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_kern_tls.h"
  40 #include "opt_mbuf_stress_test.h"
  41 #include "opt_mpath.h"
  42 #include "opt_ratelimit.h"
  43 #include "opt_route.h"
  44 #include "opt_rss.h"
  45 #include "opt_sctp.h"
  46
  47 #include <sys/param.h>
  48 #include <sys/systm.h>
  49 #include <sys/kernel.h>
  50 #include <sys/ktls.h>
  51 #include <sys/lock.h>
  52 #include <sys/malloc.h>
  53 #include <sys/mbuf.h>
  54 #include <sys/priv.h>
  55 #include <sys/proc.h>
  56 #include <sys/protosw.h>
  57 #include <sys/rmlock.h>
  58 #include <sys/sdt.h>
  59 #include <sys/socket.h>
  60 #include <sys/socketvar.h>
  61 #include <sys/sysctl.h>
  62 #include <sys/ucred.h>
  63
  64 #include <net/if.h>
  65 #include <net/if_var.h>
  66 #include <net/if_llatbl.h>
  67 #include <net/netisr.h>
  68 #include <net/pfil.h>
  69 #include <net/route.h>
  70 #include <net/route/nhop.h>
  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_fib.h>
  80 #include <netinet/in_pcb.h>
  81 #include <netinet/in_rss.h>
  82 #include <netinet/in_var.h>
  83 #include <netinet/ip_var.h>
  84 #include <netinet/ip_options.h>
  85
  86 #include <netinet/udp.h>
  87 #include <netinet/udp_var.h>
  88
  89 #if defined(SCTP) || defined(SCTP_SUPPORT)
  90 #include <netinet/sctp.h>
  91 #include <netinet/sctp_crc32.h>
  92 #endif
  93
  94 #include <netipsec/ipsec_support.h>
  95
  96 #include <machine/in_cksum.h>
  97
  98 #include <security/mac/mac_framework.h>
  99
 100 #ifdef MBUF_STRESS_TEST
 101 static int mbuf_frag_size = 0;
 102 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
 103         &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
 104 #endif
 105
 106 static void     ip_mloopback(struct ifnet *, const struct mbuf *, int);
 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 = EACCES;
 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 #if defined(SCTP) || defined(SCTP_SUPPORT)
 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 #if defined(SCTP) || defined(SCTP_SUPPORT)
 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, bool stamp_tag)
 214 {
 215 #ifdef KERN_TLS
 216         struct ktls_session *tls = NULL;
 217 #endif
 218         struct m_snd_tag *mst;
 219         int error;
 220
 221         MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
 222         mst = NULL;
 223
 224 #ifdef KERN_TLS
 225         /*
 226          * If this is an unencrypted TLS record, save a reference to
 227          * the record.  This local reference is used to call
 228          * ktls_output_eagain after the mbuf has been freed (thus
 229          * dropping the mbuf's reference) in if_output.
 230          */
 231         if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
 232                 tls = ktls_hold(m->m_next->m_epg_tls);
 233                 mst = tls->snd_tag;
 234
 235                 /*
 236                  * If a TLS session doesn't have a valid tag, it must
 237                  * have had an earlier ifp mismatch, so drop this
 238                  * packet.
 239                  */
 240                 if (mst == NULL) {
 241                         error = EAGAIN;
 242                         goto done;
 243                 }
 244                 /*
 245                  * Always stamp tags that include NIC ktls.
 246                  */
 247                 stamp_tag = true;
 248         }
 249 #endif
 250 #ifdef RATELIMIT
 251         if (inp != NULL && mst == NULL) {
 252                 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
 253                     (inp->inp_snd_tag != NULL &&
 254                     inp->inp_snd_tag->ifp != ifp))
 255                         in_pcboutput_txrtlmt(inp, ifp, m);
 256
 257                 if (inp->inp_snd_tag != NULL)
 258                         mst = inp->inp_snd_tag;
 259         }
 260 #endif
 261         if (stamp_tag && mst != NULL) {
 262                 KASSERT(m->m_pkthdr.rcvif == NULL,
 263                     ("trying to add a send tag to a forwarded packet"));
 264                 if (mst->ifp != ifp) {
 265                         error = EAGAIN;
 266                         goto done;
 267                 }
 268
 269                 /* stamp send tag on mbuf */
 270                 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
 271                 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
 272         }
 273
 274         error = (*ifp->if_output)(ifp, m, (const struct sockaddr *)gw, ro);
 275
 276 done:
 277         /* Check for route change invalidating send tags. */
 278 #ifdef KERN_TLS
 279         if (tls != NULL) {
 280                 if (error == EAGAIN)
 281                         error = ktls_output_eagain(inp, tls);
 282                 ktls_free(tls);
 283         }
 284 #endif
 285 #ifdef RATELIMIT
 286         if (error == EAGAIN)
 287                 in_pcboutput_eagain(inp);
 288 #endif
 289         return (error);
 290 }
 291
 292 /* rte<>ro_flags translation */
 293 static inline void
 294 rt_update_ro_flags(struct route *ro)
 295 {
 296         int nh_flags = ro->ro_nh->nh_flags;
 297
 298         ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
 299
 300         ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
 301         ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
 302         ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
 303 }
 304
 305 /*
 306  * IP output.  The packet in mbuf chain m contains a skeletal IP
 307  * header (with len, off, ttl, proto, tos, src, dst).
 308  * The mbuf chain containing the packet will be freed.
 309  * The mbuf opt, if present, will not be freed.
 310  * If route ro is present and has ro_rt initialized, route lookup would be
 311  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 312  * then result of route lookup is stored in ro->ro_rt.
 313  *
 314  * In the IP forwarding case, the packet will arrive with options already
 315  * inserted, so must have a NULL opt pointer.
 316  */
 317 int
 318 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
 319     struct ip_moptions *imo, struct inpcb *inp)
 320 {
 321         struct rm_priotracker in_ifa_tracker;
 322         struct ip *ip;
 323         struct ifnet *ifp = NULL;       /* keep compiler happy */
 324         struct mbuf *m0;
 325         int hlen = sizeof (struct ip);
 326         int mtu = 0;
 327         int error = 0;
 328         struct sockaddr_in *dst, sin;
 329         const struct sockaddr_in *gw;
 330         struct in_ifaddr *ia = NULL;
 331         struct in_addr src;
 332         int isbroadcast;
 333         uint16_t ip_len, ip_off;
 334         uint32_t fibnum;
 335 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 336         int no_route_but_check_spd = 0;
 337 #endif
 338
 339         M_ASSERTPKTHDR(m);
 340         NET_EPOCH_ASSERT();
 341
 342         if (inp != NULL) {
 343                 INP_LOCK_ASSERT(inp);
 344                 M_SETFIB(m, inp->inp_inc.inc_fibnum);
 345                 if ((flags & IP_NODEFAULTFLOWID) == 0) {
 346                         m->m_pkthdr.flowid = inp->inp_flowid;
 347                         M_HASHTYPE_SET(m, inp->inp_flowtype);
 348                 }
 349 #ifdef NUMA
 350                 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
 351 #endif
 352         }
 353
 354         if (opt) {
 355                 int len = 0;
 356                 m = ip_insertoptions(m, opt, &len);
 357                 if (len != 0)
 358                         hlen = len; /* ip->ip_hl is updated above */
 359         }
 360         ip = mtod(m, struct ip *);
 361         ip_len = ntohs(ip->ip_len);
 362         ip_off = ntohs(ip->ip_off);
 363
 364         if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
 365                 ip->ip_v = IPVERSION;
 366                 ip->ip_hl = hlen >> 2;
 367                 ip_fillid(ip);
 368         } else {
 369                 /* Header already set, fetch hlen from there */
 370                 hlen = ip->ip_hl << 2;
 371         }
 372         if ((flags & IP_FORWARDING) == 0)
 373                 IPSTAT_INC(ips_localout);
 374
 375         /*
 376          * dst/gw handling:
 377          *
 378          * gw is readonly but can point either to dst OR rt_gateway,
 379          * therefore we need restore gw if we're redoing lookup.
 380          */
 381         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 382         if (ro != NULL)
 383                 dst = (struct sockaddr_in *)&ro->ro_dst;
 384         else
 385                 dst = &sin;
 386         if (ro == NULL || ro->ro_nh == NULL) {
 387                 bzero(dst, sizeof(*dst));
 388                 dst->sin_family = AF_INET;
 389                 dst->sin_len = sizeof(*dst);
 390                 dst->sin_addr = ip->ip_dst;
 391         }
 392         gw = dst;
 393 again:
 394         /*
 395          * Validate route against routing table additions;
 396          * a better/more specific route might have been added.
 397          */
 398         if (inp != NULL && ro != NULL && ro->ro_nh != NULL)
 399                 NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
 400         /*
 401          * If there is a cached route,
 402          * check that it is to the same destination
 403          * and is still up.  If not, free it and try again.
 404          * The address family should also be checked in case of sharing the
 405          * cache with IPv6.
 406          * Also check whether routing cache needs invalidation.
 407          */
 408         if (ro != NULL && ro->ro_nh != NULL &&
 409             ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
 410             dst->sin_addr.s_addr != ip->ip_dst.s_addr))
 411                 RO_INVALIDATE_CACHE(ro);
 412         ia = NULL;
 413         /*
 414          * If routing to interface only, short circuit routing lookup.
 415          * The use of an all-ones broadcast address implies this; an
 416          * interface is specified by the broadcast address of an interface,
 417          * or the destination address of a ptp interface.
 418          */
 419         if (flags & IP_SENDONES) {
 420                 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
 421                                                       M_GETFIB(m)))) == NULL &&
 422                     (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 423                                                     M_GETFIB(m)))) == NULL) {
 424                         IPSTAT_INC(ips_noroute);
 425                         error = ENETUNREACH;
 426                         goto bad;
 427                 }
 428                 ip->ip_dst.s_addr = INADDR_BROADCAST;
 429                 dst->sin_addr = ip->ip_dst;
 430                 ifp = ia->ia_ifp;
 431                 mtu = ifp->if_mtu;
 432                 ip->ip_ttl = 1;
 433                 isbroadcast = 1;
 434                 src = IA_SIN(ia)->sin_addr;
 435         } else if (flags & IP_ROUTETOIF) {
 436                 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 437                                                     M_GETFIB(m)))) == NULL &&
 438                     (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
 439                                                 M_GETFIB(m)))) == NULL) {
 440                         IPSTAT_INC(ips_noroute);
 441                         error = ENETUNREACH;
 442                         goto bad;
 443                 }
 444                 ifp = ia->ia_ifp;
 445                 mtu = ifp->if_mtu;
 446                 ip->ip_ttl = 1;
 447                 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
 448                     in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
 449                 src = IA_SIN(ia)->sin_addr;
 450         } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
 451             imo != NULL && imo->imo_multicast_ifp != NULL) {
 452                 /*
 453                  * Bypass the normal routing lookup for multicast
 454                  * packets if the interface is specified.
 455                  */
 456                 ifp = imo->imo_multicast_ifp;
 457                 mtu = ifp->if_mtu;
 458                 IFP_TO_IA(ifp, ia, &in_ifa_tracker);
 459                 isbroadcast = 0;        /* fool gcc */
 460                 /* Interface may have no addresses. */
 461                 if (ia != NULL)
 462                         src = IA_SIN(ia)->sin_addr;
 463                 else
 464                         src.s_addr = INADDR_ANY;
 465         } else if (ro != NULL) {
 466                 if (ro->ro_nh == NULL) {
 467                         /*
 468                          * We want to do any cloning requested by the link
 469                          * layer, as this is probably required in all cases
 470                          * for correct operation (as it is for ARP).
 471                          */
 472                         uint32_t flowid;
 473 #ifdef RADIX_MPATH
 474                         flowid = ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr);
 475 #else
 476                         flowid = m->m_pkthdr.flowid;
 477 #endif
 478                         ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
 479                             NHR_REF, flowid);
 480
 481                         if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
 482 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 483                                 /*
 484                                  * There is no route for this packet, but it is
 485                                  * possible that a matching SPD entry exists.
 486                                  */
 487                                 no_route_but_check_spd = 1;
 488                                 goto sendit;
 489 #endif
 490                                 IPSTAT_INC(ips_noroute);
 491                                 error = EHOSTUNREACH;
 492                                 goto bad;
 493                         }
 494                 }
 495                 ia = ifatoia(ro->ro_nh->nh_ifa);
 496                 ifp = ro->ro_nh->nh_ifp;
 497                 counter_u64_add(ro->ro_nh->nh_pksent, 1);
 498                 rt_update_ro_flags(ro);
 499                 if (ro->ro_nh->nh_flags & NHF_GATEWAY)
 500                         gw = &ro->ro_nh->gw4_sa;
 501                 if (ro->ro_nh->nh_flags & NHF_HOST)
 502                         isbroadcast = (ro->ro_nh->nh_flags & NHF_BROADCAST);
 503                 else if (ifp->if_flags & IFF_BROADCAST)
 504                         isbroadcast = in_ifaddr_broadcast(gw->sin_addr, ia);
 505                 else
 506                         isbroadcast = 0;
 507                 if (ro->ro_nh->nh_flags & NHF_HOST)
 508                         mtu = ro->ro_nh->nh_mtu;
 509                 else
 510                         mtu = ifp->if_mtu;
 511                 src = IA_SIN(ia)->sin_addr;
 512         } else {
 513                 struct nhop_object *nh;
 514
 515                 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE,
 516                     m->m_pkthdr.flowid);
 517                 if (nh == NULL) {
 518 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 519                         /*
 520                          * There is no route for this packet, but it is
 521                          * possible that a matching SPD entry exists.
 522                          */
 523                         no_route_but_check_spd = 1;
 524                         goto sendit;
 525 #endif
 526                         IPSTAT_INC(ips_noroute);
 527                         error = EHOSTUNREACH;
 528                         goto bad;
 529                 }
 530                 ifp = nh->nh_ifp;
 531                 mtu = nh->nh_mtu;
 532                 /*
 533                  * We are rewriting here dst to be gw actually, contradicting
 534                  * comment at the beginning of the function. However, in this
 535                  * case we are always dealing with on stack dst.
 536                  * In case if pfil(9) sends us back to beginning of the
 537                  * function, the dst would be rewritten by ip_output_pfil().
 538                  */
 539                 MPASS(dst == &sin);
 540                 if (nh->nh_flags & NHF_GATEWAY)
 541                         dst->sin_addr = nh->gw4_sa.sin_addr;
 542                 ia = ifatoia(nh->nh_ifa);
 543                 src = IA_SIN(ia)->sin_addr;
 544                 isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
 545                     (NHF_HOST | NHF_BROADCAST)) ||
 546                     ((ifp->if_flags & IFF_BROADCAST) &&
 547                     in_ifaddr_broadcast(dst->sin_addr, ia)));
 548         }
 549
 550         /* Catch a possible divide by zero later. */
 551         KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
 552             __func__, mtu, ro,
 553             (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
 554
 555         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
 556                 m->m_flags |= M_MCAST;
 557                 /*
 558                  * IP destination address is multicast.  Make sure "gw"
 559                  * still points to the address in "ro".  (It may have been
 560                  * changed to point to a gateway address, above.)
 561                  */
 562                 gw = dst;
 563                 /*
 564                  * See if the caller provided any multicast options
 565                  */
 566                 if (imo != NULL) {
 567                         ip->ip_ttl = imo->imo_multicast_ttl;
 568                         if (imo->imo_multicast_vif != -1)
 569                                 ip->ip_src.s_addr =
 570                                     ip_mcast_src ?
 571                                     ip_mcast_src(imo->imo_multicast_vif) :
 572                                     INADDR_ANY;
 573                 } else
 574                         ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
 575                 /*
 576                  * Confirm that the outgoing interface supports multicast.
 577                  */
 578                 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
 579                         if ((ifp->if_flags & IFF_MULTICAST) == 0) {
 580                                 IPSTAT_INC(ips_noroute);
 581                                 error = ENETUNREACH;
 582                                 goto bad;
 583                         }
 584                 }
 585                 /*
 586                  * If source address not specified yet, use address
 587                  * of outgoing interface.
 588                  */
 589                 if (ip->ip_src.s_addr == INADDR_ANY)
 590                         ip->ip_src = src;
 591
 592                 if ((imo == NULL && in_mcast_loop) ||
 593                     (imo && imo->imo_multicast_loop)) {
 594                         /*
 595                          * Loop back multicast datagram if not expressly
 596                          * forbidden to do so, even if we are not a member
 597                          * of the group; ip_input() will filter it later,
 598                          * thus deferring a hash lookup and mutex acquisition
 599                          * at the expense of a cheap copy using m_copym().
 600                          */
 601                         ip_mloopback(ifp, m, hlen);
 602                 } else {
 603                         /*
 604                          * If we are acting as a multicast router, perform
 605                          * multicast forwarding as if the packet had just
 606                          * arrived on the interface to which we are about
 607                          * to send.  The multicast forwarding function
 608                          * recursively calls this function, using the
 609                          * IP_FORWARDING flag to prevent infinite recursion.
 610                          *
 611                          * Multicasts that are looped back by ip_mloopback(),
 612                          * above, will be forwarded by the ip_input() routine,
 613                          * if necessary.
 614                          */
 615                         if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
 616                                 /*
 617                                  * If rsvp daemon is not running, do not
 618                                  * set ip_moptions. This ensures that the packet
 619                                  * is multicast and not just sent down one link
 620                                  * as prescribed by rsvpd.
 621                                  */
 622                                 if (!V_rsvp_on)
 623                                         imo = NULL;
 624                                 if (ip_mforward &&
 625                                     ip_mforward(ip, ifp, m, imo) != 0) {
 626                                         m_freem(m);
 627                                         goto done;
 628                                 }
 629                         }
 630                 }
 631
 632                 /*
 633                  * Multicasts with a time-to-live of zero may be looped-
 634                  * back, above, but must not be transmitted on a network.
 635                  * Also, multicasts addressed to the loopback interface
 636                  * are not sent -- the above call to ip_mloopback() will
 637                  * loop back a copy. ip_input() will drop the copy if
 638                  * this host does not belong to the destination group on
 639                  * the loopback interface.
 640                  */
 641                 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
 642                         m_freem(m);
 643                         goto done;
 644                 }
 645
 646                 goto sendit;
 647         }
 648
 649         /*
 650          * If the source address is not specified yet, use the address
 651          * of the outoing interface.
 652          */
 653         if (ip->ip_src.s_addr == INADDR_ANY)
 654                 ip->ip_src = src;
 655
 656         /*
 657          * Look for broadcast address and
 658          * verify user is allowed to send
 659          * such a packet.
 660          */
 661         if (isbroadcast) {
 662                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 663                         error = EADDRNOTAVAIL;
 664                         goto bad;
 665                 }
 666                 if ((flags & IP_ALLOWBROADCAST) == 0) {
 667                         error = EACCES;
 668                         goto bad;
 669                 }
 670                 /* don't allow broadcast messages to be fragmented */
 671                 if (ip_len > mtu) {
 672                         error = EMSGSIZE;
 673                         goto bad;
 674                 }
 675                 m->m_flags |= M_BCAST;
 676         } else {
 677                 m->m_flags &= ~M_BCAST;
 678         }
 679
 680 sendit:
 681 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 682         if (IPSEC_ENABLED(ipv4)) {
 683                 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
 684                         if (error == EINPROGRESS)
 685                                 error = 0;
 686                         goto done;
 687                 }
 688         }
 689         /*
 690          * Check if there was a route for this packet; return error if not.
 691          */
 692         if (no_route_but_check_spd) {
 693                 IPSTAT_INC(ips_noroute);
 694                 error = EHOSTUNREACH;
 695                 goto bad;
 696         }
 697         /* Update variables that are affected by ipsec4_output(). */
 698         ip = mtod(m, struct ip *);
 699         hlen = ip->ip_hl << 2;
 700 #endif /* IPSEC */
 701
 702         /* Jump over all PFIL processing if hooks are not active. */
 703         if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
 704                 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
 705                     &error)) {
 706                 case 1: /* Finished */
 707                         goto done;
 708
 709                 case 0: /* Continue normally */
 710                         ip = mtod(m, struct ip *);
 711                         break;
 712
 713                 case -1: /* Need to try again */
 714                         /* Reset everything for a new round */
 715                         if (ro != NULL) {
 716                                 RO_NHFREE(ro);
 717                                 ro->ro_prepend = NULL;
 718                         }
 719                         gw = dst;
 720                         ip = mtod(m, struct ip *);
 721                         goto again;
 722                 }
 723         }
 724
 725         /* IN_LOOPBACK must not appear on the wire - RFC1122. */
 726         if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
 727             IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
 728                 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
 729                         IPSTAT_INC(ips_badaddr);
 730                         error = EADDRNOTAVAIL;
 731                         goto bad;
 732                 }
 733         }
 734
 735         m->m_pkthdr.csum_flags |= CSUM_IP;
 736         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
 737                 m = mb_unmapped_to_ext(m);
 738                 if (m == NULL) {
 739                         IPSTAT_INC(ips_odropped);
 740                         error = ENOBUFS;
 741                         goto bad;
 742                 }
 743                 in_delayed_cksum(m);
 744                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 745         } else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
 746                 m = mb_unmapped_to_ext(m);
 747                 if (m == NULL) {
 748                         IPSTAT_INC(ips_odropped);
 749                         error = ENOBUFS;
 750                         goto bad;
 751                 }
 752         }
 753 #if defined(SCTP) || defined(SCTP_SUPPORT)
 754         if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
 755                 m = mb_unmapped_to_ext(m);
 756                 if (m == NULL) {
 757                         IPSTAT_INC(ips_odropped);
 758                         error = ENOBUFS;
 759                         goto bad;
 760                 }
 761                 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
 762                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 763         }
 764 #endif
 765
 766         /*
 767          * If small enough for interface, or the interface will take
 768          * care of the fragmentation for us, we can just send directly.
 769          * Note that if_vxlan could have requested TSO even though the outer
 770          * frame is UDP.  It is correct to not fragment such datagrams and
 771          * instead just pass them on to the driver.
 772          */
 773         if (ip_len <= mtu ||
 774             (m->m_pkthdr.csum_flags & ifp->if_hwassist &
 775             (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
 776                 ip->ip_sum = 0;
 777                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
 778                         ip->ip_sum = in_cksum(m, hlen);
 779                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 780                 }
 781
 782                 /*
 783                  * Record statistics for this interface address.
 784                  * With CSUM_TSO the byte/packet count will be slightly
 785                  * incorrect because we count the IP+TCP headers only
 786                  * once instead of for every generated packet.
 787                  */
 788                 if (!(flags & IP_FORWARDING) && ia) {
 789                         if (m->m_pkthdr.csum_flags &
 790                             (CSUM_TSO | CSUM_INNER_TSO))
 791                                 counter_u64_add(ia->ia_ifa.ifa_opackets,
 792                                     m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
 793                         else
 794                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 795
 796                         counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
 797                 }
 798 #ifdef MBUF_STRESS_TEST
 799                 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
 800                         m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
 801 #endif
 802                 /*
 803                  * Reset layer specific mbuf flags
 804                  * to avoid confusing lower layers.
 805                  */
 806                 m_clrprotoflags(m);
 807                 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
 808                 error = ip_output_send(inp, ifp, m, gw, ro,
 809                     (flags & IP_NO_SND_TAG_RL) ? false : true);
 810                 goto done;
 811         }
 812
 813         /* Balk when DF bit is set or the interface didn't support TSO. */
 814         if ((ip_off & IP_DF) ||
 815             (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
 816                 error = EMSGSIZE;
 817                 IPSTAT_INC(ips_cantfrag);
 818                 goto bad;
 819         }
 820
 821         /*
 822          * Too large for interface; fragment if possible. If successful,
 823          * on return, m will point to a list of packets to be sent.
 824          */
 825         error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
 826         if (error)
 827                 goto bad;
 828         for (; m; m = m0) {
 829                 m0 = m->m_nextpkt;
 830                 m->m_nextpkt = 0;
 831                 if (error == 0) {
 832                         /* Record statistics for this interface address. */
 833                         if (ia != NULL) {
 834                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 835                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
 836                                     m->m_pkthdr.len);
 837                         }
 838                         /*
 839                          * Reset layer specific mbuf flags
 840                          * to avoid confusing upper layers.
 841                          */
 842                         m_clrprotoflags(m);
 843
 844                         IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
 845                             mtod(m, struct ip *), NULL);
 846                         error = ip_output_send(inp, ifp, m, gw, ro, true);
 847                 } else
 848                         m_freem(m);
 849         }
 850
 851         if (error == 0)
 852                 IPSTAT_INC(ips_fragmented);
 853
 854 done:
 855         return (error);
 856  bad:
 857         m_freem(m);
 858         goto done;
 859 }
 860
 861 /*
 862  * Create a chain of fragments which fit the given mtu. m_frag points to the
 863  * mbuf to be fragmented; on return it points to the chain with the fragments.
 864  * Return 0 if no error. If error, m_frag may contain a partially built
 865  * chain of fragments that should be freed by the caller.
 866  *
 867  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
 868  */
 869 int
 870 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
 871     u_long if_hwassist_flags)
 872 {
 873         int error = 0;
 874         int hlen = ip->ip_hl << 2;
 875         int len = (mtu - hlen) & ~7;    /* size of payload in each fragment */
 876         int off;
 877         struct mbuf *m0 = *m_frag;      /* the original packet          */
 878         int firstlen;
 879         struct mbuf **mnext;
 880         int nfrags;
 881         uint16_t ip_len, ip_off;
 882
 883         ip_len = ntohs(ip->ip_len);
 884         ip_off = ntohs(ip->ip_off);
 885
 886         if (ip_off & IP_DF) {   /* Fragmentation not allowed */
 887                 IPSTAT_INC(ips_cantfrag);
 888                 return EMSGSIZE;
 889         }
 890
 891         /*
 892          * Must be able to put at least 8 bytes per fragment.
 893          */
 894         if (len < 8)
 895                 return EMSGSIZE;
 896
 897         /*
 898          * If the interface will not calculate checksums on
 899          * fragmented packets, then do it here.
 900          */
 901         if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 902                 m0 = mb_unmapped_to_ext(m0);
 903                 if (m0 == NULL) {
 904                         error = ENOBUFS;
 905                         IPSTAT_INC(ips_odropped);
 906                         goto done;
 907                 }
 908                 in_delayed_cksum(m0);
 909                 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 910         }
 911 #if defined(SCTP) || defined(SCTP_SUPPORT)
 912         if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
 913                 m0 = mb_unmapped_to_ext(m0);
 914                 if (m0 == NULL) {
 915                         error = ENOBUFS;
 916                         IPSTAT_INC(ips_odropped);
 917                         goto done;
 918                 }
 919                 sctp_delayed_cksum(m0, hlen);
 920                 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 921         }
 922 #endif
 923         if (len > PAGE_SIZE) {
 924                 /*
 925                  * Fragment large datagrams such that each segment
 926                  * contains a multiple of PAGE_SIZE amount of data,
 927                  * plus headers. This enables a receiver to perform
 928                  * page-flipping zero-copy optimizations.
 929                  *
 930                  * XXX When does this help given that sender and receiver
 931                  * could have different page sizes, and also mtu could
 932                  * be less than the receiver's page size ?
 933                  */
 934                 int newlen;
 935
 936                 off = MIN(mtu, m0->m_pkthdr.len);
 937
 938                 /*
 939                  * firstlen (off - hlen) must be aligned on an
 940                  * 8-byte boundary
 941                  */
 942                 if (off < hlen)
 943                         goto smart_frag_failure;
 944                 off = ((off - hlen) & ~7) + hlen;
 945                 newlen = (~PAGE_MASK) & mtu;
 946                 if ((newlen + sizeof (struct ip)) > mtu) {
 947                         /* we failed, go back the default */
 948 smart_frag_failure:
 949                         newlen = len;
 950                         off = hlen + len;
 951                 }
 952                 len = newlen;
 953
 954         } else {
 955                 off = hlen + len;
 956         }
 957
 958         firstlen = off - hlen;
 959         mnext = &m0->m_nextpkt;         /* pointer to next packet */
 960
 961         /*
 962          * Loop through length of segment after first fragment,
 963          * make new header and copy data of each part and link onto chain.
 964          * Here, m0 is the original packet, m is the fragment being created.
 965          * The fragments are linked off the m_nextpkt of the original
 966          * packet, which after processing serves as the first fragment.
 967          */
 968         for (nfrags = 1; off < ip_len; off += len, nfrags++) {
 969                 struct ip *mhip;        /* ip header on the fragment */
 970                 struct mbuf *m;
 971                 int mhlen = sizeof (struct ip);
 972
 973                 m = m_gethdr(M_NOWAIT, MT_DATA);
 974                 if (m == NULL) {
 975                         error = ENOBUFS;
 976                         IPSTAT_INC(ips_odropped);
 977                         goto done;
 978                 }
 979                 /*
 980                  * Make sure the complete packet header gets copied
 981                  * from the originating mbuf to the newly created
 982                  * mbuf. This also ensures that existing firewall
 983                  * classification(s), VLAN tags and so on get copied
 984                  * to the resulting fragmented packet(s):
 985                  */
 986                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 987                         m_free(m);
 988                         error = ENOBUFS;
 989                         IPSTAT_INC(ips_odropped);
 990                         goto done;
 991                 }
 992                 /*
 993                  * In the first mbuf, leave room for the link header, then
 994                  * copy the original IP header including options. The payload
 995                  * goes into an additional mbuf chain returned by m_copym().
 996                  */
 997                 m->m_data += max_linkhdr;
 998                 mhip = mtod(m, struct ip *);
 999                 *mhip = *ip;
1000                 if (hlen > sizeof (struct ip)) {
1001                         mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1002                         mhip->ip_v = IPVERSION;
1003                         mhip->ip_hl = mhlen >> 2;
1004                 }
1005                 m->m_len = mhlen;
1006                 /* XXX do we need to add ip_off below ? */
1007                 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
1008                 if (off + len >= ip_len)
1009                         len = ip_len - off;
1010                 else
1011                         mhip->ip_off |= IP_MF;
1012                 mhip->ip_len = htons((u_short)(len + mhlen));
1013                 m->m_next = m_copym(m0, off, len, M_NOWAIT);
1014                 if (m->m_next == NULL) {        /* copy failed */
1015                         m_free(m);
1016                         error = ENOBUFS;        /* ??? */
1017                         IPSTAT_INC(ips_odropped);
1018                         goto done;
1019                 }
1020                 m->m_pkthdr.len = mhlen + len;
1021 #ifdef MAC
1022                 mac_netinet_fragment(m0, m);
1023 #endif
1024                 mhip->ip_off = htons(mhip->ip_off);
1025                 mhip->ip_sum = 0;
1026                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1027                         mhip->ip_sum = in_cksum(m, mhlen);
1028                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
1029                 }
1030                 *mnext = m;
1031                 mnext = &m->m_nextpkt;
1032         }
1033         IPSTAT_ADD(ips_ofragments, nfrags);
1034
1035         /*
1036          * Update first fragment by trimming what's been copied out
1037          * and updating header.
1038          */
1039         m_adj(m0, hlen + firstlen - ip_len);
1040         m0->m_pkthdr.len = hlen + firstlen;
1041         ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1042         ip->ip_off = htons(ip_off | IP_MF);
1043         ip->ip_sum = 0;
1044         if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1045                 ip->ip_sum = in_cksum(m0, hlen);
1046                 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1047         }
1048
1049 done:
1050         *m_frag = m0;
1051         return error;
1052 }
1053
1054 void
1055 in_delayed_cksum(struct mbuf *m)
1056 {
1057         struct ip *ip;
1058         struct udphdr *uh;
1059         uint16_t cklen, csum, offset;
1060
1061         ip = mtod(m, struct ip *);
1062         offset = ip->ip_hl << 2 ;
1063
1064         if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1065                 /* if udp header is not in the first mbuf copy udplen */
1066                 if (offset + sizeof(struct udphdr) > m->m_len) {
1067                         m_copydata(m, offset + offsetof(struct udphdr,
1068                             uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1069                         cklen = ntohs(cklen);
1070                 } else {
1071                         uh = (struct udphdr *)mtodo(m, offset);
1072                         cklen = ntohs(uh->uh_ulen);
1073                 }
1074                 csum = in_cksum_skip(m, cklen + offset, offset);
1075                 if (csum == 0)
1076                         csum = 0xffff;
1077         } else {
1078                 cklen = ntohs(ip->ip_len);
1079                 csum = in_cksum_skip(m, cklen, offset);
1080         }
1081         offset += m->m_pkthdr.csum_data;        /* checksum offset */
1082
1083         if (offset + sizeof(csum) > m->m_len)
1084                 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1085         else
1086                 *(u_short *)mtodo(m, offset) = csum;
1087 }
1088
1089 /*
1090  * IP socket option processing.
1091  */
1092 int
1093 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1094 {
1095         struct inpcb *inp = sotoinpcb(so);
1096         int     error, optval;
1097 #ifdef  RSS
1098         uint32_t rss_bucket;
1099         int retval;
1100 #endif
1101
1102         error = optval = 0;
1103         if (sopt->sopt_level != IPPROTO_IP) {
1104                 error = EINVAL;
1105
1106                 if (sopt->sopt_level == SOL_SOCKET &&
1107                     sopt->sopt_dir == SOPT_SET) {
1108                         switch (sopt->sopt_name) {
1109                         case SO_REUSEADDR:
1110                                 INP_WLOCK(inp);
1111                                 if ((so->so_options & SO_REUSEADDR) != 0)
1112                                         inp->inp_flags2 |= INP_REUSEADDR;
1113                                 else
1114                                         inp->inp_flags2 &= ~INP_REUSEADDR;
1115                                 INP_WUNLOCK(inp);
1116                                 error = 0;
1117                                 break;
1118                         case SO_REUSEPORT:
1119                                 INP_WLOCK(inp);
1120                                 if ((so->so_options & SO_REUSEPORT) != 0)
1121                                         inp->inp_flags2 |= INP_REUSEPORT;
1122                                 else
1123                                         inp->inp_flags2 &= ~INP_REUSEPORT;
1124                                 INP_WUNLOCK(inp);
1125                                 error = 0;
1126                                 break;
1127                         case SO_REUSEPORT_LB:
1128                                 INP_WLOCK(inp);
1129                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1130                                         inp->inp_flags2 |= INP_REUSEPORT_LB;
1131                                 else
1132                                         inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1133                                 INP_WUNLOCK(inp);
1134                                 error = 0;
1135                                 break;
1136                         case SO_SETFIB:
1137                                 INP_WLOCK(inp);
1138                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
1139                                 INP_WUNLOCK(inp);
1140                                 error = 0;
1141                                 break;
1142                         case SO_MAX_PACING_RATE:
1143 #ifdef RATELIMIT
1144                                 INP_WLOCK(inp);
1145                                 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1146                                 INP_WUNLOCK(inp);
1147                                 error = 0;
1148 #else
1149                                 error = EOPNOTSUPP;
1150 #endif
1151                                 break;
1152                         default:
1153                                 break;
1154                         }
1155                 }
1156                 return (error);
1157         }
1158
1159         switch (sopt->sopt_dir) {
1160         case SOPT_SET:
1161                 switch (sopt->sopt_name) {
1162                 case IP_OPTIONS:
1163 #ifdef notyet
1164                 case IP_RETOPTS:
1165 #endif
1166                 {
1167                         struct mbuf *m;
1168                         if (sopt->sopt_valsize > MLEN) {
1169                                 error = EMSGSIZE;
1170                                 break;
1171                         }
1172                         m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1173                         if (m == NULL) {
1174                                 error = ENOBUFS;
1175                                 break;
1176                         }
1177                         m->m_len = sopt->sopt_valsize;
1178                         error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1179                                             m->m_len);
1180                         if (error) {
1181                                 m_free(m);
1182                                 break;
1183                         }
1184                         INP_WLOCK(inp);
1185                         error = ip_pcbopts(inp, sopt->sopt_name, m);
1186                         INP_WUNLOCK(inp);
1187                         return (error);
1188                 }
1189
1190                 case IP_BINDANY:
1191                         if (sopt->sopt_td != NULL) {
1192                                 error = priv_check(sopt->sopt_td,
1193                                     PRIV_NETINET_BINDANY);
1194                                 if (error)
1195                                         break;
1196                         }
1197                         /* FALLTHROUGH */
1198                 case IP_BINDMULTI:
1199 #ifdef  RSS
1200                 case IP_RSS_LISTEN_BUCKET:
1201 #endif
1202                 case IP_TOS:
1203                 case IP_TTL:
1204                 case IP_MINTTL:
1205                 case IP_RECVOPTS:
1206                 case IP_RECVRETOPTS:
1207                 case IP_ORIGDSTADDR:
1208                 case IP_RECVDSTADDR:
1209                 case IP_RECVTTL:
1210                 case IP_RECVIF:
1211                 case IP_ONESBCAST:
1212                 case IP_DONTFRAG:
1213                 case IP_RECVTOS:
1214                 case IP_RECVFLOWID:
1215 #ifdef  RSS
1216                 case IP_RECVRSSBUCKETID:
1217 #endif
1218                         error = sooptcopyin(sopt, &optval, sizeof optval,
1219                                             sizeof optval);
1220                         if (error)
1221                                 break;
1222
1223                         switch (sopt->sopt_name) {
1224                         case IP_TOS:
1225                                 inp->inp_ip_tos = optval;
1226                                 break;
1227
1228                         case IP_TTL:
1229                                 inp->inp_ip_ttl = optval;
1230                                 break;
1231
1232                         case IP_MINTTL:
1233                                 if (optval >= 0 && optval <= MAXTTL)
1234                                         inp->inp_ip_minttl = optval;
1235                                 else
1236                                         error = EINVAL;
1237                                 break;
1238
1239 #define OPTSET(bit) do {                                                \
1240         INP_WLOCK(inp);                                                 \
1241         if (optval)                                                     \
1242                 inp->inp_flags |= bit;                                  \
1243         else                                                            \
1244                 inp->inp_flags &= ~bit;                                 \
1245         INP_WUNLOCK(inp);                                               \
1246 } while (0)
1247
1248 #define OPTSET2(bit, val) do {                                          \
1249         INP_WLOCK(inp);                                                 \
1250         if (val)                                                        \
1251                 inp->inp_flags2 |= bit;                                 \
1252         else                                                            \
1253                 inp->inp_flags2 &= ~bit;                                \
1254         INP_WUNLOCK(inp);                                               \
1255 } while (0)
1256
1257                         case IP_RECVOPTS:
1258                                 OPTSET(INP_RECVOPTS);
1259                                 break;
1260
1261                         case IP_RECVRETOPTS:
1262                                 OPTSET(INP_RECVRETOPTS);
1263                                 break;
1264
1265                         case IP_RECVDSTADDR:
1266                                 OPTSET(INP_RECVDSTADDR);
1267                                 break;
1268
1269                         case IP_ORIGDSTADDR:
1270                                 OPTSET2(INP_ORIGDSTADDR, optval);
1271                                 break;
1272
1273                         case IP_RECVTTL:
1274                                 OPTSET(INP_RECVTTL);
1275                                 break;
1276
1277                         case IP_RECVIF:
1278                                 OPTSET(INP_RECVIF);
1279                                 break;
1280
1281                         case IP_ONESBCAST:
1282                                 OPTSET(INP_ONESBCAST);
1283                                 break;
1284                         case IP_DONTFRAG:
1285                                 OPTSET(INP_DONTFRAG);
1286                                 break;
1287                         case IP_BINDANY:
1288                                 OPTSET(INP_BINDANY);
1289                                 break;
1290                         case IP_RECVTOS:
1291                                 OPTSET(INP_RECVTOS);
1292                                 break;
1293                         case IP_BINDMULTI:
1294                                 OPTSET2(INP_BINDMULTI, optval);
1295                                 break;
1296                         case IP_RECVFLOWID:
1297                                 OPTSET2(INP_RECVFLOWID, optval);
1298                                 break;
1299 #ifdef  RSS
1300                         case IP_RSS_LISTEN_BUCKET:
1301                                 if ((optval >= 0) &&
1302                                     (optval < rss_getnumbuckets())) {
1303                                         inp->inp_rss_listen_bucket = optval;
1304                                         OPTSET2(INP_RSS_BUCKET_SET, 1);
1305                                 } else {
1306                                         error = EINVAL;
1307                                 }
1308                                 break;
1309                         case IP_RECVRSSBUCKETID:
1310                                 OPTSET2(INP_RECVRSSBUCKETID, optval);
1311                                 break;
1312 #endif
1313                         }
1314                         break;
1315 #undef OPTSET
1316 #undef OPTSET2
1317
1318                 /*
1319                  * Multicast socket options are processed by the in_mcast
1320                  * module.
1321                  */
1322                 case IP_MULTICAST_IF:
1323                 case IP_MULTICAST_VIF:
1324                 case IP_MULTICAST_TTL:
1325                 case IP_MULTICAST_LOOP:
1326                 case IP_ADD_MEMBERSHIP:
1327                 case IP_DROP_MEMBERSHIP:
1328                 case IP_ADD_SOURCE_MEMBERSHIP:
1329                 case IP_DROP_SOURCE_MEMBERSHIP:
1330                 case IP_BLOCK_SOURCE:
1331                 case IP_UNBLOCK_SOURCE:
1332                 case IP_MSFILTER:
1333                 case MCAST_JOIN_GROUP:
1334                 case MCAST_LEAVE_GROUP:
1335                 case MCAST_JOIN_SOURCE_GROUP:
1336                 case MCAST_LEAVE_SOURCE_GROUP:
1337                 case MCAST_BLOCK_SOURCE:
1338                 case MCAST_UNBLOCK_SOURCE:
1339                         error = inp_setmoptions(inp, sopt);
1340                         break;
1341
1342                 case IP_PORTRANGE:
1343                         error = sooptcopyin(sopt, &optval, sizeof optval,
1344                                             sizeof optval);
1345                         if (error)
1346                                 break;
1347
1348                         INP_WLOCK(inp);
1349                         switch (optval) {
1350                         case IP_PORTRANGE_DEFAULT:
1351                                 inp->inp_flags &= ~(INP_LOWPORT);
1352                                 inp->inp_flags &= ~(INP_HIGHPORT);
1353                                 break;
1354
1355                         case IP_PORTRANGE_HIGH:
1356                                 inp->inp_flags &= ~(INP_LOWPORT);
1357                                 inp->inp_flags |= INP_HIGHPORT;
1358                                 break;
1359
1360                         case IP_PORTRANGE_LOW:
1361                                 inp->inp_flags &= ~(INP_HIGHPORT);
1362                                 inp->inp_flags |= INP_LOWPORT;
1363                                 break;
1364
1365                         default:
1366                                 error = EINVAL;
1367                                 break;
1368                         }
1369                         INP_WUNLOCK(inp);
1370                         break;
1371
1372 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1373                 case IP_IPSEC_POLICY:
1374                         if (IPSEC_ENABLED(ipv4)) {
1375                                 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1376                                 break;
1377                         }
1378                         /* FALLTHROUGH */
1379 #endif /* IPSEC */
1380
1381                 default:
1382                         error = ENOPROTOOPT;
1383                         break;
1384                 }
1385                 break;
1386
1387         case SOPT_GET:
1388                 switch (sopt->sopt_name) {
1389                 case IP_OPTIONS:
1390                 case IP_RETOPTS:
1391                         INP_RLOCK(inp);
1392                         if (inp->inp_options) {
1393                                 struct mbuf *options;
1394
1395                                 options = m_copym(inp->inp_options, 0,
1396                                     M_COPYALL, M_NOWAIT);
1397                                 INP_RUNLOCK(inp);
1398                                 if (options != NULL) {
1399                                         error = sooptcopyout(sopt,
1400                                                              mtod(options, char *),
1401                                                              options->m_len);
1402                                         m_freem(options);
1403                                 } else
1404                                         error = ENOMEM;
1405                         } else {
1406                                 INP_RUNLOCK(inp);
1407                                 sopt->sopt_valsize = 0;
1408                         }
1409                         break;
1410
1411                 case IP_TOS:
1412                 case IP_TTL:
1413                 case IP_MINTTL:
1414                 case IP_RECVOPTS:
1415                 case IP_RECVRETOPTS:
1416                 case IP_ORIGDSTADDR:
1417                 case IP_RECVDSTADDR:
1418                 case IP_RECVTTL:
1419                 case IP_RECVIF:
1420                 case IP_PORTRANGE:
1421                 case IP_ONESBCAST:
1422                 case IP_DONTFRAG:
1423                 case IP_BINDANY:
1424                 case IP_RECVTOS:
1425                 case IP_BINDMULTI:
1426                 case IP_FLOWID:
1427                 case IP_FLOWTYPE:
1428                 case IP_RECVFLOWID:
1429 #ifdef  RSS
1430                 case IP_RSSBUCKETID:
1431                 case IP_RECVRSSBUCKETID:
1432 #endif
1433                         switch (sopt->sopt_name) {
1434                         case IP_TOS:
1435                                 optval = inp->inp_ip_tos;
1436                                 break;
1437
1438                         case IP_TTL:
1439                                 optval = inp->inp_ip_ttl;
1440                                 break;
1441
1442                         case IP_MINTTL:
1443                                 optval = inp->inp_ip_minttl;
1444                                 break;
1445
1446 #define OPTBIT(bit)     (inp->inp_flags & bit ? 1 : 0)
1447 #define OPTBIT2(bit)    (inp->inp_flags2 & bit ? 1 : 0)
1448
1449                         case IP_RECVOPTS:
1450                                 optval = OPTBIT(INP_RECVOPTS);
1451                                 break;
1452
1453                         case IP_RECVRETOPTS:
1454                                 optval = OPTBIT(INP_RECVRETOPTS);
1455                                 break;
1456
1457                         case IP_RECVDSTADDR:
1458                                 optval = OPTBIT(INP_RECVDSTADDR);
1459                                 break;
1460
1461                         case IP_ORIGDSTADDR:
1462                                 optval = OPTBIT2(INP_ORIGDSTADDR);
1463                                 break;
1464
1465                         case IP_RECVTTL:
1466                                 optval = OPTBIT(INP_RECVTTL);
1467                                 break;
1468
1469                         case IP_RECVIF:
1470                                 optval = OPTBIT(INP_RECVIF);
1471                                 break;
1472
1473                         case IP_PORTRANGE:
1474                                 if (inp->inp_flags & INP_HIGHPORT)
1475                                         optval = IP_PORTRANGE_HIGH;
1476                                 else if (inp->inp_flags & INP_LOWPORT)
1477                                         optval = IP_PORTRANGE_LOW;
1478                                 else
1479                                         optval = 0;
1480                                 break;
1481
1482                         case IP_ONESBCAST:
1483                                 optval = OPTBIT(INP_ONESBCAST);
1484                                 break;
1485                         case IP_DONTFRAG:
1486                                 optval = OPTBIT(INP_DONTFRAG);
1487                                 break;
1488                         case IP_BINDANY:
1489                                 optval = OPTBIT(INP_BINDANY);
1490                                 break;
1491                         case IP_RECVTOS:
1492                                 optval = OPTBIT(INP_RECVTOS);
1493                                 break;
1494                         case IP_FLOWID:
1495                                 optval = inp->inp_flowid;
1496                                 break;
1497                         case IP_FLOWTYPE:
1498                                 optval = inp->inp_flowtype;
1499                                 break;
1500                         case IP_RECVFLOWID:
1501                                 optval = OPTBIT2(INP_RECVFLOWID);
1502                                 break;
1503 #ifdef  RSS
1504                         case IP_RSSBUCKETID:
1505                                 retval = rss_hash2bucket(inp->inp_flowid,
1506                                     inp->inp_flowtype,
1507                                     &rss_bucket);
1508                                 if (retval == 0)
1509                                         optval = rss_bucket;
1510                                 else
1511                                         error = EINVAL;
1512                                 break;
1513                         case IP_RECVRSSBUCKETID:
1514                                 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1515                                 break;
1516 #endif
1517                         case IP_BINDMULTI:
1518                                 optval = OPTBIT2(INP_BINDMULTI);
1519                                 break;
1520                         }
1521                         error = sooptcopyout(sopt, &optval, sizeof optval);
1522                         break;
1523
1524                 /*
1525                  * Multicast socket options are processed by the in_mcast
1526                  * module.
1527                  */
1528                 case IP_MULTICAST_IF:
1529                 case IP_MULTICAST_VIF:
1530                 case IP_MULTICAST_TTL:
1531                 case IP_MULTICAST_LOOP:
1532                 case IP_MSFILTER:
1533                         error = inp_getmoptions(inp, sopt);
1534                         break;
1535
1536 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1537                 case IP_IPSEC_POLICY:
1538                         if (IPSEC_ENABLED(ipv4)) {
1539                                 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1540                                 break;
1541                         }
1542                         /* FALLTHROUGH */
1543 #endif /* IPSEC */
1544
1545                 default:
1546                         error = ENOPROTOOPT;
1547                         break;
1548                 }
1549                 break;
1550         }
1551         return (error);
1552 }
1553
1554 /*
1555  * Routine called from ip_output() to loop back a copy of an IP multicast
1556  * packet to the input queue of a specified interface.  Note that this
1557  * calls the output routine of the loopback "driver", but with an interface
1558  * pointer that might NOT be a loopback interface -- evil, but easier than
1559  * replicating that code here.
1560  */
1561 static void
1562 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1563 {
1564         struct ip *ip;
1565         struct mbuf *copym;
1566
1567         /*
1568          * Make a deep copy of the packet because we're going to
1569          * modify the pack in order to generate checksums.
1570          */
1571         copym = m_dup(m, M_NOWAIT);
1572         if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1573                 copym = m_pullup(copym, hlen);
1574         if (copym != NULL) {
1575                 /* If needed, compute the checksum and mark it as valid. */
1576                 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1577                         in_delayed_cksum(copym);
1578                         copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1579                         copym->m_pkthdr.csum_flags |=
1580                             CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1581                         copym->m_pkthdr.csum_data = 0xffff;
1582                 }
1583                 /*
1584                  * We don't bother to fragment if the IP length is greater
1585                  * than the interface's MTU.  Can this possibly matter?
1586                  */
1587                 ip = mtod(copym, struct ip *);
1588                 ip->ip_sum = 0;
1589                 ip->ip_sum = in_cksum(copym, hlen);
1590                 if_simloop(ifp, copym, AF_INET, 0);
1591         }
1592 }