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