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 #ifdef SCTP
 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 #ifdef SCTP
 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 #ifdef SCTP
 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 {
 327 #ifdef KERN_TLS
 328         struct ktls_session *tls = NULL;
 329 #endif
 330         struct m_snd_tag *mst;
 331         int error;
 332
 333         MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
 334         mst = NULL;
 335
 336 #ifdef KERN_TLS
 337         /*
 338          * If this is an unencrypted TLS record, save a reference to
 339          * the record.  This local reference is used to call
 340          * ktls_output_eagain after the mbuf has been freed (thus
 341          * dropping the mbuf's reference) in if_output.
 342          */
 343         if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
 344                 tls = ktls_hold(m->m_next->m_epg_tls);
 345                 mst = tls->snd_tag;
 346
 347                 /*
 348                  * If a TLS session doesn't have a valid tag, it must
 349                  * have had an earlier ifp mismatch, so drop this
 350                  * packet.
 351                  */
 352                 if (mst == NULL) {
 353                         error = EAGAIN;
 354                         goto done;
 355                 }
 356                 /*
 357                  * Always stamp tags that include NIC ktls.
 358                  */
 359                 stamp_tag = true;
 360         }
 361 #endif
 362 #ifdef RATELIMIT
 363         if (inp != NULL && mst == NULL) {
 364                 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
 365                     (inp->inp_snd_tag != NULL &&
 366                     inp->inp_snd_tag->ifp != ifp))
 367                         in_pcboutput_txrtlmt(inp, ifp, m);
 368
 369                 if (inp->inp_snd_tag != NULL)
 370                         mst = inp->inp_snd_tag;
 371         }
 372 #endif
 373         if (mst != NULL) {
 374                 KASSERT(m->m_pkthdr.rcvif == NULL,
 375                     ("trying to add a send tag to a forwarded packet"));
 376                 if (mst->ifp != ifp) {
 377                         error = EAGAIN;
 378                         goto done;
 379                 }
 380
 381                 /* stamp send tag on mbuf */
 382                 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
 383                 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
 384         }
 385
 386         error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
 387
 388 done:
 389         /* Check for route change invalidating send tags. */
 390 #ifdef KERN_TLS
 391         if (tls != NULL) {
 392                 if (error == EAGAIN)
 393                         error = ktls_output_eagain(inp, tls);
 394                 ktls_free(tls);
 395         }
 396 #endif
 397 #ifdef RATELIMIT
 398         if (error == EAGAIN)
 399                 in_pcboutput_eagain(inp);
 400 #endif
 401         return (error);
 402 }
 403
 404 /*
 405  * IP6 output.
 406  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
 407  * nxt, hlim, src, dst).
 408  * This function may modify ver and hlim only.
 409  * The mbuf chain containing the packet will be freed.
 410  * The mbuf opt, if present, will not be freed.
 411  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
 412  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
 413  * then result of route lookup is stored in ro->ro_nh.
 414  *
 415  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
 416  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
 417  *
 418  * ifpp - XXX: just for statistics
 419  */
 420 /*
 421  * XXX TODO: no flowid is assigned for outbound flows?
 422  */
 423 int
 424 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 425     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 426     struct ifnet **ifpp, struct inpcb *inp)
 427 {
 428         struct ip6_hdr *ip6;
 429         struct ifnet *ifp, *origifp;
 430         struct mbuf *m = m0;
 431         struct mbuf *mprev;
 432         struct route_in6 *ro_pmtu;
 433         struct nhop_object *nh;
 434         struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
 435         struct in6_addr odst;
 436         u_char *nexthdrp;
 437         int tlen, len;
 438         int error = 0;
 439         struct in6_ifaddr *ia = NULL;
 440         u_long mtu;
 441         int alwaysfrag, dontfrag;
 442         u_int32_t optlen, plen = 0, unfragpartlen;
 443         struct ip6_exthdrs exthdrs;
 444         struct in6_addr src0, dst0;
 445         u_int32_t zone;
 446         bool hdrsplit;
 447         int sw_csum, tso;
 448         int needfiblookup;
 449         uint32_t fibnum;
 450         struct m_tag *fwd_tag = NULL;
 451         uint32_t id;
 452
 453         NET_EPOCH_ASSERT();
 454
 455         if (inp != NULL) {
 456                 INP_LOCK_ASSERT(inp);
 457                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 458                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 459                         /* Unconditionally set flowid. */
 460                         m->m_pkthdr.flowid = inp->inp_flowid;
 461                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 462                 }
 463 #ifdef NUMA
 464                 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
 465 #endif
 466         }
 467
 468 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 469         /*
 470          * IPSec checking which handles several cases.
 471          * FAST IPSEC: We re-injected the packet.
 472          * XXX: need scope argument.
 473          */
 474         if (IPSEC_ENABLED(ipv6)) {
 475                 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
 476                         if (error == EINPROGRESS)
 477                                 error = 0;
 478                         goto done;
 479                 }
 480         }
 481 #endif /* IPSEC */
 482
 483         /* Source address validation. */
 484         ip6 = mtod(m, struct ip6_hdr *);
 485         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 486             (flags & IPV6_UNSPECSRC) == 0) {
 487                 error = EOPNOTSUPP;
 488                 IP6STAT_INC(ip6s_badscope);
 489                 goto bad;
 490         }
 491         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 492                 error = EOPNOTSUPP;
 493                 IP6STAT_INC(ip6s_badscope);
 494                 goto bad;
 495         }
 496
 497         /*
 498          * If we are given packet options to add extension headers prepare them.
 499          * Calculate the total length of the extension header chain.
 500          * Keep the length of the unfragmentable part for fragmentation.
 501          */
 502         bzero(&exthdrs, sizeof(exthdrs));
 503         optlen = 0;
 504         unfragpartlen = sizeof(struct ip6_hdr);
 505         if (opt) {
 506                 /* Hop-by-Hop options header. */
 507                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
 508
 509                 /* Destination options header (1st part). */
 510                 if (opt->ip6po_rthdr) {
 511 #ifndef RTHDR_SUPPORT_IMPLEMENTED
 512                         /*
 513                          * If there is a routing header, discard the packet
 514                          * right away here. RH0/1 are obsolete and we do not
 515                          * currently support RH2/3/4.
 516                          * People trying to use RH253/254 may want to disable
 517                          * this check.
 518                          * The moment we do support any routing header (again)
 519                          * this block should check the routing type more
 520                          * selectively.
 521                          */
 522                         error = EINVAL;
 523                         goto bad;
 524 #endif
 525
 526                         /*
 527                          * Destination options header (1st part).
 528                          * This only makes sense with a routing header.
 529                          * See Section 9.2 of RFC 3542.
 530                          * Disabling this part just for MIP6 convenience is
 531                          * a bad idea.  We need to think carefully about a
 532                          * way to make the advanced API coexist with MIP6
 533                          * options, which might automatically be inserted in
 534                          * the kernel.
 535                          */
 536                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
 537                             optlen);
 538                 }
 539                 /* Routing header. */
 540                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
 541
 542                 unfragpartlen += optlen;
 543
 544                 /*
 545                  * NOTE: we don't add AH/ESP length here (done in
 546                  * ip6_ipsec_output()).
 547                  */
 548
 549                 /* Destination options header (2nd part). */
 550                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
 551         }
 552
 553         /*
 554          * If there is at least one extension header,
 555          * separate IP6 header from the payload.
 556          */
 557         hdrsplit = false;
 558         if (optlen) {
 559                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 560                         m = NULL;
 561                         goto freehdrs;
 562                 }
 563                 m = exthdrs.ip6e_ip6;
 564                 ip6 = mtod(m, struct ip6_hdr *);
 565                 hdrsplit = true;
 566         }
 567
 568         /* Adjust mbuf packet header length. */
 569         m->m_pkthdr.len += optlen;
 570         plen = m->m_pkthdr.len - sizeof(*ip6);
 571
 572         /* If this is a jumbo payload, insert a jumbo payload option. */
 573         if (plen > IPV6_MAXPACKET) {
 574                 if (!hdrsplit) {
 575                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 576                                 m = NULL;
 577                                 goto freehdrs;
 578                         }
 579                         m = exthdrs.ip6e_ip6;
 580                         ip6 = mtod(m, struct ip6_hdr *);
 581                         hdrsplit = true;
 582                 }
 583                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 584                         goto freehdrs;
 585                 ip6->ip6_plen = 0;
 586         } else
 587                 ip6->ip6_plen = htons(plen);
 588         nexthdrp = &ip6->ip6_nxt;
 589
 590         if (optlen) {
 591                 /*
 592                  * Concatenate headers and fill in next header fields.
 593                  * Here we have, on "m"
 594                  *      IPv6 payload
 595                  * and we insert headers accordingly.
 596                  * Finally, we should be getting:
 597                  *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
 598                  *
 599                  * During the header composing process "m" points to IPv6
 600                  * header.  "mprev" points to an extension header prior to esp.
 601                  */
 602                 mprev = m;
 603
 604                 /*
 605                  * We treat dest2 specially.  This makes IPsec processing
 606                  * much easier.  The goal here is to make mprev point the
 607                  * mbuf prior to dest2.
 608                  *
 609                  * Result: IPv6 dest2 payload.
 610                  * m and mprev will point to IPv6 header.
 611                  */
 612                 if (exthdrs.ip6e_dest2) {
 613                         if (!hdrsplit)
 614                                 panic("%s:%d: assumption failed: "
 615                                     "hdr not split: hdrsplit %d exthdrs %p",
 616                                     __func__, __LINE__, hdrsplit, &exthdrs);
 617                         exthdrs.ip6e_dest2->m_next = m->m_next;
 618                         m->m_next = exthdrs.ip6e_dest2;
 619                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 620                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 621                 }
 622
 623                 /*
 624                  * Result: IPv6 hbh dest1 rthdr dest2 payload.
 625                  * m will point to IPv6 header.  mprev will point to the
 626                  * extension header prior to dest2 (rthdr in the above case).
 627                  */
 628                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 629                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 630                            IPPROTO_DSTOPTS);
 631                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 632                            IPPROTO_ROUTING);
 633         }
 634
 635         IP6STAT_INC(ip6s_localout);
 636
 637         /* Route packet. */
 638         ro_pmtu = ro;
 639         if (opt && opt->ip6po_rthdr)
 640                 ro = &opt->ip6po_route;
 641         if (ro != NULL)
 642                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 643         else
 644                 dst = &sin6;
 645         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 646
 647 again:
 648         /*
 649          * If specified, try to fill in the traffic class field.
 650          * Do not override if a non-zero value is already set.
 651          * We check the diffserv field and the ECN field separately.
 652          */
 653         if (opt && opt->ip6po_tclass >= 0) {
 654                 int mask = 0;
 655
 656                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 657                         mask |= 0xfc;
 658                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 659                         mask |= 0x03;
 660                 if (mask != 0)
 661                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 662         }
 663
 664         /* Fill in or override the hop limit field, if necessary. */
 665         if (opt && opt->ip6po_hlim != -1)
 666                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 667         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 668                 if (im6o != NULL)
 669                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 670                 else
 671                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 672         }
 673
 674         if (ro == NULL || ro->ro_nh == NULL) {
 675                 bzero(dst, sizeof(*dst));
 676                 dst->sin6_family = AF_INET6;
 677                 dst->sin6_len = sizeof(*dst);
 678                 dst->sin6_addr = ip6->ip6_dst;
 679         }
 680         /*
 681          * Validate route against routing table changes.
 682          * Make sure that the address family is set in route.
 683          */
 684         nh = NULL;
 685         ifp = NULL;
 686         mtu = 0;
 687         if (ro != NULL) {
 688                 if (ro->ro_nh != NULL && inp != NULL) {
 689                         ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
 690                         NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
 691                             fibnum);
 692                 }
 693                 if (ro->ro_nh != NULL && fwd_tag == NULL &&
 694                     (!NH_IS_VALID(ro->ro_nh) ||
 695                     ro->ro_dst.sin6_family != AF_INET6 ||
 696                     !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
 697                         RO_INVALIDATE_CACHE(ro);
 698
 699                 if (ro->ro_nh != NULL && fwd_tag == NULL &&
 700                     ro->ro_dst.sin6_family == AF_INET6 &&
 701                     IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
 702                         nh = ro->ro_nh;
 703                         ifp = nh->nh_ifp;
 704                 } else {
 705                         if (ro->ro_lle)
 706                                 LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 707                         ro->ro_lle = NULL;
 708                         if (fwd_tag == NULL) {
 709                                 bzero(&dst_sa, sizeof(dst_sa));
 710                                 dst_sa.sin6_family = AF_INET6;
 711                                 dst_sa.sin6_len = sizeof(dst_sa);
 712                                 dst_sa.sin6_addr = ip6->ip6_dst;
 713                         }
 714                         error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
 715                             &nh, fibnum, m->m_pkthdr.flowid);
 716                         if (error != 0) {
 717                                 IP6STAT_INC(ip6s_noroute);
 718                                 if (ifp != NULL)
 719                                         in6_ifstat_inc(ifp, ifs6_out_discard);
 720                                 goto bad;
 721                         }
 722                         if (ifp != NULL)
 723                             mtu = ifp->if_mtu;
 724                 }
 725                 if (nh == NULL) {
 726                         /*
 727                          * If in6_selectroute() does not return a nexthop
 728                          * dst may not have been updated.
 729                          */
 730                         *dst = dst_sa;  /* XXX */
 731                 } else {
 732                         if (nh->nh_flags & NHF_HOST)
 733                             mtu = nh->nh_mtu;
 734                         ia = (struct in6_ifaddr *)(nh->nh_ifa);
 735                         counter_u64_add(nh->nh_pksent, 1);
 736                 }
 737         } else {
 738                 struct nhop6_extended nh6;
 739                 struct in6_addr kdst;
 740                 uint32_t scopeid;
 741
 742                 if (fwd_tag == NULL) {
 743                         bzero(&dst_sa, sizeof(dst_sa));
 744                         dst_sa.sin6_family = AF_INET6;
 745                         dst_sa.sin6_len = sizeof(dst_sa);
 746                         dst_sa.sin6_addr = ip6->ip6_dst;
 747                 }
 748
 749                 if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
 750                     im6o != NULL &&
 751                     (ifp = im6o->im6o_multicast_ifp) != NULL) {
 752                         /* We do not need a route lookup. */
 753                         *dst = dst_sa;  /* XXX */
 754                         goto nonh6lookup;
 755                 }
 756
 757                 in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
 758
 759                 if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
 760                     IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
 761                         if (scopeid > 0) {
 762                                 ifp = in6_getlinkifnet(scopeid);
 763                                 *dst = dst_sa;  /* XXX */
 764                                 goto nonh6lookup;
 765                         }
 766                 }
 767
 768                 error = fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0,
 769                     &nh6);
 770                 if (error != 0) {
 771                         IP6STAT_INC(ip6s_noroute);
 772                         /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
 773                         error = EHOSTUNREACH;;
 774                         goto bad;
 775                 }
 776
 777                 ifp = nh6.nh_ifp;
 778                 mtu = nh6.nh_mtu;
 779                 dst->sin6_addr = nh6.nh_addr;
 780                 ia = nh6.nh_ia;
 781                 fib6_free_nh_ext(fibnum, &nh6);
 782 nonh6lookup:
 783                 ;
 784         }
 785
 786         /* Then nh (for unicast) and ifp must be non-NULL valid values. */
 787         if ((flags & IPV6_FORWARDING) == 0) {
 788                 /* XXX: the FORWARDING flag can be set for mrouting. */
 789                 in6_ifstat_inc(ifp, ifs6_out_request);
 790         }
 791
 792         /* Setup data structures for scope ID checks. */
 793         src0 = ip6->ip6_src;
 794         bzero(&src_sa, sizeof(src_sa));
 795         src_sa.sin6_family = AF_INET6;
 796         src_sa.sin6_len = sizeof(src_sa);
 797         src_sa.sin6_addr = ip6->ip6_src;
 798
 799         dst0 = ip6->ip6_dst;
 800         /* Re-initialize to be sure. */
 801         bzero(&dst_sa, sizeof(dst_sa));
 802         dst_sa.sin6_family = AF_INET6;
 803         dst_sa.sin6_len = sizeof(dst_sa);
 804         dst_sa.sin6_addr = ip6->ip6_dst;
 805
 806         /* Check for valid scope ID. */
 807         if (in6_setscope(&src0, ifp, &zone) == 0 &&
 808             sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
 809             in6_setscope(&dst0, ifp, &zone) == 0 &&
 810             sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
 811                 /*
 812                  * The outgoing interface is in the zone of the source
 813                  * and destination addresses.
 814                  *
 815                  * Because the loopback interface cannot receive
 816                  * packets with a different scope ID than its own,
 817                  * there is a trick to pretend the outgoing packet
 818                  * was received by the real network interface, by
 819                  * setting "origifp" different from "ifp". This is
 820                  * only allowed when "ifp" is a loopback network
 821                  * interface. Refer to code in nd6_output_ifp() for
 822                  * more details.
 823                  */
 824                 origifp = ifp;
 825
 826                 /*
 827                  * We should use ia_ifp to support the case of sending
 828                  * packets to an address of our own.
 829                  */
 830                 if (ia != NULL && ia->ia_ifp)
 831                         ifp = ia->ia_ifp;
 832
 833         } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
 834             sa6_recoverscope(&src_sa) != 0 ||
 835             sa6_recoverscope(&dst_sa) != 0 ||
 836             dst_sa.sin6_scope_id == 0 ||
 837             (src_sa.sin6_scope_id != 0 &&
 838             src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
 839             (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
 840                 /*
 841                  * If the destination network interface is not a
 842                  * loopback interface, or the destination network
 843                  * address has no scope ID, or the source address has
 844                  * a scope ID set which is different from the
 845                  * destination address one, or there is no network
 846                  * interface representing this scope ID, the address
 847                  * pair is considered invalid.
 848                  */
 849                 IP6STAT_INC(ip6s_badscope);
 850                 in6_ifstat_inc(ifp, ifs6_out_discard);
 851                 if (error == 0)
 852                         error = EHOSTUNREACH; /* XXX */
 853                 goto bad;
 854         }
 855         /* All scope ID checks are successful. */
 856
 857         if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 858                 if (opt && opt->ip6po_nextroute.ro_nh) {
 859                         /*
 860                          * The nexthop is explicitly specified by the
 861                          * application.  We assume the next hop is an IPv6
 862                          * address.
 863                          */
 864                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 865                 }
 866                 else if ((nh->nh_flags & NHF_GATEWAY))
 867                         dst = &nh->gw6_sa;
 868         }
 869
 870         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 871                 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
 872         } else {
 873                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 874                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 875
 876                 /* Confirm that the outgoing interface supports multicast. */
 877                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 878                         IP6STAT_INC(ip6s_noroute);
 879                         in6_ifstat_inc(ifp, ifs6_out_discard);
 880                         error = ENETUNREACH;
 881                         goto bad;
 882                 }
 883                 if ((im6o == NULL && in6_mcast_loop) ||
 884                     (im6o && im6o->im6o_multicast_loop)) {
 885                         /*
 886                          * Loop back multicast datagram if not expressly
 887                          * forbidden to do so, even if we have not joined
 888                          * the address; protocols will filter it later,
 889                          * thus deferring a hash lookup and lock acquisition
 890                          * at the expense of an m_copym().
 891                          */
 892                         ip6_mloopback(ifp, m);
 893                 } else {
 894                         /*
 895                          * If we are acting as a multicast router, perform
 896                          * multicast forwarding as if the packet had just
 897                          * arrived on the interface to which we are about
 898                          * to send.  The multicast forwarding function
 899                          * recursively calls this function, using the
 900                          * IPV6_FORWARDING flag to prevent infinite recursion.
 901                          *
 902                          * Multicasts that are looped back by ip6_mloopback(),
 903                          * above, will be forwarded by the ip6_input() routine,
 904                          * if necessary.
 905                          */
 906                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 907                                 /*
 908                                  * XXX: ip6_mforward expects that rcvif is NULL
 909                                  * when it is called from the originating path.
 910                                  * However, it may not always be the case.
 911                                  */
 912                                 m->m_pkthdr.rcvif = NULL;
 913                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 914                                         m_freem(m);
 915                                         goto done;
 916                                 }
 917                         }
 918                 }
 919                 /*
 920                  * Multicasts with a hoplimit of zero may be looped back,
 921                  * above, but must not be transmitted on a network.
 922                  * Also, multicasts addressed to the loopback interface
 923                  * are not sent -- the above call to ip6_mloopback() will
 924                  * loop back a copy if this host actually belongs to the
 925                  * destination group on the loopback interface.
 926                  */
 927                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 928                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 929                         m_freem(m);
 930                         goto done;
 931                 }
 932         }
 933
 934         /*
 935          * Fill the outgoing inteface to tell the upper layer
 936          * to increment per-interface statistics.
 937          */
 938         if (ifpp)
 939                 *ifpp = ifp;
 940
 941         /* Determine path MTU. */
 942         if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
 943                     &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
 944                 goto bad;
 945         KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
 946             "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
 947             ifp, alwaysfrag, fibnum));
 948
 949         /*
 950          * The caller of this function may specify to use the minimum MTU
 951          * in some cases.
 952          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 953          * setting.  The logic is a bit complicated; by default, unicast
 954          * packets will follow path MTU while multicast packets will be sent at
 955          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 956          * including unicast ones will be sent at the minimum MTU.  Multicast
 957          * packets will always be sent at the minimum MTU unless
 958          * IP6PO_MINMTU_DISABLE is explicitly specified.
 959          * See RFC 3542 for more details.
 960          */
 961         if (mtu > IPV6_MMTU) {
 962                 if ((flags & IPV6_MINMTU))
 963                         mtu = IPV6_MMTU;
 964                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 965                         mtu = IPV6_MMTU;
 966                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 967                          (opt == NULL ||
 968                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 969                         mtu = IPV6_MMTU;
 970                 }
 971         }
 972
 973         /*
 974          * Clear embedded scope identifiers if necessary.
 975          * in6_clearscope() will touch the addresses only when necessary.
 976          */
 977         in6_clearscope(&ip6->ip6_src);
 978         in6_clearscope(&ip6->ip6_dst);
 979
 980         /*
 981          * If the outgoing packet contains a hop-by-hop options header,
 982          * it must be examined and processed even by the source node.
 983          * (RFC 2460, section 4.)
 984          */
 985         if (exthdrs.ip6e_hbh) {
 986                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 987                 u_int32_t dummy; /* XXX unused */
 988                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 989
 990 #ifdef DIAGNOSTIC
 991                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 992                         panic("ip6e_hbh is not contiguous");
 993 #endif
 994                 /*
 995                  *  XXX: if we have to send an ICMPv6 error to the sender,
 996                  *       we need the M_LOOP flag since icmp6_error() expects
 997                  *       the IPv6 and the hop-by-hop options header are
 998                  *       contiguous unless the flag is set.
 999                  */
1000                 m->m_flags |= M_LOOP;
1001                 m->m_pkthdr.rcvif = ifp;
1002                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1003                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1004                     &dummy, &plen) < 0) {
1005                         /* m was already freed at this point. */
1006                         error = EINVAL;/* better error? */
1007                         goto done;
1008                 }
1009                 m->m_flags &= ~M_LOOP; /* XXX */
1010                 m->m_pkthdr.rcvif = NULL;
1011         }
1012
1013         /* Jump over all PFIL processing if hooks are not active. */
1014         if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1015                 goto passout;
1016
1017         odst = ip6->ip6_dst;
1018         /* Run through list of hooks for output packets. */
1019         switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
1020         case PFIL_PASS:
1021                 ip6 = mtod(m, struct ip6_hdr *);
1022                 break;
1023         case PFIL_DROPPED:
1024                 error = EACCES;
1025                 /* FALLTHROUGH */
1026         case PFIL_CONSUMED:
1027                 goto done;
1028         }
1029
1030         needfiblookup = 0;
1031         /* See if destination IP address was changed by packet filter. */
1032         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1033                 m->m_flags |= M_SKIP_FIREWALL;
1034                 /* If destination is now ourself drop to ip6_input(). */
1035                 if (in6_localip(&ip6->ip6_dst)) {
1036                         m->m_flags |= M_FASTFWD_OURS;
1037                         if (m->m_pkthdr.rcvif == NULL)
1038                                 m->m_pkthdr.rcvif = V_loif;
1039                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1040                                 m->m_pkthdr.csum_flags |=
1041                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1042                                 m->m_pkthdr.csum_data = 0xffff;
1043                         }
1044 #ifdef SCTP
1045                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1046                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1047 #endif
1048                         error = netisr_queue(NETISR_IPV6, m);
1049                         goto done;
1050                 } else {
1051                         if (ro != NULL)
1052                                 RO_INVALIDATE_CACHE(ro);
1053                         needfiblookup = 1; /* Redo the routing table lookup. */
1054                 }
1055         }
1056         /* See if fib was changed by packet filter. */
1057         if (fibnum != M_GETFIB(m)) {
1058                 m->m_flags |= M_SKIP_FIREWALL;
1059                 fibnum = M_GETFIB(m);
1060                 if (ro != NULL)
1061                         RO_INVALIDATE_CACHE(ro);
1062                 needfiblookup = 1;
1063         }
1064         if (needfiblookup)
1065                 goto again;
1066
1067         /* See if local, if yes, send it to netisr. */
1068         if (m->m_flags & M_FASTFWD_OURS) {
1069                 if (m->m_pkthdr.rcvif == NULL)
1070                         m->m_pkthdr.rcvif = V_loif;
1071                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1072                         m->m_pkthdr.csum_flags |=
1073                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1074                         m->m_pkthdr.csum_data = 0xffff;
1075                 }
1076 #ifdef SCTP
1077                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1078                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1079 #endif
1080                 error = netisr_queue(NETISR_IPV6, m);
1081                 goto done;
1082         }
1083         /* Or forward to some other address? */
1084         if ((m->m_flags & M_IP6_NEXTHOP) &&
1085             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1086                 if (ro != NULL)
1087                         dst = (struct sockaddr_in6 *)&ro->ro_dst;
1088                 else
1089                         dst = &sin6;
1090                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1091                 m->m_flags |= M_SKIP_FIREWALL;
1092                 m->m_flags &= ~M_IP6_NEXTHOP;
1093                 m_tag_delete(m, fwd_tag);
1094                 goto again;
1095         }
1096
1097 passout:
1098         /*
1099          * Send the packet to the outgoing interface.
1100          * If necessary, do IPv6 fragmentation before sending.
1101          *
1102          * The logic here is rather complex:
1103          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1104          * 1-a: send as is if tlen <= path mtu
1105          * 1-b: fragment if tlen > path mtu
1106          *
1107          * 2: if user asks us not to fragment (dontfrag == 1)
1108          * 2-a: send as is if tlen <= interface mtu
1109          * 2-b: error if tlen > interface mtu
1110          *
1111          * 3: if we always need to attach fragment header (alwaysfrag == 1)
1112          *      always fragment
1113          *
1114          * 4: if dontfrag == 1 && alwaysfrag == 1
1115          *      error, as we cannot handle this conflicting request.
1116          */
1117         sw_csum = m->m_pkthdr.csum_flags;
1118         if (!hdrsplit) {
1119                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
1120                 sw_csum &= ~ifp->if_hwassist;
1121         } else
1122                 tso = 0;
1123         /*
1124          * If we added extension headers, we will not do TSO and calculate the
1125          * checksums ourselves for now.
1126          * XXX-BZ  Need a framework to know when the NIC can handle it, even
1127          * with ext. hdrs.
1128          */
1129         error = ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen, false);
1130         if (error != 0)
1131                 goto bad;
1132         /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1133         tlen = m->m_pkthdr.len;
1134
1135         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1136                 dontfrag = 1;
1137         else
1138                 dontfrag = 0;
1139         if (dontfrag && alwaysfrag) {   /* Case 4. */
1140                 /* Conflicting request - can't transmit. */
1141                 error = EMSGSIZE;
1142                 goto bad;
1143         }
1144         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* Case 2-b. */
1145                 /*
1146                  * Even if the DONTFRAG option is specified, we cannot send the
1147                  * packet when the data length is larger than the MTU of the
1148                  * outgoing interface.
1149                  * Notify the error by sending IPV6_PATHMTU ancillary data if
1150                  * application wanted to know the MTU value. Also return an
1151                  * error code (this is not described in the API spec).
1152                  */
1153                 if (inp != NULL)
1154                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1155                 error = EMSGSIZE;
1156                 goto bad;
1157         }
1158
1159         /* Transmit packet without fragmentation. */
1160         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */
1161                 struct in6_ifaddr *ia6;
1162
1163                 ip6 = mtod(m, struct ip6_hdr *);
1164                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1165                 if (ia6) {
1166                         /* Record statistics for this interface address. */
1167                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1168                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
1169                             m->m_pkthdr.len);
1170                         ifa_free(&ia6->ia_ifa);
1171                 }
1172                 error = ip6_output_send(inp, ifp, origifp, m, dst, ro);
1173                 goto done;
1174         }
1175
1176         /* Try to fragment the packet.  Cases 1-b and 3. */
1177         if (mtu < IPV6_MMTU) {
1178                 /* Path MTU cannot be less than IPV6_MMTU. */
1179                 error = EMSGSIZE;
1180                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1181                 goto bad;
1182         } else if (ip6->ip6_plen == 0) {
1183                 /* Jumbo payload cannot be fragmented. */
1184                 error = EMSGSIZE;
1185                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1186                 goto bad;
1187         } else {
1188                 u_char nextproto;
1189
1190                 /*
1191                  * Too large for the destination or interface;
1192                  * fragment if possible.
1193                  * Must be able to put at least 8 bytes per fragment.
1194                  */
1195                 if (mtu > IPV6_MAXPACKET)
1196                         mtu = IPV6_MAXPACKET;
1197
1198                 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1199                 if (len < 8) {
1200                         error = EMSGSIZE;
1201                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1202                         goto bad;
1203                 }
1204
1205                 /*
1206                  * If the interface will not calculate checksums on
1207                  * fragmented packets, then do it here.
1208                  * XXX-BZ handle the hw offloading case.  Need flags.
1209                  */
1210                 error = ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags,
1211                     plen, optlen, true);
1212                 if (error != 0)
1213                         goto bad;
1214
1215                 /*
1216                  * Change the next header field of the last header in the
1217                  * unfragmentable part.
1218                  */
1219                 if (exthdrs.ip6e_rthdr) {
1220                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1221                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1222                 } else if (exthdrs.ip6e_dest1) {
1223                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1224                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1225                 } else if (exthdrs.ip6e_hbh) {
1226                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1227                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1228                 } else {
1229                         ip6 = mtod(m, struct ip6_hdr *);
1230                         nextproto = ip6->ip6_nxt;
1231                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1232                 }
1233
1234                 /*
1235                  * Loop through length of segment after first fragment,
1236                  * make new header and copy data of each part and link onto
1237                  * chain.
1238                  */
1239                 m0 = m;
1240                 id = htonl(ip6_randomid());
1241                 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1242                 if (error != 0)
1243                         goto sendorfree;
1244
1245                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1246         }
1247
1248         /* Remove leading garbage. */
1249 sendorfree:
1250         m = m0->m_nextpkt;
1251         m0->m_nextpkt = 0;
1252         m_freem(m0);
1253         for (; m; m = m0) {
1254                 m0 = m->m_nextpkt;
1255                 m->m_nextpkt = 0;
1256                 if (error == 0) {
1257                         /* Record statistics for this interface address. */
1258                         if (ia) {
1259                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1260                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1261                                     m->m_pkthdr.len);
1262                         }
1263                         error = ip6_output_send(inp, ifp, origifp, m, dst, ro);
1264                 } else
1265                         m_freem(m);
1266         }
1267
1268         if (error == 0)
1269                 IP6STAT_INC(ip6s_fragmented);
1270
1271 done:
1272         return (error);
1273
1274 freehdrs:
1275         m_freem(exthdrs.ip6e_hbh);      /* m_freem() checks if mbuf is NULL. */
1276         m_freem(exthdrs.ip6e_dest1);
1277         m_freem(exthdrs.ip6e_rthdr);
1278         m_freem(exthdrs.ip6e_dest2);
1279         /* FALLTHROUGH */
1280 bad:
1281         if (m)
1282                 m_freem(m);
1283         goto done;
1284 }
1285
1286 static int
1287 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1288 {
1289         struct mbuf *m;
1290
1291         if (hlen > MCLBYTES)
1292                 return (ENOBUFS); /* XXX */
1293
1294         if (hlen > MLEN)
1295                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1296         else
1297                 m = m_get(M_NOWAIT, MT_DATA);
1298         if (m == NULL)
1299                 return (ENOBUFS);
1300         m->m_len = hlen;
1301         if (hdr)
1302                 bcopy(hdr, mtod(m, caddr_t), hlen);
1303
1304         *mp = m;
1305         return (0);
1306 }
1307
1308 /*
1309  * Insert jumbo payload option.
1310  */
1311 static int
1312 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1313 {
1314         struct mbuf *mopt;
1315         u_char *optbuf;
1316         u_int32_t v;
1317
1318 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1319
1320         /*
1321          * If there is no hop-by-hop options header, allocate new one.
1322          * If there is one but it doesn't have enough space to store the
1323          * jumbo payload option, allocate a cluster to store the whole options.
1324          * Otherwise, use it to store the options.
1325          */
1326         if (exthdrs->ip6e_hbh == NULL) {
1327                 mopt = m_get(M_NOWAIT, MT_DATA);
1328                 if (mopt == NULL)
1329                         return (ENOBUFS);
1330                 mopt->m_len = JUMBOOPTLEN;
1331                 optbuf = mtod(mopt, u_char *);
1332                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1333                 exthdrs->ip6e_hbh = mopt;
1334         } else {
1335                 struct ip6_hbh *hbh;
1336
1337                 mopt = exthdrs->ip6e_hbh;
1338                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1339                         /*
1340                          * XXX assumption:
1341                          * - exthdrs->ip6e_hbh is not referenced from places
1342                          *   other than exthdrs.
1343                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1344                          */
1345                         int oldoptlen = mopt->m_len;
1346                         struct mbuf *n;
1347
1348                         /*
1349                          * XXX: give up if the whole (new) hbh header does
1350                          * not fit even in an mbuf cluster.
1351                          */
1352                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1353                                 return (ENOBUFS);
1354
1355                         /*
1356                          * As a consequence, we must always prepare a cluster
1357                          * at this point.
1358                          */
1359                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1360                         if (n == NULL)
1361                                 return (ENOBUFS);
1362                         n->m_len = oldoptlen + JUMBOOPTLEN;
1363                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1364                             oldoptlen);
1365                         optbuf = mtod(n, caddr_t) + oldoptlen;
1366                         m_freem(mopt);
1367                         mopt = exthdrs->ip6e_hbh = n;
1368                 } else {
1369                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1370                         mopt->m_len += JUMBOOPTLEN;
1371                 }
1372                 optbuf[0] = IP6OPT_PADN;
1373                 optbuf[1] = 1;
1374
1375                 /*
1376                  * Adjust the header length according to the pad and
1377                  * the jumbo payload option.
1378                  */
1379                 hbh = mtod(mopt, struct ip6_hbh *);
1380                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1381         }
1382
1383         /* fill in the option. */
1384         optbuf[2] = IP6OPT_JUMBO;
1385         optbuf[3] = 4;
1386         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1387         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1388
1389         /* finally, adjust the packet header length */
1390         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1391
1392         return (0);
1393 #undef JUMBOOPTLEN
1394 }
1395
1396 /*
1397  * Insert fragment header and copy unfragmentable header portions.
1398  */
1399 static int
1400 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1401     struct ip6_frag **frghdrp)
1402 {
1403         struct mbuf *n, *mlast;
1404
1405         if (hlen > sizeof(struct ip6_hdr)) {
1406                 n = m_copym(m0, sizeof(struct ip6_hdr),
1407                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1408                 if (n == NULL)
1409                         return (ENOBUFS);
1410                 m->m_next = n;
1411         } else
1412                 n = m;
1413
1414         /* Search for the last mbuf of unfragmentable part. */
1415         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1416                 ;
1417
1418         if (M_WRITABLE(mlast) &&
1419             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1420                 /* use the trailing space of the last mbuf for the fragment hdr */
1421                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1422                     mlast->m_len);
1423                 mlast->m_len += sizeof(struct ip6_frag);
1424                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1425         } else {
1426                 /* allocate a new mbuf for the fragment header */
1427                 struct mbuf *mfrg;
1428
1429                 mfrg = m_get(M_NOWAIT, MT_DATA);
1430                 if (mfrg == NULL)
1431                         return (ENOBUFS);
1432                 mfrg->m_len = sizeof(struct ip6_frag);
1433                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1434                 mlast->m_next = mfrg;
1435         }
1436
1437         return (0);
1438 }
1439
1440 /*
1441  * Calculates IPv6 path mtu for destination @dst.
1442  * Resulting MTU is stored in @mtup.
1443  *
1444  * Returns 0 on success.
1445  */
1446 static int
1447 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1448 {
1449         struct nhop6_extended nh6;
1450         struct in6_addr kdst;
1451         uint32_t scopeid;
1452         struct ifnet *ifp;
1453         u_long mtu;
1454         int error;
1455
1456         in6_splitscope(dst, &kdst, &scopeid);
1457         if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1458                 return (EHOSTUNREACH);
1459
1460         ifp = nh6.nh_ifp;
1461         mtu = nh6.nh_mtu;
1462
1463         error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1464         fib6_free_nh_ext(fibnum, &nh6);
1465
1466         return (error);
1467 }
1468
1469 /*
1470  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1471  * and cached data in @ro_pmtu.
1472  * MTU from (successful) route lookup is saved (along with dst)
1473  * inside @ro_pmtu to avoid subsequent route lookups after packet
1474  * filter processing.
1475  *
1476  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1477  * Returns 0 on success.
1478  */
1479 static int
1480 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1481     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1482     int *alwaysfragp, u_int fibnum, u_int proto)
1483 {
1484         struct nhop6_basic nh6;
1485         struct in6_addr kdst;
1486         uint32_t scopeid;
1487         struct sockaddr_in6 *sa6_dst, sin6;
1488         u_long mtu;
1489
1490         mtu = 0;
1491         if (ro_pmtu == NULL || do_lookup) {
1492
1493                 /*
1494                  * Here ro_pmtu has final destination address, while
1495                  * ro might represent immediate destination.
1496                  * Use ro_pmtu destination since mtu might differ.
1497                  */
1498                 if (ro_pmtu != NULL) {
1499                         sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1500                         if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1501                                 ro_pmtu->ro_mtu = 0;
1502                 } else
1503                         sa6_dst = &sin6;
1504
1505                 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1506                         bzero(sa6_dst, sizeof(*sa6_dst));
1507                         sa6_dst->sin6_family = AF_INET6;
1508                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1509                         sa6_dst->sin6_addr = *dst;
1510
1511                         in6_splitscope(dst, &kdst, &scopeid);
1512                         if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1513                             &nh6) == 0) {
1514                                 mtu = nh6.nh_mtu;
1515                                 if (ro_pmtu != NULL)
1516                                         ro_pmtu->ro_mtu = mtu;
1517                         }
1518                 } else
1519                         mtu = ro_pmtu->ro_mtu;
1520         }
1521
1522         if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1523                 mtu = ro_pmtu->ro_nh->nh_mtu;
1524
1525         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1526 }
1527
1528 /*
1529  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1530  * hostcache data for @dst.
1531  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1532  *
1533  * Returns 0 on success.
1534  */
1535 static int
1536 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1537     u_long *mtup, int *alwaysfragp, u_int proto)
1538 {
1539         u_long mtu = 0;
1540         int alwaysfrag = 0;
1541         int error = 0;
1542
1543         if (rt_mtu > 0) {
1544                 u_int32_t ifmtu;
1545                 struct in_conninfo inc;
1546
1547                 bzero(&inc, sizeof(inc));
1548                 inc.inc_flags |= INC_ISIPV6;
1549                 inc.inc6_faddr = *dst;
1550
1551                 ifmtu = IN6_LINKMTU(ifp);
1552
1553                 /* TCP is known to react to pmtu changes so skip hc */
1554                 if (proto != IPPROTO_TCP)
1555                         mtu = tcp_hc_getmtu(&inc);
1556
1557                 if (mtu)
1558                         mtu = min(mtu, rt_mtu);
1559                 else
1560                         mtu = rt_mtu;
1561                 if (mtu == 0)
1562                         mtu = ifmtu;
1563                 else if (mtu < IPV6_MMTU) {
1564                         /*
1565                          * RFC2460 section 5, last paragraph:
1566                          * if we record ICMPv6 too big message with
1567                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1568                          * or smaller, with framgent header attached.
1569                          * (fragment header is needed regardless from the
1570                          * packet size, for translators to identify packets)
1571                          */
1572                         alwaysfrag = 1;
1573                         mtu = IPV6_MMTU;
1574                 }
1575         } else if (ifp) {
1576                 mtu = IN6_LINKMTU(ifp);
1577         } else
1578                 error = EHOSTUNREACH; /* XXX */
1579
1580         *mtup = mtu;
1581         if (alwaysfragp)
1582                 *alwaysfragp = alwaysfrag;
1583         return (error);
1584 }
1585
1586 /*
1587  * IP6 socket option processing.
1588  */
1589 int
1590 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1591 {
1592         int optdatalen, uproto;
1593         void *optdata;
1594         struct inpcb *inp = sotoinpcb(so);
1595         int error, optval;
1596         int level, op, optname;
1597         int optlen;
1598         struct thread *td;
1599 #ifdef  RSS
1600         uint32_t rss_bucket;
1601         int retval;
1602 #endif
1603
1604 /*
1605  * Don't use more than a quarter of mbuf clusters.  N.B.:
1606  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1607  * on LP64 architectures, so cast to u_long to avoid undefined
1608  * behavior.  ILP32 architectures cannot have nmbclusters
1609  * large enough to overflow for other reasons.
1610  */
1611 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1612
1613         level = sopt->sopt_level;
1614         op = sopt->sopt_dir;
1615         optname = sopt->sopt_name;
1616         optlen = sopt->sopt_valsize;
1617         td = sopt->sopt_td;
1618         error = 0;
1619         optval = 0;
1620         uproto = (int)so->so_proto->pr_protocol;
1621
1622         if (level != IPPROTO_IPV6) {
1623                 error = EINVAL;
1624
1625                 if (sopt->sopt_level == SOL_SOCKET &&
1626                     sopt->sopt_dir == SOPT_SET) {
1627                         switch (sopt->sopt_name) {
1628                         case SO_REUSEADDR:
1629                                 INP_WLOCK(inp);
1630                                 if ((so->so_options & SO_REUSEADDR) != 0)
1631                                         inp->inp_flags2 |= INP_REUSEADDR;
1632                                 else
1633                                         inp->inp_flags2 &= ~INP_REUSEADDR;
1634                                 INP_WUNLOCK(inp);
1635                                 error = 0;
1636                                 break;
1637                         case SO_REUSEPORT:
1638                                 INP_WLOCK(inp);
1639                                 if ((so->so_options & SO_REUSEPORT) != 0)
1640                                         inp->inp_flags2 |= INP_REUSEPORT;
1641                                 else
1642                                         inp->inp_flags2 &= ~INP_REUSEPORT;
1643                                 INP_WUNLOCK(inp);
1644                                 error = 0;
1645                                 break;
1646                         case SO_REUSEPORT_LB:
1647                                 INP_WLOCK(inp);
1648                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1649                                         inp->inp_flags2 |= INP_REUSEPORT_LB;
1650                                 else
1651                                         inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1652                                 INP_WUNLOCK(inp);
1653                                 error = 0;
1654                                 break;
1655                         case SO_SETFIB:
1656                                 INP_WLOCK(inp);
1657                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
1658                                 INP_WUNLOCK(inp);
1659                                 error = 0;
1660                                 break;
1661                         case SO_MAX_PACING_RATE:
1662 #ifdef RATELIMIT
1663                                 INP_WLOCK(inp);
1664                                 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1665                                 INP_WUNLOCK(inp);
1666                                 error = 0;
1667 #else
1668                                 error = EOPNOTSUPP;
1669 #endif
1670                                 break;
1671                         default:
1672                                 break;
1673                         }
1674                 }
1675         } else {                /* level == IPPROTO_IPV6 */
1676                 switch (op) {
1677
1678                 case SOPT_SET:
1679                         switch (optname) {
1680                         case IPV6_2292PKTOPTIONS:
1681 #ifdef IPV6_PKTOPTIONS
1682                         case IPV6_PKTOPTIONS:
1683 #endif
1684                         {
1685                                 struct mbuf *m;
1686
1687                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1688                                         printf("ip6_ctloutput: mbuf limit hit\n");
1689                                         error = ENOBUFS;
1690                                         break;
1691                                 }
1692
1693                                 error = soopt_getm(sopt, &m); /* XXX */
1694                                 if (error != 0)
1695                                         break;
1696                                 error = soopt_mcopyin(sopt, m); /* XXX */
1697                                 if (error != 0)
1698                                         break;
1699                                 INP_WLOCK(inp);
1700                                 error = ip6_pcbopts(&inp->in6p_outputopts, m,
1701                                     so, sopt);
1702                                 INP_WUNLOCK(inp);
1703                                 m_freem(m); /* XXX */
1704                                 break;
1705                         }
1706
1707                         /*
1708                          * Use of some Hop-by-Hop options or some
1709                          * Destination options, might require special
1710                          * privilege.  That is, normal applications
1711                          * (without special privilege) might be forbidden
1712                          * from setting certain options in outgoing packets,
1713                          * and might never see certain options in received
1714                          * packets. [RFC 2292 Section 6]
1715                          * KAME specific note:
1716                          *  KAME prevents non-privileged users from sending or
1717                          *  receiving ANY hbh/dst options in order to avoid
1718                          *  overhead of parsing options in the kernel.
1719                          */
1720                         case IPV6_RECVHOPOPTS:
1721                         case IPV6_RECVDSTOPTS:
1722                         case IPV6_RECVRTHDRDSTOPTS:
1723                                 if (td != NULL) {
1724                                         error = priv_check(td,
1725                                             PRIV_NETINET_SETHDROPTS);
1726                                         if (error)
1727                                                 break;
1728                                 }
1729                                 /* FALLTHROUGH */
1730                         case IPV6_UNICAST_HOPS:
1731                         case IPV6_HOPLIMIT:
1732
1733                         case IPV6_RECVPKTINFO:
1734                         case IPV6_RECVHOPLIMIT:
1735                         case IPV6_RECVRTHDR:
1736                         case IPV6_RECVPATHMTU:
1737                         case IPV6_RECVTCLASS:
1738                         case IPV6_RECVFLOWID:
1739 #ifdef  RSS
1740                         case IPV6_RECVRSSBUCKETID:
1741 #endif
1742                         case IPV6_V6ONLY:
1743                         case IPV6_AUTOFLOWLABEL:
1744                         case IPV6_ORIGDSTADDR:
1745                         case IPV6_BINDANY:
1746                         case IPV6_BINDMULTI:
1747 #ifdef  RSS
1748                         case IPV6_RSS_LISTEN_BUCKET:
1749 #endif
1750                                 if (optname == IPV6_BINDANY && td != NULL) {
1751                                         error = priv_check(td,
1752                                             PRIV_NETINET_BINDANY);
1753                                         if (error)
1754                                                 break;
1755                                 }
1756
1757                                 if (optlen != sizeof(int)) {
1758                                         error = EINVAL;
1759                                         break;
1760                                 }
1761                                 error = sooptcopyin(sopt, &optval,
1762                                         sizeof optval, sizeof optval);
1763                                 if (error)
1764                                         break;
1765                                 switch (optname) {
1766
1767                                 case IPV6_UNICAST_HOPS:
1768                                         if (optval < -1 || optval >= 256)
1769                                                 error = EINVAL;
1770                                         else {
1771                                                 /* -1 = kernel default */
1772                                                 inp->in6p_hops = optval;
1773                                                 if ((inp->inp_vflag &
1774                                                      INP_IPV4) != 0)
1775                                                         inp->inp_ip_ttl = optval;
1776                                         }
1777                                         break;
1778 #define OPTSET(bit) \
1779 do { \
1780         INP_WLOCK(inp); \
1781         if (optval) \
1782                 inp->inp_flags |= (bit); \
1783         else \
1784                 inp->inp_flags &= ~(bit); \
1785         INP_WUNLOCK(inp); \
1786 } while (/*CONSTCOND*/ 0)
1787 #define OPTSET2292(bit) \
1788 do { \
1789         INP_WLOCK(inp); \
1790         inp->inp_flags |= IN6P_RFC2292; \
1791         if (optval) \
1792                 inp->inp_flags |= (bit); \
1793         else \
1794                 inp->inp_flags &= ~(bit); \
1795         INP_WUNLOCK(inp); \
1796 } while (/*CONSTCOND*/ 0)
1797 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1798
1799 #define OPTSET2_N(bit, val) do {                                        \
1800         if (val)                                                        \
1801                 inp->inp_flags2 |= bit;                                 \
1802         else                                                            \
1803                 inp->inp_flags2 &= ~bit;                                \
1804 } while (0)
1805 #define OPTSET2(bit, val) do {                                          \
1806         INP_WLOCK(inp);                                                 \
1807         OPTSET2_N(bit, val);                                            \
1808         INP_WUNLOCK(inp);                                               \
1809 } while (0)
1810 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1811 #define OPTSET2292_EXCLUSIVE(bit)                                       \
1812 do {                                                                    \
1813         INP_WLOCK(inp);                                                 \
1814         if (OPTBIT(IN6P_RFC2292)) {                                     \
1815                 error = EINVAL;                                         \
1816         } else {                                                        \
1817                 if (optval)                                             \
1818                         inp->inp_flags |= (bit);                        \
1819                 else                                                    \
1820                         inp->inp_flags &= ~(bit);                       \
1821         }                                                               \
1822         INP_WUNLOCK(inp);                                               \
1823 } while (/*CONSTCOND*/ 0)
1824
1825                                 case IPV6_RECVPKTINFO:
1826                                         OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1827                                         break;
1828
1829                                 case IPV6_HOPLIMIT:
1830                                 {
1831                                         struct ip6_pktopts **optp;
1832
1833                                         /* cannot mix with RFC2292 */
1834                                         if (OPTBIT(IN6P_RFC2292)) {
1835                                                 error = EINVAL;
1836                                                 break;
1837                                         }
1838                                         INP_WLOCK(inp);
1839                                         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1840                                                 INP_WUNLOCK(inp);
1841                                                 return (ECONNRESET);
1842                                         }
1843                                         optp = &inp->in6p_outputopts;
1844                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1845                                             (u_char *)&optval, sizeof(optval),
1846                                             optp, (td != NULL) ? td->td_ucred :
1847                                             NULL, uproto);
1848                                         INP_WUNLOCK(inp);
1849                                         break;
1850                                 }
1851
1852                                 case IPV6_RECVHOPLIMIT:
1853                                         OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1854                                         break;
1855
1856                                 case IPV6_RECVHOPOPTS:
1857                                         OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1858                                         break;
1859
1860                                 case IPV6_RECVDSTOPTS:
1861                                         OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1862                                         break;
1863
1864                                 case IPV6_RECVRTHDRDSTOPTS:
1865                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1866                                         break;
1867
1868                                 case IPV6_RECVRTHDR:
1869                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1870                                         break;
1871
1872                                 case IPV6_RECVPATHMTU:
1873                                         /*
1874                                          * We ignore this option for TCP
1875                                          * sockets.
1876                                          * (RFC3542 leaves this case
1877                                          * unspecified.)
1878                                          */
1879                                         if (uproto != IPPROTO_TCP)
1880                                                 OPTSET(IN6P_MTU);
1881                                         break;
1882
1883                                 case IPV6_RECVFLOWID:
1884                                         OPTSET2(INP_RECVFLOWID, optval);
1885                                         break;
1886
1887 #ifdef  RSS
1888                                 case IPV6_RECVRSSBUCKETID:
1889                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1890                                         break;
1891 #endif
1892
1893                                 case IPV6_V6ONLY:
1894                                         INP_WLOCK(inp);
1895                                         if (inp->inp_lport ||
1896                                             !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1897                                                 /*
1898                                                  * The socket is already bound.
1899                                                  */
1900                                                 INP_WUNLOCK(inp);
1901                                                 error = EINVAL;
1902                                                 break;
1903                                         }
1904                                         if (optval) {
1905                                                 inp->inp_flags |= IN6P_IPV6_V6ONLY;
1906                                                 inp->inp_vflag &= ~INP_IPV4;
1907                                         } else {
1908                                                 inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1909                                                 inp->inp_vflag |= INP_IPV4;
1910                                         }
1911                                         INP_WUNLOCK(inp);
1912                                         break;
1913                                 case IPV6_RECVTCLASS:
1914                                         /* cannot mix with RFC2292 XXX */
1915                                         OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1916                                         break;
1917                                 case IPV6_AUTOFLOWLABEL:
1918                                         OPTSET(IN6P_AUTOFLOWLABEL);
1919                                         break;
1920
1921                                 case IPV6_ORIGDSTADDR:
1922                                         OPTSET2(INP_ORIGDSTADDR, optval);
1923                                         break;
1924                                 case IPV6_BINDANY:
1925                                         OPTSET(INP_BINDANY);
1926                                         break;
1927
1928                                 case IPV6_BINDMULTI:
1929                                         OPTSET2(INP_BINDMULTI, optval);
1930                                         break;
1931 #ifdef  RSS
1932                                 case IPV6_RSS_LISTEN_BUCKET:
1933                                         if ((optval >= 0) &&
1934                                             (optval < rss_getnumbuckets())) {
1935                                                 INP_WLOCK(inp);
1936                                                 inp->inp_rss_listen_bucket = optval;
1937                                                 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1938                                                 INP_WUNLOCK(inp);
1939                                         } else {
1940                                                 error = EINVAL;
1941                                         }
1942                                         break;
1943 #endif
1944                                 }
1945                                 break;
1946
1947                         case IPV6_TCLASS:
1948                         case IPV6_DONTFRAG:
1949                         case IPV6_USE_MIN_MTU:
1950                         case IPV6_PREFER_TEMPADDR:
1951                                 if (optlen != sizeof(optval)) {
1952                                         error = EINVAL;
1953                                         break;
1954                                 }
1955                                 error = sooptcopyin(sopt, &optval,
1956                                         sizeof optval, sizeof optval);
1957                                 if (error)
1958                                         break;
1959                                 {
1960                                         struct ip6_pktopts **optp;
1961                                         INP_WLOCK(inp);
1962                                         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1963                                                 INP_WUNLOCK(inp);
1964                                                 return (ECONNRESET);
1965                                         }
1966                                         optp = &inp->in6p_outputopts;
1967                                         error = ip6_pcbopt(optname,
1968                                             (u_char *)&optval, sizeof(optval),
1969                                             optp, (td != NULL) ? td->td_ucred :
1970                                             NULL, uproto);
1971                                         INP_WUNLOCK(inp);
1972                                         break;
1973                                 }
1974
1975                         case IPV6_2292PKTINFO:
1976                         case IPV6_2292HOPLIMIT:
1977                         case IPV6_2292HOPOPTS:
1978                         case IPV6_2292DSTOPTS:
1979                         case IPV6_2292RTHDR:
1980                                 /* RFC 2292 */
1981                                 if (optlen != sizeof(int)) {
1982                                         error = EINVAL;
1983                                         break;
1984                                 }
1985                                 error = sooptcopyin(sopt, &optval,
1986                                         sizeof optval, sizeof optval);
1987                                 if (error)
1988                                         break;
1989                                 switch (optname) {
1990                                 case IPV6_2292PKTINFO:
1991                                         OPTSET2292(IN6P_PKTINFO);
1992                                         break;
1993                                 case IPV6_2292HOPLIMIT:
1994                                         OPTSET2292(IN6P_HOPLIMIT);
1995                                         break;
1996                                 case IPV6_2292HOPOPTS:
1997                                         /*
1998                                          * Check super-user privilege.
1999                                          * See comments for IPV6_RECVHOPOPTS.
2000                                          */
2001                                         if (td != NULL) {
2002                                                 error = priv_check(td,
2003                                                     PRIV_NETINET_SETHDROPTS);
2004                                                 if (error)
2005                                                         return (error);
2006                                         }
2007                                         OPTSET2292(IN6P_HOPOPTS);
2008                                         break;
2009                                 case IPV6_2292DSTOPTS:
2010                                         if (td != NULL) {
2011                                                 error = priv_check(td,
2012                                                     PRIV_NETINET_SETHDROPTS);
2013                                                 if (error)
2014                                                         return (error);
2015                                         }
2016                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2017                                         break;
2018                                 case IPV6_2292RTHDR:
2019                                         OPTSET2292(IN6P_RTHDR);
2020                                         break;
2021                                 }
2022                                 break;
2023                         case IPV6_PKTINFO:
2024                         case IPV6_HOPOPTS:
2025                         case IPV6_RTHDR:
2026                         case IPV6_DSTOPTS:
2027                         case IPV6_RTHDRDSTOPTS:
2028                         case IPV6_NEXTHOP:
2029                         {
2030                                 /* new advanced API (RFC3542) */
2031                                 u_char *optbuf;
2032                                 u_char optbuf_storage[MCLBYTES];
2033                                 int optlen;
2034                                 struct ip6_pktopts **optp;
2035
2036                                 /* cannot mix with RFC2292 */
2037                                 if (OPTBIT(IN6P_RFC2292)) {
2038                                         error = EINVAL;
2039                                         break;
2040                                 }
2041
2042                                 /*
2043                                  * We only ensure valsize is not too large
2044                                  * here.  Further validation will be done
2045                                  * later.
2046                                  */
2047                                 error = sooptcopyin(sopt, optbuf_storage,
2048                                     sizeof(optbuf_storage), 0);
2049                                 if (error)
2050                                         break;
2051                                 optlen = sopt->sopt_valsize;
2052                                 optbuf = optbuf_storage;
2053                                 INP_WLOCK(inp);
2054                                 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2055                                         INP_WUNLOCK(inp);
2056                                         return (ECONNRESET);
2057                                 }
2058                                 optp = &inp->in6p_outputopts;
2059                                 error = ip6_pcbopt(optname, optbuf, optlen,
2060                                     optp, (td != NULL) ? td->td_ucred : NULL,
2061                                     uproto);
2062                                 INP_WUNLOCK(inp);
2063                                 break;
2064                         }
2065 #undef OPTSET
2066
2067                         case IPV6_MULTICAST_IF:
2068                         case IPV6_MULTICAST_HOPS:
2069                         case IPV6_MULTICAST_LOOP:
2070                         case IPV6_JOIN_GROUP:
2071                         case IPV6_LEAVE_GROUP:
2072                         case IPV6_MSFILTER:
2073                         case MCAST_BLOCK_SOURCE:
2074                         case MCAST_UNBLOCK_SOURCE:
2075                         case MCAST_JOIN_GROUP:
2076                         case MCAST_LEAVE_GROUP:
2077                         case MCAST_JOIN_SOURCE_GROUP:
2078                         case MCAST_LEAVE_SOURCE_GROUP:
2079                                 error = ip6_setmoptions(inp, sopt);
2080                                 break;
2081
2082                         case IPV6_PORTRANGE:
2083                                 error = sooptcopyin(sopt, &optval,
2084                                     sizeof optval, sizeof optval);
2085                                 if (error)
2086                                         break;
2087
2088                                 INP_WLOCK(inp);
2089                                 switch (optval) {
2090                                 case IPV6_PORTRANGE_DEFAULT:
2091                                         inp->inp_flags &= ~(INP_LOWPORT);
2092                                         inp->inp_flags &= ~(INP_HIGHPORT);
2093                                         break;
2094
2095                                 case IPV6_PORTRANGE_HIGH:
2096                                         inp->inp_flags &= ~(INP_LOWPORT);
2097                                         inp->inp_flags |= INP_HIGHPORT;
2098                                         break;
2099
2100                                 case IPV6_PORTRANGE_LOW:
2101                                         inp->inp_flags &= ~(INP_HIGHPORT);
2102                                         inp->inp_flags |= INP_LOWPORT;
2103                                         break;
2104
2105                                 default:
2106                                         error = EINVAL;
2107                                         break;
2108                                 }
2109                                 INP_WUNLOCK(inp);
2110                                 break;
2111
2112 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2113                         case IPV6_IPSEC_POLICY:
2114                                 if (IPSEC_ENABLED(ipv6)) {
2115                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2116                                         break;
2117                                 }
2118                                 /* FALLTHROUGH */
2119 #endif /* IPSEC */
2120
2121                         default:
2122                                 error = ENOPROTOOPT;
2123                                 break;
2124                         }
2125                         break;
2126
2127                 case SOPT_GET:
2128                         switch (optname) {
2129
2130                         case IPV6_2292PKTOPTIONS:
2131 #ifdef IPV6_PKTOPTIONS
2132                         case IPV6_PKTOPTIONS:
2133 #endif
2134                                 /*
2135                                  * RFC3542 (effectively) deprecated the
2136                                  * semantics of the 2292-style pktoptions.
2137                                  * Since it was not reliable in nature (i.e.,
2138                                  * applications had to expect the lack of some
2139                                  * information after all), it would make sense
2140                                  * to simplify this part by always returning
2141                                  * empty data.
2142                                  */
2143                                 sopt->sopt_valsize = 0;
2144                                 break;
2145
2146                         case IPV6_RECVHOPOPTS:
2147                         case IPV6_RECVDSTOPTS:
2148                         case IPV6_RECVRTHDRDSTOPTS:
2149                         case IPV6_UNICAST_HOPS:
2150                         case IPV6_RECVPKTINFO:
2151                         case IPV6_RECVHOPLIMIT:
2152                         case IPV6_RECVRTHDR:
2153                         case IPV6_RECVPATHMTU:
2154
2155                         case IPV6_V6ONLY:
2156                         case IPV6_PORTRANGE:
2157                         case IPV6_RECVTCLASS:
2158                         case IPV6_AUTOFLOWLABEL:
2159                         case IPV6_BINDANY:
2160                         case IPV6_FLOWID:
2161                         case IPV6_FLOWTYPE:
2162                         case IPV6_RECVFLOWID:
2163 #ifdef  RSS
2164                         case IPV6_RSSBUCKETID:
2165                         case IPV6_RECVRSSBUCKETID:
2166 #endif
2167                         case IPV6_BINDMULTI:
2168                                 switch (optname) {
2169
2170                                 case IPV6_RECVHOPOPTS:
2171                                         optval = OPTBIT(IN6P_HOPOPTS);
2172                                         break;
2173
2174                                 case IPV6_RECVDSTOPTS:
2175                                         optval = OPTBIT(IN6P_DSTOPTS);
2176                                         break;
2177
2178                                 case IPV6_RECVRTHDRDSTOPTS:
2179                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2180                                         break;
2181
2182                                 case IPV6_UNICAST_HOPS:
2183                                         optval = inp->in6p_hops;
2184                                         break;
2185
2186                                 case IPV6_RECVPKTINFO:
2187                                         optval = OPTBIT(IN6P_PKTINFO);
2188                                         break;
2189
2190                                 case IPV6_RECVHOPLIMIT:
2191                                         optval = OPTBIT(IN6P_HOPLIMIT);
2192                                         break;
2193
2194                                 case IPV6_RECVRTHDR:
2195                                         optval = OPTBIT(IN6P_RTHDR);
2196                                         break;
2197
2198                                 case IPV6_RECVPATHMTU:
2199                                         optval = OPTBIT(IN6P_MTU);
2200                                         break;
2201
2202                                 case IPV6_V6ONLY:
2203                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2204                                         break;
2205
2206                                 case IPV6_PORTRANGE:
2207                                     {
2208                                         int flags;
2209                                         flags = inp->inp_flags;
2210                                         if (flags & INP_HIGHPORT)
2211                                                 optval = IPV6_PORTRANGE_HIGH;
2212                                         else if (flags & INP_LOWPORT)
2213                                                 optval = IPV6_PORTRANGE_LOW;
2214                                         else
2215                                                 optval = 0;
2216                                         break;
2217                                     }
2218                                 case IPV6_RECVTCLASS:
2219                                         optval = OPTBIT(IN6P_TCLASS);
2220                                         break;
2221
2222                                 case IPV6_AUTOFLOWLABEL:
2223                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2224                                         break;
2225
2226                                 case IPV6_ORIGDSTADDR:
2227                                         optval = OPTBIT2(INP_ORIGDSTADDR);
2228                                         break;
2229
2230                                 case IPV6_BINDANY:
2231                                         optval = OPTBIT(INP_BINDANY);
2232                                         break;
2233
2234                                 case IPV6_FLOWID:
2235                                         optval = inp->inp_flowid;
2236                                         break;
2237
2238                                 case IPV6_FLOWTYPE:
2239                                         optval = inp->inp_flowtype;
2240                                         break;
2241
2242                                 case IPV6_RECVFLOWID:
2243                                         optval = OPTBIT2(INP_RECVFLOWID);
2244                                         break;
2245 #ifdef  RSS
2246                                 case IPV6_RSSBUCKETID:
2247                                         retval =
2248                                             rss_hash2bucket(inp->inp_flowid,
2249                                             inp->inp_flowtype,
2250                                             &rss_bucket);
2251                                         if (retval == 0)
2252                                                 optval = rss_bucket;
2253                                         else
2254                                                 error = EINVAL;
2255                                         break;
2256
2257                                 case IPV6_RECVRSSBUCKETID:
2258                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2259                                         break;
2260 #endif
2261
2262                                 case IPV6_BINDMULTI:
2263                                         optval = OPTBIT2(INP_BINDMULTI);
2264                                         break;
2265
2266                                 }
2267                                 if (error)
2268                                         break;
2269                                 error = sooptcopyout(sopt, &optval,
2270                                         sizeof optval);
2271                                 break;
2272
2273                         case IPV6_PATHMTU:
2274                         {
2275                                 u_long pmtu = 0;
2276                                 struct ip6_mtuinfo mtuinfo;
2277                                 struct in6_addr addr;
2278
2279                                 if (!(so->so_state & SS_ISCONNECTED))
2280                                         return (ENOTCONN);
2281                                 /*
2282                                  * XXX: we dot not consider the case of source
2283                                  * routing, or optional information to specify
2284                                  * the outgoing interface.
2285                                  * Copy faddr out of inp to avoid holding lock
2286                                  * on inp during route lookup.
2287                                  */
2288                                 INP_RLOCK(inp);
2289                                 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2290                                 INP_RUNLOCK(inp);
2291                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2292                                     &addr, &pmtu);
2293                                 if (error)
2294                                         break;
2295                                 if (pmtu > IPV6_MAXPACKET)
2296                                         pmtu = IPV6_MAXPACKET;
2297
2298                                 bzero(&mtuinfo, sizeof(mtuinfo));
2299                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2300                                 optdata = (void *)&mtuinfo;
2301                                 optdatalen = sizeof(mtuinfo);
2302                                 error = sooptcopyout(sopt, optdata,
2303                                     optdatalen);
2304                                 break;
2305                         }
2306
2307                         case IPV6_2292PKTINFO:
2308                         case IPV6_2292HOPLIMIT:
2309                         case IPV6_2292HOPOPTS:
2310                         case IPV6_2292RTHDR:
2311                         case IPV6_2292DSTOPTS:
2312                                 switch (optname) {
2313                                 case IPV6_2292PKTINFO:
2314                                         optval = OPTBIT(IN6P_PKTINFO);
2315                                         break;
2316                                 case IPV6_2292HOPLIMIT:
2317                                         optval = OPTBIT(IN6P_HOPLIMIT);
2318                                         break;
2319                                 case IPV6_2292HOPOPTS:
2320                                         optval = OPTBIT(IN6P_HOPOPTS);
2321                                         break;
2322                                 case IPV6_2292RTHDR:
2323                                         optval = OPTBIT(IN6P_RTHDR);
2324                                         break;
2325                                 case IPV6_2292DSTOPTS:
2326                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2327                                         break;
2328                                 }
2329                                 error = sooptcopyout(sopt, &optval,
2330                                     sizeof optval);
2331                                 break;
2332                         case IPV6_PKTINFO:
2333                         case IPV6_HOPOPTS:
2334                         case IPV6_RTHDR:
2335                         case IPV6_DSTOPTS:
2336                         case IPV6_RTHDRDSTOPTS:
2337                         case IPV6_NEXTHOP:
2338                         case IPV6_TCLASS:
2339                         case IPV6_DONTFRAG:
2340                         case IPV6_USE_MIN_MTU:
2341                         case IPV6_PREFER_TEMPADDR:
2342                                 error = ip6_getpcbopt(inp, optname, sopt);
2343                                 break;
2344
2345                         case IPV6_MULTICAST_IF:
2346                         case IPV6_MULTICAST_HOPS:
2347                         case IPV6_MULTICAST_LOOP:
2348                         case IPV6_MSFILTER:
2349                                 error = ip6_getmoptions(inp, sopt);
2350                                 break;
2351
2352 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2353                         case IPV6_IPSEC_POLICY:
2354                                 if (IPSEC_ENABLED(ipv6)) {
2355                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2356                                         break;
2357                                 }
2358                                 /* FALLTHROUGH */
2359 #endif /* IPSEC */
2360                         default:
2361                                 error = ENOPROTOOPT;
2362                                 break;
2363                         }
2364                         break;
2365                 }
2366         }
2367         return (error);
2368 }
2369
2370 int
2371 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2372 {
2373         int error = 0, optval, optlen;
2374         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2375         struct inpcb *inp = sotoinpcb(so);
2376         int level, op, optname;
2377
2378         level = sopt->sopt_level;
2379         op = sopt->sopt_dir;
2380         optname = sopt->sopt_name;
2381         optlen = sopt->sopt_valsize;
2382
2383         if (level != IPPROTO_IPV6) {
2384                 return (EINVAL);
2385         }
2386
2387         switch (optname) {
2388         case IPV6_CHECKSUM:
2389                 /*
2390                  * For ICMPv6 sockets, no modification allowed for checksum
2391                  * offset, permit "no change" values to help existing apps.
2392                  *
2393                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2394                  * for an ICMPv6 socket will fail."
2395                  * The current behavior does not meet RFC3542.
2396                  */
2397                 switch (op) {
2398                 case SOPT_SET:
2399                         if (optlen != sizeof(int)) {
2400                                 error = EINVAL;
2401                                 break;
2402                         }
2403                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2404                                             sizeof(optval));
2405                         if (error)
2406                                 break;
2407                         if (optval < -1 || (optval % 2) != 0) {
2408                                 /*
2409                                  * The API assumes non-negative even offset
2410                                  * values or -1 as a special value.
2411                                  */
2412                                 error = EINVAL;
2413                         } else if (so->so_proto->pr_protocol ==
2414                             IPPROTO_ICMPV6) {
2415                                 if (optval != icmp6off)
2416                                         error = EINVAL;
2417                         } else
2418                                 inp->in6p_cksum = optval;
2419                         break;
2420
2421                 case SOPT_GET:
2422                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2423                                 optval = icmp6off;
2424                         else
2425                                 optval = inp->in6p_cksum;
2426
2427                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2428                         break;
2429
2430                 default:
2431                         error = EINVAL;
2432                         break;
2433                 }
2434                 break;
2435
2436         default:
2437                 error = ENOPROTOOPT;
2438                 break;
2439         }
2440
2441         return (error);
2442 }
2443
2444 /*
2445  * Set up IP6 options in pcb for insertion in output packets or
2446  * specifying behavior of outgoing packets.
2447  */
2448 static int
2449 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2450     struct socket *so, struct sockopt *sopt)
2451 {
2452         struct ip6_pktopts *opt = *pktopt;
2453         int error = 0;
2454         struct thread *td = sopt->sopt_td;
2455
2456         /* turn off any old options. */
2457         if (opt) {
2458 #ifdef DIAGNOSTIC
2459                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2460                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2461                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2462                         printf("ip6_pcbopts: all specified options are cleared.\n");
2463 #endif
2464                 ip6_clearpktopts(opt, -1);
2465         } else {
2466                 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2467                 if (opt == NULL)
2468                         return (ENOMEM);
2469         }
2470         *pktopt = NULL;
2471
2472         if (!m || m->m_len == 0) {
2473                 /*
2474                  * Only turning off any previous options, regardless of
2475                  * whether the opt is just created or given.
2476                  */
2477                 free(opt, M_IP6OPT);
2478                 return (0);
2479         }
2480
2481         /*  set options specified by user. */
2482         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2483             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2484                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2485                 free(opt, M_IP6OPT);
2486                 return (error);
2487         }
2488         *pktopt = opt;
2489         return (0);
2490 }
2491
2492 /*
2493  * initialize ip6_pktopts.  beware that there are non-zero default values in
2494  * the struct.
2495  */
2496 void
2497 ip6_initpktopts(struct ip6_pktopts *opt)
2498 {
2499
2500         bzero(opt, sizeof(*opt));
2501         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2502         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2503         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2504         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2505 }
2506
2507 static int
2508 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2509     struct ucred *cred, int uproto)
2510 {
2511         struct ip6_pktopts *opt;
2512
2513         if (*pktopt == NULL) {
2514                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2515                     M_NOWAIT);
2516                 if (*pktopt == NULL)
2517                         return (ENOBUFS);
2518                 ip6_initpktopts(*pktopt);
2519         }
2520         opt = *pktopt;
2521
2522         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2523 }
2524
2525 #define GET_PKTOPT_VAR(field, lenexpr) do {                                     \
2526         if (pktopt && pktopt->field) {                                          \
2527                 INP_RUNLOCK(inp);                                               \
2528                 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);         \
2529                 malloc_optdata = true;                                          \
2530                 INP_RLOCK(inp);                                                 \
2531                 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {            \
2532                         INP_RUNLOCK(inp);                                       \
2533                         free(optdata, M_TEMP);                                  \
2534                         return (ECONNRESET);                                    \
2535                 }                                                               \
2536                 pktopt = inp->in6p_outputopts;                                  \
2537                 if (pktopt && pktopt->field) {                                  \
2538                         optdatalen = min(lenexpr, sopt->sopt_valsize);          \
2539                         bcopy(&pktopt->field, optdata, optdatalen);             \
2540                 } else {                                                        \
2541                         free(optdata, M_TEMP);                                  \
2542                         optdata = NULL;                                         \
2543                         malloc_optdata = false;                                 \
2544                 }                                                               \
2545         }                                                                       \
2546 } while(0)
2547
2548 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,                         \
2549         (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2550
2551 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,                        \
2552         pktopt->field->sa_len)
2553
2554 static int
2555 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2556 {
2557         void *optdata = NULL;
2558         bool malloc_optdata = false;
2559         int optdatalen = 0;
2560         int error = 0;
2561         struct in6_pktinfo null_pktinfo;
2562         int deftclass = 0, on;
2563         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2564         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2565         struct ip6_pktopts *pktopt;
2566
2567         INP_RLOCK(inp);
2568         pktopt = inp->in6p_outputopts;
2569
2570         switch (optname) {
2571         case IPV6_PKTINFO:
2572                 optdata = (void *)&null_pktinfo;
2573                 if (pktopt && pktopt->ip6po_pktinfo) {
2574                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2575                             sizeof(null_pktinfo));
2576                         in6_clearscope(&null_pktinfo.ipi6_addr);
2577                 } else {
2578                         /* XXX: we don't have to do this every time... */
2579                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2580                 }
2581                 optdatalen = sizeof(struct in6_pktinfo);
2582                 break;
2583         case IPV6_TCLASS:
2584                 if (pktopt && pktopt->ip6po_tclass >= 0)
2585                         deftclass = pktopt->ip6po_tclass;
2586                 optdata = (void *)&deftclass;
2587                 optdatalen = sizeof(int);
2588                 break;
2589         case IPV6_HOPOPTS:
2590                 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2591                 break;
2592         case IPV6_RTHDR:
2593                 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2594                 break;
2595         case IPV6_RTHDRDSTOPTS:
2596                 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2597                 break;
2598         case IPV6_DSTOPTS:
2599                 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2600                 break;
2601         case IPV6_NEXTHOP:
2602                 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2603                 break;
2604         case IPV6_USE_MIN_MTU:
2605                 if (pktopt)
2606                         defminmtu = pktopt->ip6po_minmtu;
2607                 optdata = (void *)&defminmtu;
2608                 optdatalen = sizeof(int);
2609                 break;
2610         case IPV6_DONTFRAG:
2611                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2612                         on = 1;
2613                 else
2614                         on = 0;
2615                 optdata = (void *)&on;
2616                 optdatalen = sizeof(on);
2617                 break;
2618         case IPV6_PREFER_TEMPADDR:
2619                 if (pktopt)
2620                         defpreftemp = pktopt->ip6po_prefer_tempaddr;
2621                 optdata = (void *)&defpreftemp;
2622                 optdatalen = sizeof(int);
2623                 break;
2624         default:                /* should not happen */
2625 #ifdef DIAGNOSTIC
2626                 panic("ip6_getpcbopt: unexpected option\n");
2627 #endif
2628                 INP_RUNLOCK(inp);
2629                 return (ENOPROTOOPT);
2630         }
2631         INP_RUNLOCK(inp);
2632
2633         error = sooptcopyout(sopt, optdata, optdatalen);
2634         if (malloc_optdata)
2635                 free(optdata, M_TEMP);
2636
2637         return (error);
2638 }
2639
2640 void
2641 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2642 {
2643         if (pktopt == NULL)
2644                 return;
2645
2646         if (optname == -1 || optname == IPV6_PKTINFO) {
2647                 if (pktopt->ip6po_pktinfo)
2648                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2649                 pktopt->ip6po_pktinfo = NULL;
2650         }
2651         if (optname == -1 || optname == IPV6_HOPLIMIT)
2652                 pktopt->ip6po_hlim = -1;
2653         if (optname == -1 || optname == IPV6_TCLASS)
2654                 pktopt->ip6po_tclass = -1;
2655         if (optname == -1 || optname == IPV6_NEXTHOP) {
2656                 if (pktopt->ip6po_nextroute.ro_nh) {
2657                         NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2658                         pktopt->ip6po_nextroute.ro_nh = NULL;
2659                 }
2660                 if (pktopt->ip6po_nexthop)
2661                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2662                 pktopt->ip6po_nexthop = NULL;
2663         }
2664         if (optname == -1 || optname == IPV6_HOPOPTS) {
2665                 if (pktopt->ip6po_hbh)
2666                         free(pktopt->ip6po_hbh, M_IP6OPT);
2667                 pktopt->ip6po_hbh = NULL;
2668         }
2669         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2670                 if (pktopt->ip6po_dest1)
2671                         free(pktopt->ip6po_dest1, M_IP6OPT);
2672                 pktopt->ip6po_dest1 = NULL;
2673         }
2674         if (optname == -1 || optname == IPV6_RTHDR) {
2675                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2676                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2677                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2678                 if (pktopt->ip6po_route.ro_nh) {
2679                         NH_FREE(pktopt->ip6po_route.ro_nh);
2680                         pktopt->ip6po_route.ro_nh = NULL;
2681                 }
2682         }
2683         if (optname == -1 || optname == IPV6_DSTOPTS) {
2684                 if (pktopt->ip6po_dest2)
2685                         free(pktopt->ip6po_dest2, M_IP6OPT);
2686                 pktopt->ip6po_dest2 = NULL;
2687         }
2688 }
2689
2690 #define PKTOPT_EXTHDRCPY(type) \
2691 do {\
2692         if (src->type) {\
2693                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2694                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2695                 if (dst->type == NULL)\
2696                         goto bad;\
2697                 bcopy(src->type, dst->type, hlen);\
2698         }\
2699 } while (/*CONSTCOND*/ 0)
2700
2701 static int
2702 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2703 {
2704         if (dst == NULL || src == NULL)  {
2705                 printf("ip6_clearpktopts: invalid argument\n");
2706                 return (EINVAL);
2707         }
2708
2709         dst->ip6po_hlim = src->ip6po_hlim;
2710         dst->ip6po_tclass = src->ip6po_tclass;
2711         dst->ip6po_flags = src->ip6po_flags;
2712         dst->ip6po_minmtu = src->ip6po_minmtu;
2713         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2714         if (src->ip6po_pktinfo) {
2715                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2716                     M_IP6OPT, canwait);
2717                 if (dst->ip6po_pktinfo == NULL)
2718                         goto bad;
2719                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2720         }
2721         if (src->ip6po_nexthop) {
2722                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2723                     M_IP6OPT, canwait);
2724                 if (dst->ip6po_nexthop == NULL)
2725                         goto bad;
2726                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2727                     src->ip6po_nexthop->sa_len);
2728         }
2729         PKTOPT_EXTHDRCPY(ip6po_hbh);
2730         PKTOPT_EXTHDRCPY(ip6po_dest1);
2731         PKTOPT_EXTHDRCPY(ip6po_dest2);
2732         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2733         return (0);
2734
2735   bad:
2736         ip6_clearpktopts(dst, -1);
2737         return (ENOBUFS);
2738 }
2739 #undef PKTOPT_EXTHDRCPY
2740
2741 struct ip6_pktopts *
2742 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2743 {
2744         int error;
2745         struct ip6_pktopts *dst;
2746
2747         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2748         if (dst == NULL)
2749                 return (NULL);
2750         ip6_initpktopts(dst);
2751
2752         if ((error = copypktopts(dst, src, canwait)) != 0) {
2753                 free(dst, M_IP6OPT);
2754                 return (NULL);
2755         }
2756
2757         return (dst);
2758 }
2759
2760 void
2761 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2762 {
2763         if (pktopt == NULL)
2764                 return;
2765
2766         ip6_clearpktopts(pktopt, -1);
2767
2768         free(pktopt, M_IP6OPT);
2769 }
2770
2771 /*
2772  * Set IPv6 outgoing packet options based on advanced API.
2773  */
2774 int
2775 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2776     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2777 {
2778         struct cmsghdr *cm = NULL;
2779
2780         if (control == NULL || opt == NULL)
2781                 return (EINVAL);
2782
2783         ip6_initpktopts(opt);
2784         if (stickyopt) {
2785                 int error;
2786
2787                 /*
2788                  * If stickyopt is provided, make a local copy of the options
2789                  * for this particular packet, then override them by ancillary
2790                  * objects.
2791                  * XXX: copypktopts() does not copy the cached route to a next
2792                  * hop (if any).  This is not very good in terms of efficiency,
2793                  * but we can allow this since this option should be rarely
2794                  * used.
2795                  */
2796                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2797                         return (error);
2798         }
2799
2800         /*
2801          * XXX: Currently, we assume all the optional information is stored
2802          * in a single mbuf.
2803          */
2804         if (control->m_next)
2805                 return (EINVAL);
2806
2807         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2808             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2809                 int error;
2810
2811                 if (control->m_len < CMSG_LEN(0))
2812                         return (EINVAL);
2813
2814                 cm = mtod(control, struct cmsghdr *);
2815                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2816                         return (EINVAL);
2817                 if (cm->cmsg_level != IPPROTO_IPV6)
2818                         continue;
2819
2820                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2821                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2822                 if (error)
2823                         return (error);
2824         }
2825
2826         return (0);
2827 }
2828
2829 /*
2830  * Set a particular packet option, as a sticky option or an ancillary data
2831  * item.  "len" can be 0 only when it's a sticky option.
2832  * We have 4 cases of combination of "sticky" and "cmsg":
2833  * "sticky=0, cmsg=0": impossible
2834  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2835  * "sticky=1, cmsg=0": RFC3542 socket option
2836  * "sticky=1, cmsg=1": RFC2292 socket option
2837  */
2838 static int
2839 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2840     struct ucred *cred, int sticky, int cmsg, int uproto)
2841 {
2842         int minmtupolicy, preftemp;
2843         int error;
2844
2845         if (!sticky && !cmsg) {
2846 #ifdef DIAGNOSTIC
2847                 printf("ip6_setpktopt: impossible case\n");
2848 #endif
2849                 return (EINVAL);
2850         }
2851
2852         /*
2853          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2854          * not be specified in the context of RFC3542.  Conversely,
2855          * RFC3542 types should not be specified in the context of RFC2292.
2856          */
2857         if (!cmsg) {
2858                 switch (optname) {
2859                 case IPV6_2292PKTINFO:
2860                 case IPV6_2292HOPLIMIT:
2861                 case IPV6_2292NEXTHOP:
2862                 case IPV6_2292HOPOPTS:
2863                 case IPV6_2292DSTOPTS:
2864                 case IPV6_2292RTHDR:
2865                 case IPV6_2292PKTOPTIONS:
2866                         return (ENOPROTOOPT);
2867                 }
2868         }
2869         if (sticky && cmsg) {
2870                 switch (optname) {
2871                 case IPV6_PKTINFO:
2872                 case IPV6_HOPLIMIT:
2873                 case IPV6_NEXTHOP:
2874                 case IPV6_HOPOPTS:
2875                 case IPV6_DSTOPTS:
2876                 case IPV6_RTHDRDSTOPTS:
2877                 case IPV6_RTHDR:
2878                 case IPV6_USE_MIN_MTU:
2879                 case IPV6_DONTFRAG:
2880                 case IPV6_TCLASS:
2881                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2882                         return (ENOPROTOOPT);
2883                 }
2884         }
2885
2886         switch (optname) {
2887         case IPV6_2292PKTINFO:
2888         case IPV6_PKTINFO:
2889         {
2890                 struct ifnet *ifp = NULL;
2891                 struct in6_pktinfo *pktinfo;
2892
2893                 if (len != sizeof(struct in6_pktinfo))
2894                         return (EINVAL);
2895
2896                 pktinfo = (struct in6_pktinfo *)buf;
2897
2898                 /*
2899                  * An application can clear any sticky IPV6_PKTINFO option by
2900                  * doing a "regular" setsockopt with ipi6_addr being
2901                  * in6addr_any and ipi6_ifindex being zero.
2902                  * [RFC 3542, Section 6]
2903                  */
2904                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2905                     pktinfo->ipi6_ifindex == 0 &&
2906                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2907                         ip6_clearpktopts(opt, optname);
2908                         break;
2909                 }
2910
2911                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2912                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2913                         return (EINVAL);
2914                 }
2915                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2916                         return (EINVAL);
2917                 /* validate the interface index if specified. */
2918                 if (pktinfo->ipi6_ifindex > V_if_index)
2919                          return (ENXIO);
2920                 if (pktinfo->ipi6_ifindex) {
2921                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2922                         if (ifp == NULL)
2923                                 return (ENXIO);
2924                 }
2925                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2926                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2927                         return (ENETDOWN);
2928
2929                 if (ifp != NULL &&
2930                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2931                         struct in6_ifaddr *ia;
2932
2933                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2934                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2935                         if (ia == NULL)
2936                                 return (EADDRNOTAVAIL);
2937                         ifa_free(&ia->ia_ifa);
2938                 }
2939                 /*
2940                  * We store the address anyway, and let in6_selectsrc()
2941                  * validate the specified address.  This is because ipi6_addr
2942                  * may not have enough information about its scope zone, and
2943                  * we may need additional information (such as outgoing
2944                  * interface or the scope zone of a destination address) to
2945                  * disambiguate the scope.
2946                  * XXX: the delay of the validation may confuse the
2947                  * application when it is used as a sticky option.
2948                  */
2949                 if (opt->ip6po_pktinfo == NULL) {
2950                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2951                             M_IP6OPT, M_NOWAIT);
2952                         if (opt->ip6po_pktinfo == NULL)
2953                                 return (ENOBUFS);
2954                 }
2955                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2956                 break;
2957         }
2958
2959         case IPV6_2292HOPLIMIT:
2960         case IPV6_HOPLIMIT:
2961         {
2962                 int *hlimp;
2963
2964                 /*
2965                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2966                  * to simplify the ordering among hoplimit options.
2967                  */
2968                 if (optname == IPV6_HOPLIMIT && sticky)
2969                         return (ENOPROTOOPT);
2970
2971                 if (len != sizeof(int))
2972                         return (EINVAL);
2973                 hlimp = (int *)buf;
2974                 if (*hlimp < -1 || *hlimp > 255)
2975                         return (EINVAL);
2976
2977                 opt->ip6po_hlim = *hlimp;
2978                 break;
2979         }
2980
2981         case IPV6_TCLASS:
2982         {
2983                 int tclass;
2984
2985                 if (len != sizeof(int))
2986                         return (EINVAL);
2987                 tclass = *(int *)buf;
2988                 if (tclass < -1 || tclass > 255)
2989                         return (EINVAL);
2990
2991                 opt->ip6po_tclass = tclass;
2992                 break;
2993         }
2994
2995         case IPV6_2292NEXTHOP:
2996         case IPV6_NEXTHOP:
2997                 if (cred != NULL) {
2998                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2999                         if (error)
3000                                 return (error);
3001                 }
3002
3003                 if (len == 0) { /* just remove the option */
3004                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3005                         break;
3006                 }
3007
3008                 /* check if cmsg_len is large enough for sa_len */
3009                 if (len < sizeof(struct sockaddr) || len < *buf)
3010                         return (EINVAL);
3011
3012                 switch (((struct sockaddr *)buf)->sa_family) {
3013                 case AF_INET6:
3014                 {
3015                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3016                         int error;
3017
3018                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3019                                 return (EINVAL);
3020
3021                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3022                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3023                                 return (EINVAL);
3024                         }
3025                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3026                             != 0) {
3027                                 return (error);
3028                         }
3029                         break;
3030                 }
3031                 case AF_LINK:   /* should eventually be supported */
3032                 default:
3033                         return (EAFNOSUPPORT);
3034                 }
3035
3036                 /* turn off the previous option, then set the new option. */
3037                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3038                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3039                 if (opt->ip6po_nexthop == NULL)
3040                         return (ENOBUFS);
3041                 bcopy(buf, opt->ip6po_nexthop, *buf);
3042                 break;
3043
3044         case IPV6_2292HOPOPTS:
3045         case IPV6_HOPOPTS:
3046         {
3047                 struct ip6_hbh *hbh;
3048                 int hbhlen;
3049
3050                 /*
3051                  * XXX: We don't allow a non-privileged user to set ANY HbH
3052                  * options, since per-option restriction has too much
3053                  * overhead.
3054                  */
3055                 if (cred != NULL) {
3056                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3057                         if (error)
3058                                 return (error);
3059                 }
3060
3061                 if (len == 0) {
3062                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3063                         break;  /* just remove the option */
3064                 }
3065
3066                 /* message length validation */
3067                 if (len < sizeof(struct ip6_hbh))
3068                         return (EINVAL);
3069                 hbh = (struct ip6_hbh *)buf;
3070                 hbhlen = (hbh->ip6h_len + 1) << 3;
3071                 if (len != hbhlen)
3072                         return (EINVAL);
3073
3074                 /* turn off the previous option, then set the new option. */
3075                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3076                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3077                 if (opt->ip6po_hbh == NULL)
3078                         return (ENOBUFS);
3079                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3080
3081                 break;
3082         }
3083
3084         case IPV6_2292DSTOPTS:
3085         case IPV6_DSTOPTS:
3086         case IPV6_RTHDRDSTOPTS:
3087         {
3088                 struct ip6_dest *dest, **newdest = NULL;
3089                 int destlen;
3090
3091                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3092                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3093                         if (error)
3094                                 return (error);
3095                 }
3096
3097                 if (len == 0) {
3098                         ip6_clearpktopts(opt, optname);
3099                         break;  /* just remove the option */
3100                 }
3101
3102                 /* message length validation */
3103                 if (len < sizeof(struct ip6_dest))
3104                         return (EINVAL);
3105                 dest = (struct ip6_dest *)buf;
3106                 destlen = (dest->ip6d_len + 1) << 3;
3107                 if (len != destlen)
3108                         return (EINVAL);
3109
3110                 /*
3111                  * Determine the position that the destination options header
3112                  * should be inserted; before or after the routing header.
3113                  */
3114                 switch (optname) {
3115                 case IPV6_2292DSTOPTS:
3116                         /*
3117                          * The old advacned API is ambiguous on this point.
3118                          * Our approach is to determine the position based
3119                          * according to the existence of a routing header.
3120                          * Note, however, that this depends on the order of the
3121                          * extension headers in the ancillary data; the 1st
3122                          * part of the destination options header must appear
3123                          * before the routing header in the ancillary data,
3124                          * too.
3125                          * RFC3542 solved the ambiguity by introducing
3126                          * separate ancillary data or option types.
3127                          */
3128                         if (opt->ip6po_rthdr == NULL)
3129                                 newdest = &opt->ip6po_dest1;
3130                         else
3131                                 newdest = &opt->ip6po_dest2;
3132                         break;
3133                 case IPV6_RTHDRDSTOPTS:
3134                         newdest = &opt->ip6po_dest1;
3135                         break;
3136                 case IPV6_DSTOPTS:
3137                         newdest = &opt->ip6po_dest2;
3138                         break;
3139                 }
3140
3141                 /* turn off the previous option, then set the new option. */
3142                 ip6_clearpktopts(opt, optname);
3143                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3144                 if (*newdest == NULL)
3145                         return (ENOBUFS);
3146                 bcopy(dest, *newdest, destlen);
3147
3148                 break;
3149         }
3150
3151         case IPV6_2292RTHDR:
3152         case IPV6_RTHDR:
3153         {
3154                 struct ip6_rthdr *rth;
3155                 int rthlen;
3156
3157                 if (len == 0) {
3158                         ip6_clearpktopts(opt, IPV6_RTHDR);
3159                         break;  /* just remove the option */
3160                 }
3161
3162                 /* message length validation */
3163                 if (len < sizeof(struct ip6_rthdr))
3164                         return (EINVAL);
3165                 rth = (struct ip6_rthdr *)buf;
3166                 rthlen = (rth->ip6r_len + 1) << 3;
3167                 if (len != rthlen)
3168                         return (EINVAL);
3169
3170                 switch (rth->ip6r_type) {
3171                 case IPV6_RTHDR_TYPE_0:
3172                         if (rth->ip6r_len == 0) /* must contain one addr */
3173                                 return (EINVAL);
3174                         if (rth->ip6r_len % 2) /* length must be even */
3175                                 return (EINVAL);
3176                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3177                                 return (EINVAL);
3178                         break;
3179                 default:
3180                         return (EINVAL);        /* not supported */
3181                 }
3182
3183                 /* turn off the previous option */
3184                 ip6_clearpktopts(opt, IPV6_RTHDR);
3185                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3186                 if (opt->ip6po_rthdr == NULL)
3187                         return (ENOBUFS);
3188                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3189
3190                 break;
3191         }
3192
3193         case IPV6_USE_MIN_MTU:
3194                 if (len != sizeof(int))
3195                         return (EINVAL);
3196                 minmtupolicy = *(int *)buf;
3197                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3198                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3199                     minmtupolicy != IP6PO_MINMTU_ALL) {
3200                         return (EINVAL);
3201                 }
3202                 opt->ip6po_minmtu = minmtupolicy;
3203                 break;
3204
3205         case IPV6_DONTFRAG:
3206                 if (len != sizeof(int))
3207                         return (EINVAL);
3208
3209                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3210                         /*
3211                          * we ignore this option for TCP sockets.
3212                          * (RFC3542 leaves this case unspecified.)
3213                          */
3214                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3215                 } else
3216                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3217                 break;
3218
3219         case IPV6_PREFER_TEMPADDR:
3220                 if (len != sizeof(int))
3221                         return (EINVAL);
3222                 preftemp = *(int *)buf;
3223                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3224                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3225                     preftemp != IP6PO_TEMPADDR_PREFER) {
3226                         return (EINVAL);
3227                 }
3228                 opt->ip6po_prefer_tempaddr = preftemp;
3229                 break;
3230
3231         default:
3232                 return (ENOPROTOOPT);
3233         } /* end of switch */
3234
3235         return (0);
3236 }
3237
3238 /*
3239  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3240  * packet to the input queue of a specified interface.  Note that this
3241  * calls the output routine of the loopback "driver", but with an interface
3242  * pointer that might NOT be &loif -- easier than replicating that code here.
3243  */
3244 void
3245 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3246 {
3247         struct mbuf *copym;
3248         struct ip6_hdr *ip6;
3249
3250         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3251         if (copym == NULL)
3252                 return;
3253
3254         /*
3255          * Make sure to deep-copy IPv6 header portion in case the data
3256          * is in an mbuf cluster, so that we can safely override the IPv6
3257          * header portion later.
3258          */
3259         if (!M_WRITABLE(copym) ||
3260             copym->m_len < sizeof(struct ip6_hdr)) {
3261                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3262                 if (copym == NULL)
3263                         return;
3264         }
3265         ip6 = mtod(copym, struct ip6_hdr *);
3266         /*
3267          * clear embedded scope identifiers if necessary.
3268          * in6_clearscope will touch the addresses only when necessary.
3269          */
3270         in6_clearscope(&ip6->ip6_src);
3271         in6_clearscope(&ip6->ip6_dst);
3272         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3273                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3274                     CSUM_PSEUDO_HDR;
3275                 copym->m_pkthdr.csum_data = 0xffff;
3276         }
3277         if_simloop(ifp, copym, AF_INET6, 0);
3278 }
3279
3280 /*
3281  * Chop IPv6 header off from the payload.
3282  */
3283 static int
3284 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3285 {
3286         struct mbuf *mh;
3287         struct ip6_hdr *ip6;
3288
3289         ip6 = mtod(m, struct ip6_hdr *);
3290         if (m->m_len > sizeof(*ip6)) {
3291                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3292                 if (mh == NULL) {
3293                         m_freem(m);
3294                         return ENOBUFS;
3295                 }
3296                 m_move_pkthdr(mh, m);
3297                 M_ALIGN(mh, sizeof(*ip6));
3298                 m->m_len -= sizeof(*ip6);
3299                 m->m_data += sizeof(*ip6);
3300                 mh->m_next = m;
3301                 m = mh;
3302                 m->m_len = sizeof(*ip6);
3303                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3304         }
3305         exthdrs->ip6e_ip6 = m;
3306         return 0;
3307 }
3308
3309 /*
3310  * Compute IPv6 extension header length.
3311  */
3312 int
3313 ip6_optlen(struct inpcb *inp)
3314 {
3315         int len;
3316
3317         if (!inp->in6p_outputopts)
3318                 return 0;
3319
3320         len = 0;
3321 #define elen(x) \
3322     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3323
3324         len += elen(inp->in6p_outputopts->ip6po_hbh);
3325         if (inp->in6p_outputopts->ip6po_rthdr)
3326                 /* dest1 is valid with rthdr only */
3327                 len += elen(inp->in6p_outputopts->ip6po_dest1);
3328         len += elen(inp->in6p_outputopts->ip6po_rthdr);
3329         len += elen(inp->in6p_outputopts->ip6po_dest2);
3330         return len;
3331 #undef elen
3332 }