1 /* $NetBSD: ip_output.c,v 1.104 2003/05/26 15:12:11 yamt 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 1.104 2003/05/26 15:12:11 yamt 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 u_long 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 len = 0; 190 va_start(ap, m0); 191 opt = va_arg(ap, struct mbuf *); 192 ro = va_arg(ap, struct route *); 193 flags = va_arg(ap, int); 194 imo = va_arg(ap, struct ip_moptions *); 195 if (flags & IP_RETURNMTU) 196 mtu_p = va_arg(ap, int *); 197 else 198 mtu_p = NULL; 199 va_end(ap); 200 201 MCLAIM(m, &ip_tx_mowner); 202 #ifdef IPSEC 203 so = ipsec_getsocket(m); 204 (void)ipsec_setsocket(m, NULL); 205 #endif /*IPSEC*/ 206 207 #ifdef DIAGNOSTIC 208 if ((m->m_flags & M_PKTHDR) == 0) 209 panic("ip_output no HDR"); 210 #endif 211 if (opt) { 212 m = ip_insertoptions(m, opt, &len); 213 if (len >= sizeof(struct ip)) 214 hlen = len; 215 } 216 ip = mtod(m, struct ip *); 217 /* 218 * Fill in IP header. 219 */ 220 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 221 ip->ip_v = IPVERSION; 222 ip->ip_off = htons(0); 223 ip->ip_id = htons(ip_id++); 224 ip->ip_hl = hlen >> 2; 225 ipstat.ips_localout++; 226 } else { 227 hlen = ip->ip_hl << 2; 228 } 229 /* 230 * Route packet. 231 */ 232 if (ro == 0) { 233 ro = &iproute; 234 bzero((caddr_t)ro, sizeof (*ro)); 235 } 236 dst = satosin(&ro->ro_dst); 237 /* 238 * If there is a cached route, 239 * check that it is to the same destination 240 * and is still up. If not, free it and try again. 241 * The address family should also be checked in case of sharing the 242 * cache with IPv6. 243 */ 244 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 245 dst->sin_family != AF_INET || 246 !in_hosteq(dst->sin_addr, ip->ip_dst))) { 247 RTFREE(ro->ro_rt); 248 ro->ro_rt = (struct rtentry *)0; 249 } 250 if (ro->ro_rt == 0) { 251 bzero(dst, sizeof(*dst)); 252 dst->sin_family = AF_INET; 253 dst->sin_len = sizeof(*dst); 254 dst->sin_addr = ip->ip_dst; 255 } 256 /* 257 * If routing to interface only, 258 * short circuit routing lookup. 259 */ 260 if (flags & IP_ROUTETOIF) { 261 if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) { 262 ipstat.ips_noroute++; 263 error = ENETUNREACH; 264 goto bad; 265 } 266 ifp = ia->ia_ifp; 267 mtu = ifp->if_mtu; 268 ip->ip_ttl = 1; 269 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 270 ip->ip_dst.s_addr == INADDR_BROADCAST) && 271 imo != NULL && imo->imo_multicast_ifp != NULL) { 272 ifp = imo->imo_multicast_ifp; 273 mtu = ifp->if_mtu; 274 IFP_TO_IA(ifp, ia); 275 } else { 276 if (ro->ro_rt == 0) 277 rtalloc(ro); 278 if (ro->ro_rt == 0) { 279 ipstat.ips_noroute++; 280 error = EHOSTUNREACH; 281 goto bad; 282 } 283 ia = ifatoia(ro->ro_rt->rt_ifa); 284 ifp = ro->ro_rt->rt_ifp; 285 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 286 mtu = ifp->if_mtu; 287 ro->ro_rt->rt_use++; 288 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 289 dst = satosin(ro->ro_rt->rt_gateway); 290 } 291 if (IN_MULTICAST(ip->ip_dst.s_addr) || 292 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 293 struct in_multi *inm; 294 295 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 296 M_BCAST : M_MCAST; 297 /* 298 * IP destination address is multicast. Make sure "dst" 299 * still points to the address in "ro". (It may have been 300 * changed to point to a gateway address, above.) 301 */ 302 dst = satosin(&ro->ro_dst); 303 /* 304 * See if the caller provided any multicast options 305 */ 306 if (imo != NULL) 307 ip->ip_ttl = imo->imo_multicast_ttl; 308 else 309 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 310 311 /* 312 * if we don't know the outgoing ifp yet, we can't generate 313 * output 314 */ 315 if (!ifp) { 316 ipstat.ips_noroute++; 317 error = ENETUNREACH; 318 goto bad; 319 } 320 321 /* 322 * If the packet is multicast or broadcast, confirm that 323 * the outgoing interface can transmit it. 324 */ 325 if (((m->m_flags & M_MCAST) && 326 (ifp->if_flags & IFF_MULTICAST) == 0) || 327 ((m->m_flags & M_BCAST) && 328 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 329 ipstat.ips_noroute++; 330 error = ENETUNREACH; 331 goto bad; 332 } 333 /* 334 * If source address not specified yet, use an address 335 * of outgoing interface. 336 */ 337 if (in_nullhost(ip->ip_src)) { 338 struct in_ifaddr *ia; 339 340 IFP_TO_IA(ifp, ia); 341 if (!ia) { 342 error = EADDRNOTAVAIL; 343 goto bad; 344 } 345 ip->ip_src = ia->ia_addr.sin_addr; 346 } 347 348 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 349 if (inm != NULL && 350 (imo == NULL || imo->imo_multicast_loop)) { 351 /* 352 * If we belong to the destination multicast group 353 * on the outgoing interface, and the caller did not 354 * forbid loopback, loop back a copy. 355 */ 356 ip_mloopback(ifp, m, dst); 357 } 358 #ifdef MROUTING 359 else { 360 /* 361 * If we are acting as a multicast router, perform 362 * multicast forwarding as if the packet had just 363 * arrived on the interface to which we are about 364 * to send. The multicast forwarding function 365 * recursively calls this function, using the 366 * IP_FORWARDING flag to prevent infinite recursion. 367 * 368 * Multicasts that are looped back by ip_mloopback(), 369 * above, will be forwarded by the ip_input() routine, 370 * if necessary. 371 */ 372 extern struct socket *ip_mrouter; 373 374 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 375 if (ip_mforward(m, ifp) != 0) { 376 m_freem(m); 377 goto done; 378 } 379 } 380 } 381 #endif 382 /* 383 * Multicasts with a time-to-live of zero may be looped- 384 * back, above, but must not be transmitted on a network. 385 * Also, multicasts addressed to the loopback interface 386 * are not sent -- the above call to ip_mloopback() will 387 * loop back a copy if this host actually belongs to the 388 * destination group on the loopback interface. 389 */ 390 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 391 m_freem(m); 392 goto done; 393 } 394 395 goto sendit; 396 } 397 #ifndef notdef 398 /* 399 * If source address not specified yet, use address 400 * of outgoing interface. 401 */ 402 if (in_nullhost(ip->ip_src)) 403 ip->ip_src = ia->ia_addr.sin_addr; 404 #endif 405 406 /* 407 * packets with Class-D address as source are not valid per 408 * RFC 1112 409 */ 410 if (IN_MULTICAST(ip->ip_src.s_addr)) { 411 ipstat.ips_odropped++; 412 error = EADDRNOTAVAIL; 413 goto bad; 414 } 415 416 /* 417 * Look for broadcast address and 418 * and verify user is allowed to send 419 * such a packet. 420 */ 421 if (in_broadcast(dst->sin_addr, ifp)) { 422 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 423 error = EADDRNOTAVAIL; 424 goto bad; 425 } 426 if ((flags & IP_ALLOWBROADCAST) == 0) { 427 error = EACCES; 428 goto bad; 429 } 430 /* don't allow broadcast messages to be fragmented */ 431 if (ntohs(ip->ip_len) > ifp->if_mtu) { 432 error = EMSGSIZE; 433 goto bad; 434 } 435 m->m_flags |= M_BCAST; 436 } else 437 m->m_flags &= ~M_BCAST; 438 439 sendit: 440 /* 441 * If we're doing Path MTU Discovery, we need to set DF unless 442 * the route's MTU is locked. 443 */ 444 if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL && 445 (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 446 ip->ip_off |= htons(IP_DF); 447 448 /* Remember the current ip_len */ 449 ip_len = ntohs(ip->ip_len); 450 451 #ifdef IPSEC 452 /* get SP for this packet */ 453 if (so == NULL) 454 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 455 else 456 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 457 458 if (sp == NULL) { 459 ipsecstat.out_inval++; 460 goto bad; 461 } 462 463 error = 0; 464 465 /* check policy */ 466 switch (sp->policy) { 467 case IPSEC_POLICY_DISCARD: 468 /* 469 * This packet is just discarded. 470 */ 471 ipsecstat.out_polvio++; 472 goto bad; 473 474 case IPSEC_POLICY_BYPASS: 475 case IPSEC_POLICY_NONE: 476 /* no need to do IPsec. */ 477 goto skip_ipsec; 478 479 case IPSEC_POLICY_IPSEC: 480 if (sp->req == NULL) { 481 /* XXX should be panic ? */ 482 printf("ip_output: No IPsec request specified.\n"); 483 error = EINVAL; 484 goto bad; 485 } 486 break; 487 488 case IPSEC_POLICY_ENTRUST: 489 default: 490 printf("ip_output: Invalid policy found. %d\n", sp->policy); 491 } 492 493 /* 494 * ipsec4_output() expects ip_len and ip_off in network 495 * order. They have been set to network order above. 496 */ 497 498 { 499 struct ipsec_output_state state; 500 bzero(&state, sizeof(state)); 501 state.m = m; 502 if (flags & IP_ROUTETOIF) { 503 state.ro = &iproute; 504 bzero(&iproute, sizeof(iproute)); 505 } else 506 state.ro = ro; 507 state.dst = (struct sockaddr *)dst; 508 509 /* 510 * We can't defer the checksum of payload data if 511 * we're about to encrypt/authenticate it. 512 * 513 * XXX When we support crypto offloading functions of 514 * XXX network interfaces, we need to reconsider this, 515 * XXX since it's likely that they'll support checksumming, 516 * XXX as well. 517 */ 518 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 519 in_delayed_cksum(m); 520 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 521 } 522 523 error = ipsec4_output(&state, sp, flags); 524 525 m = state.m; 526 if (flags & IP_ROUTETOIF) { 527 /* 528 * if we have tunnel mode SA, we may need to ignore 529 * IP_ROUTETOIF. 530 */ 531 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 532 flags &= ~IP_ROUTETOIF; 533 ro = state.ro; 534 } 535 } else 536 ro = state.ro; 537 dst = (struct sockaddr_in *)state.dst; 538 if (error) { 539 /* mbuf is already reclaimed in ipsec4_output. */ 540 m0 = NULL; 541 switch (error) { 542 case EHOSTUNREACH: 543 case ENETUNREACH: 544 case EMSGSIZE: 545 case ENOBUFS: 546 case ENOMEM: 547 break; 548 default: 549 printf("ip4_output (ipsec): error code %d\n", error); 550 /*fall through*/ 551 case ENOENT: 552 /* don't show these error codes to the user */ 553 error = 0; 554 break; 555 } 556 goto bad; 557 } 558 559 /* be sure to update variables that are affected by ipsec4_output() */ 560 ip = mtod(m, struct ip *); 561 #ifdef _IP_VHL 562 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 563 #else 564 hlen = ip->ip_hl << 2; 565 #endif 566 ip_len = ntohs(ip->ip_len); 567 568 if (ro->ro_rt == NULL) { 569 if ((flags & IP_ROUTETOIF) == 0) { 570 printf("ip_output: " 571 "can't update route after IPsec processing\n"); 572 error = EHOSTUNREACH; /*XXX*/ 573 goto bad; 574 } 575 } else { 576 /* nobody uses ia beyond here */ 577 if (state.encap) 578 ifp = ro->ro_rt->rt_ifp; 579 } 580 } 581 582 skip_ipsec: 583 #endif /*IPSEC*/ 584 585 #ifdef PFIL_HOOKS 586 /* 587 * Run through list of hooks for output packets. 588 */ 589 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, 590 PFIL_OUT)) != 0) 591 goto done; 592 if (m == NULL) 593 goto done; 594 595 ip = mtod(m, struct ip *); 596 #endif /* PFIL_HOOKS */ 597 598 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 599 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 600 /* 601 * If small enough for mtu of path, can just send directly. 602 */ 603 if (ip_len <= mtu) { 604 #if IFA_STATS 605 /* 606 * search for the source address structure to 607 * maintain output statistics. 608 */ 609 INADDR_TO_IA(ip->ip_src, ia); 610 if (ia) 611 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 612 #endif 613 /* 614 * Always initialize the sum to 0! Some HW assisted 615 * checksumming requires this. 616 */ 617 ip->ip_sum = 0; 618 619 /* 620 * Perform any checksums that the hardware can't do 621 * for us. 622 * 623 * XXX Does any hardware require the {th,uh}_sum 624 * XXX fields to be 0? 625 */ 626 if (sw_csum & M_CSUM_IPv4) { 627 ip->ip_sum = in_cksum(m, hlen); 628 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 629 } 630 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 631 in_delayed_cksum(m); 632 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 633 } 634 635 #ifdef IPSEC 636 /* clean ipsec history once it goes out of the node */ 637 ipsec_delaux(m); 638 #endif 639 error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt); 640 goto done; 641 } 642 643 /* 644 * We can't use HW checksumming if we're about to 645 * to fragment the packet. 646 * 647 * XXX Some hardware can do this. 648 */ 649 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 650 in_delayed_cksum(m); 651 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 652 } 653 654 /* 655 * Too large for interface; fragment if possible. 656 * Must be able to put at least 8 bytes per fragment. 657 */ 658 if (ntohs(ip->ip_off) & IP_DF) { 659 if (flags & IP_RETURNMTU) 660 *mtu_p = mtu; 661 error = EMSGSIZE; 662 ipstat.ips_cantfrag++; 663 goto bad; 664 } 665 len = (mtu - hlen) &~ 7; 666 if (len < 8) { 667 error = EMSGSIZE; 668 goto bad; 669 } 670 671 { 672 int mhlen, firstlen = len; 673 struct mbuf **mnext = &m->m_nextpkt; 674 int fragments = 0; 675 int s; 676 677 /* 678 * Loop through length of segment after first fragment, 679 * make new header and copy data of each part and link onto chain. 680 */ 681 m0 = m; 682 mhlen = sizeof (struct ip); 683 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 684 MGETHDR(m, M_DONTWAIT, MT_HEADER); 685 if (m == 0) { 686 error = ENOBUFS; 687 ipstat.ips_odropped++; 688 goto sendorfree; 689 } 690 MCLAIM(m, m0->m_owner); 691 *mnext = m; 692 mnext = &m->m_nextpkt; 693 m->m_data += max_linkhdr; 694 mhip = mtod(m, struct ip *); 695 *mhip = *ip; 696 /* we must inherit MCAST and BCAST flags */ 697 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 698 if (hlen > sizeof (struct ip)) { 699 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 700 mhip->ip_hl = mhlen >> 2; 701 } 702 m->m_len = mhlen; 703 mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF); 704 if (ip->ip_off & IP_MF) 705 mhip->ip_off |= IP_MF; 706 if (off + len >= ntohs(ip->ip_len)) 707 len = ntohs(ip->ip_len) - off; 708 else 709 mhip->ip_off |= IP_MF; 710 HTONS(mhip->ip_off); 711 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 712 m->m_next = m_copy(m0, off, len); 713 if (m->m_next == 0) { 714 error = ENOBUFS; /* ??? */ 715 ipstat.ips_odropped++; 716 goto sendorfree; 717 } 718 m->m_pkthdr.len = mhlen + len; 719 m->m_pkthdr.rcvif = (struct ifnet *)0; 720 mhip->ip_sum = 0; 721 if (sw_csum & M_CSUM_IPv4) { 722 mhip->ip_sum = in_cksum(m, mhlen); 723 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 724 } else { 725 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 726 } 727 ipstat.ips_ofragments++; 728 fragments++; 729 } 730 /* 731 * Update first fragment by trimming what's been copied out 732 * and updating header, then send each fragment (in order). 733 */ 734 m = m0; 735 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 736 m->m_pkthdr.len = hlen + firstlen; 737 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 738 ip->ip_off |= htons(IP_MF); 739 ip->ip_sum = 0; 740 if (sw_csum & M_CSUM_IPv4) { 741 ip->ip_sum = in_cksum(m, hlen); 742 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 743 } else { 744 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 745 } 746 sendorfree: 747 /* 748 * If there is no room for all the fragments, don't queue 749 * any of them. 750 */ 751 s = splnet(); 752 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments) 753 error = ENOBUFS; 754 splx(s); 755 for (m = m0; m; m = m0) { 756 m0 = m->m_nextpkt; 757 m->m_nextpkt = 0; 758 if (error == 0) { 759 #if IFA_STATS 760 /* 761 * search for the source address structure to 762 * maintain output statistics. 763 */ 764 INADDR_TO_IA(ip->ip_src, ia); 765 if (ia) { 766 ia->ia_ifa.ifa_data.ifad_outbytes += 767 ntohs(ip->ip_len); 768 } 769 #endif 770 #ifdef IPSEC 771 /* clean ipsec history once it goes out of the node */ 772 ipsec_delaux(m); 773 #endif 774 KASSERT((m->m_pkthdr.csum_flags & 775 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 776 error = (*ifp->if_output)(ifp, m, sintosa(dst), 777 ro->ro_rt); 778 } else 779 m_freem(m); 780 } 781 782 if (error == 0) 783 ipstat.ips_fragmented++; 784 } 785 done: 786 if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) { 787 RTFREE(ro->ro_rt); 788 ro->ro_rt = 0; 789 } 790 791 #ifdef IPSEC 792 if (sp != NULL) { 793 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 794 printf("DP ip_output call free SP:%p\n", sp)); 795 key_freesp(sp); 796 } 797 #endif /* IPSEC */ 798 799 return (error); 800 bad: 801 m_freem(m); 802 goto done; 803 } 804 805 /* 806 * Process a delayed payload checksum calculation. 807 */ 808 void 809 in_delayed_cksum(struct mbuf *m) 810 { 811 struct ip *ip; 812 u_int16_t csum, offset; 813 814 ip = mtod(m, struct ip *); 815 offset = ip->ip_hl << 2; 816 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 817 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 818 csum = 0xffff; 819 820 offset += m->m_pkthdr.csum_data; /* checksum offset */ 821 822 if ((offset + sizeof(u_int16_t)) > m->m_len) { 823 /* This happen when ip options were inserted 824 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 825 m->m_len, offset, ip->ip_p); 826 */ 827 m_copyback(m, offset, sizeof(csum), (caddr_t) &csum); 828 } else 829 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 830 } 831 832 /* 833 * Determine the maximum length of the options to be inserted; 834 * we would far rather allocate too much space rather than too little. 835 */ 836 837 u_int 838 ip_optlen(inp) 839 struct inpcb *inp; 840 { 841 struct mbuf *m = inp->inp_options; 842 843 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 844 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 845 else 846 return 0; 847 } 848 849 850 /* 851 * Insert IP options into preformed packet. 852 * Adjust IP destination as required for IP source routing, 853 * as indicated by a non-zero in_addr at the start of the options. 854 */ 855 static struct mbuf * 856 ip_insertoptions(m, opt, phlen) 857 struct mbuf *m; 858 struct mbuf *opt; 859 int *phlen; 860 { 861 struct ipoption *p = mtod(opt, struct ipoption *); 862 struct mbuf *n; 863 struct ip *ip = mtod(m, struct ip *); 864 unsigned optlen; 865 866 optlen = opt->m_len - sizeof(p->ipopt_dst); 867 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 868 return (m); /* XXX should fail */ 869 if (!in_nullhost(p->ipopt_dst)) 870 ip->ip_dst = p->ipopt_dst; 871 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 872 MGETHDR(n, M_DONTWAIT, MT_HEADER); 873 if (n == 0) 874 return (m); 875 MCLAIM(n, m->m_owner); 876 M_COPY_PKTHDR(n, m); 877 m->m_flags &= ~M_PKTHDR; 878 m->m_len -= sizeof(struct ip); 879 m->m_data += sizeof(struct ip); 880 n->m_next = m; 881 m = n; 882 m->m_len = optlen + sizeof(struct ip); 883 m->m_data += max_linkhdr; 884 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 885 } else { 886 m->m_data -= optlen; 887 m->m_len += optlen; 888 memmove(mtod(m, caddr_t), ip, sizeof(struct ip)); 889 } 890 m->m_pkthdr.len += optlen; 891 ip = mtod(m, struct ip *); 892 bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen); 893 *phlen = sizeof(struct ip) + optlen; 894 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 895 return (m); 896 } 897 898 /* 899 * Copy options from ip to jp, 900 * omitting those not copied during fragmentation. 901 */ 902 int 903 ip_optcopy(ip, jp) 904 struct ip *ip, *jp; 905 { 906 u_char *cp, *dp; 907 int opt, optlen, cnt; 908 909 cp = (u_char *)(ip + 1); 910 dp = (u_char *)(jp + 1); 911 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 912 for (; cnt > 0; cnt -= optlen, cp += optlen) { 913 opt = cp[0]; 914 if (opt == IPOPT_EOL) 915 break; 916 if (opt == IPOPT_NOP) { 917 /* Preserve for IP mcast tunnel's LSRR alignment. */ 918 *dp++ = IPOPT_NOP; 919 optlen = 1; 920 continue; 921 } 922 #ifdef DIAGNOSTIC 923 if (cnt < IPOPT_OLEN + sizeof(*cp)) 924 panic("malformed IPv4 option passed to ip_optcopy"); 925 #endif 926 optlen = cp[IPOPT_OLEN]; 927 #ifdef DIAGNOSTIC 928 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 929 panic("malformed IPv4 option passed to ip_optcopy"); 930 #endif 931 /* bogus lengths should have been caught by ip_dooptions */ 932 if (optlen > cnt) 933 optlen = cnt; 934 if (IPOPT_COPIED(opt)) { 935 bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen); 936 dp += optlen; 937 } 938 } 939 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 940 *dp++ = IPOPT_EOL; 941 return (optlen); 942 } 943 944 /* 945 * IP socket option processing. 946 */ 947 int 948 ip_ctloutput(op, so, level, optname, mp) 949 int op; 950 struct socket *so; 951 int level, optname; 952 struct mbuf **mp; 953 { 954 struct inpcb *inp = sotoinpcb(so); 955 struct mbuf *m = *mp; 956 int optval = 0; 957 int error = 0; 958 #ifdef IPSEC 959 #ifdef __NetBSD__ 960 struct proc *p = curproc; /*XXX*/ 961 #endif 962 #endif 963 964 if (level != IPPROTO_IP) { 965 error = EINVAL; 966 if (op == PRCO_SETOPT && *mp) 967 (void) m_free(*mp); 968 } else switch (op) { 969 970 case PRCO_SETOPT: 971 switch (optname) { 972 case IP_OPTIONS: 973 #ifdef notyet 974 case IP_RETOPTS: 975 return (ip_pcbopts(optname, &inp->inp_options, m)); 976 #else 977 return (ip_pcbopts(&inp->inp_options, m)); 978 #endif 979 980 case IP_TOS: 981 case IP_TTL: 982 case IP_RECVOPTS: 983 case IP_RECVRETOPTS: 984 case IP_RECVDSTADDR: 985 case IP_RECVIF: 986 if (m == NULL || m->m_len != sizeof(int)) 987 error = EINVAL; 988 else { 989 optval = *mtod(m, int *); 990 switch (optname) { 991 992 case IP_TOS: 993 inp->inp_ip.ip_tos = optval; 994 break; 995 996 case IP_TTL: 997 inp->inp_ip.ip_ttl = optval; 998 break; 999 #define OPTSET(bit) \ 1000 if (optval) \ 1001 inp->inp_flags |= bit; \ 1002 else \ 1003 inp->inp_flags &= ~bit; 1004 1005 case IP_RECVOPTS: 1006 OPTSET(INP_RECVOPTS); 1007 break; 1008 1009 case IP_RECVRETOPTS: 1010 OPTSET(INP_RECVRETOPTS); 1011 break; 1012 1013 case IP_RECVDSTADDR: 1014 OPTSET(INP_RECVDSTADDR); 1015 break; 1016 1017 case IP_RECVIF: 1018 OPTSET(INP_RECVIF); 1019 break; 1020 } 1021 } 1022 break; 1023 #undef OPTSET 1024 1025 case IP_MULTICAST_IF: 1026 case IP_MULTICAST_TTL: 1027 case IP_MULTICAST_LOOP: 1028 case IP_ADD_MEMBERSHIP: 1029 case IP_DROP_MEMBERSHIP: 1030 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1031 break; 1032 1033 case IP_PORTRANGE: 1034 if (m == 0 || m->m_len != sizeof(int)) 1035 error = EINVAL; 1036 else { 1037 optval = *mtod(m, int *); 1038 1039 switch (optval) { 1040 1041 case IP_PORTRANGE_DEFAULT: 1042 case IP_PORTRANGE_HIGH: 1043 inp->inp_flags &= ~(INP_LOWPORT); 1044 break; 1045 1046 case IP_PORTRANGE_LOW: 1047 inp->inp_flags |= INP_LOWPORT; 1048 break; 1049 1050 default: 1051 error = EINVAL; 1052 break; 1053 } 1054 } 1055 break; 1056 1057 #ifdef IPSEC 1058 case IP_IPSEC_POLICY: 1059 { 1060 caddr_t req = NULL; 1061 size_t len = 0; 1062 int priv = 0; 1063 1064 #ifdef __NetBSD__ 1065 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1066 priv = 0; 1067 else 1068 priv = 1; 1069 #else 1070 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1071 #endif 1072 if (m) { 1073 req = mtod(m, caddr_t); 1074 len = m->m_len; 1075 } 1076 error = ipsec4_set_policy(inp, optname, req, len, priv); 1077 break; 1078 } 1079 #endif /*IPSEC*/ 1080 1081 default: 1082 error = ENOPROTOOPT; 1083 break; 1084 } 1085 if (m) 1086 (void)m_free(m); 1087 break; 1088 1089 case PRCO_GETOPT: 1090 switch (optname) { 1091 case IP_OPTIONS: 1092 case IP_RETOPTS: 1093 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1094 MCLAIM(m, so->so_mowner); 1095 if (inp->inp_options) { 1096 m->m_len = inp->inp_options->m_len; 1097 bcopy(mtod(inp->inp_options, caddr_t), 1098 mtod(m, caddr_t), (unsigned)m->m_len); 1099 } else 1100 m->m_len = 0; 1101 break; 1102 1103 case IP_TOS: 1104 case IP_TTL: 1105 case IP_RECVOPTS: 1106 case IP_RECVRETOPTS: 1107 case IP_RECVDSTADDR: 1108 case IP_RECVIF: 1109 case IP_ERRORMTU: 1110 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1111 MCLAIM(m, so->so_mowner); 1112 m->m_len = sizeof(int); 1113 switch (optname) { 1114 1115 case IP_TOS: 1116 optval = inp->inp_ip.ip_tos; 1117 break; 1118 1119 case IP_TTL: 1120 optval = inp->inp_ip.ip_ttl; 1121 break; 1122 1123 case IP_ERRORMTU: 1124 optval = inp->inp_errormtu; 1125 break; 1126 1127 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1128 1129 case IP_RECVOPTS: 1130 optval = OPTBIT(INP_RECVOPTS); 1131 break; 1132 1133 case IP_RECVRETOPTS: 1134 optval = OPTBIT(INP_RECVRETOPTS); 1135 break; 1136 1137 case IP_RECVDSTADDR: 1138 optval = OPTBIT(INP_RECVDSTADDR); 1139 break; 1140 1141 case IP_RECVIF: 1142 optval = OPTBIT(INP_RECVIF); 1143 break; 1144 } 1145 *mtod(m, int *) = optval; 1146 break; 1147 1148 #ifdef IPSEC 1149 case IP_IPSEC_POLICY: 1150 { 1151 caddr_t req = NULL; 1152 size_t len = 0; 1153 1154 if (m) { 1155 req = mtod(m, caddr_t); 1156 len = m->m_len; 1157 } 1158 error = ipsec4_get_policy(inp, req, len, mp); 1159 break; 1160 } 1161 #endif /*IPSEC*/ 1162 1163 case IP_MULTICAST_IF: 1164 case IP_MULTICAST_TTL: 1165 case IP_MULTICAST_LOOP: 1166 case IP_ADD_MEMBERSHIP: 1167 case IP_DROP_MEMBERSHIP: 1168 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1169 if (*mp) 1170 MCLAIM(*mp, so->so_mowner); 1171 break; 1172 1173 case IP_PORTRANGE: 1174 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1175 MCLAIM(m, so->so_mowner); 1176 m->m_len = sizeof(int); 1177 1178 if (inp->inp_flags & INP_LOWPORT) 1179 optval = IP_PORTRANGE_LOW; 1180 else 1181 optval = IP_PORTRANGE_DEFAULT; 1182 1183 *mtod(m, int *) = optval; 1184 break; 1185 1186 default: 1187 error = ENOPROTOOPT; 1188 break; 1189 } 1190 break; 1191 } 1192 return (error); 1193 } 1194 1195 /* 1196 * Set up IP options in pcb for insertion in output packets. 1197 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1198 * with destination address if source routed. 1199 */ 1200 int 1201 #ifdef notyet 1202 ip_pcbopts(optname, pcbopt, m) 1203 int optname; 1204 #else 1205 ip_pcbopts(pcbopt, m) 1206 #endif 1207 struct mbuf **pcbopt; 1208 struct mbuf *m; 1209 { 1210 int cnt, optlen; 1211 u_char *cp; 1212 u_char opt; 1213 1214 /* turn off any old options */ 1215 if (*pcbopt) 1216 (void)m_free(*pcbopt); 1217 *pcbopt = 0; 1218 if (m == (struct mbuf *)0 || m->m_len == 0) { 1219 /* 1220 * Only turning off any previous options. 1221 */ 1222 if (m) 1223 (void)m_free(m); 1224 return (0); 1225 } 1226 1227 #ifndef __vax__ 1228 if (m->m_len % sizeof(int32_t)) 1229 goto bad; 1230 #endif 1231 /* 1232 * IP first-hop destination address will be stored before 1233 * actual options; move other options back 1234 * and clear it when none present. 1235 */ 1236 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1237 goto bad; 1238 cnt = m->m_len; 1239 m->m_len += sizeof(struct in_addr); 1240 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1241 memmove(cp, mtod(m, caddr_t), (unsigned)cnt); 1242 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1243 1244 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1245 opt = cp[IPOPT_OPTVAL]; 1246 if (opt == IPOPT_EOL) 1247 break; 1248 if (opt == IPOPT_NOP) 1249 optlen = 1; 1250 else { 1251 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1252 goto bad; 1253 optlen = cp[IPOPT_OLEN]; 1254 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1255 goto bad; 1256 } 1257 switch (opt) { 1258 1259 default: 1260 break; 1261 1262 case IPOPT_LSRR: 1263 case IPOPT_SSRR: 1264 /* 1265 * user process specifies route as: 1266 * ->A->B->C->D 1267 * D must be our final destination (but we can't 1268 * check that since we may not have connected yet). 1269 * A is first hop destination, which doesn't appear in 1270 * actual IP option, but is stored before the options. 1271 */ 1272 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1273 goto bad; 1274 m->m_len -= sizeof(struct in_addr); 1275 cnt -= sizeof(struct in_addr); 1276 optlen -= sizeof(struct in_addr); 1277 cp[IPOPT_OLEN] = optlen; 1278 /* 1279 * Move first hop before start of options. 1280 */ 1281 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1282 sizeof(struct in_addr)); 1283 /* 1284 * Then copy rest of options back 1285 * to close up the deleted entry. 1286 */ 1287 memmove(&cp[IPOPT_OFFSET+1], 1288 (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1289 (unsigned)cnt + sizeof(struct in_addr)); 1290 break; 1291 } 1292 } 1293 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1294 goto bad; 1295 *pcbopt = m; 1296 return (0); 1297 1298 bad: 1299 (void)m_free(m); 1300 return (EINVAL); 1301 } 1302 1303 /* 1304 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1305 */ 1306 static struct ifnet * 1307 ip_multicast_if(a, ifindexp) 1308 struct in_addr *a; 1309 int *ifindexp; 1310 { 1311 int ifindex; 1312 struct ifnet *ifp; 1313 1314 if (ifindexp) 1315 *ifindexp = 0; 1316 if (ntohl(a->s_addr) >> 24 == 0) { 1317 ifindex = ntohl(a->s_addr) & 0xffffff; 1318 if (ifindex < 0 || if_index < ifindex) 1319 return NULL; 1320 ifp = ifindex2ifnet[ifindex]; 1321 if (ifindexp) 1322 *ifindexp = ifindex; 1323 } else { 1324 INADDR_TO_IFP(*a, ifp); 1325 } 1326 return ifp; 1327 } 1328 1329 /* 1330 * Set the IP multicast options in response to user setsockopt(). 1331 */ 1332 int 1333 ip_setmoptions(optname, imop, m) 1334 int optname; 1335 struct ip_moptions **imop; 1336 struct mbuf *m; 1337 { 1338 int error = 0; 1339 u_char loop; 1340 int i; 1341 struct in_addr addr; 1342 struct ip_mreq *mreq; 1343 struct ifnet *ifp; 1344 struct ip_moptions *imo = *imop; 1345 struct route ro; 1346 struct sockaddr_in *dst; 1347 int ifindex; 1348 1349 if (imo == NULL) { 1350 /* 1351 * No multicast option buffer attached to the pcb; 1352 * allocate one and initialize to default values. 1353 */ 1354 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1355 M_WAITOK); 1356 1357 if (imo == NULL) 1358 return (ENOBUFS); 1359 *imop = imo; 1360 imo->imo_multicast_ifp = NULL; 1361 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1362 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1363 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1364 imo->imo_num_memberships = 0; 1365 } 1366 1367 switch (optname) { 1368 1369 case IP_MULTICAST_IF: 1370 /* 1371 * Select the interface for outgoing multicast packets. 1372 */ 1373 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1374 error = EINVAL; 1375 break; 1376 } 1377 addr = *(mtod(m, struct in_addr *)); 1378 /* 1379 * INADDR_ANY is used to remove a previous selection. 1380 * When no interface is selected, a default one is 1381 * chosen every time a multicast packet is sent. 1382 */ 1383 if (in_nullhost(addr)) { 1384 imo->imo_multicast_ifp = NULL; 1385 break; 1386 } 1387 /* 1388 * The selected interface is identified by its local 1389 * IP address. Find the interface and confirm that 1390 * it supports multicasting. 1391 */ 1392 ifp = ip_multicast_if(&addr, &ifindex); 1393 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1394 error = EADDRNOTAVAIL; 1395 break; 1396 } 1397 imo->imo_multicast_ifp = ifp; 1398 if (ifindex) 1399 imo->imo_multicast_addr = addr; 1400 else 1401 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1402 break; 1403 1404 case IP_MULTICAST_TTL: 1405 /* 1406 * Set the IP time-to-live for outgoing multicast packets. 1407 */ 1408 if (m == NULL || m->m_len != 1) { 1409 error = EINVAL; 1410 break; 1411 } 1412 imo->imo_multicast_ttl = *(mtod(m, u_char *)); 1413 break; 1414 1415 case IP_MULTICAST_LOOP: 1416 /* 1417 * Set the loopback flag for outgoing multicast packets. 1418 * Must be zero or one. 1419 */ 1420 if (m == NULL || m->m_len != 1 || 1421 (loop = *(mtod(m, u_char *))) > 1) { 1422 error = EINVAL; 1423 break; 1424 } 1425 imo->imo_multicast_loop = loop; 1426 break; 1427 1428 case IP_ADD_MEMBERSHIP: 1429 /* 1430 * Add a multicast group membership. 1431 * Group must be a valid IP multicast address. 1432 */ 1433 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1434 error = EINVAL; 1435 break; 1436 } 1437 mreq = mtod(m, struct ip_mreq *); 1438 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1439 error = EINVAL; 1440 break; 1441 } 1442 /* 1443 * If no interface address was provided, use the interface of 1444 * the route to the given multicast address. 1445 */ 1446 if (in_nullhost(mreq->imr_interface)) { 1447 bzero((caddr_t)&ro, sizeof(ro)); 1448 ro.ro_rt = NULL; 1449 dst = satosin(&ro.ro_dst); 1450 dst->sin_len = sizeof(*dst); 1451 dst->sin_family = AF_INET; 1452 dst->sin_addr = mreq->imr_multiaddr; 1453 rtalloc(&ro); 1454 if (ro.ro_rt == NULL) { 1455 error = EADDRNOTAVAIL; 1456 break; 1457 } 1458 ifp = ro.ro_rt->rt_ifp; 1459 rtfree(ro.ro_rt); 1460 } else { 1461 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1462 } 1463 /* 1464 * See if we found an interface, and confirm that it 1465 * supports multicast. 1466 */ 1467 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1468 error = EADDRNOTAVAIL; 1469 break; 1470 } 1471 /* 1472 * See if the membership already exists or if all the 1473 * membership slots are full. 1474 */ 1475 for (i = 0; i < imo->imo_num_memberships; ++i) { 1476 if (imo->imo_membership[i]->inm_ifp == ifp && 1477 in_hosteq(imo->imo_membership[i]->inm_addr, 1478 mreq->imr_multiaddr)) 1479 break; 1480 } 1481 if (i < imo->imo_num_memberships) { 1482 error = EADDRINUSE; 1483 break; 1484 } 1485 if (i == IP_MAX_MEMBERSHIPS) { 1486 error = ETOOMANYREFS; 1487 break; 1488 } 1489 /* 1490 * Everything looks good; add a new record to the multicast 1491 * address list for the given interface. 1492 */ 1493 if ((imo->imo_membership[i] = 1494 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1495 error = ENOBUFS; 1496 break; 1497 } 1498 ++imo->imo_num_memberships; 1499 break; 1500 1501 case IP_DROP_MEMBERSHIP: 1502 /* 1503 * Drop a multicast group membership. 1504 * Group must be a valid IP multicast address. 1505 */ 1506 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1507 error = EINVAL; 1508 break; 1509 } 1510 mreq = mtod(m, struct ip_mreq *); 1511 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1512 error = EINVAL; 1513 break; 1514 } 1515 /* 1516 * If an interface address was specified, get a pointer 1517 * to its ifnet structure. 1518 */ 1519 if (in_nullhost(mreq->imr_interface)) 1520 ifp = NULL; 1521 else { 1522 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1523 if (ifp == NULL) { 1524 error = EADDRNOTAVAIL; 1525 break; 1526 } 1527 } 1528 /* 1529 * Find the membership in the membership array. 1530 */ 1531 for (i = 0; i < imo->imo_num_memberships; ++i) { 1532 if ((ifp == NULL || 1533 imo->imo_membership[i]->inm_ifp == ifp) && 1534 in_hosteq(imo->imo_membership[i]->inm_addr, 1535 mreq->imr_multiaddr)) 1536 break; 1537 } 1538 if (i == imo->imo_num_memberships) { 1539 error = EADDRNOTAVAIL; 1540 break; 1541 } 1542 /* 1543 * Give up the multicast address record to which the 1544 * membership points. 1545 */ 1546 in_delmulti(imo->imo_membership[i]); 1547 /* 1548 * Remove the gap in the membership array. 1549 */ 1550 for (++i; i < imo->imo_num_memberships; ++i) 1551 imo->imo_membership[i-1] = imo->imo_membership[i]; 1552 --imo->imo_num_memberships; 1553 break; 1554 1555 default: 1556 error = EOPNOTSUPP; 1557 break; 1558 } 1559 1560 /* 1561 * If all options have default values, no need to keep the mbuf. 1562 */ 1563 if (imo->imo_multicast_ifp == NULL && 1564 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1565 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1566 imo->imo_num_memberships == 0) { 1567 free(*imop, M_IPMOPTS); 1568 *imop = NULL; 1569 } 1570 1571 return (error); 1572 } 1573 1574 /* 1575 * Return the IP multicast options in response to user getsockopt(). 1576 */ 1577 int 1578 ip_getmoptions(optname, imo, mp) 1579 int optname; 1580 struct ip_moptions *imo; 1581 struct mbuf **mp; 1582 { 1583 u_char *ttl; 1584 u_char *loop; 1585 struct in_addr *addr; 1586 struct in_ifaddr *ia; 1587 1588 *mp = m_get(M_WAIT, MT_SOOPTS); 1589 1590 switch (optname) { 1591 1592 case IP_MULTICAST_IF: 1593 addr = mtod(*mp, struct in_addr *); 1594 (*mp)->m_len = sizeof(struct in_addr); 1595 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1596 *addr = zeroin_addr; 1597 else if (imo->imo_multicast_addr.s_addr) { 1598 /* return the value user has set */ 1599 *addr = imo->imo_multicast_addr; 1600 } else { 1601 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1602 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1603 } 1604 return (0); 1605 1606 case IP_MULTICAST_TTL: 1607 ttl = mtod(*mp, u_char *); 1608 (*mp)->m_len = 1; 1609 *ttl = imo ? imo->imo_multicast_ttl 1610 : IP_DEFAULT_MULTICAST_TTL; 1611 return (0); 1612 1613 case IP_MULTICAST_LOOP: 1614 loop = mtod(*mp, u_char *); 1615 (*mp)->m_len = 1; 1616 *loop = imo ? imo->imo_multicast_loop 1617 : IP_DEFAULT_MULTICAST_LOOP; 1618 return (0); 1619 1620 default: 1621 return (EOPNOTSUPP); 1622 } 1623 } 1624 1625 /* 1626 * Discard the IP multicast options. 1627 */ 1628 void 1629 ip_freemoptions(imo) 1630 struct ip_moptions *imo; 1631 { 1632 int i; 1633 1634 if (imo != NULL) { 1635 for (i = 0; i < imo->imo_num_memberships; ++i) 1636 in_delmulti(imo->imo_membership[i]); 1637 free(imo, M_IPMOPTS); 1638 } 1639 } 1640 1641 /* 1642 * Routine called from ip_output() to loop back a copy of an IP multicast 1643 * packet to the input queue of a specified interface. Note that this 1644 * calls the output routine of the loopback "driver", but with an interface 1645 * pointer that might NOT be &loif -- easier than replicating that code here. 1646 */ 1647 static void 1648 ip_mloopback(ifp, m, dst) 1649 struct ifnet *ifp; 1650 struct mbuf *m; 1651 struct sockaddr_in *dst; 1652 { 1653 struct ip *ip; 1654 struct mbuf *copym; 1655 1656 copym = m_copy(m, 0, M_COPYALL); 1657 if (copym != NULL 1658 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1659 copym = m_pullup(copym, sizeof(struct ip)); 1660 if (copym != NULL) { 1661 /* 1662 * We don't bother to fragment if the IP length is greater 1663 * than the interface's MTU. Can this possibly matter? 1664 */ 1665 ip = mtod(copym, struct ip *); 1666 1667 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1668 in_delayed_cksum(copym); 1669 copym->m_pkthdr.csum_flags &= 1670 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1671 } 1672 1673 ip->ip_sum = 0; 1674 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1675 (void) looutput(ifp, copym, sintosa(dst), NULL); 1676 } 1677 } 1678