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                     inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
 152                         m->m_pkthdr.flowid = inp->inp_flowid;
 153                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 154                         m->m_flags |= M_FLOWID;
 155                 }
 156         }
 157
 158         if (ro == NULL) {
 159                 ro = &iproute;
 160                 bzero(ro, sizeof (*ro));
 161         }
 162
 163 #ifdef FLOWTABLE
 164         if (ro->ro_rt == NULL)
 165                 (void )flowtable_lookup(AF_INET, m, ro);
 166 #endif
 167
 168         if (opt) {
 169                 int len = 0;
 170                 m = ip_insertoptions(m, opt, &len);
 171                 if (len != 0)
 172                         hlen = len; /* ip->ip_hl is updated above */
 173         }
 174         ip = mtod(m, struct ip *);
 175         ip_len = ntohs(ip->ip_len);
 176         ip_off = ntohs(ip->ip_off);
 177
 178         /*
 179          * Fill in IP header.  If we are not allowing fragmentation,
 180          * then the ip_id field is meaningless, but we don't set it
 181          * to zero.  Doing so causes various problems when devices along
 182          * the path (routers, load balancers, firewalls, etc.) illegally
 183          * disable DF on our packet.  Note that a 16-bit counter
 184          * will wrap around in less than 10 seconds at 100 Mbit/s on a
 185          * medium with MTU 1500.  See Steven M. Bellovin, "A Technique
 186          * for Counting NATted Hosts", Proc. IMW'02, available at
 187          * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
 188          */
 189         if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
 190                 ip->ip_v = IPVERSION;
 191                 ip->ip_hl = hlen >> 2;
 192                 ip->ip_id = ip_newid();
 193                 IPSTAT_INC(ips_localout);
 194         } else {
 195                 /* Header already set, fetch hlen from there */
 196                 hlen = ip->ip_hl << 2;
 197         }
 198
 199         /*
 200          * dst/gw handling:
 201          *
 202          * dst can be rewritten but always points to &ro->ro_dst.
 203          * gw is readonly but can point either to dst OR rt_gateway,
 204          * therefore we need restore gw if we're redoing lookup.
 205          */
 206         gw = dst = (struct sockaddr_in *)&ro->ro_dst;
 207         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 208 again:
 209         ia = NULL;
 210         have_ia_ref = 0;
 211         /*
 212          * If there is a cached route, check that it is to the same
 213          * destination and is still up.  If not, free it and try again.
 214          * The address family should also be checked in case of sharing
 215          * the cache with IPv6.
 216          */
 217         rte = ro->ro_rt;
 218         if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
 219                     rte->rt_ifp == NULL ||
 220                     !RT_LINK_IS_UP(rte->rt_ifp) ||
 221                           dst->sin_family != AF_INET ||
 222                           dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
 223                 RO_RTFREE(ro);
 224                 ro->ro_lle = NULL;
 225                 rte = NULL;
 226                 gw = dst;
 227         }
 228         if (rte == NULL && fwd_tag == NULL) {
 229                 bzero(dst, sizeof(*dst));
 230                 dst->sin_family = AF_INET;
 231                 dst->sin_len = sizeof(*dst);
 232                 dst->sin_addr = ip->ip_dst;
 233         }
 234         /*
 235          * If routing to interface only, short circuit routing lookup.
 236          * The use of an all-ones broadcast address implies this; an
 237          * interface is specified by the broadcast address of an interface,
 238          * or the destination address of a ptp interface.
 239          */
 240         if (flags & IP_SENDONES) {
 241                 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
 242                                                       M_GETFIB(m)))) == NULL &&
 243                     (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 244                                                     M_GETFIB(m)))) == NULL) {
 245                         IPSTAT_INC(ips_noroute);
 246                         error = ENETUNREACH;
 247                         goto bad;
 248                 }
 249                 have_ia_ref = 1;
 250                 ip->ip_dst.s_addr = INADDR_BROADCAST;
 251                 dst->sin_addr = ip->ip_dst;
 252                 ifp = ia->ia_ifp;
 253                 ip->ip_ttl = 1;
 254                 isbroadcast = 1;
 255         } else if (flags & IP_ROUTETOIF) {
 256                 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 257                                                     M_GETFIB(m)))) == NULL &&
 258                     (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
 259                                                 M_GETFIB(m)))) == NULL) {
 260                         IPSTAT_INC(ips_noroute);
 261                         error = ENETUNREACH;
 262                         goto bad;
 263                 }
 264                 have_ia_ref = 1;
 265                 ifp = ia->ia_ifp;
 266                 ip->ip_ttl = 1;
 267                 isbroadcast = in_broadcast(dst->sin_addr, ifp);
 268         } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
 269             imo != NULL && imo->imo_multicast_ifp != NULL) {
 270                 /*
 271                  * Bypass the normal routing lookup for multicast
 272                  * packets if the interface is specified.
 273                  */
 274                 ifp = imo->imo_multicast_ifp;
 275                 IFP_TO_IA(ifp, ia);
 276                 if (ia)
 277                         have_ia_ref = 1;
 278                 isbroadcast = 0;        /* fool gcc */
 279         } else {
 280                 /*
 281                  * We want to do any cloning requested by the link layer,
 282                  * as this is probably required in all cases for correct
 283                  * operation (as it is for ARP).
 284                  */
 285                 if (rte == NULL) {
 286 #ifdef RADIX_MPATH
 287                         rtalloc_mpath_fib(ro,
 288                             ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
 289                             fibnum);
 290 #else
 291                         in_rtalloc_ign(ro, 0, fibnum);
 292 #endif
 293                         rte = ro->ro_rt;
 294                 }
 295                 if (rte == NULL ||
 296                     rte->rt_ifp == NULL ||
 297                     !RT_LINK_IS_UP(rte->rt_ifp)) {
 298 #ifdef IPSEC
 299                         /*
 300                          * There is no route for this packet, but it is
 301                          * possible that a matching SPD entry exists.
 302                          */
 303                         no_route_but_check_spd = 1;
 304                         mtu = 0; /* Silence GCC warning. */
 305                         goto sendit;
 306 #endif
 307                         IPSTAT_INC(ips_noroute);
 308                         error = EHOSTUNREACH;
 309                         goto bad;
 310                 }
 311                 ia = ifatoia(rte->rt_ifa);
 312                 ifp = rte->rt_ifp;
 313                 counter_u64_add(rte->rt_pksent, 1);
 314                 if (rte->rt_flags & RTF_GATEWAY)
 315                         gw = (struct sockaddr_in *)rte->rt_gateway;
 316                 if (rte->rt_flags & RTF_HOST)
 317                         isbroadcast = (rte->rt_flags & RTF_BROADCAST);
 318                 else
 319                         isbroadcast = in_broadcast(gw->sin_addr, ifp);
 320         }
 321         /*
 322          * Calculate MTU.  If we have a route that is up, use that,
 323          * otherwise use the interface's MTU.
 324          */
 325         if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
 326                 /*
 327                  * This case can happen if the user changed the MTU
 328                  * of an interface after enabling IP on it.  Because
 329                  * most netifs don't keep track of routes pointing to
 330                  * them, there is no way for one to update all its
 331                  * routes when the MTU is changed.
 332                  */
 333                 if (rte->rt_mtu > ifp->if_mtu)
 334                         rte->rt_mtu = ifp->if_mtu;
 335                 mtu = rte->rt_mtu;
 336         } else {
 337                 mtu = ifp->if_mtu;
 338         }
 339         /* Catch a possible divide by zero later. */
 340         KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
 341             __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
 342         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
 343                 m->m_flags |= M_MCAST;
 344                 /*
 345                  * IP destination address is multicast.  Make sure "gw"
 346                  * still points to the address in "ro".  (It may have been
 347                  * changed to point to a gateway address, above.)
 348                  */
 349                 gw = dst;
 350                 /*
 351                  * See if the caller provided any multicast options
 352                  */
 353                 if (imo != NULL) {
 354                         ip->ip_ttl = imo->imo_multicast_ttl;
 355                         if (imo->imo_multicast_vif != -1)
 356                                 ip->ip_src.s_addr =
 357                                     ip_mcast_src ?
 358                                     ip_mcast_src(imo->imo_multicast_vif) :
 359                                     INADDR_ANY;
 360                 } else
 361                         ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
 362                 /*
 363                  * Confirm that the outgoing interface supports multicast.
 364                  */
 365                 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
 366                         if ((ifp->if_flags & IFF_MULTICAST) == 0) {
 367                                 IPSTAT_INC(ips_noroute);
 368                                 error = ENETUNREACH;
 369                                 goto bad;
 370                         }
 371                 }
 372                 /*
 373                  * If source address not specified yet, use address
 374                  * of outgoing interface.
 375                  */
 376                 if (ip->ip_src.s_addr == INADDR_ANY) {
 377                         /* Interface may have no addresses. */
 378                         if (ia != NULL)
 379                                 ip->ip_src = IA_SIN(ia)->sin_addr;
 380                 }
 381
 382                 if ((imo == NULL && in_mcast_loop) ||
 383                     (imo && imo->imo_multicast_loop)) {
 384                         /*
 385                          * Loop back multicast datagram if not expressly
 386                          * forbidden to do so, even if we are not a member
 387                          * of the group; ip_input() will filter it later,
 388                          * thus deferring a hash lookup and mutex acquisition
 389                          * at the expense of a cheap copy using m_copym().
 390                          */
 391                         ip_mloopback(ifp, m, dst, hlen);
 392                 } else {
 393                         /*
 394                          * If we are acting as a multicast router, perform
 395                          * multicast forwarding as if the packet had just
 396                          * arrived on the interface to which we are about
 397                          * to send.  The multicast forwarding function
 398                          * recursively calls this function, using the
 399                          * IP_FORWARDING flag to prevent infinite recursion.
 400                          *
 401                          * Multicasts that are looped back by ip_mloopback(),
 402                          * above, will be forwarded by the ip_input() routine,
 403                          * if necessary.
 404                          */
 405                         if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
 406                                 /*
 407                                  * If rsvp daemon is not running, do not
 408                                  * set ip_moptions. This ensures that the packet
 409                                  * is multicast and not just sent down one link
 410                                  * as prescribed by rsvpd.
 411                                  */
 412                                 if (!V_rsvp_on)
 413                                         imo = NULL;
 414                                 if (ip_mforward &&
 415                                     ip_mforward(ip, ifp, m, imo) != 0) {
 416                                         m_freem(m);
 417                                         goto done;
 418                                 }
 419                         }
 420                 }
 421
 422                 /*
 423                  * Multicasts with a time-to-live of zero may be looped-
 424                  * back, above, but must not be transmitted on a network.
 425                  * Also, multicasts addressed to the loopback interface
 426                  * are not sent -- the above call to ip_mloopback() will
 427                  * loop back a copy. ip_input() will drop the copy if
 428                  * this host does not belong to the destination group on
 429                  * the loopback interface.
 430                  */
 431                 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
 432                         m_freem(m);
 433                         goto done;
 434                 }
 435
 436                 goto sendit;
 437         }
 438
 439         /*
 440          * If the source address is not specified yet, use the address
 441          * of the outoing interface.
 442          */
 443         if (ip->ip_src.s_addr == INADDR_ANY) {
 444                 /* Interface may have no addresses. */
 445                 if (ia != NULL) {
 446                         ip->ip_src = IA_SIN(ia)->sin_addr;
 447                 }
 448         }
 449
 450         /*
 451          * Look for broadcast address and
 452          * verify user is allowed to send
 453          * such a packet.
 454          */
 455         if (isbroadcast) {
 456                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 457                         error = EADDRNOTAVAIL;
 458                         goto bad;
 459                 }
 460                 if ((flags & IP_ALLOWBROADCAST) == 0) {
 461                         error = EACCES;
 462                         goto bad;
 463                 }
 464                 /* don't allow broadcast messages to be fragmented */
 465                 if (ip_len > mtu) {
 466                         error = EMSGSIZE;
 467                         goto bad;
 468                 }
 469                 m->m_flags |= M_BCAST;
 470         } else {
 471                 m->m_flags &= ~M_BCAST;
 472         }
 473
 474 sendit:
 475 #ifdef IPSEC
 476         switch(ip_ipsec_output(&m, inp, &flags, &error)) {
 477         case 1:
 478                 goto bad;
 479         case -1:
 480                 goto done;
 481         case 0:
 482         default:
 483                 break;  /* Continue with packet processing. */
 484         }
 485         /*
 486          * Check if there was a route for this packet; return error if not.
 487          */
 488         if (no_route_but_check_spd) {
 489                 IPSTAT_INC(ips_noroute);
 490                 error = EHOSTUNREACH;
 491                 goto bad;
 492         }
 493         /* Update variables that are affected by ipsec4_output(). */
 494         ip = mtod(m, struct ip *);
 495         hlen = ip->ip_hl << 2;
 496 #endif /* IPSEC */
 497
 498         /* Jump over all PFIL processing if hooks are not active. */
 499         if (!PFIL_HOOKED(&V_inet_pfil_hook))
 500                 goto passout;
 501
 502         /* Run through list of hooks for output packets. */
 503         odst.s_addr = ip->ip_dst.s_addr;
 504         error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
 505         if (error != 0 || m == NULL)
 506                 goto done;
 507
 508         ip = mtod(m, struct ip *);
 509         needfiblookup = 0;
 510
 511         /* See if destination IP address was changed by packet filter. */
 512         if (odst.s_addr != ip->ip_dst.s_addr) {
 513                 m->m_flags |= M_SKIP_FIREWALL;
 514                 /* If destination is now ourself drop to ip_input(). */
 515                 if (in_localip(ip->ip_dst)) {
 516                         m->m_flags |= M_FASTFWD_OURS;
 517                         if (m->m_pkthdr.rcvif == NULL)
 518                                 m->m_pkthdr.rcvif = V_loif;
 519                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 520                                 m->m_pkthdr.csum_flags |=
 521                                     CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 522                                 m->m_pkthdr.csum_data = 0xffff;
 523                         }
 524                         m->m_pkthdr.csum_flags |=
 525                             CSUM_IP_CHECKED | CSUM_IP_VALID;
 526 #ifdef SCTP
 527                         if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 528                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 529 #endif
 530                         error = netisr_queue(NETISR_IP, m);
 531                         goto done;
 532                 } else {
 533                         if (have_ia_ref)
 534                                 ifa_free(&ia->ia_ifa);
 535                         needfiblookup = 1; /* Redo the routing table lookup. */
 536                 }
 537         }
 538         /* See if fib was changed by packet filter. */
 539         if (fibnum != M_GETFIB(m)) {
 540                 m->m_flags |= M_SKIP_FIREWALL;
 541                 fibnum = M_GETFIB(m);
 542                 RO_RTFREE(ro);
 543                 needfiblookup = 1;
 544         }
 545         if (needfiblookup)
 546                 goto again;
 547
 548         /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
 549         if (m->m_flags & M_FASTFWD_OURS) {
 550                 if (m->m_pkthdr.rcvif == NULL)
 551                         m->m_pkthdr.rcvif = V_loif;
 552                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 553                         m->m_pkthdr.csum_flags |=
 554                             CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 555                         m->m_pkthdr.csum_data = 0xffff;
 556                 }
 557 #ifdef SCTP
 558                 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 559                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 560 #endif
 561                 m->m_pkthdr.csum_flags |=
 562                             CSUM_IP_CHECKED | CSUM_IP_VALID;
 563
 564                 error = netisr_queue(NETISR_IP, m);
 565                 goto done;
 566         }
 567         /* Or forward to some other address? */
 568         if ((m->m_flags & M_IP_NEXTHOP) &&
 569             (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
 570                 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
 571                 m->m_flags |= M_SKIP_FIREWALL;
 572                 m->m_flags &= ~M_IP_NEXTHOP;
 573                 m_tag_delete(m, fwd_tag);
 574                 if (have_ia_ref)
 575                         ifa_free(&ia->ia_ifa);
 576                 goto again;
 577         }
 578
 579 passout:
 580         /* 127/8 must not appear on wire - RFC1122. */
 581         if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
 582             (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
 583                 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
 584                         IPSTAT_INC(ips_badaddr);
 585                         error = EADDRNOTAVAIL;
 586                         goto bad;
 587                 }
 588         }
 589
 590         m->m_pkthdr.csum_flags |= CSUM_IP;
 591         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
 592                 in_delayed_cksum(m);
 593                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 594         }
 595 #ifdef SCTP
 596         if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
 597                 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
 598                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 599         }
 600 #endif
 601
 602         /*
 603          * If small enough for interface, or the interface will take
 604          * care of the fragmentation for us, we can just send directly.
 605          */
 606         if (ip_len <= mtu ||
 607             (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
 608                 ip->ip_sum = 0;
 609                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
 610                         ip->ip_sum = in_cksum(m, hlen);
 611                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 612                 }
 613
 614                 /*
 615                  * Record statistics for this interface address.
 616                  * With CSUM_TSO the byte/packet count will be slightly
 617                  * incorrect because we count the IP+TCP headers only
 618                  * once instead of for every generated packet.
 619                  */
 620                 if (!(flags & IP_FORWARDING) && ia) {
 621                         if (m->m_pkthdr.csum_flags & CSUM_TSO)
 622                                 counter_u64_add(ia->ia_ifa.ifa_opackets,
 623                                     m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
 624                         else
 625                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 626
 627                         counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
 628                 }
 629 #ifdef MBUF_STRESS_TEST
 630                 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
 631                         m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
 632 #endif
 633                 /*
 634                  * Reset layer specific mbuf flags
 635                  * to avoid confusing lower layers.
 636                  */
 637                 m_clrprotoflags(m);
 638                 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
 639                 error = (*ifp->if_output)(ifp, m,
 640                     (const struct sockaddr *)gw, ro);
 641                 goto done;
 642         }
 643
 644         /* Balk when DF bit is set or the interface didn't support TSO. */
 645         if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
 646                 error = EMSGSIZE;
 647                 IPSTAT_INC(ips_cantfrag);
 648                 goto bad;
 649         }
 650
 651         /*
 652          * Too large for interface; fragment if possible. If successful,
 653          * on return, m will point to a list of packets to be sent.
 654          */
 655         error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
 656         if (error)
 657                 goto bad;
 658         for (; m; m = m0) {
 659                 m0 = m->m_nextpkt;
 660                 m->m_nextpkt = 0;
 661                 if (error == 0) {
 662                         /* Record statistics for this interface address. */
 663                         if (ia != NULL) {
 664                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 665                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
 666                                     m->m_pkthdr.len);
 667                         }
 668                         /*
 669                          * Reset layer specific mbuf flags
 670                          * to avoid confusing upper layers.
 671                          */
 672                         m_clrprotoflags(m);
 673
 674                         IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
 675                         error = (*ifp->if_output)(ifp, m,
 676                             (const struct sockaddr *)gw, ro);
 677                 } else
 678                         m_freem(m);
 679         }
 680
 681         if (error == 0)
 682                 IPSTAT_INC(ips_fragmented);
 683
 684 done:
 685         if (ro == &iproute)
 686                 RO_RTFREE(ro);
 687         if (have_ia_ref)
 688                 ifa_free(&ia->ia_ifa);
 689         return (error);
 690 bad:
 691         m_freem(m);
 692         goto done;
 693 }
 694
 695 /*
 696  * Create a chain of fragments which fit the given mtu. m_frag points to the
 697  * mbuf to be fragmented; on return it points to the chain with the fragments.
 698  * Return 0 if no error. If error, m_frag may contain a partially built
 699  * chain of fragments that should be freed by the caller.
 700  *
 701  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
 702  */
 703 int
 704 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
 705     u_long if_hwassist_flags)
 706 {
 707         int error = 0;
 708         int hlen = ip->ip_hl << 2;
 709         int len = (mtu - hlen) & ~7;    /* size of payload in each fragment */
 710         int off;
 711         struct mbuf *m0 = *m_frag;      /* the original packet          */
 712         int firstlen;
 713         struct mbuf **mnext;
 714         int nfrags;
 715         uint16_t ip_len, ip_off;
 716
 717         ip_len = ntohs(ip->ip_len);
 718         ip_off = ntohs(ip->ip_off);
 719
 720         if (ip_off & IP_DF) {   /* Fragmentation not allowed */
 721                 IPSTAT_INC(ips_cantfrag);
 722                 return EMSGSIZE;
 723         }
 724
 725         /*
 726          * Must be able to put at least 8 bytes per fragment.
 727          */
 728         if (len < 8)
 729                 return EMSGSIZE;
 730
 731         /*
 732          * If the interface will not calculate checksums on
 733          * fragmented packets, then do it here.
 734          */
 735         if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 736                 in_delayed_cksum(m0);
 737                 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 738         }
 739 #ifdef SCTP
 740         if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
 741                 sctp_delayed_cksum(m0, hlen);
 742                 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 743         }
 744 #endif
 745         if (len > PAGE_SIZE) {
 746                 /*
 747                  * Fragment large datagrams such that each segment
 748                  * contains a multiple of PAGE_SIZE amount of data,
 749                  * plus headers. This enables a receiver to perform
 750                  * page-flipping zero-copy optimizations.
 751                  *
 752                  * XXX When does this help given that sender and receiver
 753                  * could have different page sizes, and also mtu could
 754                  * be less than the receiver's page size ?
 755                  */
 756                 int newlen;
 757                 struct mbuf *m;
 758
 759                 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
 760                         off += m->m_len;
 761
 762                 /*
 763                  * firstlen (off - hlen) must be aligned on an
 764                  * 8-byte boundary
 765                  */
 766                 if (off < hlen)
 767                         goto smart_frag_failure;
 768                 off = ((off - hlen) & ~7) + hlen;
 769                 newlen = (~PAGE_MASK) & mtu;
 770                 if ((newlen + sizeof (struct ip)) > mtu) {
 771                         /* we failed, go back the default */
 772 smart_frag_failure:
 773                         newlen = len;
 774                         off = hlen + len;
 775                 }
 776                 len = newlen;
 777
 778         } else {
 779                 off = hlen + len;
 780         }
 781
 782         firstlen = off - hlen;
 783         mnext = &m0->m_nextpkt;         /* pointer to next packet */
 784
 785         /*
 786          * Loop through length of segment after first fragment,
 787          * make new header and copy data of each part and link onto chain.
 788          * Here, m0 is the original packet, m is the fragment being created.
 789          * The fragments are linked off the m_nextpkt of the original
 790          * packet, which after processing serves as the first fragment.
 791          */
 792         for (nfrags = 1; off < ip_len; off += len, nfrags++) {
 793                 struct ip *mhip;        /* ip header on the fragment */
 794                 struct mbuf *m;
 795                 int mhlen = sizeof (struct ip);
 796
 797                 m = m_gethdr(M_NOWAIT, MT_DATA);
 798                 if (m == NULL) {
 799                         error = ENOBUFS;
 800                         IPSTAT_INC(ips_odropped);
 801                         goto done;
 802                 }
 803                 m->m_flags |= (m0->m_flags & M_MCAST);
 804                 /*
 805                  * In the first mbuf, leave room for the link header, then
 806                  * copy the original IP header including options. The payload
 807                  * goes into an additional mbuf chain returned by m_copym().
 808                  */
 809                 m->m_data += max_linkhdr;
 810                 mhip = mtod(m, struct ip *);
 811                 *mhip = *ip;
 812                 if (hlen > sizeof (struct ip)) {
 813                         mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
 814                         mhip->ip_v = IPVERSION;
 815                         mhip->ip_hl = mhlen >> 2;
 816                 }
 817                 m->m_len = mhlen;
 818                 /* XXX do we need to add ip_off below ? */
 819                 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
 820                 if (off + len >= ip_len)
 821                         len = ip_len - off;
 822                 else
 823                         mhip->ip_off |= IP_MF;
 824                 mhip->ip_len = htons((u_short)(len + mhlen));
 825                 m->m_next = m_copym(m0, off, len, M_NOWAIT);
 826                 if (m->m_next == NULL) {        /* copy failed */
 827                         m_free(m);
 828                         error = ENOBUFS;        /* ??? */
 829                         IPSTAT_INC(ips_odropped);
 830                         goto done;
 831                 }
 832                 m->m_pkthdr.len = mhlen + len;
 833                 m->m_pkthdr.rcvif = NULL;
 834 #ifdef MAC
 835                 mac_netinet_fragment(m0, m);
 836 #endif
 837                 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
 838                 mhip->ip_off = htons(mhip->ip_off);
 839                 mhip->ip_sum = 0;
 840                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 841                         mhip->ip_sum = in_cksum(m, mhlen);
 842                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 843                 }
 844                 *mnext = m;
 845                 mnext = &m->m_nextpkt;
 846         }
 847         IPSTAT_ADD(ips_ofragments, nfrags);
 848
 849         /*
 850          * Update first fragment by trimming what's been copied out
 851          * and updating header.
 852          */
 853         m_adj(m0, hlen + firstlen - ip_len);
 854         m0->m_pkthdr.len = hlen + firstlen;
 855         ip->ip_len = htons((u_short)m0->m_pkthdr.len);
 856         ip->ip_off = htons(ip_off | IP_MF);
 857         ip->ip_sum = 0;
 858         if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 859                 ip->ip_sum = in_cksum(m0, hlen);
 860                 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
 861         }
 862
 863 done:
 864         *m_frag = m0;
 865         return error;
 866 }
 867
 868 void
 869 in_delayed_cksum(struct mbuf *m)
 870 {
 871         struct ip *ip;
 872         uint16_t csum, offset, ip_len;
 873
 874         ip = mtod(m, struct ip *);
 875         offset = ip->ip_hl << 2 ;
 876         ip_len = ntohs(ip->ip_len);
 877         csum = in_cksum_skip(m, ip_len, offset);
 878         if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
 879                 csum = 0xffff;
 880         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 881
 882         /* find the mbuf in the chain where the checksum starts*/
 883         while ((m != NULL) && (offset >= m->m_len)) {
 884                 offset -= m->m_len;
 885                 m = m->m_next;
 886         }
 887         KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain."));
 888         KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs."));
 889         *(u_short *)(m->m_data + offset) = csum;
 890 }
 891
 892 /*
 893  * IP socket option processing.
 894  */
 895 int
 896 ip_ctloutput(struct socket *so, struct sockopt *sopt)
 897 {
 898         struct  inpcb *inp = sotoinpcb(so);
 899         int     error, optval;
 900 #ifdef  RSS
 901         uint32_t rss_bucket;
 902         int retval;
 903 #endif
 904
 905         error = optval = 0;
 906         if (sopt->sopt_level != IPPROTO_IP) {
 907                 error = EINVAL;
 908
 909                 if (sopt->sopt_level == SOL_SOCKET &&
 910                     sopt->sopt_dir == SOPT_SET) {
 911                         switch (sopt->sopt_name) {
 912                         case SO_REUSEADDR:
 913                                 INP_WLOCK(inp);
 914                                 if ((so->so_options & SO_REUSEADDR) != 0)
 915                                         inp->inp_flags2 |= INP_REUSEADDR;
 916                                 else
 917                                         inp->inp_flags2 &= ~INP_REUSEADDR;
 918                                 INP_WUNLOCK(inp);
 919                                 error = 0;
 920                                 break;
 921                         case SO_REUSEPORT:
 922                                 INP_WLOCK(inp);
 923                                 if ((so->so_options & SO_REUSEPORT) != 0)
 924                                         inp->inp_flags2 |= INP_REUSEPORT;
 925                                 else
 926                                         inp->inp_flags2 &= ~INP_REUSEPORT;
 927                                 INP_WUNLOCK(inp);
 928                                 error = 0;
 929                                 break;
 930                         case SO_SETFIB:
 931                                 INP_WLOCK(inp);
 932                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
 933                                 INP_WUNLOCK(inp);
 934                                 error = 0;
 935                                 break;
 936                         default:
 937                                 break;
 938                         }
 939                 }
 940                 return (error);
 941         }
 942
 943         switch (sopt->sopt_dir) {
 944         case SOPT_SET:
 945                 switch (sopt->sopt_name) {
 946                 case IP_OPTIONS:
 947 #ifdef notyet
 948                 case IP_RETOPTS:
 949 #endif
 950                 {
 951                         struct mbuf *m;
 952                         if (sopt->sopt_valsize > MLEN) {
 953                                 error = EMSGSIZE;
 954                                 break;
 955                         }
 956                         m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
 957                         if (m == NULL) {
 958                                 error = ENOBUFS;
 959                                 break;
 960                         }
 961                         m->m_len = sopt->sopt_valsize;
 962                         error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
 963                                             m->m_len);
 964                         if (error) {
 965                                 m_free(m);
 966                                 break;
 967                         }
 968                         INP_WLOCK(inp);
 969                         error = ip_pcbopts(inp, sopt->sopt_name, m);
 970                         INP_WUNLOCK(inp);
 971                         return (error);
 972                 }
 973
 974                 case IP_BINDANY:
 975                         if (sopt->sopt_td != NULL) {
 976                                 error = priv_check(sopt->sopt_td,
 977                                     PRIV_NETINET_BINDANY);
 978                                 if (error)
 979                                         break;
 980                         }
 981                         /* FALLTHROUGH */
 982                 case IP_BINDMULTI:
 983 #ifdef  RSS
 984                 case IP_RSS_LISTEN_BUCKET:
 985 #endif
 986                 case IP_TOS:
 987                 case IP_TTL:
 988                 case IP_MINTTL:
 989                 case IP_RECVOPTS:
 990                 case IP_RECVRETOPTS:
 991                 case IP_RECVDSTADDR:
 992                 case IP_RECVTTL:
 993                 case IP_RECVIF:
 994                 case IP_FAITH:
 995                 case IP_ONESBCAST:
 996                 case IP_DONTFRAG:
 997                 case IP_RECVTOS:
 998                 case IP_RECVFLOWID:
 999 #ifdef  RSS
1000                 case IP_RECVRSSBUCKETID:
1001 #endif
1002                         error = sooptcopyin(sopt, &optval, sizeof optval,
1003                                             sizeof optval);
1004                         if (error)
1005                                 break;
1006
1007                         switch (sopt->sopt_name) {
1008                         case IP_TOS:
1009                                 inp->inp_ip_tos = optval;
1010                                 break;
1011
1012                         case IP_TTL:
1013                                 inp->inp_ip_ttl = optval;
1014                                 break;
1015
1016                         case IP_MINTTL:
1017                                 if (optval >= 0 && optval <= MAXTTL)
1018                                         inp->inp_ip_minttl = optval;
1019                                 else
1020                                         error = EINVAL;
1021                                 break;
1022
1023 #define OPTSET(bit) do {                                                \
1024         INP_WLOCK(inp);                                                 \
1025         if (optval)                                                     \
1026                 inp->inp_flags |= bit;                                  \
1027         else                                                            \
1028                 inp->inp_flags &= ~bit;                                 \
1029         INP_WUNLOCK(inp);                                               \
1030 } while (0)
1031
1032 #define OPTSET2(bit, val) do {                                          \
1033         INP_WLOCK(inp);                                                 \
1034         if (val)                                                        \
1035                 inp->inp_flags2 |= bit;                                 \
1036         else                                                            \
1037                 inp->inp_flags2 &= ~bit;                                \
1038         INP_WUNLOCK(inp);                                               \
1039 } while (0)
1040
1041                         case IP_RECVOPTS:
1042                                 OPTSET(INP_RECVOPTS);
1043                                 break;
1044
1045                         case IP_RECVRETOPTS:
1046                                 OPTSET(INP_RECVRETOPTS);
1047                                 break;
1048
1049                         case IP_RECVDSTADDR:
1050                                 OPTSET(INP_RECVDSTADDR);
1051                                 break;
1052
1053                         case IP_RECVTTL:
1054                                 OPTSET(INP_RECVTTL);
1055                                 break;
1056
1057                         case IP_RECVIF:
1058                                 OPTSET(INP_RECVIF);
1059                                 break;
1060
1061                         case IP_FAITH:
1062                                 OPTSET(INP_FAITH);
1063                                 break;
1064
1065                         case IP_ONESBCAST:
1066                                 OPTSET(INP_ONESBCAST);
1067                                 break;
1068                         case IP_DONTFRAG:
1069                                 OPTSET(INP_DONTFRAG);
1070                                 break;
1071                         case IP_BINDANY:
1072                                 OPTSET(INP_BINDANY);
1073                                 break;
1074                         case IP_RECVTOS:
1075                                 OPTSET(INP_RECVTOS);
1076                                 break;
1077                         case IP_BINDMULTI:
1078                                 OPTSET2(INP_BINDMULTI, optval);
1079                                 break;
1080                         case IP_RECVFLOWID:
1081                                 OPTSET2(INP_RECVFLOWID, optval);
1082                                 break;
1083 #ifdef  RSS
1084                         case IP_RSS_LISTEN_BUCKET:
1085                                 if ((optval >= 0) &&
1086                                     (optval < rss_getnumbuckets())) {
1087                                         inp->inp_rss_listen_bucket = optval;
1088                                         OPTSET2(INP_RSS_BUCKET_SET, 1);
1089                                 } else {
1090                                         error = EINVAL;
1091                                 }
1092                                 break;
1093                         case IP_RECVRSSBUCKETID:
1094                                 OPTSET2(INP_RECVRSSBUCKETID, optval);
1095                                 break;
1096 #endif
1097                         }
1098                         break;
1099 #undef OPTSET
1100 #undef OPTSET2
1101
1102                 /*
1103                  * Multicast socket options are processed by the in_mcast
1104                  * module.
1105                  */
1106                 case IP_MULTICAST_IF:
1107                 case IP_MULTICAST_VIF:
1108                 case IP_MULTICAST_TTL:
1109                 case IP_MULTICAST_LOOP:
1110                 case IP_ADD_MEMBERSHIP:
1111                 case IP_DROP_MEMBERSHIP:
1112                 case IP_ADD_SOURCE_MEMBERSHIP:
1113                 case IP_DROP_SOURCE_MEMBERSHIP:
1114                 case IP_BLOCK_SOURCE:
1115                 case IP_UNBLOCK_SOURCE:
1116                 case IP_MSFILTER:
1117                 case MCAST_JOIN_GROUP:
1118                 case MCAST_LEAVE_GROUP:
1119                 case MCAST_JOIN_SOURCE_GROUP:
1120                 case MCAST_LEAVE_SOURCE_GROUP:
1121                 case MCAST_BLOCK_SOURCE:
1122                 case MCAST_UNBLOCK_SOURCE:
1123                         error = inp_setmoptions(inp, sopt);
1124                         break;
1125
1126                 case IP_PORTRANGE:
1127                         error = sooptcopyin(sopt, &optval, sizeof optval,
1128                                             sizeof optval);
1129                         if (error)
1130                                 break;
1131
1132                         INP_WLOCK(inp);
1133                         switch (optval) {
1134                         case IP_PORTRANGE_DEFAULT:
1135                                 inp->inp_flags &= ~(INP_LOWPORT);
1136                                 inp->inp_flags &= ~(INP_HIGHPORT);
1137                                 break;
1138
1139                         case IP_PORTRANGE_HIGH:
1140                                 inp->inp_flags &= ~(INP_LOWPORT);
1141                                 inp->inp_flags |= INP_HIGHPORT;
1142                                 break;
1143
1144                         case IP_PORTRANGE_LOW:
1145                                 inp->inp_flags &= ~(INP_HIGHPORT);
1146                                 inp->inp_flags |= INP_LOWPORT;
1147                                 break;
1148
1149                         default:
1150                                 error = EINVAL;
1151                                 break;
1152                         }
1153                         INP_WUNLOCK(inp);
1154                         break;
1155
1156 #ifdef IPSEC
1157                 case IP_IPSEC_POLICY:
1158                 {
1159                         caddr_t req;
1160                         struct mbuf *m;
1161
1162                         if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1163                                 break;
1164                         if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1165                                 break;
1166                         req = mtod(m, caddr_t);
1167                         error = ipsec_set_policy(inp, sopt->sopt_name, req,
1168                             m->m_len, (sopt->sopt_td != NULL) ?
1169                             sopt->sopt_td->td_ucred : NULL);
1170                         m_freem(m);
1171                         break;
1172                 }
1173 #endif /* IPSEC */
1174
1175                 default:
1176                         error = ENOPROTOOPT;
1177                         break;
1178                 }
1179                 break;
1180
1181         case SOPT_GET:
1182                 switch (sopt->sopt_name) {
1183                 case IP_OPTIONS:
1184                 case IP_RETOPTS:
1185                         if (inp->inp_options)
1186                                 error = sooptcopyout(sopt,
1187                                                      mtod(inp->inp_options,
1188                                                           char *),
1189                                                      inp->inp_options->m_len);
1190                         else
1191                                 sopt->sopt_valsize = 0;
1192                         break;
1193
1194                 case IP_TOS:
1195                 case IP_TTL:
1196                 case IP_MINTTL:
1197                 case IP_RECVOPTS:
1198                 case IP_RECVRETOPTS:
1199                 case IP_RECVDSTADDR:
1200                 case IP_RECVTTL:
1201                 case IP_RECVIF:
1202                 case IP_PORTRANGE:
1203                 case IP_FAITH:
1204                 case IP_ONESBCAST:
1205                 case IP_DONTFRAG:
1206                 case IP_BINDANY:
1207                 case IP_RECVTOS:
1208                 case IP_BINDMULTI:
1209                 case IP_FLOWID:
1210                 case IP_FLOWTYPE:
1211                 case IP_RECVFLOWID:
1212 #ifdef  RSS
1213                 case IP_RSSBUCKETID:
1214                 case IP_RECVRSSBUCKETID:
1215 #endif
1216                         switch (sopt->sopt_name) {
1217
1218                         case IP_TOS:
1219                                 optval = inp->inp_ip_tos;
1220                                 break;
1221
1222                         case IP_TTL:
1223                                 optval = inp->inp_ip_ttl;
1224                                 break;
1225
1226                         case IP_MINTTL:
1227                                 optval = inp->inp_ip_minttl;
1228                                 break;
1229
1230 #define OPTBIT(bit)     (inp->inp_flags & bit ? 1 : 0)
1231 #define OPTBIT2(bit)    (inp->inp_flags2 & bit ? 1 : 0)
1232
1233                         case IP_RECVOPTS:
1234                                 optval = OPTBIT(INP_RECVOPTS);
1235                                 break;
1236
1237                         case IP_RECVRETOPTS:
1238                                 optval = OPTBIT(INP_RECVRETOPTS);
1239                                 break;
1240
1241                         case IP_RECVDSTADDR:
1242                                 optval = OPTBIT(INP_RECVDSTADDR);
1243                                 break;
1244
1245                         case IP_RECVTTL:
1246                                 optval = OPTBIT(INP_RECVTTL);
1247                                 break;
1248
1249                         case IP_RECVIF:
1250                                 optval = OPTBIT(INP_RECVIF);
1251                                 break;
1252
1253                         case IP_PORTRANGE:
1254                                 if (inp->inp_flags & INP_HIGHPORT)
1255                                         optval = IP_PORTRANGE_HIGH;
1256                                 else if (inp->inp_flags & INP_LOWPORT)
1257                                         optval = IP_PORTRANGE_LOW;
1258                                 else
1259                                         optval = 0;
1260                                 break;
1261
1262                         case IP_FAITH:
1263                                 optval = OPTBIT(INP_FAITH);
1264                                 break;
1265
1266                         case IP_ONESBCAST:
1267                                 optval = OPTBIT(INP_ONESBCAST);
1268                                 break;
1269                         case IP_DONTFRAG:
1270                                 optval = OPTBIT(INP_DONTFRAG);
1271                                 break;
1272                         case IP_BINDANY:
1273                                 optval = OPTBIT(INP_BINDANY);
1274                                 break;
1275                         case IP_RECVTOS:
1276                                 optval = OPTBIT(INP_RECVTOS);
1277                                 break;
1278                         case IP_FLOWID:
1279                                 optval = inp->inp_flowid;
1280                                 break;
1281                         case IP_FLOWTYPE:
1282                                 optval = inp->inp_flowtype;
1283                                 break;
1284                         case IP_RECVFLOWID:
1285                                 optval = OPTBIT2(INP_RECVFLOWID);
1286                                 break;
1287 #ifdef  RSS
1288                         case IP_RSSBUCKETID:
1289                                 retval = rss_hash2bucket(inp->inp_flowid,
1290                                     inp->inp_flowtype,
1291                                     &rss_bucket);
1292                                 if (retval == 0)
1293                                         optval = rss_bucket;
1294                                 else
1295                                         error = EINVAL;
1296                                 break;
1297                         case IP_RECVRSSBUCKETID:
1298                                 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1299                                 break;
1300 #endif
1301                         case IP_BINDMULTI:
1302                                 optval = OPTBIT2(INP_BINDMULTI);
1303                                 break;
1304                         }
1305                         error = sooptcopyout(sopt, &optval, sizeof optval);
1306                         break;
1307
1308                 /*
1309                  * Multicast socket options are processed by the in_mcast
1310                  * module.
1311                  */
1312                 case IP_MULTICAST_IF:
1313                 case IP_MULTICAST_VIF:
1314                 case IP_MULTICAST_TTL:
1315                 case IP_MULTICAST_LOOP:
1316                 case IP_MSFILTER:
1317                         error = inp_getmoptions(inp, sopt);
1318                         break;
1319
1320 #ifdef IPSEC
1321                 case IP_IPSEC_POLICY:
1322                 {
1323                         struct mbuf *m = NULL;
1324                         caddr_t req = NULL;
1325                         size_t len = 0;
1326
1327                         if (m != 0) {
1328                                 req = mtod(m, caddr_t);
1329                                 len = m->m_len;
1330                         }
1331                         error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1332                         if (error == 0)
1333                                 error = soopt_mcopyout(sopt, m); /* XXX */
1334                         if (error == 0)
1335                                 m_freem(m);
1336                         break;
1337                 }
1338 #endif /* IPSEC */
1339
1340                 default:
1341                         error = ENOPROTOOPT;
1342                         break;
1343                 }
1344                 break;
1345         }
1346         return (error);
1347 }
1348
1349 /*
1350  * Routine called from ip_output() to loop back a copy of an IP multicast
1351  * packet to the input queue of a specified interface.  Note that this
1352  * calls the output routine of the loopback "driver", but with an interface
1353  * pointer that might NOT be a loopback interface -- evil, but easier than
1354  * replicating that code here.
1355  */
1356 static void
1357 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1358     int hlen)
1359 {
1360         register struct ip *ip;
1361         struct mbuf *copym;
1362
1363         /*
1364          * Make a deep copy of the packet because we're going to
1365          * modify the pack in order to generate checksums.
1366          */
1367         copym = m_dup(m, M_NOWAIT);
1368         if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1369                 copym = m_pullup(copym, hlen);
1370         if (copym != NULL) {
1371                 /* If needed, compute the checksum and mark it as valid. */
1372                 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1373                         in_delayed_cksum(copym);
1374                         copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1375                         copym->m_pkthdr.csum_flags |=
1376                             CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1377                         copym->m_pkthdr.csum_data = 0xffff;
1378                 }
1379                 /*
1380                  * We don't bother to fragment if the IP length is greater
1381                  * than the interface's MTU.  Can this possibly matter?
1382                  */
1383                 ip = mtod(copym, struct ip *);
1384                 ip->ip_sum = 0;
1385                 ip->ip_sum = in_cksum(copym, hlen);
1386 #if 1 /* XXX */
1387                 if (dst->sin_family != AF_INET) {
1388                         printf("ip_mloopback: bad address family %d\n",
1389                                                 dst->sin_family);
1390                         dst->sin_family = AF_INET;
1391                 }
1392 #endif
1393                 if_simloop(ifp, copym, dst->sin_family, 0);
1394         }
1395 }