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/malloc.h>
  48 #include <sys/mbuf.h>
  49 #include <sys/priv.h>
  50 #include <sys/proc.h>
  51 #include <sys/protosw.h>
  52 #include <sys/sdt.h>
  53 #include <sys/socket.h>
  54 #include <sys/socketvar.h>
  55 #include <sys/sysctl.h>
  56 #include <sys/ucred.h>
  57
  58 #include <net/if.h>
  59 #include <net/if_var.h>
  60 #include <net/if_llatbl.h>
  61 #include <net/netisr.h>
  62 #include <net/pfil.h>
  63 #include <net/route.h>
  64 #include <net/flowtable.h>
  65 #ifdef RADIX_MPATH
  66 #include <net/radix_mpath.h>
  67 #endif
  68 #include <net/vnet.h>
  69
  70 #include <netinet/in.h>
  71 #include <netinet/in_kdtrace.h>
  72 #include <netinet/in_systm.h>
  73 #include <netinet/ip.h>
  74 #include <netinet/in_pcb.h>
  75 #include <netinet/in_rss.h>
  76 #include <netinet/in_var.h>
  77 #include <netinet/ip_var.h>
  78 #include <netinet/ip_options.h>
  79 #ifdef SCTP
  80 #include <netinet/sctp.h>
  81 #include <netinet/sctp_crc32.h>
  82 #endif
  83
  84 #ifdef IPSEC
  85 #include <netinet/ip_ipsec.h>
  86 #include <netipsec/ipsec.h>
  87 #endif /* IPSEC*/
  88
  89 #include <machine/in_cksum.h>
  90
  91 #include <security/mac/mac_framework.h>
  92
  93 VNET_DEFINE(u_short, ip_id);
  94
  95 #ifdef MBUF_STRESS_TEST
  96 static int mbuf_frag_size = 0;
  97 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
  98         &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
  99 #endif
 100
 101 static void     ip_mloopback
 102         (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
 103
 104
 105 extern int in_mcast_loop;
 106 extern  struct protosw inetsw[];
 107
 108 /*
 109  * IP output.  The packet in mbuf chain m contains a skeletal IP
 110  * header (with len, off, ttl, proto, tos, src, dst).
 111  * The mbuf chain containing the packet will be freed.
 112  * The mbuf opt, if present, will not be freed.
 113  * If route ro is present and has ro_rt initialized, route lookup would be
 114  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 115  * then result of route lookup is stored in ro->ro_rt.
 116  *
 117  * In the IP forwarding case, the packet will arrive with options already
 118  * inserted, so must have a NULL opt pointer.
 119  */
 120 int
 121 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
 122     struct ip_moptions *imo, struct inpcb *inp)
 123 {
 124         struct ip *ip;
 125         struct ifnet *ifp = NULL;       /* keep compiler happy */
 126         struct mbuf *m0;
 127         int hlen = sizeof (struct ip);
 128         int mtu;
 129         int error = 0;
 130         struct sockaddr_in *dst;
 131         const struct sockaddr_in *gw;
 132         struct in_ifaddr *ia;
 133         int isbroadcast;
 134         uint16_t ip_len, ip_off;
 135         struct route iproute;
 136         struct rtentry *rte;    /* cache for ro->ro_rt */
 137         struct in_addr odst;
 138         struct m_tag *fwd_tag = NULL;
 139         uint32_t fibnum;
 140         int have_ia_ref;
 141         int needfiblookup;
 142 #ifdef IPSEC
 143         int no_route_but_check_spd = 0;
 144 #endif
 145         M_ASSERTPKTHDR(m);
 146
 147         if (inp != NULL) {
 148                 INP_LOCK_ASSERT(inp);
 149                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 150                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 151                         m->m_pkthdr.flowid = inp->inp_flowid;
 152                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 153                 }
 154         }
 155
 156         if (ro == NULL) {
 157                 ro = &iproute;
 158                 bzero(ro, sizeof (*ro));
 159         }
 160
 161 #ifdef FLOWTABLE
 162         if (ro->ro_rt == NULL)
 163                 (void )flowtable_lookup(AF_INET, m, ro);
 164 #endif
 165
 166         if (opt) {
 167                 int len = 0;
 168                 m = ip_insertoptions(m, opt, &len);
 169                 if (len != 0)
 170                         hlen = len; /* ip->ip_hl is updated above */
 171         }
 172         ip = mtod(m, struct ip *);
 173         ip_len = ntohs(ip->ip_len);
 174         ip_off = ntohs(ip->ip_off);
 175
 176         /*
 177          * Fill in IP header.  If we are not allowing fragmentation,
 178          * then the ip_id field is meaningless, but we don't set it
 179          * to zero.  Doing so causes various problems when devices along
 180          * the path (routers, load balancers, firewalls, etc.) illegally
 181          * disable DF on our packet.  Note that a 16-bit counter
 182          * will wrap around in less than 10 seconds at 100 Mbit/s on a
 183          * medium with MTU 1500.  See Steven M. Bellovin, "A Technique
 184          * for Counting NATted Hosts", Proc. IMW'02, available at
 185          * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
 186          */
 187         if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
 188                 ip->ip_v = IPVERSION;
 189                 ip->ip_hl = hlen >> 2;
 190                 ip->ip_id = ip_newid();
 191                 IPSTAT_INC(ips_localout);
 192         } else {
 193                 /* Header already set, fetch hlen from there */
 194                 hlen = ip->ip_hl << 2;
 195         }
 196
 197         /*
 198          * dst/gw handling:
 199          *
 200          * dst can be rewritten but always points to &ro->ro_dst.
 201          * gw is readonly but can point either to dst OR rt_gateway,
 202          * therefore we need restore gw if we're redoing lookup.
 203          */
 204         gw = dst = (struct sockaddr_in *)&ro->ro_dst;
 205         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 206 again:
 207         ia = NULL;
 208         have_ia_ref = 0;
 209         /*
 210          * If there is a cached route, check that it is to the same
 211          * destination and is still up.  If not, free it and try again.
 212          * The address family should also be checked in case of sharing
 213          * the cache with IPv6.
 214          */
 215         rte = ro->ro_rt;
 216         if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
 217                     rte->rt_ifp == NULL ||
 218                     !RT_LINK_IS_UP(rte->rt_ifp) ||
 219                           dst->sin_family != AF_INET ||
 220                           dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
 221                 RO_RTFREE(ro);
 222                 ro->ro_lle = NULL;
 223                 rte = NULL;
 224                 gw = dst;
 225         }
 226         if (rte == NULL && fwd_tag == NULL) {
 227                 bzero(dst, sizeof(*dst));
 228                 dst->sin_family = AF_INET;
 229                 dst->sin_len = sizeof(*dst);
 230                 dst->sin_addr = ip->ip_dst;
 231         }
 232         /*
 233          * If routing to interface only, short circuit routing lookup.
 234          * The use of an all-ones broadcast address implies this; an
 235          * interface is specified by the broadcast address of an interface,
 236          * or the destination address of a ptp interface.
 237          */
 238         if (flags & IP_SENDONES) {
 239                 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
 240                                                       M_GETFIB(m)))) == NULL &&
 241                     (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 242                                                     M_GETFIB(m)))) == NULL) {
 243                         IPSTAT_INC(ips_noroute);
 244                         error = ENETUNREACH;
 245                         goto bad;
 246                 }
 247                 have_ia_ref = 1;
 248                 ip->ip_dst.s_addr = INADDR_BROADCAST;
 249                 dst->sin_addr = ip->ip_dst;
 250                 ifp = ia->ia_ifp;
 251                 ip->ip_ttl = 1;
 252                 isbroadcast = 1;
 253         } else if (flags & IP_ROUTETOIF) {
 254                 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 255                                                     M_GETFIB(m)))) == NULL &&
 256                     (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
 257                                                 M_GETFIB(m)))) == NULL) {
 258                         IPSTAT_INC(ips_noroute);
 259                         error = ENETUNREACH;
 260                         goto bad;
 261                 }
 262                 have_ia_ref = 1;
 263                 ifp = ia->ia_ifp;
 264                 ip->ip_ttl = 1;
 265                 isbroadcast = in_broadcast(dst->sin_addr, ifp);
 266         } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
 267             imo != NULL && imo->imo_multicast_ifp != NULL) {
 268                 /*
 269                  * Bypass the normal routing lookup for multicast
 270                  * packets if the interface is specified.
 271                  */
 272                 ifp = imo->imo_multicast_ifp;
 273                 IFP_TO_IA(ifp, ia);
 274                 if (ia)
 275                         have_ia_ref = 1;
 276                 isbroadcast = 0;        /* fool gcc */
 277         } else {
 278                 /*
 279                  * We want to do any cloning requested by the link layer,
 280                  * as this is probably required in all cases for correct
 281                  * operation (as it is for ARP).
 282                  */
 283                 if (rte == NULL) {
 284 #ifdef RADIX_MPATH
 285                         rtalloc_mpath_fib(ro,
 286                             ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
 287                             fibnum);
 288 #else
 289                         in_rtalloc_ign(ro, 0, fibnum);
 290 #endif
 291                         rte = ro->ro_rt;
 292                 }
 293                 if (rte == NULL ||
 294                     rte->rt_ifp == NULL ||
 295                     !RT_LINK_IS_UP(rte->rt_ifp)) {
 296 #ifdef IPSEC
 297                         /*
 298                          * There is no route for this packet, but it is
 299                          * possible that a matching SPD entry exists.
 300                          */
 301                         no_route_but_check_spd = 1;
 302                         mtu = 0; /* Silence GCC warning. */
 303                         goto sendit;
 304 #endif
 305                         IPSTAT_INC(ips_noroute);
 306                         error = EHOSTUNREACH;
 307                         goto bad;
 308                 }
 309                 ia = ifatoia(rte->rt_ifa);
 310                 ifp = rte->rt_ifp;
 311                 counter_u64_add(rte->rt_pksent, 1);
 312                 if (rte->rt_flags & RTF_GATEWAY)
 313                         gw = (struct sockaddr_in *)rte->rt_gateway;
 314                 if (rte->rt_flags & RTF_HOST)
 315                         isbroadcast = (rte->rt_flags & RTF_BROADCAST);
 316                 else
 317                         isbroadcast = in_broadcast(gw->sin_addr, ifp);
 318         }
 319         /*
 320          * Calculate MTU.  If we have a route that is up, use that,
 321          * otherwise use the interface's MTU.
 322          */
 323         if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
 324                 mtu = rte->rt_mtu;
 325         else
 326                 mtu = ifp->if_mtu;
 327         /* Catch a possible divide by zero later. */
 328         KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
 329             __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
 330         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
 331                 m->m_flags |= M_MCAST;
 332                 /*
 333                  * IP destination address is multicast.  Make sure "gw"
 334                  * still points to the address in "ro".  (It may have been
 335                  * changed to point to a gateway address, above.)
 336                  */
 337                 gw = dst;
 338                 /*
 339                  * See if the caller provided any multicast options
 340                  */
 341                 if (imo != NULL) {
 342                         ip->ip_ttl = imo->imo_multicast_ttl;
 343                         if (imo->imo_multicast_vif != -1)
 344                                 ip->ip_src.s_addr =
 345                                     ip_mcast_src ?
 346                                     ip_mcast_src(imo->imo_multicast_vif) :
 347                                     INADDR_ANY;
 348                 } else
 349                         ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
 350                 /*
 351                  * Confirm that the outgoing interface supports multicast.
 352                  */
 353                 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
 354                         if ((ifp->if_flags & IFF_MULTICAST) == 0) {
 355                                 IPSTAT_INC(ips_noroute);
 356                                 error = ENETUNREACH;
 357                                 goto bad;
 358                         }
 359                 }
 360                 /*
 361                  * If source address not specified yet, use address
 362                  * of outgoing interface.
 363                  */
 364                 if (ip->ip_src.s_addr == INADDR_ANY) {
 365                         /* Interface may have no addresses. */
 366                         if (ia != NULL)
 367                                 ip->ip_src = IA_SIN(ia)->sin_addr;
 368                 }
 369
 370                 if ((imo == NULL && in_mcast_loop) ||
 371                     (imo && imo->imo_multicast_loop)) {
 372                         /*
 373                          * Loop back multicast datagram if not expressly
 374                          * forbidden to do so, even if we are not a member
 375                          * of the group; ip_input() will filter it later,
 376                          * thus deferring a hash lookup and mutex acquisition
 377                          * at the expense of a cheap copy using m_copym().
 378                          */
 379                         ip_mloopback(ifp, m, dst, hlen);
 380                 } else {
 381                         /*
 382                          * If we are acting as a multicast router, perform
 383                          * multicast forwarding as if the packet had just
 384                          * arrived on the interface to which we are about
 385                          * to send.  The multicast forwarding function
 386                          * recursively calls this function, using the
 387                          * IP_FORWARDING flag to prevent infinite recursion.
 388                          *
 389                          * Multicasts that are looped back by ip_mloopback(),
 390                          * above, will be forwarded by the ip_input() routine,
 391                          * if necessary.
 392                          */
 393                         if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
 394                                 /*
 395                                  * If rsvp daemon is not running, do not
 396                                  * set ip_moptions. This ensures that the packet
 397                                  * is multicast and not just sent down one link
 398                                  * as prescribed by rsvpd.
 399                                  */
 400                                 if (!V_rsvp_on)
 401                                         imo = NULL;
 402                                 if (ip_mforward &&
 403                                     ip_mforward(ip, ifp, m, imo) != 0) {
 404                                         m_freem(m);
 405                                         goto done;
 406                                 }
 407                         }
 408                 }
 409
 410                 /*
 411                  * Multicasts with a time-to-live of zero may be looped-
 412                  * back, above, but must not be transmitted on a network.
 413                  * Also, multicasts addressed to the loopback interface
 414                  * are not sent -- the above call to ip_mloopback() will
 415                  * loop back a copy. ip_input() will drop the copy if
 416                  * this host does not belong to the destination group on
 417                  * the loopback interface.
 418                  */
 419                 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
 420                         m_freem(m);
 421                         goto done;
 422                 }
 423
 424                 goto sendit;
 425         }
 426
 427         /*
 428          * If the source address is not specified yet, use the address
 429          * of the outoing interface.
 430          */
 431         if (ip->ip_src.s_addr == INADDR_ANY) {
 432                 /* Interface may have no addresses. */
 433                 if (ia != NULL) {
 434                         ip->ip_src = IA_SIN(ia)->sin_addr;
 435                 }
 436         }
 437
 438         /*
 439          * Look for broadcast address and
 440          * verify user is allowed to send
 441          * such a packet.
 442          */
 443         if (isbroadcast) {
 444                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 445                         error = EADDRNOTAVAIL;
 446                         goto bad;
 447                 }
 448                 if ((flags & IP_ALLOWBROADCAST) == 0) {
 449                         error = EACCES;
 450                         goto bad;
 451                 }
 452                 /* don't allow broadcast messages to be fragmented */
 453                 if (ip_len > mtu) {
 454                         error = EMSGSIZE;
 455                         goto bad;
 456                 }
 457                 m->m_flags |= M_BCAST;
 458         } else {
 459                 m->m_flags &= ~M_BCAST;
 460         }
 461
 462 sendit:
 463 #ifdef IPSEC
 464         switch(ip_ipsec_output(&m, inp, &error)) {
 465         case 1:
 466                 goto bad;
 467         case -1:
 468                 goto done;
 469         case 0:
 470         default:
 471                 break;  /* Continue with packet processing. */
 472         }
 473         /*
 474          * Check if there was a route for this packet; return error if not.
 475          */
 476         if (no_route_but_check_spd) {
 477                 IPSTAT_INC(ips_noroute);
 478                 error = EHOSTUNREACH;
 479                 goto bad;
 480         }
 481         /* Update variables that are affected by ipsec4_output(). */
 482         ip = mtod(m, struct ip *);
 483         hlen = ip->ip_hl << 2;
 484 #endif /* IPSEC */
 485
 486         /* Jump over all PFIL processing if hooks are not active. */
 487         if (!PFIL_HOOKED(&V_inet_pfil_hook))
 488                 goto passout;
 489
 490         /* Run through list of hooks for output packets. */
 491         odst.s_addr = ip->ip_dst.s_addr;
 492         error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
 493         if (error != 0 || m == NULL)
 494                 goto done;
 495
 496         ip = mtod(m, struct ip *);
 497         needfiblookup = 0;
 498
 499         /* See if destination IP address was changed by packet filter. */
 500         if (odst.s_addr != ip->ip_dst.s_addr) {
 501                 m->m_flags |= M_SKIP_FIREWALL;
 502                 /* If destination is now ourself drop to ip_input(). */
 503                 if (in_localip(ip->ip_dst)) {
 504                         m->m_flags |= M_FASTFWD_OURS;
 505                         if (m->m_pkthdr.rcvif == NULL)
 506                                 m->m_pkthdr.rcvif = V_loif;
 507                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 508                                 m->m_pkthdr.csum_flags |=
 509                                     CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 510                                 m->m_pkthdr.csum_data = 0xffff;
 511                         }
 512                         m->m_pkthdr.csum_flags |=
 513                             CSUM_IP_CHECKED | CSUM_IP_VALID;
 514 #ifdef SCTP
 515                         if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 516                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 517 #endif
 518                         error = netisr_queue(NETISR_IP, m);
 519                         goto done;
 520                 } else {
 521                         if (have_ia_ref)
 522                                 ifa_free(&ia->ia_ifa);
 523                         needfiblookup = 1; /* Redo the routing table lookup. */
 524                 }
 525         }
 526         /* See if fib was changed by packet filter. */
 527         if (fibnum != M_GETFIB(m)) {
 528                 m->m_flags |= M_SKIP_FIREWALL;
 529                 fibnum = M_GETFIB(m);
 530                 RO_RTFREE(ro);
 531                 needfiblookup = 1;
 532         }
 533         if (needfiblookup)
 534                 goto again;
 535
 536         /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
 537         if (m->m_flags & M_FASTFWD_OURS) {
 538                 if (m->m_pkthdr.rcvif == NULL)
 539                         m->m_pkthdr.rcvif = V_loif;
 540                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 541                         m->m_pkthdr.csum_flags |=
 542                             CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 543                         m->m_pkthdr.csum_data = 0xffff;
 544                 }
 545 #ifdef SCTP
 546                 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 547                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 548 #endif
 549                 m->m_pkthdr.csum_flags |=
 550                             CSUM_IP_CHECKED | CSUM_IP_VALID;
 551
 552                 error = netisr_queue(NETISR_IP, m);
 553                 goto done;
 554         }
 555         /* Or forward to some other address? */
 556         if ((m->m_flags & M_IP_NEXTHOP) &&
 557             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 558                 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
 559                 m->m_flags |= M_SKIP_FIREWALL;
 560                 m->m_flags &= ~M_IP_NEXTHOP;
 561                 m_tag_delete(m, fwd_tag);
 562                 if (have_ia_ref)
 563                         ifa_free(&ia->ia_ifa);
 564                 goto again;
 565         }
 566
 567 passout:
 568         /* 127/8 must not appear on wire - RFC1122. */
 569         if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
 570             (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
 571                 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
 572                         IPSTAT_INC(ips_badaddr);
 573                         error = EADDRNOTAVAIL;
 574                         goto bad;
 575                 }
 576         }
 577
 578         m->m_pkthdr.csum_flags |= CSUM_IP;
 579         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
 580                 in_delayed_cksum(m);
 581                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 582         }
 583 #ifdef SCTP
 584         if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
 585                 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
 586                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 587         }
 588 #endif
 589
 590         /*
 591          * If small enough for interface, or the interface will take
 592          * care of the fragmentation for us, we can just send directly.
 593          */
 594         if (ip_len <= mtu ||
 595             (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
 596                 ip->ip_sum = 0;
 597                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
 598                         ip->ip_sum = in_cksum(m, hlen);
 599                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 600                 }
 601
 602                 /*
 603                  * Record statistics for this interface address.
 604                  * With CSUM_TSO the byte/packet count will be slightly
 605                  * incorrect because we count the IP+TCP headers only
 606                  * once instead of for every generated packet.
 607                  */
 608                 if (!(flags & IP_FORWARDING) && ia) {
 609                         if (m->m_pkthdr.csum_flags & CSUM_TSO)
 610                                 counter_u64_add(ia->ia_ifa.ifa_opackets,
 611                                     m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
 612                         else
 613                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 614
 615                         counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
 616                 }
 617 #ifdef MBUF_STRESS_TEST
 618                 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
 619                         m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
 620 #endif
 621                 /*
 622                  * Reset layer specific mbuf flags
 623                  * to avoid confusing lower layers.
 624                  */
 625                 m_clrprotoflags(m);
 626                 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
 627                 error = (*ifp->if_output)(ifp, m,
 628                     (const struct sockaddr *)gw, ro);
 629                 goto done;
 630         }
 631
 632         /* Balk when DF bit is set or the interface didn't support TSO. */
 633         if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
 634                 error = EMSGSIZE;
 635                 IPSTAT_INC(ips_cantfrag);
 636                 goto bad;
 637         }
 638
 639         /*
 640          * Too large for interface; fragment if possible. If successful,
 641          * on return, m will point to a list of packets to be sent.
 642          */
 643         error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
 644         if (error)
 645                 goto bad;
 646         for (; m; m = m0) {
 647                 m0 = m->m_nextpkt;
 648                 m->m_nextpkt = 0;
 649                 if (error == 0) {
 650                         /* Record statistics for this interface address. */
 651                         if (ia != NULL) {
 652                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 653                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
 654                                     m->m_pkthdr.len);
 655                         }
 656                         /*
 657                          * Reset layer specific mbuf flags
 658                          * to avoid confusing upper layers.
 659                          */
 660                         m_clrprotoflags(m);
 661
 662                         IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
 663                         error = (*ifp->if_output)(ifp, m,
 664                             (const struct sockaddr *)gw, ro);
 665                 } else
 666                         m_freem(m);
 667         }
 668
 669         if (error == 0)
 670                 IPSTAT_INC(ips_fragmented);
 671
 672 done:
 673         if (ro == &iproute)
 674                 RO_RTFREE(ro);
 675         if (have_ia_ref)
 676                 ifa_free(&ia->ia_ifa);
 677         return (error);
 678 bad:
 679         m_freem(m);
 680         goto done;
 681 }
 682
 683 /*
 684  * Create a chain of fragments which fit the given mtu. m_frag points to the
 685  * mbuf to be fragmented; on return it points to the chain with the fragments.
 686  * Return 0 if no error. If error, m_frag may contain a partially built
 687  * chain of fragments that should be freed by the caller.
 688  *
 689  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
 690  */
 691 int
 692 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
 693     u_long if_hwassist_flags)
 694 {
 695         int error = 0;
 696         int hlen = ip->ip_hl << 2;
 697         int len = (mtu - hlen) & ~7;    /* size of payload in each fragment */
 698         int off;
 699         struct mbuf *m0 = *m_frag;      /* the original packet          */
 700         int firstlen;
 701         struct mbuf **mnext;
 702         int nfrags;
 703         uint16_t ip_len, ip_off;
 704
 705         ip_len = ntohs(ip->ip_len);
 706         ip_off = ntohs(ip->ip_off);
 707
 708         if (ip_off & IP_DF) {   /* Fragmentation not allowed */
 709                 IPSTAT_INC(ips_cantfrag);
 710                 return EMSGSIZE;
 711         }
 712
 713         /*
 714          * Must be able to put at least 8 bytes per fragment.
 715          */
 716         if (len < 8)
 717                 return EMSGSIZE;
 718
 719         /*
 720          * If the interface will not calculate checksums on
 721          * fragmented packets, then do it here.
 722          */
 723         if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 724                 in_delayed_cksum(m0);
 725                 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 726         }
 727 #ifdef SCTP
 728         if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
 729                 sctp_delayed_cksum(m0, hlen);
 730                 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 731         }
 732 #endif
 733         if (len > PAGE_SIZE) {
 734                 /*
 735                  * Fragment large datagrams such that each segment
 736                  * contains a multiple of PAGE_SIZE amount of data,
 737                  * plus headers. This enables a receiver to perform
 738                  * page-flipping zero-copy optimizations.
 739                  *
 740                  * XXX When does this help given that sender and receiver
 741                  * could have different page sizes, and also mtu could
 742                  * be less than the receiver's page size ?
 743                  */
 744                 int newlen;
 745                 struct mbuf *m;
 746
 747                 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
 748                         off += m->m_len;
 749
 750                 /*
 751                  * firstlen (off - hlen) must be aligned on an
 752                  * 8-byte boundary
 753                  */
 754                 if (off < hlen)
 755                         goto smart_frag_failure;
 756                 off = ((off - hlen) & ~7) + hlen;
 757                 newlen = (~PAGE_MASK) & mtu;
 758                 if ((newlen + sizeof (struct ip)) > mtu) {
 759                         /* we failed, go back the default */
 760 smart_frag_failure:
 761                         newlen = len;
 762                         off = hlen + len;
 763                 }
 764                 len = newlen;
 765
 766         } else {
 767                 off = hlen + len;
 768         }
 769
 770         firstlen = off - hlen;
 771         mnext = &m0->m_nextpkt;         /* pointer to next packet */
 772
 773         /*
 774          * Loop through length of segment after first fragment,
 775          * make new header and copy data of each part and link onto chain.
 776          * Here, m0 is the original packet, m is the fragment being created.
 777          * The fragments are linked off the m_nextpkt of the original
 778          * packet, which after processing serves as the first fragment.
 779          */
 780         for (nfrags = 1; off < ip_len; off += len, nfrags++) {
 781                 struct ip *mhip;        /* ip header on the fragment */
 782                 struct mbuf *m;
 783                 int mhlen = sizeof (struct ip);
 784
 785                 m = m_gethdr(M_NOWAIT, MT_DATA);
 786                 if (m == NULL) {
 787                         error = ENOBUFS;
 788                         IPSTAT_INC(ips_odropped);
 789                         goto done;
 790                 }
 791                 m->m_flags |= (m0->m_flags & M_MCAST);
 792                 /*
 793                  * In the first mbuf, leave room for the link header, then
 794                  * copy the original IP header including options. The payload
 795                  * goes into an additional mbuf chain returned by m_copym().
 796                  */
 797                 m->m_data += max_linkhdr;
 798                 mhip = mtod(m, struct ip *);
 799                 *mhip = *ip;
 800                 if (hlen > sizeof (struct ip)) {
 801                         mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
 802                         mhip->ip_v = IPVERSION;
 803                         mhip->ip_hl = mhlen >> 2;
 804                 }
 805                 m->m_len = mhlen;
 806                 /* XXX do we need to add ip_off below ? */
 807                 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
 808                 if (off + len >= ip_len)
 809                         len = ip_len - off;
 810                 else
 811                         mhip->ip_off |= IP_MF;
 812                 mhip->ip_len = htons((u_short)(len + mhlen));
 813                 m->m_next = m_copym(m0, off, len, M_NOWAIT);
 814                 if (m->m_next == NULL) {        /* copy failed */
 815                         m_free(m);
 816                         error = ENOBUFS;        /* ??? */
 817                         IPSTAT_INC(ips_odropped);
 818                         goto done;
 819                 }
 820                 m->m_pkthdr.len = mhlen + len;
 821                 m->m_pkthdr.rcvif = NULL;
 822 #ifdef MAC
 823                 mac_netinet_fragment(m0, m);
 824 #endif
 825                 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
 826                 mhip->ip_off = htons(mhip->ip_off);
 827                 mhip->ip_sum = 0;
 828                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 829                         mhip->ip_sum = in_cksum(m, mhlen);
 830                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 831                 }
 832                 *mnext = m;
 833                 mnext = &m->m_nextpkt;
 834         }
 835         IPSTAT_ADD(ips_ofragments, nfrags);
 836
 837         /*
 838          * Update first fragment by trimming what's been copied out
 839          * and updating header.
 840          */
 841         m_adj(m0, hlen + firstlen - ip_len);
 842         m0->m_pkthdr.len = hlen + firstlen;
 843         ip->ip_len = htons((u_short)m0->m_pkthdr.len);
 844         ip->ip_off = htons(ip_off | IP_MF);
 845         ip->ip_sum = 0;
 846         if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 847                 ip->ip_sum = in_cksum(m0, hlen);
 848                 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
 849         }
 850
 851 done:
 852         *m_frag = m0;
 853         return error;
 854 }
 855
 856 void
 857 in_delayed_cksum(struct mbuf *m)
 858 {
 859         struct ip *ip;
 860         uint16_t csum, offset, ip_len;
 861
 862         ip = mtod(m, struct ip *);
 863         offset = ip->ip_hl << 2 ;
 864         ip_len = ntohs(ip->ip_len);
 865         csum = in_cksum_skip(m, ip_len, offset);
 866         if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
 867                 csum = 0xffff;
 868         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 869
 870         /* find the mbuf in the chain where the checksum starts*/
 871         while ((m != NULL) && (offset >= m->m_len)) {
 872                 offset -= m->m_len;
 873                 m = m->m_next;
 874         }
 875         KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
 876         KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
 877         *(u_short *)(m->m_data + offset) = csum;
 878 }
 879
 880 /*
 881  * IP socket option processing.
 882  */
 883 int
 884 ip_ctloutput(struct socket *so, struct sockopt *sopt)
 885 {
 886         struct  inpcb *inp = sotoinpcb(so);
 887         int     error, optval;
 888 #ifdef  RSS
 889         uint32_t rss_bucket;
 890         int retval;
 891 #endif
 892
 893         error = optval = 0;
 894         if (sopt->sopt_level != IPPROTO_IP) {
 895                 error = EINVAL;
 896
 897                 if (sopt->sopt_level == SOL_SOCKET &&
 898                     sopt->sopt_dir == SOPT_SET) {
 899                         switch (sopt->sopt_name) {
 900                         case SO_REUSEADDR:
 901                                 INP_WLOCK(inp);
 902                                 if ((so->so_options & SO_REUSEADDR) != 0)
 903                                         inp->inp_flags2 |= INP_REUSEADDR;
 904                                 else
 905                                         inp->inp_flags2 &= ~INP_REUSEADDR;
 906                                 INP_WUNLOCK(inp);
 907                                 error = 0;
 908                                 break;
 909                         case SO_REUSEPORT:
 910                                 INP_WLOCK(inp);
 911                                 if ((so->so_options & SO_REUSEPORT) != 0)
 912                                         inp->inp_flags2 |= INP_REUSEPORT;
 913                                 else
 914                                         inp->inp_flags2 &= ~INP_REUSEPORT;
 915                                 INP_WUNLOCK(inp);
 916                                 error = 0;
 917                                 break;
 918                         case SO_SETFIB:
 919                                 INP_WLOCK(inp);
 920                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
 921                                 INP_WUNLOCK(inp);
 922                                 error = 0;
 923                                 break;
 924                         default:
 925                                 break;
 926                         }
 927                 }
 928                 return (error);
 929         }
 930
 931         switch (sopt->sopt_dir) {
 932         case SOPT_SET:
 933                 switch (sopt->sopt_name) {
 934                 case IP_OPTIONS:
 935 #ifdef notyet
 936                 case IP_RETOPTS:
 937 #endif
 938                 {
 939                         struct mbuf *m;
 940                         if (sopt->sopt_valsize > MLEN) {
 941                                 error = EMSGSIZE;
 942                                 break;
 943                         }
 944                         m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
 945                         if (m == NULL) {
 946                                 error = ENOBUFS;
 947                                 break;
 948                         }
 949                         m->m_len = sopt->sopt_valsize;
 950                         error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
 951                                             m->m_len);
 952                         if (error) {
 953                                 m_free(m);
 954                                 break;
 955                         }
 956                         INP_WLOCK(inp);
 957                         error = ip_pcbopts(inp, sopt->sopt_name, m);
 958                         INP_WUNLOCK(inp);
 959                         return (error);
 960                 }
 961
 962                 case IP_BINDANY:
 963                         if (sopt->sopt_td != NULL) {
 964                                 error = priv_check(sopt->sopt_td,
 965                                     PRIV_NETINET_BINDANY);
 966                                 if (error)
 967                                         break;
 968                         }
 969                         /* FALLTHROUGH */
 970                 case IP_BINDMULTI:
 971 #ifdef  RSS
 972                 case IP_RSS_LISTEN_BUCKET:
 973 #endif
 974                 case IP_TOS:
 975                 case IP_TTL:
 976                 case IP_MINTTL:
 977                 case IP_RECVOPTS:
 978                 case IP_RECVRETOPTS:
 979                 case IP_RECVDSTADDR:
 980                 case IP_RECVTTL:
 981                 case IP_RECVIF:
 982                 case IP_ONESBCAST:
 983                 case IP_DONTFRAG:
 984                 case IP_RECVTOS:
 985                 case IP_RECVFLOWID:
 986 #ifdef  RSS
 987                 case IP_RECVRSSBUCKETID:
 988 #endif
 989                         error = sooptcopyin(sopt, &optval, sizeof optval,
 990                                             sizeof optval);
 991                         if (error)
 992                                 break;
 993
 994                         switch (sopt->sopt_name) {
 995                         case IP_TOS:
 996                                 inp->inp_ip_tos = optval;
 997                                 break;
 998
 999                         case IP_TTL:
1000                                 inp->inp_ip_ttl = optval;
1001                                 break;
1002
1003                         case IP_MINTTL:
1004                                 if (optval >= 0 && optval <= MAXTTL)
1005                                         inp->inp_ip_minttl = optval;
1006                                 else
1007                                         error = EINVAL;
1008                                 break;
1009
1010 #define OPTSET(bit) do {                                                \
1011         INP_WLOCK(inp);                                                 \
1012         if (optval)                                                     \
1013                 inp->inp_flags |= bit;                                  \
1014         else                                                            \
1015                 inp->inp_flags &= ~bit;                                 \
1016         INP_WUNLOCK(inp);                                               \
1017 } while (0)
1018
1019 #define OPTSET2(bit, val) do {                                          \
1020         INP_WLOCK(inp);                                                 \
1021         if (val)                                                        \
1022                 inp->inp_flags2 |= bit;                                 \
1023         else                                                            \
1024                 inp->inp_flags2 &= ~bit;                                \
1025         INP_WUNLOCK(inp);                                               \
1026 } while (0)
1027
1028                         case IP_RECVOPTS:
1029                                 OPTSET(INP_RECVOPTS);
1030                                 break;
1031
1032                         case IP_RECVRETOPTS:
1033                                 OPTSET(INP_RECVRETOPTS);
1034                                 break;
1035
1036                         case IP_RECVDSTADDR:
1037                                 OPTSET(INP_RECVDSTADDR);
1038                                 break;
1039
1040                         case IP_RECVTTL:
1041                                 OPTSET(INP_RECVTTL);
1042                                 break;
1043
1044                         case IP_RECVIF:
1045                                 OPTSET(INP_RECVIF);
1046                                 break;
1047
1048                         case IP_ONESBCAST:
1049                                 OPTSET(INP_ONESBCAST);
1050                                 break;
1051                         case IP_DONTFRAG:
1052                                 OPTSET(INP_DONTFRAG);
1053                                 break;
1054                         case IP_BINDANY:
1055                                 OPTSET(INP_BINDANY);
1056                                 break;
1057                         case IP_RECVTOS:
1058                                 OPTSET(INP_RECVTOS);
1059                                 break;
1060                         case IP_BINDMULTI:
1061                                 OPTSET2(INP_BINDMULTI, optval);
1062                                 break;
1063                         case IP_RECVFLOWID:
1064                                 OPTSET2(INP_RECVFLOWID, optval);
1065                                 break;
1066 #ifdef  RSS
1067                         case IP_RSS_LISTEN_BUCKET:
1068                                 if ((optval >= 0) &&
1069                                     (optval < rss_getnumbuckets())) {
1070                                         inp->inp_rss_listen_bucket = optval;
1071                                         OPTSET2(INP_RSS_BUCKET_SET, 1);
1072                                 } else {
1073                                         error = EINVAL;
1074                                 }
1075                                 break;
1076                         case IP_RECVRSSBUCKETID:
1077                                 OPTSET2(INP_RECVRSSBUCKETID, optval);
1078                                 break;
1079 #endif
1080                         }
1081                         break;
1082 #undef OPTSET
1083 #undef OPTSET2
1084
1085                 /*
1086                  * Multicast socket options are processed by the in_mcast
1087                  * module.
1088                  */
1089                 case IP_MULTICAST_IF:
1090                 case IP_MULTICAST_VIF:
1091                 case IP_MULTICAST_TTL:
1092                 case IP_MULTICAST_LOOP:
1093                 case IP_ADD_MEMBERSHIP:
1094                 case IP_DROP_MEMBERSHIP:
1095                 case IP_ADD_SOURCE_MEMBERSHIP:
1096                 case IP_DROP_SOURCE_MEMBERSHIP:
1097                 case IP_BLOCK_SOURCE:
1098                 case IP_UNBLOCK_SOURCE:
1099                 case IP_MSFILTER:
1100                 case MCAST_JOIN_GROUP:
1101                 case MCAST_LEAVE_GROUP:
1102                 case MCAST_JOIN_SOURCE_GROUP:
1103                 case MCAST_LEAVE_SOURCE_GROUP:
1104                 case MCAST_BLOCK_SOURCE:
1105                 case MCAST_UNBLOCK_SOURCE:
1106                         error = inp_setmoptions(inp, sopt);
1107                         break;
1108
1109                 case IP_PORTRANGE:
1110                         error = sooptcopyin(sopt, &optval, sizeof optval,
1111                                             sizeof optval);
1112                         if (error)
1113                                 break;
1114
1115                         INP_WLOCK(inp);
1116                         switch (optval) {
1117                         case IP_PORTRANGE_DEFAULT:
1118                                 inp->inp_flags &= ~(INP_LOWPORT);
1119                                 inp->inp_flags &= ~(INP_HIGHPORT);
1120                                 break;
1121
1122                         case IP_PORTRANGE_HIGH:
1123                                 inp->inp_flags &= ~(INP_LOWPORT);
1124                                 inp->inp_flags |= INP_HIGHPORT;
1125                                 break;
1126
1127                         case IP_PORTRANGE_LOW:
1128                                 inp->inp_flags &= ~(INP_HIGHPORT);
1129                                 inp->inp_flags |= INP_LOWPORT;
1130                                 break;
1131
1132                         default:
1133                                 error = EINVAL;
1134                                 break;
1135                         }
1136                         INP_WUNLOCK(inp);
1137                         break;
1138
1139 #ifdef IPSEC
1140                 case IP_IPSEC_POLICY:
1141                 {
1142                         caddr_t req;
1143                         struct mbuf *m;
1144
1145                         if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1146                                 break;
1147                         if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1148                                 break;
1149                         req = mtod(m, caddr_t);
1150                         error = ipsec_set_policy(inp, sopt->sopt_name, req,
1151                             m->m_len, (sopt->sopt_td != NULL) ?
1152                             sopt->sopt_td->td_ucred : NULL);
1153                         m_freem(m);
1154                         break;
1155                 }
1156 #endif /* IPSEC */
1157
1158                 default:
1159                         error = ENOPROTOOPT;
1160                         break;
1161                 }
1162                 break;
1163
1164         case SOPT_GET:
1165                 switch (sopt->sopt_name) {
1166                 case IP_OPTIONS:
1167                 case IP_RETOPTS:
1168                         if (inp->inp_options)
1169                                 error = sooptcopyout(sopt,
1170                                                      mtod(inp->inp_options,
1171                                                           char *),
1172                                                      inp->inp_options->m_len);
1173                         else
1174                                 sopt->sopt_valsize = 0;
1175                         break;
1176
1177                 case IP_TOS:
1178                 case IP_TTL:
1179                 case IP_MINTTL:
1180                 case IP_RECVOPTS:
1181                 case IP_RECVRETOPTS:
1182                 case IP_RECVDSTADDR:
1183                 case IP_RECVTTL:
1184                 case IP_RECVIF:
1185                 case IP_PORTRANGE:
1186                 case IP_ONESBCAST:
1187                 case IP_DONTFRAG:
1188                 case IP_BINDANY:
1189                 case IP_RECVTOS:
1190                 case IP_BINDMULTI:
1191                 case IP_FLOWID:
1192                 case IP_FLOWTYPE:
1193                 case IP_RECVFLOWID:
1194 #ifdef  RSS
1195                 case IP_RSSBUCKETID:
1196                 case IP_RECVRSSBUCKETID:
1197 #endif
1198                         switch (sopt->sopt_name) {
1199
1200                         case IP_TOS:
1201                                 optval = inp->inp_ip_tos;
1202                                 break;
1203
1204                         case IP_TTL:
1205                                 optval = inp->inp_ip_ttl;
1206                                 break;
1207
1208                         case IP_MINTTL:
1209                                 optval = inp->inp_ip_minttl;
1210                                 break;
1211
1212 #define OPTBIT(bit)     (inp->inp_flags & bit ? 1 : 0)
1213 #define OPTBIT2(bit)    (inp->inp_flags2 & bit ? 1 : 0)
1214
1215                         case IP_RECVOPTS:
1216                                 optval = OPTBIT(INP_RECVOPTS);
1217                                 break;
1218
1219                         case IP_RECVRETOPTS:
1220                                 optval = OPTBIT(INP_RECVRETOPTS);
1221                                 break;
1222
1223                         case IP_RECVDSTADDR:
1224                                 optval = OPTBIT(INP_RECVDSTADDR);
1225                                 break;
1226
1227                         case IP_RECVTTL:
1228                                 optval = OPTBIT(INP_RECVTTL);
1229                                 break;
1230
1231                         case IP_RECVIF:
1232                                 optval = OPTBIT(INP_RECVIF);
1233                                 break;
1234
1235                         case IP_PORTRANGE:
1236                                 if (inp->inp_flags & INP_HIGHPORT)
1237                                         optval = IP_PORTRANGE_HIGH;
1238                                 else if (inp->inp_flags & INP_LOWPORT)
1239                                         optval = IP_PORTRANGE_LOW;
1240                                 else
1241                                         optval = 0;
1242                                 break;
1243
1244                         case IP_ONESBCAST:
1245                                 optval = OPTBIT(INP_ONESBCAST);
1246                                 break;
1247                         case IP_DONTFRAG:
1248                                 optval = OPTBIT(INP_DONTFRAG);
1249                                 break;
1250                         case IP_BINDANY:
1251                                 optval = OPTBIT(INP_BINDANY);
1252                                 break;
1253                         case IP_RECVTOS:
1254                                 optval = OPTBIT(INP_RECVTOS);
1255                                 break;
1256                         case IP_FLOWID:
1257                                 optval = inp->inp_flowid;
1258                                 break;
1259                         case IP_FLOWTYPE:
1260                                 optval = inp->inp_flowtype;
1261                                 break;
1262                         case IP_RECVFLOWID:
1263                                 optval = OPTBIT2(INP_RECVFLOWID);
1264                                 break;
1265 #ifdef  RSS
1266                         case IP_RSSBUCKETID:
1267                                 retval = rss_hash2bucket(inp->inp_flowid,
1268                                     inp->inp_flowtype,
1269                                     &rss_bucket);
1270                                 if (retval == 0)
1271                                         optval = rss_bucket;
1272                                 else
1273                                         error = EINVAL;
1274                                 break;
1275                         case IP_RECVRSSBUCKETID:
1276                                 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1277                                 break;
1278 #endif
1279                         case IP_BINDMULTI:
1280                                 optval = OPTBIT2(INP_BINDMULTI);
1281                                 break;
1282                         }
1283                         error = sooptcopyout(sopt, &optval, sizeof optval);
1284                         break;
1285
1286                 /*
1287                  * Multicast socket options are processed by the in_mcast
1288                  * module.
1289                  */
1290                 case IP_MULTICAST_IF:
1291                 case IP_MULTICAST_VIF:
1292                 case IP_MULTICAST_TTL:
1293                 case IP_MULTICAST_LOOP:
1294                 case IP_MSFILTER:
1295                         error = inp_getmoptions(inp, sopt);
1296                         break;
1297
1298 #ifdef IPSEC
1299                 case IP_IPSEC_POLICY:
1300                 {
1301                         struct mbuf *m = NULL;
1302                         caddr_t req = NULL;
1303                         size_t len = 0;
1304
1305                         if (m != 0) {
1306                                 req = mtod(m, caddr_t);
1307                                 len = m->m_len;
1308                         }
1309                         error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1310                         if (error == 0)
1311                                 error = soopt_mcopyout(sopt, m); /* XXX */
1312                         if (error == 0)
1313                                 m_freem(m);
1314                         break;
1315                 }
1316 #endif /* IPSEC */
1317
1318                 default:
1319                         error = ENOPROTOOPT;
1320                         break;
1321                 }
1322                 break;
1323         }
1324         return (error);
1325 }
1326
1327 /*
1328  * Routine called from ip_output() to loop back a copy of an IP multicast
1329  * packet to the input queue of a specified interface.  Note that this
1330  * calls the output routine of the loopback "driver", but with an interface
1331  * pointer that might NOT be a loopback interface -- evil, but easier than
1332  * replicating that code here.
1333  */
1334 static void
1335 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1336     int hlen)
1337 {
1338         register struct ip *ip;
1339         struct mbuf *copym;
1340
1341         /*
1342          * Make a deep copy of the packet because we're going to
1343          * modify the pack in order to generate checksums.
1344          */
1345         copym = m_dup(m, M_NOWAIT);
1346         if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1347                 copym = m_pullup(copym, hlen);
1348         if (copym != NULL) {
1349                 /* If needed, compute the checksum and mark it as valid. */
1350                 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1351                         in_delayed_cksum(copym);
1352                         copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1353                         copym->m_pkthdr.csum_flags |=
1354                             CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1355                         copym->m_pkthdr.csum_data = 0xffff;
1356                 }
1357                 /*
1358                  * We don't bother to fragment if the IP length is greater
1359                  * than the interface's MTU.  Can this possibly matter?
1360                  */
1361                 ip = mtod(copym, struct ip *);
1362                 ip->ip_sum = 0;
1363                 ip->ip_sum = in_cksum(copym, hlen);
1364 #if 1 /* XXX */
1365                 if (dst->sin_family != AF_INET) {
1366                         printf("ip_mloopback: bad address family %d\n",
1367                                                 dst->sin_family);
1368                         dst->sin_family = AF_INET;
1369                 }
1370 #endif
1371                 if_simloop(ifp, copym, dst->sin_family, 0);
1372         }
1373 }