1 /* $NetBSD: ip_output.c,v 1.312 2019/05/15 02:59:18 ozaki-r Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * 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 project 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 PROJECT 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 PROJECT 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 32 /* 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 59 * POSSIBILITY OF SUCH DAMAGE. 60 */ 61 62 /* 63 * Copyright (c) 1982, 1986, 1988, 1990, 1993 64 * The Regents of the University of California. All rights reserved. 65 * 66 * Redistribution and use in source and binary forms, with or without 67 * modification, are permitted provided that the following conditions 68 * are met: 69 * 1. Redistributions of source code must retain the above copyright 70 * notice, this list of conditions and the following disclaimer. 71 * 2. Redistributions in binary form must reproduce the above copyright 72 * notice, this list of conditions and the following disclaimer in the 73 * documentation and/or other materials provided with the distribution. 74 * 3. Neither the name of the University nor the names of its contributors 75 * may be used to endorse or promote products derived from this software 76 * without specific prior written permission. 77 * 78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 88 * SUCH DAMAGE. 89 * 90 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 91 */ 92 93 #include <sys/cdefs.h> 94 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.312 2019/05/15 02:59:18 ozaki-r Exp $"); 95 96 #ifdef _KERNEL_OPT 97 #include "opt_inet.h" 98 #include "opt_ipsec.h" 99 #include "opt_mrouting.h" 100 #include "opt_net_mpsafe.h" 101 #include "opt_mpls.h" 102 #endif 103 104 #include "arp.h" 105 106 #include <sys/param.h> 107 #include <sys/kmem.h> 108 #include <sys/mbuf.h> 109 #include <sys/socket.h> 110 #include <sys/socketvar.h> 111 #include <sys/kauth.h> 112 #include <sys/systm.h> 113 #include <sys/syslog.h> 114 115 #include <net/if.h> 116 #include <net/if_types.h> 117 #include <net/route.h> 118 #include <net/pfil.h> 119 120 #include <netinet/in.h> 121 #include <netinet/in_systm.h> 122 #include <netinet/ip.h> 123 #include <netinet/in_pcb.h> 124 #include <netinet/in_var.h> 125 #include <netinet/ip_var.h> 126 #include <netinet/ip_private.h> 127 #include <netinet/in_offload.h> 128 #include <netinet/portalgo.h> 129 #include <netinet/udp.h> 130 #include <netinet/udp_var.h> 131 132 #ifdef INET6 133 #include <netinet6/ip6_var.h> 134 #endif 135 136 #ifdef MROUTING 137 #include <netinet/ip_mroute.h> 138 #endif 139 140 #ifdef IPSEC 141 #include <netipsec/ipsec.h> 142 #include <netipsec/key.h> 143 #endif 144 145 #ifdef MPLS 146 #include <netmpls/mpls.h> 147 #include <netmpls/mpls_var.h> 148 #endif 149 150 static int ip_pcbopts(struct inpcb *, const struct sockopt *); 151 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 152 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 153 static void ip_mloopback(struct ifnet *, struct mbuf *, 154 const struct sockaddr_in *); 155 static int ip_ifaddrvalid(const struct in_ifaddr *); 156 157 extern pfil_head_t *inet_pfil_hook; /* XXX */ 158 159 int ip_do_loopback_cksum = 0; 160 161 static int 162 ip_mark_mpls(struct ifnet * const ifp, struct mbuf * const m, 163 const struct rtentry *rt) 164 { 165 int error = 0; 166 #ifdef MPLS 167 union mpls_shim msh; 168 169 if (rt == NULL || rt_gettag(rt) == NULL || 170 rt_gettag(rt)->sa_family != AF_MPLS || 171 (m->m_flags & (M_MCAST | M_BCAST)) != 0 || 172 ifp->if_type != IFT_ETHER) 173 return 0; 174 175 msh.s_addr = MPLS_GETSADDR(rt); 176 if (msh.shim.label != MPLS_LABEL_IMPLNULL) { 177 struct m_tag *mtag; 178 /* 179 * XXX tentative solution to tell ether_output 180 * it's MPLS. Need some more efficient solution. 181 */ 182 mtag = m_tag_get(PACKET_TAG_MPLS, 183 sizeof(int) /* dummy */, 184 M_NOWAIT); 185 if (mtag == NULL) 186 return ENOMEM; 187 m_tag_prepend(m, mtag); 188 } 189 #endif 190 return error; 191 } 192 193 /* 194 * Send an IP packet to a host. 195 */ 196 int 197 ip_if_output(struct ifnet * const ifp, struct mbuf * const m, 198 const struct sockaddr * const dst, const struct rtentry *rt) 199 { 200 int error = 0; 201 202 if (rt != NULL) { 203 error = rt_check_reject_route(rt, ifp); 204 if (error != 0) { 205 m_freem(m); 206 return error; 207 } 208 } 209 210 error = ip_mark_mpls(ifp, m, rt); 211 if (error != 0) { 212 m_freem(m); 213 return error; 214 } 215 216 error = if_output_lock(ifp, ifp, m, dst, rt); 217 218 return error; 219 } 220 221 /* 222 * IP output. The packet in mbuf chain m contains a skeletal IP 223 * header (with len, off, ttl, proto, tos, src, dst). 224 * The mbuf chain containing the packet will be freed. 225 * The mbuf opt, if present, will not be freed. 226 */ 227 int 228 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, int flags, 229 struct ip_moptions *imo, struct inpcb *inp) 230 { 231 struct rtentry *rt; 232 struct ip *ip; 233 struct ifnet *ifp, *mifp = NULL; 234 struct mbuf *m = m0; 235 int len, hlen, error = 0; 236 struct route iproute; 237 const struct sockaddr_in *dst; 238 struct in_ifaddr *ia = NULL; 239 struct ifaddr *ifa; 240 int isbroadcast; 241 int sw_csum; 242 u_long mtu; 243 bool natt_frag = false; 244 bool rtmtu_nolock; 245 union { 246 struct sockaddr sa; 247 struct sockaddr_in sin; 248 } udst, usrc; 249 struct sockaddr *rdst = &udst.sa; /* real IP destination, as 250 * opposed to the nexthop 251 */ 252 struct psref psref, psref_ia; 253 int bound; 254 bool bind_need_restore = false; 255 256 len = 0; 257 258 MCLAIM(m, &ip_tx_mowner); 259 260 KASSERT((m->m_flags & M_PKTHDR) != 0); 261 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 0); 262 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) != 263 (M_CSUM_TCPv4|M_CSUM_UDPv4)); 264 KASSERT(m->m_len >= sizeof(struct ip)); 265 266 hlen = sizeof(struct ip); 267 if (opt) { 268 m = ip_insertoptions(m, opt, &len); 269 hlen = len; 270 } 271 ip = mtod(m, struct ip *); 272 273 /* 274 * Fill in IP header. 275 */ 276 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 277 ip->ip_v = IPVERSION; 278 ip->ip_off = htons(0); 279 /* ip->ip_id filled in after we find out source ia */ 280 ip->ip_hl = hlen >> 2; 281 IP_STATINC(IP_STAT_LOCALOUT); 282 } else { 283 hlen = ip->ip_hl << 2; 284 } 285 286 /* 287 * Route packet. 288 */ 289 if (ro == NULL) { 290 memset(&iproute, 0, sizeof(iproute)); 291 ro = &iproute; 292 } 293 sockaddr_in_init(&udst.sin, &ip->ip_dst, 0); 294 dst = satocsin(rtcache_getdst(ro)); 295 296 /* 297 * If there is a cached route, check that it is to the same 298 * destination and is still up. If not, free it and try again. 299 * The address family should also be checked in case of sharing 300 * the cache with IPv6. 301 */ 302 if (dst && (dst->sin_family != AF_INET || 303 !in_hosteq(dst->sin_addr, ip->ip_dst))) 304 rtcache_free(ro); 305 306 /* XXX must be before rtcache operations */ 307 bound = curlwp_bind(); 308 bind_need_restore = true; 309 310 if ((rt = rtcache_validate(ro)) == NULL && 311 (rt = rtcache_update(ro, 1)) == NULL) { 312 dst = &udst.sin; 313 error = rtcache_setdst(ro, &udst.sa); 314 if (error != 0) 315 goto bad; 316 } 317 318 /* 319 * If routing to interface only, short circuit routing lookup. 320 */ 321 if (flags & IP_ROUTETOIF) { 322 ifa = ifa_ifwithladdr_psref(sintocsa(dst), &psref_ia); 323 if (ifa == NULL) { 324 IP_STATINC(IP_STAT_NOROUTE); 325 error = ENETUNREACH; 326 goto bad; 327 } 328 /* ia is already referenced by psref_ia */ 329 ia = ifatoia(ifa); 330 331 ifp = ia->ia_ifp; 332 mtu = ifp->if_mtu; 333 ip->ip_ttl = 1; 334 isbroadcast = in_broadcast(dst->sin_addr, ifp); 335 } else if (((IN_MULTICAST(ip->ip_dst.s_addr) || 336 ip->ip_dst.s_addr == INADDR_BROADCAST) || 337 (flags & IP_ROUTETOIFINDEX)) && 338 imo != NULL && imo->imo_multicast_if_index != 0) { 339 ifp = mifp = if_get_byindex(imo->imo_multicast_if_index, &psref); 340 if (ifp == NULL) { 341 IP_STATINC(IP_STAT_NOROUTE); 342 error = ENETUNREACH; 343 goto bad; 344 } 345 mtu = ifp->if_mtu; 346 ia = in_get_ia_from_ifp_psref(ifp, &psref_ia); 347 if (ia == NULL) { 348 error = EADDRNOTAVAIL; 349 goto bad; 350 } 351 if (IN_MULTICAST(ip->ip_dst.s_addr) || 352 ip->ip_dst.s_addr == INADDR_BROADCAST) { 353 isbroadcast = 0; 354 } else { 355 /* IP_ROUTETOIFINDEX */ 356 isbroadcast = in_broadcast(dst->sin_addr, ifp); 357 if ((isbroadcast == 0) && ((ifp->if_flags & 358 (IFF_LOOPBACK | IFF_POINTOPOINT)) == 0) && 359 (in_direct(dst->sin_addr, ifp) == 0)) { 360 /* gateway address required */ 361 if (rt == NULL) 362 rt = rtcache_init(ro); 363 if (rt == NULL || rt->rt_ifp != ifp) { 364 IP_STATINC(IP_STAT_NOROUTE); 365 error = EHOSTUNREACH; 366 goto bad; 367 } 368 rt->rt_use++; 369 if (rt->rt_flags & RTF_GATEWAY) 370 dst = satosin(rt->rt_gateway); 371 if (rt->rt_flags & RTF_HOST) 372 isbroadcast = 373 rt->rt_flags & RTF_BROADCAST; 374 } 375 } 376 } else { 377 if (rt == NULL) 378 rt = rtcache_init(ro); 379 if (rt == NULL) { 380 IP_STATINC(IP_STAT_NOROUTE); 381 error = EHOSTUNREACH; 382 goto bad; 383 } 384 if (ifa_is_destroying(rt->rt_ifa)) { 385 rtcache_unref(rt, ro); 386 rt = NULL; 387 IP_STATINC(IP_STAT_NOROUTE); 388 error = EHOSTUNREACH; 389 goto bad; 390 } 391 ifa_acquire(rt->rt_ifa, &psref_ia); 392 ia = ifatoia(rt->rt_ifa); 393 ifp = rt->rt_ifp; 394 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 395 mtu = ifp->if_mtu; 396 rt->rt_use++; 397 if (rt->rt_flags & RTF_GATEWAY) 398 dst = satosin(rt->rt_gateway); 399 if (rt->rt_flags & RTF_HOST) 400 isbroadcast = rt->rt_flags & RTF_BROADCAST; 401 else 402 isbroadcast = in_broadcast(dst->sin_addr, ifp); 403 } 404 rtmtu_nolock = rt && (rt->rt_rmx.rmx_locks & RTV_MTU) == 0; 405 406 if (IN_MULTICAST(ip->ip_dst.s_addr) || 407 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 408 bool inmgroup; 409 410 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 411 M_BCAST : M_MCAST; 412 /* 413 * See if the caller provided any multicast options 414 */ 415 if (imo != NULL) 416 ip->ip_ttl = imo->imo_multicast_ttl; 417 else 418 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 419 420 /* 421 * if we don't know the outgoing ifp yet, we can't generate 422 * output 423 */ 424 if (!ifp) { 425 IP_STATINC(IP_STAT_NOROUTE); 426 error = ENETUNREACH; 427 goto bad; 428 } 429 430 /* 431 * If the packet is multicast or broadcast, confirm that 432 * the outgoing interface can transmit it. 433 */ 434 if (((m->m_flags & M_MCAST) && 435 (ifp->if_flags & IFF_MULTICAST) == 0) || 436 ((m->m_flags & M_BCAST) && 437 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 438 IP_STATINC(IP_STAT_NOROUTE); 439 error = ENETUNREACH; 440 goto bad; 441 } 442 /* 443 * If source address not specified yet, use an address 444 * of outgoing interface. 445 */ 446 if (in_nullhost(ip->ip_src)) { 447 struct in_ifaddr *xia; 448 struct ifaddr *xifa; 449 struct psref _psref; 450 451 xia = in_get_ia_from_ifp_psref(ifp, &_psref); 452 if (!xia) { 453 error = EADDRNOTAVAIL; 454 goto bad; 455 } 456 xifa = &xia->ia_ifa; 457 if (xifa->ifa_getifa != NULL) { 458 ia4_release(xia, &_psref); 459 /* FIXME ifa_getifa is NOMPSAFE */ 460 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 461 if (xia == NULL) { 462 error = EADDRNOTAVAIL; 463 goto bad; 464 } 465 ia4_acquire(xia, &_psref); 466 } 467 ip->ip_src = xia->ia_addr.sin_addr; 468 ia4_release(xia, &_psref); 469 } 470 471 inmgroup = in_multi_group(ip->ip_dst, ifp, flags); 472 if (inmgroup && (imo == NULL || imo->imo_multicast_loop)) { 473 /* 474 * If we belong to the destination multicast group 475 * on the outgoing interface, and the caller did not 476 * forbid loopback, loop back a copy. 477 */ 478 ip_mloopback(ifp, m, &udst.sin); 479 } 480 #ifdef MROUTING 481 else { 482 /* 483 * If we are acting as a multicast router, perform 484 * multicast forwarding as if the packet had just 485 * arrived on the interface to which we are about 486 * to send. The multicast forwarding function 487 * recursively calls this function, using the 488 * IP_FORWARDING flag to prevent infinite recursion. 489 * 490 * Multicasts that are looped back by ip_mloopback(), 491 * above, will be forwarded by the ip_input() routine, 492 * if necessary. 493 */ 494 extern struct socket *ip_mrouter; 495 496 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 497 if (ip_mforward(m, ifp) != 0) { 498 m_freem(m); 499 goto done; 500 } 501 } 502 } 503 #endif 504 /* 505 * Multicasts with a time-to-live of zero may be looped- 506 * back, above, but must not be transmitted on a network. 507 * Also, multicasts addressed to the loopback interface 508 * are not sent -- the above call to ip_mloopback() will 509 * loop back a copy if this host actually belongs to the 510 * destination group on the loopback interface. 511 */ 512 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 513 m_freem(m); 514 goto done; 515 } 516 goto sendit; 517 } 518 519 /* 520 * If source address not specified yet, use address 521 * of outgoing interface. 522 */ 523 if (in_nullhost(ip->ip_src)) { 524 struct ifaddr *xifa; 525 526 xifa = &ia->ia_ifa; 527 if (xifa->ifa_getifa != NULL) { 528 ia4_release(ia, &psref_ia); 529 /* FIXME ifa_getifa is NOMPSAFE */ 530 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 531 if (ia == NULL) { 532 error = EADDRNOTAVAIL; 533 goto bad; 534 } 535 ia4_acquire(ia, &psref_ia); 536 } 537 ip->ip_src = ia->ia_addr.sin_addr; 538 } 539 540 /* 541 * Packets with Class-D address as source are not valid per 542 * RFC1112. 543 */ 544 if (IN_MULTICAST(ip->ip_src.s_addr)) { 545 IP_STATINC(IP_STAT_ODROPPED); 546 error = EADDRNOTAVAIL; 547 goto bad; 548 } 549 550 /* 551 * Look for broadcast address and verify user is allowed to 552 * send such a packet. 553 */ 554 if (isbroadcast) { 555 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 556 error = EADDRNOTAVAIL; 557 goto bad; 558 } 559 if ((flags & IP_ALLOWBROADCAST) == 0) { 560 error = EACCES; 561 goto bad; 562 } 563 /* don't allow broadcast messages to be fragmented */ 564 if (ntohs(ip->ip_len) > ifp->if_mtu) { 565 error = EMSGSIZE; 566 goto bad; 567 } 568 m->m_flags |= M_BCAST; 569 } else 570 m->m_flags &= ~M_BCAST; 571 572 sendit: 573 if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) { 574 if (m->m_pkthdr.len < IP_MINFRAGSIZE) { 575 ip->ip_id = 0; 576 } else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 577 ip->ip_id = ip_newid(ia); 578 } else { 579 /* 580 * TSO capable interfaces (typically?) increment 581 * ip_id for each segment. 582 * "allocate" enough ids here to increase the chance 583 * for them to be unique. 584 * 585 * note that the following calculation is not 586 * needed to be precise. wasting some ip_id is fine. 587 */ 588 589 unsigned int segsz = m->m_pkthdr.segsz; 590 unsigned int datasz = ntohs(ip->ip_len) - hlen; 591 unsigned int num = howmany(datasz, segsz); 592 593 ip->ip_id = ip_newid_range(ia, num); 594 } 595 } 596 if (ia != NULL) { 597 ia4_release(ia, &psref_ia); 598 ia = NULL; 599 } 600 601 /* 602 * If we're doing Path MTU Discovery, we need to set DF unless 603 * the route's MTU is locked. 604 */ 605 if ((flags & IP_MTUDISC) != 0 && rtmtu_nolock) { 606 ip->ip_off |= htons(IP_DF); 607 } 608 609 #ifdef IPSEC 610 if (ipsec_used) { 611 bool ipsec_done = false; 612 613 /* Perform IPsec processing, if any. */ 614 error = ipsec4_output(m, inp, flags, &mtu, &natt_frag, 615 &ipsec_done); 616 if (error || ipsec_done) 617 goto done; 618 } 619 #endif 620 621 /* 622 * Run through list of hooks for output packets. 623 */ 624 error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT); 625 if (error || m == NULL) { 626 IP_STATINC(IP_STAT_PFILDROP_OUT); 627 goto done; 628 } 629 630 ip = mtod(m, struct ip *); 631 hlen = ip->ip_hl << 2; 632 633 m->m_pkthdr.csum_data |= hlen << 16; 634 635 /* 636 * search for the source address structure to 637 * maintain output statistics, and verify address 638 * validity 639 */ 640 KASSERT(ia == NULL); 641 sockaddr_in_init(&usrc.sin, &ip->ip_src, 0); 642 ifa = ifaof_ifpforaddr_psref(&usrc.sa, ifp, &psref_ia); 643 if (ifa != NULL) 644 ia = ifatoia(ifa); 645 646 /* 647 * Ensure we only send from a valid address. 648 * A NULL address is valid because the packet could be 649 * generated from a packet filter. 650 */ 651 if (ia != NULL && (flags & IP_FORWARDING) == 0 && 652 (error = ip_ifaddrvalid(ia)) != 0) 653 { 654 ARPLOG(LOG_ERR, 655 "refusing to send from invalid address %s (pid %d)\n", 656 ARPLOGADDR(&ip->ip_src), curproc->p_pid); 657 IP_STATINC(IP_STAT_ODROPPED); 658 if (error == 1) 659 /* 660 * Address exists, but is tentative or detached. 661 * We can't send from it because it's invalid, 662 * so we drop the packet. 663 */ 664 error = 0; 665 else 666 error = EADDRNOTAVAIL; 667 goto bad; 668 } 669 670 /* Maybe skip checksums on loopback interfaces. */ 671 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 672 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 673 } 674 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 675 676 /* 677 * If small enough for mtu of path, or if using TCP segmentation 678 * offload, can just send directly. 679 */ 680 if (ntohs(ip->ip_len) <= mtu || 681 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 682 const struct sockaddr *sa; 683 684 #if IFA_STATS 685 if (ia) 686 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 687 #endif 688 /* 689 * Always initialize the sum to 0! Some HW assisted 690 * checksumming requires this. 691 */ 692 ip->ip_sum = 0; 693 694 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 695 /* 696 * Perform any checksums that the hardware can't do 697 * for us. 698 * 699 * XXX Does any hardware require the {th,uh}_sum 700 * XXX fields to be 0? 701 */ 702 if (sw_csum & M_CSUM_IPv4) { 703 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 704 ip->ip_sum = in_cksum(m, hlen); 705 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 706 } 707 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 708 if (IN_NEED_CHECKSUM(ifp, 709 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 710 in_undefer_cksum_tcpudp(m); 711 } 712 m->m_pkthdr.csum_flags &= 713 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 714 } 715 } 716 717 sa = (m->m_flags & M_MCAST) ? sintocsa(rdst) : sintocsa(dst); 718 if (__predict_false(sw_csum & M_CSUM_TSOv4)) { 719 /* 720 * TSO4 is required by a packet, but disabled for 721 * the interface. 722 */ 723 error = ip_tso_output(ifp, m, sa, rt); 724 } else 725 error = ip_if_output(ifp, m, sa, rt); 726 goto done; 727 } 728 729 /* 730 * We can't use HW checksumming if we're about to fragment the packet. 731 * 732 * XXX Some hardware can do this. 733 */ 734 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 735 if (IN_NEED_CHECKSUM(ifp, 736 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 737 in_undefer_cksum_tcpudp(m); 738 } 739 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 740 } 741 742 /* 743 * Too large for interface; fragment if possible. 744 * Must be able to put at least 8 bytes per fragment. 745 */ 746 if (ntohs(ip->ip_off) & IP_DF) { 747 if (flags & IP_RETURNMTU) { 748 KASSERT(inp != NULL); 749 inp->inp_errormtu = mtu; 750 } 751 error = EMSGSIZE; 752 IP_STATINC(IP_STAT_CANTFRAG); 753 goto bad; 754 } 755 756 error = ip_fragment(m, ifp, mtu); 757 if (error) { 758 m = NULL; 759 goto bad; 760 } 761 762 for (; m; m = m0) { 763 m0 = m->m_nextpkt; 764 m->m_nextpkt = NULL; 765 if (error) { 766 m_freem(m); 767 continue; 768 } 769 #if IFA_STATS 770 if (ia) 771 ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len); 772 #endif 773 /* 774 * If we get there, the packet has not been handled by 775 * IPsec whereas it should have. Now that it has been 776 * fragmented, re-inject it in ip_output so that IPsec 777 * processing can occur. 778 */ 779 if (natt_frag) { 780 error = ip_output(m, opt, ro, 781 flags | IP_RAWOUTPUT | IP_NOIPNEWID, 782 imo, inp); 783 } else { 784 KASSERT((m->m_pkthdr.csum_flags & 785 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 786 error = ip_if_output(ifp, m, 787 (m->m_flags & M_MCAST) ? 788 sintocsa(rdst) : sintocsa(dst), rt); 789 } 790 } 791 if (error == 0) { 792 IP_STATINC(IP_STAT_FRAGMENTED); 793 } 794 795 done: 796 ia4_release(ia, &psref_ia); 797 rtcache_unref(rt, ro); 798 if (ro == &iproute) { 799 rtcache_free(&iproute); 800 } 801 if (mifp != NULL) { 802 if_put(mifp, &psref); 803 } 804 if (bind_need_restore) 805 curlwp_bindx(bound); 806 return error; 807 808 bad: 809 m_freem(m); 810 goto done; 811 } 812 813 int 814 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 815 { 816 struct ip *ip, *mhip; 817 struct mbuf *m0; 818 int len, hlen, off; 819 int mhlen, firstlen; 820 struct mbuf **mnext; 821 int sw_csum = m->m_pkthdr.csum_flags; 822 int fragments = 0; 823 int error = 0; 824 int ipoff, ipflg; 825 826 ip = mtod(m, struct ip *); 827 hlen = ip->ip_hl << 2; 828 829 /* Preserve the offset and flags. */ 830 ipoff = ntohs(ip->ip_off) & IP_OFFMASK; 831 ipflg = ntohs(ip->ip_off) & (IP_RF|IP_DF|IP_MF); 832 833 if (ifp != NULL) 834 sw_csum &= ~ifp->if_csum_flags_tx; 835 836 len = (mtu - hlen) &~ 7; 837 if (len < 8) { 838 m_freem(m); 839 return EMSGSIZE; 840 } 841 842 firstlen = len; 843 mnext = &m->m_nextpkt; 844 845 /* 846 * Loop through length of segment after first fragment, 847 * make new header and copy data of each part and link onto chain. 848 */ 849 m0 = m; 850 mhlen = sizeof(struct ip); 851 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 852 MGETHDR(m, M_DONTWAIT, MT_HEADER); 853 if (m == NULL) { 854 error = ENOBUFS; 855 IP_STATINC(IP_STAT_ODROPPED); 856 goto sendorfree; 857 } 858 MCLAIM(m, m0->m_owner); 859 860 *mnext = m; 861 mnext = &m->m_nextpkt; 862 863 m->m_data += max_linkhdr; 864 mhip = mtod(m, struct ip *); 865 *mhip = *ip; 866 867 /* we must inherit the flags */ 868 m->m_flags |= m0->m_flags & M_COPYFLAGS; 869 870 if (hlen > sizeof(struct ip)) { 871 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip); 872 mhip->ip_hl = mhlen >> 2; 873 } 874 m->m_len = mhlen; 875 876 mhip->ip_off = ((off - hlen) >> 3) + ipoff; 877 mhip->ip_off |= ipflg; 878 if (off + len >= ntohs(ip->ip_len)) 879 len = ntohs(ip->ip_len) - off; 880 else 881 mhip->ip_off |= IP_MF; 882 HTONS(mhip->ip_off); 883 884 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 885 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 886 if (m->m_next == NULL) { 887 error = ENOBUFS; 888 IP_STATINC(IP_STAT_ODROPPED); 889 goto sendorfree; 890 } 891 892 m->m_pkthdr.len = mhlen + len; 893 m_reset_rcvif(m); 894 895 mhip->ip_sum = 0; 896 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 897 if (sw_csum & M_CSUM_IPv4) { 898 mhip->ip_sum = in_cksum(m, mhlen); 899 } else { 900 /* 901 * checksum is hw-offloaded or not necessary. 902 */ 903 m->m_pkthdr.csum_flags |= 904 m0->m_pkthdr.csum_flags & M_CSUM_IPv4; 905 m->m_pkthdr.csum_data |= mhlen << 16; 906 KASSERT(!(ifp != NULL && 907 IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) || 908 (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 909 } 910 IP_STATINC(IP_STAT_OFRAGMENTS); 911 fragments++; 912 } 913 914 /* 915 * Update first fragment by trimming what's been copied out 916 * and updating header, then send each fragment (in order). 917 */ 918 m = m0; 919 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 920 m->m_pkthdr.len = hlen + firstlen; 921 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 922 ip->ip_off |= htons(IP_MF); 923 ip->ip_sum = 0; 924 if (sw_csum & M_CSUM_IPv4) { 925 ip->ip_sum = in_cksum(m, hlen); 926 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 927 } else { 928 /* 929 * checksum is hw-offloaded or not necessary. 930 */ 931 KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) || 932 (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0); 933 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 934 sizeof(struct ip)); 935 } 936 937 sendorfree: 938 /* 939 * If there is no room for all the fragments, don't queue 940 * any of them. 941 */ 942 if (ifp != NULL) { 943 IFQ_LOCK(&ifp->if_snd); 944 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 945 error == 0) { 946 error = ENOBUFS; 947 IP_STATINC(IP_STAT_ODROPPED); 948 IFQ_INC_DROPS(&ifp->if_snd); 949 } 950 IFQ_UNLOCK(&ifp->if_snd); 951 } 952 if (error) { 953 for (m = m0; m; m = m0) { 954 m0 = m->m_nextpkt; 955 m->m_nextpkt = NULL; 956 m_freem(m); 957 } 958 } 959 960 return error; 961 } 962 963 /* 964 * Determine the maximum length of the options to be inserted; 965 * we would far rather allocate too much space rather than too little. 966 */ 967 u_int 968 ip_optlen(struct inpcb *inp) 969 { 970 struct mbuf *m = inp->inp_options; 971 972 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) { 973 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 974 } 975 return 0; 976 } 977 978 /* 979 * Insert IP options into preformed packet. 980 * Adjust IP destination as required for IP source routing, 981 * as indicated by a non-zero in_addr at the start of the options. 982 */ 983 static struct mbuf * 984 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 985 { 986 struct ipoption *p = mtod(opt, struct ipoption *); 987 struct mbuf *n; 988 struct ip *ip = mtod(m, struct ip *); 989 unsigned optlen; 990 991 optlen = opt->m_len - sizeof(p->ipopt_dst); 992 KASSERT(optlen % 4 == 0); 993 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 994 return m; /* XXX should fail */ 995 if (!in_nullhost(p->ipopt_dst)) 996 ip->ip_dst = p->ipopt_dst; 997 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 998 MGETHDR(n, M_DONTWAIT, MT_HEADER); 999 if (n == NULL) 1000 return m; 1001 MCLAIM(n, m->m_owner); 1002 m_move_pkthdr(n, m); 1003 m->m_len -= sizeof(struct ip); 1004 m->m_data += sizeof(struct ip); 1005 n->m_next = m; 1006 n->m_len = optlen + sizeof(struct ip); 1007 n->m_data += max_linkhdr; 1008 memcpy(mtod(n, void *), ip, sizeof(struct ip)); 1009 m = n; 1010 } else { 1011 m->m_data -= optlen; 1012 m->m_len += optlen; 1013 memmove(mtod(m, void *), ip, sizeof(struct ip)); 1014 } 1015 m->m_pkthdr.len += optlen; 1016 ip = mtod(m, struct ip *); 1017 memcpy(ip + 1, p->ipopt_list, optlen); 1018 *phlen = sizeof(struct ip) + optlen; 1019 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1020 return m; 1021 } 1022 1023 /* 1024 * Copy options from ipsrc to ipdst, omitting those not copied during 1025 * fragmentation. 1026 */ 1027 int 1028 ip_optcopy(struct ip *ipsrc, struct ip *ipdst) 1029 { 1030 u_char *cp, *dp; 1031 int opt, optlen, cnt; 1032 1033 cp = (u_char *)(ipsrc + 1); 1034 dp = (u_char *)(ipdst + 1); 1035 cnt = (ipsrc->ip_hl << 2) - sizeof(struct ip); 1036 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1037 opt = cp[0]; 1038 if (opt == IPOPT_EOL) 1039 break; 1040 if (opt == IPOPT_NOP) { 1041 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1042 *dp++ = IPOPT_NOP; 1043 optlen = 1; 1044 continue; 1045 } 1046 1047 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp)); 1048 optlen = cp[IPOPT_OLEN]; 1049 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt); 1050 1051 /* Invalid lengths should have been caught by ip_dooptions. */ 1052 if (optlen > cnt) 1053 optlen = cnt; 1054 if (IPOPT_COPIED(opt)) { 1055 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 1056 dp += optlen; 1057 } 1058 } 1059 1060 for (optlen = dp - (u_char *)(ipdst+1); optlen & 0x3; optlen++) { 1061 *dp++ = IPOPT_EOL; 1062 } 1063 1064 return optlen; 1065 } 1066 1067 /* 1068 * IP socket option processing. 1069 */ 1070 int 1071 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1072 { 1073 struct inpcb *inp = sotoinpcb(so); 1074 struct ip *ip = &inp->inp_ip; 1075 int inpflags = inp->inp_flags; 1076 int optval = 0, error = 0; 1077 struct in_pktinfo pktinfo; 1078 1079 KASSERT(solocked(so)); 1080 1081 if (sopt->sopt_level != IPPROTO_IP) { 1082 if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER) 1083 return 0; 1084 return ENOPROTOOPT; 1085 } 1086 1087 switch (op) { 1088 case PRCO_SETOPT: 1089 switch (sopt->sopt_name) { 1090 case IP_OPTIONS: 1091 #ifdef notyet 1092 case IP_RETOPTS: 1093 #endif 1094 error = ip_pcbopts(inp, sopt); 1095 break; 1096 1097 case IP_TOS: 1098 case IP_TTL: 1099 case IP_MINTTL: 1100 case IP_RECVOPTS: 1101 case IP_RECVRETOPTS: 1102 case IP_RECVDSTADDR: 1103 case IP_RECVIF: 1104 case IP_RECVPKTINFO: 1105 case IP_RECVTTL: 1106 error = sockopt_getint(sopt, &optval); 1107 if (error) 1108 break; 1109 1110 switch (sopt->sopt_name) { 1111 case IP_TOS: 1112 ip->ip_tos = optval; 1113 break; 1114 1115 case IP_TTL: 1116 ip->ip_ttl = optval; 1117 break; 1118 1119 case IP_MINTTL: 1120 if (optval > 0 && optval <= MAXTTL) 1121 inp->inp_ip_minttl = optval; 1122 else 1123 error = EINVAL; 1124 break; 1125 #define OPTSET(bit) \ 1126 if (optval) \ 1127 inpflags |= bit; \ 1128 else \ 1129 inpflags &= ~bit; 1130 1131 case IP_RECVOPTS: 1132 OPTSET(INP_RECVOPTS); 1133 break; 1134 1135 case IP_RECVPKTINFO: 1136 OPTSET(INP_RECVPKTINFO); 1137 break; 1138 1139 case IP_RECVRETOPTS: 1140 OPTSET(INP_RECVRETOPTS); 1141 break; 1142 1143 case IP_RECVDSTADDR: 1144 OPTSET(INP_RECVDSTADDR); 1145 break; 1146 1147 case IP_RECVIF: 1148 OPTSET(INP_RECVIF); 1149 break; 1150 1151 case IP_RECVTTL: 1152 OPTSET(INP_RECVTTL); 1153 break; 1154 } 1155 break; 1156 case IP_PKTINFO: 1157 error = sockopt_getint(sopt, &optval); 1158 if (!error) { 1159 /* Linux compatibility */ 1160 OPTSET(INP_RECVPKTINFO); 1161 break; 1162 } 1163 error = sockopt_get(sopt, &pktinfo, sizeof(pktinfo)); 1164 if (error) 1165 break; 1166 1167 if (pktinfo.ipi_ifindex == 0) { 1168 inp->inp_prefsrcip = pktinfo.ipi_addr; 1169 break; 1170 } 1171 1172 /* Solaris compatibility */ 1173 struct ifnet *ifp; 1174 struct in_ifaddr *ia; 1175 int s; 1176 1177 /* pick up primary address */ 1178 s = pserialize_read_enter(); 1179 ifp = if_byindex(pktinfo.ipi_ifindex); 1180 if (ifp == NULL) { 1181 pserialize_read_exit(s); 1182 error = EADDRNOTAVAIL; 1183 break; 1184 } 1185 ia = in_get_ia_from_ifp(ifp); 1186 if (ia == NULL) { 1187 pserialize_read_exit(s); 1188 error = EADDRNOTAVAIL; 1189 break; 1190 } 1191 inp->inp_prefsrcip = IA_SIN(ia)->sin_addr; 1192 pserialize_read_exit(s); 1193 break; 1194 break; 1195 #undef OPTSET 1196 1197 case IP_MULTICAST_IF: 1198 case IP_MULTICAST_TTL: 1199 case IP_MULTICAST_LOOP: 1200 case IP_ADD_MEMBERSHIP: 1201 case IP_DROP_MEMBERSHIP: 1202 error = ip_setmoptions(&inp->inp_moptions, sopt); 1203 break; 1204 1205 case IP_PORTRANGE: 1206 error = sockopt_getint(sopt, &optval); 1207 if (error) 1208 break; 1209 1210 switch (optval) { 1211 case IP_PORTRANGE_DEFAULT: 1212 case IP_PORTRANGE_HIGH: 1213 inpflags &= ~(INP_LOWPORT); 1214 break; 1215 1216 case IP_PORTRANGE_LOW: 1217 inpflags |= INP_LOWPORT; 1218 break; 1219 1220 default: 1221 error = EINVAL; 1222 break; 1223 } 1224 break; 1225 1226 case IP_PORTALGO: 1227 error = sockopt_getint(sopt, &optval); 1228 if (error) 1229 break; 1230 1231 error = portalgo_algo_index_select( 1232 (struct inpcb_hdr *)inp, optval); 1233 break; 1234 1235 #if defined(IPSEC) 1236 case IP_IPSEC_POLICY: 1237 if (ipsec_enabled) { 1238 error = ipsec_set_policy(inp, 1239 sopt->sopt_data, sopt->sopt_size, 1240 curlwp->l_cred); 1241 } else 1242 error = ENOPROTOOPT; 1243 break; 1244 #endif /* IPSEC */ 1245 1246 default: 1247 error = ENOPROTOOPT; 1248 break; 1249 } 1250 break; 1251 1252 case PRCO_GETOPT: 1253 switch (sopt->sopt_name) { 1254 case IP_OPTIONS: 1255 case IP_RETOPTS: { 1256 struct mbuf *mopts = inp->inp_options; 1257 1258 if (mopts) { 1259 struct mbuf *m; 1260 1261 m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT); 1262 if (m == NULL) { 1263 error = ENOBUFS; 1264 break; 1265 } 1266 error = sockopt_setmbuf(sopt, m); 1267 } 1268 break; 1269 } 1270 case IP_TOS: 1271 case IP_TTL: 1272 case IP_MINTTL: 1273 case IP_RECVOPTS: 1274 case IP_RECVRETOPTS: 1275 case IP_RECVDSTADDR: 1276 case IP_RECVIF: 1277 case IP_RECVPKTINFO: 1278 case IP_RECVTTL: 1279 case IP_ERRORMTU: 1280 switch (sopt->sopt_name) { 1281 case IP_TOS: 1282 optval = ip->ip_tos; 1283 break; 1284 1285 case IP_TTL: 1286 optval = ip->ip_ttl; 1287 break; 1288 1289 case IP_MINTTL: 1290 optval = inp->inp_ip_minttl; 1291 break; 1292 1293 case IP_ERRORMTU: 1294 optval = inp->inp_errormtu; 1295 break; 1296 1297 #define OPTBIT(bit) (inpflags & bit ? 1 : 0) 1298 1299 case IP_RECVOPTS: 1300 optval = OPTBIT(INP_RECVOPTS); 1301 break; 1302 1303 case IP_RECVPKTINFO: 1304 optval = OPTBIT(INP_RECVPKTINFO); 1305 break; 1306 1307 case IP_RECVRETOPTS: 1308 optval = OPTBIT(INP_RECVRETOPTS); 1309 break; 1310 1311 case IP_RECVDSTADDR: 1312 optval = OPTBIT(INP_RECVDSTADDR); 1313 break; 1314 1315 case IP_RECVIF: 1316 optval = OPTBIT(INP_RECVIF); 1317 break; 1318 1319 case IP_RECVTTL: 1320 optval = OPTBIT(INP_RECVTTL); 1321 break; 1322 } 1323 error = sockopt_setint(sopt, optval); 1324 break; 1325 1326 case IP_PKTINFO: 1327 switch (sopt->sopt_size) { 1328 case sizeof(int): 1329 /* Linux compatibility */ 1330 optval = OPTBIT(INP_RECVPKTINFO); 1331 error = sockopt_setint(sopt, optval); 1332 break; 1333 case sizeof(struct in_pktinfo): 1334 /* Solaris compatibility */ 1335 pktinfo.ipi_ifindex = 0; 1336 pktinfo.ipi_addr = inp->inp_prefsrcip; 1337 error = sockopt_set(sopt, &pktinfo, 1338 sizeof(pktinfo)); 1339 break; 1340 default: 1341 /* 1342 * While size is stuck at 0, and, later, if 1343 * the caller doesn't use an exactly sized 1344 * recipient for the data, default to Linux 1345 * compatibility 1346 */ 1347 optval = OPTBIT(INP_RECVPKTINFO); 1348 error = sockopt_setint(sopt, optval); 1349 break; 1350 } 1351 break; 1352 1353 #if 0 /* defined(IPSEC) */ 1354 case IP_IPSEC_POLICY: 1355 { 1356 struct mbuf *m = NULL; 1357 1358 /* XXX this will return EINVAL as sopt is empty */ 1359 error = ipsec_get_policy(inp, sopt->sopt_data, 1360 sopt->sopt_size, &m); 1361 if (error == 0) 1362 error = sockopt_setmbuf(sopt, m); 1363 break; 1364 } 1365 #endif /*IPSEC*/ 1366 1367 case IP_MULTICAST_IF: 1368 case IP_MULTICAST_TTL: 1369 case IP_MULTICAST_LOOP: 1370 case IP_ADD_MEMBERSHIP: 1371 case IP_DROP_MEMBERSHIP: 1372 error = ip_getmoptions(inp->inp_moptions, sopt); 1373 break; 1374 1375 case IP_PORTRANGE: 1376 if (inpflags & INP_LOWPORT) 1377 optval = IP_PORTRANGE_LOW; 1378 else 1379 optval = IP_PORTRANGE_DEFAULT; 1380 error = sockopt_setint(sopt, optval); 1381 break; 1382 1383 case IP_PORTALGO: 1384 optval = inp->inp_portalgo; 1385 error = sockopt_setint(sopt, optval); 1386 break; 1387 1388 default: 1389 error = ENOPROTOOPT; 1390 break; 1391 } 1392 break; 1393 } 1394 1395 if (!error) { 1396 inp->inp_flags = inpflags; 1397 } 1398 return error; 1399 } 1400 1401 static int 1402 ip_pktinfo_prepare(const struct in_pktinfo *pktinfo, struct ip_pktopts *pktopts, 1403 int *flags, kauth_cred_t cred) 1404 { 1405 struct ip_moptions *imo; 1406 int error = 0; 1407 bool addrset = false; 1408 1409 if (!in_nullhost(pktinfo->ipi_addr)) { 1410 pktopts->ippo_laddr.sin_addr = pktinfo->ipi_addr; 1411 /* EADDRNOTAVAIL? */ 1412 error = in_pcbbindableaddr(&pktopts->ippo_laddr, cred); 1413 if (error != 0) 1414 return error; 1415 addrset = true; 1416 } 1417 1418 if (pktinfo->ipi_ifindex != 0) { 1419 if (!addrset) { 1420 struct ifnet *ifp; 1421 struct in_ifaddr *ia; 1422 int s; 1423 1424 /* pick up primary address */ 1425 s = pserialize_read_enter(); 1426 ifp = if_byindex(pktinfo->ipi_ifindex); 1427 if (ifp == NULL) { 1428 pserialize_read_exit(s); 1429 return EADDRNOTAVAIL; 1430 } 1431 ia = in_get_ia_from_ifp(ifp); 1432 if (ia == NULL) { 1433 pserialize_read_exit(s); 1434 return EADDRNOTAVAIL; 1435 } 1436 pktopts->ippo_laddr.sin_addr = IA_SIN(ia)->sin_addr; 1437 pserialize_read_exit(s); 1438 } 1439 1440 /* 1441 * If specified ipi_ifindex, 1442 * use copied or locally initialized ip_moptions. 1443 * Original ip_moptions must not be modified. 1444 */ 1445 imo = &pktopts->ippo_imobuf; /* local buf in pktopts */ 1446 if (pktopts->ippo_imo != NULL) { 1447 memcpy(imo, pktopts->ippo_imo, sizeof(*imo)); 1448 } else { 1449 memset(imo, 0, sizeof(*imo)); 1450 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1451 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1452 } 1453 imo->imo_multicast_if_index = pktinfo->ipi_ifindex; 1454 pktopts->ippo_imo = imo; 1455 *flags |= IP_ROUTETOIFINDEX; 1456 } 1457 return error; 1458 } 1459 1460 /* 1461 * Set up IP outgoing packet options. Even if control is NULL, 1462 * pktopts->ippo_laddr and pktopts->ippo_imo are set and used. 1463 */ 1464 int 1465 ip_setpktopts(struct mbuf *control, struct ip_pktopts *pktopts, int *flags, 1466 struct inpcb *inp, kauth_cred_t cred) 1467 { 1468 struct cmsghdr *cm; 1469 struct in_pktinfo pktinfo; 1470 int error; 1471 1472 pktopts->ippo_imo = inp->inp_moptions; 1473 1474 struct in_addr *ia = in_nullhost(inp->inp_prefsrcip) ? &inp->inp_laddr : 1475 &inp->inp_prefsrcip; 1476 sockaddr_in_init(&pktopts->ippo_laddr, ia, 0); 1477 1478 if (control == NULL) 1479 return 0; 1480 1481 /* 1482 * XXX: Currently, we assume all the optional information is 1483 * stored in a single mbuf. 1484 */ 1485 if (control->m_next) 1486 return EINVAL; 1487 1488 for (; control->m_len > 0; 1489 control->m_data += CMSG_ALIGN(cm->cmsg_len), 1490 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 1491 cm = mtod(control, struct cmsghdr *); 1492 if ((control->m_len < sizeof(*cm)) || 1493 (cm->cmsg_len == 0) || 1494 (cm->cmsg_len > control->m_len)) { 1495 return EINVAL; 1496 } 1497 if (cm->cmsg_level != IPPROTO_IP) 1498 continue; 1499 1500 switch (cm->cmsg_type) { 1501 case IP_PKTINFO: 1502 if (cm->cmsg_len != CMSG_LEN(sizeof(pktinfo))) 1503 return EINVAL; 1504 memcpy(&pktinfo, CMSG_DATA(cm), sizeof(pktinfo)); 1505 error = ip_pktinfo_prepare(&pktinfo, pktopts, flags, 1506 cred); 1507 if (error) 1508 return error; 1509 break; 1510 case IP_SENDSRCADDR: /* FreeBSD compatibility */ 1511 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_addr))) 1512 return EINVAL; 1513 pktinfo.ipi_ifindex = 0; 1514 pktinfo.ipi_addr = 1515 ((struct in_pktinfo *)CMSG_DATA(cm))->ipi_addr; 1516 error = ip_pktinfo_prepare(&pktinfo, pktopts, flags, 1517 cred); 1518 if (error) 1519 return error; 1520 break; 1521 default: 1522 return ENOPROTOOPT; 1523 } 1524 } 1525 return 0; 1526 } 1527 1528 /* 1529 * Set up IP options in pcb for insertion in output packets. 1530 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1531 * with destination address if source routed. 1532 */ 1533 static int 1534 ip_pcbopts(struct inpcb *inp, const struct sockopt *sopt) 1535 { 1536 struct mbuf *m; 1537 const u_char *cp; 1538 u_char *dp; 1539 int cnt; 1540 1541 KASSERT(inp_locked(inp)); 1542 1543 /* Turn off any old options. */ 1544 if (inp->inp_options) { 1545 m_free(inp->inp_options); 1546 } 1547 inp->inp_options = NULL; 1548 if ((cnt = sopt->sopt_size) == 0) { 1549 /* Only turning off any previous options. */ 1550 return 0; 1551 } 1552 cp = sopt->sopt_data; 1553 1554 if (cnt % 4) { 1555 /* Must be 4-byte aligned, because there's no padding. */ 1556 return EINVAL; 1557 } 1558 1559 m = m_get(M_DONTWAIT, MT_SOOPTS); 1560 if (m == NULL) 1561 return ENOBUFS; 1562 1563 dp = mtod(m, u_char *); 1564 memset(dp, 0, sizeof(struct in_addr)); 1565 dp += sizeof(struct in_addr); 1566 m->m_len = sizeof(struct in_addr); 1567 1568 /* 1569 * IP option list according to RFC791. Each option is of the form 1570 * 1571 * [optval] [olen] [(olen - 2) data bytes] 1572 * 1573 * We validate the list and copy options to an mbuf for prepending 1574 * to data packets. The IP first-hop destination address will be 1575 * stored before actual options and is zero if unset. 1576 */ 1577 while (cnt > 0) { 1578 uint8_t optval, olen, offset; 1579 1580 optval = cp[IPOPT_OPTVAL]; 1581 1582 if (optval == IPOPT_EOL || optval == IPOPT_NOP) { 1583 olen = 1; 1584 } else { 1585 if (cnt < IPOPT_OLEN + 1) 1586 goto bad; 1587 1588 olen = cp[IPOPT_OLEN]; 1589 if (olen < IPOPT_OLEN + 1 || olen > cnt) 1590 goto bad; 1591 } 1592 1593 if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) { 1594 /* 1595 * user process specifies route as: 1596 * ->A->B->C->D 1597 * D must be our final destination (but we can't 1598 * check that since we may not have connected yet). 1599 * A is first hop destination, which doesn't appear in 1600 * actual IP option, but is stored before the options. 1601 */ 1602 if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr)) 1603 goto bad; 1604 1605 offset = cp[IPOPT_OFFSET]; 1606 memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1, 1607 sizeof(struct in_addr)); 1608 1609 cp += sizeof(struct in_addr); 1610 cnt -= sizeof(struct in_addr); 1611 olen -= sizeof(struct in_addr); 1612 1613 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1614 goto bad; 1615 1616 memcpy(dp, cp, olen); 1617 dp[IPOPT_OPTVAL] = optval; 1618 dp[IPOPT_OLEN] = olen; 1619 dp[IPOPT_OFFSET] = offset; 1620 break; 1621 } else { 1622 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1623 goto bad; 1624 1625 memcpy(dp, cp, olen); 1626 break; 1627 } 1628 1629 dp += olen; 1630 m->m_len += olen; 1631 1632 if (optval == IPOPT_EOL) 1633 break; 1634 1635 cp += olen; 1636 cnt -= olen; 1637 } 1638 1639 inp->inp_options = m; 1640 return 0; 1641 1642 bad: 1643 (void)m_free(m); 1644 return EINVAL; 1645 } 1646 1647 /* 1648 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1649 * Must be called in a pserialize critical section. 1650 */ 1651 static struct ifnet * 1652 ip_multicast_if(struct in_addr *a, int *ifindexp) 1653 { 1654 int ifindex; 1655 struct ifnet *ifp = NULL; 1656 struct in_ifaddr *ia; 1657 1658 if (ifindexp) 1659 *ifindexp = 0; 1660 if (ntohl(a->s_addr) >> 24 == 0) { 1661 ifindex = ntohl(a->s_addr) & 0xffffff; 1662 ifp = if_byindex(ifindex); 1663 if (!ifp) 1664 return NULL; 1665 if (ifindexp) 1666 *ifindexp = ifindex; 1667 } else { 1668 IN_ADDRHASH_READER_FOREACH(ia, a->s_addr) { 1669 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1670 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1671 ifp = ia->ia_ifp; 1672 if (if_is_deactivated(ifp)) 1673 ifp = NULL; 1674 break; 1675 } 1676 } 1677 } 1678 return ifp; 1679 } 1680 1681 static int 1682 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval) 1683 { 1684 u_int tval; 1685 u_char cval; 1686 int error; 1687 1688 if (sopt == NULL) 1689 return EINVAL; 1690 1691 switch (sopt->sopt_size) { 1692 case sizeof(u_char): 1693 error = sockopt_get(sopt, &cval, sizeof(u_char)); 1694 tval = cval; 1695 break; 1696 1697 case sizeof(u_int): 1698 error = sockopt_get(sopt, &tval, sizeof(u_int)); 1699 break; 1700 1701 default: 1702 error = EINVAL; 1703 } 1704 1705 if (error) 1706 return error; 1707 1708 if (tval > maxval) 1709 return EINVAL; 1710 1711 *val = tval; 1712 return 0; 1713 } 1714 1715 static int 1716 ip_get_membership(const struct sockopt *sopt, struct ifnet **ifp, 1717 struct psref *psref, struct in_addr *ia, bool add) 1718 { 1719 int error; 1720 struct ip_mreq mreq; 1721 1722 error = sockopt_get(sopt, &mreq, sizeof(mreq)); 1723 if (error) 1724 return error; 1725 1726 if (!IN_MULTICAST(mreq.imr_multiaddr.s_addr)) 1727 return EINVAL; 1728 1729 memcpy(ia, &mreq.imr_multiaddr, sizeof(*ia)); 1730 1731 if (in_nullhost(mreq.imr_interface)) { 1732 union { 1733 struct sockaddr dst; 1734 struct sockaddr_in dst4; 1735 } u; 1736 struct route ro; 1737 1738 if (!add) { 1739 *ifp = NULL; 1740 return 0; 1741 } 1742 /* 1743 * If no interface address was provided, use the interface of 1744 * the route to the given multicast address. 1745 */ 1746 struct rtentry *rt; 1747 memset(&ro, 0, sizeof(ro)); 1748 1749 sockaddr_in_init(&u.dst4, ia, 0); 1750 error = rtcache_setdst(&ro, &u.dst); 1751 if (error != 0) 1752 return error; 1753 *ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL; 1754 if (*ifp != NULL) { 1755 if (if_is_deactivated(*ifp)) 1756 *ifp = NULL; 1757 else 1758 if_acquire(*ifp, psref); 1759 } 1760 rtcache_unref(rt, &ro); 1761 rtcache_free(&ro); 1762 } else { 1763 int s = pserialize_read_enter(); 1764 *ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1765 if (!add && *ifp == NULL) { 1766 pserialize_read_exit(s); 1767 return EADDRNOTAVAIL; 1768 } 1769 if (*ifp != NULL) { 1770 if (if_is_deactivated(*ifp)) 1771 *ifp = NULL; 1772 else 1773 if_acquire(*ifp, psref); 1774 } 1775 pserialize_read_exit(s); 1776 } 1777 return 0; 1778 } 1779 1780 /* 1781 * Add a multicast group membership. 1782 * Group must be a valid IP multicast address. 1783 */ 1784 static int 1785 ip_add_membership(struct ip_moptions *imo, const struct sockopt *sopt) 1786 { 1787 struct ifnet *ifp = NULL; // XXX: gcc [ppc] 1788 struct in_addr ia; 1789 int i, error, bound; 1790 struct psref psref; 1791 1792 /* imo is protected by solock or referenced only by the caller */ 1793 1794 bound = curlwp_bind(); 1795 if (sopt->sopt_size == sizeof(struct ip_mreq)) 1796 error = ip_get_membership(sopt, &ifp, &psref, &ia, true); 1797 else { 1798 #ifdef INET6 1799 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 1800 #else 1801 error = EINVAL; 1802 #endif 1803 } 1804 1805 if (error) 1806 goto out; 1807 1808 /* 1809 * See if we found an interface, and confirm that it 1810 * supports multicast. 1811 */ 1812 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1813 error = EADDRNOTAVAIL; 1814 goto out; 1815 } 1816 1817 /* 1818 * See if the membership already exists or if all the 1819 * membership slots are full. 1820 */ 1821 for (i = 0; i < imo->imo_num_memberships; ++i) { 1822 if (imo->imo_membership[i]->inm_ifp == ifp && 1823 in_hosteq(imo->imo_membership[i]->inm_addr, ia)) 1824 break; 1825 } 1826 if (i < imo->imo_num_memberships) { 1827 error = EADDRINUSE; 1828 goto out; 1829 } 1830 1831 if (i == IP_MAX_MEMBERSHIPS) { 1832 error = ETOOMANYREFS; 1833 goto out; 1834 } 1835 1836 /* 1837 * Everything looks good; add a new record to the multicast 1838 * address list for the given interface. 1839 */ 1840 imo->imo_membership[i] = in_addmulti(&ia, ifp); 1841 if (imo->imo_membership[i] == NULL) { 1842 error = ENOBUFS; 1843 goto out; 1844 } 1845 1846 ++imo->imo_num_memberships; 1847 error = 0; 1848 out: 1849 if_put(ifp, &psref); 1850 curlwp_bindx(bound); 1851 return error; 1852 } 1853 1854 /* 1855 * Drop a multicast group membership. 1856 * Group must be a valid IP multicast address. 1857 */ 1858 static int 1859 ip_drop_membership(struct ip_moptions *imo, const struct sockopt *sopt) 1860 { 1861 struct in_addr ia = { .s_addr = 0 }; // XXX: gcc [ppc] 1862 struct ifnet *ifp = NULL; // XXX: gcc [ppc] 1863 int i, error, bound; 1864 struct psref psref; 1865 1866 /* imo is protected by solock or referenced only by the caller */ 1867 1868 bound = curlwp_bind(); 1869 if (sopt->sopt_size == sizeof(struct ip_mreq)) 1870 error = ip_get_membership(sopt, &ifp, &psref, &ia, false); 1871 else { 1872 #ifdef INET6 1873 error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia)); 1874 #else 1875 error = EINVAL; 1876 #endif 1877 } 1878 1879 if (error) 1880 goto out; 1881 1882 /* 1883 * Find the membership in the membership array. 1884 */ 1885 for (i = 0; i < imo->imo_num_memberships; ++i) { 1886 if ((ifp == NULL || 1887 imo->imo_membership[i]->inm_ifp == ifp) && 1888 in_hosteq(imo->imo_membership[i]->inm_addr, ia)) 1889 break; 1890 } 1891 if (i == imo->imo_num_memberships) { 1892 error = EADDRNOTAVAIL; 1893 goto out; 1894 } 1895 1896 /* 1897 * Give up the multicast address record to which the 1898 * membership points. 1899 */ 1900 in_delmulti(imo->imo_membership[i]); 1901 1902 /* 1903 * Remove the gap in the membership array. 1904 */ 1905 for (++i; i < imo->imo_num_memberships; ++i) 1906 imo->imo_membership[i-1] = imo->imo_membership[i]; 1907 --imo->imo_num_memberships; 1908 error = 0; 1909 out: 1910 if_put(ifp, &psref); 1911 curlwp_bindx(bound); 1912 return error; 1913 } 1914 1915 /* 1916 * Set the IP multicast options in response to user setsockopt(). 1917 */ 1918 int 1919 ip_setmoptions(struct ip_moptions **pimo, const struct sockopt *sopt) 1920 { 1921 struct ip_moptions *imo = *pimo; 1922 struct in_addr addr; 1923 struct ifnet *ifp; 1924 int ifindex, error = 0; 1925 1926 /* The passed imo isn't NULL, it should be protected by solock */ 1927 1928 if (!imo) { 1929 /* 1930 * No multicast option buffer attached to the pcb; 1931 * allocate one and initialize to default values. 1932 */ 1933 imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP); 1934 if (imo == NULL) 1935 return ENOBUFS; 1936 1937 imo->imo_multicast_if_index = 0; 1938 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1939 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1940 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1941 imo->imo_num_memberships = 0; 1942 *pimo = imo; 1943 } 1944 1945 switch (sopt->sopt_name) { 1946 case IP_MULTICAST_IF: { 1947 int s; 1948 /* 1949 * Select the interface for outgoing multicast packets. 1950 */ 1951 error = sockopt_get(sopt, &addr, sizeof(addr)); 1952 if (error) 1953 break; 1954 1955 /* 1956 * INADDR_ANY is used to remove a previous selection. 1957 * When no interface is selected, a default one is 1958 * chosen every time a multicast packet is sent. 1959 */ 1960 if (in_nullhost(addr)) { 1961 imo->imo_multicast_if_index = 0; 1962 break; 1963 } 1964 /* 1965 * The selected interface is identified by its local 1966 * IP address. Find the interface and confirm that 1967 * it supports multicasting. 1968 */ 1969 s = pserialize_read_enter(); 1970 ifp = ip_multicast_if(&addr, &ifindex); 1971 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1972 pserialize_read_exit(s); 1973 error = EADDRNOTAVAIL; 1974 break; 1975 } 1976 imo->imo_multicast_if_index = ifp->if_index; 1977 pserialize_read_exit(s); 1978 if (ifindex) 1979 imo->imo_multicast_addr = addr; 1980 else 1981 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1982 break; 1983 } 1984 1985 case IP_MULTICAST_TTL: 1986 /* 1987 * Set the IP time-to-live for outgoing multicast packets. 1988 */ 1989 error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL); 1990 break; 1991 1992 case IP_MULTICAST_LOOP: 1993 /* 1994 * Set the loopback flag for outgoing multicast packets. 1995 * Must be zero or one. 1996 */ 1997 error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1); 1998 break; 1999 2000 case IP_ADD_MEMBERSHIP: /* IPV6_JOIN_GROUP */ 2001 error = ip_add_membership(imo, sopt); 2002 break; 2003 2004 case IP_DROP_MEMBERSHIP: /* IPV6_LEAVE_GROUP */ 2005 error = ip_drop_membership(imo, sopt); 2006 break; 2007 2008 default: 2009 error = EOPNOTSUPP; 2010 break; 2011 } 2012 2013 /* 2014 * If all options have default values, no need to keep the mbuf. 2015 */ 2016 if (imo->imo_multicast_if_index == 0 && 2017 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2018 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2019 imo->imo_num_memberships == 0) { 2020 kmem_intr_free(imo, sizeof(*imo)); 2021 *pimo = NULL; 2022 } 2023 2024 return error; 2025 } 2026 2027 /* 2028 * Return the IP multicast options in response to user getsockopt(). 2029 */ 2030 int 2031 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt) 2032 { 2033 struct in_addr addr; 2034 uint8_t optval; 2035 int error = 0; 2036 2037 /* imo is protected by solock or refereced only by the caller */ 2038 2039 switch (sopt->sopt_name) { 2040 case IP_MULTICAST_IF: 2041 if (imo == NULL || imo->imo_multicast_if_index == 0) 2042 addr = zeroin_addr; 2043 else if (imo->imo_multicast_addr.s_addr) { 2044 /* return the value user has set */ 2045 addr = imo->imo_multicast_addr; 2046 } else { 2047 struct ifnet *ifp; 2048 struct in_ifaddr *ia = NULL; 2049 int s = pserialize_read_enter(); 2050 2051 ifp = if_byindex(imo->imo_multicast_if_index); 2052 if (ifp != NULL) { 2053 ia = in_get_ia_from_ifp(ifp); 2054 } 2055 addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 2056 pserialize_read_exit(s); 2057 } 2058 error = sockopt_set(sopt, &addr, sizeof(addr)); 2059 break; 2060 2061 case IP_MULTICAST_TTL: 2062 optval = imo ? imo->imo_multicast_ttl 2063 : IP_DEFAULT_MULTICAST_TTL; 2064 2065 error = sockopt_set(sopt, &optval, sizeof(optval)); 2066 break; 2067 2068 case IP_MULTICAST_LOOP: 2069 optval = imo ? imo->imo_multicast_loop 2070 : IP_DEFAULT_MULTICAST_LOOP; 2071 2072 error = sockopt_set(sopt, &optval, sizeof(optval)); 2073 break; 2074 2075 default: 2076 error = EOPNOTSUPP; 2077 } 2078 2079 return error; 2080 } 2081 2082 /* 2083 * Discard the IP multicast options. 2084 */ 2085 void 2086 ip_freemoptions(struct ip_moptions *imo) 2087 { 2088 int i; 2089 2090 /* The owner of imo (inp) should be protected by solock */ 2091 2092 if (imo != NULL) { 2093 for (i = 0; i < imo->imo_num_memberships; ++i) { 2094 struct in_multi *inm = imo->imo_membership[i]; 2095 in_delmulti(inm); 2096 /* ifp should not leave thanks to solock */ 2097 } 2098 2099 kmem_intr_free(imo, sizeof(*imo)); 2100 } 2101 } 2102 2103 /* 2104 * Routine called from ip_output() to loop back a copy of an IP multicast 2105 * packet to the input queue of a specified interface. Note that this 2106 * calls the output routine of the loopback "driver", but with an interface 2107 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 2108 */ 2109 static void 2110 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 2111 { 2112 struct ip *ip; 2113 struct mbuf *copym; 2114 2115 copym = m_copypacket(m, M_DONTWAIT); 2116 if (copym != NULL && 2117 (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 2118 copym = m_pullup(copym, sizeof(struct ip)); 2119 if (copym == NULL) 2120 return; 2121 /* 2122 * We don't bother to fragment if the IP length is greater 2123 * than the interface's MTU. Can this possibly matter? 2124 */ 2125 ip = mtod(copym, struct ip *); 2126 2127 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 2128 in_undefer_cksum_tcpudp(copym); 2129 copym->m_pkthdr.csum_flags &= 2130 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 2131 } 2132 2133 ip->ip_sum = 0; 2134 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 2135 KERNEL_LOCK_UNLESS_NET_MPSAFE(); 2136 (void)looutput(ifp, copym, sintocsa(dst), NULL); 2137 KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 2138 } 2139 2140 /* 2141 * Ensure sending address is valid. 2142 * Returns 0 on success, -1 if an error should be sent back or 1 2143 * if the packet could be dropped without error (protocol dependent). 2144 */ 2145 static int 2146 ip_ifaddrvalid(const struct in_ifaddr *ia) 2147 { 2148 2149 if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY) 2150 return 0; 2151 2152 if (ia->ia4_flags & IN_IFF_DUPLICATED) 2153 return -1; 2154 else if (ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DETACHED)) 2155 return 1; 2156 2157 return 0; 2158 } 2159