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_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/ktls.h>
  50 #include <sys/lock.h>
  51 #include <sys/malloc.h>
  52 #include <sys/mbuf.h>
  53 #include <sys/priv.h>
  54 #include <sys/proc.h>
  55 #include <sys/protosw.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_private.h>
  65 #include <net/if_vlan_var.h>
  66 #include <net/if_llatbl.h>
  67 #include <net/ethernet.h>
  68 #include <net/netisr.h>
  69 #include <net/pfil.h>
  70 #include <net/route.h>
  71 #include <net/route/nhop.h>
  72 #include <net/rss_config.h>
  73 #include <net/vnet.h>
  74
  75 #include <netinet/in.h>
  76 #include <netinet/in_fib.h>
  77 #include <netinet/in_kdtrace.h>
  78 #include <netinet/in_systm.h>
  79 #include <netinet/ip.h>
  80 #include <netinet/in_fib.h>
  81 #include <netinet/in_pcb.h>
  82 #include <netinet/in_rss.h>
  83 #include <netinet/in_var.h>
  84 #include <netinet/ip_var.h>
  85 #include <netinet/ip_options.h>
  86
  87 #include <netinet/udp.h>
  88 #include <netinet/udp_var.h>
  89
  90 #if defined(SCTP) || defined(SCTP_SUPPORT)
  91 #include <netinet/sctp.h>
  92 #include <netinet/sctp_crc32.h>
  93 #endif
  94
  95 #include <netipsec/ipsec_support.h>
  96
  97 #include <machine/in_cksum.h>
  98
  99 #include <security/mac/mac_framework.h>
 100
 101 #ifdef MBUF_STRESS_TEST
 102 static int mbuf_frag_size = 0;
 103 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
 104         &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
 105 #endif
 106
 107 static void     ip_mloopback(struct ifnet *, const struct mbuf *, int);
 108
 109 extern int in_mcast_loop;
 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         m = *mp;
 122         ip = mtod(m, struct ip *);
 123
 124         /* Run through list of hooks for output packets. */
 125         odst.s_addr = ip->ip_dst.s_addr;
 126         switch (pfil_run_hooks(V_inet_pfil_head, mp, ifp, pflags, inp)) {
 127         case PFIL_DROPPED:
 128                 *error = EACCES;
 129                 /* FALLTHROUGH */
 130         case PFIL_CONSUMED:
 131                 return 1; /* Finished */
 132         case PFIL_PASS:
 133                 *error = 0;
 134         }
 135         m = *mp;
 136         ip = mtod(m, struct ip *);
 137
 138         /* See if destination IP address was changed by packet filter. */
 139         if (odst.s_addr != ip->ip_dst.s_addr) {
 140                 m->m_flags |= M_SKIP_FIREWALL;
 141                 /* If destination is now ourself drop to ip_input(). */
 142                 if (in_localip(ip->ip_dst)) {
 143                         m->m_flags |= M_FASTFWD_OURS;
 144                         if (m->m_pkthdr.rcvif == NULL)
 145                                 m->m_pkthdr.rcvif = V_loif;
 146                         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 147                                 m->m_pkthdr.csum_flags |=
 148                                         CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 149                                 m->m_pkthdr.csum_data = 0xffff;
 150                         }
 151                         m->m_pkthdr.csum_flags |=
 152                                 CSUM_IP_CHECKED | CSUM_IP_VALID;
 153 #if defined(SCTP) || defined(SCTP_SUPPORT)
 154                         if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 155                                 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 156 #endif
 157                         *error = netisr_queue(NETISR_IP, m);
 158                         return 1; /* Finished */
 159                 }
 160
 161                 bzero(dst, sizeof(*dst));
 162                 dst->sin_family = AF_INET;
 163                 dst->sin_len = sizeof(*dst);
 164                 dst->sin_addr = ip->ip_dst;
 165
 166                 return -1; /* Reloop */
 167         }
 168         /* See if fib was changed by packet filter. */
 169         if ((*fibnum) != M_GETFIB(m)) {
 170                 m->m_flags |= M_SKIP_FIREWALL;
 171                 *fibnum = M_GETFIB(m);
 172                 return -1; /* Reloop for FIB change */
 173         }
 174
 175         /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
 176         if (m->m_flags & M_FASTFWD_OURS) {
 177                 if (m->m_pkthdr.rcvif == NULL)
 178                         m->m_pkthdr.rcvif = V_loif;
 179                 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 180                         m->m_pkthdr.csum_flags |=
 181                                 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
 182                         m->m_pkthdr.csum_data = 0xffff;
 183                 }
 184 #if defined(SCTP) || defined(SCTP_SUPPORT)
 185                 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
 186                         m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
 187 #endif
 188                 m->m_pkthdr.csum_flags |=
 189                         CSUM_IP_CHECKED | CSUM_IP_VALID;
 190
 191                 *error = netisr_queue(NETISR_IP, m);
 192                 return 1; /* Finished */
 193         }
 194         /* Or forward to some other address? */
 195         if ((m->m_flags & M_IP_NEXTHOP) &&
 196             ((fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL)) {
 197                 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
 198                 m->m_flags |= M_SKIP_FIREWALL;
 199                 m->m_flags &= ~M_IP_NEXTHOP;
 200                 m_tag_delete(m, fwd_tag);
 201
 202                 return -1; /* Reloop for CHANGE of dst */
 203         }
 204
 205         return 0;
 206 }
 207
 208 static int
 209 ip_output_send(struct inpcb *inp, struct ifnet *ifp, struct mbuf *m,
 210     const struct sockaddr *gw, struct route *ro, bool stamp_tag)
 211 {
 212 #ifdef KERN_TLS
 213         struct ktls_session *tls = NULL;
 214 #endif
 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 KERN_TLS
 222         /*
 223          * If this is an unencrypted TLS record, save a reference to
 224          * the record.  This local reference is used to call
 225          * ktls_output_eagain after the mbuf has been freed (thus
 226          * dropping the mbuf's reference) in if_output.
 227          */
 228         if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
 229                 tls = ktls_hold(m->m_next->m_epg_tls);
 230                 mst = tls->snd_tag;
 231
 232                 /*
 233                  * If a TLS session doesn't have a valid tag, it must
 234                  * have had an earlier ifp mismatch, so drop this
 235                  * packet.
 236                  */
 237                 if (mst == NULL) {
 238                         m_freem(m);
 239                         error = EAGAIN;
 240                         goto done;
 241                 }
 242                 /*
 243                  * Always stamp tags that include NIC ktls.
 244                  */
 245                 stamp_tag = true;
 246         }
 247 #endif
 248 #ifdef RATELIMIT
 249         if (inp != NULL && mst == NULL) {
 250                 if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
 251                     (inp->inp_snd_tag != NULL &&
 252                     inp->inp_snd_tag->ifp != ifp))
 253                         in_pcboutput_txrtlmt(inp, ifp, m);
 254
 255                 if (inp->inp_snd_tag != NULL)
 256                         mst = inp->inp_snd_tag;
 257         }
 258 #endif
 259         if (stamp_tag && mst != NULL) {
 260                 KASSERT(m->m_pkthdr.rcvif == NULL,
 261                     ("trying to add a send tag to a forwarded packet"));
 262                 if (mst->ifp != ifp) {
 263                         m_freem(m);
 264                         error = EAGAIN;
 265                         goto done;
 266                 }
 267
 268                 /* stamp send tag on mbuf */
 269                 m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
 270                 m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
 271         }
 272
 273         error = (*ifp->if_output)(ifp, m, gw, ro);
 274
 275 done:
 276         /* Check for route change invalidating send tags. */
 277 #ifdef KERN_TLS
 278         if (tls != NULL) {
 279                 if (error == EAGAIN)
 280                         error = ktls_output_eagain(inp, tls);
 281                 ktls_free(tls);
 282         }
 283 #endif
 284 #ifdef RATELIMIT
 285         if (error == EAGAIN)
 286                 in_pcboutput_eagain(inp);
 287 #endif
 288         return (error);
 289 }
 290
 291 /* rte<>ro_flags translation */
 292 static inline void
 293 rt_update_ro_flags(struct route *ro, const struct nhop_object *nh)
 294 {
 295         int nh_flags = nh->nh_flags;
 296
 297         ro->ro_flags &= ~ (RT_REJECT|RT_BLACKHOLE|RT_HAS_GW);
 298
 299         ro->ro_flags |= (nh_flags & NHF_REJECT) ? RT_REJECT : 0;
 300         ro->ro_flags |= (nh_flags & NHF_BLACKHOLE) ? RT_BLACKHOLE : 0;
 301         ro->ro_flags |= (nh_flags & NHF_GATEWAY) ? RT_HAS_GW : 0;
 302 }
 303
 304 /*
 305  * IP output.  The packet in mbuf chain m contains a skeletal IP
 306  * header (with len, off, ttl, proto, tos, src, dst).
 307  * The mbuf chain containing the packet will be freed.
 308  * The mbuf opt, if present, will not be freed.
 309  * If route ro is present and has ro_rt initialized, route lookup would be
 310  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
 311  * then result of route lookup is stored in ro->ro_rt.
 312  *
 313  * In the IP forwarding case, the packet will arrive with options already
 314  * inserted, so must have a NULL opt pointer.
 315  */
 316 int
 317 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
 318     struct ip_moptions *imo, struct inpcb *inp)
 319 {
 320         struct ip *ip;
 321         struct ifnet *ifp = NULL;       /* keep compiler happy */
 322         struct mbuf *m0;
 323         int hlen = sizeof (struct ip);
 324         int mtu = 0;
 325         int error = 0;
 326         int vlan_pcp = -1;
 327         struct sockaddr_in *dst;
 328         const struct sockaddr *gw;
 329         struct in_ifaddr *ia = NULL;
 330         struct in_addr src;
 331         int isbroadcast;
 332         uint16_t ip_len, ip_off;
 333         struct route iproute;
 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                 if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
 350                         vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
 351                             INP_2PCP_SHIFT;
 352 #ifdef NUMA
 353                 m->m_pkthdr.numa_domain = inp->inp_numa_domain;
 354 #endif
 355         }
 356
 357         if (opt) {
 358                 int len = 0;
 359                 m = ip_insertoptions(m, opt, &len);
 360                 if (len != 0)
 361                         hlen = len; /* ip->ip_hl is updated above */
 362         }
 363         ip = mtod(m, struct ip *);
 364         ip_len = ntohs(ip->ip_len);
 365         ip_off = ntohs(ip->ip_off);
 366
 367         if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
 368                 ip->ip_v = IPVERSION;
 369                 ip->ip_hl = hlen >> 2;
 370                 ip_fillid(ip);
 371         } else {
 372                 /* Header already set, fetch hlen from there */
 373                 hlen = ip->ip_hl << 2;
 374         }
 375         if ((flags & IP_FORWARDING) == 0)
 376                 IPSTAT_INC(ips_localout);
 377
 378         /*
 379          * dst/gw handling:
 380          *
 381          * gw is readonly but can point either to dst OR rt_gateway,
 382          * therefore we need restore gw if we're redoing lookup.
 383          */
 384         fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
 385         if (ro == NULL) {
 386                 ro = &iproute;
 387                 bzero(ro, sizeof (*ro));
 388         }
 389         dst = (struct sockaddr_in *)&ro->ro_dst;
 390         if (ro->ro_nh == NULL) {
 391                 dst->sin_family = AF_INET;
 392                 dst->sin_len = sizeof(*dst);
 393                 dst->sin_addr = ip->ip_dst;
 394         }
 395         gw = (const struct sockaddr *)dst;
 396 again:
 397         /*
 398          * Validate route against routing table additions;
 399          * a better/more specific route might have been added.
 400          */
 401         if (inp != NULL && ro->ro_nh != NULL)
 402                 NH_VALIDATE(ro, &inp->inp_rt_cookie, fibnum);
 403         /*
 404          * If there is a cached route,
 405          * check that it is to the same destination
 406          * and is still up.  If not, free it and try again.
 407          * The address family should also be checked in case of sharing the
 408          * cache with IPv6.
 409          * Also check whether routing cache needs invalidation.
 410          */
 411         if (ro->ro_nh != NULL &&
 412             ((!NH_IS_VALID(ro->ro_nh)) || dst->sin_family != AF_INET ||
 413             dst->sin_addr.s_addr != ip->ip_dst.s_addr))
 414                 RO_INVALIDATE_CACHE(ro);
 415         ia = NULL;
 416         /*
 417          * If routing to interface only, short circuit routing lookup.
 418          * The use of an all-ones broadcast address implies this; an
 419          * interface is specified by the broadcast address of an interface,
 420          * or the destination address of a ptp interface.
 421          */
 422         if (flags & IP_SENDONES) {
 423                 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst),
 424                                                       M_GETFIB(m)))) == NULL &&
 425                     (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 426                                                     M_GETFIB(m)))) == NULL) {
 427                         IPSTAT_INC(ips_noroute);
 428                         error = ENETUNREACH;
 429                         goto bad;
 430                 }
 431                 ip->ip_dst.s_addr = INADDR_BROADCAST;
 432                 dst->sin_addr = ip->ip_dst;
 433                 ifp = ia->ia_ifp;
 434                 mtu = ifp->if_mtu;
 435                 ip->ip_ttl = 1;
 436                 isbroadcast = 1;
 437                 src = IA_SIN(ia)->sin_addr;
 438         } else if (flags & IP_ROUTETOIF) {
 439                 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
 440                                                     M_GETFIB(m)))) == NULL &&
 441                     (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
 442                                                 M_GETFIB(m)))) == NULL) {
 443                         IPSTAT_INC(ips_noroute);
 444                         error = ENETUNREACH;
 445                         goto bad;
 446                 }
 447                 ifp = ia->ia_ifp;
 448                 mtu = ifp->if_mtu;
 449                 ip->ip_ttl = 1;
 450                 isbroadcast = ifp->if_flags & IFF_BROADCAST ?
 451                     in_ifaddr_broadcast(dst->sin_addr, ia) : 0;
 452                 src = IA_SIN(ia)->sin_addr;
 453         } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
 454             imo != NULL && imo->imo_multicast_ifp != NULL) {
 455                 /*
 456                  * Bypass the normal routing lookup for multicast
 457                  * packets if the interface is specified.
 458                  */
 459                 ifp = imo->imo_multicast_ifp;
 460                 mtu = ifp->if_mtu;
 461                 IFP_TO_IA(ifp, ia);
 462                 isbroadcast = 0;        /* fool gcc */
 463                 /* Interface may have no addresses. */
 464                 if (ia != NULL)
 465                         src = IA_SIN(ia)->sin_addr;
 466                 else
 467                         src.s_addr = INADDR_ANY;
 468         } else if (ro != &iproute) {
 469                 if (ro->ro_nh == NULL) {
 470                         /*
 471                          * We want to do any cloning requested by the link
 472                          * layer, as this is probably required in all cases
 473                          * for correct operation (as it is for ARP).
 474                          */
 475                         uint32_t flowid;
 476                         flowid = m->m_pkthdr.flowid;
 477                         ro->ro_nh = fib4_lookup(fibnum, dst->sin_addr, 0,
 478                             NHR_REF, flowid);
 479
 480                         if (ro->ro_nh == NULL || (!NH_IS_VALID(ro->ro_nh))) {
 481 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 482                                 /*
 483                                  * There is no route for this packet, but it is
 484                                  * possible that a matching SPD entry exists.
 485                                  */
 486                                 no_route_but_check_spd = 1;
 487                                 goto sendit;
 488 #endif
 489                                 IPSTAT_INC(ips_noroute);
 490                                 error = EHOSTUNREACH;
 491                                 goto bad;
 492                         }
 493                 }
 494                 struct nhop_object *nh = ro->ro_nh;
 495
 496                 ia = ifatoia(nh->nh_ifa);
 497                 ifp = nh->nh_ifp;
 498                 counter_u64_add(nh->nh_pksent, 1);
 499                 rt_update_ro_flags(ro, nh);
 500                 if (nh->nh_flags & NHF_GATEWAY)
 501                         gw = &nh->gw_sa;
 502                 if (nh->nh_flags & NHF_HOST)
 503                         isbroadcast = (nh->nh_flags & NHF_BROADCAST);
 504                 else if ((ifp->if_flags & IFF_BROADCAST) && (gw->sa_family == AF_INET))
 505                         isbroadcast = in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia);
 506                 else
 507                         isbroadcast = 0;
 508                 mtu = nh->nh_mtu;
 509                 src = IA_SIN(ia)->sin_addr;
 510         } else {
 511                 struct nhop_object *nh;
 512
 513                 nh = fib4_lookup(M_GETFIB(m), dst->sin_addr, 0, NHR_NONE,
 514                     m->m_pkthdr.flowid);
 515                 if (nh == NULL) {
 516 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 517                         /*
 518                          * There is no route for this packet, but it is
 519                          * possible that a matching SPD entry exists.
 520                          */
 521                         no_route_but_check_spd = 1;
 522                         goto sendit;
 523 #endif
 524                         IPSTAT_INC(ips_noroute);
 525                         error = EHOSTUNREACH;
 526                         goto bad;
 527                 }
 528                 ifp = nh->nh_ifp;
 529                 mtu = nh->nh_mtu;
 530                 rt_update_ro_flags(ro, nh);
 531                 if (nh->nh_flags & NHF_GATEWAY)
 532                         gw = &nh->gw_sa;
 533                 ia = ifatoia(nh->nh_ifa);
 534                 src = IA_SIN(ia)->sin_addr;
 535                 isbroadcast = (((nh->nh_flags & (NHF_HOST | NHF_BROADCAST)) ==
 536                     (NHF_HOST | NHF_BROADCAST)) ||
 537                     ((ifp->if_flags & IFF_BROADCAST) &&
 538                     (gw->sa_family == AF_INET) &&
 539                     in_ifaddr_broadcast(((const struct sockaddr_in *)gw)->sin_addr, ia)));
 540         }
 541
 542         /* Catch a possible divide by zero later. */
 543         KASSERT(mtu > 0, ("%s: mtu %d <= 0, ro=%p (nh_flags=0x%08x) ifp=%p",
 544             __func__, mtu, ro,
 545             (ro != NULL && ro->ro_nh != NULL) ? ro->ro_nh->nh_flags : 0, ifp));
 546
 547         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
 548                 m->m_flags |= M_MCAST;
 549                 /*
 550                  * IP destination address is multicast.  Make sure "gw"
 551                  * still points to the address in "ro".  (It may have been
 552                  * changed to point to a gateway address, above.)
 553                  */
 554                 gw = (const struct sockaddr *)dst;
 555                 /*
 556                  * See if the caller provided any multicast options
 557                  */
 558                 if (imo != NULL) {
 559                         ip->ip_ttl = imo->imo_multicast_ttl;
 560                         if (imo->imo_multicast_vif != -1)
 561                                 ip->ip_src.s_addr =
 562                                     ip_mcast_src ?
 563                                     ip_mcast_src(imo->imo_multicast_vif) :
 564                                     INADDR_ANY;
 565                 } else
 566                         ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
 567                 /*
 568                  * Confirm that the outgoing interface supports multicast.
 569                  */
 570                 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
 571                         if ((ifp->if_flags & IFF_MULTICAST) == 0) {
 572                                 IPSTAT_INC(ips_noroute);
 573                                 error = ENETUNREACH;
 574                                 goto bad;
 575                         }
 576                 }
 577                 /*
 578                  * If source address not specified yet, use address
 579                  * of outgoing interface.
 580                  */
 581                 if (ip->ip_src.s_addr == INADDR_ANY)
 582                         ip->ip_src = src;
 583
 584                 if ((imo == NULL && in_mcast_loop) ||
 585                     (imo && imo->imo_multicast_loop)) {
 586                         /*
 587                          * Loop back multicast datagram if not expressly
 588                          * forbidden to do so, even if we are not a member
 589                          * of the group; ip_input() will filter it later,
 590                          * thus deferring a hash lookup and mutex acquisition
 591                          * at the expense of a cheap copy using m_copym().
 592                          */
 593                         ip_mloopback(ifp, m, hlen);
 594                 } else {
 595                         /*
 596                          * If we are acting as a multicast router, perform
 597                          * multicast forwarding as if the packet had just
 598                          * arrived on the interface to which we are about
 599                          * to send.  The multicast forwarding function
 600                          * recursively calls this function, using the
 601                          * IP_FORWARDING flag to prevent infinite recursion.
 602                          *
 603                          * Multicasts that are looped back by ip_mloopback(),
 604                          * above, will be forwarded by the ip_input() routine,
 605                          * if necessary.
 606                          */
 607                         if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
 608                                 /*
 609                                  * If rsvp daemon is not running, do not
 610                                  * set ip_moptions. This ensures that the packet
 611                                  * is multicast and not just sent down one link
 612                                  * as prescribed by rsvpd.
 613                                  */
 614                                 if (!V_rsvp_on)
 615                                         imo = NULL;
 616                                 if (ip_mforward &&
 617                                     ip_mforward(ip, ifp, m, imo) != 0) {
 618                                         m_freem(m);
 619                                         goto done;
 620                                 }
 621                         }
 622                 }
 623
 624                 /*
 625                  * Multicasts with a time-to-live of zero may be looped-
 626                  * back, above, but must not be transmitted on a network.
 627                  * Also, multicasts addressed to the loopback interface
 628                  * are not sent -- the above call to ip_mloopback() will
 629                  * loop back a copy. ip_input() will drop the copy if
 630                  * this host does not belong to the destination group on
 631                  * the loopback interface.
 632                  */
 633                 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
 634                         m_freem(m);
 635                         goto done;
 636                 }
 637
 638                 goto sendit;
 639         }
 640
 641         /*
 642          * If the source address is not specified yet, use the address
 643          * of the outoing interface.
 644          */
 645         if (ip->ip_src.s_addr == INADDR_ANY)
 646                 ip->ip_src = src;
 647
 648         /*
 649          * Look for broadcast address and
 650          * verify user is allowed to send
 651          * such a packet.
 652          */
 653         if (isbroadcast) {
 654                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 655                         error = EADDRNOTAVAIL;
 656                         goto bad;
 657                 }
 658                 if ((flags & IP_ALLOWBROADCAST) == 0) {
 659                         error = EACCES;
 660                         goto bad;
 661                 }
 662                 /* don't allow broadcast messages to be fragmented */
 663                 if (ip_len > mtu) {
 664                         error = EMSGSIZE;
 665                         goto bad;
 666                 }
 667                 m->m_flags |= M_BCAST;
 668         } else {
 669                 m->m_flags &= ~M_BCAST;
 670         }
 671
 672 sendit:
 673 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
 674         if (IPSEC_ENABLED(ipv4)) {
 675                 if ((error = IPSEC_OUTPUT(ipv4, m, inp)) != 0) {
 676                         if (error == EINPROGRESS)
 677                                 error = 0;
 678                         goto done;
 679                 }
 680         }
 681         /*
 682          * Check if there was a route for this packet; return error if not.
 683          */
 684         if (no_route_but_check_spd) {
 685                 IPSTAT_INC(ips_noroute);
 686                 error = EHOSTUNREACH;
 687                 goto bad;
 688         }
 689         /* Update variables that are affected by ipsec4_output(). */
 690         ip = mtod(m, struct ip *);
 691         hlen = ip->ip_hl << 2;
 692 #endif /* IPSEC */
 693
 694         /* Jump over all PFIL processing if hooks are not active. */
 695         if (PFIL_HOOKED_OUT(V_inet_pfil_head)) {
 696                 switch (ip_output_pfil(&m, ifp, flags, inp, dst, &fibnum,
 697                     &error)) {
 698                 case 1: /* Finished */
 699                         goto done;
 700
 701                 case 0: /* Continue normally */
 702                         ip = mtod(m, struct ip *);
 703                         break;
 704
 705                 case -1: /* Need to try again */
 706                         /* Reset everything for a new round */
 707                         if (ro != NULL) {
 708                                 RO_NHFREE(ro);
 709                                 ro->ro_prepend = NULL;
 710                         }
 711                         gw = (const struct sockaddr *)dst;
 712                         ip = mtod(m, struct ip *);
 713                         goto again;
 714                 }
 715         }
 716
 717         if (vlan_pcp > -1)
 718                 EVL_APPLY_PRI(m, vlan_pcp);
 719
 720         /* IN_LOOPBACK must not appear on the wire - RFC1122. */
 721         if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
 722             IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
 723                 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
 724                         IPSTAT_INC(ips_badaddr);
 725                         error = EADDRNOTAVAIL;
 726                         goto bad;
 727                 }
 728         }
 729
 730         /* Ensure the packet data is mapped if the interface requires it. */
 731         if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
 732                 m = mb_unmapped_to_ext(m);
 733                 if (m == NULL) {
 734                         IPSTAT_INC(ips_odropped);
 735                         error = ENOBUFS;
 736                         goto bad;
 737                 }
 738         }
 739
 740         m->m_pkthdr.csum_flags |= CSUM_IP;
 741         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
 742                 in_delayed_cksum(m);
 743                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 744         }
 745 #if defined(SCTP) || defined(SCTP_SUPPORT)
 746         if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
 747                 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
 748                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 749         }
 750 #endif
 751
 752         /*
 753          * If small enough for interface, or the interface will take
 754          * care of the fragmentation for us, we can just send directly.
 755          * Note that if_vxlan could have requested TSO even though the outer
 756          * frame is UDP.  It is correct to not fragment such datagrams and
 757          * instead just pass them on to the driver.
 758          */
 759         if (ip_len <= mtu ||
 760             (m->m_pkthdr.csum_flags & ifp->if_hwassist &
 761             (CSUM_TSO | CSUM_INNER_TSO)) != 0) {
 762                 ip->ip_sum = 0;
 763                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
 764                         ip->ip_sum = in_cksum(m, hlen);
 765                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
 766                 }
 767
 768                 /*
 769                  * Record statistics for this interface address.
 770                  * With CSUM_TSO the byte/packet count will be slightly
 771                  * incorrect because we count the IP+TCP headers only
 772                  * once instead of for every generated packet.
 773                  */
 774                 if (!(flags & IP_FORWARDING) && ia) {
 775                         if (m->m_pkthdr.csum_flags &
 776                             (CSUM_TSO | CSUM_INNER_TSO))
 777                                 counter_u64_add(ia->ia_ifa.ifa_opackets,
 778                                     m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
 779                         else
 780                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 781
 782                         counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
 783                 }
 784 #ifdef MBUF_STRESS_TEST
 785                 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
 786                         m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
 787 #endif
 788                 /*
 789                  * Reset layer specific mbuf flags
 790                  * to avoid confusing lower layers.
 791                  */
 792                 m_clrprotoflags(m);
 793                 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
 794                 error = ip_output_send(inp, ifp, m, gw, ro,
 795                     (flags & IP_NO_SND_TAG_RL) ? false : true);
 796                 goto done;
 797         }
 798
 799         /* Balk when DF bit is set or the interface didn't support TSO. */
 800         if ((ip_off & IP_DF) ||
 801             (m->m_pkthdr.csum_flags & (CSUM_TSO | CSUM_INNER_TSO))) {
 802                 error = EMSGSIZE;
 803                 IPSTAT_INC(ips_cantfrag);
 804                 goto bad;
 805         }
 806
 807         /*
 808          * Too large for interface; fragment if possible. If successful,
 809          * on return, m will point to a list of packets to be sent.
 810          */
 811         error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
 812         if (error)
 813                 goto bad;
 814         for (; m; m = m0) {
 815                 m0 = m->m_nextpkt;
 816                 m->m_nextpkt = 0;
 817                 if (error == 0) {
 818                         /* Record statistics for this interface address. */
 819                         if (ia != NULL) {
 820                                 counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
 821                                 counter_u64_add(ia->ia_ifa.ifa_obytes,
 822                                     m->m_pkthdr.len);
 823                         }
 824                         /*
 825                          * Reset layer specific mbuf flags
 826                          * to avoid confusing upper layers.
 827                          */
 828                         m_clrprotoflags(m);
 829
 830                         IP_PROBE(send, NULL, NULL, mtod(m, struct ip *), ifp,
 831                             mtod(m, struct ip *), NULL);
 832                         error = ip_output_send(inp, ifp, m, gw, ro, true);
 833                 } else
 834                         m_freem(m);
 835         }
 836
 837         if (error == 0)
 838                 IPSTAT_INC(ips_fragmented);
 839
 840 done:
 841         return (error);
 842  bad:
 843         m_freem(m);
 844         goto done;
 845 }
 846
 847 /*
 848  * Create a chain of fragments which fit the given mtu. m_frag points to the
 849  * mbuf to be fragmented; on return it points to the chain with the fragments.
 850  * Return 0 if no error. If error, m_frag may contain a partially built
 851  * chain of fragments that should be freed by the caller.
 852  *
 853  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
 854  */
 855 int
 856 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
 857     u_long if_hwassist_flags)
 858 {
 859         int error = 0;
 860         int hlen = ip->ip_hl << 2;
 861         int len = (mtu - hlen) & ~7;    /* size of payload in each fragment */
 862         int off;
 863         struct mbuf *m0 = *m_frag;      /* the original packet          */
 864         int firstlen;
 865         struct mbuf **mnext;
 866         int nfrags;
 867         uint16_t ip_len, ip_off;
 868
 869         ip_len = ntohs(ip->ip_len);
 870         ip_off = ntohs(ip->ip_off);
 871
 872         /*
 873          * Packet shall not have "Don't Fragment" flag and have at least 8
 874          * bytes of payload.
 875          */
 876         if (__predict_false((ip_off & IP_DF) || len < 8)) {
 877                 IPSTAT_INC(ips_cantfrag);
 878                 return (EMSGSIZE);
 879         }
 880
 881         /*
 882          * If the interface will not calculate checksums on
 883          * fragmented packets, then do it here.
 884          */
 885         if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
 886                 in_delayed_cksum(m0);
 887                 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
 888         }
 889 #if defined(SCTP) || defined(SCTP_SUPPORT)
 890         if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
 891                 sctp_delayed_cksum(m0, hlen);
 892                 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
 893         }
 894 #endif
 895         if (len > PAGE_SIZE) {
 896                 /*
 897                  * Fragment large datagrams such that each segment
 898                  * contains a multiple of PAGE_SIZE amount of data,
 899                  * plus headers. This enables a receiver to perform
 900                  * page-flipping zero-copy optimizations.
 901                  *
 902                  * XXX When does this help given that sender and receiver
 903                  * could have different page sizes, and also mtu could
 904                  * be less than the receiver's page size ?
 905                  */
 906                 int newlen;
 907
 908                 off = MIN(mtu, m0->m_pkthdr.len);
 909
 910                 /*
 911                  * firstlen (off - hlen) must be aligned on an
 912                  * 8-byte boundary
 913                  */
 914                 if (off < hlen)
 915                         goto smart_frag_failure;
 916                 off = ((off - hlen) & ~7) + hlen;
 917                 newlen = (~PAGE_MASK) & mtu;
 918                 if ((newlen + sizeof (struct ip)) > mtu) {
 919                         /* we failed, go back the default */
 920 smart_frag_failure:
 921                         newlen = len;
 922                         off = hlen + len;
 923                 }
 924                 len = newlen;
 925
 926         } else {
 927                 off = hlen + len;
 928         }
 929
 930         firstlen = off - hlen;
 931         mnext = &m0->m_nextpkt;         /* pointer to next packet */
 932
 933         /*
 934          * Loop through length of segment after first fragment,
 935          * make new header and copy data of each part and link onto chain.
 936          * Here, m0 is the original packet, m is the fragment being created.
 937          * The fragments are linked off the m_nextpkt of the original
 938          * packet, which after processing serves as the first fragment.
 939          */
 940         for (nfrags = 1; off < ip_len; off += len, nfrags++) {
 941                 struct ip *mhip;        /* ip header on the fragment */
 942                 struct mbuf *m;
 943                 int mhlen = sizeof (struct ip);
 944
 945                 m = m_gethdr(M_NOWAIT, MT_DATA);
 946                 if (m == NULL) {
 947                         error = ENOBUFS;
 948                         IPSTAT_INC(ips_odropped);
 949                         goto done;
 950                 }
 951                 /*
 952                  * Make sure the complete packet header gets copied
 953                  * from the originating mbuf to the newly created
 954                  * mbuf. This also ensures that existing firewall
 955                  * classification(s), VLAN tags and so on get copied
 956                  * to the resulting fragmented packet(s):
 957                  */
 958                 if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
 959                         m_free(m);
 960                         error = ENOBUFS;
 961                         IPSTAT_INC(ips_odropped);
 962                         goto done;
 963                 }
 964                 /*
 965                  * In the first mbuf, leave room for the link header, then
 966                  * copy the original IP header including options. The payload
 967                  * goes into an additional mbuf chain returned by m_copym().
 968                  */
 969                 m->m_data += max_linkhdr;
 970                 mhip = mtod(m, struct ip *);
 971                 *mhip = *ip;
 972                 if (hlen > sizeof (struct ip)) {
 973                         mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
 974                         mhip->ip_v = IPVERSION;
 975                         mhip->ip_hl = mhlen >> 2;
 976                 }
 977                 m->m_len = mhlen;
 978                 /* XXX do we need to add ip_off below ? */
 979                 mhip->ip_off = ((off - hlen) >> 3) + ip_off;
 980                 if (off + len >= ip_len)
 981                         len = ip_len - off;
 982                 else
 983                         mhip->ip_off |= IP_MF;
 984                 mhip->ip_len = htons((u_short)(len + mhlen));
 985                 m->m_next = m_copym(m0, off, len, M_NOWAIT);
 986                 if (m->m_next == NULL) {        /* copy failed */
 987                         m_free(m);
 988                         error = ENOBUFS;        /* ??? */
 989                         IPSTAT_INC(ips_odropped);
 990                         goto done;
 991                 }
 992                 m->m_pkthdr.len = mhlen + len;
 993 #ifdef MAC
 994                 mac_netinet_fragment(m0, m);
 995 #endif
 996                 mhip->ip_off = htons(mhip->ip_off);
 997                 mhip->ip_sum = 0;
 998                 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
 999                         mhip->ip_sum = in_cksum(m, mhlen);
1000                         m->m_pkthdr.csum_flags &= ~CSUM_IP;
1001                 }
1002                 *mnext = m;
1003                 mnext = &m->m_nextpkt;
1004         }
1005         IPSTAT_ADD(ips_ofragments, nfrags);
1006
1007         /*
1008          * Update first fragment by trimming what's been copied out
1009          * and updating header.
1010          */
1011         m_adj(m0, hlen + firstlen - ip_len);
1012         m0->m_pkthdr.len = hlen + firstlen;
1013         ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1014         ip->ip_off = htons(ip_off | IP_MF);
1015         ip->ip_sum = 0;
1016         if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
1017                 ip->ip_sum = in_cksum(m0, hlen);
1018                 m0->m_pkthdr.csum_flags &= ~CSUM_IP;
1019         }
1020
1021 done:
1022         *m_frag = m0;
1023         return error;
1024 }
1025
1026 void
1027 in_delayed_cksum(struct mbuf *m)
1028 {
1029         struct ip *ip;
1030         struct udphdr *uh;
1031         uint16_t cklen, csum, offset;
1032
1033         ip = mtod(m, struct ip *);
1034         offset = ip->ip_hl << 2 ;
1035
1036         if (m->m_pkthdr.csum_flags & CSUM_UDP) {
1037                 /* if udp header is not in the first mbuf copy udplen */
1038                 if (offset + sizeof(struct udphdr) > m->m_len) {
1039                         m_copydata(m, offset + offsetof(struct udphdr,
1040                             uh_ulen), sizeof(cklen), (caddr_t)&cklen);
1041                         cklen = ntohs(cklen);
1042                 } else {
1043                         uh = (struct udphdr *)mtodo(m, offset);
1044                         cklen = ntohs(uh->uh_ulen);
1045                 }
1046                 csum = in_cksum_skip(m, cklen + offset, offset);
1047                 if (csum == 0)
1048                         csum = 0xffff;
1049         } else {
1050                 cklen = ntohs(ip->ip_len);
1051                 csum = in_cksum_skip(m, cklen, offset);
1052         }
1053         offset += m->m_pkthdr.csum_data;        /* checksum offset */
1054
1055         if (offset + sizeof(csum) > m->m_len)
1056                 m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
1057         else
1058                 *(u_short *)mtodo(m, offset) = csum;
1059 }
1060
1061 /*
1062  * IP socket option processing.
1063  */
1064 int
1065 ip_ctloutput(struct socket *so, struct sockopt *sopt)
1066 {
1067         struct inpcb *inp = sotoinpcb(so);
1068         int     error, optval;
1069 #ifdef  RSS
1070         uint32_t rss_bucket;
1071         int retval;
1072 #endif
1073
1074         error = optval = 0;
1075         if (sopt->sopt_level != IPPROTO_IP) {
1076                 error = EINVAL;
1077
1078                 if (sopt->sopt_level == SOL_SOCKET &&
1079                     sopt->sopt_dir == SOPT_SET) {
1080                         switch (sopt->sopt_name) {
1081                         case SO_REUSEADDR:
1082                                 INP_WLOCK(inp);
1083                                 if ((so->so_options & SO_REUSEADDR) != 0)
1084                                         inp->inp_flags2 |= INP_REUSEADDR;
1085                                 else
1086                                         inp->inp_flags2 &= ~INP_REUSEADDR;
1087                                 INP_WUNLOCK(inp);
1088                                 error = 0;
1089                                 break;
1090                         case SO_REUSEPORT:
1091                                 INP_WLOCK(inp);
1092                                 if ((so->so_options & SO_REUSEPORT) != 0)
1093                                         inp->inp_flags2 |= INP_REUSEPORT;
1094                                 else
1095                                         inp->inp_flags2 &= ~INP_REUSEPORT;
1096                                 INP_WUNLOCK(inp);
1097                                 error = 0;
1098                                 break;
1099                         case SO_REUSEPORT_LB:
1100                                 INP_WLOCK(inp);
1101                                 if ((so->so_options & SO_REUSEPORT_LB) != 0)
1102                                         inp->inp_flags2 |= INP_REUSEPORT_LB;
1103                                 else
1104                                         inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1105                                 INP_WUNLOCK(inp);
1106                                 error = 0;
1107                                 break;
1108                         case SO_SETFIB:
1109                                 INP_WLOCK(inp);
1110                                 inp->inp_inc.inc_fibnum = so->so_fibnum;
1111                                 INP_WUNLOCK(inp);
1112                                 error = 0;
1113                                 break;
1114                         case SO_MAX_PACING_RATE:
1115 #ifdef RATELIMIT
1116                                 INP_WLOCK(inp);
1117                                 inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1118                                 INP_WUNLOCK(inp);
1119                                 error = 0;
1120 #else
1121                                 error = EOPNOTSUPP;
1122 #endif
1123                                 break;
1124                         default:
1125                                 break;
1126                         }
1127                 }
1128                 return (error);
1129         }
1130
1131         switch (sopt->sopt_dir) {
1132         case SOPT_SET:
1133                 switch (sopt->sopt_name) {
1134                 case IP_OPTIONS:
1135 #ifdef notyet
1136                 case IP_RETOPTS:
1137 #endif
1138                 {
1139                         struct mbuf *m;
1140                         if (sopt->sopt_valsize > MLEN) {
1141                                 error = EMSGSIZE;
1142                                 break;
1143                         }
1144                         m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
1145                         if (m == NULL) {
1146                                 error = ENOBUFS;
1147                                 break;
1148                         }
1149                         m->m_len = sopt->sopt_valsize;
1150                         error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1151                                             m->m_len);
1152                         if (error) {
1153                                 m_free(m);
1154                                 break;
1155                         }
1156                         INP_WLOCK(inp);
1157                         error = ip_pcbopts(inp, sopt->sopt_name, m);
1158                         INP_WUNLOCK(inp);
1159                         return (error);
1160                 }
1161
1162                 case IP_BINDANY:
1163                         if (sopt->sopt_td != NULL) {
1164                                 error = priv_check(sopt->sopt_td,
1165                                     PRIV_NETINET_BINDANY);
1166                                 if (error)
1167                                         break;
1168                         }
1169                         /* FALLTHROUGH */
1170                 case IP_BINDMULTI:
1171 #ifdef  RSS
1172                 case IP_RSS_LISTEN_BUCKET:
1173 #endif
1174                 case IP_TOS:
1175                 case IP_TTL:
1176                 case IP_MINTTL:
1177                 case IP_RECVOPTS:
1178                 case IP_RECVRETOPTS:
1179                 case IP_ORIGDSTADDR:
1180                 case IP_RECVDSTADDR:
1181                 case IP_RECVTTL:
1182                 case IP_RECVIF:
1183                 case IP_ONESBCAST:
1184                 case IP_DONTFRAG:
1185                 case IP_RECVTOS:
1186                 case IP_RECVFLOWID:
1187 #ifdef  RSS
1188                 case IP_RECVRSSBUCKETID:
1189 #endif
1190                 case IP_VLAN_PCP:
1191                         error = sooptcopyin(sopt, &optval, sizeof optval,
1192                                             sizeof optval);
1193                         if (error)
1194                                 break;
1195
1196                         switch (sopt->sopt_name) {
1197                         case IP_TOS:
1198                                 inp->inp_ip_tos = optval;
1199                                 break;
1200
1201                         case IP_TTL:
1202                                 inp->inp_ip_ttl = optval;
1203                                 break;
1204
1205                         case IP_MINTTL:
1206                                 if (optval >= 0 && optval <= MAXTTL)
1207                                         inp->inp_ip_minttl = optval;
1208                                 else
1209                                         error = EINVAL;
1210                                 break;
1211
1212 #define OPTSET(bit) do {                                                \
1213         INP_WLOCK(inp);                                                 \
1214         if (optval)                                                     \
1215                 inp->inp_flags |= bit;                                  \
1216         else                                                            \
1217                 inp->inp_flags &= ~bit;                                 \
1218         INP_WUNLOCK(inp);                                               \
1219 } while (0)
1220
1221 #define OPTSET2(bit, val) do {                                          \
1222         INP_WLOCK(inp);                                                 \
1223         if (val)                                                        \
1224                 inp->inp_flags2 |= bit;                                 \
1225         else                                                            \
1226                 inp->inp_flags2 &= ~bit;                                \
1227         INP_WUNLOCK(inp);                                               \
1228 } while (0)
1229
1230                         case IP_RECVOPTS:
1231                                 OPTSET(INP_RECVOPTS);
1232                                 break;
1233
1234                         case IP_RECVRETOPTS:
1235                                 OPTSET(INP_RECVRETOPTS);
1236                                 break;
1237
1238                         case IP_RECVDSTADDR:
1239                                 OPTSET(INP_RECVDSTADDR);
1240                                 break;
1241
1242                         case IP_ORIGDSTADDR:
1243                                 OPTSET2(INP_ORIGDSTADDR, optval);
1244                                 break;
1245
1246                         case IP_RECVTTL:
1247                                 OPTSET(INP_RECVTTL);
1248                                 break;
1249
1250                         case IP_RECVIF:
1251                                 OPTSET(INP_RECVIF);
1252                                 break;
1253
1254                         case IP_ONESBCAST:
1255                                 OPTSET(INP_ONESBCAST);
1256                                 break;
1257                         case IP_DONTFRAG:
1258                                 OPTSET(INP_DONTFRAG);
1259                                 break;
1260                         case IP_BINDANY:
1261                                 OPTSET(INP_BINDANY);
1262                                 break;
1263                         case IP_RECVTOS:
1264                                 OPTSET(INP_RECVTOS);
1265                                 break;
1266                         case IP_BINDMULTI:
1267                                 OPTSET2(INP_BINDMULTI, optval);
1268                                 break;
1269                         case IP_RECVFLOWID:
1270                                 OPTSET2(INP_RECVFLOWID, optval);
1271                                 break;
1272 #ifdef  RSS
1273                         case IP_RSS_LISTEN_BUCKET:
1274                                 if ((optval >= 0) &&
1275                                     (optval < rss_getnumbuckets())) {
1276                                         inp->inp_rss_listen_bucket = optval;
1277                                         OPTSET2(INP_RSS_BUCKET_SET, 1);
1278                                 } else {
1279                                         error = EINVAL;
1280                                 }
1281                                 break;
1282                         case IP_RECVRSSBUCKETID:
1283                                 OPTSET2(INP_RECVRSSBUCKETID, optval);
1284                                 break;
1285 #endif
1286                         case IP_VLAN_PCP:
1287                                 if ((optval >= -1) && (optval <=
1288                                     (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1289                                         if (optval == -1) {
1290                                                 INP_WLOCK(inp);
1291                                                 inp->inp_flags2 &=
1292                                                     ~(INP_2PCP_SET |
1293                                                       INP_2PCP_MASK);
1294                                                 INP_WUNLOCK(inp);
1295                                         } else {
1296                                                 INP_WLOCK(inp);
1297                                                 inp->inp_flags2 |=
1298                                                     INP_2PCP_SET;
1299                                                 inp->inp_flags2 &=
1300                                                     ~INP_2PCP_MASK;
1301                                                 inp->inp_flags2 |=
1302                                                     optval << INP_2PCP_SHIFT;
1303                                                 INP_WUNLOCK(inp);
1304                                         }
1305                                 } else
1306                                         error = EINVAL;
1307                                 break;
1308                         }
1309                         break;
1310 #undef OPTSET
1311 #undef OPTSET2
1312
1313                 /*
1314                  * Multicast socket options are processed by the in_mcast
1315                  * module.
1316                  */
1317                 case IP_MULTICAST_IF:
1318                 case IP_MULTICAST_VIF:
1319                 case IP_MULTICAST_TTL:
1320                 case IP_MULTICAST_LOOP:
1321                 case IP_ADD_MEMBERSHIP:
1322                 case IP_DROP_MEMBERSHIP:
1323                 case IP_ADD_SOURCE_MEMBERSHIP:
1324                 case IP_DROP_SOURCE_MEMBERSHIP:
1325                 case IP_BLOCK_SOURCE:
1326                 case IP_UNBLOCK_SOURCE:
1327                 case IP_MSFILTER:
1328                 case MCAST_JOIN_GROUP:
1329                 case MCAST_LEAVE_GROUP:
1330                 case MCAST_JOIN_SOURCE_GROUP:
1331                 case MCAST_LEAVE_SOURCE_GROUP:
1332                 case MCAST_BLOCK_SOURCE:
1333                 case MCAST_UNBLOCK_SOURCE:
1334                         error = inp_setmoptions(inp, sopt);
1335                         break;
1336
1337                 case IP_PORTRANGE:
1338                         error = sooptcopyin(sopt, &optval, sizeof optval,
1339                                             sizeof optval);
1340                         if (error)
1341                                 break;
1342
1343                         INP_WLOCK(inp);
1344                         switch (optval) {
1345                         case IP_PORTRANGE_DEFAULT:
1346                                 inp->inp_flags &= ~(INP_LOWPORT);
1347                                 inp->inp_flags &= ~(INP_HIGHPORT);
1348                                 break;
1349
1350                         case IP_PORTRANGE_HIGH:
1351                                 inp->inp_flags &= ~(INP_LOWPORT);
1352                                 inp->inp_flags |= INP_HIGHPORT;
1353                                 break;
1354
1355                         case IP_PORTRANGE_LOW:
1356                                 inp->inp_flags &= ~(INP_HIGHPORT);
1357                                 inp->inp_flags |= INP_LOWPORT;
1358                                 break;
1359
1360                         default:
1361                                 error = EINVAL;
1362                                 break;
1363                         }
1364                         INP_WUNLOCK(inp);
1365                         break;
1366
1367 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1368                 case IP_IPSEC_POLICY:
1369                         if (IPSEC_ENABLED(ipv4)) {
1370                                 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1371                                 break;
1372                         }
1373                         /* FALLTHROUGH */
1374 #endif /* IPSEC */
1375
1376                 default:
1377                         error = ENOPROTOOPT;
1378                         break;
1379                 }
1380                 break;
1381
1382         case SOPT_GET:
1383                 switch (sopt->sopt_name) {
1384                 case IP_OPTIONS:
1385                 case IP_RETOPTS:
1386                         INP_RLOCK(inp);
1387                         if (inp->inp_options) {
1388                                 struct mbuf *options;
1389
1390                                 options = m_copym(inp->inp_options, 0,
1391                                     M_COPYALL, M_NOWAIT);
1392                                 INP_RUNLOCK(inp);
1393                                 if (options != NULL) {
1394                                         error = sooptcopyout(sopt,
1395                                                              mtod(options, char *),
1396                                                              options->m_len);
1397                                         m_freem(options);
1398                                 } else
1399                                         error = ENOMEM;
1400                         } else {
1401                                 INP_RUNLOCK(inp);
1402                                 sopt->sopt_valsize = 0;
1403                         }
1404                         break;
1405
1406                 case IP_TOS:
1407                 case IP_TTL:
1408                 case IP_MINTTL:
1409                 case IP_RECVOPTS:
1410                 case IP_RECVRETOPTS:
1411                 case IP_ORIGDSTADDR:
1412                 case IP_RECVDSTADDR:
1413                 case IP_RECVTTL:
1414                 case IP_RECVIF:
1415                 case IP_PORTRANGE:
1416                 case IP_ONESBCAST:
1417                 case IP_DONTFRAG:
1418                 case IP_BINDANY:
1419                 case IP_RECVTOS:
1420                 case IP_BINDMULTI:
1421                 case IP_FLOWID:
1422                 case IP_FLOWTYPE:
1423                 case IP_RECVFLOWID:
1424 #ifdef  RSS
1425                 case IP_RSSBUCKETID:
1426                 case IP_RECVRSSBUCKETID:
1427 #endif
1428                 case IP_VLAN_PCP:
1429                         switch (sopt->sopt_name) {
1430                         case IP_TOS:
1431                                 optval = inp->inp_ip_tos;
1432                                 break;
1433
1434                         case IP_TTL:
1435                                 optval = inp->inp_ip_ttl;
1436                                 break;
1437
1438                         case IP_MINTTL:
1439                                 optval = inp->inp_ip_minttl;
1440                                 break;
1441
1442 #define OPTBIT(bit)     (inp->inp_flags & bit ? 1 : 0)
1443 #define OPTBIT2(bit)    (inp->inp_flags2 & bit ? 1 : 0)
1444
1445                         case IP_RECVOPTS:
1446                                 optval = OPTBIT(INP_RECVOPTS);
1447                                 break;
1448
1449                         case IP_RECVRETOPTS:
1450                                 optval = OPTBIT(INP_RECVRETOPTS);
1451                                 break;
1452
1453                         case IP_RECVDSTADDR:
1454                                 optval = OPTBIT(INP_RECVDSTADDR);
1455                                 break;
1456
1457                         case IP_ORIGDSTADDR:
1458                                 optval = OPTBIT2(INP_ORIGDSTADDR);
1459                                 break;
1460
1461                         case IP_RECVTTL:
1462                                 optval = OPTBIT(INP_RECVTTL);
1463                                 break;
1464
1465                         case IP_RECVIF:
1466                                 optval = OPTBIT(INP_RECVIF);
1467                                 break;
1468
1469                         case IP_PORTRANGE:
1470                                 if (inp->inp_flags & INP_HIGHPORT)
1471                                         optval = IP_PORTRANGE_HIGH;
1472                                 else if (inp->inp_flags & INP_LOWPORT)
1473                                         optval = IP_PORTRANGE_LOW;
1474                                 else
1475                                         optval = 0;
1476                                 break;
1477
1478                         case IP_ONESBCAST:
1479                                 optval = OPTBIT(INP_ONESBCAST);
1480                                 break;
1481                         case IP_DONTFRAG:
1482                                 optval = OPTBIT(INP_DONTFRAG);
1483                                 break;
1484                         case IP_BINDANY:
1485                                 optval = OPTBIT(INP_BINDANY);
1486                                 break;
1487                         case IP_RECVTOS:
1488                                 optval = OPTBIT(INP_RECVTOS);
1489                                 break;
1490                         case IP_FLOWID:
1491                                 optval = inp->inp_flowid;
1492                                 break;
1493                         case IP_FLOWTYPE:
1494                                 optval = inp->inp_flowtype;
1495                                 break;
1496                         case IP_RECVFLOWID:
1497                                 optval = OPTBIT2(INP_RECVFLOWID);
1498                                 break;
1499 #ifdef  RSS
1500                         case IP_RSSBUCKETID:
1501                                 retval = rss_hash2bucket(inp->inp_flowid,
1502                                     inp->inp_flowtype,
1503                                     &rss_bucket);
1504                                 if (retval == 0)
1505                                         optval = rss_bucket;
1506                                 else
1507                                         error = EINVAL;
1508                                 break;
1509                         case IP_RECVRSSBUCKETID:
1510                                 optval = OPTBIT2(INP_RECVRSSBUCKETID);
1511                                 break;
1512 #endif
1513                         case IP_BINDMULTI:
1514                                 optval = OPTBIT2(INP_BINDMULTI);
1515                                 break;
1516                         case IP_VLAN_PCP:
1517                                 if (OPTBIT2(INP_2PCP_SET)) {
1518                                         optval = (inp->inp_flags2 &
1519                                             INP_2PCP_MASK) >> INP_2PCP_SHIFT;
1520                                 } else {
1521                                         optval = -1;
1522                                 }
1523                                 break;
1524                         }
1525                         error = sooptcopyout(sopt, &optval, sizeof optval);
1526                         break;
1527
1528                 /*
1529                  * Multicast socket options are processed by the in_mcast
1530                  * module.
1531                  */
1532                 case IP_MULTICAST_IF:
1533                 case IP_MULTICAST_VIF:
1534                 case IP_MULTICAST_TTL:
1535                 case IP_MULTICAST_LOOP:
1536                 case IP_MSFILTER:
1537                         error = inp_getmoptions(inp, sopt);
1538                         break;
1539
1540 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1541                 case IP_IPSEC_POLICY:
1542                         if (IPSEC_ENABLED(ipv4)) {
1543                                 error = IPSEC_PCBCTL(ipv4, inp, sopt);
1544                                 break;
1545                         }
1546                         /* FALLTHROUGH */
1547 #endif /* IPSEC */
1548
1549                 default:
1550                         error = ENOPROTOOPT;
1551                         break;
1552                 }
1553                 break;
1554         }
1555         return (error);
1556 }
1557
1558 /*
1559  * Routine called from ip_output() to loop back a copy of an IP multicast
1560  * packet to the input queue of a specified interface.  Note that this
1561  * calls the output routine of the loopback "driver", but with an interface
1562  * pointer that might NOT be a loopback interface -- evil, but easier than
1563  * replicating that code here.
1564  */
1565 static void
1566 ip_mloopback(struct ifnet *ifp, const struct mbuf *m, int hlen)
1567 {
1568         struct ip *ip;
1569         struct mbuf *copym;
1570
1571         /*
1572          * Make a deep copy of the packet because we're going to
1573          * modify the pack in order to generate checksums.
1574          */
1575         copym = m_dup(m, M_NOWAIT);
1576         if (copym != NULL && (!M_WRITABLE(copym) || copym->m_len < hlen))
1577                 copym = m_pullup(copym, hlen);
1578         if (copym != NULL) {
1579                 /* If needed, compute the checksum and mark it as valid. */
1580                 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1581                         in_delayed_cksum(copym);
1582                         copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1583                         copym->m_pkthdr.csum_flags |=
1584                             CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1585                         copym->m_pkthdr.csum_data = 0xffff;
1586                 }
1587                 /*
1588                  * We don't bother to fragment if the IP length is greater
1589                  * than the interface's MTU.  Can this possibly matter?
1590                  */
1591                 ip = mtod(copym, struct ip *);
1592                 ip->ip_sum = 0;
1593                 ip->ip_sum = in_cksum(copym, hlen);
1594                 if_simloop(ifp, copym, AF_INET, 0);
1595         }
1596 }