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_vlan_var.h>
  96 #include <net/if_llatbl.h>
  97 #include <net/ethernet.h>
  98 #include <net/netisr.h>
  99 #include <net/route.h>
 100 #include <net/route/nhop.h>
 101 #include <net/pfil.h>
 102 #include <net/rss_config.h>
 103 #include <net/vnet.h>
 104
 105 #include <netinet/in.h>
 106 #include <netinet/in_var.h>
 107 #include <netinet/ip_var.h>
 108 #include <netinet6/in6_fib.h>
 109 #include <netinet6/in6_var.h>
 110 #include <netinet/ip6.h>
 111 #include <netinet/icmp6.h>
 112 #include <netinet6/ip6_var.h>
 113 #include <netinet/in_pcb.h>
 114 #include <netinet/tcp_var.h>
 115 #include <netinet6/nd6.h>
 116 #include <netinet6/in6_rss.h>
 117
 118 #include <netipsec/ipsec_support.h>
 119 #if defined(SCTP) || defined(SCTP_SUPPORT)
 120 #include <netinet/sctp.h>
 121 #include <netinet/sctp_crc32.h>
 122 #endif
 123
 124 #include <netinet6/ip6protosw.h>
 125 #include <netinet6/scope6_var.h>
 126
 127 extern int in6_mcast_loop;
 128
 129 struct ip6_exthdrs {
 130         struct mbuf *ip6e_ip6;
 131         struct mbuf *ip6e_hbh;
 132         struct mbuf *ip6e_dest1;
 133         struct mbuf *ip6e_rthdr;
 134         struct mbuf *ip6e_dest2;
 135 };
 136
 137 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
 138
 139 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
 140                            struct ucred *, int);
 141 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
 142         struct socket *, struct sockopt *);
 143 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
 144 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
 145         struct ucred *, int, int, int);
 146
 147 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 148 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 149         struct ip6_frag **);
 150 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 151 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 152 static int ip6_getpmtu(struct route_in6 *, int,
 153         struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
 154         u_int);
 155 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
 156         u_long *, int *, u_int);
 157 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
 158 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 159
 160 /*
 161  * Make an extension header from option data.  hp is the source,
 162  * mp is the destination, and _ol is the optlen.
 163  */
 164 #define MAKE_EXTHDR(hp, mp, _ol)                                        \
 165     do {                                                                \
 166         if (hp) {                                                       \
 167                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 168                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 169                     ((eh)->ip6e_len + 1) << 3);                         \
 170                 if (error)                                              \
 171                         goto freehdrs;                                  \
 172                 (_ol) += (*(mp))->m_len;                                \
 173         }                                                               \
 174     } while (/*CONSTCOND*/ 0)
 175
 176 /*
 177  * Form a chain of extension headers.
 178  * m is the extension header mbuf
 179  * mp is the previous mbuf in the chain
 180  * p is the next header
 181  * i is the type of option.
 182  */
 183 #define MAKE_CHAIN(m, mp, p, i)\
 184     do {\
 185         if (m) {\
 186                 if (!hdrsplit) \
 187                         panic("%s:%d: assumption failed: "\
 188                             "hdr not split: hdrsplit %d exthdrs %p",\
 189                             __func__, __LINE__, hdrsplit, &exthdrs);\
 190                 *mtod((m), u_char *) = *(p);\
 191                 *(p) = (i);\
 192                 p = mtod((m), u_char *);\
 193                 (m)->m_next = (mp)->m_next;\
 194                 (mp)->m_next = (m);\
 195                 (mp) = (m);\
 196         }\
 197     } while (/*CONSTCOND*/ 0)
 198
 199 void
 200 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 201 {
 202         u_short csum;
 203
 204         csum = in_cksum_skip(m, offset + plen, offset);
 205         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 206                 csum = 0xffff;
 207         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 208
 209         if (offset + sizeof(csum) > m->m_len)
 210                 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
 211         else
 212                 *(u_short *)mtodo(m, offset) = csum;
 213 }
 214
 215 static int
 216 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
 217     int plen, int optlen, bool frag)
 218 {
 219
 220         KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
 221             "csum_flags %#x frag %d\n",
 222             __func__, __LINE__, plen, optlen, m, ifp, csum_flags, frag));
 223
 224         if ((csum_flags & CSUM_DELAY_DATA_IPV6) ||
 225 #if defined(SCTP) || defined(SCTP_SUPPORT)
 226             (csum_flags & CSUM_SCTP_IPV6) ||
 227 #endif
 228             (!frag && (ifp->if_capenable & IFCAP_MEXTPG) == 0)) {
 229                 m = mb_unmapped_to_ext(m);
 230                 if (m == NULL) {
 231                         if (frag)
 232                                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 233                         else
 234                                 IP6STAT_INC(ip6s_odropped);
 235                         return (ENOBUFS);
 236                 }
 237                 if (csum_flags & CSUM_DELAY_DATA_IPV6) {
 238                         in6_delayed_cksum(m, plen - optlen,
 239                             sizeof(struct ip6_hdr) + optlen);
 240                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 241                 }
 242 #if defined(SCTP) || defined(SCTP_SUPPORT)
 243                 if (csum_flags & CSUM_SCTP_IPV6) {
 244                         sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
 245                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 246                 }
 247 #endif
 248         }
 249
 250         return (0);
 251 }
 252
 253 int
 254 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
 255     int fraglen , uint32_t id)
 256 {
 257         struct mbuf *m, **mnext, *m_frgpart;
 258         struct ip6_hdr *ip6, *mhip6;
 259         struct ip6_frag *ip6f;
 260         int off;
 261         int error;
 262         int tlen = m0->m_pkthdr.len;
 263
 264         KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
 265
 266         m = m0;
 267         ip6 = mtod(m, struct ip6_hdr *);
 268         mnext = &m->m_nextpkt;
 269
 270         for (off = hlen; off < tlen; off += fraglen) {
 271                 m = m_gethdr(M_NOWAIT, MT_DATA);
 272                 if (!m) {
 273                         IP6STAT_INC(ip6s_odropped);
 274                         return (ENOBUFS);
 275                 }
 276
 277                 /*
 278                  * Make sure the complete packet header gets copied
 279                  * from the originating mbuf to the newly created
 280                  * mbuf. This also ensures that existing firewall
 281                  * classification(s), VLAN tags and so on get copied
 282                  * to the resulting fragmented packet(s):
 283                  */
 284                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 285                         m_free(m);
 286                         IP6STAT_INC(ip6s_odropped);
 287                         return (ENOBUFS);
 288                 }
 289
 290                 *mnext = m;
 291                 mnext = &m->m_nextpkt;
 292                 m->m_data += max_linkhdr;
 293                 mhip6 = mtod(m, struct ip6_hdr *);
 294                 *mhip6 = *ip6;
 295                 m->m_len = sizeof(*mhip6);
 296                 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 297                 if (error) {
 298                         IP6STAT_INC(ip6s_odropped);
 299                         return (error);
 300                 }
 301                 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 302                 if (off + fraglen >= tlen)
 303                         fraglen = tlen - off;
 304                 else
 305                         ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 306                 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
 307                     sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 308                 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
 309                         IP6STAT_INC(ip6s_odropped);
 310                         return (ENOBUFS);
 311                 }
 312                 m_cat(m, m_frgpart);
 313                 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
 314                 ip6f->ip6f_reserved = 0;
 315                 ip6f->ip6f_ident = id;
 316                 ip6f->ip6f_nxt = nextproto;
 317                 IP6STAT_INC(ip6s_ofragments);
 318                 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 319         }
 320
 321         return (0);
 322 }
 323
 324 static int
 325 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
 326     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
 327     bool stamp_tag)
 328 {
 329 #ifdef KERN_TLS
 330         struct ktls_session *tls = NULL;
 331 #endif
 332         struct m_snd_tag *mst;
 333         int error;
 334
 335         MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
 336         mst = NULL;
 337
 338 #ifdef KERN_TLS
 339         /*
 340          * If this is an unencrypted TLS record, save a reference to
 341          * the record.  This local reference is used to call
 342          * ktls_output_eagain after the mbuf has been freed (thus
 343          * dropping the mbuf's reference) in if_output.
 344          */
 345         if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
 346                 tls = ktls_hold(m->m_next->m_epg_tls);
 347                 mst = tls->snd_tag;
 348
 349                 /*
 350                  * If a TLS session doesn't have a valid tag, it must
 351                  * have had an earlier ifp mismatch, so drop this
 352                  * packet.
 353                  */
 354                 if (mst == NULL) {
 355                         error = EAGAIN;
 356                         goto done;
 357                 }
 358                 /*
 359                  * Always stamp tags that include NIC ktls.
 360                  */
 361                 stamp_tag = true;
 362         }
 363 #endif
 364 #ifdef RATELIMIT
 365         if (inp != NULL && mst == NULL) {
 366                 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
 367                     (inp->inp_snd_tag != NULL &&
 368                     inp->inp_snd_tag->ifp != ifp))
 369                         in_pcboutput_txrtlmt(inp, ifp, m);
 370
 371                 if (inp->inp_snd_tag != NULL)
 372                         mst = inp->inp_snd_tag;
 373         }
 374 #endif
 375         if (stamp_tag && mst != NULL) {
 376                 KASSERT(m->m_pkthdr.rcvif == NULL,
 377                     ("trying to add a send tag to a forwarded packet"));
 378                 if (mst->ifp != ifp) {
 379                         error = EAGAIN;
 380                         goto done;
 381                 }
 382
 383                 /* stamp send tag on mbuf */
 384                 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
 385                 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
 386         }
 387
 388         error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
 389
 390 done:
 391         /* Check for route change invalidating send tags. */
 392 #ifdef KERN_TLS
 393         if (tls != NULL) {
 394                 if (error == EAGAIN)
 395                         error = ktls_output_eagain(inp, tls);
 396                 ktls_free(tls);
 397         }
 398 #endif
 399 #ifdef RATELIMIT
 400         if (error == EAGAIN)
 401                 in_pcboutput_eagain(inp);
 402 #endif
 403         return (error);
 404 }
 405
 406 /*
 407  * IP6 output.
 408  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
 409  * nxt, hlim, src, dst).
 410  * This function may modify ver and hlim only.
 411  * The mbuf chain containing the packet will be freed.
 412  * The mbuf opt, if present, will not be freed.
 413  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
 414  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
 415  * then result of route lookup is stored in ro->ro_nh.
 416  *
 417  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
 418  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
 419  *
 420  * ifpp - XXX: just for statistics
 421  */
 422 int
 423 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 424     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 425     struct ifnet **ifpp, struct inpcb *inp)
 426 {
 427         struct ip6_hdr *ip6;
 428         struct ifnet *ifp, *origifp;
 429         struct mbuf *m = m0;
 430         struct mbuf *mprev;
 431         struct route_in6 *ro_pmtu;
 432         struct nhop_object *nh;
 433         struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
 434         struct in6_addr odst;
 435         u_char *nexthdrp;
 436         int tlen, len;
 437         int error = 0;
 438         int vlan_pcp = -1;
 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                 if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
 464                         vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
 465                             INP_2PCP_SHIFT;
 466 #ifdef NUMA
 467                 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
 468 #endif
 469         }
 470
 471 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 472         /*
 473          * IPSec checking which handles several cases.
 474          * FAST IPSEC: We re-injected the packet.
 475          * XXX: need scope argument.
 476          */
 477         if (IPSEC_ENABLED(ipv6)) {
 478                 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
 479                         if (error == EINPROGRESS)
 480                                 error = 0;
 481                         goto done;
 482                 }
 483         }
 484 #endif /* IPSEC */
 485
 486         /* Source address validation. */
 487         ip6 = mtod(m, struct ip6_hdr *);
 488         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 489             (flags & IPV6_UNSPECSRC) == 0) {
 490                 error = EOPNOTSUPP;
 491                 IP6STAT_INC(ip6s_badscope);
 492                 goto bad;
 493         }
 494         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 495                 error = EOPNOTSUPP;
 496                 IP6STAT_INC(ip6s_badscope);
 497                 goto bad;
 498         }
 499
 500         /*
 501          * If we are given packet options to add extension headers prepare them.
 502          * Calculate the total length of the extension header chain.
 503          * Keep the length of the unfragmentable part for fragmentation.
 504          */
 505         bzero(&exthdrs, sizeof(exthdrs));
 506         optlen = 0;
 507         unfragpartlen = sizeof(struct ip6_hdr);
 508         if (opt) {
 509                 /* Hop-by-Hop options header. */
 510                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
 511
 512                 /* Destination options header (1st part). */
 513                 if (opt->ip6po_rthdr) {
 514 #ifndef RTHDR_SUPPORT_IMPLEMENTED
 515                         /*
 516                          * If there is a routing header, discard the packet
 517                          * right away here. RH0/1 are obsolete and we do not
 518                          * currently support RH2/3/4.
 519                          * People trying to use RH253/254 may want to disable
 520                          * this check.
 521                          * The moment we do support any routing header (again)
 522                          * this block should check the routing type more
 523                          * selectively.
 524                          */
 525                         error = EINVAL;
 526                         goto bad;
 527 #endif
 528
 529                         /*
 530                          * Destination options header (1st part).
 531                          * This only makes sense with a routing header.
 532                          * See Section 9.2 of RFC 3542.
 533                          * Disabling this part just for MIP6 convenience is
 534                          * a bad idea.  We need to think carefully about a
 535                          * way to make the advanced API coexist with MIP6
 536                          * options, which might automatically be inserted in
 537                          * the kernel.
 538                          */
 539                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
 540                             optlen);
 541                 }
 542                 /* Routing header. */
 543                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
 544
 545                 unfragpartlen += optlen;
 546
 547                 /*
 548                  * NOTE: we don't add AH/ESP length here (done in
 549                  * ip6_ipsec_output()).
 550                  */
 551
 552                 /* Destination options header (2nd part). */
 553                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
 554         }
 555
 556         /*
 557          * If there is at least one extension header,
 558          * separate IP6 header from the payload.
 559          */
 560         hdrsplit = false;
 561         if (optlen) {
 562                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 563                         m = NULL;
 564                         goto freehdrs;
 565                 }
 566                 m = exthdrs.ip6e_ip6;
 567                 ip6 = mtod(m, struct ip6_hdr *);
 568                 hdrsplit = true;
 569         }
 570
 571         /* Adjust mbuf packet header length. */
 572         m->m_pkthdr.len += optlen;
 573         plen = m->m_pkthdr.len - sizeof(*ip6);
 574
 575         /* If this is a jumbo payload, insert a jumbo payload option. */
 576         if (plen > IPV6_MAXPACKET) {
 577                 if (!hdrsplit) {
 578                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 579                                 m = NULL;
 580                                 goto freehdrs;
 581                         }
 582                         m = exthdrs.ip6e_ip6;
 583                         ip6 = mtod(m, struct ip6_hdr *);
 584                         hdrsplit = true;
 585                 }
 586                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 587                         goto freehdrs;
 588                 ip6->ip6_plen = 0;
 589         } else
 590                 ip6->ip6_plen = htons(plen);
 591         nexthdrp = &ip6->ip6_nxt;
 592
 593         if (optlen) {
 594                 /*
 595                  * Concatenate headers and fill in next header fields.
 596                  * Here we have, on "m"
 597                  *      IPv6 payload
 598                  * and we insert headers accordingly.
 599                  * Finally, we should be getting:
 600                  *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
 601                  *
 602                  * During the header composing process "m" points to IPv6
 603                  * header.  "mprev" points to an extension header prior to esp.
 604                  */
 605                 mprev = m;
 606
 607                 /*
 608                  * We treat dest2 specially.  This makes IPsec processing
 609                  * much easier.  The goal here is to make mprev point the
 610                  * mbuf prior to dest2.
 611                  *
 612                  * Result: IPv6 dest2 payload.
 613                  * m and mprev will point to IPv6 header.
 614                  */
 615                 if (exthdrs.ip6e_dest2) {
 616                         if (!hdrsplit)
 617                                 panic("%s:%d: assumption failed: "
 618                                     "hdr not split: hdrsplit %d exthdrs %p",
 619                                     __func__, __LINE__, hdrsplit, &exthdrs);
 620                         exthdrs.ip6e_dest2->m_next = m->m_next;
 621                         m->m_next = exthdrs.ip6e_dest2;
 622                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 623                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 624                 }
 625
 626                 /*
 627                  * Result: IPv6 hbh dest1 rthdr dest2 payload.
 628                  * m will point to IPv6 header.  mprev will point to the
 629                  * extension header prior to dest2 (rthdr in the above case).
 630                  */
 631                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 632                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 633                            IPPROTO_DSTOPTS);
 634                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 635                            IPPROTO_ROUTING);
 636         }
 637
 638         IP6STAT_INC(ip6s_localout);
 639
 640         /* Route packet. */
 641         ro_pmtu = ro;
 642         if (opt && opt->ip6po_rthdr)
 643                 ro = &opt->ip6po_route;
 644         if (ro != NULL)
 645                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 646         else
 647                 dst = &sin6;
 648         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 649
 650 again:
 651         /*
 652          * If specified, try to fill in the traffic class field.
 653          * Do not override if a non-zero value is already set.
 654          * We check the diffserv field and the ECN field separately.
 655          */
 656         if (opt && opt->ip6po_tclass >= 0) {
 657                 int mask = 0;
 658
 659                 if (IPV6_DSCP(ip6) == 0)
 660                         mask |= 0xfc;
 661                 if (IPV6_ECN(ip6) == 0)
 662                         mask |= 0x03;
 663                 if (mask != 0)
 664                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 665         }
 666
 667         /* Fill in or override the hop limit field, if necessary. */
 668         if (opt && opt->ip6po_hlim != -1)
 669                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 670         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 671                 if (im6o != NULL)
 672                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 673                 else
 674                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 675         }
 676
 677         if (ro == NULL || ro->ro_nh == NULL) {
 678                 bzero(dst, sizeof(*dst));
 679                 dst->sin6_family = AF_INET6;
 680                 dst->sin6_len = sizeof(*dst);
 681                 dst->sin6_addr = ip6->ip6_dst;
 682         }
 683         /*
 684          * Validate route against routing table changes.
 685          * Make sure that the address family is set in route.
 686          */
 687         nh = NULL;
 688         ifp = NULL;
 689         mtu = 0;
 690         if (ro != NULL) {
 691                 if (ro->ro_nh != NULL && inp != NULL) {
 692                         ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
 693                         NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
 694                             fibnum);
 695                 }
 696                 if (ro->ro_nh != NULL && fwd_tag == NULL &&
 697                     (!NH_IS_VALID(ro->ro_nh) ||
 698                     ro->ro_dst.sin6_family != AF_INET6 ||
 699                     !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
 700                         RO_INVALIDATE_CACHE(ro);
 701
 702                 if (ro->ro_nh != NULL && fwd_tag == NULL &&
 703                     ro->ro_dst.sin6_family == AF_INET6 &&
 704                     IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
 705                         nh = ro->ro_nh;
 706                         ifp = nh->nh_ifp;
 707                 } else {
 708                         if (ro->ro_lle)
 709                                 LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 710                         ro->ro_lle = NULL;
 711                         if (fwd_tag == NULL) {
 712                                 bzero(&dst_sa, sizeof(dst_sa));
 713                                 dst_sa.sin6_family = AF_INET6;
 714                                 dst_sa.sin6_len = sizeof(dst_sa);
 715                                 dst_sa.sin6_addr = ip6->ip6_dst;
 716                         }
 717                         error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
 718                             &nh, fibnum, m->m_pkthdr.flowid);
 719                         if (error != 0) {
 720                                 IP6STAT_INC(ip6s_noroute);
 721                                 if (ifp != NULL)
 722                                         in6_ifstat_inc(ifp, ifs6_out_discard);
 723                                 goto bad;
 724                         }
 725                         if (ifp != NULL)
 726                             mtu = ifp->if_mtu;
 727                 }
 728                 if (nh == NULL) {
 729                         /*
 730                          * If in6_selectroute() does not return a nexthop
 731                          * dst may not have been updated.
 732                          */
 733                         *dst = dst_sa;  /* XXX */
 734                 } else {
 735                         if (nh->nh_flags & NHF_HOST)
 736                             mtu = nh->nh_mtu;
 737                         ia = (struct in6_ifaddr *)(nh->nh_ifa);
 738                         counter_u64_add(nh->nh_pksent, 1);
 739                 }
 740         } else {
 741                 struct nhop_object *nh;
 742                 struct in6_addr kdst;
 743                 uint32_t scopeid;
 744
 745                 if (fwd_tag == NULL) {
 746                         bzero(&dst_sa, sizeof(dst_sa));
 747                         dst_sa.sin6_family = AF_INET6;
 748                         dst_sa.sin6_len = sizeof(dst_sa);
 749                         dst_sa.sin6_addr = ip6->ip6_dst;
 750                 }
 751
 752                 if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
 753                     im6o != NULL &&
 754                     (ifp = im6o->im6o_multicast_ifp) != NULL) {
 755                         /* We do not need a route lookup. */
 756                         *dst = dst_sa;  /* XXX */
 757                         goto nonh6lookup;
 758                 }
 759
 760                 in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
 761
 762                 if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
 763                     IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
 764                         if (scopeid > 0) {
 765                                 ifp = in6_getlinkifnet(scopeid);
 766                                 if (ifp == NULL) {
 767                                         error = EHOSTUNREACH;
 768                                         goto bad;
 769                                 }
 770                                 *dst = dst_sa;  /* XXX */
 771                                 goto nonh6lookup;
 772                         }
 773                 }
 774
 775                 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
 776                     m->m_pkthdr.flowid);
 777                 if (nh == NULL) {
 778                         IP6STAT_INC(ip6s_noroute);
 779                         /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
 780                         error = EHOSTUNREACH;;
 781                         goto bad;
 782                 }
 783
 784                 ifp = nh->nh_ifp;
 785                 mtu = nh->nh_mtu;
 786                 ia = ifatoia6(nh->nh_ifa);
 787                 if (nh->nh_flags & NHF_GATEWAY)
 788                         dst->sin6_addr = nh->gw6_sa.sin6_addr;
 789 nonh6lookup:
 790                 ;
 791         }
 792
 793         /* Then nh (for unicast) and ifp must be non-NULL valid values. */
 794         if ((flags & IPV6_FORWARDING) == 0) {
 795                 /* XXX: the FORWARDING flag can be set for mrouting. */
 796                 in6_ifstat_inc(ifp, ifs6_out_request);
 797         }
 798
 799         /* Setup data structures for scope ID checks. */
 800         src0 = ip6->ip6_src;
 801         bzero(&src_sa, sizeof(src_sa));
 802         src_sa.sin6_family = AF_INET6;
 803         src_sa.sin6_len = sizeof(src_sa);
 804         src_sa.sin6_addr = ip6->ip6_src;
 805
 806         dst0 = ip6->ip6_dst;
 807         /* Re-initialize to be sure. */
 808         bzero(&dst_sa, sizeof(dst_sa));
 809         dst_sa.sin6_family = AF_INET6;
 810         dst_sa.sin6_len = sizeof(dst_sa);
 811         dst_sa.sin6_addr = ip6->ip6_dst;
 812
 813         /* Check for valid scope ID. */
 814         if (in6_setscope(&src0, ifp, &zone) == 0 &&
 815             sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
 816             in6_setscope(&dst0, ifp, &zone) == 0 &&
 817             sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
 818                 /*
 819                  * The outgoing interface is in the zone of the source
 820                  * and destination addresses.
 821                  *
 822                  * Because the loopback interface cannot receive
 823                  * packets with a different scope ID than its own,
 824                  * there is a trick to pretend the outgoing packet
 825                  * was received by the real network interface, by
 826                  * setting "origifp" different from "ifp". This is
 827                  * only allowed when "ifp" is a loopback network
 828                  * interface. Refer to code in nd6_output_ifp() for
 829                  * more details.
 830                  */
 831                 origifp = ifp;
 832
 833                 /*
 834                  * We should use ia_ifp to support the case of sending
 835                  * packets to an address of our own.
 836                  */
 837                 if (ia != NULL && ia->ia_ifp)
 838                         ifp = ia->ia_ifp;
 839
 840         } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
 841             sa6_recoverscope(&src_sa) != 0 ||
 842             sa6_recoverscope(&dst_sa) != 0 ||
 843             dst_sa.sin6_scope_id == 0 ||
 844             (src_sa.sin6_scope_id != 0 &&
 845             src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
 846             (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
 847                 /*
 848                  * If the destination network interface is not a
 849                  * loopback interface, or the destination network
 850                  * address has no scope ID, or the source address has
 851                  * a scope ID set which is different from the
 852                  * destination address one, or there is no network
 853                  * interface representing this scope ID, the address
 854                  * pair is considered invalid.
 855                  */
 856                 IP6STAT_INC(ip6s_badscope);
 857                 in6_ifstat_inc(ifp, ifs6_out_discard);
 858                 if (error == 0)
 859                         error = EHOSTUNREACH; /* XXX */
 860                 goto bad;
 861         }
 862         /* All scope ID checks are successful. */
 863
 864         if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 865                 if (opt && opt->ip6po_nextroute.ro_nh) {
 866                         /*
 867                          * The nexthop is explicitly specified by the
 868                          * application.  We assume the next hop is an IPv6
 869                          * address.
 870                          */
 871                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 872                 }
 873                 else if ((nh->nh_flags & NHF_GATEWAY))
 874                         dst = &nh->gw6_sa;
 875         }
 876
 877         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 878                 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
 879         } else {
 880                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 881                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 882
 883                 /* Confirm that the outgoing interface supports multicast. */
 884                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 885                         IP6STAT_INC(ip6s_noroute);
 886                         in6_ifstat_inc(ifp, ifs6_out_discard);
 887                         error = ENETUNREACH;
 888                         goto bad;
 889                 }
 890                 if ((im6o == NULL && in6_mcast_loop) ||
 891                     (im6o && im6o->im6o_multicast_loop)) {
 892                         /*
 893                          * Loop back multicast datagram if not expressly
 894                          * forbidden to do so, even if we have not joined
 895                          * the address; protocols will filter it later,
 896                          * thus deferring a hash lookup and lock acquisition
 897                          * at the expense of an m_copym().
 898                          */
 899                         ip6_mloopback(ifp, m);
 900                 } else {
 901                         /*
 902                          * If we are acting as a multicast router, perform
 903                          * multicast forwarding as if the packet had just
 904                          * arrived on the interface to which we are about
 905                          * to send.  The multicast forwarding function
 906                          * recursively calls this function, using the
 907                          * IPV6_FORWARDING flag to prevent infinite recursion.
 908                          *
 909                          * Multicasts that are looped back by ip6_mloopback(),
 910                          * above, will be forwarded by the ip6_input() routine,
 911                          * if necessary.
 912                          */
 913                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 914                                 /*
 915                                  * XXX: ip6_mforward expects that rcvif is NULL
 916                                  * when it is called from the originating path.
 917                                  * However, it may not always be the case.
 918                                  */
 919                                 m->m_pkthdr.rcvif = NULL;
 920                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 921                                         m_freem(m);
 922                                         goto done;
 923                                 }
 924                         }
 925                 }
 926                 /*
 927                  * Multicasts with a hoplimit of zero may be looped back,
 928                  * above, but must not be transmitted on a network.
 929                  * Also, multicasts addressed to the loopback interface
 930                  * are not sent -- the above call to ip6_mloopback() will
 931                  * loop back a copy if this host actually belongs to the
 932                  * destination group on the loopback interface.
 933                  */
 934                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 935                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 936                         m_freem(m);
 937                         goto done;
 938                 }
 939         }
 940
 941         /*
 942          * Fill the outgoing inteface to tell the upper layer
 943          * to increment per-interface statistics.
 944          */
 945         if (ifpp)
 946                 *ifpp = ifp;
 947
 948         /* Determine path MTU. */
 949         if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
 950                     &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
 951                 goto bad;
 952         KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
 953             "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
 954             ifp, alwaysfrag, fibnum));
 955
 956         /*
 957          * The caller of this function may specify to use the minimum MTU
 958          * in some cases.
 959          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 960          * setting.  The logic is a bit complicated; by default, unicast
 961          * packets will follow path MTU while multicast packets will be sent at
 962          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 963          * including unicast ones will be sent at the minimum MTU.  Multicast
 964          * packets will always be sent at the minimum MTU unless
 965          * IP6PO_MINMTU_DISABLE is explicitly specified.
 966          * See RFC 3542 for more details.
 967          */
 968         if (mtu > IPV6_MMTU) {
 969                 if ((flags & IPV6_MINMTU))
 970                         mtu = IPV6_MMTU;
 971                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 972                         mtu = IPV6_MMTU;
 973                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 974                          (opt == NULL ||
 975                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 976                         mtu = IPV6_MMTU;
 977                 }
 978         }
 979
 980         /*
 981          * Clear embedded scope identifiers if necessary.
 982          * in6_clearscope() will touch the addresses only when necessary.
 983          */
 984         in6_clearscope(&ip6->ip6_src);
 985         in6_clearscope(&ip6->ip6_dst);
 986
 987         /*
 988          * If the outgoing packet contains a hop-by-hop options header,
 989          * it must be examined and processed even by the source node.
 990          * (RFC 2460, section 4.)
 991          */
 992         if (exthdrs.ip6e_hbh) {
 993                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 994                 u_int32_t dummy; /* XXX unused */
 995                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 996
 997 #ifdef DIAGNOSTIC
 998                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 999                         panic("ip6e_hbh is not contiguous");
1000 #endif
1001                 /*
1002                  *  XXX: if we have to send an ICMPv6 error to the sender,
1003                  *       we need the M_LOOP flag since icmp6_error() expects
1004                  *       the IPv6 and the hop-by-hop options header are
1005                  *       contiguous unless the flag is set.
1006                  */
1007                 m->m_flags |= M_LOOP;
1008                 m->m_pkthdr.rcvif = ifp;
1009                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1010                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1011                     &dummy, &plen) < 0) {
1012                         /* m was already freed at this point. */
1013                         error = EINVAL;/* better error? */
1014                         goto done;
1015                 }
1016                 m->m_flags &= ~M_LOOP; /* XXX */
1017                 m->m_pkthdr.rcvif = NULL;
1018         }
1019
1020         /* Jump over all PFIL processing if hooks are not active. */
1021         if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1022                 goto passout;
1023
1024         odst = ip6->ip6_dst;
1025         /* Run through list of hooks for output packets. */
1026         switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
1027         case PFIL_PASS:
1028                 ip6 = mtod(m, struct ip6_hdr *);
1029                 break;
1030         case PFIL_DROPPED:
1031                 error = EACCES;
1032                 /* FALLTHROUGH */
1033         case PFIL_CONSUMED:
1034                 goto done;
1035         }
1036
1037         needfiblookup = 0;
1038         /* See if destination IP address was changed by packet filter. */
1039         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1040                 m->m_flags |= M_SKIP_FIREWALL;
1041                 /* If destination is now ourself drop to ip6_input(). */
1042                 if (in6_localip(&ip6->ip6_dst)) {
1043                         m->m_flags |= M_FASTFWD_OURS;
1044                         if (m->m_pkthdr.rcvif == NULL)
1045                                 m->m_pkthdr.rcvif = V_loif;
1046                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1047                                 m->m_pkthdr.csum_flags |=
1048                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1049                                 m->m_pkthdr.csum_data = 0xffff;
1050                         }
1051 #if defined(SCTP) || defined(SCTP_SUPPORT)
1052                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1053                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1054 #endif
1055                         error = netisr_queue(NETISR_IPV6, m);
1056                         goto done;
1057                 } else {
1058                         if (ro != NULL)
1059                                 RO_INVALIDATE_CACHE(ro);
1060                         needfiblookup = 1; /* Redo the routing table lookup. */
1061                 }
1062         }
1063         /* See if fib was changed by packet filter. */
1064         if (fibnum != M_GETFIB(m)) {
1065                 m->m_flags |= M_SKIP_FIREWALL;
1066                 fibnum = M_GETFIB(m);
1067                 if (ro != NULL)
1068                         RO_INVALIDATE_CACHE(ro);
1069                 needfiblookup = 1;
1070         }
1071         if (needfiblookup)
1072                 goto again;
1073
1074         /* See if local, if yes, send it to netisr. */
1075         if (m->m_flags & M_FASTFWD_OURS) {
1076                 if (m->m_pkthdr.rcvif == NULL)
1077                         m->m_pkthdr.rcvif = V_loif;
1078                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1079                         m->m_pkthdr.csum_flags |=
1080                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1081                         m->m_pkthdr.csum_data = 0xffff;
1082                 }
1083 #if defined(SCTP) || defined(SCTP_SUPPORT)
1084                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1085                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1086 #endif
1087                 error = netisr_queue(NETISR_IPV6, m);
1088                 goto done;
1089         }
1090         /* Or forward to some other address? */
1091         if ((m->m_flags & M_IP6_NEXTHOP) &&
1092             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1093                 if (ro != NULL)
1094                         dst = (struct sockaddr_in6 *)&ro->ro_dst;
1095                 else
1096                         dst = &sin6;
1097                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1098                 m->m_flags |= M_SKIP_FIREWALL;
1099                 m->m_flags &= ~M_IP6_NEXTHOP;
1100                 m_tag_delete(m, fwd_tag);
1101                 goto again;
1102         }
1103
1104 passout:
1105         if (vlan_pcp > -1)
1106                 EVL_APPLY_PRI(m, vlan_pcp);
1107         /*
1108          * Send the packet to the outgoing interface.
1109          * If necessary, do IPv6 fragmentation before sending.
1110          *
1111          * The logic here is rather complex:
1112          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1113          * 1-a: send as is if tlen <= path mtu
1114          * 1-b: fragment if tlen > path mtu
1115          *
1116          * 2: if user asks us not to fragment (dontfrag == 1)
1117          * 2-a: send as is if tlen <= interface mtu
1118          * 2-b: error if tlen > interface mtu
1119          *
1120          * 3: if we always need to attach fragment header (alwaysfrag == 1)
1121          *      always fragment
1122          *
1123          * 4: if dontfrag == 1 && alwaysfrag == 1
1124          *      error, as we cannot handle this conflicting request.
1125          */
1126         sw_csum = m->m_pkthdr.csum_flags;
1127         if (!hdrsplit) {
1128                 tso = ((sw_csum & ifp->if_hwassist &
1129                     (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1130                 sw_csum &= ~ifp->if_hwassist;
1131         } else
1132                 tso = 0;
1133         /*
1134          * If we added extension headers, we will not do TSO and calculate the
1135          * checksums ourselves for now.
1136          * XXX-BZ  Need a framework to know when the NIC can handle it, even
1137          * with ext. hdrs.
1138          */
1139         error = ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen, false);
1140         if (error != 0)
1141                 goto bad;
1142         /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1143         tlen = m->m_pkthdr.len;
1144
1145         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1146                 dontfrag = 1;
1147         else
1148                 dontfrag = 0;
1149         if (dontfrag && alwaysfrag) {   /* Case 4. */
1150                 /* Conflicting request - can't transmit. */
1151                 error = EMSGSIZE;
1152                 goto bad;
1153         }
1154         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* Case 2-b. */
1155                 /*
1156                  * Even if the DONTFRAG option is specified, we cannot send the
1157                  * packet when the data length is larger than the MTU of the
1158                  * outgoing interface.
1159                  * Notify the error by sending IPV6_PATHMTU ancillary data if
1160                  * application wanted to know the MTU value. Also return an
1161                  * error code (this is not described in the API spec).
1162                  */
1163                 if (inp != NULL)
1164                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1165                 error = EMSGSIZE;
1166                 goto bad;
1167         }
1168
1169         /* Transmit packet without fragmentation. */
1170         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */
1171                 struct in6_ifaddr *ia6;
1172
1173                 ip6 = mtod(m, struct ip6_hdr *);
1174                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1175                 if (ia6) {
1176                         /* Record statistics for this interface address. */
1177                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1178                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
1179                             m->m_pkthdr.len);
1180                 }
1181                 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1182                     (flags & IP_NO_SND_TAG_RL) ? false : true);
1183                 goto done;
1184         }
1185
1186         /* Try to fragment the packet.  Cases 1-b and 3. */
1187         if (mtu < IPV6_MMTU) {
1188                 /* Path MTU cannot be less than IPV6_MMTU. */
1189                 error = EMSGSIZE;
1190                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1191                 goto bad;
1192         } else if (ip6->ip6_plen == 0) {
1193                 /* Jumbo payload cannot be fragmented. */
1194                 error = EMSGSIZE;
1195                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1196                 goto bad;
1197         } else {
1198                 u_char nextproto;
1199
1200                 /*
1201                  * Too large for the destination or interface;
1202                  * fragment if possible.
1203                  * Must be able to put at least 8 bytes per fragment.
1204                  */
1205                 if (mtu > IPV6_MAXPACKET)
1206                         mtu = IPV6_MAXPACKET;
1207
1208                 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1209                 if (len < 8) {
1210                         error = EMSGSIZE;
1211                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1212                         goto bad;
1213                 }
1214
1215                 /*
1216                  * If the interface will not calculate checksums on
1217                  * fragmented packets, then do it here.
1218                  * XXX-BZ handle the hw offloading case.  Need flags.
1219                  */
1220                 error = ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags,
1221                     plen, optlen, true);
1222                 if (error != 0)
1223                         goto bad;
1224
1225                 /*
1226                  * Change the next header field of the last header in the
1227                  * unfragmentable part.
1228                  */
1229                 if (exthdrs.ip6e_rthdr) {
1230                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1231                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1232                 } else if (exthdrs.ip6e_dest1) {
1233                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1234                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1235                 } else if (exthdrs.ip6e_hbh) {
1236                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1237                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1238                 } else {
1239                         ip6 = mtod(m, struct ip6_hdr *);
1240                         nextproto = ip6->ip6_nxt;
1241                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1242                 }
1243
1244                 /*
1245                  * Loop through length of segment after first fragment,
1246                  * make new header and copy data of each part and link onto
1247                  * chain.
1248                  */
1249                 m0 = m;
1250                 id = htonl(ip6_randomid());
1251                 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1252                 if (error != 0)
1253                         goto sendorfree;
1254
1255                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1256         }
1257
1258         /* Remove leading garbage. */
1259 sendorfree:
1260         m = m0->m_nextpkt;
1261         m0->m_nextpkt = 0;
1262         m_freem(m0);
1263         for (; m; m = m0) {
1264                 m0 = m->m_nextpkt;
1265                 m->m_nextpkt = 0;
1266                 if (error == 0) {
1267                         /* Record statistics for this interface address. */
1268                         if (ia) {
1269                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1270                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1271                                     m->m_pkthdr.len);
1272                         }
1273                         if (vlan_pcp > -1)
1274                                 EVL_APPLY_PRI(m, vlan_pcp);
1275                         error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1276                             true);
1277                 } else
1278                         m_freem(m);
1279         }
1280
1281         if (error == 0)
1282                 IP6STAT_INC(ip6s_fragmented);
1283
1284 done:
1285         return (error);
1286
1287 freehdrs:
1288         m_freem(exthdrs.ip6e_hbh);      /* m_freem() checks if mbuf is NULL. */
1289         m_freem(exthdrs.ip6e_dest1);
1290         m_freem(exthdrs.ip6e_rthdr);
1291         m_freem(exthdrs.ip6e_dest2);
1292         /* FALLTHROUGH */
1293 bad:
1294         if (m)
1295                 m_freem(m);
1296         goto done;
1297 }
1298
1299 static int
1300 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1301 {
1302         struct mbuf *m;
1303
1304         if (hlen > MCLBYTES)
1305                 return (ENOBUFS); /* XXX */
1306
1307         if (hlen > MLEN)
1308                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1309         else
1310                 m = m_get(M_NOWAIT, MT_DATA);
1311         if (m == NULL)
1312                 return (ENOBUFS);
1313         m->m_len = hlen;
1314         if (hdr)
1315                 bcopy(hdr, mtod(m, caddr_t), hlen);
1316
1317         *mp = m;
1318         return (0);
1319 }
1320
1321 /*
1322  * Insert jumbo payload option.
1323  */
1324 static int
1325 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1326 {
1327         struct mbuf *mopt;
1328         u_char *optbuf;
1329         u_int32_t v;
1330
1331 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1332
1333         /*
1334          * If there is no hop-by-hop options header, allocate new one.
1335          * If there is one but it doesn't have enough space to store the
1336          * jumbo payload option, allocate a cluster to store the whole options.
1337          * Otherwise, use it to store the options.
1338          */
1339         if (exthdrs->ip6e_hbh == NULL) {
1340                 mopt = m_get(M_NOWAIT, MT_DATA);
1341                 if (mopt == NULL)
1342                         return (ENOBUFS);
1343                 mopt->m_len = JUMBOOPTLEN;
1344                 optbuf = mtod(mopt, u_char *);
1345                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1346                 exthdrs->ip6e_hbh = mopt;
1347         } else {
1348                 struct ip6_hbh *hbh;
1349
1350                 mopt = exthdrs->ip6e_hbh;
1351                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1352                         /*
1353                          * XXX assumption:
1354                          * - exthdrs->ip6e_hbh is not referenced from places
1355                          *   other than exthdrs.
1356                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1357                          */
1358                         int oldoptlen = mopt->m_len;
1359                         struct mbuf *n;
1360
1361                         /*
1362                          * XXX: give up if the whole (new) hbh header does
1363                          * not fit even in an mbuf cluster.
1364                          */
1365                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1366                                 return (ENOBUFS);
1367
1368                         /*
1369                          * As a consequence, we must always prepare a cluster
1370                          * at this point.
1371                          */
1372                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1373                         if (n == NULL)
1374                                 return (ENOBUFS);
1375                         n->m_len = oldoptlen + JUMBOOPTLEN;
1376                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1377                             oldoptlen);
1378                         optbuf = mtod(n, caddr_t) + oldoptlen;
1379                         m_freem(mopt);
1380                         mopt = exthdrs->ip6e_hbh = n;
1381                 } else {
1382                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1383                         mopt->m_len += JUMBOOPTLEN;
1384                 }
1385                 optbuf[0] = IP6OPT_PADN;
1386                 optbuf[1] = 1;
1387
1388                 /*
1389                  * Adjust the header length according to the pad and
1390                  * the jumbo payload option.
1391                  */
1392                 hbh = mtod(mopt, struct ip6_hbh *);
1393                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1394         }
1395
1396         /* fill in the option. */
1397         optbuf[2] = IP6OPT_JUMBO;
1398         optbuf[3] = 4;
1399         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1400         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1401
1402         /* finally, adjust the packet header length */
1403         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1404
1405         return (0);
1406 #undef JUMBOOPTLEN
1407 }
1408
1409 /*
1410  * Insert fragment header and copy unfragmentable header portions.
1411  */
1412 static int
1413 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1414     struct ip6_frag **frghdrp)
1415 {
1416         struct mbuf *n, *mlast;
1417
1418         if (hlen > sizeof(struct ip6_hdr)) {
1419                 n = m_copym(m0, sizeof(struct ip6_hdr),
1420                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1421                 if (n == NULL)
1422                         return (ENOBUFS);
1423                 m->m_next = n;
1424         } else
1425                 n = m;
1426
1427         /* Search for the last mbuf of unfragmentable part. */
1428         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1429                 ;
1430
1431         if (M_WRITABLE(mlast) &&
1432             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1433                 /* use the trailing space of the last mbuf for the fragment hdr */
1434                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1435                     mlast->m_len);
1436                 mlast->m_len += sizeof(struct ip6_frag);
1437                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1438         } else {
1439                 /* allocate a new mbuf for the fragment header */
1440                 struct mbuf *mfrg;
1441
1442                 mfrg = m_get(M_NOWAIT, MT_DATA);
1443                 if (mfrg == NULL)
1444                         return (ENOBUFS);
1445                 mfrg->m_len = sizeof(struct ip6_frag);
1446                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1447                 mlast->m_next = mfrg;
1448         }
1449
1450         return (0);
1451 }
1452
1453 /*
1454  * Calculates IPv6 path mtu for destination @dst.
1455  * Resulting MTU is stored in @mtup.
1456  *
1457  * Returns 0 on success.
1458  */
1459 static int
1460 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1461 {
1462         struct epoch_tracker et;
1463         struct nhop_object *nh;
1464         struct in6_addr kdst;
1465         uint32_t scopeid;
1466         int error;
1467
1468         in6_splitscope(dst, &kdst, &scopeid);
1469
1470         NET_EPOCH_ENTER(et);
1471         nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1472         if (nh != NULL)
1473                 error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1474         else
1475                 error = EHOSTUNREACH;
1476         NET_EPOCH_EXIT(et);
1477
1478         return (error);
1479 }
1480
1481 /*
1482  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1483  * and cached data in @ro_pmtu.
1484  * MTU from (successful) route lookup is saved (along with dst)
1485  * inside @ro_pmtu to avoid subsequent route lookups after packet
1486  * filter processing.
1487  *
1488  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1489  * Returns 0 on success.
1490  */
1491 static int
1492 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1493     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1494     int *alwaysfragp, u_int fibnum, u_int proto)
1495 {
1496         struct nhop_object *nh;
1497         struct in6_addr kdst;
1498         uint32_t scopeid;
1499         struct sockaddr_in6 *sa6_dst, sin6;
1500         u_long mtu;
1501
1502         NET_EPOCH_ASSERT();
1503
1504         mtu = 0;
1505         if (ro_pmtu == NULL || do_lookup) {
1506                 /*
1507                  * Here ro_pmtu has final destination address, while
1508                  * ro might represent immediate destination.
1509                  * Use ro_pmtu destination since mtu might differ.
1510                  */
1511                 if (ro_pmtu != NULL) {
1512                         sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1513                         if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1514                                 ro_pmtu->ro_mtu = 0;
1515                 } else
1516                         sa6_dst = &sin6;
1517
1518                 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1519                         bzero(sa6_dst, sizeof(*sa6_dst));
1520                         sa6_dst->sin6_family = AF_INET6;
1521                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1522                         sa6_dst->sin6_addr = *dst;
1523
1524                         in6_splitscope(dst, &kdst, &scopeid);
1525                         nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1526                         if (nh != NULL) {
1527                                 mtu = nh->nh_mtu;
1528                                 if (ro_pmtu != NULL)
1529                                         ro_pmtu->ro_mtu = mtu;
1530                         }
1531                 } else
1532                         mtu = ro_pmtu->ro_mtu;
1533         }
1534
1535         if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1536                 mtu = ro_pmtu->ro_nh->nh_mtu;
1537
1538         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1539 }
1540
1541 /*
1542  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1543  * hostcache data for @dst.
1544  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1545  *
1546  * Returns 0 on success.
1547  */
1548 static int
1549 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1550     u_long *mtup, int *alwaysfragp, u_int proto)
1551 {
1552         u_long mtu = 0;
1553         int alwaysfrag = 0;
1554         int error = 0;
1555
1556         if (rt_mtu > 0) {
1557                 u_int32_t ifmtu;
1558                 struct in_conninfo inc;
1559
1560                 bzero(&inc, sizeof(inc));
1561                 inc.inc_flags |= INC_ISIPV6;
1562                 inc.inc6_faddr = *dst;
1563
1564                 ifmtu = IN6_LINKMTU(ifp);
1565
1566                 /* TCP is known to react to pmtu changes so skip hc */
1567                 if (proto != IPPROTO_TCP)
1568                         mtu = tcp_hc_getmtu(&inc);
1569
1570                 if (mtu)
1571                         mtu = min(mtu, rt_mtu);
1572                 else
1573                         mtu = rt_mtu;
1574                 if (mtu == 0)
1575                         mtu = ifmtu;
1576                 else if (mtu < IPV6_MMTU) {
1577                         /*
1578                          * RFC2460 section 5, last paragraph:
1579                          * if we record ICMPv6 too big message with
1580                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1581                          * or smaller, with framgent header attached.
1582                          * (fragment header is needed regardless from the
1583                          * packet size, for translators to identify packets)
1584                          */
1585                         alwaysfrag = 1;
1586                         mtu = IPV6_MMTU;
1587                 }
1588         } else if (ifp) {
1589                 mtu = IN6_LINKMTU(ifp);
1590         } else
1591                 error = EHOSTUNREACH; /* XXX */
1592
1593         *mtup = mtu;
1594         if (alwaysfragp)
1595                 *alwaysfragp = alwaysfrag;
1596         return (error);
1597 }
1598
1599 /*
1600  * IP6 socket option processing.
1601  */
1602 int
1603 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1604 {
1605         int optdatalen, uproto;
1606         void *optdata;
1607         struct inpcb *inp = sotoinpcb(so);
1608         int error, optval;
1609         int level, op, optname;
1610         int optlen;
1611         struct thread *td;
1612 #ifdef  RSS
1613         uint32_t rss_bucket;
1614         int retval;
1615 #endif
1616
1617 /*
1618  * Don't use more than a quarter of mbuf clusters.  N.B.:
1619  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1620  * on LP64 architectures, so cast to u_long to avoid undefined
1621  * behavior.  ILP32 architectures cannot have nmbclusters
1622  * large enough to overflow for other reasons.
1623  */
1624 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1625
1626         level = sopt->sopt_level;
1627         op = sopt->sopt_dir;
1628         optname = sopt->sopt_name;
1629         optlen = sopt->sopt_valsize;
1630         td = sopt->sopt_td;
1631         error = 0;
1632         optval = 0;
1633         uproto = (int)so->so_proto->pr_protocol;
1634
1635         if (level != IPPROTO_IPV6) {
1636                 error = EINVAL;
1637
1638                 if (sopt->sopt_level == SOL_SOCKET &&
1639                     sopt->sopt_dir == SOPT_SET) {
1640                         switch (sopt->sopt_name) {
1641                         case SO_REUSEADDR:
1642                                 INP_WLOCK(inp);
1643                                 if ((so->so_options & SO_REUSEADDR) != 0)
1644                                         inp->inp_flags2 |= INP_REUSEADDR;
1645                                 else
1646                                         inp->inp_flags2 &= ~INP_REUSEADDR;
1647                                 INP_WUNLOCK(inp);
1648                                 error = 0;
1649                                 break;
1650                         case SO_REUSEPORT:
1651                                 INP_WLOCK(inp);
1652                                 if ((so->so_options & SO_REUSEPORT) != 0)
1653                                         inp->inp_flags2 |= INP_REUSEPORT;
1654                                 else
1655                                         inp->inp_flags2 &= ~INP_REUSEPORT;
1656                                 INP_WUNLOCK(inp);
1657                                 error = 0;
1658                                 break;
1659                         case SO_REUSEPORT_LB:
1660                                 INP_WLOCK(inp);
1661                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1662                                         inp->inp_flags2 |= INP_REUSEPORT_LB;
1663                                 else
1664                                         inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1665                                 INP_WUNLOCK(inp);
1666                                 error = 0;
1667                                 break;
1668                         case SO_SETFIB:
1669                                 INP_WLOCK(inp);
1670                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
1671                                 INP_WUNLOCK(inp);
1672                                 error = 0;
1673                                 break;
1674                         case SO_MAX_PACING_RATE:
1675 #ifdef RATELIMIT
1676                                 INP_WLOCK(inp);
1677                                 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1678                                 INP_WUNLOCK(inp);
1679                                 error = 0;
1680 #else
1681                                 error = EOPNOTSUPP;
1682 #endif
1683                                 break;
1684                         default:
1685                                 break;
1686                         }
1687                 }
1688         } else {                /* level == IPPROTO_IPV6 */
1689                 switch (op) {
1690                 case SOPT_SET:
1691                         switch (optname) {
1692                         case IPV6_2292PKTOPTIONS:
1693 #ifdef IPV6_PKTOPTIONS
1694                         case IPV6_PKTOPTIONS:
1695 #endif
1696                         {
1697                                 struct mbuf *m;
1698
1699                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1700                                         printf("ip6_ctloutput: mbuf limit hit\n");
1701                                         error = ENOBUFS;
1702                                         break;
1703                                 }
1704
1705                                 error = soopt_getm(sopt, &m); /* XXX */
1706                                 if (error != 0)
1707                                         break;
1708                                 error = soopt_mcopyin(sopt, m); /* XXX */
1709                                 if (error != 0)
1710                                         break;
1711                                 INP_WLOCK(inp);
1712                                 error = ip6_pcbopts(&inp->in6p_outputopts, m,
1713                                     so, sopt);
1714                                 INP_WUNLOCK(inp);
1715                                 m_freem(m); /* XXX */
1716                                 break;
1717                         }
1718
1719                         /*
1720                          * Use of some Hop-by-Hop options or some
1721                          * Destination options, might require special
1722                          * privilege.  That is, normal applications
1723                          * (without special privilege) might be forbidden
1724                          * from setting certain options in outgoing packets,
1725                          * and might never see certain options in received
1726                          * packets. [RFC 2292 Section 6]
1727                          * KAME specific note:
1728                          *  KAME prevents non-privileged users from sending or
1729                          *  receiving ANY hbh/dst options in order to avoid
1730                          *  overhead of parsing options in the kernel.
1731                          */
1732                         case IPV6_RECVHOPOPTS:
1733                         case IPV6_RECVDSTOPTS:
1734                         case IPV6_RECVRTHDRDSTOPTS:
1735                                 if (td != NULL) {
1736                                         error = priv_check(td,
1737                                             PRIV_NETINET_SETHDROPTS);
1738                                         if (error)
1739                                                 break;
1740                                 }
1741                                 /* FALLTHROUGH */
1742                         case IPV6_UNICAST_HOPS:
1743                         case IPV6_HOPLIMIT:
1744
1745                         case IPV6_RECVPKTINFO:
1746                         case IPV6_RECVHOPLIMIT:
1747                         case IPV6_RECVRTHDR:
1748                         case IPV6_RECVPATHMTU:
1749                         case IPV6_RECVTCLASS:
1750                         case IPV6_RECVFLOWID:
1751 #ifdef  RSS
1752                         case IPV6_RECVRSSBUCKETID:
1753 #endif
1754                         case IPV6_V6ONLY:
1755                         case IPV6_AUTOFLOWLABEL:
1756                         case IPV6_ORIGDSTADDR:
1757                         case IPV6_BINDANY:
1758                         case IPV6_BINDMULTI:
1759 #ifdef  RSS
1760                         case IPV6_RSS_LISTEN_BUCKET:
1761 #endif
1762                         case IPV6_VLAN_PCP:
1763                                 if (optname == IPV6_BINDANY && td != NULL) {
1764                                         error = priv_check(td,
1765                                             PRIV_NETINET_BINDANY);
1766                                         if (error)
1767                                                 break;
1768                                 }
1769
1770                                 if (optlen != sizeof(int)) {
1771                                         error = EINVAL;
1772                                         break;
1773                                 }
1774                                 error = sooptcopyin(sopt, &optval,
1775                                         sizeof optval, sizeof optval);
1776                                 if (error)
1777                                         break;
1778                                 switch (optname) {
1779                                 case IPV6_UNICAST_HOPS:
1780                                         if (optval < -1 || optval >= 256)
1781                                                 error = EINVAL;
1782                                         else {
1783                                                 /* -1 = kernel default */
1784                                                 inp->in6p_hops = optval;
1785                                                 if ((inp->inp_vflag &
1786                                                      INP_IPV4) != 0)
1787                                                         inp->inp_ip_ttl = optval;
1788                                         }
1789                                         break;
1790 #define OPTSET(bit) \
1791 do { \
1792         INP_WLOCK(inp); \
1793         if (optval) \
1794                 inp->inp_flags |= (bit); \
1795         else \
1796                 inp->inp_flags &= ~(bit); \
1797         INP_WUNLOCK(inp); \
1798 } while (/*CONSTCOND*/ 0)
1799 #define OPTSET2292(bit) \
1800 do { \
1801         INP_WLOCK(inp); \
1802         inp->inp_flags |= IN6P_RFC2292; \
1803         if (optval) \
1804                 inp->inp_flags |= (bit); \
1805         else \
1806                 inp->inp_flags &= ~(bit); \
1807         INP_WUNLOCK(inp); \
1808 } while (/*CONSTCOND*/ 0)
1809 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1810
1811 #define OPTSET2_N(bit, val) do {                                        \
1812         if (val)                                                        \
1813                 inp->inp_flags2 |= bit;                                 \
1814         else                                                            \
1815                 inp->inp_flags2 &= ~bit;                                \
1816 } while (0)
1817 #define OPTSET2(bit, val) do {                                          \
1818         INP_WLOCK(inp);                                                 \
1819         OPTSET2_N(bit, val);                                            \
1820         INP_WUNLOCK(inp);                                               \
1821 } while (0)
1822 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1823 #define OPTSET2292_EXCLUSIVE(bit)                                       \
1824 do {                                                                    \
1825         INP_WLOCK(inp);                                                 \
1826         if (OPTBIT(IN6P_RFC2292)) {                                     \
1827                 error = EINVAL;                                         \
1828         } else {                                                        \
1829                 if (optval)                                             \
1830                         inp->inp_flags |= (bit);                        \
1831                 else                                                    \
1832                         inp->inp_flags &= ~(bit);                       \
1833         }                                                               \
1834         INP_WUNLOCK(inp);                                               \
1835 } while (/*CONSTCOND*/ 0)
1836
1837                                 case IPV6_RECVPKTINFO:
1838                                         OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1839                                         break;
1840
1841                                 case IPV6_HOPLIMIT:
1842                                 {
1843                                         struct ip6_pktopts **optp;
1844
1845                                         /* cannot mix with RFC2292 */
1846                                         if (OPTBIT(IN6P_RFC2292)) {
1847                                                 error = EINVAL;
1848                                                 break;
1849                                         }
1850                                         INP_WLOCK(inp);
1851                                         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1852                                                 INP_WUNLOCK(inp);
1853                                                 return (ECONNRESET);
1854                                         }
1855                                         optp = &inp->in6p_outputopts;
1856                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1857                                             (u_char *)&optval, sizeof(optval),
1858                                             optp, (td != NULL) ? td->td_ucred :
1859                                             NULL, uproto);
1860                                         INP_WUNLOCK(inp);
1861                                         break;
1862                                 }
1863
1864                                 case IPV6_RECVHOPLIMIT:
1865                                         OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1866                                         break;
1867
1868                                 case IPV6_RECVHOPOPTS:
1869                                         OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1870                                         break;
1871
1872                                 case IPV6_RECVDSTOPTS:
1873                                         OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1874                                         break;
1875
1876                                 case IPV6_RECVRTHDRDSTOPTS:
1877                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1878                                         break;
1879
1880                                 case IPV6_RECVRTHDR:
1881                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1882                                         break;
1883
1884                                 case IPV6_RECVPATHMTU:
1885                                         /*
1886                                          * We ignore this option for TCP
1887                                          * sockets.
1888                                          * (RFC3542 leaves this case
1889                                          * unspecified.)
1890                                          */
1891                                         if (uproto != IPPROTO_TCP)
1892                                                 OPTSET(IN6P_MTU);
1893                                         break;
1894
1895                                 case IPV6_RECVFLOWID:
1896                                         OPTSET2(INP_RECVFLOWID, optval);
1897                                         break;
1898
1899 #ifdef  RSS
1900                                 case IPV6_RECVRSSBUCKETID:
1901                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1902                                         break;
1903 #endif
1904
1905                                 case IPV6_V6ONLY:
1906                                         INP_WLOCK(inp);
1907                                         if (inp->inp_lport ||
1908                                             !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1909                                                 /*
1910                                                  * The socket is already bound.
1911                                                  */
1912                                                 INP_WUNLOCK(inp);
1913                                                 error = EINVAL;
1914                                                 break;
1915                                         }
1916                                         if (optval) {
1917                                                 inp->inp_flags |= IN6P_IPV6_V6ONLY;
1918                                                 inp->inp_vflag &= ~INP_IPV4;
1919                                         } else {
1920                                                 inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1921                                                 inp->inp_vflag |= INP_IPV4;
1922                                         }
1923                                         INP_WUNLOCK(inp);
1924                                         break;
1925                                 case IPV6_RECVTCLASS:
1926                                         /* cannot mix with RFC2292 XXX */
1927                                         OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1928                                         break;
1929                                 case IPV6_AUTOFLOWLABEL:
1930                                         OPTSET(IN6P_AUTOFLOWLABEL);
1931                                         break;
1932
1933                                 case IPV6_ORIGDSTADDR:
1934                                         OPTSET2(INP_ORIGDSTADDR, optval);
1935                                         break;
1936                                 case IPV6_BINDANY:
1937                                         OPTSET(INP_BINDANY);
1938                                         break;
1939
1940                                 case IPV6_BINDMULTI:
1941                                         OPTSET2(INP_BINDMULTI, optval);
1942                                         break;
1943 #ifdef  RSS
1944                                 case IPV6_RSS_LISTEN_BUCKET:
1945                                         if ((optval >= 0) &&
1946                                             (optval < rss_getnumbuckets())) {
1947                                                 INP_WLOCK(inp);
1948                                                 inp->inp_rss_listen_bucket = optval;
1949                                                 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1950                                                 INP_WUNLOCK(inp);
1951                                         } else {
1952                                                 error = EINVAL;
1953                                         }
1954                                         break;
1955 #endif
1956                                 case IPV6_VLAN_PCP:
1957                                         if ((optval >= -1) && (optval <=
1958                                             (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1959                                                 if (optval == -1) {
1960                                                         INP_WLOCK(inp);
1961                                                         inp->inp_flags2 &=
1962                                                             ~(INP_2PCP_SET |
1963                                                             INP_2PCP_MASK);
1964                                                         INP_WUNLOCK(inp);
1965                                                 } else {
1966                                                         INP_WLOCK(inp);
1967                                                         inp->inp_flags2 |=
1968                                                             INP_2PCP_SET;
1969                                                         inp->inp_flags2 &=
1970                                                             ~INP_2PCP_MASK;
1971                                                         inp->inp_flags2 |=
1972                                                             optval <<
1973                                                             INP_2PCP_SHIFT;
1974                                                         INP_WUNLOCK(inp);
1975                                                 }
1976                                         } else
1977                                                 error = EINVAL;
1978                                         break;
1979                                 }
1980                                 break;
1981
1982                         case IPV6_TCLASS:
1983                         case IPV6_DONTFRAG:
1984                         case IPV6_USE_MIN_MTU:
1985                         case IPV6_PREFER_TEMPADDR:
1986                                 if (optlen != sizeof(optval)) {
1987                                         error = EINVAL;
1988                                         break;
1989                                 }
1990                                 error = sooptcopyin(sopt, &optval,
1991                                         sizeof optval, sizeof optval);
1992                                 if (error)
1993                                         break;
1994                                 {
1995                                         struct ip6_pktopts **optp;
1996                                         INP_WLOCK(inp);
1997                                         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1998                                                 INP_WUNLOCK(inp);
1999                                                 return (ECONNRESET);
2000                                         }
2001                                         optp = &inp->in6p_outputopts;
2002                                         error = ip6_pcbopt(optname,
2003                                             (u_char *)&optval, sizeof(optval),
2004                                             optp, (td != NULL) ? td->td_ucred :
2005                                             NULL, uproto);
2006                                         INP_WUNLOCK(inp);
2007                                         break;
2008                                 }
2009
2010                         case IPV6_2292PKTINFO:
2011                         case IPV6_2292HOPLIMIT:
2012                         case IPV6_2292HOPOPTS:
2013                         case IPV6_2292DSTOPTS:
2014                         case IPV6_2292RTHDR:
2015                                 /* RFC 2292 */
2016                                 if (optlen != sizeof(int)) {
2017                                         error = EINVAL;
2018                                         break;
2019                                 }
2020                                 error = sooptcopyin(sopt, &optval,
2021                                         sizeof optval, sizeof optval);
2022                                 if (error)
2023                                         break;
2024                                 switch (optname) {
2025                                 case IPV6_2292PKTINFO:
2026                                         OPTSET2292(IN6P_PKTINFO);
2027                                         break;
2028                                 case IPV6_2292HOPLIMIT:
2029                                         OPTSET2292(IN6P_HOPLIMIT);
2030                                         break;
2031                                 case IPV6_2292HOPOPTS:
2032                                         /*
2033                                          * Check super-user privilege.
2034                                          * See comments for IPV6_RECVHOPOPTS.
2035                                          */
2036                                         if (td != NULL) {
2037                                                 error = priv_check(td,
2038                                                     PRIV_NETINET_SETHDROPTS);
2039                                                 if (error)
2040                                                         return (error);
2041                                         }
2042                                         OPTSET2292(IN6P_HOPOPTS);
2043                                         break;
2044                                 case IPV6_2292DSTOPTS:
2045                                         if (td != NULL) {
2046                                                 error = priv_check(td,
2047                                                     PRIV_NETINET_SETHDROPTS);
2048                                                 if (error)
2049                                                         return (error);
2050                                         }
2051                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2052                                         break;
2053                                 case IPV6_2292RTHDR:
2054                                         OPTSET2292(IN6P_RTHDR);
2055                                         break;
2056                                 }
2057                                 break;
2058                         case IPV6_PKTINFO:
2059                         case IPV6_HOPOPTS:
2060                         case IPV6_RTHDR:
2061                         case IPV6_DSTOPTS:
2062                         case IPV6_RTHDRDSTOPTS:
2063                         case IPV6_NEXTHOP:
2064                         {
2065                                 /* new advanced API (RFC3542) */
2066                                 u_char *optbuf;
2067                                 u_char optbuf_storage[MCLBYTES];
2068                                 int optlen;
2069                                 struct ip6_pktopts **optp;
2070
2071                                 /* cannot mix with RFC2292 */
2072                                 if (OPTBIT(IN6P_RFC2292)) {
2073                                         error = EINVAL;
2074                                         break;
2075                                 }
2076
2077                                 /*
2078                                  * We only ensure valsize is not too large
2079                                  * here.  Further validation will be done
2080                                  * later.
2081                                  */
2082                                 error = sooptcopyin(sopt, optbuf_storage,
2083                                     sizeof(optbuf_storage), 0);
2084                                 if (error)
2085                                         break;
2086                                 optlen = sopt->sopt_valsize;
2087                                 optbuf = optbuf_storage;
2088                                 INP_WLOCK(inp);
2089                                 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2090                                         INP_WUNLOCK(inp);
2091                                         return (ECONNRESET);
2092                                 }
2093                                 optp = &inp->in6p_outputopts;
2094                                 error = ip6_pcbopt(optname, optbuf, optlen,
2095                                     optp, (td != NULL) ? td->td_ucred : NULL,
2096                                     uproto);
2097                                 INP_WUNLOCK(inp);
2098                                 break;
2099                         }
2100 #undef OPTSET
2101
2102                         case IPV6_MULTICAST_IF:
2103                         case IPV6_MULTICAST_HOPS:
2104                         case IPV6_MULTICAST_LOOP:
2105                         case IPV6_JOIN_GROUP:
2106                         case IPV6_LEAVE_GROUP:
2107                         case IPV6_MSFILTER:
2108                         case MCAST_BLOCK_SOURCE:
2109                         case MCAST_UNBLOCK_SOURCE:
2110                         case MCAST_JOIN_GROUP:
2111                         case MCAST_LEAVE_GROUP:
2112                         case MCAST_JOIN_SOURCE_GROUP:
2113                         case MCAST_LEAVE_SOURCE_GROUP:
2114                                 error = ip6_setmoptions(inp, sopt);
2115                                 break;
2116
2117                         case IPV6_PORTRANGE:
2118                                 error = sooptcopyin(sopt, &optval,
2119                                     sizeof optval, sizeof optval);
2120                                 if (error)
2121                                         break;
2122
2123                                 INP_WLOCK(inp);
2124                                 switch (optval) {
2125                                 case IPV6_PORTRANGE_DEFAULT:
2126                                         inp->inp_flags &= ~(INP_LOWPORT);
2127                                         inp->inp_flags &= ~(INP_HIGHPORT);
2128                                         break;
2129
2130                                 case IPV6_PORTRANGE_HIGH:
2131                                         inp->inp_flags &= ~(INP_LOWPORT);
2132                                         inp->inp_flags |= INP_HIGHPORT;
2133                                         break;
2134
2135                                 case IPV6_PORTRANGE_LOW:
2136                                         inp->inp_flags &= ~(INP_HIGHPORT);
2137                                         inp->inp_flags |= INP_LOWPORT;
2138                                         break;
2139
2140                                 default:
2141                                         error = EINVAL;
2142                                         break;
2143                                 }
2144                                 INP_WUNLOCK(inp);
2145                                 break;
2146
2147 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2148                         case IPV6_IPSEC_POLICY:
2149                                 if (IPSEC_ENABLED(ipv6)) {
2150                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2151                                         break;
2152                                 }
2153                                 /* FALLTHROUGH */
2154 #endif /* IPSEC */
2155
2156                         default:
2157                                 error = ENOPROTOOPT;
2158                                 break;
2159                         }
2160                         break;
2161
2162                 case SOPT_GET:
2163                         switch (optname) {
2164                         case IPV6_2292PKTOPTIONS:
2165 #ifdef IPV6_PKTOPTIONS
2166                         case IPV6_PKTOPTIONS:
2167 #endif
2168                                 /*
2169                                  * RFC3542 (effectively) deprecated the
2170                                  * semantics of the 2292-style pktoptions.
2171                                  * Since it was not reliable in nature (i.e.,
2172                                  * applications had to expect the lack of some
2173                                  * information after all), it would make sense
2174                                  * to simplify this part by always returning
2175                                  * empty data.
2176                                  */
2177                                 sopt->sopt_valsize = 0;
2178                                 break;
2179
2180                         case IPV6_RECVHOPOPTS:
2181                         case IPV6_RECVDSTOPTS:
2182                         case IPV6_RECVRTHDRDSTOPTS:
2183                         case IPV6_UNICAST_HOPS:
2184                         case IPV6_RECVPKTINFO:
2185                         case IPV6_RECVHOPLIMIT:
2186                         case IPV6_RECVRTHDR:
2187                         case IPV6_RECVPATHMTU:
2188
2189                         case IPV6_V6ONLY:
2190                         case IPV6_PORTRANGE:
2191                         case IPV6_RECVTCLASS:
2192                         case IPV6_AUTOFLOWLABEL:
2193                         case IPV6_BINDANY:
2194                         case IPV6_FLOWID:
2195                         case IPV6_FLOWTYPE:
2196                         case IPV6_RECVFLOWID:
2197 #ifdef  RSS
2198                         case IPV6_RSSBUCKETID:
2199                         case IPV6_RECVRSSBUCKETID:
2200 #endif
2201                         case IPV6_BINDMULTI:
2202                         case IPV6_VLAN_PCP:
2203                                 switch (optname) {
2204                                 case IPV6_RECVHOPOPTS:
2205                                         optval = OPTBIT(IN6P_HOPOPTS);
2206                                         break;
2207
2208                                 case IPV6_RECVDSTOPTS:
2209                                         optval = OPTBIT(IN6P_DSTOPTS);
2210                                         break;
2211
2212                                 case IPV6_RECVRTHDRDSTOPTS:
2213                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2214                                         break;
2215
2216                                 case IPV6_UNICAST_HOPS:
2217                                         optval = inp->in6p_hops;
2218                                         break;
2219
2220                                 case IPV6_RECVPKTINFO:
2221                                         optval = OPTBIT(IN6P_PKTINFO);
2222                                         break;
2223
2224                                 case IPV6_RECVHOPLIMIT:
2225                                         optval = OPTBIT(IN6P_HOPLIMIT);
2226                                         break;
2227
2228                                 case IPV6_RECVRTHDR:
2229                                         optval = OPTBIT(IN6P_RTHDR);
2230                                         break;
2231
2232                                 case IPV6_RECVPATHMTU:
2233                                         optval = OPTBIT(IN6P_MTU);
2234                                         break;
2235
2236                                 case IPV6_V6ONLY:
2237                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2238                                         break;
2239
2240                                 case IPV6_PORTRANGE:
2241                                     {
2242                                         int flags;
2243                                         flags = inp->inp_flags;
2244                                         if (flags & INP_HIGHPORT)
2245                                                 optval = IPV6_PORTRANGE_HIGH;
2246                                         else if (flags & INP_LOWPORT)
2247                                                 optval = IPV6_PORTRANGE_LOW;
2248                                         else
2249                                                 optval = 0;
2250                                         break;
2251                                     }
2252                                 case IPV6_RECVTCLASS:
2253                                         optval = OPTBIT(IN6P_TCLASS);
2254                                         break;
2255
2256                                 case IPV6_AUTOFLOWLABEL:
2257                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2258                                         break;
2259
2260                                 case IPV6_ORIGDSTADDR:
2261                                         optval = OPTBIT2(INP_ORIGDSTADDR);
2262                                         break;
2263
2264                                 case IPV6_BINDANY:
2265                                         optval = OPTBIT(INP_BINDANY);
2266                                         break;
2267
2268                                 case IPV6_FLOWID:
2269                                         optval = inp->inp_flowid;
2270                                         break;
2271
2272                                 case IPV6_FLOWTYPE:
2273                                         optval = inp->inp_flowtype;
2274                                         break;
2275
2276                                 case IPV6_RECVFLOWID:
2277                                         optval = OPTBIT2(INP_RECVFLOWID);
2278                                         break;
2279 #ifdef  RSS
2280                                 case IPV6_RSSBUCKETID:
2281                                         retval =
2282                                             rss_hash2bucket(inp->inp_flowid,
2283                                             inp->inp_flowtype,
2284                                             &rss_bucket);
2285                                         if (retval == 0)
2286                                                 optval = rss_bucket;
2287                                         else
2288                                                 error = EINVAL;
2289                                         break;
2290
2291                                 case IPV6_RECVRSSBUCKETID:
2292                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2293                                         break;
2294 #endif
2295
2296                                 case IPV6_BINDMULTI:
2297                                         optval = OPTBIT2(INP_BINDMULTI);
2298                                         break;
2299
2300                                 case IPV6_VLAN_PCP:
2301                                         if (OPTBIT2(INP_2PCP_SET)) {
2302                                                 optval = (inp->inp_flags2 &
2303                                                             INP_2PCP_MASK) >>
2304                                                             INP_2PCP_SHIFT;
2305                                         } else {
2306                                                 optval = -1;
2307                                         }
2308                                         break;
2309                                 }
2310
2311                                 if (error)
2312                                         break;
2313                                 error = sooptcopyout(sopt, &optval,
2314                                         sizeof optval);
2315                                 break;
2316
2317                         case IPV6_PATHMTU:
2318                         {
2319                                 u_long pmtu = 0;
2320                                 struct ip6_mtuinfo mtuinfo;
2321                                 struct in6_addr addr;
2322
2323                                 if (!(so->so_state & SS_ISCONNECTED))
2324                                         return (ENOTCONN);
2325                                 /*
2326                                  * XXX: we dot not consider the case of source
2327                                  * routing, or optional information to specify
2328                                  * the outgoing interface.
2329                                  * Copy faddr out of inp to avoid holding lock
2330                                  * on inp during route lookup.
2331                                  */
2332                                 INP_RLOCK(inp);
2333                                 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2334                                 INP_RUNLOCK(inp);
2335                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2336                                     &addr, &pmtu);
2337                                 if (error)
2338                                         break;
2339                                 if (pmtu > IPV6_MAXPACKET)
2340                                         pmtu = IPV6_MAXPACKET;
2341
2342                                 bzero(&mtuinfo, sizeof(mtuinfo));
2343                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2344                                 optdata = (void *)&mtuinfo;
2345                                 optdatalen = sizeof(mtuinfo);
2346                                 error = sooptcopyout(sopt, optdata,
2347                                     optdatalen);
2348                                 break;
2349                         }
2350
2351                         case IPV6_2292PKTINFO:
2352                         case IPV6_2292HOPLIMIT:
2353                         case IPV6_2292HOPOPTS:
2354                         case IPV6_2292RTHDR:
2355                         case IPV6_2292DSTOPTS:
2356                                 switch (optname) {
2357                                 case IPV6_2292PKTINFO:
2358                                         optval = OPTBIT(IN6P_PKTINFO);
2359                                         break;
2360                                 case IPV6_2292HOPLIMIT:
2361                                         optval = OPTBIT(IN6P_HOPLIMIT);
2362                                         break;
2363                                 case IPV6_2292HOPOPTS:
2364                                         optval = OPTBIT(IN6P_HOPOPTS);
2365                                         break;
2366                                 case IPV6_2292RTHDR:
2367                                         optval = OPTBIT(IN6P_RTHDR);
2368                                         break;
2369                                 case IPV6_2292DSTOPTS:
2370                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2371                                         break;
2372                                 }
2373                                 error = sooptcopyout(sopt, &optval,
2374                                     sizeof optval);
2375                                 break;
2376                         case IPV6_PKTINFO:
2377                         case IPV6_HOPOPTS:
2378                         case IPV6_RTHDR:
2379                         case IPV6_DSTOPTS:
2380                         case IPV6_RTHDRDSTOPTS:
2381                         case IPV6_NEXTHOP:
2382                         case IPV6_TCLASS:
2383                         case IPV6_DONTFRAG:
2384                         case IPV6_USE_MIN_MTU:
2385                         case IPV6_PREFER_TEMPADDR:
2386                                 error = ip6_getpcbopt(inp, optname, sopt);
2387                                 break;
2388
2389                         case IPV6_MULTICAST_IF:
2390                         case IPV6_MULTICAST_HOPS:
2391                         case IPV6_MULTICAST_LOOP:
2392                         case IPV6_MSFILTER:
2393                                 error = ip6_getmoptions(inp, sopt);
2394                                 break;
2395
2396 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2397                         case IPV6_IPSEC_POLICY:
2398                                 if (IPSEC_ENABLED(ipv6)) {
2399                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2400                                         break;
2401                                 }
2402                                 /* FALLTHROUGH */
2403 #endif /* IPSEC */
2404                         default:
2405                                 error = ENOPROTOOPT;
2406                                 break;
2407                         }
2408                         break;
2409                 }
2410         }
2411         return (error);
2412 }
2413
2414 int
2415 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2416 {
2417         int error = 0, optval, optlen;
2418         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2419         struct inpcb *inp = sotoinpcb(so);
2420         int level, op, optname;
2421
2422         level = sopt->sopt_level;
2423         op = sopt->sopt_dir;
2424         optname = sopt->sopt_name;
2425         optlen = sopt->sopt_valsize;
2426
2427         if (level != IPPROTO_IPV6) {
2428                 return (EINVAL);
2429         }
2430
2431         switch (optname) {
2432         case IPV6_CHECKSUM:
2433                 /*
2434                  * For ICMPv6 sockets, no modification allowed for checksum
2435                  * offset, permit "no change" values to help existing apps.
2436                  *
2437                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2438                  * for an ICMPv6 socket will fail."
2439                  * The current behavior does not meet RFC3542.
2440                  */
2441                 switch (op) {
2442                 case SOPT_SET:
2443                         if (optlen != sizeof(int)) {
2444                                 error = EINVAL;
2445                                 break;
2446                         }
2447                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2448                                             sizeof(optval));
2449                         if (error)
2450                                 break;
2451                         if (optval < -1 || (optval % 2) != 0) {
2452                                 /*
2453                                  * The API assumes non-negative even offset
2454                                  * values or -1 as a special value.
2455                                  */
2456                                 error = EINVAL;
2457                         } else if (so->so_proto->pr_protocol ==
2458                             IPPROTO_ICMPV6) {
2459                                 if (optval != icmp6off)
2460                                         error = EINVAL;
2461                         } else
2462                                 inp->in6p_cksum = optval;
2463                         break;
2464
2465                 case SOPT_GET:
2466                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2467                                 optval = icmp6off;
2468                         else
2469                                 optval = inp->in6p_cksum;
2470
2471                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2472                         break;
2473
2474                 default:
2475                         error = EINVAL;
2476                         break;
2477                 }
2478                 break;
2479
2480         default:
2481                 error = ENOPROTOOPT;
2482                 break;
2483         }
2484
2485         return (error);
2486 }
2487
2488 /*
2489  * Set up IP6 options in pcb for insertion in output packets or
2490  * specifying behavior of outgoing packets.
2491  */
2492 static int
2493 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2494     struct socket *so, struct sockopt *sopt)
2495 {
2496         struct ip6_pktopts *opt = *pktopt;
2497         int error = 0;
2498         struct thread *td = sopt->sopt_td;
2499         struct epoch_tracker et;
2500
2501         /* turn off any old options. */
2502         if (opt) {
2503 #ifdef DIAGNOSTIC
2504                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2505                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2506                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2507                         printf("ip6_pcbopts: all specified options are cleared.\n");
2508 #endif
2509                 ip6_clearpktopts(opt, -1);
2510         } else {
2511                 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2512                 if (opt == NULL)
2513                         return (ENOMEM);
2514         }
2515         *pktopt = NULL;
2516
2517         if (!m || m->m_len == 0) {
2518                 /*
2519                  * Only turning off any previous options, regardless of
2520                  * whether the opt is just created or given.
2521                  */
2522                 free(opt, M_IP6OPT);
2523                 return (0);
2524         }
2525
2526         /*  set options specified by user. */
2527         NET_EPOCH_ENTER(et);
2528         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2529             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2530                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2531                 free(opt, M_IP6OPT);
2532                 NET_EPOCH_EXIT(et);
2533                 return (error);
2534         }
2535         NET_EPOCH_EXIT(et);
2536         *pktopt = opt;
2537         return (0);
2538 }
2539
2540 /*
2541  * initialize ip6_pktopts.  beware that there are non-zero default values in
2542  * the struct.
2543  */
2544 void
2545 ip6_initpktopts(struct ip6_pktopts *opt)
2546 {
2547
2548         bzero(opt, sizeof(*opt));
2549         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2550         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2551         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2552         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2553 }
2554
2555 static int
2556 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2557     struct ucred *cred, int uproto)
2558 {
2559         struct ip6_pktopts *opt;
2560
2561         if (*pktopt == NULL) {
2562                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2563                     M_NOWAIT);
2564                 if (*pktopt == NULL)
2565                         return (ENOBUFS);
2566                 ip6_initpktopts(*pktopt);
2567         }
2568         opt = *pktopt;
2569
2570         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2571 }
2572
2573 #define GET_PKTOPT_VAR(field, lenexpr) do {                                     \
2574         if (pktopt && pktopt->field) {                                          \
2575                 INP_RUNLOCK(inp);                                               \
2576                 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);         \
2577                 malloc_optdata = true;                                          \
2578                 INP_RLOCK(inp);                                                 \
2579                 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {            \
2580                         INP_RUNLOCK(inp);                                       \
2581                         free(optdata, M_TEMP);                                  \
2582                         return (ECONNRESET);                                    \
2583                 }                                                               \
2584                 pktopt = inp->in6p_outputopts;                                  \
2585                 if (pktopt && pktopt->field) {                                  \
2586                         optdatalen = min(lenexpr, sopt->sopt_valsize);          \
2587                         bcopy(&pktopt->field, optdata, optdatalen);             \
2588                 } else {                                                        \
2589                         free(optdata, M_TEMP);                                  \
2590                         optdata = NULL;                                         \
2591                         malloc_optdata = false;                                 \
2592                 }                                                               \
2593         }                                                                       \
2594 } while(0)
2595
2596 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,                         \
2597         (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2598
2599 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,                        \
2600         pktopt->field->sa_len)
2601
2602 static int
2603 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2604 {
2605         void *optdata = NULL;
2606         bool malloc_optdata = false;
2607         int optdatalen = 0;
2608         int error = 0;
2609         struct in6_pktinfo null_pktinfo;
2610         int deftclass = 0, on;
2611         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2612         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2613         struct ip6_pktopts *pktopt;
2614
2615         INP_RLOCK(inp);
2616         pktopt = inp->in6p_outputopts;
2617
2618         switch (optname) {
2619         case IPV6_PKTINFO:
2620                 optdata = (void *)&null_pktinfo;
2621                 if (pktopt && pktopt->ip6po_pktinfo) {
2622                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2623                             sizeof(null_pktinfo));
2624                         in6_clearscope(&null_pktinfo.ipi6_addr);
2625                 } else {
2626                         /* XXX: we don't have to do this every time... */
2627                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2628                 }
2629                 optdatalen = sizeof(struct in6_pktinfo);
2630                 break;
2631         case IPV6_TCLASS:
2632                 if (pktopt && pktopt->ip6po_tclass >= 0)
2633                         deftclass = pktopt->ip6po_tclass;
2634                 optdata = (void *)&deftclass;
2635                 optdatalen = sizeof(int);
2636                 break;
2637         case IPV6_HOPOPTS:
2638                 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2639                 break;
2640         case IPV6_RTHDR:
2641                 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2642                 break;
2643         case IPV6_RTHDRDSTOPTS:
2644                 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2645                 break;
2646         case IPV6_DSTOPTS:
2647                 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2648                 break;
2649         case IPV6_NEXTHOP:
2650                 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2651                 break;
2652         case IPV6_USE_MIN_MTU:
2653                 if (pktopt)
2654                         defminmtu = pktopt->ip6po_minmtu;
2655                 optdata = (void *)&defminmtu;
2656                 optdatalen = sizeof(int);
2657                 break;
2658         case IPV6_DONTFRAG:
2659                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2660                         on = 1;
2661                 else
2662                         on = 0;
2663                 optdata = (void *)&on;
2664                 optdatalen = sizeof(on);
2665                 break;
2666         case IPV6_PREFER_TEMPADDR:
2667                 if (pktopt)
2668                         defpreftemp = pktopt->ip6po_prefer_tempaddr;
2669                 optdata = (void *)&defpreftemp;
2670                 optdatalen = sizeof(int);
2671                 break;
2672         default:                /* should not happen */
2673 #ifdef DIAGNOSTIC
2674                 panic("ip6_getpcbopt: unexpected option\n");
2675 #endif
2676                 INP_RUNLOCK(inp);
2677                 return (ENOPROTOOPT);
2678         }
2679         INP_RUNLOCK(inp);
2680
2681         error = sooptcopyout(sopt, optdata, optdatalen);
2682         if (malloc_optdata)
2683                 free(optdata, M_TEMP);
2684
2685         return (error);
2686 }
2687
2688 void
2689 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2690 {
2691         if (pktopt == NULL)
2692                 return;
2693
2694         if (optname == -1 || optname == IPV6_PKTINFO) {
2695                 if (pktopt->ip6po_pktinfo)
2696                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2697                 pktopt->ip6po_pktinfo = NULL;
2698         }
2699         if (optname == -1 || optname == IPV6_HOPLIMIT)
2700                 pktopt->ip6po_hlim = -1;
2701         if (optname == -1 || optname == IPV6_TCLASS)
2702                 pktopt->ip6po_tclass = -1;
2703         if (optname == -1 || optname == IPV6_NEXTHOP) {
2704                 if (pktopt->ip6po_nextroute.ro_nh) {
2705                         NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2706                         pktopt->ip6po_nextroute.ro_nh = NULL;
2707                 }
2708                 if (pktopt->ip6po_nexthop)
2709                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2710                 pktopt->ip6po_nexthop = NULL;
2711         }
2712         if (optname == -1 || optname == IPV6_HOPOPTS) {
2713                 if (pktopt->ip6po_hbh)
2714                         free(pktopt->ip6po_hbh, M_IP6OPT);
2715                 pktopt->ip6po_hbh = NULL;
2716         }
2717         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2718                 if (pktopt->ip6po_dest1)
2719                         free(pktopt->ip6po_dest1, M_IP6OPT);
2720                 pktopt->ip6po_dest1 = NULL;
2721         }
2722         if (optname == -1 || optname == IPV6_RTHDR) {
2723                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2724                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2725                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2726                 if (pktopt->ip6po_route.ro_nh) {
2727                         NH_FREE(pktopt->ip6po_route.ro_nh);
2728                         pktopt->ip6po_route.ro_nh = NULL;
2729                 }
2730         }
2731         if (optname == -1 || optname == IPV6_DSTOPTS) {
2732                 if (pktopt->ip6po_dest2)
2733                         free(pktopt->ip6po_dest2, M_IP6OPT);
2734                 pktopt->ip6po_dest2 = NULL;
2735         }
2736 }
2737
2738 #define PKTOPT_EXTHDRCPY(type) \
2739 do {\
2740         if (src->type) {\
2741                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2742                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2743                 if (dst->type == NULL)\
2744                         goto bad;\
2745                 bcopy(src->type, dst->type, hlen);\
2746         }\
2747 } while (/*CONSTCOND*/ 0)
2748
2749 static int
2750 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2751 {
2752         if (dst == NULL || src == NULL)  {
2753                 printf("ip6_clearpktopts: invalid argument\n");
2754                 return (EINVAL);
2755         }
2756
2757         dst->ip6po_hlim = src->ip6po_hlim;
2758         dst->ip6po_tclass = src->ip6po_tclass;
2759         dst->ip6po_flags = src->ip6po_flags;
2760         dst->ip6po_minmtu = src->ip6po_minmtu;
2761         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2762         if (src->ip6po_pktinfo) {
2763                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2764                     M_IP6OPT, canwait);
2765                 if (dst->ip6po_pktinfo == NULL)
2766                         goto bad;
2767                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2768         }
2769         if (src->ip6po_nexthop) {
2770                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2771                     M_IP6OPT, canwait);
2772                 if (dst->ip6po_nexthop == NULL)
2773                         goto bad;
2774                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2775                     src->ip6po_nexthop->sa_len);
2776         }
2777         PKTOPT_EXTHDRCPY(ip6po_hbh);
2778         PKTOPT_EXTHDRCPY(ip6po_dest1);
2779         PKTOPT_EXTHDRCPY(ip6po_dest2);
2780         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2781         return (0);
2782
2783   bad:
2784         ip6_clearpktopts(dst, -1);
2785         return (ENOBUFS);
2786 }
2787 #undef PKTOPT_EXTHDRCPY
2788
2789 struct ip6_pktopts *
2790 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2791 {
2792         int error;
2793         struct ip6_pktopts *dst;
2794
2795         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2796         if (dst == NULL)
2797                 return (NULL);
2798         ip6_initpktopts(dst);
2799
2800         if ((error = copypktopts(dst, src, canwait)) != 0) {
2801                 free(dst, M_IP6OPT);
2802                 return (NULL);
2803         }
2804
2805         return (dst);
2806 }
2807
2808 void
2809 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2810 {
2811         if (pktopt == NULL)
2812                 return;
2813
2814         ip6_clearpktopts(pktopt, -1);
2815
2816         free(pktopt, M_IP6OPT);
2817 }
2818
2819 /*
2820  * Set IPv6 outgoing packet options based on advanced API.
2821  */
2822 int
2823 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2824     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2825 {
2826         struct cmsghdr *cm = NULL;
2827
2828         if (control == NULL || opt == NULL)
2829                 return (EINVAL);
2830
2831         /*
2832          * ip6_setpktopt can call ifnet_by_index(), so it's imperative that we are
2833          * in the net epoch here.
2834          */
2835         NET_EPOCH_ASSERT();
2836
2837         ip6_initpktopts(opt);
2838         if (stickyopt) {
2839                 int error;
2840
2841                 /*
2842                  * If stickyopt is provided, make a local copy of the options
2843                  * for this particular packet, then override them by ancillary
2844                  * objects.
2845                  * XXX: copypktopts() does not copy the cached route to a next
2846                  * hop (if any).  This is not very good in terms of efficiency,
2847                  * but we can allow this since this option should be rarely
2848                  * used.
2849                  */
2850                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2851                         return (error);
2852         }
2853
2854         /*
2855          * XXX: Currently, we assume all the optional information is stored
2856          * in a single mbuf.
2857          */
2858         if (control->m_next)
2859                 return (EINVAL);
2860
2861         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2862             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2863                 int error;
2864
2865                 if (control->m_len < CMSG_LEN(0))
2866                         return (EINVAL);
2867
2868                 cm = mtod(control, struct cmsghdr *);
2869                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2870                         return (EINVAL);
2871                 if (cm->cmsg_level != IPPROTO_IPV6)
2872                         continue;
2873
2874                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2875                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2876                 if (error)
2877                         return (error);
2878         }
2879
2880         return (0);
2881 }
2882
2883 /*
2884  * Set a particular packet option, as a sticky option or an ancillary data
2885  * item.  "len" can be 0 only when it's a sticky option.
2886  * We have 4 cases of combination of "sticky" and "cmsg":
2887  * "sticky=0, cmsg=0": impossible
2888  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2889  * "sticky=1, cmsg=0": RFC3542 socket option
2890  * "sticky=1, cmsg=1": RFC2292 socket option
2891  */
2892 static int
2893 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2894     struct ucred *cred, int sticky, int cmsg, int uproto)
2895 {
2896         int minmtupolicy, preftemp;
2897         int error;
2898
2899         if (!sticky && !cmsg) {
2900 #ifdef DIAGNOSTIC
2901                 printf("ip6_setpktopt: impossible case\n");
2902 #endif
2903                 return (EINVAL);
2904         }
2905
2906         /*
2907          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2908          * not be specified in the context of RFC3542.  Conversely,
2909          * RFC3542 types should not be specified in the context of RFC2292.
2910          */
2911         if (!cmsg) {
2912                 switch (optname) {
2913                 case IPV6_2292PKTINFO:
2914                 case IPV6_2292HOPLIMIT:
2915                 case IPV6_2292NEXTHOP:
2916                 case IPV6_2292HOPOPTS:
2917                 case IPV6_2292DSTOPTS:
2918                 case IPV6_2292RTHDR:
2919                 case IPV6_2292PKTOPTIONS:
2920                         return (ENOPROTOOPT);
2921                 }
2922         }
2923         if (sticky && cmsg) {
2924                 switch (optname) {
2925                 case IPV6_PKTINFO:
2926                 case IPV6_HOPLIMIT:
2927                 case IPV6_NEXTHOP:
2928                 case IPV6_HOPOPTS:
2929                 case IPV6_DSTOPTS:
2930                 case IPV6_RTHDRDSTOPTS:
2931                 case IPV6_RTHDR:
2932                 case IPV6_USE_MIN_MTU:
2933                 case IPV6_DONTFRAG:
2934                 case IPV6_TCLASS:
2935                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2936                         return (ENOPROTOOPT);
2937                 }
2938         }
2939
2940         switch (optname) {
2941         case IPV6_2292PKTINFO:
2942         case IPV6_PKTINFO:
2943         {
2944                 struct ifnet *ifp = NULL;
2945                 struct in6_pktinfo *pktinfo;
2946
2947                 if (len != sizeof(struct in6_pktinfo))
2948                         return (EINVAL);
2949
2950                 pktinfo = (struct in6_pktinfo *)buf;
2951
2952                 /*
2953                  * An application can clear any sticky IPV6_PKTINFO option by
2954                  * doing a "regular" setsockopt with ipi6_addr being
2955                  * in6addr_any and ipi6_ifindex being zero.
2956                  * [RFC 3542, Section 6]
2957                  */
2958                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2959                     pktinfo->ipi6_ifindex == 0 &&
2960                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2961                         ip6_clearpktopts(opt, optname);
2962                         break;
2963                 }
2964
2965                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2966                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2967                         return (EINVAL);
2968                 }
2969                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2970                         return (EINVAL);
2971                 /* validate the interface index if specified. */
2972                 if (pktinfo->ipi6_ifindex > V_if_index)
2973                          return (ENXIO);
2974                 if (pktinfo->ipi6_ifindex) {
2975                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2976                         if (ifp == NULL)
2977                                 return (ENXIO);
2978                 }
2979                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2980                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2981                         return (ENETDOWN);
2982
2983                 if (ifp != NULL &&
2984                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2985                         struct in6_ifaddr *ia;
2986
2987                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2988                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2989                         if (ia == NULL)
2990                                 return (EADDRNOTAVAIL);
2991                         ifa_free(&ia->ia_ifa);
2992                 }
2993                 /*
2994                  * We store the address anyway, and let in6_selectsrc()
2995                  * validate the specified address.  This is because ipi6_addr
2996                  * may not have enough information about its scope zone, and
2997                  * we may need additional information (such as outgoing
2998                  * interface or the scope zone of a destination address) to
2999                  * disambiguate the scope.
3000                  * XXX: the delay of the validation may confuse the
3001                  * application when it is used as a sticky option.
3002                  */
3003                 if (opt->ip6po_pktinfo == NULL) {
3004                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3005                             M_IP6OPT, M_NOWAIT);
3006                         if (opt->ip6po_pktinfo == NULL)
3007                                 return (ENOBUFS);
3008                 }
3009                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3010                 break;
3011         }
3012
3013         case IPV6_2292HOPLIMIT:
3014         case IPV6_HOPLIMIT:
3015         {
3016                 int *hlimp;
3017
3018                 /*
3019                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3020                  * to simplify the ordering among hoplimit options.
3021                  */
3022                 if (optname == IPV6_HOPLIMIT && sticky)
3023                         return (ENOPROTOOPT);
3024
3025                 if (len != sizeof(int))
3026                         return (EINVAL);
3027                 hlimp = (int *)buf;
3028                 if (*hlimp < -1 || *hlimp > 255)
3029                         return (EINVAL);
3030
3031                 opt->ip6po_hlim = *hlimp;
3032                 break;
3033         }
3034
3035         case IPV6_TCLASS:
3036         {
3037                 int tclass;
3038
3039                 if (len != sizeof(int))
3040                         return (EINVAL);
3041                 tclass = *(int *)buf;
3042                 if (tclass < -1 || tclass > 255)
3043                         return (EINVAL);
3044
3045                 opt->ip6po_tclass = tclass;
3046                 break;
3047         }
3048
3049         case IPV6_2292NEXTHOP:
3050         case IPV6_NEXTHOP:
3051                 if (cred != NULL) {
3052                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3053                         if (error)
3054                                 return (error);
3055                 }
3056
3057                 if (len == 0) { /* just remove the option */
3058                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3059                         break;
3060                 }
3061
3062                 /* check if cmsg_len is large enough for sa_len */
3063                 if (len < sizeof(struct sockaddr) || len < *buf)
3064                         return (EINVAL);
3065
3066                 switch (((struct sockaddr *)buf)->sa_family) {
3067                 case AF_INET6:
3068                 {
3069                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3070                         int error;
3071
3072                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3073                                 return (EINVAL);
3074
3075                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3076                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3077                                 return (EINVAL);
3078                         }
3079                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3080                             != 0) {
3081                                 return (error);
3082                         }
3083                         break;
3084                 }
3085                 case AF_LINK:   /* should eventually be supported */
3086                 default:
3087                         return (EAFNOSUPPORT);
3088                 }
3089
3090                 /* turn off the previous option, then set the new option. */
3091                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3092                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3093                 if (opt->ip6po_nexthop == NULL)
3094                         return (ENOBUFS);
3095                 bcopy(buf, opt->ip6po_nexthop, *buf);
3096                 break;
3097
3098         case IPV6_2292HOPOPTS:
3099         case IPV6_HOPOPTS:
3100         {
3101                 struct ip6_hbh *hbh;
3102                 int hbhlen;
3103
3104                 /*
3105                  * XXX: We don't allow a non-privileged user to set ANY HbH
3106                  * options, since per-option restriction has too much
3107                  * overhead.
3108                  */
3109                 if (cred != NULL) {
3110                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3111                         if (error)
3112                                 return (error);
3113                 }
3114
3115                 if (len == 0) {
3116                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3117                         break;  /* just remove the option */
3118                 }
3119
3120                 /* message length validation */
3121                 if (len < sizeof(struct ip6_hbh))
3122                         return (EINVAL);
3123                 hbh = (struct ip6_hbh *)buf;
3124                 hbhlen = (hbh->ip6h_len + 1) << 3;
3125                 if (len != hbhlen)
3126                         return (EINVAL);
3127
3128                 /* turn off the previous option, then set the new option. */
3129                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3130                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3131                 if (opt->ip6po_hbh == NULL)
3132                         return (ENOBUFS);
3133                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3134
3135                 break;
3136         }
3137
3138         case IPV6_2292DSTOPTS:
3139         case IPV6_DSTOPTS:
3140         case IPV6_RTHDRDSTOPTS:
3141         {
3142                 struct ip6_dest *dest, **newdest = NULL;
3143                 int destlen;
3144
3145                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3146                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3147                         if (error)
3148                                 return (error);
3149                 }
3150
3151                 if (len == 0) {
3152                         ip6_clearpktopts(opt, optname);
3153                         break;  /* just remove the option */
3154                 }
3155
3156                 /* message length validation */
3157                 if (len < sizeof(struct ip6_dest))
3158                         return (EINVAL);
3159                 dest = (struct ip6_dest *)buf;
3160                 destlen = (dest->ip6d_len + 1) << 3;
3161                 if (len != destlen)
3162                         return (EINVAL);
3163
3164                 /*
3165                  * Determine the position that the destination options header
3166                  * should be inserted; before or after the routing header.
3167                  */
3168                 switch (optname) {
3169                 case IPV6_2292DSTOPTS:
3170                         /*
3171                          * The old advacned API is ambiguous on this point.
3172                          * Our approach is to determine the position based
3173                          * according to the existence of a routing header.
3174                          * Note, however, that this depends on the order of the
3175                          * extension headers in the ancillary data; the 1st
3176                          * part of the destination options header must appear
3177                          * before the routing header in the ancillary data,
3178                          * too.
3179                          * RFC3542 solved the ambiguity by introducing
3180                          * separate ancillary data or option types.
3181                          */
3182                         if (opt->ip6po_rthdr == NULL)
3183                                 newdest = &opt->ip6po_dest1;
3184                         else
3185                                 newdest = &opt->ip6po_dest2;
3186                         break;
3187                 case IPV6_RTHDRDSTOPTS:
3188                         newdest = &opt->ip6po_dest1;
3189                         break;
3190                 case IPV6_DSTOPTS:
3191                         newdest = &opt->ip6po_dest2;
3192                         break;
3193                 }
3194
3195                 /* turn off the previous option, then set the new option. */
3196                 ip6_clearpktopts(opt, optname);
3197                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3198                 if (*newdest == NULL)
3199                         return (ENOBUFS);
3200                 bcopy(dest, *newdest, destlen);
3201
3202                 break;
3203         }
3204
3205         case IPV6_2292RTHDR:
3206         case IPV6_RTHDR:
3207         {
3208                 struct ip6_rthdr *rth;
3209                 int rthlen;
3210
3211                 if (len == 0) {
3212                         ip6_clearpktopts(opt, IPV6_RTHDR);
3213                         break;  /* just remove the option */
3214                 }
3215
3216                 /* message length validation */
3217                 if (len < sizeof(struct ip6_rthdr))
3218                         return (EINVAL);
3219                 rth = (struct ip6_rthdr *)buf;
3220                 rthlen = (rth->ip6r_len + 1) << 3;
3221                 if (len != rthlen)
3222                         return (EINVAL);
3223
3224                 switch (rth->ip6r_type) {
3225                 case IPV6_RTHDR_TYPE_0:
3226                         if (rth->ip6r_len == 0) /* must contain one addr */
3227                                 return (EINVAL);
3228                         if (rth->ip6r_len % 2) /* length must be even */
3229                                 return (EINVAL);
3230                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3231                                 return (EINVAL);
3232                         break;
3233                 default:
3234                         return (EINVAL);        /* not supported */
3235                 }
3236
3237                 /* turn off the previous option */
3238                 ip6_clearpktopts(opt, IPV6_RTHDR);
3239                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3240                 if (opt->ip6po_rthdr == NULL)
3241                         return (ENOBUFS);
3242                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3243
3244                 break;
3245         }
3246
3247         case IPV6_USE_MIN_MTU:
3248                 if (len != sizeof(int))
3249                         return (EINVAL);
3250                 minmtupolicy = *(int *)buf;
3251                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3252                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3253                     minmtupolicy != IP6PO_MINMTU_ALL) {
3254                         return (EINVAL);
3255                 }
3256                 opt->ip6po_minmtu = minmtupolicy;
3257                 break;
3258
3259         case IPV6_DONTFRAG:
3260                 if (len != sizeof(int))
3261                         return (EINVAL);
3262
3263                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3264                         /*
3265                          * we ignore this option for TCP sockets.
3266                          * (RFC3542 leaves this case unspecified.)
3267                          */
3268                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3269                 } else
3270                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3271                 break;
3272
3273         case IPV6_PREFER_TEMPADDR:
3274                 if (len != sizeof(int))
3275                         return (EINVAL);
3276                 preftemp = *(int *)buf;
3277                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3278                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3279                     preftemp != IP6PO_TEMPADDR_PREFER) {
3280                         return (EINVAL);
3281                 }
3282                 opt->ip6po_prefer_tempaddr = preftemp;
3283                 break;
3284
3285         default:
3286                 return (ENOPROTOOPT);
3287         } /* end of switch */
3288
3289         return (0);
3290 }
3291
3292 /*
3293  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3294  * packet to the input queue of a specified interface.  Note that this
3295  * calls the output routine of the loopback "driver", but with an interface
3296  * pointer that might NOT be &loif -- easier than replicating that code here.
3297  */
3298 void
3299 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3300 {
3301         struct mbuf *copym;
3302         struct ip6_hdr *ip6;
3303
3304         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3305         if (copym == NULL)
3306                 return;
3307
3308         /*
3309          * Make sure to deep-copy IPv6 header portion in case the data
3310          * is in an mbuf cluster, so that we can safely override the IPv6
3311          * header portion later.
3312          */
3313         if (!M_WRITABLE(copym) ||
3314             copym->m_len < sizeof(struct ip6_hdr)) {
3315                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3316                 if (copym == NULL)
3317                         return;
3318         }
3319         ip6 = mtod(copym, struct ip6_hdr *);
3320         /*
3321          * clear embedded scope identifiers if necessary.
3322          * in6_clearscope will touch the addresses only when necessary.
3323          */
3324         in6_clearscope(&ip6->ip6_src);
3325         in6_clearscope(&ip6->ip6_dst);
3326         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3327                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3328                     CSUM_PSEUDO_HDR;
3329                 copym->m_pkthdr.csum_data = 0xffff;
3330         }
3331         if_simloop(ifp, copym, AF_INET6, 0);
3332 }
3333
3334 /*
3335  * Chop IPv6 header off from the payload.
3336  */
3337 static int
3338 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3339 {
3340         struct mbuf *mh;
3341         struct ip6_hdr *ip6;
3342
3343         ip6 = mtod(m, struct ip6_hdr *);
3344         if (m->m_len > sizeof(*ip6)) {
3345                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3346                 if (mh == NULL) {
3347                         m_freem(m);
3348                         return ENOBUFS;
3349                 }
3350                 m_move_pkthdr(mh, m);
3351                 M_ALIGN(mh, sizeof(*ip6));
3352                 m->m_len -= sizeof(*ip6);
3353                 m->m_data += sizeof(*ip6);
3354                 mh->m_next = m;
3355                 m = mh;
3356                 m->m_len = sizeof(*ip6);
3357                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3358         }
3359         exthdrs->ip6e_ip6 = m;
3360         return 0;
3361 }
3362
3363 /*
3364  * Compute IPv6 extension header length.
3365  */
3366 int
3367 ip6_optlen(struct inpcb *inp)
3368 {
3369         int len;
3370
3371         if (!inp->in6p_outputopts)
3372                 return 0;
3373
3374         len = 0;
3375 #define elen(x) \
3376     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3377
3378         len += elen(inp->in6p_outputopts->ip6po_hbh);
3379         if (inp->in6p_outputopts->ip6po_rthdr)
3380                 /* dest1 is valid with rthdr only */
3381                 len += elen(inp->in6p_outputopts->ip6po_dest1);
3382         len += elen(inp->in6p_outputopts->ip6po_rthdr);
3383         len += elen(inp->in6p_outputopts->ip6po_dest2);
3384         return len;
3385 #undef elen
3386 }