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_copy(m0, off, mtu)) == 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 /*
1402  * Don't use more than a quarter of mbuf clusters.  N.B.:
1403  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1404  * on LP64 architectures, so cast to u_long to avoid undefined
1405  * behavior.  ILP32 architectures cannot have nmbclusters
1406  * large enough to overflow for other reasons.
1407  */
1408 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1409
1410         level = sopt->sopt_level;
1411         op = sopt->sopt_dir;
1412         optname = sopt->sopt_name;
1413         optlen = sopt->sopt_valsize;
1414         td = sopt->sopt_td;
1415         error = 0;
1416         optval = 0;
1417         uproto = (int)so->so_proto->pr_protocol;
1418
1419         if (level != IPPROTO_IPV6) {
1420                 error = EINVAL;
1421
1422                 if (sopt->sopt_level == SOL_SOCKET &&
1423                     sopt->sopt_dir == SOPT_SET) {
1424                         switch (sopt->sopt_name) {
1425                         case SO_REUSEADDR:
1426                                 INP_WLOCK(in6p);
1427                                 if ((so->so_options & SO_REUSEADDR) != 0)
1428                                         in6p->inp_flags2 |= INP_REUSEADDR;
1429                                 else
1430                                         in6p->inp_flags2 &= ~INP_REUSEADDR;
1431                                 INP_WUNLOCK(in6p);
1432                                 error = 0;
1433                                 break;
1434                         case SO_REUSEPORT:
1435                                 INP_WLOCK(in6p);
1436                                 if ((so->so_options & SO_REUSEPORT) != 0)
1437                                         in6p->inp_flags2 |= INP_REUSEPORT;
1438                                 else
1439                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1440                                 INP_WUNLOCK(in6p);
1441                                 error = 0;
1442                                 break;
1443                         case SO_SETFIB:
1444                                 INP_WLOCK(in6p);
1445                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1446                                 INP_WUNLOCK(in6p);
1447                                 error = 0;
1448                                 break;
1449                         default:
1450                                 break;
1451                         }
1452                 }
1453         } else {                /* level == IPPROTO_IPV6 */
1454                 switch (op) {
1455
1456                 case SOPT_SET:
1457                         switch (optname) {
1458                         case IPV6_2292PKTOPTIONS:
1459 #ifdef IPV6_PKTOPTIONS
1460                         case IPV6_PKTOPTIONS:
1461 #endif
1462                         {
1463                                 struct mbuf *m;
1464
1465                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1466                                         printf("ip6_ctloutput: mbuf limit hit\n");
1467                                         error = ENOBUFS;
1468                                         break;
1469                                 }
1470
1471                                 error = soopt_getm(sopt, &m); /* XXX */
1472                                 if (error != 0)
1473                                         break;
1474                                 error = soopt_mcopyin(sopt, m); /* XXX */
1475                                 if (error != 0)
1476                                         break;
1477                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1478                                                     m, so, sopt);
1479                                 m_freem(m); /* XXX */
1480                                 break;
1481                         }
1482
1483                         /*
1484                          * Use of some Hop-by-Hop options or some
1485                          * Destination options, might require special
1486                          * privilege.  That is, normal applications
1487                          * (without special privilege) might be forbidden
1488                          * from setting certain options in outgoing packets,
1489                          * and might never see certain options in received
1490                          * packets. [RFC 2292 Section 6]
1491                          * KAME specific note:
1492                          *  KAME prevents non-privileged users from sending or
1493                          *  receiving ANY hbh/dst options in order to avoid
1494                          *  overhead of parsing options in the kernel.
1495                          */
1496                         case IPV6_RECVHOPOPTS:
1497                         case IPV6_RECVDSTOPTS:
1498                         case IPV6_RECVRTHDRDSTOPTS:
1499                                 if (td != NULL) {
1500                                         error = priv_check(td,
1501                                             PRIV_NETINET_SETHDROPTS);
1502                                         if (error)
1503                                                 break;
1504                                 }
1505                                 /* FALLTHROUGH */
1506                         case IPV6_UNICAST_HOPS:
1507                         case IPV6_HOPLIMIT:
1508
1509                         case IPV6_RECVPKTINFO:
1510                         case IPV6_RECVHOPLIMIT:
1511                         case IPV6_RECVRTHDR:
1512                         case IPV6_RECVPATHMTU:
1513                         case IPV6_RECVTCLASS:
1514                         case IPV6_RECVFLOWID:
1515 #ifdef  RSS
1516                         case IPV6_RECVRSSBUCKETID:
1517 #endif
1518                         case IPV6_V6ONLY:
1519                         case IPV6_AUTOFLOWLABEL:
1520                         case IPV6_BINDANY:
1521                         case IPV6_BINDMULTI:
1522 #ifdef  RSS
1523                         case IPV6_RSS_LISTEN_BUCKET:
1524 #endif
1525                                 if (optname == IPV6_BINDANY && td != NULL) {
1526                                         error = priv_check(td,
1527                                             PRIV_NETINET_BINDANY);
1528                                         if (error)
1529                                                 break;
1530                                 }
1531
1532                                 if (optlen != sizeof(int)) {
1533                                         error = EINVAL;
1534                                         break;
1535                                 }
1536                                 error = sooptcopyin(sopt, &optval,
1537                                         sizeof optval, sizeof optval);
1538                                 if (error)
1539                                         break;
1540                                 switch (optname) {
1541
1542                                 case IPV6_UNICAST_HOPS:
1543                                         if (optval < -1 || optval >= 256)
1544                                                 error = EINVAL;
1545                                         else {
1546                                                 /* -1 = kernel default */
1547                                                 in6p->in6p_hops = optval;
1548                                                 if ((in6p->inp_vflag &
1549                                                      INP_IPV4) != 0)
1550                                                         in6p->inp_ip_ttl = optval;
1551                                         }
1552                                         break;
1553 #define OPTSET(bit) \
1554 do { \
1555         INP_WLOCK(in6p); \
1556         if (optval) \
1557                 in6p->inp_flags |= (bit); \
1558         else \
1559                 in6p->inp_flags &= ~(bit); \
1560         INP_WUNLOCK(in6p); \
1561 } while (/*CONSTCOND*/ 0)
1562 #define OPTSET2292(bit) \
1563 do { \
1564         INP_WLOCK(in6p); \
1565         in6p->inp_flags |= IN6P_RFC2292; \
1566         if (optval) \
1567                 in6p->inp_flags |= (bit); \
1568         else \
1569                 in6p->inp_flags &= ~(bit); \
1570         INP_WUNLOCK(in6p); \
1571 } while (/*CONSTCOND*/ 0)
1572 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1573
1574 #define OPTSET2(bit, val) do {                                          \
1575         INP_WLOCK(in6p);                                                \
1576         if (val)                                                        \
1577                 in6p->inp_flags2 |= bit;                                \
1578         else                                                            \
1579                 in6p->inp_flags2 &= ~bit;                               \
1580         INP_WUNLOCK(in6p);                                              \
1581 } while (0)
1582 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1583
1584                                 case IPV6_RECVPKTINFO:
1585                                         /* cannot mix with RFC2292 */
1586                                         if (OPTBIT(IN6P_RFC2292)) {
1587                                                 error = EINVAL;
1588                                                 break;
1589                                         }
1590                                         OPTSET(IN6P_PKTINFO);
1591                                         break;
1592
1593                                 case IPV6_HOPLIMIT:
1594                                 {
1595                                         struct ip6_pktopts **optp;
1596
1597                                         /* cannot mix with RFC2292 */
1598                                         if (OPTBIT(IN6P_RFC2292)) {
1599                                                 error = EINVAL;
1600                                                 break;
1601                                         }
1602                                         optp = &in6p->in6p_outputopts;
1603                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1604                                             (u_char *)&optval, sizeof(optval),
1605                                             optp, (td != NULL) ? td->td_ucred :
1606                                             NULL, uproto);
1607                                         break;
1608                                 }
1609
1610                                 case IPV6_RECVHOPLIMIT:
1611                                         /* cannot mix with RFC2292 */
1612                                         if (OPTBIT(IN6P_RFC2292)) {
1613                                                 error = EINVAL;
1614                                                 break;
1615                                         }
1616                                         OPTSET(IN6P_HOPLIMIT);
1617                                         break;
1618
1619                                 case IPV6_RECVHOPOPTS:
1620                                         /* cannot mix with RFC2292 */
1621                                         if (OPTBIT(IN6P_RFC2292)) {
1622                                                 error = EINVAL;
1623                                                 break;
1624                                         }
1625                                         OPTSET(IN6P_HOPOPTS);
1626                                         break;
1627
1628                                 case IPV6_RECVDSTOPTS:
1629                                         /* cannot mix with RFC2292 */
1630                                         if (OPTBIT(IN6P_RFC2292)) {
1631                                                 error = EINVAL;
1632                                                 break;
1633                                         }
1634                                         OPTSET(IN6P_DSTOPTS);
1635                                         break;
1636
1637                                 case IPV6_RECVRTHDRDSTOPTS:
1638                                         /* cannot mix with RFC2292 */
1639                                         if (OPTBIT(IN6P_RFC2292)) {
1640                                                 error = EINVAL;
1641                                                 break;
1642                                         }
1643                                         OPTSET(IN6P_RTHDRDSTOPTS);
1644                                         break;
1645
1646                                 case IPV6_RECVRTHDR:
1647                                         /* cannot mix with RFC2292 */
1648                                         if (OPTBIT(IN6P_RFC2292)) {
1649                                                 error = EINVAL;
1650                                                 break;
1651                                         }
1652                                         OPTSET(IN6P_RTHDR);
1653                                         break;
1654
1655                                 case IPV6_RECVPATHMTU:
1656                                         /*
1657                                          * We ignore this option for TCP
1658                                          * sockets.
1659                                          * (RFC3542 leaves this case
1660                                          * unspecified.)
1661                                          */
1662                                         if (uproto != IPPROTO_TCP)
1663                                                 OPTSET(IN6P_MTU);
1664                                         break;
1665
1666                                 case IPV6_RECVFLOWID:
1667                                         OPTSET2(INP_RECVFLOWID, optval);
1668                                         break;
1669
1670 #ifdef  RSS
1671                                 case IPV6_RECVRSSBUCKETID:
1672                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1673                                         break;
1674 #endif
1675
1676                                 case IPV6_V6ONLY:
1677                                         /*
1678                                          * make setsockopt(IPV6_V6ONLY)
1679                                          * available only prior to bind(2).
1680                                          * see ipng mailing list, Jun 22 2001.
1681                                          */
1682                                         if (in6p->inp_lport ||
1683                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1684                                                 error = EINVAL;
1685                                                 break;
1686                                         }
1687                                         OPTSET(IN6P_IPV6_V6ONLY);
1688                                         if (optval)
1689                                                 in6p->inp_vflag &= ~INP_IPV4;
1690                                         else
1691                                                 in6p->inp_vflag |= INP_IPV4;
1692                                         break;
1693                                 case IPV6_RECVTCLASS:
1694                                         /* cannot mix with RFC2292 XXX */
1695                                         if (OPTBIT(IN6P_RFC2292)) {
1696                                                 error = EINVAL;
1697                                                 break;
1698                                         }
1699                                         OPTSET(IN6P_TCLASS);
1700                                         break;
1701                                 case IPV6_AUTOFLOWLABEL:
1702                                         OPTSET(IN6P_AUTOFLOWLABEL);
1703                                         break;
1704
1705                                 case IPV6_BINDANY:
1706                                         OPTSET(INP_BINDANY);
1707                                         break;
1708
1709                                 case IPV6_BINDMULTI:
1710                                         OPTSET2(INP_BINDMULTI, optval);
1711                                         break;
1712 #ifdef  RSS
1713                                 case IPV6_RSS_LISTEN_BUCKET:
1714                                         if ((optval >= 0) &&
1715                                             (optval < rss_getnumbuckets())) {
1716                                                 in6p->inp_rss_listen_bucket = optval;
1717                                                 OPTSET2(INP_RSS_BUCKET_SET, 1);
1718                                         } else {
1719                                                 error = EINVAL;
1720                                         }
1721                                         break;
1722 #endif
1723                                 }
1724                                 break;
1725
1726                         case IPV6_TCLASS:
1727                         case IPV6_DONTFRAG:
1728                         case IPV6_USE_MIN_MTU:
1729                         case IPV6_PREFER_TEMPADDR:
1730                                 if (optlen != sizeof(optval)) {
1731                                         error = EINVAL;
1732                                         break;
1733                                 }
1734                                 error = sooptcopyin(sopt, &optval,
1735                                         sizeof optval, sizeof optval);
1736                                 if (error)
1737                                         break;
1738                                 {
1739                                         struct ip6_pktopts **optp;
1740                                         optp = &in6p->in6p_outputopts;
1741                                         error = ip6_pcbopt(optname,
1742                                             (u_char *)&optval, sizeof(optval),
1743                                             optp, (td != NULL) ? td->td_ucred :
1744                                             NULL, uproto);
1745                                         break;
1746                                 }
1747
1748                         case IPV6_2292PKTINFO:
1749                         case IPV6_2292HOPLIMIT:
1750                         case IPV6_2292HOPOPTS:
1751                         case IPV6_2292DSTOPTS:
1752                         case IPV6_2292RTHDR:
1753                                 /* RFC 2292 */
1754                                 if (optlen != sizeof(int)) {
1755                                         error = EINVAL;
1756                                         break;
1757                                 }
1758                                 error = sooptcopyin(sopt, &optval,
1759                                         sizeof optval, sizeof optval);
1760                                 if (error)
1761                                         break;
1762                                 switch (optname) {
1763                                 case IPV6_2292PKTINFO:
1764                                         OPTSET2292(IN6P_PKTINFO);
1765                                         break;
1766                                 case IPV6_2292HOPLIMIT:
1767                                         OPTSET2292(IN6P_HOPLIMIT);
1768                                         break;
1769                                 case IPV6_2292HOPOPTS:
1770                                         /*
1771                                          * Check super-user privilege.
1772                                          * See comments for IPV6_RECVHOPOPTS.
1773                                          */
1774                                         if (td != NULL) {
1775                                                 error = priv_check(td,
1776                                                     PRIV_NETINET_SETHDROPTS);
1777                                                 if (error)
1778                                                         return (error);
1779                                         }
1780                                         OPTSET2292(IN6P_HOPOPTS);
1781                                         break;
1782                                 case IPV6_2292DSTOPTS:
1783                                         if (td != NULL) {
1784                                                 error = priv_check(td,
1785                                                     PRIV_NETINET_SETHDROPTS);
1786                                                 if (error)
1787                                                         return (error);
1788                                         }
1789                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1790                                         break;
1791                                 case IPV6_2292RTHDR:
1792                                         OPTSET2292(IN6P_RTHDR);
1793                                         break;
1794                                 }
1795                                 break;
1796                         case IPV6_PKTINFO:
1797                         case IPV6_HOPOPTS:
1798                         case IPV6_RTHDR:
1799                         case IPV6_DSTOPTS:
1800                         case IPV6_RTHDRDSTOPTS:
1801                         case IPV6_NEXTHOP:
1802                         {
1803                                 /* new advanced API (RFC3542) */
1804                                 u_char *optbuf;
1805                                 u_char optbuf_storage[MCLBYTES];
1806                                 int optlen;
1807                                 struct ip6_pktopts **optp;
1808
1809                                 /* cannot mix with RFC2292 */
1810                                 if (OPTBIT(IN6P_RFC2292)) {
1811                                         error = EINVAL;
1812                                         break;
1813                                 }
1814
1815                                 /*
1816                                  * We only ensure valsize is not too large
1817                                  * here.  Further validation will be done
1818                                  * later.
1819                                  */
1820                                 error = sooptcopyin(sopt, optbuf_storage,
1821                                     sizeof(optbuf_storage), 0);
1822                                 if (error)
1823                                         break;
1824                                 optlen = sopt->sopt_valsize;
1825                                 optbuf = optbuf_storage;
1826                                 optp = &in6p->in6p_outputopts;
1827                                 error = ip6_pcbopt(optname, optbuf, optlen,
1828                                     optp, (td != NULL) ? td->td_ucred : NULL,
1829                                     uproto);
1830                                 break;
1831                         }
1832 #undef OPTSET
1833
1834                         case IPV6_MULTICAST_IF:
1835                         case IPV6_MULTICAST_HOPS:
1836                         case IPV6_MULTICAST_LOOP:
1837                         case IPV6_JOIN_GROUP:
1838                         case IPV6_LEAVE_GROUP:
1839                         case IPV6_MSFILTER:
1840                         case MCAST_BLOCK_SOURCE:
1841                         case MCAST_UNBLOCK_SOURCE:
1842                         case MCAST_JOIN_GROUP:
1843                         case MCAST_LEAVE_GROUP:
1844                         case MCAST_JOIN_SOURCE_GROUP:
1845                         case MCAST_LEAVE_SOURCE_GROUP:
1846                                 error = ip6_setmoptions(in6p, sopt);
1847                                 break;
1848
1849                         case IPV6_PORTRANGE:
1850                                 error = sooptcopyin(sopt, &optval,
1851                                     sizeof optval, sizeof optval);
1852                                 if (error)
1853                                         break;
1854
1855                                 INP_WLOCK(in6p);
1856                                 switch (optval) {
1857                                 case IPV6_PORTRANGE_DEFAULT:
1858                                         in6p->inp_flags &= ~(INP_LOWPORT);
1859                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1860                                         break;
1861
1862                                 case IPV6_PORTRANGE_HIGH:
1863                                         in6p->inp_flags &= ~(INP_LOWPORT);
1864                                         in6p->inp_flags |= INP_HIGHPORT;
1865                                         break;
1866
1867                                 case IPV6_PORTRANGE_LOW:
1868                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1869                                         in6p->inp_flags |= INP_LOWPORT;
1870                                         break;
1871
1872                                 default:
1873                                         error = EINVAL;
1874                                         break;
1875                                 }
1876                                 INP_WUNLOCK(in6p);
1877                                 break;
1878
1879 #ifdef IPSEC
1880                         case IPV6_IPSEC_POLICY:
1881                         {
1882                                 caddr_t req;
1883                                 struct mbuf *m;
1884
1885                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1886                                         break;
1887                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1888                                         break;
1889                                 req = mtod(m, caddr_t);
1890                                 error = ipsec_set_policy(in6p, optname, req,
1891                                     m->m_len, (sopt->sopt_td != NULL) ?
1892                                     sopt->sopt_td->td_ucred : NULL);
1893                                 m_freem(m);
1894                                 break;
1895                         }
1896 #endif /* IPSEC */
1897
1898                         default:
1899                                 error = ENOPROTOOPT;
1900                                 break;
1901                         }
1902                         break;
1903
1904                 case SOPT_GET:
1905                         switch (optname) {
1906
1907                         case IPV6_2292PKTOPTIONS:
1908 #ifdef IPV6_PKTOPTIONS
1909                         case IPV6_PKTOPTIONS:
1910 #endif
1911                                 /*
1912                                  * RFC3542 (effectively) deprecated the
1913                                  * semantics of the 2292-style pktoptions.
1914                                  * Since it was not reliable in nature (i.e.,
1915                                  * applications had to expect the lack of some
1916                                  * information after all), it would make sense
1917                                  * to simplify this part by always returning
1918                                  * empty data.
1919                                  */
1920                                 sopt->sopt_valsize = 0;
1921                                 break;
1922
1923                         case IPV6_RECVHOPOPTS:
1924                         case IPV6_RECVDSTOPTS:
1925                         case IPV6_RECVRTHDRDSTOPTS:
1926                         case IPV6_UNICAST_HOPS:
1927                         case IPV6_RECVPKTINFO:
1928                         case IPV6_RECVHOPLIMIT:
1929                         case IPV6_RECVRTHDR:
1930                         case IPV6_RECVPATHMTU:
1931
1932                         case IPV6_V6ONLY:
1933                         case IPV6_PORTRANGE:
1934                         case IPV6_RECVTCLASS:
1935                         case IPV6_AUTOFLOWLABEL:
1936                         case IPV6_BINDANY:
1937                         case IPV6_FLOWID:
1938                         case IPV6_FLOWTYPE:
1939                         case IPV6_RECVFLOWID:
1940 #ifdef  RSS
1941                         case IPV6_RSSBUCKETID:
1942                         case IPV6_RECVRSSBUCKETID:
1943 #endif
1944                         case IPV6_BINDMULTI:
1945                                 switch (optname) {
1946
1947                                 case IPV6_RECVHOPOPTS:
1948                                         optval = OPTBIT(IN6P_HOPOPTS);
1949                                         break;
1950
1951                                 case IPV6_RECVDSTOPTS:
1952                                         optval = OPTBIT(IN6P_DSTOPTS);
1953                                         break;
1954
1955                                 case IPV6_RECVRTHDRDSTOPTS:
1956                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1957                                         break;
1958
1959                                 case IPV6_UNICAST_HOPS:
1960                                         optval = in6p->in6p_hops;
1961                                         break;
1962
1963                                 case IPV6_RECVPKTINFO:
1964                                         optval = OPTBIT(IN6P_PKTINFO);
1965                                         break;
1966
1967                                 case IPV6_RECVHOPLIMIT:
1968                                         optval = OPTBIT(IN6P_HOPLIMIT);
1969                                         break;
1970
1971                                 case IPV6_RECVRTHDR:
1972                                         optval = OPTBIT(IN6P_RTHDR);
1973                                         break;
1974
1975                                 case IPV6_RECVPATHMTU:
1976                                         optval = OPTBIT(IN6P_MTU);
1977                                         break;
1978
1979                                 case IPV6_V6ONLY:
1980                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
1981                                         break;
1982
1983                                 case IPV6_PORTRANGE:
1984                                     {
1985                                         int flags;
1986                                         flags = in6p->inp_flags;
1987                                         if (flags & INP_HIGHPORT)
1988                                                 optval = IPV6_PORTRANGE_HIGH;
1989                                         else if (flags & INP_LOWPORT)
1990                                                 optval = IPV6_PORTRANGE_LOW;
1991                                         else
1992                                                 optval = 0;
1993                                         break;
1994                                     }
1995                                 case IPV6_RECVTCLASS:
1996                                         optval = OPTBIT(IN6P_TCLASS);
1997                                         break;
1998
1999                                 case IPV6_AUTOFLOWLABEL:
2000                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2001                                         break;
2002
2003                                 case IPV6_BINDANY:
2004                                         optval = OPTBIT(INP_BINDANY);
2005                                         break;
2006
2007                                 case IPV6_FLOWID:
2008                                         optval = in6p->inp_flowid;
2009                                         break;
2010
2011                                 case IPV6_FLOWTYPE:
2012                                         optval = in6p->inp_flowtype;
2013                                         break;
2014
2015                                 case IPV6_RECVFLOWID:
2016                                         optval = OPTBIT2(INP_RECVFLOWID);
2017                                         break;
2018 #ifdef  RSS
2019                                 case IPV6_RSSBUCKETID:
2020                                         retval =
2021                                             rss_hash2bucket(in6p->inp_flowid,
2022                                             in6p->inp_flowtype,
2023                                             &rss_bucket);
2024                                         if (retval == 0)
2025                                                 optval = rss_bucket;
2026                                         else
2027                                                 error = EINVAL;
2028                                         break;
2029
2030                                 case IPV6_RECVRSSBUCKETID:
2031                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2032                                         break;
2033 #endif
2034
2035                                 case IPV6_BINDMULTI:
2036                                         optval = OPTBIT2(INP_BINDMULTI);
2037                                         break;
2038
2039                                 }
2040                                 if (error)
2041                                         break;
2042                                 error = sooptcopyout(sopt, &optval,
2043                                         sizeof optval);
2044                                 break;
2045
2046                         case IPV6_PATHMTU:
2047                         {
2048                                 u_long pmtu = 0;
2049                                 struct ip6_mtuinfo mtuinfo;
2050
2051                                 if (!(so->so_state & SS_ISCONNECTED))
2052                                         return (ENOTCONN);
2053                                 /*
2054                                  * XXX: we dot not consider the case of source
2055                                  * routing, or optional information to specify
2056                                  * the outgoing interface.
2057                                  */
2058                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2059                                     &in6p->in6p_faddr, &pmtu);
2060                                 if (error)
2061                                         break;
2062                                 if (pmtu > IPV6_MAXPACKET)
2063                                         pmtu = IPV6_MAXPACKET;
2064
2065                                 bzero(&mtuinfo, sizeof(mtuinfo));
2066                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2067                                 optdata = (void *)&mtuinfo;
2068                                 optdatalen = sizeof(mtuinfo);
2069                                 error = sooptcopyout(sopt, optdata,
2070                                     optdatalen);
2071                                 break;
2072                         }
2073
2074                         case IPV6_2292PKTINFO:
2075                         case IPV6_2292HOPLIMIT:
2076                         case IPV6_2292HOPOPTS:
2077                         case IPV6_2292RTHDR:
2078                         case IPV6_2292DSTOPTS:
2079                                 switch (optname) {
2080                                 case IPV6_2292PKTINFO:
2081                                         optval = OPTBIT(IN6P_PKTINFO);
2082                                         break;
2083                                 case IPV6_2292HOPLIMIT:
2084                                         optval = OPTBIT(IN6P_HOPLIMIT);
2085                                         break;
2086                                 case IPV6_2292HOPOPTS:
2087                                         optval = OPTBIT(IN6P_HOPOPTS);
2088                                         break;
2089                                 case IPV6_2292RTHDR:
2090                                         optval = OPTBIT(IN6P_RTHDR);
2091                                         break;
2092                                 case IPV6_2292DSTOPTS:
2093                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2094                                         break;
2095                                 }
2096                                 error = sooptcopyout(sopt, &optval,
2097                                     sizeof optval);
2098                                 break;
2099                         case IPV6_PKTINFO:
2100                         case IPV6_HOPOPTS:
2101                         case IPV6_RTHDR:
2102                         case IPV6_DSTOPTS:
2103                         case IPV6_RTHDRDSTOPTS:
2104                         case IPV6_NEXTHOP:
2105                         case IPV6_TCLASS:
2106                         case IPV6_DONTFRAG:
2107                         case IPV6_USE_MIN_MTU:
2108                         case IPV6_PREFER_TEMPADDR:
2109                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2110                                     optname, sopt);
2111                                 break;
2112
2113                         case IPV6_MULTICAST_IF:
2114                         case IPV6_MULTICAST_HOPS:
2115                         case IPV6_MULTICAST_LOOP:
2116                         case IPV6_MSFILTER:
2117                                 error = ip6_getmoptions(in6p, sopt);
2118                                 break;
2119
2120 #ifdef IPSEC
2121                         case IPV6_IPSEC_POLICY:
2122                           {
2123                                 caddr_t req = NULL;
2124                                 size_t len = 0;
2125                                 struct mbuf *m = NULL;
2126                                 struct mbuf **mp = &m;
2127                                 size_t ovalsize = sopt->sopt_valsize;
2128                                 caddr_t oval = (caddr_t)sopt->sopt_val;
2129
2130                                 error = soopt_getm(sopt, &m); /* XXX */
2131                                 if (error != 0)
2132                                         break;
2133                                 error = soopt_mcopyin(sopt, m); /* XXX */
2134                                 if (error != 0)
2135                                         break;
2136                                 sopt->sopt_valsize = ovalsize;
2137                                 sopt->sopt_val = oval;
2138                                 if (m) {
2139                                         req = mtod(m, caddr_t);
2140                                         len = m->m_len;
2141                                 }
2142                                 error = ipsec_get_policy(in6p, req, len, mp);
2143                                 if (error == 0)
2144                                         error = soopt_mcopyout(sopt, m); /* XXX */
2145                                 if (error == 0 && m)
2146                                         m_freem(m);
2147                                 break;
2148                           }
2149 #endif /* IPSEC */
2150
2151                         default:
2152                                 error = ENOPROTOOPT;
2153                                 break;
2154                         }
2155                         break;
2156                 }
2157         }
2158         return (error);
2159 }
2160
2161 int
2162 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2163 {
2164         int error = 0, optval, optlen;
2165         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2166         struct inpcb *in6p = sotoinpcb(so);
2167         int level, op, optname;
2168
2169         level = sopt->sopt_level;
2170         op = sopt->sopt_dir;
2171         optname = sopt->sopt_name;
2172         optlen = sopt->sopt_valsize;
2173
2174         if (level != IPPROTO_IPV6) {
2175                 return (EINVAL);
2176         }
2177
2178         switch (optname) {
2179         case IPV6_CHECKSUM:
2180                 /*
2181                  * For ICMPv6 sockets, no modification allowed for checksum
2182                  * offset, permit "no change" values to help existing apps.
2183                  *
2184                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2185                  * for an ICMPv6 socket will fail."
2186                  * The current behavior does not meet RFC3542.
2187                  */
2188                 switch (op) {
2189                 case SOPT_SET:
2190                         if (optlen != sizeof(int)) {
2191                                 error = EINVAL;
2192                                 break;
2193                         }
2194                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2195                                             sizeof(optval));
2196                         if (error)
2197                                 break;
2198                         if ((optval % 2) != 0) {
2199                                 /* the API assumes even offset values */
2200                                 error = EINVAL;
2201                         } else if (so->so_proto->pr_protocol ==
2202                             IPPROTO_ICMPV6) {
2203                                 if (optval != icmp6off)
2204                                         error = EINVAL;
2205                         } else
2206                                 in6p->in6p_cksum = optval;
2207                         break;
2208
2209                 case SOPT_GET:
2210                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2211                                 optval = icmp6off;
2212                         else
2213                                 optval = in6p->in6p_cksum;
2214
2215                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2216                         break;
2217
2218                 default:
2219                         error = EINVAL;
2220                         break;
2221                 }
2222                 break;
2223
2224         default:
2225                 error = ENOPROTOOPT;
2226                 break;
2227         }
2228
2229         return (error);
2230 }
2231
2232 /*
2233  * Set up IP6 options in pcb for insertion in output packets or
2234  * specifying behavior of outgoing packets.
2235  */
2236 static int
2237 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2238     struct socket *so, struct sockopt *sopt)
2239 {
2240         struct ip6_pktopts *opt = *pktopt;
2241         int error = 0;
2242         struct thread *td = sopt->sopt_td;
2243
2244         /* turn off any old options. */
2245         if (opt) {
2246 #ifdef DIAGNOSTIC
2247                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2248                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2249                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2250                         printf("ip6_pcbopts: all specified options are cleared.\n");
2251 #endif
2252                 ip6_clearpktopts(opt, -1);
2253         } else
2254                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2255         *pktopt = NULL;
2256
2257         if (!m || m->m_len == 0) {
2258                 /*
2259                  * Only turning off any previous options, regardless of
2260                  * whether the opt is just created or given.
2261                  */
2262                 free(opt, M_IP6OPT);
2263                 return (0);
2264         }
2265
2266         /*  set options specified by user. */
2267         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2268             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2269                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2270                 free(opt, M_IP6OPT);
2271                 return (error);
2272         }
2273         *pktopt = opt;
2274         return (0);
2275 }
2276
2277 /*
2278  * initialize ip6_pktopts.  beware that there are non-zero default values in
2279  * the struct.
2280  */
2281 void
2282 ip6_initpktopts(struct ip6_pktopts *opt)
2283 {
2284
2285         bzero(opt, sizeof(*opt));
2286         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2287         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2288         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2289         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2290 }
2291
2292 static int
2293 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2294     struct ucred *cred, int uproto)
2295 {
2296         struct ip6_pktopts *opt;
2297
2298         if (*pktopt == NULL) {
2299                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2300                     M_WAITOK);
2301                 ip6_initpktopts(*pktopt);
2302         }
2303         opt = *pktopt;
2304
2305         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2306 }
2307
2308 static int
2309 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2310 {
2311         void *optdata = NULL;
2312         int optdatalen = 0;
2313         struct ip6_ext *ip6e;
2314         int error = 0;
2315         struct in6_pktinfo null_pktinfo;
2316         int deftclass = 0, on;
2317         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2318         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2319
2320         switch (optname) {
2321         case IPV6_PKTINFO:
2322                 optdata = (void *)&null_pktinfo;
2323                 if (pktopt && pktopt->ip6po_pktinfo) {
2324                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2325                             sizeof(null_pktinfo));
2326                         in6_clearscope(&null_pktinfo.ipi6_addr);
2327                 } else {
2328                         /* XXX: we don't have to do this every time... */
2329                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2330                 }
2331                 optdatalen = sizeof(struct in6_pktinfo);
2332                 break;
2333         case IPV6_TCLASS:
2334                 if (pktopt && pktopt->ip6po_tclass >= 0)
2335                         optdata = (void *)&pktopt->ip6po_tclass;
2336                 else
2337                         optdata = (void *)&deftclass;
2338                 optdatalen = sizeof(int);
2339                 break;
2340         case IPV6_HOPOPTS:
2341                 if (pktopt && pktopt->ip6po_hbh) {
2342                         optdata = (void *)pktopt->ip6po_hbh;
2343                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2344                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2345                 }
2346                 break;
2347         case IPV6_RTHDR:
2348                 if (pktopt && pktopt->ip6po_rthdr) {
2349                         optdata = (void *)pktopt->ip6po_rthdr;
2350                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2351                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2352                 }
2353                 break;
2354         case IPV6_RTHDRDSTOPTS:
2355                 if (pktopt && pktopt->ip6po_dest1) {
2356                         optdata = (void *)pktopt->ip6po_dest1;
2357                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2358                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2359                 }
2360                 break;
2361         case IPV6_DSTOPTS:
2362                 if (pktopt && pktopt->ip6po_dest2) {
2363                         optdata = (void *)pktopt->ip6po_dest2;
2364                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2365                         optdatalen = (ip6e->ip6e_len + 1) << 3;
2366                 }
2367                 break;
2368         case IPV6_NEXTHOP:
2369                 if (pktopt && pktopt->ip6po_nexthop) {
2370                         optdata = (void *)pktopt->ip6po_nexthop;
2371                         optdatalen = pktopt->ip6po_nexthop->sa_len;
2372                 }
2373                 break;
2374         case IPV6_USE_MIN_MTU:
2375                 if (pktopt)
2376                         optdata = (void *)&pktopt->ip6po_minmtu;
2377                 else
2378                         optdata = (void *)&defminmtu;
2379                 optdatalen = sizeof(int);
2380                 break;
2381         case IPV6_DONTFRAG:
2382                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2383                         on = 1;
2384                 else
2385                         on = 0;
2386                 optdata = (void *)&on;
2387                 optdatalen = sizeof(on);
2388                 break;
2389         case IPV6_PREFER_TEMPADDR:
2390                 if (pktopt)
2391                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2392                 else
2393                         optdata = (void *)&defpreftemp;
2394                 optdatalen = sizeof(int);
2395                 break;
2396         default:                /* should not happen */
2397 #ifdef DIAGNOSTIC
2398                 panic("ip6_getpcbopt: unexpected option\n");
2399 #endif
2400                 return (ENOPROTOOPT);
2401         }
2402
2403         error = sooptcopyout(sopt, optdata, optdatalen);
2404
2405         return (error);
2406 }
2407
2408 void
2409 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2410 {
2411         if (pktopt == NULL)
2412                 return;
2413
2414         if (optname == -1 || optname == IPV6_PKTINFO) {
2415                 if (pktopt->ip6po_pktinfo)
2416                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2417                 pktopt->ip6po_pktinfo = NULL;
2418         }
2419         if (optname == -1 || optname == IPV6_HOPLIMIT)
2420                 pktopt->ip6po_hlim = -1;
2421         if (optname == -1 || optname == IPV6_TCLASS)
2422                 pktopt->ip6po_tclass = -1;
2423         if (optname == -1 || optname == IPV6_NEXTHOP) {
2424                 if (pktopt->ip6po_nextroute.ro_rt) {
2425                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2426                         pktopt->ip6po_nextroute.ro_rt = NULL;
2427                 }
2428                 if (pktopt->ip6po_nexthop)
2429                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2430                 pktopt->ip6po_nexthop = NULL;
2431         }
2432         if (optname == -1 || optname == IPV6_HOPOPTS) {
2433                 if (pktopt->ip6po_hbh)
2434                         free(pktopt->ip6po_hbh, M_IP6OPT);
2435                 pktopt->ip6po_hbh = NULL;
2436         }
2437         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2438                 if (pktopt->ip6po_dest1)
2439                         free(pktopt->ip6po_dest1, M_IP6OPT);
2440                 pktopt->ip6po_dest1 = NULL;
2441         }
2442         if (optname == -1 || optname == IPV6_RTHDR) {
2443                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2444                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2445                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2446                 if (pktopt->ip6po_route.ro_rt) {
2447                         RTFREE(pktopt->ip6po_route.ro_rt);
2448                         pktopt->ip6po_route.ro_rt = NULL;
2449                 }
2450         }
2451         if (optname == -1 || optname == IPV6_DSTOPTS) {
2452                 if (pktopt->ip6po_dest2)
2453                         free(pktopt->ip6po_dest2, M_IP6OPT);
2454                 pktopt->ip6po_dest2 = NULL;
2455         }
2456 }
2457
2458 #define PKTOPT_EXTHDRCPY(type) \
2459 do {\
2460         if (src->type) {\
2461                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2462                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2463                 if (dst->type == NULL && canwait == M_NOWAIT)\
2464                         goto bad;\
2465                 bcopy(src->type, dst->type, hlen);\
2466         }\
2467 } while (/*CONSTCOND*/ 0)
2468
2469 static int
2470 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2471 {
2472         if (dst == NULL || src == NULL)  {
2473                 printf("ip6_clearpktopts: invalid argument\n");
2474                 return (EINVAL);
2475         }
2476
2477         dst->ip6po_hlim = src->ip6po_hlim;
2478         dst->ip6po_tclass = src->ip6po_tclass;
2479         dst->ip6po_flags = src->ip6po_flags;
2480         dst->ip6po_minmtu = src->ip6po_minmtu;
2481         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2482         if (src->ip6po_pktinfo) {
2483                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2484                     M_IP6OPT, canwait);
2485                 if (dst->ip6po_pktinfo == NULL)
2486                         goto bad;
2487                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2488         }
2489         if (src->ip6po_nexthop) {
2490                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2491                     M_IP6OPT, canwait);
2492                 if (dst->ip6po_nexthop == NULL)
2493                         goto bad;
2494                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2495                     src->ip6po_nexthop->sa_len);
2496         }
2497         PKTOPT_EXTHDRCPY(ip6po_hbh);
2498         PKTOPT_EXTHDRCPY(ip6po_dest1);
2499         PKTOPT_EXTHDRCPY(ip6po_dest2);
2500         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2501         return (0);
2502
2503   bad:
2504         ip6_clearpktopts(dst, -1);
2505         return (ENOBUFS);
2506 }
2507 #undef PKTOPT_EXTHDRCPY
2508
2509 struct ip6_pktopts *
2510 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2511 {
2512         int error;
2513         struct ip6_pktopts *dst;
2514
2515         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2516         if (dst == NULL)
2517                 return (NULL);
2518         ip6_initpktopts(dst);
2519
2520         if ((error = copypktopts(dst, src, canwait)) != 0) {
2521                 free(dst, M_IP6OPT);
2522                 return (NULL);
2523         }
2524
2525         return (dst);
2526 }
2527
2528 void
2529 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2530 {
2531         if (pktopt == NULL)
2532                 return;
2533
2534         ip6_clearpktopts(pktopt, -1);
2535
2536         free(pktopt, M_IP6OPT);
2537 }
2538
2539 /*
2540  * Set IPv6 outgoing packet options based on advanced API.
2541  */
2542 int
2543 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2544     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2545 {
2546         struct cmsghdr *cm = NULL;
2547
2548         if (control == NULL || opt == NULL)
2549                 return (EINVAL);
2550
2551         ip6_initpktopts(opt);
2552         if (stickyopt) {
2553                 int error;
2554
2555                 /*
2556                  * If stickyopt is provided, make a local copy of the options
2557                  * for this particular packet, then override them by ancillary
2558                  * objects.
2559                  * XXX: copypktopts() does not copy the cached route to a next
2560                  * hop (if any).  This is not very good in terms of efficiency,
2561                  * but we can allow this since this option should be rarely
2562                  * used.
2563                  */
2564                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2565                         return (error);
2566         }
2567
2568         /*
2569          * XXX: Currently, we assume all the optional information is stored
2570          * in a single mbuf.
2571          */
2572         if (control->m_next)
2573                 return (EINVAL);
2574
2575         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2576             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2577                 int error;
2578
2579                 if (control->m_len < CMSG_LEN(0))
2580                         return (EINVAL);
2581
2582                 cm = mtod(control, struct cmsghdr *);
2583                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2584                         return (EINVAL);
2585                 if (cm->cmsg_level != IPPROTO_IPV6)
2586                         continue;
2587
2588                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2589                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2590                 if (error)
2591                         return (error);
2592         }
2593
2594         return (0);
2595 }
2596
2597 /*
2598  * Set a particular packet option, as a sticky option or an ancillary data
2599  * item.  "len" can be 0 only when it's a sticky option.
2600  * We have 4 cases of combination of "sticky" and "cmsg":
2601  * "sticky=0, cmsg=0": impossible
2602  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2603  * "sticky=1, cmsg=0": RFC3542 socket option
2604  * "sticky=1, cmsg=1": RFC2292 socket option
2605  */
2606 static int
2607 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2608     struct ucred *cred, int sticky, int cmsg, int uproto)
2609 {
2610         int minmtupolicy, preftemp;
2611         int error;
2612
2613         if (!sticky && !cmsg) {
2614 #ifdef DIAGNOSTIC
2615                 printf("ip6_setpktopt: impossible case\n");
2616 #endif
2617                 return (EINVAL);
2618         }
2619
2620         /*
2621          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2622          * not be specified in the context of RFC3542.  Conversely,
2623          * RFC3542 types should not be specified in the context of RFC2292.
2624          */
2625         if (!cmsg) {
2626                 switch (optname) {
2627                 case IPV6_2292PKTINFO:
2628                 case IPV6_2292HOPLIMIT:
2629                 case IPV6_2292NEXTHOP:
2630                 case IPV6_2292HOPOPTS:
2631                 case IPV6_2292DSTOPTS:
2632                 case IPV6_2292RTHDR:
2633                 case IPV6_2292PKTOPTIONS:
2634                         return (ENOPROTOOPT);
2635                 }
2636         }
2637         if (sticky && cmsg) {
2638                 switch (optname) {
2639                 case IPV6_PKTINFO:
2640                 case IPV6_HOPLIMIT:
2641                 case IPV6_NEXTHOP:
2642                 case IPV6_HOPOPTS:
2643                 case IPV6_DSTOPTS:
2644                 case IPV6_RTHDRDSTOPTS:
2645                 case IPV6_RTHDR:
2646                 case IPV6_USE_MIN_MTU:
2647                 case IPV6_DONTFRAG:
2648                 case IPV6_TCLASS:
2649                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2650                         return (ENOPROTOOPT);
2651                 }
2652         }
2653
2654         switch (optname) {
2655         case IPV6_2292PKTINFO:
2656         case IPV6_PKTINFO:
2657         {
2658                 struct ifnet *ifp = NULL;
2659                 struct in6_pktinfo *pktinfo;
2660
2661                 if (len != sizeof(struct in6_pktinfo))
2662                         return (EINVAL);
2663
2664                 pktinfo = (struct in6_pktinfo *)buf;
2665
2666                 /*
2667                  * An application can clear any sticky IPV6_PKTINFO option by
2668                  * doing a "regular" setsockopt with ipi6_addr being
2669                  * in6addr_any and ipi6_ifindex being zero.
2670                  * [RFC 3542, Section 6]
2671                  */
2672                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2673                     pktinfo->ipi6_ifindex == 0 &&
2674                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2675                         ip6_clearpktopts(opt, optname);
2676                         break;
2677                 }
2678
2679                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2680                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2681                         return (EINVAL);
2682                 }
2683                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2684                         return (EINVAL);
2685                 /* validate the interface index if specified. */
2686                 if (pktinfo->ipi6_ifindex > V_if_index)
2687                          return (ENXIO);
2688                 if (pktinfo->ipi6_ifindex) {
2689                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2690                         if (ifp == NULL)
2691                                 return (ENXIO);
2692                 }
2693                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2694                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2695                         return (ENETDOWN);
2696
2697                 if (ifp != NULL &&
2698                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2699                         struct in6_ifaddr *ia;
2700
2701                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2702                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2703                         if (ia == NULL)
2704                                 return (EADDRNOTAVAIL);
2705                         ifa_free(&ia->ia_ifa);
2706                 }
2707                 /*
2708                  * We store the address anyway, and let in6_selectsrc()
2709                  * validate the specified address.  This is because ipi6_addr
2710                  * may not have enough information about its scope zone, and
2711                  * we may need additional information (such as outgoing
2712                  * interface or the scope zone of a destination address) to
2713                  * disambiguate the scope.
2714                  * XXX: the delay of the validation may confuse the
2715                  * application when it is used as a sticky option.
2716                  */
2717                 if (opt->ip6po_pktinfo == NULL) {
2718                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2719                             M_IP6OPT, M_NOWAIT);
2720                         if (opt->ip6po_pktinfo == NULL)
2721                                 return (ENOBUFS);
2722                 }
2723                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2724                 break;
2725         }
2726
2727         case IPV6_2292HOPLIMIT:
2728         case IPV6_HOPLIMIT:
2729         {
2730                 int *hlimp;
2731
2732                 /*
2733                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2734                  * to simplify the ordering among hoplimit options.
2735                  */
2736                 if (optname == IPV6_HOPLIMIT && sticky)
2737                         return (ENOPROTOOPT);
2738
2739                 if (len != sizeof(int))
2740                         return (EINVAL);
2741                 hlimp = (int *)buf;
2742                 if (*hlimp < -1 || *hlimp > 255)
2743                         return (EINVAL);
2744
2745                 opt->ip6po_hlim = *hlimp;
2746                 break;
2747         }
2748
2749         case IPV6_TCLASS:
2750         {
2751                 int tclass;
2752
2753                 if (len != sizeof(int))
2754                         return (EINVAL);
2755                 tclass = *(int *)buf;
2756                 if (tclass < -1 || tclass > 255)
2757                         return (EINVAL);
2758
2759                 opt->ip6po_tclass = tclass;
2760                 break;
2761         }
2762
2763         case IPV6_2292NEXTHOP:
2764         case IPV6_NEXTHOP:
2765                 if (cred != NULL) {
2766                         error = priv_check_cred(cred,
2767                             PRIV_NETINET_SETHDROPTS, 0);
2768                         if (error)
2769                                 return (error);
2770                 }
2771
2772                 if (len == 0) { /* just remove the option */
2773                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2774                         break;
2775                 }
2776
2777                 /* check if cmsg_len is large enough for sa_len */
2778                 if (len < sizeof(struct sockaddr) || len < *buf)
2779                         return (EINVAL);
2780
2781                 switch (((struct sockaddr *)buf)->sa_family) {
2782                 case AF_INET6:
2783                 {
2784                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2785                         int error;
2786
2787                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2788                                 return (EINVAL);
2789
2790                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2791                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2792                                 return (EINVAL);
2793                         }
2794                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2795                             != 0) {
2796                                 return (error);
2797                         }
2798                         break;
2799                 }
2800                 case AF_LINK:   /* should eventually be supported */
2801                 default:
2802                         return (EAFNOSUPPORT);
2803                 }
2804
2805                 /* turn off the previous option, then set the new option. */
2806                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2807                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2808                 if (opt->ip6po_nexthop == NULL)
2809                         return (ENOBUFS);
2810                 bcopy(buf, opt->ip6po_nexthop, *buf);
2811                 break;
2812
2813         case IPV6_2292HOPOPTS:
2814         case IPV6_HOPOPTS:
2815         {
2816                 struct ip6_hbh *hbh;
2817                 int hbhlen;
2818
2819                 /*
2820                  * XXX: We don't allow a non-privileged user to set ANY HbH
2821                  * options, since per-option restriction has too much
2822                  * overhead.
2823                  */
2824                 if (cred != NULL) {
2825                         error = priv_check_cred(cred,
2826                             PRIV_NETINET_SETHDROPTS, 0);
2827                         if (error)
2828                                 return (error);
2829                 }
2830
2831                 if (len == 0) {
2832                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2833                         break;  /* just remove the option */
2834                 }
2835
2836                 /* message length validation */
2837                 if (len < sizeof(struct ip6_hbh))
2838                         return (EINVAL);
2839                 hbh = (struct ip6_hbh *)buf;
2840                 hbhlen = (hbh->ip6h_len + 1) << 3;
2841                 if (len != hbhlen)
2842                         return (EINVAL);
2843
2844                 /* turn off the previous option, then set the new option. */
2845                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2846                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2847                 if (opt->ip6po_hbh == NULL)
2848                         return (ENOBUFS);
2849                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2850
2851                 break;
2852         }
2853
2854         case IPV6_2292DSTOPTS:
2855         case IPV6_DSTOPTS:
2856         case IPV6_RTHDRDSTOPTS:
2857         {
2858                 struct ip6_dest *dest, **newdest = NULL;
2859                 int destlen;
2860
2861                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2862                         error = priv_check_cred(cred,
2863                             PRIV_NETINET_SETHDROPTS, 0);
2864                         if (error)
2865                                 return (error);
2866                 }
2867
2868                 if (len == 0) {
2869                         ip6_clearpktopts(opt, optname);
2870                         break;  /* just remove the option */
2871                 }
2872
2873                 /* message length validation */
2874                 if (len < sizeof(struct ip6_dest))
2875                         return (EINVAL);
2876                 dest = (struct ip6_dest *)buf;
2877                 destlen = (dest->ip6d_len + 1) << 3;
2878                 if (len != destlen)
2879                         return (EINVAL);
2880
2881                 /*
2882                  * Determine the position that the destination options header
2883                  * should be inserted; before or after the routing header.
2884                  */
2885                 switch (optname) {
2886                 case IPV6_2292DSTOPTS:
2887                         /*
2888                          * The old advacned API is ambiguous on this point.
2889                          * Our approach is to determine the position based
2890                          * according to the existence of a routing header.
2891                          * Note, however, that this depends on the order of the
2892                          * extension headers in the ancillary data; the 1st
2893                          * part of the destination options header must appear
2894                          * before the routing header in the ancillary data,
2895                          * too.
2896                          * RFC3542 solved the ambiguity by introducing
2897                          * separate ancillary data or option types.
2898                          */
2899                         if (opt->ip6po_rthdr == NULL)
2900                                 newdest = &opt->ip6po_dest1;
2901                         else
2902                                 newdest = &opt->ip6po_dest2;
2903                         break;
2904                 case IPV6_RTHDRDSTOPTS:
2905                         newdest = &opt->ip6po_dest1;
2906                         break;
2907                 case IPV6_DSTOPTS:
2908                         newdest = &opt->ip6po_dest2;
2909                         break;
2910                 }
2911
2912                 /* turn off the previous option, then set the new option. */
2913                 ip6_clearpktopts(opt, optname);
2914                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2915                 if (*newdest == NULL)
2916                         return (ENOBUFS);
2917                 bcopy(dest, *newdest, destlen);
2918
2919                 break;
2920         }
2921
2922         case IPV6_2292RTHDR:
2923         case IPV6_RTHDR:
2924         {
2925                 struct ip6_rthdr *rth;
2926                 int rthlen;
2927
2928                 if (len == 0) {
2929                         ip6_clearpktopts(opt, IPV6_RTHDR);
2930                         break;  /* just remove the option */
2931                 }
2932
2933                 /* message length validation */
2934                 if (len < sizeof(struct ip6_rthdr))
2935                         return (EINVAL);
2936                 rth = (struct ip6_rthdr *)buf;
2937                 rthlen = (rth->ip6r_len + 1) << 3;
2938                 if (len != rthlen)
2939                         return (EINVAL);
2940
2941                 switch (rth->ip6r_type) {
2942                 case IPV6_RTHDR_TYPE_0:
2943                         if (rth->ip6r_len == 0) /* must contain one addr */
2944                                 return (EINVAL);
2945                         if (rth->ip6r_len % 2) /* length must be even */
2946                                 return (EINVAL);
2947                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2948                                 return (EINVAL);
2949                         break;
2950                 default:
2951                         return (EINVAL);        /* not supported */
2952                 }
2953
2954                 /* turn off the previous option */
2955                 ip6_clearpktopts(opt, IPV6_RTHDR);
2956                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2957                 if (opt->ip6po_rthdr == NULL)
2958                         return (ENOBUFS);
2959                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2960
2961                 break;
2962         }
2963
2964         case IPV6_USE_MIN_MTU:
2965                 if (len != sizeof(int))
2966                         return (EINVAL);
2967                 minmtupolicy = *(int *)buf;
2968                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2969                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2970                     minmtupolicy != IP6PO_MINMTU_ALL) {
2971                         return (EINVAL);
2972                 }
2973                 opt->ip6po_minmtu = minmtupolicy;
2974                 break;
2975
2976         case IPV6_DONTFRAG:
2977                 if (len != sizeof(int))
2978                         return (EINVAL);
2979
2980                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2981                         /*
2982                          * we ignore this option for TCP sockets.
2983                          * (RFC3542 leaves this case unspecified.)
2984                          */
2985                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2986                 } else
2987                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2988                 break;
2989
2990         case IPV6_PREFER_TEMPADDR:
2991                 if (len != sizeof(int))
2992                         return (EINVAL);
2993                 preftemp = *(int *)buf;
2994                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2995                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2996                     preftemp != IP6PO_TEMPADDR_PREFER) {
2997                         return (EINVAL);
2998                 }
2999                 opt->ip6po_prefer_tempaddr = preftemp;
3000                 break;
3001
3002         default:
3003                 return (ENOPROTOOPT);
3004         } /* end of switch */
3005
3006         return (0);
3007 }
3008
3009 /*
3010  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3011  * packet to the input queue of a specified interface.  Note that this
3012  * calls the output routine of the loopback "driver", but with an interface
3013  * pointer that might NOT be &loif -- easier than replicating that code here.
3014  */
3015 void
3016 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3017 {
3018         struct mbuf *copym;
3019         struct ip6_hdr *ip6;
3020
3021         copym = m_copy(m, 0, M_COPYALL);
3022         if (copym == NULL)
3023                 return;
3024
3025         /*
3026          * Make sure to deep-copy IPv6 header portion in case the data
3027          * is in an mbuf cluster, so that we can safely override the IPv6
3028          * header portion later.
3029          */
3030         if (!M_WRITABLE(copym) ||
3031             copym->m_len < sizeof(struct ip6_hdr)) {
3032                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3033                 if (copym == NULL)
3034                         return;
3035         }
3036         ip6 = mtod(copym, struct ip6_hdr *);
3037         /*
3038          * clear embedded scope identifiers if necessary.
3039          * in6_clearscope will touch the addresses only when necessary.
3040          */
3041         in6_clearscope(&ip6->ip6_src);
3042         in6_clearscope(&ip6->ip6_dst);
3043         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3044                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3045                     CSUM_PSEUDO_HDR;
3046                 copym->m_pkthdr.csum_data = 0xffff;
3047         }
3048         if_simloop(ifp, copym, AF_INET6, 0);
3049 }
3050
3051 /*
3052  * Chop IPv6 header off from the payload.
3053  */
3054 static int
3055 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3056 {
3057         struct mbuf *mh;
3058         struct ip6_hdr *ip6;
3059
3060         ip6 = mtod(m, struct ip6_hdr *);
3061         if (m->m_len > sizeof(*ip6)) {
3062                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3063                 if (mh == NULL) {
3064                         m_freem(m);
3065                         return ENOBUFS;
3066                 }
3067                 m_move_pkthdr(mh, m);
3068                 M_ALIGN(mh, sizeof(*ip6));
3069                 m->m_len -= sizeof(*ip6);
3070                 m->m_data += sizeof(*ip6);
3071                 mh->m_next = m;
3072                 m = mh;
3073                 m->m_len = sizeof(*ip6);
3074                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3075         }
3076         exthdrs->ip6e_ip6 = m;
3077         return 0;
3078 }
3079
3080 /*
3081  * Compute IPv6 extension header length.
3082  */
3083 int
3084 ip6_optlen(struct inpcb *in6p)
3085 {
3086         int len;
3087
3088         if (!in6p->in6p_outputopts)
3089                 return 0;
3090
3091         len = 0;
3092 #define elen(x) \
3093     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3094
3095         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3096         if (in6p->in6p_outputopts->ip6po_rthdr)
3097                 /* dest1 is valid with rthdr only */
3098                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3099         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3100         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3101         return len;
3102 #undef elen
3103 }