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_ratelimit.h"
  71 #include "opt_ipsec.h"
  72 #include "opt_sctp.h"
  73 #include "opt_route.h"
  74 #include "opt_rss.h"
  75
  76 #include <sys/param.h>
  77 #include <sys/kernel.h>
  78 #include <sys/malloc.h>
  79 #include <sys/mbuf.h>
  80 #include <sys/errno.h>
  81 #include <sys/priv.h>
  82 #include <sys/proc.h>
  83 #include <sys/protosw.h>
  84 #include <sys/socket.h>
  85 #include <sys/socketvar.h>
  86 #include <sys/syslog.h>
  87 #include <sys/ucred.h>
  88
  89 #include <machine/in_cksum.h>
  90
  91 #include <net/if.h>
  92 #include <net/if_var.h>
  93 #include <net/if_llatbl.h>
  94 #include <net/netisr.h>
  95 #include <net/route.h>
  96 #include <net/pfil.h>
  97 #include <net/rss_config.h>
  98 #include <net/vnet.h>
  99
 100 #include <netinet/in.h>
 101 #include <netinet/in_var.h>
 102 #include <netinet/ip_var.h>
 103 #include <netinet6/in6_fib.h>
 104 #include <netinet6/in6_var.h>
 105 #include <netinet/ip6.h>
 106 #include <netinet/icmp6.h>
 107 #include <netinet6/ip6_var.h>
 108 #include <netinet/in_pcb.h>
 109 #include <netinet/tcp_var.h>
 110 #include <netinet6/nd6.h>
 111 #include <netinet6/in6_rss.h>
 112
 113 #include <netipsec/ipsec_support.h>
 114 #ifdef SCTP
 115 #include <netinet/sctp.h>
 116 #include <netinet/sctp_crc32.h>
 117 #endif
 118
 119 #include <netinet6/ip6protosw.h>
 120 #include <netinet6/scope6_var.h>
 121
 122 extern int in6_mcast_loop;
 123
 124 struct ip6_exthdrs {
 125         struct mbuf *ip6e_ip6;
 126         struct mbuf *ip6e_hbh;
 127         struct mbuf *ip6e_dest1;
 128         struct mbuf *ip6e_rthdr;
 129         struct mbuf *ip6e_dest2;
 130 };
 131
 132 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
 133
 134 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
 135                            struct ucred *, int);
 136 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
 137         struct socket *, struct sockopt *);
 138 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
 139 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
 140         struct ucred *, int, int, int);
 141
 142 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 143 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 144         struct ip6_frag **);
 145 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 146 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 147 static int ip6_getpmtu(struct route_in6 *, int,
 148         struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
 149         u_int);
 150 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
 151         u_long *, int *, u_int);
 152 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
 153 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 154
 155
 156 /*
 157  * Make an extension header from option data.  hp is the source, and
 158  * mp is the destination.
 159  */
 160 #define MAKE_EXTHDR(hp, mp)                                             \
 161     do {                                                                \
 162         if (hp) {                                                       \
 163                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 164                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 165                     ((eh)->ip6e_len + 1) << 3);                         \
 166                 if (error)                                              \
 167                         goto freehdrs;                                  \
 168         }                                                               \
 169     } while (/*CONSTCOND*/ 0)
 170
 171 /*
 172  * Form a chain of extension headers.
 173  * m is the extension header mbuf
 174  * mp is the previous mbuf in the chain
 175  * p is the next header
 176  * i is the type of option.
 177  */
 178 #define MAKE_CHAIN(m, mp, p, i)\
 179     do {\
 180         if (m) {\
 181                 if (!hdrsplit) \
 182                         panic("assumption failed: hdr not split"); \
 183                 *mtod((m), u_char *) = *(p);\
 184                 *(p) = (i);\
 185                 p = mtod((m), u_char *);\
 186                 (m)->m_next = (mp)->m_next;\
 187                 (mp)->m_next = (m);\
 188                 (mp) = (m);\
 189         }\
 190     } while (/*CONSTCOND*/ 0)
 191
 192 void
 193 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 194 {
 195         u_short csum;
 196
 197         csum = in_cksum_skip(m, offset + plen, offset);
 198         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 199                 csum = 0xffff;
 200         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 201
 202         if (offset + sizeof(csum) > m->m_len)
 203                 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
 204         else
 205                 *(u_short *)mtodo(m, offset) = csum;
 206 }
 207
 208 static int
 209 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
 210     int plen, int optlen, bool frag __unused)
 211 {
 212
 213         KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
 214             "csum_flags %#x frag %d\n",
 215             __func__, __LINE__, plen, optlen, m, ifp, csum_flags, frag));
 216
 217         if ((csum_flags & CSUM_DELAY_DATA_IPV6) ||
 218 #ifdef SCTP
 219             (csum_flags & CSUM_SCTP_IPV6) ||
 220 #endif
 221             false) {
 222                 if (csum_flags & CSUM_DELAY_DATA_IPV6) {
 223                         in6_delayed_cksum(m, plen - optlen,
 224                             sizeof(struct ip6_hdr) + optlen);
 225                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 226                 }
 227 #ifdef SCTP
 228                 if (csum_flags & CSUM_SCTP_IPV6) {
 229                         sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
 230                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
 231                 }
 232 #endif
 233         }
 234
 235         return (0);
 236 }
 237
 238 int
 239 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
 240     int fraglen , uint32_t id)
 241 {
 242         struct mbuf *m, **mnext, *m_frgpart;
 243         struct ip6_hdr *ip6, *mhip6;
 244         struct ip6_frag *ip6f;
 245         int off;
 246         int error;
 247         int tlen = m0->m_pkthdr.len;
 248
 249         KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
 250
 251         m = m0;
 252         ip6 = mtod(m, struct ip6_hdr *);
 253         mnext = &m->m_nextpkt;
 254
 255         for (off = hlen; off < tlen; off += fraglen) {
 256                 m = m_gethdr(M_NOWAIT, MT_DATA);
 257                 if (!m) {
 258                         IP6STAT_INC(ip6s_odropped);
 259                         return (ENOBUFS);
 260                 }
 261
 262                 /*
 263                  * Make sure the complete packet header gets copied
 264                  * from the originating mbuf to the newly created
 265                  * mbuf. This also ensures that existing firewall
 266                  * classification(s), VLAN tags and so on get copied
 267                  * to the resulting fragmented packet(s):
 268                  */
 269                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 270                         m_free(m);
 271                         IP6STAT_INC(ip6s_odropped);
 272                         return (ENOBUFS);
 273                 }
 274
 275                 *mnext = m;
 276                 mnext = &m->m_nextpkt;
 277                 m->m_data += max_linkhdr;
 278                 mhip6 = mtod(m, struct ip6_hdr *);
 279                 *mhip6 = *ip6;
 280                 m->m_len = sizeof(*mhip6);
 281                 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 282                 if (error) {
 283                         IP6STAT_INC(ip6s_odropped);
 284                         return (error);
 285                 }
 286                 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 287                 if (off + fraglen >= tlen)
 288                         fraglen = tlen - off;
 289                 else
 290                         ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 291                 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
 292                     sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 293                 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
 294                         IP6STAT_INC(ip6s_odropped);
 295                         return (ENOBUFS);
 296                 }
 297                 m_cat(m, m_frgpart);
 298                 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
 299                 ip6f->ip6f_reserved = 0;
 300                 ip6f->ip6f_ident = id;
 301                 ip6f->ip6f_nxt = nextproto;
 302                 IP6STAT_INC(ip6s_ofragments);
 303                 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 304         }
 305
 306         return (0);
 307 }
 308
 309 /*
 310  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 311  * header (with pri, len, nxt, hlim, src, dst).
 312  * This function may modify ver and hlim only.
 313  * The mbuf chain containing the packet will be freed.
 314  * The mbuf opt, if present, will not be freed.
 315  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
 316  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 317  * then result of route lookup is stored in ro->ro_rt.
 318  *
 319  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
 320  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 321  * which is rt_mtu.
 322  *
 323  * ifpp - XXX: just for statistics
 324  */
 325 /*
 326  * XXX TODO: no flowid is assigned for outbound flows?
 327  */
 328 int
 329 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 330     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 331     struct ifnet **ifpp, struct inpcb *inp)
 332 {
 333         struct ip6_hdr *ip6;
 334         struct ifnet *ifp, *origifp;
 335         struct mbuf *m = m0;
 336         struct mbuf *mprev = NULL;
 337         int hlen, tlen, len;
 338         struct route_in6 ip6route;
 339         struct rtentry *rt = NULL;
 340         struct sockaddr_in6 *dst, src_sa, dst_sa;
 341         struct in6_addr odst;
 342         int error = 0;
 343         struct in6_ifaddr *ia = NULL;
 344         u_long mtu;
 345         int alwaysfrag, dontfrag;
 346         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 347         struct ip6_exthdrs exthdrs;
 348         struct in6_addr src0, dst0;
 349         u_int32_t zone;
 350         struct route_in6 *ro_pmtu = NULL;
 351         int hdrsplit = 0;
 352         int sw_csum, tso;
 353         int needfiblookup;
 354         uint32_t fibnum;
 355         struct m_tag *fwd_tag = NULL;
 356         uint32_t id;
 357
 358         if (inp != NULL) {
 359                 INP_LOCK_ASSERT(inp);
 360                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 361                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 362                         /* unconditionally set flowid */
 363                         m->m_pkthdr.flowid = inp->inp_flowid;
 364                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 365                 }
 366         }
 367
 368 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 369         /*
 370          * IPSec checking which handles several cases.
 371          * FAST IPSEC: We re-injected the packet.
 372          * XXX: need scope argument.
 373          */
 374         if (IPSEC_ENABLED(ipv6)) {
 375                 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
 376                         if (error == EINPROGRESS)
 377                                 error = 0;
 378                         goto done;
 379                 }
 380         }
 381 #endif /* IPSEC */
 382
 383         bzero(&exthdrs, sizeof(exthdrs));
 384         if (opt) {
 385                 /* Hop-by-Hop options header */
 386                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 387                 /* Destination options header(1st part) */
 388                 if (opt->ip6po_rthdr) {
 389                         /*
 390                          * Destination options header(1st part)
 391                          * This only makes sense with a routing header.
 392                          * See Section 9.2 of RFC 3542.
 393                          * Disabling this part just for MIP6 convenience is
 394                          * a bad idea.  We need to think carefully about a
 395                          * way to make the advanced API coexist with MIP6
 396                          * options, which might automatically be inserted in
 397                          * the kernel.
 398                          */
 399                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 400                 }
 401                 /* Routing header */
 402                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 403                 /* Destination options header(2nd part) */
 404                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 405         }
 406
 407         /*
 408          * Calculate the total length of the extension header chain.
 409          * Keep the length of the unfragmentable part for fragmentation.
 410          */
 411         optlen = 0;
 412         if (exthdrs.ip6e_hbh)
 413                 optlen += exthdrs.ip6e_hbh->m_len;
 414         if (exthdrs.ip6e_dest1)
 415                 optlen += exthdrs.ip6e_dest1->m_len;
 416         if (exthdrs.ip6e_rthdr)
 417                 optlen += exthdrs.ip6e_rthdr->m_len;
 418         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 419
 420         /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
 421         if (exthdrs.ip6e_dest2)
 422                 optlen += exthdrs.ip6e_dest2->m_len;
 423
 424         /*
 425          * If there is at least one extension header,
 426          * separate IP6 header from the payload.
 427          */
 428         if (optlen && !hdrsplit) {
 429                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 430                         m = NULL;
 431                         goto freehdrs;
 432                 }
 433                 m = exthdrs.ip6e_ip6;
 434                 hdrsplit++;
 435         }
 436
 437         ip6 = mtod(m, struct ip6_hdr *);
 438
 439         /* adjust mbuf packet header length */
 440         m->m_pkthdr.len += optlen;
 441         plen = m->m_pkthdr.len - sizeof(*ip6);
 442
 443         /* If this is a jumbo payload, insert a jumbo payload option. */
 444         if (plen > IPV6_MAXPACKET) {
 445                 if (!hdrsplit) {
 446                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 447                                 m = NULL;
 448                                 goto freehdrs;
 449                         }
 450                         m = exthdrs.ip6e_ip6;
 451                         hdrsplit++;
 452                 }
 453                 /* adjust pointer */
 454                 ip6 = mtod(m, struct ip6_hdr *);
 455                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 456                         goto freehdrs;
 457                 ip6->ip6_plen = 0;
 458         } else
 459                 ip6->ip6_plen = htons(plen);
 460
 461         /*
 462          * Concatenate headers and fill in next header fields.
 463          * Here we have, on "m"
 464          *      IPv6 payload
 465          * and we insert headers accordingly.  Finally, we should be getting:
 466          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 467          *
 468          * during the header composing process, "m" points to IPv6 header.
 469          * "mprev" points to an extension header prior to esp.
 470          */
 471         u_char *nexthdrp = &ip6->ip6_nxt;
 472         mprev = m;
 473
 474         /*
 475          * we treat dest2 specially.  this makes IPsec processing
 476          * much easier.  the goal here is to make mprev point the
 477          * mbuf prior to dest2.
 478          *
 479          * result: IPv6 dest2 payload
 480          * m and mprev will point to IPv6 header.
 481          */
 482         if (exthdrs.ip6e_dest2) {
 483                 if (!hdrsplit)
 484                         panic("assumption failed: hdr not split");
 485                 exthdrs.ip6e_dest2->m_next = m->m_next;
 486                 m->m_next = exthdrs.ip6e_dest2;
 487                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 488                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 489         }
 490
 491         /*
 492          * result: IPv6 hbh dest1 rthdr dest2 payload
 493          * m will point to IPv6 header.  mprev will point to the
 494          * extension header prior to dest2 (rthdr in the above case).
 495          */
 496         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 497         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 498                    IPPROTO_DSTOPTS);
 499         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 500                    IPPROTO_ROUTING);
 501
 502         /*
 503          * If there is a routing header, discard the packet.
 504          */
 505         if (exthdrs.ip6e_rthdr) {
 506                  error = EINVAL;
 507                  goto bad;
 508         }
 509
 510         /* Source address validation */
 511         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 512             (flags & IPV6_UNSPECSRC) == 0) {
 513                 error = EOPNOTSUPP;
 514                 IP6STAT_INC(ip6s_badscope);
 515                 goto bad;
 516         }
 517         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 518                 error = EOPNOTSUPP;
 519                 IP6STAT_INC(ip6s_badscope);
 520                 goto bad;
 521         }
 522
 523         IP6STAT_INC(ip6s_localout);
 524
 525         /*
 526          * Route packet.
 527          */
 528         if (ro == NULL) {
 529                 ro = &ip6route;
 530                 bzero((caddr_t)ro, sizeof(*ro));
 531         }
 532         ro_pmtu = ro;
 533         if (opt && opt->ip6po_rthdr)
 534                 ro = &opt->ip6po_route;
 535         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 536         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 537 again:
 538         /*
 539          * if specified, try to fill in the traffic class field.
 540          * do not override if a non-zero value is already set.
 541          * we check the diffserv field and the ecn field separately.
 542          */
 543         if (opt && opt->ip6po_tclass >= 0) {
 544                 int mask = 0;
 545
 546                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 547                         mask |= 0xfc;
 548                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 549                         mask |= 0x03;
 550                 if (mask != 0)
 551                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 552         }
 553
 554         /* fill in or override the hop limit field, if necessary. */
 555         if (opt && opt->ip6po_hlim != -1)
 556                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 557         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 558                 if (im6o != NULL)
 559                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 560                 else
 561                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 562         }
 563         /*
 564          * Validate route against routing table additions;
 565          * a better/more specific route might have been added.
 566          * Make sure address family is set in route.
 567          */
 568         if (inp) {
 569                 ro->ro_dst.sin6_family = AF_INET6;
 570                 RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
 571         }
 572         if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
 573             ro->ro_dst.sin6_family == AF_INET6 &&
 574             IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
 575                 rt = ro->ro_rt;
 576                 ifp = ro->ro_rt->rt_ifp;
 577         } else {
 578                 if (ro->ro_lle)
 579                         LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 580                 ro->ro_lle = NULL;
 581                 if (fwd_tag == NULL) {
 582                         bzero(&dst_sa, sizeof(dst_sa));
 583                         dst_sa.sin6_family = AF_INET6;
 584                         dst_sa.sin6_len = sizeof(dst_sa);
 585                         dst_sa.sin6_addr = ip6->ip6_dst;
 586                 }
 587                 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
 588                     &rt, fibnum);
 589                 if (error != 0) {
 590                         if (ifp != NULL)
 591                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 592                         goto bad;
 593                 }
 594         }
 595         if (rt == NULL) {
 596                 /*
 597                  * If in6_selectroute() does not return a route entry,
 598                  * dst may not have been updated.
 599                  */
 600                 *dst = dst_sa;  /* XXX */
 601         }
 602
 603         /*
 604          * then rt (for unicast) and ifp must be non-NULL valid values.
 605          */
 606         if ((flags & IPV6_FORWARDING) == 0) {
 607                 /* XXX: the FORWARDING flag can be set for mrouting. */
 608                 in6_ifstat_inc(ifp, ifs6_out_request);
 609         }
 610         if (rt != NULL) {
 611                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 612                 counter_u64_add(rt->rt_pksent, 1);
 613         }
 614
 615         /* Setup data structures for scope ID checks. */
 616         src0 = ip6->ip6_src;
 617         bzero(&src_sa, sizeof(src_sa));
 618         src_sa.sin6_family = AF_INET6;
 619         src_sa.sin6_len = sizeof(src_sa);
 620         src_sa.sin6_addr = ip6->ip6_src;
 621
 622         dst0 = ip6->ip6_dst;
 623         /* re-initialize to be sure */
 624         bzero(&dst_sa, sizeof(dst_sa));
 625         dst_sa.sin6_family = AF_INET6;
 626         dst_sa.sin6_len = sizeof(dst_sa);
 627         dst_sa.sin6_addr = ip6->ip6_dst;
 628
 629         /* Check for valid scope ID. */
 630         if (in6_setscope(&src0, ifp, &zone) == 0 &&
 631             sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
 632             in6_setscope(&dst0, ifp, &zone) == 0 &&
 633             sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
 634                 /*
 635                  * The outgoing interface is in the zone of the source
 636                  * and destination addresses.
 637                  *
 638                  * Because the loopback interface cannot receive
 639                  * packets with a different scope ID than its own,
 640                  * there is a trick is to pretend the outgoing packet
 641                  * was received by the real network interface, by
 642                  * setting "origifp" different from "ifp". This is
 643                  * only allowed when "ifp" is a loopback network
 644                  * interface. Refer to code in nd6_output_ifp() for
 645                  * more details.
 646                  */
 647                 origifp = ifp;
 648
 649                 /*
 650                  * We should use ia_ifp to support the case of sending
 651                  * packets to an address of our own.
 652                  */
 653                 if (ia != NULL && ia->ia_ifp)
 654                         ifp = ia->ia_ifp;
 655
 656         } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
 657             sa6_recoverscope(&src_sa) != 0 ||
 658             sa6_recoverscope(&dst_sa) != 0 ||
 659             dst_sa.sin6_scope_id == 0 ||
 660             (src_sa.sin6_scope_id != 0 &&
 661             src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
 662             (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
 663                 /*
 664                  * If the destination network interface is not a
 665                  * loopback interface, or the destination network
 666                  * address has no scope ID, or the source address has
 667                  * a scope ID set which is different from the
 668                  * destination address one, or there is no network
 669                  * interface representing this scope ID, the address
 670                  * pair is considered invalid.
 671                  */
 672                 IP6STAT_INC(ip6s_badscope);
 673                 in6_ifstat_inc(ifp, ifs6_out_discard);
 674                 if (error == 0)
 675                         error = EHOSTUNREACH; /* XXX */
 676                 goto bad;
 677         }
 678
 679         /* All scope ID checks are successful. */
 680
 681         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 682                 if (opt && opt->ip6po_nextroute.ro_rt) {
 683                         /*
 684                          * The nexthop is explicitly specified by the
 685                          * application.  We assume the next hop is an IPv6
 686                          * address.
 687                          */
 688                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 689                 }
 690                 else if ((rt->rt_flags & RTF_GATEWAY))
 691                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 692         }
 693
 694         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 695                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 696         } else {
 697                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 698                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 699                 /*
 700                  * Confirm that the outgoing interface supports multicast.
 701                  */
 702                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 703                         IP6STAT_INC(ip6s_noroute);
 704                         in6_ifstat_inc(ifp, ifs6_out_discard);
 705                         error = ENETUNREACH;
 706                         goto bad;
 707                 }
 708                 if ((im6o == NULL && in6_mcast_loop) ||
 709                     (im6o && im6o->im6o_multicast_loop)) {
 710                         /*
 711                          * Loop back multicast datagram if not expressly
 712                          * forbidden to do so, even if we have not joined
 713                          * the address; protocols will filter it later,
 714                          * thus deferring a hash lookup and lock acquisition
 715                          * at the expense of an m_copym().
 716                          */
 717                         ip6_mloopback(ifp, m);
 718                 } else {
 719                         /*
 720                          * If we are acting as a multicast router, perform
 721                          * multicast forwarding as if the packet had just
 722                          * arrived on the interface to which we are about
 723                          * to send.  The multicast forwarding function
 724                          * recursively calls this function, using the
 725                          * IPV6_FORWARDING flag to prevent infinite recursion.
 726                          *
 727                          * Multicasts that are looped back by ip6_mloopback(),
 728                          * above, will be forwarded by the ip6_input() routine,
 729                          * if necessary.
 730                          */
 731                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 732                                 /*
 733                                  * XXX: ip6_mforward expects that rcvif is NULL
 734                                  * when it is called from the originating path.
 735                                  * However, it may not always be the case.
 736                                  */
 737                                 m->m_pkthdr.rcvif = NULL;
 738                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 739                                         m_freem(m);
 740                                         goto done;
 741                                 }
 742                         }
 743                 }
 744                 /*
 745                  * Multicasts with a hoplimit of zero may be looped back,
 746                  * above, but must not be transmitted on a network.
 747                  * Also, multicasts addressed to the loopback interface
 748                  * are not sent -- the above call to ip6_mloopback() will
 749                  * loop back a copy if this host actually belongs to the
 750                  * destination group on the loopback interface.
 751                  */
 752                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 753                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 754                         m_freem(m);
 755                         goto done;
 756                 }
 757         }
 758
 759         /*
 760          * Fill the outgoing inteface to tell the upper layer
 761          * to increment per-interface statistics.
 762          */
 763         if (ifpp)
 764                 *ifpp = ifp;
 765
 766         /* Determine path MTU. */
 767         if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
 768                     &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
 769                 goto bad;
 770
 771         /*
 772          * The caller of this function may specify to use the minimum MTU
 773          * in some cases.
 774          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 775          * setting.  The logic is a bit complicated; by default, unicast
 776          * packets will follow path MTU while multicast packets will be sent at
 777          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 778          * including unicast ones will be sent at the minimum MTU.  Multicast
 779          * packets will always be sent at the minimum MTU unless
 780          * IP6PO_MINMTU_DISABLE is explicitly specified.
 781          * See RFC 3542 for more details.
 782          */
 783         if (mtu > IPV6_MMTU) {
 784                 if ((flags & IPV6_MINMTU))
 785                         mtu = IPV6_MMTU;
 786                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 787                         mtu = IPV6_MMTU;
 788                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 789                          (opt == NULL ||
 790                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 791                         mtu = IPV6_MMTU;
 792                 }
 793         }
 794
 795         /*
 796          * clear embedded scope identifiers if necessary.
 797          * in6_clearscope will touch the addresses only when necessary.
 798          */
 799         in6_clearscope(&ip6->ip6_src);
 800         in6_clearscope(&ip6->ip6_dst);
 801
 802         /*
 803          * If the outgoing packet contains a hop-by-hop options header,
 804          * it must be examined and processed even by the source node.
 805          * (RFC 2460, section 4.)
 806          */
 807         if (exthdrs.ip6e_hbh) {
 808                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 809                 u_int32_t dummy; /* XXX unused */
 810                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 811
 812 #ifdef DIAGNOSTIC
 813                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 814                         panic("ip6e_hbh is not contiguous");
 815 #endif
 816                 /*
 817                  *  XXX: if we have to send an ICMPv6 error to the sender,
 818                  *       we need the M_LOOP flag since icmp6_error() expects
 819                  *       the IPv6 and the hop-by-hop options header are
 820                  *       contiguous unless the flag is set.
 821                  */
 822                 m->m_flags |= M_LOOP;
 823                 m->m_pkthdr.rcvif = ifp;
 824                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 825                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 826                     &dummy, &plen) < 0) {
 827                         /* m was already freed at this point */
 828                         error = EINVAL;/* better error? */
 829                         goto done;
 830                 }
 831                 m->m_flags &= ~M_LOOP; /* XXX */
 832                 m->m_pkthdr.rcvif = NULL;
 833         }
 834
 835         /* Jump over all PFIL processing if hooks are not active. */
 836         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 837                 goto passout;
 838
 839         odst = ip6->ip6_dst;
 840         /* Run through list of hooks for output packets. */
 841         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, 0, inp);
 842         if (error != 0 || m == NULL)
 843                 goto done;
 844         /* adjust pointer */
 845         ip6 = mtod(m, struct ip6_hdr *);
 846
 847         needfiblookup = 0;
 848         /* See if destination IP address was changed by packet filter. */
 849         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 850                 m->m_flags |= M_SKIP_FIREWALL;
 851                 /* If destination is now ourself drop to ip6_input(). */
 852                 if (in6_localip(&ip6->ip6_dst)) {
 853                         m->m_flags |= M_FASTFWD_OURS;
 854                         if (m->m_pkthdr.rcvif == NULL)
 855                                 m->m_pkthdr.rcvif = V_loif;
 856                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 857                                 m->m_pkthdr.csum_flags |=
 858                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 859                                 m->m_pkthdr.csum_data = 0xffff;
 860                         }
 861 #ifdef SCTP
 862                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 863                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 864 #endif
 865                         error = netisr_queue(NETISR_IPV6, m);
 866                         goto done;
 867                 } else {
 868                         RO_INVALIDATE_CACHE(ro);
 869                         needfiblookup = 1; /* Redo the routing table lookup. */
 870                 }
 871         }
 872         /* See if fib was changed by packet filter. */
 873         if (fibnum != M_GETFIB(m)) {
 874                 m->m_flags |= M_SKIP_FIREWALL;
 875                 fibnum = M_GETFIB(m);
 876                 RO_INVALIDATE_CACHE(ro);
 877                 needfiblookup = 1;
 878         }
 879         if (needfiblookup)
 880                 goto again;
 881
 882         /* See if local, if yes, send it to netisr. */
 883         if (m->m_flags & M_FASTFWD_OURS) {
 884                 if (m->m_pkthdr.rcvif == NULL)
 885                         m->m_pkthdr.rcvif = V_loif;
 886                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 887                         m->m_pkthdr.csum_flags |=
 888                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 889                         m->m_pkthdr.csum_data = 0xffff;
 890                 }
 891 #ifdef SCTP
 892                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 893                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 894 #endif
 895                 error = netisr_queue(NETISR_IPV6, m);
 896                 goto done;
 897         }
 898         /* Or forward to some other address? */
 899         if ((m->m_flags & M_IP6_NEXTHOP) &&
 900             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 901                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 902                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
 903                 m->m_flags |= M_SKIP_FIREWALL;
 904                 m->m_flags &= ~M_IP6_NEXTHOP;
 905                 m_tag_delete(m, fwd_tag);
 906                 goto again;
 907         }
 908
 909 passout:
 910         /*
 911          * Send the packet to the outgoing interface.
 912          * If necessary, do IPv6 fragmentation before sending.
 913          *
 914          * the logic here is rather complex:
 915          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 916          * 1-a: send as is if tlen <= path mtu
 917          * 1-b: fragment if tlen > path mtu
 918          *
 919          * 2: if user asks us not to fragment (dontfrag == 1)
 920          * 2-a: send as is if tlen <= interface mtu
 921          * 2-b: error if tlen > interface mtu
 922          *
 923          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 924          *      always fragment
 925          *
 926          * 4: if dontfrag == 1 && alwaysfrag == 1
 927          *      error, as we cannot handle this conflicting request
 928          */
 929         sw_csum = m->m_pkthdr.csum_flags;
 930         if (!hdrsplit) {
 931                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 932                 sw_csum &= ~ifp->if_hwassist;
 933         } else
 934                 tso = 0;
 935         /*
 936          * If we added extension headers, we will not do TSO and calculate the
 937          * checksums ourselves for now.
 938          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 939          * with ext. hdrs.
 940          */
 941         error = ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen, false);
 942         if (error != 0)
 943                 goto bad;
 944         /* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
 945         tlen = m->m_pkthdr.len;
 946
 947         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 948                 dontfrag = 1;
 949         else
 950                 dontfrag = 0;
 951         if (dontfrag && alwaysfrag) {   /* case 4 */
 952                 /* conflicting request - can't transmit */
 953                 error = EMSGSIZE;
 954                 goto bad;
 955         }
 956         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 957                 /*
 958                  * Even if the DONTFRAG option is specified, we cannot send the
 959                  * packet when the data length is larger than the MTU of the
 960                  * outgoing interface.
 961                  * Notify the error by sending IPV6_PATHMTU ancillary data if
 962                  * application wanted to know the MTU value. Also return an
 963                  * error code (this is not described in the API spec).
 964                  */
 965                 if (inp != NULL)
 966                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
 967                 error = EMSGSIZE;
 968                 goto bad;
 969         }
 970
 971         /*
 972          * transmit packet without fragmentation
 973          */
 974         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 975                 struct in6_ifaddr *ia6;
 976
 977                 ip6 = mtod(m, struct ip6_hdr *);
 978                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 979                 if (ia6) {
 980                         /* Record statistics for this interface address. */
 981                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
 982                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
 983                             m->m_pkthdr.len);
 984                         ifa_free(&ia6->ia_ifa);
 985                 }
 986 #ifdef RATELIMIT
 987                 if (inp != NULL) {
 988                         if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
 989                                 in_pcboutput_txrtlmt(inp, ifp, m);
 990                         /* stamp send tag on mbuf */
 991                         m->m_pkthdr.snd_tag = inp->inp_snd_tag;
 992                         m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
 993                 } else {
 994                         m->m_pkthdr.snd_tag = NULL;
 995                 }
 996 #endif
 997                 error = nd6_output_ifp(ifp, origifp, m, dst,
 998                     (struct route *)ro);
 999 #ifdef RATELIMIT
1000                 /* check for route change */
1001                 if (error == EAGAIN)
1002                         in_pcboutput_eagain(inp);
1003 #endif
1004                 goto done;
1005         }
1006
1007         /*
1008          * try to fragment the packet.  case 1-b and 3
1009          */
1010         if (mtu < IPV6_MMTU) {
1011                 /* path MTU cannot be less than IPV6_MMTU */
1012                 error = EMSGSIZE;
1013                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1014                 goto bad;
1015         } else if (ip6->ip6_plen == 0) {
1016                 /* jumbo payload cannot be fragmented */
1017                 error = EMSGSIZE;
1018                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1019                 goto bad;
1020         } else {
1021                 u_char nextproto;
1022
1023                 /*
1024                  * Too large for the destination or interface;
1025                  * fragment if possible.
1026                  * Must be able to put at least 8 bytes per fragment.
1027                  */
1028                 hlen = unfragpartlen;
1029                 if (mtu > IPV6_MAXPACKET)
1030                         mtu = IPV6_MAXPACKET;
1031
1032                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1033                 if (len < 8) {
1034                         error = EMSGSIZE;
1035                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1036                         goto bad;
1037                 }
1038
1039                 /*
1040                  * If the interface will not calculate checksums on
1041                  * fragmented packets, then do it here.
1042                  * XXX-BZ handle the hw offloading case.  Need flags.
1043                  */
1044                 error = ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags,
1045                     plen, optlen, true);
1046                 if (error != 0)
1047                         goto bad;
1048
1049                 /*
1050                  * Change the next header field of the last header in the
1051                  * unfragmentable part.
1052                  */
1053                 if (exthdrs.ip6e_rthdr) {
1054                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1055                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1056                 } else if (exthdrs.ip6e_dest1) {
1057                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1058                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1059                 } else if (exthdrs.ip6e_hbh) {
1060                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1061                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1062                 } else {
1063                         nextproto = ip6->ip6_nxt;
1064                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1065                 }
1066
1067                 /*
1068                  * Loop through length of segment after first fragment,
1069                  * make new header and copy data of each part and link onto
1070                  * chain.
1071                  */
1072                 m0 = m;
1073                 id = htonl(ip6_randomid());
1074                 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1075                         goto sendorfree;
1076
1077                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1078         }
1079
1080         /*
1081          * Remove leading garbages.
1082          */
1083 sendorfree:
1084         m = m0->m_nextpkt;
1085         m0->m_nextpkt = 0;
1086         m_freem(m0);
1087         for (; m; m = m0) {
1088                 m0 = m->m_nextpkt;
1089                 m->m_nextpkt = 0;
1090                 if (error == 0) {
1091                         /* Record statistics for this interface address. */
1092                         if (ia) {
1093                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1094                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1095                                     m->m_pkthdr.len);
1096                         }
1097 #ifdef RATELIMIT
1098                         if (inp != NULL) {
1099                                 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1100                                         in_pcboutput_txrtlmt(inp, ifp, m);
1101                                 /* stamp send tag on mbuf */
1102                                 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1103                                 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
1104                         } else {
1105                                 m->m_pkthdr.snd_tag = NULL;
1106                         }
1107 #endif
1108                         error = nd6_output_ifp(ifp, origifp, m, dst,
1109                             (struct route *)ro);
1110 #ifdef RATELIMIT
1111                         /* check for route change */
1112                         if (error == EAGAIN)
1113                                 in_pcboutput_eagain(inp);
1114 #endif
1115                 } else
1116                         m_freem(m);
1117         }
1118
1119         if (error == 0)
1120                 IP6STAT_INC(ip6s_fragmented);
1121
1122 done:
1123         if (ro == &ip6route)
1124                 RO_RTFREE(ro);
1125         return (error);
1126
1127 freehdrs:
1128         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1129         m_freem(exthdrs.ip6e_dest1);
1130         m_freem(exthdrs.ip6e_rthdr);
1131         m_freem(exthdrs.ip6e_dest2);
1132         /* FALLTHROUGH */
1133 bad:
1134         if (m)
1135                 m_freem(m);
1136         goto done;
1137 }
1138
1139 static int
1140 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1141 {
1142         struct mbuf *m;
1143
1144         if (hlen > MCLBYTES)
1145                 return (ENOBUFS); /* XXX */
1146
1147         if (hlen > MLEN)
1148                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1149         else
1150                 m = m_get(M_NOWAIT, MT_DATA);
1151         if (m == NULL)
1152                 return (ENOBUFS);
1153         m->m_len = hlen;
1154         if (hdr)
1155                 bcopy(hdr, mtod(m, caddr_t), hlen);
1156
1157         *mp = m;
1158         return (0);
1159 }
1160
1161 /*
1162  * Insert jumbo payload option.
1163  */
1164 static int
1165 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1166 {
1167         struct mbuf *mopt;
1168         u_char *optbuf;
1169         u_int32_t v;
1170
1171 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1172
1173         /*
1174          * If there is no hop-by-hop options header, allocate new one.
1175          * If there is one but it doesn't have enough space to store the
1176          * jumbo payload option, allocate a cluster to store the whole options.
1177          * Otherwise, use it to store the options.
1178          */
1179         if (exthdrs->ip6e_hbh == NULL) {
1180                 mopt = m_get(M_NOWAIT, MT_DATA);
1181                 if (mopt == NULL)
1182                         return (ENOBUFS);
1183                 mopt->m_len = JUMBOOPTLEN;
1184                 optbuf = mtod(mopt, u_char *);
1185                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1186                 exthdrs->ip6e_hbh = mopt;
1187         } else {
1188                 struct ip6_hbh *hbh;
1189
1190                 mopt = exthdrs->ip6e_hbh;
1191                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1192                         /*
1193                          * XXX assumption:
1194                          * - exthdrs->ip6e_hbh is not referenced from places
1195                          *   other than exthdrs.
1196                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1197                          */
1198                         int oldoptlen = mopt->m_len;
1199                         struct mbuf *n;
1200
1201                         /*
1202                          * XXX: give up if the whole (new) hbh header does
1203                          * not fit even in an mbuf cluster.
1204                          */
1205                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1206                                 return (ENOBUFS);
1207
1208                         /*
1209                          * As a consequence, we must always prepare a cluster
1210                          * at this point.
1211                          */
1212                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1213                         if (n == NULL)
1214                                 return (ENOBUFS);
1215                         n->m_len = oldoptlen + JUMBOOPTLEN;
1216                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1217                             oldoptlen);
1218                         optbuf = mtod(n, caddr_t) + oldoptlen;
1219                         m_freem(mopt);
1220                         mopt = exthdrs->ip6e_hbh = n;
1221                 } else {
1222                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1223                         mopt->m_len += JUMBOOPTLEN;
1224                 }
1225                 optbuf[0] = IP6OPT_PADN;
1226                 optbuf[1] = 1;
1227
1228                 /*
1229                  * Adjust the header length according to the pad and
1230                  * the jumbo payload option.
1231                  */
1232                 hbh = mtod(mopt, struct ip6_hbh *);
1233                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1234         }
1235
1236         /* fill in the option. */
1237         optbuf[2] = IP6OPT_JUMBO;
1238         optbuf[3] = 4;
1239         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1240         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1241
1242         /* finally, adjust the packet header length */
1243         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1244
1245         return (0);
1246 #undef JUMBOOPTLEN
1247 }
1248
1249 /*
1250  * Insert fragment header and copy unfragmentable header portions.
1251  */
1252 static int
1253 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1254     struct ip6_frag **frghdrp)
1255 {
1256         struct mbuf *n, *mlast;
1257
1258         if (hlen > sizeof(struct ip6_hdr)) {
1259                 n = m_copym(m0, sizeof(struct ip6_hdr),
1260                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1261                 if (n == NULL)
1262                         return (ENOBUFS);
1263                 m->m_next = n;
1264         } else
1265                 n = m;
1266
1267         /* Search for the last mbuf of unfragmentable part. */
1268         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1269                 ;
1270
1271         if (M_WRITABLE(mlast) &&
1272             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1273                 /* use the trailing space of the last mbuf for the fragment hdr */
1274                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1275                     mlast->m_len);
1276                 mlast->m_len += sizeof(struct ip6_frag);
1277                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1278         } else {
1279                 /* allocate a new mbuf for the fragment header */
1280                 struct mbuf *mfrg;
1281
1282                 mfrg = m_get(M_NOWAIT, MT_DATA);
1283                 if (mfrg == NULL)
1284                         return (ENOBUFS);
1285                 mfrg->m_len = sizeof(struct ip6_frag);
1286                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1287                 mlast->m_next = mfrg;
1288         }
1289
1290         return (0);
1291 }
1292
1293 /*
1294  * Calculates IPv6 path mtu for destination @dst.
1295  * Resulting MTU is stored in @mtup.
1296  *
1297  * Returns 0 on success.
1298  */
1299 static int
1300 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1301 {
1302         struct nhop6_extended nh6;
1303         struct in6_addr kdst;
1304         uint32_t scopeid;
1305         struct ifnet *ifp;
1306         u_long mtu;
1307         int error;
1308
1309         in6_splitscope(dst, &kdst, &scopeid);
1310         if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1311                 return (EHOSTUNREACH);
1312
1313         ifp = nh6.nh_ifp;
1314         mtu = nh6.nh_mtu;
1315
1316         error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1317         fib6_free_nh_ext(fibnum, &nh6);
1318
1319         return (error);
1320 }
1321
1322 /*
1323  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1324  * and cached data in @ro_pmtu.
1325  * MTU from (successful) route lookup is saved (along with dst)
1326  * inside @ro_pmtu to avoid subsequent route lookups after packet
1327  * filter processing.
1328  *
1329  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1330  * Returns 0 on success.
1331  */
1332 static int
1333 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1334     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1335     int *alwaysfragp, u_int fibnum, u_int proto)
1336 {
1337         struct nhop6_basic nh6;
1338         struct in6_addr kdst;
1339         uint32_t scopeid;
1340         struct sockaddr_in6 *sa6_dst;
1341         u_long mtu;
1342
1343         mtu = 0;
1344         if (do_lookup) {
1345
1346                 /*
1347                  * Here ro_pmtu has final destination address, while
1348                  * ro might represent immediate destination.
1349                  * Use ro_pmtu destination since mtu might differ.
1350                  */
1351                 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1352                 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1353                         ro_pmtu->ro_mtu = 0;
1354
1355                 if (ro_pmtu->ro_mtu == 0) {
1356                         bzero(sa6_dst, sizeof(*sa6_dst));
1357                         sa6_dst->sin6_family = AF_INET6;
1358                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1359                         sa6_dst->sin6_addr = *dst;
1360
1361                         in6_splitscope(dst, &kdst, &scopeid);
1362                         if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1363                             &nh6) == 0)
1364                                 ro_pmtu->ro_mtu = nh6.nh_mtu;
1365                 }
1366
1367                 mtu = ro_pmtu->ro_mtu;
1368         }
1369
1370         if (ro_pmtu->ro_rt)
1371                 mtu = ro_pmtu->ro_rt->rt_mtu;
1372
1373         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1374 }
1375
1376 /*
1377  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1378  * hostcache data for @dst.
1379  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1380  *
1381  * Returns 0 on success.
1382  */
1383 static int
1384 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1385     u_long *mtup, int *alwaysfragp, u_int proto)
1386 {
1387         u_long mtu = 0;
1388         int alwaysfrag = 0;
1389         int error = 0;
1390
1391         if (rt_mtu > 0) {
1392                 u_int32_t ifmtu;
1393                 struct in_conninfo inc;
1394
1395                 bzero(&inc, sizeof(inc));
1396                 inc.inc_flags |= INC_ISIPV6;
1397                 inc.inc6_faddr = *dst;
1398
1399                 ifmtu = IN6_LINKMTU(ifp);
1400
1401                 /* TCP is known to react to pmtu changes so skip hc */
1402                 if (proto != IPPROTO_TCP)
1403                         mtu = tcp_hc_getmtu(&inc);
1404
1405                 if (mtu)
1406                         mtu = min(mtu, rt_mtu);
1407                 else
1408                         mtu = rt_mtu;
1409                 if (mtu == 0)
1410                         mtu = ifmtu;
1411                 else if (mtu < IPV6_MMTU) {
1412                         /*
1413                          * RFC2460 section 5, last paragraph:
1414                          * if we record ICMPv6 too big message with
1415                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1416                          * or smaller, with framgent header attached.
1417                          * (fragment header is needed regardless from the
1418                          * packet size, for translators to identify packets)
1419                          */
1420                         alwaysfrag = 1;
1421                         mtu = IPV6_MMTU;
1422                 }
1423         } else if (ifp) {
1424                 mtu = IN6_LINKMTU(ifp);
1425         } else
1426                 error = EHOSTUNREACH; /* XXX */
1427
1428         *mtup = mtu;
1429         if (alwaysfragp)
1430                 *alwaysfragp = alwaysfrag;
1431         return (error);
1432 }
1433
1434 /*
1435  * IP6 socket option processing.
1436  */
1437 int
1438 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1439 {
1440         int optdatalen, uproto;
1441         void *optdata;
1442         struct inpcb *in6p = sotoinpcb(so);
1443         int error, optval;
1444         int level, op, optname;
1445         int optlen;
1446         struct thread *td;
1447 #ifdef  RSS
1448         uint32_t rss_bucket;
1449         int retval;
1450 #endif
1451
1452 /*
1453  * Don't use more than a quarter of mbuf clusters.  N.B.:
1454  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1455  * on LP64 architectures, so cast to u_long to avoid undefined
1456  * behavior.  ILP32 architectures cannot have nmbclusters
1457  * large enough to overflow for other reasons.
1458  */
1459 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1460
1461         level = sopt->sopt_level;
1462         op = sopt->sopt_dir;
1463         optname = sopt->sopt_name;
1464         optlen = sopt->sopt_valsize;
1465         td = sopt->sopt_td;
1466         error = 0;
1467         optval = 0;
1468         uproto = (int)so->so_proto->pr_protocol;
1469
1470         if (level != IPPROTO_IPV6) {
1471                 error = EINVAL;
1472
1473                 if (sopt->sopt_level == SOL_SOCKET &&
1474                     sopt->sopt_dir == SOPT_SET) {
1475                         switch (sopt->sopt_name) {
1476                         case SO_REUSEADDR:
1477                                 INP_WLOCK(in6p);
1478                                 if ((so->so_options & SO_REUSEADDR) != 0)
1479                                         in6p->inp_flags2 |= INP_REUSEADDR;
1480                                 else
1481                                         in6p->inp_flags2 &= ~INP_REUSEADDR;
1482                                 INP_WUNLOCK(in6p);
1483                                 error = 0;
1484                                 break;
1485                         case SO_REUSEPORT:
1486                                 INP_WLOCK(in6p);
1487                                 if ((so->so_options & SO_REUSEPORT) != 0)
1488                                         in6p->inp_flags2 |= INP_REUSEPORT;
1489                                 else
1490                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1491                                 INP_WUNLOCK(in6p);
1492                                 error = 0;
1493                                 break;
1494                         case SO_REUSEPORT_LB:
1495                                 INP_WLOCK(in6p);
1496                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1497                                         in6p->inp_flags2 |= INP_REUSEPORT_LB;
1498                                 else
1499                                         in6p->inp_flags2 &= ~INP_REUSEPORT_LB;
1500                                 INP_WUNLOCK(in6p);
1501                                 error = 0;
1502                                 break;
1503                         case SO_SETFIB:
1504                                 INP_WLOCK(in6p);
1505                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1506                                 INP_WUNLOCK(in6p);
1507                                 error = 0;
1508                                 break;
1509                         case SO_MAX_PACING_RATE:
1510 #ifdef RATELIMIT
1511                                 INP_WLOCK(in6p);
1512                                 in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1513                                 INP_WUNLOCK(in6p);
1514                                 error = 0;
1515 #else
1516                                 error = EOPNOTSUPP;
1517 #endif
1518                                 break;
1519                         default:
1520                                 break;
1521                         }
1522                 }
1523         } else {                /* level == IPPROTO_IPV6 */
1524                 switch (op) {
1525
1526                 case SOPT_SET:
1527                         switch (optname) {
1528                         case IPV6_2292PKTOPTIONS:
1529 #ifdef IPV6_PKTOPTIONS
1530                         case IPV6_PKTOPTIONS:
1531 #endif
1532                         {
1533                                 struct mbuf *m;
1534
1535                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1536                                         printf("ip6_ctloutput: mbuf limit hit\n");
1537                                         error = ENOBUFS;
1538                                         break;
1539                                 }
1540
1541                                 error = soopt_getm(sopt, &m); /* XXX */
1542                                 if (error != 0)
1543                                         break;
1544                                 error = soopt_mcopyin(sopt, m); /* XXX */
1545                                 if (error != 0)
1546                                         break;
1547                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1548                                                     m, so, sopt);
1549                                 m_freem(m); /* XXX */
1550                                 break;
1551                         }
1552
1553                         /*
1554                          * Use of some Hop-by-Hop options or some
1555                          * Destination options, might require special
1556                          * privilege.  That is, normal applications
1557                          * (without special privilege) might be forbidden
1558                          * from setting certain options in outgoing packets,
1559                          * and might never see certain options in received
1560                          * packets. [RFC 2292 Section 6]
1561                          * KAME specific note:
1562                          *  KAME prevents non-privileged users from sending or
1563                          *  receiving ANY hbh/dst options in order to avoid
1564                          *  overhead of parsing options in the kernel.
1565                          */
1566                         case IPV6_RECVHOPOPTS:
1567                         case IPV6_RECVDSTOPTS:
1568                         case IPV6_RECVRTHDRDSTOPTS:
1569                                 if (td != NULL) {
1570                                         error = priv_check(td,
1571                                             PRIV_NETINET_SETHDROPTS);
1572                                         if (error)
1573                                                 break;
1574                                 }
1575                                 /* FALLTHROUGH */
1576                         case IPV6_UNICAST_HOPS:
1577                         case IPV6_HOPLIMIT:
1578
1579                         case IPV6_RECVPKTINFO:
1580                         case IPV6_RECVHOPLIMIT:
1581                         case IPV6_RECVRTHDR:
1582                         case IPV6_RECVPATHMTU:
1583                         case IPV6_RECVTCLASS:
1584                         case IPV6_RECVFLOWID:
1585 #ifdef  RSS
1586                         case IPV6_RECVRSSBUCKETID:
1587 #endif
1588                         case IPV6_V6ONLY:
1589                         case IPV6_AUTOFLOWLABEL:
1590                         case IPV6_ORIGDSTADDR:
1591                         case IPV6_BINDANY:
1592                         case IPV6_BINDMULTI:
1593 #ifdef  RSS
1594                         case IPV6_RSS_LISTEN_BUCKET:
1595 #endif
1596                                 if (optname == IPV6_BINDANY && td != NULL) {
1597                                         error = priv_check(td,
1598                                             PRIV_NETINET_BINDANY);
1599                                         if (error)
1600                                                 break;
1601                                 }
1602
1603                                 if (optlen != sizeof(int)) {
1604                                         error = EINVAL;
1605                                         break;
1606                                 }
1607                                 error = sooptcopyin(sopt, &optval,
1608                                         sizeof optval, sizeof optval);
1609                                 if (error)
1610                                         break;
1611                                 switch (optname) {
1612
1613                                 case IPV6_UNICAST_HOPS:
1614                                         if (optval < -1 || optval >= 256)
1615                                                 error = EINVAL;
1616                                         else {
1617                                                 /* -1 = kernel default */
1618                                                 in6p->in6p_hops = optval;
1619                                                 if ((in6p->inp_vflag &
1620                                                      INP_IPV4) != 0)
1621                                                         in6p->inp_ip_ttl = optval;
1622                                         }
1623                                         break;
1624 #define OPTSET(bit) \
1625 do { \
1626         INP_WLOCK(in6p); \
1627         if (optval) \
1628                 in6p->inp_flags |= (bit); \
1629         else \
1630                 in6p->inp_flags &= ~(bit); \
1631         INP_WUNLOCK(in6p); \
1632 } while (/*CONSTCOND*/ 0)
1633 #define OPTSET2292(bit) \
1634 do { \
1635         INP_WLOCK(in6p); \
1636         in6p->inp_flags |= IN6P_RFC2292; \
1637         if (optval) \
1638                 in6p->inp_flags |= (bit); \
1639         else \
1640                 in6p->inp_flags &= ~(bit); \
1641         INP_WUNLOCK(in6p); \
1642 } while (/*CONSTCOND*/ 0)
1643 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1644
1645 #define OPTSET2_N(bit, val) do {                                        \
1646         if (val)                                                        \
1647                 in6p->inp_flags2 |= bit;                                \
1648         else                                                            \
1649                 in6p->inp_flags2 &= ~bit;                               \
1650 } while (0)
1651 #define OPTSET2(bit, val) do {                                          \
1652         INP_WLOCK(in6p);                                                \
1653         OPTSET2_N(bit, val);                                            \
1654         INP_WUNLOCK(in6p);                                              \
1655 } while (0)
1656 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1657 #define OPTSET2292_EXCLUSIVE(bit)                                       \
1658 do {                                                                    \
1659         INP_WLOCK(in6p);                                                \
1660         if (OPTBIT(IN6P_RFC2292)) {                                     \
1661                 error = EINVAL;                                         \
1662         } else {                                                        \
1663                 if (optval)                                             \
1664                         in6p->inp_flags |= (bit);                       \
1665                 else                                                    \
1666                         in6p->inp_flags &= ~(bit);                      \
1667         }                                                               \
1668         INP_WUNLOCK(in6p);                                              \
1669 } while (/*CONSTCOND*/ 0)
1670
1671                                 case IPV6_RECVPKTINFO:
1672                                         OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1673                                         break;
1674
1675                                 case IPV6_HOPLIMIT:
1676                                 {
1677                                         struct ip6_pktopts **optp;
1678
1679                                         /* cannot mix with RFC2292 */
1680                                         if (OPTBIT(IN6P_RFC2292)) {
1681                                                 error = EINVAL;
1682                                                 break;
1683                                         }
1684                                         INP_WLOCK(in6p);
1685                                         if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1686                                                 INP_WUNLOCK(in6p);
1687                                                 return (ECONNRESET);
1688                                         }
1689                                         optp = &in6p->in6p_outputopts;
1690                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1691                                             (u_char *)&optval, sizeof(optval),
1692                                             optp, (td != NULL) ? td->td_ucred :
1693                                             NULL, uproto);
1694                                         INP_WUNLOCK(in6p);
1695                                         break;
1696                                 }
1697
1698                                 case IPV6_RECVHOPLIMIT:
1699                                         OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1700                                         break;
1701
1702                                 case IPV6_RECVHOPOPTS:
1703                                         OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1704                                         break;
1705
1706                                 case IPV6_RECVDSTOPTS:
1707                                         OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1708                                         break;
1709
1710                                 case IPV6_RECVRTHDRDSTOPTS:
1711                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1712                                         break;
1713
1714                                 case IPV6_RECVRTHDR:
1715                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1716                                         break;
1717
1718                                 case IPV6_RECVPATHMTU:
1719                                         /*
1720                                          * We ignore this option for TCP
1721                                          * sockets.
1722                                          * (RFC3542 leaves this case
1723                                          * unspecified.)
1724                                          */
1725                                         if (uproto != IPPROTO_TCP)
1726                                                 OPTSET(IN6P_MTU);
1727                                         break;
1728
1729                                 case IPV6_RECVFLOWID:
1730                                         OPTSET2(INP_RECVFLOWID, optval);
1731                                         break;
1732
1733 #ifdef  RSS
1734                                 case IPV6_RECVRSSBUCKETID:
1735                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1736                                         break;
1737 #endif
1738
1739                                 case IPV6_V6ONLY:
1740                                         /*
1741                                          * make setsockopt(IPV6_V6ONLY)
1742                                          * available only prior to bind(2).
1743                                          * see ipng mailing list, Jun 22 2001.
1744                                          */
1745                                         if (in6p->inp_lport ||
1746                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1747                                                 error = EINVAL;
1748                                                 break;
1749                                         }
1750                                         OPTSET(IN6P_IPV6_V6ONLY);
1751                                         if (optval)
1752                                                 in6p->inp_vflag &= ~INP_IPV4;
1753                                         else
1754                                                 in6p->inp_vflag |= INP_IPV4;
1755                                         break;
1756                                 case IPV6_RECVTCLASS:
1757                                         /* cannot mix with RFC2292 XXX */
1758                                         OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1759                                         break;
1760                                 case IPV6_AUTOFLOWLABEL:
1761                                         OPTSET(IN6P_AUTOFLOWLABEL);
1762                                         break;
1763
1764                                 case IPV6_ORIGDSTADDR:
1765                                         OPTSET2(INP_ORIGDSTADDR, optval);
1766                                         break;
1767                                 case IPV6_BINDANY:
1768                                         OPTSET(INP_BINDANY);
1769                                         break;
1770
1771                                 case IPV6_BINDMULTI:
1772                                         OPTSET2(INP_BINDMULTI, optval);
1773                                         break;
1774 #ifdef  RSS
1775                                 case IPV6_RSS_LISTEN_BUCKET:
1776                                         if ((optval >= 0) &&
1777                                             (optval < rss_getnumbuckets())) {
1778                                                 INP_WLOCK(in6p);
1779                                                 in6p->inp_rss_listen_bucket = optval;
1780                                                 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1781                                                 INP_WUNLOCK(in6p);
1782                                         } else {
1783                                                 error = EINVAL;
1784                                         }
1785                                         break;
1786 #endif
1787                                 }
1788                                 break;
1789
1790                         case IPV6_TCLASS:
1791                         case IPV6_DONTFRAG:
1792                         case IPV6_USE_MIN_MTU:
1793                         case IPV6_PREFER_TEMPADDR:
1794                                 if (optlen != sizeof(optval)) {
1795                                         error = EINVAL;
1796                                         break;
1797                                 }
1798                                 error = sooptcopyin(sopt, &optval,
1799                                         sizeof optval, sizeof optval);
1800                                 if (error)
1801                                         break;
1802                                 {
1803                                         struct ip6_pktopts **optp;
1804                                         INP_WLOCK(in6p);
1805                                         if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1806                                                 INP_WUNLOCK(in6p);
1807                                                 return (ECONNRESET);
1808                                         }
1809                                         optp = &in6p->in6p_outputopts;
1810                                         error = ip6_pcbopt(optname,
1811                                             (u_char *)&optval, sizeof(optval),
1812                                             optp, (td != NULL) ? td->td_ucred :
1813                                             NULL, uproto);
1814                                         INP_WUNLOCK(in6p);
1815                                         break;
1816                                 }
1817
1818                         case IPV6_2292PKTINFO:
1819                         case IPV6_2292HOPLIMIT:
1820                         case IPV6_2292HOPOPTS:
1821                         case IPV6_2292DSTOPTS:
1822                         case IPV6_2292RTHDR:
1823                                 /* RFC 2292 */
1824                                 if (optlen != sizeof(int)) {
1825                                         error = EINVAL;
1826                                         break;
1827                                 }
1828                                 error = sooptcopyin(sopt, &optval,
1829                                         sizeof optval, sizeof optval);
1830                                 if (error)
1831                                         break;
1832                                 switch (optname) {
1833                                 case IPV6_2292PKTINFO:
1834                                         OPTSET2292(IN6P_PKTINFO);
1835                                         break;
1836                                 case IPV6_2292HOPLIMIT:
1837                                         OPTSET2292(IN6P_HOPLIMIT);
1838                                         break;
1839                                 case IPV6_2292HOPOPTS:
1840                                         /*
1841                                          * Check super-user privilege.
1842                                          * See comments for IPV6_RECVHOPOPTS.
1843                                          */
1844                                         if (td != NULL) {
1845                                                 error = priv_check(td,
1846                                                     PRIV_NETINET_SETHDROPTS);
1847                                                 if (error)
1848                                                         return (error);
1849                                         }
1850                                         OPTSET2292(IN6P_HOPOPTS);
1851                                         break;
1852                                 case IPV6_2292DSTOPTS:
1853                                         if (td != NULL) {
1854                                                 error = priv_check(td,
1855                                                     PRIV_NETINET_SETHDROPTS);
1856                                                 if (error)
1857                                                         return (error);
1858                                         }
1859                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1860                                         break;
1861                                 case IPV6_2292RTHDR:
1862                                         OPTSET2292(IN6P_RTHDR);
1863                                         break;
1864                                 }
1865                                 break;
1866                         case IPV6_PKTINFO:
1867                         case IPV6_HOPOPTS:
1868                         case IPV6_RTHDR:
1869                         case IPV6_DSTOPTS:
1870                         case IPV6_RTHDRDSTOPTS:
1871                         case IPV6_NEXTHOP:
1872                         {
1873                                 /* new advanced API (RFC3542) */
1874                                 u_char *optbuf;
1875                                 u_char optbuf_storage[MCLBYTES];
1876                                 int optlen;
1877                                 struct ip6_pktopts **optp;
1878
1879                                 /* cannot mix with RFC2292 */
1880                                 if (OPTBIT(IN6P_RFC2292)) {
1881                                         error = EINVAL;
1882                                         break;
1883                                 }
1884
1885                                 /*
1886                                  * We only ensure valsize is not too large
1887                                  * here.  Further validation will be done
1888                                  * later.
1889                                  */
1890                                 error = sooptcopyin(sopt, optbuf_storage,
1891                                     sizeof(optbuf_storage), 0);
1892                                 if (error)
1893                                         break;
1894                                 optlen = sopt->sopt_valsize;
1895                                 optbuf = optbuf_storage;
1896                                 INP_WLOCK(in6p);
1897                                 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1898                                         INP_WUNLOCK(in6p);
1899                                         return (ECONNRESET);
1900                                 }
1901                                 optp = &in6p->in6p_outputopts;
1902                                 error = ip6_pcbopt(optname, optbuf, optlen,
1903                                     optp, (td != NULL) ? td->td_ucred : NULL,
1904                                     uproto);
1905                                 INP_WUNLOCK(in6p);
1906                                 break;
1907                         }
1908 #undef OPTSET
1909
1910                         case IPV6_MULTICAST_IF:
1911                         case IPV6_MULTICAST_HOPS:
1912                         case IPV6_MULTICAST_LOOP:
1913                         case IPV6_JOIN_GROUP:
1914                         case IPV6_LEAVE_GROUP:
1915                         case IPV6_MSFILTER:
1916                         case MCAST_BLOCK_SOURCE:
1917                         case MCAST_UNBLOCK_SOURCE:
1918                         case MCAST_JOIN_GROUP:
1919                         case MCAST_LEAVE_GROUP:
1920                         case MCAST_JOIN_SOURCE_GROUP:
1921                         case MCAST_LEAVE_SOURCE_GROUP:
1922                                 error = ip6_setmoptions(in6p, sopt);
1923                                 break;
1924
1925                         case IPV6_PORTRANGE:
1926                                 error = sooptcopyin(sopt, &optval,
1927                                     sizeof optval, sizeof optval);
1928                                 if (error)
1929                                         break;
1930
1931                                 INP_WLOCK(in6p);
1932                                 switch (optval) {
1933                                 case IPV6_PORTRANGE_DEFAULT:
1934                                         in6p->inp_flags &= ~(INP_LOWPORT);
1935                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1936                                         break;
1937
1938                                 case IPV6_PORTRANGE_HIGH:
1939                                         in6p->inp_flags &= ~(INP_LOWPORT);
1940                                         in6p->inp_flags |= INP_HIGHPORT;
1941                                         break;
1942
1943                                 case IPV6_PORTRANGE_LOW:
1944                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1945                                         in6p->inp_flags |= INP_LOWPORT;
1946                                         break;
1947
1948                                 default:
1949                                         error = EINVAL;
1950                                         break;
1951                                 }
1952                                 INP_WUNLOCK(in6p);
1953                                 break;
1954
1955 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1956                         case IPV6_IPSEC_POLICY:
1957                                 if (IPSEC_ENABLED(ipv6)) {
1958                                         error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1959                                         break;
1960                                 }
1961                                 /* FALLTHROUGH */
1962 #endif /* IPSEC */
1963
1964                         default:
1965                                 error = ENOPROTOOPT;
1966                                 break;
1967                         }
1968                         break;
1969
1970                 case SOPT_GET:
1971                         switch (optname) {
1972
1973                         case IPV6_2292PKTOPTIONS:
1974 #ifdef IPV6_PKTOPTIONS
1975                         case IPV6_PKTOPTIONS:
1976 #endif
1977                                 /*
1978                                  * RFC3542 (effectively) deprecated the
1979                                  * semantics of the 2292-style pktoptions.
1980                                  * Since it was not reliable in nature (i.e.,
1981                                  * applications had to expect the lack of some
1982                                  * information after all), it would make sense
1983                                  * to simplify this part by always returning
1984                                  * empty data.
1985                                  */
1986                                 sopt->sopt_valsize = 0;
1987                                 break;
1988
1989                         case IPV6_RECVHOPOPTS:
1990                         case IPV6_RECVDSTOPTS:
1991                         case IPV6_RECVRTHDRDSTOPTS:
1992                         case IPV6_UNICAST_HOPS:
1993                         case IPV6_RECVPKTINFO:
1994                         case IPV6_RECVHOPLIMIT:
1995                         case IPV6_RECVRTHDR:
1996                         case IPV6_RECVPATHMTU:
1997
1998                         case IPV6_V6ONLY:
1999                         case IPV6_PORTRANGE:
2000                         case IPV6_RECVTCLASS:
2001                         case IPV6_AUTOFLOWLABEL:
2002                         case IPV6_BINDANY:
2003                         case IPV6_FLOWID:
2004                         case IPV6_FLOWTYPE:
2005                         case IPV6_RECVFLOWID:
2006 #ifdef  RSS
2007                         case IPV6_RSSBUCKETID:
2008                         case IPV6_RECVRSSBUCKETID:
2009 #endif
2010                         case IPV6_BINDMULTI:
2011                                 switch (optname) {
2012
2013                                 case IPV6_RECVHOPOPTS:
2014                                         optval = OPTBIT(IN6P_HOPOPTS);
2015                                         break;
2016
2017                                 case IPV6_RECVDSTOPTS:
2018                                         optval = OPTBIT(IN6P_DSTOPTS);
2019                                         break;
2020
2021                                 case IPV6_RECVRTHDRDSTOPTS:
2022                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2023                                         break;
2024
2025                                 case IPV6_UNICAST_HOPS:
2026                                         optval = in6p->in6p_hops;
2027                                         break;
2028
2029                                 case IPV6_RECVPKTINFO:
2030                                         optval = OPTBIT(IN6P_PKTINFO);
2031                                         break;
2032
2033                                 case IPV6_RECVHOPLIMIT:
2034                                         optval = OPTBIT(IN6P_HOPLIMIT);
2035                                         break;
2036
2037                                 case IPV6_RECVRTHDR:
2038                                         optval = OPTBIT(IN6P_RTHDR);
2039                                         break;
2040
2041                                 case IPV6_RECVPATHMTU:
2042                                         optval = OPTBIT(IN6P_MTU);
2043                                         break;
2044
2045                                 case IPV6_V6ONLY:
2046                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2047                                         break;
2048
2049                                 case IPV6_PORTRANGE:
2050                                     {
2051                                         int flags;
2052                                         flags = in6p->inp_flags;
2053                                         if (flags & INP_HIGHPORT)
2054                                                 optval = IPV6_PORTRANGE_HIGH;
2055                                         else if (flags & INP_LOWPORT)
2056                                                 optval = IPV6_PORTRANGE_LOW;
2057                                         else
2058                                                 optval = 0;
2059                                         break;
2060                                     }
2061                                 case IPV6_RECVTCLASS:
2062                                         optval = OPTBIT(IN6P_TCLASS);
2063                                         break;
2064
2065                                 case IPV6_AUTOFLOWLABEL:
2066                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2067                                         break;
2068
2069                                 case IPV6_ORIGDSTADDR:
2070                                         optval = OPTBIT2(INP_ORIGDSTADDR);
2071                                         break;
2072
2073                                 case IPV6_BINDANY:
2074                                         optval = OPTBIT(INP_BINDANY);
2075                                         break;
2076
2077                                 case IPV6_FLOWID:
2078                                         optval = in6p->inp_flowid;
2079                                         break;
2080
2081                                 case IPV6_FLOWTYPE:
2082                                         optval = in6p->inp_flowtype;
2083                                         break;
2084
2085                                 case IPV6_RECVFLOWID:
2086                                         optval = OPTBIT2(INP_RECVFLOWID);
2087                                         break;
2088 #ifdef  RSS
2089                                 case IPV6_RSSBUCKETID:
2090                                         retval =
2091                                             rss_hash2bucket(in6p->inp_flowid,
2092                                             in6p->inp_flowtype,
2093                                             &rss_bucket);
2094                                         if (retval == 0)
2095                                                 optval = rss_bucket;
2096                                         else
2097                                                 error = EINVAL;
2098                                         break;
2099
2100                                 case IPV6_RECVRSSBUCKETID:
2101                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2102                                         break;
2103 #endif
2104
2105                                 case IPV6_BINDMULTI:
2106                                         optval = OPTBIT2(INP_BINDMULTI);
2107                                         break;
2108
2109                                 }
2110                                 if (error)
2111                                         break;
2112                                 error = sooptcopyout(sopt, &optval,
2113                                         sizeof optval);
2114                                 break;
2115
2116                         case IPV6_PATHMTU:
2117                         {
2118                                 u_long pmtu = 0;
2119                                 struct ip6_mtuinfo mtuinfo;
2120                                 struct in6_addr addr;
2121
2122                                 if (!(so->so_state & SS_ISCONNECTED))
2123                                         return (ENOTCONN);
2124                                 /*
2125                                  * XXX: we dot not consider the case of source
2126                                  * routing, or optional information to specify
2127                                  * the outgoing interface.
2128                                  * Copy faddr out of in6p to avoid holding lock
2129                                  * on inp during route lookup.
2130                                  */
2131                                 INP_RLOCK(in6p);
2132                                 bcopy(&in6p->in6p_faddr, &addr, sizeof(addr));
2133                                 INP_RUNLOCK(in6p);
2134                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2135                                     &addr, &pmtu);
2136                                 if (error)
2137                                         break;
2138                                 if (pmtu > IPV6_MAXPACKET)
2139                                         pmtu = IPV6_MAXPACKET;
2140
2141                                 bzero(&mtuinfo, sizeof(mtuinfo));
2142                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2143                                 optdata = (void *)&mtuinfo;
2144                                 optdatalen = sizeof(mtuinfo);
2145                                 error = sooptcopyout(sopt, optdata,
2146                                     optdatalen);
2147                                 break;
2148                         }
2149
2150                         case IPV6_2292PKTINFO:
2151                         case IPV6_2292HOPLIMIT:
2152                         case IPV6_2292HOPOPTS:
2153                         case IPV6_2292RTHDR:
2154                         case IPV6_2292DSTOPTS:
2155                                 switch (optname) {
2156                                 case IPV6_2292PKTINFO:
2157                                         optval = OPTBIT(IN6P_PKTINFO);
2158                                         break;
2159                                 case IPV6_2292HOPLIMIT:
2160                                         optval = OPTBIT(IN6P_HOPLIMIT);
2161                                         break;
2162                                 case IPV6_2292HOPOPTS:
2163                                         optval = OPTBIT(IN6P_HOPOPTS);
2164                                         break;
2165                                 case IPV6_2292RTHDR:
2166                                         optval = OPTBIT(IN6P_RTHDR);
2167                                         break;
2168                                 case IPV6_2292DSTOPTS:
2169                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2170                                         break;
2171                                 }
2172                                 error = sooptcopyout(sopt, &optval,
2173                                     sizeof optval);
2174                                 break;
2175                         case IPV6_PKTINFO:
2176                         case IPV6_HOPOPTS:
2177                         case IPV6_RTHDR:
2178                         case IPV6_DSTOPTS:
2179                         case IPV6_RTHDRDSTOPTS:
2180                         case IPV6_NEXTHOP:
2181                         case IPV6_TCLASS:
2182                         case IPV6_DONTFRAG:
2183                         case IPV6_USE_MIN_MTU:
2184                         case IPV6_PREFER_TEMPADDR:
2185                                 error = ip6_getpcbopt(in6p, optname, sopt);
2186                                 break;
2187
2188                         case IPV6_MULTICAST_IF:
2189                         case IPV6_MULTICAST_HOPS:
2190                         case IPV6_MULTICAST_LOOP:
2191                         case IPV6_MSFILTER:
2192                                 error = ip6_getmoptions(in6p, sopt);
2193                                 break;
2194
2195 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2196                         case IPV6_IPSEC_POLICY:
2197                                 if (IPSEC_ENABLED(ipv6)) {
2198                                         error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2199                                         break;
2200                                 }
2201                                 /* FALLTHROUGH */
2202 #endif /* IPSEC */
2203                         default:
2204                                 error = ENOPROTOOPT;
2205                                 break;
2206                         }
2207                         break;
2208                 }
2209         }
2210         return (error);
2211 }
2212
2213 int
2214 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2215 {
2216         int error = 0, optval, optlen;
2217         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2218         struct inpcb *in6p = sotoinpcb(so);
2219         int level, op, optname;
2220
2221         level = sopt->sopt_level;
2222         op = sopt->sopt_dir;
2223         optname = sopt->sopt_name;
2224         optlen = sopt->sopt_valsize;
2225
2226         if (level != IPPROTO_IPV6) {
2227                 return (EINVAL);
2228         }
2229
2230         switch (optname) {
2231         case IPV6_CHECKSUM:
2232                 /*
2233                  * For ICMPv6 sockets, no modification allowed for checksum
2234                  * offset, permit "no change" values to help existing apps.
2235                  *
2236                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2237                  * for an ICMPv6 socket will fail."
2238                  * The current behavior does not meet RFC3542.
2239                  */
2240                 switch (op) {
2241                 case SOPT_SET:
2242                         if (optlen != sizeof(int)) {
2243                                 error = EINVAL;
2244                                 break;
2245                         }
2246                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2247                                             sizeof(optval));
2248                         if (error)
2249                                 break;
2250                         if (optval < -1 || (optval % 2) != 0) {
2251                                 /*
2252                                  * The API assumes non-negative even offset
2253                                  * values or -1 as a special value.
2254                                  */
2255                                 error = EINVAL;
2256                         } else if (so->so_proto->pr_protocol ==
2257                             IPPROTO_ICMPV6) {
2258                                 if (optval != icmp6off)
2259                                         error = EINVAL;
2260                         } else
2261                                 in6p->in6p_cksum = optval;
2262                         break;
2263
2264                 case SOPT_GET:
2265                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2266                                 optval = icmp6off;
2267                         else
2268                                 optval = in6p->in6p_cksum;
2269
2270                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2271                         break;
2272
2273                 default:
2274                         error = EINVAL;
2275                         break;
2276                 }
2277                 break;
2278
2279         default:
2280                 error = ENOPROTOOPT;
2281                 break;
2282         }
2283
2284         return (error);
2285 }
2286
2287 /*
2288  * Set up IP6 options in pcb for insertion in output packets or
2289  * specifying behavior of outgoing packets.
2290  */
2291 static int
2292 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2293     struct socket *so, struct sockopt *sopt)
2294 {
2295         struct ip6_pktopts *opt = *pktopt;
2296         int error = 0;
2297         struct thread *td = sopt->sopt_td;
2298
2299         /* turn off any old options. */
2300         if (opt) {
2301 #ifdef DIAGNOSTIC
2302                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2303                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2304                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2305                         printf("ip6_pcbopts: all specified options are cleared.\n");
2306 #endif
2307                 ip6_clearpktopts(opt, -1);
2308         } else
2309                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2310         *pktopt = NULL;
2311
2312         if (!m || m->m_len == 0) {
2313                 /*
2314                  * Only turning off any previous options, regardless of
2315                  * whether the opt is just created or given.
2316                  */
2317                 free(opt, M_IP6OPT);
2318                 return (0);
2319         }
2320
2321         /*  set options specified by user. */
2322         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2323             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2324                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2325                 free(opt, M_IP6OPT);
2326                 return (error);
2327         }
2328         *pktopt = opt;
2329         return (0);
2330 }
2331
2332 /*
2333  * initialize ip6_pktopts.  beware that there are non-zero default values in
2334  * the struct.
2335  */
2336 void
2337 ip6_initpktopts(struct ip6_pktopts *opt)
2338 {
2339
2340         bzero(opt, sizeof(*opt));
2341         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2342         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2343         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2344         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2345 }
2346
2347 static int
2348 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2349     struct ucred *cred, int uproto)
2350 {
2351         struct ip6_pktopts *opt;
2352
2353         if (*pktopt == NULL) {
2354                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2355                     M_NOWAIT);
2356                 if (*pktopt == NULL)
2357                         return (ENOBUFS);
2358                 ip6_initpktopts(*pktopt);
2359         }
2360         opt = *pktopt;
2361
2362         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2363 }
2364
2365 #define GET_PKTOPT_VAR(field, lenexpr) do {                                     \
2366         if (pktopt && pktopt->field) {                                          \
2367                 INP_RUNLOCK(in6p);                                              \
2368                 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);         \
2369                 malloc_optdata = true;                                          \
2370                 INP_RLOCK(in6p);                                                \
2371                 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {           \
2372                         INP_RUNLOCK(in6p);                                      \
2373                         free(optdata, M_TEMP);                                  \
2374                         return (ECONNRESET);                                    \
2375                 }                                                               \
2376                 pktopt = in6p->in6p_outputopts;                                 \
2377                 if (pktopt && pktopt->field) {                                  \
2378                         optdatalen = min(lenexpr, sopt->sopt_valsize);          \
2379                         bcopy(&pktopt->field, optdata, optdatalen);             \
2380                 } else {                                                        \
2381                         free(optdata, M_TEMP);                                  \
2382                         optdata = NULL;                                         \
2383                         malloc_optdata = false;                                 \
2384                 }                                                               \
2385         }                                                                       \
2386 } while(0)
2387
2388 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,                         \
2389         (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2390
2391 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,                        \
2392         pktopt->field->sa_len)
2393
2394 static int
2395 ip6_getpcbopt(struct inpcb *in6p, int optname, struct sockopt *sopt)
2396 {
2397         void *optdata = NULL;
2398         bool malloc_optdata = false;
2399         int optdatalen = 0;
2400         int error = 0;
2401         struct in6_pktinfo null_pktinfo;
2402         int deftclass = 0, on;
2403         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2404         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2405         struct ip6_pktopts *pktopt;
2406
2407         INP_RLOCK(in6p);
2408         pktopt = in6p->in6p_outputopts;
2409
2410         switch (optname) {
2411         case IPV6_PKTINFO:
2412                 optdata = (void *)&null_pktinfo;
2413                 if (pktopt && pktopt->ip6po_pktinfo) {
2414                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2415                             sizeof(null_pktinfo));
2416                         in6_clearscope(&null_pktinfo.ipi6_addr);
2417                 } else {
2418                         /* XXX: we don't have to do this every time... */
2419                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2420                 }
2421                 optdatalen = sizeof(struct in6_pktinfo);
2422                 break;
2423         case IPV6_TCLASS:
2424                 if (pktopt && pktopt->ip6po_tclass >= 0)
2425                         deftclass = pktopt->ip6po_tclass;
2426                 optdata = (void *)&deftclass;
2427                 optdatalen = sizeof(int);
2428                 break;
2429         case IPV6_HOPOPTS:
2430                 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2431                 break;
2432         case IPV6_RTHDR:
2433                 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2434                 break;
2435         case IPV6_RTHDRDSTOPTS:
2436                 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2437                 break;
2438         case IPV6_DSTOPTS:
2439                 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2440                 break;
2441         case IPV6_NEXTHOP:
2442                 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2443                 break;
2444         case IPV6_USE_MIN_MTU:
2445                 if (pktopt)
2446                         defminmtu = pktopt->ip6po_minmtu;
2447                 optdata = (void *)&defminmtu;
2448                 optdatalen = sizeof(int);
2449                 break;
2450         case IPV6_DONTFRAG:
2451                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2452                         on = 1;
2453                 else
2454                         on = 0;
2455                 optdata = (void *)&on;
2456                 optdatalen = sizeof(on);
2457                 break;
2458         case IPV6_PREFER_TEMPADDR:
2459                 if (pktopt)
2460                         defpreftemp = pktopt->ip6po_prefer_tempaddr;
2461                 optdata = (void *)&defpreftemp;
2462                 optdatalen = sizeof(int);
2463                 break;
2464         default:                /* should not happen */
2465 #ifdef DIAGNOSTIC
2466                 panic("ip6_getpcbopt: unexpected option\n");
2467 #endif
2468                 INP_RUNLOCK(in6p);
2469                 return (ENOPROTOOPT);
2470         }
2471         INP_RUNLOCK(in6p);
2472
2473         error = sooptcopyout(sopt, optdata, optdatalen);
2474         if (malloc_optdata)
2475                 free(optdata, M_TEMP);
2476
2477         return (error);
2478 }
2479
2480 void
2481 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2482 {
2483         if (pktopt == NULL)
2484                 return;
2485
2486         if (optname == -1 || optname == IPV6_PKTINFO) {
2487                 if (pktopt->ip6po_pktinfo)
2488                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2489                 pktopt->ip6po_pktinfo = NULL;
2490         }
2491         if (optname == -1 || optname == IPV6_HOPLIMIT)
2492                 pktopt->ip6po_hlim = -1;
2493         if (optname == -1 || optname == IPV6_TCLASS)
2494                 pktopt->ip6po_tclass = -1;
2495         if (optname == -1 || optname == IPV6_NEXTHOP) {
2496                 if (pktopt->ip6po_nextroute.ro_rt) {
2497                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2498                         pktopt->ip6po_nextroute.ro_rt = NULL;
2499                 }
2500                 if (pktopt->ip6po_nexthop)
2501                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2502                 pktopt->ip6po_nexthop = NULL;
2503         }
2504         if (optname == -1 || optname == IPV6_HOPOPTS) {
2505                 if (pktopt->ip6po_hbh)
2506                         free(pktopt->ip6po_hbh, M_IP6OPT);
2507                 pktopt->ip6po_hbh = NULL;
2508         }
2509         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2510                 if (pktopt->ip6po_dest1)
2511                         free(pktopt->ip6po_dest1, M_IP6OPT);
2512                 pktopt->ip6po_dest1 = NULL;
2513         }
2514         if (optname == -1 || optname == IPV6_RTHDR) {
2515                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2516                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2517                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2518                 if (pktopt->ip6po_route.ro_rt) {
2519                         RTFREE(pktopt->ip6po_route.ro_rt);
2520                         pktopt->ip6po_route.ro_rt = NULL;
2521                 }
2522         }
2523         if (optname == -1 || optname == IPV6_DSTOPTS) {
2524                 if (pktopt->ip6po_dest2)
2525                         free(pktopt->ip6po_dest2, M_IP6OPT);
2526                 pktopt->ip6po_dest2 = NULL;
2527         }
2528 }
2529
2530 #define PKTOPT_EXTHDRCPY(type) \
2531 do {\
2532         if (src->type) {\
2533                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2534                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2535                 if (dst->type == NULL)\
2536                         goto bad;\
2537                 bcopy(src->type, dst->type, hlen);\
2538         }\
2539 } while (/*CONSTCOND*/ 0)
2540
2541 static int
2542 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2543 {
2544         if (dst == NULL || src == NULL)  {
2545                 printf("ip6_clearpktopts: invalid argument\n");
2546                 return (EINVAL);
2547         }
2548
2549         dst->ip6po_hlim = src->ip6po_hlim;
2550         dst->ip6po_tclass = src->ip6po_tclass;
2551         dst->ip6po_flags = src->ip6po_flags;
2552         dst->ip6po_minmtu = src->ip6po_minmtu;
2553         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2554         if (src->ip6po_pktinfo) {
2555                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2556                     M_IP6OPT, canwait);
2557                 if (dst->ip6po_pktinfo == NULL)
2558                         goto bad;
2559                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2560         }
2561         if (src->ip6po_nexthop) {
2562                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2563                     M_IP6OPT, canwait);
2564                 if (dst->ip6po_nexthop == NULL)
2565                         goto bad;
2566                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2567                     src->ip6po_nexthop->sa_len);
2568         }
2569         PKTOPT_EXTHDRCPY(ip6po_hbh);
2570         PKTOPT_EXTHDRCPY(ip6po_dest1);
2571         PKTOPT_EXTHDRCPY(ip6po_dest2);
2572         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2573         return (0);
2574
2575   bad:
2576         ip6_clearpktopts(dst, -1);
2577         return (ENOBUFS);
2578 }
2579 #undef PKTOPT_EXTHDRCPY
2580
2581 struct ip6_pktopts *
2582 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2583 {
2584         int error;
2585         struct ip6_pktopts *dst;
2586
2587         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2588         if (dst == NULL)
2589                 return (NULL);
2590         ip6_initpktopts(dst);
2591
2592         if ((error = copypktopts(dst, src, canwait)) != 0) {
2593                 free(dst, M_IP6OPT);
2594                 return (NULL);
2595         }
2596
2597         return (dst);
2598 }
2599
2600 void
2601 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2602 {
2603         if (pktopt == NULL)
2604                 return;
2605
2606         ip6_clearpktopts(pktopt, -1);
2607
2608         free(pktopt, M_IP6OPT);
2609 }
2610
2611 /*
2612  * Set IPv6 outgoing packet options based on advanced API.
2613  */
2614 int
2615 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2616     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2617 {
2618         struct cmsghdr *cm = NULL;
2619
2620         if (control == NULL || opt == NULL)
2621                 return (EINVAL);
2622
2623         ip6_initpktopts(opt);
2624         if (stickyopt) {
2625                 int error;
2626
2627                 /*
2628                  * If stickyopt is provided, make a local copy of the options
2629                  * for this particular packet, then override them by ancillary
2630                  * objects.
2631                  * XXX: copypktopts() does not copy the cached route to a next
2632                  * hop (if any).  This is not very good in terms of efficiency,
2633                  * but we can allow this since this option should be rarely
2634                  * used.
2635                  */
2636                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2637                         return (error);
2638         }
2639
2640         /*
2641          * XXX: Currently, we assume all the optional information is stored
2642          * in a single mbuf.
2643          */
2644         if (control->m_next)
2645                 return (EINVAL);
2646
2647         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2648             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2649                 int error;
2650
2651                 if (control->m_len < CMSG_LEN(0))
2652                         return (EINVAL);
2653
2654                 cm = mtod(control, struct cmsghdr *);
2655                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2656                         return (EINVAL);
2657                 if (cm->cmsg_level != IPPROTO_IPV6)
2658                         continue;
2659
2660                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2661                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2662                 if (error)
2663                         return (error);
2664         }
2665
2666         return (0);
2667 }
2668
2669 /*
2670  * Set a particular packet option, as a sticky option or an ancillary data
2671  * item.  "len" can be 0 only when it's a sticky option.
2672  * We have 4 cases of combination of "sticky" and "cmsg":
2673  * "sticky=0, cmsg=0": impossible
2674  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2675  * "sticky=1, cmsg=0": RFC3542 socket option
2676  * "sticky=1, cmsg=1": RFC2292 socket option
2677  */
2678 static int
2679 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2680     struct ucred *cred, int sticky, int cmsg, int uproto)
2681 {
2682         int minmtupolicy, preftemp;
2683         int error;
2684
2685         if (!sticky && !cmsg) {
2686 #ifdef DIAGNOSTIC
2687                 printf("ip6_setpktopt: impossible case\n");
2688 #endif
2689                 return (EINVAL);
2690         }
2691
2692         /*
2693          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2694          * not be specified in the context of RFC3542.  Conversely,
2695          * RFC3542 types should not be specified in the context of RFC2292.
2696          */
2697         if (!cmsg) {
2698                 switch (optname) {
2699                 case IPV6_2292PKTINFO:
2700                 case IPV6_2292HOPLIMIT:
2701                 case IPV6_2292NEXTHOP:
2702                 case IPV6_2292HOPOPTS:
2703                 case IPV6_2292DSTOPTS:
2704                 case IPV6_2292RTHDR:
2705                 case IPV6_2292PKTOPTIONS:
2706                         return (ENOPROTOOPT);
2707                 }
2708         }
2709         if (sticky && cmsg) {
2710                 switch (optname) {
2711                 case IPV6_PKTINFO:
2712                 case IPV6_HOPLIMIT:
2713                 case IPV6_NEXTHOP:
2714                 case IPV6_HOPOPTS:
2715                 case IPV6_DSTOPTS:
2716                 case IPV6_RTHDRDSTOPTS:
2717                 case IPV6_RTHDR:
2718                 case IPV6_USE_MIN_MTU:
2719                 case IPV6_DONTFRAG:
2720                 case IPV6_TCLASS:
2721                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2722                         return (ENOPROTOOPT);
2723                 }
2724         }
2725
2726         switch (optname) {
2727         case IPV6_2292PKTINFO:
2728         case IPV6_PKTINFO:
2729         {
2730                 struct ifnet *ifp = NULL;
2731                 struct in6_pktinfo *pktinfo;
2732
2733                 if (len != sizeof(struct in6_pktinfo))
2734                         return (EINVAL);
2735
2736                 pktinfo = (struct in6_pktinfo *)buf;
2737
2738                 /*
2739                  * An application can clear any sticky IPV6_PKTINFO option by
2740                  * doing a "regular" setsockopt with ipi6_addr being
2741                  * in6addr_any and ipi6_ifindex being zero.
2742                  * [RFC 3542, Section 6]
2743                  */
2744                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2745                     pktinfo->ipi6_ifindex == 0 &&
2746                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2747                         ip6_clearpktopts(opt, optname);
2748                         break;
2749                 }
2750
2751                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2752                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2753                         return (EINVAL);
2754                 }
2755                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2756                         return (EINVAL);
2757                 /* validate the interface index if specified. */
2758                 if (pktinfo->ipi6_ifindex > V_if_index)
2759                          return (ENXIO);
2760                 if (pktinfo->ipi6_ifindex) {
2761                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2762                         if (ifp == NULL)
2763                                 return (ENXIO);
2764                 }
2765                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2766                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2767                         return (ENETDOWN);
2768
2769                 if (ifp != NULL &&
2770                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2771                         struct in6_ifaddr *ia;
2772
2773                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2774                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2775                         if (ia == NULL)
2776                                 return (EADDRNOTAVAIL);
2777                         ifa_free(&ia->ia_ifa);
2778                 }
2779                 /*
2780                  * We store the address anyway, and let in6_selectsrc()
2781                  * validate the specified address.  This is because ipi6_addr
2782                  * may not have enough information about its scope zone, and
2783                  * we may need additional information (such as outgoing
2784                  * interface or the scope zone of a destination address) to
2785                  * disambiguate the scope.
2786                  * XXX: the delay of the validation may confuse the
2787                  * application when it is used as a sticky option.
2788                  */
2789                 if (opt->ip6po_pktinfo == NULL) {
2790                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2791                             M_IP6OPT, M_NOWAIT);
2792                         if (opt->ip6po_pktinfo == NULL)
2793                                 return (ENOBUFS);
2794                 }
2795                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2796                 break;
2797         }
2798
2799         case IPV6_2292HOPLIMIT:
2800         case IPV6_HOPLIMIT:
2801         {
2802                 int *hlimp;
2803
2804                 /*
2805                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2806                  * to simplify the ordering among hoplimit options.
2807                  */
2808                 if (optname == IPV6_HOPLIMIT && sticky)
2809                         return (ENOPROTOOPT);
2810
2811                 if (len != sizeof(int))
2812                         return (EINVAL);
2813                 hlimp = (int *)buf;
2814                 if (*hlimp < -1 || *hlimp > 255)
2815                         return (EINVAL);
2816
2817                 opt->ip6po_hlim = *hlimp;
2818                 break;
2819         }
2820
2821         case IPV6_TCLASS:
2822         {
2823                 int tclass;
2824
2825                 if (len != sizeof(int))
2826                         return (EINVAL);
2827                 tclass = *(int *)buf;
2828                 if (tclass < -1 || tclass > 255)
2829                         return (EINVAL);
2830
2831                 opt->ip6po_tclass = tclass;
2832                 break;
2833         }
2834
2835         case IPV6_2292NEXTHOP:
2836         case IPV6_NEXTHOP:
2837                 if (cred != NULL) {
2838                         error = priv_check_cred(cred,
2839                             PRIV_NETINET_SETHDROPTS, 0);
2840                         if (error)
2841                                 return (error);
2842                 }
2843
2844                 if (len == 0) { /* just remove the option */
2845                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2846                         break;
2847                 }
2848
2849                 /* check if cmsg_len is large enough for sa_len */
2850                 if (len < sizeof(struct sockaddr) || len < *buf)
2851                         return (EINVAL);
2852
2853                 switch (((struct sockaddr *)buf)->sa_family) {
2854                 case AF_INET6:
2855                 {
2856                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2857                         int error;
2858
2859                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2860                                 return (EINVAL);
2861
2862                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2863                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2864                                 return (EINVAL);
2865                         }
2866                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2867                             != 0) {
2868                                 return (error);
2869                         }
2870                         break;
2871                 }
2872                 case AF_LINK:   /* should eventually be supported */
2873                 default:
2874                         return (EAFNOSUPPORT);
2875                 }
2876
2877                 /* turn off the previous option, then set the new option. */
2878                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2879                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2880                 if (opt->ip6po_nexthop == NULL)
2881                         return (ENOBUFS);
2882                 bcopy(buf, opt->ip6po_nexthop, *buf);
2883                 break;
2884
2885         case IPV6_2292HOPOPTS:
2886         case IPV6_HOPOPTS:
2887         {
2888                 struct ip6_hbh *hbh;
2889                 int hbhlen;
2890
2891                 /*
2892                  * XXX: We don't allow a non-privileged user to set ANY HbH
2893                  * options, since per-option restriction has too much
2894                  * overhead.
2895                  */
2896                 if (cred != NULL) {
2897                         error = priv_check_cred(cred,
2898                             PRIV_NETINET_SETHDROPTS, 0);
2899                         if (error)
2900                                 return (error);
2901                 }
2902
2903                 if (len == 0) {
2904                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2905                         break;  /* just remove the option */
2906                 }
2907
2908                 /* message length validation */
2909                 if (len < sizeof(struct ip6_hbh))
2910                         return (EINVAL);
2911                 hbh = (struct ip6_hbh *)buf;
2912                 hbhlen = (hbh->ip6h_len + 1) << 3;
2913                 if (len != hbhlen)
2914                         return (EINVAL);
2915
2916                 /* turn off the previous option, then set the new option. */
2917                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2918                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2919                 if (opt->ip6po_hbh == NULL)
2920                         return (ENOBUFS);
2921                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2922
2923                 break;
2924         }
2925
2926         case IPV6_2292DSTOPTS:
2927         case IPV6_DSTOPTS:
2928         case IPV6_RTHDRDSTOPTS:
2929         {
2930                 struct ip6_dest *dest, **newdest = NULL;
2931                 int destlen;
2932
2933                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2934                         error = priv_check_cred(cred,
2935                             PRIV_NETINET_SETHDROPTS, 0);
2936                         if (error)
2937                                 return (error);
2938                 }
2939
2940                 if (len == 0) {
2941                         ip6_clearpktopts(opt, optname);
2942                         break;  /* just remove the option */
2943                 }
2944
2945                 /* message length validation */
2946                 if (len < sizeof(struct ip6_dest))
2947                         return (EINVAL);
2948                 dest = (struct ip6_dest *)buf;
2949                 destlen = (dest->ip6d_len + 1) << 3;
2950                 if (len != destlen)
2951                         return (EINVAL);
2952
2953                 /*
2954                  * Determine the position that the destination options header
2955                  * should be inserted; before or after the routing header.
2956                  */
2957                 switch (optname) {
2958                 case IPV6_2292DSTOPTS:
2959                         /*
2960                          * The old advacned API is ambiguous on this point.
2961                          * Our approach is to determine the position based
2962                          * according to the existence of a routing header.
2963                          * Note, however, that this depends on the order of the
2964                          * extension headers in the ancillary data; the 1st
2965                          * part of the destination options header must appear
2966                          * before the routing header in the ancillary data,
2967                          * too.
2968                          * RFC3542 solved the ambiguity by introducing
2969                          * separate ancillary data or option types.
2970                          */
2971                         if (opt->ip6po_rthdr == NULL)
2972                                 newdest = &opt->ip6po_dest1;
2973                         else
2974                                 newdest = &opt->ip6po_dest2;
2975                         break;
2976                 case IPV6_RTHDRDSTOPTS:
2977                         newdest = &opt->ip6po_dest1;
2978                         break;
2979                 case IPV6_DSTOPTS:
2980                         newdest = &opt->ip6po_dest2;
2981                         break;
2982                 }
2983
2984                 /* turn off the previous option, then set the new option. */
2985                 ip6_clearpktopts(opt, optname);
2986                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2987                 if (*newdest == NULL)
2988                         return (ENOBUFS);
2989                 bcopy(dest, *newdest, destlen);
2990
2991                 break;
2992         }
2993
2994         case IPV6_2292RTHDR:
2995         case IPV6_RTHDR:
2996         {
2997                 struct ip6_rthdr *rth;
2998                 int rthlen;
2999
3000                 if (len == 0) {
3001                         ip6_clearpktopts(opt, IPV6_RTHDR);
3002                         break;  /* just remove the option */
3003                 }
3004
3005                 /* message length validation */
3006                 if (len < sizeof(struct ip6_rthdr))
3007                         return (EINVAL);
3008                 rth = (struct ip6_rthdr *)buf;
3009                 rthlen = (rth->ip6r_len + 1) << 3;
3010                 if (len != rthlen)
3011                         return (EINVAL);
3012
3013                 switch (rth->ip6r_type) {
3014                 case IPV6_RTHDR_TYPE_0:
3015                         if (rth->ip6r_len == 0) /* must contain one addr */
3016                                 return (EINVAL);
3017                         if (rth->ip6r_len % 2) /* length must be even */
3018                                 return (EINVAL);
3019                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3020                                 return (EINVAL);
3021                         break;
3022                 default:
3023                         return (EINVAL);        /* not supported */
3024                 }
3025
3026                 /* turn off the previous option */
3027                 ip6_clearpktopts(opt, IPV6_RTHDR);
3028                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3029                 if (opt->ip6po_rthdr == NULL)
3030                         return (ENOBUFS);
3031                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3032
3033                 break;
3034         }
3035
3036         case IPV6_USE_MIN_MTU:
3037                 if (len != sizeof(int))
3038                         return (EINVAL);
3039                 minmtupolicy = *(int *)buf;
3040                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3041                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3042                     minmtupolicy != IP6PO_MINMTU_ALL) {
3043                         return (EINVAL);
3044                 }
3045                 opt->ip6po_minmtu = minmtupolicy;
3046                 break;
3047
3048         case IPV6_DONTFRAG:
3049                 if (len != sizeof(int))
3050                         return (EINVAL);
3051
3052                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3053                         /*
3054                          * we ignore this option for TCP sockets.
3055                          * (RFC3542 leaves this case unspecified.)
3056                          */
3057                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3058                 } else
3059                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3060                 break;
3061
3062         case IPV6_PREFER_TEMPADDR:
3063                 if (len != sizeof(int))
3064                         return (EINVAL);
3065                 preftemp = *(int *)buf;
3066                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3067                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3068                     preftemp != IP6PO_TEMPADDR_PREFER) {
3069                         return (EINVAL);
3070                 }
3071                 opt->ip6po_prefer_tempaddr = preftemp;
3072                 break;
3073
3074         default:
3075                 return (ENOPROTOOPT);
3076         } /* end of switch */
3077
3078         return (0);
3079 }
3080
3081 /*
3082  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3083  * packet to the input queue of a specified interface.  Note that this
3084  * calls the output routine of the loopback "driver", but with an interface
3085  * pointer that might NOT be &loif -- easier than replicating that code here.
3086  */
3087 void
3088 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3089 {
3090         struct mbuf *copym;
3091         struct ip6_hdr *ip6;
3092
3093         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3094         if (copym == NULL)
3095                 return;
3096
3097         /*
3098          * Make sure to deep-copy IPv6 header portion in case the data
3099          * is in an mbuf cluster, so that we can safely override the IPv6
3100          * header portion later.
3101          */
3102         if (!M_WRITABLE(copym) ||
3103             copym->m_len < sizeof(struct ip6_hdr)) {
3104                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3105                 if (copym == NULL)
3106                         return;
3107         }
3108         ip6 = mtod(copym, struct ip6_hdr *);
3109         /*
3110          * clear embedded scope identifiers if necessary.
3111          * in6_clearscope will touch the addresses only when necessary.
3112          */
3113         in6_clearscope(&ip6->ip6_src);
3114         in6_clearscope(&ip6->ip6_dst);
3115         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3116                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3117                     CSUM_PSEUDO_HDR;
3118                 copym->m_pkthdr.csum_data = 0xffff;
3119         }
3120         if_simloop(ifp, copym, AF_INET6, 0);
3121 }
3122
3123 /*
3124  * Chop IPv6 header off from the payload.
3125  */
3126 static int
3127 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3128 {
3129         struct mbuf *mh;
3130         struct ip6_hdr *ip6;
3131
3132         ip6 = mtod(m, struct ip6_hdr *);
3133         if (m->m_len > sizeof(*ip6)) {
3134                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3135                 if (mh == NULL) {
3136                         m_freem(m);
3137                         return ENOBUFS;
3138                 }
3139                 m_move_pkthdr(mh, m);
3140                 M_ALIGN(mh, sizeof(*ip6));
3141                 m->m_len -= sizeof(*ip6);
3142                 m->m_data += sizeof(*ip6);
3143                 mh->m_next = m;
3144                 m = mh;
3145                 m->m_len = sizeof(*ip6);
3146                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3147         }
3148         exthdrs->ip6e_ip6 = m;
3149         return 0;
3150 }
3151
3152 /*
3153  * Compute IPv6 extension header length.
3154  */
3155 int
3156 ip6_optlen(struct inpcb *in6p)
3157 {
3158         int len;
3159
3160         if (!in6p->in6p_outputopts)
3161                 return 0;
3162
3163         len = 0;
3164 #define elen(x) \
3165     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3166
3167         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3168         if (in6p->in6p_outputopts->ip6po_rthdr)
3169                 /* dest1 is valid with rthdr only */
3170                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3171         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3172         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3173         return len;
3174 #undef elen
3175 }