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         KASSERT(( mtu % 8 == 0), ("Fragment length must be a multiple of 8"));
 223
 224         m = m0;
 225         ip6 = mtod(m, struct ip6_hdr *);
 226         mnext = &m->m_nextpkt;
 227
 228         for (off = hlen; off < tlen; off += mtu) {
 229                 m = m_gethdr(M_NOWAIT, MT_DATA);
 230                 if (!m) {
 231                         IP6STAT_INC(ip6s_odropped);
 232                         return (ENOBUFS);
 233                 }
 234                 m->m_flags = m0->m_flags & M_COPYFLAGS;
 235                 *mnext = m;
 236                 mnext = &m->m_nextpkt;
 237                 m->m_data += max_linkhdr;
 238                 mhip6 = mtod(m, struct ip6_hdr *);
 239                 *mhip6 = *ip6;
 240                 m->m_len = sizeof(*mhip6);
 241                 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 242                 if (error) {
 243                         IP6STAT_INC(ip6s_odropped);
 244                         return (error);
 245                 }
 246                 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 247                 if (off + mtu >= tlen)
 248                         mtu = tlen - off;
 249                 else
 250                         ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 251                 mhip6->ip6_plen = htons((u_short)(mtu + hlen +
 252                     sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 253                 if ((m_frgpart = m_copy(m0, off, mtu)) == 0) {
 254                         IP6STAT_INC(ip6s_odropped);
 255                         return (ENOBUFS);
 256                 }
 257                 m_cat(m, m_frgpart);
 258                 m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
 259                 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
 260                 m->m_pkthdr.rcvif = NULL;
 261                 ip6f->ip6f_reserved = 0;
 262                 ip6f->ip6f_ident = id;
 263                 ip6f->ip6f_nxt = nextproto;
 264                 IP6STAT_INC(ip6s_ofragments);
 265                 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 266         }
 267
 268         return (0);
 269 }
 270
 271 /*
 272  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 273  * header (with pri, len, nxt, hlim, src, dst).
 274  * This function may modify ver and hlim only.
 275  * The mbuf chain containing the packet will be freed.
 276  * The mbuf opt, if present, will not be freed.
 277  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
 278  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 279  * then result of route lookup is stored in ro->ro_rt.
 280  *
 281  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
 282  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 283  * which is rt_mtu.
 284  *
 285  * ifpp - XXX: just for statistics
 286  */
 287 int
 288 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 289     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 290     struct ifnet **ifpp, struct inpcb *inp)
 291 {
 292         struct ip6_hdr *ip6;
 293         struct ifnet *ifp, *origifp;
 294         struct mbuf *m = m0;
 295         struct mbuf *mprev = NULL;
 296         int hlen, tlen, len;
 297         struct route_in6 ip6route;
 298         struct rtentry *rt = NULL;
 299         struct sockaddr_in6 *dst, src_sa, dst_sa;
 300         struct in6_addr odst;
 301         int error = 0;
 302         struct in6_ifaddr *ia = NULL;
 303         u_long mtu;
 304         int alwaysfrag, dontfrag;
 305         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 306         struct ip6_exthdrs exthdrs;
 307         struct in6_addr finaldst, src0, dst0;
 308         u_int32_t zone;
 309         struct route_in6 *ro_pmtu = NULL;
 310         int hdrsplit = 0;
 311         int sw_csum, tso;
 312         struct m_tag *fwd_tag = NULL;
 313         uint32_t id;
 314
 315         ip6 = mtod(m, struct ip6_hdr *);
 316         if (ip6 == NULL) {
 317                 printf ("ip6 is NULL");
 318                 goto bad;
 319         }
 320
 321         if (inp != NULL)
 322                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 323
 324         finaldst = ip6->ip6_dst;
 325         bzero(&exthdrs, sizeof(exthdrs));
 326         if (opt) {
 327                 /* Hop-by-Hop options header */
 328                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 329                 /* Destination options header(1st part) */
 330                 if (opt->ip6po_rthdr) {
 331                         /*
 332                          * Destination options header(1st part)
 333                          * This only makes sense with a routing header.
 334                          * See Section 9.2 of RFC 3542.
 335                          * Disabling this part just for MIP6 convenience is
 336                          * a bad idea.  We need to think carefully about a
 337                          * way to make the advanced API coexist with MIP6
 338                          * options, which might automatically be inserted in
 339                          * the kernel.
 340                          */
 341                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 342                 }
 343                 /* Routing header */
 344                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 345                 /* Destination options header(2nd part) */
 346                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 347         }
 348
 349 #ifdef IPSEC
 350         /*
 351          * IPSec checking which handles several cases.
 352          * FAST IPSEC: We re-injected the packet.
 353          */
 354         switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp))
 355         {
 356         case 1:                 /* Bad packet */
 357                 goto freehdrs;
 358         case -1:                /* IPSec done */
 359                 goto done;
 360         case 0:                 /* No IPSec */
 361         default:
 362                 break;
 363         }
 364 #endif /* IPSEC */
 365
 366         /*
 367          * Calculate the total length of the extension header chain.
 368          * Keep the length of the unfragmentable part for fragmentation.
 369          */
 370         optlen = 0;
 371         if (exthdrs.ip6e_hbh)
 372                 optlen += exthdrs.ip6e_hbh->m_len;
 373         if (exthdrs.ip6e_dest1)
 374                 optlen += exthdrs.ip6e_dest1->m_len;
 375         if (exthdrs.ip6e_rthdr)
 376                 optlen += exthdrs.ip6e_rthdr->m_len;
 377         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 378
 379         /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
 380         if (exthdrs.ip6e_dest2)
 381                 optlen += exthdrs.ip6e_dest2->m_len;
 382
 383         /*
 384          * If there is at least one extension header,
 385          * separate IP6 header from the payload.
 386          */
 387         if (optlen && !hdrsplit) {
 388                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 389                         m = NULL;
 390                         goto freehdrs;
 391                 }
 392                 m = exthdrs.ip6e_ip6;
 393                 hdrsplit++;
 394         }
 395
 396         /* adjust pointer */
 397         ip6 = mtod(m, struct ip6_hdr *);
 398
 399         /* adjust mbuf packet header length */
 400         m->m_pkthdr.len += optlen;
 401         plen = m->m_pkthdr.len - sizeof(*ip6);
 402
 403         /* If this is a jumbo payload, insert a jumbo payload option. */
 404         if (plen > IPV6_MAXPACKET) {
 405                 if (!hdrsplit) {
 406                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 407                                 m = NULL;
 408                                 goto freehdrs;
 409                         }
 410                         m = exthdrs.ip6e_ip6;
 411                         hdrsplit++;
 412                 }
 413                 /* adjust pointer */
 414                 ip6 = mtod(m, struct ip6_hdr *);
 415                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 416                         goto freehdrs;
 417                 ip6->ip6_plen = 0;
 418         } else
 419                 ip6->ip6_plen = htons(plen);
 420
 421         /*
 422          * Concatenate headers and fill in next header fields.
 423          * Here we have, on "m"
 424          *      IPv6 payload
 425          * and we insert headers accordingly.  Finally, we should be getting:
 426          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 427          *
 428          * during the header composing process, "m" points to IPv6 header.
 429          * "mprev" points to an extension header prior to esp.
 430          */
 431         u_char *nexthdrp = &ip6->ip6_nxt;
 432         mprev = m;
 433
 434         /*
 435          * we treat dest2 specially.  this makes IPsec processing
 436          * much easier.  the goal here is to make mprev point the
 437          * mbuf prior to dest2.
 438          *
 439          * result: IPv6 dest2 payload
 440          * m and mprev will point to IPv6 header.
 441          */
 442         if (exthdrs.ip6e_dest2) {
 443                 if (!hdrsplit)
 444                         panic("assumption failed: hdr not split");
 445                 exthdrs.ip6e_dest2->m_next = m->m_next;
 446                 m->m_next = exthdrs.ip6e_dest2;
 447                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 448                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 449         }
 450
 451         /*
 452          * result: IPv6 hbh dest1 rthdr dest2 payload
 453          * m will point to IPv6 header.  mprev will point to the
 454          * extension header prior to dest2 (rthdr in the above case).
 455          */
 456         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 457         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 458                    IPPROTO_DSTOPTS);
 459         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 460                    IPPROTO_ROUTING);
 461
 462         /*
 463          * If there is a routing header, discard the packet.
 464          */
 465         if (exthdrs.ip6e_rthdr) {
 466                  error = EINVAL;
 467                  goto bad;
 468         }
 469
 470         /* Source address validation */
 471         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 472             (flags & IPV6_UNSPECSRC) == 0) {
 473                 error = EOPNOTSUPP;
 474                 IP6STAT_INC(ip6s_badscope);
 475                 goto bad;
 476         }
 477         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 478                 error = EOPNOTSUPP;
 479                 IP6STAT_INC(ip6s_badscope);
 480                 goto bad;
 481         }
 482
 483         IP6STAT_INC(ip6s_localout);
 484
 485         /*
 486          * Route packet.
 487          */
 488         if (ro == 0) {
 489                 ro = &ip6route;
 490                 bzero((caddr_t)ro, sizeof(*ro));
 491         }
 492         ro_pmtu = ro;
 493         if (opt && opt->ip6po_rthdr)
 494                 ro = &opt->ip6po_route;
 495         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 496 #ifdef FLOWTABLE
 497         if (ro->ro_rt == NULL)
 498                 (void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
 499 #endif
 500 again:
 501         /*
 502          * if specified, try to fill in the traffic class field.
 503          * do not override if a non-zero value is already set.
 504          * we check the diffserv field and the ecn field separately.
 505          */
 506         if (opt && opt->ip6po_tclass >= 0) {
 507                 int mask = 0;
 508
 509                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 510                         mask |= 0xfc;
 511                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 512                         mask |= 0x03;
 513                 if (mask != 0)
 514                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 515         }
 516
 517         /* fill in or override the hop limit field, if necessary. */
 518         if (opt && opt->ip6po_hlim != -1)
 519                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 520         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 521                 if (im6o != NULL)
 522                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 523                 else
 524                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 525         }
 526
 527         /* adjust pointer */
 528         ip6 = mtod(m, struct ip6_hdr *);
 529
 530         if (ro->ro_rt && fwd_tag == NULL) {
 531                 rt = ro->ro_rt;
 532                 ifp = ro->ro_rt->rt_ifp;
 533         } else {
 534                 if (fwd_tag == NULL) {
 535                         bzero(&dst_sa, sizeof(dst_sa));
 536                         dst_sa.sin6_family = AF_INET6;
 537                         dst_sa.sin6_len = sizeof(dst_sa);
 538                         dst_sa.sin6_addr = ip6->ip6_dst;
 539                 }
 540                 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
 541                     &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
 542                 if (error != 0) {
 543                         if (ifp != NULL)
 544                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 545                         goto bad;
 546                 }
 547         }
 548         if (rt == NULL) {
 549                 /*
 550                  * If in6_selectroute() does not return a route entry,
 551                  * dst may not have been updated.
 552                  */
 553                 *dst = dst_sa;  /* XXX */
 554         }
 555
 556         /*
 557          * then rt (for unicast) and ifp must be non-NULL valid values.
 558          */
 559         if ((flags & IPV6_FORWARDING) == 0) {
 560                 /* XXX: the FORWARDING flag can be set for mrouting. */
 561                 in6_ifstat_inc(ifp, ifs6_out_request);
 562         }
 563         if (rt != NULL) {
 564                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 565                 counter_u64_add(rt->rt_pksent, 1);
 566         }
 567
 568
 569         /*
 570          * The outgoing interface must be in the zone of source and
 571          * destination addresses.
 572          */
 573         origifp = ifp;
 574
 575         src0 = ip6->ip6_src;
 576         if (in6_setscope(&src0, origifp, &zone))
 577                 goto badscope;
 578         bzero(&src_sa, sizeof(src_sa));
 579         src_sa.sin6_family = AF_INET6;
 580         src_sa.sin6_len = sizeof(src_sa);
 581         src_sa.sin6_addr = ip6->ip6_src;
 582         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 583                 goto badscope;
 584
 585         dst0 = ip6->ip6_dst;
 586         if (in6_setscope(&dst0, origifp, &zone))
 587                 goto badscope;
 588         /* re-initialize to be sure */
 589         bzero(&dst_sa, sizeof(dst_sa));
 590         dst_sa.sin6_family = AF_INET6;
 591         dst_sa.sin6_len = sizeof(dst_sa);
 592         dst_sa.sin6_addr = ip6->ip6_dst;
 593         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 594                 goto badscope;
 595         }
 596
 597         /* We should use ia_ifp to support the case of
 598          * sending packets to an address of our own.
 599          */
 600         if (ia != NULL && ia->ia_ifp)
 601                 ifp = ia->ia_ifp;
 602
 603         /* scope check is done. */
 604         goto routefound;
 605
 606   badscope:
 607         IP6STAT_INC(ip6s_badscope);
 608         in6_ifstat_inc(origifp, ifs6_out_discard);
 609         if (error == 0)
 610                 error = EHOSTUNREACH; /* XXX */
 611         goto bad;
 612
 613   routefound:
 614         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 615                 if (opt && opt->ip6po_nextroute.ro_rt) {
 616                         /*
 617                          * The nexthop is explicitly specified by the
 618                          * application.  We assume the next hop is an IPv6
 619                          * address.
 620                          */
 621                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 622                 }
 623                 else if ((rt->rt_flags & RTF_GATEWAY))
 624                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 625         }
 626
 627         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 628                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 629         } else {
 630                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 631                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 632                 /*
 633                  * Confirm that the outgoing interface supports multicast.
 634                  */
 635                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 636                         IP6STAT_INC(ip6s_noroute);
 637                         in6_ifstat_inc(ifp, ifs6_out_discard);
 638                         error = ENETUNREACH;
 639                         goto bad;
 640                 }
 641                 if ((im6o == NULL && in6_mcast_loop) ||
 642                     (im6o && im6o->im6o_multicast_loop)) {
 643                         /*
 644                          * Loop back multicast datagram if not expressly
 645                          * forbidden to do so, even if we have not joined
 646                          * the address; protocols will filter it later,
 647                          * thus deferring a hash lookup and lock acquisition
 648                          * at the expense of an m_copym().
 649                          */
 650                         ip6_mloopback(ifp, m, dst);
 651                 } else {
 652                         /*
 653                          * If we are acting as a multicast router, perform
 654                          * multicast forwarding as if the packet had just
 655                          * arrived on the interface to which we are about
 656                          * to send.  The multicast forwarding function
 657                          * recursively calls this function, using the
 658                          * IPV6_FORWARDING flag to prevent infinite recursion.
 659                          *
 660                          * Multicasts that are looped back by ip6_mloopback(),
 661                          * above, will be forwarded by the ip6_input() routine,
 662                          * if necessary.
 663                          */
 664                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 665                                 /*
 666                                  * XXX: ip6_mforward expects that rcvif is NULL
 667                                  * when it is called from the originating path.
 668                                  * However, it may not always be the case.
 669                                  */
 670                                 m->m_pkthdr.rcvif = NULL;
 671                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 672                                         m_freem(m);
 673                                         goto done;
 674                                 }
 675                         }
 676                 }
 677                 /*
 678                  * Multicasts with a hoplimit of zero may be looped back,
 679                  * above, but must not be transmitted on a network.
 680                  * Also, multicasts addressed to the loopback interface
 681                  * are not sent -- the above call to ip6_mloopback() will
 682                  * loop back a copy if this host actually belongs to the
 683                  * destination group on the loopback interface.
 684                  */
 685                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 686                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 687                         m_freem(m);
 688                         goto done;
 689                 }
 690         }
 691
 692         /*
 693          * Fill the outgoing inteface to tell the upper layer
 694          * to increment per-interface statistics.
 695          */
 696         if (ifpp)
 697                 *ifpp = ifp;
 698
 699         /* Determine path MTU. */
 700         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 701             &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
 702                 goto bad;
 703
 704         /*
 705          * The caller of this function may specify to use the minimum MTU
 706          * in some cases.
 707          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 708          * setting.  The logic is a bit complicated; by default, unicast
 709          * packets will follow path MTU while multicast packets will be sent at
 710          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 711          * including unicast ones will be sent at the minimum MTU.  Multicast
 712          * packets will always be sent at the minimum MTU unless
 713          * IP6PO_MINMTU_DISABLE is explicitly specified.
 714          * See RFC 3542 for more details.
 715          */
 716         if (mtu > IPV6_MMTU) {
 717                 if ((flags & IPV6_MINMTU))
 718                         mtu = IPV6_MMTU;
 719                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 720                         mtu = IPV6_MMTU;
 721                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 722                          (opt == NULL ||
 723                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 724                         mtu = IPV6_MMTU;
 725                 }
 726         }
 727
 728         /*
 729          * clear embedded scope identifiers if necessary.
 730          * in6_clearscope will touch the addresses only when necessary.
 731          */
 732         in6_clearscope(&ip6->ip6_src);
 733         in6_clearscope(&ip6->ip6_dst);
 734
 735         /*
 736          * If the outgoing packet contains a hop-by-hop options header,
 737          * it must be examined and processed even by the source node.
 738          * (RFC 2460, section 4.)
 739          */
 740         if (exthdrs.ip6e_hbh) {
 741                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 742                 u_int32_t dummy; /* XXX unused */
 743                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 744
 745 #ifdef DIAGNOSTIC
 746                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 747                         panic("ip6e_hbh is not contiguous");
 748 #endif
 749                 /*
 750                  *  XXX: if we have to send an ICMPv6 error to the sender,
 751                  *       we need the M_LOOP flag since icmp6_error() expects
 752                  *       the IPv6 and the hop-by-hop options header are
 753                  *       contiguous unless the flag is set.
 754                  */
 755                 m->m_flags |= M_LOOP;
 756                 m->m_pkthdr.rcvif = ifp;
 757                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 758                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 759                     &dummy, &plen) < 0) {
 760                         /* m was already freed at this point */
 761                         error = EINVAL;/* better error? */
 762                         goto done;
 763                 }
 764                 m->m_flags &= ~M_LOOP; /* XXX */
 765                 m->m_pkthdr.rcvif = NULL;
 766         }
 767
 768         /* Jump over all PFIL processing if hooks are not active. */
 769         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 770                 goto passout;
 771
 772         odst = ip6->ip6_dst;
 773         /* Run through list of hooks for output packets. */
 774         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 775         if (error != 0 || m == NULL)
 776                 goto done;
 777         ip6 = mtod(m, struct ip6_hdr *);
 778
 779         /* See if destination IP address was changed by packet filter. */
 780         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 781                 m->m_flags |= M_SKIP_FIREWALL;
 782                 /* If destination is now ourself drop to ip6_input(). */
 783                 if (in6_localip(&ip6->ip6_dst)) {
 784                         m->m_flags |= M_FASTFWD_OURS;
 785                         if (m->m_pkthdr.rcvif == NULL)
 786                                 m->m_pkthdr.rcvif = V_loif;
 787                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 788                                 m->m_pkthdr.csum_flags |=
 789                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 790                                 m->m_pkthdr.csum_data = 0xffff;
 791                         }
 792 #ifdef SCTP
 793                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 794                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 795 #endif
 796                         error = netisr_queue(NETISR_IPV6, m);
 797                         goto done;
 798                 } else
 799                         goto again;     /* Redo the routing table lookup. */
 800         }
 801
 802         /* See if local, if yes, send it to netisr. */
 803         if (m->m_flags & M_FASTFWD_OURS) {
 804                 if (m->m_pkthdr.rcvif == NULL)
 805                         m->m_pkthdr.rcvif = V_loif;
 806                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 807                         m->m_pkthdr.csum_flags |=
 808                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 809                         m->m_pkthdr.csum_data = 0xffff;
 810                 }
 811 #ifdef SCTP
 812                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 813                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 814 #endif
 815                 error = netisr_queue(NETISR_IPV6, m);
 816                 goto done;
 817         }
 818         /* Or forward to some other address? */
 819         if ((m->m_flags & M_IP6_NEXTHOP) &&
 820             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 821                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 822                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
 823                 m->m_flags |= M_SKIP_FIREWALL;
 824                 m->m_flags &= ~M_IP6_NEXTHOP;
 825                 m_tag_delete(m, fwd_tag);
 826                 goto again;
 827         }
 828
 829 passout:
 830         /*
 831          * Send the packet to the outgoing interface.
 832          * If necessary, do IPv6 fragmentation before sending.
 833          *
 834          * the logic here is rather complex:
 835          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 836          * 1-a: send as is if tlen <= path mtu
 837          * 1-b: fragment if tlen > path mtu
 838          *
 839          * 2: if user asks us not to fragment (dontfrag == 1)
 840          * 2-a: send as is if tlen <= interface mtu
 841          * 2-b: error if tlen > interface mtu
 842          *
 843          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 844          *      always fragment
 845          *
 846          * 4: if dontfrag == 1 && alwaysfrag == 1
 847          *      error, as we cannot handle this conflicting request
 848          */
 849         sw_csum = m->m_pkthdr.csum_flags;
 850         if (!hdrsplit) {
 851                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 852                 sw_csum &= ~ifp->if_hwassist;
 853         } else
 854                 tso = 0;
 855         /*
 856          * If we added extension headers, we will not do TSO and calculate the
 857          * checksums ourselves for now.
 858          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 859          * with ext. hdrs.
 860          */
 861         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 862                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 863                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 864         }
 865 #ifdef SCTP
 866         if (sw_csum & CSUM_SCTP_IPV6) {
 867                 sw_csum &= ~CSUM_SCTP_IPV6;
 868                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 869         }
 870 #endif
 871         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 872         tlen = m->m_pkthdr.len;
 873
 874         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 875                 dontfrag = 1;
 876         else
 877                 dontfrag = 0;
 878         if (dontfrag && alwaysfrag) {   /* case 4 */
 879                 /* conflicting request - can't transmit */
 880                 error = EMSGSIZE;
 881                 goto bad;
 882         }
 883         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 884                 /*
 885                  * Even if the DONTFRAG option is specified, we cannot send the
 886                  * packet when the data length is larger than the MTU of the
 887                  * outgoing interface.
 888                  * Notify the error by sending IPV6_PATHMTU ancillary data if
 889                  * application wanted to know the MTU value. Also return an
 890                  * error code (this is not described in the API spec).
 891                  */
 892                 if (inp != NULL)
 893                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
 894                 error = EMSGSIZE;
 895                 goto bad;
 896         }
 897
 898         /*
 899          * transmit packet without fragmentation
 900          */
 901         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 902                 struct in6_ifaddr *ia6;
 903
 904                 ip6 = mtod(m, struct ip6_hdr *);
 905                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 906                 if (ia6) {
 907                         /* Record statistics for this interface address. */
 908                         ia6->ia_ifa.if_opackets++;
 909                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
 910                         ifa_free(&ia6->ia_ifa);
 911                 }
 912                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 913                 goto done;
 914         }
 915
 916         /*
 917          * try to fragment the packet.  case 1-b and 3
 918          */
 919         if (mtu < IPV6_MMTU) {
 920                 /* path MTU cannot be less than IPV6_MMTU */
 921                 error = EMSGSIZE;
 922                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 923                 goto bad;
 924         } else if (ip6->ip6_plen == 0) {
 925                 /* jumbo payload cannot be fragmented */
 926                 error = EMSGSIZE;
 927                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 928                 goto bad;
 929         } else {
 930                 u_char nextproto;
 931
 932                 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
 933
 934                 /*
 935                  * Too large for the destination or interface;
 936                  * fragment if possible.
 937                  * Must be able to put at least 8 bytes per fragment.
 938                  */
 939                 hlen = unfragpartlen;
 940                 if (mtu > IPV6_MAXPACKET)
 941                         mtu = IPV6_MAXPACKET;
 942
 943                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 944                 if (len < 8) {
 945                         error = EMSGSIZE;
 946                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 947                         goto bad;
 948                 }
 949
 950                 /*
 951                  * Verify that we have any chance at all of being able to queue
 952                  *      the packet or packet fragments
 953                  */
 954                 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
 955                     < tlen  /* - hlen */)) {
 956                         error = ENOBUFS;
 957                         IP6STAT_INC(ip6s_odropped);
 958                         goto bad;
 959                 }
 960
 961
 962                 /*
 963                  * If the interface will not calculate checksums on
 964                  * fragmented packets, then do it here.
 965                  * XXX-BZ handle the hw offloading case.  Need flags.
 966                  */
 967                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 968                         in6_delayed_cksum(m, plen, hlen);
 969                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 970                 }
 971 #ifdef SCTP
 972                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
 973                         sctp_delayed_cksum(m, hlen);
 974                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 975                 }
 976 #endif
 977                 /*
 978                  * Change the next header field of the last header in the
 979                  * unfragmentable part.
 980                  */
 981                 if (exthdrs.ip6e_rthdr) {
 982                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
 983                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
 984                 } else if (exthdrs.ip6e_dest1) {
 985                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
 986                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
 987                 } else if (exthdrs.ip6e_hbh) {
 988                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
 989                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
 990                 } else {
 991                         nextproto = ip6->ip6_nxt;
 992                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
 993                 }
 994
 995                 /*
 996                  * Loop through length of segment after first fragment,
 997                  * make new header and copy data of each part and link onto
 998                  * chain.
 999                  */
1000                 m0 = m;
1001                 id = htonl(ip6_randomid());
1002                 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1003                         goto sendorfree;
1004
1005                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1006         }
1007
1008         /*
1009          * Remove leading garbages.
1010          */
1011 sendorfree:
1012         m = m0->m_nextpkt;
1013         m0->m_nextpkt = 0;
1014         m_freem(m0);
1015         for (m0 = m; m; m = m0) {
1016                 m0 = m->m_nextpkt;
1017                 m->m_nextpkt = 0;
1018                 if (error == 0) {
1019                         /* Record statistics for this interface address. */
1020                         if (ia) {
1021                                 ia->ia_ifa.if_opackets++;
1022                                 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1023                         }
1024                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1025                 } else
1026                         m_freem(m);
1027         }
1028
1029         if (error == 0)
1030                 IP6STAT_INC(ip6s_fragmented);
1031
1032 done:
1033         if (ro == &ip6route)
1034                 RO_RTFREE(ro);
1035         if (ro_pmtu == &ip6route)
1036                 RO_RTFREE(ro_pmtu);
1037         return (error);
1038
1039 freehdrs:
1040         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1041         m_freem(exthdrs.ip6e_dest1);
1042         m_freem(exthdrs.ip6e_rthdr);
1043         m_freem(exthdrs.ip6e_dest2);
1044         /* FALLTHROUGH */
1045 bad:
1046         if (m)
1047                 m_freem(m);
1048         goto done;
1049 }
1050
1051 static int
1052 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1053 {
1054         struct mbuf *m;
1055
1056         if (hlen > MCLBYTES)
1057                 return (ENOBUFS); /* XXX */
1058
1059         if (hlen > MLEN)
1060                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1061         else
1062                 m = m_get(M_NOWAIT, MT_DATA);
1063         if (m == NULL)
1064                 return (ENOBUFS);
1065         m->m_len = hlen;
1066         if (hdr)
1067                 bcopy(hdr, mtod(m, caddr_t), hlen);
1068
1069         *mp = m;
1070         return (0);
1071 }
1072
1073 /*
1074  * Insert jumbo payload option.
1075  */
1076 static int
1077 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1078 {
1079         struct mbuf *mopt;
1080         u_char *optbuf;
1081         u_int32_t v;
1082
1083 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1084
1085         /*
1086          * If there is no hop-by-hop options header, allocate new one.
1087          * If there is one but it doesn't have enough space to store the
1088          * jumbo payload option, allocate a cluster to store the whole options.
1089          * Otherwise, use it to store the options.
1090          */
1091         if (exthdrs->ip6e_hbh == 0) {
1092                 mopt = m_get(M_NOWAIT, MT_DATA);
1093                 if (mopt == NULL)
1094                         return (ENOBUFS);
1095                 mopt->m_len = JUMBOOPTLEN;
1096                 optbuf = mtod(mopt, u_char *);
1097                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1098                 exthdrs->ip6e_hbh = mopt;
1099         } else {
1100                 struct ip6_hbh *hbh;
1101
1102                 mopt = exthdrs->ip6e_hbh;
1103                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1104                         /*
1105                          * XXX assumption:
1106                          * - exthdrs->ip6e_hbh is not referenced from places
1107                          *   other than exthdrs.
1108                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1109                          */
1110                         int oldoptlen = mopt->m_len;
1111                         struct mbuf *n;
1112
1113                         /*
1114                          * XXX: give up if the whole (new) hbh header does
1115                          * not fit even in an mbuf cluster.
1116                          */
1117                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1118                                 return (ENOBUFS);
1119
1120                         /*
1121                          * As a consequence, we must always prepare a cluster
1122                          * at this point.
1123                          */
1124                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1125                         if (n == NULL)
1126                                 return (ENOBUFS);
1127                         n->m_len = oldoptlen + JUMBOOPTLEN;
1128                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1129                             oldoptlen);
1130                         optbuf = mtod(n, caddr_t) + oldoptlen;
1131                         m_freem(mopt);
1132                         mopt = exthdrs->ip6e_hbh = n;
1133                 } else {
1134                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1135                         mopt->m_len += JUMBOOPTLEN;
1136                 }
1137                 optbuf[0] = IP6OPT_PADN;
1138                 optbuf[1] = 1;
1139
1140                 /*
1141                  * Adjust the header length according to the pad and
1142                  * the jumbo payload option.
1143                  */
1144                 hbh = mtod(mopt, struct ip6_hbh *);
1145                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1146         }
1147
1148         /* fill in the option. */
1149         optbuf[2] = IP6OPT_JUMBO;
1150         optbuf[3] = 4;
1151         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1152         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1153
1154         /* finally, adjust the packet header length */
1155         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1156
1157         return (0);
1158 #undef JUMBOOPTLEN
1159 }
1160
1161 /*
1162  * Insert fragment header and copy unfragmentable header portions.
1163  */
1164 static int
1165 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1166     struct ip6_frag **frghdrp)
1167 {
1168         struct mbuf *n, *mlast;
1169
1170         if (hlen > sizeof(struct ip6_hdr)) {
1171                 n = m_copym(m0, sizeof(struct ip6_hdr),
1172                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1173                 if (n == 0)
1174                         return (ENOBUFS);
1175                 m->m_next = n;
1176         } else
1177                 n = m;
1178
1179         /* Search for the last mbuf of unfragmentable part. */
1180         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1181                 ;
1182
1183         if ((mlast->m_flags & M_EXT) == 0 &&
1184             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1185                 /* use the trailing space of the last mbuf for the fragment hdr */
1186                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1187                     mlast->m_len);
1188                 mlast->m_len += sizeof(struct ip6_frag);
1189                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1190         } else {
1191                 /* allocate a new mbuf for the fragment header */
1192                 struct mbuf *mfrg;
1193
1194                 mfrg = m_get(M_NOWAIT, MT_DATA);
1195                 if (mfrg == NULL)
1196                         return (ENOBUFS);
1197                 mfrg->m_len = sizeof(struct ip6_frag);
1198                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1199                 mlast->m_next = mfrg;
1200         }
1201
1202         return (0);
1203 }
1204
1205 static int
1206 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1207     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1208     int *alwaysfragp, u_int fibnum)
1209 {
1210         u_int32_t mtu = 0;
1211         int alwaysfrag = 0;
1212         int error = 0;
1213
1214         if (ro_pmtu != ro) {
1215                 /* The first hop and the final destination may differ. */
1216                 struct sockaddr_in6 *sa6_dst =
1217                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1218                 if (ro_pmtu->ro_rt &&
1219                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1220                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1221                         RTFREE(ro_pmtu->ro_rt);
1222                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1223                 }
1224                 if (ro_pmtu->ro_rt == NULL) {
1225                         bzero(sa6_dst, sizeof(*sa6_dst));
1226                         sa6_dst->sin6_family = AF_INET6;
1227                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1228                         sa6_dst->sin6_addr = *dst;
1229
1230                         in6_rtalloc(ro_pmtu, fibnum);
1231                 }
1232         }
1233         if (ro_pmtu->ro_rt) {
1234                 u_int32_t ifmtu;
1235                 struct in_conninfo inc;
1236
1237                 bzero(&inc, sizeof(inc));
1238                 inc.inc_flags |= INC_ISIPV6;
1239                 inc.inc6_faddr = *dst;
1240
1241                 if (ifp == NULL)
1242                         ifp = ro_pmtu->ro_rt->rt_ifp;
1243                 ifmtu = IN6_LINKMTU(ifp);
1244                 mtu = tcp_hc_getmtu(&inc);
1245                 if (mtu)
1246                         mtu = min(mtu, ro_pmtu->ro_rt->rt_mtu);
1247                 else
1248                         mtu = ro_pmtu->ro_rt->rt_mtu;
1249                 if (mtu == 0)
1250                         mtu = ifmtu;
1251                 else if (mtu < IPV6_MMTU) {
1252                         /*
1253                          * RFC2460 section 5, last paragraph:
1254                          * if we record ICMPv6 too big message with
1255                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1256                          * or smaller, with framgent header attached.
1257                          * (fragment header is needed regardless from the
1258                          * packet size, for translators to identify packets)
1259                          */
1260                         alwaysfrag = 1;
1261                         mtu = IPV6_MMTU;
1262                 } else if (mtu > ifmtu) {
1263                         /*
1264                          * The MTU on the route is larger than the MTU on
1265                          * the interface!  This shouldn't happen, unless the
1266                          * MTU of the interface has been changed after the
1267                          * interface was brought up.  Change the MTU in the
1268                          * route to match the interface MTU (as long as the
1269                          * field isn't locked).
1270                          */
1271                         mtu = ifmtu;
1272                         ro_pmtu->ro_rt->rt_mtu = mtu;
1273                 }
1274         } else if (ifp) {
1275                 mtu = IN6_LINKMTU(ifp);
1276         } else
1277                 error = EHOSTUNREACH; /* XXX */
1278
1279         *mtup = mtu;
1280         if (alwaysfragp)
1281                 *alwaysfragp = alwaysfrag;
1282         return (error);
1283 }
1284
1285 /*
1286  * IP6 socket option processing.
1287  */
1288 int
1289 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1290 {
1291         int optdatalen, uproto;
1292         void *optdata;
1293         struct inpcb *in6p = sotoinpcb(so);
1294         int error, optval;
1295         int level, op, optname;
1296         int optlen;
1297         struct thread *td;
1298
1299         level = sopt->sopt_level;
1300         op = sopt->sopt_dir;
1301         optname = sopt->sopt_name;
1302         optlen = sopt->sopt_valsize;
1303         td = sopt->sopt_td;
1304         error = 0;
1305         optval = 0;
1306         uproto = (int)so->so_proto->pr_protocol;
1307
1308         if (level != IPPROTO_IPV6) {
1309                 error = EINVAL;
1310
1311                 if (sopt->sopt_level == SOL_SOCKET &&
1312                     sopt->sopt_dir == SOPT_SET) {
1313                         switch (sopt->sopt_name) {
1314                         case SO_REUSEADDR:
1315                                 INP_WLOCK(in6p);
1316                                 if ((so->so_options & SO_REUSEADDR) != 0)
1317                                         in6p->inp_flags2 |= INP_REUSEADDR;
1318                                 else
1319                                         in6p->inp_flags2 &= ~INP_REUSEADDR;
1320                                 INP_WUNLOCK(in6p);
1321                                 error = 0;
1322                                 break;
1323                         case SO_REUSEPORT:
1324                                 INP_WLOCK(in6p);
1325                                 if ((so->so_options & SO_REUSEPORT) != 0)
1326                                         in6p->inp_flags2 |= INP_REUSEPORT;
1327                                 else
1328                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1329                                 INP_WUNLOCK(in6p);
1330                                 error = 0;
1331                                 break;
1332                         case SO_SETFIB:
1333                                 INP_WLOCK(in6p);
1334                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1335                                 INP_WUNLOCK(in6p);
1336                                 error = 0;
1337                                 break;
1338                         default:
1339                                 break;
1340                         }
1341                 }
1342         } else {                /* level == IPPROTO_IPV6 */
1343                 switch (op) {
1344
1345                 case SOPT_SET:
1346                         switch (optname) {
1347                         case IPV6_2292PKTOPTIONS:
1348 #ifdef IPV6_PKTOPTIONS
1349                         case IPV6_PKTOPTIONS:
1350 #endif
1351                         {
1352                                 struct mbuf *m;
1353
1354                                 error = soopt_getm(sopt, &m); /* XXX */
1355                                 if (error != 0)
1356                                         break;
1357                                 error = soopt_mcopyin(sopt, m); /* XXX */
1358                                 if (error != 0)
1359                                         break;
1360                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1361                                                     m, so, sopt);
1362                                 m_freem(m); /* XXX */
1363                                 break;
1364                         }
1365
1366                         /*
1367                          * Use of some Hop-by-Hop options or some
1368                          * Destination options, might require special
1369                          * privilege.  That is, normal applications
1370                          * (without special privilege) might be forbidden
1371                          * from setting certain options in outgoing packets,
1372                          * and might never see certain options in received
1373                          * packets. [RFC 2292 Section 6]
1374                          * KAME specific note:
1375                          *  KAME prevents non-privileged users from sending or
1376                          *  receiving ANY hbh/dst options in order to avoid
1377                          *  overhead of parsing options in the kernel.
1378                          */
1379                         case IPV6_RECVHOPOPTS:
1380                         case IPV6_RECVDSTOPTS:
1381                         case IPV6_RECVRTHDRDSTOPTS:
1382                                 if (td != NULL) {
1383                                         error = priv_check(td,
1384                                             PRIV_NETINET_SETHDROPTS);
1385                                         if (error)
1386                                                 break;
1387                                 }
1388                                 /* FALLTHROUGH */
1389                         case IPV6_UNICAST_HOPS:
1390                         case IPV6_HOPLIMIT:
1391                         case IPV6_FAITH:
1392
1393                         case IPV6_RECVPKTINFO:
1394                         case IPV6_RECVHOPLIMIT:
1395                         case IPV6_RECVRTHDR:
1396                         case IPV6_RECVPATHMTU:
1397                         case IPV6_RECVTCLASS:
1398                         case IPV6_V6ONLY:
1399                         case IPV6_AUTOFLOWLABEL:
1400                         case IPV6_BINDANY:
1401                                 if (optname == IPV6_BINDANY && td != NULL) {
1402                                         error = priv_check(td,
1403                                             PRIV_NETINET_BINDANY);
1404                                         if (error)
1405                                                 break;
1406                                 }
1407
1408                                 if (optlen != sizeof(int)) {
1409                                         error = EINVAL;
1410                                         break;
1411                                 }
1412                                 error = sooptcopyin(sopt, &optval,
1413                                         sizeof optval, sizeof optval);
1414                                 if (error)
1415                                         break;
1416                                 switch (optname) {
1417
1418                                 case IPV6_UNICAST_HOPS:
1419                                         if (optval < -1 || optval >= 256)
1420                                                 error = EINVAL;
1421                                         else {
1422                                                 /* -1 = kernel default */
1423                                                 in6p->in6p_hops = optval;
1424                                                 if ((in6p->inp_vflag &
1425                                                      INP_IPV4) != 0)
1426                                                         in6p->inp_ip_ttl = optval;
1427                                         }
1428                                         break;
1429 #define OPTSET(bit) \
1430 do { \
1431         INP_WLOCK(in6p); \
1432         if (optval) \
1433                 in6p->inp_flags |= (bit); \
1434         else \
1435                 in6p->inp_flags &= ~(bit); \
1436         INP_WUNLOCK(in6p); \
1437 } while (/*CONSTCOND*/ 0)
1438 #define OPTSET2292(bit) \
1439 do { \
1440         INP_WLOCK(in6p); \
1441         in6p->inp_flags |= IN6P_RFC2292; \
1442         if (optval) \
1443                 in6p->inp_flags |= (bit); \
1444         else \
1445                 in6p->inp_flags &= ~(bit); \
1446         INP_WUNLOCK(in6p); \
1447 } while (/*CONSTCOND*/ 0)
1448 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1449
1450                                 case IPV6_RECVPKTINFO:
1451                                         /* cannot mix with RFC2292 */
1452                                         if (OPTBIT(IN6P_RFC2292)) {
1453                                                 error = EINVAL;
1454                                                 break;
1455                                         }
1456                                         OPTSET(IN6P_PKTINFO);
1457                                         break;
1458
1459                                 case IPV6_HOPLIMIT:
1460                                 {
1461                                         struct ip6_pktopts **optp;
1462
1463                                         /* cannot mix with RFC2292 */
1464                                         if (OPTBIT(IN6P_RFC2292)) {
1465                                                 error = EINVAL;
1466                                                 break;
1467                                         }
1468                                         optp = &in6p->in6p_outputopts;
1469                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1470                                             (u_char *)&optval, sizeof(optval),
1471                                             optp, (td != NULL) ? td->td_ucred :
1472                                             NULL, uproto);
1473                                         break;
1474                                 }
1475
1476                                 case IPV6_RECVHOPLIMIT:
1477                                         /* cannot mix with RFC2292 */
1478                                         if (OPTBIT(IN6P_RFC2292)) {
1479                                                 error = EINVAL;
1480                                                 break;
1481                                         }
1482                                         OPTSET(IN6P_HOPLIMIT);
1483                                         break;
1484
1485                                 case IPV6_RECVHOPOPTS:
1486                                         /* cannot mix with RFC2292 */
1487                                         if (OPTBIT(IN6P_RFC2292)) {
1488                                                 error = EINVAL;
1489                                                 break;
1490                                         }
1491                                         OPTSET(IN6P_HOPOPTS);
1492                                         break;
1493
1494                                 case IPV6_RECVDSTOPTS:
1495                                         /* cannot mix with RFC2292 */
1496                                         if (OPTBIT(IN6P_RFC2292)) {
1497                                                 error = EINVAL;
1498                                                 break;
1499                                         }
1500                                         OPTSET(IN6P_DSTOPTS);
1501                                         break;
1502
1503                                 case IPV6_RECVRTHDRDSTOPTS:
1504                                         /* cannot mix with RFC2292 */
1505                                         if (OPTBIT(IN6P_RFC2292)) {
1506                                                 error = EINVAL;
1507                                                 break;
1508                                         }
1509                                         OPTSET(IN6P_RTHDRDSTOPTS);
1510                                         break;
1511
1512                                 case IPV6_RECVRTHDR:
1513                                         /* cannot mix with RFC2292 */
1514                                         if (OPTBIT(IN6P_RFC2292)) {
1515                                                 error = EINVAL;
1516                                                 break;
1517                                         }
1518                                         OPTSET(IN6P_RTHDR);
1519                                         break;
1520
1521                                 case IPV6_FAITH:
1522                                         OPTSET(INP_FAITH);
1523                                         break;
1524
1525                                 case IPV6_RECVPATHMTU:
1526                                         /*
1527                                          * We ignore this option for TCP
1528                                          * sockets.
1529                                          * (RFC3542 leaves this case
1530                                          * unspecified.)
1531                                          */
1532                                         if (uproto != IPPROTO_TCP)
1533                                                 OPTSET(IN6P_MTU);
1534                                         break;
1535
1536                                 case IPV6_V6ONLY:
1537                                         /*
1538                                          * make setsockopt(IPV6_V6ONLY)
1539                                          * available only prior to bind(2).
1540                                          * see ipng mailing list, Jun 22 2001.
1541                                          */
1542                                         if (in6p->inp_lport ||
1543                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1544                                                 error = EINVAL;
1545                                                 break;
1546                                         }
1547                                         OPTSET(IN6P_IPV6_V6ONLY);
1548                                         if (optval)
1549                                                 in6p->inp_vflag &= ~INP_IPV4;
1550                                         else
1551                                                 in6p->inp_vflag |= INP_IPV4;
1552                                         break;
1553                                 case IPV6_RECVTCLASS:
1554                                         /* cannot mix with RFC2292 XXX */
1555                                         if (OPTBIT(IN6P_RFC2292)) {
1556                                                 error = EINVAL;
1557                                                 break;
1558                                         }
1559                                         OPTSET(IN6P_TCLASS);
1560                                         break;
1561                                 case IPV6_AUTOFLOWLABEL:
1562                                         OPTSET(IN6P_AUTOFLOWLABEL);
1563                                         break;
1564
1565                                 case IPV6_BINDANY:
1566                                         OPTSET(INP_BINDANY);
1567                                         break;
1568                                 }
1569                                 break;
1570
1571                         case IPV6_TCLASS:
1572                         case IPV6_DONTFRAG:
1573                         case IPV6_USE_MIN_MTU:
1574                         case IPV6_PREFER_TEMPADDR:
1575                                 if (optlen != sizeof(optval)) {
1576                                         error = EINVAL;
1577                                         break;
1578                                 }
1579                                 error = sooptcopyin(sopt, &optval,
1580                                         sizeof optval, sizeof optval);
1581                                 if (error)
1582                                         break;
1583                                 {
1584                                         struct ip6_pktopts **optp;
1585                                         optp = &in6p->in6p_outputopts;
1586                                         error = ip6_pcbopt(optname,
1587                                             (u_char *)&optval, sizeof(optval),
1588                                             optp, (td != NULL) ? td->td_ucred :
1589                                             NULL, uproto);
1590                                         break;
1591                                 }
1592
1593                         case IPV6_2292PKTINFO:
1594                         case IPV6_2292HOPLIMIT:
1595                         case IPV6_2292HOPOPTS:
1596                         case IPV6_2292DSTOPTS:
1597                         case IPV6_2292RTHDR:
1598                                 /* RFC 2292 */
1599                                 if (optlen != sizeof(int)) {
1600                                         error = EINVAL;
1601                                         break;
1602                                 }
1603                                 error = sooptcopyin(sopt, &optval,
1604                                         sizeof optval, sizeof optval);
1605                                 if (error)
1606                                         break;
1607                                 switch (optname) {
1608                                 case IPV6_2292PKTINFO:
1609                                         OPTSET2292(IN6P_PKTINFO);
1610                                         break;
1611                                 case IPV6_2292HOPLIMIT:
1612                                         OPTSET2292(IN6P_HOPLIMIT);
1613                                         break;
1614                                 case IPV6_2292HOPOPTS:
1615                                         /*
1616                                          * Check super-user privilege.
1617                                          * See comments for IPV6_RECVHOPOPTS.
1618                                          */
1619                                         if (td != NULL) {
1620                                                 error = priv_check(td,
1621                                                     PRIV_NETINET_SETHDROPTS);
1622                                                 if (error)
1623                                                         return (error);
1624                                         }
1625                                         OPTSET2292(IN6P_HOPOPTS);
1626                                         break;
1627                                 case IPV6_2292DSTOPTS:
1628                                         if (td != NULL) {
1629                                                 error = priv_check(td,
1630                                                     PRIV_NETINET_SETHDROPTS);
1631                                                 if (error)
1632                                                         return (error);
1633                                         }
1634                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1635                                         break;
1636                                 case IPV6_2292RTHDR:
1637                                         OPTSET2292(IN6P_RTHDR);
1638                                         break;
1639                                 }
1640                                 break;
1641                         case IPV6_PKTINFO:
1642                         case IPV6_HOPOPTS:
1643                         case IPV6_RTHDR:
1644                         case IPV6_DSTOPTS:
1645                         case IPV6_RTHDRDSTOPTS:
1646                         case IPV6_NEXTHOP:
1647                         {
1648                                 /* new advanced API (RFC3542) */
1649                                 u_char *optbuf;
1650                                 u_char optbuf_storage[MCLBYTES];
1651                                 int optlen;
1652                                 struct ip6_pktopts **optp;
1653
1654                                 /* cannot mix with RFC2292 */
1655                                 if (OPTBIT(IN6P_RFC2292)) {
1656                                         error = EINVAL;
1657                                         break;
1658                                 }
1659
1660                                 /*
1661                                  * We only ensure valsize is not too large
1662                                  * here.  Further validation will be done
1663                                  * later.
1664                                  */
1665                                 error = sooptcopyin(sopt, optbuf_storage,
1666                                     sizeof(optbuf_storage), 0);
1667                                 if (error)
1668                                         break;
1669                                 optlen = sopt->sopt_valsize;
1670                                 optbuf = optbuf_storage;
1671                                 optp = &in6p->in6p_outputopts;
1672                                 error = ip6_pcbopt(optname, optbuf, optlen,
1673                                     optp, (td != NULL) ? td->td_ucred : NULL,
1674                                     uproto);
1675                                 break;
1676                         }
1677 #undef OPTSET
1678
1679                         case IPV6_MULTICAST_IF:
1680                         case IPV6_MULTICAST_HOPS:
1681                         case IPV6_MULTICAST_LOOP:
1682                         case IPV6_JOIN_GROUP:
1683                         case IPV6_LEAVE_GROUP:
1684                         case IPV6_MSFILTER:
1685                         case MCAST_BLOCK_SOURCE:
1686                         case MCAST_UNBLOCK_SOURCE:
1687                         case MCAST_JOIN_GROUP:
1688                         case MCAST_LEAVE_GROUP:
1689                         case MCAST_JOIN_SOURCE_GROUP:
1690                         case MCAST_LEAVE_SOURCE_GROUP:
1691                                 error = ip6_setmoptions(in6p, sopt);
1692                                 break;
1693
1694                         case IPV6_PORTRANGE:
1695                                 error = sooptcopyin(sopt, &optval,
1696                                     sizeof optval, sizeof optval);
1697                                 if (error)
1698                                         break;
1699
1700                                 INP_WLOCK(in6p);
1701                                 switch (optval) {
1702                                 case IPV6_PORTRANGE_DEFAULT:
1703                                         in6p->inp_flags &= ~(INP_LOWPORT);
1704                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1705                                         break;
1706
1707                                 case IPV6_PORTRANGE_HIGH:
1708                                         in6p->inp_flags &= ~(INP_LOWPORT);
1709                                         in6p->inp_flags |= INP_HIGHPORT;
1710                                         break;
1711
1712                                 case IPV6_PORTRANGE_LOW:
1713                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1714                                         in6p->inp_flags |= INP_LOWPORT;
1715                                         break;
1716
1717                                 default:
1718                                         error = EINVAL;
1719                                         break;
1720                                 }
1721                                 INP_WUNLOCK(in6p);
1722                                 break;
1723
1724 #ifdef IPSEC
1725                         case IPV6_IPSEC_POLICY:
1726                         {
1727                                 caddr_t req;
1728                                 struct mbuf *m;
1729
1730                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1731                                         break;
1732                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1733                                         break;
1734                                 req = mtod(m, caddr_t);
1735                                 error = ipsec_set_policy(in6p, optname, req,
1736                                     m->m_len, (sopt->sopt_td != NULL) ?
1737                                     sopt->sopt_td->td_ucred : NULL);
1738                                 m_freem(m);
1739                                 break;
1740                         }
1741 #endif /* IPSEC */
1742
1743                         default:
1744                                 error = ENOPROTOOPT;
1745                                 break;
1746                         }
1747                         break;
1748
1749                 case SOPT_GET:
1750                         switch (optname) {
1751
1752                         case IPV6_2292PKTOPTIONS:
1753 #ifdef IPV6_PKTOPTIONS
1754                         case IPV6_PKTOPTIONS:
1755 #endif
1756                                 /*
1757                                  * RFC3542 (effectively) deprecated the
1758                                  * semantics of the 2292-style pktoptions.
1759                                  * Since it was not reliable in nature (i.e.,
1760                                  * applications had to expect the lack of some
1761                                  * information after all), it would make sense
1762                                  * to simplify this part by always returning
1763                                  * empty data.
1764                                  */
1765                                 sopt->sopt_valsize = 0;
1766                                 break;
1767
1768                         case IPV6_RECVHOPOPTS:
1769                         case IPV6_RECVDSTOPTS:
1770                         case IPV6_RECVRTHDRDSTOPTS:
1771                         case IPV6_UNICAST_HOPS:
1772                         case IPV6_RECVPKTINFO:
1773                         case IPV6_RECVHOPLIMIT:
1774                         case IPV6_RECVRTHDR:
1775                         case IPV6_RECVPATHMTU:
1776
1777                         case IPV6_FAITH:
1778                         case IPV6_V6ONLY:
1779                         case IPV6_PORTRANGE:
1780                         case IPV6_RECVTCLASS:
1781                         case IPV6_AUTOFLOWLABEL:
1782                         case IPV6_BINDANY:
1783                                 switch (optname) {
1784
1785                                 case IPV6_RECVHOPOPTS:
1786                                         optval = OPTBIT(IN6P_HOPOPTS);
1787                                         break;
1788
1789                                 case IPV6_RECVDSTOPTS:
1790                                         optval = OPTBIT(IN6P_DSTOPTS);
1791                                         break;
1792
1793                                 case IPV6_RECVRTHDRDSTOPTS:
1794                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1795                                         break;
1796
1797                                 case IPV6_UNICAST_HOPS:
1798                                         optval = in6p->in6p_hops;
1799                                         break;
1800
1801                                 case IPV6_RECVPKTINFO:
1802                                         optval = OPTBIT(IN6P_PKTINFO);
1803                                         break;
1804
1805                                 case IPV6_RECVHOPLIMIT:
1806                                         optval = OPTBIT(IN6P_HOPLIMIT);
1807                                         break;
1808
1809                                 case IPV6_RECVRTHDR:
1810                                         optval = OPTBIT(IN6P_RTHDR);
1811                                         break;
1812
1813                                 case IPV6_RECVPATHMTU:
1814                                         optval = OPTBIT(IN6P_MTU);
1815                                         break;
1816
1817                                 case IPV6_FAITH:
1818                                         optval = OPTBIT(INP_FAITH);
1819                                         break;
1820
1821                                 case IPV6_V6ONLY:
1822                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1823                                         break;
1824
1825                                 case IPV6_PORTRANGE:
1826                                     {
1827                                         int flags;
1828                                         flags = in6p->inp_flags;
1829                                         if (flags & INP_HIGHPORT)
1830                                                 optval = IPV6_PORTRANGE_HIGH;
1831                                         else if (flags & INP_LOWPORT)
1832                                                 optval = IPV6_PORTRANGE_LOW;
1833                                         else
1834                                                 optval = 0;
1835                                         break;
1836                                     }
1837                                 case IPV6_RECVTCLASS:
1838                                         optval = OPTBIT(IN6P_TCLASS);
1839                                         break;
1840
1841                                 case IPV6_AUTOFLOWLABEL:
1842                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1843                                         break;
1844
1845                                 case IPV6_BINDANY:
1846                                         optval = OPTBIT(INP_BINDANY);
1847                                         break;
1848                                 }
1849                                 if (error)
1850                                         break;
1851                                 error = sooptcopyout(sopt, &optval,
1852                                         sizeof optval);
1853                                 break;
1854
1855                         case IPV6_PATHMTU:
1856                         {
1857                                 u_long pmtu = 0;
1858                                 struct ip6_mtuinfo mtuinfo;
1859                                 struct route_in6 sro;
1860
1861                                 bzero(&sro, sizeof(sro));
1862
1863                                 if (!(so->so_state & SS_ISCONNECTED))
1864                                         return (ENOTCONN);
1865                                 /*
1866                                  * XXX: we dot not consider the case of source
1867                                  * routing, or optional information to specify
1868                                  * the outgoing interface.
1869                                  */
1870                                 error = ip6_getpmtu(&sro, NULL, NULL,
1871                                     &in6p->in6p_faddr, &pmtu, NULL,
1872                                     so->so_fibnum);
1873                                 if (sro.ro_rt)
1874                                         RTFREE(sro.ro_rt);
1875                                 if (error)
1876                                         break;
1877                                 if (pmtu > IPV6_MAXPACKET)
1878                                         pmtu = IPV6_MAXPACKET;
1879
1880                                 bzero(&mtuinfo, sizeof(mtuinfo));
1881                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1882                                 optdata = (void *)&mtuinfo;
1883                                 optdatalen = sizeof(mtuinfo);
1884                                 error = sooptcopyout(sopt, optdata,
1885                                     optdatalen);
1886                                 break;
1887                         }
1888
1889                         case IPV6_2292PKTINFO:
1890                         case IPV6_2292HOPLIMIT:
1891                         case IPV6_2292HOPOPTS:
1892                         case IPV6_2292RTHDR:
1893                         case IPV6_2292DSTOPTS:
1894                                 switch (optname) {
1895                                 case IPV6_2292PKTINFO:
1896                                         optval = OPTBIT(IN6P_PKTINFO);
1897                                         break;
1898                                 case IPV6_2292HOPLIMIT:
1899                                         optval = OPTBIT(IN6P_HOPLIMIT);
1900                                         break;
1901                                 case IPV6_2292HOPOPTS:
1902                                         optval = OPTBIT(IN6P_HOPOPTS);
1903                                         break;
1904                                 case IPV6_2292RTHDR:
1905                                         optval = OPTBIT(IN6P_RTHDR);
1906                                         break;
1907                                 case IPV6_2292DSTOPTS:
1908                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1909                                         break;
1910                                 }
1911                                 error = sooptcopyout(sopt, &optval,
1912                                     sizeof optval);
1913                                 break;
1914                         case IPV6_PKTINFO:
1915                         case IPV6_HOPOPTS:
1916                         case IPV6_RTHDR:
1917                         case IPV6_DSTOPTS:
1918                         case IPV6_RTHDRDSTOPTS:
1919                         case IPV6_NEXTHOP:
1920                         case IPV6_TCLASS:
1921                         case IPV6_DONTFRAG:
1922                         case IPV6_USE_MIN_MTU:
1923                         case IPV6_PREFER_TEMPADDR:
1924                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
1925                                     optname, sopt);
1926                                 break;
1927
1928                         case IPV6_MULTICAST_IF:
1929                         case IPV6_MULTICAST_HOPS:
1930                         case IPV6_MULTICAST_LOOP:
1931                         case IPV6_MSFILTER:
1932                                 error = ip6_getmoptions(in6p, sopt);
1933                                 break;
1934
1935 #ifdef IPSEC
1936                         case IPV6_IPSEC_POLICY:
1937                           {
1938                                 caddr_t req = NULL;
1939                                 size_t len = 0;
1940                                 struct mbuf *m = NULL;
1941                                 struct mbuf **mp = &m;
1942                                 size_t ovalsize = sopt->sopt_valsize;
1943                                 caddr_t oval = (caddr_t)sopt->sopt_val;
1944
1945                                 error = soopt_getm(sopt, &m); /* XXX */
1946                                 if (error != 0)
1947                                         break;
1948                                 error = soopt_mcopyin(sopt, m); /* XXX */
1949                                 if (error != 0)
1950                                         break;
1951                                 sopt->sopt_valsize = ovalsize;
1952                                 sopt->sopt_val = oval;
1953                                 if (m) {
1954                                         req = mtod(m, caddr_t);
1955                                         len = m->m_len;
1956                                 }
1957                                 error = ipsec_get_policy(in6p, req, len, mp);
1958                                 if (error == 0)
1959                                         error = soopt_mcopyout(sopt, m); /* XXX */
1960                                 if (error == 0 && m)
1961                                         m_freem(m);
1962                                 break;
1963                           }
1964 #endif /* IPSEC */
1965
1966                         default:
1967                                 error = ENOPROTOOPT;
1968                                 break;
1969                         }
1970                         break;
1971                 }
1972         }
1973         return (error);
1974 }
1975
1976 int
1977 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
1978 {
1979         int error = 0, optval, optlen;
1980         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1981         struct inpcb *in6p = sotoinpcb(so);
1982         int level, op, optname;
1983
1984         level = sopt->sopt_level;
1985         op = sopt->sopt_dir;
1986         optname = sopt->sopt_name;
1987         optlen = sopt->sopt_valsize;
1988
1989         if (level != IPPROTO_IPV6) {
1990                 return (EINVAL);
1991         }
1992
1993         switch (optname) {
1994         case IPV6_CHECKSUM:
1995                 /*
1996                  * For ICMPv6 sockets, no modification allowed for checksum
1997                  * offset, permit "no change" values to help existing apps.
1998                  *
1999                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2000                  * for an ICMPv6 socket will fail."
2001                  * The current behavior does not meet RFC3542.
2002                  */
2003                 switch (op) {
2004                 case SOPT_SET:
2005                         if (optlen != sizeof(int)) {
2006                                 error = EINVAL;
2007                                 break;
2008                         }
2009                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2010                                             sizeof(optval));
2011                         if (error)
2012                                 break;
2013                         if ((optval % 2) != 0) {
2014                                 /* the API assumes even offset values */
2015                                 error = EINVAL;
2016                         } else if (so->so_proto->pr_protocol ==
2017                             IPPROTO_ICMPV6) {
2018                                 if (optval != icmp6off)
2019                                         error = EINVAL;
2020                         } else
2021                                 in6p->in6p_cksum = optval;
2022                         break;
2023
2024                 case SOPT_GET:
2025                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2026                                 optval = icmp6off;
2027                         else
2028                                 optval = in6p->in6p_cksum;
2029
2030                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2031                         break;
2032
2033                 default:
2034                         error = EINVAL;
2035                         break;
2036                 }
2037                 break;
2038
2039         default:
2040                 error = ENOPROTOOPT;
2041                 break;
2042         }
2043
2044         return (error);
2045 }
2046
2047 /*
2048  * Set up IP6 options in pcb for insertion in output packets or
2049  * specifying behavior of outgoing packets.
2050  */
2051 static int
2052 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2053     struct socket *so, struct sockopt *sopt)
2054 {
2055         struct ip6_pktopts *opt = *pktopt;
2056         int error = 0;
2057         struct thread *td = sopt->sopt_td;
2058
2059         /* turn off any old options. */
2060         if (opt) {
2061 #ifdef DIAGNOSTIC
2062                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2063                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2064                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2065                         printf("ip6_pcbopts: all specified options are cleared.\n");
2066 #endif
2067                 ip6_clearpktopts(opt, -1);
2068         } else
2069                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2070         *pktopt = NULL;
2071
2072         if (!m || m->m_len == 0) {
2073                 /*
2074                  * Only turning off any previous options, regardless of
2075                  * whether the opt is just created or given.
2076                  */
2077                 free(opt, M_IP6OPT);
2078                 return (0);
2079         }
2080
2081         /*  set options specified by user. */
2082         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2083             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2084                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2085                 free(opt, M_IP6OPT);
2086                 return (error);
2087         }
2088         *pktopt = opt;
2089         return (0);
2090 }
2091
2092 /*
2093  * initialize ip6_pktopts.  beware that there are non-zero default values in
2094  * the struct.
2095  */
2096 void
2097 ip6_initpktopts(struct ip6_pktopts *opt)
2098 {
2099
2100         bzero(opt, sizeof(*opt));
2101         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2102         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2103         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2104         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2105 }
2106
2107 static int
2108 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2109     struct ucred *cred, int uproto)
2110 {
2111         struct ip6_pktopts *opt;
2112
2113         if (*pktopt == NULL) {
2114                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2115                     M_WAITOK);
2116                 ip6_initpktopts(*pktopt);
2117         }
2118         opt = *pktopt;
2119
2120         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2121 }
2122
2123 static int
2124 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2125 {
2126         void *optdata = NULL;
2127         int optdatalen = 0;
2128         struct ip6_ext *ip6e;
2129         int error = 0;
2130         struct in6_pktinfo null_pktinfo;
2131         int deftclass = 0, on;
2132         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2133         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2134
2135         switch (optname) {
2136         case IPV6_PKTINFO:
2137                 if (pktopt && pktopt->ip6po_pktinfo)
2138                         optdata = (void *)pktopt->ip6po_pktinfo;
2139                 else {
2140                         /* XXX: we don't have to do this every time... */
2141                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2142                         optdata = (void *)&null_pktinfo;
2143                 }
2144                 optdatalen = sizeof(struct in6_pktinfo);
2145                 break;
2146         case IPV6_TCLASS:
2147                 if (pktopt && pktopt->ip6po_tclass >= 0)
2148                         optdata = (void *)&pktopt->ip6po_tclass;
2149                 else
2150                         optdata = (void *)&deftclass;
2151                 optdatalen = sizeof(int);
2152                 break;
2153         case IPV6_HOPOPTS:
2154                 if (pktopt && pktopt->ip6po_hbh) {
2155                         optdata = (void *)pktopt->ip6po_hbh;
2156                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2157                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2158                 }
2159                 break;
2160         case IPV6_RTHDR:
2161                 if (pktopt && pktopt->ip6po_rthdr) {
2162                         optdata = (void *)pktopt->ip6po_rthdr;
2163                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2164                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2165                 }
2166                 break;
2167         case IPV6_RTHDRDSTOPTS:
2168                 if (pktopt && pktopt->ip6po_dest1) {
2169                         optdata = (void *)pktopt->ip6po_dest1;
2170                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2171                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2172                 }
2173                 break;
2174         case IPV6_DSTOPTS:
2175                 if (pktopt && pktopt->ip6po_dest2) {
2176                         optdata = (void *)pktopt->ip6po_dest2;
2177                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2178                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2179                 }
2180                 break;
2181         case IPV6_NEXTHOP:
2182                 if (pktopt && pktopt->ip6po_nexthop) {
2183                         optdata = (void *)pktopt->ip6po_nexthop;
2184                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2185                 }
2186                 break;
2187         case IPV6_USE_MIN_MTU:
2188                 if (pktopt)
2189                         optdata = (void *)&pktopt->ip6po_minmtu;
2190                 else
2191                         optdata = (void *)&defminmtu;
2192                 optdatalen = sizeof(int);
2193                 break;
2194         case IPV6_DONTFRAG:
2195                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2196                         on = 1;
2197                 else
2198                         on = 0;
2199                 optdata = (void *)&on;
2200                 optdatalen = sizeof(on);
2201                 break;
2202         case IPV6_PREFER_TEMPADDR:
2203                 if (pktopt)
2204                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2205                 else
2206                         optdata = (void *)&defpreftemp;
2207                 optdatalen = sizeof(int);
2208                 break;
2209         default:                /* should not happen */
2210 #ifdef DIAGNOSTIC
2211                 panic("ip6_getpcbopt: unexpected option\n");
2212 #endif
2213                 return (ENOPROTOOPT);
2214         }
2215
2216         error = sooptcopyout(sopt, optdata, optdatalen);
2217
2218         return (error);
2219 }
2220
2221 void
2222 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2223 {
2224         if (pktopt == NULL)
2225                 return;
2226
2227         if (optname == -1 || optname == IPV6_PKTINFO) {
2228                 if (pktopt->ip6po_pktinfo)
2229                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2230                 pktopt->ip6po_pktinfo = NULL;
2231         }
2232         if (optname == -1 || optname == IPV6_HOPLIMIT)
2233                 pktopt->ip6po_hlim = -1;
2234         if (optname == -1 || optname == IPV6_TCLASS)
2235                 pktopt->ip6po_tclass = -1;
2236         if (optname == -1 || optname == IPV6_NEXTHOP) {
2237                 if (pktopt->ip6po_nextroute.ro_rt) {
2238                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2239                         pktopt->ip6po_nextroute.ro_rt = NULL;
2240                 }
2241                 if (pktopt->ip6po_nexthop)
2242                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2243                 pktopt->ip6po_nexthop = NULL;
2244         }
2245         if (optname == -1 || optname == IPV6_HOPOPTS) {
2246                 if (pktopt->ip6po_hbh)
2247                         free(pktopt->ip6po_hbh, M_IP6OPT);
2248                 pktopt->ip6po_hbh = NULL;
2249         }
2250         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2251                 if (pktopt->ip6po_dest1)
2252                         free(pktopt->ip6po_dest1, M_IP6OPT);
2253                 pktopt->ip6po_dest1 = NULL;
2254         }
2255         if (optname == -1 || optname == IPV6_RTHDR) {
2256                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2257                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2258                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2259                 if (pktopt->ip6po_route.ro_rt) {
2260                         RTFREE(pktopt->ip6po_route.ro_rt);
2261                         pktopt->ip6po_route.ro_rt = NULL;
2262                 }
2263         }
2264         if (optname == -1 || optname == IPV6_DSTOPTS) {
2265                 if (pktopt->ip6po_dest2)
2266                         free(pktopt->ip6po_dest2, M_IP6OPT);
2267                 pktopt->ip6po_dest2 = NULL;
2268         }
2269 }
2270
2271 #define PKTOPT_EXTHDRCPY(type) \
2272 do {\
2273         if (src->type) {\
2274                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2275                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2276                 if (dst->type == NULL && canwait == M_NOWAIT)\
2277                         goto bad;\
2278                 bcopy(src->type, dst->type, hlen);\
2279         }\
2280 } while (/*CONSTCOND*/ 0)
2281
2282 static int
2283 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2284 {
2285         if (dst == NULL || src == NULL)  {
2286                 printf("ip6_clearpktopts: invalid argument\n");
2287                 return (EINVAL);
2288         }
2289
2290         dst->ip6po_hlim = src->ip6po_hlim;
2291         dst->ip6po_tclass = src->ip6po_tclass;
2292         dst->ip6po_flags = src->ip6po_flags;
2293         dst->ip6po_minmtu = src->ip6po_minmtu;
2294         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2295         if (src->ip6po_pktinfo) {
2296                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2297                     M_IP6OPT, canwait);
2298                 if (dst->ip6po_pktinfo == NULL)
2299                         goto bad;
2300                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2301         }
2302         if (src->ip6po_nexthop) {
2303                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2304                     M_IP6OPT, canwait);
2305                 if (dst->ip6po_nexthop == NULL)
2306                         goto bad;
2307                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2308                     src->ip6po_nexthop->sa_len);
2309         }
2310         PKTOPT_EXTHDRCPY(ip6po_hbh);
2311         PKTOPT_EXTHDRCPY(ip6po_dest1);
2312         PKTOPT_EXTHDRCPY(ip6po_dest2);
2313         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2314         return (0);
2315
2316   bad:
2317         ip6_clearpktopts(dst, -1);
2318         return (ENOBUFS);
2319 }
2320 #undef PKTOPT_EXTHDRCPY
2321
2322 struct ip6_pktopts *
2323 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2324 {
2325         int error;
2326         struct ip6_pktopts *dst;
2327
2328         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2329         if (dst == NULL)
2330                 return (NULL);
2331         ip6_initpktopts(dst);
2332
2333         if ((error = copypktopts(dst, src, canwait)) != 0) {
2334                 free(dst, M_IP6OPT);
2335                 return (NULL);
2336         }
2337
2338         return (dst);
2339 }
2340
2341 void
2342 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2343 {
2344         if (pktopt == NULL)
2345                 return;
2346
2347         ip6_clearpktopts(pktopt, -1);
2348
2349         free(pktopt, M_IP6OPT);
2350 }
2351
2352 /*
2353  * Set IPv6 outgoing packet options based on advanced API.
2354  */
2355 int
2356 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2357     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2358 {
2359         struct cmsghdr *cm = 0;
2360
2361         if (control == NULL || opt == NULL)
2362                 return (EINVAL);
2363
2364         ip6_initpktopts(opt);
2365         if (stickyopt) {
2366                 int error;
2367
2368                 /*
2369                  * If stickyopt is provided, make a local copy of the options
2370                  * for this particular packet, then override them by ancillary
2371                  * objects.
2372                  * XXX: copypktopts() does not copy the cached route to a next
2373                  * hop (if any).  This is not very good in terms of efficiency,
2374                  * but we can allow this since this option should be rarely
2375                  * used.
2376                  */
2377                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2378                         return (error);
2379         }
2380
2381         /*
2382          * XXX: Currently, we assume all the optional information is stored
2383          * in a single mbuf.
2384          */
2385         if (control->m_next)
2386                 return (EINVAL);
2387
2388         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2389             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2390                 int error;
2391
2392                 if (control->m_len < CMSG_LEN(0))
2393                         return (EINVAL);
2394
2395                 cm = mtod(control, struct cmsghdr *);
2396                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2397                         return (EINVAL);
2398                 if (cm->cmsg_level != IPPROTO_IPV6)
2399                         continue;
2400
2401                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2402                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2403                 if (error)
2404                         return (error);
2405         }
2406
2407         return (0);
2408 }
2409
2410 /*
2411  * Set a particular packet option, as a sticky option or an ancillary data
2412  * item.  "len" can be 0 only when it's a sticky option.
2413  * We have 4 cases of combination of "sticky" and "cmsg":
2414  * "sticky=0, cmsg=0": impossible
2415  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2416  * "sticky=1, cmsg=0": RFC3542 socket option
2417  * "sticky=1, cmsg=1": RFC2292 socket option
2418  */
2419 static int
2420 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2421     struct ucred *cred, int sticky, int cmsg, int uproto)
2422 {
2423         int minmtupolicy, preftemp;
2424         int error;
2425
2426         if (!sticky && !cmsg) {
2427 #ifdef DIAGNOSTIC
2428                 printf("ip6_setpktopt: impossible case\n");
2429 #endif
2430                 return (EINVAL);
2431         }
2432
2433         /*
2434          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2435          * not be specified in the context of RFC3542.  Conversely,
2436          * RFC3542 types should not be specified in the context of RFC2292.
2437          */
2438         if (!cmsg) {
2439                 switch (optname) {
2440                 case IPV6_2292PKTINFO:
2441                 case IPV6_2292HOPLIMIT:
2442                 case IPV6_2292NEXTHOP:
2443                 case IPV6_2292HOPOPTS:
2444                 case IPV6_2292DSTOPTS:
2445                 case IPV6_2292RTHDR:
2446                 case IPV6_2292PKTOPTIONS:
2447                         return (ENOPROTOOPT);
2448                 }
2449         }
2450         if (sticky && cmsg) {
2451                 switch (optname) {
2452                 case IPV6_PKTINFO:
2453                 case IPV6_HOPLIMIT:
2454                 case IPV6_NEXTHOP:
2455                 case IPV6_HOPOPTS:
2456                 case IPV6_DSTOPTS:
2457                 case IPV6_RTHDRDSTOPTS:
2458                 case IPV6_RTHDR:
2459                 case IPV6_USE_MIN_MTU:
2460                 case IPV6_DONTFRAG:
2461                 case IPV6_TCLASS:
2462                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2463                         return (ENOPROTOOPT);
2464                 }
2465         }
2466
2467         switch (optname) {
2468         case IPV6_2292PKTINFO:
2469         case IPV6_PKTINFO:
2470         {
2471                 struct ifnet *ifp = NULL;
2472                 struct in6_pktinfo *pktinfo;
2473
2474                 if (len != sizeof(struct in6_pktinfo))
2475                         return (EINVAL);
2476
2477                 pktinfo = (struct in6_pktinfo *)buf;
2478
2479                 /*
2480                  * An application can clear any sticky IPV6_PKTINFO option by
2481                  * doing a "regular" setsockopt with ipi6_addr being
2482                  * in6addr_any and ipi6_ifindex being zero.
2483                  * [RFC 3542, Section 6]
2484                  */
2485                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2486                     pktinfo->ipi6_ifindex == 0 &&
2487                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2488                         ip6_clearpktopts(opt, optname);
2489                         break;
2490                 }
2491
2492                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2493                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2494                         return (EINVAL);
2495                 }
2496
2497                 /* validate the interface index if specified. */
2498                 if (pktinfo->ipi6_ifindex > V_if_index ||
2499                     pktinfo->ipi6_ifindex < 0) {
2500                          return (ENXIO);
2501                 }
2502                 if (pktinfo->ipi6_ifindex) {
2503                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2504                         if (ifp == NULL)
2505                                 return (ENXIO);
2506                 }
2507
2508                 /*
2509                  * We store the address anyway, and let in6_selectsrc()
2510                  * validate the specified address.  This is because ipi6_addr
2511                  * may not have enough information about its scope zone, and
2512                  * we may need additional information (such as outgoing
2513                  * interface or the scope zone of a destination address) to
2514                  * disambiguate the scope.
2515                  * XXX: the delay of the validation may confuse the
2516                  * application when it is used as a sticky option.
2517                  */
2518                 if (opt->ip6po_pktinfo == NULL) {
2519                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2520                             M_IP6OPT, M_NOWAIT);
2521                         if (opt->ip6po_pktinfo == NULL)
2522                                 return (ENOBUFS);
2523                 }
2524                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2525                 break;
2526         }
2527
2528         case IPV6_2292HOPLIMIT:
2529         case IPV6_HOPLIMIT:
2530         {
2531                 int *hlimp;
2532
2533                 /*
2534                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2535                  * to simplify the ordering among hoplimit options.
2536                  */
2537                 if (optname == IPV6_HOPLIMIT && sticky)
2538                         return (ENOPROTOOPT);
2539
2540                 if (len != sizeof(int))
2541                         return (EINVAL);
2542                 hlimp = (int *)buf;
2543                 if (*hlimp < -1 || *hlimp > 255)
2544                         return (EINVAL);
2545
2546                 opt->ip6po_hlim = *hlimp;
2547                 break;
2548         }
2549
2550         case IPV6_TCLASS:
2551         {
2552                 int tclass;
2553
2554                 if (len != sizeof(int))
2555                         return (EINVAL);
2556                 tclass = *(int *)buf;
2557                 if (tclass < -1 || tclass > 255)
2558                         return (EINVAL);
2559
2560                 opt->ip6po_tclass = tclass;
2561                 break;
2562         }
2563
2564         case IPV6_2292NEXTHOP:
2565         case IPV6_NEXTHOP:
2566                 if (cred != NULL) {
2567                         error = priv_check_cred(cred,
2568                             PRIV_NETINET_SETHDROPTS, 0);
2569                         if (error)
2570                                 return (error);
2571                 }
2572
2573                 if (len == 0) { /* just remove the option */
2574                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2575                         break;
2576                 }
2577
2578                 /* check if cmsg_len is large enough for sa_len */
2579                 if (len < sizeof(struct sockaddr) || len < *buf)
2580                         return (EINVAL);
2581
2582                 switch (((struct sockaddr *)buf)->sa_family) {
2583                 case AF_INET6:
2584                 {
2585                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2586                         int error;
2587
2588                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2589                                 return (EINVAL);
2590
2591                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2592                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2593                                 return (EINVAL);
2594                         }
2595                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2596                             != 0) {
2597                                 return (error);
2598                         }
2599                         break;
2600                 }
2601                 case AF_LINK:   /* should eventually be supported */
2602                 default:
2603                         return (EAFNOSUPPORT);
2604                 }
2605
2606                 /* turn off the previous option, then set the new option. */
2607                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2608                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2609                 if (opt->ip6po_nexthop == NULL)
2610                         return (ENOBUFS);
2611                 bcopy(buf, opt->ip6po_nexthop, *buf);
2612                 break;
2613
2614         case IPV6_2292HOPOPTS:
2615         case IPV6_HOPOPTS:
2616         {
2617                 struct ip6_hbh *hbh;
2618                 int hbhlen;
2619
2620                 /*
2621                  * XXX: We don't allow a non-privileged user to set ANY HbH
2622                  * options, since per-option restriction has too much
2623                  * overhead.
2624                  */
2625                 if (cred != NULL) {
2626                         error = priv_check_cred(cred,
2627                             PRIV_NETINET_SETHDROPTS, 0);
2628                         if (error)
2629                                 return (error);
2630                 }
2631
2632                 if (len == 0) {
2633                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2634                         break;  /* just remove the option */
2635                 }
2636
2637                 /* message length validation */
2638                 if (len < sizeof(struct ip6_hbh))
2639                         return (EINVAL);
2640                 hbh = (struct ip6_hbh *)buf;
2641                 hbhlen = (hbh->ip6h_len + 1) << 3;
2642                 if (len != hbhlen)
2643                         return (EINVAL);
2644
2645                 /* turn off the previous option, then set the new option. */
2646                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2647                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2648                 if (opt->ip6po_hbh == NULL)
2649                         return (ENOBUFS);
2650                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2651
2652                 break;
2653         }
2654
2655         case IPV6_2292DSTOPTS:
2656         case IPV6_DSTOPTS:
2657         case IPV6_RTHDRDSTOPTS:
2658         {
2659                 struct ip6_dest *dest, **newdest = NULL;
2660                 int destlen;
2661
2662                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2663                         error = priv_check_cred(cred,
2664                             PRIV_NETINET_SETHDROPTS, 0);
2665                         if (error)
2666                                 return (error);
2667                 }
2668
2669                 if (len == 0) {
2670                         ip6_clearpktopts(opt, optname);
2671                         break;  /* just remove the option */
2672                 }
2673
2674                 /* message length validation */
2675                 if (len < sizeof(struct ip6_dest))
2676                         return (EINVAL);
2677                 dest = (struct ip6_dest *)buf;
2678                 destlen = (dest->ip6d_len + 1) << 3;
2679                 if (len != destlen)
2680                         return (EINVAL);
2681
2682                 /*
2683                  * Determine the position that the destination options header
2684                  * should be inserted; before or after the routing header.
2685                  */
2686                 switch (optname) {
2687                 case IPV6_2292DSTOPTS:
2688                         /*
2689                          * The old advacned API is ambiguous on this point.
2690                          * Our approach is to determine the position based
2691                          * according to the existence of a routing header.
2692                          * Note, however, that this depends on the order of the
2693                          * extension headers in the ancillary data; the 1st
2694                          * part of the destination options header must appear
2695                          * before the routing header in the ancillary data,
2696                          * too.
2697                          * RFC3542 solved the ambiguity by introducing
2698                          * separate ancillary data or option types.
2699                          */
2700                         if (opt->ip6po_rthdr == NULL)
2701                                 newdest = &opt->ip6po_dest1;
2702                         else
2703                                 newdest = &opt->ip6po_dest2;
2704                         break;
2705                 case IPV6_RTHDRDSTOPTS:
2706                         newdest = &opt->ip6po_dest1;
2707                         break;
2708                 case IPV6_DSTOPTS:
2709                         newdest = &opt->ip6po_dest2;
2710                         break;
2711                 }
2712
2713                 /* turn off the previous option, then set the new option. */
2714                 ip6_clearpktopts(opt, optname);
2715                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2716                 if (*newdest == NULL)
2717                         return (ENOBUFS);
2718                 bcopy(dest, *newdest, destlen);
2719
2720                 break;
2721         }
2722
2723         case IPV6_2292RTHDR:
2724         case IPV6_RTHDR:
2725         {
2726                 struct ip6_rthdr *rth;
2727                 int rthlen;
2728
2729                 if (len == 0) {
2730                         ip6_clearpktopts(opt, IPV6_RTHDR);
2731                         break;  /* just remove the option */
2732                 }
2733
2734                 /* message length validation */
2735                 if (len < sizeof(struct ip6_rthdr))
2736                         return (EINVAL);
2737                 rth = (struct ip6_rthdr *)buf;
2738                 rthlen = (rth->ip6r_len + 1) << 3;
2739                 if (len != rthlen)
2740                         return (EINVAL);
2741
2742                 switch (rth->ip6r_type) {
2743                 case IPV6_RTHDR_TYPE_0:
2744                         if (rth->ip6r_len == 0) /* must contain one addr */
2745                                 return (EINVAL);
2746                         if (rth->ip6r_len % 2) /* length must be even */
2747                                 return (EINVAL);
2748                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2749                                 return (EINVAL);
2750                         break;
2751                 default:
2752                         return (EINVAL);        /* not supported */
2753                 }
2754
2755                 /* turn off the previous option */
2756                 ip6_clearpktopts(opt, IPV6_RTHDR);
2757                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2758                 if (opt->ip6po_rthdr == NULL)
2759                         return (ENOBUFS);
2760                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2761
2762                 break;
2763         }
2764
2765         case IPV6_USE_MIN_MTU:
2766                 if (len != sizeof(int))
2767                         return (EINVAL);
2768                 minmtupolicy = *(int *)buf;
2769                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2770                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2771                     minmtupolicy != IP6PO_MINMTU_ALL) {
2772                         return (EINVAL);
2773                 }
2774                 opt->ip6po_minmtu = minmtupolicy;
2775                 break;
2776
2777         case IPV6_DONTFRAG:
2778                 if (len != sizeof(int))
2779                         return (EINVAL);
2780
2781                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2782                         /*
2783                          * we ignore this option for TCP sockets.
2784                          * (RFC3542 leaves this case unspecified.)
2785                          */
2786                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2787                 } else
2788                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2789                 break;
2790
2791         case IPV6_PREFER_TEMPADDR:
2792                 if (len != sizeof(int))
2793                         return (EINVAL);
2794                 preftemp = *(int *)buf;
2795                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2796                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2797                     preftemp != IP6PO_TEMPADDR_PREFER) {
2798                         return (EINVAL);
2799                 }
2800                 opt->ip6po_prefer_tempaddr = preftemp;
2801                 break;
2802
2803         default:
2804                 return (ENOPROTOOPT);
2805         } /* end of switch */
2806
2807         return (0);
2808 }
2809
2810 /*
2811  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2812  * packet to the input queue of a specified interface.  Note that this
2813  * calls the output routine of the loopback "driver", but with an interface
2814  * pointer that might NOT be &loif -- easier than replicating that code here.
2815  */
2816 void
2817 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2818 {
2819         struct mbuf *copym;
2820         struct ip6_hdr *ip6;
2821
2822         copym = m_copy(m, 0, M_COPYALL);
2823         if (copym == NULL)
2824                 return;
2825
2826         /*
2827          * Make sure to deep-copy IPv6 header portion in case the data
2828          * is in an mbuf cluster, so that we can safely override the IPv6
2829          * header portion later.
2830          */
2831         if ((copym->m_flags & M_EXT) != 0 ||
2832             copym->m_len < sizeof(struct ip6_hdr)) {
2833                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2834                 if (copym == NULL)
2835                         return;
2836         }
2837         ip6 = mtod(copym, struct ip6_hdr *);
2838         /*
2839          * clear embedded scope identifiers if necessary.
2840          * in6_clearscope will touch the addresses only when necessary.
2841          */
2842         in6_clearscope(&ip6->ip6_src);
2843         in6_clearscope(&ip6->ip6_dst);
2844         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
2845                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
2846                     CSUM_PSEUDO_HDR;
2847                 copym->m_pkthdr.csum_data = 0xffff;
2848         }
2849         (void)if_simloop(ifp, copym, dst->sin6_family, 0);
2850 }
2851
2852 /*
2853  * Chop IPv6 header off from the payload.
2854  */
2855 static int
2856 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2857 {
2858         struct mbuf *mh;
2859         struct ip6_hdr *ip6;
2860
2861         ip6 = mtod(m, struct ip6_hdr *);
2862         if (m->m_len > sizeof(*ip6)) {
2863                 mh = m_gethdr(M_NOWAIT, MT_DATA);
2864                 if (mh == NULL) {
2865                         m_freem(m);
2866                         return ENOBUFS;
2867                 }
2868                 m_move_pkthdr(mh, m);
2869                 MH_ALIGN(mh, sizeof(*ip6));
2870                 m->m_len -= sizeof(*ip6);
2871                 m->m_data += sizeof(*ip6);
2872                 mh->m_next = m;
2873                 m = mh;
2874                 m->m_len = sizeof(*ip6);
2875                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2876         }
2877         exthdrs->ip6e_ip6 = m;
2878         return 0;
2879 }
2880
2881 /*
2882  * Compute IPv6 extension header length.
2883  */
2884 int
2885 ip6_optlen(struct inpcb *in6p)
2886 {
2887         int len;
2888
2889         if (!in6p->in6p_outputopts)
2890                 return 0;
2891
2892         len = 0;
2893 #define elen(x) \
2894     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2895
2896         len += elen(in6p->in6p_outputopts->ip6po_hbh);
2897         if (in6p->in6p_outputopts->ip6po_rthdr)
2898                 /* dest1 is valid with rthdr only */
2899                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
2900         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2901         len += elen(in6p->in6p_outputopts->ip6po_dest2);
2902         return len;
2903 #undef elen
2904 }