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