sys/netinet6/ip6_output.c

   1 /*-
   2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. Neither the name of the project nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
  30  */
  31
  32 /*-
  33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  34  *      The Regents of the University of California.  All rights reserved.
  35  *
  36  * Redistribution and use in source and binary forms, with or without
  37  * modification, are permitted provided that the following conditions
  38  * are met:
  39  * 1. Redistributions of source code must retain the above copyright
  40  *    notice, this list of conditions and the following disclaimer.
  41  * 2. Redistributions in binary form must reproduce the above copyright
  42  *    notice, this list of conditions and the following disclaimer in the
  43  *    documentation and/or other materials provided with the distribution.
  44  * 4. Neither the name of the University nor the names of its contributors
  45  *    may be used to endorse or promote products derived from this software
  46  *    without specific prior written permission.
  47  *
  48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  58  * SUCH DAMAGE.
  59  *
  60  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  61  */
  62
  63 #include <sys/cdefs.h>
  64 __FBSDID("$FreeBSD$");
  65
  66 #include "opt_inet.h"
  67 #include "opt_inet6.h"
  68 #include "opt_ipfw.h"
  69 #include "opt_ipsec.h"
  70 #include "opt_sctp.h"
  71 #include "opt_route.h"
  72
  73 #include <sys/param.h>
  74 #include <sys/kernel.h>
  75 #include <sys/malloc.h>
  76 #include <sys/mbuf.h>
  77 #include <sys/errno.h>
  78 #include <sys/priv.h>
  79 #include <sys/proc.h>
  80 #include <sys/protosw.h>
  81 #include <sys/socket.h>
  82 #include <sys/socketvar.h>
  83 #include <sys/syslog.h>
  84 #include <sys/ucred.h>
  85
  86 #include <machine/in_cksum.h>
  87
  88 #include <net/if.h>
  89 #include <net/netisr.h>
  90 #include <net/route.h>
  91 #include <net/pfil.h>
  92 #include <net/vnet.h>
  93
  94 #include <netinet/in.h>
  95 #include <netinet/in_var.h>
  96 #include <netinet/ip_var.h>
  97 #include <netinet6/in6_var.h>
  98 #include <netinet/ip6.h>
  99 #include <netinet/icmp6.h>
 100 #include <netinet6/ip6_var.h>
 101 #include <netinet/in_pcb.h>
 102 #include <netinet/tcp_var.h>
 103 #include <netinet6/nd6.h>
 104
 105 #ifdef IPSEC
 106 #include <netipsec/ipsec.h>
 107 #include <netipsec/ipsec6.h>
 108 #include <netipsec/key.h>
 109 #include <netinet6/ip6_ipsec.h>
 110 #endif /* IPSEC */
 111 #ifdef SCTP
 112 #include <netinet/sctp.h>
 113 #include <netinet/sctp_crc32.h>
 114 #endif
 115
 116 #include <netinet6/ip6protosw.h>
 117 #include <netinet6/scope6_var.h>
 118
 119 #ifdef FLOWTABLE
 120 #include <net/flowtable.h>
 121 #endif
 122
 123 extern int in6_mcast_loop;
 124
 125 struct ip6_exthdrs {
 126         struct mbuf *ip6e_ip6;
 127         struct mbuf *ip6e_hbh;
 128         struct mbuf *ip6e_dest1;
 129         struct mbuf *ip6e_rthdr;
 130         struct mbuf *ip6e_dest2;
 131 };
 132
 133 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
 134                            struct ucred *, int);
 135 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
 136         struct socket *, struct sockopt *);
 137 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 138 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
 139         struct ucred *, int, int, int);
 140
 141 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 142 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 143         struct ip6_frag **);
 144 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 145 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 146 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
 147         struct ifnet *, struct in6_addr *, u_long *, int *, u_int);
 148 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 149
 150
 151 /*
 152  * Make an extension header from option data.  hp is the source, and
 153  * mp is the destination.
 154  */
 155 #define MAKE_EXTHDR(hp, mp)                                             \
 156     do {                                                                \
 157         if (hp) {                                                       \
 158                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 159                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 160                     ((eh)->ip6e_len + 1) << 3);                         \
 161                 if (error)                                              \
 162                         goto freehdrs;                                  \
 163         }                                                               \
 164     } while (/*CONSTCOND*/ 0)
 165
 166 /*
 167  * Form a chain of extension headers.
 168  * m is the extension header mbuf
 169  * mp is the previous mbuf in the chain
 170  * p is the next header
 171  * i is the type of option.
 172  */
 173 #define MAKE_CHAIN(m, mp, p, i)\
 174     do {\
 175         if (m) {\
 176                 if (!hdrsplit) \
 177                         panic("assumption failed: hdr not split"); \
 178                 *mtod((m), u_char *) = *(p);\
 179                 *(p) = (i);\
 180                 p = mtod((m), u_char *);\
 181                 (m)->m_next = (mp)->m_next;\
 182                 (mp)->m_next = (m);\
 183                 (mp) = (m);\
 184         }\
 185     } while (/*CONSTCOND*/ 0)
 186
 187 void
 188 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 189 {
 190         u_short csum;
 191
 192         csum = in_cksum_skip(m, offset + plen, offset);
 193         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 194                 csum = 0xffff;
 195         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 196
 197         if (offset + sizeof(u_short) > m->m_len) {
 198                 printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
 199                     "csum_flags=%b\n", __func__, m->m_len, plen, offset,
 200                     (int)m->m_pkthdr.csum_flags, CSUM_BITS);
 201                 /*
 202                  * XXX this should not happen, but if it does, the correct
 203                  * behavior may be to insert the checksum in the appropriate
 204                  * next mbuf in the chain.
 205                  */
 206                 return;
 207         }
 208         *(u_short *)(m->m_data + offset) = csum;
 209 }
 210
 211 int
 212 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
 213     int mtu, uint32_t id)
 214 {
 215         struct mbuf *m, **mnext, *m_frgpart;
 216         struct ip6_hdr *ip6, *mhip6;
 217         struct ip6_frag *ip6f;
 218         int off;
 219         int error;
 220         int tlen = m0->m_pkthdr.len;
 221
 222         m = m0;
 223         ip6 = mtod(m, struct ip6_hdr *);
 224         mnext = &m->m_nextpkt;
 225
 226         for (off = hlen; off < tlen; off += mtu) {
 227                 m = m_gethdr(M_NOWAIT, MT_DATA);
 228                 if (!m) {
 229                         IP6STAT_INC(ip6s_odropped);
 230                         return (ENOBUFS);
 231                 }
 232                 m->m_flags = m0->m_flags & M_COPYFLAGS;
 233                 *mnext = m;
 234                 mnext = &m->m_nextpkt;
 235                 m->m_data += max_linkhdr;
 236                 mhip6 = mtod(m, struct ip6_hdr *);
 237                 *mhip6 = *ip6;
 238                 m->m_len = sizeof(*mhip6);
 239                 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 240                 if (error) {
 241                         IP6STAT_INC(ip6s_odropped);
 242                         return (error);
 243                 }
 244                 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 245                 if (off + mtu >= tlen)
 246                         mtu = tlen - off;
 247                 else
 248                         ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 249                 mhip6->ip6_plen = htons((u_short)(mtu + hlen +
 250                     sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 251                 if ((m_frgpart = m_copy(m0, off, mtu)) == 0) {
 252                         IP6STAT_INC(ip6s_odropped);
 253                         return (ENOBUFS);
 254                 }
 255                 m_cat(m, m_frgpart);
 256                 m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
 257                 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
 258                 m->m_pkthdr.rcvif = NULL;
 259                 ip6f->ip6f_reserved = 0;
 260                 ip6f->ip6f_ident = id;
 261                 ip6f->ip6f_nxt = nextproto;
 262                 IP6STAT_INC(ip6s_ofragments);
 263                 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 264         }
 265
 266         return (0);
 267 }
 268
 269 /*
 270  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 271  * header (with pri, len, nxt, hlim, src, dst).
 272  * This function may modify ver and hlim only.
 273  * The mbuf chain containing the packet will be freed.
 274  * The mbuf opt, if present, will not be freed.
 275  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
 276  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 277  * then result of route lookup is stored in ro->ro_rt.
 278  *
 279  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
 280  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 281  * which is rt_mtu.
 282  *
 283  * ifpp - XXX: just for statistics
 284  */
 285 int
 286 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 287     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 288     struct ifnet **ifpp, struct inpcb *inp)
 289 {
 290         struct ip6_hdr *ip6;
 291         struct ifnet *ifp, *origifp;
 292         struct mbuf *m = m0;
 293         struct mbuf *mprev = NULL;
 294         int hlen, tlen, len;
 295         struct route_in6 ip6route;
 296         struct rtentry *rt = NULL;
 297         struct sockaddr_in6 *dst, src_sa, dst_sa;
 298         struct in6_addr odst;
 299         int error = 0;
 300         struct in6_ifaddr *ia = NULL;
 301         u_long mtu;
 302         int alwaysfrag, dontfrag;
 303         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 304         struct ip6_exthdrs exthdrs;
 305         struct in6_addr finaldst, src0, dst0;
 306         u_int32_t zone;
 307         struct route_in6 *ro_pmtu = NULL;
 308         int hdrsplit = 0;
 309         int sw_csum, tso;
 310         struct m_tag *fwd_tag = NULL;
 311         uint32_t id;
 312
 313         ip6 = mtod(m, struct ip6_hdr *);
 314         if (ip6 == NULL) {
 315                 printf ("ip6 is NULL");
 316                 goto bad;
 317         }
 318
 319         if (inp != NULL)
 320                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 321
 322         finaldst = ip6->ip6_dst;
 323         bzero(&exthdrs, sizeof(exthdrs));
 324         if (opt) {
 325                 /* Hop-by-Hop options header */
 326                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 327                 /* Destination options header(1st part) */
 328                 if (opt->ip6po_rthdr) {
 329                         /*
 330                          * Destination options header(1st part)
 331                          * This only makes sense with a routing header.
 332                          * See Section 9.2 of RFC 3542.
 333                          * Disabling this part just for MIP6 convenience is
 334                          * a bad idea.  We need to think carefully about a
 335                          * way to make the advanced API coexist with MIP6
 336                          * options, which might automatically be inserted in
 337                          * the kernel.
 338                          */
 339                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 340                 }
 341                 /* Routing header */
 342                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 343                 /* Destination options header(2nd part) */
 344                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 345         }
 346
 347 #ifdef IPSEC
 348         /*
 349          * IPSec checking which handles several cases.
 350          * FAST IPSEC: We re-injected the packet.
 351          */
 352         switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp))
 353         {
 354         case 1:                 /* Bad packet */
 355                 goto freehdrs;
 356         case -1:                /* IPSec done */
 357                 goto done;
 358         case 0:                 /* No IPSec */
 359         default:
 360                 break;
 361         }
 362 #endif /* IPSEC */
 363
 364         /*
 365          * Calculate the total length of the extension header chain.
 366          * Keep the length of the unfragmentable part for fragmentation.
 367          */
 368         optlen = 0;
 369         if (exthdrs.ip6e_hbh)
 370                 optlen += exthdrs.ip6e_hbh->m_len;
 371         if (exthdrs.ip6e_dest1)
 372                 optlen += exthdrs.ip6e_dest1->m_len;
 373         if (exthdrs.ip6e_rthdr)
 374                 optlen += exthdrs.ip6e_rthdr->m_len;
 375         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 376
 377         /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
 378         if (exthdrs.ip6e_dest2)
 379                 optlen += exthdrs.ip6e_dest2->m_len;
 380
 381         /*
 382          * If there is at least one extension header,
 383          * separate IP6 header from the payload.
 384          */
 385         if (optlen && !hdrsplit) {
 386                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 387                         m = NULL;
 388                         goto freehdrs;
 389                 }
 390                 m = exthdrs.ip6e_ip6;
 391                 hdrsplit++;
 392         }
 393
 394         /* adjust pointer */
 395         ip6 = mtod(m, struct ip6_hdr *);
 396
 397         /* adjust mbuf packet header length */
 398         m->m_pkthdr.len += optlen;
 399         plen = m->m_pkthdr.len - sizeof(*ip6);
 400
 401         /* If this is a jumbo payload, insert a jumbo payload option. */
 402         if (plen > IPV6_MAXPACKET) {
 403                 if (!hdrsplit) {
 404                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 405                                 m = NULL;
 406                                 goto freehdrs;
 407                         }
 408                         m = exthdrs.ip6e_ip6;
 409                         hdrsplit++;
 410                 }
 411                 /* adjust pointer */
 412                 ip6 = mtod(m, struct ip6_hdr *);
 413                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 414                         goto freehdrs;
 415                 ip6->ip6_plen = 0;
 416         } else
 417                 ip6->ip6_plen = htons(plen);
 418
 419         /*
 420          * Concatenate headers and fill in next header fields.
 421          * Here we have, on "m"
 422          *      IPv6 payload
 423          * and we insert headers accordingly.  Finally, we should be getting:
 424          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 425          *
 426          * during the header composing process, "m" points to IPv6 header.
 427          * "mprev" points to an extension header prior to esp.
 428          */
 429         u_char *nexthdrp = &ip6->ip6_nxt;
 430         mprev = m;
 431
 432         /*
 433          * we treat dest2 specially.  this makes IPsec processing
 434          * much easier.  the goal here is to make mprev point the
 435          * mbuf prior to dest2.
 436          *
 437          * result: IPv6 dest2 payload
 438          * m and mprev will point to IPv6 header.
 439          */
 440         if (exthdrs.ip6e_dest2) {
 441                 if (!hdrsplit)
 442                         panic("assumption failed: hdr not split");
 443                 exthdrs.ip6e_dest2->m_next = m->m_next;
 444                 m->m_next = exthdrs.ip6e_dest2;
 445                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 446                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 447         }
 448
 449         /*
 450          * result: IPv6 hbh dest1 rthdr dest2 payload
 451          * m will point to IPv6 header.  mprev will point to the
 452          * extension header prior to dest2 (rthdr in the above case).
 453          */
 454         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 455         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 456                    IPPROTO_DSTOPTS);
 457         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 458                    IPPROTO_ROUTING);
 459
 460         /*
 461          * If there is a routing header, discard the packet.
 462          */
 463         if (exthdrs.ip6e_rthdr) {
 464                  error = EINVAL;
 465                  goto bad;
 466         }
 467
 468         /* Source address validation */
 469         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 470             (flags & IPV6_UNSPECSRC) == 0) {
 471                 error = EOPNOTSUPP;
 472                 IP6STAT_INC(ip6s_badscope);
 473                 goto bad;
 474         }
 475         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 476                 error = EOPNOTSUPP;
 477                 IP6STAT_INC(ip6s_badscope);
 478                 goto bad;
 479         }
 480
 481         IP6STAT_INC(ip6s_localout);
 482
 483         /*
 484          * Route packet.
 485          */
 486         if (ro == 0) {
 487                 ro = &ip6route;
 488                 bzero((caddr_t)ro, sizeof(*ro));
 489         }
 490         ro_pmtu = ro;
 491         if (opt && opt->ip6po_rthdr)
 492                 ro = &opt->ip6po_route;
 493         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 494 #ifdef FLOWTABLE
 495         if (ro->ro_rt == NULL)
 496                 (void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
 497 #endif
 498 again:
 499         /*
 500          * if specified, try to fill in the traffic class field.
 501          * do not override if a non-zero value is already set.
 502          * we check the diffserv field and the ecn field separately.
 503          */
 504         if (opt && opt->ip6po_tclass >= 0) {
 505                 int mask = 0;
 506
 507                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 508                         mask |= 0xfc;
 509                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 510                         mask |= 0x03;
 511                 if (mask != 0)
 512                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 513         }
 514
 515         /* fill in or override the hop limit field, if necessary. */
 516         if (opt && opt->ip6po_hlim != -1)
 517                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 518         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 519                 if (im6o != NULL)
 520                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 521                 else
 522                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 523         }
 524
 525         /* adjust pointer */
 526         ip6 = mtod(m, struct ip6_hdr *);
 527
 528         if (ro->ro_rt && fwd_tag == NULL) {
 529                 rt = ro->ro_rt;
 530                 ifp = ro->ro_rt->rt_ifp;
 531         } else {
 532                 if (fwd_tag == NULL) {
 533                         bzero(&dst_sa, sizeof(dst_sa));
 534                         dst_sa.sin6_family = AF_INET6;
 535                         dst_sa.sin6_len = sizeof(dst_sa);
 536                         dst_sa.sin6_addr = ip6->ip6_dst;
 537                 }
 538                 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
 539                     &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
 540                 if (error != 0) {
 541                         if (ifp != NULL)
 542                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 543                         goto bad;
 544                 }
 545         }
 546         if (rt == NULL) {
 547                 /*
 548                  * If in6_selectroute() does not return a route entry,
 549                  * dst may not have been updated.
 550                  */
 551                 *dst = dst_sa;  /* XXX */
 552         }
 553
 554         /*
 555          * then rt (for unicast) and ifp must be non-NULL valid values.
 556          */
 557         if ((flags & IPV6_FORWARDING) == 0) {
 558                 /* XXX: the FORWARDING flag can be set for mrouting. */
 559                 in6_ifstat_inc(ifp, ifs6_out_request);
 560         }
 561         if (rt != NULL) {
 562                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 563                 counter_u64_add(rt->rt_pksent, 1);
 564         }
 565
 566
 567         /*
 568          * The outgoing interface must be in the zone of source and
 569          * destination addresses.
 570          */
 571         origifp = ifp;
 572
 573         src0 = ip6->ip6_src;
 574         if (in6_setscope(&src0, origifp, &zone))
 575                 goto badscope;
 576         bzero(&src_sa, sizeof(src_sa));
 577         src_sa.sin6_family = AF_INET6;
 578         src_sa.sin6_len = sizeof(src_sa);
 579         src_sa.sin6_addr = ip6->ip6_src;
 580         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 581                 goto badscope;
 582
 583         dst0 = ip6->ip6_dst;
 584         if (in6_setscope(&dst0, origifp, &zone))
 585                 goto badscope;
 586         /* re-initialize to be sure */
 587         bzero(&dst_sa, sizeof(dst_sa));
 588         dst_sa.sin6_family = AF_INET6;
 589         dst_sa.sin6_len = sizeof(dst_sa);
 590         dst_sa.sin6_addr = ip6->ip6_dst;
 591         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 592                 goto badscope;
 593         }
 594
 595         /* We should use ia_ifp to support the case of
 596          * sending packets to an address of our own.
 597          */
 598         if (ia != NULL && ia->ia_ifp)
 599                 ifp = ia->ia_ifp;
 600
 601         /* scope check is done. */
 602         goto routefound;
 603
 604   badscope:
 605         IP6STAT_INC(ip6s_badscope);
 606         in6_ifstat_inc(origifp, ifs6_out_discard);
 607         if (error == 0)
 608                 error = EHOSTUNREACH; /* XXX */
 609         goto bad;
 610
 611   routefound:
 612         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 613                 if (opt && opt->ip6po_nextroute.ro_rt) {
 614                         /*
 615                          * The nexthop is explicitly specified by the
 616                          * application.  We assume the next hop is an IPv6
 617                          * address.
 618                          */
 619                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 620                 }
 621                 else if ((rt->rt_flags & RTF_GATEWAY))
 622                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 623         }
 624
 625         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 626                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 627         } else {
 628                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 629                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 630                 /*
 631                  * Confirm that the outgoing interface supports multicast.
 632                  */
 633                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 634                         IP6STAT_INC(ip6s_noroute);
 635                         in6_ifstat_inc(ifp, ifs6_out_discard);
 636                         error = ENETUNREACH;
 637                         goto bad;
 638                 }
 639                 if ((im6o == NULL && in6_mcast_loop) ||
 640                     (im6o && im6o->im6o_multicast_loop)) {
 641                         /*
 642                          * Loop back multicast datagram if not expressly
 643                          * forbidden to do so, even if we have not joined
 644                          * the address; protocols will filter it later,
 645                          * thus deferring a hash lookup and lock acquisition
 646                          * at the expense of an m_copym().
 647                          */
 648                         ip6_mloopback(ifp, m, dst);
 649                 } else {
 650                         /*
 651                          * If we are acting as a multicast router, perform
 652                          * multicast forwarding as if the packet had just
 653                          * arrived on the interface to which we are about
 654                          * to send.  The multicast forwarding function
 655                          * recursively calls this function, using the
 656                          * IPV6_FORWARDING flag to prevent infinite recursion.
 657                          *
 658                          * Multicasts that are looped back by ip6_mloopback(),
 659                          * above, will be forwarded by the ip6_input() routine,
 660                          * if necessary.
 661                          */
 662                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 663                                 /*
 664                                  * XXX: ip6_mforward expects that rcvif is NULL
 665                                  * when it is called from the originating path.
 666                                  * However, it may not always be the case.
 667                                  */
 668                                 m->m_pkthdr.rcvif = NULL;
 669                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 670                                         m_freem(m);
 671                                         goto done;
 672                                 }
 673                         }
 674                 }
 675                 /*
 676                  * Multicasts with a hoplimit of zero may be looped back,
 677                  * above, but must not be transmitted on a network.
 678                  * Also, multicasts addressed to the loopback interface
 679                  * are not sent -- the above call to ip6_mloopback() will
 680                  * loop back a copy if this host actually belongs to the
 681                  * destination group on the loopback interface.
 682                  */
 683                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 684                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 685                         m_freem(m);
 686                         goto done;
 687                 }
 688         }
 689
 690         /*
 691          * Fill the outgoing inteface to tell the upper layer
 692          * to increment per-interface statistics.
 693          */
 694         if (ifpp)
 695                 *ifpp = ifp;
 696
 697         /* Determine path MTU. */
 698         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 699             &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
 700                 goto bad;
 701
 702         /*
 703          * The caller of this function may specify to use the minimum MTU
 704          * in some cases.
 705          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 706          * setting.  The logic is a bit complicated; by default, unicast
 707          * packets will follow path MTU while multicast packets will be sent at
 708          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 709          * including unicast ones will be sent at the minimum MTU.  Multicast
 710          * packets will always be sent at the minimum MTU unless
 711          * IP6PO_MINMTU_DISABLE is explicitly specified.
 712          * See RFC 3542 for more details.
 713          */
 714         if (mtu > IPV6_MMTU) {
 715                 if ((flags & IPV6_MINMTU))
 716                         mtu = IPV6_MMTU;
 717                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 718                         mtu = IPV6_MMTU;
 719                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 720                          (opt == NULL ||
 721                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 722                         mtu = IPV6_MMTU;
 723                 }
 724         }
 725
 726         /*
 727          * clear embedded scope identifiers if necessary.
 728          * in6_clearscope will touch the addresses only when necessary.
 729          */
 730         in6_clearscope(&ip6->ip6_src);
 731         in6_clearscope(&ip6->ip6_dst);
 732
 733         /*
 734          * If the outgoing packet contains a hop-by-hop options header,
 735          * it must be examined and processed even by the source node.
 736          * (RFC 2460, section 4.)
 737          */
 738         if (exthdrs.ip6e_hbh) {
 739                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 740                 u_int32_t dummy; /* XXX unused */
 741                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 742
 743 #ifdef DIAGNOSTIC
 744                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 745                         panic("ip6e_hbh is not contiguous");
 746 #endif
 747                 /*
 748                  *  XXX: if we have to send an ICMPv6 error to the sender,
 749                  *       we need the M_LOOP flag since icmp6_error() expects
 750                  *       the IPv6 and the hop-by-hop options header are
 751                  *       contiguous unless the flag is set.
 752                  */
 753                 m->m_flags |= M_LOOP;
 754                 m->m_pkthdr.rcvif = ifp;
 755                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 756                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 757                     &dummy, &plen) < 0) {
 758                         /* m was already freed at this point */
 759                         error = EINVAL;/* better error? */
 760                         goto done;
 761                 }
 762                 m->m_flags &= ~M_LOOP; /* XXX */
 763                 m->m_pkthdr.rcvif = NULL;
 764         }
 765
 766         /* Jump over all PFIL processing if hooks are not active. */
 767         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 768                 goto passout;
 769
 770         odst = ip6->ip6_dst;
 771         /* Run through list of hooks for output packets. */
 772         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 773         if (error != 0 || m == NULL)
 774                 goto done;
 775         ip6 = mtod(m, struct ip6_hdr *);
 776
 777         /* See if destination IP address was changed by packet filter. */
 778         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 779                 m->m_flags |= M_SKIP_FIREWALL;
 780                 /* If destination is now ourself drop to ip6_input(). */
 781                 if (in6_localip(&ip6->ip6_dst)) {
 782                         m->m_flags |= M_FASTFWD_OURS;
 783                         if (m->m_pkthdr.rcvif == NULL)
 784                                 m->m_pkthdr.rcvif = V_loif;
 785                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 786                                 m->m_pkthdr.csum_flags |=
 787                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 788                                 m->m_pkthdr.csum_data = 0xffff;
 789                         }
 790 #ifdef SCTP
 791                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 792                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 793 #endif
 794                         error = netisr_queue(NETISR_IPV6, m);
 795                         goto done;
 796                 } else
 797                         goto again;     /* Redo the routing table lookup. */
 798         }
 799
 800         /* See if local, if yes, send it to netisr. */
 801         if (m->m_flags & M_FASTFWD_OURS) {
 802                 if (m->m_pkthdr.rcvif == NULL)
 803                         m->m_pkthdr.rcvif = V_loif;
 804                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 805                         m->m_pkthdr.csum_flags |=
 806                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 807                         m->m_pkthdr.csum_data = 0xffff;
 808                 }
 809 #ifdef SCTP
 810                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 811                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 812 #endif
 813                 error = netisr_queue(NETISR_IPV6, m);
 814                 goto done;
 815         }
 816         /* Or forward to some other address? */
 817         if ((m->m_flags & M_IP6_NEXTHOP) &&
 818             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 819                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 820                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
 821                 m->m_flags |= M_SKIP_FIREWALL;
 822                 m->m_flags &= ~M_IP6_NEXTHOP;
 823                 m_tag_delete(m, fwd_tag);
 824                 goto again;
 825         }
 826
 827 passout:
 828         /*
 829          * Send the packet to the outgoing interface.
 830          * If necessary, do IPv6 fragmentation before sending.
 831          *
 832          * the logic here is rather complex:
 833          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 834          * 1-a: send as is if tlen <= path mtu
 835          * 1-b: fragment if tlen > path mtu
 836          *
 837          * 2: if user asks us not to fragment (dontfrag == 1)
 838          * 2-a: send as is if tlen <= interface mtu
 839          * 2-b: error if tlen > interface mtu
 840          *
 841          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 842          *      always fragment
 843          *
 844          * 4: if dontfrag == 1 && alwaysfrag == 1
 845          *      error, as we cannot handle this conflicting request
 846          */
 847         sw_csum = m->m_pkthdr.csum_flags;
 848         if (!hdrsplit) {
 849                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 850                 sw_csum &= ~ifp->if_hwassist;
 851         } else
 852                 tso = 0;
 853         /*
 854          * If we added extension headers, we will not do TSO and calculate the
 855          * checksums ourselves for now.
 856          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 857          * with ext. hdrs.
 858          */
 859         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 860                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 861                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 862         }
 863 #ifdef SCTP
 864         if (sw_csum & CSUM_SCTP_IPV6) {
 865                 sw_csum &= ~CSUM_SCTP_IPV6;
 866                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 867         }
 868 #endif
 869         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 870         tlen = m->m_pkthdr.len;
 871
 872         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 873                 dontfrag = 1;
 874         else
 875                 dontfrag = 0;
 876         if (dontfrag && alwaysfrag) {   /* case 4 */
 877                 /* conflicting request - can't transmit */
 878                 error = EMSGSIZE;
 879                 goto bad;
 880         }
 881         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 882                 /*
 883                  * Even if the DONTFRAG option is specified, we cannot send the
 884                  * packet when the data length is larger than the MTU of the
 885                  * outgoing interface.
 886                  * Notify the error by sending IPV6_PATHMTU ancillary data if
 887                  * application wanted to know the MTU value. Also return an
 888                  * error code (this is not described in the API spec).
 889                  */
 890                 if (inp != NULL)
 891                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
 892                 error = EMSGSIZE;
 893                 goto bad;
 894         }
 895
 896         /*
 897          * transmit packet without fragmentation
 898          */
 899         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 900                 struct in6_ifaddr *ia6;
 901
 902                 ip6 = mtod(m, struct ip6_hdr *);
 903                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 904                 if (ia6) {
 905                         /* Record statistics for this interface address. */
 906                         ia6->ia_ifa.if_opackets++;
 907                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
 908                         ifa_free(&ia6->ia_ifa);
 909                 }
 910                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 911                 goto done;
 912         }
 913
 914         /*
 915          * try to fragment the packet.  case 1-b and 3
 916          */
 917         if (mtu < IPV6_MMTU) {
 918                 /* path MTU cannot be less than IPV6_MMTU */
 919                 error = EMSGSIZE;
 920                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 921                 goto bad;
 922         } else if (ip6->ip6_plen == 0) {
 923                 /* jumbo payload cannot be fragmented */
 924                 error = EMSGSIZE;
 925                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 926                 goto bad;
 927         } else {
 928                 u_char nextproto;
 929
 930                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
 931
 932                 /*
 933                  * Too large for the destination or interface;
 934                  * fragment if possible.
 935                  * Must be able to put at least 8 bytes per fragment.
 936                  */
 937                 hlen = unfragpartlen;
 938                 if (mtu > IPV6_MAXPACKET)
 939                         mtu = IPV6_MAXPACKET;
 940
 941                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 942                 if (len < 8) {
 943                         error = EMSGSIZE;
 944                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 945                         goto bad;
 946                 }
 947
 948                 /*
 949                  * Verify that we have any chance at all of being able to queue
 950                  *      the packet or packet fragments
 951                  */
 952                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
 953                     < tlen  /* - hlen */)) {
 954                         error = ENOBUFS;
 955                         IP6STAT_INC(ip6s_odropped);
 956                         goto bad;
 957                 }
 958
 959
 960                 /*
 961                  * If the interface will not calculate checksums on
 962                  * fragmented packets, then do it here.
 963                  * XXX-BZ handle the hw offloading case.  Need flags.
 964                  */
 965                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 966                         in6_delayed_cksum(m, plen, hlen);
 967                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 968                 }
 969 #ifdef SCTP
 970                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
 971                         sctp_delayed_cksum(m, hlen);
 972                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 973                 }
 974 #endif
 975                 /*
 976                  * Change the next header field of the last header in the
 977                  * unfragmentable part.
 978                  */
 979                 if (exthdrs.ip6e_rthdr) {
 980                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
 981                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
 982                 } else if (exthdrs.ip6e_dest1) {
 983                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
 984                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
 985                 } else if (exthdrs.ip6e_hbh) {
 986                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
 987                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
 988                 } else {
 989                         nextproto = ip6->ip6_nxt;
 990                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
 991                 }
 992
 993                 /*
 994                  * Loop through length of segment after first fragment,
 995                  * make new header and copy data of each part and link onto
 996                  * chain.
 997                  */
 998                 m0 = m;
 999                 id = htonl(ip6_randomid());
1000                 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1001                         goto sendorfree;
1002
1003                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1004         }
1005
1006         /*
1007          * Remove leading garbages.
1008          */
1009 sendorfree:
1010         m = m0->m_nextpkt;
1011         m0->m_nextpkt = 0;
1012         m_freem(m0);
1013         for (m0 = m; m; m = m0) {
1014                 m0 = m->m_nextpkt;
1015                 m->m_nextpkt = 0;
1016                 if (error == 0) {
1017                         /* Record statistics for this interface address. */
1018                         if (ia) {
1019                                 ia->ia_ifa.if_opackets++;
1020                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1021                         }
1022                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1023                 } else
1024                         m_freem(m);
1025         }
1026
1027         if (error == 0)
1028                 IP6STAT_INC(ip6s_fragmented);
1029
1030 done:
1031         if (ro == &ip6route)
1032                 RO_RTFREE(ro);
1033         if (ro_pmtu == &ip6route)
1034                 RO_RTFREE(ro_pmtu);
1035         return (error);
1036
1037 freehdrs:
1038         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1039         m_freem(exthdrs.ip6e_dest1);
1040         m_freem(exthdrs.ip6e_rthdr);
1041         m_freem(exthdrs.ip6e_dest2);
1042         /* FALLTHROUGH */
1043 bad:
1044         if (m)
1045                 m_freem(m);
1046         goto done;
1047 }
1048
1049 static int
1050 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1051 {
1052         struct mbuf *m;
1053
1054         if (hlen > MCLBYTES)
1055                 return (ENOBUFS); /* XXX */
1056
1057         if (hlen > MLEN)
1058                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1059         else
1060                 m = m_get(M_NOWAIT, MT_DATA);
1061         if (m == NULL)
1062                 return (ENOBUFS);
1063         m->m_len = hlen;
1064         if (hdr)
1065                 bcopy(hdr, mtod(m, caddr_t), hlen);
1066
1067         *mp = m;
1068         return (0);
1069 }
1070
1071 /*
1072  * Insert jumbo payload option.
1073  */
1074 static int
1075 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1076 {
1077         struct mbuf *mopt;
1078         u_char *optbuf;
1079         u_int32_t v;
1080
1081 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1082
1083         /*
1084          * If there is no hop-by-hop options header, allocate new one.
1085          * If there is one but it doesn't have enough space to store the
1086          * jumbo payload option, allocate a cluster to store the whole options.
1087          * Otherwise, use it to store the options.
1088          */
1089         if (exthdrs->ip6e_hbh == 0) {
1090                 mopt = m_get(M_NOWAIT, MT_DATA);
1091                 if (mopt == NULL)
1092                         return (ENOBUFS);
1093                 mopt->m_len = JUMBOOPTLEN;
1094                 optbuf = mtod(mopt, u_char *);
1095                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1096                 exthdrs->ip6e_hbh = mopt;
1097         } else {
1098                 struct ip6_hbh *hbh;
1099
1100                 mopt = exthdrs->ip6e_hbh;
1101                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1102                         /*
1103                          * XXX assumption:
1104                          * - exthdrs->ip6e_hbh is not referenced from places
1105                          *   other than exthdrs.
1106                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1107                          */
1108                         int oldoptlen = mopt->m_len;
1109                         struct mbuf *n;
1110
1111                         /*
1112                          * XXX: give up if the whole (new) hbh header does
1113                          * not fit even in an mbuf cluster.
1114                          */
1115                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1116                                 return (ENOBUFS);
1117
1118                         /*
1119                          * As a consequence, we must always prepare a cluster
1120                          * at this point.
1121                          */
1122                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1123                         if (n == NULL)
1124                                 return (ENOBUFS);
1125                         n->m_len = oldoptlen + JUMBOOPTLEN;
1126                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1127                             oldoptlen);
1128                         optbuf = mtod(n, caddr_t) + oldoptlen;
1129                         m_freem(mopt);
1130                         mopt = exthdrs->ip6e_hbh = n;
1131                 } else {
1132                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1133                         mopt->m_len += JUMBOOPTLEN;
1134                 }
1135                 optbuf[0] = IP6OPT_PADN;
1136                 optbuf[1] = 1;
1137
1138                 /*
1139                  * Adjust the header length according to the pad and
1140                  * the jumbo payload option.
1141                  */
1142                 hbh = mtod(mopt, struct ip6_hbh *);
1143                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1144         }
1145
1146         /* fill in the option. */
1147         optbuf[2] = IP6OPT_JUMBO;
1148         optbuf[3] = 4;
1149         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1150         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1151
1152         /* finally, adjust the packet header length */
1153         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1154
1155         return (0);
1156 #undef JUMBOOPTLEN
1157 }
1158
1159 /*
1160  * Insert fragment header and copy unfragmentable header portions.
1161  */
1162 static int
1163 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1164     struct ip6_frag **frghdrp)
1165 {
1166         struct mbuf *n, *mlast;
1167
1168         if (hlen > sizeof(struct ip6_hdr)) {
1169                 n = m_copym(m0, sizeof(struct ip6_hdr),
1170                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1171                 if (n == 0)
1172                         return (ENOBUFS);
1173                 m->m_next = n;
1174         } else
1175                 n = m;
1176
1177         /* Search for the last mbuf of unfragmentable part. */
1178         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1179                 ;
1180
1181         if ((mlast->m_flags & M_EXT) == 0 &&
1182             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1183                 /* use the trailing space of the last mbuf for the fragment hdr */
1184                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1185                     mlast->m_len);
1186                 mlast->m_len += sizeof(struct ip6_frag);
1187                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1188         } else {
1189                 /* allocate a new mbuf for the fragment header */
1190                 struct mbuf *mfrg;
1191
1192                 mfrg = m_get(M_NOWAIT, MT_DATA);
1193                 if (mfrg == NULL)
1194                         return (ENOBUFS);
1195                 mfrg->m_len = sizeof(struct ip6_frag);
1196                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1197                 mlast->m_next = mfrg;
1198         }
1199
1200         return (0);
1201 }
1202
1203 static int
1204 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1205     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1206     int *alwaysfragp, u_int fibnum)
1207 {
1208         u_int32_t mtu = 0;
1209         int alwaysfrag = 0;
1210         int error = 0;
1211
1212         if (ro_pmtu != ro) {
1213                 /* The first hop and the final destination may differ. */
1214                 struct sockaddr_in6 *sa6_dst =
1215                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1216                 if (ro_pmtu->ro_rt &&
1217                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1218                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1219                         RTFREE(ro_pmtu->ro_rt);
1220                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1221                 }
1222                 if (ro_pmtu->ro_rt == NULL) {
1223                         bzero(sa6_dst, sizeof(*sa6_dst));
1224                         sa6_dst->sin6_family = AF_INET6;
1225                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1226                         sa6_dst->sin6_addr = *dst;
1227
1228                         in6_rtalloc(ro_pmtu, fibnum);
1229                 }
1230         }
1231         if (ro_pmtu->ro_rt) {
1232                 u_int32_t ifmtu;
1233                 struct in_conninfo inc;
1234
1235                 bzero(&inc, sizeof(inc));
1236                 inc.inc_flags |= INC_ISIPV6;
1237                 inc.inc6_faddr = *dst;
1238
1239                 if (ifp == NULL)
1240                         ifp = ro_pmtu->ro_rt->rt_ifp;
1241                 ifmtu = IN6_LINKMTU(ifp);
1242                 mtu = tcp_hc_getmtu(&inc);
1243                 if (mtu)
1244                         mtu = min(mtu, ro_pmtu->ro_rt->rt_mtu);
1245                 else
1246                         mtu = ro_pmtu->ro_rt->rt_mtu;
1247                 if (mtu == 0)
1248                         mtu = ifmtu;
1249                 else if (mtu < IPV6_MMTU) {
1250                         /*
1251                          * RFC2460 section 5, last paragraph:
1252                          * if we record ICMPv6 too big message with
1253                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1254                          * or smaller, with framgent header attached.
1255                          * (fragment header is needed regardless from the
1256                          * packet size, for translators to identify packets)
1257                          */
1258                         alwaysfrag = 1;
1259                         mtu = IPV6_MMTU;
1260                 } else if (mtu > ifmtu) {
1261                         /*
1262                          * The MTU on the route is larger than the MTU on
1263                          * the interface!  This shouldn't happen, unless the
1264                          * MTU of the interface has been changed after the
1265                          * interface was brought up.  Change the MTU in the
1266                          * route to match the interface MTU (as long as the
1267                          * field isn't locked).
1268                          */
1269                         mtu = ifmtu;
1270                         ro_pmtu->ro_rt->rt_mtu = mtu;
1271                 }
1272         } else if (ifp) {
1273                 mtu = IN6_LINKMTU(ifp);
1274         } else
1275                 error = EHOSTUNREACH; /* XXX */
1276
1277         *mtup = mtu;
1278         if (alwaysfragp)
1279                 *alwaysfragp = alwaysfrag;
1280         return (error);
1281 }
1282
1283 /*
1284  * IP6 socket option processing.
1285  */
1286 int
1287 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1288 {
1289         int optdatalen, uproto;
1290         void *optdata;
1291         struct inpcb *in6p = sotoinpcb(so);
1292         int error, optval;
1293         int level, op, optname;
1294         int optlen;
1295         struct thread *td;
1296
1297         level = sopt->sopt_level;
1298         op = sopt->sopt_dir;
1299         optname = sopt->sopt_name;
1300         optlen = sopt->sopt_valsize;
1301         td = sopt->sopt_td;
1302         error = 0;
1303         optval = 0;
1304         uproto = (int)so->so_proto->pr_protocol;
1305
1306         if (level != IPPROTO_IPV6) {
1307                 error = EINVAL;
1308
1309                 if (sopt->sopt_level == SOL_SOCKET &&
1310                     sopt->sopt_dir == SOPT_SET) {
1311                         switch (sopt->sopt_name) {
1312                         case SO_REUSEADDR:
1313                                 INP_WLOCK(in6p);
1314                                 if ((so->so_options & SO_REUSEADDR) != 0)
1315                                         in6p->inp_flags2 |= INP_REUSEADDR;
1316                                 else
1317                                         in6p->inp_flags2 &= ~INP_REUSEADDR;
1318                                 INP_WUNLOCK(in6p);
1319                                 error = 0;
1320                                 break;
1321                         case SO_REUSEPORT:
1322                                 INP_WLOCK(in6p);
1323                                 if ((so->so_options & SO_REUSEPORT) != 0)
1324                                         in6p->inp_flags2 |= INP_REUSEPORT;
1325                                 else
1326                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1327                                 INP_WUNLOCK(in6p);
1328                                 error = 0;
1329                                 break;
1330                         case SO_SETFIB:
1331                                 INP_WLOCK(in6p);
1332                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1333                                 INP_WUNLOCK(in6p);
1334                                 error = 0;
1335                                 break;
1336                         default:
1337                                 break;
1338                         }
1339                 }
1340         } else {                /* level == IPPROTO_IPV6 */
1341                 switch (op) {
1342
1343                 case SOPT_SET:
1344                         switch (optname) {
1345                         case IPV6_2292PKTOPTIONS:
1346 #ifdef IPV6_PKTOPTIONS
1347                         case IPV6_PKTOPTIONS:
1348 #endif
1349                         {
1350                                 struct mbuf *m;
1351
1352                                 error = soopt_getm(sopt, &m); /* XXX */
1353                                 if (error != 0)
1354                                         break;
1355                                 error = soopt_mcopyin(sopt, m); /* XXX */
1356                                 if (error != 0)
1357                                         break;
1358                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1359                                                     m, so, sopt);
1360                                 m_freem(m); /* XXX */
1361                                 break;
1362                         }
1363
1364                         /*
1365                          * Use of some Hop-by-Hop options or some
1366                          * Destination options, might require special
1367                          * privilege.  That is, normal applications
1368                          * (without special privilege) might be forbidden
1369                          * from setting certain options in outgoing packets,
1370                          * and might never see certain options in received
1371                          * packets. [RFC 2292 Section 6]
1372                          * KAME specific note:
1373                          *  KAME prevents non-privileged users from sending or
1374                          *  receiving ANY hbh/dst options in order to avoid
1375                          *  overhead of parsing options in the kernel.
1376                          */
1377                         case IPV6_RECVHOPOPTS:
1378                         case IPV6_RECVDSTOPTS:
1379                         case IPV6_RECVRTHDRDSTOPTS:
1380                                 if (td != NULL) {
1381                                         error = priv_check(td,
1382                                             PRIV_NETINET_SETHDROPTS);
1383                                         if (error)
1384                                                 break;
1385                                 }
1386                                 /* FALLTHROUGH */
1387                         case IPV6_UNICAST_HOPS:
1388                         case IPV6_HOPLIMIT:
1389                         case IPV6_FAITH:
1390
1391                         case IPV6_RECVPKTINFO:
1392                         case IPV6_RECVHOPLIMIT:
1393                         case IPV6_RECVRTHDR:
1394                         case IPV6_RECVPATHMTU:
1395                         case IPV6_RECVTCLASS:
1396                         case IPV6_V6ONLY:
1397                         case IPV6_AUTOFLOWLABEL:
1398                         case IPV6_BINDANY:
1399                                 if (optname == IPV6_BINDANY && td != NULL) {
1400                                         error = priv_check(td,
1401                                             PRIV_NETINET_BINDANY);
1402                                         if (error)
1403                                                 break;
1404                                 }
1405
1406                                 if (optlen != sizeof(int)) {
1407                                         error = EINVAL;
1408                                         break;
1409                                 }
1410                                 error = sooptcopyin(sopt, &optval,
1411                                         sizeof optval, sizeof optval);
1412                                 if (error)
1413                                         break;
1414                                 switch (optname) {
1415
1416                                 case IPV6_UNICAST_HOPS:
1417                                         if (optval < -1 || optval >= 256)
1418                                                 error = EINVAL;
1419                                         else {
1420                                                 /* -1 = kernel default */
1421                                                 in6p->in6p_hops = optval;
1422                                                 if ((in6p->inp_vflag &
1423                                                      INP_IPV4) != 0)
1424                                                         in6p->inp_ip_ttl = optval;
1425                                         }
1426                                         break;
1427 #define OPTSET(bit) \
1428 do { \
1429         INP_WLOCK(in6p); \
1430         if (optval) \
1431                 in6p->inp_flags |= (bit); \
1432         else \
1433                 in6p->inp_flags &= ~(bit); \
1434         INP_WUNLOCK(in6p); \
1435 } while (/*CONSTCOND*/ 0)
1436 #define OPTSET2292(bit) \
1437 do { \
1438         INP_WLOCK(in6p); \
1439         in6p->inp_flags |= IN6P_RFC2292; \
1440         if (optval) \
1441                 in6p->inp_flags |= (bit); \
1442         else \
1443                 in6p->inp_flags &= ~(bit); \
1444         INP_WUNLOCK(in6p); \
1445 } while (/*CONSTCOND*/ 0)
1446 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1447
1448                                 case IPV6_RECVPKTINFO:
1449                                         /* cannot mix with RFC2292 */
1450                                         if (OPTBIT(IN6P_RFC2292)) {
1451                                                 error = EINVAL;
1452                                                 break;
1453                                         }
1454                                         OPTSET(IN6P_PKTINFO);
1455                                         break;
1456
1457                                 case IPV6_HOPLIMIT:
1458                                 {
1459                                         struct ip6_pktopts **optp;
1460
1461                                         /* cannot mix with RFC2292 */
1462                                         if (OPTBIT(IN6P_RFC2292)) {
1463                                                 error = EINVAL;
1464                                                 break;
1465                                         }
1466                                         optp = &in6p->in6p_outputopts;
1467                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1468                                             (u_char *)&optval, sizeof(optval),
1469                                             optp, (td != NULL) ? td->td_ucred :
1470                                             NULL, uproto);
1471                                         break;
1472                                 }
1473
1474                                 case IPV6_RECVHOPLIMIT:
1475                                         /* cannot mix with RFC2292 */
1476                                         if (OPTBIT(IN6P_RFC2292)) {
1477                                                 error = EINVAL;
1478                                                 break;
1479                                         }
1480                                         OPTSET(IN6P_HOPLIMIT);
1481                                         break;
1482
1483                                 case IPV6_RECVHOPOPTS:
1484                                         /* cannot mix with RFC2292 */
1485                                         if (OPTBIT(IN6P_RFC2292)) {
1486                                                 error = EINVAL;
1487                                                 break;
1488                                         }
1489                                         OPTSET(IN6P_HOPOPTS);
1490                                         break;
1491
1492                                 case IPV6_RECVDSTOPTS:
1493                                         /* cannot mix with RFC2292 */
1494                                         if (OPTBIT(IN6P_RFC2292)) {
1495                                                 error = EINVAL;
1496                                                 break;
1497                                         }
1498                                         OPTSET(IN6P_DSTOPTS);
1499                                         break;
1500
1501                                 case IPV6_RECVRTHDRDSTOPTS:
1502                                         /* cannot mix with RFC2292 */
1503                                         if (OPTBIT(IN6P_RFC2292)) {
1504                                                 error = EINVAL;
1505                                                 break;
1506                                         }
1507                                         OPTSET(IN6P_RTHDRDSTOPTS);
1508                                         break;
1509
1510                                 case IPV6_RECVRTHDR:
1511                                         /* cannot mix with RFC2292 */
1512                                         if (OPTBIT(IN6P_RFC2292)) {
1513                                                 error = EINVAL;
1514                                                 break;
1515                                         }
1516                                         OPTSET(IN6P_RTHDR);
1517                                         break;
1518
1519                                 case IPV6_FAITH:
1520                                         OPTSET(INP_FAITH);
1521                                         break;
1522
1523                                 case IPV6_RECVPATHMTU:
1524                                         /*
1525                                          * We ignore this option for TCP
1526                                          * sockets.
1527                                          * (RFC3542 leaves this case
1528                                          * unspecified.)
1529                                          */
1530                                         if (uproto != IPPROTO_TCP)
1531                                                 OPTSET(IN6P_MTU);
1532                                         break;
1533
1534                                 case IPV6_V6ONLY:
1535                                         /*
1536                                          * make setsockopt(IPV6_V6ONLY)
1537                                          * available only prior to bind(2).
1538                                          * see ipng mailing list, Jun 22 2001.
1539                                          */
1540                                         if (in6p->inp_lport ||
1541                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1542                                                 error = EINVAL;
1543                                                 break;
1544                                         }
1545                                         OPTSET(IN6P_IPV6_V6ONLY);
1546                                         if (optval)
1547                                                 in6p->inp_vflag &= ~INP_IPV4;
1548                                         else
1549                                                 in6p->inp_vflag |= INP_IPV4;
1550                                         break;
1551                                 case IPV6_RECVTCLASS:
1552                                         /* cannot mix with RFC2292 XXX */
1553                                         if (OPTBIT(IN6P_RFC2292)) {
1554                                                 error = EINVAL;
1555                                                 break;
1556                                         }
1557                                         OPTSET(IN6P_TCLASS);
1558                                         break;
1559                                 case IPV6_AUTOFLOWLABEL:
1560                                         OPTSET(IN6P_AUTOFLOWLABEL);
1561                                         break;
1562
1563                                 case IPV6_BINDANY:
1564                                         OPTSET(INP_BINDANY);
1565                                         break;
1566                                 }
1567                                 break;
1568
1569                         case IPV6_TCLASS:
1570                         case IPV6_DONTFRAG:
1571                         case IPV6_USE_MIN_MTU:
1572                         case IPV6_PREFER_TEMPADDR:
1573                                 if (optlen != sizeof(optval)) {
1574                                         error = EINVAL;
1575                                         break;
1576                                 }
1577                                 error = sooptcopyin(sopt, &optval,
1578                                         sizeof optval, sizeof optval);
1579                                 if (error)
1580                                         break;
1581                                 {
1582                                         struct ip6_pktopts **optp;
1583                                         optp = &in6p->in6p_outputopts;
1584                                         error = ip6_pcbopt(optname,
1585                                             (u_char *)&optval, sizeof(optval),
1586                                             optp, (td != NULL) ? td->td_ucred :
1587                                             NULL, uproto);
1588                                         break;
1589                                 }
1590
1591                         case IPV6_2292PKTINFO:
1592                         case IPV6_2292HOPLIMIT:
1593                         case IPV6_2292HOPOPTS:
1594                         case IPV6_2292DSTOPTS:
1595                         case IPV6_2292RTHDR:
1596                                 /* RFC 2292 */
1597                                 if (optlen != sizeof(int)) {
1598                                         error = EINVAL;
1599                                         break;
1600                                 }
1601                                 error = sooptcopyin(sopt, &optval,
1602                                         sizeof optval, sizeof optval);
1603                                 if (error)
1604                                         break;
1605                                 switch (optname) {
1606                                 case IPV6_2292PKTINFO:
1607                                         OPTSET2292(IN6P_PKTINFO);
1608                                         break;
1609                                 case IPV6_2292HOPLIMIT:
1610                                         OPTSET2292(IN6P_HOPLIMIT);
1611                                         break;
1612                                 case IPV6_2292HOPOPTS:
1613                                         /*
1614                                          * Check super-user privilege.
1615                                          * See comments for IPV6_RECVHOPOPTS.
1616                                          */
1617                                         if (td != NULL) {
1618                                                 error = priv_check(td,
1619                                                     PRIV_NETINET_SETHDROPTS);
1620                                                 if (error)
1621                                                         return (error);
1622                                         }
1623                                         OPTSET2292(IN6P_HOPOPTS);
1624                                         break;
1625                                 case IPV6_2292DSTOPTS:
1626                                         if (td != NULL) {
1627                                                 error = priv_check(td,
1628                                                     PRIV_NETINET_SETHDROPTS);
1629                                                 if (error)
1630                                                         return (error);
1631                                         }
1632                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1633                                         break;
1634                                 case IPV6_2292RTHDR:
1635                                         OPTSET2292(IN6P_RTHDR);
1636                                         break;
1637                                 }
1638                                 break;
1639                         case IPV6_PKTINFO:
1640                         case IPV6_HOPOPTS:
1641                         case IPV6_RTHDR:
1642                         case IPV6_DSTOPTS:
1643                         case IPV6_RTHDRDSTOPTS:
1644                         case IPV6_NEXTHOP:
1645                         {
1646                                 /* new advanced API (RFC3542) */
1647                                 u_char *optbuf;
1648                                 u_char optbuf_storage[MCLBYTES];
1649                                 int optlen;
1650                                 struct ip6_pktopts **optp;
1651
1652                                 /* cannot mix with RFC2292 */
1653                                 if (OPTBIT(IN6P_RFC2292)) {
1654                                         error = EINVAL;
1655                                         break;
1656                                 }
1657
1658                                 /*
1659                                  * We only ensure valsize is not too large
1660                                  * here.  Further validation will be done
1661                                  * later.
1662                                  */
1663                                 error = sooptcopyin(sopt, optbuf_storage,
1664                                     sizeof(optbuf_storage), 0);
1665                                 if (error)
1666                                         break;
1667                                 optlen = sopt->sopt_valsize;
1668                                 optbuf = optbuf_storage;
1669                                 optp = &in6p->in6p_outputopts;
1670                                 error = ip6_pcbopt(optname, optbuf, optlen,
1671                                     optp, (td != NULL) ? td->td_ucred : NULL,
1672                                     uproto);
1673                                 break;
1674                         }
1675 #undef OPTSET
1676
1677                         case IPV6_MULTICAST_IF:
1678                         case IPV6_MULTICAST_HOPS:
1679                         case IPV6_MULTICAST_LOOP:
1680                         case IPV6_JOIN_GROUP:
1681                         case IPV6_LEAVE_GROUP:
1682                         case IPV6_MSFILTER:
1683                         case MCAST_BLOCK_SOURCE:
1684                         case MCAST_UNBLOCK_SOURCE:
1685                         case MCAST_JOIN_GROUP:
1686                         case MCAST_LEAVE_GROUP:
1687                         case MCAST_JOIN_SOURCE_GROUP:
1688                         case MCAST_LEAVE_SOURCE_GROUP:
1689                                 error = ip6_setmoptions(in6p, sopt);
1690                                 break;
1691
1692                         case IPV6_PORTRANGE:
1693                                 error = sooptcopyin(sopt, &optval,
1694                                     sizeof optval, sizeof optval);
1695                                 if (error)
1696                                         break;
1697
1698                                 INP_WLOCK(in6p);
1699                                 switch (optval) {
1700                                 case IPV6_PORTRANGE_DEFAULT:
1701                                         in6p->inp_flags &= ~(INP_LOWPORT);
1702                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1703                                         break;
1704
1705                                 case IPV6_PORTRANGE_HIGH:
1706                                         in6p->inp_flags &= ~(INP_LOWPORT);
1707                                         in6p->inp_flags |= INP_HIGHPORT;
1708                                         break;
1709
1710                                 case IPV6_PORTRANGE_LOW:
1711                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1712                                         in6p->inp_flags |= INP_LOWPORT;
1713                                         break;
1714
1715                                 default:
1716                                         error = EINVAL;
1717                                         break;
1718                                 }
1719                                 INP_WUNLOCK(in6p);
1720                                 break;
1721
1722 #ifdef IPSEC
1723                         case IPV6_IPSEC_POLICY:
1724                         {
1725                                 caddr_t req;
1726                                 struct mbuf *m;
1727
1728                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1729                                         break;
1730                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1731                                         break;
1732                                 req = mtod(m, caddr_t);
1733                                 error = ipsec_set_policy(in6p, optname, req,
1734                                     m->m_len, (sopt->sopt_td != NULL) ?
1735                                     sopt->sopt_td->td_ucred : NULL);
1736                                 m_freem(m);
1737                                 break;
1738                         }
1739 #endif /* IPSEC */
1740
1741                         default:
1742                                 error = ENOPROTOOPT;
1743                                 break;
1744                         }
1745                         break;
1746
1747                 case SOPT_GET:
1748                         switch (optname) {
1749
1750                         case IPV6_2292PKTOPTIONS:
1751 #ifdef IPV6_PKTOPTIONS
1752                         case IPV6_PKTOPTIONS:
1753 #endif
1754                                 /*
1755                                  * RFC3542 (effectively) deprecated the
1756                                  * semantics of the 2292-style pktoptions.
1757                                  * Since it was not reliable in nature (i.e.,
1758                                  * applications had to expect the lack of some
1759                                  * information after all), it would make sense
1760                                  * to simplify this part by always returning
1761                                  * empty data.
1762                                  */
1763                                 sopt->sopt_valsize = 0;
1764                                 break;
1765
1766                         case IPV6_RECVHOPOPTS:
1767                         case IPV6_RECVDSTOPTS:
1768                         case IPV6_RECVRTHDRDSTOPTS:
1769                         case IPV6_UNICAST_HOPS:
1770                         case IPV6_RECVPKTINFO:
1771                         case IPV6_RECVHOPLIMIT:
1772                         case IPV6_RECVRTHDR:
1773                         case IPV6_RECVPATHMTU:
1774
1775                         case IPV6_FAITH:
1776                         case IPV6_V6ONLY:
1777                         case IPV6_PORTRANGE:
1778                         case IPV6_RECVTCLASS:
1779                         case IPV6_AUTOFLOWLABEL:
1780                         case IPV6_BINDANY:
1781                                 switch (optname) {
1782
1783                                 case IPV6_RECVHOPOPTS:
1784                                         optval = OPTBIT(IN6P_HOPOPTS);
1785                                         break;
1786
1787                                 case IPV6_RECVDSTOPTS:
1788                                         optval = OPTBIT(IN6P_DSTOPTS);
1789                                         break;
1790
1791                                 case IPV6_RECVRTHDRDSTOPTS:
1792                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1793                                         break;
1794
1795                                 case IPV6_UNICAST_HOPS:
1796                                         optval = in6p->in6p_hops;
1797                                         break;
1798
1799                                 case IPV6_RECVPKTINFO:
1800                                         optval = OPTBIT(IN6P_PKTINFO);
1801                                         break;
1802
1803                                 case IPV6_RECVHOPLIMIT:
1804                                         optval = OPTBIT(IN6P_HOPLIMIT);
1805                                         break;
1806
1807                                 case IPV6_RECVRTHDR:
1808                                         optval = OPTBIT(IN6P_RTHDR);
1809                                         break;
1810
1811                                 case IPV6_RECVPATHMTU:
1812                                         optval = OPTBIT(IN6P_MTU);
1813                                         break;
1814
1815                                 case IPV6_FAITH:
1816                                         optval = OPTBIT(INP_FAITH);
1817                                         break;
1818
1819                                 case IPV6_V6ONLY:
1820                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1821                                         break;
1822
1823                                 case IPV6_PORTRANGE:
1824                                     {
1825                                         int flags;
1826                                         flags = in6p->inp_flags;
1827                                         if (flags & INP_HIGHPORT)
1828                                                 optval = IPV6_PORTRANGE_HIGH;
1829                                         else if (flags & INP_LOWPORT)
1830                                                 optval = IPV6_PORTRANGE_LOW;
1831                                         else
1832                                                 optval = 0;
1833                                         break;
1834                                     }
1835                                 case IPV6_RECVTCLASS:
1836                                         optval = OPTBIT(IN6P_TCLASS);
1837                                         break;
1838
1839                                 case IPV6_AUTOFLOWLABEL:
1840                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1841                                         break;
1842
1843                                 case IPV6_BINDANY:
1844                                         optval = OPTBIT(INP_BINDANY);
1845                                         break;
1846                                 }
1847                                 if (error)
1848                                         break;
1849                                 error = sooptcopyout(sopt, &optval,
1850                                         sizeof optval);
1851                                 break;
1852
1853                         case IPV6_PATHMTU:
1854                         {
1855                                 u_long pmtu = 0;
1856                                 struct ip6_mtuinfo mtuinfo;
1857                                 struct route_in6 sro;
1858
1859                                 bzero(&sro, sizeof(sro));
1860
1861                                 if (!(so->so_state & SS_ISCONNECTED))
1862                                         return (ENOTCONN);
1863                                 /*
1864                                  * XXX: we dot not consider the case of source
1865                                  * routing, or optional information to specify
1866                                  * the outgoing interface.
1867                                  */
1868                                 error = ip6_getpmtu(&sro, NULL, NULL,
1869                                     &in6p->in6p_faddr, &pmtu, NULL,
1870                                     so->so_fibnum);
1871                                 if (sro.ro_rt)
1872                                         RTFREE(sro.ro_rt);
1873                                 if (error)
1874                                         break;
1875                                 if (pmtu > IPV6_MAXPACKET)
1876                                         pmtu = IPV6_MAXPACKET;
1877
1878                                 bzero(&mtuinfo, sizeof(mtuinfo));
1879                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1880                                 optdata = (void *)&mtuinfo;
1881                                 optdatalen = sizeof(mtuinfo);
1882                                 error = sooptcopyout(sopt, optdata,
1883                                     optdatalen);
1884                                 break;
1885                         }
1886
1887                         case IPV6_2292PKTINFO:
1888                         case IPV6_2292HOPLIMIT:
1889                         case IPV6_2292HOPOPTS:
1890                         case IPV6_2292RTHDR:
1891                         case IPV6_2292DSTOPTS:
1892                                 switch (optname) {
1893                                 case IPV6_2292PKTINFO:
1894                                         optval = OPTBIT(IN6P_PKTINFO);
1895                                         break;
1896                                 case IPV6_2292HOPLIMIT:
1897                                         optval = OPTBIT(IN6P_HOPLIMIT);
1898                                         break;
1899                                 case IPV6_2292HOPOPTS:
1900                                         optval = OPTBIT(IN6P_HOPOPTS);
1901                                         break;
1902                                 case IPV6_2292RTHDR:
1903                                         optval = OPTBIT(IN6P_RTHDR);
1904                                         break;
1905                                 case IPV6_2292DSTOPTS:
1906                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1907                                         break;
1908                                 }
1909                                 error = sooptcopyout(sopt, &optval,
1910                                     sizeof optval);
1911                                 break;
1912                         case IPV6_PKTINFO:
1913                         case IPV6_HOPOPTS:
1914                         case IPV6_RTHDR:
1915                         case IPV6_DSTOPTS:
1916                         case IPV6_RTHDRDSTOPTS:
1917                         case IPV6_NEXTHOP:
1918                         case IPV6_TCLASS:
1919                         case IPV6_DONTFRAG:
1920                         case IPV6_USE_MIN_MTU:
1921                         case IPV6_PREFER_TEMPADDR:
1922                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
1923                                     optname, sopt);
1924                                 break;
1925
1926                         case IPV6_MULTICAST_IF:
1927                         case IPV6_MULTICAST_HOPS:
1928                         case IPV6_MULTICAST_LOOP:
1929                         case IPV6_MSFILTER:
1930                                 error = ip6_getmoptions(in6p, sopt);
1931                                 break;
1932
1933 #ifdef IPSEC
1934                         case IPV6_IPSEC_POLICY:
1935                           {
1936                                 caddr_t req = NULL;
1937                                 size_t len = 0;
1938                                 struct mbuf *m = NULL;
1939                                 struct mbuf **mp = &m;
1940                                 size_t ovalsize = sopt->sopt_valsize;
1941                                 caddr_t oval = (caddr_t)sopt->sopt_val;
1942
1943                                 error = soopt_getm(sopt, &m); /* XXX */
1944                                 if (error != 0)
1945                                         break;
1946                                 error = soopt_mcopyin(sopt, m); /* XXX */
1947                                 if (error != 0)
1948                                         break;
1949                                 sopt->sopt_valsize = ovalsize;
1950                                 sopt->sopt_val = oval;
1951                                 if (m) {
1952                                         req = mtod(m, caddr_t);
1953                                         len = m->m_len;
1954                                 }
1955                                 error = ipsec_get_policy(in6p, req, len, mp);
1956                                 if (error == 0)
1957                                         error = soopt_mcopyout(sopt, m); /* XXX */
1958                                 if (error == 0 && m)
1959                                         m_freem(m);
1960                                 break;
1961                           }
1962 #endif /* IPSEC */
1963
1964                         default:
1965                                 error = ENOPROTOOPT;
1966                                 break;
1967                         }
1968                         break;
1969                 }
1970         }
1971         return (error);
1972 }
1973
1974 int
1975 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
1976 {
1977         int error = 0, optval, optlen;
1978         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1979         struct inpcb *in6p = sotoinpcb(so);
1980         int level, op, optname;
1981
1982         level = sopt->sopt_level;
1983         op = sopt->sopt_dir;
1984         optname = sopt->sopt_name;
1985         optlen = sopt->sopt_valsize;
1986
1987         if (level != IPPROTO_IPV6) {
1988                 return (EINVAL);
1989         }
1990
1991         switch (optname) {
1992         case IPV6_CHECKSUM:
1993                 /*
1994                  * For ICMPv6 sockets, no modification allowed for checksum
1995                  * offset, permit "no change" values to help existing apps.
1996                  *
1997                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
1998                  * for an ICMPv6 socket will fail."
1999                  * The current behavior does not meet RFC3542.
2000                  */
2001                 switch (op) {
2002                 case SOPT_SET:
2003                         if (optlen != sizeof(int)) {
2004                                 error = EINVAL;
2005                                 break;
2006                         }
2007                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2008                                             sizeof(optval));
2009                         if (error)
2010                                 break;
2011                         if ((optval % 2) != 0) {
2012                                 /* the API assumes even offset values */
2013                                 error = EINVAL;
2014                         } else if (so->so_proto->pr_protocol ==
2015                             IPPROTO_ICMPV6) {
2016                                 if (optval != icmp6off)
2017                                         error = EINVAL;
2018                         } else
2019                                 in6p->in6p_cksum = optval;
2020                         break;
2021
2022                 case SOPT_GET:
2023                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2024                                 optval = icmp6off;
2025                         else
2026                                 optval = in6p->in6p_cksum;
2027
2028                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2029                         break;
2030
2031                 default:
2032                         error = EINVAL;
2033                         break;
2034                 }
2035                 break;
2036
2037         default:
2038                 error = ENOPROTOOPT;
2039                 break;
2040         }
2041
2042         return (error);
2043 }
2044
2045 /*
2046  * Set up IP6 options in pcb for insertion in output packets or
2047  * specifying behavior of outgoing packets.
2048  */
2049 static int
2050 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2051     struct socket *so, struct sockopt *sopt)
2052 {
2053         struct ip6_pktopts *opt = *pktopt;
2054         int error = 0;
2055         struct thread *td = sopt->sopt_td;
2056
2057         /* turn off any old options. */
2058         if (opt) {
2059 #ifdef DIAGNOSTIC
2060                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2061                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2062                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2063                         printf("ip6_pcbopts: all specified options are cleared.\n");
2064 #endif
2065                 ip6_clearpktopts(opt, -1);
2066         } else
2067                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2068         *pktopt = NULL;
2069
2070         if (!m || m->m_len == 0) {
2071                 /*
2072                  * Only turning off any previous options, regardless of
2073                  * whether the opt is just created or given.
2074                  */
2075                 free(opt, M_IP6OPT);
2076                 return (0);
2077         }
2078
2079         /*  set options specified by user. */
2080         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2081             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2082                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2083                 free(opt, M_IP6OPT);
2084                 return (error);
2085         }
2086         *pktopt = opt;
2087         return (0);
2088 }
2089
2090 /*
2091  * initialize ip6_pktopts.  beware that there are non-zero default values in
2092  * the struct.
2093  */
2094 void
2095 ip6_initpktopts(struct ip6_pktopts *opt)
2096 {
2097
2098         bzero(opt, sizeof(*opt));
2099         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2100         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2101         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2102         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2103 }
2104
2105 static int
2106 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2107     struct ucred *cred, int uproto)
2108 {
2109         struct ip6_pktopts *opt;
2110
2111         if (*pktopt == NULL) {
2112                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2113                     M_WAITOK);
2114                 ip6_initpktopts(*pktopt);
2115         }
2116         opt = *pktopt;
2117
2118         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2119 }
2120
2121 static int
2122 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2123 {
2124         void *optdata = NULL;
2125         int optdatalen = 0;
2126         struct ip6_ext *ip6e;
2127         int error = 0;
2128         struct in6_pktinfo null_pktinfo;
2129         int deftclass = 0, on;
2130         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2131         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2132
2133         switch (optname) {
2134         case IPV6_PKTINFO:
2135                 if (pktopt && pktopt->ip6po_pktinfo)
2136                         optdata = (void *)pktopt->ip6po_pktinfo;
2137                 else {
2138                         /* XXX: we don't have to do this every time... */
2139                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2140                         optdata = (void *)&null_pktinfo;
2141                 }
2142                 optdatalen = sizeof(struct in6_pktinfo);
2143                 break;
2144         case IPV6_TCLASS:
2145                 if (pktopt && pktopt->ip6po_tclass >= 0)
2146                         optdata = (void *)&pktopt->ip6po_tclass;
2147                 else
2148                         optdata = (void *)&deftclass;
2149                 optdatalen = sizeof(int);
2150                 break;
2151         case IPV6_HOPOPTS:
2152                 if (pktopt && pktopt->ip6po_hbh) {
2153                         optdata = (void *)pktopt->ip6po_hbh;
2154                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2155                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2156                 }
2157                 break;
2158         case IPV6_RTHDR:
2159                 if (pktopt && pktopt->ip6po_rthdr) {
2160                         optdata = (void *)pktopt->ip6po_rthdr;
2161                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2162                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2163                 }
2164                 break;
2165         case IPV6_RTHDRDSTOPTS:
2166                 if (pktopt && pktopt->ip6po_dest1) {
2167                         optdata = (void *)pktopt->ip6po_dest1;
2168                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2169                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2170                 }
2171                 break;
2172         case IPV6_DSTOPTS:
2173                 if (pktopt && pktopt->ip6po_dest2) {
2174                         optdata = (void *)pktopt->ip6po_dest2;
2175                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2176                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2177                 }
2178                 break;
2179         case IPV6_NEXTHOP:
2180                 if (pktopt && pktopt->ip6po_nexthop) {
2181                         optdata = (void *)pktopt->ip6po_nexthop;
2182                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2183                 }
2184                 break;
2185         case IPV6_USE_MIN_MTU:
2186                 if (pktopt)
2187                         optdata = (void *)&pktopt->ip6po_minmtu;
2188                 else
2189                         optdata = (void *)&defminmtu;
2190                 optdatalen = sizeof(int);
2191                 break;
2192         case IPV6_DONTFRAG:
2193                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2194                         on = 1;
2195                 else
2196                         on = 0;
2197                 optdata = (void *)&on;
2198                 optdatalen = sizeof(on);
2199                 break;
2200         case IPV6_PREFER_TEMPADDR:
2201                 if (pktopt)
2202                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2203                 else
2204                         optdata = (void *)&defpreftemp;
2205                 optdatalen = sizeof(int);
2206                 break;
2207         default:                /* should not happen */
2208 #ifdef DIAGNOSTIC
2209                 panic("ip6_getpcbopt: unexpected option\n");
2210 #endif
2211                 return (ENOPROTOOPT);
2212         }
2213
2214         error = sooptcopyout(sopt, optdata, optdatalen);
2215
2216         return (error);
2217 }
2218
2219 void
2220 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2221 {
2222         if (pktopt == NULL)
2223                 return;
2224
2225         if (optname == -1 || optname == IPV6_PKTINFO) {
2226                 if (pktopt->ip6po_pktinfo)
2227                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2228                 pktopt->ip6po_pktinfo = NULL;
2229         }
2230         if (optname == -1 || optname == IPV6_HOPLIMIT)
2231                 pktopt->ip6po_hlim = -1;
2232         if (optname == -1 || optname == IPV6_TCLASS)
2233                 pktopt->ip6po_tclass = -1;
2234         if (optname == -1 || optname == IPV6_NEXTHOP) {
2235                 if (pktopt->ip6po_nextroute.ro_rt) {
2236                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2237                         pktopt->ip6po_nextroute.ro_rt = NULL;
2238                 }
2239                 if (pktopt->ip6po_nexthop)
2240                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2241                 pktopt->ip6po_nexthop = NULL;
2242         }
2243         if (optname == -1 || optname == IPV6_HOPOPTS) {
2244                 if (pktopt->ip6po_hbh)
2245                         free(pktopt->ip6po_hbh, M_IP6OPT);
2246                 pktopt->ip6po_hbh = NULL;
2247         }
2248         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2249                 if (pktopt->ip6po_dest1)
2250                         free(pktopt->ip6po_dest1, M_IP6OPT);
2251                 pktopt->ip6po_dest1 = NULL;
2252         }
2253         if (optname == -1 || optname == IPV6_RTHDR) {
2254                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2255                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2256                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2257                 if (pktopt->ip6po_route.ro_rt) {
2258                         RTFREE(pktopt->ip6po_route.ro_rt);
2259                         pktopt->ip6po_route.ro_rt = NULL;
2260                 }
2261         }
2262         if (optname == -1 || optname == IPV6_DSTOPTS) {
2263                 if (pktopt->ip6po_dest2)
2264                         free(pktopt->ip6po_dest2, M_IP6OPT);
2265                 pktopt->ip6po_dest2 = NULL;
2266         }
2267 }
2268
2269 #define PKTOPT_EXTHDRCPY(type) \
2270 do {\
2271         if (src->type) {\
2272                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2273                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2274                 if (dst->type == NULL && canwait == M_NOWAIT)\
2275                         goto bad;\
2276                 bcopy(src->type, dst->type, hlen);\
2277         }\
2278 } while (/*CONSTCOND*/ 0)
2279
2280 static int
2281 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2282 {
2283         if (dst == NULL || src == NULL)  {
2284                 printf("ip6_clearpktopts: invalid argument\n");
2285                 return (EINVAL);
2286         }
2287
2288         dst->ip6po_hlim = src->ip6po_hlim;
2289         dst->ip6po_tclass = src->ip6po_tclass;
2290         dst->ip6po_flags = src->ip6po_flags;
2291         dst->ip6po_minmtu = src->ip6po_minmtu;
2292         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2293         if (src->ip6po_pktinfo) {
2294                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2295                     M_IP6OPT, canwait);
2296                 if (dst->ip6po_pktinfo == NULL)
2297                         goto bad;
2298                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2299         }
2300         if (src->ip6po_nexthop) {
2301                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2302                     M_IP6OPT, canwait);
2303                 if (dst->ip6po_nexthop == NULL)
2304                         goto bad;
2305                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2306                     src->ip6po_nexthop->sa_len);
2307         }
2308         PKTOPT_EXTHDRCPY(ip6po_hbh);
2309         PKTOPT_EXTHDRCPY(ip6po_dest1);
2310         PKTOPT_EXTHDRCPY(ip6po_dest2);
2311         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2312         return (0);
2313
2314   bad:
2315         ip6_clearpktopts(dst, -1);
2316         return (ENOBUFS);
2317 }
2318 #undef PKTOPT_EXTHDRCPY
2319
2320 struct ip6_pktopts *
2321 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2322 {
2323         int error;
2324         struct ip6_pktopts *dst;
2325
2326         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2327         if (dst == NULL)
2328                 return (NULL);
2329         ip6_initpktopts(dst);
2330
2331         if ((error = copypktopts(dst, src, canwait)) != 0) {
2332                 free(dst, M_IP6OPT);
2333                 return (NULL);
2334         }
2335
2336         return (dst);
2337 }
2338
2339 void
2340 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2341 {
2342         if (pktopt == NULL)
2343                 return;
2344
2345         ip6_clearpktopts(pktopt, -1);
2346
2347         free(pktopt, M_IP6OPT);
2348 }
2349
2350 /*
2351  * Set IPv6 outgoing packet options based on advanced API.
2352  */
2353 int
2354 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2355     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2356 {
2357         struct cmsghdr *cm = 0;
2358
2359         if (control == NULL || opt == NULL)
2360                 return (EINVAL);
2361
2362         ip6_initpktopts(opt);
2363         if (stickyopt) {
2364                 int error;
2365
2366                 /*
2367                  * If stickyopt is provided, make a local copy of the options
2368                  * for this particular packet, then override them by ancillary
2369                  * objects.
2370                  * XXX: copypktopts() does not copy the cached route to a next
2371                  * hop (if any).  This is not very good in terms of efficiency,
2372                  * but we can allow this since this option should be rarely
2373                  * used.
2374                  */
2375                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2376                         return (error);
2377         }
2378
2379         /*
2380          * XXX: Currently, we assume all the optional information is stored
2381          * in a single mbuf.
2382          */
2383         if (control->m_next)
2384                 return (EINVAL);
2385
2386         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2387             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2388                 int error;
2389
2390                 if (control->m_len < CMSG_LEN(0))
2391                         return (EINVAL);
2392
2393                 cm = mtod(control, struct cmsghdr *);
2394                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2395                         return (EINVAL);
2396                 if (cm->cmsg_level != IPPROTO_IPV6)
2397                         continue;
2398
2399                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2400                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2401                 if (error)
2402                         return (error);
2403         }
2404
2405         return (0);
2406 }
2407
2408 /*
2409  * Set a particular packet option, as a sticky option or an ancillary data
2410  * item.  "len" can be 0 only when it's a sticky option.
2411  * We have 4 cases of combination of "sticky" and "cmsg":
2412  * "sticky=0, cmsg=0": impossible
2413  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2414  * "sticky=1, cmsg=0": RFC3542 socket option
2415  * "sticky=1, cmsg=1": RFC2292 socket option
2416  */
2417 static int
2418 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2419     struct ucred *cred, int sticky, int cmsg, int uproto)
2420 {
2421         int minmtupolicy, preftemp;
2422         int error;
2423
2424         if (!sticky && !cmsg) {
2425 #ifdef DIAGNOSTIC
2426                 printf("ip6_setpktopt: impossible case\n");
2427 #endif
2428                 return (EINVAL);
2429         }
2430
2431         /*
2432          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2433          * not be specified in the context of RFC3542.  Conversely,
2434          * RFC3542 types should not be specified in the context of RFC2292.
2435          */
2436         if (!cmsg) {
2437                 switch (optname) {
2438                 case IPV6_2292PKTINFO:
2439                 case IPV6_2292HOPLIMIT:
2440                 case IPV6_2292NEXTHOP:
2441                 case IPV6_2292HOPOPTS:
2442                 case IPV6_2292DSTOPTS:
2443                 case IPV6_2292RTHDR:
2444                 case IPV6_2292PKTOPTIONS:
2445                         return (ENOPROTOOPT);
2446                 }
2447         }
2448         if (sticky && cmsg) {
2449                 switch (optname) {
2450                 case IPV6_PKTINFO:
2451                 case IPV6_HOPLIMIT:
2452                 case IPV6_NEXTHOP:
2453                 case IPV6_HOPOPTS:
2454                 case IPV6_DSTOPTS:
2455                 case IPV6_RTHDRDSTOPTS:
2456                 case IPV6_RTHDR:
2457                 case IPV6_USE_MIN_MTU:
2458                 case IPV6_DONTFRAG:
2459                 case IPV6_TCLASS:
2460                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2461                         return (ENOPROTOOPT);
2462                 }
2463         }
2464
2465         switch (optname) {
2466         case IPV6_2292PKTINFO:
2467         case IPV6_PKTINFO:
2468         {
2469                 struct ifnet *ifp = NULL;
2470                 struct in6_pktinfo *pktinfo;
2471
2472                 if (len != sizeof(struct in6_pktinfo))
2473                         return (EINVAL);
2474
2475                 pktinfo = (struct in6_pktinfo *)buf;
2476
2477                 /*
2478                  * An application can clear any sticky IPV6_PKTINFO option by
2479                  * doing a "regular" setsockopt with ipi6_addr being
2480                  * in6addr_any and ipi6_ifindex being zero.
2481                  * [RFC 3542, Section 6]
2482                  */
2483                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2484                     pktinfo->ipi6_ifindex == 0 &&
2485                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2486                         ip6_clearpktopts(opt, optname);
2487                         break;
2488                 }
2489
2490                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2491                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2492                         return (EINVAL);
2493                 }
2494
2495                 /* validate the interface index if specified. */
2496                 if (pktinfo->ipi6_ifindex > V_if_index ||
2497                     pktinfo->ipi6_ifindex < 0) {
2498                          return (ENXIO);
2499                 }
2500                 if (pktinfo->ipi6_ifindex) {
2501                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2502                         if (ifp == NULL)
2503                                 return (ENXIO);
2504                 }
2505
2506                 /*
2507                  * We store the address anyway, and let in6_selectsrc()
2508                  * validate the specified address.  This is because ipi6_addr
2509                  * may not have enough information about its scope zone, and
2510                  * we may need additional information (such as outgoing
2511                  * interface or the scope zone of a destination address) to
2512                  * disambiguate the scope.
2513                  * XXX: the delay of the validation may confuse the
2514                  * application when it is used as a sticky option.
2515                  */
2516                 if (opt->ip6po_pktinfo == NULL) {
2517                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2518                             M_IP6OPT, M_NOWAIT);
2519                         if (opt->ip6po_pktinfo == NULL)
2520                                 return (ENOBUFS);
2521                 }
2522                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2523                 break;
2524         }
2525
2526         case IPV6_2292HOPLIMIT:
2527         case IPV6_HOPLIMIT:
2528         {
2529                 int *hlimp;
2530
2531                 /*
2532                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2533                  * to simplify the ordering among hoplimit options.
2534                  */
2535                 if (optname == IPV6_HOPLIMIT && sticky)
2536                         return (ENOPROTOOPT);
2537
2538                 if (len != sizeof(int))
2539                         return (EINVAL);
2540                 hlimp = (int *)buf;
2541                 if (*hlimp < -1 || *hlimp > 255)
2542                         return (EINVAL);
2543
2544                 opt->ip6po_hlim = *hlimp;
2545                 break;
2546         }
2547
2548         case IPV6_TCLASS:
2549         {
2550                 int tclass;
2551
2552                 if (len != sizeof(int))
2553                         return (EINVAL);
2554                 tclass = *(int *)buf;
2555                 if (tclass < -1 || tclass > 255)
2556                         return (EINVAL);
2557
2558                 opt->ip6po_tclass = tclass;
2559                 break;
2560         }
2561
2562         case IPV6_2292NEXTHOP:
2563         case IPV6_NEXTHOP:
2564                 if (cred != NULL) {
2565                         error = priv_check_cred(cred,
2566                             PRIV_NETINET_SETHDROPTS, 0);
2567                         if (error)
2568                                 return (error);
2569                 }
2570
2571                 if (len == 0) { /* just remove the option */
2572                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2573                         break;
2574                 }
2575
2576                 /* check if cmsg_len is large enough for sa_len */
2577                 if (len < sizeof(struct sockaddr) || len < *buf)
2578                         return (EINVAL);
2579
2580                 switch (((struct sockaddr *)buf)->sa_family) {
2581                 case AF_INET6:
2582                 {
2583                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2584                         int error;
2585
2586                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2587                                 return (EINVAL);
2588
2589                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2590                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2591                                 return (EINVAL);
2592                         }
2593                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2594                             != 0) {
2595                                 return (error);
2596                         }
2597                         break;
2598                 }
2599                 case AF_LINK:   /* should eventually be supported */
2600                 default:
2601                         return (EAFNOSUPPORT);
2602                 }
2603
2604                 /* turn off the previous option, then set the new option. */
2605                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2606                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2607                 if (opt->ip6po_nexthop == NULL)
2608                         return (ENOBUFS);
2609                 bcopy(buf, opt->ip6po_nexthop, *buf);
2610                 break;
2611
2612         case IPV6_2292HOPOPTS:
2613         case IPV6_HOPOPTS:
2614         {
2615                 struct ip6_hbh *hbh;
2616                 int hbhlen;
2617
2618                 /*
2619                  * XXX: We don't allow a non-privileged user to set ANY HbH
2620                  * options, since per-option restriction has too much
2621                  * overhead.
2622                  */
2623                 if (cred != NULL) {
2624                         error = priv_check_cred(cred,
2625                             PRIV_NETINET_SETHDROPTS, 0);
2626                         if (error)
2627                                 return (error);
2628                 }
2629
2630                 if (len == 0) {
2631                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2632                         break;  /* just remove the option */
2633                 }
2634
2635                 /* message length validation */
2636                 if (len < sizeof(struct ip6_hbh))
2637                         return (EINVAL);
2638                 hbh = (struct ip6_hbh *)buf;
2639                 hbhlen = (hbh->ip6h_len + 1) << 3;
2640                 if (len != hbhlen)
2641                         return (EINVAL);
2642
2643                 /* turn off the previous option, then set the new option. */
2644                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2645                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2646                 if (opt->ip6po_hbh == NULL)
2647                         return (ENOBUFS);
2648                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2649
2650                 break;
2651         }
2652
2653         case IPV6_2292DSTOPTS:
2654         case IPV6_DSTOPTS:
2655         case IPV6_RTHDRDSTOPTS:
2656         {
2657                 struct ip6_dest *dest, **newdest = NULL;
2658                 int destlen;
2659
2660                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2661                         error = priv_check_cred(cred,
2662                             PRIV_NETINET_SETHDROPTS, 0);
2663                         if (error)
2664                                 return (error);
2665                 }
2666
2667                 if (len == 0) {
2668                         ip6_clearpktopts(opt, optname);
2669                         break;  /* just remove the option */
2670                 }
2671
2672                 /* message length validation */
2673                 if (len < sizeof(struct ip6_dest))
2674                         return (EINVAL);
2675                 dest = (struct ip6_dest *)buf;
2676                 destlen = (dest->ip6d_len + 1) << 3;
2677                 if (len != destlen)
2678                         return (EINVAL);
2679
2680                 /*
2681                  * Determine the position that the destination options header
2682                  * should be inserted; before or after the routing header.
2683                  */
2684                 switch (optname) {
2685                 case IPV6_2292DSTOPTS:
2686                         /*
2687                          * The old advacned API is ambiguous on this point.
2688                          * Our approach is to determine the position based
2689                          * according to the existence of a routing header.
2690                          * Note, however, that this depends on the order of the
2691                          * extension headers in the ancillary data; the 1st
2692                          * part of the destination options header must appear
2693                          * before the routing header in the ancillary data,
2694                          * too.
2695                          * RFC3542 solved the ambiguity by introducing
2696                          * separate ancillary data or option types.
2697                          */
2698                         if (opt->ip6po_rthdr == NULL)
2699                                 newdest = &opt->ip6po_dest1;
2700                         else
2701                                 newdest = &opt->ip6po_dest2;
2702                         break;
2703                 case IPV6_RTHDRDSTOPTS:
2704                         newdest = &opt->ip6po_dest1;
2705                         break;
2706                 case IPV6_DSTOPTS:
2707                         newdest = &opt->ip6po_dest2;
2708                         break;
2709                 }
2710
2711                 /* turn off the previous option, then set the new option. */
2712                 ip6_clearpktopts(opt, optname);
2713                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2714                 if (*newdest == NULL)
2715                         return (ENOBUFS);
2716                 bcopy(dest, *newdest, destlen);
2717
2718                 break;
2719         }
2720
2721         case IPV6_2292RTHDR:
2722         case IPV6_RTHDR:
2723         {
2724                 struct ip6_rthdr *rth;
2725                 int rthlen;
2726
2727                 if (len == 0) {
2728                         ip6_clearpktopts(opt, IPV6_RTHDR);
2729                         break;  /* just remove the option */
2730                 }
2731
2732                 /* message length validation */
2733                 if (len < sizeof(struct ip6_rthdr))
2734                         return (EINVAL);
2735                 rth = (struct ip6_rthdr *)buf;
2736                 rthlen = (rth->ip6r_len + 1) << 3;
2737                 if (len != rthlen)
2738                         return (EINVAL);
2739
2740                 switch (rth->ip6r_type) {
2741                 case IPV6_RTHDR_TYPE_0:
2742                         if (rth->ip6r_len == 0) /* must contain one addr */
2743                                 return (EINVAL);
2744                         if (rth->ip6r_len % 2) /* length must be even */
2745                                 return (EINVAL);
2746                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2747                                 return (EINVAL);
2748                         break;
2749                 default:
2750                         return (EINVAL);        /* not supported */
2751                 }
2752
2753                 /* turn off the previous option */
2754                 ip6_clearpktopts(opt, IPV6_RTHDR);
2755                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2756                 if (opt->ip6po_rthdr == NULL)
2757                         return (ENOBUFS);
2758                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2759
2760                 break;
2761         }
2762
2763         case IPV6_USE_MIN_MTU:
2764                 if (len != sizeof(int))
2765                         return (EINVAL);
2766                 minmtupolicy = *(int *)buf;
2767                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2768                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2769                     minmtupolicy != IP6PO_MINMTU_ALL) {
2770                         return (EINVAL);
2771                 }
2772                 opt->ip6po_minmtu = minmtupolicy;
2773                 break;
2774
2775         case IPV6_DONTFRAG:
2776                 if (len != sizeof(int))
2777                         return (EINVAL);
2778
2779                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2780                         /*
2781                          * we ignore this option for TCP sockets.
2782                          * (RFC3542 leaves this case unspecified.)
2783                          */
2784                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2785                 } else
2786                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2787                 break;
2788
2789         case IPV6_PREFER_TEMPADDR:
2790                 if (len != sizeof(int))
2791                         return (EINVAL);
2792                 preftemp = *(int *)buf;
2793                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2794                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2795                     preftemp != IP6PO_TEMPADDR_PREFER) {
2796                         return (EINVAL);
2797                 }
2798                 opt->ip6po_prefer_tempaddr = preftemp;
2799                 break;
2800
2801         default:
2802                 return (ENOPROTOOPT);
2803         } /* end of switch */
2804
2805         return (0);
2806 }
2807
2808 /*
2809  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2810  * packet to the input queue of a specified interface.  Note that this
2811  * calls the output routine of the loopback "driver", but with an interface
2812  * pointer that might NOT be &loif -- easier than replicating that code here.
2813  */
2814 void
2815 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2816 {
2817         struct mbuf *copym;
2818         struct ip6_hdr *ip6;
2819
2820         copym = m_copy(m, 0, M_COPYALL);
2821         if (copym == NULL)
2822                 return;
2823
2824         /*
2825          * Make sure to deep-copy IPv6 header portion in case the data
2826          * is in an mbuf cluster, so that we can safely override the IPv6
2827          * header portion later.
2828          */
2829         if ((copym->m_flags & M_EXT) != 0 ||
2830             copym->m_len < sizeof(struct ip6_hdr)) {
2831                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2832                 if (copym == NULL)
2833                         return;
2834         }
2835         ip6 = mtod(copym, struct ip6_hdr *);
2836         /*
2837          * clear embedded scope identifiers if necessary.
2838          * in6_clearscope will touch the addresses only when necessary.
2839          */
2840         in6_clearscope(&ip6->ip6_src);
2841         in6_clearscope(&ip6->ip6_dst);
2842         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
2843                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
2844                     CSUM_PSEUDO_HDR;
2845                 copym->m_pkthdr.csum_data = 0xffff;
2846         }
2847         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
2848 }
2849
2850 /*
2851  * Chop IPv6 header off from the payload.
2852  */
2853 static int
2854 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2855 {
2856         struct mbuf *mh;
2857         struct ip6_hdr *ip6;
2858
2859         ip6 = mtod(m, struct ip6_hdr *);
2860         if (m->m_len > sizeof(*ip6)) {
2861                 mh = m_gethdr(M_NOWAIT, MT_DATA);
2862                 if (mh == NULL) {
2863                         m_freem(m);
2864                         return ENOBUFS;
2865                 }
2866                 m_move_pkthdr(mh, m);
2867                 MH_ALIGN(mh, sizeof(*ip6));
2868                 m->m_len -= sizeof(*ip6);
2869                 m->m_data += sizeof(*ip6);
2870                 mh->m_next = m;
2871                 m = mh;
2872                 m->m_len = sizeof(*ip6);
2873                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2874         }
2875         exthdrs->ip6e_ip6 = m;
2876         return 0;
2877 }
2878
2879 /*
2880  * Compute IPv6 extension header length.
2881  */
2882 int
2883 ip6_optlen(struct inpcb *in6p)
2884 {
2885         int len;
2886
2887         if (!in6p->in6p_outputopts)
2888                 return 0;
2889
2890         len = 0;
2891 #define elen(x) \
2892     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2893
2894         len += elen(in6p->in6p_outputopts->ip6po_hbh);
2895         if (in6p->in6p_outputopts->ip6po_rthdr)
2896                 /* dest1 is valid with rthdr only */
2897                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
2898         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2899         len += elen(in6p->in6p_outputopts->ip6po_dest2);
2900         return len;
2901 #undef elen
2902 }