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 int
 209 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
 210     int fraglen , uint32_t id)
 211 {
 212         struct mbuf *m, **mnext, *m_frgpart;
 213         struct ip6_hdr *ip6, *mhip6;
 214         struct ip6_frag *ip6f;
 215         int off;
 216         int error;
 217         int tlen = m0->m_pkthdr.len;
 218
 219         KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
 220
 221         m = m0;
 222         ip6 = mtod(m, struct ip6_hdr *);
 223         mnext = &m->m_nextpkt;
 224
 225         for (off = hlen; off < tlen; off += fraglen) {
 226                 m = m_gethdr(M_NOWAIT, MT_DATA);
 227                 if (!m) {
 228                         IP6STAT_INC(ip6s_odropped);
 229                         return (ENOBUFS);
 230                 }
 231
 232                 /*
 233                  * Make sure the complete packet header gets copied
 234                  * from the originating mbuf to the newly created
 235                  * mbuf. This also ensures that existing firewall
 236                  * classification(s), VLAN tags and so on get copied
 237                  * to the resulting fragmented packet(s):
 238                  */
 239                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 240                         m_free(m);
 241                         IP6STAT_INC(ip6s_odropped);
 242                         return (ENOBUFS);
 243                 }
 244
 245                 *mnext = m;
 246                 mnext = &m->m_nextpkt;
 247                 m->m_data += max_linkhdr;
 248                 mhip6 = mtod(m, struct ip6_hdr *);
 249                 *mhip6 = *ip6;
 250                 m->m_len = sizeof(*mhip6);
 251                 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 252                 if (error) {
 253                         IP6STAT_INC(ip6s_odropped);
 254                         return (error);
 255                 }
 256                 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 257                 if (off + fraglen >= tlen)
 258                         fraglen = tlen - off;
 259                 else
 260                         ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 261                 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
 262                     sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 263                 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
 264                         IP6STAT_INC(ip6s_odropped);
 265                         return (ENOBUFS);
 266                 }
 267                 m_cat(m, m_frgpart);
 268                 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
 269                 ip6f->ip6f_reserved = 0;
 270                 ip6f->ip6f_ident = id;
 271                 ip6f->ip6f_nxt = nextproto;
 272                 IP6STAT_INC(ip6s_ofragments);
 273                 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 274         }
 275
 276         return (0);
 277 }
 278
 279 static int
 280 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
 281     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro)
 282 {
 283         struct m_snd_tag *mst;
 284         int error;
 285
 286         MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
 287         mst = NULL;
 288
 289 #ifdef RATELIMIT
 290         if (inp != NULL) {
 291                 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
 292                     (inp->inp_snd_tag != NULL &&
 293                     inp->inp_snd_tag->ifp != ifp))
 294                         in_pcboutput_txrtlmt(inp, ifp, m);
 295
 296                 if (inp->inp_snd_tag != NULL)
 297                         mst = inp->inp_snd_tag;
 298         }
 299 #endif
 300         if (mst != NULL) {
 301                 KASSERT(m->m_pkthdr.rcvif == NULL,
 302                     ("trying to add a send tag to a forwarded packet"));
 303                 if (mst->ifp != ifp) {
 304                         error = EAGAIN;
 305                         goto done;
 306                 }
 307
 308                 /* stamp send tag on mbuf */
 309                 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
 310                 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
 311         }
 312
 313         error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
 314
 315 done:
 316         /* Check for route change invalidating send tags. */
 317 #ifdef RATELIMIT
 318         if (error == EAGAIN)
 319                 in_pcboutput_eagain(inp);
 320 #endif
 321         return (error);
 322 }
 323
 324 /*
 325  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 326  * header (with pri, len, nxt, hlim, src, dst).
 327  * This function may modify ver and hlim only.
 328  * The mbuf chain containing the packet will be freed.
 329  * The mbuf opt, if present, will not be freed.
 330  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
 331  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 332  * then result of route lookup is stored in ro->ro_rt.
 333  *
 334  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
 335  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 336  * which is rt_mtu.
 337  *
 338  * ifpp - XXX: just for statistics
 339  */
 340 /*
 341  * XXX TODO: no flowid is assigned for outbound flows?
 342  */
 343 int
 344 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 345     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 346     struct ifnet **ifpp, struct inpcb *inp)
 347 {
 348         struct ip6_hdr *ip6;
 349         struct ifnet *ifp, *origifp;
 350         struct mbuf *m = m0;
 351         struct mbuf *mprev = NULL;
 352         int hlen, tlen, len;
 353         struct route_in6 ip6route;
 354         struct rtentry *rt = NULL;
 355         struct sockaddr_in6 *dst, src_sa, dst_sa;
 356         struct in6_addr odst;
 357         int error = 0;
 358         struct in6_ifaddr *ia = NULL;
 359         u_long mtu;
 360         int alwaysfrag, dontfrag;
 361         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 362         struct ip6_exthdrs exthdrs;
 363         struct in6_addr src0, dst0;
 364         u_int32_t zone;
 365         struct route_in6 *ro_pmtu = NULL;
 366         int hdrsplit = 0;
 367         int sw_csum, tso;
 368         int needfiblookup;
 369         uint32_t fibnum;
 370         struct m_tag *fwd_tag = NULL;
 371         uint32_t id;
 372
 373         if (inp != NULL) {
 374                 INP_LOCK_ASSERT(inp);
 375                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 376                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 377                         /* unconditionally set flowid */
 378                         m->m_pkthdr.flowid = inp->inp_flowid;
 379                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 380                 }
 381 #ifdef NUMA
 382                 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
 383 #endif
 384         }
 385
 386 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 387         /*
 388          * IPSec checking which handles several cases.
 389          * FAST IPSEC: We re-injected the packet.
 390          * XXX: need scope argument.
 391          */
 392         if (IPSEC_ENABLED(ipv6)) {
 393                 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
 394                         if (error == EINPROGRESS)
 395                                 error = 0;
 396                         goto done;
 397                 }
 398         }
 399 #endif /* IPSEC */
 400
 401         bzero(&exthdrs, sizeof(exthdrs));
 402         if (opt) {
 403                 /* Hop-by-Hop options header */
 404                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 405                 /* Destination options header(1st part) */
 406                 if (opt->ip6po_rthdr) {
 407                         /*
 408                          * Destination options header(1st part)
 409                          * This only makes sense with a routing header.
 410                          * See Section 9.2 of RFC 3542.
 411                          * Disabling this part just for MIP6 convenience is
 412                          * a bad idea.  We need to think carefully about a
 413                          * way to make the advanced API coexist with MIP6
 414                          * options, which might automatically be inserted in
 415                          * the kernel.
 416                          */
 417                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 418                 }
 419                 /* Routing header */
 420                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 421                 /* Destination options header(2nd part) */
 422                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 423         }
 424
 425         /*
 426          * Calculate the total length of the extension header chain.
 427          * Keep the length of the unfragmentable part for fragmentation.
 428          */
 429         optlen = 0;
 430         if (exthdrs.ip6e_hbh)
 431                 optlen += exthdrs.ip6e_hbh->m_len;
 432         if (exthdrs.ip6e_dest1)
 433                 optlen += exthdrs.ip6e_dest1->m_len;
 434         if (exthdrs.ip6e_rthdr)
 435                 optlen += exthdrs.ip6e_rthdr->m_len;
 436         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 437
 438         /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
 439         if (exthdrs.ip6e_dest2)
 440                 optlen += exthdrs.ip6e_dest2->m_len;
 441
 442         /*
 443          * If there is at least one extension header,
 444          * separate IP6 header from the payload.
 445          */
 446         if (optlen && !hdrsplit) {
 447                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 448                         m = NULL;
 449                         goto freehdrs;
 450                 }
 451                 m = exthdrs.ip6e_ip6;
 452                 hdrsplit++;
 453         }
 454
 455         ip6 = mtod(m, struct ip6_hdr *);
 456
 457         /* adjust mbuf packet header length */
 458         m->m_pkthdr.len += optlen;
 459         plen = m->m_pkthdr.len - sizeof(*ip6);
 460
 461         /* If this is a jumbo payload, insert a jumbo payload option. */
 462         if (plen > IPV6_MAXPACKET) {
 463                 if (!hdrsplit) {
 464                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 465                                 m = NULL;
 466                                 goto freehdrs;
 467                         }
 468                         m = exthdrs.ip6e_ip6;
 469                         hdrsplit++;
 470                 }
 471                 /* adjust pointer */
 472                 ip6 = mtod(m, struct ip6_hdr *);
 473                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 474                         goto freehdrs;
 475                 ip6->ip6_plen = 0;
 476         } else
 477                 ip6->ip6_plen = htons(plen);
 478
 479         /*
 480          * Concatenate headers and fill in next header fields.
 481          * Here we have, on "m"
 482          *      IPv6 payload
 483          * and we insert headers accordingly.  Finally, we should be getting:
 484          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 485          *
 486          * during the header composing process, "m" points to IPv6 header.
 487          * "mprev" points to an extension header prior to esp.
 488          */
 489         u_char *nexthdrp = &ip6->ip6_nxt;
 490         mprev = m;
 491
 492         /*
 493          * we treat dest2 specially.  this makes IPsec processing
 494          * much easier.  the goal here is to make mprev point the
 495          * mbuf prior to dest2.
 496          *
 497          * result: IPv6 dest2 payload
 498          * m and mprev will point to IPv6 header.
 499          */
 500         if (exthdrs.ip6e_dest2) {
 501                 if (!hdrsplit)
 502                         panic("assumption failed: hdr not split");
 503                 exthdrs.ip6e_dest2->m_next = m->m_next;
 504                 m->m_next = exthdrs.ip6e_dest2;
 505                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 506                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 507         }
 508
 509         /*
 510          * result: IPv6 hbh dest1 rthdr dest2 payload
 511          * m will point to IPv6 header.  mprev will point to the
 512          * extension header prior to dest2 (rthdr in the above case).
 513          */
 514         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 515         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 516                    IPPROTO_DSTOPTS);
 517         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 518                    IPPROTO_ROUTING);
 519
 520         /*
 521          * If there is a routing header, discard the packet.
 522          */
 523         if (exthdrs.ip6e_rthdr) {
 524                  error = EINVAL;
 525                  goto bad;
 526         }
 527
 528         /* Source address validation */
 529         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 530             (flags & IPV6_UNSPECSRC) == 0) {
 531                 error = EOPNOTSUPP;
 532                 IP6STAT_INC(ip6s_badscope);
 533                 goto bad;
 534         }
 535         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 536                 error = EOPNOTSUPP;
 537                 IP6STAT_INC(ip6s_badscope);
 538                 goto bad;
 539         }
 540
 541         IP6STAT_INC(ip6s_localout);
 542
 543         /*
 544          * Route packet.
 545          */
 546         if (ro == NULL) {
 547                 ro = &ip6route;
 548                 bzero((caddr_t)ro, sizeof(*ro));
 549         }
 550         ro_pmtu = ro;
 551         if (opt && opt->ip6po_rthdr)
 552                 ro = &opt->ip6po_route;
 553         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 554         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 555 again:
 556         /*
 557          * if specified, try to fill in the traffic class field.
 558          * do not override if a non-zero value is already set.
 559          * we check the diffserv field and the ecn field separately.
 560          */
 561         if (opt && opt->ip6po_tclass >= 0) {
 562                 int mask = 0;
 563
 564                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 565                         mask |= 0xfc;
 566                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 567                         mask |= 0x03;
 568                 if (mask != 0)
 569                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 570         }
 571
 572         /* fill in or override the hop limit field, if necessary. */
 573         if (opt && opt->ip6po_hlim != -1)
 574                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 575         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 576                 if (im6o != NULL)
 577                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 578                 else
 579                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 580         }
 581         /*
 582          * Validate route against routing table additions;
 583          * a better/more specific route might have been added.
 584          * Make sure address family is set in route.
 585          */
 586         if (inp) {
 587                 ro->ro_dst.sin6_family = AF_INET6;
 588                 RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
 589         }
 590         if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
 591             ro->ro_dst.sin6_family == AF_INET6 &&
 592             IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
 593                 rt = ro->ro_rt;
 594                 ifp = ro->ro_rt->rt_ifp;
 595         } else {
 596                 if (ro->ro_lle)
 597                         LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 598                 ro->ro_lle = NULL;
 599                 if (fwd_tag == NULL) {
 600                         bzero(&dst_sa, sizeof(dst_sa));
 601                         dst_sa.sin6_family = AF_INET6;
 602                         dst_sa.sin6_len = sizeof(dst_sa);
 603                         dst_sa.sin6_addr = ip6->ip6_dst;
 604                 }
 605                 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
 606                     &rt, fibnum);
 607                 if (error != 0) {
 608                         if (ifp != NULL)
 609                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 610                         goto bad;
 611                 }
 612         }
 613         if (rt == NULL) {
 614                 /*
 615                  * If in6_selectroute() does not return a route entry,
 616                  * dst may not have been updated.
 617                  */
 618                 *dst = dst_sa;  /* XXX */
 619         }
 620
 621         /*
 622          * then rt (for unicast) and ifp must be non-NULL valid values.
 623          */
 624         if ((flags & IPV6_FORWARDING) == 0) {
 625                 /* XXX: the FORWARDING flag can be set for mrouting. */
 626                 in6_ifstat_inc(ifp, ifs6_out_request);
 627         }
 628         if (rt != NULL) {
 629                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 630                 counter_u64_add(rt->rt_pksent, 1);
 631         }
 632
 633         /* Setup data structures for scope ID checks. */
 634         src0 = ip6->ip6_src;
 635         bzero(&src_sa, sizeof(src_sa));
 636         src_sa.sin6_family = AF_INET6;
 637         src_sa.sin6_len = sizeof(src_sa);
 638         src_sa.sin6_addr = ip6->ip6_src;
 639
 640         dst0 = ip6->ip6_dst;
 641         /* re-initialize to be sure */
 642         bzero(&dst_sa, sizeof(dst_sa));
 643         dst_sa.sin6_family = AF_INET6;
 644         dst_sa.sin6_len = sizeof(dst_sa);
 645         dst_sa.sin6_addr = ip6->ip6_dst;
 646
 647         /* Check for valid scope ID. */
 648         if (in6_setscope(&src0, ifp, &zone) == 0 &&
 649             sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
 650             in6_setscope(&dst0, ifp, &zone) == 0 &&
 651             sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
 652                 /*
 653                  * The outgoing interface is in the zone of the source
 654                  * and destination addresses.
 655                  *
 656                  * Because the loopback interface cannot receive
 657                  * packets with a different scope ID than its own,
 658                  * there is a trick is to pretend the outgoing packet
 659                  * was received by the real network interface, by
 660                  * setting "origifp" different from "ifp". This is
 661                  * only allowed when "ifp" is a loopback network
 662                  * interface. Refer to code in nd6_output_ifp() for
 663                  * more details.
 664                  */
 665                 origifp = ifp;
 666
 667                 /*
 668                  * We should use ia_ifp to support the case of sending
 669                  * packets to an address of our own.
 670                  */
 671                 if (ia != NULL && ia->ia_ifp)
 672                         ifp = ia->ia_ifp;
 673
 674         } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
 675             sa6_recoverscope(&src_sa) != 0 ||
 676             sa6_recoverscope(&dst_sa) != 0 ||
 677             dst_sa.sin6_scope_id == 0 ||
 678             (src_sa.sin6_scope_id != 0 &&
 679             src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
 680             (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
 681                 /*
 682                  * If the destination network interface is not a
 683                  * loopback interface, or the destination network
 684                  * address has no scope ID, or the source address has
 685                  * a scope ID set which is different from the
 686                  * destination address one, or there is no network
 687                  * interface representing this scope ID, the address
 688                  * pair is considered invalid.
 689                  */
 690                 IP6STAT_INC(ip6s_badscope);
 691                 in6_ifstat_inc(ifp, ifs6_out_discard);
 692                 if (error == 0)
 693                         error = EHOSTUNREACH; /* XXX */
 694                 goto bad;
 695         }
 696
 697         /* All scope ID checks are successful. */
 698
 699         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 700                 if (opt && opt->ip6po_nextroute.ro_rt) {
 701                         /*
 702                          * The nexthop is explicitly specified by the
 703                          * application.  We assume the next hop is an IPv6
 704                          * address.
 705                          */
 706                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 707                 }
 708                 else if ((rt->rt_flags & RTF_GATEWAY))
 709                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 710         }
 711
 712         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 713                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 714         } else {
 715                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 716                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 717                 /*
 718                  * Confirm that the outgoing interface supports multicast.
 719                  */
 720                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 721                         IP6STAT_INC(ip6s_noroute);
 722                         in6_ifstat_inc(ifp, ifs6_out_discard);
 723                         error = ENETUNREACH;
 724                         goto bad;
 725                 }
 726                 if ((im6o == NULL && in6_mcast_loop) ||
 727                     (im6o && im6o->im6o_multicast_loop)) {
 728                         /*
 729                          * Loop back multicast datagram if not expressly
 730                          * forbidden to do so, even if we have not joined
 731                          * the address; protocols will filter it later,
 732                          * thus deferring a hash lookup and lock acquisition
 733                          * at the expense of an m_copym().
 734                          */
 735                         ip6_mloopback(ifp, m);
 736                 } else {
 737                         /*
 738                          * If we are acting as a multicast router, perform
 739                          * multicast forwarding as if the packet had just
 740                          * arrived on the interface to which we are about
 741                          * to send.  The multicast forwarding function
 742                          * recursively calls this function, using the
 743                          * IPV6_FORWARDING flag to prevent infinite recursion.
 744                          *
 745                          * Multicasts that are looped back by ip6_mloopback(),
 746                          * above, will be forwarded by the ip6_input() routine,
 747                          * if necessary.
 748                          */
 749                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 750                                 /*
 751                                  * XXX: ip6_mforward expects that rcvif is NULL
 752                                  * when it is called from the originating path.
 753                                  * However, it may not always be the case.
 754                                  */
 755                                 m->m_pkthdr.rcvif = NULL;
 756                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 757                                         m_freem(m);
 758                                         goto done;
 759                                 }
 760                         }
 761                 }
 762                 /*
 763                  * Multicasts with a hoplimit of zero may be looped back,
 764                  * above, but must not be transmitted on a network.
 765                  * Also, multicasts addressed to the loopback interface
 766                  * are not sent -- the above call to ip6_mloopback() will
 767                  * loop back a copy if this host actually belongs to the
 768                  * destination group on the loopback interface.
 769                  */
 770                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 771                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 772                         m_freem(m);
 773                         goto done;
 774                 }
 775         }
 776
 777         /*
 778          * Fill the outgoing inteface to tell the upper layer
 779          * to increment per-interface statistics.
 780          */
 781         if (ifpp)
 782                 *ifpp = ifp;
 783
 784         /* Determine path MTU. */
 785         if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
 786                     &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
 787                 goto bad;
 788
 789         /*
 790          * The caller of this function may specify to use the minimum MTU
 791          * in some cases.
 792          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 793          * setting.  The logic is a bit complicated; by default, unicast
 794          * packets will follow path MTU while multicast packets will be sent at
 795          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 796          * including unicast ones will be sent at the minimum MTU.  Multicast
 797          * packets will always be sent at the minimum MTU unless
 798          * IP6PO_MINMTU_DISABLE is explicitly specified.
 799          * See RFC 3542 for more details.
 800          */
 801         if (mtu > IPV6_MMTU) {
 802                 if ((flags & IPV6_MINMTU))
 803                         mtu = IPV6_MMTU;
 804                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 805                         mtu = IPV6_MMTU;
 806                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 807                          (opt == NULL ||
 808                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 809                         mtu = IPV6_MMTU;
 810                 }
 811         }
 812
 813         /*
 814          * clear embedded scope identifiers if necessary.
 815          * in6_clearscope will touch the addresses only when necessary.
 816          */
 817         in6_clearscope(&ip6->ip6_src);
 818         in6_clearscope(&ip6->ip6_dst);
 819
 820         /*
 821          * If the outgoing packet contains a hop-by-hop options header,
 822          * it must be examined and processed even by the source node.
 823          * (RFC 2460, section 4.)
 824          */
 825         if (exthdrs.ip6e_hbh) {
 826                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 827                 u_int32_t dummy; /* XXX unused */
 828                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 829
 830 #ifdef DIAGNOSTIC
 831                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 832                         panic("ip6e_hbh is not contiguous");
 833 #endif
 834                 /*
 835                  *  XXX: if we have to send an ICMPv6 error to the sender,
 836                  *       we need the M_LOOP flag since icmp6_error() expects
 837                  *       the IPv6 and the hop-by-hop options header are
 838                  *       contiguous unless the flag is set.
 839                  */
 840                 m->m_flags |= M_LOOP;
 841                 m->m_pkthdr.rcvif = ifp;
 842                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 843                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 844                     &dummy, &plen) < 0) {
 845                         /* m was already freed at this point */
 846                         error = EINVAL;/* better error? */
 847                         goto done;
 848                 }
 849                 m->m_flags &= ~M_LOOP; /* XXX */
 850                 m->m_pkthdr.rcvif = NULL;
 851         }
 852
 853         /* Jump over all PFIL processing if hooks are not active. */
 854         if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
 855                 goto passout;
 856
 857         odst = ip6->ip6_dst;
 858         /* Run through list of hooks for output packets. */
 859         switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
 860         case PFIL_PASS:
 861                 ip6 = mtod(m, struct ip6_hdr *);
 862                 break;
 863         case PFIL_DROPPED:
 864                 error = EPERM;
 865                 /* FALLTHROUGH */
 866         case PFIL_CONSUMED:
 867                 goto done;
 868         }
 869
 870         needfiblookup = 0;
 871         /* See if destination IP address was changed by packet filter. */
 872         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 873                 m->m_flags |= M_SKIP_FIREWALL;
 874                 /* If destination is now ourself drop to ip6_input(). */
 875                 if (in6_localip(&ip6->ip6_dst)) {
 876                         m->m_flags |= M_FASTFWD_OURS;
 877                         if (m->m_pkthdr.rcvif == NULL)
 878                                 m->m_pkthdr.rcvif = V_loif;
 879                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 880                                 m->m_pkthdr.csum_flags |=
 881                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 882                                 m->m_pkthdr.csum_data = 0xffff;
 883                         }
 884 #ifdef SCTP
 885                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 886                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 887 #endif
 888                         error = netisr_queue(NETISR_IPV6, m);
 889                         goto done;
 890                 } else {
 891                         RO_INVALIDATE_CACHE(ro);
 892                         needfiblookup = 1; /* Redo the routing table lookup. */
 893                 }
 894         }
 895         /* See if fib was changed by packet filter. */
 896         if (fibnum != M_GETFIB(m)) {
 897                 m->m_flags |= M_SKIP_FIREWALL;
 898                 fibnum = M_GETFIB(m);
 899                 RO_INVALIDATE_CACHE(ro);
 900                 needfiblookup = 1;
 901         }
 902         if (needfiblookup)
 903                 goto again;
 904
 905         /* See if local, if yes, send it to netisr. */
 906         if (m->m_flags & M_FASTFWD_OURS) {
 907                 if (m->m_pkthdr.rcvif == NULL)
 908                         m->m_pkthdr.rcvif = V_loif;
 909                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 910                         m->m_pkthdr.csum_flags |=
 911                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 912                         m->m_pkthdr.csum_data = 0xffff;
 913                 }
 914 #ifdef SCTP
 915                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 916                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 917 #endif
 918                 error = netisr_queue(NETISR_IPV6, m);
 919                 goto done;
 920         }
 921         /* Or forward to some other address? */
 922         if ((m->m_flags & M_IP6_NEXTHOP) &&
 923             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 924                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 925                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
 926                 m->m_flags |= M_SKIP_FIREWALL;
 927                 m->m_flags &= ~M_IP6_NEXTHOP;
 928                 m_tag_delete(m, fwd_tag);
 929                 goto again;
 930         }
 931
 932 passout:
 933         /*
 934          * Send the packet to the outgoing interface.
 935          * If necessary, do IPv6 fragmentation before sending.
 936          *
 937          * the logic here is rather complex:
 938          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 939          * 1-a: send as is if tlen <= path mtu
 940          * 1-b: fragment if tlen > path mtu
 941          *
 942          * 2: if user asks us not to fragment (dontfrag == 1)
 943          * 2-a: send as is if tlen <= interface mtu
 944          * 2-b: error if tlen > interface mtu
 945          *
 946          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 947          *      always fragment
 948          *
 949          * 4: if dontfrag == 1 && alwaysfrag == 1
 950          *      error, as we cannot handle this conflicting request
 951          */
 952         sw_csum = m->m_pkthdr.csum_flags;
 953         if (!hdrsplit) {
 954                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 955                 sw_csum &= ~ifp->if_hwassist;
 956         } else
 957                 tso = 0;
 958         /*
 959          * If we added extension headers, we will not do TSO and calculate the
 960          * checksums ourselves for now.
 961          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 962          * with ext. hdrs.
 963          */
 964         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 965                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 966                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 967         }
 968 #ifdef SCTP
 969         if (sw_csum & CSUM_SCTP_IPV6) {
 970                 sw_csum &= ~CSUM_SCTP_IPV6;
 971                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 972         }
 973 #endif
 974         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 975         tlen = m->m_pkthdr.len;
 976
 977         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 978                 dontfrag = 1;
 979         else
 980                 dontfrag = 0;
 981         if (dontfrag && alwaysfrag) {   /* case 4 */
 982                 /* conflicting request - can't transmit */
 983                 error = EMSGSIZE;
 984                 goto bad;
 985         }
 986         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 987                 /*
 988                  * Even if the DONTFRAG option is specified, we cannot send the
 989                  * packet when the data length is larger than the MTU of the
 990                  * outgoing interface.
 991                  * Notify the error by sending IPV6_PATHMTU ancillary data if
 992                  * application wanted to know the MTU value. Also return an
 993                  * error code (this is not described in the API spec).
 994                  */
 995                 if (inp != NULL)
 996                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
 997                 error = EMSGSIZE;
 998                 goto bad;
 999         }
1000
1001         /*
1002          * transmit packet without fragmentation
1003          */
1004         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1005                 struct in6_ifaddr *ia6;
1006
1007                 ip6 = mtod(m, struct ip6_hdr *);
1008                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1009                 if (ia6) {
1010                         /* Record statistics for this interface address. */
1011                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1012                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
1013                             m->m_pkthdr.len);
1014                         ifa_free(&ia6->ia_ifa);
1015                 }
1016                 error = ip6_output_send(inp, ifp, origifp, m, dst, ro);
1017                 goto done;
1018         }
1019
1020         /*
1021          * try to fragment the packet.  case 1-b and 3
1022          */
1023         if (mtu < IPV6_MMTU) {
1024                 /* path MTU cannot be less than IPV6_MMTU */
1025                 error = EMSGSIZE;
1026                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1027                 goto bad;
1028         } else if (ip6->ip6_plen == 0) {
1029                 /* jumbo payload cannot be fragmented */
1030                 error = EMSGSIZE;
1031                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1032                 goto bad;
1033         } else {
1034                 u_char nextproto;
1035
1036                 /*
1037                  * Too large for the destination or interface;
1038                  * fragment if possible.
1039                  * Must be able to put at least 8 bytes per fragment.
1040                  */
1041                 hlen = unfragpartlen;
1042                 if (mtu > IPV6_MAXPACKET)
1043                         mtu = IPV6_MAXPACKET;
1044
1045                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1046                 if (len < 8) {
1047                         error = EMSGSIZE;
1048                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1049                         goto bad;
1050                 }
1051
1052                 /*
1053                  * If the interface will not calculate checksums on
1054                  * fragmented packets, then do it here.
1055                  * XXX-BZ handle the hw offloading case.  Need flags.
1056                  */
1057                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1058                         in6_delayed_cksum(m, plen, hlen);
1059                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1060                 }
1061 #ifdef SCTP
1062                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1063                         sctp_delayed_cksum(m, hlen);
1064                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1065                 }
1066 #endif
1067                 /*
1068                  * Change the next header field of the last header in the
1069                  * unfragmentable part.
1070                  */
1071                 if (exthdrs.ip6e_rthdr) {
1072                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1073                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1074                 } else if (exthdrs.ip6e_dest1) {
1075                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1076                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1077                 } else if (exthdrs.ip6e_hbh) {
1078                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1079                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1080                 } else {
1081                         nextproto = ip6->ip6_nxt;
1082                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1083                 }
1084
1085                 /*
1086                  * Loop through length of segment after first fragment,
1087                  * make new header and copy data of each part and link onto
1088                  * chain.
1089                  */
1090                 m0 = m;
1091                 id = htonl(ip6_randomid());
1092                 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1093                         goto sendorfree;
1094
1095                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1096         }
1097
1098         /*
1099          * Remove leading garbages.
1100          */
1101 sendorfree:
1102         m = m0->m_nextpkt;
1103         m0->m_nextpkt = 0;
1104         m_freem(m0);
1105         for (; m; m = m0) {
1106                 m0 = m->m_nextpkt;
1107                 m->m_nextpkt = 0;
1108                 if (error == 0) {
1109                         /* Record statistics for this interface address. */
1110                         if (ia) {
1111                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1112                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1113                                     m->m_pkthdr.len);
1114                         }
1115                         error = ip6_output_send(inp, ifp, origifp, m, dst, ro);
1116                 } else
1117                         m_freem(m);
1118         }
1119
1120         if (error == 0)
1121                 IP6STAT_INC(ip6s_fragmented);
1122
1123 done:
1124         if (ro == &ip6route)
1125                 RO_RTFREE(ro);
1126         return (error);
1127
1128 freehdrs:
1129         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1130         m_freem(exthdrs.ip6e_dest1);
1131         m_freem(exthdrs.ip6e_rthdr);
1132         m_freem(exthdrs.ip6e_dest2);
1133         /* FALLTHROUGH */
1134 bad:
1135         if (m)
1136                 m_freem(m);
1137         goto done;
1138 }
1139
1140 static int
1141 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1142 {
1143         struct mbuf *m;
1144
1145         if (hlen > MCLBYTES)
1146                 return (ENOBUFS); /* XXX */
1147
1148         if (hlen > MLEN)
1149                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1150         else
1151                 m = m_get(M_NOWAIT, MT_DATA);
1152         if (m == NULL)
1153                 return (ENOBUFS);
1154         m->m_len = hlen;
1155         if (hdr)
1156                 bcopy(hdr, mtod(m, caddr_t), hlen);
1157
1158         *mp = m;
1159         return (0);
1160 }
1161
1162 /*
1163  * Insert jumbo payload option.
1164  */
1165 static int
1166 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1167 {
1168         struct mbuf *mopt;
1169         u_char *optbuf;
1170         u_int32_t v;
1171
1172 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1173
1174         /*
1175          * If there is no hop-by-hop options header, allocate new one.
1176          * If there is one but it doesn't have enough space to store the
1177          * jumbo payload option, allocate a cluster to store the whole options.
1178          * Otherwise, use it to store the options.
1179          */
1180         if (exthdrs->ip6e_hbh == NULL) {
1181                 mopt = m_get(M_NOWAIT, MT_DATA);
1182                 if (mopt == NULL)
1183                         return (ENOBUFS);
1184                 mopt->m_len = JUMBOOPTLEN;
1185                 optbuf = mtod(mopt, u_char *);
1186                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1187                 exthdrs->ip6e_hbh = mopt;
1188         } else {
1189                 struct ip6_hbh *hbh;
1190
1191                 mopt = exthdrs->ip6e_hbh;
1192                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1193                         /*
1194                          * XXX assumption:
1195                          * - exthdrs->ip6e_hbh is not referenced from places
1196                          *   other than exthdrs.
1197                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1198                          */
1199                         int oldoptlen = mopt->m_len;
1200                         struct mbuf *n;
1201
1202                         /*
1203                          * XXX: give up if the whole (new) hbh header does
1204                          * not fit even in an mbuf cluster.
1205                          */
1206                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1207                                 return (ENOBUFS);
1208
1209                         /*
1210                          * As a consequence, we must always prepare a cluster
1211                          * at this point.
1212                          */
1213                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1214                         if (n == NULL)
1215                                 return (ENOBUFS);
1216                         n->m_len = oldoptlen + JUMBOOPTLEN;
1217                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1218                             oldoptlen);
1219                         optbuf = mtod(n, caddr_t) + oldoptlen;
1220                         m_freem(mopt);
1221                         mopt = exthdrs->ip6e_hbh = n;
1222                 } else {
1223                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1224                         mopt->m_len += JUMBOOPTLEN;
1225                 }
1226                 optbuf[0] = IP6OPT_PADN;
1227                 optbuf[1] = 1;
1228
1229                 /*
1230                  * Adjust the header length according to the pad and
1231                  * the jumbo payload option.
1232                  */
1233                 hbh = mtod(mopt, struct ip6_hbh *);
1234                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1235         }
1236
1237         /* fill in the option. */
1238         optbuf[2] = IP6OPT_JUMBO;
1239         optbuf[3] = 4;
1240         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1241         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1242
1243         /* finally, adjust the packet header length */
1244         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1245
1246         return (0);
1247 #undef JUMBOOPTLEN
1248 }
1249
1250 /*
1251  * Insert fragment header and copy unfragmentable header portions.
1252  */
1253 static int
1254 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1255     struct ip6_frag **frghdrp)
1256 {
1257         struct mbuf *n, *mlast;
1258
1259         if (hlen > sizeof(struct ip6_hdr)) {
1260                 n = m_copym(m0, sizeof(struct ip6_hdr),
1261                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1262                 if (n == NULL)
1263                         return (ENOBUFS);
1264                 m->m_next = n;
1265         } else
1266                 n = m;
1267
1268         /* Search for the last mbuf of unfragmentable part. */
1269         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1270                 ;
1271
1272         if (M_WRITABLE(mlast) &&
1273             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1274                 /* use the trailing space of the last mbuf for the fragment hdr */
1275                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1276                     mlast->m_len);
1277                 mlast->m_len += sizeof(struct ip6_frag);
1278                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1279         } else {
1280                 /* allocate a new mbuf for the fragment header */
1281                 struct mbuf *mfrg;
1282
1283                 mfrg = m_get(M_NOWAIT, MT_DATA);
1284                 if (mfrg == NULL)
1285                         return (ENOBUFS);
1286                 mfrg->m_len = sizeof(struct ip6_frag);
1287                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1288                 mlast->m_next = mfrg;
1289         }
1290
1291         return (0);
1292 }
1293
1294 /*
1295  * Calculates IPv6 path mtu for destination @dst.
1296  * Resulting MTU is stored in @mtup.
1297  *
1298  * Returns 0 on success.
1299  */
1300 static int
1301 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1302 {
1303         struct nhop6_extended nh6;
1304         struct in6_addr kdst;
1305         uint32_t scopeid;
1306         struct ifnet *ifp;
1307         u_long mtu;
1308         int error;
1309
1310         in6_splitscope(dst, &kdst, &scopeid);
1311         if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1312                 return (EHOSTUNREACH);
1313
1314         ifp = nh6.nh_ifp;
1315         mtu = nh6.nh_mtu;
1316
1317         error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1318         fib6_free_nh_ext(fibnum, &nh6);
1319
1320         return (error);
1321 }
1322
1323 /*
1324  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1325  * and cached data in @ro_pmtu.
1326  * MTU from (successful) route lookup is saved (along with dst)
1327  * inside @ro_pmtu to avoid subsequent route lookups after packet
1328  * filter processing.
1329  *
1330  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1331  * Returns 0 on success.
1332  */
1333 static int
1334 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1335     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1336     int *alwaysfragp, u_int fibnum, u_int proto)
1337 {
1338         struct nhop6_basic nh6;
1339         struct in6_addr kdst;
1340         uint32_t scopeid;
1341         struct sockaddr_in6 *sa6_dst;
1342         u_long mtu;
1343
1344         mtu = 0;
1345         if (do_lookup) {
1346
1347                 /*
1348                  * Here ro_pmtu has final destination address, while
1349                  * ro might represent immediate destination.
1350                  * Use ro_pmtu destination since mtu might differ.
1351                  */
1352                 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1353                 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1354                         ro_pmtu->ro_mtu = 0;
1355
1356                 if (ro_pmtu->ro_mtu == 0) {
1357                         bzero(sa6_dst, sizeof(*sa6_dst));
1358                         sa6_dst->sin6_family = AF_INET6;
1359                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1360                         sa6_dst->sin6_addr = *dst;
1361
1362                         in6_splitscope(dst, &kdst, &scopeid);
1363                         if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1364                             &nh6) == 0)
1365                                 ro_pmtu->ro_mtu = nh6.nh_mtu;
1366                 }
1367
1368                 mtu = ro_pmtu->ro_mtu;
1369         }
1370
1371         if (ro_pmtu->ro_rt)
1372                 mtu = ro_pmtu->ro_rt->rt_mtu;
1373
1374         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1375 }
1376
1377 /*
1378  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1379  * hostcache data for @dst.
1380  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1381  *
1382  * Returns 0 on success.
1383  */
1384 static int
1385 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1386     u_long *mtup, int *alwaysfragp, u_int proto)
1387 {
1388         u_long mtu = 0;
1389         int alwaysfrag = 0;
1390         int error = 0;
1391
1392         if (rt_mtu > 0) {
1393                 u_int32_t ifmtu;
1394                 struct in_conninfo inc;
1395
1396                 bzero(&inc, sizeof(inc));
1397                 inc.inc_flags |= INC_ISIPV6;
1398                 inc.inc6_faddr = *dst;
1399
1400                 ifmtu = IN6_LINKMTU(ifp);
1401
1402                 /* TCP is known to react to pmtu changes so skip hc */
1403                 if (proto != IPPROTO_TCP)
1404                         mtu = tcp_hc_getmtu(&inc);
1405
1406                 if (mtu)
1407                         mtu = min(mtu, rt_mtu);
1408                 else
1409                         mtu = rt_mtu;
1410                 if (mtu == 0)
1411                         mtu = ifmtu;
1412                 else if (mtu < IPV6_MMTU) {
1413                         /*
1414                          * RFC2460 section 5, last paragraph:
1415                          * if we record ICMPv6 too big message with
1416                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1417                          * or smaller, with framgent header attached.
1418                          * (fragment header is needed regardless from the
1419                          * packet size, for translators to identify packets)
1420                          */
1421                         alwaysfrag = 1;
1422                         mtu = IPV6_MMTU;
1423                 }
1424         } else if (ifp) {
1425                 mtu = IN6_LINKMTU(ifp);
1426         } else
1427                 error = EHOSTUNREACH; /* XXX */
1428
1429         *mtup = mtu;
1430         if (alwaysfragp)
1431                 *alwaysfragp = alwaysfrag;
1432         return (error);
1433 }
1434
1435 /*
1436  * IP6 socket option processing.
1437  */
1438 int
1439 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1440 {
1441         int optdatalen, uproto;
1442         void *optdata;
1443         struct inpcb *in6p = sotoinpcb(so);
1444         int error, optval;
1445         int level, op, optname;
1446         int optlen;
1447         struct thread *td;
1448 #ifdef  RSS
1449         uint32_t rss_bucket;
1450         int retval;
1451 #endif
1452
1453 /*
1454  * Don't use more than a quarter of mbuf clusters.  N.B.:
1455  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1456  * on LP64 architectures, so cast to u_long to avoid undefined
1457  * behavior.  ILP32 architectures cannot have nmbclusters
1458  * large enough to overflow for other reasons.
1459  */
1460 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1461
1462         level = sopt->sopt_level;
1463         op = sopt->sopt_dir;
1464         optname = sopt->sopt_name;
1465         optlen = sopt->sopt_valsize;
1466         td = sopt->sopt_td;
1467         error = 0;
1468         optval = 0;
1469         uproto = (int)so->so_proto->pr_protocol;
1470
1471         if (level != IPPROTO_IPV6) {
1472                 error = EINVAL;
1473
1474                 if (sopt->sopt_level == SOL_SOCKET &&
1475                     sopt->sopt_dir == SOPT_SET) {
1476                         switch (sopt->sopt_name) {
1477                         case SO_REUSEADDR:
1478                                 INP_WLOCK(in6p);
1479                                 if ((so->so_options & SO_REUSEADDR) != 0)
1480                                         in6p->inp_flags2 |= INP_REUSEADDR;
1481                                 else
1482                                         in6p->inp_flags2 &= ~INP_REUSEADDR;
1483                                 INP_WUNLOCK(in6p);
1484                                 error = 0;
1485                                 break;
1486                         case SO_REUSEPORT:
1487                                 INP_WLOCK(in6p);
1488                                 if ((so->so_options & SO_REUSEPORT) != 0)
1489                                         in6p->inp_flags2 |= INP_REUSEPORT;
1490                                 else
1491                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1492                                 INP_WUNLOCK(in6p);
1493                                 error = 0;
1494                                 break;
1495                         case SO_REUSEPORT_LB:
1496                                 INP_WLOCK(in6p);
1497                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1498                                         in6p->inp_flags2 |= INP_REUSEPORT_LB;
1499                                 else
1500                                         in6p->inp_flags2 &= ~INP_REUSEPORT_LB;
1501                                 INP_WUNLOCK(in6p);
1502                                 error = 0;
1503                                 break;
1504                         case SO_SETFIB:
1505                                 INP_WLOCK(in6p);
1506                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1507                                 INP_WUNLOCK(in6p);
1508                                 error = 0;
1509                                 break;
1510                         case SO_MAX_PACING_RATE:
1511 #ifdef RATELIMIT
1512                                 INP_WLOCK(in6p);
1513                                 in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1514                                 INP_WUNLOCK(in6p);
1515                                 error = 0;
1516 #else
1517                                 error = EOPNOTSUPP;
1518 #endif
1519                                 break;
1520                         default:
1521                                 break;
1522                         }
1523                 }
1524         } else {                /* level == IPPROTO_IPV6 */
1525                 switch (op) {
1526
1527                 case SOPT_SET:
1528                         switch (optname) {
1529                         case IPV6_2292PKTOPTIONS:
1530 #ifdef IPV6_PKTOPTIONS
1531                         case IPV6_PKTOPTIONS:
1532 #endif
1533                         {
1534                                 struct mbuf *m;
1535
1536                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1537                                         printf("ip6_ctloutput: mbuf limit hit\n");
1538                                         error = ENOBUFS;
1539                                         break;
1540                                 }
1541
1542                                 error = soopt_getm(sopt, &m); /* XXX */
1543                                 if (error != 0)
1544                                         break;
1545                                 error = soopt_mcopyin(sopt, m); /* XXX */
1546                                 if (error != 0)
1547                                         break;
1548                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1549                                                     m, so, sopt);
1550                                 m_freem(m); /* XXX */
1551                                 break;
1552                         }
1553
1554                         /*
1555                          * Use of some Hop-by-Hop options or some
1556                          * Destination options, might require special
1557                          * privilege.  That is, normal applications
1558                          * (without special privilege) might be forbidden
1559                          * from setting certain options in outgoing packets,
1560                          * and might never see certain options in received
1561                          * packets. [RFC 2292 Section 6]
1562                          * KAME specific note:
1563                          *  KAME prevents non-privileged users from sending or
1564                          *  receiving ANY hbh/dst options in order to avoid
1565                          *  overhead of parsing options in the kernel.
1566                          */
1567                         case IPV6_RECVHOPOPTS:
1568                         case IPV6_RECVDSTOPTS:
1569                         case IPV6_RECVRTHDRDSTOPTS:
1570                                 if (td != NULL) {
1571                                         error = priv_check(td,
1572                                             PRIV_NETINET_SETHDROPTS);
1573                                         if (error)
1574                                                 break;
1575                                 }
1576                                 /* FALLTHROUGH */
1577                         case IPV6_UNICAST_HOPS:
1578                         case IPV6_HOPLIMIT:
1579
1580                         case IPV6_RECVPKTINFO:
1581                         case IPV6_RECVHOPLIMIT:
1582                         case IPV6_RECVRTHDR:
1583                         case IPV6_RECVPATHMTU:
1584                         case IPV6_RECVTCLASS:
1585                         case IPV6_RECVFLOWID:
1586 #ifdef  RSS
1587                         case IPV6_RECVRSSBUCKETID:
1588 #endif
1589                         case IPV6_V6ONLY:
1590                         case IPV6_AUTOFLOWLABEL:
1591                         case IPV6_ORIGDSTADDR:
1592                         case IPV6_BINDANY:
1593                         case IPV6_BINDMULTI:
1594 #ifdef  RSS
1595                         case IPV6_RSS_LISTEN_BUCKET:
1596 #endif
1597                                 if (optname == IPV6_BINDANY && td != NULL) {
1598                                         error = priv_check(td,
1599                                             PRIV_NETINET_BINDANY);
1600                                         if (error)
1601                                                 break;
1602                                 }
1603
1604                                 if (optlen != sizeof(int)) {
1605                                         error = EINVAL;
1606                                         break;
1607                                 }
1608                                 error = sooptcopyin(sopt, &optval,
1609                                         sizeof optval, sizeof optval);
1610                                 if (error)
1611                                         break;
1612                                 switch (optname) {
1613
1614                                 case IPV6_UNICAST_HOPS:
1615                                         if (optval < -1 || optval >= 256)
1616                                                 error = EINVAL;
1617                                         else {
1618                                                 /* -1 = kernel default */
1619                                                 in6p->in6p_hops = optval;
1620                                                 if ((in6p->inp_vflag &
1621                                                      INP_IPV4) != 0)
1622                                                         in6p->inp_ip_ttl = optval;
1623                                         }
1624                                         break;
1625 #define OPTSET(bit) \
1626 do { \
1627         INP_WLOCK(in6p); \
1628         if (optval) \
1629                 in6p->inp_flags |= (bit); \
1630         else \
1631                 in6p->inp_flags &= ~(bit); \
1632         INP_WUNLOCK(in6p); \
1633 } while (/*CONSTCOND*/ 0)
1634 #define OPTSET2292(bit) \
1635 do { \
1636         INP_WLOCK(in6p); \
1637         in6p->inp_flags |= IN6P_RFC2292; \
1638         if (optval) \
1639                 in6p->inp_flags |= (bit); \
1640         else \
1641                 in6p->inp_flags &= ~(bit); \
1642         INP_WUNLOCK(in6p); \
1643 } while (/*CONSTCOND*/ 0)
1644 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1645
1646 #define OPTSET2_N(bit, val) do {                                        \
1647         if (val)                                                        \
1648                 in6p->inp_flags2 |= bit;                                \
1649         else                                                            \
1650                 in6p->inp_flags2 &= ~bit;                               \
1651 } while (0)
1652 #define OPTSET2(bit, val) do {                                          \
1653         INP_WLOCK(in6p);                                                \
1654         OPTSET2_N(bit, val);                                            \
1655         INP_WUNLOCK(in6p);                                              \
1656 } while (0)
1657 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1658 #define OPTSET2292_EXCLUSIVE(bit)                                       \
1659 do {                                                                    \
1660         INP_WLOCK(in6p);                                                \
1661         if (OPTBIT(IN6P_RFC2292)) {                                     \
1662                 error = EINVAL;                                         \
1663         } else {                                                        \
1664                 if (optval)                                             \
1665                         in6p->inp_flags |= (bit);                       \
1666                 else                                                    \
1667                         in6p->inp_flags &= ~(bit);                      \
1668         }                                                               \
1669         INP_WUNLOCK(in6p);                                              \
1670 } while (/*CONSTCOND*/ 0)
1671
1672                                 case IPV6_RECVPKTINFO:
1673                                         OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1674                                         break;
1675
1676                                 case IPV6_HOPLIMIT:
1677                                 {
1678                                         struct ip6_pktopts **optp;
1679
1680                                         /* cannot mix with RFC2292 */
1681                                         if (OPTBIT(IN6P_RFC2292)) {
1682                                                 error = EINVAL;
1683                                                 break;
1684                                         }
1685                                         INP_WLOCK(in6p);
1686                                         if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1687                                                 INP_WUNLOCK(in6p);
1688                                                 return (ECONNRESET);
1689                                         }
1690                                         optp = &in6p->in6p_outputopts;
1691                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1692                                             (u_char *)&optval, sizeof(optval),
1693                                             optp, (td != NULL) ? td->td_ucred :
1694                                             NULL, uproto);
1695                                         INP_WUNLOCK(in6p);
1696                                         break;
1697                                 }
1698
1699                                 case IPV6_RECVHOPLIMIT:
1700                                         OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1701                                         break;
1702
1703                                 case IPV6_RECVHOPOPTS:
1704                                         OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1705                                         break;
1706
1707                                 case IPV6_RECVDSTOPTS:
1708                                         OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1709                                         break;
1710
1711                                 case IPV6_RECVRTHDRDSTOPTS:
1712                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1713                                         break;
1714
1715                                 case IPV6_RECVRTHDR:
1716                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1717                                         break;
1718
1719                                 case IPV6_RECVPATHMTU:
1720                                         /*
1721                                          * We ignore this option for TCP
1722                                          * sockets.
1723                                          * (RFC3542 leaves this case
1724                                          * unspecified.)
1725                                          */
1726                                         if (uproto != IPPROTO_TCP)
1727                                                 OPTSET(IN6P_MTU);
1728                                         break;
1729
1730                                 case IPV6_RECVFLOWID:
1731                                         OPTSET2(INP_RECVFLOWID, optval);
1732                                         break;
1733
1734 #ifdef  RSS
1735                                 case IPV6_RECVRSSBUCKETID:
1736                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1737                                         break;
1738 #endif
1739
1740                                 case IPV6_V6ONLY:
1741                                         /*
1742                                          * make setsockopt(IPV6_V6ONLY)
1743                                          * available only prior to bind(2).
1744                                          * see ipng mailing list, Jun 22 2001.
1745                                          */
1746                                         if (in6p->inp_lport ||
1747                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1748                                                 error = EINVAL;
1749                                                 break;
1750                                         }
1751                                         OPTSET(IN6P_IPV6_V6ONLY);
1752                                         if (optval)
1753                                                 in6p->inp_vflag &= ~INP_IPV4;
1754                                         else
1755                                                 in6p->inp_vflag |= INP_IPV4;
1756                                         break;
1757                                 case IPV6_RECVTCLASS:
1758                                         /* cannot mix with RFC2292 XXX */
1759                                         OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1760                                         break;
1761                                 case IPV6_AUTOFLOWLABEL:
1762                                         OPTSET(IN6P_AUTOFLOWLABEL);
1763                                         break;
1764
1765                                 case IPV6_ORIGDSTADDR:
1766                                         OPTSET2(INP_ORIGDSTADDR, optval);
1767                                         break;
1768                                 case IPV6_BINDANY:
1769                                         OPTSET(INP_BINDANY);
1770                                         break;
1771
1772                                 case IPV6_BINDMULTI:
1773                                         OPTSET2(INP_BINDMULTI, optval);
1774                                         break;
1775 #ifdef  RSS
1776                                 case IPV6_RSS_LISTEN_BUCKET:
1777                                         if ((optval >= 0) &&
1778                                             (optval < rss_getnumbuckets())) {
1779                                                 INP_WLOCK(in6p);
1780                                                 in6p->inp_rss_listen_bucket = optval;
1781                                                 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1782                                                 INP_WUNLOCK(in6p);
1783                                         } else {
1784                                                 error = EINVAL;
1785                                         }
1786                                         break;
1787 #endif
1788                                 }
1789                                 break;
1790
1791                         case IPV6_TCLASS:
1792                         case IPV6_DONTFRAG:
1793                         case IPV6_USE_MIN_MTU:
1794                         case IPV6_PREFER_TEMPADDR:
1795                                 if (optlen != sizeof(optval)) {
1796                                         error = EINVAL;
1797                                         break;
1798                                 }
1799                                 error = sooptcopyin(sopt, &optval,
1800                                         sizeof optval, sizeof optval);
1801                                 if (error)
1802                                         break;
1803                                 {
1804                                         struct ip6_pktopts **optp;
1805                                         INP_WLOCK(in6p);
1806                                         if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1807                                                 INP_WUNLOCK(in6p);
1808                                                 return (ECONNRESET);
1809                                         }
1810                                         optp = &in6p->in6p_outputopts;
1811                                         error = ip6_pcbopt(optname,
1812                                             (u_char *)&optval, sizeof(optval),
1813                                             optp, (td != NULL) ? td->td_ucred :
1814                                             NULL, uproto);
1815                                         INP_WUNLOCK(in6p);
1816                                         break;
1817                                 }
1818
1819                         case IPV6_2292PKTINFO:
1820                         case IPV6_2292HOPLIMIT:
1821                         case IPV6_2292HOPOPTS:
1822                         case IPV6_2292DSTOPTS:
1823                         case IPV6_2292RTHDR:
1824                                 /* RFC 2292 */
1825                                 if (optlen != sizeof(int)) {
1826                                         error = EINVAL;
1827                                         break;
1828                                 }
1829                                 error = sooptcopyin(sopt, &optval,
1830                                         sizeof optval, sizeof optval);
1831                                 if (error)
1832                                         break;
1833                                 switch (optname) {
1834                                 case IPV6_2292PKTINFO:
1835                                         OPTSET2292(IN6P_PKTINFO);
1836                                         break;
1837                                 case IPV6_2292HOPLIMIT:
1838                                         OPTSET2292(IN6P_HOPLIMIT);
1839                                         break;
1840                                 case IPV6_2292HOPOPTS:
1841                                         /*
1842                                          * Check super-user privilege.
1843                                          * See comments for IPV6_RECVHOPOPTS.
1844                                          */
1845                                         if (td != NULL) {
1846                                                 error = priv_check(td,
1847                                                     PRIV_NETINET_SETHDROPTS);
1848                                                 if (error)
1849                                                         return (error);
1850                                         }
1851                                         OPTSET2292(IN6P_HOPOPTS);
1852                                         break;
1853                                 case IPV6_2292DSTOPTS:
1854                                         if (td != NULL) {
1855                                                 error = priv_check(td,
1856                                                     PRIV_NETINET_SETHDROPTS);
1857                                                 if (error)
1858                                                         return (error);
1859                                         }
1860                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1861                                         break;
1862                                 case IPV6_2292RTHDR:
1863                                         OPTSET2292(IN6P_RTHDR);
1864                                         break;
1865                                 }
1866                                 break;
1867                         case IPV6_PKTINFO:
1868                         case IPV6_HOPOPTS:
1869                         case IPV6_RTHDR:
1870                         case IPV6_DSTOPTS:
1871                         case IPV6_RTHDRDSTOPTS:
1872                         case IPV6_NEXTHOP:
1873                         {
1874                                 /* new advanced API (RFC3542) */
1875                                 u_char *optbuf;
1876                                 u_char optbuf_storage[MCLBYTES];
1877                                 int optlen;
1878                                 struct ip6_pktopts **optp;
1879
1880                                 /* cannot mix with RFC2292 */
1881                                 if (OPTBIT(IN6P_RFC2292)) {
1882                                         error = EINVAL;
1883                                         break;
1884                                 }
1885
1886                                 /*
1887                                  * We only ensure valsize is not too large
1888                                  * here.  Further validation will be done
1889                                  * later.
1890                                  */
1891                                 error = sooptcopyin(sopt, optbuf_storage,
1892                                     sizeof(optbuf_storage), 0);
1893                                 if (error)
1894                                         break;
1895                                 optlen = sopt->sopt_valsize;
1896                                 optbuf = optbuf_storage;
1897                                 INP_WLOCK(in6p);
1898                                 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1899                                         INP_WUNLOCK(in6p);
1900                                         return (ECONNRESET);
1901                                 }
1902                                 optp = &in6p->in6p_outputopts;
1903                                 error = ip6_pcbopt(optname, optbuf, optlen,
1904                                     optp, (td != NULL) ? td->td_ucred : NULL,
1905                                     uproto);
1906                                 INP_WUNLOCK(in6p);
1907                                 break;
1908                         }
1909 #undef OPTSET
1910
1911                         case IPV6_MULTICAST_IF:
1912                         case IPV6_MULTICAST_HOPS:
1913                         case IPV6_MULTICAST_LOOP:
1914                         case IPV6_JOIN_GROUP:
1915                         case IPV6_LEAVE_GROUP:
1916                         case IPV6_MSFILTER:
1917                         case MCAST_BLOCK_SOURCE:
1918                         case MCAST_UNBLOCK_SOURCE:
1919                         case MCAST_JOIN_GROUP:
1920                         case MCAST_LEAVE_GROUP:
1921                         case MCAST_JOIN_SOURCE_GROUP:
1922                         case MCAST_LEAVE_SOURCE_GROUP:
1923                                 error = ip6_setmoptions(in6p, sopt);
1924                                 break;
1925
1926                         case IPV6_PORTRANGE:
1927                                 error = sooptcopyin(sopt, &optval,
1928                                     sizeof optval, sizeof optval);
1929                                 if (error)
1930                                         break;
1931
1932                                 INP_WLOCK(in6p);
1933                                 switch (optval) {
1934                                 case IPV6_PORTRANGE_DEFAULT:
1935                                         in6p->inp_flags &= ~(INP_LOWPORT);
1936                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1937                                         break;
1938
1939                                 case IPV6_PORTRANGE_HIGH:
1940                                         in6p->inp_flags &= ~(INP_LOWPORT);
1941                                         in6p->inp_flags |= INP_HIGHPORT;
1942                                         break;
1943
1944                                 case IPV6_PORTRANGE_LOW:
1945                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1946                                         in6p->inp_flags |= INP_LOWPORT;
1947                                         break;
1948
1949                                 default:
1950                                         error = EINVAL;
1951                                         break;
1952                                 }
1953                                 INP_WUNLOCK(in6p);
1954                                 break;
1955
1956 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1957                         case IPV6_IPSEC_POLICY:
1958                                 if (IPSEC_ENABLED(ipv6)) {
1959                                         error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1960                                         break;
1961                                 }
1962                                 /* FALLTHROUGH */
1963 #endif /* IPSEC */
1964
1965                         default:
1966                                 error = ENOPROTOOPT;
1967                                 break;
1968                         }
1969                         break;
1970
1971                 case SOPT_GET:
1972                         switch (optname) {
1973
1974                         case IPV6_2292PKTOPTIONS:
1975 #ifdef IPV6_PKTOPTIONS
1976                         case IPV6_PKTOPTIONS:
1977 #endif
1978                                 /*
1979                                  * RFC3542 (effectively) deprecated the
1980                                  * semantics of the 2292-style pktoptions.
1981                                  * Since it was not reliable in nature (i.e.,
1982                                  * applications had to expect the lack of some
1983                                  * information after all), it would make sense
1984                                  * to simplify this part by always returning
1985                                  * empty data.
1986                                  */
1987                                 sopt->sopt_valsize = 0;
1988                                 break;
1989
1990                         case IPV6_RECVHOPOPTS:
1991                         case IPV6_RECVDSTOPTS:
1992                         case IPV6_RECVRTHDRDSTOPTS:
1993                         case IPV6_UNICAST_HOPS:
1994                         case IPV6_RECVPKTINFO:
1995                         case IPV6_RECVHOPLIMIT:
1996                         case IPV6_RECVRTHDR:
1997                         case IPV6_RECVPATHMTU:
1998
1999                         case IPV6_V6ONLY:
2000                         case IPV6_PORTRANGE:
2001                         case IPV6_RECVTCLASS:
2002                         case IPV6_AUTOFLOWLABEL:
2003                         case IPV6_BINDANY:
2004                         case IPV6_FLOWID:
2005                         case IPV6_FLOWTYPE:
2006                         case IPV6_RECVFLOWID:
2007 #ifdef  RSS
2008                         case IPV6_RSSBUCKETID:
2009                         case IPV6_RECVRSSBUCKETID:
2010 #endif
2011                         case IPV6_BINDMULTI:
2012                                 switch (optname) {
2013
2014                                 case IPV6_RECVHOPOPTS:
2015                                         optval = OPTBIT(IN6P_HOPOPTS);
2016                                         break;
2017
2018                                 case IPV6_RECVDSTOPTS:
2019                                         optval = OPTBIT(IN6P_DSTOPTS);
2020                                         break;
2021
2022                                 case IPV6_RECVRTHDRDSTOPTS:
2023                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2024                                         break;
2025
2026                                 case IPV6_UNICAST_HOPS:
2027                                         optval = in6p->in6p_hops;
2028                                         break;
2029
2030                                 case IPV6_RECVPKTINFO:
2031                                         optval = OPTBIT(IN6P_PKTINFO);
2032                                         break;
2033
2034                                 case IPV6_RECVHOPLIMIT:
2035                                         optval = OPTBIT(IN6P_HOPLIMIT);
2036                                         break;
2037
2038                                 case IPV6_RECVRTHDR:
2039                                         optval = OPTBIT(IN6P_RTHDR);
2040                                         break;
2041
2042                                 case IPV6_RECVPATHMTU:
2043                                         optval = OPTBIT(IN6P_MTU);
2044                                         break;
2045
2046                                 case IPV6_V6ONLY:
2047                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2048                                         break;
2049
2050                                 case IPV6_PORTRANGE:
2051                                     {
2052                                         int flags;
2053                                         flags = in6p->inp_flags;
2054                                         if (flags & INP_HIGHPORT)
2055                                                 optval = IPV6_PORTRANGE_HIGH;
2056                                         else if (flags & INP_LOWPORT)
2057                                                 optval = IPV6_PORTRANGE_LOW;
2058                                         else
2059                                                 optval = 0;
2060                                         break;
2061                                     }
2062                                 case IPV6_RECVTCLASS:
2063                                         optval = OPTBIT(IN6P_TCLASS);
2064                                         break;
2065
2066                                 case IPV6_AUTOFLOWLABEL:
2067                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2068                                         break;
2069
2070                                 case IPV6_ORIGDSTADDR:
2071                                         optval = OPTBIT2(INP_ORIGDSTADDR);
2072                                         break;
2073
2074                                 case IPV6_BINDANY:
2075                                         optval = OPTBIT(INP_BINDANY);
2076                                         break;
2077
2078                                 case IPV6_FLOWID:
2079                                         optval = in6p->inp_flowid;
2080                                         break;
2081
2082                                 case IPV6_FLOWTYPE:
2083                                         optval = in6p->inp_flowtype;
2084                                         break;
2085
2086                                 case IPV6_RECVFLOWID:
2087                                         optval = OPTBIT2(INP_RECVFLOWID);
2088                                         break;
2089 #ifdef  RSS
2090                                 case IPV6_RSSBUCKETID:
2091                                         retval =
2092                                             rss_hash2bucket(in6p->inp_flowid,
2093                                             in6p->inp_flowtype,
2094                                             &rss_bucket);
2095                                         if (retval == 0)
2096                                                 optval = rss_bucket;
2097                                         else
2098                                                 error = EINVAL;
2099                                         break;
2100
2101                                 case IPV6_RECVRSSBUCKETID:
2102                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2103                                         break;
2104 #endif
2105
2106                                 case IPV6_BINDMULTI:
2107                                         optval = OPTBIT2(INP_BINDMULTI);
2108                                         break;
2109
2110                                 }
2111                                 if (error)
2112                                         break;
2113                                 error = sooptcopyout(sopt, &optval,
2114                                         sizeof optval);
2115                                 break;
2116
2117                         case IPV6_PATHMTU:
2118                         {
2119                                 u_long pmtu = 0;
2120                                 struct ip6_mtuinfo mtuinfo;
2121                                 struct in6_addr addr;
2122
2123                                 if (!(so->so_state & SS_ISCONNECTED))
2124                                         return (ENOTCONN);
2125                                 /*
2126                                  * XXX: we dot not consider the case of source
2127                                  * routing, or optional information to specify
2128                                  * the outgoing interface.
2129                                  * Copy faddr out of in6p to avoid holding lock
2130                                  * on inp during route lookup.
2131                                  */
2132                                 INP_RLOCK(in6p);
2133                                 bcopy(&in6p->in6p_faddr, &addr, sizeof(addr));
2134                                 INP_RUNLOCK(in6p);
2135                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2136                                     &addr, &pmtu);
2137                                 if (error)
2138                                         break;
2139                                 if (pmtu > IPV6_MAXPACKET)
2140                                         pmtu = IPV6_MAXPACKET;
2141
2142                                 bzero(&mtuinfo, sizeof(mtuinfo));
2143                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2144                                 optdata = (void *)&mtuinfo;
2145                                 optdatalen = sizeof(mtuinfo);
2146                                 error = sooptcopyout(sopt, optdata,
2147                                     optdatalen);
2148                                 break;
2149                         }
2150
2151                         case IPV6_2292PKTINFO:
2152                         case IPV6_2292HOPLIMIT:
2153                         case IPV6_2292HOPOPTS:
2154                         case IPV6_2292RTHDR:
2155                         case IPV6_2292DSTOPTS:
2156                                 switch (optname) {
2157                                 case IPV6_2292PKTINFO:
2158                                         optval = OPTBIT(IN6P_PKTINFO);
2159                                         break;
2160                                 case IPV6_2292HOPLIMIT:
2161                                         optval = OPTBIT(IN6P_HOPLIMIT);
2162                                         break;
2163                                 case IPV6_2292HOPOPTS:
2164                                         optval = OPTBIT(IN6P_HOPOPTS);
2165                                         break;
2166                                 case IPV6_2292RTHDR:
2167                                         optval = OPTBIT(IN6P_RTHDR);
2168                                         break;
2169                                 case IPV6_2292DSTOPTS:
2170                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2171                                         break;
2172                                 }
2173                                 error = sooptcopyout(sopt, &optval,
2174                                     sizeof optval);
2175                                 break;
2176                         case IPV6_PKTINFO:
2177                         case IPV6_HOPOPTS:
2178                         case IPV6_RTHDR:
2179                         case IPV6_DSTOPTS:
2180                         case IPV6_RTHDRDSTOPTS:
2181                         case IPV6_NEXTHOP:
2182                         case IPV6_TCLASS:
2183                         case IPV6_DONTFRAG:
2184                         case IPV6_USE_MIN_MTU:
2185                         case IPV6_PREFER_TEMPADDR:
2186                                 error = ip6_getpcbopt(in6p, optname, sopt);
2187                                 break;
2188
2189                         case IPV6_MULTICAST_IF:
2190                         case IPV6_MULTICAST_HOPS:
2191                         case IPV6_MULTICAST_LOOP:
2192                         case IPV6_MSFILTER:
2193                                 error = ip6_getmoptions(in6p, sopt);
2194                                 break;
2195
2196 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2197                         case IPV6_IPSEC_POLICY:
2198                                 if (IPSEC_ENABLED(ipv6)) {
2199                                         error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2200                                         break;
2201                                 }
2202                                 /* FALLTHROUGH */
2203 #endif /* IPSEC */
2204                         default:
2205                                 error = ENOPROTOOPT;
2206                                 break;
2207                         }
2208                         break;
2209                 }
2210         }
2211         return (error);
2212 }
2213
2214 int
2215 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2216 {
2217         int error = 0, optval, optlen;
2218         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2219         struct inpcb *in6p = sotoinpcb(so);
2220         int level, op, optname;
2221
2222         level = sopt->sopt_level;
2223         op = sopt->sopt_dir;
2224         optname = sopt->sopt_name;
2225         optlen = sopt->sopt_valsize;
2226
2227         if (level != IPPROTO_IPV6) {
2228                 return (EINVAL);
2229         }
2230
2231         switch (optname) {
2232         case IPV6_CHECKSUM:
2233                 /*
2234                  * For ICMPv6 sockets, no modification allowed for checksum
2235                  * offset, permit "no change" values to help existing apps.
2236                  *
2237                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2238                  * for an ICMPv6 socket will fail."
2239                  * The current behavior does not meet RFC3542.
2240                  */
2241                 switch (op) {
2242                 case SOPT_SET:
2243                         if (optlen != sizeof(int)) {
2244                                 error = EINVAL;
2245                                 break;
2246                         }
2247                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2248                                             sizeof(optval));
2249                         if (error)
2250                                 break;
2251                         if (optval < -1 || (optval % 2) != 0) {
2252                                 /*
2253                                  * The API assumes non-negative even offset
2254                                  * values or -1 as a special value.
2255                                  */
2256                                 error = EINVAL;
2257                         } else if (so->so_proto->pr_protocol ==
2258                             IPPROTO_ICMPV6) {
2259                                 if (optval != icmp6off)
2260                                         error = EINVAL;
2261                         } else
2262                                 in6p->in6p_cksum = optval;
2263                         break;
2264
2265                 case SOPT_GET:
2266                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2267                                 optval = icmp6off;
2268                         else
2269                                 optval = in6p->in6p_cksum;
2270
2271                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2272                         break;
2273
2274                 default:
2275                         error = EINVAL;
2276                         break;
2277                 }
2278                 break;
2279
2280         default:
2281                 error = ENOPROTOOPT;
2282                 break;
2283         }
2284
2285         return (error);
2286 }
2287
2288 /*
2289  * Set up IP6 options in pcb for insertion in output packets or
2290  * specifying behavior of outgoing packets.
2291  */
2292 static int
2293 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2294     struct socket *so, struct sockopt *sopt)
2295 {
2296         struct ip6_pktopts *opt = *pktopt;
2297         int error = 0;
2298         struct thread *td = sopt->sopt_td;
2299
2300         /* turn off any old options. */
2301         if (opt) {
2302 #ifdef DIAGNOSTIC
2303                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2304                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2305                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2306                         printf("ip6_pcbopts: all specified options are cleared.\n");
2307 #endif
2308                 ip6_clearpktopts(opt, -1);
2309         } else
2310                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2311         *pktopt = NULL;
2312
2313         if (!m || m->m_len == 0) {
2314                 /*
2315                  * Only turning off any previous options, regardless of
2316                  * whether the opt is just created or given.
2317                  */
2318                 free(opt, M_IP6OPT);
2319                 return (0);
2320         }
2321
2322         /*  set options specified by user. */
2323         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2324             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2325                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2326                 free(opt, M_IP6OPT);
2327                 return (error);
2328         }
2329         *pktopt = opt;
2330         return (0);
2331 }
2332
2333 /*
2334  * initialize ip6_pktopts.  beware that there are non-zero default values in
2335  * the struct.
2336  */
2337 void
2338 ip6_initpktopts(struct ip6_pktopts *opt)
2339 {
2340
2341         bzero(opt, sizeof(*opt));
2342         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2343         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2344         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2345         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2346 }
2347
2348 static int
2349 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2350     struct ucred *cred, int uproto)
2351 {
2352         struct ip6_pktopts *opt;
2353
2354         if (*pktopt == NULL) {
2355                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2356                     M_NOWAIT);
2357                 if (*pktopt == NULL)
2358                         return (ENOBUFS);
2359                 ip6_initpktopts(*pktopt);
2360         }
2361         opt = *pktopt;
2362
2363         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2364 }
2365
2366 #define GET_PKTOPT_VAR(field, lenexpr) do {                                     \
2367         if (pktopt && pktopt->field) {                                          \
2368                 INP_RUNLOCK(in6p);                                              \
2369                 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);         \
2370                 malloc_optdata = true;                                          \
2371                 INP_RLOCK(in6p);                                                \
2372                 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {           \
2373                         INP_RUNLOCK(in6p);                                      \
2374                         free(optdata, M_TEMP);                                  \
2375                         return (ECONNRESET);                                    \
2376                 }                                                               \
2377                 pktopt = in6p->in6p_outputopts;                                 \
2378                 if (pktopt && pktopt->field) {                                  \
2379                         optdatalen = min(lenexpr, sopt->sopt_valsize);          \
2380                         bcopy(&pktopt->field, optdata, optdatalen);             \
2381                 } else {                                                        \
2382                         free(optdata, M_TEMP);                                  \
2383                         optdata = NULL;                                         \
2384                         malloc_optdata = false;                                 \
2385                 }                                                               \
2386         }                                                                       \
2387 } while(0)
2388
2389 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,                         \
2390         (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2391
2392 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,                        \
2393         pktopt->field->sa_len)
2394
2395 static int
2396 ip6_getpcbopt(struct inpcb *in6p, int optname, struct sockopt *sopt)
2397 {
2398         void *optdata = NULL;
2399         bool malloc_optdata = false;
2400         int optdatalen = 0;
2401         int error = 0;
2402         struct in6_pktinfo null_pktinfo;
2403         int deftclass = 0, on;
2404         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2405         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2406         struct ip6_pktopts *pktopt;
2407
2408         INP_RLOCK(in6p);
2409         pktopt = in6p->in6p_outputopts;
2410
2411         switch (optname) {
2412         case IPV6_PKTINFO:
2413                 optdata = (void *)&null_pktinfo;
2414                 if (pktopt && pktopt->ip6po_pktinfo) {
2415                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2416                             sizeof(null_pktinfo));
2417                         in6_clearscope(&null_pktinfo.ipi6_addr);
2418                 } else {
2419                         /* XXX: we don't have to do this every time... */
2420                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2421                 }
2422                 optdatalen = sizeof(struct in6_pktinfo);
2423                 break;
2424         case IPV6_TCLASS:
2425                 if (pktopt && pktopt->ip6po_tclass >= 0)
2426                         deftclass = pktopt->ip6po_tclass;
2427                 optdata = (void *)&deftclass;
2428                 optdatalen = sizeof(int);
2429                 break;
2430         case IPV6_HOPOPTS:
2431                 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2432                 break;
2433         case IPV6_RTHDR:
2434                 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2435                 break;
2436         case IPV6_RTHDRDSTOPTS:
2437                 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2438                 break;
2439         case IPV6_DSTOPTS:
2440                 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2441                 break;
2442         case IPV6_NEXTHOP:
2443                 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2444                 break;
2445         case IPV6_USE_MIN_MTU:
2446                 if (pktopt)
2447                         defminmtu = pktopt->ip6po_minmtu;
2448                 optdata = (void *)&defminmtu;
2449                 optdatalen = sizeof(int);
2450                 break;
2451         case IPV6_DONTFRAG:
2452                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2453                         on = 1;
2454                 else
2455                         on = 0;
2456                 optdata = (void *)&on;
2457                 optdatalen = sizeof(on);
2458                 break;
2459         case IPV6_PREFER_TEMPADDR:
2460                 if (pktopt)
2461                         defpreftemp = pktopt->ip6po_prefer_tempaddr;
2462                 optdata = (void *)&defpreftemp;
2463                 optdatalen = sizeof(int);
2464                 break;
2465         default:                /* should not happen */
2466 #ifdef DIAGNOSTIC
2467                 panic("ip6_getpcbopt: unexpected option\n");
2468 #endif
2469                 INP_RUNLOCK(in6p);
2470                 return (ENOPROTOOPT);
2471         }
2472         INP_RUNLOCK(in6p);
2473
2474         error = sooptcopyout(sopt, optdata, optdatalen);
2475         if (malloc_optdata)
2476                 free(optdata, M_TEMP);
2477
2478         return (error);
2479 }
2480
2481 void
2482 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2483 {
2484         if (pktopt == NULL)
2485                 return;
2486
2487         if (optname == -1 || optname == IPV6_PKTINFO) {
2488                 if (pktopt->ip6po_pktinfo)
2489                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2490                 pktopt->ip6po_pktinfo = NULL;
2491         }
2492         if (optname == -1 || optname == IPV6_HOPLIMIT)
2493                 pktopt->ip6po_hlim = -1;
2494         if (optname == -1 || optname == IPV6_TCLASS)
2495                 pktopt->ip6po_tclass = -1;
2496         if (optname == -1 || optname == IPV6_NEXTHOP) {
2497                 if (pktopt->ip6po_nextroute.ro_rt) {
2498                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2499                         pktopt->ip6po_nextroute.ro_rt = NULL;
2500                 }
2501                 if (pktopt->ip6po_nexthop)
2502                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2503                 pktopt->ip6po_nexthop = NULL;
2504         }
2505         if (optname == -1 || optname == IPV6_HOPOPTS) {
2506                 if (pktopt->ip6po_hbh)
2507                         free(pktopt->ip6po_hbh, M_IP6OPT);
2508                 pktopt->ip6po_hbh = NULL;
2509         }
2510         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2511                 if (pktopt->ip6po_dest1)
2512                         free(pktopt->ip6po_dest1, M_IP6OPT);
2513                 pktopt->ip6po_dest1 = NULL;
2514         }
2515         if (optname == -1 || optname == IPV6_RTHDR) {
2516                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2517                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2518                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2519                 if (pktopt->ip6po_route.ro_rt) {
2520                         RTFREE(pktopt->ip6po_route.ro_rt);
2521                         pktopt->ip6po_route.ro_rt = NULL;
2522                 }
2523         }
2524         if (optname == -1 || optname == IPV6_DSTOPTS) {
2525                 if (pktopt->ip6po_dest2)
2526                         free(pktopt->ip6po_dest2, M_IP6OPT);
2527                 pktopt->ip6po_dest2 = NULL;
2528         }
2529 }
2530
2531 #define PKTOPT_EXTHDRCPY(type) \
2532 do {\
2533         if (src->type) {\
2534                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2535                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2536                 if (dst->type == NULL)\
2537                         goto bad;\
2538                 bcopy(src->type, dst->type, hlen);\
2539         }\
2540 } while (/*CONSTCOND*/ 0)
2541
2542 static int
2543 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2544 {
2545         if (dst == NULL || src == NULL)  {
2546                 printf("ip6_clearpktopts: invalid argument\n");
2547                 return (EINVAL);
2548         }
2549
2550         dst->ip6po_hlim = src->ip6po_hlim;
2551         dst->ip6po_tclass = src->ip6po_tclass;
2552         dst->ip6po_flags = src->ip6po_flags;
2553         dst->ip6po_minmtu = src->ip6po_minmtu;
2554         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2555         if (src->ip6po_pktinfo) {
2556                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2557                     M_IP6OPT, canwait);
2558                 if (dst->ip6po_pktinfo == NULL)
2559                         goto bad;
2560                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2561         }
2562         if (src->ip6po_nexthop) {
2563                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2564                     M_IP6OPT, canwait);
2565                 if (dst->ip6po_nexthop == NULL)
2566                         goto bad;
2567                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2568                     src->ip6po_nexthop->sa_len);
2569         }
2570         PKTOPT_EXTHDRCPY(ip6po_hbh);
2571         PKTOPT_EXTHDRCPY(ip6po_dest1);
2572         PKTOPT_EXTHDRCPY(ip6po_dest2);
2573         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2574         return (0);
2575
2576   bad:
2577         ip6_clearpktopts(dst, -1);
2578         return (ENOBUFS);
2579 }
2580 #undef PKTOPT_EXTHDRCPY
2581
2582 struct ip6_pktopts *
2583 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2584 {
2585         int error;
2586         struct ip6_pktopts *dst;
2587
2588         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2589         if (dst == NULL)
2590                 return (NULL);
2591         ip6_initpktopts(dst);
2592
2593         if ((error = copypktopts(dst, src, canwait)) != 0) {
2594                 free(dst, M_IP6OPT);
2595                 return (NULL);
2596         }
2597
2598         return (dst);
2599 }
2600
2601 void
2602 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2603 {
2604         if (pktopt == NULL)
2605                 return;
2606
2607         ip6_clearpktopts(pktopt, -1);
2608
2609         free(pktopt, M_IP6OPT);
2610 }
2611
2612 /*
2613  * Set IPv6 outgoing packet options based on advanced API.
2614  */
2615 int
2616 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2617     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2618 {
2619         struct cmsghdr *cm = NULL;
2620
2621         if (control == NULL || opt == NULL)
2622                 return (EINVAL);
2623
2624         ip6_initpktopts(opt);
2625         if (stickyopt) {
2626                 int error;
2627
2628                 /*
2629                  * If stickyopt is provided, make a local copy of the options
2630                  * for this particular packet, then override them by ancillary
2631                  * objects.
2632                  * XXX: copypktopts() does not copy the cached route to a next
2633                  * hop (if any).  This is not very good in terms of efficiency,
2634                  * but we can allow this since this option should be rarely
2635                  * used.
2636                  */
2637                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2638                         return (error);
2639         }
2640
2641         /*
2642          * XXX: Currently, we assume all the optional information is stored
2643          * in a single mbuf.
2644          */
2645         if (control->m_next)
2646                 return (EINVAL);
2647
2648         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2649             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2650                 int error;
2651
2652                 if (control->m_len < CMSG_LEN(0))
2653                         return (EINVAL);
2654
2655                 cm = mtod(control, struct cmsghdr *);
2656                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2657                         return (EINVAL);
2658                 if (cm->cmsg_level != IPPROTO_IPV6)
2659                         continue;
2660
2661                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2662                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2663                 if (error)
2664                         return (error);
2665         }
2666
2667         return (0);
2668 }
2669
2670 /*
2671  * Set a particular packet option, as a sticky option or an ancillary data
2672  * item.  "len" can be 0 only when it's a sticky option.
2673  * We have 4 cases of combination of "sticky" and "cmsg":
2674  * "sticky=0, cmsg=0": impossible
2675  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2676  * "sticky=1, cmsg=0": RFC3542 socket option
2677  * "sticky=1, cmsg=1": RFC2292 socket option
2678  */
2679 static int
2680 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2681     struct ucred *cred, int sticky, int cmsg, int uproto)
2682 {
2683         int minmtupolicy, preftemp;
2684         int error;
2685
2686         if (!sticky && !cmsg) {
2687 #ifdef DIAGNOSTIC
2688                 printf("ip6_setpktopt: impossible case\n");
2689 #endif
2690                 return (EINVAL);
2691         }
2692
2693         /*
2694          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2695          * not be specified in the context of RFC3542.  Conversely,
2696          * RFC3542 types should not be specified in the context of RFC2292.
2697          */
2698         if (!cmsg) {
2699                 switch (optname) {
2700                 case IPV6_2292PKTINFO:
2701                 case IPV6_2292HOPLIMIT:
2702                 case IPV6_2292NEXTHOP:
2703                 case IPV6_2292HOPOPTS:
2704                 case IPV6_2292DSTOPTS:
2705                 case IPV6_2292RTHDR:
2706                 case IPV6_2292PKTOPTIONS:
2707                         return (ENOPROTOOPT);
2708                 }
2709         }
2710         if (sticky && cmsg) {
2711                 switch (optname) {
2712                 case IPV6_PKTINFO:
2713                 case IPV6_HOPLIMIT:
2714                 case IPV6_NEXTHOP:
2715                 case IPV6_HOPOPTS:
2716                 case IPV6_DSTOPTS:
2717                 case IPV6_RTHDRDSTOPTS:
2718                 case IPV6_RTHDR:
2719                 case IPV6_USE_MIN_MTU:
2720                 case IPV6_DONTFRAG:
2721                 case IPV6_TCLASS:
2722                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2723                         return (ENOPROTOOPT);
2724                 }
2725         }
2726
2727         switch (optname) {
2728         case IPV6_2292PKTINFO:
2729         case IPV6_PKTINFO:
2730         {
2731                 struct ifnet *ifp = NULL;
2732                 struct in6_pktinfo *pktinfo;
2733
2734                 if (len != sizeof(struct in6_pktinfo))
2735                         return (EINVAL);
2736
2737                 pktinfo = (struct in6_pktinfo *)buf;
2738
2739                 /*
2740                  * An application can clear any sticky IPV6_PKTINFO option by
2741                  * doing a "regular" setsockopt with ipi6_addr being
2742                  * in6addr_any and ipi6_ifindex being zero.
2743                  * [RFC 3542, Section 6]
2744                  */
2745                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2746                     pktinfo->ipi6_ifindex == 0 &&
2747                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2748                         ip6_clearpktopts(opt, optname);
2749                         break;
2750                 }
2751
2752                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2753                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2754                         return (EINVAL);
2755                 }
2756                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2757                         return (EINVAL);
2758                 /* validate the interface index if specified. */
2759                 if (pktinfo->ipi6_ifindex > V_if_index)
2760                          return (ENXIO);
2761                 if (pktinfo->ipi6_ifindex) {
2762                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2763                         if (ifp == NULL)
2764                                 return (ENXIO);
2765                 }
2766                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2767                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2768                         return (ENETDOWN);
2769
2770                 if (ifp != NULL &&
2771                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2772                         struct in6_ifaddr *ia;
2773
2774                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2775                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2776                         if (ia == NULL)
2777                                 return (EADDRNOTAVAIL);
2778                         ifa_free(&ia->ia_ifa);
2779                 }
2780                 /*
2781                  * We store the address anyway, and let in6_selectsrc()
2782                  * validate the specified address.  This is because ipi6_addr
2783                  * may not have enough information about its scope zone, and
2784                  * we may need additional information (such as outgoing
2785                  * interface or the scope zone of a destination address) to
2786                  * disambiguate the scope.
2787                  * XXX: the delay of the validation may confuse the
2788                  * application when it is used as a sticky option.
2789                  */
2790                 if (opt->ip6po_pktinfo == NULL) {
2791                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2792                             M_IP6OPT, M_NOWAIT);
2793                         if (opt->ip6po_pktinfo == NULL)
2794                                 return (ENOBUFS);
2795                 }
2796                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2797                 break;
2798         }
2799
2800         case IPV6_2292HOPLIMIT:
2801         case IPV6_HOPLIMIT:
2802         {
2803                 int *hlimp;
2804
2805                 /*
2806                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2807                  * to simplify the ordering among hoplimit options.
2808                  */
2809                 if (optname == IPV6_HOPLIMIT && sticky)
2810                         return (ENOPROTOOPT);
2811
2812                 if (len != sizeof(int))
2813                         return (EINVAL);
2814                 hlimp = (int *)buf;
2815                 if (*hlimp < -1 || *hlimp > 255)
2816                         return (EINVAL);
2817
2818                 opt->ip6po_hlim = *hlimp;
2819                 break;
2820         }
2821
2822         case IPV6_TCLASS:
2823         {
2824                 int tclass;
2825
2826                 if (len != sizeof(int))
2827                         return (EINVAL);
2828                 tclass = *(int *)buf;
2829                 if (tclass < -1 || tclass > 255)
2830                         return (EINVAL);
2831
2832                 opt->ip6po_tclass = tclass;
2833                 break;
2834         }
2835
2836         case IPV6_2292NEXTHOP:
2837         case IPV6_NEXTHOP:
2838                 if (cred != NULL) {
2839                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
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, PRIV_NETINET_SETHDROPTS);
2898                         if (error)
2899                                 return (error);
2900                 }
2901
2902                 if (len == 0) {
2903                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2904                         break;  /* just remove the option */
2905                 }
2906
2907                 /* message length validation */
2908                 if (len < sizeof(struct ip6_hbh))
2909                         return (EINVAL);
2910                 hbh = (struct ip6_hbh *)buf;
2911                 hbhlen = (hbh->ip6h_len + 1) << 3;
2912                 if (len != hbhlen)
2913                         return (EINVAL);
2914
2915                 /* turn off the previous option, then set the new option. */
2916                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2917                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2918                 if (opt->ip6po_hbh == NULL)
2919                         return (ENOBUFS);
2920                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2921
2922                 break;
2923         }
2924
2925         case IPV6_2292DSTOPTS:
2926         case IPV6_DSTOPTS:
2927         case IPV6_RTHDRDSTOPTS:
2928         {
2929                 struct ip6_dest *dest, **newdest = NULL;
2930                 int destlen;
2931
2932                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2933                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2934                         if (error)
2935                                 return (error);
2936                 }
2937
2938                 if (len == 0) {
2939                         ip6_clearpktopts(opt, optname);
2940                         break;  /* just remove the option */
2941                 }
2942
2943                 /* message length validation */
2944                 if (len < sizeof(struct ip6_dest))
2945                         return (EINVAL);
2946                 dest = (struct ip6_dest *)buf;
2947                 destlen = (dest->ip6d_len + 1) << 3;
2948                 if (len != destlen)
2949                         return (EINVAL);
2950
2951                 /*
2952                  * Determine the position that the destination options header
2953                  * should be inserted; before or after the routing header.
2954                  */
2955                 switch (optname) {
2956                 case IPV6_2292DSTOPTS:
2957                         /*
2958                          * The old advacned API is ambiguous on this point.
2959                          * Our approach is to determine the position based
2960                          * according to the existence of a routing header.
2961                          * Note, however, that this depends on the order of the
2962                          * extension headers in the ancillary data; the 1st
2963                          * part of the destination options header must appear
2964                          * before the routing header in the ancillary data,
2965                          * too.
2966                          * RFC3542 solved the ambiguity by introducing
2967                          * separate ancillary data or option types.
2968                          */
2969                         if (opt->ip6po_rthdr == NULL)
2970                                 newdest = &opt->ip6po_dest1;
2971                         else
2972                                 newdest = &opt->ip6po_dest2;
2973                         break;
2974                 case IPV6_RTHDRDSTOPTS:
2975                         newdest = &opt->ip6po_dest1;
2976                         break;
2977                 case IPV6_DSTOPTS:
2978                         newdest = &opt->ip6po_dest2;
2979                         break;
2980                 }
2981
2982                 /* turn off the previous option, then set the new option. */
2983                 ip6_clearpktopts(opt, optname);
2984                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2985                 if (*newdest == NULL)
2986                         return (ENOBUFS);
2987                 bcopy(dest, *newdest, destlen);
2988
2989                 break;
2990         }
2991
2992         case IPV6_2292RTHDR:
2993         case IPV6_RTHDR:
2994         {
2995                 struct ip6_rthdr *rth;
2996                 int rthlen;
2997
2998                 if (len == 0) {
2999                         ip6_clearpktopts(opt, IPV6_RTHDR);
3000                         break;  /* just remove the option */
3001                 }
3002
3003                 /* message length validation */
3004                 if (len < sizeof(struct ip6_rthdr))
3005                         return (EINVAL);
3006                 rth = (struct ip6_rthdr *)buf;
3007                 rthlen = (rth->ip6r_len + 1) << 3;
3008                 if (len != rthlen)
3009                         return (EINVAL);
3010
3011                 switch (rth->ip6r_type) {
3012                 case IPV6_RTHDR_TYPE_0:
3013                         if (rth->ip6r_len == 0) /* must contain one addr */
3014                                 return (EINVAL);
3015                         if (rth->ip6r_len % 2) /* length must be even */
3016                                 return (EINVAL);
3017                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3018                                 return (EINVAL);
3019                         break;
3020                 default:
3021                         return (EINVAL);        /* not supported */
3022                 }
3023
3024                 /* turn off the previous option */
3025                 ip6_clearpktopts(opt, IPV6_RTHDR);
3026                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3027                 if (opt->ip6po_rthdr == NULL)
3028                         return (ENOBUFS);
3029                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3030
3031                 break;
3032         }
3033
3034         case IPV6_USE_MIN_MTU:
3035                 if (len != sizeof(int))
3036                         return (EINVAL);
3037                 minmtupolicy = *(int *)buf;
3038                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3039                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3040                     minmtupolicy != IP6PO_MINMTU_ALL) {
3041                         return (EINVAL);
3042                 }
3043                 opt->ip6po_minmtu = minmtupolicy;
3044                 break;
3045
3046         case IPV6_DONTFRAG:
3047                 if (len != sizeof(int))
3048                         return (EINVAL);
3049
3050                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3051                         /*
3052                          * we ignore this option for TCP sockets.
3053                          * (RFC3542 leaves this case unspecified.)
3054                          */
3055                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3056                 } else
3057                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3058                 break;
3059
3060         case IPV6_PREFER_TEMPADDR:
3061                 if (len != sizeof(int))
3062                         return (EINVAL);
3063                 preftemp = *(int *)buf;
3064                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3065                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3066                     preftemp != IP6PO_TEMPADDR_PREFER) {
3067                         return (EINVAL);
3068                 }
3069                 opt->ip6po_prefer_tempaddr = preftemp;
3070                 break;
3071
3072         default:
3073                 return (ENOPROTOOPT);
3074         } /* end of switch */
3075
3076         return (0);
3077 }
3078
3079 /*
3080  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3081  * packet to the input queue of a specified interface.  Note that this
3082  * calls the output routine of the loopback "driver", but with an interface
3083  * pointer that might NOT be &loif -- easier than replicating that code here.
3084  */
3085 void
3086 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3087 {
3088         struct mbuf *copym;
3089         struct ip6_hdr *ip6;
3090
3091         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3092         if (copym == NULL)
3093                 return;
3094
3095         /*
3096          * Make sure to deep-copy IPv6 header portion in case the data
3097          * is in an mbuf cluster, so that we can safely override the IPv6
3098          * header portion later.
3099          */
3100         if (!M_WRITABLE(copym) ||
3101             copym->m_len < sizeof(struct ip6_hdr)) {
3102                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3103                 if (copym == NULL)
3104                         return;
3105         }
3106         ip6 = mtod(copym, struct ip6_hdr *);
3107         /*
3108          * clear embedded scope identifiers if necessary.
3109          * in6_clearscope will touch the addresses only when necessary.
3110          */
3111         in6_clearscope(&ip6->ip6_src);
3112         in6_clearscope(&ip6->ip6_dst);
3113         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3114                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3115                     CSUM_PSEUDO_HDR;
3116                 copym->m_pkthdr.csum_data = 0xffff;
3117         }
3118         if_simloop(ifp, copym, AF_INET6, 0);
3119 }
3120
3121 /*
3122  * Chop IPv6 header off from the payload.
3123  */
3124 static int
3125 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3126 {
3127         struct mbuf *mh;
3128         struct ip6_hdr *ip6;
3129
3130         ip6 = mtod(m, struct ip6_hdr *);
3131         if (m->m_len > sizeof(*ip6)) {
3132                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3133                 if (mh == NULL) {
3134                         m_freem(m);
3135                         return ENOBUFS;
3136                 }
3137                 m_move_pkthdr(mh, m);
3138                 M_ALIGN(mh, sizeof(*ip6));
3139                 m->m_len -= sizeof(*ip6);
3140                 m->m_data += sizeof(*ip6);
3141                 mh->m_next = m;
3142                 m = mh;
3143                 m->m_len = sizeof(*ip6);
3144                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3145         }
3146         exthdrs->ip6e_ip6 = m;
3147         return 0;
3148 }
3149
3150 /*
3151  * Compute IPv6 extension header length.
3152  */
3153 int
3154 ip6_optlen(struct inpcb *in6p)
3155 {
3156         int len;
3157
3158         if (!in6p->in6p_outputopts)
3159                 return 0;
3160
3161         len = 0;
3162 #define elen(x) \
3163     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3164
3165         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3166         if (in6p->in6p_outputopts->ip6po_rthdr)
3167                 /* dest1 is valid with rthdr only */
3168                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3169         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3170         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3171         return len;
3172 #undef elen
3173 }