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_ipsec.h"
  69 #include "opt_sctp.h"
  70 #include "opt_route.h"
  71 #include "opt_rss.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/if_var.h>
  90 #include <net/if_llatbl.h>
  91 #include <net/netisr.h>
  92 #include <net/route.h>
  93 #include <net/pfil.h>
  94 #include <net/rss_config.h>
  95 #include <net/vnet.h>
  96
  97 #include <netinet/in.h>
  98 #include <netinet/in_var.h>
  99 #include <netinet/ip_var.h>
 100 #include <netinet6/in6_fib.h>
 101 #include <netinet6/in6_var.h>
 102 #include <netinet/ip6.h>
 103 #include <netinet/icmp6.h>
 104 #include <netinet6/ip6_var.h>
 105 #include <netinet/in_pcb.h>
 106 #include <netinet/tcp_var.h>
 107 #include <netinet6/nd6.h>
 108 #include <netinet6/in6_rss.h>
 109
 110 #ifdef IPSEC
 111 #include <netipsec/ipsec.h>
 112 #include <netipsec/ipsec6.h>
 113 #include <netipsec/key.h>
 114 #include <netinet6/ip6_ipsec.h>
 115 #endif /* IPSEC */
 116 #ifdef SCTP
 117 #include <netinet/sctp.h>
 118 #include <netinet/sctp_crc32.h>
 119 #endif
 120
 121 #include <netinet6/ip6protosw.h>
 122 #include <netinet6/scope6_var.h>
 123
 124 #ifdef FLOWTABLE
 125 #include <net/flowtable.h>
 126 #endif
 127
 128 extern int in6_mcast_loop;
 129
 130 struct ip6_exthdrs {
 131         struct mbuf *ip6e_ip6;
 132         struct mbuf *ip6e_hbh;
 133         struct mbuf *ip6e_dest1;
 134         struct mbuf *ip6e_rthdr;
 135         struct mbuf *ip6e_dest2;
 136 };
 137
 138 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
 139
 140 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
 141                            struct ucred *, int);
 142 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
 143         struct socket *, struct sockopt *);
 144 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 145 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
 146         struct ucred *, int, int, int);
 147
 148 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 149 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 150         struct ip6_frag **);
 151 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 152 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 153 static int ip6_getpmtu(struct route_in6 *, int,
 154         struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
 155         u_int);
 156 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
 157         u_long *, int *, u_int);
 158 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
 159 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 160
 161
 162 /*
 163  * Make an extension header from option data.  hp is the source, and
 164  * mp is the destination.
 165  */
 166 #define MAKE_EXTHDR(hp, mp)                                             \
 167     do {                                                                \
 168         if (hp) {                                                       \
 169                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 170                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 171                     ((eh)->ip6e_len + 1) << 3);                         \
 172                 if (error)                                              \
 173                         goto freehdrs;                                  \
 174         }                                                               \
 175     } while (/*CONSTCOND*/ 0)
 176
 177 /*
 178  * Form a chain of extension headers.
 179  * m is the extension header mbuf
 180  * mp is the previous mbuf in the chain
 181  * p is the next header
 182  * i is the type of option.
 183  */
 184 #define MAKE_CHAIN(m, mp, p, i)\
 185     do {\
 186         if (m) {\
 187                 if (!hdrsplit) \
 188                         panic("assumption failed: hdr not split"); \
 189                 *mtod((m), u_char *) = *(p);\
 190                 *(p) = (i);\
 191                 p = mtod((m), u_char *);\
 192                 (m)->m_next = (mp)->m_next;\
 193                 (mp)->m_next = (m);\
 194                 (mp) = (m);\
 195         }\
 196     } while (/*CONSTCOND*/ 0)
 197
 198 void
 199 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 200 {
 201         u_short csum;
 202
 203         csum = in_cksum_skip(m, offset + plen, offset);
 204         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 205                 csum = 0xffff;
 206         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 207
 208         if (offset + sizeof(u_short) > m->m_len) {
 209                 printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
 210                     "csum_flags=%b\n", __func__, m->m_len, plen, offset,
 211                     (int)m->m_pkthdr.csum_flags, CSUM_BITS);
 212                 /*
 213                  * XXX this should not happen, but if it does, the correct
 214                  * behavior may be to insert the checksum in the appropriate
 215                  * next mbuf in the chain.
 216                  */
 217                 return;
 218         }
 219         *(u_short *)(m->m_data + offset) = csum;
 220 }
 221
 222 int
 223 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
 224     int mtu, uint32_t id)
 225 {
 226         struct mbuf *m, **mnext, *m_frgpart;
 227         struct ip6_hdr *ip6, *mhip6;
 228         struct ip6_frag *ip6f;
 229         int off;
 230         int error;
 231         int tlen = m0->m_pkthdr.len;
 232
 233         m = m0;
 234         ip6 = mtod(m, struct ip6_hdr *);
 235         mnext = &m->m_nextpkt;
 236
 237         for (off = hlen; off < tlen; off += mtu) {
 238                 m = m_gethdr(M_NOWAIT, MT_DATA);
 239                 if (!m) {
 240                         IP6STAT_INC(ip6s_odropped);
 241                         return (ENOBUFS);
 242                 }
 243                 m->m_flags = m0->m_flags & M_COPYFLAGS;
 244                 *mnext = m;
 245                 mnext = &m->m_nextpkt;
 246                 m->m_data += max_linkhdr;
 247                 mhip6 = mtod(m, struct ip6_hdr *);
 248                 *mhip6 = *ip6;
 249                 m->m_len = sizeof(*mhip6);
 250                 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 251                 if (error) {
 252                         IP6STAT_INC(ip6s_odropped);
 253                         return (error);
 254                 }
 255                 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 256                 if (off + mtu >= tlen)
 257                         mtu = tlen - off;
 258                 else
 259                         ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 260                 mhip6->ip6_plen = htons((u_short)(mtu + hlen +
 261                     sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 262                 if ((m_frgpart = m_copym(m0, off, mtu, M_NOWAIT)) == NULL) {
 263                         IP6STAT_INC(ip6s_odropped);
 264                         return (ENOBUFS);
 265                 }
 266                 m_cat(m, m_frgpart);
 267                 m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
 268                 m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
 269                 m->m_pkthdr.rcvif = NULL;
 270                 ip6f->ip6f_reserved = 0;
 271                 ip6f->ip6f_ident = id;
 272                 ip6f->ip6f_nxt = nextproto;
 273                 IP6STAT_INC(ip6s_ofragments);
 274                 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 275         }
 276
 277         return (0);
 278 }
 279
 280 /*
 281  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 282  * header (with pri, len, nxt, hlim, src, dst).
 283  * This function may modify ver and hlim only.
 284  * The mbuf chain containing the packet will be freed.
 285  * The mbuf opt, if present, will not be freed.
 286  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
 287  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 288  * then result of route lookup is stored in ro->ro_rt.
 289  *
 290  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
 291  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 292  * which is rt_mtu.
 293  *
 294  * ifpp - XXX: just for statistics
 295  */
 296 /*
 297  * XXX TODO: no flowid is assigned for outbound flows?
 298  */
 299 int
 300 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 301     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 302     struct ifnet **ifpp, struct inpcb *inp)
 303 {
 304         struct ip6_hdr *ip6;
 305         struct ifnet *ifp, *origifp;
 306         struct mbuf *m = m0;
 307         struct mbuf *mprev = NULL;
 308         int hlen, tlen, len;
 309         struct route_in6 ip6route;
 310         struct rtentry *rt = NULL;
 311         struct sockaddr_in6 *dst, src_sa, dst_sa;
 312         struct in6_addr odst;
 313         int error = 0;
 314         struct in6_ifaddr *ia = NULL;
 315         u_long mtu;
 316         int alwaysfrag, dontfrag;
 317         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 318         struct ip6_exthdrs exthdrs;
 319         struct in6_addr src0, dst0;
 320         u_int32_t zone;
 321         struct route_in6 *ro_pmtu = NULL;
 322         int hdrsplit = 0;
 323         int sw_csum, tso;
 324         int needfiblookup;
 325         uint32_t fibnum;
 326         struct m_tag *fwd_tag = NULL;
 327         uint32_t id;
 328
 329         if (inp != NULL) {
 330                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 331                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 332                         /* unconditionally set flowid */
 333                         m->m_pkthdr.flowid = inp->inp_flowid;
 334                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 335                 }
 336         }
 337
 338         bzero(&exthdrs, sizeof(exthdrs));
 339         if (opt) {
 340                 /* Hop-by-Hop options header */
 341                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 342                 /* Destination options header(1st part) */
 343                 if (opt->ip6po_rthdr) {
 344                         /*
 345                          * Destination options header(1st part)
 346                          * This only makes sense with a routing header.
 347                          * See Section 9.2 of RFC 3542.
 348                          * Disabling this part just for MIP6 convenience is
 349                          * a bad idea.  We need to think carefully about a
 350                          * way to make the advanced API coexist with MIP6
 351                          * options, which might automatically be inserted in
 352                          * the kernel.
 353                          */
 354                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 355                 }
 356                 /* Routing header */
 357                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 358                 /* Destination options header(2nd part) */
 359                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 360         }
 361
 362 #ifdef IPSEC
 363         /*
 364          * IPSec checking which handles several cases.
 365          * FAST IPSEC: We re-injected the packet.
 366          * XXX: need scope argument.
 367          */
 368         switch(ip6_ipsec_output(&m, inp, &error))
 369         {
 370         case 1:                 /* Bad packet */
 371                 goto freehdrs;
 372         case -1:                /* IPSec done */
 373                 goto done;
 374         case 0:                 /* No IPSec */
 375         default:
 376                 break;
 377         }
 378 #endif /* IPSEC */
 379
 380         /*
 381          * Calculate the total length of the extension header chain.
 382          * Keep the length of the unfragmentable part for fragmentation.
 383          */
 384         optlen = 0;
 385         if (exthdrs.ip6e_hbh)
 386                 optlen += exthdrs.ip6e_hbh->m_len;
 387         if (exthdrs.ip6e_dest1)
 388                 optlen += exthdrs.ip6e_dest1->m_len;
 389         if (exthdrs.ip6e_rthdr)
 390                 optlen += exthdrs.ip6e_rthdr->m_len;
 391         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 392
 393         /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
 394         if (exthdrs.ip6e_dest2)
 395                 optlen += exthdrs.ip6e_dest2->m_len;
 396
 397         /*
 398          * If there is at least one extension header,
 399          * separate IP6 header from the payload.
 400          */
 401         if (optlen && !hdrsplit) {
 402                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 403                         m = NULL;
 404                         goto freehdrs;
 405                 }
 406                 m = exthdrs.ip6e_ip6;
 407                 hdrsplit++;
 408         }
 409
 410         ip6 = mtod(m, struct ip6_hdr *);
 411
 412         /* adjust mbuf packet header length */
 413         m->m_pkthdr.len += optlen;
 414         plen = m->m_pkthdr.len - sizeof(*ip6);
 415
 416         /* If this is a jumbo payload, insert a jumbo payload option. */
 417         if (plen > IPV6_MAXPACKET) {
 418                 if (!hdrsplit) {
 419                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 420                                 m = NULL;
 421                                 goto freehdrs;
 422                         }
 423                         m = exthdrs.ip6e_ip6;
 424                         hdrsplit++;
 425                 }
 426                 /* adjust pointer */
 427                 ip6 = mtod(m, struct ip6_hdr *);
 428                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 429                         goto freehdrs;
 430                 ip6->ip6_plen = 0;
 431         } else
 432                 ip6->ip6_plen = htons(plen);
 433
 434         /*
 435          * Concatenate headers and fill in next header fields.
 436          * Here we have, on "m"
 437          *      IPv6 payload
 438          * and we insert headers accordingly.  Finally, we should be getting:
 439          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 440          *
 441          * during the header composing process, "m" points to IPv6 header.
 442          * "mprev" points to an extension header prior to esp.
 443          */
 444         u_char *nexthdrp = &ip6->ip6_nxt;
 445         mprev = m;
 446
 447         /*
 448          * we treat dest2 specially.  this makes IPsec processing
 449          * much easier.  the goal here is to make mprev point the
 450          * mbuf prior to dest2.
 451          *
 452          * result: IPv6 dest2 payload
 453          * m and mprev will point to IPv6 header.
 454          */
 455         if (exthdrs.ip6e_dest2) {
 456                 if (!hdrsplit)
 457                         panic("assumption failed: hdr not split");
 458                 exthdrs.ip6e_dest2->m_next = m->m_next;
 459                 m->m_next = exthdrs.ip6e_dest2;
 460                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 461                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 462         }
 463
 464         /*
 465          * result: IPv6 hbh dest1 rthdr dest2 payload
 466          * m will point to IPv6 header.  mprev will point to the
 467          * extension header prior to dest2 (rthdr in the above case).
 468          */
 469         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 470         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 471                    IPPROTO_DSTOPTS);
 472         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 473                    IPPROTO_ROUTING);
 474
 475         /*
 476          * If there is a routing header, discard the packet.
 477          */
 478         if (exthdrs.ip6e_rthdr) {
 479                  error = EINVAL;
 480                  goto bad;
 481         }
 482
 483         /* Source address validation */
 484         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 485             (flags & IPV6_UNSPECSRC) == 0) {
 486                 error = EOPNOTSUPP;
 487                 IP6STAT_INC(ip6s_badscope);
 488                 goto bad;
 489         }
 490         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 491                 error = EOPNOTSUPP;
 492                 IP6STAT_INC(ip6s_badscope);
 493                 goto bad;
 494         }
 495
 496         IP6STAT_INC(ip6s_localout);
 497
 498         /*
 499          * Route packet.
 500          */
 501         if (ro == NULL) {
 502                 ro = &ip6route;
 503                 bzero((caddr_t)ro, sizeof(*ro));
 504         } else
 505                 ro->ro_flags |= RT_LLE_CACHE;
 506         ro_pmtu = ro;
 507         if (opt && opt->ip6po_rthdr)
 508                 ro = &opt->ip6po_route;
 509         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 510 #ifdef FLOWTABLE
 511         if (ro->ro_rt == NULL)
 512                 (void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
 513 #endif
 514         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 515 again:
 516         /*
 517          * if specified, try to fill in the traffic class field.
 518          * do not override if a non-zero value is already set.
 519          * we check the diffserv field and the ecn field separately.
 520          */
 521         if (opt && opt->ip6po_tclass >= 0) {
 522                 int mask = 0;
 523
 524                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 525                         mask |= 0xfc;
 526                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 527                         mask |= 0x03;
 528                 if (mask != 0)
 529                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 530         }
 531
 532         /* fill in or override the hop limit field, if necessary. */
 533         if (opt && opt->ip6po_hlim != -1)
 534                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 535         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 536                 if (im6o != NULL)
 537                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 538                 else
 539                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 540         }
 541         /*
 542          * Validate route against routing table additions;
 543          * a better/more specific route might have been added.
 544          * Make sure address family is set in route.
 545          */
 546         if (inp) {
 547                 ro->ro_dst.sin6_family = AF_INET6;
 548                 RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
 549         }
 550         if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
 551             ro->ro_dst.sin6_family == AF_INET6 &&
 552             IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
 553                 rt = ro->ro_rt;
 554                 ifp = ro->ro_rt->rt_ifp;
 555         } else {
 556                 if (ro->ro_lle)
 557                         LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 558                 ro->ro_lle = NULL;
 559                 if (fwd_tag == NULL) {
 560                         bzero(&dst_sa, sizeof(dst_sa));
 561                         dst_sa.sin6_family = AF_INET6;
 562                         dst_sa.sin6_len = sizeof(dst_sa);
 563                         dst_sa.sin6_addr = ip6->ip6_dst;
 564                 }
 565                 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
 566                     &rt, fibnum);
 567                 if (error != 0) {
 568                         if (ifp != NULL)
 569                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 570                         goto bad;
 571                 }
 572         }
 573         if (rt == NULL) {
 574                 /*
 575                  * If in6_selectroute() does not return a route entry,
 576                  * dst may not have been updated.
 577                  */
 578                 *dst = dst_sa;  /* XXX */
 579         }
 580
 581         /*
 582          * then rt (for unicast) and ifp must be non-NULL valid values.
 583          */
 584         if ((flags & IPV6_FORWARDING) == 0) {
 585                 /* XXX: the FORWARDING flag can be set for mrouting. */
 586                 in6_ifstat_inc(ifp, ifs6_out_request);
 587         }
 588         if (rt != NULL) {
 589                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 590                 counter_u64_add(rt->rt_pksent, 1);
 591         }
 592
 593
 594         /*
 595          * The outgoing interface must be in the zone of source and
 596          * destination addresses.
 597          */
 598         origifp = ifp;
 599
 600         src0 = ip6->ip6_src;
 601         if (in6_setscope(&src0, origifp, &zone))
 602                 goto badscope;
 603         bzero(&src_sa, sizeof(src_sa));
 604         src_sa.sin6_family = AF_INET6;
 605         src_sa.sin6_len = sizeof(src_sa);
 606         src_sa.sin6_addr = ip6->ip6_src;
 607         if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
 608                 goto badscope;
 609
 610         dst0 = ip6->ip6_dst;
 611         if (in6_setscope(&dst0, origifp, &zone))
 612                 goto badscope;
 613         /* re-initialize to be sure */
 614         bzero(&dst_sa, sizeof(dst_sa));
 615         dst_sa.sin6_family = AF_INET6;
 616         dst_sa.sin6_len = sizeof(dst_sa);
 617         dst_sa.sin6_addr = ip6->ip6_dst;
 618         if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
 619                 goto badscope;
 620         }
 621
 622         /* We should use ia_ifp to support the case of
 623          * sending packets to an address of our own.
 624          */
 625         if (ia != NULL && ia->ia_ifp)
 626                 ifp = ia->ia_ifp;
 627
 628         /* scope check is done. */
 629         goto routefound;
 630
 631   badscope:
 632         IP6STAT_INC(ip6s_badscope);
 633         in6_ifstat_inc(origifp, ifs6_out_discard);
 634         if (error == 0)
 635                 error = EHOSTUNREACH; /* XXX */
 636         goto bad;
 637
 638   routefound:
 639         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 640                 if (opt && opt->ip6po_nextroute.ro_rt) {
 641                         /*
 642                          * The nexthop is explicitly specified by the
 643                          * application.  We assume the next hop is an IPv6
 644                          * address.
 645                          */
 646                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 647                 }
 648                 else if ((rt->rt_flags & RTF_GATEWAY))
 649                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 650         }
 651
 652         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 653                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 654         } else {
 655                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 656                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 657                 /*
 658                  * Confirm that the outgoing interface supports multicast.
 659                  */
 660                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 661                         IP6STAT_INC(ip6s_noroute);
 662                         in6_ifstat_inc(ifp, ifs6_out_discard);
 663                         error = ENETUNREACH;
 664                         goto bad;
 665                 }
 666                 if ((im6o == NULL && in6_mcast_loop) ||
 667                     (im6o && im6o->im6o_multicast_loop)) {
 668                         /*
 669                          * Loop back multicast datagram if not expressly
 670                          * forbidden to do so, even if we have not joined
 671                          * the address; protocols will filter it later,
 672                          * thus deferring a hash lookup and lock acquisition
 673                          * at the expense of an m_copym().
 674                          */
 675                         ip6_mloopback(ifp, m);
 676                 } else {
 677                         /*
 678                          * If we are acting as a multicast router, perform
 679                          * multicast forwarding as if the packet had just
 680                          * arrived on the interface to which we are about
 681                          * to send.  The multicast forwarding function
 682                          * recursively calls this function, using the
 683                          * IPV6_FORWARDING flag to prevent infinite recursion.
 684                          *
 685                          * Multicasts that are looped back by ip6_mloopback(),
 686                          * above, will be forwarded by the ip6_input() routine,
 687                          * if necessary.
 688                          */
 689                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 690                                 /*
 691                                  * XXX: ip6_mforward expects that rcvif is NULL
 692                                  * when it is called from the originating path.
 693                                  * However, it may not always be the case.
 694                                  */
 695                                 m->m_pkthdr.rcvif = NULL;
 696                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 697                                         m_freem(m);
 698                                         goto done;
 699                                 }
 700                         }
 701                 }
 702                 /*
 703                  * Multicasts with a hoplimit of zero may be looped back,
 704                  * above, but must not be transmitted on a network.
 705                  * Also, multicasts addressed to the loopback interface
 706                  * are not sent -- the above call to ip6_mloopback() will
 707                  * loop back a copy if this host actually belongs to the
 708                  * destination group on the loopback interface.
 709                  */
 710                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 711                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 712                         m_freem(m);
 713                         goto done;
 714                 }
 715         }
 716
 717         /*
 718          * Fill the outgoing inteface to tell the upper layer
 719          * to increment per-interface statistics.
 720          */
 721         if (ifpp)
 722                 *ifpp = ifp;
 723
 724         /* Determine path MTU. */
 725         if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
 726                     &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
 727                 goto bad;
 728
 729         /*
 730          * The caller of this function may specify to use the minimum MTU
 731          * in some cases.
 732          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 733          * setting.  The logic is a bit complicated; by default, unicast
 734          * packets will follow path MTU while multicast packets will be sent at
 735          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 736          * including unicast ones will be sent at the minimum MTU.  Multicast
 737          * packets will always be sent at the minimum MTU unless
 738          * IP6PO_MINMTU_DISABLE is explicitly specified.
 739          * See RFC 3542 for more details.
 740          */
 741         if (mtu > IPV6_MMTU) {
 742                 if ((flags & IPV6_MINMTU))
 743                         mtu = IPV6_MMTU;
 744                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 745                         mtu = IPV6_MMTU;
 746                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 747                          (opt == NULL ||
 748                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 749                         mtu = IPV6_MMTU;
 750                 }
 751         }
 752
 753         /*
 754          * clear embedded scope identifiers if necessary.
 755          * in6_clearscope will touch the addresses only when necessary.
 756          */
 757         in6_clearscope(&ip6->ip6_src);
 758         in6_clearscope(&ip6->ip6_dst);
 759
 760         /*
 761          * If the outgoing packet contains a hop-by-hop options header,
 762          * it must be examined and processed even by the source node.
 763          * (RFC 2460, section 4.)
 764          */
 765         if (exthdrs.ip6e_hbh) {
 766                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 767                 u_int32_t dummy; /* XXX unused */
 768                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 769
 770 #ifdef DIAGNOSTIC
 771                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 772                         panic("ip6e_hbh is not contiguous");
 773 #endif
 774                 /*
 775                  *  XXX: if we have to send an ICMPv6 error to the sender,
 776                  *       we need the M_LOOP flag since icmp6_error() expects
 777                  *       the IPv6 and the hop-by-hop options header are
 778                  *       contiguous unless the flag is set.
 779                  */
 780                 m->m_flags |= M_LOOP;
 781                 m->m_pkthdr.rcvif = ifp;
 782                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 783                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 784                     &dummy, &plen) < 0) {
 785                         /* m was already freed at this point */
 786                         error = EINVAL;/* better error? */
 787                         goto done;
 788                 }
 789                 m->m_flags &= ~M_LOOP; /* XXX */
 790                 m->m_pkthdr.rcvif = NULL;
 791         }
 792
 793         /* Jump over all PFIL processing if hooks are not active. */
 794         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 795                 goto passout;
 796
 797         odst = ip6->ip6_dst;
 798         /* Run through list of hooks for output packets. */
 799         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
 800         if (error != 0 || m == NULL)
 801                 goto done;
 802         /* adjust pointer */
 803         ip6 = mtod(m, struct ip6_hdr *);
 804
 805         needfiblookup = 0;
 806         /* See if destination IP address was changed by packet filter. */
 807         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 808                 m->m_flags |= M_SKIP_FIREWALL;
 809                 /* If destination is now ourself drop to ip6_input(). */
 810                 if (in6_localip(&ip6->ip6_dst)) {
 811                         m->m_flags |= M_FASTFWD_OURS;
 812                         if (m->m_pkthdr.rcvif == NULL)
 813                                 m->m_pkthdr.rcvif = V_loif;
 814                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 815                                 m->m_pkthdr.csum_flags |=
 816                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 817                                 m->m_pkthdr.csum_data = 0xffff;
 818                         }
 819 #ifdef SCTP
 820                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 821                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 822 #endif
 823                         error = netisr_queue(NETISR_IPV6, m);
 824                         goto done;
 825                 } else {
 826                         RO_RTFREE(ro);
 827                         needfiblookup = 1; /* Redo the routing table lookup. */
 828                         if (ro->ro_lle)
 829                                 LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 830                         ro->ro_lle = NULL;
 831                 }
 832         }
 833         /* See if fib was changed by packet filter. */
 834         if (fibnum != M_GETFIB(m)) {
 835                 m->m_flags |= M_SKIP_FIREWALL;
 836                 fibnum = M_GETFIB(m);
 837                 RO_RTFREE(ro);
 838                 needfiblookup = 1;
 839                 if (ro->ro_lle)
 840                         LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 841                 ro->ro_lle = NULL;
 842         }
 843         if (needfiblookup)
 844                 goto again;
 845
 846         /* See if local, if yes, send it to netisr. */
 847         if (m->m_flags & M_FASTFWD_OURS) {
 848                 if (m->m_pkthdr.rcvif == NULL)
 849                         m->m_pkthdr.rcvif = V_loif;
 850                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 851                         m->m_pkthdr.csum_flags |=
 852                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 853                         m->m_pkthdr.csum_data = 0xffff;
 854                 }
 855 #ifdef SCTP
 856                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 857                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 858 #endif
 859                 error = netisr_queue(NETISR_IPV6, m);
 860                 goto done;
 861         }
 862         /* Or forward to some other address? */
 863         if ((m->m_flags & M_IP6_NEXTHOP) &&
 864             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 865                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 866                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
 867                 m->m_flags |= M_SKIP_FIREWALL;
 868                 m->m_flags &= ~M_IP6_NEXTHOP;
 869                 m_tag_delete(m, fwd_tag);
 870                 goto again;
 871         }
 872
 873 passout:
 874         /*
 875          * Send the packet to the outgoing interface.
 876          * If necessary, do IPv6 fragmentation before sending.
 877          *
 878          * the logic here is rather complex:
 879          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 880          * 1-a: send as is if tlen <= path mtu
 881          * 1-b: fragment if tlen > path mtu
 882          *
 883          * 2: if user asks us not to fragment (dontfrag == 1)
 884          * 2-a: send as is if tlen <= interface mtu
 885          * 2-b: error if tlen > interface mtu
 886          *
 887          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 888          *      always fragment
 889          *
 890          * 4: if dontfrag == 1 && alwaysfrag == 1
 891          *      error, as we cannot handle this conflicting request
 892          */
 893         sw_csum = m->m_pkthdr.csum_flags;
 894         if (!hdrsplit) {
 895                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 896                 sw_csum &= ~ifp->if_hwassist;
 897         } else
 898                 tso = 0;
 899         /*
 900          * If we added extension headers, we will not do TSO and calculate the
 901          * checksums ourselves for now.
 902          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 903          * with ext. hdrs.
 904          */
 905         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 906                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 907                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 908         }
 909 #ifdef SCTP
 910         if (sw_csum & CSUM_SCTP_IPV6) {
 911                 sw_csum &= ~CSUM_SCTP_IPV6;
 912                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 913         }
 914 #endif
 915         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 916         tlen = m->m_pkthdr.len;
 917
 918         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 919                 dontfrag = 1;
 920         else
 921                 dontfrag = 0;
 922         if (dontfrag && alwaysfrag) {   /* case 4 */
 923                 /* conflicting request - can't transmit */
 924                 error = EMSGSIZE;
 925                 goto bad;
 926         }
 927         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 928                 /*
 929                  * Even if the DONTFRAG option is specified, we cannot send the
 930                  * packet when the data length is larger than the MTU of the
 931                  * outgoing interface.
 932                  * Notify the error by sending IPV6_PATHMTU ancillary data if
 933                  * application wanted to know the MTU value. Also return an
 934                  * error code (this is not described in the API spec).
 935                  */
 936                 if (inp != NULL)
 937                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
 938                 error = EMSGSIZE;
 939                 goto bad;
 940         }
 941
 942         /*
 943          * transmit packet without fragmentation
 944          */
 945         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 946                 struct in6_ifaddr *ia6;
 947
 948                 ip6 = mtod(m, struct ip6_hdr *);
 949                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 950                 if (ia6) {
 951                         /* Record statistics for this interface address. */
 952                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
 953                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
 954                             m->m_pkthdr.len);
 955                         ifa_free(&ia6->ia_ifa);
 956                 }
 957                 error = nd6_output_ifp(ifp, origifp, m, dst,
 958                     (struct route *)ro);
 959                 goto done;
 960         }
 961
 962         /*
 963          * try to fragment the packet.  case 1-b and 3
 964          */
 965         if (mtu < IPV6_MMTU) {
 966                 /* path MTU cannot be less than IPV6_MMTU */
 967                 error = EMSGSIZE;
 968                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 969                 goto bad;
 970         } else if (ip6->ip6_plen == 0) {
 971                 /* jumbo payload cannot be fragmented */
 972                 error = EMSGSIZE;
 973                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 974                 goto bad;
 975         } else {
 976                 u_char nextproto;
 977
 978                 /*
 979                  * Too large for the destination or interface;
 980                  * fragment if possible.
 981                  * Must be able to put at least 8 bytes per fragment.
 982                  */
 983                 hlen = unfragpartlen;
 984                 if (mtu > IPV6_MAXPACKET)
 985                         mtu = IPV6_MAXPACKET;
 986
 987                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 988                 if (len < 8) {
 989                         error = EMSGSIZE;
 990                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 991                         goto bad;
 992                 }
 993
 994                 /*
 995                  * If the interface will not calculate checksums on
 996                  * fragmented packets, then do it here.
 997                  * XXX-BZ handle the hw offloading case.  Need flags.
 998                  */
 999                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1000                         in6_delayed_cksum(m, plen, hlen);
1001                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1002                 }
1003 #ifdef SCTP
1004                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1005                         sctp_delayed_cksum(m, hlen);
1006                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1007                 }
1008 #endif
1009                 /*
1010                  * Change the next header field of the last header in the
1011                  * unfragmentable part.
1012                  */
1013                 if (exthdrs.ip6e_rthdr) {
1014                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1015                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1016                 } else if (exthdrs.ip6e_dest1) {
1017                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1018                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1019                 } else if (exthdrs.ip6e_hbh) {
1020                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1021                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1022                 } else {
1023                         nextproto = ip6->ip6_nxt;
1024                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1025                 }
1026
1027                 /*
1028                  * Loop through length of segment after first fragment,
1029                  * make new header and copy data of each part and link onto
1030                  * chain.
1031                  */
1032                 m0 = m;
1033                 id = htonl(ip6_randomid());
1034                 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1035                         goto sendorfree;
1036
1037                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1038         }
1039
1040         /*
1041          * Remove leading garbages.
1042          */
1043 sendorfree:
1044         m = m0->m_nextpkt;
1045         m0->m_nextpkt = 0;
1046         m_freem(m0);
1047         for (m0 = m; m; m = m0) {
1048                 m0 = m->m_nextpkt;
1049                 m->m_nextpkt = 0;
1050                 if (error == 0) {
1051                         /* Record statistics for this interface address. */
1052                         if (ia) {
1053                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1054                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1055                                     m->m_pkthdr.len);
1056                         }
1057                         error = nd6_output_ifp(ifp, origifp, m, dst,
1058                             (struct route *)ro);
1059                 } else
1060                         m_freem(m);
1061         }
1062
1063         if (error == 0)
1064                 IP6STAT_INC(ip6s_fragmented);
1065
1066 done:
1067         /*
1068          * Release the route if using our private route, or if
1069          * (with flowtable) we don't have our own reference.
1070          */
1071         if (ro == &ip6route ||
1072             (ro != NULL && ro->ro_flags & RT_NORTREF))
1073                 RO_RTFREE(ro);
1074         return (error);
1075
1076 freehdrs:
1077         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1078         m_freem(exthdrs.ip6e_dest1);
1079         m_freem(exthdrs.ip6e_rthdr);
1080         m_freem(exthdrs.ip6e_dest2);
1081         /* FALLTHROUGH */
1082 bad:
1083         if (m)
1084                 m_freem(m);
1085         goto done;
1086 }
1087
1088 static int
1089 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1090 {
1091         struct mbuf *m;
1092
1093         if (hlen > MCLBYTES)
1094                 return (ENOBUFS); /* XXX */
1095
1096         if (hlen > MLEN)
1097                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1098         else
1099                 m = m_get(M_NOWAIT, MT_DATA);
1100         if (m == NULL)
1101                 return (ENOBUFS);
1102         m->m_len = hlen;
1103         if (hdr)
1104                 bcopy(hdr, mtod(m, caddr_t), hlen);
1105
1106         *mp = m;
1107         return (0);
1108 }
1109
1110 /*
1111  * Insert jumbo payload option.
1112  */
1113 static int
1114 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1115 {
1116         struct mbuf *mopt;
1117         u_char *optbuf;
1118         u_int32_t v;
1119
1120 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1121
1122         /*
1123          * If there is no hop-by-hop options header, allocate new one.
1124          * If there is one but it doesn't have enough space to store the
1125          * jumbo payload option, allocate a cluster to store the whole options.
1126          * Otherwise, use it to store the options.
1127          */
1128         if (exthdrs->ip6e_hbh == NULL) {
1129                 mopt = m_get(M_NOWAIT, MT_DATA);
1130                 if (mopt == NULL)
1131                         return (ENOBUFS);
1132                 mopt->m_len = JUMBOOPTLEN;
1133                 optbuf = mtod(mopt, u_char *);
1134                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1135                 exthdrs->ip6e_hbh = mopt;
1136         } else {
1137                 struct ip6_hbh *hbh;
1138
1139                 mopt = exthdrs->ip6e_hbh;
1140                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1141                         /*
1142                          * XXX assumption:
1143                          * - exthdrs->ip6e_hbh is not referenced from places
1144                          *   other than exthdrs.
1145                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1146                          */
1147                         int oldoptlen = mopt->m_len;
1148                         struct mbuf *n;
1149
1150                         /*
1151                          * XXX: give up if the whole (new) hbh header does
1152                          * not fit even in an mbuf cluster.
1153                          */
1154                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1155                                 return (ENOBUFS);
1156
1157                         /*
1158                          * As a consequence, we must always prepare a cluster
1159                          * at this point.
1160                          */
1161                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1162                         if (n == NULL)
1163                                 return (ENOBUFS);
1164                         n->m_len = oldoptlen + JUMBOOPTLEN;
1165                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1166                             oldoptlen);
1167                         optbuf = mtod(n, caddr_t) + oldoptlen;
1168                         m_freem(mopt);
1169                         mopt = exthdrs->ip6e_hbh = n;
1170                 } else {
1171                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1172                         mopt->m_len += JUMBOOPTLEN;
1173                 }
1174                 optbuf[0] = IP6OPT_PADN;
1175                 optbuf[1] = 1;
1176
1177                 /*
1178                  * Adjust the header length according to the pad and
1179                  * the jumbo payload option.
1180                  */
1181                 hbh = mtod(mopt, struct ip6_hbh *);
1182                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1183         }
1184
1185         /* fill in the option. */
1186         optbuf[2] = IP6OPT_JUMBO;
1187         optbuf[3] = 4;
1188         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1189         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1190
1191         /* finally, adjust the packet header length */
1192         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1193
1194         return (0);
1195 #undef JUMBOOPTLEN
1196 }
1197
1198 /*
1199  * Insert fragment header and copy unfragmentable header portions.
1200  */
1201 static int
1202 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1203     struct ip6_frag **frghdrp)
1204 {
1205         struct mbuf *n, *mlast;
1206
1207         if (hlen > sizeof(struct ip6_hdr)) {
1208                 n = m_copym(m0, sizeof(struct ip6_hdr),
1209                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1210                 if (n == NULL)
1211                         return (ENOBUFS);
1212                 m->m_next = n;
1213         } else
1214                 n = m;
1215
1216         /* Search for the last mbuf of unfragmentable part. */
1217         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1218                 ;
1219
1220         if (M_WRITABLE(mlast) &&
1221             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1222                 /* use the trailing space of the last mbuf for the fragment hdr */
1223                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1224                     mlast->m_len);
1225                 mlast->m_len += sizeof(struct ip6_frag);
1226                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1227         } else {
1228                 /* allocate a new mbuf for the fragment header */
1229                 struct mbuf *mfrg;
1230
1231                 mfrg = m_get(M_NOWAIT, MT_DATA);
1232                 if (mfrg == NULL)
1233                         return (ENOBUFS);
1234                 mfrg->m_len = sizeof(struct ip6_frag);
1235                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1236                 mlast->m_next = mfrg;
1237         }
1238
1239         return (0);
1240 }
1241
1242 /*
1243  * Calculates IPv6 path mtu for destination @dst.
1244  * Resulting MTU is stored in @mtup.
1245  *
1246  * Returns 0 on success.
1247  */
1248 static int
1249 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1250 {
1251         struct nhop6_extended nh6;
1252         struct in6_addr kdst;
1253         uint32_t scopeid;
1254         struct ifnet *ifp;
1255         u_long mtu;
1256         int error;
1257
1258         in6_splitscope(dst, &kdst, &scopeid);
1259         if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1260                 return (EHOSTUNREACH);
1261
1262         ifp = nh6.nh_ifp;
1263         mtu = nh6.nh_mtu;
1264
1265         error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1266         fib6_free_nh_ext(fibnum, &nh6);
1267
1268         return (error);
1269 }
1270
1271 /*
1272  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1273  * and cached data in @ro_pmtu.
1274  * MTU from (successful) route lookup is saved (along with dst)
1275  * inside @ro_pmtu to avoid subsequent route lookups after packet
1276  * filter processing.
1277  *
1278  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1279  * Returns 0 on success.
1280  */
1281 static int
1282 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1283     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1284     int *alwaysfragp, u_int fibnum, u_int proto)
1285 {
1286         struct nhop6_basic nh6;
1287         struct in6_addr kdst;
1288         uint32_t scopeid;
1289         struct sockaddr_in6 *sa6_dst;
1290         u_long mtu;
1291
1292         mtu = 0;
1293         if (do_lookup) {
1294
1295                 /*
1296                  * Here ro_pmtu has final destination address, while
1297                  * ro might represent immediate destination.
1298                  * Use ro_pmtu destination since mtu might differ.
1299                  */
1300                 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1301                 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1302                         ro_pmtu->ro_mtu = 0;
1303
1304                 if (ro_pmtu->ro_mtu == 0) {
1305                         bzero(sa6_dst, sizeof(*sa6_dst));
1306                         sa6_dst->sin6_family = AF_INET6;
1307                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1308                         sa6_dst->sin6_addr = *dst;
1309
1310                         in6_splitscope(dst, &kdst, &scopeid);
1311                         if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1312                             &nh6) == 0)
1313                                 ro_pmtu->ro_mtu = nh6.nh_mtu;
1314                 }
1315
1316                 mtu = ro_pmtu->ro_mtu;
1317         }
1318
1319         if (ro_pmtu->ro_rt)
1320                 mtu = ro_pmtu->ro_rt->rt_mtu;
1321
1322         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1323 }
1324
1325 /*
1326  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1327  * hostcache data for @dst.
1328  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1329  *
1330  * Returns 0 on success.
1331  */
1332 static int
1333 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1334     u_long *mtup, int *alwaysfragp, u_int proto)
1335 {
1336         u_long mtu = 0;
1337         int alwaysfrag = 0;
1338         int error = 0;
1339
1340         if (rt_mtu > 0) {
1341                 u_int32_t ifmtu;
1342                 struct in_conninfo inc;
1343
1344                 bzero(&inc, sizeof(inc));
1345                 inc.inc_flags |= INC_ISIPV6;
1346                 inc.inc6_faddr = *dst;
1347
1348                 ifmtu = IN6_LINKMTU(ifp);
1349
1350                 /* TCP is known to react to pmtu changes so skip hc */
1351                 if (proto != IPPROTO_TCP)
1352                         mtu = tcp_hc_getmtu(&inc);
1353
1354                 if (mtu)
1355                         mtu = min(mtu, rt_mtu);
1356                 else
1357                         mtu = rt_mtu;
1358                 if (mtu == 0)
1359                         mtu = ifmtu;
1360                 else if (mtu < IPV6_MMTU) {
1361                         /*
1362                          * RFC2460 section 5, last paragraph:
1363                          * if we record ICMPv6 too big message with
1364                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1365                          * or smaller, with framgent header attached.
1366                          * (fragment header is needed regardless from the
1367                          * packet size, for translators to identify packets)
1368                          */
1369                         alwaysfrag = 1;
1370                         mtu = IPV6_MMTU;
1371                 }
1372         } else if (ifp) {
1373                 mtu = IN6_LINKMTU(ifp);
1374         } else
1375                 error = EHOSTUNREACH; /* XXX */
1376
1377         *mtup = mtu;
1378         if (alwaysfragp)
1379                 *alwaysfragp = alwaysfrag;
1380         return (error);
1381 }
1382
1383 /*
1384  * IP6 socket option processing.
1385  */
1386 int
1387 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1388 {
1389         int optdatalen, uproto;
1390         void *optdata;
1391         struct inpcb *in6p = sotoinpcb(so);
1392         int error, optval;
1393         int level, op, optname;
1394         int optlen;
1395         struct thread *td;
1396 #ifdef  RSS
1397         uint32_t rss_bucket;
1398         int retval;
1399 #endif
1400
1401         level = sopt->sopt_level;
1402         op = sopt->sopt_dir;
1403         optname = sopt->sopt_name;
1404         optlen = sopt->sopt_valsize;
1405         td = sopt->sopt_td;
1406         error = 0;
1407         optval = 0;
1408         uproto = (int)so->so_proto->pr_protocol;
1409
1410         if (level != IPPROTO_IPV6) {
1411                 error = EINVAL;
1412
1413                 if (sopt->sopt_level == SOL_SOCKET &&
1414                     sopt->sopt_dir == SOPT_SET) {
1415                         switch (sopt->sopt_name) {
1416                         case SO_REUSEADDR:
1417                                 INP_WLOCK(in6p);
1418                                 if ((so->so_options & SO_REUSEADDR) != 0)
1419                                         in6p->inp_flags2 |= INP_REUSEADDR;
1420                                 else
1421                                         in6p->inp_flags2 &= ~INP_REUSEADDR;
1422                                 INP_WUNLOCK(in6p);
1423                                 error = 0;
1424                                 break;
1425                         case SO_REUSEPORT:
1426                                 INP_WLOCK(in6p);
1427                                 if ((so->so_options & SO_REUSEPORT) != 0)
1428                                         in6p->inp_flags2 |= INP_REUSEPORT;
1429                                 else
1430                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1431                                 INP_WUNLOCK(in6p);
1432                                 error = 0;
1433                                 break;
1434                         case SO_SETFIB:
1435                                 INP_WLOCK(in6p);
1436                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1437                                 INP_WUNLOCK(in6p);
1438                                 error = 0;
1439                                 break;
1440                         default:
1441                                 break;
1442                         }
1443                 }
1444         } else {                /* level == IPPROTO_IPV6 */
1445                 switch (op) {
1446
1447                 case SOPT_SET:
1448                         switch (optname) {
1449                         case IPV6_2292PKTOPTIONS:
1450 #ifdef IPV6_PKTOPTIONS
1451                         case IPV6_PKTOPTIONS:
1452 #endif
1453                         {
1454                                 struct mbuf *m;
1455
1456                                 error = soopt_getm(sopt, &m); /* XXX */
1457                                 if (error != 0)
1458                                         break;
1459                                 error = soopt_mcopyin(sopt, m); /* XXX */
1460                                 if (error != 0)
1461                                         break;
1462                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1463                                                     m, so, sopt);
1464                                 m_freem(m); /* XXX */
1465                                 break;
1466                         }
1467
1468                         /*
1469                          * Use of some Hop-by-Hop options or some
1470                          * Destination options, might require special
1471                          * privilege.  That is, normal applications
1472                          * (without special privilege) might be forbidden
1473                          * from setting certain options in outgoing packets,
1474                          * and might never see certain options in received
1475                          * packets. [RFC 2292 Section 6]
1476                          * KAME specific note:
1477                          *  KAME prevents non-privileged users from sending or
1478                          *  receiving ANY hbh/dst options in order to avoid
1479                          *  overhead of parsing options in the kernel.
1480                          */
1481                         case IPV6_RECVHOPOPTS:
1482                         case IPV6_RECVDSTOPTS:
1483                         case IPV6_RECVRTHDRDSTOPTS:
1484                                 if (td != NULL) {
1485                                         error = priv_check(td,
1486                                             PRIV_NETINET_SETHDROPTS);
1487                                         if (error)
1488                                                 break;
1489                                 }
1490                                 /* FALLTHROUGH */
1491                         case IPV6_UNICAST_HOPS:
1492                         case IPV6_HOPLIMIT:
1493
1494                         case IPV6_RECVPKTINFO:
1495                         case IPV6_RECVHOPLIMIT:
1496                         case IPV6_RECVRTHDR:
1497                         case IPV6_RECVPATHMTU:
1498                         case IPV6_RECVTCLASS:
1499                         case IPV6_RECVFLOWID:
1500 #ifdef  RSS
1501                         case IPV6_RECVRSSBUCKETID:
1502 #endif
1503                         case IPV6_V6ONLY:
1504                         case IPV6_AUTOFLOWLABEL:
1505                         case IPV6_BINDANY:
1506                         case IPV6_BINDMULTI:
1507 #ifdef  RSS
1508                         case IPV6_RSS_LISTEN_BUCKET:
1509 #endif
1510                                 if (optname == IPV6_BINDANY && td != NULL) {
1511                                         error = priv_check(td,
1512                                             PRIV_NETINET_BINDANY);
1513                                         if (error)
1514                                                 break;
1515                                 }
1516
1517                                 if (optlen != sizeof(int)) {
1518                                         error = EINVAL;
1519                                         break;
1520                                 }
1521                                 error = sooptcopyin(sopt, &optval,
1522                                         sizeof optval, sizeof optval);
1523                                 if (error)
1524                                         break;
1525                                 switch (optname) {
1526
1527                                 case IPV6_UNICAST_HOPS:
1528                                         if (optval < -1 || optval >= 256)
1529                                                 error = EINVAL;
1530                                         else {
1531                                                 /* -1 = kernel default */
1532                                                 in6p->in6p_hops = optval;
1533                                                 if ((in6p->inp_vflag &
1534                                                      INP_IPV4) != 0)
1535                                                         in6p->inp_ip_ttl = optval;
1536                                         }
1537                                         break;
1538 #define OPTSET(bit) \
1539 do { \
1540         INP_WLOCK(in6p); \
1541         if (optval) \
1542                 in6p->inp_flags |= (bit); \
1543         else \
1544                 in6p->inp_flags &= ~(bit); \
1545         INP_WUNLOCK(in6p); \
1546 } while (/*CONSTCOND*/ 0)
1547 #define OPTSET2292(bit) \
1548 do { \
1549         INP_WLOCK(in6p); \
1550         in6p->inp_flags |= IN6P_RFC2292; \
1551         if (optval) \
1552                 in6p->inp_flags |= (bit); \
1553         else \
1554                 in6p->inp_flags &= ~(bit); \
1555         INP_WUNLOCK(in6p); \
1556 } while (/*CONSTCOND*/ 0)
1557 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1558
1559 #define OPTSET2(bit, val) do {                                          \
1560         INP_WLOCK(in6p);                                                \
1561         if (val)                                                        \
1562                 in6p->inp_flags2 |= bit;                                \
1563         else                                                            \
1564                 in6p->inp_flags2 &= ~bit;                               \
1565         INP_WUNLOCK(in6p);                                              \
1566 } while (0)
1567 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1568
1569                                 case IPV6_RECVPKTINFO:
1570                                         /* cannot mix with RFC2292 */
1571                                         if (OPTBIT(IN6P_RFC2292)) {
1572                                                 error = EINVAL;
1573                                                 break;
1574                                         }
1575                                         OPTSET(IN6P_PKTINFO);
1576                                         break;
1577
1578                                 case IPV6_HOPLIMIT:
1579                                 {
1580                                         struct ip6_pktopts **optp;
1581
1582                                         /* cannot mix with RFC2292 */
1583                                         if (OPTBIT(IN6P_RFC2292)) {
1584                                                 error = EINVAL;
1585                                                 break;
1586                                         }
1587                                         optp = &in6p->in6p_outputopts;
1588                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1589                                             (u_char *)&optval, sizeof(optval),
1590                                             optp, (td != NULL) ? td->td_ucred :
1591                                             NULL, uproto);
1592                                         break;
1593                                 }
1594
1595                                 case IPV6_RECVHOPLIMIT:
1596                                         /* cannot mix with RFC2292 */
1597                                         if (OPTBIT(IN6P_RFC2292)) {
1598                                                 error = EINVAL;
1599                                                 break;
1600                                         }
1601                                         OPTSET(IN6P_HOPLIMIT);
1602                                         break;
1603
1604                                 case IPV6_RECVHOPOPTS:
1605                                         /* cannot mix with RFC2292 */
1606                                         if (OPTBIT(IN6P_RFC2292)) {
1607                                                 error = EINVAL;
1608                                                 break;
1609                                         }
1610                                         OPTSET(IN6P_HOPOPTS);
1611                                         break;
1612
1613                                 case IPV6_RECVDSTOPTS:
1614                                         /* cannot mix with RFC2292 */
1615                                         if (OPTBIT(IN6P_RFC2292)) {
1616                                                 error = EINVAL;
1617                                                 break;
1618                                         }
1619                                         OPTSET(IN6P_DSTOPTS);
1620                                         break;
1621
1622                                 case IPV6_RECVRTHDRDSTOPTS:
1623                                         /* cannot mix with RFC2292 */
1624                                         if (OPTBIT(IN6P_RFC2292)) {
1625                                                 error = EINVAL;
1626                                                 break;
1627                                         }
1628                                         OPTSET(IN6P_RTHDRDSTOPTS);
1629                                         break;
1630
1631                                 case IPV6_RECVRTHDR:
1632                                         /* cannot mix with RFC2292 */
1633                                         if (OPTBIT(IN6P_RFC2292)) {
1634                                                 error = EINVAL;
1635                                                 break;
1636                                         }
1637                                         OPTSET(IN6P_RTHDR);
1638                                         break;
1639
1640                                 case IPV6_RECVPATHMTU:
1641                                         /*
1642                                          * We ignore this option for TCP
1643                                          * sockets.
1644                                          * (RFC3542 leaves this case
1645                                          * unspecified.)
1646                                          */
1647                                         if (uproto != IPPROTO_TCP)
1648                                                 OPTSET(IN6P_MTU);
1649                                         break;
1650
1651                                 case IPV6_RECVFLOWID:
1652                                         OPTSET2(INP_RECVFLOWID, optval);
1653                                         break;
1654
1655 #ifdef  RSS
1656                                 case IPV6_RECVRSSBUCKETID:
1657                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1658                                         break;
1659 #endif
1660
1661                                 case IPV6_V6ONLY:
1662                                         /*
1663                                          * make setsockopt(IPV6_V6ONLY)
1664                                          * available only prior to bind(2).
1665                                          * see ipng mailing list, Jun 22 2001.
1666                                          */
1667                                         if (in6p->inp_lport ||
1668                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1669                                                 error = EINVAL;
1670                                                 break;
1671                                         }
1672                                         OPTSET(IN6P_IPV6_V6ONLY);
1673                                         if (optval)
1674                                                 in6p->inp_vflag &= ~INP_IPV4;
1675                                         else
1676                                                 in6p->inp_vflag |= INP_IPV4;
1677                                         break;
1678                                 case IPV6_RECVTCLASS:
1679                                         /* cannot mix with RFC2292 XXX */
1680                                         if (OPTBIT(IN6P_RFC2292)) {
1681                                                 error = EINVAL;
1682                                                 break;
1683                                         }
1684                                         OPTSET(IN6P_TCLASS);
1685                                         break;
1686                                 case IPV6_AUTOFLOWLABEL:
1687                                         OPTSET(IN6P_AUTOFLOWLABEL);
1688                                         break;
1689
1690                                 case IPV6_BINDANY:
1691                                         OPTSET(INP_BINDANY);
1692                                         break;
1693
1694                                 case IPV6_BINDMULTI:
1695                                         OPTSET2(INP_BINDMULTI, optval);
1696                                         break;
1697 #ifdef  RSS
1698                                 case IPV6_RSS_LISTEN_BUCKET:
1699                                         if ((optval >= 0) &&
1700                                             (optval < rss_getnumbuckets())) {
1701                                                 in6p->inp_rss_listen_bucket = optval;
1702                                                 OPTSET2(INP_RSS_BUCKET_SET, 1);
1703                                         } else {
1704                                                 error = EINVAL;
1705                                         }
1706                                         break;
1707 #endif
1708                                 }
1709                                 break;
1710
1711                         case IPV6_TCLASS:
1712                         case IPV6_DONTFRAG:
1713                         case IPV6_USE_MIN_MTU:
1714                         case IPV6_PREFER_TEMPADDR:
1715                                 if (optlen != sizeof(optval)) {
1716                                         error = EINVAL;
1717                                         break;
1718                                 }
1719                                 error = sooptcopyin(sopt, &optval,
1720                                         sizeof optval, sizeof optval);
1721                                 if (error)
1722                                         break;
1723                                 {
1724                                         struct ip6_pktopts **optp;
1725                                         optp = &in6p->in6p_outputopts;
1726                                         error = ip6_pcbopt(optname,
1727                                             (u_char *)&optval, sizeof(optval),
1728                                             optp, (td != NULL) ? td->td_ucred :
1729                                             NULL, uproto);
1730                                         break;
1731                                 }
1732
1733                         case IPV6_2292PKTINFO:
1734                         case IPV6_2292HOPLIMIT:
1735                         case IPV6_2292HOPOPTS:
1736                         case IPV6_2292DSTOPTS:
1737                         case IPV6_2292RTHDR:
1738                                 /* RFC 2292 */
1739                                 if (optlen != sizeof(int)) {
1740                                         error = EINVAL;
1741                                         break;
1742                                 }
1743                                 error = sooptcopyin(sopt, &optval,
1744                                         sizeof optval, sizeof optval);
1745                                 if (error)
1746                                         break;
1747                                 switch (optname) {
1748                                 case IPV6_2292PKTINFO:
1749                                         OPTSET2292(IN6P_PKTINFO);
1750                                         break;
1751                                 case IPV6_2292HOPLIMIT:
1752                                         OPTSET2292(IN6P_HOPLIMIT);
1753                                         break;
1754                                 case IPV6_2292HOPOPTS:
1755                                         /*
1756                                          * Check super-user privilege.
1757                                          * See comments for IPV6_RECVHOPOPTS.
1758                                          */
1759                                         if (td != NULL) {
1760                                                 error = priv_check(td,
1761                                                     PRIV_NETINET_SETHDROPTS);
1762                                                 if (error)
1763                                                         return (error);
1764                                         }
1765                                         OPTSET2292(IN6P_HOPOPTS);
1766                                         break;
1767                                 case IPV6_2292DSTOPTS:
1768                                         if (td != NULL) {
1769                                                 error = priv_check(td,
1770                                                     PRIV_NETINET_SETHDROPTS);
1771                                                 if (error)
1772                                                         return (error);
1773                                         }
1774                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1775                                         break;
1776                                 case IPV6_2292RTHDR:
1777                                         OPTSET2292(IN6P_RTHDR);
1778                                         break;
1779                                 }
1780                                 break;
1781                         case IPV6_PKTINFO:
1782                         case IPV6_HOPOPTS:
1783                         case IPV6_RTHDR:
1784                         case IPV6_DSTOPTS:
1785                         case IPV6_RTHDRDSTOPTS:
1786                         case IPV6_NEXTHOP:
1787                         {
1788                                 /* new advanced API (RFC3542) */
1789                                 u_char *optbuf;
1790                                 u_char optbuf_storage[MCLBYTES];
1791                                 int optlen;
1792                                 struct ip6_pktopts **optp;
1793
1794                                 /* cannot mix with RFC2292 */
1795                                 if (OPTBIT(IN6P_RFC2292)) {
1796                                         error = EINVAL;
1797                                         break;
1798                                 }
1799
1800                                 /*
1801                                  * We only ensure valsize is not too large
1802                                  * here.  Further validation will be done
1803                                  * later.
1804                                  */
1805                                 error = sooptcopyin(sopt, optbuf_storage,
1806                                     sizeof(optbuf_storage), 0);
1807                                 if (error)
1808                                         break;
1809                                 optlen = sopt->sopt_valsize;
1810                                 optbuf = optbuf_storage;
1811                                 optp = &in6p->in6p_outputopts;
1812                                 error = ip6_pcbopt(optname, optbuf, optlen,
1813                                     optp, (td != NULL) ? td->td_ucred : NULL,
1814                                     uproto);
1815                                 break;
1816                         }
1817 #undef OPTSET
1818
1819                         case IPV6_MULTICAST_IF:
1820                         case IPV6_MULTICAST_HOPS:
1821                         case IPV6_MULTICAST_LOOP:
1822                         case IPV6_JOIN_GROUP:
1823                         case IPV6_LEAVE_GROUP:
1824                         case IPV6_MSFILTER:
1825                         case MCAST_BLOCK_SOURCE:
1826                         case MCAST_UNBLOCK_SOURCE:
1827                         case MCAST_JOIN_GROUP:
1828                         case MCAST_LEAVE_GROUP:
1829                         case MCAST_JOIN_SOURCE_GROUP:
1830                         case MCAST_LEAVE_SOURCE_GROUP:
1831                                 error = ip6_setmoptions(in6p, sopt);
1832                                 break;
1833
1834                         case IPV6_PORTRANGE:
1835                                 error = sooptcopyin(sopt, &optval,
1836                                     sizeof optval, sizeof optval);
1837                                 if (error)
1838                                         break;
1839
1840                                 INP_WLOCK(in6p);
1841                                 switch (optval) {
1842                                 case IPV6_PORTRANGE_DEFAULT:
1843                                         in6p->inp_flags &= ~(INP_LOWPORT);
1844                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1845                                         break;
1846
1847                                 case IPV6_PORTRANGE_HIGH:
1848                                         in6p->inp_flags &= ~(INP_LOWPORT);
1849                                         in6p->inp_flags |= INP_HIGHPORT;
1850                                         break;
1851
1852                                 case IPV6_PORTRANGE_LOW:
1853                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1854                                         in6p->inp_flags |= INP_LOWPORT;
1855                                         break;
1856
1857                                 default:
1858                                         error = EINVAL;
1859                                         break;
1860                                 }
1861                                 INP_WUNLOCK(in6p);
1862                                 break;
1863
1864 #ifdef IPSEC
1865                         case IPV6_IPSEC_POLICY:
1866                         {
1867                                 caddr_t req;
1868                                 struct mbuf *m;
1869
1870                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1871                                         break;
1872                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1873                                         break;
1874                                 req = mtod(m, caddr_t);
1875                                 error = ipsec_set_policy(in6p, optname, req,
1876                                     m->m_len, (sopt->sopt_td != NULL) ?
1877                                     sopt->sopt_td->td_ucred : NULL);
1878                                 m_freem(m);
1879                                 break;
1880                         }
1881 #endif /* IPSEC */
1882
1883                         default:
1884                                 error = ENOPROTOOPT;
1885                                 break;
1886                         }
1887                         break;
1888
1889                 case SOPT_GET:
1890                         switch (optname) {
1891
1892                         case IPV6_2292PKTOPTIONS:
1893 #ifdef IPV6_PKTOPTIONS
1894                         case IPV6_PKTOPTIONS:
1895 #endif
1896                                 /*
1897                                  * RFC3542 (effectively) deprecated the
1898                                  * semantics of the 2292-style pktoptions.
1899                                  * Since it was not reliable in nature (i.e.,
1900                                  * applications had to expect the lack of some
1901                                  * information after all), it would make sense
1902                                  * to simplify this part by always returning
1903                                  * empty data.
1904                                  */
1905                                 sopt->sopt_valsize = 0;
1906                                 break;
1907
1908                         case IPV6_RECVHOPOPTS:
1909                         case IPV6_RECVDSTOPTS:
1910                         case IPV6_RECVRTHDRDSTOPTS:
1911                         case IPV6_UNICAST_HOPS:
1912                         case IPV6_RECVPKTINFO:
1913                         case IPV6_RECVHOPLIMIT:
1914                         case IPV6_RECVRTHDR:
1915                         case IPV6_RECVPATHMTU:
1916
1917                         case IPV6_V6ONLY:
1918                         case IPV6_PORTRANGE:
1919                         case IPV6_RECVTCLASS:
1920                         case IPV6_AUTOFLOWLABEL:
1921                         case IPV6_BINDANY:
1922                         case IPV6_FLOWID:
1923                         case IPV6_FLOWTYPE:
1924                         case IPV6_RECVFLOWID:
1925 #ifdef  RSS
1926                         case IPV6_RSSBUCKETID:
1927                         case IPV6_RECVRSSBUCKETID:
1928 #endif
1929                         case IPV6_BINDMULTI:
1930                                 switch (optname) {
1931
1932                                 case IPV6_RECVHOPOPTS:
1933                                         optval = OPTBIT(IN6P_HOPOPTS);
1934                                         break;
1935
1936                                 case IPV6_RECVDSTOPTS:
1937                                         optval = OPTBIT(IN6P_DSTOPTS);
1938                                         break;
1939
1940                                 case IPV6_RECVRTHDRDSTOPTS:
1941                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1942                                         break;
1943
1944                                 case IPV6_UNICAST_HOPS:
1945                                         optval = in6p->in6p_hops;
1946                                         break;
1947
1948                                 case IPV6_RECVPKTINFO:
1949                                         optval = OPTBIT(IN6P_PKTINFO);
1950                                         break;
1951
1952                                 case IPV6_RECVHOPLIMIT:
1953                                         optval = OPTBIT(IN6P_HOPLIMIT);
1954                                         break;
1955
1956                                 case IPV6_RECVRTHDR:
1957                                         optval = OPTBIT(IN6P_RTHDR);
1958                                         break;
1959
1960                                 case IPV6_RECVPATHMTU:
1961                                         optval = OPTBIT(IN6P_MTU);
1962                                         break;
1963
1964                                 case IPV6_V6ONLY:
1965                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1966                                         break;
1967
1968                                 case IPV6_PORTRANGE:
1969                                     {
1970                                         int flags;
1971                                         flags = in6p->inp_flags;
1972                                         if (flags & INP_HIGHPORT)
1973                                                 optval = IPV6_PORTRANGE_HIGH;
1974                                         else if (flags & INP_LOWPORT)
1975                                                 optval = IPV6_PORTRANGE_LOW;
1976                                         else
1977                                                 optval = 0;
1978                                         break;
1979                                     }
1980                                 case IPV6_RECVTCLASS:
1981                                         optval = OPTBIT(IN6P_TCLASS);
1982                                         break;
1983
1984                                 case IPV6_AUTOFLOWLABEL:
1985                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1986                                         break;
1987
1988                                 case IPV6_BINDANY:
1989                                         optval = OPTBIT(INP_BINDANY);
1990                                         break;
1991
1992                                 case IPV6_FLOWID:
1993                                         optval = in6p->inp_flowid;
1994                                         break;
1995
1996                                 case IPV6_FLOWTYPE:
1997                                         optval = in6p->inp_flowtype;
1998                                         break;
1999
2000                                 case IPV6_RECVFLOWID:
2001                                         optval = OPTBIT2(INP_RECVFLOWID);
2002                                         break;
2003 #ifdef  RSS
2004                                 case IPV6_RSSBUCKETID:
2005                                         retval =
2006                                             rss_hash2bucket(in6p->inp_flowid,
2007                                             in6p->inp_flowtype,
2008                                             &rss_bucket);
2009                                         if (retval == 0)
2010                                                 optval = rss_bucket;
2011                                         else
2012                                                 error = EINVAL;
2013                                         break;
2014
2015                                 case IPV6_RECVRSSBUCKETID:
2016                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2017                                         break;
2018 #endif
2019
2020                                 case IPV6_BINDMULTI:
2021                                         optval = OPTBIT2(INP_BINDMULTI);
2022                                         break;
2023
2024                                 }
2025                                 if (error)
2026                                         break;
2027                                 error = sooptcopyout(sopt, &optval,
2028                                         sizeof optval);
2029                                 break;
2030
2031                         case IPV6_PATHMTU:
2032                         {
2033                                 u_long pmtu = 0;
2034                                 struct ip6_mtuinfo mtuinfo;
2035
2036                                 if (!(so->so_state & SS_ISCONNECTED))
2037                                         return (ENOTCONN);
2038                                 /*
2039                                  * XXX: we dot not consider the case of source
2040                                  * routing, or optional information to specify
2041                                  * the outgoing interface.
2042                                  */
2043                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2044                                     &in6p->in6p_faddr, &pmtu);
2045                                 if (error)
2046                                         break;
2047                                 if (pmtu > IPV6_MAXPACKET)
2048                                         pmtu = IPV6_MAXPACKET;
2049
2050                                 bzero(&mtuinfo, sizeof(mtuinfo));
2051                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2052                                 optdata = (void *)&mtuinfo;
2053                                 optdatalen = sizeof(mtuinfo);
2054                                 error = sooptcopyout(sopt, optdata,
2055                                     optdatalen);
2056                                 break;
2057                         }
2058
2059                         case IPV6_2292PKTINFO:
2060                         case IPV6_2292HOPLIMIT:
2061                         case IPV6_2292HOPOPTS:
2062                         case IPV6_2292RTHDR:
2063                         case IPV6_2292DSTOPTS:
2064                                 switch (optname) {
2065                                 case IPV6_2292PKTINFO:
2066                                         optval = OPTBIT(IN6P_PKTINFO);
2067                                         break;
2068                                 case IPV6_2292HOPLIMIT:
2069                                         optval = OPTBIT(IN6P_HOPLIMIT);
2070                                         break;
2071                                 case IPV6_2292HOPOPTS:
2072                                         optval = OPTBIT(IN6P_HOPOPTS);
2073                                         break;
2074                                 case IPV6_2292RTHDR:
2075                                         optval = OPTBIT(IN6P_RTHDR);
2076                                         break;
2077                                 case IPV6_2292DSTOPTS:
2078                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2079                                         break;
2080                                 }
2081                                 error = sooptcopyout(sopt, &optval,
2082                                     sizeof optval);
2083                                 break;
2084                         case IPV6_PKTINFO:
2085                         case IPV6_HOPOPTS:
2086                         case IPV6_RTHDR:
2087                         case IPV6_DSTOPTS:
2088                         case IPV6_RTHDRDSTOPTS:
2089                         case IPV6_NEXTHOP:
2090                         case IPV6_TCLASS:
2091                         case IPV6_DONTFRAG:
2092                         case IPV6_USE_MIN_MTU:
2093                         case IPV6_PREFER_TEMPADDR:
2094                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2095                                     optname, sopt);
2096                                 break;
2097
2098                         case IPV6_MULTICAST_IF:
2099                         case IPV6_MULTICAST_HOPS:
2100                         case IPV6_MULTICAST_LOOP:
2101                         case IPV6_MSFILTER:
2102                                 error = ip6_getmoptions(in6p, sopt);
2103                                 break;
2104
2105 #ifdef IPSEC
2106                         case IPV6_IPSEC_POLICY:
2107                           {
2108                                 caddr_t req = NULL;
2109                                 size_t len = 0;
2110                                 struct mbuf *m = NULL;
2111                                 struct mbuf **mp = &m;
2112                                 size_t ovalsize = sopt->sopt_valsize;
2113                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2114
2115                                 error = soopt_getm(sopt, &m); /* XXX */
2116                                 if (error != 0)
2117                                         break;
2118                                 error = soopt_mcopyin(sopt, m); /* XXX */
2119                                 if (error != 0)
2120                                         break;
2121                                 sopt->sopt_valsize = ovalsize;
2122                                 sopt->sopt_val = oval;
2123                                 if (m) {
2124                                         req = mtod(m, caddr_t);
2125                                         len = m->m_len;
2126                                 }
2127                                 error = ipsec_get_policy(in6p, req, len, mp);
2128                                 if (error == 0)
2129                                         error = soopt_mcopyout(sopt, m); /* XXX */
2130                                 if (error == 0 && m)
2131                                         m_freem(m);
2132                                 break;
2133                           }
2134 #endif /* IPSEC */
2135
2136                         default:
2137                                 error = ENOPROTOOPT;
2138                                 break;
2139                         }
2140                         break;
2141                 }
2142         }
2143         return (error);
2144 }
2145
2146 int
2147 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2148 {
2149         int error = 0, optval, optlen;
2150         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2151         struct inpcb *in6p = sotoinpcb(so);
2152         int level, op, optname;
2153
2154         level = sopt->sopt_level;
2155         op = sopt->sopt_dir;
2156         optname = sopt->sopt_name;
2157         optlen = sopt->sopt_valsize;
2158
2159         if (level != IPPROTO_IPV6) {
2160                 return (EINVAL);
2161         }
2162
2163         switch (optname) {
2164         case IPV6_CHECKSUM:
2165                 /*
2166                  * For ICMPv6 sockets, no modification allowed for checksum
2167                  * offset, permit "no change" values to help existing apps.
2168                  *
2169                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2170                  * for an ICMPv6 socket will fail."
2171                  * The current behavior does not meet RFC3542.
2172                  */
2173                 switch (op) {
2174                 case SOPT_SET:
2175                         if (optlen != sizeof(int)) {
2176                                 error = EINVAL;
2177                                 break;
2178                         }
2179                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2180                                             sizeof(optval));
2181                         if (error)
2182                                 break;
2183                         if ((optval % 2) != 0) {
2184                                 /* the API assumes even offset values */
2185                                 error = EINVAL;
2186                         } else if (so->so_proto->pr_protocol ==
2187                             IPPROTO_ICMPV6) {
2188                                 if (optval != icmp6off)
2189                                         error = EINVAL;
2190                         } else
2191                                 in6p->in6p_cksum = optval;
2192                         break;
2193
2194                 case SOPT_GET:
2195                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2196                                 optval = icmp6off;
2197                         else
2198                                 optval = in6p->in6p_cksum;
2199
2200                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2201                         break;
2202
2203                 default:
2204                         error = EINVAL;
2205                         break;
2206                 }
2207                 break;
2208
2209         default:
2210                 error = ENOPROTOOPT;
2211                 break;
2212         }
2213
2214         return (error);
2215 }
2216
2217 /*
2218  * Set up IP6 options in pcb for insertion in output packets or
2219  * specifying behavior of outgoing packets.
2220  */
2221 static int
2222 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2223     struct socket *so, struct sockopt *sopt)
2224 {
2225         struct ip6_pktopts *opt = *pktopt;
2226         int error = 0;
2227         struct thread *td = sopt->sopt_td;
2228
2229         /* turn off any old options. */
2230         if (opt) {
2231 #ifdef DIAGNOSTIC
2232                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2233                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2234                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2235                         printf("ip6_pcbopts: all specified options are cleared.\n");
2236 #endif
2237                 ip6_clearpktopts(opt, -1);
2238         } else
2239                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2240         *pktopt = NULL;
2241
2242         if (!m || m->m_len == 0) {
2243                 /*
2244                  * Only turning off any previous options, regardless of
2245                  * whether the opt is just created or given.
2246                  */
2247                 free(opt, M_IP6OPT);
2248                 return (0);
2249         }
2250
2251         /*  set options specified by user. */
2252         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2253             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2254                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2255                 free(opt, M_IP6OPT);
2256                 return (error);
2257         }
2258         *pktopt = opt;
2259         return (0);
2260 }
2261
2262 /*
2263  * initialize ip6_pktopts.  beware that there are non-zero default values in
2264  * the struct.
2265  */
2266 void
2267 ip6_initpktopts(struct ip6_pktopts *opt)
2268 {
2269
2270         bzero(opt, sizeof(*opt));
2271         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2272         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2273         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2274         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2275 }
2276
2277 static int
2278 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2279     struct ucred *cred, int uproto)
2280 {
2281         struct ip6_pktopts *opt;
2282
2283         if (*pktopt == NULL) {
2284                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2285                     M_WAITOK);
2286                 ip6_initpktopts(*pktopt);
2287         }
2288         opt = *pktopt;
2289
2290         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2291 }
2292
2293 static int
2294 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2295 {
2296         void *optdata = NULL;
2297         int optdatalen = 0;
2298         struct ip6_ext *ip6e;
2299         int error = 0;
2300         struct in6_pktinfo null_pktinfo;
2301         int deftclass = 0, on;
2302         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2303         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2304
2305         switch (optname) {
2306         case IPV6_PKTINFO:
2307                 optdata = (void *)&null_pktinfo;
2308                 if (pktopt && pktopt->ip6po_pktinfo) {
2309                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2310                             sizeof(null_pktinfo));
2311                         in6_clearscope(&null_pktinfo.ipi6_addr);
2312                 } else {
2313                         /* XXX: we don't have to do this every time... */
2314                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2315                 }
2316                 optdatalen = sizeof(struct in6_pktinfo);
2317                 break;
2318         case IPV6_TCLASS:
2319                 if (pktopt && pktopt->ip6po_tclass >= 0)
2320                         optdata = (void *)&pktopt->ip6po_tclass;
2321                 else
2322                         optdata = (void *)&deftclass;
2323                 optdatalen = sizeof(int);
2324                 break;
2325         case IPV6_HOPOPTS:
2326                 if (pktopt && pktopt->ip6po_hbh) {
2327                         optdata = (void *)pktopt->ip6po_hbh;
2328                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2329                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2330                 }
2331                 break;
2332         case IPV6_RTHDR:
2333                 if (pktopt && pktopt->ip6po_rthdr) {
2334                         optdata = (void *)pktopt->ip6po_rthdr;
2335                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2336                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2337                 }
2338                 break;
2339         case IPV6_RTHDRDSTOPTS:
2340                 if (pktopt && pktopt->ip6po_dest1) {
2341                         optdata = (void *)pktopt->ip6po_dest1;
2342                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2343                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2344                 }
2345                 break;
2346         case IPV6_DSTOPTS:
2347                 if (pktopt && pktopt->ip6po_dest2) {
2348                         optdata = (void *)pktopt->ip6po_dest2;
2349                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2350                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2351                 }
2352                 break;
2353         case IPV6_NEXTHOP:
2354                 if (pktopt && pktopt->ip6po_nexthop) {
2355                         optdata = (void *)pktopt->ip6po_nexthop;
2356                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2357                 }
2358                 break;
2359         case IPV6_USE_MIN_MTU:
2360                 if (pktopt)
2361                         optdata = (void *)&pktopt->ip6po_minmtu;
2362                 else
2363                         optdata = (void *)&defminmtu;
2364                 optdatalen = sizeof(int);
2365                 break;
2366         case IPV6_DONTFRAG:
2367                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2368                         on = 1;
2369                 else
2370                         on = 0;
2371                 optdata = (void *)&on;
2372                 optdatalen = sizeof(on);
2373                 break;
2374         case IPV6_PREFER_TEMPADDR:
2375                 if (pktopt)
2376                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2377                 else
2378                         optdata = (void *)&defpreftemp;
2379                 optdatalen = sizeof(int);
2380                 break;
2381         default:                /* should not happen */
2382 #ifdef DIAGNOSTIC
2383                 panic("ip6_getpcbopt: unexpected option\n");
2384 #endif
2385                 return (ENOPROTOOPT);
2386         }
2387
2388         error = sooptcopyout(sopt, optdata, optdatalen);
2389
2390         return (error);
2391 }
2392
2393 void
2394 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2395 {
2396         if (pktopt == NULL)
2397                 return;
2398
2399         if (optname == -1 || optname == IPV6_PKTINFO) {
2400                 if (pktopt->ip6po_pktinfo)
2401                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2402                 pktopt->ip6po_pktinfo = NULL;
2403         }
2404         if (optname == -1 || optname == IPV6_HOPLIMIT)
2405                 pktopt->ip6po_hlim = -1;
2406         if (optname == -1 || optname == IPV6_TCLASS)
2407                 pktopt->ip6po_tclass = -1;
2408         if (optname == -1 || optname == IPV6_NEXTHOP) {
2409                 if (pktopt->ip6po_nextroute.ro_rt) {
2410                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2411                         pktopt->ip6po_nextroute.ro_rt = NULL;
2412                 }
2413                 if (pktopt->ip6po_nexthop)
2414                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2415                 pktopt->ip6po_nexthop = NULL;
2416         }
2417         if (optname == -1 || optname == IPV6_HOPOPTS) {
2418                 if (pktopt->ip6po_hbh)
2419                         free(pktopt->ip6po_hbh, M_IP6OPT);
2420                 pktopt->ip6po_hbh = NULL;
2421         }
2422         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2423                 if (pktopt->ip6po_dest1)
2424                         free(pktopt->ip6po_dest1, M_IP6OPT);
2425                 pktopt->ip6po_dest1 = NULL;
2426         }
2427         if (optname == -1 || optname == IPV6_RTHDR) {
2428                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2429                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2430                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2431                 if (pktopt->ip6po_route.ro_rt) {
2432                         RTFREE(pktopt->ip6po_route.ro_rt);
2433                         pktopt->ip6po_route.ro_rt = NULL;
2434                 }
2435         }
2436         if (optname == -1 || optname == IPV6_DSTOPTS) {
2437                 if (pktopt->ip6po_dest2)
2438                         free(pktopt->ip6po_dest2, M_IP6OPT);
2439                 pktopt->ip6po_dest2 = NULL;
2440         }
2441 }
2442
2443 #define PKTOPT_EXTHDRCPY(type) \
2444 do {\
2445         if (src->type) {\
2446                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2447                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2448                 if (dst->type == NULL && canwait == M_NOWAIT)\
2449                         goto bad;\
2450                 bcopy(src->type, dst->type, hlen);\
2451         }\
2452 } while (/*CONSTCOND*/ 0)
2453
2454 static int
2455 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2456 {
2457         if (dst == NULL || src == NULL)  {
2458                 printf("ip6_clearpktopts: invalid argument\n");
2459                 return (EINVAL);
2460         }
2461
2462         dst->ip6po_hlim = src->ip6po_hlim;
2463         dst->ip6po_tclass = src->ip6po_tclass;
2464         dst->ip6po_flags = src->ip6po_flags;
2465         dst->ip6po_minmtu = src->ip6po_minmtu;
2466         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2467         if (src->ip6po_pktinfo) {
2468                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2469                     M_IP6OPT, canwait);
2470                 if (dst->ip6po_pktinfo == NULL)
2471                         goto bad;
2472                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2473         }
2474         if (src->ip6po_nexthop) {
2475                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2476                     M_IP6OPT, canwait);
2477                 if (dst->ip6po_nexthop == NULL)
2478                         goto bad;
2479                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2480                     src->ip6po_nexthop->sa_len);
2481         }
2482         PKTOPT_EXTHDRCPY(ip6po_hbh);
2483         PKTOPT_EXTHDRCPY(ip6po_dest1);
2484         PKTOPT_EXTHDRCPY(ip6po_dest2);
2485         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2486         return (0);
2487
2488   bad:
2489         ip6_clearpktopts(dst, -1);
2490         return (ENOBUFS);
2491 }
2492 #undef PKTOPT_EXTHDRCPY
2493
2494 struct ip6_pktopts *
2495 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2496 {
2497         int error;
2498         struct ip6_pktopts *dst;
2499
2500         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2501         if (dst == NULL)
2502                 return (NULL);
2503         ip6_initpktopts(dst);
2504
2505         if ((error = copypktopts(dst, src, canwait)) != 0) {
2506                 free(dst, M_IP6OPT);
2507                 return (NULL);
2508         }
2509
2510         return (dst);
2511 }
2512
2513 void
2514 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2515 {
2516         if (pktopt == NULL)
2517                 return;
2518
2519         ip6_clearpktopts(pktopt, -1);
2520
2521         free(pktopt, M_IP6OPT);
2522 }
2523
2524 /*
2525  * Set IPv6 outgoing packet options based on advanced API.
2526  */
2527 int
2528 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2529     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2530 {
2531         struct cmsghdr *cm = NULL;
2532
2533         if (control == NULL || opt == NULL)
2534                 return (EINVAL);
2535
2536         ip6_initpktopts(opt);
2537         if (stickyopt) {
2538                 int error;
2539
2540                 /*
2541                  * If stickyopt is provided, make a local copy of the options
2542                  * for this particular packet, then override them by ancillary
2543                  * objects.
2544                  * XXX: copypktopts() does not copy the cached route to a next
2545                  * hop (if any).  This is not very good in terms of efficiency,
2546                  * but we can allow this since this option should be rarely
2547                  * used.
2548                  */
2549                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2550                         return (error);
2551         }
2552
2553         /*
2554          * XXX: Currently, we assume all the optional information is stored
2555          * in a single mbuf.
2556          */
2557         if (control->m_next)
2558                 return (EINVAL);
2559
2560         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2561             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2562                 int error;
2563
2564                 if (control->m_len < CMSG_LEN(0))
2565                         return (EINVAL);
2566
2567                 cm = mtod(control, struct cmsghdr *);
2568                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2569                         return (EINVAL);
2570                 if (cm->cmsg_level != IPPROTO_IPV6)
2571                         continue;
2572
2573                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2574                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2575                 if (error)
2576                         return (error);
2577         }
2578
2579         return (0);
2580 }
2581
2582 /*
2583  * Set a particular packet option, as a sticky option or an ancillary data
2584  * item.  "len" can be 0 only when it's a sticky option.
2585  * We have 4 cases of combination of "sticky" and "cmsg":
2586  * "sticky=0, cmsg=0": impossible
2587  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2588  * "sticky=1, cmsg=0": RFC3542 socket option
2589  * "sticky=1, cmsg=1": RFC2292 socket option
2590  */
2591 static int
2592 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2593     struct ucred *cred, int sticky, int cmsg, int uproto)
2594 {
2595         int minmtupolicy, preftemp;
2596         int error;
2597
2598         if (!sticky && !cmsg) {
2599 #ifdef DIAGNOSTIC
2600                 printf("ip6_setpktopt: impossible case\n");
2601 #endif
2602                 return (EINVAL);
2603         }
2604
2605         /*
2606          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2607          * not be specified in the context of RFC3542.  Conversely,
2608          * RFC3542 types should not be specified in the context of RFC2292.
2609          */
2610         if (!cmsg) {
2611                 switch (optname) {
2612                 case IPV6_2292PKTINFO:
2613                 case IPV6_2292HOPLIMIT:
2614                 case IPV6_2292NEXTHOP:
2615                 case IPV6_2292HOPOPTS:
2616                 case IPV6_2292DSTOPTS:
2617                 case IPV6_2292RTHDR:
2618                 case IPV6_2292PKTOPTIONS:
2619                         return (ENOPROTOOPT);
2620                 }
2621         }
2622         if (sticky && cmsg) {
2623                 switch (optname) {
2624                 case IPV6_PKTINFO:
2625                 case IPV6_HOPLIMIT:
2626                 case IPV6_NEXTHOP:
2627                 case IPV6_HOPOPTS:
2628                 case IPV6_DSTOPTS:
2629                 case IPV6_RTHDRDSTOPTS:
2630                 case IPV6_RTHDR:
2631                 case IPV6_USE_MIN_MTU:
2632                 case IPV6_DONTFRAG:
2633                 case IPV6_TCLASS:
2634                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2635                         return (ENOPROTOOPT);
2636                 }
2637         }
2638
2639         switch (optname) {
2640         case IPV6_2292PKTINFO:
2641         case IPV6_PKTINFO:
2642         {
2643                 struct ifnet *ifp = NULL;
2644                 struct in6_pktinfo *pktinfo;
2645
2646                 if (len != sizeof(struct in6_pktinfo))
2647                         return (EINVAL);
2648
2649                 pktinfo = (struct in6_pktinfo *)buf;
2650
2651                 /*
2652                  * An application can clear any sticky IPV6_PKTINFO option by
2653                  * doing a "regular" setsockopt with ipi6_addr being
2654                  * in6addr_any and ipi6_ifindex being zero.
2655                  * [RFC 3542, Section 6]
2656                  */
2657                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2658                     pktinfo->ipi6_ifindex == 0 &&
2659                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2660                         ip6_clearpktopts(opt, optname);
2661                         break;
2662                 }
2663
2664                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2665                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2666                         return (EINVAL);
2667                 }
2668                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2669                         return (EINVAL);
2670                 /* validate the interface index if specified. */
2671                 if (pktinfo->ipi6_ifindex > V_if_index)
2672                          return (ENXIO);
2673                 if (pktinfo->ipi6_ifindex) {
2674                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2675                         if (ifp == NULL)
2676                                 return (ENXIO);
2677                 }
2678                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2679                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2680                         return (ENETDOWN);
2681
2682                 if (ifp != NULL &&
2683                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2684                         struct in6_ifaddr *ia;
2685
2686                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2687                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2688                         if (ia == NULL)
2689                                 return (EADDRNOTAVAIL);
2690                         ifa_free(&ia->ia_ifa);
2691                 }
2692                 /*
2693                  * We store the address anyway, and let in6_selectsrc()
2694                  * validate the specified address.  This is because ipi6_addr
2695                  * may not have enough information about its scope zone, and
2696                  * we may need additional information (such as outgoing
2697                  * interface or the scope zone of a destination address) to
2698                  * disambiguate the scope.
2699                  * XXX: the delay of the validation may confuse the
2700                  * application when it is used as a sticky option.
2701                  */
2702                 if (opt->ip6po_pktinfo == NULL) {
2703                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2704                             M_IP6OPT, M_NOWAIT);
2705                         if (opt->ip6po_pktinfo == NULL)
2706                                 return (ENOBUFS);
2707                 }
2708                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2709                 break;
2710         }
2711
2712         case IPV6_2292HOPLIMIT:
2713         case IPV6_HOPLIMIT:
2714         {
2715                 int *hlimp;
2716
2717                 /*
2718                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2719                  * to simplify the ordering among hoplimit options.
2720                  */
2721                 if (optname == IPV6_HOPLIMIT && sticky)
2722                         return (ENOPROTOOPT);
2723
2724                 if (len != sizeof(int))
2725                         return (EINVAL);
2726                 hlimp = (int *)buf;
2727                 if (*hlimp < -1 || *hlimp > 255)
2728                         return (EINVAL);
2729
2730                 opt->ip6po_hlim = *hlimp;
2731                 break;
2732         }
2733
2734         case IPV6_TCLASS:
2735         {
2736                 int tclass;
2737
2738                 if (len != sizeof(int))
2739                         return (EINVAL);
2740                 tclass = *(int *)buf;
2741                 if (tclass < -1 || tclass > 255)
2742                         return (EINVAL);
2743
2744                 opt->ip6po_tclass = tclass;
2745                 break;
2746         }
2747
2748         case IPV6_2292NEXTHOP:
2749         case IPV6_NEXTHOP:
2750                 if (cred != NULL) {
2751                         error = priv_check_cred(cred,
2752                             PRIV_NETINET_SETHDROPTS, 0);
2753                         if (error)
2754                                 return (error);
2755                 }
2756
2757                 if (len == 0) { /* just remove the option */
2758                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2759                         break;
2760                 }
2761
2762                 /* check if cmsg_len is large enough for sa_len */
2763                 if (len < sizeof(struct sockaddr) || len < *buf)
2764                         return (EINVAL);
2765
2766                 switch (((struct sockaddr *)buf)->sa_family) {
2767                 case AF_INET6:
2768                 {
2769                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2770                         int error;
2771
2772                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2773                                 return (EINVAL);
2774
2775                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2776                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2777                                 return (EINVAL);
2778                         }
2779                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2780                             != 0) {
2781                                 return (error);
2782                         }
2783                         break;
2784                 }
2785                 case AF_LINK:   /* should eventually be supported */
2786                 default:
2787                         return (EAFNOSUPPORT);
2788                 }
2789
2790                 /* turn off the previous option, then set the new option. */
2791                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2792                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2793                 if (opt->ip6po_nexthop == NULL)
2794                         return (ENOBUFS);
2795                 bcopy(buf, opt->ip6po_nexthop, *buf);
2796                 break;
2797
2798         case IPV6_2292HOPOPTS:
2799         case IPV6_HOPOPTS:
2800         {
2801                 struct ip6_hbh *hbh;
2802                 int hbhlen;
2803
2804                 /*
2805                  * XXX: We don't allow a non-privileged user to set ANY HbH
2806                  * options, since per-option restriction has too much
2807                  * overhead.
2808                  */
2809                 if (cred != NULL) {
2810                         error = priv_check_cred(cred,
2811                             PRIV_NETINET_SETHDROPTS, 0);
2812                         if (error)
2813                                 return (error);
2814                 }
2815
2816                 if (len == 0) {
2817                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2818                         break;  /* just remove the option */
2819                 }
2820
2821                 /* message length validation */
2822                 if (len < sizeof(struct ip6_hbh))
2823                         return (EINVAL);
2824                 hbh = (struct ip6_hbh *)buf;
2825                 hbhlen = (hbh->ip6h_len + 1) << 3;
2826                 if (len != hbhlen)
2827                         return (EINVAL);
2828
2829                 /* turn off the previous option, then set the new option. */
2830                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2831                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2832                 if (opt->ip6po_hbh == NULL)
2833                         return (ENOBUFS);
2834                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2835
2836                 break;
2837         }
2838
2839         case IPV6_2292DSTOPTS:
2840         case IPV6_DSTOPTS:
2841         case IPV6_RTHDRDSTOPTS:
2842         {
2843                 struct ip6_dest *dest, **newdest = NULL;
2844                 int destlen;
2845
2846                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2847                         error = priv_check_cred(cred,
2848                             PRIV_NETINET_SETHDROPTS, 0);
2849                         if (error)
2850                                 return (error);
2851                 }
2852
2853                 if (len == 0) {
2854                         ip6_clearpktopts(opt, optname);
2855                         break;  /* just remove the option */
2856                 }
2857
2858                 /* message length validation */
2859                 if (len < sizeof(struct ip6_dest))
2860                         return (EINVAL);
2861                 dest = (struct ip6_dest *)buf;
2862                 destlen = (dest->ip6d_len + 1) << 3;
2863                 if (len != destlen)
2864                         return (EINVAL);
2865
2866                 /*
2867                  * Determine the position that the destination options header
2868                  * should be inserted; before or after the routing header.
2869                  */
2870                 switch (optname) {
2871                 case IPV6_2292DSTOPTS:
2872                         /*
2873                          * The old advacned API is ambiguous on this point.
2874                          * Our approach is to determine the position based
2875                          * according to the existence of a routing header.
2876                          * Note, however, that this depends on the order of the
2877                          * extension headers in the ancillary data; the 1st
2878                          * part of the destination options header must appear
2879                          * before the routing header in the ancillary data,
2880                          * too.
2881                          * RFC3542 solved the ambiguity by introducing
2882                          * separate ancillary data or option types.
2883                          */
2884                         if (opt->ip6po_rthdr == NULL)
2885                                 newdest = &opt->ip6po_dest1;
2886                         else
2887                                 newdest = &opt->ip6po_dest2;
2888                         break;
2889                 case IPV6_RTHDRDSTOPTS:
2890                         newdest = &opt->ip6po_dest1;
2891                         break;
2892                 case IPV6_DSTOPTS:
2893                         newdest = &opt->ip6po_dest2;
2894                         break;
2895                 }
2896
2897                 /* turn off the previous option, then set the new option. */
2898                 ip6_clearpktopts(opt, optname);
2899                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2900                 if (*newdest == NULL)
2901                         return (ENOBUFS);
2902                 bcopy(dest, *newdest, destlen);
2903
2904                 break;
2905         }
2906
2907         case IPV6_2292RTHDR:
2908         case IPV6_RTHDR:
2909         {
2910                 struct ip6_rthdr *rth;
2911                 int rthlen;
2912
2913                 if (len == 0) {
2914                         ip6_clearpktopts(opt, IPV6_RTHDR);
2915                         break;  /* just remove the option */
2916                 }
2917
2918                 /* message length validation */
2919                 if (len < sizeof(struct ip6_rthdr))
2920                         return (EINVAL);
2921                 rth = (struct ip6_rthdr *)buf;
2922                 rthlen = (rth->ip6r_len + 1) << 3;
2923                 if (len != rthlen)
2924                         return (EINVAL);
2925
2926                 switch (rth->ip6r_type) {
2927                 case IPV6_RTHDR_TYPE_0:
2928                         if (rth->ip6r_len == 0) /* must contain one addr */
2929                                 return (EINVAL);
2930                         if (rth->ip6r_len % 2) /* length must be even */
2931                                 return (EINVAL);
2932                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2933                                 return (EINVAL);
2934                         break;
2935                 default:
2936                         return (EINVAL);        /* not supported */
2937                 }
2938
2939                 /* turn off the previous option */
2940                 ip6_clearpktopts(opt, IPV6_RTHDR);
2941                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2942                 if (opt->ip6po_rthdr == NULL)
2943                         return (ENOBUFS);
2944                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2945
2946                 break;
2947         }
2948
2949         case IPV6_USE_MIN_MTU:
2950                 if (len != sizeof(int))
2951                         return (EINVAL);
2952                 minmtupolicy = *(int *)buf;
2953                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2954                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2955                     minmtupolicy != IP6PO_MINMTU_ALL) {
2956                         return (EINVAL);
2957                 }
2958                 opt->ip6po_minmtu = minmtupolicy;
2959                 break;
2960
2961         case IPV6_DONTFRAG:
2962                 if (len != sizeof(int))
2963                         return (EINVAL);
2964
2965                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2966                         /*
2967                          * we ignore this option for TCP sockets.
2968                          * (RFC3542 leaves this case unspecified.)
2969                          */
2970                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2971                 } else
2972                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2973                 break;
2974
2975         case IPV6_PREFER_TEMPADDR:
2976                 if (len != sizeof(int))
2977                         return (EINVAL);
2978                 preftemp = *(int *)buf;
2979                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2980                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2981                     preftemp != IP6PO_TEMPADDR_PREFER) {
2982                         return (EINVAL);
2983                 }
2984                 opt->ip6po_prefer_tempaddr = preftemp;
2985                 break;
2986
2987         default:
2988                 return (ENOPROTOOPT);
2989         } /* end of switch */
2990
2991         return (0);
2992 }
2993
2994 /*
2995  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2996  * packet to the input queue of a specified interface.  Note that this
2997  * calls the output routine of the loopback "driver", but with an interface
2998  * pointer that might NOT be &loif -- easier than replicating that code here.
2999  */
3000 void
3001 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3002 {
3003         struct mbuf *copym;
3004         struct ip6_hdr *ip6;
3005
3006         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3007         if (copym == NULL)
3008                 return;
3009
3010         /*
3011          * Make sure to deep-copy IPv6 header portion in case the data
3012          * is in an mbuf cluster, so that we can safely override the IPv6
3013          * header portion later.
3014          */
3015         if (!M_WRITABLE(copym) ||
3016             copym->m_len < sizeof(struct ip6_hdr)) {
3017                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3018                 if (copym == NULL)
3019                         return;
3020         }
3021         ip6 = mtod(copym, struct ip6_hdr *);
3022         /*
3023          * clear embedded scope identifiers if necessary.
3024          * in6_clearscope will touch the addresses only when necessary.
3025          */
3026         in6_clearscope(&ip6->ip6_src);
3027         in6_clearscope(&ip6->ip6_dst);
3028         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3029                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3030                     CSUM_PSEUDO_HDR;
3031                 copym->m_pkthdr.csum_data = 0xffff;
3032         }
3033         if_simloop(ifp, copym, AF_INET6, 0);
3034 }
3035
3036 /*
3037  * Chop IPv6 header off from the payload.
3038  */
3039 static int
3040 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3041 {
3042         struct mbuf *mh;
3043         struct ip6_hdr *ip6;
3044
3045         ip6 = mtod(m, struct ip6_hdr *);
3046         if (m->m_len > sizeof(*ip6)) {
3047                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3048                 if (mh == NULL) {
3049                         m_freem(m);
3050                         return ENOBUFS;
3051                 }
3052                 m_move_pkthdr(mh, m);
3053                 M_ALIGN(mh, sizeof(*ip6));
3054                 m->m_len -= sizeof(*ip6);
3055                 m->m_data += sizeof(*ip6);
3056                 mh->m_next = m;
3057                 m = mh;
3058                 m->m_len = sizeof(*ip6);
3059                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3060         }
3061         exthdrs->ip6e_ip6 = m;
3062         return 0;
3063 }
3064
3065 /*
3066  * Compute IPv6 extension header length.
3067  */
3068 int
3069 ip6_optlen(struct inpcb *in6p)
3070 {
3071         int len;
3072
3073         if (!in6p->in6p_outputopts)
3074                 return 0;
3075
3076         len = 0;
3077 #define elen(x) \
3078     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3079
3080         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3081         if (in6p->in6p_outputopts->ip6po_rthdr)
3082                 /* dest1 is valid with rthdr only */
3083                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3084         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3085         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3086         return len;
3087 #undef elen
3088 }