1 /* $NetBSD: ip_output.c,v 1.95 2002/02/07 21:47:45 thorpej 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 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69 /* 70 * Copyright (c) 1982, 1986, 1988, 1990, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. All advertising materials mentioning features or use of this software 82 * must display the following acknowledgement: 83 * This product includes software developed by the University of 84 * California, Berkeley and its contributors. 85 * 4. Neither the name of the University nor the names of its contributors 86 * may be used to endorse or promote products derived from this software 87 * without specific prior written permission. 88 * 89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 99 * SUCH DAMAGE. 100 * 101 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 102 */ 103 104 #include <sys/cdefs.h> 105 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.95 2002/02/07 21:47:45 thorpej Exp $"); 106 107 #include "opt_pfil_hooks.h" 108 #include "opt_ipsec.h" 109 #include "opt_mrouting.h" 110 111 #include <sys/param.h> 112 #include <sys/malloc.h> 113 #include <sys/mbuf.h> 114 #include <sys/errno.h> 115 #include <sys/protosw.h> 116 #include <sys/socket.h> 117 #include <sys/socketvar.h> 118 #include <sys/systm.h> 119 #include <sys/proc.h> 120 121 #include <net/if.h> 122 #include <net/route.h> 123 #include <net/pfil.h> 124 125 #include <netinet/in.h> 126 #include <netinet/in_systm.h> 127 #include <netinet/ip.h> 128 #include <netinet/in_pcb.h> 129 #include <netinet/in_var.h> 130 #include <netinet/ip_var.h> 131 132 #ifdef MROUTING 133 #include <netinet/ip_mroute.h> 134 #endif 135 136 #include <machine/stdarg.h> 137 138 #ifdef IPSEC 139 #include <netinet6/ipsec.h> 140 #include <netkey/key.h> 141 #include <netkey/key_debug.h> 142 #endif /*IPSEC*/ 143 144 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *)); 145 static struct ifnet *ip_multicast_if __P((struct in_addr *, int *)); 146 static void ip_mloopback 147 __P((struct ifnet *, struct mbuf *, struct sockaddr_in *)); 148 149 #ifdef PFIL_HOOKS 150 extern struct pfil_head inet_pfil_hook; /* XXX */ 151 #endif 152 153 /* 154 * IP output. The packet in mbuf chain m contains a skeletal IP 155 * header (with len, off, ttl, proto, tos, src, dst). 156 * The mbuf chain containing the packet will be freed. 157 * The mbuf opt, if present, will not be freed. 158 */ 159 int 160 #if __STDC__ 161 ip_output(struct mbuf *m0, ...) 162 #else 163 ip_output(m0, va_alist) 164 struct mbuf *m0; 165 va_dcl 166 #endif 167 { 168 struct ip *ip, *mhip; 169 struct ifnet *ifp; 170 struct mbuf *m = m0; 171 int hlen = sizeof (struct ip); 172 int len, off, error = 0; 173 struct route iproute; 174 struct sockaddr_in *dst; 175 struct in_ifaddr *ia; 176 struct mbuf *opt; 177 struct route *ro; 178 int flags, sw_csum; 179 int *mtu_p; 180 int mtu; 181 struct ip_moptions *imo; 182 va_list ap; 183 #ifdef IPSEC 184 struct socket *so; 185 struct secpolicy *sp = NULL; 186 #endif /*IPSEC*/ 187 u_int16_t ip_len; 188 189 va_start(ap, m0); 190 opt = va_arg(ap, struct mbuf *); 191 ro = va_arg(ap, struct route *); 192 flags = va_arg(ap, int); 193 imo = va_arg(ap, struct ip_moptions *); 194 if (flags & IP_RETURNMTU) 195 mtu_p = va_arg(ap, int *); 196 else 197 mtu_p = NULL; 198 va_end(ap); 199 200 #ifdef IPSEC 201 so = ipsec_getsocket(m); 202 (void)ipsec_setsocket(m, NULL); 203 #endif /*IPSEC*/ 204 205 #ifdef DIAGNOSTIC 206 if ((m->m_flags & M_PKTHDR) == 0) 207 panic("ip_output no HDR"); 208 #endif 209 if (opt) { 210 m = ip_insertoptions(m, opt, &len); 211 hlen = len; 212 } 213 ip = mtod(m, struct ip *); 214 /* 215 * Fill in IP header. 216 */ 217 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 218 ip->ip_v = IPVERSION; 219 ip->ip_off = 0; 220 ip->ip_id = htons(ip_id++); 221 ip->ip_hl = hlen >> 2; 222 ipstat.ips_localout++; 223 } else { 224 hlen = ip->ip_hl << 2; 225 } 226 /* 227 * Route packet. 228 */ 229 if (ro == 0) { 230 ro = &iproute; 231 bzero((caddr_t)ro, sizeof (*ro)); 232 } 233 dst = satosin(&ro->ro_dst); 234 /* 235 * If there is a cached route, 236 * check that it is to the same destination 237 * and is still up. If not, free it and try again. 238 * The address family should also be checked in case of sharing the 239 * cache with IPv6. 240 */ 241 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 242 dst->sin_family != AF_INET || 243 !in_hosteq(dst->sin_addr, ip->ip_dst))) { 244 RTFREE(ro->ro_rt); 245 ro->ro_rt = (struct rtentry *)0; 246 } 247 if (ro->ro_rt == 0) { 248 bzero(dst, sizeof(*dst)); 249 dst->sin_family = AF_INET; 250 dst->sin_len = sizeof(*dst); 251 dst->sin_addr = ip->ip_dst; 252 } 253 /* 254 * If routing to interface only, 255 * short circuit routing lookup. 256 */ 257 if (flags & IP_ROUTETOIF) { 258 if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) { 259 ipstat.ips_noroute++; 260 error = ENETUNREACH; 261 goto bad; 262 } 263 ifp = ia->ia_ifp; 264 mtu = ifp->if_mtu; 265 ip->ip_ttl = 1; 266 } else { 267 if (ro->ro_rt == 0) 268 rtalloc(ro); 269 if (ro->ro_rt == 0) { 270 ipstat.ips_noroute++; 271 error = EHOSTUNREACH; 272 goto bad; 273 } 274 ia = ifatoia(ro->ro_rt->rt_ifa); 275 ifp = ro->ro_rt->rt_ifp; 276 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 277 mtu = ifp->if_mtu; 278 ro->ro_rt->rt_use++; 279 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 280 dst = satosin(ro->ro_rt->rt_gateway); 281 } 282 if (IN_MULTICAST(ip->ip_dst.s_addr) || 283 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 284 struct in_multi *inm; 285 286 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 287 M_BCAST : M_MCAST; 288 /* 289 * IP destination address is multicast. Make sure "dst" 290 * still points to the address in "ro". (It may have been 291 * changed to point to a gateway address, above.) 292 */ 293 dst = satosin(&ro->ro_dst); 294 /* 295 * See if the caller provided any multicast options 296 */ 297 if (imo != NULL) { 298 ip->ip_ttl = imo->imo_multicast_ttl; 299 if (imo->imo_multicast_ifp != NULL) { 300 ifp = imo->imo_multicast_ifp; 301 mtu = ifp->if_mtu; 302 } 303 } else 304 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 305 /* 306 * If the packet is multicast or broadcast, confirm that 307 * the outgoing interface can transmit it. 308 */ 309 if (((m->m_flags & M_MCAST) && 310 (ifp->if_flags & IFF_MULTICAST) == 0) || 311 ((m->m_flags & M_BCAST) && 312 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 313 ipstat.ips_noroute++; 314 error = ENETUNREACH; 315 goto bad; 316 } 317 /* 318 * If source address not specified yet, use an address 319 * of outgoing interface. 320 */ 321 if (in_nullhost(ip->ip_src)) { 322 struct in_ifaddr *ia; 323 324 IFP_TO_IA(ifp, ia); 325 if (!ia) { 326 error = EADDRNOTAVAIL; 327 goto bad; 328 } 329 ip->ip_src = ia->ia_addr.sin_addr; 330 } 331 332 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 333 if (inm != NULL && 334 (imo == NULL || imo->imo_multicast_loop)) { 335 /* 336 * If we belong to the destination multicast group 337 * on the outgoing interface, and the caller did not 338 * forbid loopback, loop back a copy. 339 */ 340 ip_mloopback(ifp, m, dst); 341 } 342 #ifdef MROUTING 343 else { 344 /* 345 * If we are acting as a multicast router, perform 346 * multicast forwarding as if the packet had just 347 * arrived on the interface to which we are about 348 * to send. The multicast forwarding function 349 * recursively calls this function, using the 350 * IP_FORWARDING flag to prevent infinite recursion. 351 * 352 * Multicasts that are looped back by ip_mloopback(), 353 * above, will be forwarded by the ip_input() routine, 354 * if necessary. 355 */ 356 extern struct socket *ip_mrouter; 357 358 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 359 if (ip_mforward(m, ifp) != 0) { 360 m_freem(m); 361 goto done; 362 } 363 } 364 } 365 #endif 366 /* 367 * Multicasts with a time-to-live of zero may be looped- 368 * back, above, but must not be transmitted on a network. 369 * Also, multicasts addressed to the loopback interface 370 * are not sent -- the above call to ip_mloopback() will 371 * loop back a copy if this host actually belongs to the 372 * destination group on the loopback interface. 373 */ 374 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 375 m_freem(m); 376 goto done; 377 } 378 379 goto sendit; 380 } 381 #ifndef notdef 382 /* 383 * If source address not specified yet, use address 384 * of outgoing interface. 385 */ 386 if (in_nullhost(ip->ip_src)) 387 ip->ip_src = ia->ia_addr.sin_addr; 388 #endif 389 390 /* 391 * packets with Class-D address as source are not valid per 392 * RFC 1112 393 */ 394 if (IN_MULTICAST(ip->ip_src.s_addr)) { 395 ipstat.ips_odropped++; 396 error = EADDRNOTAVAIL; 397 goto bad; 398 } 399 400 /* 401 * Look for broadcast address and 402 * and verify user is allowed to send 403 * such a packet. 404 */ 405 if (in_broadcast(dst->sin_addr, ifp)) { 406 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 407 error = EADDRNOTAVAIL; 408 goto bad; 409 } 410 if ((flags & IP_ALLOWBROADCAST) == 0) { 411 error = EACCES; 412 goto bad; 413 } 414 /* don't allow broadcast messages to be fragmented */ 415 if ((u_int16_t)ip->ip_len > ifp->if_mtu) { 416 error = EMSGSIZE; 417 goto bad; 418 } 419 m->m_flags |= M_BCAST; 420 } else 421 m->m_flags &= ~M_BCAST; 422 423 sendit: 424 /* 425 * If we're doing Path MTU Discovery, we need to set DF unless 426 * the route's MTU is locked. 427 */ 428 if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL && 429 (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 430 ip->ip_off |= IP_DF; 431 432 /* 433 * Remember the current ip_len and ip_off, and swap them into 434 * network order. 435 */ 436 ip_len = ip->ip_len; 437 438 HTONS(ip->ip_len); 439 HTONS(ip->ip_off); 440 441 #ifdef IPSEC 442 /* get SP for this packet */ 443 if (so == NULL) 444 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 445 else 446 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 447 448 if (sp == NULL) { 449 ipsecstat.out_inval++; 450 goto bad; 451 } 452 453 error = 0; 454 455 /* check policy */ 456 switch (sp->policy) { 457 case IPSEC_POLICY_DISCARD: 458 /* 459 * This packet is just discarded. 460 */ 461 ipsecstat.out_polvio++; 462 goto bad; 463 464 case IPSEC_POLICY_BYPASS: 465 case IPSEC_POLICY_NONE: 466 /* no need to do IPsec. */ 467 goto skip_ipsec; 468 469 case IPSEC_POLICY_IPSEC: 470 if (sp->req == NULL) { 471 /* XXX should be panic ? */ 472 printf("ip_output: No IPsec request specified.\n"); 473 error = EINVAL; 474 goto bad; 475 } 476 break; 477 478 case IPSEC_POLICY_ENTRUST: 479 default: 480 printf("ip_output: Invalid policy found. %d\n", sp->policy); 481 } 482 483 /* 484 * ipsec4_output() expects ip_len and ip_off in network 485 * order. They have been set to network order above. 486 */ 487 488 { 489 struct ipsec_output_state state; 490 bzero(&state, sizeof(state)); 491 state.m = m; 492 if (flags & IP_ROUTETOIF) { 493 state.ro = &iproute; 494 bzero(&iproute, sizeof(iproute)); 495 } else 496 state.ro = ro; 497 state.dst = (struct sockaddr *)dst; 498 499 /* 500 * We can't defer the checksum of payload data if 501 * we're about to encrypt/authenticate it. 502 * 503 * XXX When we support crypto offloading functions of 504 * XXX network interfaces, we need to reconsider this, 505 * XXX since it's likely that they'll support checksumming, 506 * XXX as well. 507 */ 508 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 509 in_delayed_cksum(m); 510 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 511 } 512 513 error = ipsec4_output(&state, sp, flags); 514 515 m = state.m; 516 if (flags & IP_ROUTETOIF) { 517 /* 518 * if we have tunnel mode SA, we may need to ignore 519 * IP_ROUTETOIF. 520 */ 521 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 522 flags &= ~IP_ROUTETOIF; 523 ro = state.ro; 524 } 525 } else 526 ro = state.ro; 527 dst = (struct sockaddr_in *)state.dst; 528 if (error) { 529 /* mbuf is already reclaimed in ipsec4_output. */ 530 m0 = NULL; 531 switch (error) { 532 case EHOSTUNREACH: 533 case ENETUNREACH: 534 case EMSGSIZE: 535 case ENOBUFS: 536 case ENOMEM: 537 break; 538 default: 539 printf("ip4_output (ipsec): error code %d\n", error); 540 /*fall through*/ 541 case ENOENT: 542 /* don't show these error codes to the user */ 543 error = 0; 544 break; 545 } 546 goto bad; 547 } 548 549 /* be sure to update variables that are affected by ipsec4_output() */ 550 ip = mtod(m, struct ip *); 551 #ifdef _IP_VHL 552 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 553 #else 554 hlen = ip->ip_hl << 2; 555 #endif 556 ip_len = ntohs(ip->ip_len); 557 558 if (ro->ro_rt == NULL) { 559 if ((flags & IP_ROUTETOIF) == 0) { 560 printf("ip_output: " 561 "can't update route after IPsec processing\n"); 562 error = EHOSTUNREACH; /*XXX*/ 563 goto bad; 564 } 565 } else { 566 /* nobody uses ia beyond here */ 567 if (state.encap) 568 ifp = ro->ro_rt->rt_ifp; 569 } 570 } 571 572 skip_ipsec: 573 #endif /*IPSEC*/ 574 575 #ifdef PFIL_HOOKS 576 /* 577 * Run through list of hooks for output packets. 578 */ 579 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, 580 PFIL_OUT)) != 0) 581 goto done; 582 if (m == NULL) 583 goto done; 584 585 ip = mtod(m, struct ip *); 586 #endif /* PFIL_HOOKS */ 587 588 /* 589 * If small enough for mtu of path, can just send directly. 590 */ 591 if (ip_len <= mtu) { 592 #if IFA_STATS 593 /* 594 * search for the source address structure to 595 * maintain output statistics. 596 */ 597 INADDR_TO_IA(ip->ip_src, ia); 598 if (ia) 599 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 600 #endif 601 /* 602 * Always initialize the sum to 0! Some HW assisted 603 * checksumming requires this. 604 */ 605 ip->ip_sum = 0; 606 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 607 608 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 609 610 /* 611 * Perform any checksums that the hardware can't do 612 * for us. 613 * 614 * XXX Does any hardware require the {th,uh}_sum 615 * XXX fields to be 0? 616 */ 617 if (sw_csum & M_CSUM_IPv4) 618 ip->ip_sum = in_cksum(m, hlen); 619 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 620 in_delayed_cksum(m); 621 sw_csum &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 622 } 623 m->m_pkthdr.csum_flags &= ifp->if_csum_flags_tx; 624 625 #ifdef IPSEC 626 /* clean ipsec history once it goes out of the node */ 627 ipsec_delaux(m); 628 #endif 629 error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt); 630 goto done; 631 } 632 633 /* 634 * We can't use HW checksumming if we're about to 635 * to fragment the packet. 636 * 637 * XXX Some hardware can do this. 638 */ 639 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 640 in_delayed_cksum(m); 641 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 642 } 643 644 /* 645 * Too large for interface; fragment if possible. 646 * Must be able to put at least 8 bytes per fragment. 647 * 648 * Note we swap ip_len and ip_off into host order to make 649 * the logic below a little simpler. 650 */ 651 652 NTOHS(ip->ip_len); 653 NTOHS(ip->ip_off); 654 655 if (ip->ip_off & IP_DF) { 656 if (flags & IP_RETURNMTU) 657 *mtu_p = mtu; 658 error = EMSGSIZE; 659 ipstat.ips_cantfrag++; 660 goto bad; 661 } 662 len = (mtu - hlen) &~ 7; 663 if (len < 8) { 664 error = EMSGSIZE; 665 goto bad; 666 } 667 668 { 669 int mhlen, firstlen = len; 670 struct mbuf **mnext = &m->m_nextpkt; 671 int fragments = 0; 672 int s; 673 674 /* 675 * Loop through length of segment after first fragment, 676 * make new header and copy data of each part and link onto chain. 677 */ 678 m0 = m; 679 mhlen = sizeof (struct ip); 680 for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) { 681 MGETHDR(m, M_DONTWAIT, MT_HEADER); 682 if (m == 0) { 683 error = ENOBUFS; 684 ipstat.ips_odropped++; 685 goto sendorfree; 686 } 687 *mnext = m; 688 mnext = &m->m_nextpkt; 689 m->m_data += max_linkhdr; 690 mhip = mtod(m, struct ip *); 691 *mhip = *ip; 692 /* we must inherit MCAST and BCAST flags */ 693 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 694 if (hlen > sizeof (struct ip)) { 695 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 696 mhip->ip_hl = mhlen >> 2; 697 } 698 m->m_len = mhlen; 699 mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF); 700 if (ip->ip_off & IP_MF) 701 mhip->ip_off |= IP_MF; 702 if (off + len >= (u_int16_t)ip->ip_len) 703 len = (u_int16_t)ip->ip_len - off; 704 else 705 mhip->ip_off |= IP_MF; 706 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 707 m->m_next = m_copy(m0, off, len); 708 if (m->m_next == 0) { 709 error = ENOBUFS; /* ??? */ 710 ipstat.ips_odropped++; 711 goto sendorfree; 712 } 713 m->m_pkthdr.len = mhlen + len; 714 m->m_pkthdr.rcvif = (struct ifnet *)0; 715 HTONS(mhip->ip_off); 716 mhip->ip_sum = 0; 717 mhip->ip_sum = in_cksum(m, mhlen); 718 ipstat.ips_ofragments++; 719 fragments++; 720 } 721 /* 722 * Update first fragment by trimming what's been copied out 723 * and updating header, then send each fragment (in order). 724 */ 725 m = m0; 726 m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len); 727 m->m_pkthdr.len = hlen + firstlen; 728 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 729 ip->ip_off |= IP_MF; 730 HTONS(ip->ip_off); 731 ip->ip_sum = 0; 732 ip->ip_sum = in_cksum(m, hlen); 733 sendorfree: 734 /* 735 * If there is no room for all the fragments, don't queue 736 * any of them. 737 */ 738 s = splnet(); 739 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments) 740 error = ENOBUFS; 741 splx(s); 742 for (m = m0; m; m = m0) { 743 m0 = m->m_nextpkt; 744 m->m_nextpkt = 0; 745 if (error == 0) { 746 #if IFA_STATS 747 /* 748 * search for the source address structure to 749 * maintain output statistics. 750 */ 751 INADDR_TO_IA(ip->ip_src, ia); 752 if (ia) { 753 ia->ia_ifa.ifa_data.ifad_outbytes += 754 ntohs(ip->ip_len); 755 } 756 #endif 757 #ifdef IPSEC 758 /* clean ipsec history once it goes out of the node */ 759 ipsec_delaux(m); 760 #endif 761 error = (*ifp->if_output)(ifp, m, sintosa(dst), 762 ro->ro_rt); 763 } else 764 m_freem(m); 765 } 766 767 if (error == 0) 768 ipstat.ips_fragmented++; 769 } 770 done: 771 if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) { 772 RTFREE(ro->ro_rt); 773 ro->ro_rt = 0; 774 } 775 776 #ifdef IPSEC 777 if (sp != NULL) { 778 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 779 printf("DP ip_output call free SP:%p\n", sp)); 780 key_freesp(sp); 781 } 782 #endif /* IPSEC */ 783 784 return (error); 785 bad: 786 m_freem(m); 787 goto done; 788 } 789 790 /* 791 * Process a delayed payload checksum calculation. 792 */ 793 void 794 in_delayed_cksum(struct mbuf *m) 795 { 796 struct ip *ip; 797 u_int16_t csum, offset; 798 799 ip = mtod(m, struct ip *); 800 offset = ip->ip_hl << 2; 801 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 802 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 803 csum = 0xffff; 804 805 offset += m->m_pkthdr.csum_data; /* checksum offset */ 806 807 if ((offset + sizeof(u_int16_t)) > m->m_len) { 808 /* This happen when ip options were inserted 809 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 810 m->m_len, offset, ip->ip_p); 811 */ 812 m_copyback(m, offset, sizeof(csum), (caddr_t) &csum); 813 } else 814 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 815 } 816 817 /* 818 * Determine the maximum length of the options to be inserted; 819 * we would far rather allocate too much space rather than too little. 820 */ 821 822 u_int 823 ip_optlen(inp) 824 struct inpcb *inp; 825 { 826 struct mbuf *m = inp->inp_options; 827 828 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 829 return(m->m_len - offsetof(struct ipoption, ipopt_dst)); 830 else 831 return 0; 832 } 833 834 835 /* 836 * Insert IP options into preformed packet. 837 * Adjust IP destination as required for IP source routing, 838 * as indicated by a non-zero in_addr at the start of the options. 839 */ 840 static struct mbuf * 841 ip_insertoptions(m, opt, phlen) 842 struct mbuf *m; 843 struct mbuf *opt; 844 int *phlen; 845 { 846 struct ipoption *p = mtod(opt, struct ipoption *); 847 struct mbuf *n; 848 struct ip *ip = mtod(m, struct ip *); 849 unsigned optlen; 850 851 optlen = opt->m_len - sizeof(p->ipopt_dst); 852 if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET) 853 return (m); /* XXX should fail */ 854 if (!in_nullhost(p->ipopt_dst)) 855 ip->ip_dst = p->ipopt_dst; 856 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 857 MGETHDR(n, M_DONTWAIT, MT_HEADER); 858 if (n == 0) 859 return (m); 860 M_COPY_PKTHDR(n, m); 861 m->m_flags &= ~M_PKTHDR; 862 m->m_len -= sizeof(struct ip); 863 m->m_data += sizeof(struct ip); 864 n->m_next = m; 865 m = n; 866 m->m_len = optlen + sizeof(struct ip); 867 m->m_data += max_linkhdr; 868 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 869 } else { 870 m->m_data -= optlen; 871 m->m_len += optlen; 872 memmove(mtod(m, caddr_t), ip, sizeof(struct ip)); 873 } 874 m->m_pkthdr.len += optlen; 875 ip = mtod(m, struct ip *); 876 bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen); 877 *phlen = sizeof(struct ip) + optlen; 878 ip->ip_len += optlen; 879 return (m); 880 } 881 882 /* 883 * Copy options from ip to jp, 884 * omitting those not copied during fragmentation. 885 */ 886 int 887 ip_optcopy(ip, jp) 888 struct ip *ip, *jp; 889 { 890 u_char *cp, *dp; 891 int opt, optlen, cnt; 892 893 cp = (u_char *)(ip + 1); 894 dp = (u_char *)(jp + 1); 895 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 896 for (; cnt > 0; cnt -= optlen, cp += optlen) { 897 opt = cp[0]; 898 if (opt == IPOPT_EOL) 899 break; 900 if (opt == IPOPT_NOP) { 901 /* Preserve for IP mcast tunnel's LSRR alignment. */ 902 *dp++ = IPOPT_NOP; 903 optlen = 1; 904 continue; 905 } 906 #ifdef DIAGNOSTIC 907 if (cnt < IPOPT_OLEN + sizeof(*cp)) 908 panic("malformed IPv4 option passed to ip_optcopy"); 909 #endif 910 optlen = cp[IPOPT_OLEN]; 911 #ifdef DIAGNOSTIC 912 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 913 panic("malformed IPv4 option passed to ip_optcopy"); 914 #endif 915 /* bogus lengths should have been caught by ip_dooptions */ 916 if (optlen > cnt) 917 optlen = cnt; 918 if (IPOPT_COPIED(opt)) { 919 bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen); 920 dp += optlen; 921 } 922 } 923 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 924 *dp++ = IPOPT_EOL; 925 return (optlen); 926 } 927 928 /* 929 * IP socket option processing. 930 */ 931 int 932 ip_ctloutput(op, so, level, optname, mp) 933 int op; 934 struct socket *so; 935 int level, optname; 936 struct mbuf **mp; 937 { 938 struct inpcb *inp = sotoinpcb(so); 939 struct mbuf *m = *mp; 940 int optval = 0; 941 int error = 0; 942 #ifdef IPSEC 943 #ifdef __NetBSD__ 944 struct proc *p = curproc; /*XXX*/ 945 #endif 946 #endif 947 948 if (level != IPPROTO_IP) { 949 error = EINVAL; 950 if (op == PRCO_SETOPT && *mp) 951 (void) m_free(*mp); 952 } else switch (op) { 953 954 case PRCO_SETOPT: 955 switch (optname) { 956 case IP_OPTIONS: 957 #ifdef notyet 958 case IP_RETOPTS: 959 return (ip_pcbopts(optname, &inp->inp_options, m)); 960 #else 961 return (ip_pcbopts(&inp->inp_options, m)); 962 #endif 963 964 case IP_TOS: 965 case IP_TTL: 966 case IP_RECVOPTS: 967 case IP_RECVRETOPTS: 968 case IP_RECVDSTADDR: 969 case IP_RECVIF: 970 if (m == NULL || m->m_len != sizeof(int)) 971 error = EINVAL; 972 else { 973 optval = *mtod(m, int *); 974 switch (optname) { 975 976 case IP_TOS: 977 inp->inp_ip.ip_tos = optval; 978 break; 979 980 case IP_TTL: 981 inp->inp_ip.ip_ttl = optval; 982 break; 983 #define OPTSET(bit) \ 984 if (optval) \ 985 inp->inp_flags |= bit; \ 986 else \ 987 inp->inp_flags &= ~bit; 988 989 case IP_RECVOPTS: 990 OPTSET(INP_RECVOPTS); 991 break; 992 993 case IP_RECVRETOPTS: 994 OPTSET(INP_RECVRETOPTS); 995 break; 996 997 case IP_RECVDSTADDR: 998 OPTSET(INP_RECVDSTADDR); 999 break; 1000 1001 case IP_RECVIF: 1002 OPTSET(INP_RECVIF); 1003 break; 1004 } 1005 } 1006 break; 1007 #undef OPTSET 1008 1009 case IP_MULTICAST_IF: 1010 case IP_MULTICAST_TTL: 1011 case IP_MULTICAST_LOOP: 1012 case IP_ADD_MEMBERSHIP: 1013 case IP_DROP_MEMBERSHIP: 1014 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1015 break; 1016 1017 case IP_PORTRANGE: 1018 if (m == 0 || m->m_len != sizeof(int)) 1019 error = EINVAL; 1020 else { 1021 optval = *mtod(m, int *); 1022 1023 switch (optval) { 1024 1025 case IP_PORTRANGE_DEFAULT: 1026 case IP_PORTRANGE_HIGH: 1027 inp->inp_flags &= ~(INP_LOWPORT); 1028 break; 1029 1030 case IP_PORTRANGE_LOW: 1031 inp->inp_flags |= INP_LOWPORT; 1032 break; 1033 1034 default: 1035 error = EINVAL; 1036 break; 1037 } 1038 } 1039 break; 1040 1041 #ifdef IPSEC 1042 case IP_IPSEC_POLICY: 1043 { 1044 caddr_t req = NULL; 1045 size_t len = 0; 1046 int priv = 0; 1047 1048 #ifdef __NetBSD__ 1049 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1050 priv = 0; 1051 else 1052 priv = 1; 1053 #else 1054 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1055 #endif 1056 if (m) { 1057 req = mtod(m, caddr_t); 1058 len = m->m_len; 1059 } 1060 error = ipsec4_set_policy(inp, optname, req, len, priv); 1061 break; 1062 } 1063 #endif /*IPSEC*/ 1064 1065 default: 1066 error = ENOPROTOOPT; 1067 break; 1068 } 1069 if (m) 1070 (void)m_free(m); 1071 break; 1072 1073 case PRCO_GETOPT: 1074 switch (optname) { 1075 case IP_OPTIONS: 1076 case IP_RETOPTS: 1077 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1078 if (inp->inp_options) { 1079 m->m_len = inp->inp_options->m_len; 1080 bcopy(mtod(inp->inp_options, caddr_t), 1081 mtod(m, caddr_t), (unsigned)m->m_len); 1082 } else 1083 m->m_len = 0; 1084 break; 1085 1086 case IP_TOS: 1087 case IP_TTL: 1088 case IP_RECVOPTS: 1089 case IP_RECVRETOPTS: 1090 case IP_RECVDSTADDR: 1091 case IP_RECVIF: 1092 case IP_ERRORMTU: 1093 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1094 m->m_len = sizeof(int); 1095 switch (optname) { 1096 1097 case IP_TOS: 1098 optval = inp->inp_ip.ip_tos; 1099 break; 1100 1101 case IP_TTL: 1102 optval = inp->inp_ip.ip_ttl; 1103 break; 1104 1105 case IP_ERRORMTU: 1106 optval = inp->inp_errormtu; 1107 break; 1108 1109 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1110 1111 case IP_RECVOPTS: 1112 optval = OPTBIT(INP_RECVOPTS); 1113 break; 1114 1115 case IP_RECVRETOPTS: 1116 optval = OPTBIT(INP_RECVRETOPTS); 1117 break; 1118 1119 case IP_RECVDSTADDR: 1120 optval = OPTBIT(INP_RECVDSTADDR); 1121 break; 1122 1123 case IP_RECVIF: 1124 optval = OPTBIT(INP_RECVIF); 1125 break; 1126 } 1127 *mtod(m, int *) = optval; 1128 break; 1129 1130 #ifdef IPSEC 1131 case IP_IPSEC_POLICY: 1132 { 1133 caddr_t req = NULL; 1134 size_t len = 0; 1135 1136 if (m) { 1137 req = mtod(m, caddr_t); 1138 len = m->m_len; 1139 } 1140 error = ipsec4_get_policy(inp, req, len, mp); 1141 break; 1142 } 1143 #endif /*IPSEC*/ 1144 1145 case IP_MULTICAST_IF: 1146 case IP_MULTICAST_TTL: 1147 case IP_MULTICAST_LOOP: 1148 case IP_ADD_MEMBERSHIP: 1149 case IP_DROP_MEMBERSHIP: 1150 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1151 break; 1152 1153 case IP_PORTRANGE: 1154 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1155 m->m_len = sizeof(int); 1156 1157 if (inp->inp_flags & INP_LOWPORT) 1158 optval = IP_PORTRANGE_LOW; 1159 else 1160 optval = IP_PORTRANGE_DEFAULT; 1161 1162 *mtod(m, int *) = optval; 1163 break; 1164 1165 default: 1166 error = ENOPROTOOPT; 1167 break; 1168 } 1169 break; 1170 } 1171 return (error); 1172 } 1173 1174 /* 1175 * Set up IP options in pcb for insertion in output packets. 1176 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1177 * with destination address if source routed. 1178 */ 1179 int 1180 #ifdef notyet 1181 ip_pcbopts(optname, pcbopt, m) 1182 int optname; 1183 #else 1184 ip_pcbopts(pcbopt, m) 1185 #endif 1186 struct mbuf **pcbopt; 1187 struct mbuf *m; 1188 { 1189 int cnt, optlen; 1190 u_char *cp; 1191 u_char opt; 1192 1193 /* turn off any old options */ 1194 if (*pcbopt) 1195 (void)m_free(*pcbopt); 1196 *pcbopt = 0; 1197 if (m == (struct mbuf *)0 || m->m_len == 0) { 1198 /* 1199 * Only turning off any previous options. 1200 */ 1201 if (m) 1202 (void)m_free(m); 1203 return (0); 1204 } 1205 1206 #ifndef __vax__ 1207 if (m->m_len % sizeof(int32_t)) 1208 goto bad; 1209 #endif 1210 /* 1211 * IP first-hop destination address will be stored before 1212 * actual options; move other options back 1213 * and clear it when none present. 1214 */ 1215 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1216 goto bad; 1217 cnt = m->m_len; 1218 m->m_len += sizeof(struct in_addr); 1219 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1220 memmove(cp, mtod(m, caddr_t), (unsigned)cnt); 1221 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1222 1223 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1224 opt = cp[IPOPT_OPTVAL]; 1225 if (opt == IPOPT_EOL) 1226 break; 1227 if (opt == IPOPT_NOP) 1228 optlen = 1; 1229 else { 1230 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1231 goto bad; 1232 optlen = cp[IPOPT_OLEN]; 1233 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1234 goto bad; 1235 } 1236 switch (opt) { 1237 1238 default: 1239 break; 1240 1241 case IPOPT_LSRR: 1242 case IPOPT_SSRR: 1243 /* 1244 * user process specifies route as: 1245 * ->A->B->C->D 1246 * D must be our final destination (but we can't 1247 * check that since we may not have connected yet). 1248 * A is first hop destination, which doesn't appear in 1249 * actual IP option, but is stored before the options. 1250 */ 1251 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1252 goto bad; 1253 m->m_len -= sizeof(struct in_addr); 1254 cnt -= sizeof(struct in_addr); 1255 optlen -= sizeof(struct in_addr); 1256 cp[IPOPT_OLEN] = optlen; 1257 /* 1258 * Move first hop before start of options. 1259 */ 1260 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1261 sizeof(struct in_addr)); 1262 /* 1263 * Then copy rest of options back 1264 * to close up the deleted entry. 1265 */ 1266 memmove(&cp[IPOPT_OFFSET+1], 1267 (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1268 (unsigned)cnt + sizeof(struct in_addr)); 1269 break; 1270 } 1271 } 1272 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1273 goto bad; 1274 *pcbopt = m; 1275 return (0); 1276 1277 bad: 1278 (void)m_free(m); 1279 return (EINVAL); 1280 } 1281 1282 /* 1283 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1284 */ 1285 static struct ifnet * 1286 ip_multicast_if(a, ifindexp) 1287 struct in_addr *a; 1288 int *ifindexp; 1289 { 1290 int ifindex; 1291 struct ifnet *ifp; 1292 1293 if (ifindexp) 1294 *ifindexp = 0; 1295 if (ntohl(a->s_addr) >> 24 == 0) { 1296 ifindex = ntohl(a->s_addr) & 0xffffff; 1297 if (ifindex < 0 || if_index < ifindex) 1298 return NULL; 1299 ifp = ifindex2ifnet[ifindex]; 1300 if (ifindexp) 1301 *ifindexp = ifindex; 1302 } else { 1303 INADDR_TO_IFP(*a, ifp); 1304 } 1305 return ifp; 1306 } 1307 1308 /* 1309 * Set the IP multicast options in response to user setsockopt(). 1310 */ 1311 int 1312 ip_setmoptions(optname, imop, m) 1313 int optname; 1314 struct ip_moptions **imop; 1315 struct mbuf *m; 1316 { 1317 int error = 0; 1318 u_char loop; 1319 int i; 1320 struct in_addr addr; 1321 struct ip_mreq *mreq; 1322 struct ifnet *ifp; 1323 struct ip_moptions *imo = *imop; 1324 struct route ro; 1325 struct sockaddr_in *dst; 1326 int ifindex; 1327 1328 if (imo == NULL) { 1329 /* 1330 * No multicast option buffer attached to the pcb; 1331 * allocate one and initialize to default values. 1332 */ 1333 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1334 M_WAITOK); 1335 1336 if (imo == NULL) 1337 return (ENOBUFS); 1338 *imop = imo; 1339 imo->imo_multicast_ifp = NULL; 1340 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1341 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1342 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1343 imo->imo_num_memberships = 0; 1344 } 1345 1346 switch (optname) { 1347 1348 case IP_MULTICAST_IF: 1349 /* 1350 * Select the interface for outgoing multicast packets. 1351 */ 1352 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1353 error = EINVAL; 1354 break; 1355 } 1356 addr = *(mtod(m, struct in_addr *)); 1357 /* 1358 * INADDR_ANY is used to remove a previous selection. 1359 * When no interface is selected, a default one is 1360 * chosen every time a multicast packet is sent. 1361 */ 1362 if (in_nullhost(addr)) { 1363 imo->imo_multicast_ifp = NULL; 1364 break; 1365 } 1366 /* 1367 * The selected interface is identified by its local 1368 * IP address. Find the interface and confirm that 1369 * it supports multicasting. 1370 */ 1371 ifp = ip_multicast_if(&addr, &ifindex); 1372 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1373 error = EADDRNOTAVAIL; 1374 break; 1375 } 1376 imo->imo_multicast_ifp = ifp; 1377 if (ifindex) 1378 imo->imo_multicast_addr = addr; 1379 else 1380 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1381 break; 1382 1383 case IP_MULTICAST_TTL: 1384 /* 1385 * Set the IP time-to-live for outgoing multicast packets. 1386 */ 1387 if (m == NULL || m->m_len != 1) { 1388 error = EINVAL; 1389 break; 1390 } 1391 imo->imo_multicast_ttl = *(mtod(m, u_char *)); 1392 break; 1393 1394 case IP_MULTICAST_LOOP: 1395 /* 1396 * Set the loopback flag for outgoing multicast packets. 1397 * Must be zero or one. 1398 */ 1399 if (m == NULL || m->m_len != 1 || 1400 (loop = *(mtod(m, u_char *))) > 1) { 1401 error = EINVAL; 1402 break; 1403 } 1404 imo->imo_multicast_loop = loop; 1405 break; 1406 1407 case IP_ADD_MEMBERSHIP: 1408 /* 1409 * Add a multicast group membership. 1410 * Group must be a valid IP multicast address. 1411 */ 1412 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1413 error = EINVAL; 1414 break; 1415 } 1416 mreq = mtod(m, struct ip_mreq *); 1417 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1418 error = EINVAL; 1419 break; 1420 } 1421 /* 1422 * If no interface address was provided, use the interface of 1423 * the route to the given multicast address. 1424 */ 1425 if (in_nullhost(mreq->imr_interface)) { 1426 bzero((caddr_t)&ro, sizeof(ro)); 1427 ro.ro_rt = NULL; 1428 dst = satosin(&ro.ro_dst); 1429 dst->sin_len = sizeof(*dst); 1430 dst->sin_family = AF_INET; 1431 dst->sin_addr = mreq->imr_multiaddr; 1432 rtalloc(&ro); 1433 if (ro.ro_rt == NULL) { 1434 error = EADDRNOTAVAIL; 1435 break; 1436 } 1437 ifp = ro.ro_rt->rt_ifp; 1438 rtfree(ro.ro_rt); 1439 } else { 1440 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1441 } 1442 /* 1443 * See if we found an interface, and confirm that it 1444 * supports multicast. 1445 */ 1446 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1447 error = EADDRNOTAVAIL; 1448 break; 1449 } 1450 /* 1451 * See if the membership already exists or if all the 1452 * membership slots are full. 1453 */ 1454 for (i = 0; i < imo->imo_num_memberships; ++i) { 1455 if (imo->imo_membership[i]->inm_ifp == ifp && 1456 in_hosteq(imo->imo_membership[i]->inm_addr, 1457 mreq->imr_multiaddr)) 1458 break; 1459 } 1460 if (i < imo->imo_num_memberships) { 1461 error = EADDRINUSE; 1462 break; 1463 } 1464 if (i == IP_MAX_MEMBERSHIPS) { 1465 error = ETOOMANYREFS; 1466 break; 1467 } 1468 /* 1469 * Everything looks good; add a new record to the multicast 1470 * address list for the given interface. 1471 */ 1472 if ((imo->imo_membership[i] = 1473 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1474 error = ENOBUFS; 1475 break; 1476 } 1477 ++imo->imo_num_memberships; 1478 break; 1479 1480 case IP_DROP_MEMBERSHIP: 1481 /* 1482 * Drop a multicast group membership. 1483 * Group must be a valid IP multicast address. 1484 */ 1485 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1486 error = EINVAL; 1487 break; 1488 } 1489 mreq = mtod(m, struct ip_mreq *); 1490 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1491 error = EINVAL; 1492 break; 1493 } 1494 /* 1495 * If an interface address was specified, get a pointer 1496 * to its ifnet structure. 1497 */ 1498 if (in_nullhost(mreq->imr_interface)) 1499 ifp = NULL; 1500 else { 1501 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1502 if (ifp == NULL) { 1503 error = EADDRNOTAVAIL; 1504 break; 1505 } 1506 } 1507 /* 1508 * Find the membership in the membership array. 1509 */ 1510 for (i = 0; i < imo->imo_num_memberships; ++i) { 1511 if ((ifp == NULL || 1512 imo->imo_membership[i]->inm_ifp == ifp) && 1513 in_hosteq(imo->imo_membership[i]->inm_addr, 1514 mreq->imr_multiaddr)) 1515 break; 1516 } 1517 if (i == imo->imo_num_memberships) { 1518 error = EADDRNOTAVAIL; 1519 break; 1520 } 1521 /* 1522 * Give up the multicast address record to which the 1523 * membership points. 1524 */ 1525 in_delmulti(imo->imo_membership[i]); 1526 /* 1527 * Remove the gap in the membership array. 1528 */ 1529 for (++i; i < imo->imo_num_memberships; ++i) 1530 imo->imo_membership[i-1] = imo->imo_membership[i]; 1531 --imo->imo_num_memberships; 1532 break; 1533 1534 default: 1535 error = EOPNOTSUPP; 1536 break; 1537 } 1538 1539 /* 1540 * If all options have default values, no need to keep the mbuf. 1541 */ 1542 if (imo->imo_multicast_ifp == NULL && 1543 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1544 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1545 imo->imo_num_memberships == 0) { 1546 free(*imop, M_IPMOPTS); 1547 *imop = NULL; 1548 } 1549 1550 return (error); 1551 } 1552 1553 /* 1554 * Return the IP multicast options in response to user getsockopt(). 1555 */ 1556 int 1557 ip_getmoptions(optname, imo, mp) 1558 int optname; 1559 struct ip_moptions *imo; 1560 struct mbuf **mp; 1561 { 1562 u_char *ttl; 1563 u_char *loop; 1564 struct in_addr *addr; 1565 struct in_ifaddr *ia; 1566 1567 *mp = m_get(M_WAIT, MT_SOOPTS); 1568 1569 switch (optname) { 1570 1571 case IP_MULTICAST_IF: 1572 addr = mtod(*mp, struct in_addr *); 1573 (*mp)->m_len = sizeof(struct in_addr); 1574 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1575 *addr = zeroin_addr; 1576 else if (imo->imo_multicast_addr.s_addr) { 1577 /* return the value user has set */ 1578 *addr = imo->imo_multicast_addr; 1579 } else { 1580 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1581 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1582 } 1583 return (0); 1584 1585 case IP_MULTICAST_TTL: 1586 ttl = mtod(*mp, u_char *); 1587 (*mp)->m_len = 1; 1588 *ttl = imo ? imo->imo_multicast_ttl 1589 : IP_DEFAULT_MULTICAST_TTL; 1590 return (0); 1591 1592 case IP_MULTICAST_LOOP: 1593 loop = mtod(*mp, u_char *); 1594 (*mp)->m_len = 1; 1595 *loop = imo ? imo->imo_multicast_loop 1596 : IP_DEFAULT_MULTICAST_LOOP; 1597 return (0); 1598 1599 default: 1600 return (EOPNOTSUPP); 1601 } 1602 } 1603 1604 /* 1605 * Discard the IP multicast options. 1606 */ 1607 void 1608 ip_freemoptions(imo) 1609 struct ip_moptions *imo; 1610 { 1611 int i; 1612 1613 if (imo != NULL) { 1614 for (i = 0; i < imo->imo_num_memberships; ++i) 1615 in_delmulti(imo->imo_membership[i]); 1616 free(imo, M_IPMOPTS); 1617 } 1618 } 1619 1620 /* 1621 * Routine called from ip_output() to loop back a copy of an IP multicast 1622 * packet to the input queue of a specified interface. Note that this 1623 * calls the output routine of the loopback "driver", but with an interface 1624 * pointer that might NOT be &loif -- easier than replicating that code here. 1625 */ 1626 static void 1627 ip_mloopback(ifp, m, dst) 1628 struct ifnet *ifp; 1629 struct mbuf *m; 1630 struct sockaddr_in *dst; 1631 { 1632 struct ip *ip; 1633 struct mbuf *copym; 1634 1635 copym = m_copy(m, 0, M_COPYALL); 1636 if (copym != NULL 1637 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1638 copym = m_pullup(copym, sizeof(struct ip)); 1639 if (copym != NULL) { 1640 /* 1641 * We don't bother to fragment if the IP length is greater 1642 * than the interface's MTU. Can this possibly matter? 1643 */ 1644 ip = mtod(copym, struct ip *); 1645 HTONS(ip->ip_len); 1646 HTONS(ip->ip_off); 1647 1648 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1649 in_delayed_cksum(copym); 1650 copym->m_pkthdr.csum_flags &= 1651 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1652 } 1653 1654 ip->ip_sum = 0; 1655 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1656 (void) looutput(ifp, copym, sintosa(dst), NULL); 1657 } 1658 } 1659