sys/netinet6/ip6_output.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-3-Clause
   3  *
   4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 3. Neither the name of the project nor the names of its contributors
  16  *    may be used to endorse or promote products derived from this software
  17  *    without specific prior written permission.
  18  *
  19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  29  * SUCH DAMAGE.
  30  *
  31  *      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
  32  */
  33
  34 /*-
  35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. Neither the name of the University nor the names of its contributors
  47  *    may be used to endorse or promote products derived from this software
  48  *    without specific prior written permission.
  49  *
  50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  60  * SUCH DAMAGE.
  61  *
  62  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  63  */
  64
  65 #include <sys/cdefs.h>
  66 __FBSDID("$FreeBSD$");
  67
  68 #include "opt_inet.h"
  69 #include "opt_inet6.h"
  70 #include "opt_ipsec.h"
  71 #include "opt_kern_tls.h"
  72 #include "opt_ratelimit.h"
  73 #include "opt_route.h"
  74 #include "opt_rss.h"
  75 #include "opt_sctp.h"
  76
  77 #include <sys/param.h>
  78 #include <sys/kernel.h>
  79 #include <sys/ktls.h>
  80 #include <sys/malloc.h>
  81 #include <sys/mbuf.h>
  82 #include <sys/errno.h>
  83 #include <sys/priv.h>
  84 #include <sys/proc.h>
  85 #include <sys/protosw.h>
  86 #include <sys/socket.h>
  87 #include <sys/socketvar.h>
  88 #include <sys/syslog.h>
  89 #include <sys/ucred.h>
  90
  91 #include <machine/in_cksum.h>
  92
  93 #include <net/if.h>
  94 #include <net/if_var.h>
  95 #include <net/if_llatbl.h>
  96 #include <net/netisr.h>
  97 #include <net/route.h>
  98 #include <net/route/nhop.h>
  99 #include <net/pfil.h>
 100 #include <net/rss_config.h>
 101 #include <net/vnet.h>
 102
 103 #include <netinet/in.h>
 104 #include <netinet/in_var.h>
 105 #include <netinet/ip_var.h>
 106 #include <netinet6/in6_fib.h>
 107 #include <netinet6/in6_var.h>
 108 #include <netinet/ip6.h>
 109 #include <netinet/icmp6.h>
 110 #include <netinet6/ip6_var.h>
 111 #include <netinet/in_pcb.h>
 112 #include <netinet/tcp_var.h>
 113 #include <netinet6/nd6.h>
 114 #include <netinet6/in6_rss.h>
 115
 116 #include <netipsec/ipsec_support.h>
 117 #if defined(SCTP) || defined(SCTP_SUPPORT)
 118 #include <netinet/sctp.h>
 119 #include <netinet/sctp_crc32.h>
 120 #endif
 121
 122 #include <netinet6/ip6protosw.h>
 123 #include <netinet6/scope6_var.h>
 124
 125 extern int in6_mcast_loop;
 126
 127 struct ip6_exthdrs {
 128         struct mbuf *ip6e_ip6;
 129         struct mbuf *ip6e_hbh;
 130         struct mbuf *ip6e_dest1;
 131         struct mbuf *ip6e_rthdr;
 132         struct mbuf *ip6e_dest2;
 133 };
 134
 135 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
 136
 137 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
 138                            struct ucred *, int);
 139 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
 140         struct socket *, struct sockopt *);
 141 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
 142 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
 143         struct ucred *, int, int, int);
 144
 145 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 146 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 147         struct ip6_frag **);
 148 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 149 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 150 static int ip6_getpmtu(struct route_in6 *, int,
 151         struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
 152         u_int);
 153 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
 154         u_long *, int *, u_int);
 155 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
 156 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 157
 158
 159 /*
 160  * Make an extension header from option data.  hp is the source,
 161  * mp is the destination, and _ol is the optlen.
 162  */
 163 #define MAKE_EXTHDR(hp, mp, _ol)                                        \
 164     do {                                                                \
 165         if (hp) {                                                       \
 166                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 167                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 168                     ((eh)->ip6e_len + 1) << 3);                         \
 169                 if (error)                                              \
 170                         goto freehdrs;                                  \
 171                 (_ol) += (*(mp))->m_len;                                \
 172         }                                                               \
 173     } while (/*CONSTCOND*/ 0)
 174
 175 /*
 176  * Form a chain of extension headers.
 177  * m is the extension header mbuf
 178  * mp is the previous mbuf in the chain
 179  * p is the next header
 180  * i is the type of option.
 181  */
 182 #define MAKE_CHAIN(m, mp, p, i)\
 183     do {\
 184         if (m) {\
 185                 if (!hdrsplit) \
 186                         panic("%s:%d: assumption failed: "\
 187                             "hdr not split: hdrsplit %d exthdrs %p",\
 188                             __func__, __LINE__, hdrsplit, &exthdrs);\
 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(csum) > m->m_len)
 209                 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
 210         else
 211                 *(u_short *)mtodo(m, offset) = csum;
 212 }
 213
 214 static int
 215 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
 216     int plen, int optlen, bool frag)
 217 {
 218
 219         KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
 220             "csum_flags %#x frag %d\n",
 221             __func__, __LINE__, plen, optlen, m, ifp, csum_flags, frag));
 222
 223         if ((csum_flags & CSUM_DELAY_DATA_IPV6) ||
 224 #if defined(SCTP) || defined(SCTP_SUPPORT)
 225             (csum_flags & CSUM_SCTP_IPV6) ||
 226 #endif
 227             (!frag && (ifp->if_capenable & IFCAP_NOMAP) == 0)) {
 228                 m = mb_unmapped_to_ext(m);
 229                 if (m == NULL) {
 230                         if (frag)
 231                                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 232                         else
 233                                 IP6STAT_INC(ip6s_odropped);
 234                         return (ENOBUFS);
 235                 }
 236                 if (csum_flags & CSUM_DELAY_DATA_IPV6) {
 237                         in6_delayed_cksum(m, plen - optlen,
 238                             sizeof(struct ip6_hdr) + optlen);
 239                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 240                 }
 241 #if defined(SCTP) || defined(SCTP_SUPPORT)
 242                 if (csum_flags & CSUM_SCTP_IPV6) {
 243                         sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
 244                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 245                 }
 246 #endif
 247         }
 248
 249         return (0);
 250 }
 251
 252 int
 253 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
 254     int fraglen , uint32_t id)
 255 {
 256         struct mbuf *m, **mnext, *m_frgpart;
 257         struct ip6_hdr *ip6, *mhip6;
 258         struct ip6_frag *ip6f;
 259         int off;
 260         int error;
 261         int tlen = m0->m_pkthdr.len;
 262
 263         KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
 264
 265         m = m0;
 266         ip6 = mtod(m, struct ip6_hdr *);
 267         mnext = &m->m_nextpkt;
 268
 269         for (off = hlen; off < tlen; off += fraglen) {
 270                 m = m_gethdr(M_NOWAIT, MT_DATA);
 271                 if (!m) {
 272                         IP6STAT_INC(ip6s_odropped);
 273                         return (ENOBUFS);
 274                 }
 275
 276                 /*
 277                  * Make sure the complete packet header gets copied
 278                  * from the originating mbuf to the newly created
 279                  * mbuf. This also ensures that existing firewall
 280                  * classification(s), VLAN tags and so on get copied
 281                  * to the resulting fragmented packet(s):
 282                  */
 283                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 284                         m_free(m);
 285                         IP6STAT_INC(ip6s_odropped);
 286                         return (ENOBUFS);
 287                 }
 288
 289                 *mnext = m;
 290                 mnext = &m->m_nextpkt;
 291                 m->m_data += max_linkhdr;
 292                 mhip6 = mtod(m, struct ip6_hdr *);
 293                 *mhip6 = *ip6;
 294                 m->m_len = sizeof(*mhip6);
 295                 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 296                 if (error) {
 297                         IP6STAT_INC(ip6s_odropped);
 298                         return (error);
 299                 }
 300                 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 301                 if (off + fraglen >= tlen)
 302                         fraglen = tlen - off;
 303                 else
 304                         ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 305                 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
 306                     sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 307                 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
 308                         IP6STAT_INC(ip6s_odropped);
 309                         return (ENOBUFS);
 310                 }
 311                 m_cat(m, m_frgpart);
 312                 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
 313                 ip6f->ip6f_reserved = 0;
 314                 ip6f->ip6f_ident = id;
 315                 ip6f->ip6f_nxt = nextproto;
 316                 IP6STAT_INC(ip6s_ofragments);
 317                 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 318         }
 319
 320         return (0);
 321 }
 322
 323 static int
 324 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
 325     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
 326     bool stamp_tag)
 327 {
 328 #ifdef KERN_TLS
 329         struct ktls_session *tls = NULL;
 330 #endif
 331         struct m_snd_tag *mst;
 332         int error;
 333
 334         MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
 335         mst = NULL;
 336
 337 #ifdef KERN_TLS
 338         /*
 339          * If this is an unencrypted TLS record, save a reference to
 340          * the record.  This local reference is used to call
 341          * ktls_output_eagain after the mbuf has been freed (thus
 342          * dropping the mbuf's reference) in if_output.
 343          */
 344         if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
 345                 tls = ktls_hold(m->m_next->m_epg_tls);
 346                 mst = tls->snd_tag;
 347
 348                 /*
 349                  * If a TLS session doesn't have a valid tag, it must
 350                  * have had an earlier ifp mismatch, so drop this
 351                  * packet.
 352                  */
 353                 if (mst == NULL) {
 354                         error = EAGAIN;
 355                         goto done;
 356                 }
 357                 /*
 358                  * Always stamp tags that include NIC ktls.
 359                  */
 360                 stamp_tag = true;
 361         }
 362 #endif
 363 #ifdef RATELIMIT
 364         if (inp != NULL && mst == NULL) {
 365                 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
 366                     (inp->inp_snd_tag != NULL &&
 367                     inp->inp_snd_tag->ifp != ifp))
 368                         in_pcboutput_txrtlmt(inp, ifp, m);
 369
 370                 if (inp->inp_snd_tag != NULL)
 371                         mst = inp->inp_snd_tag;
 372         }
 373 #endif
 374         if (stamp_tag && mst != NULL) {
 375                 KASSERT(m->m_pkthdr.rcvif == NULL,
 376                     ("trying to add a send tag to a forwarded packet"));
 377                 if (mst->ifp != ifp) {
 378                         error = EAGAIN;
 379                         goto done;
 380                 }
 381
 382                 /* stamp send tag on mbuf */
 383                 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
 384                 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
 385         }
 386
 387         error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
 388
 389 done:
 390         /* Check for route change invalidating send tags. */
 391 #ifdef KERN_TLS
 392         if (tls != NULL) {
 393                 if (error == EAGAIN)
 394                         error = ktls_output_eagain(inp, tls);
 395                 ktls_free(tls);
 396         }
 397 #endif
 398 #ifdef RATELIMIT
 399         if (error == EAGAIN)
 400                 in_pcboutput_eagain(inp);
 401 #endif
 402         return (error);
 403 }
 404
 405 /*
 406  * IP6 output.
 407  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
 408  * nxt, hlim, src, dst).
 409  * This function may modify ver and hlim only.
 410  * The mbuf chain containing the packet will be freed.
 411  * The mbuf opt, if present, will not be freed.
 412  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
 413  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
 414  * then result of route lookup is stored in ro->ro_nh.
 415  *
 416  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
 417  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
 418  *
 419  * ifpp - XXX: just for statistics
 420  */
 421 /*
 422  * XXX TODO: no flowid is assigned for outbound flows?
 423  */
 424 int
 425 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 426     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 427     struct ifnet **ifpp, struct inpcb *inp)
 428 {
 429         struct ip6_hdr *ip6;
 430         struct ifnet *ifp, *origifp;
 431         struct mbuf *m = m0;
 432         struct mbuf *mprev;
 433         struct route_in6 *ro_pmtu;
 434         struct nhop_object *nh;
 435         struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
 436         struct in6_addr odst;
 437         u_char *nexthdrp;
 438         int tlen, len;
 439         int error = 0;
 440         struct in6_ifaddr *ia = NULL;
 441         u_long mtu;
 442         int alwaysfrag, dontfrag;
 443         u_int32_t optlen, plen = 0, unfragpartlen;
 444         struct ip6_exthdrs exthdrs;
 445         struct in6_addr src0, dst0;
 446         u_int32_t zone;
 447         bool hdrsplit;
 448         int sw_csum, tso;
 449         int needfiblookup;
 450         uint32_t fibnum;
 451         struct m_tag *fwd_tag = NULL;
 452         uint32_t id;
 453
 454         NET_EPOCH_ASSERT();
 455
 456         if (inp != NULL) {
 457                 INP_LOCK_ASSERT(inp);
 458                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 459                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 460                         /* Unconditionally set flowid. */
 461                         m->m_pkthdr.flowid = inp->inp_flowid;
 462                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 463                 }
 464 #ifdef NUMA
 465                 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
 466 #endif
 467         }
 468
 469 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 470         /*
 471          * IPSec checking which handles several cases.
 472          * FAST IPSEC: We re-injected the packet.
 473          * XXX: need scope argument.
 474          */
 475         if (IPSEC_ENABLED(ipv6)) {
 476                 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
 477                         if (error == EINPROGRESS)
 478                                 error = 0;
 479                         goto done;
 480                 }
 481         }
 482 #endif /* IPSEC */
 483
 484         /* Source address validation. */
 485         ip6 = mtod(m, struct ip6_hdr *);
 486         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 487             (flags & IPV6_UNSPECSRC) == 0) {
 488                 error = EOPNOTSUPP;
 489                 IP6STAT_INC(ip6s_badscope);
 490                 goto bad;
 491         }
 492         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 493                 error = EOPNOTSUPP;
 494                 IP6STAT_INC(ip6s_badscope);
 495                 goto bad;
 496         }
 497
 498         /*
 499          * If we are given packet options to add extension headers prepare them.
 500          * Calculate the total length of the extension header chain.
 501          * Keep the length of the unfragmentable part for fragmentation.
 502          */
 503         bzero(&exthdrs, sizeof(exthdrs));
 504         optlen = 0;
 505         unfragpartlen = sizeof(struct ip6_hdr);
 506         if (opt) {
 507                 /* Hop-by-Hop options header. */
 508                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
 509
 510                 /* Destination options header (1st part). */
 511                 if (opt->ip6po_rthdr) {
 512 #ifndef RTHDR_SUPPORT_IMPLEMENTED
 513                         /*
 514                          * If there is a routing header, discard the packet
 515                          * right away here. RH0/1 are obsolete and we do not
 516                          * currently support RH2/3/4.
 517                          * People trying to use RH253/254 may want to disable
 518                          * this check.
 519                          * The moment we do support any routing header (again)
 520                          * this block should check the routing type more
 521                          * selectively.
 522                          */
 523                         error = EINVAL;
 524                         goto bad;
 525 #endif
 526
 527                         /*
 528                          * Destination options header (1st part).
 529                          * This only makes sense with a routing header.
 530                          * See Section 9.2 of RFC 3542.
 531                          * Disabling this part just for MIP6 convenience is
 532                          * a bad idea.  We need to think carefully about a
 533                          * way to make the advanced API coexist with MIP6
 534                          * options, which might automatically be inserted in
 535                          * the kernel.
 536                          */
 537                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
 538                             optlen);
 539                 }
 540                 /* Routing header. */
 541                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
 542
 543                 unfragpartlen += optlen;
 544
 545                 /*
 546                  * NOTE: we don't add AH/ESP length here (done in
 547                  * ip6_ipsec_output()).
 548                  */
 549
 550                 /* Destination options header (2nd part). */
 551                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
 552         }
 553
 554         /*
 555          * If there is at least one extension header,
 556          * separate IP6 header from the payload.
 557          */
 558         hdrsplit = false;
 559         if (optlen) {
 560                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 561                         m = NULL;
 562                         goto freehdrs;
 563                 }
 564                 m = exthdrs.ip6e_ip6;
 565                 ip6 = mtod(m, struct ip6_hdr *);
 566                 hdrsplit = true;
 567         }
 568
 569         /* Adjust mbuf packet header length. */
 570         m->m_pkthdr.len += optlen;
 571         plen = m->m_pkthdr.len - sizeof(*ip6);
 572
 573         /* If this is a jumbo payload, insert a jumbo payload option. */
 574         if (plen > IPV6_MAXPACKET) {
 575                 if (!hdrsplit) {
 576                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 577                                 m = NULL;
 578                                 goto freehdrs;
 579                         }
 580                         m = exthdrs.ip6e_ip6;
 581                         ip6 = mtod(m, struct ip6_hdr *);
 582                         hdrsplit = true;
 583                 }
 584                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 585                         goto freehdrs;
 586                 ip6->ip6_plen = 0;
 587         } else
 588                 ip6->ip6_plen = htons(plen);
 589         nexthdrp = &ip6->ip6_nxt;
 590
 591         if (optlen) {
 592                 /*
 593                  * Concatenate headers and fill in next header fields.
 594                  * Here we have, on "m"
 595                  *      IPv6 payload
 596                  * and we insert headers accordingly.
 597                  * Finally, we should be getting:
 598                  *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
 599                  *
 600                  * During the header composing process "m" points to IPv6
 601                  * header.  "mprev" points to an extension header prior to esp.
 602                  */
 603                 mprev = m;
 604
 605                 /*
 606                  * We treat dest2 specially.  This makes IPsec processing
 607                  * much easier.  The goal here is to make mprev point the
 608                  * mbuf prior to dest2.
 609                  *
 610                  * Result: IPv6 dest2 payload.
 611                  * m and mprev will point to IPv6 header.
 612                  */
 613                 if (exthdrs.ip6e_dest2) {
 614                         if (!hdrsplit)
 615                                 panic("%s:%d: assumption failed: "
 616                                     "hdr not split: hdrsplit %d exthdrs %p",
 617                                     __func__, __LINE__, hdrsplit, &exthdrs);
 618                         exthdrs.ip6e_dest2->m_next = m->m_next;
 619                         m->m_next = exthdrs.ip6e_dest2;
 620                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 621                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 622                 }
 623
 624                 /*
 625                  * Result: IPv6 hbh dest1 rthdr dest2 payload.
 626                  * m will point to IPv6 header.  mprev will point to the
 627                  * extension header prior to dest2 (rthdr in the above case).
 628                  */
 629                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 630                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 631                            IPPROTO_DSTOPTS);
 632                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 633                            IPPROTO_ROUTING);
 634         }
 635
 636         IP6STAT_INC(ip6s_localout);
 637
 638         /* Route packet. */
 639         ro_pmtu = ro;
 640         if (opt && opt->ip6po_rthdr)
 641                 ro = &opt->ip6po_route;
 642         if (ro != NULL)
 643                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 644         else
 645                 dst = &sin6;
 646         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 647
 648 again:
 649         /*
 650          * If specified, try to fill in the traffic class field.
 651          * Do not override if a non-zero value is already set.
 652          * We check the diffserv field and the ECN field separately.
 653          */
 654         if (opt && opt->ip6po_tclass >= 0) {
 655                 int mask = 0;
 656
 657                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 658                         mask |= 0xfc;
 659                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 660                         mask |= 0x03;
 661                 if (mask != 0)
 662                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 663         }
 664
 665         /* Fill in or override the hop limit field, if necessary. */
 666         if (opt && opt->ip6po_hlim != -1)
 667                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 668         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 669                 if (im6o != NULL)
 670                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 671                 else
 672                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 673         }
 674
 675         if (ro == NULL || ro->ro_nh == NULL) {
 676                 bzero(dst, sizeof(*dst));
 677                 dst->sin6_family = AF_INET6;
 678                 dst->sin6_len = sizeof(*dst);
 679                 dst->sin6_addr = ip6->ip6_dst;
 680         }
 681         /*
 682          * Validate route against routing table changes.
 683          * Make sure that the address family is set in route.
 684          */
 685         nh = NULL;
 686         ifp = NULL;
 687         mtu = 0;
 688         if (ro != NULL) {
 689                 if (ro->ro_nh != NULL && inp != NULL) {
 690                         ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
 691                         NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
 692                             fibnum);
 693                 }
 694                 if (ro->ro_nh != NULL && fwd_tag == NULL &&
 695                     (!NH_IS_VALID(ro->ro_nh) ||
 696                     ro->ro_dst.sin6_family != AF_INET6 ||
 697                     !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
 698                         RO_INVALIDATE_CACHE(ro);
 699
 700                 if (ro->ro_nh != NULL && fwd_tag == NULL &&
 701                     ro->ro_dst.sin6_family == AF_INET6 &&
 702                     IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
 703                         nh = ro->ro_nh;
 704                         ifp = nh->nh_ifp;
 705                 } else {
 706                         if (ro->ro_lle)
 707                                 LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 708                         ro->ro_lle = NULL;
 709                         if (fwd_tag == NULL) {
 710                                 bzero(&dst_sa, sizeof(dst_sa));
 711                                 dst_sa.sin6_family = AF_INET6;
 712                                 dst_sa.sin6_len = sizeof(dst_sa);
 713                                 dst_sa.sin6_addr = ip6->ip6_dst;
 714                         }
 715                         error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
 716                             &nh, fibnum, m->m_pkthdr.flowid);
 717                         if (error != 0) {
 718                                 IP6STAT_INC(ip6s_noroute);
 719                                 if (ifp != NULL)
 720                                         in6_ifstat_inc(ifp, ifs6_out_discard);
 721                                 goto bad;
 722                         }
 723                         if (ifp != NULL)
 724                             mtu = ifp->if_mtu;
 725                 }
 726                 if (nh == NULL) {
 727                         /*
 728                          * If in6_selectroute() does not return a nexthop
 729                          * dst may not have been updated.
 730                          */
 731                         *dst = dst_sa;  /* XXX */
 732                 } else {
 733                         if (nh->nh_flags & NHF_HOST)
 734                             mtu = nh->nh_mtu;
 735                         ia = (struct in6_ifaddr *)(nh->nh_ifa);
 736                         counter_u64_add(nh->nh_pksent, 1);
 737                 }
 738         } else {
 739                 struct nhop_object *nh;
 740                 struct in6_addr kdst;
 741                 uint32_t scopeid;
 742
 743                 if (fwd_tag == NULL) {
 744                         bzero(&dst_sa, sizeof(dst_sa));
 745                         dst_sa.sin6_family = AF_INET6;
 746                         dst_sa.sin6_len = sizeof(dst_sa);
 747                         dst_sa.sin6_addr = ip6->ip6_dst;
 748                 }
 749
 750                 if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
 751                     im6o != NULL &&
 752                     (ifp = im6o->im6o_multicast_ifp) != NULL) {
 753                         /* We do not need a route lookup. */
 754                         *dst = dst_sa;  /* XXX */
 755                         goto nonh6lookup;
 756                 }
 757
 758                 in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
 759
 760                 if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
 761                     IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
 762                         if (scopeid > 0) {
 763                                 ifp = in6_getlinkifnet(scopeid);
 764                                 if (ifp == NULL) {
 765                                         error = EHOSTUNREACH;
 766                                         goto bad;
 767                                 }
 768                                 *dst = dst_sa;  /* XXX */
 769                                 goto nonh6lookup;
 770                         }
 771                 }
 772
 773                 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
 774                 if (nh == NULL) {
 775                         IP6STAT_INC(ip6s_noroute);
 776                         /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
 777                         error = EHOSTUNREACH;;
 778                         goto bad;
 779                 }
 780
 781                 ifp = nh->nh_ifp;
 782                 mtu = nh->nh_mtu;
 783                 ia = ifatoia6(nh->nh_ifa);
 784                 if (nh->nh_flags & NHF_GATEWAY)
 785                         dst->sin6_addr = nh->gw6_sa.sin6_addr;
 786 nonh6lookup:
 787                 ;
 788         }
 789
 790         /* Then nh (for unicast) and ifp must be non-NULL valid values. */
 791         if ((flags & IPV6_FORWARDING) == 0) {
 792                 /* XXX: the FORWARDING flag can be set for mrouting. */
 793                 in6_ifstat_inc(ifp, ifs6_out_request);
 794         }
 795
 796         /* Setup data structures for scope ID checks. */
 797         src0 = ip6->ip6_src;
 798         bzero(&src_sa, sizeof(src_sa));
 799         src_sa.sin6_family = AF_INET6;
 800         src_sa.sin6_len = sizeof(src_sa);
 801         src_sa.sin6_addr = ip6->ip6_src;
 802
 803         dst0 = ip6->ip6_dst;
 804         /* Re-initialize to be sure. */
 805         bzero(&dst_sa, sizeof(dst_sa));
 806         dst_sa.sin6_family = AF_INET6;
 807         dst_sa.sin6_len = sizeof(dst_sa);
 808         dst_sa.sin6_addr = ip6->ip6_dst;
 809
 810         /* Check for valid scope ID. */
 811         if (in6_setscope(&src0, ifp, &zone) == 0 &&
 812             sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
 813             in6_setscope(&dst0, ifp, &zone) == 0 &&
 814             sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
 815                 /*
 816                  * The outgoing interface is in the zone of the source
 817                  * and destination addresses.
 818                  *
 819                  * Because the loopback interface cannot receive
 820                  * packets with a different scope ID than its own,
 821                  * there is a trick to pretend the outgoing packet
 822                  * was received by the real network interface, by
 823                  * setting "origifp" different from "ifp". This is
 824                  * only allowed when "ifp" is a loopback network
 825                  * interface. Refer to code in nd6_output_ifp() for
 826                  * more details.
 827                  */
 828                 origifp = ifp;
 829
 830                 /*
 831                  * We should use ia_ifp to support the case of sending
 832                  * packets to an address of our own.
 833                  */
 834                 if (ia != NULL && ia->ia_ifp)
 835                         ifp = ia->ia_ifp;
 836
 837         } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
 838             sa6_recoverscope(&src_sa) != 0 ||
 839             sa6_recoverscope(&dst_sa) != 0 ||
 840             dst_sa.sin6_scope_id == 0 ||
 841             (src_sa.sin6_scope_id != 0 &&
 842             src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
 843             (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
 844                 /*
 845                  * If the destination network interface is not a
 846                  * loopback interface, or the destination network
 847                  * address has no scope ID, or the source address has
 848                  * a scope ID set which is different from the
 849                  * destination address one, or there is no network
 850                  * interface representing this scope ID, the address
 851                  * pair is considered invalid.
 852                  */
 853                 IP6STAT_INC(ip6s_badscope);
 854                 in6_ifstat_inc(ifp, ifs6_out_discard);
 855                 if (error == 0)
 856                         error = EHOSTUNREACH; /* XXX */
 857                 goto bad;
 858         }
 859         /* All scope ID checks are successful. */
 860
 861         if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 862                 if (opt && opt->ip6po_nextroute.ro_nh) {
 863                         /*
 864                          * The nexthop is explicitly specified by the
 865                          * application.  We assume the next hop is an IPv6
 866                          * address.
 867                          */
 868                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 869                 }
 870                 else if ((nh->nh_flags & NHF_GATEWAY))
 871                         dst = &nh->gw6_sa;
 872         }
 873
 874         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 875                 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
 876         } else {
 877                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 878                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 879
 880                 /* Confirm that the outgoing interface supports multicast. */
 881                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 882                         IP6STAT_INC(ip6s_noroute);
 883                         in6_ifstat_inc(ifp, ifs6_out_discard);
 884                         error = ENETUNREACH;
 885                         goto bad;
 886                 }
 887                 if ((im6o == NULL && in6_mcast_loop) ||
 888                     (im6o && im6o->im6o_multicast_loop)) {
 889                         /*
 890                          * Loop back multicast datagram if not expressly
 891                          * forbidden to do so, even if we have not joined
 892                          * the address; protocols will filter it later,
 893                          * thus deferring a hash lookup and lock acquisition
 894                          * at the expense of an m_copym().
 895                          */
 896                         ip6_mloopback(ifp, m);
 897                 } else {
 898                         /*
 899                          * If we are acting as a multicast router, perform
 900                          * multicast forwarding as if the packet had just
 901                          * arrived on the interface to which we are about
 902                          * to send.  The multicast forwarding function
 903                          * recursively calls this function, using the
 904                          * IPV6_FORWARDING flag to prevent infinite recursion.
 905                          *
 906                          * Multicasts that are looped back by ip6_mloopback(),
 907                          * above, will be forwarded by the ip6_input() routine,
 908                          * if necessary.
 909                          */
 910                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 911                                 /*
 912                                  * XXX: ip6_mforward expects that rcvif is NULL
 913                                  * when it is called from the originating path.
 914                                  * However, it may not always be the case.
 915                                  */
 916                                 m->m_pkthdr.rcvif = NULL;
 917                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 918                                         m_freem(m);
 919                                         goto done;
 920                                 }
 921                         }
 922                 }
 923                 /*
 924                  * Multicasts with a hoplimit of zero may be looped back,
 925                  * above, but must not be transmitted on a network.
 926                  * Also, multicasts addressed to the loopback interface
 927                  * are not sent -- the above call to ip6_mloopback() will
 928                  * loop back a copy if this host actually belongs to the
 929                  * destination group on the loopback interface.
 930                  */
 931                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 932                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 933                         m_freem(m);
 934                         goto done;
 935                 }
 936         }
 937
 938         /*
 939          * Fill the outgoing inteface to tell the upper layer
 940          * to increment per-interface statistics.
 941          */
 942         if (ifpp)
 943                 *ifpp = ifp;
 944
 945         /* Determine path MTU. */
 946         if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
 947                     &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
 948                 goto bad;
 949         KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
 950             "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
 951             ifp, alwaysfrag, fibnum));
 952
 953         /*
 954          * The caller of this function may specify to use the minimum MTU
 955          * in some cases.
 956          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 957          * setting.  The logic is a bit complicated; by default, unicast
 958          * packets will follow path MTU while multicast packets will be sent at
 959          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 960          * including unicast ones will be sent at the minimum MTU.  Multicast
 961          * packets will always be sent at the minimum MTU unless
 962          * IP6PO_MINMTU_DISABLE is explicitly specified.
 963          * See RFC 3542 for more details.
 964          */
 965         if (mtu > IPV6_MMTU) {
 966                 if ((flags & IPV6_MINMTU))
 967                         mtu = IPV6_MMTU;
 968                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 969                         mtu = IPV6_MMTU;
 970                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 971                          (opt == NULL ||
 972                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 973                         mtu = IPV6_MMTU;
 974                 }
 975         }
 976
 977         /*
 978          * Clear embedded scope identifiers if necessary.
 979          * in6_clearscope() will touch the addresses only when necessary.
 980          */
 981         in6_clearscope(&ip6->ip6_src);
 982         in6_clearscope(&ip6->ip6_dst);
 983
 984         /*
 985          * If the outgoing packet contains a hop-by-hop options header,
 986          * it must be examined and processed even by the source node.
 987          * (RFC 2460, section 4.)
 988          */
 989         if (exthdrs.ip6e_hbh) {
 990                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 991                 u_int32_t dummy; /* XXX unused */
 992                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 993
 994 #ifdef DIAGNOSTIC
 995                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 996                         panic("ip6e_hbh is not contiguous");
 997 #endif
 998                 /*
 999                  *  XXX: if we have to send an ICMPv6 error to the sender,
1000                  *       we need the M_LOOP flag since icmp6_error() expects
1001                  *       the IPv6 and the hop-by-hop options header are
1002                  *       contiguous unless the flag is set.
1003                  */
1004                 m->m_flags |= M_LOOP;
1005                 m->m_pkthdr.rcvif = ifp;
1006                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1007                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1008                     &dummy, &plen) < 0) {
1009                         /* m was already freed at this point. */
1010                         error = EINVAL;/* better error? */
1011                         goto done;
1012                 }
1013                 m->m_flags &= ~M_LOOP; /* XXX */
1014                 m->m_pkthdr.rcvif = NULL;
1015         }
1016
1017         /* Jump over all PFIL processing if hooks are not active. */
1018         if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1019                 goto passout;
1020
1021         odst = ip6->ip6_dst;
1022         /* Run through list of hooks for output packets. */
1023         switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
1024         case PFIL_PASS:
1025                 ip6 = mtod(m, struct ip6_hdr *);
1026                 break;
1027         case PFIL_DROPPED:
1028                 error = EACCES;
1029                 /* FALLTHROUGH */
1030         case PFIL_CONSUMED:
1031                 goto done;
1032         }
1033
1034         needfiblookup = 0;
1035         /* See if destination IP address was changed by packet filter. */
1036         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1037                 m->m_flags |= M_SKIP_FIREWALL;
1038                 /* If destination is now ourself drop to ip6_input(). */
1039                 if (in6_localip(&ip6->ip6_dst)) {
1040                         m->m_flags |= M_FASTFWD_OURS;
1041                         if (m->m_pkthdr.rcvif == NULL)
1042                                 m->m_pkthdr.rcvif = V_loif;
1043                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1044                                 m->m_pkthdr.csum_flags |=
1045                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1046                                 m->m_pkthdr.csum_data = 0xffff;
1047                         }
1048 #if defined(SCTP) || defined(SCTP_SUPPORT)
1049                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1050                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1051 #endif
1052                         error = netisr_queue(NETISR_IPV6, m);
1053                         goto done;
1054                 } else {
1055                         if (ro != NULL)
1056                                 RO_INVALIDATE_CACHE(ro);
1057                         needfiblookup = 1; /* Redo the routing table lookup. */
1058                 }
1059         }
1060         /* See if fib was changed by packet filter. */
1061         if (fibnum != M_GETFIB(m)) {
1062                 m->m_flags |= M_SKIP_FIREWALL;
1063                 fibnum = M_GETFIB(m);
1064                 if (ro != NULL)
1065                         RO_INVALIDATE_CACHE(ro);
1066                 needfiblookup = 1;
1067         }
1068         if (needfiblookup)
1069                 goto again;
1070
1071         /* See if local, if yes, send it to netisr. */
1072         if (m->m_flags & M_FASTFWD_OURS) {
1073                 if (m->m_pkthdr.rcvif == NULL)
1074                         m->m_pkthdr.rcvif = V_loif;
1075                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1076                         m->m_pkthdr.csum_flags |=
1077                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1078                         m->m_pkthdr.csum_data = 0xffff;
1079                 }
1080 #if defined(SCTP) || defined(SCTP_SUPPORT)
1081                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1082                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1083 #endif
1084                 error = netisr_queue(NETISR_IPV6, m);
1085                 goto done;
1086         }
1087         /* Or forward to some other address? */
1088         if ((m->m_flags & M_IP6_NEXTHOP) &&
1089             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1090                 if (ro != NULL)
1091                         dst = (struct sockaddr_in6 *)&ro->ro_dst;
1092                 else
1093                         dst = &sin6;
1094                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1095                 m->m_flags |= M_SKIP_FIREWALL;
1096                 m->m_flags &= ~M_IP6_NEXTHOP;
1097                 m_tag_delete(m, fwd_tag);
1098                 goto again;
1099         }
1100
1101 passout:
1102         /*
1103          * Send the packet to the outgoing interface.
1104          * If necessary, do IPv6 fragmentation before sending.
1105          *
1106          * The logic here is rather complex:
1107          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1108          * 1-a: send as is if tlen <= path mtu
1109          * 1-b: fragment if tlen > path mtu
1110          *
1111          * 2: if user asks us not to fragment (dontfrag == 1)
1112          * 2-a: send as is if tlen <= interface mtu
1113          * 2-b: error if tlen > interface mtu
1114          *
1115          * 3: if we always need to attach fragment header (alwaysfrag == 1)
1116          *      always fragment
1117          *
1118          * 4: if dontfrag == 1 && alwaysfrag == 1
1119          *      error, as we cannot handle this conflicting request.
1120          */
1121         sw_csum = m->m_pkthdr.csum_flags;
1122         if (!hdrsplit) {
1123                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
1124                 sw_csum &= ~ifp->if_hwassist;
1125         } else
1126                 tso = 0;
1127         /*
1128          * If we added extension headers, we will not do TSO and calculate the
1129          * checksums ourselves for now.
1130          * XXX-BZ  Need a framework to know when the NIC can handle it, even
1131          * with ext. hdrs.
1132          */
1133         error = ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen, false);
1134         if (error != 0)
1135                 goto bad;
1136         /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1137         tlen = m->m_pkthdr.len;
1138
1139         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1140                 dontfrag = 1;
1141         else
1142                 dontfrag = 0;
1143         if (dontfrag && alwaysfrag) {   /* Case 4. */
1144                 /* Conflicting request - can't transmit. */
1145                 error = EMSGSIZE;
1146                 goto bad;
1147         }
1148         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* Case 2-b. */
1149                 /*
1150                  * Even if the DONTFRAG option is specified, we cannot send the
1151                  * packet when the data length is larger than the MTU of the
1152                  * outgoing interface.
1153                  * Notify the error by sending IPV6_PATHMTU ancillary data if
1154                  * application wanted to know the MTU value. Also return an
1155                  * error code (this is not described in the API spec).
1156                  */
1157                 if (inp != NULL)
1158                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1159                 error = EMSGSIZE;
1160                 goto bad;
1161         }
1162
1163         /* Transmit packet without fragmentation. */
1164         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */
1165                 struct in6_ifaddr *ia6;
1166
1167                 ip6 = mtod(m, struct ip6_hdr *);
1168                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1169                 if (ia6) {
1170                         /* Record statistics for this interface address. */
1171                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1172                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
1173                             m->m_pkthdr.len);
1174                         ifa_free(&ia6->ia_ifa);
1175                 }
1176                 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1177                     (flags & IP_NO_SND_TAG_RL) ? false : true);
1178                 goto done;
1179         }
1180
1181         /* Try to fragment the packet.  Cases 1-b and 3. */
1182         if (mtu < IPV6_MMTU) {
1183                 /* Path MTU cannot be less than IPV6_MMTU. */
1184                 error = EMSGSIZE;
1185                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1186                 goto bad;
1187         } else if (ip6->ip6_plen == 0) {
1188                 /* Jumbo payload cannot be fragmented. */
1189                 error = EMSGSIZE;
1190                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1191                 goto bad;
1192         } else {
1193                 u_char nextproto;
1194
1195                 /*
1196                  * Too large for the destination or interface;
1197                  * fragment if possible.
1198                  * Must be able to put at least 8 bytes per fragment.
1199                  */
1200                 if (mtu > IPV6_MAXPACKET)
1201                         mtu = IPV6_MAXPACKET;
1202
1203                 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1204                 if (len < 8) {
1205                         error = EMSGSIZE;
1206                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1207                         goto bad;
1208                 }
1209
1210                 /*
1211                  * If the interface will not calculate checksums on
1212                  * fragmented packets, then do it here.
1213                  * XXX-BZ handle the hw offloading case.  Need flags.
1214                  */
1215                 error = ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags,
1216                     plen, optlen, true);
1217                 if (error != 0)
1218                         goto bad;
1219
1220                 /*
1221                  * Change the next header field of the last header in the
1222                  * unfragmentable part.
1223                  */
1224                 if (exthdrs.ip6e_rthdr) {
1225                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1226                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1227                 } else if (exthdrs.ip6e_dest1) {
1228                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1229                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1230                 } else if (exthdrs.ip6e_hbh) {
1231                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1232                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1233                 } else {
1234                         ip6 = mtod(m, struct ip6_hdr *);
1235                         nextproto = ip6->ip6_nxt;
1236                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1237                 }
1238
1239                 /*
1240                  * Loop through length of segment after first fragment,
1241                  * make new header and copy data of each part and link onto
1242                  * chain.
1243                  */
1244                 m0 = m;
1245                 id = htonl(ip6_randomid());
1246                 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1247                 if (error != 0)
1248                         goto sendorfree;
1249
1250                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1251         }
1252
1253         /* Remove leading garbage. */
1254 sendorfree:
1255         m = m0->m_nextpkt;
1256         m0->m_nextpkt = 0;
1257         m_freem(m0);
1258         for (; m; m = m0) {
1259                 m0 = m->m_nextpkt;
1260                 m->m_nextpkt = 0;
1261                 if (error == 0) {
1262                         /* Record statistics for this interface address. */
1263                         if (ia) {
1264                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1265                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1266                                     m->m_pkthdr.len);
1267                         }
1268                         error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1269                             true);
1270                 } else
1271                         m_freem(m);
1272         }
1273
1274         if (error == 0)
1275                 IP6STAT_INC(ip6s_fragmented);
1276
1277 done:
1278         return (error);
1279
1280 freehdrs:
1281         m_freem(exthdrs.ip6e_hbh);      /* m_freem() checks if mbuf is NULL. */
1282         m_freem(exthdrs.ip6e_dest1);
1283         m_freem(exthdrs.ip6e_rthdr);
1284         m_freem(exthdrs.ip6e_dest2);
1285         /* FALLTHROUGH */
1286 bad:
1287         if (m)
1288                 m_freem(m);
1289         goto done;
1290 }
1291
1292 static int
1293 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1294 {
1295         struct mbuf *m;
1296
1297         if (hlen > MCLBYTES)
1298                 return (ENOBUFS); /* XXX */
1299
1300         if (hlen > MLEN)
1301                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1302         else
1303                 m = m_get(M_NOWAIT, MT_DATA);
1304         if (m == NULL)
1305                 return (ENOBUFS);
1306         m->m_len = hlen;
1307         if (hdr)
1308                 bcopy(hdr, mtod(m, caddr_t), hlen);
1309
1310         *mp = m;
1311         return (0);
1312 }
1313
1314 /*
1315  * Insert jumbo payload option.
1316  */
1317 static int
1318 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1319 {
1320         struct mbuf *mopt;
1321         u_char *optbuf;
1322         u_int32_t v;
1323
1324 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1325
1326         /*
1327          * If there is no hop-by-hop options header, allocate new one.
1328          * If there is one but it doesn't have enough space to store the
1329          * jumbo payload option, allocate a cluster to store the whole options.
1330          * Otherwise, use it to store the options.
1331          */
1332         if (exthdrs->ip6e_hbh == NULL) {
1333                 mopt = m_get(M_NOWAIT, MT_DATA);
1334                 if (mopt == NULL)
1335                         return (ENOBUFS);
1336                 mopt->m_len = JUMBOOPTLEN;
1337                 optbuf = mtod(mopt, u_char *);
1338                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1339                 exthdrs->ip6e_hbh = mopt;
1340         } else {
1341                 struct ip6_hbh *hbh;
1342
1343                 mopt = exthdrs->ip6e_hbh;
1344                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1345                         /*
1346                          * XXX assumption:
1347                          * - exthdrs->ip6e_hbh is not referenced from places
1348                          *   other than exthdrs.
1349                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1350                          */
1351                         int oldoptlen = mopt->m_len;
1352                         struct mbuf *n;
1353
1354                         /*
1355                          * XXX: give up if the whole (new) hbh header does
1356                          * not fit even in an mbuf cluster.
1357                          */
1358                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1359                                 return (ENOBUFS);
1360
1361                         /*
1362                          * As a consequence, we must always prepare a cluster
1363                          * at this point.
1364                          */
1365                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1366                         if (n == NULL)
1367                                 return (ENOBUFS);
1368                         n->m_len = oldoptlen + JUMBOOPTLEN;
1369                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1370                             oldoptlen);
1371                         optbuf = mtod(n, caddr_t) + oldoptlen;
1372                         m_freem(mopt);
1373                         mopt = exthdrs->ip6e_hbh = n;
1374                 } else {
1375                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1376                         mopt->m_len += JUMBOOPTLEN;
1377                 }
1378                 optbuf[0] = IP6OPT_PADN;
1379                 optbuf[1] = 1;
1380
1381                 /*
1382                  * Adjust the header length according to the pad and
1383                  * the jumbo payload option.
1384                  */
1385                 hbh = mtod(mopt, struct ip6_hbh *);
1386                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1387         }
1388
1389         /* fill in the option. */
1390         optbuf[2] = IP6OPT_JUMBO;
1391         optbuf[3] = 4;
1392         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1393         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1394
1395         /* finally, adjust the packet header length */
1396         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1397
1398         return (0);
1399 #undef JUMBOOPTLEN
1400 }
1401
1402 /*
1403  * Insert fragment header and copy unfragmentable header portions.
1404  */
1405 static int
1406 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1407     struct ip6_frag **frghdrp)
1408 {
1409         struct mbuf *n, *mlast;
1410
1411         if (hlen > sizeof(struct ip6_hdr)) {
1412                 n = m_copym(m0, sizeof(struct ip6_hdr),
1413                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1414                 if (n == NULL)
1415                         return (ENOBUFS);
1416                 m->m_next = n;
1417         } else
1418                 n = m;
1419
1420         /* Search for the last mbuf of unfragmentable part. */
1421         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1422                 ;
1423
1424         if (M_WRITABLE(mlast) &&
1425             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1426                 /* use the trailing space of the last mbuf for the fragment hdr */
1427                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1428                     mlast->m_len);
1429                 mlast->m_len += sizeof(struct ip6_frag);
1430                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1431         } else {
1432                 /* allocate a new mbuf for the fragment header */
1433                 struct mbuf *mfrg;
1434
1435                 mfrg = m_get(M_NOWAIT, MT_DATA);
1436                 if (mfrg == NULL)
1437                         return (ENOBUFS);
1438                 mfrg->m_len = sizeof(struct ip6_frag);
1439                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1440                 mlast->m_next = mfrg;
1441         }
1442
1443         return (0);
1444 }
1445
1446 /*
1447  * Calculates IPv6 path mtu for destination @dst.
1448  * Resulting MTU is stored in @mtup.
1449  *
1450  * Returns 0 on success.
1451  */
1452 static int
1453 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1454 {
1455         struct epoch_tracker et;
1456         struct nhop_object *nh;
1457         struct in6_addr kdst;
1458         uint32_t scopeid;
1459         int error;
1460
1461         in6_splitscope(dst, &kdst, &scopeid);
1462
1463         NET_EPOCH_ENTER(et);
1464         nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1465         if (nh != NULL)
1466                 error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1467         else
1468                 error = EHOSTUNREACH;
1469         NET_EPOCH_EXIT(et);
1470
1471         return (error);
1472 }
1473
1474 /*
1475  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1476  * and cached data in @ro_pmtu.
1477  * MTU from (successful) route lookup is saved (along with dst)
1478  * inside @ro_pmtu to avoid subsequent route lookups after packet
1479  * filter processing.
1480  *
1481  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1482  * Returns 0 on success.
1483  */
1484 static int
1485 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1486     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1487     int *alwaysfragp, u_int fibnum, u_int proto)
1488 {
1489         struct nhop_object *nh;
1490         struct in6_addr kdst;
1491         uint32_t scopeid;
1492         struct sockaddr_in6 *sa6_dst, sin6;
1493         u_long mtu;
1494
1495         NET_EPOCH_ASSERT();
1496
1497         mtu = 0;
1498         if (ro_pmtu == NULL || do_lookup) {
1499
1500                 /*
1501                  * Here ro_pmtu has final destination address, while
1502                  * ro might represent immediate destination.
1503                  * Use ro_pmtu destination since mtu might differ.
1504                  */
1505                 if (ro_pmtu != NULL) {
1506                         sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1507                         if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1508                                 ro_pmtu->ro_mtu = 0;
1509                 } else
1510                         sa6_dst = &sin6;
1511
1512                 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1513                         bzero(sa6_dst, sizeof(*sa6_dst));
1514                         sa6_dst->sin6_family = AF_INET6;
1515                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1516                         sa6_dst->sin6_addr = *dst;
1517
1518                         in6_splitscope(dst, &kdst, &scopeid);
1519                         nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1520                         if (nh != NULL) {
1521                                 mtu = nh->nh_mtu;
1522                                 if (ro_pmtu != NULL)
1523                                         ro_pmtu->ro_mtu = mtu;
1524                         }
1525                 } else
1526                         mtu = ro_pmtu->ro_mtu;
1527         }
1528
1529         if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1530                 mtu = ro_pmtu->ro_nh->nh_mtu;
1531
1532         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1533 }
1534
1535 /*
1536  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1537  * hostcache data for @dst.
1538  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1539  *
1540  * Returns 0 on success.
1541  */
1542 static int
1543 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1544     u_long *mtup, int *alwaysfragp, u_int proto)
1545 {
1546         u_long mtu = 0;
1547         int alwaysfrag = 0;
1548         int error = 0;
1549
1550         if (rt_mtu > 0) {
1551                 u_int32_t ifmtu;
1552                 struct in_conninfo inc;
1553
1554                 bzero(&inc, sizeof(inc));
1555                 inc.inc_flags |= INC_ISIPV6;
1556                 inc.inc6_faddr = *dst;
1557
1558                 ifmtu = IN6_LINKMTU(ifp);
1559
1560                 /* TCP is known to react to pmtu changes so skip hc */
1561                 if (proto != IPPROTO_TCP)
1562                         mtu = tcp_hc_getmtu(&inc);
1563
1564                 if (mtu)
1565                         mtu = min(mtu, rt_mtu);
1566                 else
1567                         mtu = rt_mtu;
1568                 if (mtu == 0)
1569                         mtu = ifmtu;
1570                 else if (mtu < IPV6_MMTU) {
1571                         /*
1572                          * RFC2460 section 5, last paragraph:
1573                          * if we record ICMPv6 too big message with
1574                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1575                          * or smaller, with framgent header attached.
1576                          * (fragment header is needed regardless from the
1577                          * packet size, for translators to identify packets)
1578                          */
1579                         alwaysfrag = 1;
1580                         mtu = IPV6_MMTU;
1581                 }
1582         } else if (ifp) {
1583                 mtu = IN6_LINKMTU(ifp);
1584         } else
1585                 error = EHOSTUNREACH; /* XXX */
1586
1587         *mtup = mtu;
1588         if (alwaysfragp)
1589                 *alwaysfragp = alwaysfrag;
1590         return (error);
1591 }
1592
1593 /*
1594  * IP6 socket option processing.
1595  */
1596 int
1597 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1598 {
1599         int optdatalen, uproto;
1600         void *optdata;
1601         struct inpcb *inp = sotoinpcb(so);
1602         int error, optval;
1603         int level, op, optname;
1604         int optlen;
1605         struct thread *td;
1606 #ifdef  RSS
1607         uint32_t rss_bucket;
1608         int retval;
1609 #endif
1610
1611 /*
1612  * Don't use more than a quarter of mbuf clusters.  N.B.:
1613  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1614  * on LP64 architectures, so cast to u_long to avoid undefined
1615  * behavior.  ILP32 architectures cannot have nmbclusters
1616  * large enough to overflow for other reasons.
1617  */
1618 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1619
1620         level = sopt->sopt_level;
1621         op = sopt->sopt_dir;
1622         optname = sopt->sopt_name;
1623         optlen = sopt->sopt_valsize;
1624         td = sopt->sopt_td;
1625         error = 0;
1626         optval = 0;
1627         uproto = (int)so->so_proto->pr_protocol;
1628
1629         if (level != IPPROTO_IPV6) {
1630                 error = EINVAL;
1631
1632                 if (sopt->sopt_level == SOL_SOCKET &&
1633                     sopt->sopt_dir == SOPT_SET) {
1634                         switch (sopt->sopt_name) {
1635                         case SO_REUSEADDR:
1636                                 INP_WLOCK(inp);
1637                                 if ((so->so_options & SO_REUSEADDR) != 0)
1638                                         inp->inp_flags2 |= INP_REUSEADDR;
1639                                 else
1640                                         inp->inp_flags2 &= ~INP_REUSEADDR;
1641                                 INP_WUNLOCK(inp);
1642                                 error = 0;
1643                                 break;
1644                         case SO_REUSEPORT:
1645                                 INP_WLOCK(inp);
1646                                 if ((so->so_options & SO_REUSEPORT) != 0)
1647                                         inp->inp_flags2 |= INP_REUSEPORT;
1648                                 else
1649                                         inp->inp_flags2 &= ~INP_REUSEPORT;
1650                                 INP_WUNLOCK(inp);
1651                                 error = 0;
1652                                 break;
1653                         case SO_REUSEPORT_LB:
1654                                 INP_WLOCK(inp);
1655                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1656                                         inp->inp_flags2 |= INP_REUSEPORT_LB;
1657                                 else
1658                                         inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1659                                 INP_WUNLOCK(inp);
1660                                 error = 0;
1661                                 break;
1662                         case SO_SETFIB:
1663                                 INP_WLOCK(inp);
1664                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
1665                                 INP_WUNLOCK(inp);
1666                                 error = 0;
1667                                 break;
1668                         case SO_MAX_PACING_RATE:
1669 #ifdef RATELIMIT
1670                                 INP_WLOCK(inp);
1671                                 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1672                                 INP_WUNLOCK(inp);
1673                                 error = 0;
1674 #else
1675                                 error = EOPNOTSUPP;
1676 #endif
1677                                 break;
1678                         default:
1679                                 break;
1680                         }
1681                 }
1682         } else {                /* level == IPPROTO_IPV6 */
1683                 switch (op) {
1684
1685                 case SOPT_SET:
1686                         switch (optname) {
1687                         case IPV6_2292PKTOPTIONS:
1688 #ifdef IPV6_PKTOPTIONS
1689                         case IPV6_PKTOPTIONS:
1690 #endif
1691                         {
1692                                 struct mbuf *m;
1693
1694                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1695                                         printf("ip6_ctloutput: mbuf limit hit\n");
1696                                         error = ENOBUFS;
1697                                         break;
1698                                 }
1699
1700                                 error = soopt_getm(sopt, &m); /* XXX */
1701                                 if (error != 0)
1702                                         break;
1703                                 error = soopt_mcopyin(sopt, m); /* XXX */
1704                                 if (error != 0)
1705                                         break;
1706                                 INP_WLOCK(inp);
1707                                 error = ip6_pcbopts(&inp->in6p_outputopts, m,
1708                                     so, sopt);
1709                                 INP_WUNLOCK(inp);
1710                                 m_freem(m); /* XXX */
1711                                 break;
1712                         }
1713
1714                         /*
1715                          * Use of some Hop-by-Hop options or some
1716                          * Destination options, might require special
1717                          * privilege.  That is, normal applications
1718                          * (without special privilege) might be forbidden
1719                          * from setting certain options in outgoing packets,
1720                          * and might never see certain options in received
1721                          * packets. [RFC 2292 Section 6]
1722                          * KAME specific note:
1723                          *  KAME prevents non-privileged users from sending or
1724                          *  receiving ANY hbh/dst options in order to avoid
1725                          *  overhead of parsing options in the kernel.
1726                          */
1727                         case IPV6_RECVHOPOPTS:
1728                         case IPV6_RECVDSTOPTS:
1729                         case IPV6_RECVRTHDRDSTOPTS:
1730                                 if (td != NULL) {
1731                                         error = priv_check(td,
1732                                             PRIV_NETINET_SETHDROPTS);
1733                                         if (error)
1734                                                 break;
1735                                 }
1736                                 /* FALLTHROUGH */
1737                         case IPV6_UNICAST_HOPS:
1738                         case IPV6_HOPLIMIT:
1739
1740                         case IPV6_RECVPKTINFO:
1741                         case IPV6_RECVHOPLIMIT:
1742                         case IPV6_RECVRTHDR:
1743                         case IPV6_RECVPATHMTU:
1744                         case IPV6_RECVTCLASS:
1745                         case IPV6_RECVFLOWID:
1746 #ifdef  RSS
1747                         case IPV6_RECVRSSBUCKETID:
1748 #endif
1749                         case IPV6_V6ONLY:
1750                         case IPV6_AUTOFLOWLABEL:
1751                         case IPV6_ORIGDSTADDR:
1752                         case IPV6_BINDANY:
1753                         case IPV6_BINDMULTI:
1754 #ifdef  RSS
1755                         case IPV6_RSS_LISTEN_BUCKET:
1756 #endif
1757                                 if (optname == IPV6_BINDANY && td != NULL) {
1758                                         error = priv_check(td,
1759                                             PRIV_NETINET_BINDANY);
1760                                         if (error)
1761                                                 break;
1762                                 }
1763
1764                                 if (optlen != sizeof(int)) {
1765                                         error = EINVAL;
1766                                         break;
1767                                 }
1768                                 error = sooptcopyin(sopt, &optval,
1769                                         sizeof optval, sizeof optval);
1770                                 if (error)
1771                                         break;
1772                                 switch (optname) {
1773
1774                                 case IPV6_UNICAST_HOPS:
1775                                         if (optval < -1 || optval >= 256)
1776                                                 error = EINVAL;
1777                                         else {
1778                                                 /* -1 = kernel default */
1779                                                 inp->in6p_hops = optval;
1780                                                 if ((inp->inp_vflag &
1781                                                      INP_IPV4) != 0)
1782                                                         inp->inp_ip_ttl = optval;
1783                                         }
1784                                         break;
1785 #define OPTSET(bit) \
1786 do { \
1787         INP_WLOCK(inp); \
1788         if (optval) \
1789                 inp->inp_flags |= (bit); \
1790         else \
1791                 inp->inp_flags &= ~(bit); \
1792         INP_WUNLOCK(inp); \
1793 } while (/*CONSTCOND*/ 0)
1794 #define OPTSET2292(bit) \
1795 do { \
1796         INP_WLOCK(inp); \
1797         inp->inp_flags |= IN6P_RFC2292; \
1798         if (optval) \
1799                 inp->inp_flags |= (bit); \
1800         else \
1801                 inp->inp_flags &= ~(bit); \
1802         INP_WUNLOCK(inp); \
1803 } while (/*CONSTCOND*/ 0)
1804 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1805
1806 #define OPTSET2_N(bit, val) do {                                        \
1807         if (val)                                                        \
1808                 inp->inp_flags2 |= bit;                                 \
1809         else                                                            \
1810                 inp->inp_flags2 &= ~bit;                                \
1811 } while (0)
1812 #define OPTSET2(bit, val) do {                                          \
1813         INP_WLOCK(inp);                                                 \
1814         OPTSET2_N(bit, val);                                            \
1815         INP_WUNLOCK(inp);                                               \
1816 } while (0)
1817 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1818 #define OPTSET2292_EXCLUSIVE(bit)                                       \
1819 do {                                                                    \
1820         INP_WLOCK(inp);                                                 \
1821         if (OPTBIT(IN6P_RFC2292)) {                                     \
1822                 error = EINVAL;                                         \
1823         } else {                                                        \
1824                 if (optval)                                             \
1825                         inp->inp_flags |= (bit);                        \
1826                 else                                                    \
1827                         inp->inp_flags &= ~(bit);                       \
1828         }                                                               \
1829         INP_WUNLOCK(inp);                                               \
1830 } while (/*CONSTCOND*/ 0)
1831
1832                                 case IPV6_RECVPKTINFO:
1833                                         OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1834                                         break;
1835
1836                                 case IPV6_HOPLIMIT:
1837                                 {
1838                                         struct ip6_pktopts **optp;
1839
1840                                         /* cannot mix with RFC2292 */
1841                                         if (OPTBIT(IN6P_RFC2292)) {
1842                                                 error = EINVAL;
1843                                                 break;
1844                                         }
1845                                         INP_WLOCK(inp);
1846                                         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1847                                                 INP_WUNLOCK(inp);
1848                                                 return (ECONNRESET);
1849                                         }
1850                                         optp = &inp->in6p_outputopts;
1851                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1852                                             (u_char *)&optval, sizeof(optval),
1853                                             optp, (td != NULL) ? td->td_ucred :
1854                                             NULL, uproto);
1855                                         INP_WUNLOCK(inp);
1856                                         break;
1857                                 }
1858
1859                                 case IPV6_RECVHOPLIMIT:
1860                                         OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1861                                         break;
1862
1863                                 case IPV6_RECVHOPOPTS:
1864                                         OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1865                                         break;
1866
1867                                 case IPV6_RECVDSTOPTS:
1868                                         OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1869                                         break;
1870
1871                                 case IPV6_RECVRTHDRDSTOPTS:
1872                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1873                                         break;
1874
1875                                 case IPV6_RECVRTHDR:
1876                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1877                                         break;
1878
1879                                 case IPV6_RECVPATHMTU:
1880                                         /*
1881                                          * We ignore this option for TCP
1882                                          * sockets.
1883                                          * (RFC3542 leaves this case
1884                                          * unspecified.)
1885                                          */
1886                                         if (uproto != IPPROTO_TCP)
1887                                                 OPTSET(IN6P_MTU);
1888                                         break;
1889
1890                                 case IPV6_RECVFLOWID:
1891                                         OPTSET2(INP_RECVFLOWID, optval);
1892                                         break;
1893
1894 #ifdef  RSS
1895                                 case IPV6_RECVRSSBUCKETID:
1896                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1897                                         break;
1898 #endif
1899
1900                                 case IPV6_V6ONLY:
1901                                         INP_WLOCK(inp);
1902                                         if (inp->inp_lport ||
1903                                             !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1904                                                 /*
1905                                                  * The socket is already bound.
1906                                                  */
1907                                                 INP_WUNLOCK(inp);
1908                                                 error = EINVAL;
1909                                                 break;
1910                                         }
1911                                         if (optval) {
1912                                                 inp->inp_flags |= IN6P_IPV6_V6ONLY;
1913                                                 inp->inp_vflag &= ~INP_IPV4;
1914                                         } else {
1915                                                 inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1916                                                 inp->inp_vflag |= INP_IPV4;
1917                                         }
1918                                         INP_WUNLOCK(inp);
1919                                         break;
1920                                 case IPV6_RECVTCLASS:
1921                                         /* cannot mix with RFC2292 XXX */
1922                                         OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1923                                         break;
1924                                 case IPV6_AUTOFLOWLABEL:
1925                                         OPTSET(IN6P_AUTOFLOWLABEL);
1926                                         break;
1927
1928                                 case IPV6_ORIGDSTADDR:
1929                                         OPTSET2(INP_ORIGDSTADDR, optval);
1930                                         break;
1931                                 case IPV6_BINDANY:
1932                                         OPTSET(INP_BINDANY);
1933                                         break;
1934
1935                                 case IPV6_BINDMULTI:
1936                                         OPTSET2(INP_BINDMULTI, optval);
1937                                         break;
1938 #ifdef  RSS
1939                                 case IPV6_RSS_LISTEN_BUCKET:
1940                                         if ((optval >= 0) &&
1941                                             (optval < rss_getnumbuckets())) {
1942                                                 INP_WLOCK(inp);
1943                                                 inp->inp_rss_listen_bucket = optval;
1944                                                 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1945                                                 INP_WUNLOCK(inp);
1946                                         } else {
1947                                                 error = EINVAL;
1948                                         }
1949                                         break;
1950 #endif
1951                                 }
1952                                 break;
1953
1954                         case IPV6_TCLASS:
1955                         case IPV6_DONTFRAG:
1956                         case IPV6_USE_MIN_MTU:
1957                         case IPV6_PREFER_TEMPADDR:
1958                                 if (optlen != sizeof(optval)) {
1959                                         error = EINVAL;
1960                                         break;
1961                                 }
1962                                 error = sooptcopyin(sopt, &optval,
1963                                         sizeof optval, sizeof optval);
1964                                 if (error)
1965                                         break;
1966                                 {
1967                                         struct ip6_pktopts **optp;
1968                                         INP_WLOCK(inp);
1969                                         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1970                                                 INP_WUNLOCK(inp);
1971                                                 return (ECONNRESET);
1972                                         }
1973                                         optp = &inp->in6p_outputopts;
1974                                         error = ip6_pcbopt(optname,
1975                                             (u_char *)&optval, sizeof(optval),
1976                                             optp, (td != NULL) ? td->td_ucred :
1977                                             NULL, uproto);
1978                                         INP_WUNLOCK(inp);
1979                                         break;
1980                                 }
1981
1982                         case IPV6_2292PKTINFO:
1983                         case IPV6_2292HOPLIMIT:
1984                         case IPV6_2292HOPOPTS:
1985                         case IPV6_2292DSTOPTS:
1986                         case IPV6_2292RTHDR:
1987                                 /* RFC 2292 */
1988                                 if (optlen != sizeof(int)) {
1989                                         error = EINVAL;
1990                                         break;
1991                                 }
1992                                 error = sooptcopyin(sopt, &optval,
1993                                         sizeof optval, sizeof optval);
1994                                 if (error)
1995                                         break;
1996                                 switch (optname) {
1997                                 case IPV6_2292PKTINFO:
1998                                         OPTSET2292(IN6P_PKTINFO);
1999                                         break;
2000                                 case IPV6_2292HOPLIMIT:
2001                                         OPTSET2292(IN6P_HOPLIMIT);
2002                                         break;
2003                                 case IPV6_2292HOPOPTS:
2004                                         /*
2005                                          * Check super-user privilege.
2006                                          * See comments for IPV6_RECVHOPOPTS.
2007                                          */
2008                                         if (td != NULL) {
2009                                                 error = priv_check(td,
2010                                                     PRIV_NETINET_SETHDROPTS);
2011                                                 if (error)
2012                                                         return (error);
2013                                         }
2014                                         OPTSET2292(IN6P_HOPOPTS);
2015                                         break;
2016                                 case IPV6_2292DSTOPTS:
2017                                         if (td != NULL) {
2018                                                 error = priv_check(td,
2019                                                     PRIV_NETINET_SETHDROPTS);
2020                                                 if (error)
2021                                                         return (error);
2022                                         }
2023                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2024                                         break;
2025                                 case IPV6_2292RTHDR:
2026                                         OPTSET2292(IN6P_RTHDR);
2027                                         break;
2028                                 }
2029                                 break;
2030                         case IPV6_PKTINFO:
2031                         case IPV6_HOPOPTS:
2032                         case IPV6_RTHDR:
2033                         case IPV6_DSTOPTS:
2034                         case IPV6_RTHDRDSTOPTS:
2035                         case IPV6_NEXTHOP:
2036                         {
2037                                 /* new advanced API (RFC3542) */
2038                                 u_char *optbuf;
2039                                 u_char optbuf_storage[MCLBYTES];
2040                                 int optlen;
2041                                 struct ip6_pktopts **optp;
2042
2043                                 /* cannot mix with RFC2292 */
2044                                 if (OPTBIT(IN6P_RFC2292)) {
2045                                         error = EINVAL;
2046                                         break;
2047                                 }
2048
2049                                 /*
2050                                  * We only ensure valsize is not too large
2051                                  * here.  Further validation will be done
2052                                  * later.
2053                                  */
2054                                 error = sooptcopyin(sopt, optbuf_storage,
2055                                     sizeof(optbuf_storage), 0);
2056                                 if (error)
2057                                         break;
2058                                 optlen = sopt->sopt_valsize;
2059                                 optbuf = optbuf_storage;
2060                                 INP_WLOCK(inp);
2061                                 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2062                                         INP_WUNLOCK(inp);
2063                                         return (ECONNRESET);
2064                                 }
2065                                 optp = &inp->in6p_outputopts;
2066                                 error = ip6_pcbopt(optname, optbuf, optlen,
2067                                     optp, (td != NULL) ? td->td_ucred : NULL,
2068                                     uproto);
2069                                 INP_WUNLOCK(inp);
2070                                 break;
2071                         }
2072 #undef OPTSET
2073
2074                         case IPV6_MULTICAST_IF:
2075                         case IPV6_MULTICAST_HOPS:
2076                         case IPV6_MULTICAST_LOOP:
2077                         case IPV6_JOIN_GROUP:
2078                         case IPV6_LEAVE_GROUP:
2079                         case IPV6_MSFILTER:
2080                         case MCAST_BLOCK_SOURCE:
2081                         case MCAST_UNBLOCK_SOURCE:
2082                         case MCAST_JOIN_GROUP:
2083                         case MCAST_LEAVE_GROUP:
2084                         case MCAST_JOIN_SOURCE_GROUP:
2085                         case MCAST_LEAVE_SOURCE_GROUP:
2086                                 error = ip6_setmoptions(inp, sopt);
2087                                 break;
2088
2089                         case IPV6_PORTRANGE:
2090                                 error = sooptcopyin(sopt, &optval,
2091                                     sizeof optval, sizeof optval);
2092                                 if (error)
2093                                         break;
2094
2095                                 INP_WLOCK(inp);
2096                                 switch (optval) {
2097                                 case IPV6_PORTRANGE_DEFAULT:
2098                                         inp->inp_flags &= ~(INP_LOWPORT);
2099                                         inp->inp_flags &= ~(INP_HIGHPORT);
2100                                         break;
2101
2102                                 case IPV6_PORTRANGE_HIGH:
2103                                         inp->inp_flags &= ~(INP_LOWPORT);
2104                                         inp->inp_flags |= INP_HIGHPORT;
2105                                         break;
2106
2107                                 case IPV6_PORTRANGE_LOW:
2108                                         inp->inp_flags &= ~(INP_HIGHPORT);
2109                                         inp->inp_flags |= INP_LOWPORT;
2110                                         break;
2111
2112                                 default:
2113                                         error = EINVAL;
2114                                         break;
2115                                 }
2116                                 INP_WUNLOCK(inp);
2117                                 break;
2118
2119 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2120                         case IPV6_IPSEC_POLICY:
2121                                 if (IPSEC_ENABLED(ipv6)) {
2122                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2123                                         break;
2124                                 }
2125                                 /* FALLTHROUGH */
2126 #endif /* IPSEC */
2127
2128                         default:
2129                                 error = ENOPROTOOPT;
2130                                 break;
2131                         }
2132                         break;
2133
2134                 case SOPT_GET:
2135                         switch (optname) {
2136
2137                         case IPV6_2292PKTOPTIONS:
2138 #ifdef IPV6_PKTOPTIONS
2139                         case IPV6_PKTOPTIONS:
2140 #endif
2141                                 /*
2142                                  * RFC3542 (effectively) deprecated the
2143                                  * semantics of the 2292-style pktoptions.
2144                                  * Since it was not reliable in nature (i.e.,
2145                                  * applications had to expect the lack of some
2146                                  * information after all), it would make sense
2147                                  * to simplify this part by always returning
2148                                  * empty data.
2149                                  */
2150                                 sopt->sopt_valsize = 0;
2151                                 break;
2152
2153                         case IPV6_RECVHOPOPTS:
2154                         case IPV6_RECVDSTOPTS:
2155                         case IPV6_RECVRTHDRDSTOPTS:
2156                         case IPV6_UNICAST_HOPS:
2157                         case IPV6_RECVPKTINFO:
2158                         case IPV6_RECVHOPLIMIT:
2159                         case IPV6_RECVRTHDR:
2160                         case IPV6_RECVPATHMTU:
2161
2162                         case IPV6_V6ONLY:
2163                         case IPV6_PORTRANGE:
2164                         case IPV6_RECVTCLASS:
2165                         case IPV6_AUTOFLOWLABEL:
2166                         case IPV6_BINDANY:
2167                         case IPV6_FLOWID:
2168                         case IPV6_FLOWTYPE:
2169                         case IPV6_RECVFLOWID:
2170 #ifdef  RSS
2171                         case IPV6_RSSBUCKETID:
2172                         case IPV6_RECVRSSBUCKETID:
2173 #endif
2174                         case IPV6_BINDMULTI:
2175                                 switch (optname) {
2176
2177                                 case IPV6_RECVHOPOPTS:
2178                                         optval = OPTBIT(IN6P_HOPOPTS);
2179                                         break;
2180
2181                                 case IPV6_RECVDSTOPTS:
2182                                         optval = OPTBIT(IN6P_DSTOPTS);
2183                                         break;
2184
2185                                 case IPV6_RECVRTHDRDSTOPTS:
2186                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2187                                         break;
2188
2189                                 case IPV6_UNICAST_HOPS:
2190                                         optval = inp->in6p_hops;
2191                                         break;
2192
2193                                 case IPV6_RECVPKTINFO:
2194                                         optval = OPTBIT(IN6P_PKTINFO);
2195                                         break;
2196
2197                                 case IPV6_RECVHOPLIMIT:
2198                                         optval = OPTBIT(IN6P_HOPLIMIT);
2199                                         break;
2200
2201                                 case IPV6_RECVRTHDR:
2202                                         optval = OPTBIT(IN6P_RTHDR);
2203                                         break;
2204
2205                                 case IPV6_RECVPATHMTU:
2206                                         optval = OPTBIT(IN6P_MTU);
2207                                         break;
2208
2209                                 case IPV6_V6ONLY:
2210                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2211                                         break;
2212
2213                                 case IPV6_PORTRANGE:
2214                                     {
2215                                         int flags;
2216                                         flags = inp->inp_flags;
2217                                         if (flags & INP_HIGHPORT)
2218                                                 optval = IPV6_PORTRANGE_HIGH;
2219                                         else if (flags & INP_LOWPORT)
2220                                                 optval = IPV6_PORTRANGE_LOW;
2221                                         else
2222                                                 optval = 0;
2223                                         break;
2224                                     }
2225                                 case IPV6_RECVTCLASS:
2226                                         optval = OPTBIT(IN6P_TCLASS);
2227                                         break;
2228
2229                                 case IPV6_AUTOFLOWLABEL:
2230                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2231                                         break;
2232
2233                                 case IPV6_ORIGDSTADDR:
2234                                         optval = OPTBIT2(INP_ORIGDSTADDR);
2235                                         break;
2236
2237                                 case IPV6_BINDANY:
2238                                         optval = OPTBIT(INP_BINDANY);
2239                                         break;
2240
2241                                 case IPV6_FLOWID:
2242                                         optval = inp->inp_flowid;
2243                                         break;
2244
2245                                 case IPV6_FLOWTYPE:
2246                                         optval = inp->inp_flowtype;
2247                                         break;
2248
2249                                 case IPV6_RECVFLOWID:
2250                                         optval = OPTBIT2(INP_RECVFLOWID);
2251                                         break;
2252 #ifdef  RSS
2253                                 case IPV6_RSSBUCKETID:
2254                                         retval =
2255                                             rss_hash2bucket(inp->inp_flowid,
2256                                             inp->inp_flowtype,
2257                                             &rss_bucket);
2258                                         if (retval == 0)
2259                                                 optval = rss_bucket;
2260                                         else
2261                                                 error = EINVAL;
2262                                         break;
2263
2264                                 case IPV6_RECVRSSBUCKETID:
2265                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2266                                         break;
2267 #endif
2268
2269                                 case IPV6_BINDMULTI:
2270                                         optval = OPTBIT2(INP_BINDMULTI);
2271                                         break;
2272
2273                                 }
2274                                 if (error)
2275                                         break;
2276                                 error = sooptcopyout(sopt, &optval,
2277                                         sizeof optval);
2278                                 break;
2279
2280                         case IPV6_PATHMTU:
2281                         {
2282                                 u_long pmtu = 0;
2283                                 struct ip6_mtuinfo mtuinfo;
2284                                 struct in6_addr addr;
2285
2286                                 if (!(so->so_state & SS_ISCONNECTED))
2287                                         return (ENOTCONN);
2288                                 /*
2289                                  * XXX: we dot not consider the case of source
2290                                  * routing, or optional information to specify
2291                                  * the outgoing interface.
2292                                  * Copy faddr out of inp to avoid holding lock
2293                                  * on inp during route lookup.
2294                                  */
2295                                 INP_RLOCK(inp);
2296                                 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2297                                 INP_RUNLOCK(inp);
2298                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2299                                     &addr, &pmtu);
2300                                 if (error)
2301                                         break;
2302                                 if (pmtu > IPV6_MAXPACKET)
2303                                         pmtu = IPV6_MAXPACKET;
2304
2305                                 bzero(&mtuinfo, sizeof(mtuinfo));
2306                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2307                                 optdata = (void *)&mtuinfo;
2308                                 optdatalen = sizeof(mtuinfo);
2309                                 error = sooptcopyout(sopt, optdata,
2310                                     optdatalen);
2311                                 break;
2312                         }
2313
2314                         case IPV6_2292PKTINFO:
2315                         case IPV6_2292HOPLIMIT:
2316                         case IPV6_2292HOPOPTS:
2317                         case IPV6_2292RTHDR:
2318                         case IPV6_2292DSTOPTS:
2319                                 switch (optname) {
2320                                 case IPV6_2292PKTINFO:
2321                                         optval = OPTBIT(IN6P_PKTINFO);
2322                                         break;
2323                                 case IPV6_2292HOPLIMIT:
2324                                         optval = OPTBIT(IN6P_HOPLIMIT);
2325                                         break;
2326                                 case IPV6_2292HOPOPTS:
2327                                         optval = OPTBIT(IN6P_HOPOPTS);
2328                                         break;
2329                                 case IPV6_2292RTHDR:
2330                                         optval = OPTBIT(IN6P_RTHDR);
2331                                         break;
2332                                 case IPV6_2292DSTOPTS:
2333                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2334                                         break;
2335                                 }
2336                                 error = sooptcopyout(sopt, &optval,
2337                                     sizeof optval);
2338                                 break;
2339                         case IPV6_PKTINFO:
2340                         case IPV6_HOPOPTS:
2341                         case IPV6_RTHDR:
2342                         case IPV6_DSTOPTS:
2343                         case IPV6_RTHDRDSTOPTS:
2344                         case IPV6_NEXTHOP:
2345                         case IPV6_TCLASS:
2346                         case IPV6_DONTFRAG:
2347                         case IPV6_USE_MIN_MTU:
2348                         case IPV6_PREFER_TEMPADDR:
2349                                 error = ip6_getpcbopt(inp, optname, sopt);
2350                                 break;
2351
2352                         case IPV6_MULTICAST_IF:
2353                         case IPV6_MULTICAST_HOPS:
2354                         case IPV6_MULTICAST_LOOP:
2355                         case IPV6_MSFILTER:
2356                                 error = ip6_getmoptions(inp, sopt);
2357                                 break;
2358
2359 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2360                         case IPV6_IPSEC_POLICY:
2361                                 if (IPSEC_ENABLED(ipv6)) {
2362                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2363                                         break;
2364                                 }
2365                                 /* FALLTHROUGH */
2366 #endif /* IPSEC */
2367                         default:
2368                                 error = ENOPROTOOPT;
2369                                 break;
2370                         }
2371                         break;
2372                 }
2373         }
2374         return (error);
2375 }
2376
2377 int
2378 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2379 {
2380         int error = 0, optval, optlen;
2381         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2382         struct inpcb *inp = sotoinpcb(so);
2383         int level, op, optname;
2384
2385         level = sopt->sopt_level;
2386         op = sopt->sopt_dir;
2387         optname = sopt->sopt_name;
2388         optlen = sopt->sopt_valsize;
2389
2390         if (level != IPPROTO_IPV6) {
2391                 return (EINVAL);
2392         }
2393
2394         switch (optname) {
2395         case IPV6_CHECKSUM:
2396                 /*
2397                  * For ICMPv6 sockets, no modification allowed for checksum
2398                  * offset, permit "no change" values to help existing apps.
2399                  *
2400                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2401                  * for an ICMPv6 socket will fail."
2402                  * The current behavior does not meet RFC3542.
2403                  */
2404                 switch (op) {
2405                 case SOPT_SET:
2406                         if (optlen != sizeof(int)) {
2407                                 error = EINVAL;
2408                                 break;
2409                         }
2410                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2411                                             sizeof(optval));
2412                         if (error)
2413                                 break;
2414                         if (optval < -1 || (optval % 2) != 0) {
2415                                 /*
2416                                  * The API assumes non-negative even offset
2417                                  * values or -1 as a special value.
2418                                  */
2419                                 error = EINVAL;
2420                         } else if (so->so_proto->pr_protocol ==
2421                             IPPROTO_ICMPV6) {
2422                                 if (optval != icmp6off)
2423                                         error = EINVAL;
2424                         } else
2425                                 inp->in6p_cksum = optval;
2426                         break;
2427
2428                 case SOPT_GET:
2429                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2430                                 optval = icmp6off;
2431                         else
2432                                 optval = inp->in6p_cksum;
2433
2434                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2435                         break;
2436
2437                 default:
2438                         error = EINVAL;
2439                         break;
2440                 }
2441                 break;
2442
2443         default:
2444                 error = ENOPROTOOPT;
2445                 break;
2446         }
2447
2448         return (error);
2449 }
2450
2451 /*
2452  * Set up IP6 options in pcb for insertion in output packets or
2453  * specifying behavior of outgoing packets.
2454  */
2455 static int
2456 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2457     struct socket *so, struct sockopt *sopt)
2458 {
2459         struct ip6_pktopts *opt = *pktopt;
2460         int error = 0;
2461         struct thread *td = sopt->sopt_td;
2462
2463         /* turn off any old options. */
2464         if (opt) {
2465 #ifdef DIAGNOSTIC
2466                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2467                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2468                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2469                         printf("ip6_pcbopts: all specified options are cleared.\n");
2470 #endif
2471                 ip6_clearpktopts(opt, -1);
2472         } else {
2473                 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2474                 if (opt == NULL)
2475                         return (ENOMEM);
2476         }
2477         *pktopt = NULL;
2478
2479         if (!m || m->m_len == 0) {
2480                 /*
2481                  * Only turning off any previous options, regardless of
2482                  * whether the opt is just created or given.
2483                  */
2484                 free(opt, M_IP6OPT);
2485                 return (0);
2486         }
2487
2488         /*  set options specified by user. */
2489         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2490             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2491                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2492                 free(opt, M_IP6OPT);
2493                 return (error);
2494         }
2495         *pktopt = opt;
2496         return (0);
2497 }
2498
2499 /*
2500  * initialize ip6_pktopts.  beware that there are non-zero default values in
2501  * the struct.
2502  */
2503 void
2504 ip6_initpktopts(struct ip6_pktopts *opt)
2505 {
2506
2507         bzero(opt, sizeof(*opt));
2508         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2509         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2510         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2511         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2512 }
2513
2514 static int
2515 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2516     struct ucred *cred, int uproto)
2517 {
2518         struct ip6_pktopts *opt;
2519
2520         if (*pktopt == NULL) {
2521                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2522                     M_NOWAIT);
2523                 if (*pktopt == NULL)
2524                         return (ENOBUFS);
2525                 ip6_initpktopts(*pktopt);
2526         }
2527         opt = *pktopt;
2528
2529         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2530 }
2531
2532 #define GET_PKTOPT_VAR(field, lenexpr) do {                                     \
2533         if (pktopt && pktopt->field) {                                          \
2534                 INP_RUNLOCK(inp);                                               \
2535                 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);         \
2536                 malloc_optdata = true;                                          \
2537                 INP_RLOCK(inp);                                                 \
2538                 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {            \
2539                         INP_RUNLOCK(inp);                                       \
2540                         free(optdata, M_TEMP);                                  \
2541                         return (ECONNRESET);                                    \
2542                 }                                                               \
2543                 pktopt = inp->in6p_outputopts;                                  \
2544                 if (pktopt && pktopt->field) {                                  \
2545                         optdatalen = min(lenexpr, sopt->sopt_valsize);          \
2546                         bcopy(&pktopt->field, optdata, optdatalen);             \
2547                 } else {                                                        \
2548                         free(optdata, M_TEMP);                                  \
2549                         optdata = NULL;                                         \
2550                         malloc_optdata = false;                                 \
2551                 }                                                               \
2552         }                                                                       \
2553 } while(0)
2554
2555 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,                         \
2556         (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2557
2558 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,                        \
2559         pktopt->field->sa_len)
2560
2561 static int
2562 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2563 {
2564         void *optdata = NULL;
2565         bool malloc_optdata = false;
2566         int optdatalen = 0;
2567         int error = 0;
2568         struct in6_pktinfo null_pktinfo;
2569         int deftclass = 0, on;
2570         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2571         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2572         struct ip6_pktopts *pktopt;
2573
2574         INP_RLOCK(inp);
2575         pktopt = inp->in6p_outputopts;
2576
2577         switch (optname) {
2578         case IPV6_PKTINFO:
2579                 optdata = (void *)&null_pktinfo;
2580                 if (pktopt && pktopt->ip6po_pktinfo) {
2581                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2582                             sizeof(null_pktinfo));
2583                         in6_clearscope(&null_pktinfo.ipi6_addr);
2584                 } else {
2585                         /* XXX: we don't have to do this every time... */
2586                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2587                 }
2588                 optdatalen = sizeof(struct in6_pktinfo);
2589                 break;
2590         case IPV6_TCLASS:
2591                 if (pktopt && pktopt->ip6po_tclass >= 0)
2592                         deftclass = pktopt->ip6po_tclass;
2593                 optdata = (void *)&deftclass;
2594                 optdatalen = sizeof(int);
2595                 break;
2596         case IPV6_HOPOPTS:
2597                 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2598                 break;
2599         case IPV6_RTHDR:
2600                 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2601                 break;
2602         case IPV6_RTHDRDSTOPTS:
2603                 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2604                 break;
2605         case IPV6_DSTOPTS:
2606                 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2607                 break;
2608         case IPV6_NEXTHOP:
2609                 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2610                 break;
2611         case IPV6_USE_MIN_MTU:
2612                 if (pktopt)
2613                         defminmtu = pktopt->ip6po_minmtu;
2614                 optdata = (void *)&defminmtu;
2615                 optdatalen = sizeof(int);
2616                 break;
2617         case IPV6_DONTFRAG:
2618                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2619                         on = 1;
2620                 else
2621                         on = 0;
2622                 optdata = (void *)&on;
2623                 optdatalen = sizeof(on);
2624                 break;
2625         case IPV6_PREFER_TEMPADDR:
2626                 if (pktopt)
2627                         defpreftemp = pktopt->ip6po_prefer_tempaddr;
2628                 optdata = (void *)&defpreftemp;
2629                 optdatalen = sizeof(int);
2630                 break;
2631         default:                /* should not happen */
2632 #ifdef DIAGNOSTIC
2633                 panic("ip6_getpcbopt: unexpected option\n");
2634 #endif
2635                 INP_RUNLOCK(inp);
2636                 return (ENOPROTOOPT);
2637         }
2638         INP_RUNLOCK(inp);
2639
2640         error = sooptcopyout(sopt, optdata, optdatalen);
2641         if (malloc_optdata)
2642                 free(optdata, M_TEMP);
2643
2644         return (error);
2645 }
2646
2647 void
2648 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2649 {
2650         if (pktopt == NULL)
2651                 return;
2652
2653         if (optname == -1 || optname == IPV6_PKTINFO) {
2654                 if (pktopt->ip6po_pktinfo)
2655                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2656                 pktopt->ip6po_pktinfo = NULL;
2657         }
2658         if (optname == -1 || optname == IPV6_HOPLIMIT)
2659                 pktopt->ip6po_hlim = -1;
2660         if (optname == -1 || optname == IPV6_TCLASS)
2661                 pktopt->ip6po_tclass = -1;
2662         if (optname == -1 || optname == IPV6_NEXTHOP) {
2663                 if (pktopt->ip6po_nextroute.ro_nh) {
2664                         NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2665                         pktopt->ip6po_nextroute.ro_nh = NULL;
2666                 }
2667                 if (pktopt->ip6po_nexthop)
2668                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2669                 pktopt->ip6po_nexthop = NULL;
2670         }
2671         if (optname == -1 || optname == IPV6_HOPOPTS) {
2672                 if (pktopt->ip6po_hbh)
2673                         free(pktopt->ip6po_hbh, M_IP6OPT);
2674                 pktopt->ip6po_hbh = NULL;
2675         }
2676         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2677                 if (pktopt->ip6po_dest1)
2678                         free(pktopt->ip6po_dest1, M_IP6OPT);
2679                 pktopt->ip6po_dest1 = NULL;
2680         }
2681         if (optname == -1 || optname == IPV6_RTHDR) {
2682                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2683                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2684                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2685                 if (pktopt->ip6po_route.ro_nh) {
2686                         NH_FREE(pktopt->ip6po_route.ro_nh);
2687                         pktopt->ip6po_route.ro_nh = NULL;
2688                 }
2689         }
2690         if (optname == -1 || optname == IPV6_DSTOPTS) {
2691                 if (pktopt->ip6po_dest2)
2692                         free(pktopt->ip6po_dest2, M_IP6OPT);
2693                 pktopt->ip6po_dest2 = NULL;
2694         }
2695 }
2696
2697 #define PKTOPT_EXTHDRCPY(type) \
2698 do {\
2699         if (src->type) {\
2700                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2701                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2702                 if (dst->type == NULL)\
2703                         goto bad;\
2704                 bcopy(src->type, dst->type, hlen);\
2705         }\
2706 } while (/*CONSTCOND*/ 0)
2707
2708 static int
2709 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2710 {
2711         if (dst == NULL || src == NULL)  {
2712                 printf("ip6_clearpktopts: invalid argument\n");
2713                 return (EINVAL);
2714         }
2715
2716         dst->ip6po_hlim = src->ip6po_hlim;
2717         dst->ip6po_tclass = src->ip6po_tclass;
2718         dst->ip6po_flags = src->ip6po_flags;
2719         dst->ip6po_minmtu = src->ip6po_minmtu;
2720         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2721         if (src->ip6po_pktinfo) {
2722                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2723                     M_IP6OPT, canwait);
2724                 if (dst->ip6po_pktinfo == NULL)
2725                         goto bad;
2726                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2727         }
2728         if (src->ip6po_nexthop) {
2729                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2730                     M_IP6OPT, canwait);
2731                 if (dst->ip6po_nexthop == NULL)
2732                         goto bad;
2733                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2734                     src->ip6po_nexthop->sa_len);
2735         }
2736         PKTOPT_EXTHDRCPY(ip6po_hbh);
2737         PKTOPT_EXTHDRCPY(ip6po_dest1);
2738         PKTOPT_EXTHDRCPY(ip6po_dest2);
2739         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2740         return (0);
2741
2742   bad:
2743         ip6_clearpktopts(dst, -1);
2744         return (ENOBUFS);
2745 }
2746 #undef PKTOPT_EXTHDRCPY
2747
2748 struct ip6_pktopts *
2749 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2750 {
2751         int error;
2752         struct ip6_pktopts *dst;
2753
2754         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2755         if (dst == NULL)
2756                 return (NULL);
2757         ip6_initpktopts(dst);
2758
2759         if ((error = copypktopts(dst, src, canwait)) != 0) {
2760                 free(dst, M_IP6OPT);
2761                 return (NULL);
2762         }
2763
2764         return (dst);
2765 }
2766
2767 void
2768 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2769 {
2770         if (pktopt == NULL)
2771                 return;
2772
2773         ip6_clearpktopts(pktopt, -1);
2774
2775         free(pktopt, M_IP6OPT);
2776 }
2777
2778 /*
2779  * Set IPv6 outgoing packet options based on advanced API.
2780  */
2781 int
2782 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2783     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2784 {
2785         struct cmsghdr *cm = NULL;
2786
2787         if (control == NULL || opt == NULL)
2788                 return (EINVAL);
2789
2790         ip6_initpktopts(opt);
2791         if (stickyopt) {
2792                 int error;
2793
2794                 /*
2795                  * If stickyopt is provided, make a local copy of the options
2796                  * for this particular packet, then override them by ancillary
2797                  * objects.
2798                  * XXX: copypktopts() does not copy the cached route to a next
2799                  * hop (if any).  This is not very good in terms of efficiency,
2800                  * but we can allow this since this option should be rarely
2801                  * used.
2802                  */
2803                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2804                         return (error);
2805         }
2806
2807         /*
2808          * XXX: Currently, we assume all the optional information is stored
2809          * in a single mbuf.
2810          */
2811         if (control->m_next)
2812                 return (EINVAL);
2813
2814         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2815             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2816                 int error;
2817
2818                 if (control->m_len < CMSG_LEN(0))
2819                         return (EINVAL);
2820
2821                 cm = mtod(control, struct cmsghdr *);
2822                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2823                         return (EINVAL);
2824                 if (cm->cmsg_level != IPPROTO_IPV6)
2825                         continue;
2826
2827                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2828                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2829                 if (error)
2830                         return (error);
2831         }
2832
2833         return (0);
2834 }
2835
2836 /*
2837  * Set a particular packet option, as a sticky option or an ancillary data
2838  * item.  "len" can be 0 only when it's a sticky option.
2839  * We have 4 cases of combination of "sticky" and "cmsg":
2840  * "sticky=0, cmsg=0": impossible
2841  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2842  * "sticky=1, cmsg=0": RFC3542 socket option
2843  * "sticky=1, cmsg=1": RFC2292 socket option
2844  */
2845 static int
2846 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2847     struct ucred *cred, int sticky, int cmsg, int uproto)
2848 {
2849         int minmtupolicy, preftemp;
2850         int error;
2851
2852         if (!sticky && !cmsg) {
2853 #ifdef DIAGNOSTIC
2854                 printf("ip6_setpktopt: impossible case\n");
2855 #endif
2856                 return (EINVAL);
2857         }
2858
2859         /*
2860          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2861          * not be specified in the context of RFC3542.  Conversely,
2862          * RFC3542 types should not be specified in the context of RFC2292.
2863          */
2864         if (!cmsg) {
2865                 switch (optname) {
2866                 case IPV6_2292PKTINFO:
2867                 case IPV6_2292HOPLIMIT:
2868                 case IPV6_2292NEXTHOP:
2869                 case IPV6_2292HOPOPTS:
2870                 case IPV6_2292DSTOPTS:
2871                 case IPV6_2292RTHDR:
2872                 case IPV6_2292PKTOPTIONS:
2873                         return (ENOPROTOOPT);
2874                 }
2875         }
2876         if (sticky && cmsg) {
2877                 switch (optname) {
2878                 case IPV6_PKTINFO:
2879                 case IPV6_HOPLIMIT:
2880                 case IPV6_NEXTHOP:
2881                 case IPV6_HOPOPTS:
2882                 case IPV6_DSTOPTS:
2883                 case IPV6_RTHDRDSTOPTS:
2884                 case IPV6_RTHDR:
2885                 case IPV6_USE_MIN_MTU:
2886                 case IPV6_DONTFRAG:
2887                 case IPV6_TCLASS:
2888                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2889                         return (ENOPROTOOPT);
2890                 }
2891         }
2892
2893         switch (optname) {
2894         case IPV6_2292PKTINFO:
2895         case IPV6_PKTINFO:
2896         {
2897                 struct ifnet *ifp = NULL;
2898                 struct in6_pktinfo *pktinfo;
2899
2900                 if (len != sizeof(struct in6_pktinfo))
2901                         return (EINVAL);
2902
2903                 pktinfo = (struct in6_pktinfo *)buf;
2904
2905                 /*
2906                  * An application can clear any sticky IPV6_PKTINFO option by
2907                  * doing a "regular" setsockopt with ipi6_addr being
2908                  * in6addr_any and ipi6_ifindex being zero.
2909                  * [RFC 3542, Section 6]
2910                  */
2911                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2912                     pktinfo->ipi6_ifindex == 0 &&
2913                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2914                         ip6_clearpktopts(opt, optname);
2915                         break;
2916                 }
2917
2918                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2919                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2920                         return (EINVAL);
2921                 }
2922                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2923                         return (EINVAL);
2924                 /* validate the interface index if specified. */
2925                 if (pktinfo->ipi6_ifindex > V_if_index)
2926                          return (ENXIO);
2927                 if (pktinfo->ipi6_ifindex) {
2928                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2929                         if (ifp == NULL)
2930                                 return (ENXIO);
2931                 }
2932                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2933                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2934                         return (ENETDOWN);
2935
2936                 if (ifp != NULL &&
2937                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2938                         struct in6_ifaddr *ia;
2939
2940                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2941                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2942                         if (ia == NULL)
2943                                 return (EADDRNOTAVAIL);
2944                         ifa_free(&ia->ia_ifa);
2945                 }
2946                 /*
2947                  * We store the address anyway, and let in6_selectsrc()
2948                  * validate the specified address.  This is because ipi6_addr
2949                  * may not have enough information about its scope zone, and
2950                  * we may need additional information (such as outgoing
2951                  * interface or the scope zone of a destination address) to
2952                  * disambiguate the scope.
2953                  * XXX: the delay of the validation may confuse the
2954                  * application when it is used as a sticky option.
2955                  */
2956                 if (opt->ip6po_pktinfo == NULL) {
2957                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2958                             M_IP6OPT, M_NOWAIT);
2959                         if (opt->ip6po_pktinfo == NULL)
2960                                 return (ENOBUFS);
2961                 }
2962                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2963                 break;
2964         }
2965
2966         case IPV6_2292HOPLIMIT:
2967         case IPV6_HOPLIMIT:
2968         {
2969                 int *hlimp;
2970
2971                 /*
2972                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2973                  * to simplify the ordering among hoplimit options.
2974                  */
2975                 if (optname == IPV6_HOPLIMIT && sticky)
2976                         return (ENOPROTOOPT);
2977
2978                 if (len != sizeof(int))
2979                         return (EINVAL);
2980                 hlimp = (int *)buf;
2981                 if (*hlimp < -1 || *hlimp > 255)
2982                         return (EINVAL);
2983
2984                 opt->ip6po_hlim = *hlimp;
2985                 break;
2986         }
2987
2988         case IPV6_TCLASS:
2989         {
2990                 int tclass;
2991
2992                 if (len != sizeof(int))
2993                         return (EINVAL);
2994                 tclass = *(int *)buf;
2995                 if (tclass < -1 || tclass > 255)
2996                         return (EINVAL);
2997
2998                 opt->ip6po_tclass = tclass;
2999                 break;
3000         }
3001
3002         case IPV6_2292NEXTHOP:
3003         case IPV6_NEXTHOP:
3004                 if (cred != NULL) {
3005                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3006                         if (error)
3007                                 return (error);
3008                 }
3009
3010                 if (len == 0) { /* just remove the option */
3011                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3012                         break;
3013                 }
3014
3015                 /* check if cmsg_len is large enough for sa_len */
3016                 if (len < sizeof(struct sockaddr) || len < *buf)
3017                         return (EINVAL);
3018
3019                 switch (((struct sockaddr *)buf)->sa_family) {
3020                 case AF_INET6:
3021                 {
3022                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3023                         int error;
3024
3025                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3026                                 return (EINVAL);
3027
3028                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3029                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3030                                 return (EINVAL);
3031                         }
3032                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3033                             != 0) {
3034                                 return (error);
3035                         }
3036                         break;
3037                 }
3038                 case AF_LINK:   /* should eventually be supported */
3039                 default:
3040                         return (EAFNOSUPPORT);
3041                 }
3042
3043                 /* turn off the previous option, then set the new option. */
3044                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3045                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3046                 if (opt->ip6po_nexthop == NULL)
3047                         return (ENOBUFS);
3048                 bcopy(buf, opt->ip6po_nexthop, *buf);
3049                 break;
3050
3051         case IPV6_2292HOPOPTS:
3052         case IPV6_HOPOPTS:
3053         {
3054                 struct ip6_hbh *hbh;
3055                 int hbhlen;
3056
3057                 /*
3058                  * XXX: We don't allow a non-privileged user to set ANY HbH
3059                  * options, since per-option restriction has too much
3060                  * overhead.
3061                  */
3062                 if (cred != NULL) {
3063                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3064                         if (error)
3065                                 return (error);
3066                 }
3067
3068                 if (len == 0) {
3069                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3070                         break;  /* just remove the option */
3071                 }
3072
3073                 /* message length validation */
3074                 if (len < sizeof(struct ip6_hbh))
3075                         return (EINVAL);
3076                 hbh = (struct ip6_hbh *)buf;
3077                 hbhlen = (hbh->ip6h_len + 1) << 3;
3078                 if (len != hbhlen)
3079                         return (EINVAL);
3080
3081                 /* turn off the previous option, then set the new option. */
3082                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3083                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3084                 if (opt->ip6po_hbh == NULL)
3085                         return (ENOBUFS);
3086                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3087
3088                 break;
3089         }
3090
3091         case IPV6_2292DSTOPTS:
3092         case IPV6_DSTOPTS:
3093         case IPV6_RTHDRDSTOPTS:
3094         {
3095                 struct ip6_dest *dest, **newdest = NULL;
3096                 int destlen;
3097
3098                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3099                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3100                         if (error)
3101                                 return (error);
3102                 }
3103
3104                 if (len == 0) {
3105                         ip6_clearpktopts(opt, optname);
3106                         break;  /* just remove the option */
3107                 }
3108
3109                 /* message length validation */
3110                 if (len < sizeof(struct ip6_dest))
3111                         return (EINVAL);
3112                 dest = (struct ip6_dest *)buf;
3113                 destlen = (dest->ip6d_len + 1) << 3;
3114                 if (len != destlen)
3115                         return (EINVAL);
3116
3117                 /*
3118                  * Determine the position that the destination options header
3119                  * should be inserted; before or after the routing header.
3120                  */
3121                 switch (optname) {
3122                 case IPV6_2292DSTOPTS:
3123                         /*
3124                          * The old advacned API is ambiguous on this point.
3125                          * Our approach is to determine the position based
3126                          * according to the existence of a routing header.
3127                          * Note, however, that this depends on the order of the
3128                          * extension headers in the ancillary data; the 1st
3129                          * part of the destination options header must appear
3130                          * before the routing header in the ancillary data,
3131                          * too.
3132                          * RFC3542 solved the ambiguity by introducing
3133                          * separate ancillary data or option types.
3134                          */
3135                         if (opt->ip6po_rthdr == NULL)
3136                                 newdest = &opt->ip6po_dest1;
3137                         else
3138                                 newdest = &opt->ip6po_dest2;
3139                         break;
3140                 case IPV6_RTHDRDSTOPTS:
3141                         newdest = &opt->ip6po_dest1;
3142                         break;
3143                 case IPV6_DSTOPTS:
3144                         newdest = &opt->ip6po_dest2;
3145                         break;
3146                 }
3147
3148                 /* turn off the previous option, then set the new option. */
3149                 ip6_clearpktopts(opt, optname);
3150                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3151                 if (*newdest == NULL)
3152                         return (ENOBUFS);
3153                 bcopy(dest, *newdest, destlen);
3154
3155                 break;
3156         }
3157
3158         case IPV6_2292RTHDR:
3159         case IPV6_RTHDR:
3160         {
3161                 struct ip6_rthdr *rth;
3162                 int rthlen;
3163
3164                 if (len == 0) {
3165                         ip6_clearpktopts(opt, IPV6_RTHDR);
3166                         break;  /* just remove the option */
3167                 }
3168
3169                 /* message length validation */
3170                 if (len < sizeof(struct ip6_rthdr))
3171                         return (EINVAL);
3172                 rth = (struct ip6_rthdr *)buf;
3173                 rthlen = (rth->ip6r_len + 1) << 3;
3174                 if (len != rthlen)
3175                         return (EINVAL);
3176
3177                 switch (rth->ip6r_type) {
3178                 case IPV6_RTHDR_TYPE_0:
3179                         if (rth->ip6r_len == 0) /* must contain one addr */
3180                                 return (EINVAL);
3181                         if (rth->ip6r_len % 2) /* length must be even */
3182                                 return (EINVAL);
3183                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3184                                 return (EINVAL);
3185                         break;
3186                 default:
3187                         return (EINVAL);        /* not supported */
3188                 }
3189
3190                 /* turn off the previous option */
3191                 ip6_clearpktopts(opt, IPV6_RTHDR);
3192                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3193                 if (opt->ip6po_rthdr == NULL)
3194                         return (ENOBUFS);
3195                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3196
3197                 break;
3198         }
3199
3200         case IPV6_USE_MIN_MTU:
3201                 if (len != sizeof(int))
3202                         return (EINVAL);
3203                 minmtupolicy = *(int *)buf;
3204                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3205                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3206                     minmtupolicy != IP6PO_MINMTU_ALL) {
3207                         return (EINVAL);
3208                 }
3209                 opt->ip6po_minmtu = minmtupolicy;
3210                 break;
3211
3212         case IPV6_DONTFRAG:
3213                 if (len != sizeof(int))
3214                         return (EINVAL);
3215
3216                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3217                         /*
3218                          * we ignore this option for TCP sockets.
3219                          * (RFC3542 leaves this case unspecified.)
3220                          */
3221                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3222                 } else
3223                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3224                 break;
3225
3226         case IPV6_PREFER_TEMPADDR:
3227                 if (len != sizeof(int))
3228                         return (EINVAL);
3229                 preftemp = *(int *)buf;
3230                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3231                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3232                     preftemp != IP6PO_TEMPADDR_PREFER) {
3233                         return (EINVAL);
3234                 }
3235                 opt->ip6po_prefer_tempaddr = preftemp;
3236                 break;
3237
3238         default:
3239                 return (ENOPROTOOPT);
3240         } /* end of switch */
3241
3242         return (0);
3243 }
3244
3245 /*
3246  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3247  * packet to the input queue of a specified interface.  Note that this
3248  * calls the output routine of the loopback "driver", but with an interface
3249  * pointer that might NOT be &loif -- easier than replicating that code here.
3250  */
3251 void
3252 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3253 {
3254         struct mbuf *copym;
3255         struct ip6_hdr *ip6;
3256
3257         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3258         if (copym == NULL)
3259                 return;
3260
3261         /*
3262          * Make sure to deep-copy IPv6 header portion in case the data
3263          * is in an mbuf cluster, so that we can safely override the IPv6
3264          * header portion later.
3265          */
3266         if (!M_WRITABLE(copym) ||
3267             copym->m_len < sizeof(struct ip6_hdr)) {
3268                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3269                 if (copym == NULL)
3270                         return;
3271         }
3272         ip6 = mtod(copym, struct ip6_hdr *);
3273         /*
3274          * clear embedded scope identifiers if necessary.
3275          * in6_clearscope will touch the addresses only when necessary.
3276          */
3277         in6_clearscope(&ip6->ip6_src);
3278         in6_clearscope(&ip6->ip6_dst);
3279         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3280                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3281                     CSUM_PSEUDO_HDR;
3282                 copym->m_pkthdr.csum_data = 0xffff;
3283         }
3284         if_simloop(ifp, copym, AF_INET6, 0);
3285 }
3286
3287 /*
3288  * Chop IPv6 header off from the payload.
3289  */
3290 static int
3291 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3292 {
3293         struct mbuf *mh;
3294         struct ip6_hdr *ip6;
3295
3296         ip6 = mtod(m, struct ip6_hdr *);
3297         if (m->m_len > sizeof(*ip6)) {
3298                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3299                 if (mh == NULL) {
3300                         m_freem(m);
3301                         return ENOBUFS;
3302                 }
3303                 m_move_pkthdr(mh, m);
3304                 M_ALIGN(mh, sizeof(*ip6));
3305                 m->m_len -= sizeof(*ip6);
3306                 m->m_data += sizeof(*ip6);
3307                 mh->m_next = m;
3308                 m = mh;
3309                 m->m_len = sizeof(*ip6);
3310                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3311         }
3312         exthdrs->ip6e_ip6 = m;
3313         return 0;
3314 }
3315
3316 /*
3317  * Compute IPv6 extension header length.
3318  */
3319 int
3320 ip6_optlen(struct inpcb *inp)
3321 {
3322         int len;
3323
3324         if (!inp->in6p_outputopts)
3325                 return 0;
3326
3327         len = 0;
3328 #define elen(x) \
3329     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3330
3331         len += elen(inp->in6p_outputopts->ip6po_hbh);
3332         if (inp->in6p_outputopts->ip6po_rthdr)
3333                 /* dest1 is valid with rthdr only */
3334                 len += elen(inp->in6p_outputopts->ip6po_dest1);
3335         len += elen(inp->in6p_outputopts->ip6po_rthdr);
3336         len += elen(inp->in6p_outputopts->ip6po_dest2);
3337         return len;
3338 #undef elen
3339 }