sys/netinet/ip_output.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-3-Clause
   3  *
   4  * Copyright (c) 1982, 1986, 1988, 1990, 1993
   5  *      The Regents of the University of California.  All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 3. Neither the name of the University nor the names of its contributors
  16  *    may be used to endorse or promote products derived from this software
  17  *    without specific prior written permission.
  18  *
  19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  29  * SUCH DAMAGE.
  30  *
  31  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  32  */
  33
  34 #include <sys/cdefs.h>
  35 __FBSDID("$FreeBSD$");
  36
  37 #include "opt_inet.h"
  38 #include "opt_ratelimit.h"
  39 #include "opt_ipsec.h"
  40 #include "opt_mbuf_stress_test.h"
  41 #include "opt_mpath.h"
  42 #include "opt_route.h"
  43 #include "opt_sctp.h"
  44 #include "opt_rss.h"
  45
  46 #include <sys/param.h>
  47 #include <sys/systm.h>
  48 #include <sys/kernel.h>
  49 #include <sys/lock.h>
  50 #include <sys/malloc.h>
  51 #include <sys/mbuf.h>
  52 #include <sys/priv.h>
  53 #include <sys/proc.h>
  54 #include <sys/protosw.h>
  55 #include <sys/rmlock.h>
  56 #include <sys/sdt.h>
  57 #include <sys/socket.h>
  58 #include <sys/socketvar.h>
  59 #include <sys/sysctl.h>
  60 #include <sys/ucred.h>
  61
  62 #include <net/if.h>
  63 #include <net/if_var.h>
  64 #include <net/if_llatbl.h>
  65 #include <net/netisr.h>
  66 #include <net/pfil.h>
  67 #include <net/route.h>
  68 #ifdef RADIX_MPATH
  69 #include <net/radix_mpath.h>
  70 #endif
  71 #include <net/rss_config.h>
  72 #include <net/vnet.h>
  73
  74 #include <netinet/in.h>
  75 #include <netinet/in_kdtrace.h>
  76 #include <netinet/in_systm.h>
  77 #include <netinet/ip.h>
  78 #include <netinet/in_pcb.h>
  79 #include <netinet/in_rss.h>
  80 #include <netinet/in_var.h>
  81 #include <netinet/ip_var.h>
  82 #include <netinet/ip_options.h>
  83
  84 #include <netinet/udp.h>
  85 #include <netinet/udp_var.h>
  86
  87 #ifdef SCTP
  88 #include <netinet/sctp.h>
  89 #include <netinet/sctp_crc32.h>
  90 #endif
  91
  92 #include <netipsec/ipsec_support.h>
  93
  94 #include <machine/in_cksum.h>
  95
  96 #include <security/mac/mac_framework.h>
  97
  98 #ifdef MBUF_STRESS_TEST
  99 static int mbuf_frag_size = 0;
 100 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
 101         &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
 102 #endif
 103
 104 static void     ip_mloopback(struct ifnet *, const struct mbuf *, int);
 105
 106
 107 extern int in_mcast_loop;
 108 extern  struct protosw inetsw[];
 109
 110 static inline int
 111 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, struct inpcb *inp,
 112     struct sockaddr_in *dst, int *fibnum, int *error)
 113 {
 114         struct m_tag *fwd_tag = NULL;
 115         struct mbuf *m;
 116         struct in_addr odst;
 117         struct ip *ip;
 118
 119         m = *mp;
 120         ip = mtod(m, struct ip *);
 121
 122         /* Run through list of hooks for output packets. */
 123         odst.s_addr = ip->ip_dst.s_addr;
 124         *error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp, PFIL_OUT, 0, inp);
 125         m = *mp;
 126         if ((*error) != 0 || m == NULL)
 127                 return 1; /* Finished */
 128
 129         ip = mtod(m, struct ip *);
 130
 131         /* See if destination IP address was changed by packet filter. */
 132         if (odst.s_addr != ip->ip_dst.s_addr) {
 133                 m->m_flags |= M_SKIP_FIREWALL;
 134                 /* If destination is now ourself drop to ip_input(). */
 135                 if (in_localip(ip->ip_dst)) {
 136                         m->m_flags |= M_FASTFWD_OURS;
 137                         if (m->m_pkthdr.rcvif == NULL)
 138                                 m->m_pkthdr.rcvif = V_loif;
 139                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 140                                 m->m_pkthdr.csum_flags |=
 141                                         CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 142                                 m->m_pkthdr.csum_data = 0xffff;
 143                         }
 144                         m->m_pkthdr.csum_flags |=
 145                                 CSUM_IP_CHECKED | CSUM_IP_VALID;
 146 #ifdef SCTP
 147                         if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 148                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 149 #endif
 150                         *error = netisr_queue(NETISR_IP, m);
 151                         return 1; /* Finished */
 152                 }
 153
 154                 bzero(dst, sizeof(*dst));
 155                 dst->sin_family = AF_INET;
 156                 dst->sin_len = sizeof(*dst);
 157                 dst->sin_addr = ip->ip_dst;
 158
 159                 return -1; /* Reloop */
 160         }
 161         /* See if fib was changed by packet filter. */
 162         if ((*fibnum) != M_GETFIB(m)) {
 163                 m->m_flags |= M_SKIP_FIREWALL;
 164                 *fibnum = M_GETFIB(m);
 165                 return -1; /* Reloop for FIB change */
 166         }
 167
 168         /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
 169         if (m->m_flags & M_FASTFWD_OURS) {
 170                 if (m->m_pkthdr.rcvif == NULL)
 171                         m->m_pkthdr.rcvif = V_loif;
 172                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 173                         m->m_pkthdr.csum_flags |=
 174                                 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 175                         m->m_pkthdr.csum_data = 0xffff;
 176                 }
 177 #ifdef SCTP
 178                 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 179                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 180 #endif
 181                 m->m_pkthdr.csum_flags |=
 182                         CSUM_IP_CHECKED | CSUM_IP_VALID;
 183
 184                 *error = netisr_queue(NETISR_IP, m);
 185                 return 1; /* Finished */
 186         }
 187         /* Or forward to some other address? */
 188         if ((m->m_flags & M_IP_NEXTHOP) &&
 189             ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
 190                 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
 191                 m->m_flags |= M_SKIP_FIREWALL;
 192                 m->m_flags &= ~M_IP_NEXTHOP;
 193                 m_tag_delete(m, fwd_tag);
 194
 195                 return -1; /* Reloop for CHANGE of dst */
 196         }
 197
 198         return 0;
 199 }
 200
 201 /*
 202  * IP output.  The packet in mbuf chain m contains a skeletal IP
 203  * header (with len, off, ttl, proto, tos, src, dst).
 204  * The mbuf chain containing the packet will be freed.
 205  * The mbuf opt, if present, will not be freed.
 206  * If route ro is present and has ro_rt initialized, route lookup would be
 207  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 208  * then result of route lookup is stored in ro->ro_rt.
 209  *
 210  * In the IP forwarding case, the packet will arrive with options already
 211  * inserted, so must have a NULL opt pointer.
 212  */
 213 int
 214 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
 215     struct ip_moptions *imo, struct inpcb *inp)
 216 {
 217         struct rm_priotracker in_ifa_tracker;
 218         struct ip *ip;
 219         struct ifnet *ifp = NULL;       /* keep compiler happy */
 220         struct mbuf *m0;
 221         int hlen = sizeof (struct ip);
 222         int mtu;
 223         int error = 0;
 224         struct sockaddr_in *dst;
 225         const struct sockaddr_in *gw;
 226         struct in_ifaddr *ia;
 227         int isbroadcast;
 228         uint16_t ip_len, ip_off;
 229         struct route iproute;
 230         struct rtentry *rte;    /* cache for ro->ro_rt */
 231         uint32_t fibnum;
 232 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 233         int no_route_but_check_spd = 0;
 234 #endif
 235         M_ASSERTPKTHDR(m);
 236
 237         if (inp != NULL) {
 238                 INP_LOCK_ASSERT(inp);
 239                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 240                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 241                         m->m_pkthdr.flowid = inp->inp_flowid;
 242                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 243                 }
 244         }
 245
 246         if (ro == NULL) {
 247                 ro = &iproute;
 248                 bzero(ro, sizeof (*ro));
 249         }
 250
 251         if (opt) {
 252                 int len = 0;
 253                 m = ip_insertoptions(m, opt, &len);
 254                 if (len != 0)
 255                         hlen = len; /* ip->ip_hl is updated above */
 256         }
 257         ip = mtod(m, struct ip *);
 258         ip_len = ntohs(ip->ip_len);
 259         ip_off = ntohs(ip->ip_off);
 260
 261         if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
 262                 ip->ip_v = IPVERSION;
 263                 ip->ip_hl = hlen >> 2;
 264                 ip_fillid(ip);
 265                 IPSTAT_INC(ips_localout);
 266         } else {
 267                 /* Header already set, fetch hlen from there */
 268                 hlen = ip->ip_hl << 2;
 269         }
 270
 271         /*
 272          * dst/gw handling:
 273          *
 274          * dst can be rewritten but always points to &ro->ro_dst.
 275          * gw is readonly but can point either to dst OR rt_gateway,
 276          * therefore we need restore gw if we're redoing lookup.
 277          */
 278         gw = dst = (struct sockaddr_in *)&ro->ro_dst;
 279         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 280         rte = ro->ro_rt;
 281         if (rte == NULL) {
 282                 bzero(dst, sizeof(*dst));
 283                 dst->sin_family = AF_INET;
 284                 dst->sin_len = sizeof(*dst);
 285                 dst->sin_addr = ip->ip_dst;
 286         }
 287         NET_EPOCH_ENTER();
 288 again:
 289         /*
 290          * Validate route against routing table additions;
 291          * a better/more specific route might have been added.
 292          */
 293         if (inp)
 294                 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
 295         /*
 296          * If there is a cached route,
 297          * check that it is to the same destination
 298          * and is still up.  If not, free it and try again.
 299          * The address family should also be checked in case of sharing the
 300          * cache with IPv6.
 301          * Also check whether routing cache needs invalidation.
 302          */
 303         rte = ro->ro_rt;
 304         if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
 305                     rte->rt_ifp == NULL ||
 306                     !RT_LINK_IS_UP(rte->rt_ifp) ||
 307                           dst->sin_family != AF_INET ||
 308                           dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
 309                 RO_INVALIDATE_CACHE(ro);
 310                 rte = NULL;
 311         }
 312         ia = NULL;
 313         /*
 314          * If routing to interface only, short circuit routing lookup.
 315          * The use of an all-ones broadcast address implies this; an
 316          * interface is specified by the broadcast address of an interface,
 317          * or the destination address of a ptp interface.
 318          */
 319         if (flags & IP_SENDONES) {
 320                 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
 321                                                       M_GETFIB(m)))) == NULL &&
 322                     (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 323                                                     M_GETFIB(m)))) == NULL) {
 324                         IPSTAT_INC(ips_noroute);
 325                         error = ENETUNREACH;
 326                         goto bad;
 327                 }
 328                 ip->ip_dst.s_addr = INADDR_BROADCAST;
 329                 dst->sin_addr = ip->ip_dst;
 330                 ifp = ia->ia_ifp;
 331                 ip->ip_ttl = 1;
 332                 isbroadcast = 1;
 333         } else if (flags & IP_ROUTETOIF) {
 334                 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 335                                                     M_GETFIB(m)))) == NULL &&
 336                     (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
 337                                                 M_GETFIB(m)))) == NULL) {
 338                         IPSTAT_INC(ips_noroute);
 339                         error = ENETUNREACH;
 340                         goto bad;
 341                 }
 342                 ifp = ia->ia_ifp;
 343                 ip->ip_ttl = 1;
 344                 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
 345                     in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
 346         } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
 347             imo != NULL && imo->imo_multicast_ifp != NULL) {
 348                 /*
 349                  * Bypass the normal routing lookup for multicast
 350                  * packets if the interface is specified.
 351                  */
 352                 ifp = imo->imo_multicast_ifp;
 353                 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
 354                 isbroadcast = 0;        /* fool gcc */
 355         } else {
 356                 /*
 357                  * We want to do any cloning requested by the link layer,
 358                  * as this is probably required in all cases for correct
 359                  * operation (as it is for ARP).
 360                  */
 361                 if (rte == NULL) {
 362 #ifdef RADIX_MPATH
 363                         rtalloc_mpath_fib(ro,
 364                             ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
 365                             fibnum);
 366 #else
 367                         in_rtalloc_ign(ro, 0, fibnum);
 368 #endif
 369                         rte = ro->ro_rt;
 370                 }
 371                 if (rte == NULL ||
 372                     (rte->rt_flags & RTF_UP) == 0 ||
 373                     rte->rt_ifp == NULL ||
 374                     !RT_LINK_IS_UP(rte->rt_ifp)) {
 375 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 376                         /*
 377                          * There is no route for this packet, but it is
 378                          * possible that a matching SPD entry exists.
 379                          */
 380                         no_route_but_check_spd = 1;
 381                         mtu = 0; /* Silence GCC warning. */
 382                         goto sendit;
 383 #endif
 384                         IPSTAT_INC(ips_noroute);
 385                         error = EHOSTUNREACH;
 386                         goto bad;
 387                 }
 388                 ia = ifatoia(rte->rt_ifa);
 389                 ifp = rte->rt_ifp;
 390                 counter_u64_add(rte->rt_pksent, 1);
 391                 rt_update_ro_flags(ro);
 392                 if (rte->rt_flags & RTF_GATEWAY)
 393                         gw = (struct sockaddr_in *)rte->rt_gateway;
 394                 if (rte->rt_flags & RTF_HOST)
 395                         isbroadcast = (rte->rt_flags & RTF_BROADCAST);
 396                 else if (ifp->if_flags & IFF_BROADCAST)
 397                         isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
 398                 else
 399                         isbroadcast = 0;
 400         }
 401
 402         /*
 403          * Calculate MTU.  If we have a route that is up, use that,
 404          * otherwise use the interface's MTU.
 405          */
 406         if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST)))
 407                 mtu = rte->rt_mtu;
 408         else
 409                 mtu = ifp->if_mtu;
 410         /* Catch a possible divide by zero later. */
 411         KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
 412             __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
 413
 414         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
 415                 m->m_flags |= M_MCAST;
 416                 /*
 417                  * IP destination address is multicast.  Make sure "gw"
 418                  * still points to the address in "ro".  (It may have been
 419                  * changed to point to a gateway address, above.)
 420                  */
 421                 gw = dst;
 422                 /*
 423                  * See if the caller provided any multicast options
 424                  */
 425                 if (imo != NULL) {
 426                         ip->ip_ttl = imo->imo_multicast_ttl;
 427                         if (imo->imo_multicast_vif != -1)
 428                                 ip->ip_src.s_addr =
 429                                     ip_mcast_src ?
 430                                     ip_mcast_src(imo->imo_multicast_vif) :
 431                                     INADDR_ANY;
 432                 } else
 433                         ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
 434                 /*
 435                  * Confirm that the outgoing interface supports multicast.
 436                  */
 437                 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
 438                         if ((ifp->if_flags & IFF_MULTICAST) == 0) {
 439                                 IPSTAT_INC(ips_noroute);
 440                                 error = ENETUNREACH;
 441                                 goto bad;
 442                         }
 443                 }
 444                 /*
 445                  * If source address not specified yet, use address
 446                  * of outgoing interface.
 447                  */
 448                 if (ip->ip_src.s_addr == INADDR_ANY) {
 449                         /* Interface may have no addresses. */
 450                         if (ia != NULL)
 451                                 ip->ip_src = IA_SIN(ia)->sin_addr;
 452                 }
 453
 454                 if ((imo == NULL && in_mcast_loop) ||
 455                     (imo && imo->imo_multicast_loop)) {
 456                         /*
 457                          * Loop back multicast datagram if not expressly
 458                          * forbidden to do so, even if we are not a member
 459                          * of the group; ip_input() will filter it later,
 460                          * thus deferring a hash lookup and mutex acquisition
 461                          * at the expense of a cheap copy using m_copym().
 462                          */
 463                         ip_mloopback(ifp, m, hlen);
 464                 } else {
 465                         /*
 466                          * If we are acting as a multicast router, perform
 467                          * multicast forwarding as if the packet had just
 468                          * arrived on the interface to which we are about
 469                          * to send.  The multicast forwarding function
 470                          * recursively calls this function, using the
 471                          * IP_FORWARDING flag to prevent infinite recursion.
 472                          *
 473                          * Multicasts that are looped back by ip_mloopback(),
 474                          * above, will be forwarded by the ip_input() routine,
 475                          * if necessary.
 476                          */
 477                         if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
 478                                 /*
 479                                  * If rsvp daemon is not running, do not
 480                                  * set ip_moptions. This ensures that the packet
 481                                  * is multicast and not just sent down one link
 482                                  * as prescribed by rsvpd.
 483                                  */
 484                                 if (!V_rsvp_on)
 485                                         imo = NULL;
 486                                 if (ip_mforward &&
 487                                     ip_mforward(ip, ifp, m, imo) != 0) {
 488                                         m_freem(m);
 489                                         goto done;
 490                                 }
 491                         }
 492                 }
 493
 494                 /*
 495                  * Multicasts with a time-to-live of zero may be looped-
 496                  * back, above, but must not be transmitted on a network.
 497                  * Also, multicasts addressed to the loopback interface
 498                  * are not sent -- the above call to ip_mloopback() will
 499                  * loop back a copy. ip_input() will drop the copy if
 500                  * this host does not belong to the destination group on
 501                  * the loopback interface.
 502                  */
 503                 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
 504                         m_freem(m);
 505                         goto done;
 506                 }
 507
 508                 goto sendit;
 509         }
 510
 511         /*
 512          * If the source address is not specified yet, use the address
 513          * of the outoing interface.
 514          */
 515         if (ip->ip_src.s_addr == INADDR_ANY) {
 516                 /* Interface may have no addresses. */
 517                 if (ia != NULL) {
 518                         ip->ip_src = IA_SIN(ia)->sin_addr;
 519                 }
 520         }
 521
 522         /*
 523          * Look for broadcast address and
 524          * verify user is allowed to send
 525          * such a packet.
 526          */
 527         if (isbroadcast) {
 528                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 529                         error = EADDRNOTAVAIL;
 530                         goto bad;
 531                 }
 532                 if ((flags & IP_ALLOWBROADCAST) == 0) {
 533                         error = EACCES;
 534                         goto bad;
 535                 }
 536                 /* don't allow broadcast messages to be fragmented */
 537                 if (ip_len > mtu) {
 538                         error = EMSGSIZE;
 539                         goto bad;
 540                 }
 541                 m->m_flags |= M_BCAST;
 542         } else {
 543                 m->m_flags &= ~M_BCAST;
 544         }
 545
 546 sendit:
 547 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 548         if (IPSEC_ENABLED(ipv4)) {
 549                 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
 550                         if (error == EINPROGRESS)
 551                                 error = 0;
 552                         goto done;
 553                 }
 554         }
 555         /*
 556          * Check if there was a route for this packet; return error if not.
 557          */
 558         if (no_route_but_check_spd) {
 559                 IPSTAT_INC(ips_noroute);
 560                 error = EHOSTUNREACH;
 561                 goto bad;
 562         }
 563         /* Update variables that are affected by ipsec4_output(). */
 564         ip = mtod(m, struct ip *);
 565         hlen = ip->ip_hl << 2;
 566 #endif /* IPSEC */
 567
 568         /* Jump over all PFIL processing if hooks are not active. */
 569         if (PFIL_HOOKED(&V_inet_pfil_hook)) {
 570                 switch (ip_output_pfil(&m, ifp, inp, dst, &fibnum, &error)) {
 571                 case 1: /* Finished */
 572                         goto done;
 573
 574                 case 0: /* Continue normally */
 575                         ip = mtod(m, struct ip *);
 576                         break;
 577
 578                 case -1: /* Need to try again */
 579                         /* Reset everything for a new round */
 580                         RO_RTFREE(ro);
 581                         ro->ro_prepend = NULL;
 582                         rte = NULL;
 583                         gw = dst;
 584                         ip = mtod(m, struct ip *);
 585                         goto again;
 586
 587                 }
 588         }
 589
 590         /* 127/8 must not appear on wire - RFC1122. */
 591         if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
 592             (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
 593                 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
 594                         IPSTAT_INC(ips_badaddr);
 595                         error = EADDRNOTAVAIL;
 596                         goto bad;
 597                 }
 598         }
 599
 600         m->m_pkthdr.csum_flags |= CSUM_IP;
 601         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
 602                 in_delayed_cksum(m);
 603                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 604         }
 605 #ifdef SCTP
 606         if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
 607                 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
 608                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 609         }
 610 #endif
 611
 612         /*
 613          * If small enough for interface, or the interface will take
 614          * care of the fragmentation for us, we can just send directly.
 615          */
 616         if (ip_len <= mtu ||
 617             (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
 618                 ip->ip_sum = 0;
 619                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
 620                         ip->ip_sum = in_cksum(m, hlen);
 621                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 622                 }
 623
 624                 /*
 625                  * Record statistics for this interface address.
 626                  * With CSUM_TSO the byte/packet count will be slightly
 627                  * incorrect because we count the IP+TCP headers only
 628                  * once instead of for every generated packet.
 629                  */
 630                 if (!(flags & IP_FORWARDING) && ia) {
 631                         if (m->m_pkthdr.csum_flags & CSUM_TSO)
 632                                 counter_u64_add(ia->ia_ifa.ifa_opackets,
 633                                     m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
 634                         else
 635                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 636
 637                         counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
 638                 }
 639 #ifdef MBUF_STRESS_TEST
 640                 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
 641                         m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
 642 #endif
 643                 /*
 644                  * Reset layer specific mbuf flags
 645                  * to avoid confusing lower layers.
 646                  */
 647                 m_clrprotoflags(m);
 648                 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
 649 #ifdef RATELIMIT
 650                 if (inp != NULL) {
 651                         if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
 652                                 in_pcboutput_txrtlmt(inp, ifp, m);
 653                         /* stamp send tag on mbuf */
 654                         m->m_pkthdr.snd_tag = inp->inp_snd_tag;
 655                 } else {
 656                         m->m_pkthdr.snd_tag = NULL;
 657                 }
 658 #endif
 659                 error = (*ifp->if_output)(ifp, m,
 660                     (const struct sockaddr *)gw, ro);
 661 #ifdef RATELIMIT
 662                 /* check for route change */
 663                 if (error == EAGAIN)
 664                         in_pcboutput_eagain(inp);
 665 #endif
 666                 goto done;
 667         }
 668
 669         /* Balk when DF bit is set or the interface didn't support TSO. */
 670         if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
 671                 error = EMSGSIZE;
 672                 IPSTAT_INC(ips_cantfrag);
 673                 goto bad;
 674         }
 675
 676         /*
 677          * Too large for interface; fragment if possible. If successful,
 678          * on return, m will point to a list of packets to be sent.
 679          */
 680         error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
 681         if (error)
 682                 goto bad;
 683         for (; m; m = m0) {
 684                 m0 = m->m_nextpkt;
 685                 m->m_nextpkt = 0;
 686                 if (error == 0) {
 687                         /* Record statistics for this interface address. */
 688                         if (ia != NULL) {
 689                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 690                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
 691                                     m->m_pkthdr.len);
 692                         }
 693                         /*
 694                          * Reset layer specific mbuf flags
 695                          * to avoid confusing upper layers.
 696                          */
 697                         m_clrprotoflags(m);
 698
 699                         IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
 700                             mtod(m, struct ip *), NULL);
 701 #ifdef RATELIMIT
 702                         if (inp != NULL) {
 703                                 if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
 704                                         in_pcboutput_txrtlmt(inp, ifp, m);
 705                                 /* stamp send tag on mbuf */
 706                                 m->m_pkthdr.snd_tag = inp->inp_snd_tag;
 707                         } else {
 708                                 m->m_pkthdr.snd_tag = NULL;
 709                         }
 710 #endif
 711                         error = (*ifp->if_output)(ifp, m,
 712                             (const struct sockaddr *)gw, ro);
 713 #ifdef RATELIMIT
 714                         /* check for route change */
 715                         if (error == EAGAIN)
 716                                 in_pcboutput_eagain(inp);
 717 #endif
 718                 } else
 719                         m_freem(m);
 720         }
 721
 722         if (error == 0)
 723                 IPSTAT_INC(ips_fragmented);
 724
 725 done:
 726         if (ro == &iproute)
 727                 RO_RTFREE(ro);
 728         else if (rte == NULL)
 729                 /*
 730                  * If the caller supplied a route but somehow the reference
 731                  * to it has been released need to prevent the caller
 732                  * calling RTFREE on it again.
 733                  */
 734                 ro->ro_rt = NULL;
 735         NET_EPOCH_EXIT();
 736         return (error);
 737  bad:
 738         m_freem(m);
 739         goto done;
 740 }
 741
 742 /*
 743  * Create a chain of fragments which fit the given mtu. m_frag points to the
 744  * mbuf to be fragmented; on return it points to the chain with the fragments.
 745  * Return 0 if no error. If error, m_frag may contain a partially built
 746  * chain of fragments that should be freed by the caller.
 747  *
 748  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
 749  */
 750 int
 751 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
 752     u_long if_hwassist_flags)
 753 {
 754         int error = 0;
 755         int hlen = ip->ip_hl << 2;
 756         int len = (mtu - hlen) & ~7;    /* size of payload in each fragment */
 757         int off;
 758         struct mbuf *m0 = *m_frag;      /* the original packet          */
 759         int firstlen;
 760         struct mbuf **mnext;
 761         int nfrags;
 762         uint16_t ip_len, ip_off;
 763
 764         ip_len = ntohs(ip->ip_len);
 765         ip_off = ntohs(ip->ip_off);
 766
 767         if (ip_off & IP_DF) {   /* Fragmentation not allowed */
 768                 IPSTAT_INC(ips_cantfrag);
 769                 return EMSGSIZE;
 770         }
 771
 772         /*
 773          * Must be able to put at least 8 bytes per fragment.
 774          */
 775         if (len < 8)
 776                 return EMSGSIZE;
 777
 778         /*
 779          * If the interface will not calculate checksums on
 780          * fragmented packets, then do it here.
 781          */
 782         if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 783                 in_delayed_cksum(m0);
 784                 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 785         }
 786 #ifdef SCTP
 787         if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
 788                 sctp_delayed_cksum(m0, hlen);
 789                 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 790         }
 791 #endif
 792         if (len > PAGE_SIZE) {
 793                 /*
 794                  * Fragment large datagrams such that each segment
 795                  * contains a multiple of PAGE_SIZE amount of data,
 796                  * plus headers. This enables a receiver to perform
 797                  * page-flipping zero-copy optimizations.
 798                  *
 799                  * XXX When does this help given that sender and receiver
 800                  * could have different page sizes, and also mtu could
 801                  * be less than the receiver's page size ?
 802                  */
 803                 int newlen;
 804
 805                 off = MIN(mtu, m0->m_pkthdr.len);
 806
 807                 /*
 808                  * firstlen (off - hlen) must be aligned on an
 809                  * 8-byte boundary
 810                  */
 811                 if (off < hlen)
 812                         goto smart_frag_failure;
 813                 off = ((off - hlen) & ~7) + hlen;
 814                 newlen = (~PAGE_MASK) & mtu;
 815                 if ((newlen + sizeof (struct ip)) > mtu) {
 816                         /* we failed, go back the default */
 817 smart_frag_failure:
 818                         newlen = len;
 819                         off = hlen + len;
 820                 }
 821                 len = newlen;
 822
 823         } else {
 824                 off = hlen + len;
 825         }
 826
 827         firstlen = off - hlen;
 828         mnext = &m0->m_nextpkt;         /* pointer to next packet */
 829
 830         /*
 831          * Loop through length of segment after first fragment,
 832          * make new header and copy data of each part and link onto chain.
 833          * Here, m0 is the original packet, m is the fragment being created.
 834          * The fragments are linked off the m_nextpkt of the original
 835          * packet, which after processing serves as the first fragment.
 836          */
 837         for (nfrags = 1; off < ip_len; off += len, nfrags++) {
 838                 struct ip *mhip;        /* ip header on the fragment */
 839                 struct mbuf *m;
 840                 int mhlen = sizeof (struct ip);
 841
 842                 m = m_gethdr(M_NOWAIT, MT_DATA);
 843                 if (m == NULL) {
 844                         error = ENOBUFS;
 845                         IPSTAT_INC(ips_odropped);
 846                         goto done;
 847                 }
 848                 /*
 849                  * Make sure the complete packet header gets copied
 850                  * from the originating mbuf to the newly created
 851                  * mbuf. This also ensures that existing firewall
 852                  * classification(s), VLAN tags and so on get copied
 853                  * to the resulting fragmented packet(s):
 854                  */
 855                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 856                         m_free(m);
 857                         error = ENOBUFS;
 858                         IPSTAT_INC(ips_odropped);
 859                         goto done;
 860                 }
 861                 /*
 862                  * In the first mbuf, leave room for the link header, then
 863                  * copy the original IP header including options. The payload
 864                  * goes into an additional mbuf chain returned by m_copym().
 865                  */
 866                 m->m_data += max_linkhdr;
 867                 mhip = mtod(m, struct ip *);
 868                 *mhip = *ip;
 869                 if (hlen > sizeof (struct ip)) {
 870                         mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
 871                         mhip->ip_v = IPVERSION;
 872                         mhip->ip_hl = mhlen >> 2;
 873                 }
 874                 m->m_len = mhlen;
 875                 /* XXX do we need to add ip_off below ? */
 876                 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
 877                 if (off + len >= ip_len)
 878                         len = ip_len - off;
 879                 else
 880                         mhip->ip_off |= IP_MF;
 881                 mhip->ip_len = htons((u_short)(len + mhlen));
 882                 m->m_next = m_copym(m0, off, len, M_NOWAIT);
 883                 if (m->m_next == NULL) {        /* copy failed */
 884                         m_free(m);
 885                         error = ENOBUFS;        /* ??? */
 886                         IPSTAT_INC(ips_odropped);
 887                         goto done;
 888                 }
 889                 m->m_pkthdr.len = mhlen + len;
 890 #ifdef MAC
 891                 mac_netinet_fragment(m0, m);
 892 #endif
 893                 mhip->ip_off = htons(mhip->ip_off);
 894                 mhip->ip_sum = 0;
 895                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 896                         mhip->ip_sum = in_cksum(m, mhlen);
 897                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 898                 }
 899                 *mnext = m;
 900                 mnext = &m->m_nextpkt;
 901         }
 902         IPSTAT_ADD(ips_ofragments, nfrags);
 903
 904         /*
 905          * Update first fragment by trimming what's been copied out
 906          * and updating header.
 907          */
 908         m_adj(m0, hlen + firstlen - ip_len);
 909         m0->m_pkthdr.len = hlen + firstlen;
 910         ip->ip_len = htons((u_short)m0->m_pkthdr.len);
 911         ip->ip_off = htons(ip_off | IP_MF);
 912         ip->ip_sum = 0;
 913         if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 914                 ip->ip_sum = in_cksum(m0, hlen);
 915                 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
 916         }
 917
 918 done:
 919         *m_frag = m0;
 920         return error;
 921 }
 922
 923 void
 924 in_delayed_cksum(struct mbuf *m)
 925 {
 926         struct ip *ip;
 927         struct udphdr *uh;
 928         uint16_t cklen, csum, offset;
 929
 930         ip = mtod(m, struct ip *);
 931         offset = ip->ip_hl << 2 ;
 932
 933         if (m->m_pkthdr.csum_flags & CSUM_UDP) {
 934                 /* if udp header is not in the first mbuf copy udplen */
 935                 if (offset + sizeof(struct udphdr) > m->m_len)
 936                         m_copydata(m, offset + offsetof(struct udphdr,
 937                             uh_ulen), sizeof(cklen), (caddr_t)&cklen);
 938                 else {
 939                         uh = (struct udphdr *)mtodo(m, offset);
 940                         cklen = ntohs(uh->uh_ulen);
 941                 }
 942                 csum = in_cksum_skip(m, cklen + offset, offset);
 943                 if (csum == 0)
 944                         csum = 0xffff;
 945         } else {
 946                 cklen = ntohs(ip->ip_len);
 947                 csum = in_cksum_skip(m, cklen, offset);
 948         }
 949         offset += m->m_pkthdr.csum_data;        /* checksum offset */
 950
 951         if (offset + sizeof(csum) > m->m_len)
 952                 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
 953         else
 954                 *(u_short *)mtodo(m, offset) = csum;
 955 }
 956
 957 /*
 958  * IP socket option processing.
 959  */
 960 int
 961 ip_ctloutput(struct socket *so, struct sockopt *sopt)
 962 {
 963         struct  inpcb *inp = sotoinpcb(so);
 964         int     error, optval;
 965 #ifdef  RSS
 966         uint32_t rss_bucket;
 967         int retval;
 968 #endif
 969
 970         error = optval = 0;
 971         if (sopt->sopt_level != IPPROTO_IP) {
 972                 error = EINVAL;
 973
 974                 if (sopt->sopt_level == SOL_SOCKET &&
 975                     sopt->sopt_dir == SOPT_SET) {
 976                         switch (sopt->sopt_name) {
 977                         case SO_REUSEADDR:
 978                                 INP_WLOCK(inp);
 979                                 if ((so->so_options & SO_REUSEADDR) != 0)
 980                                         inp->inp_flags2 |= INP_REUSEADDR;
 981                                 else
 982                                         inp->inp_flags2 &= ~INP_REUSEADDR;
 983                                 INP_WUNLOCK(inp);
 984                                 error = 0;
 985                                 break;
 986                         case SO_REUSEPORT:
 987                                 INP_WLOCK(inp);
 988                                 if ((so->so_options & SO_REUSEPORT) != 0)
 989                                         inp->inp_flags2 |= INP_REUSEPORT;
 990                                 else
 991                                         inp->inp_flags2 &= ~INP_REUSEPORT;
 992                                 INP_WUNLOCK(inp);
 993                                 error = 0;
 994                                 break;
 995                         case SO_REUSEPORT_LB:
 996                                 INP_WLOCK(inp);
 997                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
 998                                         inp->inp_flags2 |= INP_REUSEPORT_LB;
 999                                 else
1000                                         inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1001                                 INP_WUNLOCK(inp);
1002                                 error = 0;
1003                                 break;
1004                         case SO_SETFIB:
1005                                 INP_WLOCK(inp);
1006                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
1007                                 INP_WUNLOCK(inp);
1008                                 error = 0;
1009                                 break;
1010                         case SO_MAX_PACING_RATE:
1011 #ifdef RATELIMIT
1012                                 INP_WLOCK(inp);
1013                                 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1014                                 INP_WUNLOCK(inp);
1015                                 error = 0;
1016 #else
1017                                 error = EOPNOTSUPP;
1018 #endif
1019                                 break;
1020                         default:
1021                                 break;
1022                         }
1023                 }
1024                 return (error);
1025         }
1026
1027         switch (sopt->sopt_dir) {
1028         case SOPT_SET:
1029                 switch (sopt->sopt_name) {
1030                 case IP_OPTIONS:
1031 #ifdef notyet
1032                 case IP_RETOPTS:
1033 #endif
1034                 {
1035                         struct mbuf *m;
1036                         if (sopt->sopt_valsize > MLEN) {
1037                                 error = EMSGSIZE;
1038                                 break;
1039                         }
1040                         m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1041                         if (m == NULL) {
1042                                 error = ENOBUFS;
1043                                 break;
1044                         }
1045                         m->m_len = sopt->sopt_valsize;
1046                         error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1047                                             m->m_len);
1048                         if (error) {
1049                                 m_free(m);
1050                                 break;
1051                         }
1052                         INP_WLOCK(inp);
1053                         error = ip_pcbopts(inp, sopt->sopt_name, m);
1054                         INP_WUNLOCK(inp);
1055                         return (error);
1056                 }
1057
1058                 case IP_BINDANY:
1059                         if (sopt->sopt_td != NULL) {
1060                                 error = priv_check(sopt->sopt_td,
1061                                     PRIV_NETINET_BINDANY);
1062                                 if (error)
1063                                         break;
1064                         }
1065                         /* FALLTHROUGH */
1066                 case IP_BINDMULTI:
1067 #ifdef  RSS
1068                 case IP_RSS_LISTEN_BUCKET:
1069 #endif
1070                 case IP_TOS:
1071                 case IP_TTL:
1072                 case IP_MINTTL:
1073                 case IP_RECVOPTS:
1074                 case IP_RECVRETOPTS:
1075                 case IP_ORIGDSTADDR:
1076                 case IP_RECVDSTADDR:
1077                 case IP_RECVTTL:
1078                 case IP_RECVIF:
1079                 case IP_ONESBCAST:
1080                 case IP_DONTFRAG:
1081                 case IP_RECVTOS:
1082                 case IP_RECVFLOWID:
1083 #ifdef  RSS
1084                 case IP_RECVRSSBUCKETID:
1085 #endif
1086                         error = sooptcopyin(sopt, &optval, sizeof optval,
1087                                             sizeof optval);
1088                         if (error)
1089                                 break;
1090
1091                         switch (sopt->sopt_name) {
1092                         case IP_TOS:
1093                                 inp->inp_ip_tos = optval;
1094                                 break;
1095
1096                         case IP_TTL:
1097                                 inp->inp_ip_ttl = optval;
1098                                 break;
1099
1100                         case IP_MINTTL:
1101                                 if (optval >= 0 && optval <= MAXTTL)
1102                                         inp->inp_ip_minttl = optval;
1103                                 else
1104                                         error = EINVAL;
1105                                 break;
1106
1107 #define OPTSET(bit) do {                                                \
1108         INP_WLOCK(inp);                                                 \
1109         if (optval)                                                     \
1110                 inp->inp_flags |= bit;                                  \
1111         else                                                            \
1112                 inp->inp_flags &= ~bit;                                 \
1113         INP_WUNLOCK(inp);                                               \
1114 } while (0)
1115
1116 #define OPTSET2(bit, val) do {                                          \
1117         INP_WLOCK(inp);                                                 \
1118         if (val)                                                        \
1119                 inp->inp_flags2 |= bit;                                 \
1120         else                                                            \
1121                 inp->inp_flags2 &= ~bit;                                \
1122         INP_WUNLOCK(inp);                                               \
1123 } while (0)
1124
1125                         case IP_RECVOPTS:
1126                                 OPTSET(INP_RECVOPTS);
1127                                 break;
1128
1129                         case IP_RECVRETOPTS:
1130                                 OPTSET(INP_RECVRETOPTS);
1131                                 break;
1132
1133                         case IP_RECVDSTADDR:
1134                                 OPTSET(INP_RECVDSTADDR);
1135                                 break;
1136
1137                         case IP_ORIGDSTADDR:
1138                                 OPTSET2(INP_ORIGDSTADDR, optval);
1139                                 break;
1140
1141                         case IP_RECVTTL:
1142                                 OPTSET(INP_RECVTTL);
1143                                 break;
1144
1145                         case IP_RECVIF:
1146                                 OPTSET(INP_RECVIF);
1147                                 break;
1148
1149                         case IP_ONESBCAST:
1150                                 OPTSET(INP_ONESBCAST);
1151                                 break;
1152                         case IP_DONTFRAG:
1153                                 OPTSET(INP_DONTFRAG);
1154                                 break;
1155                         case IP_BINDANY:
1156                                 OPTSET(INP_BINDANY);
1157                                 break;
1158                         case IP_RECVTOS:
1159                                 OPTSET(INP_RECVTOS);
1160                                 break;
1161                         case IP_BINDMULTI:
1162                                 OPTSET2(INP_BINDMULTI, optval);
1163                                 break;
1164                         case IP_RECVFLOWID:
1165                                 OPTSET2(INP_RECVFLOWID, optval);
1166                                 break;
1167 #ifdef  RSS
1168                         case IP_RSS_LISTEN_BUCKET:
1169                                 if ((optval >= 0) &&
1170                                     (optval < rss_getnumbuckets())) {
1171                                         inp->inp_rss_listen_bucket = optval;
1172                                         OPTSET2(INP_RSS_BUCKET_SET, 1);
1173                                 } else {
1174                                         error = EINVAL;
1175                                 }
1176                                 break;
1177                         case IP_RECVRSSBUCKETID:
1178                                 OPTSET2(INP_RECVRSSBUCKETID, optval);
1179                                 break;
1180 #endif
1181                         }
1182                         break;
1183 #undef OPTSET
1184 #undef OPTSET2
1185
1186                 /*
1187                  * Multicast socket options are processed by the in_mcast
1188                  * module.
1189                  */
1190                 case IP_MULTICAST_IF:
1191                 case IP_MULTICAST_VIF:
1192                 case IP_MULTICAST_TTL:
1193                 case IP_MULTICAST_LOOP:
1194                 case IP_ADD_MEMBERSHIP:
1195                 case IP_DROP_MEMBERSHIP:
1196                 case IP_ADD_SOURCE_MEMBERSHIP:
1197                 case IP_DROP_SOURCE_MEMBERSHIP:
1198                 case IP_BLOCK_SOURCE:
1199                 case IP_UNBLOCK_SOURCE:
1200                 case IP_MSFILTER:
1201                 case MCAST_JOIN_GROUP:
1202                 case MCAST_LEAVE_GROUP:
1203                 case MCAST_JOIN_SOURCE_GROUP:
1204                 case MCAST_LEAVE_SOURCE_GROUP:
1205                 case MCAST_BLOCK_SOURCE:
1206                 case MCAST_UNBLOCK_SOURCE:
1207                         error = inp_setmoptions(inp, sopt);
1208                         break;
1209
1210                 case IP_PORTRANGE:
1211                         error = sooptcopyin(sopt, &optval, sizeof optval,
1212                                             sizeof optval);
1213                         if (error)
1214                                 break;
1215
1216                         INP_WLOCK(inp);
1217                         switch (optval) {
1218                         case IP_PORTRANGE_DEFAULT:
1219                                 inp->inp_flags &= ~(INP_LOWPORT);
1220                                 inp->inp_flags &= ~(INP_HIGHPORT);
1221                                 break;
1222
1223                         case IP_PORTRANGE_HIGH:
1224                                 inp->inp_flags &= ~(INP_LOWPORT);
1225                                 inp->inp_flags |= INP_HIGHPORT;
1226                                 break;
1227
1228                         case IP_PORTRANGE_LOW:
1229                                 inp->inp_flags &= ~(INP_HIGHPORT);
1230                                 inp->inp_flags |= INP_LOWPORT;
1231                                 break;
1232
1233                         default:
1234                                 error = EINVAL;
1235                                 break;
1236                         }
1237                         INP_WUNLOCK(inp);
1238                         break;
1239
1240 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1241                 case IP_IPSEC_POLICY:
1242                         if (IPSEC_ENABLED(ipv4)) {
1243                                 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1244                                 break;
1245                         }
1246                         /* FALLTHROUGH */
1247 #endif /* IPSEC */
1248
1249                 default:
1250                         error = ENOPROTOOPT;
1251                         break;
1252                 }
1253                 break;
1254
1255         case SOPT_GET:
1256                 switch (sopt->sopt_name) {
1257                 case IP_OPTIONS:
1258                 case IP_RETOPTS:
1259                         INP_RLOCK(inp);
1260                         if (inp->inp_options) {
1261                                 struct mbuf *options;
1262
1263                                 options = m_dup(inp->inp_options, M_NOWAIT);
1264                                 INP_RUNLOCK(inp);
1265                                 if (options != NULL) {
1266                                         error = sooptcopyout(sopt,
1267                                                              mtod(options, char *),
1268                                                              options->m_len);
1269                                         m_freem(options);
1270                                 } else
1271                                         error = ENOMEM;
1272                         } else {
1273                                 INP_RUNLOCK(inp);
1274                                 sopt->sopt_valsize = 0;
1275                         }
1276                         break;
1277
1278                 case IP_TOS:
1279                 case IP_TTL:
1280                 case IP_MINTTL:
1281                 case IP_RECVOPTS:
1282                 case IP_RECVRETOPTS:
1283                 case IP_ORIGDSTADDR:
1284                 case IP_RECVDSTADDR:
1285                 case IP_RECVTTL:
1286                 case IP_RECVIF:
1287                 case IP_PORTRANGE:
1288                 case IP_ONESBCAST:
1289                 case IP_DONTFRAG:
1290                 case IP_BINDANY:
1291                 case IP_RECVTOS:
1292                 case IP_BINDMULTI:
1293                 case IP_FLOWID:
1294                 case IP_FLOWTYPE:
1295                 case IP_RECVFLOWID:
1296 #ifdef  RSS
1297                 case IP_RSSBUCKETID:
1298                 case IP_RECVRSSBUCKETID:
1299 #endif
1300                         switch (sopt->sopt_name) {
1301
1302                         case IP_TOS:
1303                                 optval = inp->inp_ip_tos;
1304                                 break;
1305
1306                         case IP_TTL:
1307                                 optval = inp->inp_ip_ttl;
1308                                 break;
1309
1310                         case IP_MINTTL:
1311                                 optval = inp->inp_ip_minttl;
1312                                 break;
1313
1314 #define OPTBIT(bit)     (inp->inp_flags & bit ? 1 : 0)
1315 #define OPTBIT2(bit)    (inp->inp_flags2 & bit ? 1 : 0)
1316
1317                         case IP_RECVOPTS:
1318                                 optval = OPTBIT(INP_RECVOPTS);
1319                                 break;
1320
1321                         case IP_RECVRETOPTS:
1322                                 optval = OPTBIT(INP_RECVRETOPTS);
1323                                 break;
1324
1325                         case IP_RECVDSTADDR:
1326                                 optval = OPTBIT(INP_RECVDSTADDR);
1327                                 break;
1328
1329                         case IP_ORIGDSTADDR:
1330                                 optval = OPTBIT2(INP_ORIGDSTADDR);
1331                                 break;
1332
1333                         case IP_RECVTTL:
1334                                 optval = OPTBIT(INP_RECVTTL);
1335                                 break;
1336
1337                         case IP_RECVIF:
1338                                 optval = OPTBIT(INP_RECVIF);
1339                                 break;
1340
1341                         case IP_PORTRANGE:
1342                                 if (inp->inp_flags & INP_HIGHPORT)
1343                                         optval = IP_PORTRANGE_HIGH;
1344                                 else if (inp->inp_flags & INP_LOWPORT)
1345                                         optval = IP_PORTRANGE_LOW;
1346                                 else
1347                                         optval = 0;
1348                                 break;
1349
1350                         case IP_ONESBCAST:
1351                                 optval = OPTBIT(INP_ONESBCAST);
1352                                 break;
1353                         case IP_DONTFRAG:
1354                                 optval = OPTBIT(INP_DONTFRAG);
1355                                 break;
1356                         case IP_BINDANY:
1357                                 optval = OPTBIT(INP_BINDANY);
1358                                 break;
1359                         case IP_RECVTOS:
1360                                 optval = OPTBIT(INP_RECVTOS);
1361                                 break;
1362                         case IP_FLOWID:
1363                                 optval = inp->inp_flowid;
1364                                 break;
1365                         case IP_FLOWTYPE:
1366                                 optval = inp->inp_flowtype;
1367                                 break;
1368                         case IP_RECVFLOWID:
1369                                 optval = OPTBIT2(INP_RECVFLOWID);
1370                                 break;
1371 #ifdef  RSS
1372                         case IP_RSSBUCKETID:
1373                                 retval = rss_hash2bucket(inp->inp_flowid,
1374                                     inp->inp_flowtype,
1375                                     &rss_bucket);
1376                                 if (retval == 0)
1377                                         optval = rss_bucket;
1378                                 else
1379                                         error = EINVAL;
1380                                 break;
1381                         case IP_RECVRSSBUCKETID:
1382                                 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1383                                 break;
1384 #endif
1385                         case IP_BINDMULTI:
1386                                 optval = OPTBIT2(INP_BINDMULTI);
1387                                 break;
1388                         }
1389                         error = sooptcopyout(sopt, &optval, sizeof optval);
1390                         break;
1391
1392                 /*
1393                  * Multicast socket options are processed by the in_mcast
1394                  * module.
1395                  */
1396                 case IP_MULTICAST_IF:
1397                 case IP_MULTICAST_VIF:
1398                 case IP_MULTICAST_TTL:
1399                 case IP_MULTICAST_LOOP:
1400                 case IP_MSFILTER:
1401                         error = inp_getmoptions(inp, sopt);
1402                         break;
1403
1404 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1405                 case IP_IPSEC_POLICY:
1406                         if (IPSEC_ENABLED(ipv4)) {
1407                                 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1408                                 break;
1409                         }
1410                         /* FALLTHROUGH */
1411 #endif /* IPSEC */
1412
1413                 default:
1414                         error = ENOPROTOOPT;
1415                         break;
1416                 }
1417                 break;
1418         }
1419         return (error);
1420 }
1421
1422 /*
1423  * Routine called from ip_output() to loop back a copy of an IP multicast
1424  * packet to the input queue of a specified interface.  Note that this
1425  * calls the output routine of the loopback "driver", but with an interface
1426  * pointer that might NOT be a loopback interface -- evil, but easier than
1427  * replicating that code here.
1428  */
1429 static void
1430 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1431 {
1432         struct ip *ip;
1433         struct mbuf *copym;
1434
1435         /*
1436          * Make a deep copy of the packet because we're going to
1437          * modify the pack in order to generate checksums.
1438          */
1439         copym = m_dup(m, M_NOWAIT);
1440         if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1441                 copym = m_pullup(copym, hlen);
1442         if (copym != NULL) {
1443                 /* If needed, compute the checksum and mark it as valid. */
1444                 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1445                         in_delayed_cksum(copym);
1446                         copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1447                         copym->m_pkthdr.csum_flags |=
1448                             CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1449                         copym->m_pkthdr.csum_data = 0xffff;
1450                 }
1451                 /*
1452                  * We don't bother to fragment if the IP length is greater
1453                  * than the interface's MTU.  Can this possibly matter?
1454                  */
1455                 ip = mtod(copym, struct ip *);
1456                 ip->ip_sum = 0;
1457                 ip->ip_sum = in_cksum(copym, hlen);
1458                 if_simloop(ifp, copym, AF_INET, 0);
1459         }
1460 }