sys/netinet6/ip6_output.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-3-Clause
   3  *
   4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 3. Neither the name of the project nor the names of its contributors
  16  *    may be used to endorse or promote products derived from this software
  17  *    without specific prior written permission.
  18  *
  19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  29  * SUCH DAMAGE.
  30  *
  31  *      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
  32  */
  33
  34 /*-
  35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. Neither the name of the University nor the names of its contributors
  47  *    may be used to endorse or promote products derived from this software
  48  *    without specific prior written permission.
  49  *
  50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  60  * SUCH DAMAGE.
  61  *
  62  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  63  */
  64
  65 #include <sys/cdefs.h>
  66 __FBSDID("$FreeBSD$");
  67
  68 #include "opt_inet.h"
  69 #include "opt_inet6.h"
  70 #include "opt_ipsec.h"
  71 #include "opt_ratelimit.h"
  72 #include "opt_route.h"
  73 #include "opt_rss.h"
  74 #include "opt_sctp.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                 m = mb_unmapped_to_ext(m);
 967                 if (m == NULL) {
 968                         error = ENOBUFS;
 969                         IP6STAT_INC(ip6s_odropped);
 970                         goto bad;
 971                 }
 972                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 973         } else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
 974                 m = mb_unmapped_to_ext(m);
 975                 if (m == NULL) {
 976                         error = ENOBUFS;
 977                         IP6STAT_INC(ip6s_odropped);
 978                         goto bad;
 979                 }
 980         }
 981 #ifdef SCTP
 982         if (sw_csum & CSUM_SCTP_IPV6) {
 983                 sw_csum &= ~CSUM_SCTP_IPV6;
 984                 m = mb_unmapped_to_ext(m);
 985                 if (m == NULL) {
 986                         error = ENOBUFS;
 987                         IP6STAT_INC(ip6s_odropped);
 988                         goto bad;
 989                 }
 990                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 991         }
 992 #endif
 993         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 994         tlen = m->m_pkthdr.len;
 995
 996         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 997                 dontfrag = 1;
 998         else
 999                 dontfrag = 0;
1000         if (dontfrag && alwaysfrag) {   /* case 4 */
1001                 /* conflicting request - can't transmit */
1002                 error = EMSGSIZE;
1003                 goto bad;
1004         }
1005         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
1006                 /*
1007                  * Even if the DONTFRAG option is specified, we cannot send the
1008                  * packet when the data length is larger than the MTU of the
1009                  * outgoing interface.
1010                  * Notify the error by sending IPV6_PATHMTU ancillary data if
1011                  * application wanted to know the MTU value. Also return an
1012                  * error code (this is not described in the API spec).
1013                  */
1014                 if (inp != NULL)
1015                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1016                 error = EMSGSIZE;
1017                 goto bad;
1018         }
1019
1020         /*
1021          * transmit packet without fragmentation
1022          */
1023         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
1024                 struct in6_ifaddr *ia6;
1025
1026                 ip6 = mtod(m, struct ip6_hdr *);
1027                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1028                 if (ia6) {
1029                         /* Record statistics for this interface address. */
1030                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1031                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
1032                             m->m_pkthdr.len);
1033                         ifa_free(&ia6->ia_ifa);
1034                 }
1035                 error = ip6_output_send(inp, ifp, origifp, m, dst, ro);
1036                 goto done;
1037         }
1038
1039         /*
1040          * try to fragment the packet.  case 1-b and 3
1041          */
1042         if (mtu < IPV6_MMTU) {
1043                 /* path MTU cannot be less than IPV6_MMTU */
1044                 error = EMSGSIZE;
1045                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1046                 goto bad;
1047         } else if (ip6->ip6_plen == 0) {
1048                 /* jumbo payload cannot be fragmented */
1049                 error = EMSGSIZE;
1050                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1051                 goto bad;
1052         } else {
1053                 u_char nextproto;
1054
1055                 /*
1056                  * Too large for the destination or interface;
1057                  * fragment if possible.
1058                  * Must be able to put at least 8 bytes per fragment.
1059                  */
1060                 hlen = unfragpartlen;
1061                 if (mtu > IPV6_MAXPACKET)
1062                         mtu = IPV6_MAXPACKET;
1063
1064                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1065                 if (len < 8) {
1066                         error = EMSGSIZE;
1067                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1068                         goto bad;
1069                 }
1070
1071                 /*
1072                  * If the interface will not calculate checksums on
1073                  * fragmented packets, then do it here.
1074                  * XXX-BZ handle the hw offloading case.  Need flags.
1075                  */
1076                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1077                         m = mb_unmapped_to_ext(m);
1078                         if (m == NULL) {
1079                                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1080                                 error = ENOBUFS;
1081                                 goto bad;
1082                         }
1083                         in6_delayed_cksum(m, plen, hlen);
1084                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1085                 }
1086 #ifdef SCTP
1087                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1088                         m = mb_unmapped_to_ext(m);
1089                         if (m == NULL) {
1090                                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1091                                 error = ENOBUFS;
1092                                 goto bad;
1093                         }
1094                         sctp_delayed_cksum(m, hlen);
1095                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1096                 }
1097 #endif
1098                 /*
1099                  * Change the next header field of the last header in the
1100                  * unfragmentable part.
1101                  */
1102                 if (exthdrs.ip6e_rthdr) {
1103                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1104                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1105                 } else if (exthdrs.ip6e_dest1) {
1106                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1107                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1108                 } else if (exthdrs.ip6e_hbh) {
1109                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1110                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1111                 } else {
1112                         nextproto = ip6->ip6_nxt;
1113                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1114                 }
1115
1116                 /*
1117                  * Loop through length of segment after first fragment,
1118                  * make new header and copy data of each part and link onto
1119                  * chain.
1120                  */
1121                 m0 = m;
1122                 id = htonl(ip6_randomid());
1123                 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1124                         goto sendorfree;
1125
1126                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1127         }
1128
1129         /*
1130          * Remove leading garbages.
1131          */
1132 sendorfree:
1133         m = m0->m_nextpkt;
1134         m0->m_nextpkt = 0;
1135         m_freem(m0);
1136         for (; m; m = m0) {
1137                 m0 = m->m_nextpkt;
1138                 m->m_nextpkt = 0;
1139                 if (error == 0) {
1140                         /* Record statistics for this interface address. */
1141                         if (ia) {
1142                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1143                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1144                                     m->m_pkthdr.len);
1145                         }
1146                         error = ip6_output_send(inp, ifp, origifp, m, dst, ro);
1147                 } else
1148                         m_freem(m);
1149         }
1150
1151         if (error == 0)
1152                 IP6STAT_INC(ip6s_fragmented);
1153
1154 done:
1155         if (ro == &ip6route)
1156                 RO_RTFREE(ro);
1157         return (error);
1158
1159 freehdrs:
1160         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1161         m_freem(exthdrs.ip6e_dest1);
1162         m_freem(exthdrs.ip6e_rthdr);
1163         m_freem(exthdrs.ip6e_dest2);
1164         /* FALLTHROUGH */
1165 bad:
1166         if (m)
1167                 m_freem(m);
1168         goto done;
1169 }
1170
1171 static int
1172 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1173 {
1174         struct mbuf *m;
1175
1176         if (hlen > MCLBYTES)
1177                 return (ENOBUFS); /* XXX */
1178
1179         if (hlen > MLEN)
1180                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1181         else
1182                 m = m_get(M_NOWAIT, MT_DATA);
1183         if (m == NULL)
1184                 return (ENOBUFS);
1185         m->m_len = hlen;
1186         if (hdr)
1187                 bcopy(hdr, mtod(m, caddr_t), hlen);
1188
1189         *mp = m;
1190         return (0);
1191 }
1192
1193 /*
1194  * Insert jumbo payload option.
1195  */
1196 static int
1197 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1198 {
1199         struct mbuf *mopt;
1200         u_char *optbuf;
1201         u_int32_t v;
1202
1203 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1204
1205         /*
1206          * If there is no hop-by-hop options header, allocate new one.
1207          * If there is one but it doesn't have enough space to store the
1208          * jumbo payload option, allocate a cluster to store the whole options.
1209          * Otherwise, use it to store the options.
1210          */
1211         if (exthdrs->ip6e_hbh == NULL) {
1212                 mopt = m_get(M_NOWAIT, MT_DATA);
1213                 if (mopt == NULL)
1214                         return (ENOBUFS);
1215                 mopt->m_len = JUMBOOPTLEN;
1216                 optbuf = mtod(mopt, u_char *);
1217                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1218                 exthdrs->ip6e_hbh = mopt;
1219         } else {
1220                 struct ip6_hbh *hbh;
1221
1222                 mopt = exthdrs->ip6e_hbh;
1223                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1224                         /*
1225                          * XXX assumption:
1226                          * - exthdrs->ip6e_hbh is not referenced from places
1227                          *   other than exthdrs.
1228                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1229                          */
1230                         int oldoptlen = mopt->m_len;
1231                         struct mbuf *n;
1232
1233                         /*
1234                          * XXX: give up if the whole (new) hbh header does
1235                          * not fit even in an mbuf cluster.
1236                          */
1237                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1238                                 return (ENOBUFS);
1239
1240                         /*
1241                          * As a consequence, we must always prepare a cluster
1242                          * at this point.
1243                          */
1244                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1245                         if (n == NULL)
1246                                 return (ENOBUFS);
1247                         n->m_len = oldoptlen + JUMBOOPTLEN;
1248                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1249                             oldoptlen);
1250                         optbuf = mtod(n, caddr_t) + oldoptlen;
1251                         m_freem(mopt);
1252                         mopt = exthdrs->ip6e_hbh = n;
1253                 } else {
1254                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1255                         mopt->m_len += JUMBOOPTLEN;
1256                 }
1257                 optbuf[0] = IP6OPT_PADN;
1258                 optbuf[1] = 1;
1259
1260                 /*
1261                  * Adjust the header length according to the pad and
1262                  * the jumbo payload option.
1263                  */
1264                 hbh = mtod(mopt, struct ip6_hbh *);
1265                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1266         }
1267
1268         /* fill in the option. */
1269         optbuf[2] = IP6OPT_JUMBO;
1270         optbuf[3] = 4;
1271         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1272         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1273
1274         /* finally, adjust the packet header length */
1275         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1276
1277         return (0);
1278 #undef JUMBOOPTLEN
1279 }
1280
1281 /*
1282  * Insert fragment header and copy unfragmentable header portions.
1283  */
1284 static int
1285 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1286     struct ip6_frag **frghdrp)
1287 {
1288         struct mbuf *n, *mlast;
1289
1290         if (hlen > sizeof(struct ip6_hdr)) {
1291                 n = m_copym(m0, sizeof(struct ip6_hdr),
1292                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1293                 if (n == NULL)
1294                         return (ENOBUFS);
1295                 m->m_next = n;
1296         } else
1297                 n = m;
1298
1299         /* Search for the last mbuf of unfragmentable part. */
1300         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1301                 ;
1302
1303         if (M_WRITABLE(mlast) &&
1304             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1305                 /* use the trailing space of the last mbuf for the fragment hdr */
1306                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1307                     mlast->m_len);
1308                 mlast->m_len += sizeof(struct ip6_frag);
1309                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1310         } else {
1311                 /* allocate a new mbuf for the fragment header */
1312                 struct mbuf *mfrg;
1313
1314                 mfrg = m_get(M_NOWAIT, MT_DATA);
1315                 if (mfrg == NULL)
1316                         return (ENOBUFS);
1317                 mfrg->m_len = sizeof(struct ip6_frag);
1318                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1319                 mlast->m_next = mfrg;
1320         }
1321
1322         return (0);
1323 }
1324
1325 /*
1326  * Calculates IPv6 path mtu for destination @dst.
1327  * Resulting MTU is stored in @mtup.
1328  *
1329  * Returns 0 on success.
1330  */
1331 static int
1332 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1333 {
1334         struct nhop6_extended nh6;
1335         struct in6_addr kdst;
1336         uint32_t scopeid;
1337         struct ifnet *ifp;
1338         u_long mtu;
1339         int error;
1340
1341         in6_splitscope(dst, &kdst, &scopeid);
1342         if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1343                 return (EHOSTUNREACH);
1344
1345         ifp = nh6.nh_ifp;
1346         mtu = nh6.nh_mtu;
1347
1348         error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1349         fib6_free_nh_ext(fibnum, &nh6);
1350
1351         return (error);
1352 }
1353
1354 /*
1355  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1356  * and cached data in @ro_pmtu.
1357  * MTU from (successful) route lookup is saved (along with dst)
1358  * inside @ro_pmtu to avoid subsequent route lookups after packet
1359  * filter processing.
1360  *
1361  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1362  * Returns 0 on success.
1363  */
1364 static int
1365 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1366     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1367     int *alwaysfragp, u_int fibnum, u_int proto)
1368 {
1369         struct nhop6_basic nh6;
1370         struct in6_addr kdst;
1371         uint32_t scopeid;
1372         struct sockaddr_in6 *sa6_dst;
1373         u_long mtu;
1374
1375         mtu = 0;
1376         if (do_lookup) {
1377
1378                 /*
1379                  * Here ro_pmtu has final destination address, while
1380                  * ro might represent immediate destination.
1381                  * Use ro_pmtu destination since mtu might differ.
1382                  */
1383                 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1384                 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1385                         ro_pmtu->ro_mtu = 0;
1386
1387                 if (ro_pmtu->ro_mtu == 0) {
1388                         bzero(sa6_dst, sizeof(*sa6_dst));
1389                         sa6_dst->sin6_family = AF_INET6;
1390                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1391                         sa6_dst->sin6_addr = *dst;
1392
1393                         in6_splitscope(dst, &kdst, &scopeid);
1394                         if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1395                             &nh6) == 0)
1396                                 ro_pmtu->ro_mtu = nh6.nh_mtu;
1397                 }
1398
1399                 mtu = ro_pmtu->ro_mtu;
1400         }
1401
1402         if (ro_pmtu->ro_rt)
1403                 mtu = ro_pmtu->ro_rt->rt_mtu;
1404
1405         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1406 }
1407
1408 /*
1409  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1410  * hostcache data for @dst.
1411  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1412  *
1413  * Returns 0 on success.
1414  */
1415 static int
1416 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1417     u_long *mtup, int *alwaysfragp, u_int proto)
1418 {
1419         u_long mtu = 0;
1420         int alwaysfrag = 0;
1421         int error = 0;
1422
1423         if (rt_mtu > 0) {
1424                 u_int32_t ifmtu;
1425                 struct in_conninfo inc;
1426
1427                 bzero(&inc, sizeof(inc));
1428                 inc.inc_flags |= INC_ISIPV6;
1429                 inc.inc6_faddr = *dst;
1430
1431                 ifmtu = IN6_LINKMTU(ifp);
1432
1433                 /* TCP is known to react to pmtu changes so skip hc */
1434                 if (proto != IPPROTO_TCP)
1435                         mtu = tcp_hc_getmtu(&inc);
1436
1437                 if (mtu)
1438                         mtu = min(mtu, rt_mtu);
1439                 else
1440                         mtu = rt_mtu;
1441                 if (mtu == 0)
1442                         mtu = ifmtu;
1443                 else if (mtu < IPV6_MMTU) {
1444                         /*
1445                          * RFC2460 section 5, last paragraph:
1446                          * if we record ICMPv6 too big message with
1447                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1448                          * or smaller, with framgent header attached.
1449                          * (fragment header is needed regardless from the
1450                          * packet size, for translators to identify packets)
1451                          */
1452                         alwaysfrag = 1;
1453                         mtu = IPV6_MMTU;
1454                 }
1455         } else if (ifp) {
1456                 mtu = IN6_LINKMTU(ifp);
1457         } else
1458                 error = EHOSTUNREACH; /* XXX */
1459
1460         *mtup = mtu;
1461         if (alwaysfragp)
1462                 *alwaysfragp = alwaysfrag;
1463         return (error);
1464 }
1465
1466 /*
1467  * IP6 socket option processing.
1468  */
1469 int
1470 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1471 {
1472         int optdatalen, uproto;
1473         void *optdata;
1474         struct inpcb *inp = sotoinpcb(so);
1475         int error, optval;
1476         int level, op, optname;
1477         int optlen;
1478         struct thread *td;
1479 #ifdef  RSS
1480         uint32_t rss_bucket;
1481         int retval;
1482 #endif
1483
1484 /*
1485  * Don't use more than a quarter of mbuf clusters.  N.B.:
1486  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1487  * on LP64 architectures, so cast to u_long to avoid undefined
1488  * behavior.  ILP32 architectures cannot have nmbclusters
1489  * large enough to overflow for other reasons.
1490  */
1491 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1492
1493         level = sopt->sopt_level;
1494         op = sopt->sopt_dir;
1495         optname = sopt->sopt_name;
1496         optlen = sopt->sopt_valsize;
1497         td = sopt->sopt_td;
1498         error = 0;
1499         optval = 0;
1500         uproto = (int)so->so_proto->pr_protocol;
1501
1502         if (level != IPPROTO_IPV6) {
1503                 error = EINVAL;
1504
1505                 if (sopt->sopt_level == SOL_SOCKET &&
1506                     sopt->sopt_dir == SOPT_SET) {
1507                         switch (sopt->sopt_name) {
1508                         case SO_REUSEADDR:
1509                                 INP_WLOCK(inp);
1510                                 if ((so->so_options & SO_REUSEADDR) != 0)
1511                                         inp->inp_flags2 |= INP_REUSEADDR;
1512                                 else
1513                                         inp->inp_flags2 &= ~INP_REUSEADDR;
1514                                 INP_WUNLOCK(inp);
1515                                 error = 0;
1516                                 break;
1517                         case SO_REUSEPORT:
1518                                 INP_WLOCK(inp);
1519                                 if ((so->so_options & SO_REUSEPORT) != 0)
1520                                         inp->inp_flags2 |= INP_REUSEPORT;
1521                                 else
1522                                         inp->inp_flags2 &= ~INP_REUSEPORT;
1523                                 INP_WUNLOCK(inp);
1524                                 error = 0;
1525                                 break;
1526                         case SO_REUSEPORT_LB:
1527                                 INP_WLOCK(inp);
1528                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1529                                         inp->inp_flags2 |= INP_REUSEPORT_LB;
1530                                 else
1531                                         inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1532                                 INP_WUNLOCK(inp);
1533                                 error = 0;
1534                                 break;
1535                         case SO_SETFIB:
1536                                 INP_WLOCK(inp);
1537                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
1538                                 INP_WUNLOCK(inp);
1539                                 error = 0;
1540                                 break;
1541                         case SO_MAX_PACING_RATE:
1542 #ifdef RATELIMIT
1543                                 INP_WLOCK(inp);
1544                                 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1545                                 INP_WUNLOCK(inp);
1546                                 error = 0;
1547 #else
1548                                 error = EOPNOTSUPP;
1549 #endif
1550                                 break;
1551                         default:
1552                                 break;
1553                         }
1554                 }
1555         } else {                /* level == IPPROTO_IPV6 */
1556                 switch (op) {
1557
1558                 case SOPT_SET:
1559                         switch (optname) {
1560                         case IPV6_2292PKTOPTIONS:
1561 #ifdef IPV6_PKTOPTIONS
1562                         case IPV6_PKTOPTIONS:
1563 #endif
1564                         {
1565                                 struct mbuf *m;
1566
1567                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1568                                         printf("ip6_ctloutput: mbuf limit hit\n");
1569                                         error = ENOBUFS;
1570                                         break;
1571                                 }
1572
1573                                 error = soopt_getm(sopt, &m); /* XXX */
1574                                 if (error != 0)
1575                                         break;
1576                                 error = soopt_mcopyin(sopt, m); /* XXX */
1577                                 if (error != 0)
1578                                         break;
1579                                 error = ip6_pcbopts(&inp->in6p_outputopts,
1580                                                     m, so, sopt);
1581                                 m_freem(m); /* XXX */
1582                                 break;
1583                         }
1584
1585                         /*
1586                          * Use of some Hop-by-Hop options or some
1587                          * Destination options, might require special
1588                          * privilege.  That is, normal applications
1589                          * (without special privilege) might be forbidden
1590                          * from setting certain options in outgoing packets,
1591                          * and might never see certain options in received
1592                          * packets. [RFC 2292 Section 6]
1593                          * KAME specific note:
1594                          *  KAME prevents non-privileged users from sending or
1595                          *  receiving ANY hbh/dst options in order to avoid
1596                          *  overhead of parsing options in the kernel.
1597                          */
1598                         case IPV6_RECVHOPOPTS:
1599                         case IPV6_RECVDSTOPTS:
1600                         case IPV6_RECVRTHDRDSTOPTS:
1601                                 if (td != NULL) {
1602                                         error = priv_check(td,
1603                                             PRIV_NETINET_SETHDROPTS);
1604                                         if (error)
1605                                                 break;
1606                                 }
1607                                 /* FALLTHROUGH */
1608                         case IPV6_UNICAST_HOPS:
1609                         case IPV6_HOPLIMIT:
1610
1611                         case IPV6_RECVPKTINFO:
1612                         case IPV6_RECVHOPLIMIT:
1613                         case IPV6_RECVRTHDR:
1614                         case IPV6_RECVPATHMTU:
1615                         case IPV6_RECVTCLASS:
1616                         case IPV6_RECVFLOWID:
1617 #ifdef  RSS
1618                         case IPV6_RECVRSSBUCKETID:
1619 #endif
1620                         case IPV6_V6ONLY:
1621                         case IPV6_AUTOFLOWLABEL:
1622                         case IPV6_ORIGDSTADDR:
1623                         case IPV6_BINDANY:
1624                         case IPV6_BINDMULTI:
1625 #ifdef  RSS
1626                         case IPV6_RSS_LISTEN_BUCKET:
1627 #endif
1628                                 if (optname == IPV6_BINDANY && td != NULL) {
1629                                         error = priv_check(td,
1630                                             PRIV_NETINET_BINDANY);
1631                                         if (error)
1632                                                 break;
1633                                 }
1634
1635                                 if (optlen != sizeof(int)) {
1636                                         error = EINVAL;
1637                                         break;
1638                                 }
1639                                 error = sooptcopyin(sopt, &optval,
1640                                         sizeof optval, sizeof optval);
1641                                 if (error)
1642                                         break;
1643                                 switch (optname) {
1644
1645                                 case IPV6_UNICAST_HOPS:
1646                                         if (optval < -1 || optval >= 256)
1647                                                 error = EINVAL;
1648                                         else {
1649                                                 /* -1 = kernel default */
1650                                                 inp->in6p_hops = optval;
1651                                                 if ((inp->inp_vflag &
1652                                                      INP_IPV4) != 0)
1653                                                         inp->inp_ip_ttl = optval;
1654                                         }
1655                                         break;
1656 #define OPTSET(bit) \
1657 do { \
1658         INP_WLOCK(inp); \
1659         if (optval) \
1660                 inp->inp_flags |= (bit); \
1661         else \
1662                 inp->inp_flags &= ~(bit); \
1663         INP_WUNLOCK(inp); \
1664 } while (/*CONSTCOND*/ 0)
1665 #define OPTSET2292(bit) \
1666 do { \
1667         INP_WLOCK(inp); \
1668         inp->inp_flags |= IN6P_RFC2292; \
1669         if (optval) \
1670                 inp->inp_flags |= (bit); \
1671         else \
1672                 inp->inp_flags &= ~(bit); \
1673         INP_WUNLOCK(inp); \
1674 } while (/*CONSTCOND*/ 0)
1675 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1676
1677 #define OPTSET2_N(bit, val) do {                                        \
1678         if (val)                                                        \
1679                 inp->inp_flags2 |= bit;                                 \
1680         else                                                            \
1681                 inp->inp_flags2 &= ~bit;                                \
1682 } while (0)
1683 #define OPTSET2(bit, val) do {                                          \
1684         INP_WLOCK(inp);                                                 \
1685         OPTSET2_N(bit, val);                                            \
1686         INP_WUNLOCK(inp);                                               \
1687 } while (0)
1688 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1689 #define OPTSET2292_EXCLUSIVE(bit)                                       \
1690 do {                                                                    \
1691         INP_WLOCK(inp);                                                 \
1692         if (OPTBIT(IN6P_RFC2292)) {                                     \
1693                 error = EINVAL;                                         \
1694         } else {                                                        \
1695                 if (optval)                                             \
1696                         inp->inp_flags |= (bit);                        \
1697                 else                                                    \
1698                         inp->inp_flags &= ~(bit);                       \
1699         }                                                               \
1700         INP_WUNLOCK(inp);                                               \
1701 } while (/*CONSTCOND*/ 0)
1702
1703                                 case IPV6_RECVPKTINFO:
1704                                         OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1705                                         break;
1706
1707                                 case IPV6_HOPLIMIT:
1708                                 {
1709                                         struct ip6_pktopts **optp;
1710
1711                                         /* cannot mix with RFC2292 */
1712                                         if (OPTBIT(IN6P_RFC2292)) {
1713                                                 error = EINVAL;
1714                                                 break;
1715                                         }
1716                                         INP_WLOCK(inp);
1717                                         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1718                                                 INP_WUNLOCK(inp);
1719                                                 return (ECONNRESET);
1720                                         }
1721                                         optp = &inp->in6p_outputopts;
1722                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1723                                             (u_char *)&optval, sizeof(optval),
1724                                             optp, (td != NULL) ? td->td_ucred :
1725                                             NULL, uproto);
1726                                         INP_WUNLOCK(inp);
1727                                         break;
1728                                 }
1729
1730                                 case IPV6_RECVHOPLIMIT:
1731                                         OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1732                                         break;
1733
1734                                 case IPV6_RECVHOPOPTS:
1735                                         OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1736                                         break;
1737
1738                                 case IPV6_RECVDSTOPTS:
1739                                         OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1740                                         break;
1741
1742                                 case IPV6_RECVRTHDRDSTOPTS:
1743                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1744                                         break;
1745
1746                                 case IPV6_RECVRTHDR:
1747                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1748                                         break;
1749
1750                                 case IPV6_RECVPATHMTU:
1751                                         /*
1752                                          * We ignore this option for TCP
1753                                          * sockets.
1754                                          * (RFC3542 leaves this case
1755                                          * unspecified.)
1756                                          */
1757                                         if (uproto != IPPROTO_TCP)
1758                                                 OPTSET(IN6P_MTU);
1759                                         break;
1760
1761                                 case IPV6_RECVFLOWID:
1762                                         OPTSET2(INP_RECVFLOWID, optval);
1763                                         break;
1764
1765 #ifdef  RSS
1766                                 case IPV6_RECVRSSBUCKETID:
1767                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1768                                         break;
1769 #endif
1770
1771                                 case IPV6_V6ONLY:
1772                                         /*
1773                                          * make setsockopt(IPV6_V6ONLY)
1774                                          * available only prior to bind(2).
1775                                          * see ipng mailing list, Jun 22 2001.
1776                                          */
1777                                         if (inp->inp_lport ||
1778                                             !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1779                                                 error = EINVAL;
1780                                                 break;
1781                                         }
1782                                         OPTSET(IN6P_IPV6_V6ONLY);
1783                                         if (optval)
1784                                                 inp->inp_vflag &= ~INP_IPV4;
1785                                         else
1786                                                 inp->inp_vflag |= INP_IPV4;
1787                                         break;
1788                                 case IPV6_RECVTCLASS:
1789                                         /* cannot mix with RFC2292 XXX */
1790                                         OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1791                                         break;
1792                                 case IPV6_AUTOFLOWLABEL:
1793                                         OPTSET(IN6P_AUTOFLOWLABEL);
1794                                         break;
1795
1796                                 case IPV6_ORIGDSTADDR:
1797                                         OPTSET2(INP_ORIGDSTADDR, optval);
1798                                         break;
1799                                 case IPV6_BINDANY:
1800                                         OPTSET(INP_BINDANY);
1801                                         break;
1802
1803                                 case IPV6_BINDMULTI:
1804                                         OPTSET2(INP_BINDMULTI, optval);
1805                                         break;
1806 #ifdef  RSS
1807                                 case IPV6_RSS_LISTEN_BUCKET:
1808                                         if ((optval >= 0) &&
1809                                             (optval < rss_getnumbuckets())) {
1810                                                 INP_WLOCK(inp);
1811                                                 inp->inp_rss_listen_bucket = optval;
1812                                                 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1813                                                 INP_WUNLOCK(inp);
1814                                         } else {
1815                                                 error = EINVAL;
1816                                         }
1817                                         break;
1818 #endif
1819                                 }
1820                                 break;
1821
1822                         case IPV6_TCLASS:
1823                         case IPV6_DONTFRAG:
1824                         case IPV6_USE_MIN_MTU:
1825                         case IPV6_PREFER_TEMPADDR:
1826                                 if (optlen != sizeof(optval)) {
1827                                         error = EINVAL;
1828                                         break;
1829                                 }
1830                                 error = sooptcopyin(sopt, &optval,
1831                                         sizeof optval, sizeof optval);
1832                                 if (error)
1833                                         break;
1834                                 {
1835                                         struct ip6_pktopts **optp;
1836                                         INP_WLOCK(inp);
1837                                         if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1838                                                 INP_WUNLOCK(inp);
1839                                                 return (ECONNRESET);
1840                                         }
1841                                         optp = &inp->in6p_outputopts;
1842                                         error = ip6_pcbopt(optname,
1843                                             (u_char *)&optval, sizeof(optval),
1844                                             optp, (td != NULL) ? td->td_ucred :
1845                                             NULL, uproto);
1846                                         INP_WUNLOCK(inp);
1847                                         break;
1848                                 }
1849
1850                         case IPV6_2292PKTINFO:
1851                         case IPV6_2292HOPLIMIT:
1852                         case IPV6_2292HOPOPTS:
1853                         case IPV6_2292DSTOPTS:
1854                         case IPV6_2292RTHDR:
1855                                 /* RFC 2292 */
1856                                 if (optlen != sizeof(int)) {
1857                                         error = EINVAL;
1858                                         break;
1859                                 }
1860                                 error = sooptcopyin(sopt, &optval,
1861                                         sizeof optval, sizeof optval);
1862                                 if (error)
1863                                         break;
1864                                 switch (optname) {
1865                                 case IPV6_2292PKTINFO:
1866                                         OPTSET2292(IN6P_PKTINFO);
1867                                         break;
1868                                 case IPV6_2292HOPLIMIT:
1869                                         OPTSET2292(IN6P_HOPLIMIT);
1870                                         break;
1871                                 case IPV6_2292HOPOPTS:
1872                                         /*
1873                                          * Check super-user privilege.
1874                                          * See comments for IPV6_RECVHOPOPTS.
1875                                          */
1876                                         if (td != NULL) {
1877                                                 error = priv_check(td,
1878                                                     PRIV_NETINET_SETHDROPTS);
1879                                                 if (error)
1880                                                         return (error);
1881                                         }
1882                                         OPTSET2292(IN6P_HOPOPTS);
1883                                         break;
1884                                 case IPV6_2292DSTOPTS:
1885                                         if (td != NULL) {
1886                                                 error = priv_check(td,
1887                                                     PRIV_NETINET_SETHDROPTS);
1888                                                 if (error)
1889                                                         return (error);
1890                                         }
1891                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1892                                         break;
1893                                 case IPV6_2292RTHDR:
1894                                         OPTSET2292(IN6P_RTHDR);
1895                                         break;
1896                                 }
1897                                 break;
1898                         case IPV6_PKTINFO:
1899                         case IPV6_HOPOPTS:
1900                         case IPV6_RTHDR:
1901                         case IPV6_DSTOPTS:
1902                         case IPV6_RTHDRDSTOPTS:
1903                         case IPV6_NEXTHOP:
1904                         {
1905                                 /* new advanced API (RFC3542) */
1906                                 u_char *optbuf;
1907                                 u_char optbuf_storage[MCLBYTES];
1908                                 int optlen;
1909                                 struct ip6_pktopts **optp;
1910
1911                                 /* cannot mix with RFC2292 */
1912                                 if (OPTBIT(IN6P_RFC2292)) {
1913                                         error = EINVAL;
1914                                         break;
1915                                 }
1916
1917                                 /*
1918                                  * We only ensure valsize is not too large
1919                                  * here.  Further validation will be done
1920                                  * later.
1921                                  */
1922                                 error = sooptcopyin(sopt, optbuf_storage,
1923                                     sizeof(optbuf_storage), 0);
1924                                 if (error)
1925                                         break;
1926                                 optlen = sopt->sopt_valsize;
1927                                 optbuf = optbuf_storage;
1928                                 INP_WLOCK(inp);
1929                                 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1930                                         INP_WUNLOCK(inp);
1931                                         return (ECONNRESET);
1932                                 }
1933                                 optp = &inp->in6p_outputopts;
1934                                 error = ip6_pcbopt(optname, optbuf, optlen,
1935                                     optp, (td != NULL) ? td->td_ucred : NULL,
1936                                     uproto);
1937                                 INP_WUNLOCK(inp);
1938                                 break;
1939                         }
1940 #undef OPTSET
1941
1942                         case IPV6_MULTICAST_IF:
1943                         case IPV6_MULTICAST_HOPS:
1944                         case IPV6_MULTICAST_LOOP:
1945                         case IPV6_JOIN_GROUP:
1946                         case IPV6_LEAVE_GROUP:
1947                         case IPV6_MSFILTER:
1948                         case MCAST_BLOCK_SOURCE:
1949                         case MCAST_UNBLOCK_SOURCE:
1950                         case MCAST_JOIN_GROUP:
1951                         case MCAST_LEAVE_GROUP:
1952                         case MCAST_JOIN_SOURCE_GROUP:
1953                         case MCAST_LEAVE_SOURCE_GROUP:
1954                                 error = ip6_setmoptions(inp, sopt);
1955                                 break;
1956
1957                         case IPV6_PORTRANGE:
1958                                 error = sooptcopyin(sopt, &optval,
1959                                     sizeof optval, sizeof optval);
1960                                 if (error)
1961                                         break;
1962
1963                                 INP_WLOCK(inp);
1964                                 switch (optval) {
1965                                 case IPV6_PORTRANGE_DEFAULT:
1966                                         inp->inp_flags &= ~(INP_LOWPORT);
1967                                         inp->inp_flags &= ~(INP_HIGHPORT);
1968                                         break;
1969
1970                                 case IPV6_PORTRANGE_HIGH:
1971                                         inp->inp_flags &= ~(INP_LOWPORT);
1972                                         inp->inp_flags |= INP_HIGHPORT;
1973                                         break;
1974
1975                                 case IPV6_PORTRANGE_LOW:
1976                                         inp->inp_flags &= ~(INP_HIGHPORT);
1977                                         inp->inp_flags |= INP_LOWPORT;
1978                                         break;
1979
1980                                 default:
1981                                         error = EINVAL;
1982                                         break;
1983                                 }
1984                                 INP_WUNLOCK(inp);
1985                                 break;
1986
1987 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1988                         case IPV6_IPSEC_POLICY:
1989                                 if (IPSEC_ENABLED(ipv6)) {
1990                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
1991                                         break;
1992                                 }
1993                                 /* FALLTHROUGH */
1994 #endif /* IPSEC */
1995
1996                         default:
1997                                 error = ENOPROTOOPT;
1998                                 break;
1999                         }
2000                         break;
2001
2002                 case SOPT_GET:
2003                         switch (optname) {
2004
2005                         case IPV6_2292PKTOPTIONS:
2006 #ifdef IPV6_PKTOPTIONS
2007                         case IPV6_PKTOPTIONS:
2008 #endif
2009                                 /*
2010                                  * RFC3542 (effectively) deprecated the
2011                                  * semantics of the 2292-style pktoptions.
2012                                  * Since it was not reliable in nature (i.e.,
2013                                  * applications had to expect the lack of some
2014                                  * information after all), it would make sense
2015                                  * to simplify this part by always returning
2016                                  * empty data.
2017                                  */
2018                                 sopt->sopt_valsize = 0;
2019                                 break;
2020
2021                         case IPV6_RECVHOPOPTS:
2022                         case IPV6_RECVDSTOPTS:
2023                         case IPV6_RECVRTHDRDSTOPTS:
2024                         case IPV6_UNICAST_HOPS:
2025                         case IPV6_RECVPKTINFO:
2026                         case IPV6_RECVHOPLIMIT:
2027                         case IPV6_RECVRTHDR:
2028                         case IPV6_RECVPATHMTU:
2029
2030                         case IPV6_V6ONLY:
2031                         case IPV6_PORTRANGE:
2032                         case IPV6_RECVTCLASS:
2033                         case IPV6_AUTOFLOWLABEL:
2034                         case IPV6_BINDANY:
2035                         case IPV6_FLOWID:
2036                         case IPV6_FLOWTYPE:
2037                         case IPV6_RECVFLOWID:
2038 #ifdef  RSS
2039                         case IPV6_RSSBUCKETID:
2040                         case IPV6_RECVRSSBUCKETID:
2041 #endif
2042                         case IPV6_BINDMULTI:
2043                                 switch (optname) {
2044
2045                                 case IPV6_RECVHOPOPTS:
2046                                         optval = OPTBIT(IN6P_HOPOPTS);
2047                                         break;
2048
2049                                 case IPV6_RECVDSTOPTS:
2050                                         optval = OPTBIT(IN6P_DSTOPTS);
2051                                         break;
2052
2053                                 case IPV6_RECVRTHDRDSTOPTS:
2054                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2055                                         break;
2056
2057                                 case IPV6_UNICAST_HOPS:
2058                                         optval = inp->in6p_hops;
2059                                         break;
2060
2061                                 case IPV6_RECVPKTINFO:
2062                                         optval = OPTBIT(IN6P_PKTINFO);
2063                                         break;
2064
2065                                 case IPV6_RECVHOPLIMIT:
2066                                         optval = OPTBIT(IN6P_HOPLIMIT);
2067                                         break;
2068
2069                                 case IPV6_RECVRTHDR:
2070                                         optval = OPTBIT(IN6P_RTHDR);
2071                                         break;
2072
2073                                 case IPV6_RECVPATHMTU:
2074                                         optval = OPTBIT(IN6P_MTU);
2075                                         break;
2076
2077                                 case IPV6_V6ONLY:
2078                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2079                                         break;
2080
2081                                 case IPV6_PORTRANGE:
2082                                     {
2083                                         int flags;
2084                                         flags = inp->inp_flags;
2085                                         if (flags & INP_HIGHPORT)
2086                                                 optval = IPV6_PORTRANGE_HIGH;
2087                                         else if (flags & INP_LOWPORT)
2088                                                 optval = IPV6_PORTRANGE_LOW;
2089                                         else
2090                                                 optval = 0;
2091                                         break;
2092                                     }
2093                                 case IPV6_RECVTCLASS:
2094                                         optval = OPTBIT(IN6P_TCLASS);
2095                                         break;
2096
2097                                 case IPV6_AUTOFLOWLABEL:
2098                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2099                                         break;
2100
2101                                 case IPV6_ORIGDSTADDR:
2102                                         optval = OPTBIT2(INP_ORIGDSTADDR);
2103                                         break;
2104
2105                                 case IPV6_BINDANY:
2106                                         optval = OPTBIT(INP_BINDANY);
2107                                         break;
2108
2109                                 case IPV6_FLOWID:
2110                                         optval = inp->inp_flowid;
2111                                         break;
2112
2113                                 case IPV6_FLOWTYPE:
2114                                         optval = inp->inp_flowtype;
2115                                         break;
2116
2117                                 case IPV6_RECVFLOWID:
2118                                         optval = OPTBIT2(INP_RECVFLOWID);
2119                                         break;
2120 #ifdef  RSS
2121                                 case IPV6_RSSBUCKETID:
2122                                         retval =
2123                                             rss_hash2bucket(inp->inp_flowid,
2124                                             inp->inp_flowtype,
2125                                             &rss_bucket);
2126                                         if (retval == 0)
2127                                                 optval = rss_bucket;
2128                                         else
2129                                                 error = EINVAL;
2130                                         break;
2131
2132                                 case IPV6_RECVRSSBUCKETID:
2133                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2134                                         break;
2135 #endif
2136
2137                                 case IPV6_BINDMULTI:
2138                                         optval = OPTBIT2(INP_BINDMULTI);
2139                                         break;
2140
2141                                 }
2142                                 if (error)
2143                                         break;
2144                                 error = sooptcopyout(sopt, &optval,
2145                                         sizeof optval);
2146                                 break;
2147
2148                         case IPV6_PATHMTU:
2149                         {
2150                                 u_long pmtu = 0;
2151                                 struct ip6_mtuinfo mtuinfo;
2152                                 struct in6_addr addr;
2153
2154                                 if (!(so->so_state & SS_ISCONNECTED))
2155                                         return (ENOTCONN);
2156                                 /*
2157                                  * XXX: we dot not consider the case of source
2158                                  * routing, or optional information to specify
2159                                  * the outgoing interface.
2160                                  * Copy faddr out of inp to avoid holding lock
2161                                  * on inp during route lookup.
2162                                  */
2163                                 INP_RLOCK(inp);
2164                                 bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2165                                 INP_RUNLOCK(inp);
2166                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2167                                     &addr, &pmtu);
2168                                 if (error)
2169                                         break;
2170                                 if (pmtu > IPV6_MAXPACKET)
2171                                         pmtu = IPV6_MAXPACKET;
2172
2173                                 bzero(&mtuinfo, sizeof(mtuinfo));
2174                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2175                                 optdata = (void *)&mtuinfo;
2176                                 optdatalen = sizeof(mtuinfo);
2177                                 error = sooptcopyout(sopt, optdata,
2178                                     optdatalen);
2179                                 break;
2180                         }
2181
2182                         case IPV6_2292PKTINFO:
2183                         case IPV6_2292HOPLIMIT:
2184                         case IPV6_2292HOPOPTS:
2185                         case IPV6_2292RTHDR:
2186                         case IPV6_2292DSTOPTS:
2187                                 switch (optname) {
2188                                 case IPV6_2292PKTINFO:
2189                                         optval = OPTBIT(IN6P_PKTINFO);
2190                                         break;
2191                                 case IPV6_2292HOPLIMIT:
2192                                         optval = OPTBIT(IN6P_HOPLIMIT);
2193                                         break;
2194                                 case IPV6_2292HOPOPTS:
2195                                         optval = OPTBIT(IN6P_HOPOPTS);
2196                                         break;
2197                                 case IPV6_2292RTHDR:
2198                                         optval = OPTBIT(IN6P_RTHDR);
2199                                         break;
2200                                 case IPV6_2292DSTOPTS:
2201                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2202                                         break;
2203                                 }
2204                                 error = sooptcopyout(sopt, &optval,
2205                                     sizeof optval);
2206                                 break;
2207                         case IPV6_PKTINFO:
2208                         case IPV6_HOPOPTS:
2209                         case IPV6_RTHDR:
2210                         case IPV6_DSTOPTS:
2211                         case IPV6_RTHDRDSTOPTS:
2212                         case IPV6_NEXTHOP:
2213                         case IPV6_TCLASS:
2214                         case IPV6_DONTFRAG:
2215                         case IPV6_USE_MIN_MTU:
2216                         case IPV6_PREFER_TEMPADDR:
2217                                 error = ip6_getpcbopt(inp, optname, sopt);
2218                                 break;
2219
2220                         case IPV6_MULTICAST_IF:
2221                         case IPV6_MULTICAST_HOPS:
2222                         case IPV6_MULTICAST_LOOP:
2223                         case IPV6_MSFILTER:
2224                                 error = ip6_getmoptions(inp, sopt);
2225                                 break;
2226
2227 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2228                         case IPV6_IPSEC_POLICY:
2229                                 if (IPSEC_ENABLED(ipv6)) {
2230                                         error = IPSEC_PCBCTL(ipv6, inp, sopt);
2231                                         break;
2232                                 }
2233                                 /* FALLTHROUGH */
2234 #endif /* IPSEC */
2235                         default:
2236                                 error = ENOPROTOOPT;
2237                                 break;
2238                         }
2239                         break;
2240                 }
2241         }
2242         return (error);
2243 }
2244
2245 int
2246 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2247 {
2248         int error = 0, optval, optlen;
2249         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2250         struct inpcb *inp = sotoinpcb(so);
2251         int level, op, optname;
2252
2253         level = sopt->sopt_level;
2254         op = sopt->sopt_dir;
2255         optname = sopt->sopt_name;
2256         optlen = sopt->sopt_valsize;
2257
2258         if (level != IPPROTO_IPV6) {
2259                 return (EINVAL);
2260         }
2261
2262         switch (optname) {
2263         case IPV6_CHECKSUM:
2264                 /*
2265                  * For ICMPv6 sockets, no modification allowed for checksum
2266                  * offset, permit "no change" values to help existing apps.
2267                  *
2268                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2269                  * for an ICMPv6 socket will fail."
2270                  * The current behavior does not meet RFC3542.
2271                  */
2272                 switch (op) {
2273                 case SOPT_SET:
2274                         if (optlen != sizeof(int)) {
2275                                 error = EINVAL;
2276                                 break;
2277                         }
2278                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2279                                             sizeof(optval));
2280                         if (error)
2281                                 break;
2282                         if (optval < -1 || (optval % 2) != 0) {
2283                                 /*
2284                                  * The API assumes non-negative even offset
2285                                  * values or -1 as a special value.
2286                                  */
2287                                 error = EINVAL;
2288                         } else if (so->so_proto->pr_protocol ==
2289                             IPPROTO_ICMPV6) {
2290                                 if (optval != icmp6off)
2291                                         error = EINVAL;
2292                         } else
2293                                 inp->in6p_cksum = optval;
2294                         break;
2295
2296                 case SOPT_GET:
2297                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2298                                 optval = icmp6off;
2299                         else
2300                                 optval = inp->in6p_cksum;
2301
2302                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2303                         break;
2304
2305                 default:
2306                         error = EINVAL;
2307                         break;
2308                 }
2309                 break;
2310
2311         default:
2312                 error = ENOPROTOOPT;
2313                 break;
2314         }
2315
2316         return (error);
2317 }
2318
2319 /*
2320  * Set up IP6 options in pcb for insertion in output packets or
2321  * specifying behavior of outgoing packets.
2322  */
2323 static int
2324 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2325     struct socket *so, struct sockopt *sopt)
2326 {
2327         struct ip6_pktopts *opt = *pktopt;
2328         int error = 0;
2329         struct thread *td = sopt->sopt_td;
2330
2331         /* turn off any old options. */
2332         if (opt) {
2333 #ifdef DIAGNOSTIC
2334                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2335                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2336                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2337                         printf("ip6_pcbopts: all specified options are cleared.\n");
2338 #endif
2339                 ip6_clearpktopts(opt, -1);
2340         } else
2341                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2342         *pktopt = NULL;
2343
2344         if (!m || m->m_len == 0) {
2345                 /*
2346                  * Only turning off any previous options, regardless of
2347                  * whether the opt is just created or given.
2348                  */
2349                 free(opt, M_IP6OPT);
2350                 return (0);
2351         }
2352
2353         /*  set options specified by user. */
2354         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2355             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2356                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2357                 free(opt, M_IP6OPT);
2358                 return (error);
2359         }
2360         *pktopt = opt;
2361         return (0);
2362 }
2363
2364 /*
2365  * initialize ip6_pktopts.  beware that there are non-zero default values in
2366  * the struct.
2367  */
2368 void
2369 ip6_initpktopts(struct ip6_pktopts *opt)
2370 {
2371
2372         bzero(opt, sizeof(*opt));
2373         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2374         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2375         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2376         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2377 }
2378
2379 static int
2380 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2381     struct ucred *cred, int uproto)
2382 {
2383         struct ip6_pktopts *opt;
2384
2385         if (*pktopt == NULL) {
2386                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2387                     M_NOWAIT);
2388                 if (*pktopt == NULL)
2389                         return (ENOBUFS);
2390                 ip6_initpktopts(*pktopt);
2391         }
2392         opt = *pktopt;
2393
2394         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2395 }
2396
2397 #define GET_PKTOPT_VAR(field, lenexpr) do {                                     \
2398         if (pktopt && pktopt->field) {                                          \
2399                 INP_RUNLOCK(inp);                                               \
2400                 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);         \
2401                 malloc_optdata = true;                                          \
2402                 INP_RLOCK(inp);                                                 \
2403                 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {            \
2404                         INP_RUNLOCK(inp);                                       \
2405                         free(optdata, M_TEMP);                                  \
2406                         return (ECONNRESET);                                    \
2407                 }                                                               \
2408                 pktopt = inp->in6p_outputopts;                                  \
2409                 if (pktopt && pktopt->field) {                                  \
2410                         optdatalen = min(lenexpr, sopt->sopt_valsize);          \
2411                         bcopy(&pktopt->field, optdata, optdatalen);             \
2412                 } else {                                                        \
2413                         free(optdata, M_TEMP);                                  \
2414                         optdata = NULL;                                         \
2415                         malloc_optdata = false;                                 \
2416                 }                                                               \
2417         }                                                                       \
2418 } while(0)
2419
2420 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,                         \
2421         (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2422
2423 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,                        \
2424         pktopt->field->sa_len)
2425
2426 static int
2427 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2428 {
2429         void *optdata = NULL;
2430         bool malloc_optdata = false;
2431         int optdatalen = 0;
2432         int error = 0;
2433         struct in6_pktinfo null_pktinfo;
2434         int deftclass = 0, on;
2435         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2436         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2437         struct ip6_pktopts *pktopt;
2438
2439         INP_RLOCK(inp);
2440         pktopt = inp->in6p_outputopts;
2441
2442         switch (optname) {
2443         case IPV6_PKTINFO:
2444                 optdata = (void *)&null_pktinfo;
2445                 if (pktopt && pktopt->ip6po_pktinfo) {
2446                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2447                             sizeof(null_pktinfo));
2448                         in6_clearscope(&null_pktinfo.ipi6_addr);
2449                 } else {
2450                         /* XXX: we don't have to do this every time... */
2451                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2452                 }
2453                 optdatalen = sizeof(struct in6_pktinfo);
2454                 break;
2455         case IPV6_TCLASS:
2456                 if (pktopt && pktopt->ip6po_tclass >= 0)
2457                         deftclass = pktopt->ip6po_tclass;
2458                 optdata = (void *)&deftclass;
2459                 optdatalen = sizeof(int);
2460                 break;
2461         case IPV6_HOPOPTS:
2462                 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2463                 break;
2464         case IPV6_RTHDR:
2465                 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2466                 break;
2467         case IPV6_RTHDRDSTOPTS:
2468                 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2469                 break;
2470         case IPV6_DSTOPTS:
2471                 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2472                 break;
2473         case IPV6_NEXTHOP:
2474                 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2475                 break;
2476         case IPV6_USE_MIN_MTU:
2477                 if (pktopt)
2478                         defminmtu = pktopt->ip6po_minmtu;
2479                 optdata = (void *)&defminmtu;
2480                 optdatalen = sizeof(int);
2481                 break;
2482         case IPV6_DONTFRAG:
2483                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2484                         on = 1;
2485                 else
2486                         on = 0;
2487                 optdata = (void *)&on;
2488                 optdatalen = sizeof(on);
2489                 break;
2490         case IPV6_PREFER_TEMPADDR:
2491                 if (pktopt)
2492                         defpreftemp = pktopt->ip6po_prefer_tempaddr;
2493                 optdata = (void *)&defpreftemp;
2494                 optdatalen = sizeof(int);
2495                 break;
2496         default:                /* should not happen */
2497 #ifdef DIAGNOSTIC
2498                 panic("ip6_getpcbopt: unexpected option\n");
2499 #endif
2500                 INP_RUNLOCK(inp);
2501                 return (ENOPROTOOPT);
2502         }
2503         INP_RUNLOCK(inp);
2504
2505         error = sooptcopyout(sopt, optdata, optdatalen);
2506         if (malloc_optdata)
2507                 free(optdata, M_TEMP);
2508
2509         return (error);
2510 }
2511
2512 void
2513 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2514 {
2515         if (pktopt == NULL)
2516                 return;
2517
2518         if (optname == -1 || optname == IPV6_PKTINFO) {
2519                 if (pktopt->ip6po_pktinfo)
2520                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2521                 pktopt->ip6po_pktinfo = NULL;
2522         }
2523         if (optname == -1 || optname == IPV6_HOPLIMIT)
2524                 pktopt->ip6po_hlim = -1;
2525         if (optname == -1 || optname == IPV6_TCLASS)
2526                 pktopt->ip6po_tclass = -1;
2527         if (optname == -1 || optname == IPV6_NEXTHOP) {
2528                 if (pktopt->ip6po_nextroute.ro_rt) {
2529                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2530                         pktopt->ip6po_nextroute.ro_rt = NULL;
2531                 }
2532                 if (pktopt->ip6po_nexthop)
2533                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2534                 pktopt->ip6po_nexthop = NULL;
2535         }
2536         if (optname == -1 || optname == IPV6_HOPOPTS) {
2537                 if (pktopt->ip6po_hbh)
2538                         free(pktopt->ip6po_hbh, M_IP6OPT);
2539                 pktopt->ip6po_hbh = NULL;
2540         }
2541         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2542                 if (pktopt->ip6po_dest1)
2543                         free(pktopt->ip6po_dest1, M_IP6OPT);
2544                 pktopt->ip6po_dest1 = NULL;
2545         }
2546         if (optname == -1 || optname == IPV6_RTHDR) {
2547                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2548                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2549                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2550                 if (pktopt->ip6po_route.ro_rt) {
2551                         RTFREE(pktopt->ip6po_route.ro_rt);
2552                         pktopt->ip6po_route.ro_rt = NULL;
2553                 }
2554         }
2555         if (optname == -1 || optname == IPV6_DSTOPTS) {
2556                 if (pktopt->ip6po_dest2)
2557                         free(pktopt->ip6po_dest2, M_IP6OPT);
2558                 pktopt->ip6po_dest2 = NULL;
2559         }
2560 }
2561
2562 #define PKTOPT_EXTHDRCPY(type) \
2563 do {\
2564         if (src->type) {\
2565                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2566                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2567                 if (dst->type == NULL)\
2568                         goto bad;\
2569                 bcopy(src->type, dst->type, hlen);\
2570         }\
2571 } while (/*CONSTCOND*/ 0)
2572
2573 static int
2574 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2575 {
2576         if (dst == NULL || src == NULL)  {
2577                 printf("ip6_clearpktopts: invalid argument\n");
2578                 return (EINVAL);
2579         }
2580
2581         dst->ip6po_hlim = src->ip6po_hlim;
2582         dst->ip6po_tclass = src->ip6po_tclass;
2583         dst->ip6po_flags = src->ip6po_flags;
2584         dst->ip6po_minmtu = src->ip6po_minmtu;
2585         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2586         if (src->ip6po_pktinfo) {
2587                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2588                     M_IP6OPT, canwait);
2589                 if (dst->ip6po_pktinfo == NULL)
2590                         goto bad;
2591                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2592         }
2593         if (src->ip6po_nexthop) {
2594                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2595                     M_IP6OPT, canwait);
2596                 if (dst->ip6po_nexthop == NULL)
2597                         goto bad;
2598                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2599                     src->ip6po_nexthop->sa_len);
2600         }
2601         PKTOPT_EXTHDRCPY(ip6po_hbh);
2602         PKTOPT_EXTHDRCPY(ip6po_dest1);
2603         PKTOPT_EXTHDRCPY(ip6po_dest2);
2604         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2605         return (0);
2606
2607   bad:
2608         ip6_clearpktopts(dst, -1);
2609         return (ENOBUFS);
2610 }
2611 #undef PKTOPT_EXTHDRCPY
2612
2613 struct ip6_pktopts *
2614 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2615 {
2616         int error;
2617         struct ip6_pktopts *dst;
2618
2619         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2620         if (dst == NULL)
2621                 return (NULL);
2622         ip6_initpktopts(dst);
2623
2624         if ((error = copypktopts(dst, src, canwait)) != 0) {
2625                 free(dst, M_IP6OPT);
2626                 return (NULL);
2627         }
2628
2629         return (dst);
2630 }
2631
2632 void
2633 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2634 {
2635         if (pktopt == NULL)
2636                 return;
2637
2638         ip6_clearpktopts(pktopt, -1);
2639
2640         free(pktopt, M_IP6OPT);
2641 }
2642
2643 /*
2644  * Set IPv6 outgoing packet options based on advanced API.
2645  */
2646 int
2647 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2648     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2649 {
2650         struct cmsghdr *cm = NULL;
2651
2652         if (control == NULL || opt == NULL)
2653                 return (EINVAL);
2654
2655         ip6_initpktopts(opt);
2656         if (stickyopt) {
2657                 int error;
2658
2659                 /*
2660                  * If stickyopt is provided, make a local copy of the options
2661                  * for this particular packet, then override them by ancillary
2662                  * objects.
2663                  * XXX: copypktopts() does not copy the cached route to a next
2664                  * hop (if any).  This is not very good in terms of efficiency,
2665                  * but we can allow this since this option should be rarely
2666                  * used.
2667                  */
2668                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2669                         return (error);
2670         }
2671
2672         /*
2673          * XXX: Currently, we assume all the optional information is stored
2674          * in a single mbuf.
2675          */
2676         if (control->m_next)
2677                 return (EINVAL);
2678
2679         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2680             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2681                 int error;
2682
2683                 if (control->m_len < CMSG_LEN(0))
2684                         return (EINVAL);
2685
2686                 cm = mtod(control, struct cmsghdr *);
2687                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2688                         return (EINVAL);
2689                 if (cm->cmsg_level != IPPROTO_IPV6)
2690                         continue;
2691
2692                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2693                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2694                 if (error)
2695                         return (error);
2696         }
2697
2698         return (0);
2699 }
2700
2701 /*
2702  * Set a particular packet option, as a sticky option or an ancillary data
2703  * item.  "len" can be 0 only when it's a sticky option.
2704  * We have 4 cases of combination of "sticky" and "cmsg":
2705  * "sticky=0, cmsg=0": impossible
2706  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2707  * "sticky=1, cmsg=0": RFC3542 socket option
2708  * "sticky=1, cmsg=1": RFC2292 socket option
2709  */
2710 static int
2711 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2712     struct ucred *cred, int sticky, int cmsg, int uproto)
2713 {
2714         int minmtupolicy, preftemp;
2715         int error;
2716
2717         if (!sticky && !cmsg) {
2718 #ifdef DIAGNOSTIC
2719                 printf("ip6_setpktopt: impossible case\n");
2720 #endif
2721                 return (EINVAL);
2722         }
2723
2724         /*
2725          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2726          * not be specified in the context of RFC3542.  Conversely,
2727          * RFC3542 types should not be specified in the context of RFC2292.
2728          */
2729         if (!cmsg) {
2730                 switch (optname) {
2731                 case IPV6_2292PKTINFO:
2732                 case IPV6_2292HOPLIMIT:
2733                 case IPV6_2292NEXTHOP:
2734                 case IPV6_2292HOPOPTS:
2735                 case IPV6_2292DSTOPTS:
2736                 case IPV6_2292RTHDR:
2737                 case IPV6_2292PKTOPTIONS:
2738                         return (ENOPROTOOPT);
2739                 }
2740         }
2741         if (sticky && cmsg) {
2742                 switch (optname) {
2743                 case IPV6_PKTINFO:
2744                 case IPV6_HOPLIMIT:
2745                 case IPV6_NEXTHOP:
2746                 case IPV6_HOPOPTS:
2747                 case IPV6_DSTOPTS:
2748                 case IPV6_RTHDRDSTOPTS:
2749                 case IPV6_RTHDR:
2750                 case IPV6_USE_MIN_MTU:
2751                 case IPV6_DONTFRAG:
2752                 case IPV6_TCLASS:
2753                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2754                         return (ENOPROTOOPT);
2755                 }
2756         }
2757
2758         switch (optname) {
2759         case IPV6_2292PKTINFO:
2760         case IPV6_PKTINFO:
2761         {
2762                 struct ifnet *ifp = NULL;
2763                 struct in6_pktinfo *pktinfo;
2764
2765                 if (len != sizeof(struct in6_pktinfo))
2766                         return (EINVAL);
2767
2768                 pktinfo = (struct in6_pktinfo *)buf;
2769
2770                 /*
2771                  * An application can clear any sticky IPV6_PKTINFO option by
2772                  * doing a "regular" setsockopt with ipi6_addr being
2773                  * in6addr_any and ipi6_ifindex being zero.
2774                  * [RFC 3542, Section 6]
2775                  */
2776                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2777                     pktinfo->ipi6_ifindex == 0 &&
2778                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2779                         ip6_clearpktopts(opt, optname);
2780                         break;
2781                 }
2782
2783                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2784                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2785                         return (EINVAL);
2786                 }
2787                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2788                         return (EINVAL);
2789                 /* validate the interface index if specified. */
2790                 if (pktinfo->ipi6_ifindex > V_if_index)
2791                          return (ENXIO);
2792                 if (pktinfo->ipi6_ifindex) {
2793                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2794                         if (ifp == NULL)
2795                                 return (ENXIO);
2796                 }
2797                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2798                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2799                         return (ENETDOWN);
2800
2801                 if (ifp != NULL &&
2802                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2803                         struct in6_ifaddr *ia;
2804
2805                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2806                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2807                         if (ia == NULL)
2808                                 return (EADDRNOTAVAIL);
2809                         ifa_free(&ia->ia_ifa);
2810                 }
2811                 /*
2812                  * We store the address anyway, and let in6_selectsrc()
2813                  * validate the specified address.  This is because ipi6_addr
2814                  * may not have enough information about its scope zone, and
2815                  * we may need additional information (such as outgoing
2816                  * interface or the scope zone of a destination address) to
2817                  * disambiguate the scope.
2818                  * XXX: the delay of the validation may confuse the
2819                  * application when it is used as a sticky option.
2820                  */
2821                 if (opt->ip6po_pktinfo == NULL) {
2822                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2823                             M_IP6OPT, M_NOWAIT);
2824                         if (opt->ip6po_pktinfo == NULL)
2825                                 return (ENOBUFS);
2826                 }
2827                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2828                 break;
2829         }
2830
2831         case IPV6_2292HOPLIMIT:
2832         case IPV6_HOPLIMIT:
2833         {
2834                 int *hlimp;
2835
2836                 /*
2837                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2838                  * to simplify the ordering among hoplimit options.
2839                  */
2840                 if (optname == IPV6_HOPLIMIT && sticky)
2841                         return (ENOPROTOOPT);
2842
2843                 if (len != sizeof(int))
2844                         return (EINVAL);
2845                 hlimp = (int *)buf;
2846                 if (*hlimp < -1 || *hlimp > 255)
2847                         return (EINVAL);
2848
2849                 opt->ip6po_hlim = *hlimp;
2850                 break;
2851         }
2852
2853         case IPV6_TCLASS:
2854         {
2855                 int tclass;
2856
2857                 if (len != sizeof(int))
2858                         return (EINVAL);
2859                 tclass = *(int *)buf;
2860                 if (tclass < -1 || tclass > 255)
2861                         return (EINVAL);
2862
2863                 opt->ip6po_tclass = tclass;
2864                 break;
2865         }
2866
2867         case IPV6_2292NEXTHOP:
2868         case IPV6_NEXTHOP:
2869                 if (cred != NULL) {
2870                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2871                         if (error)
2872                                 return (error);
2873                 }
2874
2875                 if (len == 0) { /* just remove the option */
2876                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2877                         break;
2878                 }
2879
2880                 /* check if cmsg_len is large enough for sa_len */
2881                 if (len < sizeof(struct sockaddr) || len < *buf)
2882                         return (EINVAL);
2883
2884                 switch (((struct sockaddr *)buf)->sa_family) {
2885                 case AF_INET6:
2886                 {
2887                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2888                         int error;
2889
2890                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2891                                 return (EINVAL);
2892
2893                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2894                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2895                                 return (EINVAL);
2896                         }
2897                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2898                             != 0) {
2899                                 return (error);
2900                         }
2901                         break;
2902                 }
2903                 case AF_LINK:   /* should eventually be supported */
2904                 default:
2905                         return (EAFNOSUPPORT);
2906                 }
2907
2908                 /* turn off the previous option, then set the new option. */
2909                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2910                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2911                 if (opt->ip6po_nexthop == NULL)
2912                         return (ENOBUFS);
2913                 bcopy(buf, opt->ip6po_nexthop, *buf);
2914                 break;
2915
2916         case IPV6_2292HOPOPTS:
2917         case IPV6_HOPOPTS:
2918         {
2919                 struct ip6_hbh *hbh;
2920                 int hbhlen;
2921
2922                 /*
2923                  * XXX: We don't allow a non-privileged user to set ANY HbH
2924                  * options, since per-option restriction has too much
2925                  * overhead.
2926                  */
2927                 if (cred != NULL) {
2928                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2929                         if (error)
2930                                 return (error);
2931                 }
2932
2933                 if (len == 0) {
2934                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2935                         break;  /* just remove the option */
2936                 }
2937
2938                 /* message length validation */
2939                 if (len < sizeof(struct ip6_hbh))
2940                         return (EINVAL);
2941                 hbh = (struct ip6_hbh *)buf;
2942                 hbhlen = (hbh->ip6h_len + 1) << 3;
2943                 if (len != hbhlen)
2944                         return (EINVAL);
2945
2946                 /* turn off the previous option, then set the new option. */
2947                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2948                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2949                 if (opt->ip6po_hbh == NULL)
2950                         return (ENOBUFS);
2951                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2952
2953                 break;
2954         }
2955
2956         case IPV6_2292DSTOPTS:
2957         case IPV6_DSTOPTS:
2958         case IPV6_RTHDRDSTOPTS:
2959         {
2960                 struct ip6_dest *dest, **newdest = NULL;
2961                 int destlen;
2962
2963                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2964                         error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2965                         if (error)
2966                                 return (error);
2967                 }
2968
2969                 if (len == 0) {
2970                         ip6_clearpktopts(opt, optname);
2971                         break;  /* just remove the option */
2972                 }
2973
2974                 /* message length validation */
2975                 if (len < sizeof(struct ip6_dest))
2976                         return (EINVAL);
2977                 dest = (struct ip6_dest *)buf;
2978                 destlen = (dest->ip6d_len + 1) << 3;
2979                 if (len != destlen)
2980                         return (EINVAL);
2981
2982                 /*
2983                  * Determine the position that the destination options header
2984                  * should be inserted; before or after the routing header.
2985                  */
2986                 switch (optname) {
2987                 case IPV6_2292DSTOPTS:
2988                         /*
2989                          * The old advacned API is ambiguous on this point.
2990                          * Our approach is to determine the position based
2991                          * according to the existence of a routing header.
2992                          * Note, however, that this depends on the order of the
2993                          * extension headers in the ancillary data; the 1st
2994                          * part of the destination options header must appear
2995                          * before the routing header in the ancillary data,
2996                          * too.
2997                          * RFC3542 solved the ambiguity by introducing
2998                          * separate ancillary data or option types.
2999                          */
3000                         if (opt->ip6po_rthdr == NULL)
3001                                 newdest = &opt->ip6po_dest1;
3002                         else
3003                                 newdest = &opt->ip6po_dest2;
3004                         break;
3005                 case IPV6_RTHDRDSTOPTS:
3006                         newdest = &opt->ip6po_dest1;
3007                         break;
3008                 case IPV6_DSTOPTS:
3009                         newdest = &opt->ip6po_dest2;
3010                         break;
3011                 }
3012
3013                 /* turn off the previous option, then set the new option. */
3014                 ip6_clearpktopts(opt, optname);
3015                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3016                 if (*newdest == NULL)
3017                         return (ENOBUFS);
3018                 bcopy(dest, *newdest, destlen);
3019
3020                 break;
3021         }
3022
3023         case IPV6_2292RTHDR:
3024         case IPV6_RTHDR:
3025         {
3026                 struct ip6_rthdr *rth;
3027                 int rthlen;
3028
3029                 if (len == 0) {
3030                         ip6_clearpktopts(opt, IPV6_RTHDR);
3031                         break;  /* just remove the option */
3032                 }
3033
3034                 /* message length validation */
3035                 if (len < sizeof(struct ip6_rthdr))
3036                         return (EINVAL);
3037                 rth = (struct ip6_rthdr *)buf;
3038                 rthlen = (rth->ip6r_len + 1) << 3;
3039                 if (len != rthlen)
3040                         return (EINVAL);
3041
3042                 switch (rth->ip6r_type) {
3043                 case IPV6_RTHDR_TYPE_0:
3044                         if (rth->ip6r_len == 0) /* must contain one addr */
3045                                 return (EINVAL);
3046                         if (rth->ip6r_len % 2) /* length must be even */
3047                                 return (EINVAL);
3048                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3049                                 return (EINVAL);
3050                         break;
3051                 default:
3052                         return (EINVAL);        /* not supported */
3053                 }
3054
3055                 /* turn off the previous option */
3056                 ip6_clearpktopts(opt, IPV6_RTHDR);
3057                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3058                 if (opt->ip6po_rthdr == NULL)
3059                         return (ENOBUFS);
3060                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3061
3062                 break;
3063         }
3064
3065         case IPV6_USE_MIN_MTU:
3066                 if (len != sizeof(int))
3067                         return (EINVAL);
3068                 minmtupolicy = *(int *)buf;
3069                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3070                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3071                     minmtupolicy != IP6PO_MINMTU_ALL) {
3072                         return (EINVAL);
3073                 }
3074                 opt->ip6po_minmtu = minmtupolicy;
3075                 break;
3076
3077         case IPV6_DONTFRAG:
3078                 if (len != sizeof(int))
3079                         return (EINVAL);
3080
3081                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3082                         /*
3083                          * we ignore this option for TCP sockets.
3084                          * (RFC3542 leaves this case unspecified.)
3085                          */
3086                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3087                 } else
3088                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3089                 break;
3090
3091         case IPV6_PREFER_TEMPADDR:
3092                 if (len != sizeof(int))
3093                         return (EINVAL);
3094                 preftemp = *(int *)buf;
3095                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3096                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3097                     preftemp != IP6PO_TEMPADDR_PREFER) {
3098                         return (EINVAL);
3099                 }
3100                 opt->ip6po_prefer_tempaddr = preftemp;
3101                 break;
3102
3103         default:
3104                 return (ENOPROTOOPT);
3105         } /* end of switch */
3106
3107         return (0);
3108 }
3109
3110 /*
3111  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3112  * packet to the input queue of a specified interface.  Note that this
3113  * calls the output routine of the loopback "driver", but with an interface
3114  * pointer that might NOT be &loif -- easier than replicating that code here.
3115  */
3116 void
3117 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3118 {
3119         struct mbuf *copym;
3120         struct ip6_hdr *ip6;
3121
3122         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3123         if (copym == NULL)
3124                 return;
3125
3126         /*
3127          * Make sure to deep-copy IPv6 header portion in case the data
3128          * is in an mbuf cluster, so that we can safely override the IPv6
3129          * header portion later.
3130          */
3131         if (!M_WRITABLE(copym) ||
3132             copym->m_len < sizeof(struct ip6_hdr)) {
3133                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3134                 if (copym == NULL)
3135                         return;
3136         }
3137         ip6 = mtod(copym, struct ip6_hdr *);
3138         /*
3139          * clear embedded scope identifiers if necessary.
3140          * in6_clearscope will touch the addresses only when necessary.
3141          */
3142         in6_clearscope(&ip6->ip6_src);
3143         in6_clearscope(&ip6->ip6_dst);
3144         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3145                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3146                     CSUM_PSEUDO_HDR;
3147                 copym->m_pkthdr.csum_data = 0xffff;
3148         }
3149         if_simloop(ifp, copym, AF_INET6, 0);
3150 }
3151
3152 /*
3153  * Chop IPv6 header off from the payload.
3154  */
3155 static int
3156 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3157 {
3158         struct mbuf *mh;
3159         struct ip6_hdr *ip6;
3160
3161         ip6 = mtod(m, struct ip6_hdr *);
3162         if (m->m_len > sizeof(*ip6)) {
3163                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3164                 if (mh == NULL) {
3165                         m_freem(m);
3166                         return ENOBUFS;
3167                 }
3168                 m_move_pkthdr(mh, m);
3169                 M_ALIGN(mh, sizeof(*ip6));
3170                 m->m_len -= sizeof(*ip6);
3171                 m->m_data += sizeof(*ip6);
3172                 mh->m_next = m;
3173                 m = mh;
3174                 m->m_len = sizeof(*ip6);
3175                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3176         }
3177         exthdrs->ip6e_ip6 = m;
3178         return 0;
3179 }
3180
3181 /*
3182  * Compute IPv6 extension header length.
3183  */
3184 int
3185 ip6_optlen(struct inpcb *inp)
3186 {
3187         int len;
3188
3189         if (!inp->in6p_outputopts)
3190                 return 0;
3191
3192         len = 0;
3193 #define elen(x) \
3194     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3195
3196         len += elen(inp->in6p_outputopts->ip6po_hbh);
3197         if (inp->in6p_outputopts->ip6po_rthdr)
3198                 /* dest1 is valid with rthdr only */
3199                 len += elen(inp->in6p_outputopts->ip6po_dest1);
3200         len += elen(inp->in6p_outputopts->ip6po_rthdr);
3201         len += elen(inp->in6p_outputopts->ip6po_dest2);
3202         return len;
3203 #undef elen
3204 }