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 void
 216 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
 217     int plen, int optlen)
 218 {
 219
 220         KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
 221             "csum_flags %#x",
 222             __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
 223
 224         if (csum_flags & CSUM_DELAY_DATA_IPV6) {
 225                 in6_delayed_cksum(m, plen - optlen,
 226                     sizeof(struct ip6_hdr) + optlen);
 227                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 228         }
 229 #if defined(SCTP) || defined(SCTP_SUPPORT)
 230         if (csum_flags & CSUM_SCTP_IPV6) {
 231                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
 232                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 233         }
 234 #endif
 235 }
 236
 237 int
 238 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
 239     int fraglen , uint32_t id)
 240 {
 241         struct mbuf *m, **mnext, *m_frgpart;
 242         struct ip6_hdr *ip6, *mhip6;
 243         struct ip6_frag *ip6f;
 244         int off;
 245         int error;
 246         int tlen = m0->m_pkthdr.len;
 247
 248         KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
 249
 250         m = m0;
 251         ip6 = mtod(m, struct ip6_hdr *);
 252         mnext = &m->m_nextpkt;
 253
 254         for (off = hlen; off < tlen; off += fraglen) {
 255                 m = m_gethdr(M_NOWAIT, MT_DATA);
 256                 if (!m) {
 257                         IP6STAT_INC(ip6s_odropped);
 258                         return (ENOBUFS);
 259                 }
 260
 261                 /*
 262                  * Make sure the complete packet header gets copied
 263                  * from the originating mbuf to the newly created
 264                  * mbuf. This also ensures that existing firewall
 265                  * classification(s), VLAN tags and so on get copied
 266                  * to the resulting fragmented packet(s):
 267                  */
 268                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 269                         m_free(m);
 270                         IP6STAT_INC(ip6s_odropped);
 271                         return (ENOBUFS);
 272                 }
 273
 274                 *mnext = m;
 275                 mnext = &m->m_nextpkt;
 276                 m->m_data += max_linkhdr;
 277                 mhip6 = mtod(m, struct ip6_hdr *);
 278                 *mhip6 = *ip6;
 279                 m->m_len = sizeof(*mhip6);
 280                 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 281                 if (error) {
 282                         IP6STAT_INC(ip6s_odropped);
 283                         return (error);
 284                 }
 285                 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 286                 if (off + fraglen >= tlen)
 287                         fraglen = tlen - off;
 288                 else
 289                         ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 290                 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
 291                     sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 292                 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
 293                         IP6STAT_INC(ip6s_odropped);
 294                         return (ENOBUFS);
 295                 }
 296                 m_cat(m, m_frgpart);
 297                 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
 298                 ip6f->ip6f_reserved = 0;
 299                 ip6f->ip6f_ident = id;
 300                 ip6f->ip6f_nxt = nextproto;
 301                 IP6STAT_INC(ip6s_ofragments);
 302                 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 303         }
 304
 305         return (0);
 306 }
 307
 308 static int
 309 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
 310     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
 311     bool stamp_tag)
 312 {
 313 #ifdef KERN_TLS
 314         struct ktls_session *tls = NULL;
 315 #endif
 316         struct m_snd_tag *mst;
 317         int error;
 318
 319         MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
 320         mst = NULL;
 321
 322 #ifdef KERN_TLS
 323         /*
 324          * If this is an unencrypted TLS record, save a reference to
 325          * the record.  This local reference is used to call
 326          * ktls_output_eagain after the mbuf has been freed (thus
 327          * dropping the mbuf's reference) in if_output.
 328          */
 329         if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
 330                 tls = ktls_hold(m->m_next->m_epg_tls);
 331                 mst = tls->snd_tag;
 332
 333                 /*
 334                  * If a TLS session doesn't have a valid tag, it must
 335                  * have had an earlier ifp mismatch, so drop this
 336                  * packet.
 337                  */
 338                 if (mst == NULL) {
 339                         m_freem(m);
 340                         error = EAGAIN;
 341                         goto done;
 342                 }
 343                 /*
 344                  * Always stamp tags that include NIC ktls.
 345                  */
 346                 stamp_tag = true;
 347         }
 348 #endif
 349 #ifdef RATELIMIT
 350         if (inp != NULL && mst == NULL) {
 351                 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
 352                     (inp->inp_snd_tag != NULL &&
 353                     inp->inp_snd_tag->ifp != ifp))
 354                         in_pcboutput_txrtlmt(inp, ifp, m);
 355
 356                 if (inp->inp_snd_tag != NULL)
 357                         mst = inp->inp_snd_tag;
 358         }
 359 #endif
 360         if (stamp_tag && mst != NULL) {
 361                 KASSERT(m->m_pkthdr.rcvif == NULL,
 362                     ("trying to add a send tag to a forwarded packet"));
 363                 if (mst->ifp != ifp) {
 364                         m_freem(m);
 365                         error = EAGAIN;
 366                         goto done;
 367                 }
 368
 369                 /* stamp send tag on mbuf */
 370                 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
 371                 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
 372         }
 373
 374         error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
 375
 376 done:
 377         /* Check for route change invalidating send tags. */
 378 #ifdef KERN_TLS
 379         if (tls != NULL) {
 380                 if (error == EAGAIN)
 381                         error = ktls_output_eagain(inp, tls);
 382                 ktls_free(tls);
 383         }
 384 #endif
 385 #ifdef RATELIMIT
 386         if (error == EAGAIN)
 387                 in_pcboutput_eagain(inp);
 388 #endif
 389         return (error);
 390 }
 391
 392 /*
 393  * IP6 output.
 394  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
 395  * nxt, hlim, src, dst).
 396  * This function may modify ver and hlim only.
 397  * The mbuf chain containing the packet will be freed.
 398  * The mbuf opt, if present, will not be freed.
 399  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
 400  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
 401  * then result of route lookup is stored in ro->ro_nh.
 402  *
 403  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
 404  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
 405  *
 406  * ifpp - XXX: just for statistics
 407  */
 408 int
 409 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 410     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 411     struct ifnet **ifpp, struct inpcb *inp)
 412 {
 413         struct ip6_hdr *ip6;
 414         struct ifnet *ifp, *origifp;
 415         struct mbuf *m = m0;
 416         struct mbuf *mprev;
 417         struct route_in6 *ro_pmtu;
 418         struct nhop_object *nh;
 419         struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
 420         struct in6_addr odst;
 421         u_char *nexthdrp;
 422         int tlen, len;
 423         int error = 0;
 424         int vlan_pcp = -1;
 425         struct in6_ifaddr *ia = NULL;
 426         u_long mtu;
 427         int alwaysfrag, dontfrag;
 428         u_int32_t optlen, plen = 0, unfragpartlen;
 429         struct ip6_exthdrs exthdrs;
 430         struct in6_addr src0, dst0;
 431         u_int32_t zone;
 432         bool hdrsplit;
 433         int sw_csum, tso;
 434         int needfiblookup;
 435         uint32_t fibnum;
 436         struct m_tag *fwd_tag = NULL;
 437         uint32_t id;
 438
 439         NET_EPOCH_ASSERT();
 440
 441         if (inp != NULL) {
 442                 INP_LOCK_ASSERT(inp);
 443                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 444                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 445                         /* Unconditionally set flowid. */
 446                         m->m_pkthdr.flowid = inp->inp_flowid;
 447                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 448                 }
 449                 if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
 450                         vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
 451                             INP_2PCP_SHIFT;
 452 #ifdef NUMA
 453                 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
 454 #endif
 455         }
 456
 457 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 458         /*
 459          * IPSec checking which handles several cases.
 460          * FAST IPSEC: We re-injected the packet.
 461          * XXX: need scope argument.
 462          */
 463         if (IPSEC_ENABLED(ipv6)) {
 464                 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
 465                         if (error == EINPROGRESS)
 466                                 error = 0;
 467                         goto done;
 468                 }
 469         }
 470 #endif /* IPSEC */
 471
 472         /* Source address validation. */
 473         ip6 = mtod(m, struct ip6_hdr *);
 474         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 475             (flags & IPV6_UNSPECSRC) == 0) {
 476                 error = EOPNOTSUPP;
 477                 IP6STAT_INC(ip6s_badscope);
 478                 goto bad;
 479         }
 480         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 481                 error = EOPNOTSUPP;
 482                 IP6STAT_INC(ip6s_badscope);
 483                 goto bad;
 484         }
 485
 486         /*
 487          * If we are given packet options to add extension headers prepare them.
 488          * Calculate the total length of the extension header chain.
 489          * Keep the length of the unfragmentable part for fragmentation.
 490          */
 491         bzero(&exthdrs, sizeof(exthdrs));
 492         optlen = 0;
 493         unfragpartlen = sizeof(struct ip6_hdr);
 494         if (opt) {
 495                 /* Hop-by-Hop options header. */
 496                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
 497
 498                 /* Destination options header (1st part). */
 499                 if (opt->ip6po_rthdr) {
 500 #ifndef RTHDR_SUPPORT_IMPLEMENTED
 501                         /*
 502                          * If there is a routing header, discard the packet
 503                          * right away here. RH0/1 are obsolete and we do not
 504                          * currently support RH2/3/4.
 505                          * People trying to use RH253/254 may want to disable
 506                          * this check.
 507                          * The moment we do support any routing header (again)
 508                          * this block should check the routing type more
 509                          * selectively.
 510                          */
 511                         error = EINVAL;
 512                         goto bad;
 513 #endif
 514
 515                         /*
 516                          * Destination options header (1st part).
 517                          * This only makes sense with a routing header.
 518                          * See Section 9.2 of RFC 3542.
 519                          * Disabling this part just for MIP6 convenience is
 520                          * a bad idea.  We need to think carefully about a
 521                          * way to make the advanced API coexist with MIP6
 522                          * options, which might automatically be inserted in
 523                          * the kernel.
 524                          */
 525                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
 526                             optlen);
 527                 }
 528                 /* Routing header. */
 529                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
 530
 531                 unfragpartlen += optlen;
 532
 533                 /*
 534                  * NOTE: we don't add AH/ESP length here (done in
 535                  * ip6_ipsec_output()).
 536                  */
 537
 538                 /* Destination options header (2nd part). */
 539                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
 540         }
 541
 542         /*
 543          * If there is at least one extension header,
 544          * separate IP6 header from the payload.
 545          */
 546         hdrsplit = false;
 547         if (optlen) {
 548                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 549                         m = NULL;
 550                         goto freehdrs;
 551                 }
 552                 m = exthdrs.ip6e_ip6;
 553                 ip6 = mtod(m, struct ip6_hdr *);
 554                 hdrsplit = true;
 555         }
 556
 557         /* Adjust mbuf packet header length. */
 558         m->m_pkthdr.len += optlen;
 559         plen = m->m_pkthdr.len - sizeof(*ip6);
 560
 561         /* If this is a jumbo payload, insert a jumbo payload option. */
 562         if (plen > IPV6_MAXPACKET) {
 563                 if (!hdrsplit) {
 564                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 565                                 m = NULL;
 566                                 goto freehdrs;
 567                         }
 568                         m = exthdrs.ip6e_ip6;
 569                         ip6 = mtod(m, struct ip6_hdr *);
 570                         hdrsplit = true;
 571                 }
 572                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 573                         goto freehdrs;
 574                 ip6->ip6_plen = 0;
 575         } else
 576                 ip6->ip6_plen = htons(plen);
 577         nexthdrp = &ip6->ip6_nxt;
 578
 579         if (optlen) {
 580                 /*
 581                  * Concatenate headers and fill in next header fields.
 582                  * Here we have, on "m"
 583                  *      IPv6 payload
 584                  * and we insert headers accordingly.
 585                  * Finally, we should be getting:
 586                  *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
 587                  *
 588                  * During the header composing process "m" points to IPv6
 589                  * header.  "mprev" points to an extension header prior to esp.
 590                  */
 591                 mprev = m;
 592
 593                 /*
 594                  * We treat dest2 specially.  This makes IPsec processing
 595                  * much easier.  The goal here is to make mprev point the
 596                  * mbuf prior to dest2.
 597                  *
 598                  * Result: IPv6 dest2 payload.
 599                  * m and mprev will point to IPv6 header.
 600                  */
 601                 if (exthdrs.ip6e_dest2) {
 602                         if (!hdrsplit)
 603                                 panic("%s:%d: assumption failed: "
 604                                     "hdr not split: hdrsplit %d exthdrs %p",
 605                                     __func__, __LINE__, hdrsplit, &exthdrs);
 606                         exthdrs.ip6e_dest2->m_next = m->m_next;
 607                         m->m_next = exthdrs.ip6e_dest2;
 608                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 609                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 610                 }
 611
 612                 /*
 613                  * Result: IPv6 hbh dest1 rthdr dest2 payload.
 614                  * m will point to IPv6 header.  mprev will point to the
 615                  * extension header prior to dest2 (rthdr in the above case).
 616                  */
 617                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 618                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 619                            IPPROTO_DSTOPTS);
 620                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 621                            IPPROTO_ROUTING);
 622         }
 623
 624         IP6STAT_INC(ip6s_localout);
 625
 626         /* Route packet. */
 627         ro_pmtu = ro;
 628         if (opt && opt->ip6po_rthdr)
 629                 ro = &opt->ip6po_route;
 630         if (ro != NULL)
 631                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 632         else
 633                 dst = &sin6;
 634         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 635
 636 again:
 637         /*
 638          * If specified, try to fill in the traffic class field.
 639          * Do not override if a non-zero value is already set.
 640          * We check the diffserv field and the ECN field separately.
 641          */
 642         if (opt && opt->ip6po_tclass >= 0) {
 643                 int mask = 0;
 644
 645                 if (IPV6_DSCP(ip6) == 0)
 646                         mask |= 0xfc;
 647                 if (IPV6_ECN(ip6) == 0)
 648                         mask |= 0x03;
 649                 if (mask != 0)
 650                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 651         }
 652
 653         /* Fill in or override the hop limit field, if necessary. */
 654         if (opt && opt->ip6po_hlim != -1)
 655                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 656         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 657                 if (im6o != NULL)
 658                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 659                 else
 660                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 661         }
 662
 663         if (ro == NULL || ro->ro_nh == NULL) {
 664                 bzero(dst, sizeof(*dst));
 665                 dst->sin6_family = AF_INET6;
 666                 dst->sin6_len = sizeof(*dst);
 667                 dst->sin6_addr = ip6->ip6_dst;
 668         }
 669         /*
 670          * Validate route against routing table changes.
 671          * Make sure that the address family is set in route.
 672          */
 673         nh = NULL;
 674         ifp = NULL;
 675         mtu = 0;
 676         if (ro != NULL) {
 677                 if (ro->ro_nh != NULL && inp != NULL) {
 678                         ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
 679                         NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
 680                             fibnum);
 681                 }
 682                 if (ro->ro_nh != NULL && fwd_tag == NULL &&
 683                     (!NH_IS_VALID(ro->ro_nh) ||
 684                     ro->ro_dst.sin6_family != AF_INET6 ||
 685                     !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
 686                         RO_INVALIDATE_CACHE(ro);
 687
 688                 if (ro->ro_nh != NULL && fwd_tag == NULL &&
 689                     ro->ro_dst.sin6_family == AF_INET6 &&
 690                     IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
 691                         nh = ro->ro_nh;
 692                         ifp = nh->nh_ifp;
 693                 } else {
 694                         if (ro->ro_lle)
 695                                 LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 696                         ro->ro_lle = NULL;
 697                         if (fwd_tag == NULL) {
 698                                 bzero(&dst_sa, sizeof(dst_sa));
 699                                 dst_sa.sin6_family = AF_INET6;
 700                                 dst_sa.sin6_len = sizeof(dst_sa);
 701                                 dst_sa.sin6_addr = ip6->ip6_dst;
 702                         }
 703                         error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
 704                             &nh, fibnum, m->m_pkthdr.flowid);
 705                         if (error != 0) {
 706                                 IP6STAT_INC(ip6s_noroute);
 707                                 if (ifp != NULL)
 708                                         in6_ifstat_inc(ifp, ifs6_out_discard);
 709                                 goto bad;
 710                         }
 711                         if (ifp != NULL)
 712                             mtu = ifp->if_mtu;
 713                 }
 714                 if (nh == NULL) {
 715                         /*
 716                          * If in6_selectroute() does not return a nexthop
 717                          * dst may not have been updated.
 718                          */
 719                         *dst = dst_sa;  /* XXX */
 720                 } else {
 721                         if (nh->nh_flags & NHF_HOST)
 722                             mtu = nh->nh_mtu;
 723                         ia = (struct in6_ifaddr *)(nh->nh_ifa);
 724                         counter_u64_add(nh->nh_pksent, 1);
 725                 }
 726         } else {
 727                 struct nhop_object *nh;
 728                 struct in6_addr kdst;
 729                 uint32_t scopeid;
 730
 731                 if (fwd_tag == NULL) {
 732                         bzero(&dst_sa, sizeof(dst_sa));
 733                         dst_sa.sin6_family = AF_INET6;
 734                         dst_sa.sin6_len = sizeof(dst_sa);
 735                         dst_sa.sin6_addr = ip6->ip6_dst;
 736                 }
 737
 738                 if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
 739                     im6o != NULL &&
 740                     (ifp = im6o->im6o_multicast_ifp) != NULL) {
 741                         /* We do not need a route lookup. */
 742                         *dst = dst_sa;  /* XXX */
 743                         goto nonh6lookup;
 744                 }
 745
 746                 in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
 747
 748                 if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
 749                     IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
 750                         if (scopeid > 0) {
 751                                 ifp = in6_getlinkifnet(scopeid);
 752                                 if (ifp == NULL) {
 753                                         error = EHOSTUNREACH;
 754                                         goto bad;
 755                                 }
 756                                 *dst = dst_sa;  /* XXX */
 757                                 goto nonh6lookup;
 758                         }
 759                 }
 760
 761                 nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
 762                     m->m_pkthdr.flowid);
 763                 if (nh == NULL) {
 764                         IP6STAT_INC(ip6s_noroute);
 765                         /* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
 766                         error = EHOSTUNREACH;;
 767                         goto bad;
 768                 }
 769
 770                 ifp = nh->nh_ifp;
 771                 mtu = nh->nh_mtu;
 772                 ia = ifatoia6(nh->nh_ifa);
 773                 if (nh->nh_flags & NHF_GATEWAY)
 774                         dst->sin6_addr = nh->gw6_sa.sin6_addr;
 775 nonh6lookup:
 776                 ;
 777         }
 778
 779         /* Then nh (for unicast) and ifp must be non-NULL valid values. */
 780         if ((flags & IPV6_FORWARDING) == 0) {
 781                 /* XXX: the FORWARDING flag can be set for mrouting. */
 782                 in6_ifstat_inc(ifp, ifs6_out_request);
 783         }
 784
 785         /* Setup data structures for scope ID checks. */
 786         src0 = ip6->ip6_src;
 787         bzero(&src_sa, sizeof(src_sa));
 788         src_sa.sin6_family = AF_INET6;
 789         src_sa.sin6_len = sizeof(src_sa);
 790         src_sa.sin6_addr = ip6->ip6_src;
 791
 792         dst0 = ip6->ip6_dst;
 793         /* Re-initialize to be sure. */
 794         bzero(&dst_sa, sizeof(dst_sa));
 795         dst_sa.sin6_family = AF_INET6;
 796         dst_sa.sin6_len = sizeof(dst_sa);
 797         dst_sa.sin6_addr = ip6->ip6_dst;
 798
 799         /* Check for valid scope ID. */
 800         if (in6_setscope(&src0, ifp, &zone) == 0 &&
 801             sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
 802             in6_setscope(&dst0, ifp, &zone) == 0 &&
 803             sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
 804                 /*
 805                  * The outgoing interface is in the zone of the source
 806                  * and destination addresses.
 807                  *
 808                  * Because the loopback interface cannot receive
 809                  * packets with a different scope ID than its own,
 810                  * there is a trick to pretend the outgoing packet
 811                  * was received by the real network interface, by
 812                  * setting "origifp" different from "ifp". This is
 813                  * only allowed when "ifp" is a loopback network
 814                  * interface. Refer to code in nd6_output_ifp() for
 815                  * more details.
 816                  */
 817                 origifp = ifp;
 818
 819                 /*
 820                  * We should use ia_ifp to support the case of sending
 821                  * packets to an address of our own.
 822                  */
 823                 if (ia != NULL && ia->ia_ifp)
 824                         ifp = ia->ia_ifp;
 825
 826         } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
 827             sa6_recoverscope(&src_sa) != 0 ||
 828             sa6_recoverscope(&dst_sa) != 0 ||
 829             dst_sa.sin6_scope_id == 0 ||
 830             (src_sa.sin6_scope_id != 0 &&
 831             src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
 832             (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
 833                 /*
 834                  * If the destination network interface is not a
 835                  * loopback interface, or the destination network
 836                  * address has no scope ID, or the source address has
 837                  * a scope ID set which is different from the
 838                  * destination address one, or there is no network
 839                  * interface representing this scope ID, the address
 840                  * pair is considered invalid.
 841                  */
 842                 IP6STAT_INC(ip6s_badscope);
 843                 in6_ifstat_inc(ifp, ifs6_out_discard);
 844                 if (error == 0)
 845                         error = EHOSTUNREACH; /* XXX */
 846                 goto bad;
 847         }
 848         /* All scope ID checks are successful. */
 849
 850         if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 851                 if (opt && opt->ip6po_nextroute.ro_nh) {
 852                         /*
 853                          * The nexthop is explicitly specified by the
 854                          * application.  We assume the next hop is an IPv6
 855                          * address.
 856                          */
 857                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 858                 }
 859                 else if ((nh->nh_flags & NHF_GATEWAY))
 860                         dst = &nh->gw6_sa;
 861         }
 862
 863         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 864                 m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
 865         } else {
 866                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 867                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 868
 869                 /* Confirm that the outgoing interface supports multicast. */
 870                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 871                         IP6STAT_INC(ip6s_noroute);
 872                         in6_ifstat_inc(ifp, ifs6_out_discard);
 873                         error = ENETUNREACH;
 874                         goto bad;
 875                 }
 876                 if ((im6o == NULL && in6_mcast_loop) ||
 877                     (im6o && im6o->im6o_multicast_loop)) {
 878                         /*
 879                          * Loop back multicast datagram if not expressly
 880                          * forbidden to do so, even if we have not joined
 881                          * the address; protocols will filter it later,
 882                          * thus deferring a hash lookup and lock acquisition
 883                          * at the expense of an m_copym().
 884                          */
 885                         ip6_mloopback(ifp, m);
 886                 } else {
 887                         /*
 888                          * If we are acting as a multicast router, perform
 889                          * multicast forwarding as if the packet had just
 890                          * arrived on the interface to which we are about
 891                          * to send.  The multicast forwarding function
 892                          * recursively calls this function, using the
 893                          * IPV6_FORWARDING flag to prevent infinite recursion.
 894                          *
 895                          * Multicasts that are looped back by ip6_mloopback(),
 896                          * above, will be forwarded by the ip6_input() routine,
 897                          * if necessary.
 898                          */
 899                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 900                                 /*
 901                                  * XXX: ip6_mforward expects that rcvif is NULL
 902                                  * when it is called from the originating path.
 903                                  * However, it may not always be the case.
 904                                  */
 905                                 m->m_pkthdr.rcvif = NULL;
 906                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 907                                         m_freem(m);
 908                                         goto done;
 909                                 }
 910                         }
 911                 }
 912                 /*
 913                  * Multicasts with a hoplimit of zero may be looped back,
 914                  * above, but must not be transmitted on a network.
 915                  * Also, multicasts addressed to the loopback interface
 916                  * are not sent -- the above call to ip6_mloopback() will
 917                  * loop back a copy if this host actually belongs to the
 918                  * destination group on the loopback interface.
 919                  */
 920                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 921                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 922                         m_freem(m);
 923                         goto done;
 924                 }
 925         }
 926
 927         /*
 928          * Fill the outgoing inteface to tell the upper layer
 929          * to increment per-interface statistics.
 930          */
 931         if (ifpp)
 932                 *ifpp = ifp;
 933
 934         /* Determine path MTU. */
 935         if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
 936                     &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
 937                 goto bad;
 938         KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
 939             "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
 940             ifp, alwaysfrag, fibnum));
 941
 942         /*
 943          * The caller of this function may specify to use the minimum MTU
 944          * in some cases.
 945          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 946          * setting.  The logic is a bit complicated; by default, unicast
 947          * packets will follow path MTU while multicast packets will be sent at
 948          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 949          * including unicast ones will be sent at the minimum MTU.  Multicast
 950          * packets will always be sent at the minimum MTU unless
 951          * IP6PO_MINMTU_DISABLE is explicitly specified.
 952          * See RFC 3542 for more details.
 953          */
 954         if (mtu > IPV6_MMTU) {
 955                 if ((flags & IPV6_MINMTU))
 956                         mtu = IPV6_MMTU;
 957                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 958                         mtu = IPV6_MMTU;
 959                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 960                          (opt == NULL ||
 961                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 962                         mtu = IPV6_MMTU;
 963                 }
 964         }
 965
 966         /*
 967          * Clear embedded scope identifiers if necessary.
 968          * in6_clearscope() will touch the addresses only when necessary.
 969          */
 970         in6_clearscope(&ip6->ip6_src);
 971         in6_clearscope(&ip6->ip6_dst);
 972
 973         /*
 974          * If the outgoing packet contains a hop-by-hop options header,
 975          * it must be examined and processed even by the source node.
 976          * (RFC 2460, section 4.)
 977          */
 978         if (exthdrs.ip6e_hbh) {
 979                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 980                 u_int32_t dummy; /* XXX unused */
 981                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 982
 983 #ifdef DIAGNOSTIC
 984                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 985                         panic("ip6e_hbh is not contiguous");
 986 #endif
 987                 /*
 988                  *  XXX: if we have to send an ICMPv6 error to the sender,
 989                  *       we need the M_LOOP flag since icmp6_error() expects
 990                  *       the IPv6 and the hop-by-hop options header are
 991                  *       contiguous unless the flag is set.
 992                  */
 993                 m->m_flags |= M_LOOP;
 994                 m->m_pkthdr.rcvif = ifp;
 995                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 996                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 997                     &dummy, &plen) < 0) {
 998                         /* m was already freed at this point. */
 999                         error = EINVAL;/* better error? */
1000                         goto done;
1001                 }
1002                 m->m_flags &= ~M_LOOP; /* XXX */
1003                 m->m_pkthdr.rcvif = NULL;
1004         }
1005
1006         /* Jump over all PFIL processing if hooks are not active. */
1007         if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1008                 goto passout;
1009
1010         odst = ip6->ip6_dst;
1011         /* Run through list of hooks for output packets. */
1012         switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
1013         case PFIL_PASS:
1014                 ip6 = mtod(m, struct ip6_hdr *);
1015                 break;
1016         case PFIL_DROPPED:
1017                 error = EACCES;
1018                 /* FALLTHROUGH */
1019         case PFIL_CONSUMED:
1020                 goto done;
1021         }
1022
1023         needfiblookup = 0;
1024         /* See if destination IP address was changed by packet filter. */
1025         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1026                 m->m_flags |= M_SKIP_FIREWALL;
1027                 /* If destination is now ourself drop to ip6_input(). */
1028                 if (in6_localip(&ip6->ip6_dst)) {
1029                         m->m_flags |= M_FASTFWD_OURS;
1030                         if (m->m_pkthdr.rcvif == NULL)
1031                                 m->m_pkthdr.rcvif = V_loif;
1032                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1033                                 m->m_pkthdr.csum_flags |=
1034                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1035                                 m->m_pkthdr.csum_data = 0xffff;
1036                         }
1037 #if defined(SCTP) || defined(SCTP_SUPPORT)
1038                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1039                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1040 #endif
1041                         error = netisr_queue(NETISR_IPV6, m);
1042                         goto done;
1043                 } else {
1044                         if (ro != NULL)
1045                                 RO_INVALIDATE_CACHE(ro);
1046                         needfiblookup = 1; /* Redo the routing table lookup. */
1047                 }
1048         }
1049         /* See if fib was changed by packet filter. */
1050         if (fibnum != M_GETFIB(m)) {
1051                 m->m_flags |= M_SKIP_FIREWALL;
1052                 fibnum = M_GETFIB(m);
1053                 if (ro != NULL)
1054                         RO_INVALIDATE_CACHE(ro);
1055                 needfiblookup = 1;
1056         }
1057         if (needfiblookup)
1058                 goto again;
1059
1060         /* See if local, if yes, send it to netisr. */
1061         if (m->m_flags & M_FASTFWD_OURS) {
1062                 if (m->m_pkthdr.rcvif == NULL)
1063                         m->m_pkthdr.rcvif = V_loif;
1064                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1065                         m->m_pkthdr.csum_flags |=
1066                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1067                         m->m_pkthdr.csum_data = 0xffff;
1068                 }
1069 #if defined(SCTP) || defined(SCTP_SUPPORT)
1070                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1071                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1072 #endif
1073                 error = netisr_queue(NETISR_IPV6, m);
1074                 goto done;
1075         }
1076         /* Or forward to some other address? */
1077         if ((m->m_flags & M_IP6_NEXTHOP) &&
1078             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1079                 if (ro != NULL)
1080                         dst = (struct sockaddr_in6 *)&ro->ro_dst;
1081                 else
1082                         dst = &sin6;
1083                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1084                 m->m_flags |= M_SKIP_FIREWALL;
1085                 m->m_flags &= ~M_IP6_NEXTHOP;
1086                 m_tag_delete(m, fwd_tag);
1087                 goto again;
1088         }
1089
1090 passout:
1091         if (vlan_pcp > -1)
1092                 EVL_APPLY_PRI(m, vlan_pcp);
1093
1094         /* Ensure the packet data is mapped if the interface requires it. */
1095         if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1096                 m = mb_unmapped_to_ext(m);
1097                 if (m == NULL) {
1098                         IP6STAT_INC(ip6s_odropped);
1099                         return (ENOBUFS);
1100                 }
1101         }
1102
1103         /*
1104          * Send the packet to the outgoing interface.
1105          * If necessary, do IPv6 fragmentation before sending.
1106          *
1107          * The logic here is rather complex:
1108          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1109          * 1-a: send as is if tlen <= path mtu
1110          * 1-b: fragment if tlen > path mtu
1111          *
1112          * 2: if user asks us not to fragment (dontfrag == 1)
1113          * 2-a: send as is if tlen <= interface mtu
1114          * 2-b: error if tlen > interface mtu
1115          *
1116          * 3: if we always need to attach fragment header (alwaysfrag == 1)
1117          *      always fragment
1118          *
1119          * 4: if dontfrag == 1 && alwaysfrag == 1
1120          *      error, as we cannot handle this conflicting request.
1121          */
1122         sw_csum = m->m_pkthdr.csum_flags;
1123         if (!hdrsplit) {
1124                 tso = ((sw_csum & ifp->if_hwassist &
1125                     (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1126                 sw_csum &= ~ifp->if_hwassist;
1127         } else
1128                 tso = 0;
1129         /*
1130          * If we added extension headers, we will not do TSO and calculate the
1131          * checksums ourselves for now.
1132          * XXX-BZ  Need a framework to know when the NIC can handle it, even
1133          * with ext. hdrs.
1134          */
1135         ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1136         /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1137         tlen = m->m_pkthdr.len;
1138
1139         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1140                 dontfrag = 1;
1141         else
1142                 dontfrag = 0;
1143         if (dontfrag && alwaysfrag) {   /* Case 4. */
1144                 /* Conflicting request - can't transmit. */
1145                 error = EMSGSIZE;
1146                 goto bad;
1147         }
1148         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* Case 2-b. */
1149                 /*
1150                  * Even if the DONTFRAG option is specified, we cannot send the
1151                  * packet when the data length is larger than the MTU of the
1152                  * outgoing interface.
1153                  * Notify the error by sending IPV6_PATHMTU ancillary data if
1154                  * application wanted to know the MTU value. Also return an
1155                  * error code (this is not described in the API spec).
1156                  */
1157                 if (inp != NULL)
1158                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1159                 error = EMSGSIZE;
1160                 goto bad;
1161         }
1162
1163         /* Transmit packet without fragmentation. */
1164         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* Cases 1-a and 2-a. */
1165                 struct in6_ifaddr *ia6;
1166
1167                 ip6 = mtod(m, struct ip6_hdr *);
1168                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1169                 if (ia6) {
1170                         /* Record statistics for this interface address. */
1171                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1172                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
1173                             m->m_pkthdr.len);
1174                 }
1175                 error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1176                     (flags & IP_NO_SND_TAG_RL) ? false : true);
1177                 goto done;
1178         }
1179
1180         /* Try to fragment the packet.  Cases 1-b and 3. */
1181         if (mtu < IPV6_MMTU) {
1182                 /* Path MTU cannot be less than IPV6_MMTU. */
1183                 error = EMSGSIZE;
1184                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1185                 goto bad;
1186         } else if (ip6->ip6_plen == 0) {
1187                 /* Jumbo payload cannot be fragmented. */
1188                 error = EMSGSIZE;
1189                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1190                 goto bad;
1191         } else {
1192                 u_char nextproto;
1193
1194                 /*
1195                  * Too large for the destination or interface;
1196                  * fragment if possible.
1197                  * Must be able to put at least 8 bytes per fragment.
1198                  */
1199                 if (mtu > IPV6_MAXPACKET)
1200                         mtu = IPV6_MAXPACKET;
1201
1202                 len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1203                 if (len < 8) {
1204                         error = EMSGSIZE;
1205                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1206                         goto bad;
1207                 }
1208
1209                 /*
1210                  * If the interface will not calculate checksums on
1211                  * fragmented packets, then do it here.
1212                  * XXX-BZ handle the hw offloading case.  Need flags.
1213                  */
1214                 ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1215                     optlen);
1216
1217                 /*
1218                  * Change the next header field of the last header in the
1219                  * unfragmentable part.
1220                  */
1221                 if (exthdrs.ip6e_rthdr) {
1222                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1223                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1224                 } else if (exthdrs.ip6e_dest1) {
1225                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1226                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1227                 } else if (exthdrs.ip6e_hbh) {
1228                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1229                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1230                 } else {
1231                         ip6 = mtod(m, struct ip6_hdr *);
1232                         nextproto = ip6->ip6_nxt;
1233                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1234                 }
1235
1236                 /*
1237                  * Loop through length of segment after first fragment,
1238                  * make new header and copy data of each part and link onto
1239                  * chain.
1240                  */
1241                 m0 = m;
1242                 id = htonl(ip6_randomid());
1243                 error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1244                 if (error != 0)
1245                         goto sendorfree;
1246
1247                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1248         }
1249
1250         /* Remove leading garbage. */
1251 sendorfree:
1252         m = m0->m_nextpkt;
1253         m0->m_nextpkt = 0;
1254         m_freem(m0);
1255         for (; m; m = m0) {
1256                 m0 = m->m_nextpkt;
1257                 m->m_nextpkt = 0;
1258                 if (error == 0) {
1259                         /* Record statistics for this interface address. */
1260                         if (ia) {
1261                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1262                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1263                                     m->m_pkthdr.len);
1264                         }
1265                         if (vlan_pcp > -1)
1266                                 EVL_APPLY_PRI(m, vlan_pcp);
1267                         error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1268                             true);
1269                 } else
1270                         m_freem(m);
1271         }
1272
1273         if (error == 0)
1274                 IP6STAT_INC(ip6s_fragmented);
1275
1276 done:
1277         return (error);
1278
1279 freehdrs:
1280         m_freem(exthdrs.ip6e_hbh);      /* m_freem() checks if mbuf is NULL. */
1281         m_freem(exthdrs.ip6e_dest1);
1282         m_freem(exthdrs.ip6e_rthdr);
1283         m_freem(exthdrs.ip6e_dest2);
1284         /* FALLTHROUGH */
1285 bad:
1286         if (m)
1287                 m_freem(m);
1288         goto done;
1289 }
1290
1291 static int
1292 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1293 {
1294         struct mbuf *m;
1295
1296         if (hlen > MCLBYTES)
1297                 return (ENOBUFS); /* XXX */
1298
1299         if (hlen > MLEN)
1300                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1301         else
1302                 m = m_get(M_NOWAIT, MT_DATA);
1303         if (m == NULL)
1304                 return (ENOBUFS);
1305         m->m_len = hlen;
1306         if (hdr)
1307                 bcopy(hdr, mtod(m, caddr_t), hlen);
1308
1309         *mp = m;
1310         return (0);
1311 }
1312
1313 /*
1314  * Insert jumbo payload option.
1315  */
1316 static int
1317 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1318 {
1319         struct mbuf *mopt;
1320         u_char *optbuf;
1321         u_int32_t v;
1322
1323 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1324
1325         /*
1326          * If there is no hop-by-hop options header, allocate new one.
1327          * If there is one but it doesn't have enough space to store the
1328          * jumbo payload option, allocate a cluster to store the whole options.
1329          * Otherwise, use it to store the options.
1330          */
1331         if (exthdrs->ip6e_hbh == NULL) {
1332                 mopt = m_get(M_NOWAIT, MT_DATA);
1333                 if (mopt == NULL)
1334                         return (ENOBUFS);
1335                 mopt->m_len = JUMBOOPTLEN;
1336                 optbuf = mtod(mopt, u_char *);
1337                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1338                 exthdrs->ip6e_hbh = mopt;
1339         } else {
1340                 struct ip6_hbh *hbh;
1341
1342                 mopt = exthdrs->ip6e_hbh;
1343                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1344                         /*
1345                          * XXX assumption:
1346                          * - exthdrs->ip6e_hbh is not referenced from places
1347                          *   other than exthdrs.
1348                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1349                          */
1350                         int oldoptlen = mopt->m_len;
1351                         struct mbuf *n;
1352
1353                         /*
1354                          * XXX: give up if the whole (new) hbh header does
1355                          * not fit even in an mbuf cluster.
1356                          */
1357                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1358                                 return (ENOBUFS);
1359
1360                         /*
1361                          * As a consequence, we must always prepare a cluster
1362                          * at this point.
1363                          */
1364                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1365                         if (n == NULL)
1366                                 return (ENOBUFS);
1367                         n->m_len = oldoptlen + JUMBOOPTLEN;
1368                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1369                             oldoptlen);
1370                         optbuf = mtod(n, caddr_t) + oldoptlen;
1371                         m_freem(mopt);
1372                         mopt = exthdrs->ip6e_hbh = n;
1373                 } else {
1374                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1375                         mopt->m_len += JUMBOOPTLEN;
1376                 }
1377                 optbuf[0] = IP6OPT_PADN;
1378                 optbuf[1] = 1;
1379
1380                 /*
1381                  * Adjust the header length according to the pad and
1382                  * the jumbo payload option.
1383                  */
1384                 hbh = mtod(mopt, struct ip6_hbh *);
1385                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1386         }
1387
1388         /* fill in the option. */
1389         optbuf[2] = IP6OPT_JUMBO;
1390         optbuf[3] = 4;
1391         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1392         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1393
1394         /* finally, adjust the packet header length */
1395         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1396
1397         return (0);
1398 #undef JUMBOOPTLEN
1399 }
1400
1401 /*
1402  * Insert fragment header and copy unfragmentable header portions.
1403  */
1404 static int
1405 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1406     struct ip6_frag **frghdrp)
1407 {
1408         struct mbuf *n, *mlast;
1409
1410         if (hlen > sizeof(struct ip6_hdr)) {
1411                 n = m_copym(m0, sizeof(struct ip6_hdr),
1412                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1413                 if (n == NULL)
1414                         return (ENOBUFS);
1415                 m->m_next = n;
1416         } else
1417                 n = m;
1418
1419         /* Search for the last mbuf of unfragmentable part. */
1420         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1421                 ;
1422
1423         if (M_WRITABLE(mlast) &&
1424             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1425                 /* use the trailing space of the last mbuf for the fragment hdr */
1426                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1427                     mlast->m_len);
1428                 mlast->m_len += sizeof(struct ip6_frag);
1429                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1430         } else {
1431                 /* allocate a new mbuf for the fragment header */
1432                 struct mbuf *mfrg;
1433
1434                 mfrg = m_get(M_NOWAIT, MT_DATA);
1435                 if (mfrg == NULL)
1436                         return (ENOBUFS);
1437                 mfrg->m_len = sizeof(struct ip6_frag);
1438                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1439                 mlast->m_next = mfrg;
1440         }
1441
1442         return (0);
1443 }
1444
1445 /*
1446  * Calculates IPv6 path mtu for destination @dst.
1447  * Resulting MTU is stored in @mtup.
1448  *
1449  * Returns 0 on success.
1450  */
1451 static int
1452 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1453 {
1454         struct epoch_tracker et;
1455         struct nhop_object *nh;
1456         struct in6_addr kdst;
1457         uint32_t scopeid;
1458         int error;
1459
1460         in6_splitscope(dst, &kdst, &scopeid);
1461
1462         NET_EPOCH_ENTER(et);
1463         nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1464         if (nh != NULL)
1465                 error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1466         else
1467                 error = EHOSTUNREACH;
1468         NET_EPOCH_EXIT(et);
1469
1470         return (error);
1471 }
1472
1473 /*
1474  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1475  * and cached data in @ro_pmtu.
1476  * MTU from (successful) route lookup is saved (along with dst)
1477  * inside @ro_pmtu to avoid subsequent route lookups after packet
1478  * filter processing.
1479  *
1480  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1481  * Returns 0 on success.
1482  */
1483 static int
1484 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1485     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1486     int *alwaysfragp, u_int fibnum, u_int proto)
1487 {
1488         struct nhop_object *nh;
1489         struct in6_addr kdst;
1490         uint32_t scopeid;
1491         struct sockaddr_in6 *sa6_dst, sin6;
1492         u_long mtu;
1493
1494         NET_EPOCH_ASSERT();
1495
1496         mtu = 0;
1497         if (ro_pmtu == NULL || do_lookup) {
1498                 /*
1499                  * Here ro_pmtu has final destination address, while
1500                  * ro might represent immediate destination.
1501                  * Use ro_pmtu destination since mtu might differ.
1502                  */
1503                 if (ro_pmtu != NULL) {
1504                         sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1505                         if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1506                                 ro_pmtu->ro_mtu = 0;
1507                 } else
1508                         sa6_dst = &sin6;
1509
1510                 if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1511                         bzero(sa6_dst, sizeof(*sa6_dst));
1512                         sa6_dst->sin6_family = AF_INET6;
1513                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1514                         sa6_dst->sin6_addr = *dst;
1515
1516                         in6_splitscope(dst, &kdst, &scopeid);
1517                         nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1518                         if (nh != NULL) {
1519                                 mtu = nh->nh_mtu;
1520                                 if (ro_pmtu != NULL)
1521                                         ro_pmtu->ro_mtu = mtu;
1522                         }
1523                 } else
1524                         mtu = ro_pmtu->ro_mtu;
1525         }
1526
1527         if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1528                 mtu = ro_pmtu->ro_nh->nh_mtu;
1529
1530         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1531 }
1532
1533 /*
1534  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1535  * hostcache data for @dst.
1536  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1537  *
1538  * Returns 0 on success.
1539  */
1540 static int
1541 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1542     u_long *mtup, int *alwaysfragp, u_int proto)
1543 {
1544         u_long mtu = 0;
1545         int alwaysfrag = 0;
1546         int error = 0;
1547
1548         if (rt_mtu > 0) {
1549                 u_int32_t ifmtu;
1550                 struct in_conninfo inc;
1551
1552                 bzero(&inc, sizeof(inc));
1553                 inc.inc_flags |= INC_ISIPV6;
1554                 inc.inc6_faddr = *dst;
1555
1556                 ifmtu = IN6_LINKMTU(ifp);
1557
1558                 /* TCP is known to react to pmtu changes so skip hc */
1559                 if (proto != IPPROTO_TCP)
1560                         mtu = tcp_hc_getmtu(&inc);
1561
1562                 if (mtu)
1563                         mtu = min(mtu, rt_mtu);
1564                 else
1565                         mtu = rt_mtu;
1566                 if (mtu == 0)
1567                         mtu = ifmtu;
1568                 else if (mtu < IPV6_MMTU) {
1569                         /*
1570                          * RFC2460 section 5, last paragraph:
1571                          * if we record ICMPv6 too big message with
1572                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1573                          * or smaller, with framgent header attached.
1574                          * (fragment header is needed regardless from the
1575                          * packet size, for translators to identify packets)
1576                          */
1577                         alwaysfrag = 1;
1578                         mtu = IPV6_MMTU;
1579                 }
1580         } else if (ifp) {
1581                 mtu = IN6_LINKMTU(ifp);
1582         } else
1583                 error = EHOSTUNREACH; /* XXX */
1584
1585         *mtup = mtu;
1586         if (alwaysfragp)
1587                 *alwaysfragp = alwaysfrag;
1588         return (error);
1589 }
1590
1591 /*
1592  * IP6 socket option processing.
1593  */
1594 int
1595 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1596 {
1597         int optdatalen, uproto;
1598         void *optdata;
1599         struct inpcb *inp = sotoinpcb(so);
1600         int error, optval;
1601         int level, op, optname;
1602         int optlen;
1603         struct thread *td;
1604 #ifdef  RSS
1605         uint32_t rss_bucket;
1606         int retval;
1607 #endif
1608
1609 /*
1610  * Don't use more than a quarter of mbuf clusters.  N.B.:
1611  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1612  * on LP64 architectures, so cast to u_long to avoid undefined
1613  * behavior.  ILP32 architectures cannot have nmbclusters
1614  * large enough to overflow for other reasons.
1615  */
1616 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1617
1618         level = sopt->sopt_level;
1619         op = sopt->sopt_dir;
1620         optname = sopt->sopt_name;
1621         optlen = sopt->sopt_valsize;
1622         td = sopt->sopt_td;
1623         error = 0;
1624         optval = 0;
1625         uproto = (int)so->so_proto->pr_protocol;
1626
1627         if (level != IPPROTO_IPV6) {
1628                 error = EINVAL;
1629
1630                 if (sopt->sopt_level == SOL_SOCKET &&
1631                     sopt->sopt_dir == SOPT_SET) {
1632                         switch (sopt->sopt_name) {
1633                         case SO_REUSEADDR:
1634                                 INP_WLOCK(inp);
1635                                 if ((so->so_options & SO_REUSEADDR) != 0)
1636                                         inp->inp_flags2 |= INP_REUSEADDR;
1637                                 else
1638                                         inp->inp_flags2 &= ~INP_REUSEADDR;
1639                                 INP_WUNLOCK(inp);
1640                                 error = 0;
1641                                 break;
1642                         case SO_REUSEPORT:
1643                                 INP_WLOCK(inp);
1644                                 if ((so->so_options & SO_REUSEPORT) != 0)
1645                                         inp->inp_flags2 |= INP_REUSEPORT;
1646                                 else
1647                                         inp->inp_flags2 &= ~INP_REUSEPORT;
1648                                 INP_WUNLOCK(inp);
1649                                 error = 0;
1650                                 break;
1651                         case SO_REUSEPORT_LB:
1652                                 INP_WLOCK(inp);
1653                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1654                                         inp->inp_flags2 |= INP_REUSEPORT_LB;
1655                                 else
1656                                         inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1657                                 INP_WUNLOCK(inp);
1658                                 error = 0;
1659                                 break;
1660                         case SO_SETFIB:
1661                                 INP_WLOCK(inp);
1662                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
1663                                 INP_WUNLOCK(inp);
1664                                 error = 0;
1665                                 break;
1666                         case SO_MAX_PACING_RATE:
1667 #ifdef RATELIMIT
1668                                 INP_WLOCK(inp);
1669                                 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1670                                 INP_WUNLOCK(inp);
1671                                 error = 0;
1672 #else
1673                                 error = EOPNOTSUPP;
1674 #endif
1675                                 break;
1676                         default:
1677                                 break;
1678                         }
1679                 }
1680         } else {                /* level == IPPROTO_IPV6 */
1681                 switch (op) {
1682                 case SOPT_SET:
1683                         switch (optname) {
1684                         case IPV6_2292PKTOPTIONS:
1685 #ifdef IPV6_PKTOPTIONS
1686                         case IPV6_PKTOPTIONS:
1687 #endif
1688                         {
1689                                 struct mbuf *m;
1690
1691                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1692                                         printf("ip6_ctloutput: mbuf limit hit\n");
1693                                         error = ENOBUFS;
1694                                         break;
1695                                 }
1696
1697                                 error = soopt_getm(sopt, &m); /* XXX */
1698                                 if (error != 0)
1699                                         break;
1700                                 error = soopt_mcopyin(sopt, m); /* XXX */
1701                                 if (error != 0)
1702                                         break;
1703                                 INP_WLOCK(inp);
1704                                 error = ip6_pcbopts(&inp->in6p_outputopts, m,
1705                                     so, sopt);
1706                                 INP_WUNLOCK(inp);
1707                                 m_freem(m); /* XXX */
1708                                 break;
1709                         }
1710
1711                         /*
1712                          * Use of some Hop-by-Hop options or some
1713                          * Destination options, might require special
1714                          * privilege.  That is, normal applications
1715                          * (without special privilege) might be forbidden
1716                          * from setting certain options in outgoing packets,
1717                          * and might never see certain options in received
1718                          * packets. [RFC 2292 Section 6]
1719                          * KAME specific note:
1720                          *  KAME prevents non-privileged users from sending or
1721                          *  receiving ANY hbh/dst options in order to avoid
1722                          *  overhead of parsing options in the kernel.
1723                          */
1724                         case IPV6_RECVHOPOPTS:
1725                         case IPV6_RECVDSTOPTS:
1726                         case IPV6_RECVRTHDRDSTOPTS:
1727                                 if (td != NULL) {
1728                                         error = priv_check(td,
1729                                             PRIV_NETINET_SETHDROPTS);
1730                                         if (error)
1731                                                 break;
1732                                 }
1733                                 /* FALLTHROUGH */
1734                         case IPV6_UNICAST_HOPS:
1735                         case IPV6_HOPLIMIT:
1736
1737                         case IPV6_RECVPKTINFO:
1738                         case IPV6_RECVHOPLIMIT:
1739                         case IPV6_RECVRTHDR:
1740                         case IPV6_RECVPATHMTU:
1741                         case IPV6_RECVTCLASS:
1742                         case IPV6_RECVFLOWID:
1743 #ifdef  RSS
1744                         case IPV6_RECVRSSBUCKETID:
1745 #endif
1746                         case IPV6_V6ONLY:
1747                         case IPV6_AUTOFLOWLABEL:
1748                         case IPV6_ORIGDSTADDR:
1749                         case IPV6_BINDANY:
1750                         case IPV6_BINDMULTI:
1751 #ifdef  RSS
1752                         case IPV6_RSS_LISTEN_BUCKET:
1753 #endif
1754                         case IPV6_VLAN_PCP:
1755                                 if (optname == IPV6_BINDANY && td != NULL) {
1756                                         error = priv_check(td,
1757                                             PRIV_NETINET_BINDANY);
1758                                         if (error)
1759                                                 break;
1760                                 }
1761
1762                                 if (optlen != sizeof(int)) {
1763                                         error = EINVAL;
1764                                         break;
1765                                 }
1766                                 error = sooptcopyin(sopt, &optval,
1767                                         sizeof optval, sizeof optval);
1768                                 if (error)
1769                                         break;
1770                                 switch (optname) {
1771                                 case IPV6_UNICAST_HOPS:
1772                                         if (optval < -1 || optval >= 256)
1773                                                 error = EINVAL;
1774                                         else {
1775                                                 /* -1 = kernel default */
1776                                                 inp->in6p_hops = optval;
1777                                                 if ((inp->inp_vflag &
1778                                                      INP_IPV4) != 0)
1779                                                         inp->inp_ip_ttl = optval;
1780                                         }
1781                                         break;
1782 #define OPTSET(bit) \
1783 do { \
1784         INP_WLOCK(inp); \
1785         if (optval) \
1786                 inp->inp_flags |= (bit); \
1787         else \
1788                 inp->inp_flags &= ~(bit); \
1789         INP_WUNLOCK(inp); \
1790 } while (/*CONSTCOND*/ 0)
1791 #define OPTSET2292(bit) \
1792 do { \
1793         INP_WLOCK(inp); \
1794         inp->inp_flags |= IN6P_RFC2292; \
1795         if (optval) \
1796                 inp->inp_flags |= (bit); \
1797         else \
1798                 inp->inp_flags &= ~(bit); \
1799         INP_WUNLOCK(inp); \
1800 } while (/*CONSTCOND*/ 0)
1801 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1802
1803 #define OPTSET2_N(bit, val) do {                                        \
1804         if (val)                                                        \
1805                 inp->inp_flags2 |= bit;                                 \
1806         else                                                            \
1807                 inp->inp_flags2 &= ~bit;                                \
1808 } while (0)
1809 #define OPTSET2(bit, val) do {                                          \
1810         INP_WLOCK(inp);                                                 \
1811         OPTSET2_N(bit, val);                                            \
1812         INP_WUNLOCK(inp);                                               \
1813 } while (0)
1814 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1815 #define OPTSET2292_EXCLUSIVE(bit)                                       \
1816 do {                                                                    \
1817         INP_WLOCK(inp);                                                 \
1818         if (OPTBIT(IN6P_RFC2292)) {                                     \
1819                 error = EINVAL;                                         \
1820         } else {                                                        \
1821                 if (optval)                                             \
1822                         inp->inp_flags |= (bit);                        \
1823                 else                                                    \
1824                         inp->inp_flags &= ~(bit);                       \
1825         }                                                               \
1826         INP_WUNLOCK(inp);                                               \
1827 } while (/*CONSTCOND*/ 0)
1828
1829                                 case IPV6_RECVPKTINFO:
1830                                         OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1831                                         break;
1832
1833                                 case IPV6_HOPLIMIT:
1834                                 {
1835                                         struct ip6_pktopts **optp;
1836
1837                                         /* cannot mix with RFC2292 */
1838                                         if (OPTBIT(IN6P_RFC2292)) {
1839                                                 error = EINVAL;
1840                                                 break;
1841                                         }
1842                                         INP_WLOCK(inp);
1843                                         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1844                                                 INP_WUNLOCK(inp);
1845                                                 return (ECONNRESET);
1846                                         }
1847                                         optp = &inp->in6p_outputopts;
1848                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1849                                             (u_char *)&optval, sizeof(optval),
1850                                             optp, (td != NULL) ? td->td_ucred :
1851                                             NULL, uproto);
1852                                         INP_WUNLOCK(inp);
1853                                         break;
1854                                 }
1855
1856                                 case IPV6_RECVHOPLIMIT:
1857                                         OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1858                                         break;
1859
1860                                 case IPV6_RECVHOPOPTS:
1861                                         OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1862                                         break;
1863
1864                                 case IPV6_RECVDSTOPTS:
1865                                         OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1866                                         break;
1867
1868                                 case IPV6_RECVRTHDRDSTOPTS:
1869                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1870                                         break;
1871
1872                                 case IPV6_RECVRTHDR:
1873                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1874                                         break;
1875
1876                                 case IPV6_RECVPATHMTU:
1877                                         /*
1878                                          * We ignore this option for TCP
1879                                          * sockets.
1880                                          * (RFC3542 leaves this case
1881                                          * unspecified.)
1882                                          */
1883                                         if (uproto != IPPROTO_TCP)
1884                                                 OPTSET(IN6P_MTU);
1885                                         break;
1886
1887                                 case IPV6_RECVFLOWID:
1888                                         OPTSET2(INP_RECVFLOWID, optval);
1889                                         break;
1890
1891 #ifdef  RSS
1892                                 case IPV6_RECVRSSBUCKETID:
1893                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1894                                         break;
1895 #endif
1896
1897                                 case IPV6_V6ONLY:
1898                                         INP_WLOCK(inp);
1899                                         if (inp->inp_lport ||
1900                                             !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1901                                                 /*
1902                                                  * The socket is already bound.
1903                                                  */
1904                                                 INP_WUNLOCK(inp);
1905                                                 error = EINVAL;
1906                                                 break;
1907                                         }
1908                                         if (optval) {
1909                                                 inp->inp_flags |= IN6P_IPV6_V6ONLY;
1910                                                 inp->inp_vflag &= ~INP_IPV4;
1911                                         } else {
1912                                                 inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1913                                                 inp->inp_vflag |= INP_IPV4;
1914                                         }
1915                                         INP_WUNLOCK(inp);
1916                                         break;
1917                                 case IPV6_RECVTCLASS:
1918                                         /* cannot mix with RFC2292 XXX */
1919                                         OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1920                                         break;
1921                                 case IPV6_AUTOFLOWLABEL:
1922                                         OPTSET(IN6P_AUTOFLOWLABEL);
1923                                         break;
1924
1925                                 case IPV6_ORIGDSTADDR:
1926                                         OPTSET2(INP_ORIGDSTADDR, optval);
1927                                         break;
1928                                 case IPV6_BINDANY:
1929                                         OPTSET(INP_BINDANY);
1930                                         break;
1931
1932                                 case IPV6_BINDMULTI:
1933                                         OPTSET2(INP_BINDMULTI, optval);
1934                                         break;
1935 #ifdef  RSS
1936                                 case IPV6_RSS_LISTEN_BUCKET:
1937                                         if ((optval >= 0) &&
1938                                             (optval < rss_getnumbuckets())) {
1939                                                 INP_WLOCK(inp);
1940                                                 inp->inp_rss_listen_bucket = optval;
1941                                                 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1942                                                 INP_WUNLOCK(inp);
1943                                         } else {
1944                                                 error = EINVAL;
1945                                         }
1946                                         break;
1947 #endif
1948                                 case IPV6_VLAN_PCP:
1949                                         if ((optval >= -1) && (optval <=
1950                                             (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1951                                                 if (optval == -1) {
1952                                                         INP_WLOCK(inp);
1953                                                         inp->inp_flags2 &=
1954                                                             ~(INP_2PCP_SET |
1955                                                             INP_2PCP_MASK);
1956                                                         INP_WUNLOCK(inp);
1957                                                 } else {
1958                                                         INP_WLOCK(inp);
1959                                                         inp->inp_flags2 |=
1960                                                             INP_2PCP_SET;
1961                                                         inp->inp_flags2 &=
1962                                                             ~INP_2PCP_MASK;
1963                                                         inp->inp_flags2 |=
1964                                                             optval <<
1965                                                             INP_2PCP_SHIFT;
1966                                                         INP_WUNLOCK(inp);
1967                                                 }
1968                                         } else
1969                                                 error = EINVAL;
1970                                         break;
1971                                 }
1972                                 break;
1973
1974                         case IPV6_TCLASS:
1975                         case IPV6_DONTFRAG:
1976                         case IPV6_USE_MIN_MTU:
1977                         case IPV6_PREFER_TEMPADDR:
1978                                 if (optlen != sizeof(optval)) {
1979                                         error = EINVAL;
1980                                         break;
1981                                 }
1982                                 error = sooptcopyin(sopt, &optval,
1983                                         sizeof optval, sizeof optval);
1984                                 if (error)
1985                                         break;
1986                                 {
1987                                         struct ip6_pktopts **optp;
1988                                         INP_WLOCK(inp);
1989                                         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1990                                                 INP_WUNLOCK(inp);
1991                                                 return (ECONNRESET);
1992                                         }
1993                                         optp = &inp->in6p_outputopts;
1994                                         error = ip6_pcbopt(optname,
1995                                             (u_char *)&optval, sizeof(optval),
1996                                             optp, (td != NULL) ? td->td_ucred :
1997                                             NULL, uproto);
1998                                         INP_WUNLOCK(inp);
1999                                         break;
2000                                 }
2001
2002                         case IPV6_2292PKTINFO:
2003                         case IPV6_2292HOPLIMIT:
2004                         case IPV6_2292HOPOPTS:
2005                         case IPV6_2292DSTOPTS:
2006                         case IPV6_2292RTHDR:
2007                                 /* RFC 2292 */
2008                                 if (optlen != sizeof(int)) {
2009                                         error = EINVAL;
2010                                         break;
2011                                 }
2012                                 error = sooptcopyin(sopt, &optval,
2013                                         sizeof optval, sizeof optval);
2014                                 if (error)
2015                                         break;
2016                                 switch (optname) {
2017                                 case IPV6_2292PKTINFO:
2018                                         OPTSET2292(IN6P_PKTINFO);
2019                                         break;
2020                                 case IPV6_2292HOPLIMIT:
2021                                         OPTSET2292(IN6P_HOPLIMIT);
2022                                         break;
2023                                 case IPV6_2292HOPOPTS:
2024                                         /*
2025                                          * Check super-user privilege.
2026                                          * See comments for IPV6_RECVHOPOPTS.
2027                                          */
2028                                         if (td != NULL) {
2029                                                 error = priv_check(td,
2030                                                     PRIV_NETINET_SETHDROPTS);
2031                                                 if (error)
2032                                                         return (error);
2033                                         }
2034                                         OPTSET2292(IN6P_HOPOPTS);
2035                                         break;
2036                                 case IPV6_2292DSTOPTS:
2037                                         if (td != NULL) {
2038                                                 error = priv_check(td,
2039                                                     PRIV_NETINET_SETHDROPTS);
2040                                                 if (error)
2041                                                         return (error);
2042                                         }
2043                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2044                                         break;
2045                                 case IPV6_2292RTHDR:
2046                                         OPTSET2292(IN6P_RTHDR);
2047                                         break;
2048                                 }
2049                                 break;
2050                         case IPV6_PKTINFO:
2051                         case IPV6_HOPOPTS:
2052                         case IPV6_RTHDR:
2053                         case IPV6_DSTOPTS:
2054                         case IPV6_RTHDRDSTOPTS:
2055                         case IPV6_NEXTHOP:
2056                         {
2057                                 /* new advanced API (RFC3542) */
2058                                 u_char *optbuf;
2059                                 u_char optbuf_storage[MCLBYTES];
2060                                 int optlen;
2061                                 struct ip6_pktopts **optp;
2062
2063                                 /* cannot mix with RFC2292 */
2064                                 if (OPTBIT(IN6P_RFC2292)) {
2065                                         error = EINVAL;
2066                                         break;
2067                                 }
2068
2069                                 /*
2070                                  * We only ensure valsize is not too large
2071                                  * here.  Further validation will be done
2072                                  * later.
2073                                  */
2074                                 error = sooptcopyin(sopt, optbuf_storage,
2075                                     sizeof(optbuf_storage), 0);
2076                                 if (error)
2077                                         break;
2078                                 optlen = sopt->sopt_valsize;
2079                                 optbuf = optbuf_storage;
2080                                 INP_WLOCK(inp);
2081                                 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2082                                         INP_WUNLOCK(inp);
2083                                         return (ECONNRESET);
2084                                 }
2085                                 optp = &inp->in6p_outputopts;
2086                                 error = ip6_pcbopt(optname, optbuf, optlen,
2087                                     optp, (td != NULL) ? td->td_ucred : NULL,
2088                                     uproto);
2089                                 INP_WUNLOCK(inp);
2090                                 break;
2091                         }
2092 #undef OPTSET
2093
2094                         case IPV6_MULTICAST_IF:
2095                         case IPV6_MULTICAST_HOPS:
2096                         case IPV6_MULTICAST_LOOP:
2097                         case IPV6_JOIN_GROUP:
2098                         case IPV6_LEAVE_GROUP:
2099                         case IPV6_MSFILTER:
2100                         case MCAST_BLOCK_SOURCE:
2101                         case MCAST_UNBLOCK_SOURCE:
2102                         case MCAST_JOIN_GROUP:
2103                         case MCAST_LEAVE_GROUP:
2104                         case MCAST_JOIN_SOURCE_GROUP:
2105                         case MCAST_LEAVE_SOURCE_GROUP:
2106                                 error = ip6_setmoptions(inp, sopt);
2107                                 break;
2108
2109                         case IPV6_PORTRANGE:
2110                                 error = sooptcopyin(sopt, &optval,
2111                                     sizeof optval, sizeof optval);
2112                                 if (error)
2113                                         break;
2114
2115                                 INP_WLOCK(inp);
2116                                 switch (optval) {
2117                                 case IPV6_PORTRANGE_DEFAULT:
2118                                         inp->inp_flags &= ~(INP_LOWPORT);
2119                                         inp->inp_flags &= ~(INP_HIGHPORT);
2120                                         break;
2121
2122                                 case IPV6_PORTRANGE_HIGH:
2123                                         inp->inp_flags &= ~(INP_LOWPORT);
2124                                         inp->inp_flags |= INP_HIGHPORT;
2125                                         break;
2126
2127                                 case IPV6_PORTRANGE_LOW:
2128                                         inp->inp_flags &= ~(INP_HIGHPORT);
2129                                         inp->inp_flags |= INP_LOWPORT;
2130                                         break;
2131
2132                                 default:
2133                                         error = EINVAL;
2134                                         break;
2135                                 }
2136                                 INP_WUNLOCK(inp);
2137                                 break;
2138
2139 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2140                         case IPV6_IPSEC_POLICY:
2141                                 if (IPSEC_ENABLED(ipv6)) {
2142                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2143                                         break;
2144                                 }
2145                                 /* FALLTHROUGH */
2146 #endif /* IPSEC */
2147
2148                         default:
2149                                 error = ENOPROTOOPT;
2150                                 break;
2151                         }
2152                         break;
2153
2154                 case SOPT_GET:
2155                         switch (optname) {
2156                         case IPV6_2292PKTOPTIONS:
2157 #ifdef IPV6_PKTOPTIONS
2158                         case IPV6_PKTOPTIONS:
2159 #endif
2160                                 /*
2161                                  * RFC3542 (effectively) deprecated the
2162                                  * semantics of the 2292-style pktoptions.
2163                                  * Since it was not reliable in nature (i.e.,
2164                                  * applications had to expect the lack of some
2165                                  * information after all), it would make sense
2166                                  * to simplify this part by always returning
2167                                  * empty data.
2168                                  */
2169                                 sopt->sopt_valsize = 0;
2170                                 break;
2171
2172                         case IPV6_RECVHOPOPTS:
2173                         case IPV6_RECVDSTOPTS:
2174                         case IPV6_RECVRTHDRDSTOPTS:
2175                         case IPV6_UNICAST_HOPS:
2176                         case IPV6_RECVPKTINFO:
2177                         case IPV6_RECVHOPLIMIT:
2178                         case IPV6_RECVRTHDR:
2179                         case IPV6_RECVPATHMTU:
2180
2181                         case IPV6_V6ONLY:
2182                         case IPV6_PORTRANGE:
2183                         case IPV6_RECVTCLASS:
2184                         case IPV6_AUTOFLOWLABEL:
2185                         case IPV6_BINDANY:
2186                         case IPV6_FLOWID:
2187                         case IPV6_FLOWTYPE:
2188                         case IPV6_RECVFLOWID:
2189 #ifdef  RSS
2190                         case IPV6_RSSBUCKETID:
2191                         case IPV6_RECVRSSBUCKETID:
2192 #endif
2193                         case IPV6_BINDMULTI:
2194                         case IPV6_VLAN_PCP:
2195                                 switch (optname) {
2196                                 case IPV6_RECVHOPOPTS:
2197                                         optval = OPTBIT(IN6P_HOPOPTS);
2198                                         break;
2199
2200                                 case IPV6_RECVDSTOPTS:
2201                                         optval = OPTBIT(IN6P_DSTOPTS);
2202                                         break;
2203
2204                                 case IPV6_RECVRTHDRDSTOPTS:
2205                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2206                                         break;
2207
2208                                 case IPV6_UNICAST_HOPS:
2209                                         optval = inp->in6p_hops;
2210                                         break;
2211
2212                                 case IPV6_RECVPKTINFO:
2213                                         optval = OPTBIT(IN6P_PKTINFO);
2214                                         break;
2215
2216                                 case IPV6_RECVHOPLIMIT:
2217                                         optval = OPTBIT(IN6P_HOPLIMIT);
2218                                         break;
2219
2220                                 case IPV6_RECVRTHDR:
2221                                         optval = OPTBIT(IN6P_RTHDR);
2222                                         break;
2223
2224                                 case IPV6_RECVPATHMTU:
2225                                         optval = OPTBIT(IN6P_MTU);
2226                                         break;
2227
2228                                 case IPV6_V6ONLY:
2229                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2230                                         break;
2231
2232                                 case IPV6_PORTRANGE:
2233                                     {
2234                                         int flags;
2235                                         flags = inp->inp_flags;
2236                                         if (flags & INP_HIGHPORT)
2237                                                 optval = IPV6_PORTRANGE_HIGH;
2238                                         else if (flags & INP_LOWPORT)
2239                                                 optval = IPV6_PORTRANGE_LOW;
2240                                         else
2241                                                 optval = 0;
2242                                         break;
2243                                     }
2244                                 case IPV6_RECVTCLASS:
2245                                         optval = OPTBIT(IN6P_TCLASS);
2246                                         break;
2247
2248                                 case IPV6_AUTOFLOWLABEL:
2249                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2250                                         break;
2251
2252                                 case IPV6_ORIGDSTADDR:
2253                                         optval = OPTBIT2(INP_ORIGDSTADDR);
2254                                         break;
2255
2256                                 case IPV6_BINDANY:
2257                                         optval = OPTBIT(INP_BINDANY);
2258                                         break;
2259
2260                                 case IPV6_FLOWID:
2261                                         optval = inp->inp_flowid;
2262                                         break;
2263
2264                                 case IPV6_FLOWTYPE:
2265                                         optval = inp->inp_flowtype;
2266                                         break;
2267
2268                                 case IPV6_RECVFLOWID:
2269                                         optval = OPTBIT2(INP_RECVFLOWID);
2270                                         break;
2271 #ifdef  RSS
2272                                 case IPV6_RSSBUCKETID:
2273                                         retval =
2274                                             rss_hash2bucket(inp->inp_flowid,
2275                                             inp->inp_flowtype,
2276                                             &rss_bucket);
2277                                         if (retval == 0)
2278                                                 optval = rss_bucket;
2279                                         else
2280                                                 error = EINVAL;
2281                                         break;
2282
2283                                 case IPV6_RECVRSSBUCKETID:
2284                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2285                                         break;
2286 #endif
2287
2288                                 case IPV6_BINDMULTI:
2289                                         optval = OPTBIT2(INP_BINDMULTI);
2290                                         break;
2291
2292                                 case IPV6_VLAN_PCP:
2293                                         if (OPTBIT2(INP_2PCP_SET)) {
2294                                                 optval = (inp->inp_flags2 &
2295                                                             INP_2PCP_MASK) >>
2296                                                             INP_2PCP_SHIFT;
2297                                         } else {
2298                                                 optval = -1;
2299                                         }
2300                                         break;
2301                                 }
2302
2303                                 if (error)
2304                                         break;
2305                                 error = sooptcopyout(sopt, &optval,
2306                                         sizeof optval);
2307                                 break;
2308
2309                         case IPV6_PATHMTU:
2310                         {
2311                                 u_long pmtu = 0;
2312                                 struct ip6_mtuinfo mtuinfo;
2313                                 struct in6_addr addr;
2314
2315                                 if (!(so->so_state & SS_ISCONNECTED))
2316                                         return (ENOTCONN);
2317                                 /*
2318                                  * XXX: we dot not consider the case of source
2319                                  * routing, or optional information to specify
2320                                  * the outgoing interface.
2321                                  * Copy faddr out of inp to avoid holding lock
2322                                  * on inp during route lookup.
2323                                  */
2324                                 INP_RLOCK(inp);
2325                                 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2326                                 INP_RUNLOCK(inp);
2327                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2328                                     &addr, &pmtu);
2329                                 if (error)
2330                                         break;
2331                                 if (pmtu > IPV6_MAXPACKET)
2332                                         pmtu = IPV6_MAXPACKET;
2333
2334                                 bzero(&mtuinfo, sizeof(mtuinfo));
2335                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2336                                 optdata = (void *)&mtuinfo;
2337                                 optdatalen = sizeof(mtuinfo);
2338                                 error = sooptcopyout(sopt, optdata,
2339                                     optdatalen);
2340                                 break;
2341                         }
2342
2343                         case IPV6_2292PKTINFO:
2344                         case IPV6_2292HOPLIMIT:
2345                         case IPV6_2292HOPOPTS:
2346                         case IPV6_2292RTHDR:
2347                         case IPV6_2292DSTOPTS:
2348                                 switch (optname) {
2349                                 case IPV6_2292PKTINFO:
2350                                         optval = OPTBIT(IN6P_PKTINFO);
2351                                         break;
2352                                 case IPV6_2292HOPLIMIT:
2353                                         optval = OPTBIT(IN6P_HOPLIMIT);
2354                                         break;
2355                                 case IPV6_2292HOPOPTS:
2356                                         optval = OPTBIT(IN6P_HOPOPTS);
2357                                         break;
2358                                 case IPV6_2292RTHDR:
2359                                         optval = OPTBIT(IN6P_RTHDR);
2360                                         break;
2361                                 case IPV6_2292DSTOPTS:
2362                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2363                                         break;
2364                                 }
2365                                 error = sooptcopyout(sopt, &optval,
2366                                     sizeof optval);
2367                                 break;
2368                         case IPV6_PKTINFO:
2369                         case IPV6_HOPOPTS:
2370                         case IPV6_RTHDR:
2371                         case IPV6_DSTOPTS:
2372                         case IPV6_RTHDRDSTOPTS:
2373                         case IPV6_NEXTHOP:
2374                         case IPV6_TCLASS:
2375                         case IPV6_DONTFRAG:
2376                         case IPV6_USE_MIN_MTU:
2377                         case IPV6_PREFER_TEMPADDR:
2378                                 error = ip6_getpcbopt(inp, optname, sopt);
2379                                 break;
2380
2381                         case IPV6_MULTICAST_IF:
2382                         case IPV6_MULTICAST_HOPS:
2383                         case IPV6_MULTICAST_LOOP:
2384                         case IPV6_MSFILTER:
2385                                 error = ip6_getmoptions(inp, sopt);
2386                                 break;
2387
2388 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2389                         case IPV6_IPSEC_POLICY:
2390                                 if (IPSEC_ENABLED(ipv6)) {
2391                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2392                                         break;
2393                                 }
2394                                 /* FALLTHROUGH */
2395 #endif /* IPSEC */
2396                         default:
2397                                 error = ENOPROTOOPT;
2398                                 break;
2399                         }
2400                         break;
2401                 }
2402         }
2403         return (error);
2404 }
2405
2406 int
2407 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2408 {
2409         int error = 0, optval, optlen;
2410         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2411         struct inpcb *inp = sotoinpcb(so);
2412         int level, op, optname;
2413
2414         level = sopt->sopt_level;
2415         op = sopt->sopt_dir;
2416         optname = sopt->sopt_name;
2417         optlen = sopt->sopt_valsize;
2418
2419         if (level != IPPROTO_IPV6) {
2420                 return (EINVAL);
2421         }
2422
2423         switch (optname) {
2424         case IPV6_CHECKSUM:
2425                 /*
2426                  * For ICMPv6 sockets, no modification allowed for checksum
2427                  * offset, permit "no change" values to help existing apps.
2428                  *
2429                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2430                  * for an ICMPv6 socket will fail."
2431                  * The current behavior does not meet RFC3542.
2432                  */
2433                 switch (op) {
2434                 case SOPT_SET:
2435                         if (optlen != sizeof(int)) {
2436                                 error = EINVAL;
2437                                 break;
2438                         }
2439                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2440                                             sizeof(optval));
2441                         if (error)
2442                                 break;
2443                         if (optval < -1 || (optval % 2) != 0) {
2444                                 /*
2445                                  * The API assumes non-negative even offset
2446                                  * values or -1 as a special value.
2447                                  */
2448                                 error = EINVAL;
2449                         } else if (so->so_proto->pr_protocol ==
2450                             IPPROTO_ICMPV6) {
2451                                 if (optval != icmp6off)
2452                                         error = EINVAL;
2453                         } else
2454                                 inp->in6p_cksum = optval;
2455                         break;
2456
2457                 case SOPT_GET:
2458                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2459                                 optval = icmp6off;
2460                         else
2461                                 optval = inp->in6p_cksum;
2462
2463                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2464                         break;
2465
2466                 default:
2467                         error = EINVAL;
2468                         break;
2469                 }
2470                 break;
2471
2472         default:
2473                 error = ENOPROTOOPT;
2474                 break;
2475         }
2476
2477         return (error);
2478 }
2479
2480 /*
2481  * Set up IP6 options in pcb for insertion in output packets or
2482  * specifying behavior of outgoing packets.
2483  */
2484 static int
2485 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2486     struct socket *so, struct sockopt *sopt)
2487 {
2488         struct ip6_pktopts *opt = *pktopt;
2489         int error = 0;
2490         struct thread *td = sopt->sopt_td;
2491         struct epoch_tracker et;
2492
2493         /* turn off any old options. */
2494         if (opt) {
2495 #ifdef DIAGNOSTIC
2496                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2497                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2498                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2499                         printf("ip6_pcbopts: all specified options are cleared.\n");
2500 #endif
2501                 ip6_clearpktopts(opt, -1);
2502         } else {
2503                 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2504                 if (opt == NULL)
2505                         return (ENOMEM);
2506         }
2507         *pktopt = NULL;
2508
2509         if (!m || m->m_len == 0) {
2510                 /*
2511                  * Only turning off any previous options, regardless of
2512                  * whether the opt is just created or given.
2513                  */
2514                 free(opt, M_IP6OPT);
2515                 return (0);
2516         }
2517
2518         /*  set options specified by user. */
2519         NET_EPOCH_ENTER(et);
2520         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2521             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2522                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2523                 free(opt, M_IP6OPT);
2524                 NET_EPOCH_EXIT(et);
2525                 return (error);
2526         }
2527         NET_EPOCH_EXIT(et);
2528         *pktopt = opt;
2529         return (0);
2530 }
2531
2532 /*
2533  * initialize ip6_pktopts.  beware that there are non-zero default values in
2534  * the struct.
2535  */
2536 void
2537 ip6_initpktopts(struct ip6_pktopts *opt)
2538 {
2539
2540         bzero(opt, sizeof(*opt));
2541         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2542         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2543         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2544         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2545 }
2546
2547 static int
2548 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2549     struct ucred *cred, int uproto)
2550 {
2551         struct epoch_tracker et;
2552         struct ip6_pktopts *opt;
2553         int ret;
2554
2555         if (*pktopt == NULL) {
2556                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2557                     M_NOWAIT);
2558                 if (*pktopt == NULL)
2559                         return (ENOBUFS);
2560                 ip6_initpktopts(*pktopt);
2561         }
2562         opt = *pktopt;
2563
2564         NET_EPOCH_ENTER(et);
2565         ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2566         NET_EPOCH_EXIT(et);
2567
2568         return (ret);
2569 }
2570
2571 #define GET_PKTOPT_VAR(field, lenexpr) do {                                     \
2572         if (pktopt && pktopt->field) {                                          \
2573                 INP_RUNLOCK(inp);                                               \
2574                 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);         \
2575                 malloc_optdata = true;                                          \
2576                 INP_RLOCK(inp);                                                 \
2577                 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {            \
2578                         INP_RUNLOCK(inp);                                       \
2579                         free(optdata, M_TEMP);                                  \
2580                         return (ECONNRESET);                                    \
2581                 }                                                               \
2582                 pktopt = inp->in6p_outputopts;                                  \
2583                 if (pktopt && pktopt->field) {                                  \
2584                         optdatalen = min(lenexpr, sopt->sopt_valsize);          \
2585                         bcopy(pktopt->field, optdata, optdatalen);              \
2586                 } else {                                                        \
2587                         free(optdata, M_TEMP);                                  \
2588                         optdata = NULL;                                         \
2589                         malloc_optdata = false;                                 \
2590                 }                                                               \
2591         }                                                                       \
2592 } while(0)
2593
2594 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,                         \
2595         (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2596
2597 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,                        \
2598         pktopt->field->sa_len)
2599
2600 static int
2601 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2602 {
2603         void *optdata = NULL;
2604         bool malloc_optdata = false;
2605         int optdatalen = 0;
2606         int error = 0;
2607         struct in6_pktinfo null_pktinfo;
2608         int deftclass = 0, on;
2609         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2610         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2611         struct ip6_pktopts *pktopt;
2612
2613         INP_RLOCK(inp);
2614         pktopt = inp->in6p_outputopts;
2615
2616         switch (optname) {
2617         case IPV6_PKTINFO:
2618                 optdata = (void *)&null_pktinfo;
2619                 if (pktopt && pktopt->ip6po_pktinfo) {
2620                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2621                             sizeof(null_pktinfo));
2622                         in6_clearscope(&null_pktinfo.ipi6_addr);
2623                 } else {
2624                         /* XXX: we don't have to do this every time... */
2625                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2626                 }
2627                 optdatalen = sizeof(struct in6_pktinfo);
2628                 break;
2629         case IPV6_TCLASS:
2630                 if (pktopt && pktopt->ip6po_tclass >= 0)
2631                         deftclass = pktopt->ip6po_tclass;
2632                 optdata = (void *)&deftclass;
2633                 optdatalen = sizeof(int);
2634                 break;
2635         case IPV6_HOPOPTS:
2636                 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2637                 break;
2638         case IPV6_RTHDR:
2639                 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2640                 break;
2641         case IPV6_RTHDRDSTOPTS:
2642                 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2643                 break;
2644         case IPV6_DSTOPTS:
2645                 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2646                 break;
2647         case IPV6_NEXTHOP:
2648                 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2649                 break;
2650         case IPV6_USE_MIN_MTU:
2651                 if (pktopt)
2652                         defminmtu = pktopt->ip6po_minmtu;
2653                 optdata = (void *)&defminmtu;
2654                 optdatalen = sizeof(int);
2655                 break;
2656         case IPV6_DONTFRAG:
2657                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2658                         on = 1;
2659                 else
2660                         on = 0;
2661                 optdata = (void *)&on;
2662                 optdatalen = sizeof(on);
2663                 break;
2664         case IPV6_PREFER_TEMPADDR:
2665                 if (pktopt)
2666                         defpreftemp = pktopt->ip6po_prefer_tempaddr;
2667                 optdata = (void *)&defpreftemp;
2668                 optdatalen = sizeof(int);
2669                 break;
2670         default:                /* should not happen */
2671 #ifdef DIAGNOSTIC
2672                 panic("ip6_getpcbopt: unexpected option\n");
2673 #endif
2674                 INP_RUNLOCK(inp);
2675                 return (ENOPROTOOPT);
2676         }
2677         INP_RUNLOCK(inp);
2678
2679         error = sooptcopyout(sopt, optdata, optdatalen);
2680         if (malloc_optdata)
2681                 free(optdata, M_TEMP);
2682
2683         return (error);
2684 }
2685
2686 void
2687 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2688 {
2689         if (pktopt == NULL)
2690                 return;
2691
2692         if (optname == -1 || optname == IPV6_PKTINFO) {
2693                 if (pktopt->ip6po_pktinfo)
2694                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2695                 pktopt->ip6po_pktinfo = NULL;
2696         }
2697         if (optname == -1 || optname == IPV6_HOPLIMIT)
2698                 pktopt->ip6po_hlim = -1;
2699         if (optname == -1 || optname == IPV6_TCLASS)
2700                 pktopt->ip6po_tclass = -1;
2701         if (optname == -1 || optname == IPV6_NEXTHOP) {
2702                 if (pktopt->ip6po_nextroute.ro_nh) {
2703                         NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2704                         pktopt->ip6po_nextroute.ro_nh = NULL;
2705                 }
2706                 if (pktopt->ip6po_nexthop)
2707                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2708                 pktopt->ip6po_nexthop = NULL;
2709         }
2710         if (optname == -1 || optname == IPV6_HOPOPTS) {
2711                 if (pktopt->ip6po_hbh)
2712                         free(pktopt->ip6po_hbh, M_IP6OPT);
2713                 pktopt->ip6po_hbh = NULL;
2714         }
2715         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2716                 if (pktopt->ip6po_dest1)
2717                         free(pktopt->ip6po_dest1, M_IP6OPT);
2718                 pktopt->ip6po_dest1 = NULL;
2719         }
2720         if (optname == -1 || optname == IPV6_RTHDR) {
2721                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2722                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2723                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2724                 if (pktopt->ip6po_route.ro_nh) {
2725                         NH_FREE(pktopt->ip6po_route.ro_nh);
2726                         pktopt->ip6po_route.ro_nh = NULL;
2727                 }
2728         }
2729         if (optname == -1 || optname == IPV6_DSTOPTS) {
2730                 if (pktopt->ip6po_dest2)
2731                         free(pktopt->ip6po_dest2, M_IP6OPT);
2732                 pktopt->ip6po_dest2 = NULL;
2733         }
2734 }
2735
2736 #define PKTOPT_EXTHDRCPY(type) \
2737 do {\
2738         if (src->type) {\
2739                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2740                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2741                 if (dst->type == NULL)\
2742                         goto bad;\
2743                 bcopy(src->type, dst->type, hlen);\
2744         }\
2745 } while (/*CONSTCOND*/ 0)
2746
2747 static int
2748 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2749 {
2750         if (dst == NULL || src == NULL)  {
2751                 printf("ip6_clearpktopts: invalid argument\n");
2752                 return (EINVAL);
2753         }
2754
2755         dst->ip6po_hlim = src->ip6po_hlim;
2756         dst->ip6po_tclass = src->ip6po_tclass;
2757         dst->ip6po_flags = src->ip6po_flags;
2758         dst->ip6po_minmtu = src->ip6po_minmtu;
2759         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2760         if (src->ip6po_pktinfo) {
2761                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2762                     M_IP6OPT, canwait);
2763                 if (dst->ip6po_pktinfo == NULL)
2764                         goto bad;
2765                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2766         }
2767         if (src->ip6po_nexthop) {
2768                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2769                     M_IP6OPT, canwait);
2770                 if (dst->ip6po_nexthop == NULL)
2771                         goto bad;
2772                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2773                     src->ip6po_nexthop->sa_len);
2774         }
2775         PKTOPT_EXTHDRCPY(ip6po_hbh);
2776         PKTOPT_EXTHDRCPY(ip6po_dest1);
2777         PKTOPT_EXTHDRCPY(ip6po_dest2);
2778         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2779         return (0);
2780
2781   bad:
2782         ip6_clearpktopts(dst, -1);
2783         return (ENOBUFS);
2784 }
2785 #undef PKTOPT_EXTHDRCPY
2786
2787 struct ip6_pktopts *
2788 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2789 {
2790         int error;
2791         struct ip6_pktopts *dst;
2792
2793         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2794         if (dst == NULL)
2795                 return (NULL);
2796         ip6_initpktopts(dst);
2797
2798         if ((error = copypktopts(dst, src, canwait)) != 0) {
2799                 free(dst, M_IP6OPT);
2800                 return (NULL);
2801         }
2802
2803         return (dst);
2804 }
2805
2806 void
2807 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2808 {
2809         if (pktopt == NULL)
2810                 return;
2811
2812         ip6_clearpktopts(pktopt, -1);
2813
2814         free(pktopt, M_IP6OPT);
2815 }
2816
2817 /*
2818  * Set IPv6 outgoing packet options based on advanced API.
2819  */
2820 int
2821 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2822     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2823 {
2824         struct cmsghdr *cm = NULL;
2825
2826         if (control == NULL || opt == NULL)
2827                 return (EINVAL);
2828
2829         /*
2830          * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2831          * are in the network epoch here.
2832          */
2833         NET_EPOCH_ASSERT();
2834
2835         ip6_initpktopts(opt);
2836         if (stickyopt) {
2837                 int error;
2838
2839                 /*
2840                  * If stickyopt is provided, make a local copy of the options
2841                  * for this particular packet, then override them by ancillary
2842                  * objects.
2843                  * XXX: copypktopts() does not copy the cached route to a next
2844                  * hop (if any).  This is not very good in terms of efficiency,
2845                  * but we can allow this since this option should be rarely
2846                  * used.
2847                  */
2848                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2849                         return (error);
2850         }
2851
2852         /*
2853          * XXX: Currently, we assume all the optional information is stored
2854          * in a single mbuf.
2855          */
2856         if (control->m_next)
2857                 return (EINVAL);
2858
2859         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2860             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2861                 int error;
2862
2863                 if (control->m_len < CMSG_LEN(0))
2864                         return (EINVAL);
2865
2866                 cm = mtod(control, struct cmsghdr *);
2867                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2868                         return (EINVAL);
2869                 if (cm->cmsg_level != IPPROTO_IPV6)
2870                         continue;
2871
2872                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2873                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2874                 if (error)
2875                         return (error);
2876         }
2877
2878         return (0);
2879 }
2880
2881 /*
2882  * Set a particular packet option, as a sticky option or an ancillary data
2883  * item.  "len" can be 0 only when it's a sticky option.
2884  * We have 4 cases of combination of "sticky" and "cmsg":
2885  * "sticky=0, cmsg=0": impossible
2886  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2887  * "sticky=1, cmsg=0": RFC3542 socket option
2888  * "sticky=1, cmsg=1": RFC2292 socket option
2889  */
2890 static int
2891 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2892     struct ucred *cred, int sticky, int cmsg, int uproto)
2893 {
2894         int minmtupolicy, preftemp;
2895         int error;
2896
2897         NET_EPOCH_ASSERT();
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) {
2973                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2974                         if (ifp == NULL)
2975                                 return (ENXIO);
2976                 }
2977                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2978                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2979                         return (ENETDOWN);
2980
2981                 if (ifp != NULL &&
2982                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2983                         struct in6_ifaddr *ia;
2984
2985                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2986                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2987                         if (ia == NULL)
2988                                 return (EADDRNOTAVAIL);
2989                         ifa_free(&ia->ia_ifa);
2990                 }
2991                 /*
2992                  * We store the address anyway, and let in6_selectsrc()
2993                  * validate the specified address.  This is because ipi6_addr
2994                  * may not have enough information about its scope zone, and
2995                  * we may need additional information (such as outgoing
2996                  * interface or the scope zone of a destination address) to
2997                  * disambiguate the scope.
2998                  * XXX: the delay of the validation may confuse the
2999                  * application when it is used as a sticky option.
3000                  */
3001                 if (opt->ip6po_pktinfo == NULL) {
3002                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3003                             M_IP6OPT, M_NOWAIT);
3004                         if (opt->ip6po_pktinfo == NULL)
3005                                 return (ENOBUFS);
3006                 }
3007                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3008                 break;
3009         }
3010
3011         case IPV6_2292HOPLIMIT:
3012         case IPV6_HOPLIMIT:
3013         {
3014                 int *hlimp;
3015
3016                 /*
3017                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3018                  * to simplify the ordering among hoplimit options.
3019                  */
3020                 if (optname == IPV6_HOPLIMIT && sticky)
3021                         return (ENOPROTOOPT);
3022
3023                 if (len != sizeof(int))
3024                         return (EINVAL);
3025                 hlimp = (int *)buf;
3026                 if (*hlimp < -1 || *hlimp > 255)
3027                         return (EINVAL);
3028
3029                 opt->ip6po_hlim = *hlimp;
3030                 break;
3031         }
3032
3033         case IPV6_TCLASS:
3034         {
3035                 int tclass;
3036
3037                 if (len != sizeof(int))
3038                         return (EINVAL);
3039                 tclass = *(int *)buf;
3040                 if (tclass < -1 || tclass > 255)
3041                         return (EINVAL);
3042
3043                 opt->ip6po_tclass = tclass;
3044                 break;
3045         }
3046
3047         case IPV6_2292NEXTHOP:
3048         case IPV6_NEXTHOP:
3049                 if (cred != NULL) {
3050                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3051                         if (error)
3052                                 return (error);
3053                 }
3054
3055                 if (len == 0) { /* just remove the option */
3056                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3057                         break;
3058                 }
3059
3060                 /* check if cmsg_len is large enough for sa_len */
3061                 if (len < sizeof(struct sockaddr) || len < *buf)
3062                         return (EINVAL);
3063
3064                 switch (((struct sockaddr *)buf)->sa_family) {
3065                 case AF_INET6:
3066                 {
3067                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3068                         int error;
3069
3070                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3071                                 return (EINVAL);
3072
3073                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3074                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3075                                 return (EINVAL);
3076                         }
3077                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3078                             != 0) {
3079                                 return (error);
3080                         }
3081                         break;
3082                 }
3083                 case AF_LINK:   /* should eventually be supported */
3084                 default:
3085                         return (EAFNOSUPPORT);
3086                 }
3087
3088                 /* turn off the previous option, then set the new option. */
3089                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3090                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3091                 if (opt->ip6po_nexthop == NULL)
3092                         return (ENOBUFS);
3093                 bcopy(buf, opt->ip6po_nexthop, *buf);
3094                 break;
3095
3096         case IPV6_2292HOPOPTS:
3097         case IPV6_HOPOPTS:
3098         {
3099                 struct ip6_hbh *hbh;
3100                 int hbhlen;
3101
3102                 /*
3103                  * XXX: We don't allow a non-privileged user to set ANY HbH
3104                  * options, since per-option restriction has too much
3105                  * overhead.
3106                  */
3107                 if (cred != NULL) {
3108                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3109                         if (error)
3110                                 return (error);
3111                 }
3112
3113                 if (len == 0) {
3114                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3115                         break;  /* just remove the option */
3116                 }
3117
3118                 /* message length validation */
3119                 if (len < sizeof(struct ip6_hbh))
3120                         return (EINVAL);
3121                 hbh = (struct ip6_hbh *)buf;
3122                 hbhlen = (hbh->ip6h_len + 1) << 3;
3123                 if (len != hbhlen)
3124                         return (EINVAL);
3125
3126                 /* turn off the previous option, then set the new option. */
3127                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3128                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3129                 if (opt->ip6po_hbh == NULL)
3130                         return (ENOBUFS);
3131                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3132
3133                 break;
3134         }
3135
3136         case IPV6_2292DSTOPTS:
3137         case IPV6_DSTOPTS:
3138         case IPV6_RTHDRDSTOPTS:
3139         {
3140                 struct ip6_dest *dest, **newdest = NULL;
3141                 int destlen;
3142
3143                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3144                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3145                         if (error)
3146                                 return (error);
3147                 }
3148
3149                 if (len == 0) {
3150                         ip6_clearpktopts(opt, optname);
3151                         break;  /* just remove the option */
3152                 }
3153
3154                 /* message length validation */
3155                 if (len < sizeof(struct ip6_dest))
3156                         return (EINVAL);
3157                 dest = (struct ip6_dest *)buf;
3158                 destlen = (dest->ip6d_len + 1) << 3;
3159                 if (len != destlen)
3160                         return (EINVAL);
3161
3162                 /*
3163                  * Determine the position that the destination options header
3164                  * should be inserted; before or after the routing header.
3165                  */
3166                 switch (optname) {
3167                 case IPV6_2292DSTOPTS:
3168                         /*
3169                          * The old advacned API is ambiguous on this point.
3170                          * Our approach is to determine the position based
3171                          * according to the existence of a routing header.
3172                          * Note, however, that this depends on the order of the
3173                          * extension headers in the ancillary data; the 1st
3174                          * part of the destination options header must appear
3175                          * before the routing header in the ancillary data,
3176                          * too.
3177                          * RFC3542 solved the ambiguity by introducing
3178                          * separate ancillary data or option types.
3179                          */
3180                         if (opt->ip6po_rthdr == NULL)
3181                                 newdest = &opt->ip6po_dest1;
3182                         else
3183                                 newdest = &opt->ip6po_dest2;
3184                         break;
3185                 case IPV6_RTHDRDSTOPTS:
3186                         newdest = &opt->ip6po_dest1;
3187                         break;
3188                 case IPV6_DSTOPTS:
3189                         newdest = &opt->ip6po_dest2;
3190                         break;
3191                 }
3192
3193                 /* turn off the previous option, then set the new option. */
3194                 ip6_clearpktopts(opt, optname);
3195                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3196                 if (*newdest == NULL)
3197                         return (ENOBUFS);
3198                 bcopy(dest, *newdest, destlen);
3199
3200                 break;
3201         }
3202
3203         case IPV6_2292RTHDR:
3204         case IPV6_RTHDR:
3205         {
3206                 struct ip6_rthdr *rth;
3207                 int rthlen;
3208
3209                 if (len == 0) {
3210                         ip6_clearpktopts(opt, IPV6_RTHDR);
3211                         break;  /* just remove the option */
3212                 }
3213
3214                 /* message length validation */
3215                 if (len < sizeof(struct ip6_rthdr))
3216                         return (EINVAL);
3217                 rth = (struct ip6_rthdr *)buf;
3218                 rthlen = (rth->ip6r_len + 1) << 3;
3219                 if (len != rthlen)
3220                         return (EINVAL);
3221
3222                 switch (rth->ip6r_type) {
3223                 case IPV6_RTHDR_TYPE_0:
3224                         if (rth->ip6r_len == 0) /* must contain one addr */
3225                                 return (EINVAL);
3226                         if (rth->ip6r_len % 2) /* length must be even */
3227                                 return (EINVAL);
3228                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3229                                 return (EINVAL);
3230                         break;
3231                 default:
3232                         return (EINVAL);        /* not supported */
3233                 }
3234
3235                 /* turn off the previous option */
3236                 ip6_clearpktopts(opt, IPV6_RTHDR);
3237                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3238                 if (opt->ip6po_rthdr == NULL)
3239                         return (ENOBUFS);
3240                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3241
3242                 break;
3243         }
3244
3245         case IPV6_USE_MIN_MTU:
3246                 if (len != sizeof(int))
3247                         return (EINVAL);
3248                 minmtupolicy = *(int *)buf;
3249                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3250                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3251                     minmtupolicy != IP6PO_MINMTU_ALL) {
3252                         return (EINVAL);
3253                 }
3254                 opt->ip6po_minmtu = minmtupolicy;
3255                 break;
3256
3257         case IPV6_DONTFRAG:
3258                 if (len != sizeof(int))
3259                         return (EINVAL);
3260
3261                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3262                         /*
3263                          * we ignore this option for TCP sockets.
3264                          * (RFC3542 leaves this case unspecified.)
3265                          */
3266                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3267                 } else
3268                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3269                 break;
3270
3271         case IPV6_PREFER_TEMPADDR:
3272                 if (len != sizeof(int))
3273                         return (EINVAL);
3274                 preftemp = *(int *)buf;
3275                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3276                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3277                     preftemp != IP6PO_TEMPADDR_PREFER) {
3278                         return (EINVAL);
3279                 }
3280                 opt->ip6po_prefer_tempaddr = preftemp;
3281                 break;
3282
3283         default:
3284                 return (ENOPROTOOPT);
3285         } /* end of switch */
3286
3287         return (0);
3288 }
3289
3290 /*
3291  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3292  * packet to the input queue of a specified interface.  Note that this
3293  * calls the output routine of the loopback "driver", but with an interface
3294  * pointer that might NOT be &loif -- easier than replicating that code here.
3295  */
3296 void
3297 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3298 {
3299         struct mbuf *copym;
3300         struct ip6_hdr *ip6;
3301
3302         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3303         if (copym == NULL)
3304                 return;
3305
3306         /*
3307          * Make sure to deep-copy IPv6 header portion in case the data
3308          * is in an mbuf cluster, so that we can safely override the IPv6
3309          * header portion later.
3310          */
3311         if (!M_WRITABLE(copym) ||
3312             copym->m_len < sizeof(struct ip6_hdr)) {
3313                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3314                 if (copym == NULL)
3315                         return;
3316         }
3317         ip6 = mtod(copym, struct ip6_hdr *);
3318         /*
3319          * clear embedded scope identifiers if necessary.
3320          * in6_clearscope will touch the addresses only when necessary.
3321          */
3322         in6_clearscope(&ip6->ip6_src);
3323         in6_clearscope(&ip6->ip6_dst);
3324         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3325                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3326                     CSUM_PSEUDO_HDR;
3327                 copym->m_pkthdr.csum_data = 0xffff;
3328         }
3329         if_simloop(ifp, copym, AF_INET6, 0);
3330 }
3331
3332 /*
3333  * Chop IPv6 header off from the payload.
3334  */
3335 static int
3336 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3337 {
3338         struct mbuf *mh;
3339         struct ip6_hdr *ip6;
3340
3341         ip6 = mtod(m, struct ip6_hdr *);
3342         if (m->m_len > sizeof(*ip6)) {
3343                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3344                 if (mh == NULL) {
3345                         m_freem(m);
3346                         return ENOBUFS;
3347                 }
3348                 m_move_pkthdr(mh, m);
3349                 M_ALIGN(mh, sizeof(*ip6));
3350                 m->m_len -= sizeof(*ip6);
3351                 m->m_data += sizeof(*ip6);
3352                 mh->m_next = m;
3353                 m = mh;
3354                 m->m_len = sizeof(*ip6);
3355                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3356         }
3357         exthdrs->ip6e_ip6 = m;
3358         return 0;
3359 }
3360
3361 /*
3362  * Compute IPv6 extension header length.
3363  */
3364 int
3365 ip6_optlen(struct inpcb *inp)
3366 {
3367         int len;
3368
3369         if (!inp->in6p_outputopts)
3370                 return 0;
3371
3372         len = 0;
3373 #define elen(x) \
3374     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3375
3376         len += elen(inp->in6p_outputopts->ip6po_hbh);
3377         if (inp->in6p_outputopts->ip6po_rthdr)
3378                 /* dest1 is valid with rthdr only */
3379                 len += elen(inp->in6p_outputopts->ip6po_dest1);
3380         len += elen(inp->in6p_outputopts->ip6po_rthdr);
3381         len += elen(inp->in6p_outputopts->ip6po_dest2);
3382         return len;
3383 #undef elen
3384 }