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_ipsec.h"
  39 #include "opt_mbuf_stress_test.h"
  40 #include "opt_mpath.h"
  41 #include "opt_ratelimit.h"
  42 #include "opt_route.h"
  43 #include "opt_rss.h"
  44 #include "opt_sctp.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_fib.h>
  76 #include <netinet/in_kdtrace.h>
  77 #include <netinet/in_systm.h>
  78 #include <netinet/ip.h>
  79 #include <netinet/in_pcb.h>
  80 #include <netinet/in_rss.h>
  81 #include <netinet/in_var.h>
  82 #include <netinet/ip_var.h>
  83 #include <netinet/ip_options.h>
  84
  85 #include <netinet/udp.h>
  86 #include <netinet/udp_var.h>
  87
  88 #ifdef SCTP
  89 #include <netinet/sctp.h>
  90 #include <netinet/sctp_crc32.h>
  91 #endif
  92
  93 #include <netipsec/ipsec_support.h>
  94
  95 #include <machine/in_cksum.h>
  96
  97 #include <security/mac/mac_framework.h>
  98
  99 #ifdef MBUF_STRESS_TEST
 100 static int mbuf_frag_size = 0;
 101 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
 102         &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
 103 #endif
 104
 105 static void     ip_mloopback(struct ifnet *, const struct mbuf *, int);
 106
 107
 108 extern int in_mcast_loop;
 109 extern  struct protosw inetsw[];
 110
 111 static inline int
 112 ip_output_pfil(struct mbuf **mp, struct ifnet *ifp, int flags,
 113     struct inpcb *inp, struct sockaddr_in *dst, int *fibnum, int *error)
 114 {
 115         struct m_tag *fwd_tag = NULL;
 116         struct mbuf *m;
 117         struct in_addr odst;
 118         struct ip *ip;
 119         int pflags = PFIL_OUT;
 120
 121         if (flags & IP_FORWARDING)
 122                 pflags |= PFIL_FWD;
 123
 124         m = *mp;
 125         ip = mtod(m, struct ip *);
 126
 127         /* Run through list of hooks for output packets. */
 128         odst.s_addr = ip->ip_dst.s_addr;
 129         switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
 130         case PFIL_DROPPED:
 131                 *error = EPERM;
 132                 /* FALLTHROUGH */
 133         case PFIL_CONSUMED:
 134                 return 1; /* Finished */
 135         case PFIL_PASS:
 136                 *error = 0;
 137         }
 138         m = *mp;
 139         ip = mtod(m, struct ip *);
 140
 141         /* See if destination IP address was changed by packet filter. */
 142         if (odst.s_addr != ip->ip_dst.s_addr) {
 143                 m->m_flags |= M_SKIP_FIREWALL;
 144                 /* If destination is now ourself drop to ip_input(). */
 145                 if (in_localip(ip->ip_dst)) {
 146                         m->m_flags |= M_FASTFWD_OURS;
 147                         if (m->m_pkthdr.rcvif == NULL)
 148                                 m->m_pkthdr.rcvif = V_loif;
 149                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 150                                 m->m_pkthdr.csum_flags |=
 151                                         CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 152                                 m->m_pkthdr.csum_data = 0xffff;
 153                         }
 154                         m->m_pkthdr.csum_flags |=
 155                                 CSUM_IP_CHECKED | CSUM_IP_VALID;
 156 #ifdef SCTP
 157                         if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 158                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 159 #endif
 160                         *error = netisr_queue(NETISR_IP, m);
 161                         return 1; /* Finished */
 162                 }
 163
 164                 bzero(dst, sizeof(*dst));
 165                 dst->sin_family = AF_INET;
 166                 dst->sin_len = sizeof(*dst);
 167                 dst->sin_addr = ip->ip_dst;
 168
 169                 return -1; /* Reloop */
 170         }
 171         /* See if fib was changed by packet filter. */
 172         if ((*fibnum) != M_GETFIB(m)) {
 173                 m->m_flags |= M_SKIP_FIREWALL;
 174                 *fibnum = M_GETFIB(m);
 175                 return -1; /* Reloop for FIB change */
 176         }
 177
 178         /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
 179         if (m->m_flags & M_FASTFWD_OURS) {
 180                 if (m->m_pkthdr.rcvif == NULL)
 181                         m->m_pkthdr.rcvif = V_loif;
 182                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 183                         m->m_pkthdr.csum_flags |=
 184                                 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 185                         m->m_pkthdr.csum_data = 0xffff;
 186                 }
 187 #ifdef SCTP
 188                 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 189                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 190 #endif
 191                 m->m_pkthdr.csum_flags |=
 192                         CSUM_IP_CHECKED | CSUM_IP_VALID;
 193
 194                 *error = netisr_queue(NETISR_IP, m);
 195                 return 1; /* Finished */
 196         }
 197         /* Or forward to some other address? */
 198         if ((m->m_flags & M_IP_NEXTHOP) &&
 199             ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
 200                 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
 201                 m->m_flags |= M_SKIP_FIREWALL;
 202                 m->m_flags &= ~M_IP_NEXTHOP;
 203                 m_tag_delete(m, fwd_tag);
 204
 205                 return -1; /* Reloop for CHANGE of dst */
 206         }
 207
 208         return 0;
 209 }
 210
 211 static int
 212 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
 213     const struct sockaddr_in *gw, struct route *ro)
 214 {
 215         struct m_snd_tag *mst;
 216         int error;
 217
 218         MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
 219         mst = NULL;
 220
 221 #ifdef RATELIMIT
 222         if (inp != NULL) {
 223                 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
 224                     (inp->inp_snd_tag != NULL &&
 225                     inp->inp_snd_tag->ifp != ifp))
 226                         in_pcboutput_txrtlmt(inp, ifp, m);
 227
 228                 if (inp->inp_snd_tag != NULL)
 229                         mst = inp->inp_snd_tag;
 230         }
 231 #endif
 232         if (mst != NULL) {
 233                 KASSERT(m->m_pkthdr.rcvif == NULL,
 234                     ("trying to add a send tag to a forwarded packet"));
 235                 if (mst->ifp != ifp) {
 236                         error = EAGAIN;
 237                         goto done;
 238                 }
 239
 240                 /* stamp send tag on mbuf */
 241                 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
 242                 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
 243         }
 244
 245         error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
 246
 247 done:
 248         /* Check for route change invalidating send tags. */
 249 #ifdef RATELIMIT
 250         if (error == EAGAIN)
 251                 in_pcboutput_eagain(inp);
 252 #endif
 253         return (error);
 254 }
 255
 256 /*
 257  * IP output.  The packet in mbuf chain m contains a skeletal IP
 258  * header (with len, off, ttl, proto, tos, src, dst).
 259  * The mbuf chain containing the packet will be freed.
 260  * The mbuf opt, if present, will not be freed.
 261  * If route ro is present and has ro_rt initialized, route lookup would be
 262  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 263  * then result of route lookup is stored in ro->ro_rt.
 264  *
 265  * In the IP forwarding case, the packet will arrive with options already
 266  * inserted, so must have a NULL opt pointer.
 267  */
 268 int
 269 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
 270     struct ip_moptions *imo, struct inpcb *inp)
 271 {
 272         struct rm_priotracker in_ifa_tracker;
 273         struct epoch_tracker et;
 274         struct ip *ip;
 275         struct ifnet *ifp = NULL;       /* keep compiler happy */
 276         struct mbuf *m0;
 277         int hlen = sizeof (struct ip);
 278         int mtu;
 279         int error = 0;
 280         struct sockaddr_in *dst, sin;
 281         const struct sockaddr_in *gw;
 282         struct in_ifaddr *ia;
 283         struct in_addr src;
 284         int isbroadcast;
 285         uint16_t ip_len, ip_off;
 286         uint32_t fibnum;
 287 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 288         int no_route_but_check_spd = 0;
 289 #endif
 290
 291         M_ASSERTPKTHDR(m);
 292
 293         if (inp != NULL) {
 294                 INP_LOCK_ASSERT(inp);
 295                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 296                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 297                         m->m_pkthdr.flowid = inp->inp_flowid;
 298                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 299                 }
 300 #ifdef NUMA
 301                 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
 302 #endif
 303         }
 304
 305         if (opt) {
 306                 int len = 0;
 307                 m = ip_insertoptions(m, opt, &len);
 308                 if (len != 0)
 309                         hlen = len; /* ip->ip_hl is updated above */
 310         }
 311         ip = mtod(m, struct ip *);
 312         ip_len = ntohs(ip->ip_len);
 313         ip_off = ntohs(ip->ip_off);
 314
 315         if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
 316                 ip->ip_v = IPVERSION;
 317                 ip->ip_hl = hlen >> 2;
 318                 ip_fillid(ip);
 319         } else {
 320                 /* Header already set, fetch hlen from there */
 321                 hlen = ip->ip_hl << 2;
 322         }
 323         if ((flags & IP_FORWARDING) == 0)
 324                 IPSTAT_INC(ips_localout);
 325
 326         /*
 327          * dst/gw handling:
 328          *
 329          * gw is readonly but can point either to dst OR rt_gateway,
 330          * therefore we need restore gw if we're redoing lookup.
 331          */
 332         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 333         if (ro != NULL)
 334                 dst = (struct sockaddr_in *)&ro->ro_dst;
 335         else
 336                 dst = &sin;
 337         if (ro == NULL || ro->ro_rt == NULL) {
 338                 bzero(dst, sizeof(*dst));
 339                 dst->sin_family = AF_INET;
 340                 dst->sin_len = sizeof(*dst);
 341                 dst->sin_addr = ip->ip_dst;
 342         }
 343         gw = dst;
 344         NET_EPOCH_ENTER(et);
 345 again:
 346         /*
 347          * Validate route against routing table additions;
 348          * a better/more specific route might have been added.
 349          */
 350         if (inp != NULL && ro != NULL && ro->ro_rt != NULL)
 351                 RT_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
 352         /*
 353          * If there is a cached route,
 354          * check that it is to the same destination
 355          * and is still up.  If not, free it and try again.
 356          * The address family should also be checked in case of sharing the
 357          * cache with IPv6.
 358          * Also check whether routing cache needs invalidation.
 359          */
 360         if (ro != NULL && ro->ro_rt != NULL &&
 361             ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
 362             ro->ro_rt->rt_ifp == NULL || !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
 363             dst->sin_family != AF_INET ||
 364             dst->sin_addr.s_addr != ip->ip_dst.s_addr))
 365                 RO_INVALIDATE_CACHE(ro);
 366         ia = NULL;
 367         /*
 368          * If routing to interface only, short circuit routing lookup.
 369          * The use of an all-ones broadcast address implies this; an
 370          * interface is specified by the broadcast address of an interface,
 371          * or the destination address of a ptp interface.
 372          */
 373         if (flags & IP_SENDONES) {
 374                 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
 375                                                       M_GETFIB(m)))) == NULL &&
 376                     (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 377                                                     M_GETFIB(m)))) == NULL) {
 378                         IPSTAT_INC(ips_noroute);
 379                         error = ENETUNREACH;
 380                         goto bad;
 381                 }
 382                 ip->ip_dst.s_addr = INADDR_BROADCAST;
 383                 dst->sin_addr = ip->ip_dst;
 384                 ifp = ia->ia_ifp;
 385                 mtu = ifp->if_mtu;
 386                 ip->ip_ttl = 1;
 387                 isbroadcast = 1;
 388                 src = IA_SIN(ia)->sin_addr;
 389         } else if (flags & IP_ROUTETOIF) {
 390                 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 391                                                     M_GETFIB(m)))) == NULL &&
 392                     (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
 393                                                 M_GETFIB(m)))) == NULL) {
 394                         IPSTAT_INC(ips_noroute);
 395                         error = ENETUNREACH;
 396                         goto bad;
 397                 }
 398                 ifp = ia->ia_ifp;
 399                 mtu = ifp->if_mtu;
 400                 ip->ip_ttl = 1;
 401                 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
 402                     in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
 403                 src = IA_SIN(ia)->sin_addr;
 404         } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
 405             imo != NULL && imo->imo_multicast_ifp != NULL) {
 406                 /*
 407                  * Bypass the normal routing lookup for multicast
 408                  * packets if the interface is specified.
 409                  */
 410                 ifp = imo->imo_multicast_ifp;
 411                 mtu = ifp->if_mtu;
 412                 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
 413                 isbroadcast = 0;        /* fool gcc */
 414                 /* Interface may have no addresses. */
 415                 if (ia != NULL)
 416                         src = IA_SIN(ia)->sin_addr;
 417                 else
 418                         src.s_addr = INADDR_ANY;
 419         } else if (ro != NULL) {
 420                 if (ro->ro_rt == NULL) {
 421                         /*
 422                          * We want to do any cloning requested by the link
 423                          * layer, as this is probably required in all cases
 424                          * for correct operation (as it is for ARP).
 425                          */
 426 #ifdef RADIX_MPATH
 427                         rtalloc_mpath_fib(ro,
 428                             ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
 429                             fibnum);
 430 #else
 431                         in_rtalloc_ign(ro, 0, fibnum);
 432 #endif
 433                         if (ro->ro_rt == NULL ||
 434                             (ro->ro_rt->rt_flags & RTF_UP) == 0 ||
 435                             ro->ro_rt->rt_ifp == NULL ||
 436                             !RT_LINK_IS_UP(ro->ro_rt->rt_ifp)) {
 437 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 438                                 /*
 439                                  * There is no route for this packet, but it is
 440                                  * possible that a matching SPD entry exists.
 441                                  */
 442                                 no_route_but_check_spd = 1;
 443                                 mtu = 0; /* Silence GCC warning. */
 444                                 goto sendit;
 445 #endif
 446                                 IPSTAT_INC(ips_noroute);
 447                                 error = EHOSTUNREACH;
 448                                 goto bad;
 449                         }
 450                 }
 451                 ia = ifatoia(ro->ro_rt->rt_ifa);
 452                 ifp = ro->ro_rt->rt_ifp;
 453                 counter_u64_add(ro->ro_rt->rt_pksent, 1);
 454                 rt_update_ro_flags(ro);
 455                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
 456                         gw = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
 457                 if (ro->ro_rt->rt_flags & RTF_HOST)
 458                         isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
 459                 else if (ifp->if_flags & IFF_BROADCAST)
 460                         isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
 461                 else
 462                         isbroadcast = 0;
 463                 if (ro->ro_rt->rt_flags & RTF_HOST)
 464                         mtu = ro->ro_rt->rt_mtu;
 465                 else
 466                         mtu = ifp->if_mtu;
 467                 src = IA_SIN(ia)->sin_addr;
 468         } else {
 469                 struct nhop4_extended nh;
 470
 471                 bzero(&nh, sizeof(nh));
 472                 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh) !=
 473                     0) {
 474 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 475                         /*
 476                          * There is no route for this packet, but it is
 477                          * possible that a matching SPD entry exists.
 478                          */
 479                         no_route_but_check_spd = 1;
 480                         mtu = 0; /* Silence GCC warning. */
 481                         goto sendit;
 482 #endif
 483                         IPSTAT_INC(ips_noroute);
 484                         error = EHOSTUNREACH;
 485                         goto bad;
 486                 }
 487                 ifp = nh.nh_ifp;
 488                 mtu = nh.nh_mtu;
 489                 /*
 490                  * We are rewriting here dst to be gw actually, contradicting
 491                  * comment at the beginning of the function. However, in this
 492                  * case we are always dealing with on stack dst.
 493                  * In case if pfil(9) sends us back to beginning of the
 494                  * function, the dst would be rewritten by ip_output_pfil().
 495                  */
 496                 MPASS(dst == &sin);
 497                 dst->sin_addr = nh.nh_addr;
 498                 ia = nh.nh_ia;
 499                 src = nh.nh_src;
 500                 isbroadcast = (((nh.nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
 501                     (NHF_HOST | NHF_BROADCAST)) ||
 502                     ((ifp->if_flags & IFF_BROADCAST) &&
 503                     in_ifaddr_broadcast(dst->sin_addr, ia)));
 504         }
 505
 506         /* Catch a possible divide by zero later. */
 507         KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (rt_flags=0x%08x) ifp=%p",
 508             __func__, mtu, ro,
 509             (ro != NULL && ro->ro_rt != NULL) ? ro->ro_rt->rt_flags : 0, ifp));
 510
 511         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
 512                 m->m_flags |= M_MCAST;
 513                 /*
 514                  * IP destination address is multicast.  Make sure "gw"
 515                  * still points to the address in "ro".  (It may have been
 516                  * changed to point to a gateway address, above.)
 517                  */
 518                 gw = dst;
 519                 /*
 520                  * See if the caller provided any multicast options
 521                  */
 522                 if (imo != NULL) {
 523                         ip->ip_ttl = imo->imo_multicast_ttl;
 524                         if (imo->imo_multicast_vif != -1)
 525                                 ip->ip_src.s_addr =
 526                                     ip_mcast_src ?
 527                                     ip_mcast_src(imo->imo_multicast_vif) :
 528                                     INADDR_ANY;
 529                 } else
 530                         ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
 531                 /*
 532                  * Confirm that the outgoing interface supports multicast.
 533                  */
 534                 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
 535                         if ((ifp->if_flags & IFF_MULTICAST) == 0) {
 536                                 IPSTAT_INC(ips_noroute);
 537                                 error = ENETUNREACH;
 538                                 goto bad;
 539                         }
 540                 }
 541                 /*
 542                  * If source address not specified yet, use address
 543                  * of outgoing interface.
 544                  */
 545                 if (ip->ip_src.s_addr == INADDR_ANY)
 546                         ip->ip_src = src;
 547
 548                 if ((imo == NULL && in_mcast_loop) ||
 549                     (imo && imo->imo_multicast_loop)) {
 550                         /*
 551                          * Loop back multicast datagram if not expressly
 552                          * forbidden to do so, even if we are not a member
 553                          * of the group; ip_input() will filter it later,
 554                          * thus deferring a hash lookup and mutex acquisition
 555                          * at the expense of a cheap copy using m_copym().
 556                          */
 557                         ip_mloopback(ifp, m, hlen);
 558                 } else {
 559                         /*
 560                          * If we are acting as a multicast router, perform
 561                          * multicast forwarding as if the packet had just
 562                          * arrived on the interface to which we are about
 563                          * to send.  The multicast forwarding function
 564                          * recursively calls this function, using the
 565                          * IP_FORWARDING flag to prevent infinite recursion.
 566                          *
 567                          * Multicasts that are looped back by ip_mloopback(),
 568                          * above, will be forwarded by the ip_input() routine,
 569                          * if necessary.
 570                          */
 571                         if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
 572                                 /*
 573                                  * If rsvp daemon is not running, do not
 574                                  * set ip_moptions. This ensures that the packet
 575                                  * is multicast and not just sent down one link
 576                                  * as prescribed by rsvpd.
 577                                  */
 578                                 if (!V_rsvp_on)
 579                                         imo = NULL;
 580                                 if (ip_mforward &&
 581                                     ip_mforward(ip, ifp, m, imo) != 0) {
 582                                         m_freem(m);
 583                                         goto done;
 584                                 }
 585                         }
 586                 }
 587
 588                 /*
 589                  * Multicasts with a time-to-live of zero may be looped-
 590                  * back, above, but must not be transmitted on a network.
 591                  * Also, multicasts addressed to the loopback interface
 592                  * are not sent -- the above call to ip_mloopback() will
 593                  * loop back a copy. ip_input() will drop the copy if
 594                  * this host does not belong to the destination group on
 595                  * the loopback interface.
 596                  */
 597                 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
 598                         m_freem(m);
 599                         goto done;
 600                 }
 601
 602                 goto sendit;
 603         }
 604
 605         /*
 606          * If the source address is not specified yet, use the address
 607          * of the outoing interface.
 608          */
 609         if (ip->ip_src.s_addr == INADDR_ANY)
 610                 ip->ip_src = src;
 611
 612         /*
 613          * Look for broadcast address and
 614          * verify user is allowed to send
 615          * such a packet.
 616          */
 617         if (isbroadcast) {
 618                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 619                         error = EADDRNOTAVAIL;
 620                         goto bad;
 621                 }
 622                 if ((flags & IP_ALLOWBROADCAST) == 0) {
 623                         error = EACCES;
 624                         goto bad;
 625                 }
 626                 /* don't allow broadcast messages to be fragmented */
 627                 if (ip_len > mtu) {
 628                         error = EMSGSIZE;
 629                         goto bad;
 630                 }
 631                 m->m_flags |= M_BCAST;
 632         } else {
 633                 m->m_flags &= ~M_BCAST;
 634         }
 635
 636 sendit:
 637 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 638         if (IPSEC_ENABLED(ipv4)) {
 639                 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
 640                         if (error == EINPROGRESS)
 641                                 error = 0;
 642                         goto done;
 643                 }
 644         }
 645         /*
 646          * Check if there was a route for this packet; return error if not.
 647          */
 648         if (no_route_but_check_spd) {
 649                 IPSTAT_INC(ips_noroute);
 650                 error = EHOSTUNREACH;
 651                 goto bad;
 652         }
 653         /* Update variables that are affected by ipsec4_output(). */
 654         ip = mtod(m, struct ip *);
 655         hlen = ip->ip_hl << 2;
 656 #endif /* IPSEC */
 657
 658         /* Jump over all PFIL processing if hooks are not active. */
 659         if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
 660                 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
 661                     &error)) {
 662                 case 1: /* Finished */
 663                         goto done;
 664
 665                 case 0: /* Continue normally */
 666                         ip = mtod(m, struct ip *);
 667                         break;
 668
 669                 case -1: /* Need to try again */
 670                         /* Reset everything for a new round */
 671                         if (ro != NULL) {
 672                                 RO_RTFREE(ro);
 673                                 ro->ro_prepend = NULL;
 674                         }
 675                         gw = dst;
 676                         ip = mtod(m, struct ip *);
 677                         goto again;
 678
 679                 }
 680         }
 681
 682         /* IN_LOOPBACK must not appear on the wire - RFC1122. */
 683         if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
 684             IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
 685                 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
 686                         IPSTAT_INC(ips_badaddr);
 687                         error = EADDRNOTAVAIL;
 688                         goto bad;
 689                 }
 690         }
 691
 692         m->m_pkthdr.csum_flags |= CSUM_IP;
 693         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
 694                 m = mb_unmapped_to_ext(m);
 695                 if (m == NULL) {
 696                         IPSTAT_INC(ips_odropped);
 697                         error = ENOBUFS;
 698                         goto bad;
 699                 }
 700                 in_delayed_cksum(m);
 701                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 702         } else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
 703                 m = mb_unmapped_to_ext(m);
 704                 if (m == NULL) {
 705                         IPSTAT_INC(ips_odropped);
 706                         error = ENOBUFS;
 707                         goto bad;
 708                 }
 709         }
 710 #ifdef SCTP
 711         if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
 712                 m = mb_unmapped_to_ext(m);
 713                 if (m == NULL) {
 714                         IPSTAT_INC(ips_odropped);
 715                         error = ENOBUFS;
 716                         goto bad;
 717                 }
 718                 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
 719                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 720         }
 721 #endif
 722
 723         /*
 724          * If small enough for interface, or the interface will take
 725          * care of the fragmentation for us, we can just send directly.
 726          */
 727         if (ip_len <= mtu ||
 728             (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) {
 729                 ip->ip_sum = 0;
 730                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
 731                         ip->ip_sum = in_cksum(m, hlen);
 732                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 733                 }
 734
 735                 /*
 736                  * Record statistics for this interface address.
 737                  * With CSUM_TSO the byte/packet count will be slightly
 738                  * incorrect because we count the IP+TCP headers only
 739                  * once instead of for every generated packet.
 740                  */
 741                 if (!(flags & IP_FORWARDING) && ia) {
 742                         if (m->m_pkthdr.csum_flags & CSUM_TSO)
 743                                 counter_u64_add(ia->ia_ifa.ifa_opackets,
 744                                     m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
 745                         else
 746                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 747
 748                         counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
 749                 }
 750 #ifdef MBUF_STRESS_TEST
 751                 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
 752                         m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
 753 #endif
 754                 /*
 755                  * Reset layer specific mbuf flags
 756                  * to avoid confusing lower layers.
 757                  */
 758                 m_clrprotoflags(m);
 759                 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
 760                 error = ip_output_send(inp, ifp, m, gw, ro);
 761                 goto done;
 762         }
 763
 764         /* Balk when DF bit is set or the interface didn't support TSO. */
 765         if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
 766                 error = EMSGSIZE;
 767                 IPSTAT_INC(ips_cantfrag);
 768                 goto bad;
 769         }
 770
 771         /*
 772          * Too large for interface; fragment if possible. If successful,
 773          * on return, m will point to a list of packets to be sent.
 774          */
 775         error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
 776         if (error)
 777                 goto bad;
 778         for (; m; m = m0) {
 779                 m0 = m->m_nextpkt;
 780                 m->m_nextpkt = 0;
 781                 if (error == 0) {
 782                         /* Record statistics for this interface address. */
 783                         if (ia != NULL) {
 784                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 785                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
 786                                     m->m_pkthdr.len);
 787                         }
 788                         /*
 789                          * Reset layer specific mbuf flags
 790                          * to avoid confusing upper layers.
 791                          */
 792                         m_clrprotoflags(m);
 793
 794                         IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
 795                             mtod(m, struct ip *), NULL);
 796                         error = ip_output_send(inp, ifp, m, gw, ro);
 797                 } else
 798                         m_freem(m);
 799         }
 800
 801         if (error == 0)
 802                 IPSTAT_INC(ips_fragmented);
 803
 804 done:
 805         NET_EPOCH_EXIT(et);
 806         return (error);
 807  bad:
 808         m_freem(m);
 809         goto done;
 810 }
 811
 812 /*
 813  * Create a chain of fragments which fit the given mtu. m_frag points to the
 814  * mbuf to be fragmented; on return it points to the chain with the fragments.
 815  * Return 0 if no error. If error, m_frag may contain a partially built
 816  * chain of fragments that should be freed by the caller.
 817  *
 818  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
 819  */
 820 int
 821 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
 822     u_long if_hwassist_flags)
 823 {
 824         int error = 0;
 825         int hlen = ip->ip_hl << 2;
 826         int len = (mtu - hlen) & ~7;    /* size of payload in each fragment */
 827         int off;
 828         struct mbuf *m0 = *m_frag;      /* the original packet          */
 829         int firstlen;
 830         struct mbuf **mnext;
 831         int nfrags;
 832         uint16_t ip_len, ip_off;
 833
 834         ip_len = ntohs(ip->ip_len);
 835         ip_off = ntohs(ip->ip_off);
 836
 837         if (ip_off & IP_DF) {   /* Fragmentation not allowed */
 838                 IPSTAT_INC(ips_cantfrag);
 839                 return EMSGSIZE;
 840         }
 841
 842         /*
 843          * Must be able to put at least 8 bytes per fragment.
 844          */
 845         if (len < 8)
 846                 return EMSGSIZE;
 847
 848         /*
 849          * If the interface will not calculate checksums on
 850          * fragmented packets, then do it here.
 851          */
 852         if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 853                 m0 = mb_unmapped_to_ext(m0);
 854                 if (m0 == NULL) {
 855                         error = ENOBUFS;
 856                         IPSTAT_INC(ips_odropped);
 857                         goto done;
 858                 }
 859                 in_delayed_cksum(m0);
 860                 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 861         }
 862 #ifdef SCTP
 863         if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
 864                 m0 = mb_unmapped_to_ext(m0);
 865                 if (m0 == NULL) {
 866                         error = ENOBUFS;
 867                         IPSTAT_INC(ips_odropped);
 868                         goto done;
 869                 }
 870                 sctp_delayed_cksum(m0, hlen);
 871                 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 872         }
 873 #endif
 874         if (len > PAGE_SIZE) {
 875                 /*
 876                  * Fragment large datagrams such that each segment
 877                  * contains a multiple of PAGE_SIZE amount of data,
 878                  * plus headers. This enables a receiver to perform
 879                  * page-flipping zero-copy optimizations.
 880                  *
 881                  * XXX When does this help given that sender and receiver
 882                  * could have different page sizes, and also mtu could
 883                  * be less than the receiver's page size ?
 884                  */
 885                 int newlen;
 886
 887                 off = MIN(mtu, m0->m_pkthdr.len);
 888
 889                 /*
 890                  * firstlen (off - hlen) must be aligned on an
 891                  * 8-byte boundary
 892                  */
 893                 if (off < hlen)
 894                         goto smart_frag_failure;
 895                 off = ((off - hlen) & ~7) + hlen;
 896                 newlen = (~PAGE_MASK) & mtu;
 897                 if ((newlen + sizeof (struct ip)) > mtu) {
 898                         /* we failed, go back the default */
 899 smart_frag_failure:
 900                         newlen = len;
 901                         off = hlen + len;
 902                 }
 903                 len = newlen;
 904
 905         } else {
 906                 off = hlen + len;
 907         }
 908
 909         firstlen = off - hlen;
 910         mnext = &m0->m_nextpkt;         /* pointer to next packet */
 911
 912         /*
 913          * Loop through length of segment after first fragment,
 914          * make new header and copy data of each part and link onto chain.
 915          * Here, m0 is the original packet, m is the fragment being created.
 916          * The fragments are linked off the m_nextpkt of the original
 917          * packet, which after processing serves as the first fragment.
 918          */
 919         for (nfrags = 1; off < ip_len; off += len, nfrags++) {
 920                 struct ip *mhip;        /* ip header on the fragment */
 921                 struct mbuf *m;
 922                 int mhlen = sizeof (struct ip);
 923
 924                 m = m_gethdr(M_NOWAIT, MT_DATA);
 925                 if (m == NULL) {
 926                         error = ENOBUFS;
 927                         IPSTAT_INC(ips_odropped);
 928                         goto done;
 929                 }
 930                 /*
 931                  * Make sure the complete packet header gets copied
 932                  * from the originating mbuf to the newly created
 933                  * mbuf. This also ensures that existing firewall
 934                  * classification(s), VLAN tags and so on get copied
 935                  * to the resulting fragmented packet(s):
 936                  */
 937                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 938                         m_free(m);
 939                         error = ENOBUFS;
 940                         IPSTAT_INC(ips_odropped);
 941                         goto done;
 942                 }
 943                 /*
 944                  * In the first mbuf, leave room for the link header, then
 945                  * copy the original IP header including options. The payload
 946                  * goes into an additional mbuf chain returned by m_copym().
 947                  */
 948                 m->m_data += max_linkhdr;
 949                 mhip = mtod(m, struct ip *);
 950                 *mhip = *ip;
 951                 if (hlen > sizeof (struct ip)) {
 952                         mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
 953                         mhip->ip_v = IPVERSION;
 954                         mhip->ip_hl = mhlen >> 2;
 955                 }
 956                 m->m_len = mhlen;
 957                 /* XXX do we need to add ip_off below ? */
 958                 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
 959                 if (off + len >= ip_len)
 960                         len = ip_len - off;
 961                 else
 962                         mhip->ip_off |= IP_MF;
 963                 mhip->ip_len = htons((u_short)(len + mhlen));
 964                 m->m_next = m_copym(m0, off, len, M_NOWAIT);
 965                 if (m->m_next == NULL) {        /* copy failed */
 966                         m_free(m);
 967                         error = ENOBUFS;        /* ??? */
 968                         IPSTAT_INC(ips_odropped);
 969                         goto done;
 970                 }
 971                 m->m_pkthdr.len = mhlen + len;
 972 #ifdef MAC
 973                 mac_netinet_fragment(m0, m);
 974 #endif
 975                 mhip->ip_off = htons(mhip->ip_off);
 976                 mhip->ip_sum = 0;
 977                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 978                         mhip->ip_sum = in_cksum(m, mhlen);
 979                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 980                 }
 981                 *mnext = m;
 982                 mnext = &m->m_nextpkt;
 983         }
 984         IPSTAT_ADD(ips_ofragments, nfrags);
 985
 986         /*
 987          * Update first fragment by trimming what's been copied out
 988          * and updating header.
 989          */
 990         m_adj(m0, hlen + firstlen - ip_len);
 991         m0->m_pkthdr.len = hlen + firstlen;
 992         ip->ip_len = htons((u_short)m0->m_pkthdr.len);
 993         ip->ip_off = htons(ip_off | IP_MF);
 994         ip->ip_sum = 0;
 995         if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 996                 ip->ip_sum = in_cksum(m0, hlen);
 997                 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
 998         }
 999
1000 done:
1001         *m_frag = m0;
1002         return error;
1003 }
1004
1005 void
1006 in_delayed_cksum(struct mbuf *m)
1007 {
1008         struct ip *ip;
1009         struct udphdr *uh;
1010         uint16_t cklen, csum, offset;
1011
1012         ip = mtod(m, struct ip *);
1013         offset = ip->ip_hl << 2 ;
1014
1015         if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1016                 /* if udp header is not in the first mbuf copy udplen */
1017                 if (offset + sizeof(struct udphdr) > m->m_len) {
1018                         m_copydata(m, offset + offsetof(struct udphdr,
1019                             uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1020                         cklen = ntohs(cklen);
1021                 } else {
1022                         uh = (struct udphdr *)mtodo(m, offset);
1023                         cklen = ntohs(uh->uh_ulen);
1024                 }
1025                 csum = in_cksum_skip(m, cklen + offset, offset);
1026                 if (csum == 0)
1027                         csum = 0xffff;
1028         } else {
1029                 cklen = ntohs(ip->ip_len);
1030                 csum = in_cksum_skip(m, cklen, offset);
1031         }
1032         offset += m->m_pkthdr.csum_data;        /* checksum offset */
1033
1034         if (offset + sizeof(csum) > m->m_len)
1035                 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1036         else
1037                 *(u_short *)mtodo(m, offset) = csum;
1038 }
1039
1040 /*
1041  * IP socket option processing.
1042  */
1043 int
1044 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1045 {
1046         struct  inpcb *inp = sotoinpcb(so);
1047         int     error, optval;
1048 #ifdef  RSS
1049         uint32_t rss_bucket;
1050         int retval;
1051 #endif
1052
1053         error = optval = 0;
1054         if (sopt->sopt_level != IPPROTO_IP) {
1055                 error = EINVAL;
1056
1057                 if (sopt->sopt_level == SOL_SOCKET &&
1058                     sopt->sopt_dir == SOPT_SET) {
1059                         switch (sopt->sopt_name) {
1060                         case SO_REUSEADDR:
1061                                 INP_WLOCK(inp);
1062                                 if ((so->so_options & SO_REUSEADDR) != 0)
1063                                         inp->inp_flags2 |= INP_REUSEADDR;
1064                                 else
1065                                         inp->inp_flags2 &= ~INP_REUSEADDR;
1066                                 INP_WUNLOCK(inp);
1067                                 error = 0;
1068                                 break;
1069                         case SO_REUSEPORT:
1070                                 INP_WLOCK(inp);
1071                                 if ((so->so_options & SO_REUSEPORT) != 0)
1072                                         inp->inp_flags2 |= INP_REUSEPORT;
1073                                 else
1074                                         inp->inp_flags2 &= ~INP_REUSEPORT;
1075                                 INP_WUNLOCK(inp);
1076                                 error = 0;
1077                                 break;
1078                         case SO_REUSEPORT_LB:
1079                                 INP_WLOCK(inp);
1080                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1081                                         inp->inp_flags2 |= INP_REUSEPORT_LB;
1082                                 else
1083                                         inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1084                                 INP_WUNLOCK(inp);
1085                                 error = 0;
1086                                 break;
1087                         case SO_SETFIB:
1088                                 INP_WLOCK(inp);
1089                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
1090                                 INP_WUNLOCK(inp);
1091                                 error = 0;
1092                                 break;
1093                         case SO_MAX_PACING_RATE:
1094 #ifdef RATELIMIT
1095                                 INP_WLOCK(inp);
1096                                 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1097                                 INP_WUNLOCK(inp);
1098                                 error = 0;
1099 #else
1100                                 error = EOPNOTSUPP;
1101 #endif
1102                                 break;
1103                         default:
1104                                 break;
1105                         }
1106                 }
1107                 return (error);
1108         }
1109
1110         switch (sopt->sopt_dir) {
1111         case SOPT_SET:
1112                 switch (sopt->sopt_name) {
1113                 case IP_OPTIONS:
1114 #ifdef notyet
1115                 case IP_RETOPTS:
1116 #endif
1117                 {
1118                         struct mbuf *m;
1119                         if (sopt->sopt_valsize > MLEN) {
1120                                 error = EMSGSIZE;
1121                                 break;
1122                         }
1123                         m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1124                         if (m == NULL) {
1125                                 error = ENOBUFS;
1126                                 break;
1127                         }
1128                         m->m_len = sopt->sopt_valsize;
1129                         error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1130                                             m->m_len);
1131                         if (error) {
1132                                 m_free(m);
1133                                 break;
1134                         }
1135                         INP_WLOCK(inp);
1136                         error = ip_pcbopts(inp, sopt->sopt_name, m);
1137                         INP_WUNLOCK(inp);
1138                         return (error);
1139                 }
1140
1141                 case IP_BINDANY:
1142                         if (sopt->sopt_td != NULL) {
1143                                 error = priv_check(sopt->sopt_td,
1144                                     PRIV_NETINET_BINDANY);
1145                                 if (error)
1146                                         break;
1147                         }
1148                         /* FALLTHROUGH */
1149                 case IP_BINDMULTI:
1150 #ifdef  RSS
1151                 case IP_RSS_LISTEN_BUCKET:
1152 #endif
1153                 case IP_TOS:
1154                 case IP_TTL:
1155                 case IP_MINTTL:
1156                 case IP_RECVOPTS:
1157                 case IP_RECVRETOPTS:
1158                 case IP_ORIGDSTADDR:
1159                 case IP_RECVDSTADDR:
1160                 case IP_RECVTTL:
1161                 case IP_RECVIF:
1162                 case IP_ONESBCAST:
1163                 case IP_DONTFRAG:
1164                 case IP_RECVTOS:
1165                 case IP_RECVFLOWID:
1166 #ifdef  RSS
1167                 case IP_RECVRSSBUCKETID:
1168 #endif
1169                         error = sooptcopyin(sopt, &optval, sizeof optval,
1170                                             sizeof optval);
1171                         if (error)
1172                                 break;
1173
1174                         switch (sopt->sopt_name) {
1175                         case IP_TOS:
1176                                 inp->inp_ip_tos = optval;
1177                                 break;
1178
1179                         case IP_TTL:
1180                                 inp->inp_ip_ttl = optval;
1181                                 break;
1182
1183                         case IP_MINTTL:
1184                                 if (optval >= 0 && optval <= MAXTTL)
1185                                         inp->inp_ip_minttl = optval;
1186                                 else
1187                                         error = EINVAL;
1188                                 break;
1189
1190 #define OPTSET(bit) do {                                                \
1191         INP_WLOCK(inp);                                                 \
1192         if (optval)                                                     \
1193                 inp->inp_flags |= bit;                                  \
1194         else                                                            \
1195                 inp->inp_flags &= ~bit;                                 \
1196         INP_WUNLOCK(inp);                                               \
1197 } while (0)
1198
1199 #define OPTSET2(bit, val) do {                                          \
1200         INP_WLOCK(inp);                                                 \
1201         if (val)                                                        \
1202                 inp->inp_flags2 |= bit;                                 \
1203         else                                                            \
1204                 inp->inp_flags2 &= ~bit;                                \
1205         INP_WUNLOCK(inp);                                               \
1206 } while (0)
1207
1208                         case IP_RECVOPTS:
1209                                 OPTSET(INP_RECVOPTS);
1210                                 break;
1211
1212                         case IP_RECVRETOPTS:
1213                                 OPTSET(INP_RECVRETOPTS);
1214                                 break;
1215
1216                         case IP_RECVDSTADDR:
1217                                 OPTSET(INP_RECVDSTADDR);
1218                                 break;
1219
1220                         case IP_ORIGDSTADDR:
1221                                 OPTSET2(INP_ORIGDSTADDR, optval);
1222                                 break;
1223
1224                         case IP_RECVTTL:
1225                                 OPTSET(INP_RECVTTL);
1226                                 break;
1227
1228                         case IP_RECVIF:
1229                                 OPTSET(INP_RECVIF);
1230                                 break;
1231
1232                         case IP_ONESBCAST:
1233                                 OPTSET(INP_ONESBCAST);
1234                                 break;
1235                         case IP_DONTFRAG:
1236                                 OPTSET(INP_DONTFRAG);
1237                                 break;
1238                         case IP_BINDANY:
1239                                 OPTSET(INP_BINDANY);
1240                                 break;
1241                         case IP_RECVTOS:
1242                                 OPTSET(INP_RECVTOS);
1243                                 break;
1244                         case IP_BINDMULTI:
1245                                 OPTSET2(INP_BINDMULTI, optval);
1246                                 break;
1247                         case IP_RECVFLOWID:
1248                                 OPTSET2(INP_RECVFLOWID, optval);
1249                                 break;
1250 #ifdef  RSS
1251                         case IP_RSS_LISTEN_BUCKET:
1252                                 if ((optval >= 0) &&
1253                                     (optval < rss_getnumbuckets())) {
1254                                         inp->inp_rss_listen_bucket = optval;
1255                                         OPTSET2(INP_RSS_BUCKET_SET, 1);
1256                                 } else {
1257                                         error = EINVAL;
1258                                 }
1259                                 break;
1260                         case IP_RECVRSSBUCKETID:
1261                                 OPTSET2(INP_RECVRSSBUCKETID, optval);
1262                                 break;
1263 #endif
1264                         }
1265                         break;
1266 #undef OPTSET
1267 #undef OPTSET2
1268
1269                 /*
1270                  * Multicast socket options are processed by the in_mcast
1271                  * module.
1272                  */
1273                 case IP_MULTICAST_IF:
1274                 case IP_MULTICAST_VIF:
1275                 case IP_MULTICAST_TTL:
1276                 case IP_MULTICAST_LOOP:
1277                 case IP_ADD_MEMBERSHIP:
1278                 case IP_DROP_MEMBERSHIP:
1279                 case IP_ADD_SOURCE_MEMBERSHIP:
1280                 case IP_DROP_SOURCE_MEMBERSHIP:
1281                 case IP_BLOCK_SOURCE:
1282                 case IP_UNBLOCK_SOURCE:
1283                 case IP_MSFILTER:
1284                 case MCAST_JOIN_GROUP:
1285                 case MCAST_LEAVE_GROUP:
1286                 case MCAST_JOIN_SOURCE_GROUP:
1287                 case MCAST_LEAVE_SOURCE_GROUP:
1288                 case MCAST_BLOCK_SOURCE:
1289                 case MCAST_UNBLOCK_SOURCE:
1290                         error = inp_setmoptions(inp, sopt);
1291                         break;
1292
1293                 case IP_PORTRANGE:
1294                         error = sooptcopyin(sopt, &optval, sizeof optval,
1295                                             sizeof optval);
1296                         if (error)
1297                                 break;
1298
1299                         INP_WLOCK(inp);
1300                         switch (optval) {
1301                         case IP_PORTRANGE_DEFAULT:
1302                                 inp->inp_flags &= ~(INP_LOWPORT);
1303                                 inp->inp_flags &= ~(INP_HIGHPORT);
1304                                 break;
1305
1306                         case IP_PORTRANGE_HIGH:
1307                                 inp->inp_flags &= ~(INP_LOWPORT);
1308                                 inp->inp_flags |= INP_HIGHPORT;
1309                                 break;
1310
1311                         case IP_PORTRANGE_LOW:
1312                                 inp->inp_flags &= ~(INP_HIGHPORT);
1313                                 inp->inp_flags |= INP_LOWPORT;
1314                                 break;
1315
1316                         default:
1317                                 error = EINVAL;
1318                                 break;
1319                         }
1320                         INP_WUNLOCK(inp);
1321                         break;
1322
1323 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1324                 case IP_IPSEC_POLICY:
1325                         if (IPSEC_ENABLED(ipv4)) {
1326                                 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1327                                 break;
1328                         }
1329                         /* FALLTHROUGH */
1330 #endif /* IPSEC */
1331
1332                 default:
1333                         error = ENOPROTOOPT;
1334                         break;
1335                 }
1336                 break;
1337
1338         case SOPT_GET:
1339                 switch (sopt->sopt_name) {
1340                 case IP_OPTIONS:
1341                 case IP_RETOPTS:
1342                         INP_RLOCK(inp);
1343                         if (inp->inp_options) {
1344                                 struct mbuf *options;
1345
1346                                 options = m_copym(inp->inp_options, 0,
1347                                     M_COPYALL, M_NOWAIT);
1348                                 INP_RUNLOCK(inp);
1349                                 if (options != NULL) {
1350                                         error = sooptcopyout(sopt,
1351                                                              mtod(options, char *),
1352                                                              options->m_len);
1353                                         m_freem(options);
1354                                 } else
1355                                         error = ENOMEM;
1356                         } else {
1357                                 INP_RUNLOCK(inp);
1358                                 sopt->sopt_valsize = 0;
1359                         }
1360                         break;
1361
1362                 case IP_TOS:
1363                 case IP_TTL:
1364                 case IP_MINTTL:
1365                 case IP_RECVOPTS:
1366                 case IP_RECVRETOPTS:
1367                 case IP_ORIGDSTADDR:
1368                 case IP_RECVDSTADDR:
1369                 case IP_RECVTTL:
1370                 case IP_RECVIF:
1371                 case IP_PORTRANGE:
1372                 case IP_ONESBCAST:
1373                 case IP_DONTFRAG:
1374                 case IP_BINDANY:
1375                 case IP_RECVTOS:
1376                 case IP_BINDMULTI:
1377                 case IP_FLOWID:
1378                 case IP_FLOWTYPE:
1379                 case IP_RECVFLOWID:
1380 #ifdef  RSS
1381                 case IP_RSSBUCKETID:
1382                 case IP_RECVRSSBUCKETID:
1383 #endif
1384                         switch (sopt->sopt_name) {
1385
1386                         case IP_TOS:
1387                                 optval = inp->inp_ip_tos;
1388                                 break;
1389
1390                         case IP_TTL:
1391                                 optval = inp->inp_ip_ttl;
1392                                 break;
1393
1394                         case IP_MINTTL:
1395                                 optval = inp->inp_ip_minttl;
1396                                 break;
1397
1398 #define OPTBIT(bit)     (inp->inp_flags & bit ? 1 : 0)
1399 #define OPTBIT2(bit)    (inp->inp_flags2 & bit ? 1 : 0)
1400
1401                         case IP_RECVOPTS:
1402                                 optval = OPTBIT(INP_RECVOPTS);
1403                                 break;
1404
1405                         case IP_RECVRETOPTS:
1406                                 optval = OPTBIT(INP_RECVRETOPTS);
1407                                 break;
1408
1409                         case IP_RECVDSTADDR:
1410                                 optval = OPTBIT(INP_RECVDSTADDR);
1411                                 break;
1412
1413                         case IP_ORIGDSTADDR:
1414                                 optval = OPTBIT2(INP_ORIGDSTADDR);
1415                                 break;
1416
1417                         case IP_RECVTTL:
1418                                 optval = OPTBIT(INP_RECVTTL);
1419                                 break;
1420
1421                         case IP_RECVIF:
1422                                 optval = OPTBIT(INP_RECVIF);
1423                                 break;
1424
1425                         case IP_PORTRANGE:
1426                                 if (inp->inp_flags & INP_HIGHPORT)
1427                                         optval = IP_PORTRANGE_HIGH;
1428                                 else if (inp->inp_flags & INP_LOWPORT)
1429                                         optval = IP_PORTRANGE_LOW;
1430                                 else
1431                                         optval = 0;
1432                                 break;
1433
1434                         case IP_ONESBCAST:
1435                                 optval = OPTBIT(INP_ONESBCAST);
1436                                 break;
1437                         case IP_DONTFRAG:
1438                                 optval = OPTBIT(INP_DONTFRAG);
1439                                 break;
1440                         case IP_BINDANY:
1441                                 optval = OPTBIT(INP_BINDANY);
1442                                 break;
1443                         case IP_RECVTOS:
1444                                 optval = OPTBIT(INP_RECVTOS);
1445                                 break;
1446                         case IP_FLOWID:
1447                                 optval = inp->inp_flowid;
1448                                 break;
1449                         case IP_FLOWTYPE:
1450                                 optval = inp->inp_flowtype;
1451                                 break;
1452                         case IP_RECVFLOWID:
1453                                 optval = OPTBIT2(INP_RECVFLOWID);
1454                                 break;
1455 #ifdef  RSS
1456                         case IP_RSSBUCKETID:
1457                                 retval = rss_hash2bucket(inp->inp_flowid,
1458                                     inp->inp_flowtype,
1459                                     &rss_bucket);
1460                                 if (retval == 0)
1461                                         optval = rss_bucket;
1462                                 else
1463                                         error = EINVAL;
1464                                 break;
1465                         case IP_RECVRSSBUCKETID:
1466                                 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1467                                 break;
1468 #endif
1469                         case IP_BINDMULTI:
1470                                 optval = OPTBIT2(INP_BINDMULTI);
1471                                 break;
1472                         }
1473                         error = sooptcopyout(sopt, &optval, sizeof optval);
1474                         break;
1475
1476                 /*
1477                  * Multicast socket options are processed by the in_mcast
1478                  * module.
1479                  */
1480                 case IP_MULTICAST_IF:
1481                 case IP_MULTICAST_VIF:
1482                 case IP_MULTICAST_TTL:
1483                 case IP_MULTICAST_LOOP:
1484                 case IP_MSFILTER:
1485                         error = inp_getmoptions(inp, sopt);
1486                         break;
1487
1488 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1489                 case IP_IPSEC_POLICY:
1490                         if (IPSEC_ENABLED(ipv4)) {
1491                                 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1492                                 break;
1493                         }
1494                         /* FALLTHROUGH */
1495 #endif /* IPSEC */
1496
1497                 default:
1498                         error = ENOPROTOOPT;
1499                         break;
1500                 }
1501                 break;
1502         }
1503         return (error);
1504 }
1505
1506 /*
1507  * Routine called from ip_output() to loop back a copy of an IP multicast
1508  * packet to the input queue of a specified interface.  Note that this
1509  * calls the output routine of the loopback "driver", but with an interface
1510  * pointer that might NOT be a loopback interface -- evil, but easier than
1511  * replicating that code here.
1512  */
1513 static void
1514 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1515 {
1516         struct ip *ip;
1517         struct mbuf *copym;
1518
1519         /*
1520          * Make a deep copy of the packet because we're going to
1521          * modify the pack in order to generate checksums.
1522          */
1523         copym = m_dup(m, M_NOWAIT);
1524         if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1525                 copym = m_pullup(copym, hlen);
1526         if (copym != NULL) {
1527                 /* If needed, compute the checksum and mark it as valid. */
1528                 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1529                         in_delayed_cksum(copym);
1530                         copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1531                         copym->m_pkthdr.csum_flags |=
1532                             CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1533                         copym->m_pkthdr.csum_data = 0xffff;
1534                 }
1535                 /*
1536                  * We don't bother to fragment if the IP length is greater
1537                  * than the interface's MTU.  Can this possibly matter?
1538                  */
1539                 ip = mtod(copym, struct ip *);
1540                 ip->ip_sum = 0;
1541                 ip->ip_sum = in_cksum(copym, hlen);
1542                 if_simloop(ifp, copym, AF_INET, 0);
1543         }
1544 }