origifp, &zone))
 676                 goto badscope;
 677         bzero(&src_sa, sizeof(src_sa));
 678         src_sa.sin6_family = AF_INET6;
 679         src_sa.sin6_len = sizeof(src_sa);
 680         src_sa.sin6_addr = ip6->ip6_src;
 681         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 682                 goto badscope;
 683
 684         dst0 = ip6->ip6_dst;
 685         if (in6_setscope(&dst0, origifp, &zone))
 686                 goto badscope;
 687         /* re-initialize to be sure */
 688         bzero(&dst_sa, sizeof(dst_sa));
 689         dst_sa.sin6_family = AF_INET6;
 690         dst_sa.sin6_len = sizeof(dst_sa);
 691         dst_sa.sin6_addr = ip6->ip6_dst;
 692         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 693                 goto badscope;
 694         }
 695
 696         /* We should use ia_ifp to support the case of
 697          * sending packets to an address of our own.
 698          */
 699         if (ia != NULL && ia->ia_ifp)
 700                 ifp = ia->ia_ifp;
 701
 702         /* scope check is done. */
 703         goto routefound;
 704
 705   badscope:
 706         V_ip6stat.ip6s_badscope++;
 707         in6_ifstat_inc(origifp, ifs6_out_discard);
 708         if (error == 0)
 709                 error = EHOSTUNREACH; /* XXX */
 710         goto bad;
 711
 712   routefound:
 713         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 714                 if (opt && opt->ip6po_nextroute.ro_rt) {
 715                         /*
 716                          * The nexthop is explicitly specified by the
 717                          * application.  We assume the next hop is an IPv6
 718                          * address.
 719                          */
 720                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 721                 }
 722                 else if ((rt->rt_flags & RTF_GATEWAY))
 723                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 724         }
 725
 726         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 727                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 728         } else {
 729                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 730                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 731                 /*
 732                  * Confirm that the outgoing interface supports multicast.
 733                  */
 734                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 735                         V_ip6stat.ip6s_noroute++;
 736                         in6_ifstat_inc(ifp, ifs6_out_discard);
 737                         error = ENETUNREACH;
 738                         goto bad;
 739                 }
 740                 if ((im6o == NULL && in6_mcast_loop) ||
 741                     (im6o && im6o->im6o_multicast_loop)) {
 742                         /*
 743                          * Loop back multicast datagram if not expressly
 744                          * forbidden to do so, even if we have not joined
 745                          * the address; protocols will filter it later,
 746                          * thus deferring a hash lookup and lock acquisition
 747                          * at the expense of an m_copym().
 748                          */
 749                         ip6_mloopback(ifp, m, dst);
 750                 } else {
 751                         /*
 752                          * If we are acting as a multicast router, perform
 753                          * multicast forwarding as if the packet had just
 754                          * arrived on the interface to which we are about
 755                          * to send.  The multicast forwarding function
 756                          * recursively calls this function, using the
 757                          * IPV6_FORWARDING flag to prevent infinite recursion.
 758                          *
 759                          * Multicasts that are looped back by ip6_mloopback(),
 760                          * above, will be forwarded by the ip6_input() routine,
 761                          * if necessary.
 762                          */
 763                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 764                                 /*
 765                                  * XXX: ip6_mforward expects that rcvif is NULL
 766                                  * when it is called from the originating path.
 767                                  * However, it is not always the case, since
 768                                  * some versions of MGETHDR() does not
 769                                  * initialize the field.
 770                                  */
 771                                 m->m_pkthdr.rcvif = NULL;
 772                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 773                                         m_freem(m);
 774                                         goto done;
 775                                 }
 776                         }
 777                 }
 778                 /*
 779                  * Multicasts with a hoplimit of zero may be looped back,
 780                  * above, but must not be transmitted on a network.
 781                  * Also, multicasts addressed to the loopback interface
 782                  * are not sent -- the above call to ip6_mloopback() will
 783                  * loop back a copy if this host actually belongs to the
 784                  * destination group on the loopback interface.
 785                  */
 786                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 787                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 788                         m_freem(m);
 789                         goto done;
 790                 }
 791         }
 792
 793         /*
 794          * Fill the outgoing inteface to tell the upper layer
 795          * to increment per-interface statistics.
 796          */
 797         if (ifpp)
 798                 *ifpp = ifp;
 799
 800         /* Determine path MTU. */
 801         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 802             &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
 803                 goto bad;
 804
 805         /*
 806          * The caller of this function may specify to use the minimum MTU
 807          * in some cases.
 808          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 809          * setting.  The logic is a bit complicated; by default, unicast
 810          * packets will follow path MTU while multicast packets will be sent at
 811          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 812          * including unicast ones will be sent at the minimum MTU.  Multicast
 813          * packets will always be sent at the minimum MTU unless
 814          * IP6PO_MINMTU_DISABLE is explicitly specified.
 815          * See RFC 3542 for more details.
 816          */
 817         if (mtu > IPV6_MMTU) {
 818                 if ((flags & IPV6_MINMTU))
 819                         mtu = IPV6_MMTU;
 820                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 821                         mtu = IPV6_MMTU;
 822                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 823                          (opt == NULL ||
 824                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 825                         mtu = IPV6_MMTU;
 826                 }
 827         }
 828
 829         /*
 830          * clear embedded scope identifiers if necessary.
 831          * in6_clearscope will touch the addresses only when necessary.
 832          */
 833         in6_clearscope(&ip6->ip6_src);
 834         in6_clearscope(&ip6->ip6_dst);
 835
 836         /*
 837          * If the outgoing packet contains a hop-by-hop options header,
 838          * it must be examined and processed even by the source node.
 839          * (RFC 2460, section 4.)
 840          */
 841         if (exthdrs.ip6e_hbh) {
 842                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 843                 u_int32_t dummy; /* XXX unused */
 844                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 845
 846 #ifdef DIAGNOSTIC
 847                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 848                         panic("ip6e_hbh is not contiguous");
 849 #endif
 850                 /*
 851                  *  XXX: if we have to send an ICMPv6 error to the sender,
 852                  *       we need the M_LOOP flag since icmp6_error() expects
 853                  *       the IPv6 and the hop-by-hop options header are
 854                  *       contiguous unless the flag is set.
 855                  */
 856                 m->m_flags |= M_LOOP;
 857                 m->m_pkthdr.rcvif = ifp;
 858                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 859                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 860                     &dummy, &plen) < 0) {
 861                         /* m was already freed at this point */
 862                         error = EINVAL;/* better error? */
 863                         goto done;
 864                 }
 865                 m->m_flags &= ~M_LOOP; /* XXX */
 866                 m->m_pkthdr.rcvif = NULL;
 867         }
 868
 869         /* Jump over all PFIL processing if hooks are not active. */
 870         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 871                 goto passout;
 872
 873         odst = ip6->ip6_dst;
 874         /* Run through list of hooks for output packets. */
 875         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 876         if (error != 0 || m == NULL)
 877                 goto done;
 878         ip6 = mtod(m, struct ip6_hdr *);
 879
 880         /* See if destination IP address was changed by packet filter. */
 881         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 882                 m->m_flags |= M_SKIP_FIREWALL;
 883                 /* If destination is now ourself drop to ip6_input(). */
 884                 if (in6_localip(&ip6->ip6_dst)) {
 885                         m->m_flags |= M_FASTFWD_OURS;
 886                         if (m->m_pkthdr.rcvif == NULL)
 887                                 m->m_pkthdr.rcvif = V_loif;
 888                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 889                                 m->m_pkthdr.csum_flags |=
 890                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 891                                 m->m_pkthdr.csum_data = 0xffff;
 892                         }
 893 #ifdef SCTP
 894                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 895                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 896 #endif
 897                         error = netisr_queue(NETISR_IPV6, m);
 898                         goto done;
 899                 } else
 900                         goto again;     /* Redo the routing table lookup. */
 901         }
 902
 903 #ifdef IPFIREWALL_FORWARD
 904         /* See if local, if yes, send it to netisr. */
 905         if (m->m_flags & M_FASTFWD_OURS) {
 906                 if (m->m_pkthdr.rcvif == NULL)
 907                         m->m_pkthdr.rcvif = V_loif;
 908                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 909                         m->m_pkthdr.csum_flags |=
 910                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 911                         m->m_pkthdr.csum_data = 0xffff;
 912                 }
 913 #ifdef SCTP
 914                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 915                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 916 #endif
 917                 error = netisr_queue(NETISR_IPV6, m);
 918                 goto done;
 919         }
 920         /* Or forward to some other address? */
 921         fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
 922         if (fwd_tag) {
 923                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 924                 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6));
 925                 m->m_flags |= M_SKIP_FIREWALL;
 926                 m_tag_delete(m, fwd_tag);
 927                 goto again;
 928         }
 929 #endif /* IPFIREWALL_FORWARD */
 930
 931 passout:
 932         /*
 933          * Send the packet to the outgoing interface.
 934          * If necessary, do IPv6 fragmentation before sending.
 935          *
 936          * the logic here is rather complex:
 937          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 938          * 1-a: send as is if tlen <= path mtu
 939          * 1-b: fragment if tlen > path mtu
 940          *
 941          * 2: if user asks us not to fragment (dontfrag == 1)
 942          * 2-a: send as is if tlen <= interface mtu
 943          * 2-b: error if tlen > interface mtu
 944          *
 945          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 946          *      always fragment
 947          *
 948          * 4: if dontfrag == 1 && alwaysfrag == 1
 949          *      error, as we cannot handle this conflicting request
 950          */
 951         sw_csum = m->m_pkthdr.csum_flags;
 952         if (!hdrsplit) {
 953                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 954                 sw_csum &= ~ifp->if_hwassist;
 955         } else
 956                 tso = 0;
 957         /*
 958          * If we added extension headers, we will not do TSO and calculate the
 959          * checksums ourselves for now.
 960          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 961          * with ext. hdrs.
 962          */
 963         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 964                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 965                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 966         }
 967 #ifdef SCTP
 968         if (sw_csum & CSUM_SCTP_IPV6) {
 969                 sw_csum &= ~CSUM_SCTP_IPV6;
 970                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 971         }
 972 #endif
 973         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 974         tlen = m->m_pkthdr.len;
 975
 976         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 977                 dontfrag = 1;
 978         else
 979                 dontfrag = 0;
 980         if (dontfrag && alwaysfrag) {   /* case 4 */
 981                 /* conflicting request - can't transmit */
 982                 error = EMSGSIZE;
 983                 goto bad;
 984         }
 985         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 986                 /*
 987                  * Even if the DONTFRAG option is specified, we cannot send the
 988                  * packet when the data length is larger than the MTU of the
 989                  * outgoing interface.
 990                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 991                  * well as returning an error code (the latter is not described
 992                  * in the API spec.)
 993                  */
 994                 u_int32_t mtu32;
 995                 struct ip6ctlparam ip6cp;
 996
 997                 mtu32 = (u_int32_t)mtu;
 998                 bzero(&ip6cp, sizeof(ip6cp));
 999                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1000                 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1001                     (void *)&ip6cp);
1002
1003                 error = EMSGSIZE;
1004                 goto bad;
1005         }
1006
1007         /*
1008          * transmit packet without fragmentation
1009          */
1010         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1011                 struct in6_ifaddr *ia6;
1012
1013                 ip6 = mtod(m, struct ip6_hdr *);
1014                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1015                 if (ia6) {
1016                         /* Record statistics for this interface address. */
1017                         ia6->ia_ifa.if_opackets++;
1018                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1019                         ifa_free(&ia6->ia_ifa);
1020                 }
1021                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1022                 goto done;
1023         }
1024
1025         /*
1026          * try to fragment the packet.  case 1-b and 3
1027          */
1028         if (mtu < IPV6_MMTU) {
1029                 /* path MTU cannot be less than IPV6_MMTU */
1030                 error = EMSGSIZE;
1031                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1032                 goto bad;
1033         } else if (ip6->ip6_plen == 0) {
1034                 /* jumbo payload cannot be fragmented */
1035                 error = EMSGSIZE;
1036                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1037                 goto bad;
1038         } else {
1039                 struct mbuf **mnext, *m_frgpart;
1040                 struct ip6_frag *ip6f;
1041                 u_int32_t id = htonl(ip6_randomid());
1042                 u_char nextproto;
1043
1044                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1045
1046                 /*
1047                  * Too large for the destination or interface;
1048                  * fragment if possible.
1049                  * Must be able to put at least 8 bytes per fragment.
1050                  */
1051                 hlen = unfragpartlen;
1052                 if (mtu > IPV6_MAXPACKET)
1053                         mtu = IPV6_MAXPACKET;
1054
1055                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1056                 if (len < 8) {
1057                         error = EMSGSIZE;
1058                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1059                         goto bad;
1060                 }
1061
1062                 /*
1063                  * Verify that we have any chance at all of being able to queue
1064                  *      the packet or packet fragments
1065                  */
1066                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1067                     < tlen  /* - hlen */)) {
1068                         error = ENOBUFS;
1069                         V_ip6stat.ip6s_odropped++;
1070                         goto bad;
1071                 }
1072
1073
1074                 /*
1075                  * If the interface will not calculate checksums on
1076                  * fragmented packets, then do it here.
1077                  * XXX-BZ handle the hw offloading case.  Need flags.
1078                  */
1079                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1080                         in6_delayed_cksum(m, plen, hlen);
1081                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1082                 }
1083 #ifdef SCTP
1084                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1085                         sctp_delayed_cksum(m, hlen);
1086                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1087                 }
1088 #endif
1089                 mnext = &m->m_nextpkt;
1090
1091                 /*
1092                  * Change the next header field of the last header in the
1093                  * unfragmentable part.
1094                  */
1095                 if (exthdrs.ip6e_rthdr) {
1096                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1097                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1098                 } else if (exthdrs.ip6e_dest1) {
1099                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1100                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1101                 } else if (exthdrs.ip6e_hbh) {
1102                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1103                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1104                 } else {
1105                         nextproto = ip6->ip6_nxt;
1106                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1107                 }
1108
1109                 /*
1110                  * Loop through length of segment after first fragment,
1111                  * make new header and copy data of each part and link onto
1112                  * chain.
1113                  */
1114                 m0 = m;
1115                 for (off = hlen; off < tlen; off += len) {
1116                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1117                         if (!m) {
1118                                 error = ENOBUFS;
1119                                 V_ip6stat.ip6s_odropped++;
1120                                 goto sendorfree;
1121                         }
1122                         m->m_pkthdr.rcvif = NULL;
1123                         m->m_flags = m0->m_flags & M_COPYFLAGS; /* incl. FIB */
1124                         *mnext = m;
1125                         mnext = &m->m_nextpkt;
1126                         m->m_data += max_linkhdr;
1127                         mhip6 = mtod(m, struct ip6_hdr *);
1128                         *mhip6 = *ip6;
1129                         m->m_len = sizeof(*mhip6);
1130                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1131                         if (error) {
1132                                 V_ip6stat.ip6s_odropped++;
1133                                 goto sendorfree;
1134                         }
1135                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1136                         if (off + len >= tlen)
1137                                 len = tlen - off;
1138                         else
1139                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1140                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1141                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1142                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1143                                 error = ENOBUFS;
1144                                 V_ip6stat.ip6s_odropped++;
1145                                 goto sendorfree;
1146                         }
1147                         m_cat(m, m_frgpart);
1148                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1149                         m->m_pkthdr.rcvif = NULL;
1150                         ip6f->ip6f_reserved = 0;
1151                         ip6f->ip6f_ident = id;
1152                         ip6f->ip6f_nxt = nextproto;
1153                         V_ip6stat.ip6s_ofragments++;
1154                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1155                 }
1156
1157                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1158         }
1159
1160         /*
1161          * Remove leading garbages.
1162          */
1163 sendorfree:
1164         m = m0->m_nextpkt;
1165         m0->m_nextpkt = 0;
1166         m_freem(m0);
1167         for (m0 = m; m; m = m0) {
1168                 m0 = m->m_nextpkt;
1169                 m->m_nextpkt = 0;
1170                 if (error == 0) {
1171                         /* Record statistics for this interface address. */
1172                         if (ia) {
1173                                 ia->ia_ifa.if_opackets++;
1174                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1175                         }
1176                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1177                 } else
1178                         m_freem(m);
1179         }
1180
1181         if (error == 0)
1182                 V_ip6stat.ip6s_fragmented++;
1183
1184 done:
1185         if (ro == &ip6route && ro->ro_rt && flevalid == 0) {
1186                 /* brace necessary for RTFREE */
1187                 RTFREE(ro->ro_rt);
1188         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt &&
1189             ((flevalid == 0) || (ro_pmtu != ro))) {
1190                 RTFREE(ro_pmtu->ro_rt);
1191         }
1192 #ifdef IPSEC
1193         if (sp != NULL)
1194                 KEY_FREESP(&sp);
1195 #endif
1196
1197         return (error);
1198
1199 freehdrs:
1200         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1201         m_freem(exthdrs.ip6e_dest1);
1202         m_freem(exthdrs.ip6e_rthdr);
1203         m_freem(exthdrs.ip6e_dest2);
1204         /* FALLTHROUGH */
1205 bad:
1206         if (m)
1207                 m_freem(m);
1208         goto done;
1209 }
1210
1211 static int
1212 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1213 {
1214         struct mbuf *m;
1215
1216         if (hlen > MCLBYTES)
1217                 return (ENOBUFS); /* XXX */
1218
1219         MGET(m, M_DONTWAIT, MT_DATA);
1220         if (!m)
1221                 return (ENOBUFS);
1222
1223         if (hlen > MLEN) {
1224                 MCLGET(m, M_DONTWAIT);
1225                 if ((m->m_flags & M_EXT) == 0) {
1226                         m_free(m);
1227                         return (ENOBUFS);
1228                 }
1229         }
1230         m->m_len = hlen;
1231         if (hdr)
1232                 bcopy(hdr, mtod(m, caddr_t), hlen);
1233
1234         *mp = m;
1235         return (0);
1236 }
1237
1238 /*
1239  * Insert jumbo payload option.
1240  */
1241 static int
1242 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1243 {
1244         struct mbuf *mopt;
1245         u_char *optbuf;
1246         u_int32_t v;
1247
1248 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1249
1250         /*
1251          * If there is no hop-by-hop options header, allocate new one.
1252          * If there is one but it doesn't have enough space to store the
1253          * jumbo payload option, allocate a cluster to store the whole options.
1254          * Otherwise, use it to store the options.
1255          */
1256         if (exthdrs->ip6e_hbh == 0) {
1257                 MGET(mopt, M_DONTWAIT, MT_DATA);
1258                 if (mopt == 0)
1259                         return (ENOBUFS);
1260                 mopt->m_len = JUMBOOPTLEN;
1261                 optbuf = mtod(mopt, u_char *);
1262                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1263                 exthdrs->ip6e_hbh = mopt;
1264         } else {
1265                 struct ip6_hbh *hbh;
1266
1267                 mopt = exthdrs->ip6e_hbh;
1268                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1269                         /*
1270                          * XXX assumption:
1271                          * - exthdrs->ip6e_hbh is not referenced from places
1272                          *   other than exthdrs.
1273                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1274                          */
1275                         int oldoptlen = mopt->m_len;
1276                         struct mbuf *n;
1277
1278                         /*
1279                          * XXX: give up if the whole (new) hbh header does
1280                          * not fit even in an mbuf cluster.
1281                          */
1282                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1283                                 return (ENOBUFS);
1284
1285                         /*
1286                          * As a consequence, we must always prepare a cluster
1287                          * at this point.
1288                          */
1289                         MGET(n, M_DONTWAIT, MT_DATA);
1290                         if (n) {
1291                                 MCLGET(n, M_DONTWAIT);
1292                                 if ((n->m_flags & M_EXT) == 0) {
1293                                         m_freem(n);
1294                                         n = NULL;
1295                                 }
1296                         }
1297                         if (!n)
1298                                 return (ENOBUFS);
1299                         n->m_len = oldoptlen + JUMBOOPTLEN;
1300                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1301                             oldoptlen);
1302                         optbuf = mtod(n, caddr_t) + oldoptlen;
1303                         m_freem(mopt);
1304                         mopt = exthdrs->ip6e_hbh = n;
1305                 } else {
1306                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1307                         mopt->m_len += JUMBOOPTLEN;
1308                 }
1309                 optbuf[0] = IP6OPT_PADN;
1310                 optbuf[1] = 1;
1311
1312                 /*
1313                  * Adjust the header length according to the pad and
1314                  * the jumbo payload option.
1315                  */
1316                 hbh = mtod(mopt, struct ip6_hbh *);
1317                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1318         }
1319
1320         /* fill in the option. */
1321         optbuf[2] = IP6OPT_JUMBO;
1322         optbuf[3] = 4;
1323         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1324         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1325
1326         /* finally, adjust the packet header length */
1327         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1328
1329         return (0);
1330 #undef JUMBOOPTLEN
1331 }
1332
1333 /*
1334  * Insert fragment header and copy unfragmentable header portions.
1335  */
1336 static int
1337 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1338     struct ip6_frag **frghdrp)
1339 {
1340         struct mbuf *n, *mlast;
1341
1342         if (hlen > sizeof(struct ip6_hdr)) {
1343                 n = m_copym(m0, sizeof(struct ip6_hdr),
1344                     hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1345                 if (n == 0)
1346                         return (ENOBUFS);
1347                 m->m_next = n;
1348         } else
1349                 n = m;
1350
1351         /* Search for the last mbuf of unfragmentable part. */
1352         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1353                 ;
1354
1355         if ((mlast->m_flags & M_EXT) == 0 &&
1356             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1357                 /* use the trailing space of the last mbuf for the fragment hdr */
1358                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1359                     mlast->m_len);
1360                 mlast->m_len += sizeof(struct ip6_frag);
1361                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1362         } else {
1363                 /* allocate a new mbuf for the fragment header */
1364                 struct mbuf *mfrg;
1365
1366                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1367                 if (mfrg == 0)
1368                         return (ENOBUFS);
1369                 mfrg->m_len = sizeof(struct ip6_frag);
1370                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1371                 mlast->m_next = mfrg;
1372         }
1373
1374         return (0);
1375 }
1376
1377 static int
1378 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1379     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1380     int *alwaysfragp, u_int fibnum)
1381 {
1382         u_int32_t mtu = 0;
1383         int alwaysfrag = 0;
1384         int error = 0;
1385
1386         if (ro_pmtu != ro) {
1387                 /* The first hop and the final destination may differ. */
1388                 struct sockaddr_in6 *sa6_dst =
1389                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1390                 if (ro_pmtu->ro_rt &&
1391                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1392                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1393                         RTFREE(ro_pmtu->ro_rt);
1394                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1395                 }
1396                 if (ro_pmtu->ro_rt == NULL) {
1397                         bzero(sa6_dst, sizeof(*sa6_dst));
1398                         sa6_dst->sin6_family = AF_INET6;
1399                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1400                         sa6_dst->sin6_addr = *dst;
1401
1402                         in6_rtalloc(ro_pmtu, fibnum);
1403                 }
1404         }
1405         if (ro_pmtu->ro_rt) {
1406                 u_int32_t ifmtu;
1407                 struct in_conninfo inc;
1408
1409                 bzero(&inc, sizeof(inc));
1410                 inc.inc_flags |= INC_ISIPV6;
1411                 inc.inc6_faddr = *dst;
1412
1413                 if (ifp == NULL)
1414                         ifp = ro_pmtu->ro_rt->rt_ifp;
1415                 ifmtu = IN6_LINKMTU(ifp);
1416                 mtu = tcp_hc_getmtu(&inc);
1417                 if (mtu)
1418                         mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1419                 else
1420                         mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1421                 if (mtu == 0)
1422                         mtu = ifmtu;
1423                 else if (mtu < IPV6_MMTU) {
1424                         /*
1425                          * RFC2460 section 5, last paragraph:
1426                          * if we record ICMPv6 too big message with
1427                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1428                          * or smaller, with framgent header attached.
1429                          * (fragment header is needed regardless from the
1430                          * packet size, for translators to identify packets)
1431                          */
1432                         alwaysfrag = 1;
1433                         mtu = IPV6_MMTU;
1434                 } else if (mtu > ifmtu) {
1435                         /*
1436                          * The MTU on the route is larger than the MTU on
1437                          * the interface!  This shouldn't happen, unless the
1438                          * MTU of the interface has been changed after the
1439                          * interface was brought up.  Change the MTU in the
1440                          * route to match the interface MTU (as long as the
1441                          * field isn't locked).
1442                          */
1443                         mtu = ifmtu;
1444                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1445                 }
1446         } else if (ifp) {
1447                 mtu = IN6_LINKMTU(ifp);
1448         } else
1449                 error = EHOSTUNREACH; /* XXX */
1450
1451         *mtup = mtu;
1452         if (alwaysfragp)
1453                 *alwaysfragp = alwaysfrag;
1454         return (error);
1455 }
1456
1457 /*
1458  * IP6 socket option processing.
1459  */
1460 int
1461 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1462 {
1463         int optdatalen, uproto;
1464         void *optdata;
1465         struct inpcb *in6p = sotoinpcb(so);
1466         int error, optval;
1467         int level, op, optname;
1468         int optlen;
1469         struct thread *td;
1470
1471         level = sopt->sopt_level;
1472         op = sopt->sopt_dir;
1473         optname = sopt->sopt_name;
1474         optlen = sopt->sopt_valsize;
1475         td = sopt->sopt_td;
1476         error = 0;
1477         optval = 0;
1478         uproto = (int)so->so_proto->pr_protocol;
1479
1480         if (level != IPPROTO_IPV6) {
1481                 error = EINVAL;
1482
1483                 if (sopt->sopt_level == SOL_SOCKET &&
1484                     sopt->sopt_dir == SOPT_SET) {
1485                         switch (sopt->sopt_name) {
1486                         case SO_REUSEADDR:
1487                                 INP_WLOCK(in6p);
1488                                 if (IN_MULTICAST(ntohl(in6p->inp_laddr.s_addr))) {
1489                                         if ((so->so_options &
1490                                             (SO_REUSEADDR | SO_REUSEPORT)) != 0)
1491                                                 in6p->inp_flags2 |= INP_REUSEPORT;
1492                                         else
1493                                                 in6p->inp_flags2 &= ~INP_REUSEPORT;
1494                                 }
1495                                 INP_WUNLOCK(in6p);
1496                                 error = 0;
1497                                 break;
1498                         case SO_REUSEPORT:
1499                                 INP_WLOCK(in6p);
1500                                 if ((so->so_options & SO_REUSEPORT) != 0)
1501                                         in6p->inp_flags2 |= INP_REUSEPORT;
1502                                 else
1503                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1504                                 INP_WUNLOCK(in6p);
1505                                 error = 0;
1506                                 break;
1507                         case SO_SETFIB:
1508                                 INP_WLOCK(in6p);
1509                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1510                                 INP_WUNLOCK(in6p);
1511                                 error = 0;
1512                                 break;
1513                         default:
1514                                 break;
1515                         }
1516                 }
1517         } else {                /* level == IPPROTO_IPV6 */
1518                 switch (op) {
1519
1520                 case SOPT_SET:
1521                         switch (optname) {
1522                         case IPV6_2292PKTOPTIONS:
1523 #ifdef IPV6_PKTOPTIONS
1524                         case IPV6_PKTOPTIONS:
1525 #endif
1526                         {
1527                                 struct mbuf *m;
1528
1529                                 error = soopt_getm(sopt, &m); /* XXX */
1530                                 if (error != 0)
1531                                         break;
1532                                 error = soopt_mcopyin(sopt, m); /* XXX */
1533                                 if (error != 0)
1534                                         break;
1535                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1536                                                     m, so, sopt);
1537                                 m_freem(m); /* XXX */
1538                                 break;
1539                         }
1540
1541                         /*
1542                          * Use of some Hop-by-Hop options or some
1543                          * Destination options, might require special
1544                          * privilege.  That is, normal applications
1545                          * (without special privilege) might be forbidden
1546                          * from setting certain options in outgoing packets,
1547                          * and might never see certain options in received
1548                          * packets. [RFC 2292 Section 6]
1549                          * KAME specific note:
1550                          *  KAME prevents non-privileged users from sending or
1551                          *  receiving ANY hbh/dst options in order to avoid
1552                          *  overhead of parsing options in the kernel.
1553                          */
1554                         case IPV6_RECVHOPOPTS:
1555                         case IPV6_RECVDSTOPTS:
1556                         case IPV6_RECVRTHDRDSTOPTS:
1557                                 if (td != NULL) {
1558                                         error = priv_check(td,
1559                                             PRIV_NETINET_SETHDROPTS);
1560                                         if (error)
1561                                                 break;
1562                                 }
1563                                 /* FALLTHROUGH */
1564                         case IPV6_UNICAST_HOPS:
1565                         case IPV6_HOPLIMIT:
1566                         case IPV6_FAITH:
1567
1568                         case IPV6_RECVPKTINFO:
1569                         case IPV6_RECVHOPLIMIT:
1570                         case IPV6_RECVRTHDR:
1571                         case IPV6_RECVPATHMTU:
1572                         case IPV6_RECVTCLASS:
1573                         case IPV6_V6ONLY:
1574                         case IPV6_AUTOFLOWLABEL:
1575                         case IPV6_BINDANY:
1576                                 if (optname == IPV6_BINDANY && td != NULL) {
1577                                         error = priv_check(td,
1578                                             PRIV_NETINET_BINDANY);
1579                                         if (error)
1580                                                 break;
1581                                 }
1582
1583                                 if (optlen != sizeof(int)) {
1584                                         error = EINVAL;
1585                                         break;
1586                                 }
1587                                 error = sooptcopyin(sopt, &optval,
1588                                         sizeof optval, sizeof optval);
1589                                 if (error)
1590                                         break;
1591                                 switch (optname) {
1592
1593                                 case IPV6_UNICAST_HOPS:
1594                                         if (optval < -1 || optval >= 256)
1595                                                 error = EINVAL;
1596                                         else {
1597                                                 /* -1 = kernel default */
1598                                                 in6p->in6p_hops = optval;
1599                                                 if ((in6p->inp_vflag &
1600                                                      INP_IPV4) != 0)
1601                                                         in6p->inp_ip_ttl = optval;
1602                                         }
1603                                         break;
1604 #define OPTSET(bit) \
1605 do { \
1606         if (optval) \
1607                 in6p->inp_flags |= (bit); \
1608         else \
1609                 in6p->inp_flags &= ~(bit); \
1610 } while (/*CONSTCOND*/ 0)
1611 #define OPTSET2292(bit) \
1612 do { \
1613         in6p->inp_flags |= IN6P_RFC2292; \
1614         if (optval) \
1615                 in6p->inp_flags |= (bit); \
1616         else \
1617                 in6p->inp_flags &= ~(bit); \
1618 } while (/*CONSTCOND*/ 0)
1619 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1620
1621                                 case IPV6_RECVPKTINFO:
1622                                         /* cannot mix with RFC2292 */
1623                                         if (OPTBIT(IN6P_RFC2292)) {
1624                                                 error = EINVAL;
1625                                                 break;
1626                                         }
1627                                         OPTSET(IN6P_PKTINFO);
1628                                         break;
1629
1630                                 case IPV6_HOPLIMIT:
1631                                 {
1632                                         struct ip6_pktopts **optp;
1633
1634                                         /* cannot mix with RFC2292 */
1635                                         if (OPTBIT(IN6P_RFC2292)) {
1636                                                 error = EINVAL;
1637                                                 break;
1638                                         }
1639                                         optp = &in6p->in6p_outputopts;
1640                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1641                                             (u_char *)&optval, sizeof(optval),
1642                                             optp, (td != NULL) ? td->td_ucred :
1643                                             NULL, uproto);
1644                                         break;
1645                                 }
1646
1647                                 case IPV6_RECVHOPLIMIT:
1648                                         /* cannot mix with RFC2292 */
1649                                         if (OPTBIT(IN6P_RFC2292)) {
1650                                                 error = EINVAL;
1651                                                 break;
1652                                         }
1653                                         OPTSET(IN6P_HOPLIMIT);
1654                                         break;
1655
1656                                 case IPV6_RECVHOPOPTS:
1657                                         /* cannot mix with RFC2292 */
1658                                         if (OPTBIT(IN6P_RFC2292)) {
1659                                                 error = EINVAL;
1660                                                 break;
1661                                         }
1662                                         OPTSET(IN6P_HOPOPTS);
1663                                         break;
1664
1665                                 case IPV6_RECVDSTOPTS:
1666                                         /* cannot mix with RFC2292 */
1667                                         if (OPTBIT(IN6P_RFC2292)) {
1668                                                 error = EINVAL;
1669                                                 break;
1670                                         }
1671                                         OPTSET(IN6P_DSTOPTS);
1672                                         break;
1673
1674                                 case IPV6_RECVRTHDRDSTOPTS:
1675                                         /* cannot mix with RFC2292 */
1676                                         if (OPTBIT(IN6P_RFC2292)) {
1677                                                 error = EINVAL;
1678                                                 break;
1679                                         }
1680                                         OPTSET(IN6P_RTHDRDSTOPTS);
1681                                         break;
1682
1683                                 case IPV6_RECVRTHDR:
1684                                         /* cannot mix with RFC2292 */
1685                                         if (OPTBIT(IN6P_RFC2292)) {
1686                                                 error = EINVAL;
1687                                                 break;
1688                                         }
1689                                         OPTSET(IN6P_RTHDR);
1690                                         break;
1691
1692                                 case IPV6_FAITH:
1693                                         OPTSET(INP_FAITH);
1694                                         break;
1695
1696                                 case IPV6_RECVPATHMTU:
1697                                         /*
1698                                          * We ignore this option for TCP
1699                                          * sockets.
1700                                          * (RFC3542 leaves this case
1701                                          * unspecified.)
1702                                          */
1703                                         if (uproto != IPPROTO_TCP)
1704                                                 OPTSET(IN6P_MTU);
1705                                         break;
1706
1707                                 case IPV6_V6ONLY:
1708                                         /*
1709                                          * make setsockopt(IPV6_V6ONLY)
1710                                          * available only prior to bind(2).
1711                                          * see ipng mailing list, Jun 22 2001.
1712                                          */
1713                                         if (in6p->inp_lport ||
1714                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1715                                                 error = EINVAL;
1716                                                 break;
1717                                         }
1718                                         OPTSET(IN6P_IPV6_V6ONLY);
1719                                         if (optval)
1720                                                 in6p->inp_vflag &= ~INP_IPV4;
1721                                         else
1722                                                 in6p->inp_vflag |= INP_IPV4;
1723                                         break;
1724                                 case IPV6_RECVTCLASS:
1725                                         /* cannot mix with RFC2292 XXX */
1726                                         if (OPTBIT(IN6P_RFC2292)) {
1727                                                 error = EINVAL;
1728                                                 break;
1729                                         }
1730                                         OPTSET(IN6P_TCLASS);
1731                                         break;
1732                                 case IPV6_AUTOFLOWLABEL:
1733                                         OPTSET(IN6P_AUTOFLOWLABEL);
1734                                         break;
1735
1736                                 case IPV6_BINDANY:
1737                                         OPTSET(INP_BINDANY);
1738                                         break;
1739                                 }
1740                                 break;
1741
1742                         case IPV6_TCLASS:
1743                         case IPV6_DONTFRAG:
1744                         case IPV6_USE_MIN_MTU:
1745                         case IPV6_PREFER_TEMPADDR:
1746                                 if (optlen != sizeof(optval)) {
1747                                         error = EINVAL;
1748                                         break;
1749                                 }
1750                                 error = sooptcopyin(sopt, &optval,
1751                                         sizeof optval, sizeof optval);
1752                                 if (error)
1753                                         break;
1754                                 {
1755                                         struct ip6_pktopts **optp;
1756                                         optp = &in6p->in6p_outputopts;
1757                                         error = ip6_pcbopt(optname,
1758                                             (u_char *)&optval, sizeof(optval),
1759                                             optp, (td != NULL) ? td->td_ucred :
1760                                             NULL, uproto);
1761                                         break;
1762                                 }
1763
1764                         case IPV6_2292PKTINFO:
1765                         case IPV6_2292HOPLIMIT:
1766                         case IPV6_2292HOPOPTS:
1767                         case IPV6_2292DSTOPTS:
1768                         case IPV6_2292RTHDR:
1769                                 /* RFC 2292 */
1770                                 if (optlen != sizeof(int)) {
1771                                         error = EINVAL;
1772                                         break;
1773                                 }
1774                                 error = sooptcopyin(sopt, &optval,
1775                                         sizeof optval, sizeof optval);
1776                                 if (error)
1777                                         break;
1778                                 switch (optname) {
1779                                 case IPV6_2292PKTINFO:
1780                                         OPTSET2292(IN6P_PKTINFO);
1781                                         break;
1782                                 case IPV6_2292HOPLIMIT:
1783                                         OPTSET2292(IN6P_HOPLIMIT);
1784                                         break;
1785                                 case IPV6_2292HOPOPTS:
1786                                         /*
1787                                          * Check super-user privilege.
1788                                          * See comments for IPV6_RECVHOPOPTS.
1789                                          */
1790                                         if (td != NULL) {
1791                                                 error = priv_check(td,
1792                                                     PRIV_NETINET_SETHDROPTS);
1793                                                 if (error)
1794                                                         return (error);
1795                                         }
1796                                         OPTSET2292(IN6P_HOPOPTS);
1797                                         break;
1798                                 case IPV6_2292DSTOPTS:
1799                                         if (td != NULL) {
1800                                                 error = priv_check(td,
1801                                                     PRIV_NETINET_SETHDROPTS);
1802                                                 if (error)
1803                                                         return (error);
1804                                         }
1805                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1806                                         break;
1807                                 case IPV6_2292RTHDR:
1808                                         OPTSET2292(IN6P_RTHDR);
1809                                         break;
1810                                 }
1811                                 break;
1812                         case IPV6_PKTINFO:
1813                         case IPV6_HOPOPTS:
1814                         case IPV6_RTHDR:
1815                         case IPV6_DSTOPTS:
1816                         case IPV6_RTHDRDSTOPTS:
1817                         case IPV6_NEXTHOP:
1818                         {
1819                                 /* new advanced API (RFC3542) */
1820                                 u_char *optbuf;
1821                                 u_char optbuf_storage[MCLBYTES];
1822                                 int optlen;
1823                                 struct ip6_pktopts **optp;
1824
1825                                 /* cannot mix with RFC2292 */
1826                                 if (OPTBIT(IN6P_RFC2292)) {
1827                                         error = EINVAL;
1828                                         break;
1829                                 }
1830
1831                                 /*
1832                                  * We only ensure valsize is not too large
1833                                  * here.  Further validation will be done
1834                                  * later.
1835                                  */
1836                                 error = sooptcopyin(sopt, optbuf_storage,
1837                                     sizeof(optbuf_storage), 0);
1838                                 if (error)
1839                                         break;
1840                                 optlen = sopt->sopt_valsize;
1841                                 optbuf = optbuf_storage;
1842                                 optp = &in6p->in6p_outputopts;
1843                                 error = ip6_pcbopt(optname, optbuf, optlen,
1844                                     optp, (td != NULL) ? td->td_ucred : NULL,
1845                                     uproto);
1846                                 break;
1847                         }
1848 #undef OPTSET
1849
1850                         case IPV6_MULTICAST_IF:
1851                         case IPV6_MULTICAST_HOPS:
1852                         case IPV6_MULTICAST_LOOP:
1853                         case IPV6_JOIN_GROUP:
1854                         case IPV6_LEAVE_GROUP:
1855                         case IPV6_MSFILTER:
1856                         case MCAST_BLOCK_SOURCE:
1857                         case MCAST_UNBLOCK_SOURCE:
1858                         case MCAST_JOIN_GROUP:
1859                         case MCAST_LEAVE_GROUP:
1860                         case MCAST_JOIN_SOURCE_GROUP:
1861                         case MCAST_LEAVE_SOURCE_GROUP:
1862                                 error = ip6_setmoptions(in6p, sopt);
1863                                 break;
1864
1865                         case IPV6_PORTRANGE:
1866                                 error = sooptcopyin(sopt, &optval,
1867                                     sizeof optval, sizeof optval);
1868                                 if (error)
1869                                         break;
1870
1871                                 switch (optval) {
1872                                 case IPV6_PORTRANGE_DEFAULT:
1873                                         in6p->inp_flags &= ~(INP_LOWPORT);
1874                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1875                                         break;
1876
1877                                 case IPV6_PORTRANGE_HIGH:
1878                                         in6p->inp_flags &= ~(INP_LOWPORT);
1879                                         in6p->inp_flags |= INP_HIGHPORT;
1880                                         break;
1881
1882                                 case IPV6_PORTRANGE_LOW:
1883                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1884                                         in6p->inp_flags |= INP_LOWPORT;
1885                                         break;
1886
1887                                 default:
1888                                         error = EINVAL;
1889                                         break;
1890                                 }
1891                                 break;
1892
1893 #ifdef IPSEC
1894                         case IPV6_IPSEC_POLICY:
1895                         {
1896                                 caddr_t req;
1897                                 struct mbuf *m;
1898
1899                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1900                                         break;
1901                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1902                                         break;
1903                                 req = mtod(m, caddr_t);
1904                                 error = ipsec_set_policy(in6p, optname, req,
1905                                     m->m_len, (sopt->sopt_td != NULL) ?
1906                                     sopt->sopt_td->td_ucred : NULL);
1907                                 m_freem(m);
1908                                 break;
1909                         }
1910 #endif /* IPSEC */
1911
1912                         default:
1913                                 error = ENOPROTOOPT;
1914                                 break;
1915                         }
1916                         break;
1917
1918                 case SOPT_GET:
1919                         switch (optname) {
1920
1921                         case IPV6_2292PKTOPTIONS:
1922 #ifdef IPV6_PKTOPTIONS
1923                         case IPV6_PKTOPTIONS:
1924 #endif
1925                                 /*
1926                                  * RFC3542 (effectively) deprecated the
1927                                  * semantics of the 2292-style pktoptions.
1928                                  * Since it was not reliable in nature (i.e.,
1929                                  * applications had to expect the lack of some
1930                                  * information after all), it would make sense
1931                                  * to simplify this part by always returning
1932                                  * empty data.
1933                                  */
1934                                 sopt->sopt_valsize = 0;
1935                                 break;
1936
1937                         case IPV6_RECVHOPOPTS:
1938                         case IPV6_RECVDSTOPTS:
1939                         case IPV6_RECVRTHDRDSTOPTS:
1940                         case IPV6_UNICAST_HOPS:
1941                         case IPV6_RECVPKTINFO:
1942                         case IPV6_RECVHOPLIMIT:
1943                         case IPV6_RECVRTHDR:
1944                         case IPV6_RECVPATHMTU:
1945
1946                         case IPV6_FAITH:
1947                         case IPV6_V6ONLY:
1948                         case IPV6_PORTRANGE:
1949                         case IPV6_RECVTCLASS:
1950                         case IPV6_AUTOFLOWLABEL:
1951                         case IPV6_BINDANY:
1952                                 switch (optname) {
1953
1954                                 case IPV6_RECVHOPOPTS:
1955                                         optval = OPTBIT(IN6P_HOPOPTS);
1956                                         break;
1957
1958                                 case IPV6_RECVDSTOPTS:
1959                                         optval = OPTBIT(IN6P_DSTOPTS);
1960                                         break;
1961
1962                                 case IPV6_RECVRTHDRDSTOPTS:
1963                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1964                                         break;
1965
1966                                 case IPV6_UNICAST_HOPS:
1967                                         optval = in6p->in6p_hops;
1968                                         break;
1969
1970                                 case IPV6_RECVPKTINFO:
1971                                         optval = OPTBIT(IN6P_PKTINFO);
1972                                         break;
1973
1974                                 case IPV6_RECVHOPLIMIT:
1975                                         optval = OPTBIT(IN6P_HOPLIMIT);
1976                                         break;
1977
1978                                 case IPV6_RECVRTHDR:
1979                                         optval = OPTBIT(IN6P_RTHDR);
1980                                         break;
1981
1982                                 case IPV6_RECVPATHMTU:
1983                                         optval = OPTBIT(IN6P_MTU);
1984                                         break;
1985
1986                                 case IPV6_FAITH:
1987                                         optval = OPTBIT(INP_FAITH);
1988                                         break;
1989
1990                                 case IPV6_V6ONLY:
1991                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1992                                         break;
1993
1994                                 case IPV6_PORTRANGE:
1995                                     {
1996                                         int flags;
1997                                         flags = in6p->inp_flags;
1998                                         if (flags & INP_HIGHPORT)
1999                                                 optval = IPV6_PORTRANGE_HIGH;
2000                                         else if (flags & INP_LOWPORT)
2001                                                 optval = IPV6_PORTRANGE_LOW;
2002                                         else
2003                                                 optval = 0;
2004                                         break;
2005                                     }
2006                                 case IPV6_RECVTCLASS:
2007                                         optval = OPTBIT(IN6P_TCLASS);
2008                                         break;
2009
2010                                 case IPV6_AUTOFLOWLABEL:
2011                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2012                                         break;
2013
2014                                 case IPV6_BINDANY:
2015                                         optval = OPTBIT(INP_BINDANY);
2016                                         break;
2017                                 }
2018                                 if (error)
2019                                         break;
2020                                 error = sooptcopyout(sopt, &optval,
2021                                         sizeof optval);
2022                                 break;
2023
2024                         case IPV6_PATHMTU:
2025                         {
2026                                 u_long pmtu = 0;
2027                                 struct ip6_mtuinfo mtuinfo;
2028                                 struct route_in6 sro;
2029
2030                                 bzero(&sro, sizeof(sro));
2031
2032                                 if (!(so->so_state & SS_ISCONNECTED))
2033                                         return (ENOTCONN);
2034                                 /*
2035                                  * XXX: we dot not consider the case of source
2036                                  * routing, or optional information to specify
2037                                  * the outgoing interface.
2038                                  */
2039                                 error = ip6_getpmtu(&sro, NULL, NULL,
2040                                     &in6p->in6p_faddr, &pmtu, NULL,
2041                                     so->so_fibnum);
2042                                 if (sro.ro_rt)
2043                                         RTFREE(sro.ro_rt);
2044                                 if (error)
2045                                         break;
2046                                 if (pmtu > IPV6_MAXPACKET)
2047                                         pmtu = IPV6_MAXPACKET;
2048
2049                                 bzero(&mtuinfo, sizeof(mtuinfo));
2050                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2051                                 optdata = (void *)&mtuinfo;
2052                                 optdatalen = sizeof(mtuinfo);
2053                                 error = sooptcopyout(sopt, optdata,
2054                                     optdatalen);
2055                                 break;
2056                         }
2057
2058                         case IPV6_2292PKTINFO:
2059                         case IPV6_2292HOPLIMIT:
2060                         case IPV6_2292HOPOPTS:
2061                         case IPV6_2292RTHDR:
2062                         case IPV6_2292DSTOPTS:
2063                                 switch (optname) {
2064                                 case IPV6_2292PKTINFO:
2065                                         optval = OPTBIT(IN6P_PKTINFO);
2066                                         break;
2067                                 case IPV6_2292HOPLIMIT:
2068                                         optval = OPTBIT(IN6P_HOPLIMIT);
2069                                         break;
2070                                 case IPV6_2292HOPOPTS:
2071                                         optval = OPTBIT(IN6P_HOPOPTS);
2072                                         break;
2073                                 case IPV6_2292RTHDR:
2074                                         optval = OPTBIT(IN6P_RTHDR);
2075                                         break;
2076                                 case IPV6_2292DSTOPTS:
2077                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2078                                         break;
2079                                 }
2080                                 error = sooptcopyout(sopt, &optval,
2081                                     sizeof optval);
2082                                 break;
2083                         case IPV6_PKTINFO:
2084                         case IPV6_HOPOPTS:
2085                         case IPV6_RTHDR:
2086                         case IPV6_DSTOPTS:
2087                         case IPV6_RTHDRDSTOPTS:
2088                         case IPV6_NEXTHOP:
2089                         case IPV6_TCLASS:
2090                         case IPV6_DONTFRAG:
2091                         case IPV6_USE_MIN_MTU:
2092                         case IPV6_PREFER_TEMPADDR:
2093                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2094                                     optname, sopt);
2095                                 break;
2096
2097                         case IPV6_MULTICAST_IF:
2098                         case IPV6_MULTICAST_HOPS:
2099                         case IPV6_MULTICAST_LOOP:
2100                         case IPV6_MSFILTER:
2101                                 error = ip6_getmoptions(in6p, sopt);
2102                                 break;
2103
2104 #ifdef IPSEC
2105                         case IPV6_IPSEC_POLICY:
2106                           {
2107                                 caddr_t req = NULL;
2108                                 size_t len = 0;
2109                                 struct mbuf *m = NULL;
2110                                 struct mbuf **mp = &m;
2111                                 size_t ovalsize = sopt->sopt_valsize;
2112                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2113
2114                                 error = soopt_getm(sopt, &m); /* XXX */
2115                                 if (error != 0)
2116                                         break;
2117                                 error = soopt_mcopyin(sopt, m); /* XXX */
2118                                 if (error != 0)
2119                                         break;
2120                                 sopt->sopt_valsize = ovalsize;
2121                                 sopt->sopt_val = oval;
2122                                 if (m) {
2123                                         req = mtod(m, caddr_t);
2124                                         len = m->m_len;
2125                                 }
2126                                 error = ipsec_get_policy(in6p, req, len, mp);
2127                                 if (error == 0)
2128                                         error = soopt_mcopyout(sopt, m); /* XXX */
2129                                 if (error == 0 && m)
2130                                         m_freem(m);
2131                                 break;
2132                           }
2133 #endif /* IPSEC */
2134
2135                         default:
2136                                 error = ENOPROTOOPT;
2137                                 break;
2138                         }
2139                         break;
2140                 }
2141         }
2142         return (error);
2143 }
2144
2145 int
2146 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2147 {
2148         int error = 0, optval, optlen;
2149         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2150         struct inpcb *in6p = sotoinpcb(so);
2151         int level, op, optname;
2152
2153         level = sopt->sopt_level;
2154         op = sopt->sopt_dir;
2155         optname = sopt->sopt_name;
2156         optlen = sopt->sopt_valsize;
2157
2158         if (level != IPPROTO_IPV6) {
2159                 return (EINVAL);
2160         }
2161
2162         switch (optname) {
2163         case IPV6_CHECKSUM:
2164                 /*
2165                  * For ICMPv6 sockets, no modification allowed for checksum
2166                  * offset, permit "no change" values to help existing apps.
2167                  *
2168                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2169                  * for an ICMPv6 socket will fail."
2170                  * The current behavior does not meet RFC3542.
2171                  */
2172                 switch (op) {
2173                 case SOPT_SET:
2174                         if (optlen != sizeof(int)) {
2175                                 error = EINVAL;
2176                                 break;
2177                         }
2178                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2179                                             sizeof(optval));
2180                         if (error)
2181                                 break;
2182                         if ((optval % 2) != 0) {
2183                                 /* the API assumes even offset values */
2184                                 error = EINVAL;
2185                         } else if (so->so_proto->pr_protocol ==
2186                             IPPROTO_ICMPV6) {
2187                                 if (optval != icmp6off)
2188                                         error = EINVAL;
2189                         } else
2190                                 in6p->in6p_cksum = optval;
2191                         break;
2192
2193                 case SOPT_GET:
2194                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2195                                 optval = icmp6off;
2196                         else
2197                                 optval = in6p->in6p_cksum;
2198
2199                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2200                         break;
2201
2202                 default:
2203                         error = EINVAL;
2204                         break;
2205                 }
2206                 break;
2207
2208         default:
2209                 error = ENOPROTOOPT;
2210                 break;
2211         }
2212
2213         return (error);
2214 }
2215
2216 /*
2217  * Set up IP6 options in pcb for insertion in output packets or
2218  * specifying behavior of outgoing packets.
2219  */
2220 static int
2221 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2222     struct socket *so, struct sockopt *sopt)
2223 {
2224         struct ip6_pktopts *opt = *pktopt;
2225         int error = 0;
2226         struct thread *td = sopt->sopt_td;
2227
2228         /* turn off any old options. */
2229         if (opt) {
2230 #ifdef DIAGNOSTIC
2231                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2232                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2233                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2234                         printf("ip6_pcbopts: all specified options are cleared.\n");
2235 #endif
2236                 ip6_clearpktopts(opt, -1);
2237         } else
2238                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2239         *pktopt = NULL;
2240
2241         if (!m || m->m_len == 0) {
2242                 /*
2243                  * Only turning off any previous options, regardless of
2244                  * whether the opt is just created or given.
2245                  */
2246                 free(opt, M_IP6OPT);
2247                 return (0);
2248         }
2249
2250         /*  set options specified by user. */
2251         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2252             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2253                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2254                 free(opt, M_IP6OPT);
2255                 return (error);
2256         }
2257         *pktopt = opt;
2258         return (0);
2259 }
2260
2261 /*
2262  * initialize ip6_pktopts.  beware that there are non-zero default values in
2263  * the struct.
2264  */
2265 void
2266 ip6_initpktopts(struct ip6_pktopts *opt)
2267 {
2268
2269         bzero(opt, sizeof(*opt));
2270         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2271         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2272         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2273         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2274 }
2275
2276 static int
2277 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2278     struct ucred *cred, int uproto)
2279 {
2280         struct ip6_pktopts *opt;
2281
2282         if (*pktopt == NULL) {
2283                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2284                     M_WAITOK);
2285                 ip6_initpktopts(*pktopt);
2286         }
2287         opt = *pktopt;
2288
2289         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2290 }
2291
2292 static int
2293 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2294 {
2295         void *optdata = NULL;
2296         int optdatalen = 0;
2297         struct ip6_ext *ip6e;
2298         int error = 0;
2299         struct in6_pktinfo null_pktinfo;
2300         int deftclass = 0, on;
2301         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2302         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2303
2304         switch (optname) {
2305         case IPV6_PKTINFO:
2306                 if (pktopt && pktopt->ip6po_pktinfo)
2307                         optdata = (void *)pktopt->ip6po_pktinfo;
2308                 else {
2309                         /* XXX: we don't have to do this every time... */
2310                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2311                         optdata = (void *)&null_pktinfo;
2312                 }
2313                 optdatalen = sizeof(struct in6_pktinfo);
2314                 break;
2315         case IPV6_TCLASS:
2316                 if (pktopt && pktopt->ip6po_tclass >= 0)
2317                         optdata = (void *)&pktopt->ip6po_tclass;
2318                 else
2319                         optdata = (void *)&deftclass;
2320                 optdatalen = sizeof(int);
2321                 break;
2322         case IPV6_HOPOPTS:
2323                 if (pktopt && pktopt->ip6po_hbh) {
2324                         optdata = (void *)pktopt->ip6po_hbh;
2325                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2326                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2327                 }
2328                 break;
2329         case IPV6_RTHDR:
2330                 if (pktopt && pktopt->ip6po_rthdr) {
2331                         optdata = (void *)pktopt->ip6po_rthdr;
2332                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2333                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2334                 }
2335                 break;
2336         case IPV6_RTHDRDSTOPTS:
2337                 if (pktopt && pktopt->ip6po_dest1) {
2338                         optdata = (void *)pktopt->ip6po_dest1;
2339                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2340                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2341                 }
2342                 break;
2343         case IPV6_DSTOPTS:
2344                 if (pktopt && pktopt->ip6po_dest2) {
2345                         optdata = (void *)pktopt->ip6po_dest2;
2346                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2347                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2348                 }
2349                 break;
2350         case IPV6_NEXTHOP:
2351                 if (pktopt && pktopt->ip6po_nexthop) {
2352                         optdata = (void *)pktopt->ip6po_nexthop;
2353                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2354                 }
2355                 break;
2356         case IPV6_USE_MIN_MTU:
2357                 if (pktopt)
2358                         optdata = (void *)&pktopt->ip6po_minmtu;
2359                 else
2360                         optdata = (void *)&defminmtu;
2361                 optdatalen = sizeof(int);
2362                 break;
2363         case IPV6_DONTFRAG:
2364                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2365                         on = 1;
2366                 else
2367                         on = 0;
2368                 optdata = (void *)&on;
2369                 optdatalen = sizeof(on);
2370                 break;
2371         case IPV6_PREFER_TEMPADDR:
2372                 if (pktopt)
2373                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2374                 else
2375                         optdata = (void *)&defpreftemp;
2376                 optdatalen = sizeof(int);
2377                 break;
2378         default:                /* should not happen */
2379 #ifdef DIAGNOSTIC
2380                 panic("ip6_getpcbopt: unexpected option\n");
2381 #endif
2382                 return (ENOPROTOOPT);
2383         }
2384
2385         error = sooptcopyout(sopt, optdata, optdatalen);
2386
2387         return (error);
2388 }
2389
2390 void
2391 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2392 {
2393         if (pktopt == NULL)
2394                 return;
2395
2396         if (optname == -1 || optname == IPV6_PKTINFO) {
2397                 if (pktopt->ip6po_pktinfo)
2398                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2399                 pktopt->ip6po_pktinfo = NULL;
2400         }
2401         if (optname == -1 || optname == IPV6_HOPLIMIT)
2402                 pktopt->ip6po_hlim = -1;
2403         if (optname == -1 || optname == IPV6_TCLASS)
2404                 pktopt->ip6po_tclass = -1;
2405         if (optname == -1 || optname == IPV6_NEXTHOP) {
2406                 if (pktopt->ip6po_nextroute.ro_rt) {
2407                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2408                         pktopt->ip6po_nextroute.ro_rt = NULL;
2409                 }
2410                 if (pktopt->ip6po_nexthop)
2411                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2412                 pktopt->ip6po_nexthop = NULL;
2413         }
2414         if (optname == -1 || optname == IPV6_HOPOPTS) {
2415                 if (pktopt->ip6po_hbh)
2416                         free(pktopt->ip6po_hbh, M_IP6OPT);
2417                 pktopt->ip6po_hbh = NULL;
2418         }
2419         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2420                 if (pktopt->ip6po_dest1)
2421                         free(pktopt->ip6po_dest1, M_IP6OPT);
2422                 pktopt->ip6po_dest1 = NULL;
2423         }
2424         if (optname == -1 || optname == IPV6_RTHDR) {
2425                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2426                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2427                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2428                 if (pktopt->ip6po_route.ro_rt) {
2429                         RTFREE(pktopt->ip6po_route.ro_rt);
2430                         pktopt->ip6po_route.ro_rt = NULL;
2431                 }
2432         }
2433         if (optname == -1 || optname == IPV6_DSTOPTS) {
2434                 if (pktopt->ip6po_dest2)
2435                         free(pktopt->ip6po_dest2, M_IP6OPT);
2436                 pktopt->ip6po_dest2 = NULL;
2437         }
2438 }
2439
2440 #define PKTOPT_EXTHDRCPY(type) \
2441 do {\
2442         if (src->type) {\
2443                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2444                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2445                 if (dst->type == NULL && canwait == M_NOWAIT)\
2446                         goto bad;\
2447                 bcopy(src->type, dst->type, hlen);\
2448         }\
2449 } while (/*CONSTCOND*/ 0)
2450
2451 static int
2452 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2453 {
2454         if (dst == NULL || src == NULL)  {
2455                 printf("ip6_clearpktopts: invalid argument\n");
2456                 return (EINVAL);
2457         }
2458
2459         dst->ip6po_hlim = src->ip6po_hlim;
2460         dst->ip6po_tclass = src->ip6po_tclass;
2461         dst->ip6po_flags = src->ip6po_flags;
2462         dst->ip6po_minmtu = src->ip6po_minmtu;
2463         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2464         if (src->ip6po_pktinfo) {
2465                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2466                     M_IP6OPT, canwait);
2467                 if (dst->ip6po_pktinfo == NULL)
2468                         goto bad;
2469                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2470         }
2471         if (src->ip6po_nexthop) {
2472                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2473                     M_IP6OPT, canwait);
2474                 if (dst->ip6po_nexthop == NULL)
2475                         goto bad;
2476                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2477                     src->ip6po_nexthop->sa_len);
2478         }
2479         PKTOPT_EXTHDRCPY(ip6po_hbh);
2480         PKTOPT_EXTHDRCPY(ip6po_dest1);
2481         PKTOPT_EXTHDRCPY(ip6po_dest2);
2482         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2483         return (0);
2484
2485   bad:
2486         ip6_clearpktopts(dst, -1);
2487         return (ENOBUFS);
2488 }
2489 #undef PKTOPT_EXTHDRCPY
2490
2491 struct ip6_pktopts *
2492 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2493 {
2494         int error;
2495         struct ip6_pktopts *dst;
2496
2497         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2498         if (dst == NULL)
2499                 return (NULL);
2500         ip6_initpktopts(dst);
2501
2502         if ((error = copypktopts(dst, src, canwait)) != 0) {
2503                 free(dst, M_IP6OPT);
2504                 return (NULL);
2505         }
2506
2507         return (dst);
2508 }
2509
2510 void
2511 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2512 {
2513         if (pktopt == NULL)
2514                 return;
2515
2516         ip6_clearpktopts(pktopt, -1);
2517
2518         free(pktopt, M_IP6OPT);
2519 }
2520
2521 /*
2522  * Set IPv6 outgoing packet options based on advanced API.
2523  */
2524 int
2525 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2526     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2527 {
2528         struct cmsghdr *cm = 0;
2529
2530         if (control == NULL || opt == NULL)
2531                 return (EINVAL);
2532
2533         ip6_initpktopts(opt);
2534         if (stickyopt) {
2535                 int error;
2536
2537                 /*
2538                  * If stickyopt is provided, make a local copy of the options
2539                  * for this particular packet, then override them by ancillary
2540                  * objects.
2541                  * XXX: copypktopts() does not copy the cached route to a next
2542                  * hop (if any).  This is not very good in terms of efficiency,
2543                  * but we can allow this since this option should be rarely
2544                  * used.
2545                  */
2546                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2547                         return (error);
2548         }
2549
2550         /*
2551          * XXX: Currently, we assume all the optional information is stored
2552          * in a single mbuf.
2553          */
2554         if (control->m_next)
2555                 return (EINVAL);
2556
2557         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2558             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2559                 int error;
2560
2561                 if (control->m_len < CMSG_LEN(0))
2562                         return (EINVAL);
2563
2564                 cm = mtod(control, struct cmsghdr *);
2565                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2566                         return (EINVAL);
2567                 if (cm->cmsg_level != IPPROTO_IPV6)
2568                         continue;
2569
2570                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2571                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2572                 if (error)
2573                         return (error);
2574         }
2575
2576         return (0);
2577 }
2578
2579 /*
2580  * Set a particular packet option, as a sticky option or an ancillary data
2581  * item.  "len" can be 0 only when it's a sticky option.
2582  * We have 4 cases of combination of "sticky" and "cmsg":
2583  * "sticky=0, cmsg=0": impossible
2584  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2585  * "sticky=1, cmsg=0": RFC3542 socket option
2586  * "sticky=1, cmsg=1": RFC2292 socket option
2587  */
2588 static int
2589 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2590     struct ucred *cred, int sticky, int cmsg, int uproto)
2591 {
2592         int minmtupolicy, preftemp;
2593         int error;
2594
2595         if (!sticky && !cmsg) {
2596 #ifdef DIAGNOSTIC
2597                 printf("ip6_setpktopt: impossible case\n");
2598 #endif
2599                 return (EINVAL);
2600         }
2601
2602         /*
2603          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2604          * not be specified in the context of RFC3542.  Conversely,
2605          * RFC3542 types should not be specified in the context of RFC2292.
2606          */
2607         if (!cmsg) {
2608                 switch (optname) {
2609                 case IPV6_2292PKTINFO:
2610                 case IPV6_2292HOPLIMIT:
2611                 case IPV6_2292NEXTHOP:
2612                 case IPV6_2292HOPOPTS:
2613                 case IPV6_2292DSTOPTS:
2614                 case IPV6_2292RTHDR:
2615                 case IPV6_2292PKTOPTIONS:
2616                         return (ENOPROTOOPT);
2617                 }
2618         }
2619         if (sticky && cmsg) {
2620                 switch (optname) {
2621                 case IPV6_PKTINFO:
2622                 case IPV6_HOPLIMIT:
2623                 case IPV6_NEXTHOP:
2624                 case IPV6_HOPOPTS:
2625                 case IPV6_DSTOPTS:
2626                 case IPV6_RTHDRDSTOPTS:
2627                 case IPV6_RTHDR:
2628                 case IPV6_USE_MIN_MTU:
2629                 case IPV6_DONTFRAG:
2630                 case IPV6_TCLASS:
2631                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2632                         return (ENOPROTOOPT);
2633                 }
2634         }
2635
2636         switch (optname) {
2637         case IPV6_2292PKTINFO:
2638         case IPV6_PKTINFO:
2639         {
2640                 struct ifnet *ifp = NULL;
2641                 struct in6_pktinfo *pktinfo;
2642
2643                 if (len != sizeof(struct in6_pktinfo))
2644                         return (EINVAL);
2645
2646                 pktinfo = (struct in6_pktinfo *)buf;
2647
2648                 /*
2649                  * An application can clear any sticky IPV6_PKTINFO option by
2650                  * doing a "regular" setsockopt with ipi6_addr being
2651                  * in6addr_any and ipi6_ifindex being zero.
2652                  * [RFC 3542, Section 6]
2653                  */
2654                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2655                     pktinfo->ipi6_ifindex == 0 &&
2656                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2657                         ip6_clearpktopts(opt, optname);
2658                         break;
2659                 }
2660
2661                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2662                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2663                         return (EINVAL);
2664                 }
2665
2666                 /* validate the interface index if specified. */
2667                 if (pktinfo->ipi6_ifindex > V_if_index ||
2668                     pktinfo->ipi6_ifindex < 0) {
2669                          return (ENXIO);
2670                 }
2671                 if (pktinfo->ipi6_ifindex) {
2672                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2673                         if (ifp == NULL)
2674                                 return (ENXIO);
2675                 }
2676
2677                 /*
2678                  * We store the address anyway, and let in6_selectsrc()
2679                  * validate the specified address.  This is because ipi6_addr
2680                  * may not have enough information about its scope zone, and
2681                  * we may need additional information (such as outgoing
2682                  * interface or the scope zone of a destination address) to
2683                  * disambiguate the scope.
2684                  * XXX: the delay of the validation may confuse the
2685                  * application when it is used as a sticky option.
2686                  */
2687                 if (opt->ip6po_pktinfo == NULL) {
2688                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2689                             M_IP6OPT, M_NOWAIT);
2690                         if (opt->ip6po_pktinfo == NULL)
2691                                 return (ENOBUFS);
2692                 }
2693                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2694                 break;
2695         }
2696
2697         case IPV6_2292HOPLIMIT:
2698         case IPV6_HOPLIMIT:
2699         {
2700                 int *hlimp;
2701
2702                 /*
2703                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2704                  * to simplify the ordering among hoplimit options.
2705                  */
2706                 if (optname == IPV6_HOPLIMIT && sticky)
2707                         return (ENOPROTOOPT);
2708
2709                 if (len != sizeof(int))
2710                         return (EINVAL);
2711                 hlimp = (int *)buf;
2712                 if (*hlimp < -1 || *hlimp > 255)
2713                         return (EINVAL);
2714
2715                 opt->ip6po_hlim = *hlimp;
2716                 break;
2717         }
2718
2719         case IPV6_TCLASS:
2720         {
2721                 int tclass;
2722
2723                 if (len != sizeof(int))
2724                         return (EINVAL);
2725                 tclass = *(int *)buf;
2726                 if (tclass < -1 || tclass > 255)
2727                         return (EINVAL);
2728
2729                 opt->ip6po_tclass = tclass;
2730                 break;
2731         }
2732
2733         case IPV6_2292NEXTHOP:
2734         case IPV6_NEXTHOP:
2735                 if (cred != NULL) {
2736                         error = priv_check_cred(cred,
2737                             PRIV_NETINET_SETHDROPTS, 0);
2738                         if (error)
2739                                 return (error);
2740                 }
2741
2742                 if (len == 0) { /* just remove the option */
2743                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2744                         break;
2745                 }
2746
2747                 /* check if cmsg_len is large enough for sa_len */
2748                 if (len < sizeof(struct sockaddr) || len < *buf)
2749                         return (EINVAL);
2750
2751                 switch (((struct sockaddr *)buf)->sa_family) {
2752                 case AF_INET6:
2753                 {
2754                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2755                         int error;
2756
2757                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2758                                 return (EINVAL);
2759
2760                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2761                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2762                                 return (EINVAL);
2763                         }
2764                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2765                             != 0) {
2766                                 return (error);
2767                         }
2768                         break;
2769                 }
2770                 case AF_LINK:   /* should eventually be supported */
2771                 default:
2772                         return (EAFNOSUPPORT);
2773                 }
2774
2775                 /* turn off the previous option, then set the new option. */
2776                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2777                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2778                 if (opt->ip6po_nexthop == NULL)
2779                         return (ENOBUFS);
2780                 bcopy(buf, opt->ip6po_nexthop, *buf);
2781                 break;
2782
2783         case IPV6_2292HOPOPTS:
2784         case IPV6_HOPOPTS:
2785         {
2786                 struct ip6_hbh *hbh;
2787                 int hbhlen;
2788
2789                 /*
2790                  * XXX: We don't allow a non-privileged user to set ANY HbH
2791                  * options, since per-option restriction has too much
2792                  * overhead.
2793                  */
2794                 if (cred != NULL) {
2795                         error = priv_check_cred(cred,
2796                             PRIV_NETINET_SETHDROPTS, 0);
2797                         if (error)
2798                                 return (error);
2799                 }
2800
2801                 if (len == 0) {
2802                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2803                         break;  /* just remove the option */
2804                 }
2805
2806                 /* message length validation */
2807                 if (len < sizeof(struct ip6_hbh))
2808                         return (EINVAL);
2809                 hbh = (struct ip6_hbh *)buf;
2810                 hbhlen = (hbh->ip6h_len + 1) << 3;
2811                 if (len != hbhlen)
2812                         return (EINVAL);
2813
2814                 /* turn off the previous option, then set the new option. */
2815                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2816                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2817                 if (opt->ip6po_hbh == NULL)
2818                         return (ENOBUFS);
2819                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2820
2821                 break;
2822         }
2823
2824         case IPV6_2292DSTOPTS:
2825         case IPV6_DSTOPTS:
2826         case IPV6_RTHDRDSTOPTS:
2827         {
2828                 struct ip6_dest *dest, **newdest = NULL;
2829                 int destlen;
2830
2831                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2832                         error = priv_check_cred(cred,
2833                             PRIV_NETINET_SETHDROPTS, 0);
2834                         if (error)
2835                                 return (error);
2836                 }
2837
2838                 if (len == 0) {
2839                         ip6_clearpktopts(opt, optname);
2840                         break;  /* just remove the option */
2841                 }
2842
2843                 /* message length validation */
2844                 if (len < sizeof(struct ip6_dest))
2845                         return (EINVAL);
2846                 dest = (struct ip6_dest *)buf;
2847                 destlen = (dest->ip6d_len + 1) << 3;
2848                 if (len != destlen)
2849                         return (EINVAL);
2850
2851                 /*
2852                  * Determine the position that the destination options header
2853                  * should be inserted; before or after the routing header.
2854                  */
2855                 switch (optname) {
2856                 case IPV6_2292DSTOPTS:
2857                         /*
2858                          * The old advacned API is ambiguous on this point.
2859                          * Our approach is to determine the position based
2860                          * according to the existence of a routing header.
2861                          * Note, however, that this depends on the order of the
2862                          * extension headers in the ancillary data; the 1st
2863                          * part of the destination options header must appear
2864                          * before the routing header in the ancillary data,
2865                          * too.
2866                          * RFC3542 solved the ambiguity by introducing
2867                          * separate ancillary data or option types.
2868                          */
2869                         if (opt->ip6po_rthdr == NULL)
2870                                 newdest = &opt->ip6po_dest1;
2871                         else
2872                                 newdest = &opt->ip6po_dest2;
2873                         break;
2874                 case IPV6_RTHDRDSTOPTS:
2875                         newdest = &opt->ip6po_dest1;
2876                         break;
2877                 case IPV6_DSTOPTS:
2878                         newdest = &opt->ip6po_dest2;
2879                         break;
2880                 }
2881
2882                 /* turn off the previous option, then set the new option. */
2883                 ip6_clearpktopts(opt, optname);
2884                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2885                 if (*newdest == NULL)
2886                         return (ENOBUFS);
2887                 bcopy(dest, *newdest, destlen);
2888
2889                 break;
2890         }
2891
2892         case IPV6_2292RTHDR:
2893         case IPV6_RTHDR:
2894         {
2895                 struct ip6_rthdr *rth;
2896                 int rthlen;
2897
2898                 if (len == 0) {
2899                         ip6_clearpktopts(opt, IPV6_RTHDR);
2900                         break;  /* just remove the option */
2901                 }
2902
2903                 /* message length validation */
2904                 if (len < sizeof(struct ip6_rthdr))
2905                         return (EINVAL);
2906                 rth = (struct ip6_rthdr *)buf;
2907                 rthlen = (rth->ip6r_len + 1) << 3;
2908                 if (len != rthlen)
2909                         return (EINVAL);
2910
2911                 switch (rth->ip6r_type) {
2912                 case IPV6_RTHDR_TYPE_0:
2913                         if (rth->ip6r_len == 0) /* must contain one addr */
2914                                 return (EINVAL);
2915                         if (rth->ip6r_len % 2) /* length must be even */
2916                                 return (EINVAL);
2917                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2918                                 return (EINVAL);
2919                         break;
2920                 default:
2921                         return (EINVAL);        /* not supported */
2922                 }
2923
2924                 /* turn off the previous option */
2925                 ip6_clearpktopts(opt, IPV6_RTHDR);
2926                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2927                 if (opt->ip6po_rthdr == NULL)
2928                         return (ENOBUFS);
2929                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2930
2931                 break;
2932         }
2933
2934         case IPV6_USE_MIN_MTU:
2935                 if (len != sizeof(int))
2936                         return (EINVAL);
2937                 minmtupolicy = *(int *)buf;
2938                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2939                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2940                     minmtupolicy != IP6PO_MINMTU_ALL) {
2941                         return (EINVAL);
2942                 }
2943                 opt->ip6po_minmtu = minmtupolicy;
2944                 break;
2945
2946         case IPV6_DONTFRAG:
2947                 if (len != sizeof(int))
2948                         return (EINVAL);
2949
2950                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2951                         /*
2952                          * we ignore this option for TCP sockets.
2953                          * (RFC3542 leaves this case unspecified.)
2954                          */
2955                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2956                 } else
2957                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2958                 break;
2959
2960         case IPV6_PREFER_TEMPADDR:
2961                 if (len != sizeof(int))
2962                         return (EINVAL);
2963                 preftemp = *(int *)buf;
2964                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2965                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2966                     preftemp != IP6PO_TEMPADDR_PREFER) {
2967                         return (EINVAL);
2968                 }
2969                 opt->ip6po_prefer_tempaddr = preftemp;
2970                 break;
2971
2972         default:
2973                 return (ENOPROTOOPT);
2974         } /* end of switch */
2975
2976         return (0);
2977 }
2978
2979 /*
2980  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2981  * packet to the input queue of a specified interface.  Note that this
2982  * calls the output routine of the loopback "driver", but with an interface
2983  * pointer that might NOT be &loif -- easier than replicating that code here.
2984  */
2985 void
2986 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2987 {
2988         struct mbuf *copym;
2989         struct ip6_hdr *ip6;
2990
2991         copym = m_copy(m, 0, M_COPYALL);
2992         if (copym == NULL)
2993                 return;
2994
2995         /*
2996          * Make sure to deep-copy IPv6 header portion in case the data
2997          * is in an mbuf cluster, so that we can safely override the IPv6
2998          * header portion later.
2999          */
3000         if ((copym->m_flags & M_EXT) != 0 ||
3001             copym->m_len < sizeof(struct ip6_hdr)) {
3002                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3003                 if (copym == NULL)
3004                         return;
3005         }
3006
3007 #ifdef DIAGNOSTIC
3008         if (copym->m_len < sizeof(*ip6)) {
3009                 m_freem(copym);
3010                 return;
3011         }
3012 #endif
3013
3014         ip6 = mtod(copym, struct ip6_hdr *);
3015         /*
3016          * clear embedded scope identifiers if necessary.
3017          * in6_clearscope will touch the addresses only when necessary.
3018          */
3019         in6_clearscope(&ip6->ip6_src);
3020         in6_clearscope(&ip6->ip6_dst);
3021
3022         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3023 }
3024
3025 /*
3026  * Chop IPv6 header off from the payload.
3027  */
3028 static int
3029 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3030 {
3031         struct mbuf *mh;
3032         struct ip6_hdr *ip6;
3033
3034         ip6 = mtod(m, struct ip6_hdr *);
3035         if (m->m_len > sizeof(*ip6)) {
3036                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3037                 if (mh == 0) {
3038                         m_freem(m);
3039                         return ENOBUFS;
3040                 }
3041                 M_MOVE_PKTHDR(mh, m);
3042                 MH_ALIGN(mh, sizeof(*ip6));
3043                 m->m_len -= sizeof(*ip6);
3044                 m->m_data += sizeof(*ip6);
3045                 mh->m_next = m;
3046                 m = mh;
3047                 m->m_len = sizeof(*ip6);
3048                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3049         }
3050         exthdrs->ip6e_ip6 = m;
3051         return 0;
3052 }
3053
3054 /*
3055  * Compute IPv6 extension header length.
3056  */
3057 int
3058 ip6_optlen(struct inpcb *in6p)
3059 {
3060         int len;
3061
3062         if (!in6p->in6p_outputopts)
3063                 return 0;
3064
3065         len = 0;
3066 #define elen(x) \
3067     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3068
3069         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3070         if (in6p->in6p_outputopts->ip6po_rthdr)
3071                 /* dest1 is valid with rthdr only */
3072                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3073         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3074         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3075         return len;
3076 #undef elen
3077 }