sys/netinet6/ip6_output.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-3-Clause
   3  *
   4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 3. Neither the name of the project nor the names of its contributors
  16  *    may be used to endorse or promote products derived from this software
  17  *    without specific prior written permission.
  18  *
  19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  29  * SUCH DAMAGE.
  30  *
  31  *      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
  32  */
  33
  34 /*-
  35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. Neither the name of the University nor the names of its contributors
  47  *    may be used to endorse or promote products derived from this software
  48  *    without specific prior written permission.
  49  *
  50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  60  * SUCH DAMAGE.
  61  *
  62  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  63  */
  64
  65 #include <sys/cdefs.h>
  66 __FBSDID("$FreeBSD$");
  67
  68 #include "opt_inet.h"
  69 #include "opt_inet6.h"
  70 #include "opt_ratelimit.h"
  71 #include "opt_ipsec.h"
  72 #include "opt_sctp.h"
  73 #include "opt_route.h"
  74 #include "opt_rss.h"
  75
  76 #include <sys/param.h>
  77 #include <sys/kernel.h>
  78 #include <sys/malloc.h>
  79 #include <sys/mbuf.h>
  80 #include <sys/errno.h>
  81 #include <sys/priv.h>
  82 #include <sys/proc.h>
  83 #include <sys/protosw.h>
  84 #include <sys/socket.h>
  85 #include <sys/socketvar.h>
  86 #include <sys/syslog.h>
  87 #include <sys/ucred.h>
  88
  89 #include <machine/in_cksum.h>
  90
  91 #include <net/if.h>
  92 #include <net/if_var.h>
  93 #include <net/if_llatbl.h>
  94 #include <net/netisr.h>
  95 #include <net/route.h>
  96 #include <net/pfil.h>
  97 #include <net/rss_config.h>
  98 #include <net/vnet.h>
  99
 100 #include <netinet/in.h>
 101 #include <netinet/in_var.h>
 102 #include <netinet/ip_var.h>
 103 #include <netinet6/in6_fib.h>
 104 #include <netinet6/in6_var.h>
 105 #include <netinet/ip6.h>
 106 #include <netinet/icmp6.h>
 107 #include <netinet6/ip6_var.h>
 108 #include <netinet/in_pcb.h>
 109 #include <netinet/tcp_var.h>
 110 #include <netinet6/nd6.h>
 111 #include <netinet6/in6_rss.h>
 112
 113 #include <netipsec/ipsec_support.h>
 114 #ifdef SCTP
 115 #include <netinet/sctp.h>
 116 #include <netinet/sctp_crc32.h>
 117 #endif
 118
 119 #include <netinet6/ip6protosw.h>
 120 #include <netinet6/scope6_var.h>
 121
 122 extern int in6_mcast_loop;
 123
 124 struct ip6_exthdrs {
 125         struct mbuf *ip6e_ip6;
 126         struct mbuf *ip6e_hbh;
 127         struct mbuf *ip6e_dest1;
 128         struct mbuf *ip6e_rthdr;
 129         struct mbuf *ip6e_dest2;
 130 };
 131
 132 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
 133
 134 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
 135                            struct ucred *, int);
 136 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
 137         struct socket *, struct sockopt *);
 138 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
 139 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
 140         struct ucred *, int, int, int);
 141
 142 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 143 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 144         struct ip6_frag **);
 145 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 146 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 147 static int ip6_getpmtu(struct route_in6 *, int,
 148         struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
 149         u_int);
 150 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
 151         u_long *, int *, u_int);
 152 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
 153 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 154
 155
 156 /*
 157  * Make an extension header from option data.  hp is the source, and
 158  * mp is the destination.
 159  */
 160 #define MAKE_EXTHDR(hp, mp)                                             \
 161     do {                                                                \
 162         if (hp) {                                                       \
 163                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 164                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 165                     ((eh)->ip6e_len + 1) << 3);                         \
 166                 if (error)                                              \
 167                         goto freehdrs;                                  \
 168         }                                                               \
 169     } while (/*CONSTCOND*/ 0)
 170
 171 /*
 172  * Form a chain of extension headers.
 173  * m is the extension header mbuf
 174  * mp is the previous mbuf in the chain
 175  * p is the next header
 176  * i is the type of option.
 177  */
 178 #define MAKE_CHAIN(m, mp, p, i)\
 179     do {\
 180         if (m) {\
 181                 if (!hdrsplit) \
 182                         panic("assumption failed: hdr not split"); \
 183                 *mtod((m), u_char *) = *(p);\
 184                 *(p) = (i);\
 185                 p = mtod((m), u_char *);\
 186                 (m)->m_next = (mp)->m_next;\
 187                 (mp)->m_next = (m);\
 188                 (mp) = (m);\
 189         }\
 190     } while (/*CONSTCOND*/ 0)
 191
 192 void
 193 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
 194 {
 195         u_short csum;
 196
 197         csum = in_cksum_skip(m, offset + plen, offset);
 198         if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
 199                 csum = 0xffff;
 200         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 201
 202         if (offset + sizeof(csum) > m->m_len)
 203                 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
 204         else
 205                 *(u_short *)mtodo(m, offset) = csum;
 206 }
 207
 208 int
 209 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
 210     int fraglen , uint32_t id)
 211 {
 212         struct mbuf *m, **mnext, *m_frgpart;
 213         struct ip6_hdr *ip6, *mhip6;
 214         struct ip6_frag *ip6f;
 215         int off;
 216         int error;
 217         int tlen = m0->m_pkthdr.len;
 218
 219         KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
 220
 221         m = m0;
 222         ip6 = mtod(m, struct ip6_hdr *);
 223         mnext = &m->m_nextpkt;
 224
 225         for (off = hlen; off < tlen; off += fraglen) {
 226                 m = m_gethdr(M_NOWAIT, MT_DATA);
 227                 if (!m) {
 228                         IP6STAT_INC(ip6s_odropped);
 229                         return (ENOBUFS);
 230                 }
 231
 232                 /*
 233                  * Make sure the complete packet header gets copied
 234                  * from the originating mbuf to the newly created
 235                  * mbuf. This also ensures that existing firewall
 236                  * classification(s), VLAN tags and so on get copied
 237                  * to the resulting fragmented packet(s):
 238                  */
 239                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 240                         m_free(m);
 241                         IP6STAT_INC(ip6s_odropped);
 242                         return (ENOBUFS);
 243                 }
 244
 245                 *mnext = m;
 246                 mnext = &m->m_nextpkt;
 247                 m->m_data += max_linkhdr;
 248                 mhip6 = mtod(m, struct ip6_hdr *);
 249                 *mhip6 = *ip6;
 250                 m->m_len = sizeof(*mhip6);
 251                 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 252                 if (error) {
 253                         IP6STAT_INC(ip6s_odropped);
 254                         return (error);
 255                 }
 256                 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 257                 if (off + fraglen >= tlen)
 258                         fraglen = tlen - off;
 259                 else
 260                         ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 261                 mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
 262                     sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 263                 if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
 264                         IP6STAT_INC(ip6s_odropped);
 265                         return (ENOBUFS);
 266                 }
 267                 m_cat(m, m_frgpart);
 268                 m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
 269                 ip6f->ip6f_reserved = 0;
 270                 ip6f->ip6f_ident = id;
 271                 ip6f->ip6f_nxt = nextproto;
 272                 IP6STAT_INC(ip6s_ofragments);
 273                 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 274         }
 275
 276         return (0);
 277 }
 278
 279 /*
 280  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 281  * header (with pri, len, nxt, hlim, src, dst).
 282  * This function may modify ver and hlim only.
 283  * The mbuf chain containing the packet will be freed.
 284  * The mbuf opt, if present, will not be freed.
 285  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
 286  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 287  * then result of route lookup is stored in ro->ro_rt.
 288  *
 289  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
 290  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 291  * which is rt_mtu.
 292  *
 293  * ifpp - XXX: just for statistics
 294  */
 295 /*
 296  * XXX TODO: no flowid is assigned for outbound flows?
 297  */
 298 int
 299 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
 300     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
 301     struct ifnet **ifpp, struct inpcb *inp)
 302 {
 303         struct ip6_hdr *ip6;
 304         struct ifnet *ifp, *origifp;
 305         struct mbuf *m = m0;
 306         struct mbuf *mprev = NULL;
 307         int hlen, tlen, len;
 308         struct route_in6 ip6route;
 309         struct rtentry *rt = NULL;
 310         struct sockaddr_in6 *dst, src_sa, dst_sa;
 311         struct in6_addr odst;
 312         int error = 0;
 313         struct in6_ifaddr *ia = NULL;
 314         u_long mtu;
 315         int alwaysfrag, dontfrag;
 316         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 317         struct ip6_exthdrs exthdrs;
 318         struct in6_addr src0, dst0;
 319         u_int32_t zone;
 320         struct route_in6 *ro_pmtu = NULL;
 321         int hdrsplit = 0;
 322         int sw_csum, tso;
 323         int needfiblookup;
 324         uint32_t fibnum;
 325         struct m_tag *fwd_tag = NULL;
 326         uint32_t id;
 327
 328         if (inp != NULL) {
 329                 INP_LOCK_ASSERT(inp);
 330                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 331                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 332                         /* unconditionally set flowid */
 333                         m->m_pkthdr.flowid = inp->inp_flowid;
 334                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 335                 }
 336         }
 337
 338 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 339         /*
 340          * IPSec checking which handles several cases.
 341          * FAST IPSEC: We re-injected the packet.
 342          * XXX: need scope argument.
 343          */
 344         if (IPSEC_ENABLED(ipv6)) {
 345                 if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
 346                         if (error == EINPROGRESS)
 347                                 error = 0;
 348                         goto done;
 349                 }
 350         }
 351 #endif /* IPSEC */
 352
 353         bzero(&exthdrs, sizeof(exthdrs));
 354         if (opt) {
 355                 /* Hop-by-Hop options header */
 356                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 357                 /* Destination options header(1st part) */
 358                 if (opt->ip6po_rthdr) {
 359                         /*
 360                          * Destination options header(1st part)
 361                          * This only makes sense with a routing header.
 362                          * See Section 9.2 of RFC 3542.
 363                          * Disabling this part just for MIP6 convenience is
 364                          * a bad idea.  We need to think carefully about a
 365                          * way to make the advanced API coexist with MIP6
 366                          * options, which might automatically be inserted in
 367                          * the kernel.
 368                          */
 369                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 370                 }
 371                 /* Routing header */
 372                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 373                 /* Destination options header(2nd part) */
 374                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 375         }
 376
 377         /*
 378          * Calculate the total length of the extension header chain.
 379          * Keep the length of the unfragmentable part for fragmentation.
 380          */
 381         optlen = 0;
 382         if (exthdrs.ip6e_hbh)
 383                 optlen += exthdrs.ip6e_hbh->m_len;
 384         if (exthdrs.ip6e_dest1)
 385                 optlen += exthdrs.ip6e_dest1->m_len;
 386         if (exthdrs.ip6e_rthdr)
 387                 optlen += exthdrs.ip6e_rthdr->m_len;
 388         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 389
 390         /* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
 391         if (exthdrs.ip6e_dest2)
 392                 optlen += exthdrs.ip6e_dest2->m_len;
 393
 394         /*
 395          * If there is at least one extension header,
 396          * separate IP6 header from the payload.
 397          */
 398         if (optlen && !hdrsplit) {
 399                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 400                         m = NULL;
 401                         goto freehdrs;
 402                 }
 403                 m = exthdrs.ip6e_ip6;
 404                 hdrsplit++;
 405         }
 406
 407         ip6 = mtod(m, struct ip6_hdr *);
 408
 409         /* adjust mbuf packet header length */
 410         m->m_pkthdr.len += optlen;
 411         plen = m->m_pkthdr.len - sizeof(*ip6);
 412
 413         /* If this is a jumbo payload, insert a jumbo payload option. */
 414         if (plen > IPV6_MAXPACKET) {
 415                 if (!hdrsplit) {
 416                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 417                                 m = NULL;
 418                                 goto freehdrs;
 419                         }
 420                         m = exthdrs.ip6e_ip6;
 421                         hdrsplit++;
 422                 }
 423                 /* adjust pointer */
 424                 ip6 = mtod(m, struct ip6_hdr *);
 425                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 426                         goto freehdrs;
 427                 ip6->ip6_plen = 0;
 428         } else
 429                 ip6->ip6_plen = htons(plen);
 430
 431         /*
 432          * Concatenate headers and fill in next header fields.
 433          * Here we have, on "m"
 434          *      IPv6 payload
 435          * and we insert headers accordingly.  Finally, we should be getting:
 436          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 437          *
 438          * during the header composing process, "m" points to IPv6 header.
 439          * "mprev" points to an extension header prior to esp.
 440          */
 441         u_char *nexthdrp = &ip6->ip6_nxt;
 442         mprev = m;
 443
 444         /*
 445          * we treat dest2 specially.  this makes IPsec processing
 446          * much easier.  the goal here is to make mprev point the
 447          * mbuf prior to dest2.
 448          *
 449          * result: IPv6 dest2 payload
 450          * m and mprev will point to IPv6 header.
 451          */
 452         if (exthdrs.ip6e_dest2) {
 453                 if (!hdrsplit)
 454                         panic("assumption failed: hdr not split");
 455                 exthdrs.ip6e_dest2->m_next = m->m_next;
 456                 m->m_next = exthdrs.ip6e_dest2;
 457                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 458                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 459         }
 460
 461         /*
 462          * result: IPv6 hbh dest1 rthdr dest2 payload
 463          * m will point to IPv6 header.  mprev will point to the
 464          * extension header prior to dest2 (rthdr in the above case).
 465          */
 466         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 467         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
 468                    IPPROTO_DSTOPTS);
 469         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
 470                    IPPROTO_ROUTING);
 471
 472         /*
 473          * If there is a routing header, discard the packet.
 474          */
 475         if (exthdrs.ip6e_rthdr) {
 476                  error = EINVAL;
 477                  goto bad;
 478         }
 479
 480         /* Source address validation */
 481         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 482             (flags & IPV6_UNSPECSRC) == 0) {
 483                 error = EOPNOTSUPP;
 484                 IP6STAT_INC(ip6s_badscope);
 485                 goto bad;
 486         }
 487         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 488                 error = EOPNOTSUPP;
 489                 IP6STAT_INC(ip6s_badscope);
 490                 goto bad;
 491         }
 492
 493         IP6STAT_INC(ip6s_localout);
 494
 495         /*
 496          * Route packet.
 497          */
 498         if (ro == NULL) {
 499                 ro = &ip6route;
 500                 bzero((caddr_t)ro, sizeof(*ro));
 501         }
 502         ro_pmtu = ro;
 503         if (opt && opt->ip6po_rthdr)
 504                 ro = &opt->ip6po_route;
 505         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 506         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 507 again:
 508         /*
 509          * if specified, try to fill in the traffic class field.
 510          * do not override if a non-zero value is already set.
 511          * we check the diffserv field and the ecn field separately.
 512          */
 513         if (opt && opt->ip6po_tclass >= 0) {
 514                 int mask = 0;
 515
 516                 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
 517                         mask |= 0xfc;
 518                 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
 519                         mask |= 0x03;
 520                 if (mask != 0)
 521                         ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
 522         }
 523
 524         /* fill in or override the hop limit field, if necessary. */
 525         if (opt && opt->ip6po_hlim != -1)
 526                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 527         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 528                 if (im6o != NULL)
 529                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 530                 else
 531                         ip6->ip6_hlim = V_ip6_defmcasthlim;
 532         }
 533         /*
 534          * Validate route against routing table additions;
 535          * a better/more specific route might have been added.
 536          * Make sure address family is set in route.
 537          */
 538         if (inp) {
 539                 ro->ro_dst.sin6_family = AF_INET6;
 540                 RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
 541         }
 542         if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
 543             ro->ro_dst.sin6_family == AF_INET6 &&
 544             IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
 545                 rt = ro->ro_rt;
 546                 ifp = ro->ro_rt->rt_ifp;
 547         } else {
 548                 if (ro->ro_lle)
 549                         LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 550                 ro->ro_lle = NULL;
 551                 if (fwd_tag == NULL) {
 552                         bzero(&dst_sa, sizeof(dst_sa));
 553                         dst_sa.sin6_family = AF_INET6;
 554                         dst_sa.sin6_len = sizeof(dst_sa);
 555                         dst_sa.sin6_addr = ip6->ip6_dst;
 556                 }
 557                 error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
 558                     &rt, fibnum);
 559                 if (error != 0) {
 560                         if (ifp != NULL)
 561                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 562                         goto bad;
 563                 }
 564         }
 565         if (rt == NULL) {
 566                 /*
 567                  * If in6_selectroute() does not return a route entry,
 568                  * dst may not have been updated.
 569                  */
 570                 *dst = dst_sa;  /* XXX */
 571         }
 572
 573         /*
 574          * then rt (for unicast) and ifp must be non-NULL valid values.
 575          */
 576         if ((flags & IPV6_FORWARDING) == 0) {
 577                 /* XXX: the FORWARDING flag can be set for mrouting. */
 578                 in6_ifstat_inc(ifp, ifs6_out_request);
 579         }
 580         if (rt != NULL) {
 581                 ia = (struct in6_ifaddr *)(rt->rt_ifa);
 582                 counter_u64_add(rt->rt_pksent, 1);
 583         }
 584
 585         /* Setup data structures for scope ID checks. */
 586         src0 = ip6->ip6_src;
 587         bzero(&src_sa, sizeof(src_sa));
 588         src_sa.sin6_family = AF_INET6;
 589         src_sa.sin6_len = sizeof(src_sa);
 590         src_sa.sin6_addr = ip6->ip6_src;
 591
 592         dst0 = ip6->ip6_dst;
 593         /* re-initialize to be sure */
 594         bzero(&dst_sa, sizeof(dst_sa));
 595         dst_sa.sin6_family = AF_INET6;
 596         dst_sa.sin6_len = sizeof(dst_sa);
 597         dst_sa.sin6_addr = ip6->ip6_dst;
 598
 599         /* Check for valid scope ID. */
 600         if (in6_setscope(&src0, ifp, &zone) == 0 &&
 601             sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
 602             in6_setscope(&dst0, ifp, &zone) == 0 &&
 603             sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
 604                 /*
 605                  * The outgoing interface is in the zone of the source
 606                  * and destination addresses.
 607                  *
 608                  * Because the loopback interface cannot receive
 609                  * packets with a different scope ID than its own,
 610                  * there is a trick is to pretend the outgoing packet
 611                  * was received by the real network interface, by
 612                  * setting "origifp" different from "ifp". This is
 613                  * only allowed when "ifp" is a loopback network
 614                  * interface. Refer to code in nd6_output_ifp() for
 615                  * more details.
 616                  */
 617                 origifp = ifp;
 618
 619                 /*
 620                  * We should use ia_ifp to support the case of sending
 621                  * packets to an address of our own.
 622                  */
 623                 if (ia != NULL && ia->ia_ifp)
 624                         ifp = ia->ia_ifp;
 625
 626         } else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
 627             sa6_recoverscope(&src_sa) != 0 ||
 628             sa6_recoverscope(&dst_sa) != 0 ||
 629             dst_sa.sin6_scope_id == 0 ||
 630             (src_sa.sin6_scope_id != 0 &&
 631             src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
 632             (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
 633                 /*
 634                  * If the destination network interface is not a
 635                  * loopback interface, or the destination network
 636                  * address has no scope ID, or the source address has
 637                  * a scope ID set which is different from the
 638                  * destination address one, or there is no network
 639                  * interface representing this scope ID, the address
 640                  * pair is considered invalid.
 641                  */
 642                 IP6STAT_INC(ip6s_badscope);
 643                 in6_ifstat_inc(ifp, ifs6_out_discard);
 644                 if (error == 0)
 645                         error = EHOSTUNREACH; /* XXX */
 646                 goto bad;
 647         }
 648
 649         /* All scope ID checks are successful. */
 650
 651         if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 652                 if (opt && opt->ip6po_nextroute.ro_rt) {
 653                         /*
 654                          * The nexthop is explicitly specified by the
 655                          * application.  We assume the next hop is an IPv6
 656                          * address.
 657                          */
 658                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
 659                 }
 660                 else if ((rt->rt_flags & RTF_GATEWAY))
 661                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
 662         }
 663
 664         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 665                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
 666         } else {
 667                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 668                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 669                 /*
 670                  * Confirm that the outgoing interface supports multicast.
 671                  */
 672                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 673                         IP6STAT_INC(ip6s_noroute);
 674                         in6_ifstat_inc(ifp, ifs6_out_discard);
 675                         error = ENETUNREACH;
 676                         goto bad;
 677                 }
 678                 if ((im6o == NULL && in6_mcast_loop) ||
 679                     (im6o && im6o->im6o_multicast_loop)) {
 680                         /*
 681                          * Loop back multicast datagram if not expressly
 682                          * forbidden to do so, even if we have not joined
 683                          * the address; protocols will filter it later,
 684                          * thus deferring a hash lookup and lock acquisition
 685                          * at the expense of an m_copym().
 686                          */
 687                         ip6_mloopback(ifp, m);
 688                 } else {
 689                         /*
 690                          * If we are acting as a multicast router, perform
 691                          * multicast forwarding as if the packet had just
 692                          * arrived on the interface to which we are about
 693                          * to send.  The multicast forwarding function
 694                          * recursively calls this function, using the
 695                          * IPV6_FORWARDING flag to prevent infinite recursion.
 696                          *
 697                          * Multicasts that are looped back by ip6_mloopback(),
 698                          * above, will be forwarded by the ip6_input() routine,
 699                          * if necessary.
 700                          */
 701                         if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
 702                                 /*
 703                                  * XXX: ip6_mforward expects that rcvif is NULL
 704                                  * when it is called from the originating path.
 705                                  * However, it may not always be the case.
 706                                  */
 707                                 m->m_pkthdr.rcvif = NULL;
 708                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 709                                         m_freem(m);
 710                                         goto done;
 711                                 }
 712                         }
 713                 }
 714                 /*
 715                  * Multicasts with a hoplimit of zero may be looped back,
 716                  * above, but must not be transmitted on a network.
 717                  * Also, multicasts addressed to the loopback interface
 718                  * are not sent -- the above call to ip6_mloopback() will
 719                  * loop back a copy if this host actually belongs to the
 720                  * destination group on the loopback interface.
 721                  */
 722                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
 723                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 724                         m_freem(m);
 725                         goto done;
 726                 }
 727         }
 728
 729         /*
 730          * Fill the outgoing inteface to tell the upper layer
 731          * to increment per-interface statistics.
 732          */
 733         if (ifpp)
 734                 *ifpp = ifp;
 735
 736         /* Determine path MTU. */
 737         if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
 738                     &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
 739                 goto bad;
 740
 741         /*
 742          * The caller of this function may specify to use the minimum MTU
 743          * in some cases.
 744          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 745          * setting.  The logic is a bit complicated; by default, unicast
 746          * packets will follow path MTU while multicast packets will be sent at
 747          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 748          * including unicast ones will be sent at the minimum MTU.  Multicast
 749          * packets will always be sent at the minimum MTU unless
 750          * IP6PO_MINMTU_DISABLE is explicitly specified.
 751          * See RFC 3542 for more details.
 752          */
 753         if (mtu > IPV6_MMTU) {
 754                 if ((flags & IPV6_MINMTU))
 755                         mtu = IPV6_MMTU;
 756                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 757                         mtu = IPV6_MMTU;
 758                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 759                          (opt == NULL ||
 760                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 761                         mtu = IPV6_MMTU;
 762                 }
 763         }
 764
 765         /*
 766          * clear embedded scope identifiers if necessary.
 767          * in6_clearscope will touch the addresses only when necessary.
 768          */
 769         in6_clearscope(&ip6->ip6_src);
 770         in6_clearscope(&ip6->ip6_dst);
 771
 772         /*
 773          * If the outgoing packet contains a hop-by-hop options header,
 774          * it must be examined and processed even by the source node.
 775          * (RFC 2460, section 4.)
 776          */
 777         if (exthdrs.ip6e_hbh) {
 778                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 779                 u_int32_t dummy; /* XXX unused */
 780                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
 781
 782 #ifdef DIAGNOSTIC
 783                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 784                         panic("ip6e_hbh is not contiguous");
 785 #endif
 786                 /*
 787                  *  XXX: if we have to send an ICMPv6 error to the sender,
 788                  *       we need the M_LOOP flag since icmp6_error() expects
 789                  *       the IPv6 and the hop-by-hop options header are
 790                  *       contiguous unless the flag is set.
 791                  */
 792                 m->m_flags |= M_LOOP;
 793                 m->m_pkthdr.rcvif = ifp;
 794                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 795                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 796                     &dummy, &plen) < 0) {
 797                         /* m was already freed at this point */
 798                         error = EINVAL;/* better error? */
 799                         goto done;
 800                 }
 801                 m->m_flags &= ~M_LOOP; /* XXX */
 802                 m->m_pkthdr.rcvif = NULL;
 803         }
 804
 805         /* Jump over all PFIL processing if hooks are not active. */
 806         if (!PFIL_HOOKED(&V_inet6_pfil_hook))
 807                 goto passout;
 808
 809         odst = ip6->ip6_dst;
 810         /* Run through list of hooks for output packets. */
 811         error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, 0, inp);
 812         if (error != 0 || m == NULL)
 813                 goto done;
 814         /* adjust pointer */
 815         ip6 = mtod(m, struct ip6_hdr *);
 816
 817         needfiblookup = 0;
 818         /* See if destination IP address was changed by packet filter. */
 819         if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
 820                 m->m_flags |= M_SKIP_FIREWALL;
 821                 /* If destination is now ourself drop to ip6_input(). */
 822                 if (in6_localip(&ip6->ip6_dst)) {
 823                         m->m_flags |= M_FASTFWD_OURS;
 824                         if (m->m_pkthdr.rcvif == NULL)
 825                                 m->m_pkthdr.rcvif = V_loif;
 826                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 827                                 m->m_pkthdr.csum_flags |=
 828                                     CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 829                                 m->m_pkthdr.csum_data = 0xffff;
 830                         }
 831 #ifdef SCTP
 832                         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 833                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 834 #endif
 835                         error = netisr_queue(NETISR_IPV6, m);
 836                         goto done;
 837                 } else {
 838                         RO_INVALIDATE_CACHE(ro);
 839                         needfiblookup = 1; /* Redo the routing table lookup. */
 840                 }
 841         }
 842         /* See if fib was changed by packet filter. */
 843         if (fibnum != M_GETFIB(m)) {
 844                 m->m_flags |= M_SKIP_FIREWALL;
 845                 fibnum = M_GETFIB(m);
 846                 RO_INVALIDATE_CACHE(ro);
 847                 needfiblookup = 1;
 848         }
 849         if (needfiblookup)
 850                 goto again;
 851
 852         /* See if local, if yes, send it to netisr. */
 853         if (m->m_flags & M_FASTFWD_OURS) {
 854                 if (m->m_pkthdr.rcvif == NULL)
 855                         m->m_pkthdr.rcvif = V_loif;
 856                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 857                         m->m_pkthdr.csum_flags |=
 858                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 859                         m->m_pkthdr.csum_data = 0xffff;
 860                 }
 861 #ifdef SCTP
 862                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 863                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 864 #endif
 865                 error = netisr_queue(NETISR_IPV6, m);
 866                 goto done;
 867         }
 868         /* Or forward to some other address? */
 869         if ((m->m_flags & M_IP6_NEXTHOP) &&
 870             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 871                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 872                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
 873                 m->m_flags |= M_SKIP_FIREWALL;
 874                 m->m_flags &= ~M_IP6_NEXTHOP;
 875                 m_tag_delete(m, fwd_tag);
 876                 goto again;
 877         }
 878
 879 passout:
 880         /*
 881          * Send the packet to the outgoing interface.
 882          * If necessary, do IPv6 fragmentation before sending.
 883          *
 884          * the logic here is rather complex:
 885          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 886          * 1-a: send as is if tlen <= path mtu
 887          * 1-b: fragment if tlen > path mtu
 888          *
 889          * 2: if user asks us not to fragment (dontfrag == 1)
 890          * 2-a: send as is if tlen <= interface mtu
 891          * 2-b: error if tlen > interface mtu
 892          *
 893          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 894          *      always fragment
 895          *
 896          * 4: if dontfrag == 1 && alwaysfrag == 1
 897          *      error, as we cannot handle this conflicting request
 898          */
 899         sw_csum = m->m_pkthdr.csum_flags;
 900         if (!hdrsplit) {
 901                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 902                 sw_csum &= ~ifp->if_hwassist;
 903         } else
 904                 tso = 0;
 905         /*
 906          * If we added extension headers, we will not do TSO and calculate the
 907          * checksums ourselves for now.
 908          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 909          * with ext. hdrs.
 910          */
 911         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 912                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 913                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 914         }
 915 #ifdef SCTP
 916         if (sw_csum & CSUM_SCTP_IPV6) {
 917                 sw_csum &= ~CSUM_SCTP_IPV6;
 918                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 919         }
 920 #endif
 921         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 922         tlen = m->m_pkthdr.len;
 923
 924         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 925                 dontfrag = 1;
 926         else
 927                 dontfrag = 0;
 928         if (dontfrag && alwaysfrag) {   /* case 4 */
 929                 /* conflicting request - can't transmit */
 930                 error = EMSGSIZE;
 931                 goto bad;
 932         }
 933         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 934                 /*
 935                  * Even if the DONTFRAG option is specified, we cannot send the
 936                  * packet when the data length is larger than the MTU of the
 937                  * outgoing interface.
 938                  * Notify the error by sending IPV6_PATHMTU ancillary data if
 939                  * application wanted to know the MTU value. Also return an
 940                  * error code (this is not described in the API spec).
 941                  */
 942                 if (inp != NULL)
 943                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
 944                 error = EMSGSIZE;
 945                 goto bad;
 946         }
 947
 948         /*
 949          * transmit packet without fragmentation
 950          */
 951         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 952                 struct in6_ifaddr *ia6;
 953
 954                 ip6 = mtod(m, struct ip6_hdr *);
 955                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 956                 if (ia6) {
 957                         /* Record statistics for this interface address. */
 958                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
 959                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
 960                             m->m_pkthdr.len);
 961                         ifa_free(&ia6->ia_ifa);
 962                 }
 963 #ifdef RATELIMIT
 964                 if (inp != NULL) {
 965                         if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
 966                                 in_pcboutput_txrtlmt(inp, ifp, m);
 967                         /* stamp send tag on mbuf */
 968                         m->m_pkthdr.snd_tag = inp->inp_snd_tag;
 969                 } else {
 970                         m->m_pkthdr.snd_tag = NULL;
 971                 }
 972 #endif
 973                 error = nd6_output_ifp(ifp, origifp, m, dst,
 974                     (struct route *)ro);
 975 #ifdef RATELIMIT
 976                 /* check for route change */
 977                 if (error == EAGAIN)
 978                         in_pcboutput_eagain(inp);
 979 #endif
 980                 goto done;
 981         }
 982
 983         /*
 984          * try to fragment the packet.  case 1-b and 3
 985          */
 986         if (mtu < IPV6_MMTU) {
 987                 /* path MTU cannot be less than IPV6_MMTU */
 988                 error = EMSGSIZE;
 989                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 990                 goto bad;
 991         } else if (ip6->ip6_plen == 0) {
 992                 /* jumbo payload cannot be fragmented */
 993                 error = EMSGSIZE;
 994                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 995                 goto bad;
 996         } else {
 997                 u_char nextproto;
 998
 999                 /*
1000                  * Too large for the destination or interface;
1001                  * fragment if possible.
1002                  * Must be able to put at least 8 bytes per fragment.
1003                  */
1004                 hlen = unfragpartlen;
1005                 if (mtu > IPV6_MAXPACKET)
1006                         mtu = IPV6_MAXPACKET;
1007
1008                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1009                 if (len < 8) {
1010                         error = EMSGSIZE;
1011                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1012                         goto bad;
1013                 }
1014
1015                 /*
1016                  * If the interface will not calculate checksums on
1017                  * fragmented packets, then do it here.
1018                  * XXX-BZ handle the hw offloading case.  Need flags.
1019                  */
1020                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1021                         in6_delayed_cksum(m, plen, hlen);
1022                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1023                 }
1024 #ifdef SCTP
1025                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1026                         sctp_delayed_cksum(m, hlen);
1027                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1028                 }
1029 #endif
1030                 /*
1031                  * Change the next header field of the last header in the
1032                  * unfragmentable part.
1033                  */
1034                 if (exthdrs.ip6e_rthdr) {
1035                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1036                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1037                 } else if (exthdrs.ip6e_dest1) {
1038                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1039                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1040                 } else if (exthdrs.ip6e_hbh) {
1041                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1042                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1043                 } else {
1044                         nextproto = ip6->ip6_nxt;
1045                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1046                 }
1047
1048                 /*
1049                  * Loop through length of segment after first fragment,
1050                  * make new header and copy data of each part and link onto
1051                  * chain.
1052                  */
1053                 m0 = m;
1054                 id = htonl(ip6_randomid());
1055                 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1056                         goto sendorfree;
1057
1058                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1059         }
1060
1061         /*
1062          * Remove leading garbages.
1063          */
1064 sendorfree:
1065         m = m0->m_nextpkt;
1066         m0->m_nextpkt = 0;
1067         m_freem(m0);
1068         for (; m; m = m0) {
1069                 m0 = m->m_nextpkt;
1070                 m->m_nextpkt = 0;
1071                 if (error == 0) {
1072                         /* Record statistics for this interface address. */
1073                         if (ia) {
1074                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1075                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1076                                     m->m_pkthdr.len);
1077                         }
1078 #ifdef RATELIMIT
1079                         if (inp != NULL) {
1080                                 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1081                                         in_pcboutput_txrtlmt(inp, ifp, m);
1082                                 /* stamp send tag on mbuf */
1083                                 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1084                         } else {
1085                                 m->m_pkthdr.snd_tag = NULL;
1086                         }
1087 #endif
1088                         error = nd6_output_ifp(ifp, origifp, m, dst,
1089                             (struct route *)ro);
1090 #ifdef RATELIMIT
1091                         /* check for route change */
1092                         if (error == EAGAIN)
1093                                 in_pcboutput_eagain(inp);
1094 #endif
1095                 } else
1096                         m_freem(m);
1097         }
1098
1099         if (error == 0)
1100                 IP6STAT_INC(ip6s_fragmented);
1101
1102 done:
1103         if (ro == &ip6route)
1104                 RO_RTFREE(ro);
1105         return (error);
1106
1107 freehdrs:
1108         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1109         m_freem(exthdrs.ip6e_dest1);
1110         m_freem(exthdrs.ip6e_rthdr);
1111         m_freem(exthdrs.ip6e_dest2);
1112         /* FALLTHROUGH */
1113 bad:
1114         if (m)
1115                 m_freem(m);
1116         goto done;
1117 }
1118
1119 static int
1120 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1121 {
1122         struct mbuf *m;
1123
1124         if (hlen > MCLBYTES)
1125                 return (ENOBUFS); /* XXX */
1126
1127         if (hlen > MLEN)
1128                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1129         else
1130                 m = m_get(M_NOWAIT, MT_DATA);
1131         if (m == NULL)
1132                 return (ENOBUFS);
1133         m->m_len = hlen;
1134         if (hdr)
1135                 bcopy(hdr, mtod(m, caddr_t), hlen);
1136
1137         *mp = m;
1138         return (0);
1139 }
1140
1141 /*
1142  * Insert jumbo payload option.
1143  */
1144 static int
1145 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1146 {
1147         struct mbuf *mopt;
1148         u_char *optbuf;
1149         u_int32_t v;
1150
1151 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1152
1153         /*
1154          * If there is no hop-by-hop options header, allocate new one.
1155          * If there is one but it doesn't have enough space to store the
1156          * jumbo payload option, allocate a cluster to store the whole options.
1157          * Otherwise, use it to store the options.
1158          */
1159         if (exthdrs->ip6e_hbh == NULL) {
1160                 mopt = m_get(M_NOWAIT, MT_DATA);
1161                 if (mopt == NULL)
1162                         return (ENOBUFS);
1163                 mopt->m_len = JUMBOOPTLEN;
1164                 optbuf = mtod(mopt, u_char *);
1165                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1166                 exthdrs->ip6e_hbh = mopt;
1167         } else {
1168                 struct ip6_hbh *hbh;
1169
1170                 mopt = exthdrs->ip6e_hbh;
1171                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1172                         /*
1173                          * XXX assumption:
1174                          * - exthdrs->ip6e_hbh is not referenced from places
1175                          *   other than exthdrs.
1176                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1177                          */
1178                         int oldoptlen = mopt->m_len;
1179                         struct mbuf *n;
1180
1181                         /*
1182                          * XXX: give up if the whole (new) hbh header does
1183                          * not fit even in an mbuf cluster.
1184                          */
1185                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1186                                 return (ENOBUFS);
1187
1188                         /*
1189                          * As a consequence, we must always prepare a cluster
1190                          * at this point.
1191                          */
1192                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1193                         if (n == NULL)
1194                                 return (ENOBUFS);
1195                         n->m_len = oldoptlen + JUMBOOPTLEN;
1196                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1197                             oldoptlen);
1198                         optbuf = mtod(n, caddr_t) + oldoptlen;
1199                         m_freem(mopt);
1200                         mopt = exthdrs->ip6e_hbh = n;
1201                 } else {
1202                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1203                         mopt->m_len += JUMBOOPTLEN;
1204                 }
1205                 optbuf[0] = IP6OPT_PADN;
1206                 optbuf[1] = 1;
1207
1208                 /*
1209                  * Adjust the header length according to the pad and
1210                  * the jumbo payload option.
1211                  */
1212                 hbh = mtod(mopt, struct ip6_hbh *);
1213                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1214         }
1215
1216         /* fill in the option. */
1217         optbuf[2] = IP6OPT_JUMBO;
1218         optbuf[3] = 4;
1219         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1220         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1221
1222         /* finally, adjust the packet header length */
1223         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1224
1225         return (0);
1226 #undef JUMBOOPTLEN
1227 }
1228
1229 /*
1230  * Insert fragment header and copy unfragmentable header portions.
1231  */
1232 static int
1233 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1234     struct ip6_frag **frghdrp)
1235 {
1236         struct mbuf *n, *mlast;
1237
1238         if (hlen > sizeof(struct ip6_hdr)) {
1239                 n = m_copym(m0, sizeof(struct ip6_hdr),
1240                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1241                 if (n == NULL)
1242                         return (ENOBUFS);
1243                 m->m_next = n;
1244         } else
1245                 n = m;
1246
1247         /* Search for the last mbuf of unfragmentable part. */
1248         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1249                 ;
1250
1251         if (M_WRITABLE(mlast) &&
1252             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1253                 /* use the trailing space of the last mbuf for the fragment hdr */
1254                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1255                     mlast->m_len);
1256                 mlast->m_len += sizeof(struct ip6_frag);
1257                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1258         } else {
1259                 /* allocate a new mbuf for the fragment header */
1260                 struct mbuf *mfrg;
1261
1262                 mfrg = m_get(M_NOWAIT, MT_DATA);
1263                 if (mfrg == NULL)
1264                         return (ENOBUFS);
1265                 mfrg->m_len = sizeof(struct ip6_frag);
1266                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1267                 mlast->m_next = mfrg;
1268         }
1269
1270         return (0);
1271 }
1272
1273 /*
1274  * Calculates IPv6 path mtu for destination @dst.
1275  * Resulting MTU is stored in @mtup.
1276  *
1277  * Returns 0 on success.
1278  */
1279 static int
1280 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1281 {
1282         struct nhop6_extended nh6;
1283         struct in6_addr kdst;
1284         uint32_t scopeid;
1285         struct ifnet *ifp;
1286         u_long mtu;
1287         int error;
1288
1289         in6_splitscope(dst, &kdst, &scopeid);
1290         if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1291                 return (EHOSTUNREACH);
1292
1293         ifp = nh6.nh_ifp;
1294         mtu = nh6.nh_mtu;
1295
1296         error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1297         fib6_free_nh_ext(fibnum, &nh6);
1298
1299         return (error);
1300 }
1301
1302 /*
1303  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1304  * and cached data in @ro_pmtu.
1305  * MTU from (successful) route lookup is saved (along with dst)
1306  * inside @ro_pmtu to avoid subsequent route lookups after packet
1307  * filter processing.
1308  *
1309  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1310  * Returns 0 on success.
1311  */
1312 static int
1313 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1314     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1315     int *alwaysfragp, u_int fibnum, u_int proto)
1316 {
1317         struct nhop6_basic nh6;
1318         struct in6_addr kdst;
1319         uint32_t scopeid;
1320         struct sockaddr_in6 *sa6_dst;
1321         u_long mtu;
1322
1323         mtu = 0;
1324         if (do_lookup) {
1325
1326                 /*
1327                  * Here ro_pmtu has final destination address, while
1328                  * ro might represent immediate destination.
1329                  * Use ro_pmtu destination since mtu might differ.
1330                  */
1331                 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1332                 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1333                         ro_pmtu->ro_mtu = 0;
1334
1335                 if (ro_pmtu->ro_mtu == 0) {
1336                         bzero(sa6_dst, sizeof(*sa6_dst));
1337                         sa6_dst->sin6_family = AF_INET6;
1338                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1339                         sa6_dst->sin6_addr = *dst;
1340
1341                         in6_splitscope(dst, &kdst, &scopeid);
1342                         if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1343                             &nh6) == 0)
1344                                 ro_pmtu->ro_mtu = nh6.nh_mtu;
1345                 }
1346
1347                 mtu = ro_pmtu->ro_mtu;
1348         }
1349
1350         if (ro_pmtu->ro_rt)
1351                 mtu = ro_pmtu->ro_rt->rt_mtu;
1352
1353         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1354 }
1355
1356 /*
1357  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1358  * hostcache data for @dst.
1359  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1360  *
1361  * Returns 0 on success.
1362  */
1363 static int
1364 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1365     u_long *mtup, int *alwaysfragp, u_int proto)
1366 {
1367         u_long mtu = 0;
1368         int alwaysfrag = 0;
1369         int error = 0;
1370
1371         if (rt_mtu > 0) {
1372                 u_int32_t ifmtu;
1373                 struct in_conninfo inc;
1374
1375                 bzero(&inc, sizeof(inc));
1376                 inc.inc_flags |= INC_ISIPV6;
1377                 inc.inc6_faddr = *dst;
1378
1379                 ifmtu = IN6_LINKMTU(ifp);
1380
1381                 /* TCP is known to react to pmtu changes so skip hc */
1382                 if (proto != IPPROTO_TCP)
1383                         mtu = tcp_hc_getmtu(&inc);
1384
1385                 if (mtu)
1386                         mtu = min(mtu, rt_mtu);
1387                 else
1388                         mtu = rt_mtu;
1389                 if (mtu == 0)
1390                         mtu = ifmtu;
1391                 else if (mtu < IPV6_MMTU) {
1392                         /*
1393                          * RFC2460 section 5, last paragraph:
1394                          * if we record ICMPv6 too big message with
1395                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1396                          * or smaller, with framgent header attached.
1397                          * (fragment header is needed regardless from the
1398                          * packet size, for translators to identify packets)
1399                          */
1400                         alwaysfrag = 1;
1401                         mtu = IPV6_MMTU;
1402                 }
1403         } else if (ifp) {
1404                 mtu = IN6_LINKMTU(ifp);
1405         } else
1406                 error = EHOSTUNREACH; /* XXX */
1407
1408         *mtup = mtu;
1409         if (alwaysfragp)
1410                 *alwaysfragp = alwaysfrag;
1411         return (error);
1412 }
1413
1414 /*
1415  * IP6 socket option processing.
1416  */
1417 int
1418 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1419 {
1420         int optdatalen, uproto;
1421         void *optdata;
1422         struct inpcb *in6p = sotoinpcb(so);
1423         int error, optval;
1424         int level, op, optname;
1425         int optlen;
1426         struct thread *td;
1427 #ifdef  RSS
1428         uint32_t rss_bucket;
1429         int retval;
1430 #endif
1431
1432 /*
1433  * Don't use more than a quarter of mbuf clusters.  N.B.:
1434  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1435  * on LP64 architectures, so cast to u_long to avoid undefined
1436  * behavior.  ILP32 architectures cannot have nmbclusters
1437  * large enough to overflow for other reasons.
1438  */
1439 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1440
1441         level = sopt->sopt_level;
1442         op = sopt->sopt_dir;
1443         optname = sopt->sopt_name;
1444         optlen = sopt->sopt_valsize;
1445         td = sopt->sopt_td;
1446         error = 0;
1447         optval = 0;
1448         uproto = (int)so->so_proto->pr_protocol;
1449
1450         if (level != IPPROTO_IPV6) {
1451                 error = EINVAL;
1452
1453                 if (sopt->sopt_level == SOL_SOCKET &&
1454                     sopt->sopt_dir == SOPT_SET) {
1455                         switch (sopt->sopt_name) {
1456                         case SO_REUSEADDR:
1457                                 INP_WLOCK(in6p);
1458                                 if ((so->so_options & SO_REUSEADDR) != 0)
1459                                         in6p->inp_flags2 |= INP_REUSEADDR;
1460                                 else
1461                                         in6p->inp_flags2 &= ~INP_REUSEADDR;
1462                                 INP_WUNLOCK(in6p);
1463                                 error = 0;
1464                                 break;
1465                         case SO_REUSEPORT:
1466                                 INP_WLOCK(in6p);
1467                                 if ((so->so_options & SO_REUSEPORT) != 0)
1468                                         in6p->inp_flags2 |= INP_REUSEPORT;
1469                                 else
1470                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1471                                 INP_WUNLOCK(in6p);
1472                                 error = 0;
1473                                 break;
1474                         case SO_REUSEPORT_LB:
1475                                 INP_WLOCK(in6p);
1476                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1477                                         in6p->inp_flags2 |= INP_REUSEPORT_LB;
1478                                 else
1479                                         in6p->inp_flags2 &= ~INP_REUSEPORT_LB;
1480                                 INP_WUNLOCK(in6p);
1481                                 error = 0;
1482                                 break;
1483                         case SO_SETFIB:
1484                                 INP_WLOCK(in6p);
1485                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1486                                 INP_WUNLOCK(in6p);
1487                                 error = 0;
1488                                 break;
1489                         case SO_MAX_PACING_RATE:
1490 #ifdef RATELIMIT
1491                                 INP_WLOCK(in6p);
1492                                 in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1493                                 INP_WUNLOCK(in6p);
1494                                 error = 0;
1495 #else
1496                                 error = EOPNOTSUPP;
1497 #endif
1498                                 break;
1499                         default:
1500                                 break;
1501                         }
1502                 }
1503         } else {                /* level == IPPROTO_IPV6 */
1504                 switch (op) {
1505
1506                 case SOPT_SET:
1507                         switch (optname) {
1508                         case IPV6_2292PKTOPTIONS:
1509 #ifdef IPV6_PKTOPTIONS
1510                         case IPV6_PKTOPTIONS:
1511 #endif
1512                         {
1513                                 struct mbuf *m;
1514
1515                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1516                                         printf("ip6_ctloutput: mbuf limit hit\n");
1517                                         error = ENOBUFS;
1518                                         break;
1519                                 }
1520
1521                                 error = soopt_getm(sopt, &m); /* XXX */
1522                                 if (error != 0)
1523                                         break;
1524                                 error = soopt_mcopyin(sopt, m); /* XXX */
1525                                 if (error != 0)
1526                                         break;
1527                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1528                                                     m, so, sopt);
1529                                 m_freem(m); /* XXX */
1530                                 break;
1531                         }
1532
1533                         /*
1534                          * Use of some Hop-by-Hop options or some
1535                          * Destination options, might require special
1536                          * privilege.  That is, normal applications
1537                          * (without special privilege) might be forbidden
1538                          * from setting certain options in outgoing packets,
1539                          * and might never see certain options in received
1540                          * packets. [RFC 2292 Section 6]
1541                          * KAME specific note:
1542                          *  KAME prevents non-privileged users from sending or
1543                          *  receiving ANY hbh/dst options in order to avoid
1544                          *  overhead of parsing options in the kernel.
1545                          */
1546                         case IPV6_RECVHOPOPTS:
1547                         case IPV6_RECVDSTOPTS:
1548                         case IPV6_RECVRTHDRDSTOPTS:
1549                                 if (td != NULL) {
1550                                         error = priv_check(td,
1551                                             PRIV_NETINET_SETHDROPTS);
1552                                         if (error)
1553                                                 break;
1554                                 }
1555                                 /* FALLTHROUGH */
1556                         case IPV6_UNICAST_HOPS:
1557                         case IPV6_HOPLIMIT:
1558
1559                         case IPV6_RECVPKTINFO:
1560                         case IPV6_RECVHOPLIMIT:
1561                         case IPV6_RECVRTHDR:
1562                         case IPV6_RECVPATHMTU:
1563                         case IPV6_RECVTCLASS:
1564                         case IPV6_RECVFLOWID:
1565 #ifdef  RSS
1566                         case IPV6_RECVRSSBUCKETID:
1567 #endif
1568                         case IPV6_V6ONLY:
1569                         case IPV6_AUTOFLOWLABEL:
1570                         case IPV6_ORIGDSTADDR:
1571                         case IPV6_BINDANY:
1572                         case IPV6_BINDMULTI:
1573 #ifdef  RSS
1574                         case IPV6_RSS_LISTEN_BUCKET:
1575 #endif
1576                                 if (optname == IPV6_BINDANY && td != NULL) {
1577                                         error = priv_check(td,
1578                                             PRIV_NETINET_BINDANY);
1579                                         if (error)
1580                                                 break;
1581                                 }
1582
1583                                 if (optlen != sizeof(int)) {
1584                                         error = EINVAL;
1585                                         break;
1586                                 }
1587                                 error = sooptcopyin(sopt, &optval,
1588                                         sizeof optval, sizeof optval);
1589                                 if (error)
1590                                         break;
1591                                 switch (optname) {
1592
1593                                 case IPV6_UNICAST_HOPS:
1594                                         if (optval < -1 || optval >= 256)
1595                                                 error = EINVAL;
1596                                         else {
1597                                                 /* -1 = kernel default */
1598                                                 in6p->in6p_hops = optval;
1599                                                 if ((in6p->inp_vflag &
1600                                                      INP_IPV4) != 0)
1601                                                         in6p->inp_ip_ttl = optval;
1602                                         }
1603                                         break;
1604 #define OPTSET(bit) \
1605 do { \
1606         INP_WLOCK(in6p); \
1607         if (optval) \
1608                 in6p->inp_flags |= (bit); \
1609         else \
1610                 in6p->inp_flags &= ~(bit); \
1611         INP_WUNLOCK(in6p); \
1612 } while (/*CONSTCOND*/ 0)
1613 #define OPTSET2292(bit) \
1614 do { \
1615         INP_WLOCK(in6p); \
1616         in6p->inp_flags |= IN6P_RFC2292; \
1617         if (optval) \
1618                 in6p->inp_flags |= (bit); \
1619         else \
1620                 in6p->inp_flags &= ~(bit); \
1621         INP_WUNLOCK(in6p); \
1622 } while (/*CONSTCOND*/ 0)
1623 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1624
1625 #define OPTSET2_N(bit, val) do {                                        \
1626         if (val)                                                        \
1627                 in6p->inp_flags2 |= bit;                                \
1628         else                                                            \
1629                 in6p->inp_flags2 &= ~bit;                               \
1630 } while (0)
1631 #define OPTSET2(bit, val) do {                                          \
1632         INP_WLOCK(in6p);                                                \
1633         OPTSET2_N(bit, val);                                            \
1634         INP_WUNLOCK(in6p);                                              \
1635 } while (0)
1636 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1637 #define OPTSET2292_EXCLUSIVE(bit)                                       \
1638 do {                                                                    \
1639         INP_WLOCK(in6p);                                                \
1640         if (OPTBIT(IN6P_RFC2292)) {                                     \
1641                 error = EINVAL;                                         \
1642         } else {                                                        \
1643                 if (optval)                                             \
1644                         in6p->inp_flags |= (bit);                       \
1645                 else                                                    \
1646                         in6p->inp_flags &= ~(bit);                      \
1647         }                                                               \
1648         INP_WUNLOCK(in6p);                                              \
1649 } while (/*CONSTCOND*/ 0)
1650
1651                                 case IPV6_RECVPKTINFO:
1652                                         OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1653                                         break;
1654
1655                                 case IPV6_HOPLIMIT:
1656                                 {
1657                                         struct ip6_pktopts **optp;
1658
1659                                         /* cannot mix with RFC2292 */
1660                                         if (OPTBIT(IN6P_RFC2292)) {
1661                                                 error = EINVAL;
1662                                                 break;
1663                                         }
1664                                         INP_WLOCK(in6p);
1665                                         if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1666                                                 INP_WUNLOCK(in6p);
1667                                                 return (ECONNRESET);
1668                                         }
1669                                         optp = &in6p->in6p_outputopts;
1670                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1671                                             (u_char *)&optval, sizeof(optval),
1672                                             optp, (td != NULL) ? td->td_ucred :
1673                                             NULL, uproto);
1674                                         INP_WUNLOCK(in6p);
1675                                         break;
1676                                 }
1677
1678                                 case IPV6_RECVHOPLIMIT:
1679                                         OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1680                                         break;
1681
1682                                 case IPV6_RECVHOPOPTS:
1683                                         OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1684                                         break;
1685
1686                                 case IPV6_RECVDSTOPTS:
1687                                         OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1688                                         break;
1689
1690                                 case IPV6_RECVRTHDRDSTOPTS:
1691                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1692                                         break;
1693
1694                                 case IPV6_RECVRTHDR:
1695                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1696                                         break;
1697
1698                                 case IPV6_RECVPATHMTU:
1699                                         /*
1700                                          * We ignore this option for TCP
1701                                          * sockets.
1702                                          * (RFC3542 leaves this case
1703                                          * unspecified.)
1704                                          */
1705                                         if (uproto != IPPROTO_TCP)
1706                                                 OPTSET(IN6P_MTU);
1707                                         break;
1708
1709                                 case IPV6_RECVFLOWID:
1710                                         OPTSET2(INP_RECVFLOWID, optval);
1711                                         break;
1712
1713 #ifdef  RSS
1714                                 case IPV6_RECVRSSBUCKETID:
1715                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1716                                         break;
1717 #endif
1718
1719                                 case IPV6_V6ONLY:
1720                                         /*
1721                                          * make setsockopt(IPV6_V6ONLY)
1722                                          * available only prior to bind(2).
1723                                          * see ipng mailing list, Jun 22 2001.
1724                                          */
1725                                         if (in6p->inp_lport ||
1726                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1727                                                 error = EINVAL;
1728                                                 break;
1729                                         }
1730                                         OPTSET(IN6P_IPV6_V6ONLY);
1731                                         if (optval)
1732                                                 in6p->inp_vflag &= ~INP_IPV4;
1733                                         else
1734                                                 in6p->inp_vflag |= INP_IPV4;
1735                                         break;
1736                                 case IPV6_RECVTCLASS:
1737                                         /* cannot mix with RFC2292 XXX */
1738                                         OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1739                                         break;
1740                                 case IPV6_AUTOFLOWLABEL:
1741                                         OPTSET(IN6P_AUTOFLOWLABEL);
1742                                         break;
1743
1744                                 case IPV6_ORIGDSTADDR:
1745                                         OPTSET2(INP_ORIGDSTADDR, optval);
1746                                         break;
1747                                 case IPV6_BINDANY:
1748                                         OPTSET(INP_BINDANY);
1749                                         break;
1750
1751                                 case IPV6_BINDMULTI:
1752                                         OPTSET2(INP_BINDMULTI, optval);
1753                                         break;
1754 #ifdef  RSS
1755                                 case IPV6_RSS_LISTEN_BUCKET:
1756                                         if ((optval >= 0) &&
1757                                             (optval < rss_getnumbuckets())) {
1758                                                 INP_WLOCK(in6p);
1759                                                 in6p->inp_rss_listen_bucket = optval;
1760                                                 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1761                                                 INP_WUNLOCK(in6p);
1762                                         } else {
1763                                                 error = EINVAL;
1764                                         }
1765                                         break;
1766 #endif
1767                                 }
1768                                 break;
1769
1770                         case IPV6_TCLASS:
1771                         case IPV6_DONTFRAG:
1772                         case IPV6_USE_MIN_MTU:
1773                         case IPV6_PREFER_TEMPADDR:
1774                                 if (optlen != sizeof(optval)) {
1775                                         error = EINVAL;
1776                                         break;
1777                                 }
1778                                 error = sooptcopyin(sopt, &optval,
1779                                         sizeof optval, sizeof optval);
1780                                 if (error)
1781                                         break;
1782                                 {
1783                                         struct ip6_pktopts **optp;
1784                                         INP_WLOCK(in6p);
1785                                         if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1786                                                 INP_WUNLOCK(in6p);
1787                                                 return (ECONNRESET);
1788                                         }
1789                                         optp = &in6p->in6p_outputopts;
1790                                         error = ip6_pcbopt(optname,
1791                                             (u_char *)&optval, sizeof(optval),
1792                                             optp, (td != NULL) ? td->td_ucred :
1793                                             NULL, uproto);
1794                                         INP_WUNLOCK(in6p);
1795                                         break;
1796                                 }
1797
1798                         case IPV6_2292PKTINFO:
1799                         case IPV6_2292HOPLIMIT:
1800                         case IPV6_2292HOPOPTS:
1801                         case IPV6_2292DSTOPTS:
1802                         case IPV6_2292RTHDR:
1803                                 /* RFC 2292 */
1804                                 if (optlen != sizeof(int)) {
1805                                         error = EINVAL;
1806                                         break;
1807                                 }
1808                                 error = sooptcopyin(sopt, &optval,
1809                                         sizeof optval, sizeof optval);
1810                                 if (error)
1811                                         break;
1812                                 switch (optname) {
1813                                 case IPV6_2292PKTINFO:
1814                                         OPTSET2292(IN6P_PKTINFO);
1815                                         break;
1816                                 case IPV6_2292HOPLIMIT:
1817                                         OPTSET2292(IN6P_HOPLIMIT);
1818                                         break;
1819                                 case IPV6_2292HOPOPTS:
1820                                         /*
1821                                          * Check super-user privilege.
1822                                          * See comments for IPV6_RECVHOPOPTS.
1823                                          */
1824                                         if (td != NULL) {
1825                                                 error = priv_check(td,
1826                                                     PRIV_NETINET_SETHDROPTS);
1827                                                 if (error)
1828                                                         return (error);
1829                                         }
1830                                         OPTSET2292(IN6P_HOPOPTS);
1831                                         break;
1832                                 case IPV6_2292DSTOPTS:
1833                                         if (td != NULL) {
1834                                                 error = priv_check(td,
1835                                                     PRIV_NETINET_SETHDROPTS);
1836                                                 if (error)
1837                                                         return (error);
1838                                         }
1839                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1840                                         break;
1841                                 case IPV6_2292RTHDR:
1842                                         OPTSET2292(IN6P_RTHDR);
1843                                         break;
1844                                 }
1845                                 break;
1846                         case IPV6_PKTINFO:
1847                         case IPV6_HOPOPTS:
1848                         case IPV6_RTHDR:
1849                         case IPV6_DSTOPTS:
1850                         case IPV6_RTHDRDSTOPTS:
1851                         case IPV6_NEXTHOP:
1852                         {
1853                                 /* new advanced API (RFC3542) */
1854                                 u_char *optbuf;
1855                                 u_char optbuf_storage[MCLBYTES];
1856                                 int optlen;
1857                                 struct ip6_pktopts **optp;
1858
1859                                 /* cannot mix with RFC2292 */
1860                                 if (OPTBIT(IN6P_RFC2292)) {
1861                                         error = EINVAL;
1862                                         break;
1863                                 }
1864
1865                                 /*
1866                                  * We only ensure valsize is not too large
1867                                  * here.  Further validation will be done
1868                                  * later.
1869                                  */
1870                                 error = sooptcopyin(sopt, optbuf_storage,
1871                                     sizeof(optbuf_storage), 0);
1872                                 if (error)
1873                                         break;
1874                                 optlen = sopt->sopt_valsize;
1875                                 optbuf = optbuf_storage;
1876                                 INP_WLOCK(in6p);
1877                                 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1878                                         INP_WUNLOCK(in6p);
1879                                         return (ECONNRESET);
1880                                 }
1881                                 optp = &in6p->in6p_outputopts;
1882                                 error = ip6_pcbopt(optname, optbuf, optlen,
1883                                     optp, (td != NULL) ? td->td_ucred : NULL,
1884                                     uproto);
1885                                 INP_WUNLOCK(in6p);
1886                                 break;
1887                         }
1888 #undef OPTSET
1889
1890                         case IPV6_MULTICAST_IF:
1891                         case IPV6_MULTICAST_HOPS:
1892                         case IPV6_MULTICAST_LOOP:
1893                         case IPV6_JOIN_GROUP:
1894                         case IPV6_LEAVE_GROUP:
1895                         case IPV6_MSFILTER:
1896                         case MCAST_BLOCK_SOURCE:
1897                         case MCAST_UNBLOCK_SOURCE:
1898                         case MCAST_JOIN_GROUP:
1899                         case MCAST_LEAVE_GROUP:
1900                         case MCAST_JOIN_SOURCE_GROUP:
1901                         case MCAST_LEAVE_SOURCE_GROUP:
1902                                 error = ip6_setmoptions(in6p, sopt);
1903                                 break;
1904
1905                         case IPV6_PORTRANGE:
1906                                 error = sooptcopyin(sopt, &optval,
1907                                     sizeof optval, sizeof optval);
1908                                 if (error)
1909                                         break;
1910
1911                                 INP_WLOCK(in6p);
1912                                 switch (optval) {
1913                                 case IPV6_PORTRANGE_DEFAULT:
1914                                         in6p->inp_flags &= ~(INP_LOWPORT);
1915                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1916                                         break;
1917
1918                                 case IPV6_PORTRANGE_HIGH:
1919                                         in6p->inp_flags &= ~(INP_LOWPORT);
1920                                         in6p->inp_flags |= INP_HIGHPORT;
1921                                         break;
1922
1923                                 case IPV6_PORTRANGE_LOW:
1924                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1925                                         in6p->inp_flags |= INP_LOWPORT;
1926                                         break;
1927
1928                                 default:
1929                                         error = EINVAL;
1930                                         break;
1931                                 }
1932                                 INP_WUNLOCK(in6p);
1933                                 break;
1934
1935 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1936                         case IPV6_IPSEC_POLICY:
1937                                 if (IPSEC_ENABLED(ipv6)) {
1938                                         error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1939                                         break;
1940                                 }
1941                                 /* FALLTHROUGH */
1942 #endif /* IPSEC */
1943
1944                         default:
1945                                 error = ENOPROTOOPT;
1946                                 break;
1947                         }
1948                         break;
1949
1950                 case SOPT_GET:
1951                         switch (optname) {
1952
1953                         case IPV6_2292PKTOPTIONS:
1954 #ifdef IPV6_PKTOPTIONS
1955                         case IPV6_PKTOPTIONS:
1956 #endif
1957                                 /*
1958                                  * RFC3542 (effectively) deprecated the
1959                                  * semantics of the 2292-style pktoptions.
1960                                  * Since it was not reliable in nature (i.e.,
1961                                  * applications had to expect the lack of some
1962                                  * information after all), it would make sense
1963                                  * to simplify this part by always returning
1964                                  * empty data.
1965                                  */
1966                                 sopt->sopt_valsize = 0;
1967                                 break;
1968
1969                         case IPV6_RECVHOPOPTS:
1970                         case IPV6_RECVDSTOPTS:
1971                         case IPV6_RECVRTHDRDSTOPTS:
1972                         case IPV6_UNICAST_HOPS:
1973                         case IPV6_RECVPKTINFO:
1974                         case IPV6_RECVHOPLIMIT:
1975                         case IPV6_RECVRTHDR:
1976                         case IPV6_RECVPATHMTU:
1977
1978                         case IPV6_V6ONLY:
1979                         case IPV6_PORTRANGE:
1980                         case IPV6_RECVTCLASS:
1981                         case IPV6_AUTOFLOWLABEL:
1982                         case IPV6_BINDANY:
1983                         case IPV6_FLOWID:
1984                         case IPV6_FLOWTYPE:
1985                         case IPV6_RECVFLOWID:
1986 #ifdef  RSS
1987                         case IPV6_RSSBUCKETID:
1988                         case IPV6_RECVRSSBUCKETID:
1989 #endif
1990                         case IPV6_BINDMULTI:
1991                                 switch (optname) {
1992
1993                                 case IPV6_RECVHOPOPTS:
1994                                         optval = OPTBIT(IN6P_HOPOPTS);
1995                                         break;
1996
1997                                 case IPV6_RECVDSTOPTS:
1998                                         optval = OPTBIT(IN6P_DSTOPTS);
1999                                         break;
2000
2001                                 case IPV6_RECVRTHDRDSTOPTS:
2002                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2003                                         break;
2004
2005                                 case IPV6_UNICAST_HOPS:
2006                                         optval = in6p->in6p_hops;
2007                                         break;
2008
2009                                 case IPV6_RECVPKTINFO:
2010                                         optval = OPTBIT(IN6P_PKTINFO);
2011                                         break;
2012
2013                                 case IPV6_RECVHOPLIMIT:
2014                                         optval = OPTBIT(IN6P_HOPLIMIT);
2015                                         break;
2016
2017                                 case IPV6_RECVRTHDR:
2018                                         optval = OPTBIT(IN6P_RTHDR);
2019                                         break;
2020
2021                                 case IPV6_RECVPATHMTU:
2022                                         optval = OPTBIT(IN6P_MTU);
2023                                         break;
2024
2025                                 case IPV6_V6ONLY:
2026                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2027                                         break;
2028
2029                                 case IPV6_PORTRANGE:
2030                                     {
2031                                         int flags;
2032                                         flags = in6p->inp_flags;
2033                                         if (flags & INP_HIGHPORT)
2034                                                 optval = IPV6_PORTRANGE_HIGH;
2035                                         else if (flags & INP_LOWPORT)
2036                                                 optval = IPV6_PORTRANGE_LOW;
2037                                         else
2038                                                 optval = 0;
2039                                         break;
2040                                     }
2041                                 case IPV6_RECVTCLASS:
2042                                         optval = OPTBIT(IN6P_TCLASS);
2043                                         break;
2044
2045                                 case IPV6_AUTOFLOWLABEL:
2046                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2047                                         break;
2048
2049                                 case IPV6_ORIGDSTADDR:
2050                                         optval = OPTBIT2(INP_ORIGDSTADDR);
2051                                         break;
2052
2053                                 case IPV6_BINDANY:
2054                                         optval = OPTBIT(INP_BINDANY);
2055                                         break;
2056
2057                                 case IPV6_FLOWID:
2058                                         optval = in6p->inp_flowid;
2059                                         break;
2060
2061                                 case IPV6_FLOWTYPE:
2062                                         optval = in6p->inp_flowtype;
2063                                         break;
2064
2065                                 case IPV6_RECVFLOWID:
2066                                         optval = OPTBIT2(INP_RECVFLOWID);
2067                                         break;
2068 #ifdef  RSS
2069                                 case IPV6_RSSBUCKETID:
2070                                         retval =
2071                                             rss_hash2bucket(in6p->inp_flowid,
2072                                             in6p->inp_flowtype,
2073                                             &rss_bucket);
2074                                         if (retval == 0)
2075                                                 optval = rss_bucket;
2076                                         else
2077                                                 error = EINVAL;
2078                                         break;
2079
2080                                 case IPV6_RECVRSSBUCKETID:
2081                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2082                                         break;
2083 #endif
2084
2085                                 case IPV6_BINDMULTI:
2086                                         optval = OPTBIT2(INP_BINDMULTI);
2087                                         break;
2088
2089                                 }
2090                                 if (error)
2091                                         break;
2092                                 error = sooptcopyout(sopt, &optval,
2093                                         sizeof optval);
2094                                 break;
2095
2096                         case IPV6_PATHMTU:
2097                         {
2098                                 u_long pmtu = 0;
2099                                 struct ip6_mtuinfo mtuinfo;
2100                                 struct in6_addr addr;
2101
2102                                 if (!(so->so_state & SS_ISCONNECTED))
2103                                         return (ENOTCONN);
2104                                 /*
2105                                  * XXX: we dot not consider the case of source
2106                                  * routing, or optional information to specify
2107                                  * the outgoing interface.
2108                                  * Copy faddr out of in6p to avoid holding lock
2109                                  * on inp during route lookup.
2110                                  */
2111                                 INP_RLOCK(in6p);
2112                                 bcopy(&in6p->in6p_faddr, &addr, sizeof(addr));
2113                                 INP_RUNLOCK(in6p);
2114                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2115                                     &addr, &pmtu);
2116                                 if (error)
2117                                         break;
2118                                 if (pmtu > IPV6_MAXPACKET)
2119                                         pmtu = IPV6_MAXPACKET;
2120
2121                                 bzero(&mtuinfo, sizeof(mtuinfo));
2122                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2123                                 optdata = (void *)&mtuinfo;
2124                                 optdatalen = sizeof(mtuinfo);
2125                                 error = sooptcopyout(sopt, optdata,
2126                                     optdatalen);
2127                                 break;
2128                         }
2129
2130                         case IPV6_2292PKTINFO:
2131                         case IPV6_2292HOPLIMIT:
2132                         case IPV6_2292HOPOPTS:
2133                         case IPV6_2292RTHDR:
2134                         case IPV6_2292DSTOPTS:
2135                                 switch (optname) {
2136                                 case IPV6_2292PKTINFO:
2137                                         optval = OPTBIT(IN6P_PKTINFO);
2138                                         break;
2139                                 case IPV6_2292HOPLIMIT:
2140                                         optval = OPTBIT(IN6P_HOPLIMIT);
2141                                         break;
2142                                 case IPV6_2292HOPOPTS:
2143                                         optval = OPTBIT(IN6P_HOPOPTS);
2144                                         break;
2145                                 case IPV6_2292RTHDR:
2146                                         optval = OPTBIT(IN6P_RTHDR);
2147                                         break;
2148                                 case IPV6_2292DSTOPTS:
2149                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2150                                         break;
2151                                 }
2152                                 error = sooptcopyout(sopt, &optval,
2153                                     sizeof optval);
2154                                 break;
2155                         case IPV6_PKTINFO:
2156                         case IPV6_HOPOPTS:
2157                         case IPV6_RTHDR:
2158                         case IPV6_DSTOPTS:
2159                         case IPV6_RTHDRDSTOPTS:
2160                         case IPV6_NEXTHOP:
2161                         case IPV6_TCLASS:
2162                         case IPV6_DONTFRAG:
2163                         case IPV6_USE_MIN_MTU:
2164                         case IPV6_PREFER_TEMPADDR:
2165                                 error = ip6_getpcbopt(in6p, optname, sopt);
2166                                 break;
2167
2168                         case IPV6_MULTICAST_IF:
2169                         case IPV6_MULTICAST_HOPS:
2170                         case IPV6_MULTICAST_LOOP:
2171                         case IPV6_MSFILTER:
2172                                 error = ip6_getmoptions(in6p, sopt);
2173                                 break;
2174
2175 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2176                         case IPV6_IPSEC_POLICY:
2177                                 if (IPSEC_ENABLED(ipv6)) {
2178                                         error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2179                                         break;
2180                                 }
2181                                 /* FALLTHROUGH */
2182 #endif /* IPSEC */
2183                         default:
2184                                 error = ENOPROTOOPT;
2185                                 break;
2186                         }
2187                         break;
2188                 }
2189         }
2190         return (error);
2191 }
2192
2193 int
2194 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2195 {
2196         int error = 0, optval, optlen;
2197         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2198         struct inpcb *in6p = sotoinpcb(so);
2199         int level, op, optname;
2200
2201         level = sopt->sopt_level;
2202         op = sopt->sopt_dir;
2203         optname = sopt->sopt_name;
2204         optlen = sopt->sopt_valsize;
2205
2206         if (level != IPPROTO_IPV6) {
2207                 return (EINVAL);
2208         }
2209
2210         switch (optname) {
2211         case IPV6_CHECKSUM:
2212                 /*
2213                  * For ICMPv6 sockets, no modification allowed for checksum
2214                  * offset, permit "no change" values to help existing apps.
2215                  *
2216                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2217                  * for an ICMPv6 socket will fail."
2218                  * The current behavior does not meet RFC3542.
2219                  */
2220                 switch (op) {
2221                 case SOPT_SET:
2222                         if (optlen != sizeof(int)) {
2223                                 error = EINVAL;
2224                                 break;
2225                         }
2226                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2227                                             sizeof(optval));
2228                         if (error)
2229                                 break;
2230                         if (optval < -1 || (optval % 2) != 0) {
2231                                 /*
2232                                  * The API assumes non-negative even offset
2233                                  * values or -1 as a special value.
2234                                  */
2235                                 error = EINVAL;
2236                         } else if (so->so_proto->pr_protocol ==
2237                             IPPROTO_ICMPV6) {
2238                                 if (optval != icmp6off)
2239                                         error = EINVAL;
2240                         } else
2241                                 in6p->in6p_cksum = optval;
2242                         break;
2243
2244                 case SOPT_GET:
2245                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2246                                 optval = icmp6off;
2247                         else
2248                                 optval = in6p->in6p_cksum;
2249
2250                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2251                         break;
2252
2253                 default:
2254                         error = EINVAL;
2255                         break;
2256                 }
2257                 break;
2258
2259         default:
2260                 error = ENOPROTOOPT;
2261                 break;
2262         }
2263
2264         return (error);
2265 }
2266
2267 /*
2268  * Set up IP6 options in pcb for insertion in output packets or
2269  * specifying behavior of outgoing packets.
2270  */
2271 static int
2272 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2273     struct socket *so, struct sockopt *sopt)
2274 {
2275         struct ip6_pktopts *opt = *pktopt;
2276         int error = 0;
2277         struct thread *td = sopt->sopt_td;
2278
2279         /* turn off any old options. */
2280         if (opt) {
2281 #ifdef DIAGNOSTIC
2282                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2283                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2284                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2285                         printf("ip6_pcbopts: all specified options are cleared.\n");
2286 #endif
2287                 ip6_clearpktopts(opt, -1);
2288         } else
2289                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2290         *pktopt = NULL;
2291
2292         if (!m || m->m_len == 0) {
2293                 /*
2294                  * Only turning off any previous options, regardless of
2295                  * whether the opt is just created or given.
2296                  */
2297                 free(opt, M_IP6OPT);
2298                 return (0);
2299         }
2300
2301         /*  set options specified by user. */
2302         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2303             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2304                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2305                 free(opt, M_IP6OPT);
2306                 return (error);
2307         }
2308         *pktopt = opt;
2309         return (0);
2310 }
2311
2312 /*
2313  * initialize ip6_pktopts.  beware that there are non-zero default values in
2314  * the struct.
2315  */
2316 void
2317 ip6_initpktopts(struct ip6_pktopts *opt)
2318 {
2319
2320         bzero(opt, sizeof(*opt));
2321         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2322         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2323         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2324         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2325 }
2326
2327 static int
2328 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2329     struct ucred *cred, int uproto)
2330 {
2331         struct ip6_pktopts *opt;
2332
2333         if (*pktopt == NULL) {
2334                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2335                     M_NOWAIT);
2336                 if (*pktopt == NULL)
2337                         return (ENOBUFS);
2338                 ip6_initpktopts(*pktopt);
2339         }
2340         opt = *pktopt;
2341
2342         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2343 }
2344
2345 #define GET_PKTOPT_VAR(field, lenexpr) do {                                     \
2346         if (pktopt && pktopt->field) {                                          \
2347                 INP_RUNLOCK(in6p);                                              \
2348                 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);         \
2349                 malloc_optdata = true;                                          \
2350                 INP_RLOCK(in6p);                                                \
2351                 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {           \
2352                         INP_RUNLOCK(in6p);                                      \
2353                         free(optdata, M_TEMP);                                  \
2354                         return (ECONNRESET);                                    \
2355                 }                                                               \
2356                 pktopt = in6p->in6p_outputopts;                                 \
2357                 if (pktopt && pktopt->field) {                                  \
2358                         optdatalen = min(lenexpr, sopt->sopt_valsize);          \
2359                         bcopy(&pktopt->field, optdata, optdatalen);             \
2360                 } else {                                                        \
2361                         free(optdata, M_TEMP);                                  \
2362                         optdata = NULL;                                         \
2363                         malloc_optdata = false;                                 \
2364                 }                                                               \
2365         }                                                                       \
2366 } while(0)
2367
2368 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,                         \
2369         (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2370
2371 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,                        \
2372         pktopt->field->sa_len)
2373
2374 static int
2375 ip6_getpcbopt(struct inpcb *in6p, int optname, struct sockopt *sopt)
2376 {
2377         void *optdata = NULL;
2378         bool malloc_optdata = false;
2379         int optdatalen = 0;
2380         int error = 0;
2381         struct in6_pktinfo null_pktinfo;
2382         int deftclass = 0, on;
2383         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2384         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2385         struct ip6_pktopts *pktopt;
2386
2387         INP_RLOCK(in6p);
2388         pktopt = in6p->in6p_outputopts;
2389
2390         switch (optname) {
2391         case IPV6_PKTINFO:
2392                 optdata = (void *)&null_pktinfo;
2393                 if (pktopt && pktopt->ip6po_pktinfo) {
2394                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2395                             sizeof(null_pktinfo));
2396                         in6_clearscope(&null_pktinfo.ipi6_addr);
2397                 } else {
2398                         /* XXX: we don't have to do this every time... */
2399                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2400                 }
2401                 optdatalen = sizeof(struct in6_pktinfo);
2402                 break;
2403         case IPV6_TCLASS:
2404                 if (pktopt && pktopt->ip6po_tclass >= 0)
2405                         deftclass = pktopt->ip6po_tclass;
2406                 optdata = (void *)&deftclass;
2407                 optdatalen = sizeof(int);
2408                 break;
2409         case IPV6_HOPOPTS:
2410                 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2411                 break;
2412         case IPV6_RTHDR:
2413                 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2414                 break;
2415         case IPV6_RTHDRDSTOPTS:
2416                 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2417                 break;
2418         case IPV6_DSTOPTS:
2419                 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2420                 break;
2421         case IPV6_NEXTHOP:
2422                 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2423                 break;
2424         case IPV6_USE_MIN_MTU:
2425                 if (pktopt)
2426                         defminmtu = pktopt->ip6po_minmtu;
2427                 optdata = (void *)&defminmtu;
2428                 optdatalen = sizeof(int);
2429                 break;
2430         case IPV6_DONTFRAG:
2431                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2432                         on = 1;
2433                 else
2434                         on = 0;
2435                 optdata = (void *)&on;
2436                 optdatalen = sizeof(on);
2437                 break;
2438         case IPV6_PREFER_TEMPADDR:
2439                 if (pktopt)
2440                         defpreftemp = pktopt->ip6po_prefer_tempaddr;
2441                 optdata = (void *)&defpreftemp;
2442                 optdatalen = sizeof(int);
2443                 break;
2444         default:                /* should not happen */
2445 #ifdef DIAGNOSTIC
2446                 panic("ip6_getpcbopt: unexpected option\n");
2447 #endif
2448                 INP_RUNLOCK(in6p);
2449                 return (ENOPROTOOPT);
2450         }
2451         INP_RUNLOCK(in6p);
2452
2453         error = sooptcopyout(sopt, optdata, optdatalen);
2454         if (malloc_optdata)
2455                 free(optdata, M_TEMP);
2456
2457         return (error);
2458 }
2459
2460 void
2461 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2462 {
2463         if (pktopt == NULL)
2464                 return;
2465
2466         if (optname == -1 || optname == IPV6_PKTINFO) {
2467                 if (pktopt->ip6po_pktinfo)
2468                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2469                 pktopt->ip6po_pktinfo = NULL;
2470         }
2471         if (optname == -1 || optname == IPV6_HOPLIMIT)
2472                 pktopt->ip6po_hlim = -1;
2473         if (optname == -1 || optname == IPV6_TCLASS)
2474                 pktopt->ip6po_tclass = -1;
2475         if (optname == -1 || optname == IPV6_NEXTHOP) {
2476                 if (pktopt->ip6po_nextroute.ro_rt) {
2477                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2478                         pktopt->ip6po_nextroute.ro_rt = NULL;
2479                 }
2480                 if (pktopt->ip6po_nexthop)
2481                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2482                 pktopt->ip6po_nexthop = NULL;
2483         }
2484         if (optname == -1 || optname == IPV6_HOPOPTS) {
2485                 if (pktopt->ip6po_hbh)
2486                         free(pktopt->ip6po_hbh, M_IP6OPT);
2487                 pktopt->ip6po_hbh = NULL;
2488         }
2489         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2490                 if (pktopt->ip6po_dest1)
2491                         free(pktopt->ip6po_dest1, M_IP6OPT);
2492                 pktopt->ip6po_dest1 = NULL;
2493         }
2494         if (optname == -1 || optname == IPV6_RTHDR) {
2495                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2496                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2497                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2498                 if (pktopt->ip6po_route.ro_rt) {
2499                         RTFREE(pktopt->ip6po_route.ro_rt);
2500                         pktopt->ip6po_route.ro_rt = NULL;
2501                 }
2502         }
2503         if (optname == -1 || optname == IPV6_DSTOPTS) {
2504                 if (pktopt->ip6po_dest2)
2505                         free(pktopt->ip6po_dest2, M_IP6OPT);
2506                 pktopt->ip6po_dest2 = NULL;
2507         }
2508 }
2509
2510 #define PKTOPT_EXTHDRCPY(type) \
2511 do {\
2512         if (src->type) {\
2513                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2514                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2515                 if (dst->type == NULL)\
2516                         goto bad;\
2517                 bcopy(src->type, dst->type, hlen);\
2518         }\
2519 } while (/*CONSTCOND*/ 0)
2520
2521 static int
2522 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2523 {
2524         if (dst == NULL || src == NULL)  {
2525                 printf("ip6_clearpktopts: invalid argument\n");
2526                 return (EINVAL);
2527         }
2528
2529         dst->ip6po_hlim = src->ip6po_hlim;
2530         dst->ip6po_tclass = src->ip6po_tclass;
2531         dst->ip6po_flags = src->ip6po_flags;
2532         dst->ip6po_minmtu = src->ip6po_minmtu;
2533         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2534         if (src->ip6po_pktinfo) {
2535                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2536                     M_IP6OPT, canwait);
2537                 if (dst->ip6po_pktinfo == NULL)
2538                         goto bad;
2539                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2540         }
2541         if (src->ip6po_nexthop) {
2542                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2543                     M_IP6OPT, canwait);
2544                 if (dst->ip6po_nexthop == NULL)
2545                         goto bad;
2546                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2547                     src->ip6po_nexthop->sa_len);
2548         }
2549         PKTOPT_EXTHDRCPY(ip6po_hbh);
2550         PKTOPT_EXTHDRCPY(ip6po_dest1);
2551         PKTOPT_EXTHDRCPY(ip6po_dest2);
2552         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2553         return (0);
2554
2555   bad:
2556         ip6_clearpktopts(dst, -1);
2557         return (ENOBUFS);
2558 }
2559 #undef PKTOPT_EXTHDRCPY
2560
2561 struct ip6_pktopts *
2562 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2563 {
2564         int error;
2565         struct ip6_pktopts *dst;
2566
2567         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2568         if (dst == NULL)
2569                 return (NULL);
2570         ip6_initpktopts(dst);
2571
2572         if ((error = copypktopts(dst, src, canwait)) != 0) {
2573                 free(dst, M_IP6OPT);
2574                 return (NULL);
2575         }
2576
2577         return (dst);
2578 }
2579
2580 void
2581 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2582 {
2583         if (pktopt == NULL)
2584                 return;
2585
2586         ip6_clearpktopts(pktopt, -1);
2587
2588         free(pktopt, M_IP6OPT);
2589 }
2590
2591 /*
2592  * Set IPv6 outgoing packet options based on advanced API.
2593  */
2594 int
2595 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2596     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2597 {
2598         struct cmsghdr *cm = NULL;
2599
2600         if (control == NULL || opt == NULL)
2601                 return (EINVAL);
2602
2603         ip6_initpktopts(opt);
2604         if (stickyopt) {
2605                 int error;
2606
2607                 /*
2608                  * If stickyopt is provided, make a local copy of the options
2609                  * for this particular packet, then override them by ancillary
2610                  * objects.
2611                  * XXX: copypktopts() does not copy the cached route to a next
2612                  * hop (if any).  This is not very good in terms of efficiency,
2613                  * but we can allow this since this option should be rarely
2614                  * used.
2615                  */
2616                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2617                         return (error);
2618         }
2619
2620         /*
2621          * XXX: Currently, we assume all the optional information is stored
2622          * in a single mbuf.
2623          */
2624         if (control->m_next)
2625                 return (EINVAL);
2626
2627         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2628             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2629                 int error;
2630
2631                 if (control->m_len < CMSG_LEN(0))
2632                         return (EINVAL);
2633
2634                 cm = mtod(control, struct cmsghdr *);
2635                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2636                         return (EINVAL);
2637                 if (cm->cmsg_level != IPPROTO_IPV6)
2638                         continue;
2639
2640                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2641                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2642                 if (error)
2643                         return (error);
2644         }
2645
2646         return (0);
2647 }
2648
2649 /*
2650  * Set a particular packet option, as a sticky option or an ancillary data
2651  * item.  "len" can be 0 only when it's a sticky option.
2652  * We have 4 cases of combination of "sticky" and "cmsg":
2653  * "sticky=0, cmsg=0": impossible
2654  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2655  * "sticky=1, cmsg=0": RFC3542 socket option
2656  * "sticky=1, cmsg=1": RFC2292 socket option
2657  */
2658 static int
2659 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2660     struct ucred *cred, int sticky, int cmsg, int uproto)
2661 {
2662         int minmtupolicy, preftemp;
2663         int error;
2664
2665         if (!sticky && !cmsg) {
2666 #ifdef DIAGNOSTIC
2667                 printf("ip6_setpktopt: impossible case\n");
2668 #endif
2669                 return (EINVAL);
2670         }
2671
2672         /*
2673          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2674          * not be specified in the context of RFC3542.  Conversely,
2675          * RFC3542 types should not be specified in the context of RFC2292.
2676          */
2677         if (!cmsg) {
2678                 switch (optname) {
2679                 case IPV6_2292PKTINFO:
2680                 case IPV6_2292HOPLIMIT:
2681                 case IPV6_2292NEXTHOP:
2682                 case IPV6_2292HOPOPTS:
2683                 case IPV6_2292DSTOPTS:
2684                 case IPV6_2292RTHDR:
2685                 case IPV6_2292PKTOPTIONS:
2686                         return (ENOPROTOOPT);
2687                 }
2688         }
2689         if (sticky && cmsg) {
2690                 switch (optname) {
2691                 case IPV6_PKTINFO:
2692                 case IPV6_HOPLIMIT:
2693                 case IPV6_NEXTHOP:
2694                 case IPV6_HOPOPTS:
2695                 case IPV6_DSTOPTS:
2696                 case IPV6_RTHDRDSTOPTS:
2697                 case IPV6_RTHDR:
2698                 case IPV6_USE_MIN_MTU:
2699                 case IPV6_DONTFRAG:
2700                 case IPV6_TCLASS:
2701                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2702                         return (ENOPROTOOPT);
2703                 }
2704         }
2705
2706         switch (optname) {
2707         case IPV6_2292PKTINFO:
2708         case IPV6_PKTINFO:
2709         {
2710                 struct ifnet *ifp = NULL;
2711                 struct in6_pktinfo *pktinfo;
2712
2713                 if (len != sizeof(struct in6_pktinfo))
2714                         return (EINVAL);
2715
2716                 pktinfo = (struct in6_pktinfo *)buf;
2717
2718                 /*
2719                  * An application can clear any sticky IPV6_PKTINFO option by
2720                  * doing a "regular" setsockopt with ipi6_addr being
2721                  * in6addr_any and ipi6_ifindex being zero.
2722                  * [RFC 3542, Section 6]
2723                  */
2724                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2725                     pktinfo->ipi6_ifindex == 0 &&
2726                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2727                         ip6_clearpktopts(opt, optname);
2728                         break;
2729                 }
2730
2731                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2732                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2733                         return (EINVAL);
2734                 }
2735                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2736                         return (EINVAL);
2737                 /* validate the interface index if specified. */
2738                 if (pktinfo->ipi6_ifindex > V_if_index)
2739                          return (ENXIO);
2740                 if (pktinfo->ipi6_ifindex) {
2741                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2742                         if (ifp == NULL)
2743                                 return (ENXIO);
2744                 }
2745                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2746                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2747                         return (ENETDOWN);
2748
2749                 if (ifp != NULL &&
2750                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2751                         struct in6_ifaddr *ia;
2752
2753                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2754                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2755                         if (ia == NULL)
2756                                 return (EADDRNOTAVAIL);
2757                         ifa_free(&ia->ia_ifa);
2758                 }
2759                 /*
2760                  * We store the address anyway, and let in6_selectsrc()
2761                  * validate the specified address.  This is because ipi6_addr
2762                  * may not have enough information about its scope zone, and
2763                  * we may need additional information (such as outgoing
2764                  * interface or the scope zone of a destination address) to
2765                  * disambiguate the scope.
2766                  * XXX: the delay of the validation may confuse the
2767                  * application when it is used as a sticky option.
2768                  */
2769                 if (opt->ip6po_pktinfo == NULL) {
2770                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2771                             M_IP6OPT, M_NOWAIT);
2772                         if (opt->ip6po_pktinfo == NULL)
2773                                 return (ENOBUFS);
2774                 }
2775                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2776                 break;
2777         }
2778
2779         case IPV6_2292HOPLIMIT:
2780         case IPV6_HOPLIMIT:
2781         {
2782                 int *hlimp;
2783
2784                 /*
2785                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2786                  * to simplify the ordering among hoplimit options.
2787                  */
2788                 if (optname == IPV6_HOPLIMIT && sticky)
2789                         return (ENOPROTOOPT);
2790
2791                 if (len != sizeof(int))
2792                         return (EINVAL);
2793                 hlimp = (int *)buf;
2794                 if (*hlimp < -1 || *hlimp > 255)
2795                         return (EINVAL);
2796
2797                 opt->ip6po_hlim = *hlimp;
2798                 break;
2799         }
2800
2801         case IPV6_TCLASS:
2802         {
2803                 int tclass;
2804
2805                 if (len != sizeof(int))
2806                         return (EINVAL);
2807                 tclass = *(int *)buf;
2808                 if (tclass < -1 || tclass > 255)
2809                         return (EINVAL);
2810
2811                 opt->ip6po_tclass = tclass;
2812                 break;
2813         }
2814
2815         case IPV6_2292NEXTHOP:
2816         case IPV6_NEXTHOP:
2817                 if (cred != NULL) {
2818                         error = priv_check_cred(cred,
2819                             PRIV_NETINET_SETHDROPTS, 0);
2820                         if (error)
2821                                 return (error);
2822                 }
2823
2824                 if (len == 0) { /* just remove the option */
2825                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2826                         break;
2827                 }
2828
2829                 /* check if cmsg_len is large enough for sa_len */
2830                 if (len < sizeof(struct sockaddr) || len < *buf)
2831                         return (EINVAL);
2832
2833                 switch (((struct sockaddr *)buf)->sa_family) {
2834                 case AF_INET6:
2835                 {
2836                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2837                         int error;
2838
2839                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2840                                 return (EINVAL);
2841
2842                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2843                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2844                                 return (EINVAL);
2845                         }
2846                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2847                             != 0) {
2848                                 return (error);
2849                         }
2850                         break;
2851                 }
2852                 case AF_LINK:   /* should eventually be supported */
2853                 default:
2854                         return (EAFNOSUPPORT);
2855                 }
2856
2857                 /* turn off the previous option, then set the new option. */
2858                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2859                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2860                 if (opt->ip6po_nexthop == NULL)
2861                         return (ENOBUFS);
2862                 bcopy(buf, opt->ip6po_nexthop, *buf);
2863                 break;
2864
2865         case IPV6_2292HOPOPTS:
2866         case IPV6_HOPOPTS:
2867         {
2868                 struct ip6_hbh *hbh;
2869                 int hbhlen;
2870
2871                 /*
2872                  * XXX: We don't allow a non-privileged user to set ANY HbH
2873                  * options, since per-option restriction has too much
2874                  * overhead.
2875                  */
2876                 if (cred != NULL) {
2877                         error = priv_check_cred(cred,
2878                             PRIV_NETINET_SETHDROPTS, 0);
2879                         if (error)
2880                                 return (error);
2881                 }
2882
2883                 if (len == 0) {
2884                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2885                         break;  /* just remove the option */
2886                 }
2887
2888                 /* message length validation */
2889                 if (len < sizeof(struct ip6_hbh))
2890                         return (EINVAL);
2891                 hbh = (struct ip6_hbh *)buf;
2892                 hbhlen = (hbh->ip6h_len + 1) << 3;
2893                 if (len != hbhlen)
2894                         return (EINVAL);
2895
2896                 /* turn off the previous option, then set the new option. */
2897                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2898                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2899                 if (opt->ip6po_hbh == NULL)
2900                         return (ENOBUFS);
2901                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2902
2903                 break;
2904         }
2905
2906         case IPV6_2292DSTOPTS:
2907         case IPV6_DSTOPTS:
2908         case IPV6_RTHDRDSTOPTS:
2909         {
2910                 struct ip6_dest *dest, **newdest = NULL;
2911                 int destlen;
2912
2913                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2914                         error = priv_check_cred(cred,
2915                             PRIV_NETINET_SETHDROPTS, 0);
2916                         if (error)
2917                                 return (error);
2918                 }
2919
2920                 if (len == 0) {
2921                         ip6_clearpktopts(opt, optname);
2922                         break;  /* just remove the option */
2923                 }
2924
2925                 /* message length validation */
2926                 if (len < sizeof(struct ip6_dest))
2927                         return (EINVAL);
2928                 dest = (struct ip6_dest *)buf;
2929                 destlen = (dest->ip6d_len + 1) << 3;
2930                 if (len != destlen)
2931                         return (EINVAL);
2932
2933                 /*
2934                  * Determine the position that the destination options header
2935                  * should be inserted; before or after the routing header.
2936                  */
2937                 switch (optname) {
2938                 case IPV6_2292DSTOPTS:
2939                         /*
2940                          * The old advacned API is ambiguous on this point.
2941                          * Our approach is to determine the position based
2942                          * according to the existence of a routing header.
2943                          * Note, however, that this depends on the order of the
2944                          * extension headers in the ancillary data; the 1st
2945                          * part of the destination options header must appear
2946                          * before the routing header in the ancillary data,
2947                          * too.
2948                          * RFC3542 solved the ambiguity by introducing
2949                          * separate ancillary data or option types.
2950                          */
2951                         if (opt->ip6po_rthdr == NULL)
2952                                 newdest = &opt->ip6po_dest1;
2953                         else
2954                                 newdest = &opt->ip6po_dest2;
2955                         break;
2956                 case IPV6_RTHDRDSTOPTS:
2957                         newdest = &opt->ip6po_dest1;
2958                         break;
2959                 case IPV6_DSTOPTS:
2960                         newdest = &opt->ip6po_dest2;
2961                         break;
2962                 }
2963
2964                 /* turn off the previous option, then set the new option. */
2965                 ip6_clearpktopts(opt, optname);
2966                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2967                 if (*newdest == NULL)
2968                         return (ENOBUFS);
2969                 bcopy(dest, *newdest, destlen);
2970
2971                 break;
2972         }
2973
2974         case IPV6_2292RTHDR:
2975         case IPV6_RTHDR:
2976         {
2977                 struct ip6_rthdr *rth;
2978                 int rthlen;
2979
2980                 if (len == 0) {
2981                         ip6_clearpktopts(opt, IPV6_RTHDR);
2982                         break;  /* just remove the option */
2983                 }
2984
2985                 /* message length validation */
2986                 if (len < sizeof(struct ip6_rthdr))
2987                         return (EINVAL);
2988                 rth = (struct ip6_rthdr *)buf;
2989                 rthlen = (rth->ip6r_len + 1) << 3;
2990                 if (len != rthlen)
2991                         return (EINVAL);
2992
2993                 switch (rth->ip6r_type) {
2994                 case IPV6_RTHDR_TYPE_0:
2995                         if (rth->ip6r_len == 0) /* must contain one addr */
2996                                 return (EINVAL);
2997                         if (rth->ip6r_len % 2) /* length must be even */
2998                                 return (EINVAL);
2999                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3000                                 return (EINVAL);
3001                         break;
3002                 default:
3003                         return (EINVAL);        /* not supported */
3004                 }
3005
3006                 /* turn off the previous option */
3007                 ip6_clearpktopts(opt, IPV6_RTHDR);
3008                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3009                 if (opt->ip6po_rthdr == NULL)
3010                         return (ENOBUFS);
3011                 bcopy(rth, opt->ip6po_rthdr, rthlen);
3012
3013                 break;
3014         }
3015
3016         case IPV6_USE_MIN_MTU:
3017                 if (len != sizeof(int))
3018                         return (EINVAL);
3019                 minmtupolicy = *(int *)buf;
3020                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3021                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
3022                     minmtupolicy != IP6PO_MINMTU_ALL) {
3023                         return (EINVAL);
3024                 }
3025                 opt->ip6po_minmtu = minmtupolicy;
3026                 break;
3027
3028         case IPV6_DONTFRAG:
3029                 if (len != sizeof(int))
3030                         return (EINVAL);
3031
3032                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3033                         /*
3034                          * we ignore this option for TCP sockets.
3035                          * (RFC3542 leaves this case unspecified.)
3036                          */
3037                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3038                 } else
3039                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3040                 break;
3041
3042         case IPV6_PREFER_TEMPADDR:
3043                 if (len != sizeof(int))
3044                         return (EINVAL);
3045                 preftemp = *(int *)buf;
3046                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3047                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3048                     preftemp != IP6PO_TEMPADDR_PREFER) {
3049                         return (EINVAL);
3050                 }
3051                 opt->ip6po_prefer_tempaddr = preftemp;
3052                 break;
3053
3054         default:
3055                 return (ENOPROTOOPT);
3056         } /* end of switch */
3057
3058         return (0);
3059 }
3060
3061 /*
3062  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3063  * packet to the input queue of a specified interface.  Note that this
3064  * calls the output routine of the loopback "driver", but with an interface
3065  * pointer that might NOT be &loif -- easier than replicating that code here.
3066  */
3067 void
3068 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3069 {
3070         struct mbuf *copym;
3071         struct ip6_hdr *ip6;
3072
3073         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3074         if (copym == NULL)
3075                 return;
3076
3077         /*
3078          * Make sure to deep-copy IPv6 header portion in case the data
3079          * is in an mbuf cluster, so that we can safely override the IPv6
3080          * header portion later.
3081          */
3082         if (!M_WRITABLE(copym) ||
3083             copym->m_len < sizeof(struct ip6_hdr)) {
3084                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3085                 if (copym == NULL)
3086                         return;
3087         }
3088         ip6 = mtod(copym, struct ip6_hdr *);
3089         /*
3090          * clear embedded scope identifiers if necessary.
3091          * in6_clearscope will touch the addresses only when necessary.
3092          */
3093         in6_clearscope(&ip6->ip6_src);
3094         in6_clearscope(&ip6->ip6_dst);
3095         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3096                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3097                     CSUM_PSEUDO_HDR;
3098                 copym->m_pkthdr.csum_data = 0xffff;
3099         }
3100         if_simloop(ifp, copym, AF_INET6, 0);
3101 }
3102
3103 /*
3104  * Chop IPv6 header off from the payload.
3105  */
3106 static int
3107 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3108 {
3109         struct mbuf *mh;
3110         struct ip6_hdr *ip6;
3111
3112         ip6 = mtod(m, struct ip6_hdr *);
3113         if (m->m_len > sizeof(*ip6)) {
3114                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3115                 if (mh == NULL) {
3116                         m_freem(m);
3117                         return ENOBUFS;
3118                 }
3119                 m_move_pkthdr(mh, m);
3120                 M_ALIGN(mh, sizeof(*ip6));
3121                 m->m_len -= sizeof(*ip6);
3122                 m->m_data += sizeof(*ip6);
3123                 mh->m_next = m;
3124                 m = mh;
3125                 m->m_len = sizeof(*ip6);
3126                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3127         }
3128         exthdrs->ip6e_ip6 = m;
3129         return 0;
3130 }
3131
3132 /*
3133  * Compute IPv6 extension header length.
3134  */
3135 int
3136 ip6_optlen(struct inpcb *in6p)
3137 {
3138         int len;
3139
3140         if (!in6p->in6p_outputopts)
3141                 return 0;
3142
3143         len = 0;
3144 #define elen(x) \
3145     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3146
3147         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3148         if (in6p->in6p_outputopts->ip6po_rthdr)
3149                 /* dest1 is valid with rthdr only */
3150                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3151         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3152         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3153         return len;
3154 #undef elen
3155 }