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_RTFREE(ro);
 808                         needfiblookup = 1; /* Redo the routing table lookup. */
 809                         if (ro->ro_lle)
 810                                 LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 811                         ro->ro_lle = NULL;
 812                 }
 813         }
 814         /* See if fib was changed by packet filter. */
 815         if (fibnum != M_GETFIB(m)) {
 816                 m->m_flags |= M_SKIP_FIREWALL;
 817                 fibnum = M_GETFIB(m);
 818                 RO_RTFREE(ro);
 819                 needfiblookup = 1;
 820                 if (ro->ro_lle)
 821                         LLE_FREE(ro->ro_lle);   /* zeros ro_lle */
 822                 ro->ro_lle = NULL;
 823         }
 824         if (needfiblookup)
 825                 goto again;
 826
 827         /* See if local, if yes, send it to netisr. */
 828         if (m->m_flags & M_FASTFWD_OURS) {
 829                 if (m->m_pkthdr.rcvif == NULL)
 830                         m->m_pkthdr.rcvif = V_loif;
 831                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 832                         m->m_pkthdr.csum_flags |=
 833                             CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
 834                         m->m_pkthdr.csum_data = 0xffff;
 835                 }
 836 #ifdef SCTP
 837                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
 838                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 839 #endif
 840                 error = netisr_queue(NETISR_IPV6, m);
 841                 goto done;
 842         }
 843         /* Or forward to some other address? */
 844         if ((m->m_flags & M_IP6_NEXTHOP) &&
 845             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 846                 dst = (struct sockaddr_in6 *)&ro->ro_dst;
 847                 bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
 848                 m->m_flags |= M_SKIP_FIREWALL;
 849                 m->m_flags &= ~M_IP6_NEXTHOP;
 850                 m_tag_delete(m, fwd_tag);
 851                 goto again;
 852         }
 853
 854 passout:
 855         /*
 856          * Send the packet to the outgoing interface.
 857          * If necessary, do IPv6 fragmentation before sending.
 858          *
 859          * the logic here is rather complex:
 860          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 861          * 1-a: send as is if tlen <= path mtu
 862          * 1-b: fragment if tlen > path mtu
 863          *
 864          * 2: if user asks us not to fragment (dontfrag == 1)
 865          * 2-a: send as is if tlen <= interface mtu
 866          * 2-b: error if tlen > interface mtu
 867          *
 868          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 869          *      always fragment
 870          *
 871          * 4: if dontfrag == 1 && alwaysfrag == 1
 872          *      error, as we cannot handle this conflicting request
 873          */
 874         sw_csum = m->m_pkthdr.csum_flags;
 875         if (!hdrsplit) {
 876                 tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
 877                 sw_csum &= ~ifp->if_hwassist;
 878         } else
 879                 tso = 0;
 880         /*
 881          * If we added extension headers, we will not do TSO and calculate the
 882          * checksums ourselves for now.
 883          * XXX-BZ  Need a framework to know when the NIC can handle it, even
 884          * with ext. hdrs.
 885          */
 886         if (sw_csum & CSUM_DELAY_DATA_IPV6) {
 887                 sw_csum &= ~CSUM_DELAY_DATA_IPV6;
 888                 in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
 889         }
 890 #ifdef SCTP
 891         if (sw_csum & CSUM_SCTP_IPV6) {
 892                 sw_csum &= ~CSUM_SCTP_IPV6;
 893                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
 894         }
 895 #endif
 896         m->m_pkthdr.csum_flags &= ifp->if_hwassist;
 897         tlen = m->m_pkthdr.len;
 898
 899         if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
 900                 dontfrag = 1;
 901         else
 902                 dontfrag = 0;
 903         if (dontfrag && alwaysfrag) {   /* case 4 */
 904                 /* conflicting request - can't transmit */
 905                 error = EMSGSIZE;
 906                 goto bad;
 907         }
 908         if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {      /* case 2-b */
 909                 /*
 910                  * Even if the DONTFRAG option is specified, we cannot send the
 911                  * packet when the data length is larger than the MTU of the
 912                  * outgoing interface.
 913                  * Notify the error by sending IPV6_PATHMTU ancillary data if
 914                  * application wanted to know the MTU value. Also return an
 915                  * error code (this is not described in the API spec).
 916                  */
 917                 if (inp != NULL)
 918                         ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
 919                 error = EMSGSIZE;
 920                 goto bad;
 921         }
 922
 923         /*
 924          * transmit packet without fragmentation
 925          */
 926         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 927                 struct in6_ifaddr *ia6;
 928
 929                 ip6 = mtod(m, struct ip6_hdr *);
 930                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 931                 if (ia6) {
 932                         /* Record statistics for this interface address. */
 933                         counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
 934                         counter_u64_add(ia6->ia_ifa.ifa_obytes,
 935                             m->m_pkthdr.len);
 936                         ifa_free(&ia6->ia_ifa);
 937                 }
 938 #ifdef RATELIMIT
 939                 if (inp != NULL) {
 940                         if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
 941                                 in_pcboutput_txrtlmt(inp, ifp, m);
 942                         /* stamp send tag on mbuf */
 943                         m->m_pkthdr.snd_tag = inp->inp_snd_tag;
 944                 } else {
 945                         m->m_pkthdr.snd_tag = NULL;
 946                 }
 947 #endif
 948                 error = nd6_output_ifp(ifp, origifp, m, dst,
 949                     (struct route *)ro);
 950 #ifdef RATELIMIT
 951                 /* check for route change */
 952                 if (error == EAGAIN)
 953                         in_pcboutput_eagain(inp);
 954 #endif
 955                 goto done;
 956         }
 957
 958         /*
 959          * try to fragment the packet.  case 1-b and 3
 960          */
 961         if (mtu < IPV6_MMTU) {
 962                 /* path MTU cannot be less than IPV6_MMTU */
 963                 error = EMSGSIZE;
 964                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 965                 goto bad;
 966         } else if (ip6->ip6_plen == 0) {
 967                 /* jumbo payload cannot be fragmented */
 968                 error = EMSGSIZE;
 969                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 970                 goto bad;
 971         } else {
 972                 u_char nextproto;
 973
 974                 /*
 975                  * Too large for the destination or interface;
 976                  * fragment if possible.
 977                  * Must be able to put at least 8 bytes per fragment.
 978                  */
 979                 hlen = unfragpartlen;
 980                 if (mtu > IPV6_MAXPACKET)
 981                         mtu = IPV6_MAXPACKET;
 982
 983                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 984                 if (len < 8) {
 985                         error = EMSGSIZE;
 986                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 987                         goto bad;
 988                 }
 989
 990                 /*
 991                  * If the interface will not calculate checksums on
 992                  * fragmented packets, then do it here.
 993                  * XXX-BZ handle the hw offloading case.  Need flags.
 994                  */
 995                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
 996                         in6_delayed_cksum(m, plen, hlen);
 997                         m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
 998                 }
 999 #ifdef SCTP
1000                 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1001                         sctp_delayed_cksum(m, hlen);
1002                         m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1003                 }
1004 #endif
1005                 /*
1006                  * Change the next header field of the last header in the
1007                  * unfragmentable part.
1008                  */
1009                 if (exthdrs.ip6e_rthdr) {
1010                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1011                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1012                 } else if (exthdrs.ip6e_dest1) {
1013                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1014                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1015                 } else if (exthdrs.ip6e_hbh) {
1016                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1017                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1018                 } else {
1019                         nextproto = ip6->ip6_nxt;
1020                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1021                 }
1022
1023                 /*
1024                  * Loop through length of segment after first fragment,
1025                  * make new header and copy data of each part and link onto
1026                  * chain.
1027                  */
1028                 m0 = m;
1029                 id = htonl(ip6_randomid());
1030                 if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1031                         goto sendorfree;
1032
1033                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1034         }
1035
1036         /*
1037          * Remove leading garbages.
1038          */
1039 sendorfree:
1040         m = m0->m_nextpkt;
1041         m0->m_nextpkt = 0;
1042         m_freem(m0);
1043         for (; m; m = m0) {
1044                 m0 = m->m_nextpkt;
1045                 m->m_nextpkt = 0;
1046                 if (error == 0) {
1047                         /* Record statistics for this interface address. */
1048                         if (ia) {
1049                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1050                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
1051                                     m->m_pkthdr.len);
1052                         }
1053 #ifdef RATELIMIT
1054                         if (inp != NULL) {
1055                                 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1056                                         in_pcboutput_txrtlmt(inp, ifp, m);
1057                                 /* stamp send tag on mbuf */
1058                                 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1059                         } else {
1060                                 m->m_pkthdr.snd_tag = NULL;
1061                         }
1062 #endif
1063                         error = nd6_output_ifp(ifp, origifp, m, dst,
1064                             (struct route *)ro);
1065 #ifdef RATELIMIT
1066                         /* check for route change */
1067                         if (error == EAGAIN)
1068                                 in_pcboutput_eagain(inp);
1069 #endif
1070                 } else
1071                         m_freem(m);
1072         }
1073
1074         if (error == 0)
1075                 IP6STAT_INC(ip6s_fragmented);
1076
1077 done:
1078         if (ro == &ip6route)
1079                 RO_RTFREE(ro);
1080         return (error);
1081
1082 freehdrs:
1083         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1084         m_freem(exthdrs.ip6e_dest1);
1085         m_freem(exthdrs.ip6e_rthdr);
1086         m_freem(exthdrs.ip6e_dest2);
1087         /* FALLTHROUGH */
1088 bad:
1089         if (m)
1090                 m_freem(m);
1091         goto done;
1092 }
1093
1094 static int
1095 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1096 {
1097         struct mbuf *m;
1098
1099         if (hlen > MCLBYTES)
1100                 return (ENOBUFS); /* XXX */
1101
1102         if (hlen > MLEN)
1103                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1104         else
1105                 m = m_get(M_NOWAIT, MT_DATA);
1106         if (m == NULL)
1107                 return (ENOBUFS);
1108         m->m_len = hlen;
1109         if (hdr)
1110                 bcopy(hdr, mtod(m, caddr_t), hlen);
1111
1112         *mp = m;
1113         return (0);
1114 }
1115
1116 /*
1117  * Insert jumbo payload option.
1118  */
1119 static int
1120 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1121 {
1122         struct mbuf *mopt;
1123         u_char *optbuf;
1124         u_int32_t v;
1125
1126 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1127
1128         /*
1129          * If there is no hop-by-hop options header, allocate new one.
1130          * If there is one but it doesn't have enough space to store the
1131          * jumbo payload option, allocate a cluster to store the whole options.
1132          * Otherwise, use it to store the options.
1133          */
1134         if (exthdrs->ip6e_hbh == NULL) {
1135                 mopt = m_get(M_NOWAIT, MT_DATA);
1136                 if (mopt == NULL)
1137                         return (ENOBUFS);
1138                 mopt->m_len = JUMBOOPTLEN;
1139                 optbuf = mtod(mopt, u_char *);
1140                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1141                 exthdrs->ip6e_hbh = mopt;
1142         } else {
1143                 struct ip6_hbh *hbh;
1144
1145                 mopt = exthdrs->ip6e_hbh;
1146                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1147                         /*
1148                          * XXX assumption:
1149                          * - exthdrs->ip6e_hbh is not referenced from places
1150                          *   other than exthdrs.
1151                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1152                          */
1153                         int oldoptlen = mopt->m_len;
1154                         struct mbuf *n;
1155
1156                         /*
1157                          * XXX: give up if the whole (new) hbh header does
1158                          * not fit even in an mbuf cluster.
1159                          */
1160                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1161                                 return (ENOBUFS);
1162
1163                         /*
1164                          * As a consequence, we must always prepare a cluster
1165                          * at this point.
1166                          */
1167                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
1168                         if (n == NULL)
1169                                 return (ENOBUFS);
1170                         n->m_len = oldoptlen + JUMBOOPTLEN;
1171                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1172                             oldoptlen);
1173                         optbuf = mtod(n, caddr_t) + oldoptlen;
1174                         m_freem(mopt);
1175                         mopt = exthdrs->ip6e_hbh = n;
1176                 } else {
1177                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1178                         mopt->m_len += JUMBOOPTLEN;
1179                 }
1180                 optbuf[0] = IP6OPT_PADN;
1181                 optbuf[1] = 1;
1182
1183                 /*
1184                  * Adjust the header length according to the pad and
1185                  * the jumbo payload option.
1186                  */
1187                 hbh = mtod(mopt, struct ip6_hbh *);
1188                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1189         }
1190
1191         /* fill in the option. */
1192         optbuf[2] = IP6OPT_JUMBO;
1193         optbuf[3] = 4;
1194         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1195         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1196
1197         /* finally, adjust the packet header length */
1198         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1199
1200         return (0);
1201 #undef JUMBOOPTLEN
1202 }
1203
1204 /*
1205  * Insert fragment header and copy unfragmentable header portions.
1206  */
1207 static int
1208 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1209     struct ip6_frag **frghdrp)
1210 {
1211         struct mbuf *n, *mlast;
1212
1213         if (hlen > sizeof(struct ip6_hdr)) {
1214                 n = m_copym(m0, sizeof(struct ip6_hdr),
1215                     hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1216                 if (n == NULL)
1217                         return (ENOBUFS);
1218                 m->m_next = n;
1219         } else
1220                 n = m;
1221
1222         /* Search for the last mbuf of unfragmentable part. */
1223         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1224                 ;
1225
1226         if (M_WRITABLE(mlast) &&
1227             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1228                 /* use the trailing space of the last mbuf for the fragment hdr */
1229                 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1230                     mlast->m_len);
1231                 mlast->m_len += sizeof(struct ip6_frag);
1232                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1233         } else {
1234                 /* allocate a new mbuf for the fragment header */
1235                 struct mbuf *mfrg;
1236
1237                 mfrg = m_get(M_NOWAIT, MT_DATA);
1238                 if (mfrg == NULL)
1239                         return (ENOBUFS);
1240                 mfrg->m_len = sizeof(struct ip6_frag);
1241                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1242                 mlast->m_next = mfrg;
1243         }
1244
1245         return (0);
1246 }
1247
1248 /*
1249  * Calculates IPv6 path mtu for destination @dst.
1250  * Resulting MTU is stored in @mtup.
1251  *
1252  * Returns 0 on success.
1253  */
1254 static int
1255 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1256 {
1257         struct nhop6_extended nh6;
1258         struct in6_addr kdst;
1259         uint32_t scopeid;
1260         struct ifnet *ifp;
1261         u_long mtu;
1262         int error;
1263
1264         in6_splitscope(dst, &kdst, &scopeid);
1265         if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1266                 return (EHOSTUNREACH);
1267
1268         ifp = nh6.nh_ifp;
1269         mtu = nh6.nh_mtu;
1270
1271         error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1272         fib6_free_nh_ext(fibnum, &nh6);
1273
1274         return (error);
1275 }
1276
1277 /*
1278  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1279  * and cached data in @ro_pmtu.
1280  * MTU from (successful) route lookup is saved (along with dst)
1281  * inside @ro_pmtu to avoid subsequent route lookups after packet
1282  * filter processing.
1283  *
1284  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1285  * Returns 0 on success.
1286  */
1287 static int
1288 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1289     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1290     int *alwaysfragp, u_int fibnum, u_int proto)
1291 {
1292         struct nhop6_basic nh6;
1293         struct in6_addr kdst;
1294         uint32_t scopeid;
1295         struct sockaddr_in6 *sa6_dst;
1296         u_long mtu;
1297
1298         mtu = 0;
1299         if (do_lookup) {
1300
1301                 /*
1302                  * Here ro_pmtu has final destination address, while
1303                  * ro might represent immediate destination.
1304                  * Use ro_pmtu destination since mtu might differ.
1305                  */
1306                 sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1307                 if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1308                         ro_pmtu->ro_mtu = 0;
1309
1310                 if (ro_pmtu->ro_mtu == 0) {
1311                         bzero(sa6_dst, sizeof(*sa6_dst));
1312                         sa6_dst->sin6_family = AF_INET6;
1313                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1314                         sa6_dst->sin6_addr = *dst;
1315
1316                         in6_splitscope(dst, &kdst, &scopeid);
1317                         if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1318                             &nh6) == 0)
1319                                 ro_pmtu->ro_mtu = nh6.nh_mtu;
1320                 }
1321
1322                 mtu = ro_pmtu->ro_mtu;
1323         }
1324
1325         if (ro_pmtu->ro_rt)
1326                 mtu = ro_pmtu->ro_rt->rt_mtu;
1327
1328         return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1329 }
1330
1331 /*
1332  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1333  * hostcache data for @dst.
1334  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1335  *
1336  * Returns 0 on success.
1337  */
1338 static int
1339 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1340     u_long *mtup, int *alwaysfragp, u_int proto)
1341 {
1342         u_long mtu = 0;
1343         int alwaysfrag = 0;
1344         int error = 0;
1345
1346         if (rt_mtu > 0) {
1347                 u_int32_t ifmtu;
1348                 struct in_conninfo inc;
1349
1350                 bzero(&inc, sizeof(inc));
1351                 inc.inc_flags |= INC_ISIPV6;
1352                 inc.inc6_faddr = *dst;
1353
1354                 ifmtu = IN6_LINKMTU(ifp);
1355
1356                 /* TCP is known to react to pmtu changes so skip hc */
1357                 if (proto != IPPROTO_TCP)
1358                         mtu = tcp_hc_getmtu(&inc);
1359
1360                 if (mtu)
1361                         mtu = min(mtu, rt_mtu);
1362                 else
1363                         mtu = rt_mtu;
1364                 if (mtu == 0)
1365                         mtu = ifmtu;
1366                 else if (mtu < IPV6_MMTU) {
1367                         /*
1368                          * RFC2460 section 5, last paragraph:
1369                          * if we record ICMPv6 too big message with
1370                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1371                          * or smaller, with framgent header attached.
1372                          * (fragment header is needed regardless from the
1373                          * packet size, for translators to identify packets)
1374                          */
1375                         alwaysfrag = 1;
1376                         mtu = IPV6_MMTU;
1377                 }
1378         } else if (ifp) {
1379                 mtu = IN6_LINKMTU(ifp);
1380         } else
1381                 error = EHOSTUNREACH; /* XXX */
1382
1383         *mtup = mtu;
1384         if (alwaysfragp)
1385                 *alwaysfragp = alwaysfrag;
1386         return (error);
1387 }
1388
1389 /*
1390  * IP6 socket option processing.
1391  */
1392 int
1393 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1394 {
1395         int optdatalen, uproto;
1396         void *optdata;
1397         struct inpcb *in6p = sotoinpcb(so);
1398         int error, optval;
1399         int level, op, optname;
1400         int optlen;
1401         struct thread *td;
1402 #ifdef  RSS
1403         uint32_t rss_bucket;
1404         int retval;
1405 #endif
1406
1407 /*
1408  * Don't use more than a quarter of mbuf clusters.  N.B.:
1409  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1410  * on LP64 architectures, so cast to u_long to avoid undefined
1411  * behavior.  ILP32 architectures cannot have nmbclusters
1412  * large enough to overflow for other reasons.
1413  */
1414 #define IPV6_PKTOPTIONS_MBUF_LIMIT      ((u_long)nmbclusters * MCLBYTES / 4)
1415
1416         level = sopt->sopt_level;
1417         op = sopt->sopt_dir;
1418         optname = sopt->sopt_name;
1419         optlen = sopt->sopt_valsize;
1420         td = sopt->sopt_td;
1421         error = 0;
1422         optval = 0;
1423         uproto = (int)so->so_proto->pr_protocol;
1424
1425         if (level != IPPROTO_IPV6) {
1426                 error = EINVAL;
1427
1428                 if (sopt->sopt_level == SOL_SOCKET &&
1429                     sopt->sopt_dir == SOPT_SET) {
1430                         switch (sopt->sopt_name) {
1431                         case SO_REUSEADDR:
1432                                 INP_WLOCK(in6p);
1433                                 if ((so->so_options & SO_REUSEADDR) != 0)
1434                                         in6p->inp_flags2 |= INP_REUSEADDR;
1435                                 else
1436                                         in6p->inp_flags2 &= ~INP_REUSEADDR;
1437                                 INP_WUNLOCK(in6p);
1438                                 error = 0;
1439                                 break;
1440                         case SO_REUSEPORT:
1441                                 INP_WLOCK(in6p);
1442                                 if ((so->so_options & SO_REUSEPORT) != 0)
1443                                         in6p->inp_flags2 |= INP_REUSEPORT;
1444                                 else
1445                                         in6p->inp_flags2 &= ~INP_REUSEPORT;
1446                                 INP_WUNLOCK(in6p);
1447                                 error = 0;
1448                                 break;
1449                         case SO_REUSEPORT_LB:
1450                                 INP_WLOCK(in6p);
1451                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1452                                         in6p->inp_flags2 |= INP_REUSEPORT_LB;
1453                                 else
1454                                         in6p->inp_flags2 &= ~INP_REUSEPORT_LB;
1455                                 INP_WUNLOCK(in6p);
1456                                 error = 0;
1457                                 break;
1458                         case SO_SETFIB:
1459                                 INP_WLOCK(in6p);
1460                                 in6p->inp_inc.inc_fibnum = so->so_fibnum;
1461                                 INP_WUNLOCK(in6p);
1462                                 error = 0;
1463                                 break;
1464                         case SO_MAX_PACING_RATE:
1465 #ifdef RATELIMIT
1466                                 INP_WLOCK(in6p);
1467                                 in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1468                                 INP_WUNLOCK(in6p);
1469                                 error = 0;
1470 #else
1471                                 error = EOPNOTSUPP;
1472 #endif
1473                                 break;
1474                         default:
1475                                 break;
1476                         }
1477                 }
1478         } else {                /* level == IPPROTO_IPV6 */
1479                 switch (op) {
1480
1481                 case SOPT_SET:
1482                         switch (optname) {
1483                         case IPV6_2292PKTOPTIONS:
1484 #ifdef IPV6_PKTOPTIONS
1485                         case IPV6_PKTOPTIONS:
1486 #endif
1487                         {
1488                                 struct mbuf *m;
1489
1490                                 if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1491                                         printf("ip6_ctloutput: mbuf limit hit\n");
1492                                         error = ENOBUFS;
1493                                         break;
1494                                 }
1495
1496                                 error = soopt_getm(sopt, &m); /* XXX */
1497                                 if (error != 0)
1498                                         break;
1499                                 error = soopt_mcopyin(sopt, m); /* XXX */
1500                                 if (error != 0)
1501                                         break;
1502                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1503                                                     m, so, sopt);
1504                                 m_freem(m); /* XXX */
1505                                 break;
1506                         }
1507
1508                         /*
1509                          * Use of some Hop-by-Hop options or some
1510                          * Destination options, might require special
1511                          * privilege.  That is, normal applications
1512                          * (without special privilege) might be forbidden
1513                          * from setting certain options in outgoing packets,
1514                          * and might never see certain options in received
1515                          * packets. [RFC 2292 Section 6]
1516                          * KAME specific note:
1517                          *  KAME prevents non-privileged users from sending or
1518                          *  receiving ANY hbh/dst options in order to avoid
1519                          *  overhead of parsing options in the kernel.
1520                          */
1521                         case IPV6_RECVHOPOPTS:
1522                         case IPV6_RECVDSTOPTS:
1523                         case IPV6_RECVRTHDRDSTOPTS:
1524                                 if (td != NULL) {
1525                                         error = priv_check(td,
1526                                             PRIV_NETINET_SETHDROPTS);
1527                                         if (error)
1528                                                 break;
1529                                 }
1530                                 /* FALLTHROUGH */
1531                         case IPV6_UNICAST_HOPS:
1532                         case IPV6_HOPLIMIT:
1533
1534                         case IPV6_RECVPKTINFO:
1535                         case IPV6_RECVHOPLIMIT:
1536                         case IPV6_RECVRTHDR:
1537                         case IPV6_RECVPATHMTU:
1538                         case IPV6_RECVTCLASS:
1539                         case IPV6_RECVFLOWID:
1540 #ifdef  RSS
1541                         case IPV6_RECVRSSBUCKETID:
1542 #endif
1543                         case IPV6_V6ONLY:
1544                         case IPV6_AUTOFLOWLABEL:
1545                         case IPV6_ORIGDSTADDR:
1546                         case IPV6_BINDANY:
1547                         case IPV6_BINDMULTI:
1548 #ifdef  RSS
1549                         case IPV6_RSS_LISTEN_BUCKET:
1550 #endif
1551                                 if (optname == IPV6_BINDANY && td != NULL) {
1552                                         error = priv_check(td,
1553                                             PRIV_NETINET_BINDANY);
1554                                         if (error)
1555                                                 break;
1556                                 }
1557
1558                                 if (optlen != sizeof(int)) {
1559                                         error = EINVAL;
1560                                         break;
1561                                 }
1562                                 error = sooptcopyin(sopt, &optval,
1563                                         sizeof optval, sizeof optval);
1564                                 if (error)
1565                                         break;
1566                                 switch (optname) {
1567
1568                                 case IPV6_UNICAST_HOPS:
1569                                         if (optval < -1 || optval >= 256)
1570                                                 error = EINVAL;
1571                                         else {
1572                                                 /* -1 = kernel default */
1573                                                 in6p->in6p_hops = optval;
1574                                                 if ((in6p->inp_vflag &
1575                                                      INP_IPV4) != 0)
1576                                                         in6p->inp_ip_ttl = optval;
1577                                         }
1578                                         break;
1579 #define OPTSET(bit) \
1580 do { \
1581         INP_WLOCK(in6p); \
1582         if (optval) \
1583                 in6p->inp_flags |= (bit); \
1584         else \
1585                 in6p->inp_flags &= ~(bit); \
1586         INP_WUNLOCK(in6p); \
1587 } while (/*CONSTCOND*/ 0)
1588 #define OPTSET2292(bit) \
1589 do { \
1590         INP_WLOCK(in6p); \
1591         in6p->inp_flags |= IN6P_RFC2292; \
1592         if (optval) \
1593                 in6p->inp_flags |= (bit); \
1594         else \
1595                 in6p->inp_flags &= ~(bit); \
1596         INP_WUNLOCK(in6p); \
1597 } while (/*CONSTCOND*/ 0)
1598 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1599
1600 #define OPTSET2_N(bit, val) do {                                        \
1601         if (val)                                                        \
1602                 in6p->inp_flags2 |= bit;                                \
1603         else                                                            \
1604                 in6p->inp_flags2 &= ~bit;                               \
1605 } while (0)
1606 #define OPTSET2(bit, val) do {                                          \
1607         INP_WLOCK(in6p);                                                \
1608         OPTSET2_N(bit, val);                                            \
1609         INP_WUNLOCK(in6p);                                              \
1610 } while (0)
1611 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1612 #define OPTSET2292_EXCLUSIVE(bit)                                       \
1613 do {                                                                    \
1614         INP_WLOCK(in6p);                                                \
1615         if (OPTBIT(IN6P_RFC2292)) {                                     \
1616                 error = EINVAL;                                         \
1617         } else {                                                        \
1618                 if (optval)                                             \
1619                         in6p->inp_flags |= (bit);                       \
1620                 else                                                    \
1621                         in6p->inp_flags &= ~(bit);                      \
1622         }                                                               \
1623         INP_WUNLOCK(in6p);                                              \
1624 } while (/*CONSTCOND*/ 0)
1625
1626                                 case IPV6_RECVPKTINFO:
1627                                         OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1628                                         break;
1629
1630                                 case IPV6_HOPLIMIT:
1631                                 {
1632                                         struct ip6_pktopts **optp;
1633
1634                                         /* cannot mix with RFC2292 */
1635                                         if (OPTBIT(IN6P_RFC2292)) {
1636                                                 error = EINVAL;
1637                                                 break;
1638                                         }
1639                                         INP_WLOCK(in6p);
1640                                         if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1641                                                 INP_WUNLOCK(in6p);
1642                                                 return (ECONNRESET);
1643                                         }
1644                                         optp = &in6p->in6p_outputopts;
1645                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1646                                             (u_char *)&optval, sizeof(optval),
1647                                             optp, (td != NULL) ? td->td_ucred :
1648                                             NULL, uproto);
1649                                         INP_WUNLOCK(in6p);
1650                                         break;
1651                                 }
1652
1653                                 case IPV6_RECVHOPLIMIT:
1654                                         OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1655                                         break;
1656
1657                                 case IPV6_RECVHOPOPTS:
1658                                         OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1659                                         break;
1660
1661                                 case IPV6_RECVDSTOPTS:
1662                                         OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1663                                         break;
1664
1665                                 case IPV6_RECVRTHDRDSTOPTS:
1666                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1667                                         break;
1668
1669                                 case IPV6_RECVRTHDR:
1670                                         OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1671                                         break;
1672
1673                                 case IPV6_RECVPATHMTU:
1674                                         /*
1675                                          * We ignore this option for TCP
1676                                          * sockets.
1677                                          * (RFC3542 leaves this case
1678                                          * unspecified.)
1679                                          */
1680                                         if (uproto != IPPROTO_TCP)
1681                                                 OPTSET(IN6P_MTU);
1682                                         break;
1683
1684                                 case IPV6_RECVFLOWID:
1685                                         OPTSET2(INP_RECVFLOWID, optval);
1686                                         break;
1687
1688 #ifdef  RSS
1689                                 case IPV6_RECVRSSBUCKETID:
1690                                         OPTSET2(INP_RECVRSSBUCKETID, optval);
1691                                         break;
1692 #endif
1693
1694                                 case IPV6_V6ONLY:
1695                                         /*
1696                                          * make setsockopt(IPV6_V6ONLY)
1697                                          * available only prior to bind(2).
1698                                          * see ipng mailing list, Jun 22 2001.
1699                                          */
1700                                         if (in6p->inp_lport ||
1701                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1702                                                 error = EINVAL;
1703                                                 break;
1704                                         }
1705                                         OPTSET(IN6P_IPV6_V6ONLY);
1706                                         if (optval)
1707                                                 in6p->inp_vflag &= ~INP_IPV4;
1708                                         else
1709                                                 in6p->inp_vflag |= INP_IPV4;
1710                                         break;
1711                                 case IPV6_RECVTCLASS:
1712                                         /* cannot mix with RFC2292 XXX */
1713                                         OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1714                                         break;
1715                                 case IPV6_AUTOFLOWLABEL:
1716                                         OPTSET(IN6P_AUTOFLOWLABEL);
1717                                         break;
1718
1719                                 case IPV6_ORIGDSTADDR:
1720                                         OPTSET2(INP_ORIGDSTADDR, optval);
1721                                         break;
1722                                 case IPV6_BINDANY:
1723                                         OPTSET(INP_BINDANY);
1724                                         break;
1725
1726                                 case IPV6_BINDMULTI:
1727                                         OPTSET2(INP_BINDMULTI, optval);
1728                                         break;
1729 #ifdef  RSS
1730                                 case IPV6_RSS_LISTEN_BUCKET:
1731                                         if ((optval >= 0) &&
1732                                             (optval < rss_getnumbuckets())) {
1733                                                 INP_WLOCK(in6p);
1734                                                 in6p->inp_rss_listen_bucket = optval;
1735                                                 OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1736                                                 INP_WUNLOCK(in6p);
1737                                         } else {
1738                                                 error = EINVAL;
1739                                         }
1740                                         break;
1741 #endif
1742                                 }
1743                                 break;
1744
1745                         case IPV6_TCLASS:
1746                         case IPV6_DONTFRAG:
1747                         case IPV6_USE_MIN_MTU:
1748                         case IPV6_PREFER_TEMPADDR:
1749                                 if (optlen != sizeof(optval)) {
1750                                         error = EINVAL;
1751                                         break;
1752                                 }
1753                                 error = sooptcopyin(sopt, &optval,
1754                                         sizeof optval, sizeof optval);
1755                                 if (error)
1756                                         break;
1757                                 {
1758                                         struct ip6_pktopts **optp;
1759                                         INP_WLOCK(in6p);
1760                                         if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1761                                                 INP_WUNLOCK(in6p);
1762                                                 return (ECONNRESET);
1763                                         }
1764                                         optp = &in6p->in6p_outputopts;
1765                                         error = ip6_pcbopt(optname,
1766                                             (u_char *)&optval, sizeof(optval),
1767                                             optp, (td != NULL) ? td->td_ucred :
1768                                             NULL, uproto);
1769                                         INP_WUNLOCK(in6p);
1770                                         break;
1771                                 }
1772
1773                         case IPV6_2292PKTINFO:
1774                         case IPV6_2292HOPLIMIT:
1775                         case IPV6_2292HOPOPTS:
1776                         case IPV6_2292DSTOPTS:
1777                         case IPV6_2292RTHDR:
1778                                 /* RFC 2292 */
1779                                 if (optlen != sizeof(int)) {
1780                                         error = EINVAL;
1781                                         break;
1782                                 }
1783                                 error = sooptcopyin(sopt, &optval,
1784                                         sizeof optval, sizeof optval);
1785                                 if (error)
1786                                         break;
1787                                 switch (optname) {
1788                                 case IPV6_2292PKTINFO:
1789                                         OPTSET2292(IN6P_PKTINFO);
1790                                         break;
1791                                 case IPV6_2292HOPLIMIT:
1792                                         OPTSET2292(IN6P_HOPLIMIT);
1793                                         break;
1794                                 case IPV6_2292HOPOPTS:
1795                                         /*
1796                                          * Check super-user privilege.
1797                                          * See comments for IPV6_RECVHOPOPTS.
1798                                          */
1799                                         if (td != NULL) {
1800                                                 error = priv_check(td,
1801                                                     PRIV_NETINET_SETHDROPTS);
1802                                                 if (error)
1803                                                         return (error);
1804                                         }
1805                                         OPTSET2292(IN6P_HOPOPTS);
1806                                         break;
1807                                 case IPV6_2292DSTOPTS:
1808                                         if (td != NULL) {
1809                                                 error = priv_check(td,
1810                                                     PRIV_NETINET_SETHDROPTS);
1811                                                 if (error)
1812                                                         return (error);
1813                                         }
1814                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1815                                         break;
1816                                 case IPV6_2292RTHDR:
1817                                         OPTSET2292(IN6P_RTHDR);
1818                                         break;
1819                                 }
1820                                 break;
1821                         case IPV6_PKTINFO:
1822                         case IPV6_HOPOPTS:
1823                         case IPV6_RTHDR:
1824                         case IPV6_DSTOPTS:
1825                         case IPV6_RTHDRDSTOPTS:
1826                         case IPV6_NEXTHOP:
1827                         {
1828                                 /* new advanced API (RFC3542) */
1829                                 u_char *optbuf;
1830                                 u_char optbuf_storage[MCLBYTES];
1831                                 int optlen;
1832                                 struct ip6_pktopts **optp;
1833
1834                                 /* cannot mix with RFC2292 */
1835                                 if (OPTBIT(IN6P_RFC2292)) {
1836                                         error = EINVAL;
1837                                         break;
1838                                 }
1839
1840                                 /*
1841                                  * We only ensure valsize is not too large
1842                                  * here.  Further validation will be done
1843                                  * later.
1844                                  */
1845                                 error = sooptcopyin(sopt, optbuf_storage,
1846                                     sizeof(optbuf_storage), 0);
1847                                 if (error)
1848                                         break;
1849                                 optlen = sopt->sopt_valsize;
1850                                 optbuf = optbuf_storage;
1851                                 INP_WLOCK(in6p);
1852                                 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1853                                         INP_WUNLOCK(in6p);
1854                                         return (ECONNRESET);
1855                                 }
1856                                 optp = &in6p->in6p_outputopts;
1857                                 error = ip6_pcbopt(optname, optbuf, optlen,
1858                                     optp, (td != NULL) ? td->td_ucred : NULL,
1859                                     uproto);
1860                                 INP_WUNLOCK(in6p);
1861                                 break;
1862                         }
1863 #undef OPTSET
1864
1865                         case IPV6_MULTICAST_IF:
1866                         case IPV6_MULTICAST_HOPS:
1867                         case IPV6_MULTICAST_LOOP:
1868                         case IPV6_JOIN_GROUP:
1869                         case IPV6_LEAVE_GROUP:
1870                         case IPV6_MSFILTER:
1871                         case MCAST_BLOCK_SOURCE:
1872                         case MCAST_UNBLOCK_SOURCE:
1873                         case MCAST_JOIN_GROUP:
1874                         case MCAST_LEAVE_GROUP:
1875                         case MCAST_JOIN_SOURCE_GROUP:
1876                         case MCAST_LEAVE_SOURCE_GROUP:
1877                                 error = ip6_setmoptions(in6p, sopt);
1878                                 break;
1879
1880                         case IPV6_PORTRANGE:
1881                                 error = sooptcopyin(sopt, &optval,
1882                                     sizeof optval, sizeof optval);
1883                                 if (error)
1884                                         break;
1885
1886                                 INP_WLOCK(in6p);
1887                                 switch (optval) {
1888                                 case IPV6_PORTRANGE_DEFAULT:
1889                                         in6p->inp_flags &= ~(INP_LOWPORT);
1890                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1891                                         break;
1892
1893                                 case IPV6_PORTRANGE_HIGH:
1894                                         in6p->inp_flags &= ~(INP_LOWPORT);
1895                                         in6p->inp_flags |= INP_HIGHPORT;
1896                                         break;
1897
1898                                 case IPV6_PORTRANGE_LOW:
1899                                         in6p->inp_flags &= ~(INP_HIGHPORT);
1900                                         in6p->inp_flags |= INP_LOWPORT;
1901                                         break;
1902
1903                                 default:
1904                                         error = EINVAL;
1905                                         break;
1906                                 }
1907                                 INP_WUNLOCK(in6p);
1908                                 break;
1909
1910 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1911                         case IPV6_IPSEC_POLICY:
1912                                 if (IPSEC_ENABLED(ipv6)) {
1913                                         error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1914                                         break;
1915                                 }
1916                                 /* FALLTHROUGH */
1917 #endif /* IPSEC */
1918
1919                         default:
1920                                 error = ENOPROTOOPT;
1921                                 break;
1922                         }
1923                         break;
1924
1925                 case SOPT_GET:
1926                         switch (optname) {
1927
1928                         case IPV6_2292PKTOPTIONS:
1929 #ifdef IPV6_PKTOPTIONS
1930                         case IPV6_PKTOPTIONS:
1931 #endif
1932                                 /*
1933                                  * RFC3542 (effectively) deprecated the
1934                                  * semantics of the 2292-style pktoptions.
1935                                  * Since it was not reliable in nature (i.e.,
1936                                  * applications had to expect the lack of some
1937                                  * information after all), it would make sense
1938                                  * to simplify this part by always returning
1939                                  * empty data.
1940                                  */
1941                                 sopt->sopt_valsize = 0;
1942                                 break;
1943
1944                         case IPV6_RECVHOPOPTS:
1945                         case IPV6_RECVDSTOPTS:
1946                         case IPV6_RECVRTHDRDSTOPTS:
1947                         case IPV6_UNICAST_HOPS:
1948                         case IPV6_RECVPKTINFO:
1949                         case IPV6_RECVHOPLIMIT:
1950                         case IPV6_RECVRTHDR:
1951                         case IPV6_RECVPATHMTU:
1952
1953                         case IPV6_V6ONLY:
1954                         case IPV6_PORTRANGE:
1955                         case IPV6_RECVTCLASS:
1956                         case IPV6_AUTOFLOWLABEL:
1957                         case IPV6_BINDANY:
1958                         case IPV6_FLOWID:
1959                         case IPV6_FLOWTYPE:
1960                         case IPV6_RECVFLOWID:
1961 #ifdef  RSS
1962                         case IPV6_RSSBUCKETID:
1963                         case IPV6_RECVRSSBUCKETID:
1964 #endif
1965                         case IPV6_BINDMULTI:
1966                                 switch (optname) {
1967
1968                                 case IPV6_RECVHOPOPTS:
1969                                         optval = OPTBIT(IN6P_HOPOPTS);
1970                                         break;
1971
1972                                 case IPV6_RECVDSTOPTS:
1973                                         optval = OPTBIT(IN6P_DSTOPTS);
1974                                         break;
1975
1976                                 case IPV6_RECVRTHDRDSTOPTS:
1977                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1978                                         break;
1979
1980                                 case IPV6_UNICAST_HOPS:
1981                                         optval = in6p->in6p_hops;
1982                                         break;
1983
1984                                 case IPV6_RECVPKTINFO:
1985                                         optval = OPTBIT(IN6P_PKTINFO);
1986                                         break;
1987
1988                                 case IPV6_RECVHOPLIMIT:
1989                                         optval = OPTBIT(IN6P_HOPLIMIT);
1990                                         break;
1991
1992                                 case IPV6_RECVRTHDR:
1993                                         optval = OPTBIT(IN6P_RTHDR);
1994                                         break;
1995
1996                                 case IPV6_RECVPATHMTU:
1997                                         optval = OPTBIT(IN6P_MTU);
1998                                         break;
1999
2000                                 case IPV6_V6ONLY:
2001                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2002                                         break;
2003
2004                                 case IPV6_PORTRANGE:
2005                                     {
2006                                         int flags;
2007                                         flags = in6p->inp_flags;
2008                                         if (flags & INP_HIGHPORT)
2009                                                 optval = IPV6_PORTRANGE_HIGH;
2010                                         else if (flags & INP_LOWPORT)
2011                                                 optval = IPV6_PORTRANGE_LOW;
2012                                         else
2013                                                 optval = 0;
2014                                         break;
2015                                     }
2016                                 case IPV6_RECVTCLASS:
2017                                         optval = OPTBIT(IN6P_TCLASS);
2018                                         break;
2019
2020                                 case IPV6_AUTOFLOWLABEL:
2021                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2022                                         break;
2023
2024                                 case IPV6_ORIGDSTADDR:
2025                                         optval = OPTBIT2(INP_ORIGDSTADDR);
2026                                         break;
2027
2028                                 case IPV6_BINDANY:
2029                                         optval = OPTBIT(INP_BINDANY);
2030                                         break;
2031
2032                                 case IPV6_FLOWID:
2033                                         optval = in6p->inp_flowid;
2034                                         break;
2035
2036                                 case IPV6_FLOWTYPE:
2037                                         optval = in6p->inp_flowtype;
2038                                         break;
2039
2040                                 case IPV6_RECVFLOWID:
2041                                         optval = OPTBIT2(INP_RECVFLOWID);
2042                                         break;
2043 #ifdef  RSS
2044                                 case IPV6_RSSBUCKETID:
2045                                         retval =
2046                                             rss_hash2bucket(in6p->inp_flowid,
2047                                             in6p->inp_flowtype,
2048                                             &rss_bucket);
2049                                         if (retval == 0)
2050                                                 optval = rss_bucket;
2051                                         else
2052                                                 error = EINVAL;
2053                                         break;
2054
2055                                 case IPV6_RECVRSSBUCKETID:
2056                                         optval = OPTBIT2(INP_RECVRSSBUCKETID);
2057                                         break;
2058 #endif
2059
2060                                 case IPV6_BINDMULTI:
2061                                         optval = OPTBIT2(INP_BINDMULTI);
2062                                         break;
2063
2064                                 }
2065                                 if (error)
2066                                         break;
2067                                 error = sooptcopyout(sopt, &optval,
2068                                         sizeof optval);
2069                                 break;
2070
2071                         case IPV6_PATHMTU:
2072                         {
2073                                 u_long pmtu = 0;
2074                                 struct ip6_mtuinfo mtuinfo;
2075                                 struct in6_addr addr;
2076
2077                                 if (!(so->so_state & SS_ISCONNECTED))
2078                                         return (ENOTCONN);
2079                                 /*
2080                                  * XXX: we dot not consider the case of source
2081                                  * routing, or optional information to specify
2082                                  * the outgoing interface.
2083                                  * Copy faddr out of in6p to avoid holding lock
2084                                  * on inp during route lookup.
2085                                  */
2086                                 INP_RLOCK(in6p);
2087                                 bcopy(&in6p->in6p_faddr, &addr, sizeof(addr));
2088                                 INP_RUNLOCK(in6p);
2089                                 error = ip6_getpmtu_ctl(so->so_fibnum,
2090                                     &addr, &pmtu);
2091                                 if (error)
2092                                         break;
2093                                 if (pmtu > IPV6_MAXPACKET)
2094                                         pmtu = IPV6_MAXPACKET;
2095
2096                                 bzero(&mtuinfo, sizeof(mtuinfo));
2097                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2098                                 optdata = (void *)&mtuinfo;
2099                                 optdatalen = sizeof(mtuinfo);
2100                                 error = sooptcopyout(sopt, optdata,
2101                                     optdatalen);
2102                                 break;
2103                         }
2104
2105                         case IPV6_2292PKTINFO:
2106                         case IPV6_2292HOPLIMIT:
2107                         case IPV6_2292HOPOPTS:
2108                         case IPV6_2292RTHDR:
2109                         case IPV6_2292DSTOPTS:
2110                                 switch (optname) {
2111                                 case IPV6_2292PKTINFO:
2112                                         optval = OPTBIT(IN6P_PKTINFO);
2113                                         break;
2114                                 case IPV6_2292HOPLIMIT:
2115                                         optval = OPTBIT(IN6P_HOPLIMIT);
2116                                         break;
2117                                 case IPV6_2292HOPOPTS:
2118                                         optval = OPTBIT(IN6P_HOPOPTS);
2119                                         break;
2120                                 case IPV6_2292RTHDR:
2121                                         optval = OPTBIT(IN6P_RTHDR);
2122                                         break;
2123                                 case IPV6_2292DSTOPTS:
2124                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2125                                         break;
2126                                 }
2127                                 error = sooptcopyout(sopt, &optval,
2128                                     sizeof optval);
2129                                 break;
2130                         case IPV6_PKTINFO:
2131                         case IPV6_HOPOPTS:
2132                         case IPV6_RTHDR:
2133                         case IPV6_DSTOPTS:
2134                         case IPV6_RTHDRDSTOPTS:
2135                         case IPV6_NEXTHOP:
2136                         case IPV6_TCLASS:
2137                         case IPV6_DONTFRAG:
2138                         case IPV6_USE_MIN_MTU:
2139                         case IPV6_PREFER_TEMPADDR:
2140                                 error = ip6_getpcbopt(in6p, optname, sopt);
2141                                 break;
2142
2143                         case IPV6_MULTICAST_IF:
2144                         case IPV6_MULTICAST_HOPS:
2145                         case IPV6_MULTICAST_LOOP:
2146                         case IPV6_MSFILTER:
2147                                 error = ip6_getmoptions(in6p, sopt);
2148                                 break;
2149
2150 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2151                         case IPV6_IPSEC_POLICY:
2152                                 if (IPSEC_ENABLED(ipv6)) {
2153                                         error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2154                                         break;
2155                                 }
2156                                 /* FALLTHROUGH */
2157 #endif /* IPSEC */
2158                         default:
2159                                 error = ENOPROTOOPT;
2160                                 break;
2161                         }
2162                         break;
2163                 }
2164         }
2165         return (error);
2166 }
2167
2168 int
2169 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2170 {
2171         int error = 0, optval, optlen;
2172         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2173         struct inpcb *in6p = sotoinpcb(so);
2174         int level, op, optname;
2175
2176         level = sopt->sopt_level;
2177         op = sopt->sopt_dir;
2178         optname = sopt->sopt_name;
2179         optlen = sopt->sopt_valsize;
2180
2181         if (level != IPPROTO_IPV6) {
2182                 return (EINVAL);
2183         }
2184
2185         switch (optname) {
2186         case IPV6_CHECKSUM:
2187                 /*
2188                  * For ICMPv6 sockets, no modification allowed for checksum
2189                  * offset, permit "no change" values to help existing apps.
2190                  *
2191                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2192                  * for an ICMPv6 socket will fail."
2193                  * The current behavior does not meet RFC3542.
2194                  */
2195                 switch (op) {
2196                 case SOPT_SET:
2197                         if (optlen != sizeof(int)) {
2198                                 error = EINVAL;
2199                                 break;
2200                         }
2201                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2202                                             sizeof(optval));
2203                         if (error)
2204                                 break;
2205                         if ((optval % 2) != 0) {
2206                                 /* the API assumes even offset values */
2207                                 error = EINVAL;
2208                         } else if (so->so_proto->pr_protocol ==
2209                             IPPROTO_ICMPV6) {
2210                                 if (optval != icmp6off)
2211                                         error = EINVAL;
2212                         } else
2213                                 in6p->in6p_cksum = optval;
2214                         break;
2215
2216                 case SOPT_GET:
2217                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2218                                 optval = icmp6off;
2219                         else
2220                                 optval = in6p->in6p_cksum;
2221
2222                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2223                         break;
2224
2225                 default:
2226                         error = EINVAL;
2227                         break;
2228                 }
2229                 break;
2230
2231         default:
2232                 error = ENOPROTOOPT;
2233                 break;
2234         }
2235
2236         return (error);
2237 }
2238
2239 /*
2240  * Set up IP6 options in pcb for insertion in output packets or
2241  * specifying behavior of outgoing packets.
2242  */
2243 static int
2244 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2245     struct socket *so, struct sockopt *sopt)
2246 {
2247         struct ip6_pktopts *opt = *pktopt;
2248         int error = 0;
2249         struct thread *td = sopt->sopt_td;
2250
2251         /* turn off any old options. */
2252         if (opt) {
2253 #ifdef DIAGNOSTIC
2254                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2255                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2256                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2257                         printf("ip6_pcbopts: all specified options are cleared.\n");
2258 #endif
2259                 ip6_clearpktopts(opt, -1);
2260         } else
2261                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2262         *pktopt = NULL;
2263
2264         if (!m || m->m_len == 0) {
2265                 /*
2266                  * Only turning off any previous options, regardless of
2267                  * whether the opt is just created or given.
2268                  */
2269                 free(opt, M_IP6OPT);
2270                 return (0);
2271         }
2272
2273         /*  set options specified by user. */
2274         if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2275             td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2276                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2277                 free(opt, M_IP6OPT);
2278                 return (error);
2279         }
2280         *pktopt = opt;
2281         return (0);
2282 }
2283
2284 /*
2285  * initialize ip6_pktopts.  beware that there are non-zero default values in
2286  * the struct.
2287  */
2288 void
2289 ip6_initpktopts(struct ip6_pktopts *opt)
2290 {
2291
2292         bzero(opt, sizeof(*opt));
2293         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2294         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2295         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2296         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2297 }
2298
2299 static int
2300 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2301     struct ucred *cred, int uproto)
2302 {
2303         struct ip6_pktopts *opt;
2304
2305         if (*pktopt == NULL) {
2306                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2307                     M_NOWAIT);
2308                 if (*pktopt == NULL)
2309                         return (ENOBUFS);
2310                 ip6_initpktopts(*pktopt);
2311         }
2312         opt = *pktopt;
2313
2314         return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2315 }
2316
2317 #define GET_PKTOPT_VAR(field, lenexpr) do {                                     \
2318         if (pktopt && pktopt->field) {                                          \
2319                 INP_RUNLOCK(in6p);                                              \
2320                 optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);         \
2321                 malloc_optdata = true;                                          \
2322                 INP_RLOCK(in6p);                                                \
2323                 if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {           \
2324                         INP_RUNLOCK(in6p);                                      \
2325                         free(optdata, M_TEMP);                                  \
2326                         return (ECONNRESET);                                    \
2327                 }                                                               \
2328                 pktopt = in6p->in6p_outputopts;                                 \
2329                 if (pktopt && pktopt->field) {                                  \
2330                         optdatalen = min(lenexpr, sopt->sopt_valsize);          \
2331                         bcopy(&pktopt->field, optdata, optdatalen);             \
2332                 } else {                                                        \
2333                         free(optdata, M_TEMP);                                  \
2334                         optdata = NULL;                                         \
2335                         malloc_optdata = false;                                 \
2336                 }                                                               \
2337         }                                                                       \
2338 } while(0)
2339
2340 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,                         \
2341         (((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2342
2343 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,                        \
2344         pktopt->field->sa_len)
2345
2346 static int
2347 ip6_getpcbopt(struct inpcb *in6p, int optname, struct sockopt *sopt)
2348 {
2349         void *optdata = NULL;
2350         bool malloc_optdata = false;
2351         int optdatalen = 0;
2352         int error = 0;
2353         struct in6_pktinfo null_pktinfo;
2354         int deftclass = 0, on;
2355         int defminmtu = IP6PO_MINMTU_MCASTONLY;
2356         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2357         struct ip6_pktopts *pktopt;
2358
2359         INP_RLOCK(in6p);
2360         pktopt = in6p->in6p_outputopts;
2361
2362         switch (optname) {
2363         case IPV6_PKTINFO:
2364                 optdata = (void *)&null_pktinfo;
2365                 if (pktopt && pktopt->ip6po_pktinfo) {
2366                         bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2367                             sizeof(null_pktinfo));
2368                         in6_clearscope(&null_pktinfo.ipi6_addr);
2369                 } else {
2370                         /* XXX: we don't have to do this every time... */
2371                         bzero(&null_pktinfo, sizeof(null_pktinfo));
2372                 }
2373                 optdatalen = sizeof(struct in6_pktinfo);
2374                 break;
2375         case IPV6_TCLASS:
2376                 if (pktopt && pktopt->ip6po_tclass >= 0)
2377                         deftclass = pktopt->ip6po_tclass;
2378                 optdata = (void *)&deftclass;
2379                 optdatalen = sizeof(int);
2380                 break;
2381         case IPV6_HOPOPTS:
2382                 GET_PKTOPT_EXT_HDR(ip6po_hbh);
2383                 break;
2384         case IPV6_RTHDR:
2385                 GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2386                 break;
2387         case IPV6_RTHDRDSTOPTS:
2388                 GET_PKTOPT_EXT_HDR(ip6po_dest1);
2389                 break;
2390         case IPV6_DSTOPTS:
2391                 GET_PKTOPT_EXT_HDR(ip6po_dest2);
2392                 break;
2393         case IPV6_NEXTHOP:
2394                 GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2395                 break;
2396         case IPV6_USE_MIN_MTU:
2397                 if (pktopt)
2398                         defminmtu = pktopt->ip6po_minmtu;
2399                 optdata = (void *)&defminmtu;
2400                 optdatalen = sizeof(int);
2401                 break;
2402         case IPV6_DONTFRAG:
2403                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2404                         on = 1;
2405                 else
2406                         on = 0;
2407                 optdata = (void *)&on;
2408                 optdatalen = sizeof(on);
2409                 break;
2410         case IPV6_PREFER_TEMPADDR:
2411                 if (pktopt)
2412                         defpreftemp = pktopt->ip6po_prefer_tempaddr;
2413                 optdata = (void *)&defpreftemp;
2414                 optdatalen = sizeof(int);
2415                 break;
2416         default:                /* should not happen */
2417 #ifdef DIAGNOSTIC
2418                 panic("ip6_getpcbopt: unexpected option\n");
2419 #endif
2420                 INP_RUNLOCK(in6p);
2421                 return (ENOPROTOOPT);
2422         }
2423         INP_RUNLOCK(in6p);
2424
2425         error = sooptcopyout(sopt, optdata, optdatalen);
2426         if (malloc_optdata)
2427                 free(optdata, M_TEMP);
2428
2429         return (error);
2430 }
2431
2432 void
2433 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2434 {
2435         if (pktopt == NULL)
2436                 return;
2437
2438         if (optname == -1 || optname == IPV6_PKTINFO) {
2439                 if (pktopt->ip6po_pktinfo)
2440                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
2441                 pktopt->ip6po_pktinfo = NULL;
2442         }
2443         if (optname == -1 || optname == IPV6_HOPLIMIT)
2444                 pktopt->ip6po_hlim = -1;
2445         if (optname == -1 || optname == IPV6_TCLASS)
2446                 pktopt->ip6po_tclass = -1;
2447         if (optname == -1 || optname == IPV6_NEXTHOP) {
2448                 if (pktopt->ip6po_nextroute.ro_rt) {
2449                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2450                         pktopt->ip6po_nextroute.ro_rt = NULL;
2451                 }
2452                 if (pktopt->ip6po_nexthop)
2453                         free(pktopt->ip6po_nexthop, M_IP6OPT);
2454                 pktopt->ip6po_nexthop = NULL;
2455         }
2456         if (optname == -1 || optname == IPV6_HOPOPTS) {
2457                 if (pktopt->ip6po_hbh)
2458                         free(pktopt->ip6po_hbh, M_IP6OPT);
2459                 pktopt->ip6po_hbh = NULL;
2460         }
2461         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2462                 if (pktopt->ip6po_dest1)
2463                         free(pktopt->ip6po_dest1, M_IP6OPT);
2464                 pktopt->ip6po_dest1 = NULL;
2465         }
2466         if (optname == -1 || optname == IPV6_RTHDR) {
2467                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2468                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2469                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2470                 if (pktopt->ip6po_route.ro_rt) {
2471                         RTFREE(pktopt->ip6po_route.ro_rt);
2472                         pktopt->ip6po_route.ro_rt = NULL;
2473                 }
2474         }
2475         if (optname == -1 || optname == IPV6_DSTOPTS) {
2476                 if (pktopt->ip6po_dest2)
2477                         free(pktopt->ip6po_dest2, M_IP6OPT);
2478                 pktopt->ip6po_dest2 = NULL;
2479         }
2480 }
2481
2482 #define PKTOPT_EXTHDRCPY(type) \
2483 do {\
2484         if (src->type) {\
2485                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2486                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2487                 if (dst->type == NULL)\
2488                         goto bad;\
2489                 bcopy(src->type, dst->type, hlen);\
2490         }\
2491 } while (/*CONSTCOND*/ 0)
2492
2493 static int
2494 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2495 {
2496         if (dst == NULL || src == NULL)  {
2497                 printf("ip6_clearpktopts: invalid argument\n");
2498                 return (EINVAL);
2499         }
2500
2501         dst->ip6po_hlim = src->ip6po_hlim;
2502         dst->ip6po_tclass = src->ip6po_tclass;
2503         dst->ip6po_flags = src->ip6po_flags;
2504         dst->ip6po_minmtu = src->ip6po_minmtu;
2505         dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2506         if (src->ip6po_pktinfo) {
2507                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2508                     M_IP6OPT, canwait);
2509                 if (dst->ip6po_pktinfo == NULL)
2510                         goto bad;
2511                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2512         }
2513         if (src->ip6po_nexthop) {
2514                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2515                     M_IP6OPT, canwait);
2516                 if (dst->ip6po_nexthop == NULL)
2517                         goto bad;
2518                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2519                     src->ip6po_nexthop->sa_len);
2520         }
2521         PKTOPT_EXTHDRCPY(ip6po_hbh);
2522         PKTOPT_EXTHDRCPY(ip6po_dest1);
2523         PKTOPT_EXTHDRCPY(ip6po_dest2);
2524         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2525         return (0);
2526
2527   bad:
2528         ip6_clearpktopts(dst, -1);
2529         return (ENOBUFS);
2530 }
2531 #undef PKTOPT_EXTHDRCPY
2532
2533 struct ip6_pktopts *
2534 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2535 {
2536         int error;
2537         struct ip6_pktopts *dst;
2538
2539         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2540         if (dst == NULL)
2541                 return (NULL);
2542         ip6_initpktopts(dst);
2543
2544         if ((error = copypktopts(dst, src, canwait)) != 0) {
2545                 free(dst, M_IP6OPT);
2546                 return (NULL);
2547         }
2548
2549         return (dst);
2550 }
2551
2552 void
2553 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2554 {
2555         if (pktopt == NULL)
2556                 return;
2557
2558         ip6_clearpktopts(pktopt, -1);
2559
2560         free(pktopt, M_IP6OPT);
2561 }
2562
2563 /*
2564  * Set IPv6 outgoing packet options based on advanced API.
2565  */
2566 int
2567 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2568     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2569 {
2570         struct cmsghdr *cm = NULL;
2571
2572         if (control == NULL || opt == NULL)
2573                 return (EINVAL);
2574
2575         ip6_initpktopts(opt);
2576         if (stickyopt) {
2577                 int error;
2578
2579                 /*
2580                  * If stickyopt is provided, make a local copy of the options
2581                  * for this particular packet, then override them by ancillary
2582                  * objects.
2583                  * XXX: copypktopts() does not copy the cached route to a next
2584                  * hop (if any).  This is not very good in terms of efficiency,
2585                  * but we can allow this since this option should be rarely
2586                  * used.
2587                  */
2588                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2589                         return (error);
2590         }
2591
2592         /*
2593          * XXX: Currently, we assume all the optional information is stored
2594          * in a single mbuf.
2595          */
2596         if (control->m_next)
2597                 return (EINVAL);
2598
2599         for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2600             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2601                 int error;
2602
2603                 if (control->m_len < CMSG_LEN(0))
2604                         return (EINVAL);
2605
2606                 cm = mtod(control, struct cmsghdr *);
2607                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2608                         return (EINVAL);
2609                 if (cm->cmsg_level != IPPROTO_IPV6)
2610                         continue;
2611
2612                 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2613                     cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2614                 if (error)
2615                         return (error);
2616         }
2617
2618         return (0);
2619 }
2620
2621 /*
2622  * Set a particular packet option, as a sticky option or an ancillary data
2623  * item.  "len" can be 0 only when it's a sticky option.
2624  * We have 4 cases of combination of "sticky" and "cmsg":
2625  * "sticky=0, cmsg=0": impossible
2626  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2627  * "sticky=1, cmsg=0": RFC3542 socket option
2628  * "sticky=1, cmsg=1": RFC2292 socket option
2629  */
2630 static int
2631 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2632     struct ucred *cred, int sticky, int cmsg, int uproto)
2633 {
2634         int minmtupolicy, preftemp;
2635         int error;
2636
2637         if (!sticky && !cmsg) {
2638 #ifdef DIAGNOSTIC
2639                 printf("ip6_setpktopt: impossible case\n");
2640 #endif
2641                 return (EINVAL);
2642         }
2643
2644         /*
2645          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2646          * not be specified in the context of RFC3542.  Conversely,
2647          * RFC3542 types should not be specified in the context of RFC2292.
2648          */
2649         if (!cmsg) {
2650                 switch (optname) {
2651                 case IPV6_2292PKTINFO:
2652                 case IPV6_2292HOPLIMIT:
2653                 case IPV6_2292NEXTHOP:
2654                 case IPV6_2292HOPOPTS:
2655                 case IPV6_2292DSTOPTS:
2656                 case IPV6_2292RTHDR:
2657                 case IPV6_2292PKTOPTIONS:
2658                         return (ENOPROTOOPT);
2659                 }
2660         }
2661         if (sticky && cmsg) {
2662                 switch (optname) {
2663                 case IPV6_PKTINFO:
2664                 case IPV6_HOPLIMIT:
2665                 case IPV6_NEXTHOP:
2666                 case IPV6_HOPOPTS:
2667                 case IPV6_DSTOPTS:
2668                 case IPV6_RTHDRDSTOPTS:
2669                 case IPV6_RTHDR:
2670                 case IPV6_USE_MIN_MTU:
2671                 case IPV6_DONTFRAG:
2672                 case IPV6_TCLASS:
2673                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2674                         return (ENOPROTOOPT);
2675                 }
2676         }
2677
2678         switch (optname) {
2679         case IPV6_2292PKTINFO:
2680         case IPV6_PKTINFO:
2681         {
2682                 struct ifnet *ifp = NULL;
2683                 struct in6_pktinfo *pktinfo;
2684
2685                 if (len != sizeof(struct in6_pktinfo))
2686                         return (EINVAL);
2687
2688                 pktinfo = (struct in6_pktinfo *)buf;
2689
2690                 /*
2691                  * An application can clear any sticky IPV6_PKTINFO option by
2692                  * doing a "regular" setsockopt with ipi6_addr being
2693                  * in6addr_any and ipi6_ifindex being zero.
2694                  * [RFC 3542, Section 6]
2695                  */
2696                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2697                     pktinfo->ipi6_ifindex == 0 &&
2698                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2699                         ip6_clearpktopts(opt, optname);
2700                         break;
2701                 }
2702
2703                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2704                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2705                         return (EINVAL);
2706                 }
2707                 if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2708                         return (EINVAL);
2709                 /* validate the interface index if specified. */
2710                 if (pktinfo->ipi6_ifindex > V_if_index)
2711                          return (ENXIO);
2712                 if (pktinfo->ipi6_ifindex) {
2713                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2714                         if (ifp == NULL)
2715                                 return (ENXIO);
2716                 }
2717                 if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2718                     (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2719                         return (ENETDOWN);
2720
2721                 if (ifp != NULL &&
2722                     !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2723                         struct in6_ifaddr *ia;
2724
2725                         in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2726                         ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2727                         if (ia == NULL)
2728                                 return (EADDRNOTAVAIL);
2729                         ifa_free(&ia->ia_ifa);
2730                 }
2731                 /*
2732                  * We store the address anyway, and let in6_selectsrc()
2733                  * validate the specified address.  This is because ipi6_addr
2734                  * may not have enough information about its scope zone, and
2735                  * we may need additional information (such as outgoing
2736                  * interface or the scope zone of a destination address) to
2737                  * disambiguate the scope.
2738                  * XXX: the delay of the validation may confuse the
2739                  * application when it is used as a sticky option.
2740                  */
2741                 if (opt->ip6po_pktinfo == NULL) {
2742                         opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2743                             M_IP6OPT, M_NOWAIT);
2744                         if (opt->ip6po_pktinfo == NULL)
2745                                 return (ENOBUFS);
2746                 }
2747                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2748                 break;
2749         }
2750
2751         case IPV6_2292HOPLIMIT:
2752         case IPV6_HOPLIMIT:
2753         {
2754                 int *hlimp;
2755
2756                 /*
2757                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2758                  * to simplify the ordering among hoplimit options.
2759                  */
2760                 if (optname == IPV6_HOPLIMIT && sticky)
2761                         return (ENOPROTOOPT);
2762
2763                 if (len != sizeof(int))
2764                         return (EINVAL);
2765                 hlimp = (int *)buf;
2766                 if (*hlimp < -1 || *hlimp > 255)
2767                         return (EINVAL);
2768
2769                 opt->ip6po_hlim = *hlimp;
2770                 break;
2771         }
2772
2773         case IPV6_TCLASS:
2774         {
2775                 int tclass;
2776
2777                 if (len != sizeof(int))
2778                         return (EINVAL);
2779                 tclass = *(int *)buf;
2780                 if (tclass < -1 || tclass > 255)
2781                         return (EINVAL);
2782
2783                 opt->ip6po_tclass = tclass;
2784                 break;
2785         }
2786
2787         case IPV6_2292NEXTHOP:
2788         case IPV6_NEXTHOP:
2789                 if (cred != NULL) {
2790                         error = priv_check_cred(cred,
2791                             PRIV_NETINET_SETHDROPTS, 0);
2792                         if (error)
2793                                 return (error);
2794                 }
2795
2796                 if (len == 0) { /* just remove the option */
2797                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2798                         break;
2799                 }
2800
2801                 /* check if cmsg_len is large enough for sa_len */
2802                 if (len < sizeof(struct sockaddr) || len < *buf)
2803                         return (EINVAL);
2804
2805                 switch (((struct sockaddr *)buf)->sa_family) {
2806                 case AF_INET6:
2807                 {
2808                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2809                         int error;
2810
2811                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2812                                 return (EINVAL);
2813
2814                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2815                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2816                                 return (EINVAL);
2817                         }
2818                         if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2819                             != 0) {
2820                                 return (error);
2821                         }
2822                         break;
2823                 }
2824                 case AF_LINK:   /* should eventually be supported */
2825                 default:
2826                         return (EAFNOSUPPORT);
2827                 }
2828
2829                 /* turn off the previous option, then set the new option. */
2830                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2831                 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2832                 if (opt->ip6po_nexthop == NULL)
2833                         return (ENOBUFS);
2834                 bcopy(buf, opt->ip6po_nexthop, *buf);
2835                 break;
2836
2837         case IPV6_2292HOPOPTS:
2838         case IPV6_HOPOPTS:
2839         {
2840                 struct ip6_hbh *hbh;
2841                 int hbhlen;
2842
2843                 /*
2844                  * XXX: We don't allow a non-privileged user to set ANY HbH
2845                  * options, since per-option restriction has too much
2846                  * overhead.
2847                  */
2848                 if (cred != NULL) {
2849                         error = priv_check_cred(cred,
2850                             PRIV_NETINET_SETHDROPTS, 0);
2851                         if (error)
2852                                 return (error);
2853                 }
2854
2855                 if (len == 0) {
2856                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2857                         break;  /* just remove the option */
2858                 }
2859
2860                 /* message length validation */
2861                 if (len < sizeof(struct ip6_hbh))
2862                         return (EINVAL);
2863                 hbh = (struct ip6_hbh *)buf;
2864                 hbhlen = (hbh->ip6h_len + 1) << 3;
2865                 if (len != hbhlen)
2866                         return (EINVAL);
2867
2868                 /* turn off the previous option, then set the new option. */
2869                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2870                 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2871                 if (opt->ip6po_hbh == NULL)
2872                         return (ENOBUFS);
2873                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2874
2875                 break;
2876         }
2877
2878         case IPV6_2292DSTOPTS:
2879         case IPV6_DSTOPTS:
2880         case IPV6_RTHDRDSTOPTS:
2881         {
2882                 struct ip6_dest *dest, **newdest = NULL;
2883                 int destlen;
2884
2885                 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2886                         error = priv_check_cred(cred,
2887                             PRIV_NETINET_SETHDROPTS, 0);
2888                         if (error)
2889                                 return (error);
2890                 }
2891
2892                 if (len == 0) {
2893                         ip6_clearpktopts(opt, optname);
2894                         break;  /* just remove the option */
2895                 }
2896
2897                 /* message length validation */
2898                 if (len < sizeof(struct ip6_dest))
2899                         return (EINVAL);
2900                 dest = (struct ip6_dest *)buf;
2901                 destlen = (dest->ip6d_len + 1) << 3;
2902                 if (len != destlen)
2903                         return (EINVAL);
2904
2905                 /*
2906                  * Determine the position that the destination options header
2907                  * should be inserted; before or after the routing header.
2908                  */
2909                 switch (optname) {
2910                 case IPV6_2292DSTOPTS:
2911                         /*
2912                          * The old advacned API is ambiguous on this point.
2913                          * Our approach is to determine the position based
2914                          * according to the existence of a routing header.
2915                          * Note, however, that this depends on the order of the
2916                          * extension headers in the ancillary data; the 1st
2917                          * part of the destination options header must appear
2918                          * before the routing header in the ancillary data,
2919                          * too.
2920                          * RFC3542 solved the ambiguity by introducing
2921                          * separate ancillary data or option types.
2922                          */
2923                         if (opt->ip6po_rthdr == NULL)
2924                                 newdest = &opt->ip6po_dest1;
2925                         else
2926                                 newdest = &opt->ip6po_dest2;
2927                         break;
2928                 case IPV6_RTHDRDSTOPTS:
2929                         newdest = &opt->ip6po_dest1;
2930                         break;
2931                 case IPV6_DSTOPTS:
2932                         newdest = &opt->ip6po_dest2;
2933                         break;
2934                 }
2935
2936                 /* turn off the previous option, then set the new option. */
2937                 ip6_clearpktopts(opt, optname);
2938                 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2939                 if (*newdest == NULL)
2940                         return (ENOBUFS);
2941                 bcopy(dest, *newdest, destlen);
2942
2943                 break;
2944         }
2945
2946         case IPV6_2292RTHDR:
2947         case IPV6_RTHDR:
2948         {
2949                 struct ip6_rthdr *rth;
2950                 int rthlen;
2951
2952                 if (len == 0) {
2953                         ip6_clearpktopts(opt, IPV6_RTHDR);
2954                         break;  /* just remove the option */
2955                 }
2956
2957                 /* message length validation */
2958                 if (len < sizeof(struct ip6_rthdr))
2959                         return (EINVAL);
2960                 rth = (struct ip6_rthdr *)buf;
2961                 rthlen = (rth->ip6r_len + 1) << 3;
2962                 if (len != rthlen)
2963                         return (EINVAL);
2964
2965                 switch (rth->ip6r_type) {
2966                 case IPV6_RTHDR_TYPE_0:
2967                         if (rth->ip6r_len == 0) /* must contain one addr */
2968                                 return (EINVAL);
2969                         if (rth->ip6r_len % 2) /* length must be even */
2970                                 return (EINVAL);
2971                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2972                                 return (EINVAL);
2973                         break;
2974                 default:
2975                         return (EINVAL);        /* not supported */
2976                 }
2977
2978                 /* turn off the previous option */
2979                 ip6_clearpktopts(opt, IPV6_RTHDR);
2980                 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2981                 if (opt->ip6po_rthdr == NULL)
2982                         return (ENOBUFS);
2983                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2984
2985                 break;
2986         }
2987
2988         case IPV6_USE_MIN_MTU:
2989                 if (len != sizeof(int))
2990                         return (EINVAL);
2991                 minmtupolicy = *(int *)buf;
2992                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2993                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2994                     minmtupolicy != IP6PO_MINMTU_ALL) {
2995                         return (EINVAL);
2996                 }
2997                 opt->ip6po_minmtu = minmtupolicy;
2998                 break;
2999
3000         case IPV6_DONTFRAG:
3001                 if (len != sizeof(int))
3002                         return (EINVAL);
3003
3004                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3005                         /*
3006                          * we ignore this option for TCP sockets.
3007                          * (RFC3542 leaves this case unspecified.)
3008                          */
3009                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3010                 } else
3011                         opt->ip6po_flags |= IP6PO_DONTFRAG;
3012                 break;
3013
3014         case IPV6_PREFER_TEMPADDR:
3015                 if (len != sizeof(int))
3016                         return (EINVAL);
3017                 preftemp = *(int *)buf;
3018                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3019                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3020                     preftemp != IP6PO_TEMPADDR_PREFER) {
3021                         return (EINVAL);
3022                 }
3023                 opt->ip6po_prefer_tempaddr = preftemp;
3024                 break;
3025
3026         default:
3027                 return (ENOPROTOOPT);
3028         } /* end of switch */
3029
3030         return (0);
3031 }
3032
3033 /*
3034  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3035  * packet to the input queue of a specified interface.  Note that this
3036  * calls the output routine of the loopback "driver", but with an interface
3037  * pointer that might NOT be &loif -- easier than replicating that code here.
3038  */
3039 void
3040 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3041 {
3042         struct mbuf *copym;
3043         struct ip6_hdr *ip6;
3044
3045         copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3046         if (copym == NULL)
3047                 return;
3048
3049         /*
3050          * Make sure to deep-copy IPv6 header portion in case the data
3051          * is in an mbuf cluster, so that we can safely override the IPv6
3052          * header portion later.
3053          */
3054         if (!M_WRITABLE(copym) ||
3055             copym->m_len < sizeof(struct ip6_hdr)) {
3056                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3057                 if (copym == NULL)
3058                         return;
3059         }
3060         ip6 = mtod(copym, struct ip6_hdr *);
3061         /*
3062          * clear embedded scope identifiers if necessary.
3063          * in6_clearscope will touch the addresses only when necessary.
3064          */
3065         in6_clearscope(&ip6->ip6_src);
3066         in6_clearscope(&ip6->ip6_dst);
3067         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3068                 copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3069                     CSUM_PSEUDO_HDR;
3070                 copym->m_pkthdr.csum_data = 0xffff;
3071         }
3072         if_simloop(ifp, copym, AF_INET6, 0);
3073 }
3074
3075 /*
3076  * Chop IPv6 header off from the payload.
3077  */
3078 static int
3079 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3080 {
3081         struct mbuf *mh;
3082         struct ip6_hdr *ip6;
3083
3084         ip6 = mtod(m, struct ip6_hdr *);
3085         if (m->m_len > sizeof(*ip6)) {
3086                 mh = m_gethdr(M_NOWAIT, MT_DATA);
3087                 if (mh == NULL) {
3088                         m_freem(m);
3089                         return ENOBUFS;
3090                 }
3091                 m_move_pkthdr(mh, m);
3092                 M_ALIGN(mh, sizeof(*ip6));
3093                 m->m_len -= sizeof(*ip6);
3094                 m->m_data += sizeof(*ip6);
3095                 mh->m_next = m;
3096                 m = mh;
3097                 m->m_len = sizeof(*ip6);
3098                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3099         }
3100         exthdrs->ip6e_ip6 = m;
3101         return 0;
3102 }
3103
3104 /*
3105  * Compute IPv6 extension header length.
3106  */
3107 int
3108 ip6_optlen(struct inpcb *in6p)
3109 {
3110         int len;
3111
3112         if (!in6p->in6p_outputopts)
3113                 return 0;
3114
3115         len = 0;
3116 #define elen(x) \
3117     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3118
3119         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3120         if (in6p->in6p_outputopts->ip6po_rthdr)
3121                 /* dest1 is valid with rthdr only */
3122                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3123         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3124         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3125         return len;
3126 #undef elen
3127 }