ip_input.c revision 1.82.2.2.4.1
1 /* $NetBSD: ip_input.c,v 1.82.2.2.4.1 1999/06/28 06:37:00 itojun Exp $ */ 2 3 /*- 4 * Copyright (c) 1998 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Public Access Networks Corporation ("Panix"). It was developed under 9 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the NetBSD 22 * Foundation, Inc. and its contributors. 23 * 4. Neither the name of The NetBSD Foundation nor the names of its 24 * contributors may be used to endorse or promote products derived 25 * from this software without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 28 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 29 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 30 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 31 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 32 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 33 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 34 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 35 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 36 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 37 * POSSIBILITY OF SUCH DAMAGE. 38 */ 39 40 /* 41 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 42 * All rights reserved. 43 * 44 * Redistribution and use in source and binary forms, with or without 45 * modification, are permitted provided that the following conditions 46 * are met: 47 * 1. Redistributions of source code must retain the above copyright 48 * notice, this list of conditions and the following disclaimer. 49 * 2. Redistributions in binary form must reproduce the above copyright 50 * notice, this list of conditions and the following disclaimer in the 51 * documentation and/or other materials provided with the distribution. 52 * 3. Neither the name of the project nor the names of its contributors 53 * may be used to endorse or promote products derived from this software 54 * without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 57 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 58 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 59 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 60 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 61 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 62 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 63 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 64 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 65 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 66 * SUCH DAMAGE. 67 */ 68 69 /* 70 * Copyright (c) 1982, 1986, 1988, 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_input.c 8.2 (Berkeley) 1/4/94 102 */ 103 104 #include "opt_gateway.h" 105 #include "opt_pfil_hooks.h" 106 #include "opt_mrouting.h" 107 108 #include <sys/param.h> 109 #include <sys/systm.h> 110 #include <sys/malloc.h> 111 #include <sys/mbuf.h> 112 #include <sys/domain.h> 113 #include <sys/protosw.h> 114 #include <sys/socket.h> 115 #include <sys/socketvar.h> 116 #include <sys/errno.h> 117 #include <sys/time.h> 118 #include <sys/kernel.h> 119 #include <sys/proc.h> 120 #include <sys/pool.h> 121 122 #include <vm/vm.h> 123 #include <sys/sysctl.h> 124 125 #include <net/if.h> 126 #include <net/if_dl.h> 127 #include <net/route.h> 128 #include <net/pfil.h> 129 130 #include <netinet/in.h> 131 #include <netinet/in_systm.h> 132 #include <netinet/ip.h> 133 #include <netinet/in_pcb.h> 134 #include <netinet/in_var.h> 135 #include <netinet/ip_var.h> 136 #include <netinet/ip_icmp.h> 137 /* just for gif_ttl */ 138 #include <netinet/in_gif.h> 139 #include "gif.h" 140 141 #ifdef IPSEC 142 #include <netinet6/ipsec.h> 143 #include <netinet6/ah.h> 144 #include <netinet6/esp.h> 145 #include <netkey/key.h> 146 #include <netkey/key_debug.h> 147 #endif 148 149 #ifndef IPFORWARDING 150 #ifdef GATEWAY 151 #define IPFORWARDING 1 /* forward IP packets not for us */ 152 #else /* GATEWAY */ 153 #define IPFORWARDING 0 /* don't forward IP packets not for us */ 154 #endif /* GATEWAY */ 155 #endif /* IPFORWARDING */ 156 #ifndef IPSENDREDIRECTS 157 #define IPSENDREDIRECTS 1 158 #endif 159 #ifndef IPFORWSRCRT 160 #define IPFORWSRCRT 1 /* forward source-routed packets */ 161 #endif 162 #ifndef IPALLOWSRCRT 163 #define IPALLOWSRCRT 1 /* allow source-routed packets */ 164 #endif 165 #ifndef IPMTUDISC 166 #define IPMTUDISC 0 167 #endif 168 #ifndef IPMTUDISCTIMEOUT 169 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 170 #endif 171 172 /* 173 * Note: DIRECTED_BROADCAST is handled this way so that previous 174 * configuration using this option will Just Work. 175 */ 176 #ifndef IPDIRECTEDBCAST 177 #ifdef DIRECTED_BROADCAST 178 #define IPDIRECTEDBCAST 1 179 #else 180 #define IPDIRECTEDBCAST 0 181 #endif /* DIRECTED_BROADCAST */ 182 #endif /* IPDIRECTEDBCAST */ 183 int ipforwarding = IPFORWARDING; 184 int ipsendredirects = IPSENDREDIRECTS; 185 int ip_defttl = IPDEFTTL; 186 int ip_forwsrcrt = IPFORWSRCRT; 187 int ip_directedbcast = IPDIRECTEDBCAST; 188 int ip_allowsrcrt = IPALLOWSRCRT; 189 int ip_mtudisc = IPMTUDISC; 190 u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 191 #ifdef DIAGNOSTIC 192 int ipprintfs = 0; 193 #endif 194 195 struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 196 197 extern struct domain inetdomain; 198 extern struct protosw inetsw[]; 199 u_char ip_protox[IPPROTO_MAX]; 200 int ipqmaxlen = IFQ_MAXLEN; 201 struct in_ifaddrhead in_ifaddr; 202 struct in_ifaddrhashhead *in_ifaddrhashtbl; 203 struct ifqueue ipintrq; 204 struct ipstat ipstat; 205 u_int16_t ip_id; 206 int ip_defttl; 207 208 struct ipqhead ipq; 209 int ipq_locked; 210 211 static __inline int ipq_lock_try __P((void)); 212 static __inline void ipq_unlock __P((void)); 213 214 static __inline int 215 ipq_lock_try() 216 { 217 int s; 218 219 s = splimp(); 220 if (ipq_locked) { 221 splx(s); 222 return (0); 223 } 224 ipq_locked = 1; 225 splx(s); 226 return (1); 227 } 228 229 static __inline void 230 ipq_unlock() 231 { 232 int s; 233 234 s = splimp(); 235 ipq_locked = 0; 236 splx(s); 237 } 238 239 #ifdef DIAGNOSTIC 240 #define IPQ_LOCK() \ 241 do { \ 242 if (ipq_lock_try() == 0) { \ 243 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \ 244 panic("ipq_lock"); \ 245 } \ 246 } while (0) 247 #define IPQ_LOCK_CHECK() \ 248 do { \ 249 if (ipq_locked == 0) { \ 250 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \ 251 panic("ipq lock check"); \ 252 } \ 253 } while (0) 254 #else 255 #define IPQ_LOCK() (void) ipq_lock_try() 256 #define IPQ_LOCK_CHECK() /* nothing */ 257 #endif 258 259 #define IPQ_UNLOCK() ipq_unlock() 260 261 struct pool ipqent_pool; 262 263 /* 264 * We need to save the IP options in case a protocol wants to respond 265 * to an incoming packet over the same route if the packet got here 266 * using IP source routing. This allows connection establishment and 267 * maintenance when the remote end is on a network that is not known 268 * to us. 269 */ 270 int ip_nhops = 0; 271 static struct ip_srcrt { 272 struct in_addr dst; /* final destination */ 273 char nop; /* one NOP to align */ 274 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 275 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 276 } ip_srcrt; 277 278 static void save_rte __P((u_char *, struct in_addr)); 279 280 /* 281 * IP initialization: fill in IP protocol switch table. 282 * All protocols not implemented in kernel go to raw IP protocol handler. 283 */ 284 void 285 ip_init() 286 { 287 register struct protosw *pr; 288 register int i; 289 290 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", 291 0, NULL, NULL, M_IPQ); 292 293 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 294 if (pr == 0) 295 panic("ip_init"); 296 for (i = 0; i < IPPROTO_MAX; i++) 297 ip_protox[i] = pr - inetsw; 298 for (pr = inetdomain.dom_protosw; 299 pr < inetdomain.dom_protoswNPROTOSW; pr++) 300 if (pr->pr_domain->dom_family == PF_INET && 301 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 302 ip_protox[pr->pr_protocol] = pr - inetsw; 303 LIST_INIT(&ipq); 304 ip_id = time.tv_sec & 0xffff; 305 ipintrq.ifq_maxlen = ipqmaxlen; 306 TAILQ_INIT(&in_ifaddr); 307 in_ifaddrhashtbl = 308 hashinit(IN_IFADDR_HASH_SIZE, M_IFADDR, M_WAITOK, &in_ifaddrhash); 309 if (ip_mtudisc != 0) 310 ip_mtudisc_timeout_q = 311 rt_timer_queue_create(ip_mtudisc_timeout); 312 #ifdef GATEWAY 313 ipflow_init(); 314 #endif 315 } 316 317 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 318 struct route ipforward_rt; 319 320 /* 321 * IP software interrupt routine 322 */ 323 void 324 ipintr() 325 { 326 int s; 327 struct mbuf *m; 328 329 while (1) { 330 s = splimp(); 331 IF_DEQUEUE(&ipintrq, m); 332 splx(s); 333 if (m == 0) 334 return; 335 ip_input(m); 336 } 337 } 338 339 /* 340 * Ip input routine. Checksum and byte swap header. If fragmented 341 * try to reassemble. Process options. Pass to next level. 342 */ 343 void 344 ip_input(struct mbuf *m) 345 { 346 register struct ip *ip = NULL; 347 register struct ipq *fp; 348 register struct in_ifaddr *ia; 349 register struct ifaddr *ifa; 350 struct ipqent *ipqe; 351 int hlen = 0, mff, len; 352 #ifdef PFIL_HOOKS 353 struct packet_filter_hook *pfh; 354 struct mbuf *m0; 355 int rv; 356 #endif /* PFIL_HOOKS */ 357 358 #ifdef DIAGNOSTIC 359 if ((m->m_flags & M_PKTHDR) == 0) 360 panic("ipintr no HDR"); 361 #endif 362 #ifdef IPSEC 363 /* 364 * should the inner packet be considered authentic? 365 * see comment in ah4_input(). 366 */ 367 if (m) { 368 m->m_flags &= ~M_AUTHIPHDR; 369 m->m_flags &= ~M_AUTHIPDGM; 370 } 371 #endif 372 /* 373 * If no IP addresses have been set yet but the interfaces 374 * are receiving, can't do anything with incoming packets yet. 375 */ 376 if (in_ifaddr.tqh_first == 0) 377 goto bad; 378 ipstat.ips_total++; 379 if (m->m_len < sizeof (struct ip) && 380 (m = m_pullup(m, sizeof (struct ip))) == 0) { 381 ipstat.ips_toosmall++; 382 return; 383 } 384 ip = mtod(m, struct ip *); 385 if (ip->ip_v != IPVERSION) { 386 ipstat.ips_badvers++; 387 goto bad; 388 } 389 hlen = ip->ip_hl << 2; 390 if (hlen < sizeof(struct ip)) { /* minimum header length */ 391 ipstat.ips_badhlen++; 392 goto bad; 393 } 394 if (hlen > m->m_len) { 395 if ((m = m_pullup(m, hlen)) == 0) { 396 ipstat.ips_badhlen++; 397 return; 398 } 399 ip = mtod(m, struct ip *); 400 } 401 if (in_cksum(m, hlen) != 0) { 402 ipstat.ips_badsum++; 403 goto bad; 404 } 405 406 /* 407 * Convert fields to host representation. 408 */ 409 NTOHS(ip->ip_len); 410 NTOHS(ip->ip_off); 411 len = ip->ip_len; 412 413 /* 414 * Check for additional length bogosity 415 */ 416 if (len < hlen) 417 { 418 ipstat.ips_badlen++; 419 goto bad; 420 } 421 422 /* 423 * Check that the amount of data in the buffers 424 * is as at least much as the IP header would have us expect. 425 * Trim mbufs if longer than we expect. 426 * Drop packet if shorter than we expect. 427 */ 428 if (m->m_pkthdr.len < len) { 429 ipstat.ips_tooshort++; 430 goto bad; 431 } 432 if (m->m_pkthdr.len > len) { 433 if (m->m_len == m->m_pkthdr.len) { 434 m->m_len = len; 435 m->m_pkthdr.len = len; 436 } else 437 m_adj(m, len - m->m_pkthdr.len); 438 } 439 440 /* 441 * Assume that we can create a fast-forward IP flow entry 442 * based on this packet. 443 */ 444 m->m_flags |= M_CANFASTFWD; 445 446 #ifdef PFIL_HOOKS 447 /* 448 * Run through list of hooks for input packets. If there are any 449 * filters which require that additional packets in the flow are 450 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 451 * Note that filters must _never_ set this flag, as another filter 452 * in the list may have previously cleared it. 453 */ 454 m0 = m; 455 for (pfh = pfil_hook_get(PFIL_IN); pfh; pfh = pfh->pfil_link.tqe_next) 456 if (pfh->pfil_func) { 457 rv = pfh->pfil_func(ip, hlen, m->m_pkthdr.rcvif, 0, &m0); 458 if (rv) 459 return; 460 m = m0; 461 if (m == NULL) 462 return; 463 ip = mtod(m, struct ip *); 464 } 465 #endif /* PFIL_HOOKS */ 466 467 /* 468 * Process options and, if not destined for us, 469 * ship it on. ip_dooptions returns 1 when an 470 * error was detected (causing an icmp message 471 * to be sent and the original packet to be freed). 472 */ 473 ip_nhops = 0; /* for source routed packets */ 474 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 475 return; 476 477 /* 478 * Check our list of addresses, to see if the packet is for us. 479 */ 480 INADDR_TO_IA(ip->ip_dst, ia); 481 if (ia != NULL) 482 goto ours; 483 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 484 for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; 485 ifa != NULL; ifa = ifa->ifa_list.tqe_next) { 486 if (ifa->ifa_addr->sa_family != AF_INET) continue; 487 ia = ifatoia(ifa); 488 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 489 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 490 /* 491 * Look for all-0's host part (old broadcast addr), 492 * either for subnet or net. 493 */ 494 ip->ip_dst.s_addr == ia->ia_subnet || 495 ip->ip_dst.s_addr == ia->ia_net) 496 goto ours; 497 /* 498 * An interface with IP address zero accepts 499 * all packets that arrive on that interface. 500 */ 501 if (in_nullhost(ia->ia_addr.sin_addr)) 502 goto ours; 503 } 504 } 505 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 506 struct in_multi *inm; 507 #ifdef MROUTING 508 extern struct socket *ip_mrouter; 509 510 if (m->m_flags & M_EXT) { 511 if ((m = m_pullup(m, hlen)) == 0) { 512 ipstat.ips_toosmall++; 513 return; 514 } 515 ip = mtod(m, struct ip *); 516 } 517 518 if (ip_mrouter) { 519 /* 520 * If we are acting as a multicast router, all 521 * incoming multicast packets are passed to the 522 * kernel-level multicast forwarding function. 523 * The packet is returned (relatively) intact; if 524 * ip_mforward() returns a non-zero value, the packet 525 * must be discarded, else it may be accepted below. 526 * 527 * (The IP ident field is put in the same byte order 528 * as expected when ip_mforward() is called from 529 * ip_output().) 530 */ 531 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 532 ipstat.ips_cantforward++; 533 m_freem(m); 534 return; 535 } 536 537 /* 538 * The process-level routing demon needs to receive 539 * all multicast IGMP packets, whether or not this 540 * host belongs to their destination groups. 541 */ 542 if (ip->ip_p == IPPROTO_IGMP) 543 goto ours; 544 ipstat.ips_forward++; 545 } 546 #endif 547 /* 548 * See if we belong to the destination multicast group on the 549 * arrival interface. 550 */ 551 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 552 if (inm == NULL) { 553 ipstat.ips_cantforward++; 554 m_freem(m); 555 return; 556 } 557 goto ours; 558 } 559 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 560 in_nullhost(ip->ip_dst)) 561 goto ours; 562 563 /* 564 * Not for us; forward if possible and desirable. 565 */ 566 if (ipforwarding == 0) { 567 ipstat.ips_cantforward++; 568 m_freem(m); 569 } else 570 ip_forward(m, 0); 571 return; 572 573 ours: 574 /* 575 * If offset or IP_MF are set, must reassemble. 576 * Otherwise, nothing need be done. 577 * (We could look in the reassembly queue to see 578 * if the packet was previously fragmented, 579 * but it's not worth the time; just let them time out.) 580 */ 581 if (ip->ip_off & ~(IP_DF|IP_RF)) { 582 /* 583 * Look for queue of fragments 584 * of this datagram. 585 */ 586 IPQ_LOCK(); 587 for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) 588 if (ip->ip_id == fp->ipq_id && 589 in_hosteq(ip->ip_src, fp->ipq_src) && 590 in_hosteq(ip->ip_dst, fp->ipq_dst) && 591 ip->ip_p == fp->ipq_p) 592 goto found; 593 fp = 0; 594 found: 595 596 /* 597 * Adjust ip_len to not reflect header, 598 * set ipqe_mff if more fragments are expected, 599 * convert offset of this to bytes. 600 */ 601 ip->ip_len -= hlen; 602 mff = (ip->ip_off & IP_MF) != 0; 603 if (mff) { 604 /* 605 * Make sure that fragments have a data length 606 * that's a non-zero multiple of 8 bytes. 607 */ 608 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 609 ipstat.ips_badfrags++; 610 IPQ_UNLOCK(); 611 goto bad; 612 } 613 } 614 ip->ip_off <<= 3; 615 616 /* 617 * If datagram marked as having more fragments 618 * or if this is not the first fragment, 619 * attempt reassembly; if it succeeds, proceed. 620 */ 621 if (mff || ip->ip_off) { 622 ipstat.ips_fragments++; 623 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 624 if (ipqe == NULL) { 625 ipstat.ips_rcvmemdrop++; 626 IPQ_UNLOCK(); 627 goto bad; 628 } 629 ipqe->ipqe_mff = mff; 630 ipqe->ipqe_m = m; 631 ipqe->ipqe_ip = ip; 632 m = ip_reass(ipqe, fp); 633 if (m == 0) { 634 IPQ_UNLOCK(); 635 return; 636 } 637 ipstat.ips_reassembled++; 638 ip = mtod(m, struct ip *); 639 hlen = ip->ip_hl << 2; 640 ip->ip_len += hlen; 641 } else 642 if (fp) 643 ip_freef(fp); 644 IPQ_UNLOCK(); 645 } 646 647 /* 648 * Switch out to protocol's input routine. 649 */ 650 #if IFA_STATS 651 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len; 652 #endif 653 ipstat.ips_delivered++; 654 { 655 int off = hlen, nh = ip->ip_p; 656 657 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 658 return; 659 } 660 bad: 661 m_freem(m); 662 } 663 664 /* 665 * Take incoming datagram fragment and try to 666 * reassemble it into whole datagram. If a chain for 667 * reassembly of this datagram already exists, then it 668 * is given as fp; otherwise have to make a chain. 669 */ 670 struct mbuf * 671 ip_reass(ipqe, fp) 672 register struct ipqent *ipqe; 673 register struct ipq *fp; 674 { 675 register struct mbuf *m = ipqe->ipqe_m; 676 register struct ipqent *nq, *p, *q; 677 struct ip *ip; 678 struct mbuf *t; 679 int hlen = ipqe->ipqe_ip->ip_hl << 2; 680 int i, next; 681 682 IPQ_LOCK_CHECK(); 683 684 /* 685 * Presence of header sizes in mbufs 686 * would confuse code below. 687 */ 688 m->m_data += hlen; 689 m->m_len -= hlen; 690 691 /* 692 * If first fragment to arrive, create a reassembly queue. 693 */ 694 if (fp == 0) { 695 MALLOC(fp, struct ipq *, sizeof (struct ipq), 696 M_FTABLE, M_NOWAIT); 697 if (fp == NULL) 698 goto dropfrag; 699 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 700 fp->ipq_ttl = IPFRAGTTL; 701 fp->ipq_p = ipqe->ipqe_ip->ip_p; 702 fp->ipq_id = ipqe->ipqe_ip->ip_id; 703 LIST_INIT(&fp->ipq_fragq); 704 fp->ipq_src = ipqe->ipqe_ip->ip_src; 705 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 706 p = NULL; 707 goto insert; 708 } 709 710 /* 711 * Find a segment which begins after this one does. 712 */ 713 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 714 p = q, q = q->ipqe_q.le_next) 715 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) 716 break; 717 718 /* 719 * If there is a preceding segment, it may provide some of 720 * our data already. If so, drop the data from the incoming 721 * segment. If it provides all of our data, drop us. 722 */ 723 if (p != NULL) { 724 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len - 725 ipqe->ipqe_ip->ip_off; 726 if (i > 0) { 727 if (i >= ipqe->ipqe_ip->ip_len) 728 goto dropfrag; 729 m_adj(ipqe->ipqe_m, i); 730 ipqe->ipqe_ip->ip_off += i; 731 ipqe->ipqe_ip->ip_len -= i; 732 } 733 } 734 735 /* 736 * While we overlap succeeding segments trim them or, 737 * if they are completely covered, dequeue them. 738 */ 739 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len > 740 q->ipqe_ip->ip_off; q = nq) { 741 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) - 742 q->ipqe_ip->ip_off; 743 if (i < q->ipqe_ip->ip_len) { 744 q->ipqe_ip->ip_len -= i; 745 q->ipqe_ip->ip_off += i; 746 m_adj(q->ipqe_m, i); 747 break; 748 } 749 nq = q->ipqe_q.le_next; 750 m_freem(q->ipqe_m); 751 LIST_REMOVE(q, ipqe_q); 752 pool_put(&ipqent_pool, q); 753 } 754 755 insert: 756 /* 757 * Stick new segment in its place; 758 * check for complete reassembly. 759 */ 760 if (p == NULL) { 761 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 762 } else { 763 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 764 } 765 next = 0; 766 for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; 767 p = q, q = q->ipqe_q.le_next) { 768 if (q->ipqe_ip->ip_off != next) 769 return (0); 770 next += q->ipqe_ip->ip_len; 771 } 772 if (p->ipqe_mff) 773 return (0); 774 775 /* 776 * Reassembly is complete. Check for a bogus message size and 777 * concatenate fragments. 778 */ 779 q = fp->ipq_fragq.lh_first; 780 ip = q->ipqe_ip; 781 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 782 ipstat.ips_toolong++; 783 ip_freef(fp); 784 return (0); 785 } 786 m = q->ipqe_m; 787 t = m->m_next; 788 m->m_next = 0; 789 m_cat(m, t); 790 nq = q->ipqe_q.le_next; 791 pool_put(&ipqent_pool, q); 792 for (q = nq; q != NULL; q = nq) { 793 t = q->ipqe_m; 794 nq = q->ipqe_q.le_next; 795 pool_put(&ipqent_pool, q); 796 m_cat(m, t); 797 } 798 799 /* 800 * Create header for new ip packet by 801 * modifying header of first packet; 802 * dequeue and discard fragment reassembly header. 803 * Make header visible. 804 */ 805 ip->ip_len = next; 806 ip->ip_ttl = 0; /* xxx */ 807 ip->ip_sum = 0; 808 ip->ip_src = fp->ipq_src; 809 ip->ip_dst = fp->ipq_dst; 810 LIST_REMOVE(fp, ipq_q); 811 FREE(fp, M_FTABLE); 812 m->m_len += (ip->ip_hl << 2); 813 m->m_data -= (ip->ip_hl << 2); 814 /* some debugging cruft by sklower, below, will go away soon */ 815 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 816 register int plen = 0; 817 for (t = m; t; t = t->m_next) 818 plen += t->m_len; 819 m->m_pkthdr.len = plen; 820 } 821 return (m); 822 823 dropfrag: 824 ipstat.ips_fragdropped++; 825 m_freem(m); 826 pool_put(&ipqent_pool, ipqe); 827 return (0); 828 } 829 830 /* 831 * Free a fragment reassembly header and all 832 * associated datagrams. 833 */ 834 void 835 ip_freef(fp) 836 struct ipq *fp; 837 { 838 register struct ipqent *q, *p; 839 840 IPQ_LOCK_CHECK(); 841 842 for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { 843 p = q->ipqe_q.le_next; 844 m_freem(q->ipqe_m); 845 LIST_REMOVE(q, ipqe_q); 846 pool_put(&ipqent_pool, q); 847 } 848 LIST_REMOVE(fp, ipq_q); 849 FREE(fp, M_FTABLE); 850 } 851 852 /* 853 * IP timer processing; 854 * if a timer expires on a reassembly 855 * queue, discard it. 856 */ 857 void 858 ip_slowtimo() 859 { 860 register struct ipq *fp, *nfp; 861 int s = splsoftnet(); 862 863 IPQ_LOCK(); 864 for (fp = ipq.lh_first; fp != NULL; fp = nfp) { 865 nfp = fp->ipq_q.le_next; 866 if (--fp->ipq_ttl == 0) { 867 ipstat.ips_fragtimeout++; 868 ip_freef(fp); 869 } 870 } 871 IPQ_UNLOCK(); 872 #ifdef GATEWAY 873 ipflow_slowtimo(); 874 #endif 875 splx(s); 876 } 877 878 /* 879 * Drain off all datagram fragments. 880 */ 881 void 882 ip_drain() 883 { 884 885 /* 886 * We may be called from a device's interrupt context. If 887 * the ipq is already busy, just bail out now. 888 */ 889 if (ipq_lock_try() == 0) 890 return; 891 892 while (ipq.lh_first != NULL) { 893 ipstat.ips_fragdropped++; 894 ip_freef(ipq.lh_first); 895 } 896 897 IPQ_UNLOCK(); 898 } 899 900 /* 901 * Do option processing on a datagram, 902 * possibly discarding it if bad options are encountered, 903 * or forwarding it if source-routed. 904 * Returns 1 if packet has been forwarded/freed, 905 * 0 if the packet should be processed further. 906 */ 907 int 908 ip_dooptions(m) 909 struct mbuf *m; 910 { 911 register struct ip *ip = mtod(m, struct ip *); 912 register u_char *cp; 913 register struct ip_timestamp *ipt; 914 register struct in_ifaddr *ia; 915 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 916 struct in_addr *sin, dst; 917 n_time ntime; 918 919 dst = ip->ip_dst; 920 cp = (u_char *)(ip + 1); 921 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 922 for (; cnt > 0; cnt -= optlen, cp += optlen) { 923 opt = cp[IPOPT_OPTVAL]; 924 if (opt == IPOPT_EOL) 925 break; 926 if (opt == IPOPT_NOP) 927 optlen = 1; 928 else { 929 optlen = cp[IPOPT_OLEN]; 930 if (optlen <= 0 || optlen > cnt) { 931 code = &cp[IPOPT_OLEN] - (u_char *)ip; 932 goto bad; 933 } 934 } 935 switch (opt) { 936 937 default: 938 break; 939 940 /* 941 * Source routing with record. 942 * Find interface with current destination address. 943 * If none on this machine then drop if strictly routed, 944 * or do nothing if loosely routed. 945 * Record interface address and bring up next address 946 * component. If strictly routed make sure next 947 * address is on directly accessible net. 948 */ 949 case IPOPT_LSRR: 950 case IPOPT_SSRR: 951 if (ip_allowsrcrt == 0) { 952 type = ICMP_UNREACH; 953 code = ICMP_UNREACH_NET_PROHIB; 954 goto bad; 955 } 956 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 957 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 958 goto bad; 959 } 960 ipaddr.sin_addr = ip->ip_dst; 961 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 962 if (ia == 0) { 963 if (opt == IPOPT_SSRR) { 964 type = ICMP_UNREACH; 965 code = ICMP_UNREACH_SRCFAIL; 966 goto bad; 967 } 968 /* 969 * Loose routing, and not at next destination 970 * yet; nothing to do except forward. 971 */ 972 break; 973 } 974 off--; /* 0 origin */ 975 if (off > optlen - sizeof(struct in_addr)) { 976 /* 977 * End of source route. Should be for us. 978 */ 979 save_rte(cp, ip->ip_src); 980 break; 981 } 982 /* 983 * locate outgoing interface 984 */ 985 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 986 sizeof(ipaddr.sin_addr)); 987 if (opt == IPOPT_SSRR) { 988 #define INA struct in_ifaddr * 989 #define SA struct sockaddr * 990 ia = (INA)ifa_ifwithladdr((SA)&ipaddr); 991 } else 992 ia = ip_rtaddr(ipaddr.sin_addr); 993 if (ia == 0) { 994 type = ICMP_UNREACH; 995 code = ICMP_UNREACH_SRCFAIL; 996 goto bad; 997 } 998 ip->ip_dst = ipaddr.sin_addr; 999 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1000 (caddr_t)(cp + off), sizeof(struct in_addr)); 1001 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1002 /* 1003 * Let ip_intr's mcast routing check handle mcast pkts 1004 */ 1005 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1006 break; 1007 1008 case IPOPT_RR: 1009 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1010 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1011 goto bad; 1012 } 1013 /* 1014 * If no space remains, ignore. 1015 */ 1016 off--; /* 0 origin */ 1017 if (off > optlen - sizeof(struct in_addr)) 1018 break; 1019 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 1020 sizeof(ipaddr.sin_addr)); 1021 /* 1022 * locate outgoing interface; if we're the destination, 1023 * use the incoming interface (should be same). 1024 */ 1025 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1026 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1027 type = ICMP_UNREACH; 1028 code = ICMP_UNREACH_HOST; 1029 goto bad; 1030 } 1031 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1032 (caddr_t)(cp + off), sizeof(struct in_addr)); 1033 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1034 break; 1035 1036 case IPOPT_TS: 1037 code = cp - (u_char *)ip; 1038 ipt = (struct ip_timestamp *)cp; 1039 if (ipt->ipt_len < 5) 1040 goto bad; 1041 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1042 if (++ipt->ipt_oflw == 0) 1043 goto bad; 1044 break; 1045 } 1046 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1047 switch (ipt->ipt_flg) { 1048 1049 case IPOPT_TS_TSONLY: 1050 break; 1051 1052 case IPOPT_TS_TSANDADDR: 1053 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1054 sizeof(struct in_addr) > ipt->ipt_len) 1055 goto bad; 1056 ipaddr.sin_addr = dst; 1057 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1058 m->m_pkthdr.rcvif); 1059 if (ia == 0) 1060 continue; 1061 bcopy((caddr_t)&ia->ia_addr.sin_addr, 1062 (caddr_t)sin, sizeof(struct in_addr)); 1063 ipt->ipt_ptr += sizeof(struct in_addr); 1064 break; 1065 1066 case IPOPT_TS_PRESPEC: 1067 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1068 sizeof(struct in_addr) > ipt->ipt_len) 1069 goto bad; 1070 bcopy((caddr_t)sin, (caddr_t)&ipaddr.sin_addr, 1071 sizeof(struct in_addr)); 1072 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1073 continue; 1074 ipt->ipt_ptr += sizeof(struct in_addr); 1075 break; 1076 1077 default: 1078 goto bad; 1079 } 1080 ntime = iptime(); 1081 bcopy((caddr_t)&ntime, (caddr_t)cp + ipt->ipt_ptr - 1, 1082 sizeof(n_time)); 1083 ipt->ipt_ptr += sizeof(n_time); 1084 } 1085 } 1086 if (forward) { 1087 if (ip_forwsrcrt == 0) { 1088 type = ICMP_UNREACH; 1089 code = ICMP_UNREACH_SRCFAIL; 1090 goto bad; 1091 } 1092 ip_forward(m, 1); 1093 return (1); 1094 } 1095 return (0); 1096 bad: 1097 icmp_error(m, type, code, 0, 0); 1098 ipstat.ips_badoptions++; 1099 return (1); 1100 } 1101 1102 /* 1103 * Given address of next destination (final or next hop), 1104 * return internet address info of interface to be used to get there. 1105 */ 1106 struct in_ifaddr * 1107 ip_rtaddr(dst) 1108 struct in_addr dst; 1109 { 1110 register struct sockaddr_in *sin; 1111 1112 sin = satosin(&ipforward_rt.ro_dst); 1113 1114 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { 1115 if (ipforward_rt.ro_rt) { 1116 RTFREE(ipforward_rt.ro_rt); 1117 ipforward_rt.ro_rt = 0; 1118 } 1119 sin->sin_family = AF_INET; 1120 sin->sin_len = sizeof(*sin); 1121 sin->sin_addr = dst; 1122 1123 rtalloc(&ipforward_rt); 1124 } 1125 if (ipforward_rt.ro_rt == 0) 1126 return ((struct in_ifaddr *)0); 1127 return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); 1128 } 1129 1130 /* 1131 * Save incoming source route for use in replies, 1132 * to be picked up later by ip_srcroute if the receiver is interested. 1133 */ 1134 void 1135 save_rte(option, dst) 1136 u_char *option; 1137 struct in_addr dst; 1138 { 1139 unsigned olen; 1140 1141 olen = option[IPOPT_OLEN]; 1142 #ifdef DIAGNOSTIC 1143 if (ipprintfs) 1144 printf("save_rte: olen %d\n", olen); 1145 #endif /* 0 */ 1146 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1147 return; 1148 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 1149 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1150 ip_srcrt.dst = dst; 1151 } 1152 1153 /* 1154 * Retrieve incoming source route for use in replies, 1155 * in the same form used by setsockopt. 1156 * The first hop is placed before the options, will be removed later. 1157 */ 1158 struct mbuf * 1159 ip_srcroute() 1160 { 1161 register struct in_addr *p, *q; 1162 register struct mbuf *m; 1163 1164 if (ip_nhops == 0) 1165 return ((struct mbuf *)0); 1166 m = m_get(M_DONTWAIT, MT_SOOPTS); 1167 if (m == 0) 1168 return ((struct mbuf *)0); 1169 1170 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1171 1172 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1173 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1174 OPTSIZ; 1175 #ifdef DIAGNOSTIC 1176 if (ipprintfs) 1177 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1178 #endif 1179 1180 /* 1181 * First save first hop for return route 1182 */ 1183 p = &ip_srcrt.route[ip_nhops - 1]; 1184 *(mtod(m, struct in_addr *)) = *p--; 1185 #ifdef DIAGNOSTIC 1186 if (ipprintfs) 1187 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1188 #endif 1189 1190 /* 1191 * Copy option fields and padding (nop) to mbuf. 1192 */ 1193 ip_srcrt.nop = IPOPT_NOP; 1194 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1195 bcopy((caddr_t)&ip_srcrt.nop, 1196 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 1197 q = (struct in_addr *)(mtod(m, caddr_t) + 1198 sizeof(struct in_addr) + OPTSIZ); 1199 #undef OPTSIZ 1200 /* 1201 * Record return path as an IP source route, 1202 * reversing the path (pointers are now aligned). 1203 */ 1204 while (p >= ip_srcrt.route) { 1205 #ifdef DIAGNOSTIC 1206 if (ipprintfs) 1207 printf(" %x", ntohl(q->s_addr)); 1208 #endif 1209 *q++ = *p--; 1210 } 1211 /* 1212 * Last hop goes to final destination. 1213 */ 1214 *q = ip_srcrt.dst; 1215 #ifdef DIAGNOSTIC 1216 if (ipprintfs) 1217 printf(" %x\n", ntohl(q->s_addr)); 1218 #endif 1219 return (m); 1220 } 1221 1222 /* 1223 * Strip out IP options, at higher 1224 * level protocol in the kernel. 1225 * Second argument is buffer to which options 1226 * will be moved, and return value is their length. 1227 * XXX should be deleted; last arg currently ignored. 1228 */ 1229 void 1230 ip_stripoptions(m, mopt) 1231 register struct mbuf *m; 1232 struct mbuf *mopt; 1233 { 1234 register int i; 1235 struct ip *ip = mtod(m, struct ip *); 1236 register caddr_t opts; 1237 int olen; 1238 1239 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1240 opts = (caddr_t)(ip + 1); 1241 i = m->m_len - (sizeof (struct ip) + olen); 1242 bcopy(opts + olen, opts, (unsigned)i); 1243 m->m_len -= olen; 1244 if (m->m_flags & M_PKTHDR) 1245 m->m_pkthdr.len -= olen; 1246 ip->ip_len -= olen; 1247 ip->ip_hl = sizeof (struct ip) >> 2; 1248 } 1249 1250 int inetctlerrmap[PRC_NCMDS] = { 1251 0, 0, 0, 0, 1252 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1253 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1254 EMSGSIZE, EHOSTUNREACH, 0, 0, 1255 0, 0, 0, 0, 1256 ENOPROTOOPT 1257 }; 1258 1259 /* 1260 * Forward a packet. If some error occurs return the sender 1261 * an icmp packet. Note we can't always generate a meaningful 1262 * icmp message because icmp doesn't have a large enough repertoire 1263 * of codes and types. 1264 * 1265 * If not forwarding, just drop the packet. This could be confusing 1266 * if ipforwarding was zero but some routing protocol was advancing 1267 * us as a gateway to somewhere. However, we must let the routing 1268 * protocol deal with that. 1269 * 1270 * The srcrt parameter indicates whether the packet is being forwarded 1271 * via a source route. 1272 */ 1273 void 1274 ip_forward(m, srcrt) 1275 struct mbuf *m; 1276 int srcrt; 1277 { 1278 register struct ip *ip = mtod(m, struct ip *); 1279 register struct sockaddr_in *sin; 1280 register struct rtentry *rt; 1281 int error, type = 0, code = 0; 1282 struct mbuf *mcopy; 1283 n_long dest; 1284 struct ifnet *destifp; 1285 #ifdef IPSEC 1286 struct ifnet dummyifp; 1287 #endif 1288 1289 dest = 0; 1290 #ifdef DIAGNOSTIC 1291 if (ipprintfs) 1292 printf("forward: src %2.2x dst %2.2x ttl %x\n", 1293 ntohl(ip->ip_src.s_addr), 1294 ntohl(ip->ip_dst.s_addr), ip->ip_ttl); 1295 #endif 1296 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) { 1297 ipstat.ips_cantforward++; 1298 m_freem(m); 1299 return; 1300 } 1301 if (ip->ip_ttl <= IPTTLDEC) { 1302 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1303 return; 1304 } 1305 ip->ip_ttl -= IPTTLDEC; 1306 1307 sin = satosin(&ipforward_rt.ro_dst); 1308 if ((rt = ipforward_rt.ro_rt) == 0 || 1309 !in_hosteq(ip->ip_dst, sin->sin_addr)) { 1310 if (ipforward_rt.ro_rt) { 1311 RTFREE(ipforward_rt.ro_rt); 1312 ipforward_rt.ro_rt = 0; 1313 } 1314 sin->sin_family = AF_INET; 1315 sin->sin_len = sizeof(struct sockaddr_in); 1316 sin->sin_addr = ip->ip_dst; 1317 1318 rtalloc(&ipforward_rt); 1319 if (ipforward_rt.ro_rt == 0) { 1320 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1321 return; 1322 } 1323 rt = ipforward_rt.ro_rt; 1324 } 1325 1326 /* 1327 * Save at most 68 bytes of the packet in case 1328 * we need to generate an ICMP message to the src. 1329 */ 1330 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 68)); 1331 1332 /* 1333 * If forwarding packet using same interface that it came in on, 1334 * perhaps should send a redirect to sender to shortcut a hop. 1335 * Only send redirect if source is sending directly to us, 1336 * and if packet was not source routed (or has any options). 1337 * Also, don't send redirect if forwarding using a default route 1338 * or a route modified by a redirect. 1339 */ 1340 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1341 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1342 !in_nullhost(satosin(rt_key(rt))->sin_addr) && 1343 ipsendredirects && !srcrt) { 1344 if (rt->rt_ifa && 1345 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1346 ifatoia(rt->rt_ifa)->ia_subnet) { 1347 if (rt->rt_flags & RTF_GATEWAY) 1348 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1349 else 1350 dest = ip->ip_dst.s_addr; 1351 /* 1352 * Router requirements says to only send host 1353 * redirects. 1354 */ 1355 type = ICMP_REDIRECT; 1356 code = ICMP_REDIRECT_HOST; 1357 #ifdef DIAGNOSTIC 1358 if (ipprintfs) 1359 printf("redirect (%d) to %x\n", code, 1360 (u_int32_t)dest); 1361 #endif 1362 } 1363 } 1364 1365 #ifdef IPSEC 1366 m->m_pkthdr.rcvif = NULL; 1367 #endif /*IPSEC*/ 1368 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1369 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0); 1370 if (error) 1371 ipstat.ips_cantforward++; 1372 else { 1373 ipstat.ips_forward++; 1374 if (type) 1375 ipstat.ips_redirectsent++; 1376 else { 1377 if (mcopy) { 1378 #ifdef GATEWAY 1379 if (mcopy->m_flags & M_CANFASTFWD) 1380 ipflow_create(&ipforward_rt, mcopy); 1381 #endif 1382 m_freem(mcopy); 1383 } 1384 return; 1385 } 1386 } 1387 if (mcopy == NULL) 1388 return; 1389 destifp = NULL; 1390 1391 switch (error) { 1392 1393 case 0: /* forwarded, but need redirect */ 1394 /* type, code set above */ 1395 break; 1396 1397 case ENETUNREACH: /* shouldn't happen, checked above */ 1398 case EHOSTUNREACH: 1399 case ENETDOWN: 1400 case EHOSTDOWN: 1401 default: 1402 type = ICMP_UNREACH; 1403 code = ICMP_UNREACH_HOST; 1404 break; 1405 1406 case EMSGSIZE: 1407 type = ICMP_UNREACH; 1408 code = ICMP_UNREACH_NEEDFRAG; 1409 #ifndef IPSEC 1410 if (ipforward_rt.ro_rt) 1411 destifp = ipforward_rt.ro_rt->rt_ifp; 1412 #else 1413 /* 1414 * If the packet is routed over IPsec tunnel, tell the 1415 * originator the tunnel MTU. 1416 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1417 * XXX quickhack!!! 1418 */ 1419 if (ipforward_rt.ro_rt) { 1420 struct secpolicy *sp; 1421 int ipsecerror; 1422 int ipsechdr; 1423 struct route *ro; 1424 1425 sp = ipsec4_getpolicybyaddr(mcopy, 1426 IP_FORWARDING, 1427 &ipsecerror); 1428 1429 if (sp == NULL) 1430 destifp = ipforward_rt.ro_rt->rt_ifp; 1431 else { 1432 /* count IPsec header size */ 1433 ipsechdr = ipsec4_hdrsiz(mcopy, NULL); 1434 1435 /* 1436 * find the correct route for outer IPv4 1437 * header, compute tunnel MTU. 1438 * 1439 * XXX BUG ALERT 1440 * The "dummyifp" code relies upon the fact 1441 * that icmp_error() touches only ifp->if_mtu. 1442 */ 1443 /*XXX*/ 1444 destifp = NULL; 1445 if (sp->req != NULL 1446 && sp->req->sa != NULL) { 1447 ro = &sp->req->sa->saidx->sa_route; 1448 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1449 dummyifp.if_mtu = 1450 ro->ro_rt->rt_ifp->if_mtu; 1451 dummyifp.if_mtu -= ipsechdr; 1452 destifp = &dummyifp; 1453 } 1454 } 1455 1456 key_freesp(sp); 1457 } 1458 } 1459 #endif /*IPSEC*/ 1460 ipstat.ips_cantfrag++; 1461 break; 1462 1463 case ENOBUFS: 1464 type = ICMP_SOURCEQUENCH; 1465 code = 0; 1466 break; 1467 } 1468 icmp_error(mcopy, type, code, dest, destifp); 1469 } 1470 1471 void 1472 ip_savecontrol(inp, mp, ip, m) 1473 register struct inpcb *inp; 1474 register struct mbuf **mp; 1475 register struct ip *ip; 1476 register struct mbuf *m; 1477 { 1478 1479 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1480 struct timeval tv; 1481 1482 microtime(&tv); 1483 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1484 SCM_TIMESTAMP, SOL_SOCKET); 1485 if (*mp) 1486 mp = &(*mp)->m_next; 1487 } 1488 if (inp->inp_flags & INP_RECVDSTADDR) { 1489 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1490 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1491 if (*mp) 1492 mp = &(*mp)->m_next; 1493 } 1494 #ifdef notyet 1495 /* 1496 * XXX 1497 * Moving these out of udp_input() made them even more broken 1498 * than they already were. 1499 * - fenner (at) parc.xerox.com 1500 */ 1501 /* options were tossed already */ 1502 if (inp->inp_flags & INP_RECVOPTS) { 1503 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1504 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1505 if (*mp) 1506 mp = &(*mp)->m_next; 1507 } 1508 /* ip_srcroute doesn't do what we want here, need to fix */ 1509 if (inp->inp_flags & INP_RECVRETOPTS) { 1510 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1511 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1512 if (*mp) 1513 mp = &(*mp)->m_next; 1514 } 1515 #endif 1516 if (inp->inp_flags & INP_RECVIF) { 1517 struct sockaddr_dl sdl; 1518 1519 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1520 sdl.sdl_family = AF_LINK; 1521 sdl.sdl_index = m->m_pkthdr.rcvif ? 1522 m->m_pkthdr.rcvif->if_index : 0; 1523 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1524 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1525 IP_RECVIF, IPPROTO_IP); 1526 if (*mp) 1527 mp = &(*mp)->m_next; 1528 } 1529 } 1530 1531 int 1532 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 1533 int *name; 1534 u_int namelen; 1535 void *oldp; 1536 size_t *oldlenp; 1537 void *newp; 1538 size_t newlen; 1539 { 1540 extern int subnetsarelocal; 1541 1542 int error, old; 1543 1544 /* All sysctl names at this level are terminal. */ 1545 if (namelen != 1) 1546 return (ENOTDIR); 1547 1548 switch (name[0]) { 1549 case IPCTL_FORWARDING: 1550 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); 1551 case IPCTL_SENDREDIRECTS: 1552 return (sysctl_int(oldp, oldlenp, newp, newlen, 1553 &ipsendredirects)); 1554 case IPCTL_DEFTTL: 1555 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); 1556 #ifdef notyet 1557 case IPCTL_DEFMTU: 1558 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1559 #endif 1560 case IPCTL_FORWSRCRT: 1561 /* Don't allow this to change in a secure environment. */ 1562 if (securelevel > 0) 1563 return (sysctl_rdint(oldp, oldlenp, newp, 1564 ip_forwsrcrt)); 1565 else 1566 return (sysctl_int(oldp, oldlenp, newp, newlen, 1567 &ip_forwsrcrt)); 1568 case IPCTL_DIRECTEDBCAST: 1569 return (sysctl_int(oldp, oldlenp, newp, newlen, 1570 &ip_directedbcast)); 1571 case IPCTL_ALLOWSRCRT: 1572 return (sysctl_int(oldp, oldlenp, newp, newlen, 1573 &ip_allowsrcrt)); 1574 case IPCTL_SUBNETSARELOCAL: 1575 return (sysctl_int(oldp, oldlenp, newp, newlen, 1576 &subnetsarelocal)); 1577 case IPCTL_MTUDISC: 1578 error = sysctl_int(oldp, oldlenp, newp, newlen, 1579 &ip_mtudisc); 1580 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1581 ip_mtudisc_timeout_q = 1582 rt_timer_queue_create(ip_mtudisc_timeout); 1583 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1584 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE); 1585 ip_mtudisc_timeout_q = NULL; 1586 } 1587 return error; 1588 case IPCTL_ANONPORTMIN: 1589 old = anonportmin; 1590 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); 1591 if (anonportmin >= anonportmax || anonportmin > 65535 1592 #ifndef IPNOPRIVPORTS 1593 || anonportmin < IPPORT_RESERVED 1594 #endif 1595 ) { 1596 anonportmin = old; 1597 return (EINVAL); 1598 } 1599 return (error); 1600 case IPCTL_ANONPORTMAX: 1601 old = anonportmax; 1602 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); 1603 if (anonportmin >= anonportmax || anonportmax > 65535 1604 #ifndef IPNOPRIVPORTS 1605 || anonportmax < IPPORT_RESERVED 1606 #endif 1607 ) { 1608 anonportmax = old; 1609 return (EINVAL); 1610 } 1611 return (error); 1612 case IPCTL_MTUDISCTIMEOUT: 1613 error = sysctl_int(oldp, oldlenp, newp, newlen, 1614 &ip_mtudisc_timeout); 1615 if (ip_mtudisc_timeout_q != NULL) 1616 rt_timer_queue_change(ip_mtudisc_timeout_q, 1617 ip_mtudisc_timeout); 1618 return (error); 1619 #ifdef GATEWAY 1620 case IPCTL_MAXFLOWS: 1621 { 1622 int s; 1623 1624 error = sysctl_int(oldp, oldlenp, newp, newlen, 1625 &ip_maxflows); 1626 s = splsoftnet(); 1627 ipflow_reap(0); 1628 splx(s); 1629 return (error); 1630 } 1631 #endif 1632 #if NGIF > 0 1633 case IPCTL_GIF_TTL: 1634 return(sysctl_int(oldp, oldlenp, newp, newlen, 1635 &gif_ttl)); 1636 #endif 1637 default: 1638 return (EOPNOTSUPP); 1639 } 1640 /* NOTREACHED */ 1641 } 1642
Indexes created Wed Oct 15 16:09:53 GMT 2025