ip_reass.c revision 96432
1/* 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 34 * $FreeBSD: head/sys/netinet/ip_input.c 96432 2002-05-12 00:22:38Z dd $ 35 */ 36 37#define _IP_VHL 38 39#include "opt_bootp.h" 40#include "opt_ipfw.h" 41#include "opt_ipdn.h" 42#include "opt_ipdivert.h" 43#include "opt_ipfilter.h" 44#include "opt_ipstealth.h" 45#include "opt_ipsec.h" 46#include "opt_pfil_hooks.h" 47#include "opt_random_ip_id.h" 48 49#include <sys/param.h> 50#include <sys/systm.h> 51#include <sys/mbuf.h> 52#include <sys/malloc.h> 53#include <sys/domain.h> 54#include <sys/protosw.h> 55#include <sys/socket.h> 56#include <sys/time.h> 57#include <sys/kernel.h> 58#include <sys/syslog.h> 59#include <sys/sysctl.h> 60 61#include <net/pfil.h> 62#include <net/if.h> 63#include <net/if_types.h> 64#include <net/if_var.h> 65#include <net/if_dl.h> 66#include <net/route.h> 67#include <net/netisr.h> 68#include <net/intrq.h> 69 70#include <netinet/in.h> 71#include <netinet/in_systm.h> 72#include <netinet/in_var.h> 73#include <netinet/ip.h> 74#include <netinet/in_pcb.h> 75#include <netinet/ip_var.h> 76#include <netinet/ip_icmp.h> 77#include <machine/in_cksum.h> 78 79#include <sys/socketvar.h> 80 81#include <netinet/ip_fw.h> 82#include <netinet/ip_dummynet.h> 83 84#ifdef IPSEC 85#include <netinet6/ipsec.h> 86#include <netkey/key.h> 87#endif 88 89int rsvp_on = 0; 90static int ip_rsvp_on; 91struct socket *ip_rsvpd; 92 93int ipforwarding = 0; 94SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 95 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 96 97static int ipsendredirects = 1; /* XXX */ 98SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 99 &ipsendredirects, 0, "Enable sending IP redirects"); 100 101int ip_defttl = IPDEFTTL; 102SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 103 &ip_defttl, 0, "Maximum TTL on IP packets"); 104 105static int ip_dosourceroute = 0; 106SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 107 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 108 109static int ip_acceptsourceroute = 0; 110SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 111 CTLFLAG_RW, &ip_acceptsourceroute, 0, 112 "Enable accepting source routed IP packets"); 113 114static int ip_keepfaith = 0; 115SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 116 &ip_keepfaith, 0, 117 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 118 119static int ip_nfragpackets = 0; 120static int ip_maxfragpackets; /* initialized in ip_init() */ 121SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 122 &ip_maxfragpackets, 0, 123 "Maximum number of IPv4 fragment reassembly queue entries"); 124 125/* 126 * XXX - Setting ip_checkinterface mostly implements the receive side of 127 * the Strong ES model described in RFC 1122, but since the routing table 128 * and transmit implementation do not implement the Strong ES model, 129 * setting this to 1 results in an odd hybrid. 130 * 131 * XXX - ip_checkinterface currently must be disabled if you use ipnat 132 * to translate the destination address to another local interface. 133 * 134 * XXX - ip_checkinterface must be disabled if you add IP aliases 135 * to the loopback interface instead of the interface where the 136 * packets for those addresses are received. 137 */ 138static int ip_checkinterface = 1; 139SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 140 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 141 142#ifdef DIAGNOSTIC 143static int ipprintfs = 0; 144#endif 145 146static int ipqmaxlen = IFQ_MAXLEN; 147 148extern struct domain inetdomain; 149extern struct protosw inetsw[]; 150u_char ip_protox[IPPROTO_MAX]; 151struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 152struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 153u_long in_ifaddrhmask; /* mask for hash table */ 154 155SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 156 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 157SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 158 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 159 160struct ipstat ipstat; 161SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 162 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 163 164/* Packet reassembly stuff */ 165#define IPREASS_NHASH_LOG2 6 166#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 167#define IPREASS_HMASK (IPREASS_NHASH - 1) 168#define IPREASS_HASH(x,y) \ 169 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 170 171static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 172static int nipq = 0; /* total # of reass queues */ 173static int maxnipq; 174 175#ifdef IPCTL_DEFMTU 176SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 177 &ip_mtu, 0, "Default MTU"); 178#endif 179 180#ifdef IPSTEALTH 181static int ipstealth = 0; 182SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 183 &ipstealth, 0, ""); 184#endif 185 186 187/* Firewall hooks */ 188ip_fw_chk_t *ip_fw_chk_ptr; 189int fw_enable = 1 ; 190 191/* Dummynet hooks */ 192ip_dn_io_t *ip_dn_io_ptr; 193 194 195/* 196 * We need to save the IP options in case a protocol wants to respond 197 * to an incoming packet over the same route if the packet got here 198 * using IP source routing. This allows connection establishment and 199 * maintenance when the remote end is on a network that is not known 200 * to us. 201 */ 202static int ip_nhops = 0; 203static struct ip_srcrt { 204 struct in_addr dst; /* final destination */ 205 char nop; /* one NOP to align */ 206 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 207 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 208} ip_srcrt; 209 210struct sockaddr_in *ip_fw_fwd_addr; 211 212static void save_rte(u_char *, struct in_addr); 213static int ip_dooptions(struct mbuf *, int); 214static void ip_forward(struct mbuf *, int); 215static void ip_freef(struct ipqhead *, struct ipq *); 216#ifdef IPDIVERT 217static struct mbuf *ip_reass(struct mbuf *, struct ipqhead *, struct ipq *, u_int32_t *, u_int16_t *); 218#else 219static struct mbuf *ip_reass(struct mbuf *, struct ipqhead *, struct ipq *); 220#endif 221static void ipintr(void); 222 223/* 224 * IP initialization: fill in IP protocol switch table. 225 * All protocols not implemented in kernel go to raw IP protocol handler. 226 */ 227void 228ip_init() 229{ 230 register struct protosw *pr; 231 register int i; 232 233 TAILQ_INIT(&in_ifaddrhead); 234 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 235 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 236 if (pr == 0) 237 panic("ip_init"); 238 for (i = 0; i < IPPROTO_MAX; i++) 239 ip_protox[i] = pr - inetsw; 240 for (pr = inetdomain.dom_protosw; 241 pr < inetdomain.dom_protoswNPROTOSW; pr++) 242 if (pr->pr_domain->dom_family == PF_INET && 243 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 244 ip_protox[pr->pr_protocol] = pr - inetsw; 245 246 for (i = 0; i < IPREASS_NHASH; i++) 247 TAILQ_INIT(&ipq[i]); 248 249 maxnipq = nmbclusters / 4; 250 ip_maxfragpackets = nmbclusters / 4; 251 252#ifndef RANDOM_IP_ID 253 ip_id = time_second & 0xffff; 254#endif 255 ipintrq.ifq_maxlen = ipqmaxlen; 256 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 257 ipintrq_present = 1; 258 259 register_netisr(NETISR_IP, ipintr); 260} 261 262static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 263struct route ipforward_rt; 264 265/* 266 * Ip input routine. Checksum and byte swap header. If fragmented 267 * try to reassemble. Process options. Pass to next level. 268 */ 269void 270ip_input(struct mbuf *m) 271{ 272 struct ip *ip; 273 struct ipq *fp; 274 struct in_ifaddr *ia = NULL; 275 struct ifaddr *ifa; 276 int i, hlen, checkif; 277 u_short sum; 278 u_int16_t divert_cookie; /* firewall cookie */ 279 struct in_addr pkt_dst; 280#ifdef IPDIVERT 281 u_int32_t divert_info = 0; /* packet divert/tee info */ 282#endif 283 struct ip_fw *rule = NULL; 284#ifdef PFIL_HOOKS 285 struct packet_filter_hook *pfh; 286 struct mbuf *m0; 287 int rv; 288#endif /* PFIL_HOOKS */ 289 290#ifdef IPDIVERT 291 /* Get and reset firewall cookie */ 292 divert_cookie = ip_divert_cookie; 293 ip_divert_cookie = 0; 294#else 295 divert_cookie = 0; 296#endif 297 298 /* 299 * dummynet packet are prepended a vestigial mbuf with 300 * m_type = MT_DUMMYNET and m_data pointing to the matching 301 * rule. 302 */ 303 if (m->m_type == MT_DUMMYNET) { 304 rule = (struct ip_fw *)(m->m_data) ; 305 m = m->m_next ; 306 ip = mtod(m, struct ip *); 307 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 308 goto iphack ; 309 } else 310 rule = NULL ; 311 312#ifdef DIAGNOSTIC 313 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 314 panic("ip_input no HDR"); 315#endif 316 ipstat.ips_total++; 317 318 if (m->m_pkthdr.len < sizeof(struct ip)) 319 goto tooshort; 320 321 if (m->m_len < sizeof (struct ip) && 322 (m = m_pullup(m, sizeof (struct ip))) == 0) { 323 ipstat.ips_toosmall++; 324 return; 325 } 326 ip = mtod(m, struct ip *); 327 328 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 329 ipstat.ips_badvers++; 330 goto bad; 331 } 332 333 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 334 if (hlen < sizeof(struct ip)) { /* minimum header length */ 335 ipstat.ips_badhlen++; 336 goto bad; 337 } 338 if (hlen > m->m_len) { 339 if ((m = m_pullup(m, hlen)) == 0) { 340 ipstat.ips_badhlen++; 341 return; 342 } 343 ip = mtod(m, struct ip *); 344 } 345 346 /* 127/8 must not appear on wire - RFC1122 */ 347 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 348 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 349 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 350 ipstat.ips_badaddr++; 351 goto bad; 352 } 353 } 354 355 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 356 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 357 } else { 358 if (hlen == sizeof(struct ip)) { 359 sum = in_cksum_hdr(ip); 360 } else { 361 sum = in_cksum(m, hlen); 362 } 363 } 364 if (sum) { 365 ipstat.ips_badsum++; 366 goto bad; 367 } 368 369 /* 370 * Convert fields to host representation. 371 */ 372 ip->ip_len = ntohs(ip->ip_len); 373 if (ip->ip_len < hlen) { 374 ipstat.ips_badlen++; 375 goto bad; 376 } 377 ip->ip_off = ntohs(ip->ip_off); 378 379 /* 380 * Check that the amount of data in the buffers 381 * is as at least much as the IP header would have us expect. 382 * Trim mbufs if longer than we expect. 383 * Drop packet if shorter than we expect. 384 */ 385 if (m->m_pkthdr.len < ip->ip_len) { 386tooshort: 387 ipstat.ips_tooshort++; 388 goto bad; 389 } 390 if (m->m_pkthdr.len > ip->ip_len) { 391 if (m->m_len == m->m_pkthdr.len) { 392 m->m_len = ip->ip_len; 393 m->m_pkthdr.len = ip->ip_len; 394 } else 395 m_adj(m, ip->ip_len - m->m_pkthdr.len); 396 } 397 398#ifdef IPSEC 399 if (ipsec_gethist(m, NULL)) 400 goto pass; 401#endif 402 403 /* 404 * IpHack's section. 405 * Right now when no processing on packet has done 406 * and it is still fresh out of network we do our black 407 * deals with it. 408 * - Firewall: deny/allow/divert 409 * - Xlate: translate packet's addr/port (NAT). 410 * - Pipe: pass pkt through dummynet. 411 * - Wrap: fake packet's addr/port <unimpl.> 412 * - Encapsulate: put it in another IP and send out. <unimp.> 413 */ 414 415iphack: 416 417#ifdef PFIL_HOOKS 418 /* 419 * Run through list of hooks for input packets. If there are any 420 * filters which require that additional packets in the flow are 421 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 422 * Note that filters must _never_ set this flag, as another filter 423 * in the list may have previously cleared it. 424 */ 425 m0 = m; 426 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 427 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 428 if (pfh->pfil_func) { 429 rv = pfh->pfil_func(ip, hlen, 430 m->m_pkthdr.rcvif, 0, &m0); 431 if (rv) 432 return; 433 m = m0; 434 if (m == NULL) 435 return; 436 ip = mtod(m, struct ip *); 437 } 438#endif /* PFIL_HOOKS */ 439 440 if (fw_enable && IPFW_LOADED) { 441#ifdef IPFIREWALL_FORWARD 442 /* 443 * If we've been forwarded from the output side, then 444 * skip the firewall a second time 445 */ 446 if (ip_fw_fwd_addr) 447 goto ours; 448#endif /* IPFIREWALL_FORWARD */ 449 /* 450 * See the comment in ip_output for the return values 451 * produced by the firewall. 452 */ 453 i = ip_fw_chk_ptr(&m, NULL /* oif */, &divert_cookie, 454 &rule, &ip_fw_fwd_addr); 455 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */ 456 if (m) 457 m_freem(m); 458 return; 459 } 460 ip = mtod(m, struct ip *); /* just in case m changed */ 461 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 462 goto pass; 463 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) { 464 /* Send packet to the appropriate pipe */ 465 ip_dn_io_ptr(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 466 0); 467 return; 468 } 469#ifdef IPDIVERT 470 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 471 /* Divert or tee packet */ 472 divert_info = i; 473 goto ours; 474 } 475#endif 476#ifdef IPFIREWALL_FORWARD 477 if (i == 0 && ip_fw_fwd_addr != NULL) 478 goto pass; 479#endif 480 /* 481 * if we get here, the packet must be dropped 482 */ 483 m_freem(m); 484 return; 485 } 486pass: 487 488 /* 489 * Process options and, if not destined for us, 490 * ship it on. ip_dooptions returns 1 when an 491 * error was detected (causing an icmp message 492 * to be sent and the original packet to be freed). 493 */ 494 ip_nhops = 0; /* for source routed packets */ 495 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) { 496#ifdef IPFIREWALL_FORWARD 497 ip_fw_fwd_addr = NULL; 498#endif 499 return; 500 } 501 502 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 503 * matter if it is destined to another node, or whether it is 504 * a multicast one, RSVP wants it! and prevents it from being forwarded 505 * anywhere else. Also checks if the rsvp daemon is running before 506 * grabbing the packet. 507 */ 508 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 509 goto ours; 510 511 /* 512 * Check our list of addresses, to see if the packet is for us. 513 * If we don't have any addresses, assume any unicast packet 514 * we receive might be for us (and let the upper layers deal 515 * with it). 516 */ 517 if (TAILQ_EMPTY(&in_ifaddrhead) && 518 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 519 goto ours; 520 521 /* 522 * Cache the destination address of the packet; this may be 523 * changed by use of 'ipfw fwd'. 524 */ 525 pkt_dst = ip_fw_fwd_addr == NULL ? 526 ip->ip_dst : ip_fw_fwd_addr->sin_addr; 527 528 /* 529 * Enable a consistency check between the destination address 530 * and the arrival interface for a unicast packet (the RFC 1122 531 * strong ES model) if IP forwarding is disabled and the packet 532 * is not locally generated and the packet is not subject to 533 * 'ipfw fwd'. 534 * 535 * XXX - Checking also should be disabled if the destination 536 * address is ipnat'ed to a different interface. 537 * 538 * XXX - Checking is incompatible with IP aliases added 539 * to the loopback interface instead of the interface where 540 * the packets are received. 541 */ 542 checkif = ip_checkinterface && (ipforwarding == 0) && 543 m->m_pkthdr.rcvif != NULL && 544 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 545 (ip_fw_fwd_addr == NULL); 546 547 /* 548 * Check for exact addresses in the hash bucket. 549 */ 550 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 551 /* 552 * If the address matches, verify that the packet 553 * arrived via the correct interface if checking is 554 * enabled. 555 */ 556 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 557 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 558 goto ours; 559 } 560 /* 561 * Check for broadcast addresses. 562 * 563 * Only accept broadcast packets that arrive via the matching 564 * interface. Reception of forwarded directed broadcasts would 565 * be handled via ip_forward() and ether_output() with the loopback 566 * into the stack for SIMPLEX interfaces handled by ether_output(). 567 */ 568 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 569 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 570 if (ifa->ifa_addr->sa_family != AF_INET) 571 continue; 572 ia = ifatoia(ifa); 573 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 574 pkt_dst.s_addr) 575 goto ours; 576 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 577 goto ours; 578#ifdef BOOTP_COMPAT 579 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 580 goto ours; 581#endif 582 } 583 } 584 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 585 struct in_multi *inm; 586 if (ip_mrouter) { 587 /* 588 * If we are acting as a multicast router, all 589 * incoming multicast packets are passed to the 590 * kernel-level multicast forwarding function. 591 * The packet is returned (relatively) intact; if 592 * ip_mforward() returns a non-zero value, the packet 593 * must be discarded, else it may be accepted below. 594 */ 595 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 596 ipstat.ips_cantforward++; 597 m_freem(m); 598 return; 599 } 600 601 /* 602 * The process-level routing daemon needs to receive 603 * all multicast IGMP packets, whether or not this 604 * host belongs to their destination groups. 605 */ 606 if (ip->ip_p == IPPROTO_IGMP) 607 goto ours; 608 ipstat.ips_forward++; 609 } 610 /* 611 * See if we belong to the destination multicast group on the 612 * arrival interface. 613 */ 614 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 615 if (inm == NULL) { 616 ipstat.ips_notmember++; 617 m_freem(m); 618 return; 619 } 620 goto ours; 621 } 622 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 623 goto ours; 624 if (ip->ip_dst.s_addr == INADDR_ANY) 625 goto ours; 626 627 /* 628 * FAITH(Firewall Aided Internet Translator) 629 */ 630 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 631 if (ip_keepfaith) { 632 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 633 goto ours; 634 } 635 m_freem(m); 636 return; 637 } 638 639 /* 640 * Not for us; forward if possible and desirable. 641 */ 642 if (ipforwarding == 0) { 643 ipstat.ips_cantforward++; 644 m_freem(m); 645 } else { 646#ifdef IPSEC 647 /* 648 * Enforce inbound IPsec SPD. 649 */ 650 if (ipsec4_in_reject(m, NULL)) { 651 ipsecstat.in_polvio++; 652 goto bad; 653 } 654#endif /* IPSEC */ 655 ip_forward(m, 0); 656 } 657#ifdef IPFIREWALL_FORWARD 658 ip_fw_fwd_addr = NULL; 659#endif 660 return; 661 662ours: 663#ifdef IPSTEALTH 664 /* 665 * IPSTEALTH: Process non-routing options only 666 * if the packet is destined for us. 667 */ 668 if (ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) { 669#ifdef IPFIREWALL_FORWARD 670 ip_fw_fwd_addr = NULL; 671#endif 672 return; 673 } 674#endif /* IPSTEALTH */ 675 676 /* Count the packet in the ip address stats */ 677 if (ia != NULL) { 678 ia->ia_ifa.if_ipackets++; 679 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 680 } 681 682 /* 683 * If offset or IP_MF are set, must reassemble. 684 * Otherwise, nothing need be done. 685 * (We could look in the reassembly queue to see 686 * if the packet was previously fragmented, 687 * but it's not worth the time; just let them time out.) 688 */ 689 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 690 691 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 692 /* 693 * Look for queue of fragments 694 * of this datagram. 695 */ 696 TAILQ_FOREACH(fp, &ipq[sum], ipq_list) 697 if (ip->ip_id == fp->ipq_id && 698 ip->ip_src.s_addr == fp->ipq_src.s_addr && 699 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 700 ip->ip_p == fp->ipq_p) 701 goto found; 702 703 fp = 0; 704 705 /* check if there's a place for the new queue */ 706 if (nipq > maxnipq) { 707 /* 708 * drop something from the tail of the current queue 709 * before proceeding further 710 */ 711 struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead); 712 if (q == NULL) { /* gak */ 713 for (i = 0; i < IPREASS_NHASH; i++) { 714 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 715 if (r) { 716 ip_freef(&ipq[i], r); 717 break; 718 } 719 } 720 } else 721 ip_freef(&ipq[sum], q); 722 } 723found: 724 /* 725 * Adjust ip_len to not reflect header, 726 * convert offset of this to bytes. 727 */ 728 ip->ip_len -= hlen; 729 if (ip->ip_off & IP_MF) { 730 /* 731 * Make sure that fragments have a data length 732 * that's a non-zero multiple of 8 bytes. 733 */ 734 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 735 ipstat.ips_toosmall++; /* XXX */ 736 goto bad; 737 } 738 m->m_flags |= M_FRAG; 739 } 740 ip->ip_off <<= 3; 741 742 /* 743 * Attempt reassembly; if it succeeds, proceed. 744 */ 745 ipstat.ips_fragments++; 746 m->m_pkthdr.header = ip; 747#ifdef IPDIVERT 748 m = ip_reass(m, 749 &ipq[sum], fp, &divert_info, &divert_cookie); 750#else 751 m = ip_reass(m, &ipq[sum], fp); 752#endif 753 if (m == 0) { 754#ifdef IPFIREWALL_FORWARD 755 ip_fw_fwd_addr = NULL; 756#endif 757 return; 758 } 759 ipstat.ips_reassembled++; 760 ip = mtod(m, struct ip *); 761 /* Get the header length of the reassembled packet */ 762 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 763#ifdef IPDIVERT 764 /* Restore original checksum before diverting packet */ 765 if (divert_info != 0) { 766 ip->ip_len += hlen; 767 ip->ip_len = htons(ip->ip_len); 768 ip->ip_off = htons(ip->ip_off); 769 ip->ip_sum = 0; 770 if (hlen == sizeof(struct ip)) 771 ip->ip_sum = in_cksum_hdr(ip); 772 else 773 ip->ip_sum = in_cksum(m, hlen); 774 ip->ip_off = ntohs(ip->ip_off); 775 ip->ip_len = ntohs(ip->ip_len); 776 ip->ip_len -= hlen; 777 } 778#endif 779 } else 780 ip->ip_len -= hlen; 781 782#ifdef IPDIVERT 783 /* 784 * Divert or tee packet to the divert protocol if required. 785 * 786 * If divert_info is zero then cookie should be too, so we shouldn't 787 * need to clear them here. Assume divert_packet() does so also. 788 */ 789 if (divert_info != 0) { 790 struct mbuf *clone = NULL; 791 792 /* Clone packet if we're doing a 'tee' */ 793 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 794 clone = m_dup(m, M_DONTWAIT); 795 796 /* Restore packet header fields to original values */ 797 ip->ip_len += hlen; 798 ip->ip_len = htons(ip->ip_len); 799 ip->ip_off = htons(ip->ip_off); 800 801 /* Deliver packet to divert input routine */ 802 ip_divert_cookie = divert_cookie; 803 divert_packet(m, 1, divert_info & 0xffff); 804 ipstat.ips_delivered++; 805 806 /* If 'tee', continue with original packet */ 807 if (clone == NULL) 808 return; 809 m = clone; 810 ip = mtod(m, struct ip *); 811 ip->ip_len += hlen; 812 divert_info = 0; 813 goto pass; 814 } 815#endif 816 817#ifdef IPSEC 818 /* 819 * enforce IPsec policy checking if we are seeing last header. 820 * note that we do not visit this with protocols with pcb layer 821 * code - like udp/tcp/raw ip. 822 */ 823 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 824 ipsec4_in_reject(m, NULL)) { 825 ipsecstat.in_polvio++; 826 goto bad; 827 } 828#endif 829 830 /* 831 * Switch out to protocol's input routine. 832 */ 833 ipstat.ips_delivered++; 834 { 835 int off = hlen; 836 837 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off); 838#ifdef IPFIREWALL_FORWARD 839 ip_fw_fwd_addr = NULL; /* tcp needed it */ 840#endif 841 return; 842 } 843bad: 844#ifdef IPFIREWALL_FORWARD 845 ip_fw_fwd_addr = NULL; 846#endif 847 m_freem(m); 848} 849 850/* 851 * IP software interrupt routine - to go away sometime soon 852 */ 853static void 854ipintr(void) 855{ 856 struct mbuf *m; 857 858 while (1) { 859 IF_DEQUEUE(&ipintrq, m); 860 if (m == 0) 861 return; 862 ip_input(m); 863 } 864} 865 866/* 867 * Take incoming datagram fragment and try to reassemble it into 868 * whole datagram. If a chain for reassembly of this datagram already 869 * exists, then it is given as fp; otherwise have to make a chain. 870 * 871 * When IPDIVERT enabled, keep additional state with each packet that 872 * tells us if we need to divert or tee the packet we're building. 873 */ 874 875static struct mbuf * 876#ifdef IPDIVERT 877ip_reass(m, head, fp, divinfo, divcookie) 878#else 879ip_reass(m, head, fp) 880#endif 881 struct mbuf *m; 882 struct ipqhead *head; 883 struct ipq *fp; 884#ifdef IPDIVERT 885 u_int32_t *divinfo; 886 u_int16_t *divcookie; 887#endif 888{ 889 struct ip *ip = mtod(m, struct ip *); 890 register struct mbuf *p, *q, *nq; 891 struct mbuf *t; 892 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 893 int i, next; 894 895 /* 896 * Presence of header sizes in mbufs 897 * would confuse code below. 898 */ 899 m->m_data += hlen; 900 m->m_len -= hlen; 901 902 /* 903 * If first fragment to arrive, create a reassembly queue. 904 */ 905 if (fp == 0) { 906 /* 907 * Enforce upper bound on number of fragmented packets 908 * for which we attempt reassembly; 909 * If maxfrag is 0, never accept fragments. 910 * If maxfrag is -1, accept all fragments without limitation. 911 */ 912 if ((ip_maxfragpackets >= 0) && (ip_nfragpackets >= ip_maxfragpackets)) 913 goto dropfrag; 914 ip_nfragpackets++; 915 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 916 goto dropfrag; 917 fp = mtod(t, struct ipq *); 918 TAILQ_INSERT_HEAD(head, fp, ipq_list); 919 nipq++; 920 fp->ipq_ttl = IPFRAGTTL; 921 fp->ipq_p = ip->ip_p; 922 fp->ipq_id = ip->ip_id; 923 fp->ipq_src = ip->ip_src; 924 fp->ipq_dst = ip->ip_dst; 925 fp->ipq_frags = m; 926 m->m_nextpkt = NULL; 927#ifdef IPDIVERT 928 fp->ipq_div_info = 0; 929 fp->ipq_div_cookie = 0; 930#endif 931 goto inserted; 932 } 933 934#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 935 936 /* 937 * Find a segment which begins after this one does. 938 */ 939 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 940 if (GETIP(q)->ip_off > ip->ip_off) 941 break; 942 943 /* 944 * If there is a preceding segment, it may provide some of 945 * our data already. If so, drop the data from the incoming 946 * segment. If it provides all of our data, drop us, otherwise 947 * stick new segment in the proper place. 948 * 949 * If some of the data is dropped from the the preceding 950 * segment, then it's checksum is invalidated. 951 */ 952 if (p) { 953 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 954 if (i > 0) { 955 if (i >= ip->ip_len) 956 goto dropfrag; 957 m_adj(m, i); 958 m->m_pkthdr.csum_flags = 0; 959 ip->ip_off += i; 960 ip->ip_len -= i; 961 } 962 m->m_nextpkt = p->m_nextpkt; 963 p->m_nextpkt = m; 964 } else { 965 m->m_nextpkt = fp->ipq_frags; 966 fp->ipq_frags = m; 967 } 968 969 /* 970 * While we overlap succeeding segments trim them or, 971 * if they are completely covered, dequeue them. 972 */ 973 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 974 q = nq) { 975 i = (ip->ip_off + ip->ip_len) - 976 GETIP(q)->ip_off; 977 if (i < GETIP(q)->ip_len) { 978 GETIP(q)->ip_len -= i; 979 GETIP(q)->ip_off += i; 980 m_adj(q, i); 981 q->m_pkthdr.csum_flags = 0; 982 break; 983 } 984 nq = q->m_nextpkt; 985 m->m_nextpkt = nq; 986 m_freem(q); 987 } 988 989inserted: 990 991#ifdef IPDIVERT 992 /* 993 * Transfer firewall instructions to the fragment structure. 994 * Any fragment diverting causes the whole packet to divert. 995 */ 996 fp->ipq_div_info = *divinfo; 997 fp->ipq_div_cookie = *divcookie; 998 *divinfo = 0; 999 *divcookie = 0; 1000#endif 1001 1002 /* 1003 * Check for complete reassembly. 1004 */ 1005 next = 0; 1006 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1007 if (GETIP(q)->ip_off != next) 1008 return (0); 1009 next += GETIP(q)->ip_len; 1010 } 1011 /* Make sure the last packet didn't have the IP_MF flag */ 1012 if (p->m_flags & M_FRAG) 1013 return (0); 1014 1015 /* 1016 * Reassembly is complete. Make sure the packet is a sane size. 1017 */ 1018 q = fp->ipq_frags; 1019 ip = GETIP(q); 1020 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1021 ipstat.ips_toolong++; 1022 ip_freef(head, fp); 1023 return (0); 1024 } 1025 1026 /* 1027 * Concatenate fragments. 1028 */ 1029 m = q; 1030 t = m->m_next; 1031 m->m_next = 0; 1032 m_cat(m, t); 1033 nq = q->m_nextpkt; 1034 q->m_nextpkt = 0; 1035 for (q = nq; q != NULL; q = nq) { 1036 nq = q->m_nextpkt; 1037 q->m_nextpkt = NULL; 1038 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1039 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1040 m_cat(m, q); 1041 } 1042 1043#ifdef IPDIVERT 1044 /* 1045 * Extract firewall instructions from the fragment structure. 1046 */ 1047 *divinfo = fp->ipq_div_info; 1048 *divcookie = fp->ipq_div_cookie; 1049#endif 1050 1051 /* 1052 * Create header for new ip packet by 1053 * modifying header of first packet; 1054 * dequeue and discard fragment reassembly header. 1055 * Make header visible. 1056 */ 1057 ip->ip_len = next; 1058 ip->ip_src = fp->ipq_src; 1059 ip->ip_dst = fp->ipq_dst; 1060 TAILQ_REMOVE(head, fp, ipq_list); 1061 nipq--; 1062 (void) m_free(dtom(fp)); 1063 ip_nfragpackets--; 1064 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1065 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1066 /* some debugging cruft by sklower, below, will go away soon */ 1067 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1068 register int plen = 0; 1069 for (t = m; t; t = t->m_next) 1070 plen += t->m_len; 1071 m->m_pkthdr.len = plen; 1072 } 1073 return (m); 1074 1075dropfrag: 1076#ifdef IPDIVERT 1077 *divinfo = 0; 1078 *divcookie = 0; 1079#endif 1080 ipstat.ips_fragdropped++; 1081 m_freem(m); 1082 return (0); 1083 1084#undef GETIP 1085} 1086 1087/* 1088 * Free a fragment reassembly header and all 1089 * associated datagrams. 1090 */ 1091static void 1092ip_freef(fhp, fp) 1093 struct ipqhead *fhp; 1094 struct ipq *fp; 1095{ 1096 register struct mbuf *q; 1097 1098 while (fp->ipq_frags) { 1099 q = fp->ipq_frags; 1100 fp->ipq_frags = q->m_nextpkt; 1101 m_freem(q); 1102 } 1103 TAILQ_REMOVE(fhp, fp, ipq_list); 1104 (void) m_free(dtom(fp)); 1105 ip_nfragpackets--; 1106 nipq--; 1107} 1108 1109/* 1110 * IP timer processing; 1111 * if a timer expires on a reassembly 1112 * queue, discard it. 1113 */ 1114void 1115ip_slowtimo() 1116{ 1117 register struct ipq *fp; 1118 int s = splnet(); 1119 int i; 1120 1121 for (i = 0; i < IPREASS_NHASH; i++) { 1122 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1123 struct ipq *fpp; 1124 1125 fpp = fp; 1126 fp = TAILQ_NEXT(fp, ipq_list); 1127 if(--fpp->ipq_ttl == 0) { 1128 ipstat.ips_fragtimeout++; 1129 ip_freef(&ipq[i], fpp); 1130 } 1131 } 1132 } 1133 /* 1134 * If we are over the maximum number of fragments 1135 * (due to the limit being lowered), drain off 1136 * enough to get down to the new limit. 1137 */ 1138 for (i = 0; i < IPREASS_NHASH; i++) { 1139 if (ip_maxfragpackets >= 0) { 1140 while (ip_nfragpackets > ip_maxfragpackets && 1141 !TAILQ_EMPTY(&ipq[i])) { 1142 ipstat.ips_fragdropped++; 1143 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1144 } 1145 } 1146 } 1147 ipflow_slowtimo(); 1148 splx(s); 1149} 1150 1151/* 1152 * Drain off all datagram fragments. 1153 */ 1154void 1155ip_drain() 1156{ 1157 int i; 1158 1159 for (i = 0; i < IPREASS_NHASH; i++) { 1160 while(!TAILQ_EMPTY(&ipq[i])) { 1161 ipstat.ips_fragdropped++; 1162 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1163 } 1164 } 1165 in_rtqdrain(); 1166} 1167 1168/* 1169 * Do option processing on a datagram, 1170 * possibly discarding it if bad options are encountered, 1171 * or forwarding it if source-routed. 1172 * The pass argument is used when operating in the IPSTEALTH 1173 * mode to tell what options to process: 1174 * [LS]SRR (pass 0) or the others (pass 1). 1175 * The reason for as many as two passes is that when doing IPSTEALTH, 1176 * non-routing options should be processed only if the packet is for us. 1177 * Returns 1 if packet has been forwarded/freed, 1178 * 0 if the packet should be processed further. 1179 */ 1180static int 1181ip_dooptions(m, pass) 1182 struct mbuf *m; 1183 int pass; 1184{ 1185 register struct ip *ip = mtod(m, struct ip *); 1186 register u_char *cp; 1187 register struct in_ifaddr *ia; 1188 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1189 struct in_addr *sin, dst; 1190 n_time ntime; 1191 1192 dst = ip->ip_dst; 1193 cp = (u_char *)(ip + 1); 1194 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1195 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1196 opt = cp[IPOPT_OPTVAL]; 1197 if (opt == IPOPT_EOL) 1198 break; 1199 if (opt == IPOPT_NOP) 1200 optlen = 1; 1201 else { 1202 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1203 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1204 goto bad; 1205 } 1206 optlen = cp[IPOPT_OLEN]; 1207 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1208 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1209 goto bad; 1210 } 1211 } 1212 switch (opt) { 1213 1214 default: 1215 break; 1216 1217 /* 1218 * Source routing with record. 1219 * Find interface with current destination address. 1220 * If none on this machine then drop if strictly routed, 1221 * or do nothing if loosely routed. 1222 * Record interface address and bring up next address 1223 * component. If strictly routed make sure next 1224 * address is on directly accessible net. 1225 */ 1226 case IPOPT_LSRR: 1227 case IPOPT_SSRR: 1228#ifdef IPSTEALTH 1229 if (ipstealth && pass > 0) 1230 break; 1231#endif 1232 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1233 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1234 goto bad; 1235 } 1236 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1237 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1238 goto bad; 1239 } 1240 ipaddr.sin_addr = ip->ip_dst; 1241 ia = (struct in_ifaddr *) 1242 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1243 if (ia == 0) { 1244 if (opt == IPOPT_SSRR) { 1245 type = ICMP_UNREACH; 1246 code = ICMP_UNREACH_SRCFAIL; 1247 goto bad; 1248 } 1249 if (!ip_dosourceroute) 1250 goto nosourcerouting; 1251 /* 1252 * Loose routing, and not at next destination 1253 * yet; nothing to do except forward. 1254 */ 1255 break; 1256 } 1257 off--; /* 0 origin */ 1258 if (off > optlen - (int)sizeof(struct in_addr)) { 1259 /* 1260 * End of source route. Should be for us. 1261 */ 1262 if (!ip_acceptsourceroute) 1263 goto nosourcerouting; 1264 save_rte(cp, ip->ip_src); 1265 break; 1266 } 1267#ifdef IPSTEALTH 1268 if (ipstealth) 1269 goto dropit; 1270#endif 1271 if (!ip_dosourceroute) { 1272 if (ipforwarding) { 1273 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1274 /* 1275 * Acting as a router, so generate ICMP 1276 */ 1277nosourcerouting: 1278 strcpy(buf, inet_ntoa(ip->ip_dst)); 1279 log(LOG_WARNING, 1280 "attempted source route from %s to %s\n", 1281 inet_ntoa(ip->ip_src), buf); 1282 type = ICMP_UNREACH; 1283 code = ICMP_UNREACH_SRCFAIL; 1284 goto bad; 1285 } else { 1286 /* 1287 * Not acting as a router, so silently drop. 1288 */ 1289#ifdef IPSTEALTH 1290dropit: 1291#endif 1292 ipstat.ips_cantforward++; 1293 m_freem(m); 1294 return (1); 1295 } 1296 } 1297 1298 /* 1299 * locate outgoing interface 1300 */ 1301 (void)memcpy(&ipaddr.sin_addr, cp + off, 1302 sizeof(ipaddr.sin_addr)); 1303 1304 if (opt == IPOPT_SSRR) { 1305#define INA struct in_ifaddr * 1306#define SA struct sockaddr * 1307 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1308 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1309 } else 1310 ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt); 1311 if (ia == 0) { 1312 type = ICMP_UNREACH; 1313 code = ICMP_UNREACH_SRCFAIL; 1314 goto bad; 1315 } 1316 ip->ip_dst = ipaddr.sin_addr; 1317 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1318 sizeof(struct in_addr)); 1319 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1320 /* 1321 * Let ip_intr's mcast routing check handle mcast pkts 1322 */ 1323 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1324 break; 1325 1326 case IPOPT_RR: 1327#ifdef IPSTEALTH 1328 if (ipstealth && pass == 0) 1329 break; 1330#endif 1331 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1332 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1333 goto bad; 1334 } 1335 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1336 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1337 goto bad; 1338 } 1339 /* 1340 * If no space remains, ignore. 1341 */ 1342 off--; /* 0 origin */ 1343 if (off > optlen - (int)sizeof(struct in_addr)) 1344 break; 1345 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1346 sizeof(ipaddr.sin_addr)); 1347 /* 1348 * locate outgoing interface; if we're the destination, 1349 * use the incoming interface (should be same). 1350 */ 1351 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1352 (ia = ip_rtaddr(ipaddr.sin_addr, 1353 &ipforward_rt)) == 0) { 1354 type = ICMP_UNREACH; 1355 code = ICMP_UNREACH_HOST; 1356 goto bad; 1357 } 1358 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1359 sizeof(struct in_addr)); 1360 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1361 break; 1362 1363 case IPOPT_TS: 1364#ifdef IPSTEALTH 1365 if (ipstealth && pass == 0) 1366 break; 1367#endif 1368 code = cp - (u_char *)ip; 1369 if (optlen < 4 || optlen > 40) { 1370 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1371 goto bad; 1372 } 1373 if ((off = cp[IPOPT_OFFSET]) < 5) { 1374 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1375 goto bad; 1376 } 1377 if (off > optlen - (int)sizeof(int32_t)) { 1378 cp[IPOPT_OFFSET + 1] += (1 << 4); 1379 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1380 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1381 goto bad; 1382 } 1383 break; 1384 } 1385 off--; /* 0 origin */ 1386 sin = (struct in_addr *)(cp + off); 1387 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1388 1389 case IPOPT_TS_TSONLY: 1390 break; 1391 1392 case IPOPT_TS_TSANDADDR: 1393 if (off + sizeof(n_time) + 1394 sizeof(struct in_addr) > optlen) { 1395 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1396 goto bad; 1397 } 1398 ipaddr.sin_addr = dst; 1399 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1400 m->m_pkthdr.rcvif); 1401 if (ia == 0) 1402 continue; 1403 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1404 sizeof(struct in_addr)); 1405 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1406 break; 1407 1408 case IPOPT_TS_PRESPEC: 1409 if (off + sizeof(n_time) + 1410 sizeof(struct in_addr) > optlen) { 1411 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1412 goto bad; 1413 } 1414 (void)memcpy(&ipaddr.sin_addr, sin, 1415 sizeof(struct in_addr)); 1416 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1417 continue; 1418 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1419 break; 1420 1421 default: 1422 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1423 goto bad; 1424 } 1425 ntime = iptime(); 1426 (void)memcpy(cp + off, &ntime, sizeof(n_time)); 1427 cp[IPOPT_OFFSET] += sizeof(n_time); 1428 } 1429 } 1430 if (forward && ipforwarding) { 1431 ip_forward(m, 1); 1432 return (1); 1433 } 1434 return (0); 1435bad: 1436 icmp_error(m, type, code, 0, 0); 1437 ipstat.ips_badoptions++; 1438 return (1); 1439} 1440 1441/* 1442 * Given address of next destination (final or next hop), 1443 * return internet address info of interface to be used to get there. 1444 */ 1445struct in_ifaddr * 1446ip_rtaddr(dst, rt) 1447 struct in_addr dst; 1448 struct route *rt; 1449{ 1450 register struct sockaddr_in *sin; 1451 1452 sin = (struct sockaddr_in *)&rt->ro_dst; 1453 1454 if (rt->ro_rt == 0 || 1455 !(rt->ro_rt->rt_flags & RTF_UP) || 1456 dst.s_addr != sin->sin_addr.s_addr) { 1457 if (rt->ro_rt) { 1458 RTFREE(rt->ro_rt); 1459 rt->ro_rt = 0; 1460 } 1461 sin->sin_family = AF_INET; 1462 sin->sin_len = sizeof(*sin); 1463 sin->sin_addr = dst; 1464 1465 rtalloc_ign(rt, RTF_PRCLONING); 1466 } 1467 if (rt->ro_rt == 0) 1468 return ((struct in_ifaddr *)0); 1469 return (ifatoia(rt->ro_rt->rt_ifa)); 1470} 1471 1472/* 1473 * Save incoming source route for use in replies, 1474 * to be picked up later by ip_srcroute if the receiver is interested. 1475 */ 1476void 1477save_rte(option, dst) 1478 u_char *option; 1479 struct in_addr dst; 1480{ 1481 unsigned olen; 1482 1483 olen = option[IPOPT_OLEN]; 1484#ifdef DIAGNOSTIC 1485 if (ipprintfs) 1486 printf("save_rte: olen %d\n", olen); 1487#endif 1488 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1489 return; 1490 bcopy(option, ip_srcrt.srcopt, olen); 1491 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1492 ip_srcrt.dst = dst; 1493} 1494 1495/* 1496 * Retrieve incoming source route for use in replies, 1497 * in the same form used by setsockopt. 1498 * The first hop is placed before the options, will be removed later. 1499 */ 1500struct mbuf * 1501ip_srcroute() 1502{ 1503 register struct in_addr *p, *q; 1504 register struct mbuf *m; 1505 1506 if (ip_nhops == 0) 1507 return ((struct mbuf *)0); 1508 m = m_get(M_DONTWAIT, MT_HEADER); 1509 if (m == 0) 1510 return ((struct mbuf *)0); 1511 1512#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1513 1514 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1515 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1516 OPTSIZ; 1517#ifdef DIAGNOSTIC 1518 if (ipprintfs) 1519 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1520#endif 1521 1522 /* 1523 * First save first hop for return route 1524 */ 1525 p = &ip_srcrt.route[ip_nhops - 1]; 1526 *(mtod(m, struct in_addr *)) = *p--; 1527#ifdef DIAGNOSTIC 1528 if (ipprintfs) 1529 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1530#endif 1531 1532 /* 1533 * Copy option fields and padding (nop) to mbuf. 1534 */ 1535 ip_srcrt.nop = IPOPT_NOP; 1536 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1537 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1538 &ip_srcrt.nop, OPTSIZ); 1539 q = (struct in_addr *)(mtod(m, caddr_t) + 1540 sizeof(struct in_addr) + OPTSIZ); 1541#undef OPTSIZ 1542 /* 1543 * Record return path as an IP source route, 1544 * reversing the path (pointers are now aligned). 1545 */ 1546 while (p >= ip_srcrt.route) { 1547#ifdef DIAGNOSTIC 1548 if (ipprintfs) 1549 printf(" %lx", (u_long)ntohl(q->s_addr)); 1550#endif 1551 *q++ = *p--; 1552 } 1553 /* 1554 * Last hop goes to final destination. 1555 */ 1556 *q = ip_srcrt.dst; 1557#ifdef DIAGNOSTIC 1558 if (ipprintfs) 1559 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1560#endif 1561 return (m); 1562} 1563 1564/* 1565 * Strip out IP options, at higher 1566 * level protocol in the kernel. 1567 * Second argument is buffer to which options 1568 * will be moved, and return value is their length. 1569 * XXX should be deleted; last arg currently ignored. 1570 */ 1571void 1572ip_stripoptions(m, mopt) 1573 register struct mbuf *m; 1574 struct mbuf *mopt; 1575{ 1576 register int i; 1577 struct ip *ip = mtod(m, struct ip *); 1578 register caddr_t opts; 1579 int olen; 1580 1581 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1582 opts = (caddr_t)(ip + 1); 1583 i = m->m_len - (sizeof (struct ip) + olen); 1584 bcopy(opts + olen, opts, (unsigned)i); 1585 m->m_len -= olen; 1586 if (m->m_flags & M_PKTHDR) 1587 m->m_pkthdr.len -= olen; 1588 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1589} 1590 1591u_char inetctlerrmap[PRC_NCMDS] = { 1592 0, 0, 0, 0, 1593 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1594 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1595 EMSGSIZE, EHOSTUNREACH, 0, 0, 1596 0, 0, 0, 0, 1597 ENOPROTOOPT, ECONNREFUSED 1598}; 1599 1600/* 1601 * Forward a packet. If some error occurs return the sender 1602 * an icmp packet. Note we can't always generate a meaningful 1603 * icmp message because icmp doesn't have a large enough repertoire 1604 * of codes and types. 1605 * 1606 * If not forwarding, just drop the packet. This could be confusing 1607 * if ipforwarding was zero but some routing protocol was advancing 1608 * us as a gateway to somewhere. However, we must let the routing 1609 * protocol deal with that. 1610 * 1611 * The srcrt parameter indicates whether the packet is being forwarded 1612 * via a source route. 1613 */ 1614static void 1615ip_forward(m, srcrt) 1616 struct mbuf *m; 1617 int srcrt; 1618{ 1619 register struct ip *ip = mtod(m, struct ip *); 1620 register struct rtentry *rt; 1621 int error, type = 0, code = 0; 1622 struct mbuf *mcopy; 1623 n_long dest; 1624 struct in_addr pkt_dst; 1625 struct ifnet *destifp; 1626#ifdef IPSEC 1627 struct ifnet dummyifp; 1628#endif 1629 1630 dest = 0; 1631 /* 1632 * Cache the destination address of the packet; this may be 1633 * changed by use of 'ipfw fwd'. 1634 */ 1635 pkt_dst = ip_fw_fwd_addr == NULL ? 1636 ip->ip_dst : ip_fw_fwd_addr->sin_addr; 1637 1638#ifdef DIAGNOSTIC 1639 if (ipprintfs) 1640 printf("forward: src %lx dst %lx ttl %x\n", 1641 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr, 1642 ip->ip_ttl); 1643#endif 1644 1645 1646 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) { 1647 ipstat.ips_cantforward++; 1648 m_freem(m); 1649 return; 1650 } 1651#ifdef IPSTEALTH 1652 if (!ipstealth) { 1653#endif 1654 if (ip->ip_ttl <= IPTTLDEC) { 1655 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1656 dest, 0); 1657 return; 1658 } 1659#ifdef IPSTEALTH 1660 } 1661#endif 1662 1663 if (ip_rtaddr(pkt_dst, &ipforward_rt) == 0) { 1664 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1665 return; 1666 } else 1667 rt = ipforward_rt.ro_rt; 1668 1669 /* 1670 * Save the IP header and at most 8 bytes of the payload, 1671 * in case we need to generate an ICMP message to the src. 1672 * 1673 * We don't use m_copy() because it might return a reference 1674 * to a shared cluster. Both this function and ip_output() 1675 * assume exclusive access to the IP header in `m', so any 1676 * data in a cluster may change before we reach icmp_error(). 1677 */ 1678 MGET(mcopy, M_DONTWAIT, m->m_type); 1679 if (mcopy != NULL) { 1680 M_COPY_PKTHDR(mcopy, m); 1681 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1682 (int)ip->ip_len); 1683 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1684 } 1685 1686#ifdef IPSTEALTH 1687 if (!ipstealth) { 1688#endif 1689 ip->ip_ttl -= IPTTLDEC; 1690#ifdef IPSTEALTH 1691 } 1692#endif 1693 1694 /* 1695 * If forwarding packet using same interface that it came in on, 1696 * perhaps should send a redirect to sender to shortcut a hop. 1697 * Only send redirect if source is sending directly to us, 1698 * and if packet was not source routed (or has any options). 1699 * Also, don't send redirect if forwarding using a default route 1700 * or a route modified by a redirect. 1701 */ 1702 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1703 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1704 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1705 ipsendredirects && !srcrt && !ip_fw_fwd_addr) { 1706#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1707 u_long src = ntohl(ip->ip_src.s_addr); 1708 1709 if (RTA(rt) && 1710 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1711 if (rt->rt_flags & RTF_GATEWAY) 1712 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1713 else 1714 dest = pkt_dst.s_addr; 1715 /* Router requirements says to only send host redirects */ 1716 type = ICMP_REDIRECT; 1717 code = ICMP_REDIRECT_HOST; 1718#ifdef DIAGNOSTIC 1719 if (ipprintfs) 1720 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1721#endif 1722 } 1723 } 1724 1725 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1726 IP_FORWARDING, 0); 1727 if (error) 1728 ipstat.ips_cantforward++; 1729 else { 1730 ipstat.ips_forward++; 1731 if (type) 1732 ipstat.ips_redirectsent++; 1733 else { 1734 if (mcopy) { 1735 ipflow_create(&ipforward_rt, mcopy); 1736 m_freem(mcopy); 1737 } 1738 return; 1739 } 1740 } 1741 if (mcopy == NULL) 1742 return; 1743 destifp = NULL; 1744 1745 switch (error) { 1746 1747 case 0: /* forwarded, but need redirect */ 1748 /* type, code set above */ 1749 break; 1750 1751 case ENETUNREACH: /* shouldn't happen, checked above */ 1752 case EHOSTUNREACH: 1753 case ENETDOWN: 1754 case EHOSTDOWN: 1755 default: 1756 type = ICMP_UNREACH; 1757 code = ICMP_UNREACH_HOST; 1758 break; 1759 1760 case EMSGSIZE: 1761 type = ICMP_UNREACH; 1762 code = ICMP_UNREACH_NEEDFRAG; 1763#ifndef IPSEC 1764 if (ipforward_rt.ro_rt) 1765 destifp = ipforward_rt.ro_rt->rt_ifp; 1766#else 1767 /* 1768 * If the packet is routed over IPsec tunnel, tell the 1769 * originator the tunnel MTU. 1770 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1771 * XXX quickhack!!! 1772 */ 1773 if (ipforward_rt.ro_rt) { 1774 struct secpolicy *sp = NULL; 1775 int ipsecerror; 1776 int ipsechdr; 1777 struct route *ro; 1778 1779 sp = ipsec4_getpolicybyaddr(mcopy, 1780 IPSEC_DIR_OUTBOUND, 1781 IP_FORWARDING, 1782 &ipsecerror); 1783 1784 if (sp == NULL) 1785 destifp = ipforward_rt.ro_rt->rt_ifp; 1786 else { 1787 /* count IPsec header size */ 1788 ipsechdr = ipsec4_hdrsiz(mcopy, 1789 IPSEC_DIR_OUTBOUND, 1790 NULL); 1791 1792 /* 1793 * find the correct route for outer IPv4 1794 * header, compute tunnel MTU. 1795 * 1796 * XXX BUG ALERT 1797 * The "dummyifp" code relies upon the fact 1798 * that icmp_error() touches only ifp->if_mtu. 1799 */ 1800 /*XXX*/ 1801 destifp = NULL; 1802 if (sp->req != NULL 1803 && sp->req->sav != NULL 1804 && sp->req->sav->sah != NULL) { 1805 ro = &sp->req->sav->sah->sa_route; 1806 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1807 dummyifp.if_mtu = 1808 ro->ro_rt->rt_ifp->if_mtu; 1809 dummyifp.if_mtu -= ipsechdr; 1810 destifp = &dummyifp; 1811 } 1812 } 1813 1814 key_freesp(sp); 1815 } 1816 } 1817#endif /*IPSEC*/ 1818 ipstat.ips_cantfrag++; 1819 break; 1820 1821 case ENOBUFS: 1822 type = ICMP_SOURCEQUENCH; 1823 code = 0; 1824 break; 1825 1826 case EACCES: /* ipfw denied packet */ 1827 m_freem(mcopy); 1828 return; 1829 } 1830 icmp_error(mcopy, type, code, dest, destifp); 1831} 1832 1833void 1834ip_savecontrol(inp, mp, ip, m) 1835 register struct inpcb *inp; 1836 register struct mbuf **mp; 1837 register struct ip *ip; 1838 register struct mbuf *m; 1839{ 1840 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1841 struct timeval tv; 1842 1843 microtime(&tv); 1844 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1845 SCM_TIMESTAMP, SOL_SOCKET); 1846 if (*mp) 1847 mp = &(*mp)->m_next; 1848 } 1849 if (inp->inp_flags & INP_RECVDSTADDR) { 1850 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1851 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1852 if (*mp) 1853 mp = &(*mp)->m_next; 1854 } 1855#ifdef notyet 1856 /* XXX 1857 * Moving these out of udp_input() made them even more broken 1858 * than they already were. 1859 */ 1860 /* options were tossed already */ 1861 if (inp->inp_flags & INP_RECVOPTS) { 1862 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1863 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1864 if (*mp) 1865 mp = &(*mp)->m_next; 1866 } 1867 /* ip_srcroute doesn't do what we want here, need to fix */ 1868 if (inp->inp_flags & INP_RECVRETOPTS) { 1869 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1870 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1871 if (*mp) 1872 mp = &(*mp)->m_next; 1873 } 1874#endif 1875 if (inp->inp_flags & INP_RECVIF) { 1876 struct ifnet *ifp; 1877 struct sdlbuf { 1878 struct sockaddr_dl sdl; 1879 u_char pad[32]; 1880 } sdlbuf; 1881 struct sockaddr_dl *sdp; 1882 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1883 1884 if (((ifp = m->m_pkthdr.rcvif)) 1885 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1886 sdp = (struct sockaddr_dl *) 1887 (ifaddr_byindex(ifp->if_index)->ifa_addr); 1888 /* 1889 * Change our mind and don't try copy. 1890 */ 1891 if ((sdp->sdl_family != AF_LINK) 1892 || (sdp->sdl_len > sizeof(sdlbuf))) { 1893 goto makedummy; 1894 } 1895 bcopy(sdp, sdl2, sdp->sdl_len); 1896 } else { 1897makedummy: 1898 sdl2->sdl_len 1899 = offsetof(struct sockaddr_dl, sdl_data[0]); 1900 sdl2->sdl_family = AF_LINK; 1901 sdl2->sdl_index = 0; 1902 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1903 } 1904 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1905 IP_RECVIF, IPPROTO_IP); 1906 if (*mp) 1907 mp = &(*mp)->m_next; 1908 } 1909} 1910 1911int 1912ip_rsvp_init(struct socket *so) 1913{ 1914 if (so->so_type != SOCK_RAW || 1915 so->so_proto->pr_protocol != IPPROTO_RSVP) 1916 return EOPNOTSUPP; 1917 1918 if (ip_rsvpd != NULL) 1919 return EADDRINUSE; 1920 1921 ip_rsvpd = so; 1922 /* 1923 * This may seem silly, but we need to be sure we don't over-increment 1924 * the RSVP counter, in case something slips up. 1925 */ 1926 if (!ip_rsvp_on) { 1927 ip_rsvp_on = 1; 1928 rsvp_on++; 1929 } 1930 1931 return 0; 1932} 1933 1934int 1935ip_rsvp_done(void) 1936{ 1937 ip_rsvpd = NULL; 1938 /* 1939 * This may seem silly, but we need to be sure we don't over-decrement 1940 * the RSVP counter, in case something slips up. 1941 */ 1942 if (ip_rsvp_on) { 1943 ip_rsvp_on = 0; 1944 rsvp_on--; 1945 } 1946 return 0; 1947} 1948