sys/netinet/ip_output.c

   1 /*-
   2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 4. Neither the name of the University nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  30  */
  31
  32 #include <sys/cdefs.h>
  33 __FBSDID("$FreeBSD$");
  34
  35 #include "opt_inet.h"
  36 #include "opt_ipfw.h"
  37 #include "opt_ipsec.h"
  38 #include "opt_mbuf_stress_test.h"
  39 #include "opt_mpath.h"
  40 #include "opt_route.h"
  41 #include "opt_sctp.h"
  42 #include "opt_rss.h"
  43
  44 #include <sys/param.h>
  45 #include <sys/systm.h>
  46 #include <sys/kernel.h>
  47 #include <sys/lock.h>
  48 #include <sys/malloc.h>
  49 #include <sys/mbuf.h>
  50 #include <sys/priv.h>
  51 #include <sys/proc.h>
  52 #include <sys/protosw.h>
  53 #include <sys/rmlock.h>
  54 #include <sys/sdt.h>
  55 #include <sys/socket.h>
  56 #include <sys/socketvar.h>
  57 #include <sys/sysctl.h>
  58 #include <sys/ucred.h>
  59
  60 #include <net/if.h>
  61 #include <net/if_var.h>
  62 #include <net/if_llatbl.h>
  63 #include <net/netisr.h>
  64 #include <net/pfil.h>
  65 #include <net/route.h>
  66 #include <net/flowtable.h>
  67 #ifdef RADIX_MPATH
  68 #include <net/radix_mpath.h>
  69 #endif
  70 #include <net/rss_config.h>
  71 #include <net/vnet.h>
  72
  73 #include <netinet/in.h>
  74 #include <netinet/in_kdtrace.h>
  75 #include <netinet/in_systm.h>
  76 #include <netinet/ip.h>
  77 #include <netinet/in_pcb.h>
  78 #include <netinet/in_rss.h>
  79 #include <netinet/in_var.h>
  80 #include <netinet/ip_var.h>
  81 #include <netinet/ip_options.h>
  82 #ifdef SCTP
  83 #include <netinet/sctp.h>
  84 #include <netinet/sctp_crc32.h>
  85 #endif
  86
  87 #ifdef IPSEC
  88 #include <netinet/ip_ipsec.h>
  89 #include <netipsec/ipsec.h>
  90 #endif /* IPSEC*/
  91
  92 #include <machine/in_cksum.h>
  93
  94 #include <security/mac/mac_framework.h>
  95
  96 #ifdef MBUF_STRESS_TEST
  97 static int mbuf_frag_size = 0;
  98 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
  99         &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
 100 #endif
 101
 102 static void     ip_mloopback(struct ifnet *, const struct mbuf *, int);
 103
 104
 105 extern int in_mcast_loop;
 106 extern  struct protosw inetsw[];
 107
 108 static inline int
 109 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
 110     struct sockaddr_in *dst, int *fibnum, int *error)
 111 {
 112         struct m_tag *fwd_tag = NULL;
 113         struct mbuf *m;
 114         struct in_addr odst;
 115         struct ip *ip;
 116
 117         m = *mp;
 118         ip = mtod(m, struct ip *);
 119
 120         /* Run through list of hooks for output packets. */
 121         odst.s_addr = ip->ip_dst.s_addr;
 122         *error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, inp);
 123         m = *mp;
 124         if ((*error) != 0 || m == NULL)
 125                 return 1; /* Finished */
 126
 127         ip = mtod(m, struct ip *);
 128
 129         /* See if destination IP address was changed by packet filter. */
 130         if (odst.s_addr != ip->ip_dst.s_addr) {
 131                 m->m_flags |= M_SKIP_FIREWALL;
 132                 /* If destination is now ourself drop to ip_input(). */
 133                 if (in_localip(ip->ip_dst)) {
 134                         m->m_flags |= M_FASTFWD_OURS;
 135                         if (m->m_pkthdr.rcvif == NULL)
 136                                 m->m_pkthdr.rcvif = V_loif;
 137                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 138                                 m->m_pkthdr.csum_flags |=
 139                                         CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 140                                 m->m_pkthdr.csum_data = 0xffff;
 141                         }
 142                         m->m_pkthdr.csum_flags |=
 143                                 CSUM_IP_CHECKED | CSUM_IP_VALID;
 144 #ifdef SCTP
 145                         if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 146                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 147 #endif
 148                         *error = netisr_queue(NETISR_IP, m);
 149                         return 1; /* Finished */
 150                 }
 151
 152                 bzero(dst, sizeof(*dst));
 153                 dst->sin_family = AF_INET;
 154                 dst->sin_len = sizeof(*dst);
 155                 dst->sin_addr = ip->ip_dst;
 156
 157                 return -1; /* Reloop */
 158         }
 159         /* See if fib was changed by packet filter. */
 160         if ((*fibnum) != M_GETFIB(m)) {
 161                 m->m_flags |= M_SKIP_FIREWALL;
 162                 *fibnum = M_GETFIB(m);
 163                 return -1; /* Reloop for FIB change */
 164         }
 165
 166         /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
 167         if (m->m_flags & M_FASTFWD_OURS) {
 168                 if (m->m_pkthdr.rcvif == NULL)
 169                         m->m_pkthdr.rcvif = V_loif;
 170                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 171                         m->m_pkthdr.csum_flags |=
 172                                 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 173                         m->m_pkthdr.csum_data = 0xffff;
 174                 }
 175 #ifdef SCTP
 176                 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 177                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 178 #endif
 179                 m->m_pkthdr.csum_flags |=
 180                         CSUM_IP_CHECKED | CSUM_IP_VALID;
 181
 182                 *error = netisr_queue(NETISR_IP, m);
 183                 return 1; /* Finished */
 184         }
 185         /* Or forward to some other address? */
 186         if ((m->m_flags & M_IP_NEXTHOP) &&
 187             ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
 188                 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
 189                 m->m_flags |= M_SKIP_FIREWALL;
 190                 m->m_flags &= ~M_IP_NEXTHOP;
 191                 m_tag_delete(m, fwd_tag);
 192
 193                 return -1; /* Reloop for CHANGE of dst */
 194         }
 195
 196         return 0;
 197 }
 198
 199 /*
 200  * IP output.  The packet in mbuf chain m contains a skeletal IP
 201  * header (with len, off, ttl, proto, tos, src, dst).
 202  * The mbuf chain containing the packet will be freed.
 203  * The mbuf opt, if present, will not be freed.
 204  * If route ro is present and has ro_rt initialized, route lookup would be
 205  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 206  * then result of route lookup is stored in ro->ro_rt.
 207  *
 208  * In the IP forwarding case, the packet will arrive with options already
 209  * inserted, so must have a NULL opt pointer.
 210  */
 211 int
 212 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
 213     struct ip_moptions *imo, struct inpcb *inp)
 214 {
 215         struct rm_priotracker in_ifa_tracker;
 216         struct ip *ip;
 217         struct ifnet *ifp = NULL;       /* keep compiler happy */
 218         struct mbuf *m0;
 219         int hlen = sizeof (struct ip);
 220         int mtu;
 221         int error = 0;
 222         struct sockaddr_in *dst;
 223         const struct sockaddr_in *gw;
 224         struct in_ifaddr *ia;
 225         int isbroadcast;
 226         uint16_t ip_len, ip_off;
 227         struct route iproute;
 228         struct rtentry *rte;    /* cache for ro->ro_rt */
 229         uint32_t fibnum;
 230         int have_ia_ref;
 231 #ifdef IPSEC
 232         int no_route_but_check_spd = 0;
 233 #endif
 234         M_ASSERTPKTHDR(m);
 235
 236         if (inp != NULL) {
 237                 INP_LOCK_ASSERT(inp);
 238                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 239                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 240                         m->m_pkthdr.flowid = inp->inp_flowid;
 241                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 242                 }
 243         }
 244
 245         if (ro == NULL) {
 246                 ro = &iproute;
 247                 bzero(ro, sizeof (*ro));
 248         }
 249
 250 #ifdef FLOWTABLE
 251         if (ro->ro_rt == NULL)
 252                 (void )flowtable_lookup(AF_INET, m, ro);
 253 #endif
 254
 255         if (opt) {
 256                 int len = 0;
 257                 m = ip_insertoptions(m, opt, &len);
 258                 if (len != 0)
 259                         hlen = len; /* ip->ip_hl is updated above */
 260         }
 261         ip = mtod(m, struct ip *);
 262         ip_len = ntohs(ip->ip_len);
 263         ip_off = ntohs(ip->ip_off);
 264
 265         if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
 266                 ip->ip_v = IPVERSION;
 267                 ip->ip_hl = hlen >> 2;
 268                 ip_fillid(ip);
 269                 IPSTAT_INC(ips_localout);
 270         } else {
 271                 /* Header already set, fetch hlen from there */
 272                 hlen = ip->ip_hl << 2;
 273         }
 274
 275         /*
 276          * dst/gw handling:
 277          *
 278          * dst can be rewritten but always points to &ro->ro_dst.
 279          * gw is readonly but can point either to dst OR rt_gateway,
 280          * therefore we need restore gw if we're redoing lookup.
 281          */
 282         gw = dst = (struct sockaddr_in *)&ro->ro_dst;
 283         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 284         rte = ro->ro_rt;
 285         if (rte == NULL) {
 286                 bzero(dst, sizeof(*dst));
 287                 dst->sin_family = AF_INET;
 288                 dst->sin_len = sizeof(*dst);
 289                 dst->sin_addr = ip->ip_dst;
 290         }
 291 again:
 292         /*
 293          * Validate route against routing table additions;
 294          * a better/more specific route might have been added.
 295          */
 296         if (inp)
 297                 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
 298         /*
 299          * If there is a cached route,
 300          * check that it is to the same destination
 301          * and is still up.  If not, free it and try again.
 302          * The address family should also be checked in case of sharing the
 303          * cache with IPv6.
 304          * Also check whether routing cache needs invalidation.
 305          */
 306         rte = ro->ro_rt;
 307         if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
 308                     rte->rt_ifp == NULL ||
 309                     !RT_LINK_IS_UP(rte->rt_ifp) ||
 310                           dst->sin_family != AF_INET ||
 311                           dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
 312                 RTFREE(rte);
 313                 rte = ro->ro_rt = (struct rtentry *)NULL;
 314         }
 315         ia = NULL;
 316         have_ia_ref = 0;
 317         /*
 318          * If routing to interface only, short circuit routing lookup.
 319          * The use of an all-ones broadcast address implies this; an
 320          * interface is specified by the broadcast address of an interface,
 321          * or the destination address of a ptp interface.
 322          */
 323         if (flags & IP_SENDONES) {
 324                 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
 325                                                       M_GETFIB(m)))) == NULL &&
 326                     (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 327                                                     M_GETFIB(m)))) == NULL) {
 328                         IPSTAT_INC(ips_noroute);
 329                         error = ENETUNREACH;
 330                         goto bad;
 331                 }
 332                 have_ia_ref = 1;
 333                 ip->ip_dst.s_addr = INADDR_BROADCAST;
 334                 dst->sin_addr = ip->ip_dst;
 335                 ifp = ia->ia_ifp;
 336                 ip->ip_ttl = 1;
 337                 isbroadcast = 1;
 338         } else if (flags & IP_ROUTETOIF) {
 339                 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 340                                                     M_GETFIB(m)))) == NULL &&
 341                     (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
 342                                                 M_GETFIB(m)))) == NULL) {
 343                         IPSTAT_INC(ips_noroute);
 344                         error = ENETUNREACH;
 345                         goto bad;
 346                 }
 347                 have_ia_ref = 1;
 348                 ifp = ia->ia_ifp;
 349                 ip->ip_ttl = 1;
 350                 isbroadcast = in_broadcast(dst->sin_addr, ifp);
 351         } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
 352             imo != NULL && imo->imo_multicast_ifp != NULL) {
 353                 /*
 354                  * Bypass the normal routing lookup for multicast
 355                  * packets if the interface is specified.
 356                  */
 357                 ifp = imo->imo_multicast_ifp;
 358                 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
 359                 if (ia)
 360                         have_ia_ref = 1;
 361                 isbroadcast = 0;        /* fool gcc */
 362         } else {
 363                 /*
 364                  * We want to do any cloning requested by the link layer,
 365                  * as this is probably required in all cases for correct
 366                  * operation (as it is for ARP).
 367                  */
 368                 if (rte == NULL) {
 369 #ifdef RADIX_MPATH
 370                         rtalloc_mpath_fib(ro,
 371                             ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
 372                             fibnum);
 373 #else
 374                         in_rtalloc_ign(ro, 0, fibnum);
 375 #endif
 376                         rte = ro->ro_rt;
 377                 }
 378                 if (rte == NULL ||
 379                     (rte->rt_flags & RTF_UP) == 0 ||
 380                     rte->rt_ifp == NULL ||
 381                     !RT_LINK_IS_UP(rte->rt_ifp)) {
 382 #ifdef IPSEC
 383                         /*
 384                          * There is no route for this packet, but it is
 385                          * possible that a matching SPD entry exists.
 386                          */
 387                         no_route_but_check_spd = 1;
 388                         mtu = 0; /* Silence GCC warning. */
 389                         goto sendit;
 390 #endif
 391                         IPSTAT_INC(ips_noroute);
 392                         error = EHOSTUNREACH;
 393                         goto bad;
 394                 }
 395                 ia = ifatoia(rte->rt_ifa);
 396                 ifp = rte->rt_ifp;
 397                 counter_u64_add(rte->rt_pksent, 1);
 398                 rt_update_ro_flags(ro);
 399                 if (rte->rt_flags & RTF_GATEWAY)
 400                         gw = (struct sockaddr_in *)rte->rt_gateway;
 401                 if (rte->rt_flags & RTF_HOST)
 402                         isbroadcast = (rte->rt_flags & RTF_BROADCAST);
 403                 else
 404                         isbroadcast = in_broadcast(gw->sin_addr, ifp);
 405         }
 406
 407         /*
 408          * Calculate MTU.  If we have a route that is up, use that,
 409          * otherwise use the interface's MTU.
 410          */
 411         if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
 412                 mtu = rte->rt_mtu;
 413         else
 414                 mtu = ifp->if_mtu;
 415         /* Catch a possible divide by zero later. */
 416         KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
 417             __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
 418
 419         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
 420                 m->m_flags |= M_MCAST;
 421                 /*
 422                  * IP destination address is multicast.  Make sure "gw"
 423                  * still points to the address in "ro".  (It may have been
 424                  * changed to point to a gateway address, above.)
 425                  */
 426                 gw = dst;
 427                 /*
 428                  * See if the caller provided any multicast options
 429                  */
 430                 if (imo != NULL) {
 431                         ip->ip_ttl = imo->imo_multicast_ttl;
 432                         if (imo->imo_multicast_vif != -1)
 433                                 ip->ip_src.s_addr =
 434                                     ip_mcast_src ?
 435                                     ip_mcast_src(imo->imo_multicast_vif) :
 436                                     INADDR_ANY;
 437                 } else
 438                         ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
 439                 /*
 440                  * Confirm that the outgoing interface supports multicast.
 441                  */
 442                 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
 443                         if ((ifp->if_flags & IFF_MULTICAST) == 0) {
 444                                 IPSTAT_INC(ips_noroute);
 445                                 error = ENETUNREACH;
 446                                 goto bad;
 447                         }
 448                 }
 449                 /*
 450                  * If source address not specified yet, use address
 451                  * of outgoing interface.
 452                  */
 453                 if (ip->ip_src.s_addr == INADDR_ANY) {
 454                         /* Interface may have no addresses. */
 455                         if (ia != NULL)
 456                                 ip->ip_src = IA_SIN(ia)->sin_addr;
 457                 }
 458
 459                 if ((imo == NULL && in_mcast_loop) ||
 460                     (imo && imo->imo_multicast_loop)) {
 461                         /*
 462                          * Loop back multicast datagram if not expressly
 463                          * forbidden to do so, even if we are not a member
 464                          * of the group; ip_input() will filter it later,
 465                          * thus deferring a hash lookup and mutex acquisition
 466                          * at the expense of a cheap copy using m_copym().
 467                          */
 468                         ip_mloopback(ifp, m, hlen);
 469                 } else {
 470                         /*
 471                          * If we are acting as a multicast router, perform
 472                          * multicast forwarding as if the packet had just
 473                          * arrived on the interface to which we are about
 474                          * to send.  The multicast forwarding function
 475                          * recursively calls this function, using the
 476                          * IP_FORWARDING flag to prevent infinite recursion.
 477                          *
 478                          * Multicasts that are looped back by ip_mloopback(),
 479                          * above, will be forwarded by the ip_input() routine,
 480                          * if necessary.
 481                          */
 482                         if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
 483                                 /*
 484                                  * If rsvp daemon is not running, do not
 485                                  * set ip_moptions. This ensures that the packet
 486                                  * is multicast and not just sent down one link
 487                                  * as prescribed by rsvpd.
 488                                  */
 489                                 if (!V_rsvp_on)
 490                                         imo = NULL;
 491                                 if (ip_mforward &&
 492                                     ip_mforward(ip, ifp, m, imo) != 0) {
 493                                         m_freem(m);
 494                                         goto done;
 495                                 }
 496                         }
 497                 }
 498
 499                 /*
 500                  * Multicasts with a time-to-live of zero may be looped-
 501                  * back, above, but must not be transmitted on a network.
 502                  * Also, multicasts addressed to the loopback interface
 503                  * are not sent -- the above call to ip_mloopback() will
 504                  * loop back a copy. ip_input() will drop the copy if
 505                  * this host does not belong to the destination group on
 506                  * the loopback interface.
 507                  */
 508                 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
 509                         m_freem(m);
 510                         goto done;
 511                 }
 512
 513                 goto sendit;
 514         }
 515
 516         /*
 517          * If the source address is not specified yet, use the address
 518          * of the outoing interface.
 519          */
 520         if (ip->ip_src.s_addr == INADDR_ANY) {
 521                 /* Interface may have no addresses. */
 522                 if (ia != NULL) {
 523                         ip->ip_src = IA_SIN(ia)->sin_addr;
 524                 }
 525         }
 526
 527         /*
 528          * Look for broadcast address and
 529          * verify user is allowed to send
 530          * such a packet.
 531          */
 532         if (isbroadcast) {
 533                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 534                         error = EADDRNOTAVAIL;
 535                         goto bad;
 536                 }
 537                 if ((flags & IP_ALLOWBROADCAST) == 0) {
 538                         error = EACCES;
 539                         goto bad;
 540                 }
 541                 /* don't allow broadcast messages to be fragmented */
 542                 if (ip_len > mtu) {
 543                         error = EMSGSIZE;
 544                         goto bad;
 545                 }
 546                 m->m_flags |= M_BCAST;
 547         } else {
 548                 m->m_flags &= ~M_BCAST;
 549         }
 550
 551 sendit:
 552 #ifdef IPSEC
 553         switch(ip_ipsec_output(&m, inp, &error)) {
 554         case 1:
 555                 goto bad;
 556         case -1:
 557                 goto done;
 558         case 0:
 559         default:
 560                 break;  /* Continue with packet processing. */
 561         }
 562         /*
 563          * Check if there was a route for this packet; return error if not.
 564          */
 565         if (no_route_but_check_spd) {
 566                 IPSTAT_INC(ips_noroute);
 567                 error = EHOSTUNREACH;
 568                 goto bad;
 569         }
 570         /* Update variables that are affected by ipsec4_output(). */
 571         ip = mtod(m, struct ip *);
 572         hlen = ip->ip_hl << 2;
 573 #endif /* IPSEC */
 574
 575         /* Jump over all PFIL processing if hooks are not active. */
 576         if (PFIL_HOOKED(&V_inet_pfil_hook)) {
 577                 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
 578                 case 1: /* Finished */
 579                         goto done;
 580
 581                 case 0: /* Continue normally */
 582                         ip = mtod(m, struct ip *);
 583                         break;
 584
 585                 case -1: /* Need to try again */
 586                         /* Reset everything for a new round */
 587                         RO_RTFREE(ro);
 588                         if (have_ia_ref)
 589                                 ifa_free(&ia->ia_ifa);
 590                         ro->ro_prepend = NULL;
 591                         rte = NULL;
 592                         gw = dst;
 593                         ip = mtod(m, struct ip *);
 594                         goto again;
 595
 596                 }
 597         }
 598
 599         /* 127/8 must not appear on wire - RFC1122. */
 600         if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
 601             (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
 602                 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
 603                         IPSTAT_INC(ips_badaddr);
 604                         error = EADDRNOTAVAIL;
 605                         goto bad;
 606                 }
 607         }
 608
 609         m->m_pkthdr.csum_flags |= CSUM_IP;
 610         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
 611                 in_delayed_cksum(m);
 612                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 613         }
 614 #ifdef SCTP
 615         if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
 616                 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
 617                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 618         }
 619 #endif
 620
 621         /*
 622          * If small enough for interface, or the interface will take
 623          * care of the fragmentation for us, we can just send directly.
 624          */
 625         if (ip_len <= mtu ||
 626             (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
 627                 ip->ip_sum = 0;
 628                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
 629                         ip->ip_sum = in_cksum(m, hlen);
 630                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 631                 }
 632
 633                 /*
 634                  * Record statistics for this interface address.
 635                  * With CSUM_TSO the byte/packet count will be slightly
 636                  * incorrect because we count the IP+TCP headers only
 637                  * once instead of for every generated packet.
 638                  */
 639                 if (!(flags & IP_FORWARDING) && ia) {
 640                         if (m->m_pkthdr.csum_flags & CSUM_TSO)
 641                                 counter_u64_add(ia->ia_ifa.ifa_opackets,
 642                                     m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
 643                         else
 644                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 645
 646                         counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
 647                 }
 648 #ifdef MBUF_STRESS_TEST
 649                 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
 650                         m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
 651 #endif
 652                 /*
 653                  * Reset layer specific mbuf flags
 654                  * to avoid confusing lower layers.
 655                  */
 656                 m_clrprotoflags(m);
 657                 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
 658                 error = (*ifp->if_output)(ifp, m,
 659                     (const struct sockaddr *)gw, ro);
 660                 goto done;
 661         }
 662
 663         /* Balk when DF bit is set or the interface didn't support TSO. */
 664         if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
 665                 error = EMSGSIZE;
 666                 IPSTAT_INC(ips_cantfrag);
 667                 goto bad;
 668         }
 669
 670         /*
 671          * Too large for interface; fragment if possible. If successful,
 672          * on return, m will point to a list of packets to be sent.
 673          */
 674         error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
 675         if (error)
 676                 goto bad;
 677         for (; m; m = m0) {
 678                 m0 = m->m_nextpkt;
 679                 m->m_nextpkt = 0;
 680                 if (error == 0) {
 681                         /* Record statistics for this interface address. */
 682                         if (ia != NULL) {
 683                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 684                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
 685                                     m->m_pkthdr.len);
 686                         }
 687                         /*
 688                          * Reset layer specific mbuf flags
 689                          * to avoid confusing upper layers.
 690                          */
 691                         m_clrprotoflags(m);
 692
 693                         IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
 694                         error = (*ifp->if_output)(ifp, m,
 695                             (const struct sockaddr *)gw, ro);
 696                 } else
 697                         m_freem(m);
 698         }
 699
 700         if (error == 0)
 701                 IPSTAT_INC(ips_fragmented);
 702
 703 done:
 704         if (ro == &iproute)
 705                 RO_RTFREE(ro);
 706         else if (rte == NULL)
 707                 /*
 708                  * If the caller supplied a route but somehow the reference
 709                  * to it has been released need to prevent the caller
 710                  * calling RTFREE on it again.
 711                  */
 712                 ro->ro_rt = NULL;
 713         if (have_ia_ref)
 714                 ifa_free(&ia->ia_ifa);
 715         return (error);
 716 bad:
 717         m_freem(m);
 718         goto done;
 719 }
 720
 721 /*
 722  * Create a chain of fragments which fit the given mtu. m_frag points to the
 723  * mbuf to be fragmented; on return it points to the chain with the fragments.
 724  * Return 0 if no error. If error, m_frag may contain a partially built
 725  * chain of fragments that should be freed by the caller.
 726  *
 727  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
 728  */
 729 int
 730 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
 731     u_long if_hwassist_flags)
 732 {
 733         int error = 0;
 734         int hlen = ip->ip_hl << 2;
 735         int len = (mtu - hlen) & ~7;    /* size of payload in each fragment */
 736         int off;
 737         struct mbuf *m0 = *m_frag;      /* the original packet          */
 738         int firstlen;
 739         struct mbuf **mnext;
 740         int nfrags;
 741         uint16_t ip_len, ip_off;
 742
 743         ip_len = ntohs(ip->ip_len);
 744         ip_off = ntohs(ip->ip_off);
 745
 746         if (ip_off & IP_DF) {   /* Fragmentation not allowed */
 747                 IPSTAT_INC(ips_cantfrag);
 748                 return EMSGSIZE;
 749         }
 750
 751         /*
 752          * Must be able to put at least 8 bytes per fragment.
 753          */
 754         if (len < 8)
 755                 return EMSGSIZE;
 756
 757         /*
 758          * If the interface will not calculate checksums on
 759          * fragmented packets, then do it here.
 760          */
 761         if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 762                 in_delayed_cksum(m0);
 763                 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 764         }
 765 #ifdef SCTP
 766         if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
 767                 sctp_delayed_cksum(m0, hlen);
 768                 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 769         }
 770 #endif
 771         if (len > PAGE_SIZE) {
 772                 /*
 773                  * Fragment large datagrams such that each segment
 774                  * contains a multiple of PAGE_SIZE amount of data,
 775                  * plus headers. This enables a receiver to perform
 776                  * page-flipping zero-copy optimizations.
 777                  *
 778                  * XXX When does this help given that sender and receiver
 779                  * could have different page sizes, and also mtu could
 780                  * be less than the receiver's page size ?
 781                  */
 782                 int newlen;
 783
 784                 off = MIN(mtu, m0->m_pkthdr.len);
 785
 786                 /*
 787                  * firstlen (off - hlen) must be aligned on an
 788                  * 8-byte boundary
 789                  */
 790                 if (off < hlen)
 791                         goto smart_frag_failure;
 792                 off = ((off - hlen) & ~7) + hlen;
 793                 newlen = (~PAGE_MASK) & mtu;
 794                 if ((newlen + sizeof (struct ip)) > mtu) {
 795                         /* we failed, go back the default */
 796 smart_frag_failure:
 797                         newlen = len;
 798                         off = hlen + len;
 799                 }
 800                 len = newlen;
 801
 802         } else {
 803                 off = hlen + len;
 804         }
 805
 806         firstlen = off - hlen;
 807         mnext = &m0->m_nextpkt;         /* pointer to next packet */
 808
 809         /*
 810          * Loop through length of segment after first fragment,
 811          * make new header and copy data of each part and link onto chain.
 812          * Here, m0 is the original packet, m is the fragment being created.
 813          * The fragments are linked off the m_nextpkt of the original
 814          * packet, which after processing serves as the first fragment.
 815          */
 816         for (nfrags = 1; off < ip_len; off += len, nfrags++) {
 817                 struct ip *mhip;        /* ip header on the fragment */
 818                 struct mbuf *m;
 819                 int mhlen = sizeof (struct ip);
 820
 821                 m = m_gethdr(M_NOWAIT, MT_DATA);
 822                 if (m == NULL) {
 823                         error = ENOBUFS;
 824                         IPSTAT_INC(ips_odropped);
 825                         goto done;
 826                 }
 827                 /*
 828                  * Make sure the complete packet header gets copied
 829                  * from the originating mbuf to the newly created
 830                  * mbuf. This also ensures that existing firewall
 831                  * classification(s), VLAN tags and so on get copied
 832                  * to the resulting fragmented packet(s):
 833                  */
 834                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 835                         m_free(m);
 836                         error = ENOBUFS;
 837                         IPSTAT_INC(ips_odropped);
 838                         goto done;
 839                 }
 840                 /*
 841                  * In the first mbuf, leave room for the link header, then
 842                  * copy the original IP header including options. The payload
 843                  * goes into an additional mbuf chain returned by m_copym().
 844                  */
 845                 m->m_data += max_linkhdr;
 846                 mhip = mtod(m, struct ip *);
 847                 *mhip = *ip;
 848                 if (hlen > sizeof (struct ip)) {
 849                         mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
 850                         mhip->ip_v = IPVERSION;
 851                         mhip->ip_hl = mhlen >> 2;
 852                 }
 853                 m->m_len = mhlen;
 854                 /* XXX do we need to add ip_off below ? */
 855                 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
 856                 if (off + len >= ip_len)
 857                         len = ip_len - off;
 858                 else
 859                         mhip->ip_off |= IP_MF;
 860                 mhip->ip_len = htons((u_short)(len + mhlen));
 861                 m->m_next = m_copym(m0, off, len, M_NOWAIT);
 862                 if (m->m_next == NULL) {        /* copy failed */
 863                         m_free(m);
 864                         error = ENOBUFS;        /* ??? */
 865                         IPSTAT_INC(ips_odropped);
 866                         goto done;
 867                 }
 868                 m->m_pkthdr.len = mhlen + len;
 869 #ifdef MAC
 870                 mac_netinet_fragment(m0, m);
 871 #endif
 872                 mhip->ip_off = htons(mhip->ip_off);
 873                 mhip->ip_sum = 0;
 874                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 875                         mhip->ip_sum = in_cksum(m, mhlen);
 876                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 877                 }
 878                 *mnext = m;
 879                 mnext = &m->m_nextpkt;
 880         }
 881         IPSTAT_ADD(ips_ofragments, nfrags);
 882
 883         /*
 884          * Update first fragment by trimming what's been copied out
 885          * and updating header.
 886          */
 887         m_adj(m0, hlen + firstlen - ip_len);
 888         m0->m_pkthdr.len = hlen + firstlen;
 889         ip->ip_len = htons((u_short)m0->m_pkthdr.len);
 890         ip->ip_off = htons(ip_off | IP_MF);
 891         ip->ip_sum = 0;
 892         if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 893                 ip->ip_sum = in_cksum(m0, hlen);
 894                 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
 895         }
 896
 897 done:
 898         *m_frag = m0;
 899         return error;
 900 }
 901
 902 void
 903 in_delayed_cksum(struct mbuf *m)
 904 {
 905         struct ip *ip;
 906         uint16_t csum, offset, ip_len;
 907
 908         ip = mtod(m, struct ip *);
 909         offset = ip->ip_hl << 2 ;
 910         ip_len = ntohs(ip->ip_len);
 911         csum = in_cksum_skip(m, ip_len, offset);
 912         if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
 913                 csum = 0xffff;
 914         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 915
 916         /* find the mbuf in the chain where the checksum starts*/
 917         while ((m != NULL) && (offset >= m->m_len)) {
 918                 offset -= m->m_len;
 919                 m = m->m_next;
 920         }
 921         KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
 922         KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
 923         *(u_short *)(m->m_data + offset) = csum;
 924 }
 925
 926 /*
 927  * IP socket option processing.
 928  */
 929 int
 930 ip_ctloutput(struct socket *so, struct sockopt *sopt)
 931 {
 932         struct  inpcb *inp = sotoinpcb(so);
 933         int     error, optval;
 934 #ifdef  RSS
 935         uint32_t rss_bucket;
 936         int retval;
 937 #endif
 938
 939         error = optval = 0;
 940         if (sopt->sopt_level != IPPROTO_IP) {
 941                 error = EINVAL;
 942
 943                 if (sopt->sopt_level == SOL_SOCKET &&
 944                     sopt->sopt_dir == SOPT_SET) {
 945                         switch (sopt->sopt_name) {
 946                         case SO_REUSEADDR:
 947                                 INP_WLOCK(inp);
 948                                 if ((so->so_options & SO_REUSEADDR) != 0)
 949                                         inp->inp_flags2 |= INP_REUSEADDR;
 950                                 else
 951                                         inp->inp_flags2 &= ~INP_REUSEADDR;
 952                                 INP_WUNLOCK(inp);
 953                                 error = 0;
 954                                 break;
 955                         case SO_REUSEPORT:
 956                                 INP_WLOCK(inp);
 957                                 if ((so->so_options & SO_REUSEPORT) != 0)
 958                                         inp->inp_flags2 |= INP_REUSEPORT;
 959                                 else
 960                                         inp->inp_flags2 &= ~INP_REUSEPORT;
 961                                 INP_WUNLOCK(inp);
 962                                 error = 0;
 963                                 break;
 964                         case SO_SETFIB:
 965                                 INP_WLOCK(inp);
 966                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
 967                                 INP_WUNLOCK(inp);
 968                                 error = 0;
 969                                 break;
 970                         default:
 971                                 break;
 972                         }
 973                 }
 974                 return (error);
 975         }
 976
 977         switch (sopt->sopt_dir) {
 978         case SOPT_SET:
 979                 switch (sopt->sopt_name) {
 980                 case IP_OPTIONS:
 981 #ifdef notyet
 982                 case IP_RETOPTS:
 983 #endif
 984                 {
 985                         struct mbuf *m;
 986                         if (sopt->sopt_valsize > MLEN) {
 987                                 error = EMSGSIZE;
 988                                 break;
 989                         }
 990                         m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
 991                         if (m == NULL) {
 992                                 error = ENOBUFS;
 993                                 break;
 994                         }
 995                         m->m_len = sopt->sopt_valsize;
 996                         error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
 997                                             m->m_len);
 998                         if (error) {
 999                                 m_free(m);
1000                                 break;
1001                         }
1002                         INP_WLOCK(inp);
1003                         error = ip_pcbopts(inp, sopt->sopt_name, m);
1004                         INP_WUNLOCK(inp);
1005                         return (error);
1006                 }
1007
1008                 case IP_BINDANY:
1009                         if (sopt->sopt_td != NULL) {
1010                                 error = priv_check(sopt->sopt_td,
1011                                     PRIV_NETINET_BINDANY);
1012                                 if (error)
1013                                         break;
1014                         }
1015                         /* FALLTHROUGH */
1016                 case IP_BINDMULTI:
1017 #ifdef  RSS
1018                 case IP_RSS_LISTEN_BUCKET:
1019 #endif
1020                 case IP_TOS:
1021                 case IP_TTL:
1022                 case IP_MINTTL:
1023                 case IP_RECVOPTS:
1024                 case IP_RECVRETOPTS:
1025                 case IP_RECVDSTADDR:
1026                 case IP_RECVTTL:
1027                 case IP_RECVIF:
1028                 case IP_ONESBCAST:
1029                 case IP_DONTFRAG:
1030                 case IP_RECVTOS:
1031                 case IP_RECVFLOWID:
1032 #ifdef  RSS
1033                 case IP_RECVRSSBUCKETID:
1034 #endif
1035                         error = sooptcopyin(sopt, &optval, sizeof optval,
1036                                             sizeof optval);
1037                         if (error)
1038                                 break;
1039
1040                         switch (sopt->sopt_name) {
1041                         case IP_TOS:
1042                                 inp->inp_ip_tos = optval;
1043                                 break;
1044
1045                         case IP_TTL:
1046                                 inp->inp_ip_ttl = optval;
1047                                 break;
1048
1049                         case IP_MINTTL:
1050                                 if (optval >= 0 && optval <= MAXTTL)
1051                                         inp->inp_ip_minttl = optval;
1052                                 else
1053                                         error = EINVAL;
1054                                 break;
1055
1056 #define OPTSET(bit) do {                                                \
1057         INP_WLOCK(inp);                                                 \
1058         if (optval)                                                     \
1059                 inp->inp_flags |= bit;                                  \
1060         else                                                            \
1061                 inp->inp_flags &= ~bit;                                 \
1062         INP_WUNLOCK(inp);                                               \
1063 } while (0)
1064
1065 #define OPTSET2(bit, val) do {                                          \
1066         INP_WLOCK(inp);                                                 \
1067         if (val)                                                        \
1068                 inp->inp_flags2 |= bit;                                 \
1069         else                                                            \
1070                 inp->inp_flags2 &= ~bit;                                \
1071         INP_WUNLOCK(inp);                                               \
1072 } while (0)
1073
1074                         case IP_RECVOPTS:
1075                                 OPTSET(INP_RECVOPTS);
1076                                 break;
1077
1078                         case IP_RECVRETOPTS:
1079                                 OPTSET(INP_RECVRETOPTS);
1080                                 break;
1081
1082                         case IP_RECVDSTADDR:
1083                                 OPTSET(INP_RECVDSTADDR);
1084                                 break;
1085
1086                         case IP_RECVTTL:
1087                                 OPTSET(INP_RECVTTL);
1088                                 break;
1089
1090                         case IP_RECVIF:
1091                                 OPTSET(INP_RECVIF);
1092                                 break;
1093
1094                         case IP_ONESBCAST:
1095                                 OPTSET(INP_ONESBCAST);
1096                                 break;
1097                         case IP_DONTFRAG:
1098                                 OPTSET(INP_DONTFRAG);
1099                                 break;
1100                         case IP_BINDANY:
1101                                 OPTSET(INP_BINDANY);
1102                                 break;
1103                         case IP_RECVTOS:
1104                                 OPTSET(INP_RECVTOS);
1105                                 break;
1106                         case IP_BINDMULTI:
1107                                 OPTSET2(INP_BINDMULTI, optval);
1108                                 break;
1109                         case IP_RECVFLOWID:
1110                                 OPTSET2(INP_RECVFLOWID, optval);
1111                                 break;
1112 #ifdef  RSS
1113                         case IP_RSS_LISTEN_BUCKET:
1114                                 if ((optval >= 0) &&
1115                                     (optval < rss_getnumbuckets())) {
1116                                         inp->inp_rss_listen_bucket = optval;
1117                                         OPTSET2(INP_RSS_BUCKET_SET, 1);
1118                                 } else {
1119                                         error = EINVAL;
1120                                 }
1121                                 break;
1122                         case IP_RECVRSSBUCKETID:
1123                                 OPTSET2(INP_RECVRSSBUCKETID, optval);
1124                                 break;
1125 #endif
1126                         }
1127                         break;
1128 #undef OPTSET
1129 #undef OPTSET2
1130
1131                 /*
1132                  * Multicast socket options are processed by the in_mcast
1133                  * module.
1134                  */
1135                 case IP_MULTICAST_IF:
1136                 case IP_MULTICAST_VIF:
1137                 case IP_MULTICAST_TTL:
1138                 case IP_MULTICAST_LOOP:
1139                 case IP_ADD_MEMBERSHIP:
1140                 case IP_DROP_MEMBERSHIP:
1141                 case IP_ADD_SOURCE_MEMBERSHIP:
1142                 case IP_DROP_SOURCE_MEMBERSHIP:
1143                 case IP_BLOCK_SOURCE:
1144                 case IP_UNBLOCK_SOURCE:
1145                 case IP_MSFILTER:
1146                 case MCAST_JOIN_GROUP:
1147                 case MCAST_LEAVE_GROUP:
1148                 case MCAST_JOIN_SOURCE_GROUP:
1149                 case MCAST_LEAVE_SOURCE_GROUP:
1150                 case MCAST_BLOCK_SOURCE:
1151                 case MCAST_UNBLOCK_SOURCE:
1152                         error = inp_setmoptions(inp, sopt);
1153                         break;
1154
1155                 case IP_PORTRANGE:
1156                         error = sooptcopyin(sopt, &optval, sizeof optval,
1157                                             sizeof optval);
1158                         if (error)
1159                                 break;
1160
1161                         INP_WLOCK(inp);
1162                         switch (optval) {
1163                         case IP_PORTRANGE_DEFAULT:
1164                                 inp->inp_flags &= ~(INP_LOWPORT);
1165                                 inp->inp_flags &= ~(INP_HIGHPORT);
1166                                 break;
1167
1168                         case IP_PORTRANGE_HIGH:
1169                                 inp->inp_flags &= ~(INP_LOWPORT);
1170                                 inp->inp_flags |= INP_HIGHPORT;
1171                                 break;
1172
1173                         case IP_PORTRANGE_LOW:
1174                                 inp->inp_flags &= ~(INP_HIGHPORT);
1175                                 inp->inp_flags |= INP_LOWPORT;
1176                                 break;
1177
1178                         default:
1179                                 error = EINVAL;
1180                                 break;
1181                         }
1182                         INP_WUNLOCK(inp);
1183                         break;
1184
1185 #ifdef IPSEC
1186                 case IP_IPSEC_POLICY:
1187                 {
1188                         caddr_t req;
1189                         struct mbuf *m;
1190
1191                         if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1192                                 break;
1193                         if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1194                                 break;
1195                         req = mtod(m, caddr_t);
1196                         error = ipsec_set_policy(inp, sopt->sopt_name, req,
1197                             m->m_len, (sopt->sopt_td != NULL) ?
1198                             sopt->sopt_td->td_ucred : NULL);
1199                         m_freem(m);
1200                         break;
1201                 }
1202 #endif /* IPSEC */
1203
1204                 default:
1205                         error = ENOPROTOOPT;
1206                         break;
1207                 }
1208                 break;
1209
1210         case SOPT_GET:
1211                 switch (sopt->sopt_name) {
1212                 case IP_OPTIONS:
1213                 case IP_RETOPTS:
1214                         if (inp->inp_options)
1215                                 error = sooptcopyout(sopt,
1216                                                      mtod(inp->inp_options,
1217                                                           char *),
1218                                                      inp->inp_options->m_len);
1219                         else
1220                                 sopt->sopt_valsize = 0;
1221                         break;
1222
1223                 case IP_TOS:
1224                 case IP_TTL:
1225                 case IP_MINTTL:
1226                 case IP_RECVOPTS:
1227                 case IP_RECVRETOPTS:
1228                 case IP_RECVDSTADDR:
1229                 case IP_RECVTTL:
1230                 case IP_RECVIF:
1231                 case IP_PORTRANGE:
1232                 case IP_ONESBCAST:
1233                 case IP_DONTFRAG:
1234                 case IP_BINDANY:
1235                 case IP_RECVTOS:
1236                 case IP_BINDMULTI:
1237                 case IP_FLOWID:
1238                 case IP_FLOWTYPE:
1239                 case IP_RECVFLOWID:
1240 #ifdef  RSS
1241                 case IP_RSSBUCKETID:
1242                 case IP_RECVRSSBUCKETID:
1243 #endif
1244                         switch (sopt->sopt_name) {
1245
1246                         case IP_TOS:
1247                                 optval = inp->inp_ip_tos;
1248                                 break;
1249
1250                         case IP_TTL:
1251                                 optval = inp->inp_ip_ttl;
1252                                 break;
1253
1254                         case IP_MINTTL:
1255                                 optval = inp->inp_ip_minttl;
1256                                 break;
1257
1258 #define OPTBIT(bit)     (inp->inp_flags & bit ? 1 : 0)
1259 #define OPTBIT2(bit)    (inp->inp_flags2 & bit ? 1 : 0)
1260
1261                         case IP_RECVOPTS:
1262                                 optval = OPTBIT(INP_RECVOPTS);
1263                                 break;
1264
1265                         case IP_RECVRETOPTS:
1266                                 optval = OPTBIT(INP_RECVRETOPTS);
1267                                 break;
1268
1269                         case IP_RECVDSTADDR:
1270                                 optval = OPTBIT(INP_RECVDSTADDR);
1271                                 break;
1272
1273                         case IP_RECVTTL:
1274                                 optval = OPTBIT(INP_RECVTTL);
1275                                 break;
1276
1277                         case IP_RECVIF:
1278                                 optval = OPTBIT(INP_RECVIF);
1279                                 break;
1280
1281                         case IP_PORTRANGE:
1282                                 if (inp->inp_flags & INP_HIGHPORT)
1283                                         optval = IP_PORTRANGE_HIGH;
1284                                 else if (inp->inp_flags & INP_LOWPORT)
1285                                         optval = IP_PORTRANGE_LOW;
1286                                 else
1287                                         optval = 0;
1288                                 break;
1289
1290                         case IP_ONESBCAST:
1291                                 optval = OPTBIT(INP_ONESBCAST);
1292                                 break;
1293                         case IP_DONTFRAG:
1294                                 optval = OPTBIT(INP_DONTFRAG);
1295                                 break;
1296                         case IP_BINDANY:
1297                                 optval = OPTBIT(INP_BINDANY);
1298                                 break;
1299                         case IP_RECVTOS:
1300                                 optval = OPTBIT(INP_RECVTOS);
1301                                 break;
1302                         case IP_FLOWID:
1303                                 optval = inp->inp_flowid;
1304                                 break;
1305                         case IP_FLOWTYPE:
1306                                 optval = inp->inp_flowtype;
1307                                 break;
1308                         case IP_RECVFLOWID:
1309                                 optval = OPTBIT2(INP_RECVFLOWID);
1310                                 break;
1311 #ifdef  RSS
1312                         case IP_RSSBUCKETID:
1313                                 retval = rss_hash2bucket(inp->inp_flowid,
1314                                     inp->inp_flowtype,
1315                                     &rss_bucket);
1316                                 if (retval == 0)
1317                                         optval = rss_bucket;
1318                                 else
1319                                         error = EINVAL;
1320                                 break;
1321                         case IP_RECVRSSBUCKETID:
1322                                 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1323                                 break;
1324 #endif
1325                         case IP_BINDMULTI:
1326                                 optval = OPTBIT2(INP_BINDMULTI);
1327                                 break;
1328                         }
1329                         error = sooptcopyout(sopt, &optval, sizeof optval);
1330                         break;
1331
1332                 /*
1333                  * Multicast socket options are processed by the in_mcast
1334                  * module.
1335                  */
1336                 case IP_MULTICAST_IF:
1337                 case IP_MULTICAST_VIF:
1338                 case IP_MULTICAST_TTL:
1339                 case IP_MULTICAST_LOOP:
1340                 case IP_MSFILTER:
1341                         error = inp_getmoptions(inp, sopt);
1342                         break;
1343
1344 #ifdef IPSEC
1345                 case IP_IPSEC_POLICY:
1346                 {
1347                         struct mbuf *m = NULL;
1348                         caddr_t req = NULL;
1349                         size_t len = 0;
1350
1351                         if (m != NULL) {
1352                                 req = mtod(m, caddr_t);
1353                                 len = m->m_len;
1354                         }
1355                         error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1356                         if (error == 0)
1357                                 error = soopt_mcopyout(sopt, m); /* XXX */
1358                         if (error == 0)
1359                                 m_freem(m);
1360                         break;
1361                 }
1362 #endif /* IPSEC */
1363
1364                 default:
1365                         error = ENOPROTOOPT;
1366                         break;
1367                 }
1368                 break;
1369         }
1370         return (error);
1371 }
1372
1373 /*
1374  * Routine called from ip_output() to loop back a copy of an IP multicast
1375  * packet to the input queue of a specified interface.  Note that this
1376  * calls the output routine of the loopback "driver", but with an interface
1377  * pointer that might NOT be a loopback interface -- evil, but easier than
1378  * replicating that code here.
1379  */
1380 static void
1381 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1382 {
1383         struct ip *ip;
1384         struct mbuf *copym;
1385
1386         /*
1387          * Make a deep copy of the packet because we're going to
1388          * modify the pack in order to generate checksums.
1389          */
1390         copym = m_dup(m, M_NOWAIT);
1391         if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1392                 copym = m_pullup(copym, hlen);
1393         if (copym != NULL) {
1394                 /* If needed, compute the checksum and mark it as valid. */
1395                 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1396                         in_delayed_cksum(copym);
1397                         copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1398                         copym->m_pkthdr.csum_flags |=
1399                             CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1400                         copym->m_pkthdr.csum_data = 0xffff;
1401                 }
1402                 /*
1403                  * We don't bother to fragment if the IP length is greater
1404                  * than the interface's MTU.  Can this possibly matter?
1405                  */
1406                 ip = mtod(copym, struct ip *);
1407                 ip->ip_sum = 0;
1408                 ip->ip_sum = in_cksum(copym, hlen);
1409                 if_simloop(ifp, copym, AF_INET, 0);
1410         }
1411 }