ip_input.c revision 112973
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 112973 2003-04-02 10:28:47Z mdodd $ 35 */ 36 37#include "opt_bootp.h" 38#include "opt_ipfw.h" 39#include "opt_ipdn.h" 40#include "opt_ipdivert.h" 41#include "opt_ipfilter.h" 42#include "opt_ipstealth.h" 43#include "opt_ipsec.h" 44#include "opt_mac.h" 45#include "opt_pfil_hooks.h" 46#include "opt_random_ip_id.h" 47 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/mac.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 69#include <netinet/in.h> 70#include <netinet/in_systm.h> 71#include <netinet/in_var.h> 72#include <netinet/ip.h> 73#include <netinet/in_pcb.h> 74#include <netinet/ip_var.h> 75#include <netinet/ip_icmp.h> 76#include <machine/in_cksum.h> 77 78#include <sys/socketvar.h> 79 80#include <netinet/ip_fw.h> 81#include <netinet/ip_dummynet.h> 82 83#ifdef IPSEC 84#include <netinet6/ipsec.h> 85#include <netkey/key.h> 86#endif 87 88#ifdef FAST_IPSEC 89#include <netipsec/ipsec.h> 90#include <netipsec/key.h> 91#endif 92 93int rsvp_on = 0; 94 95int ipforwarding = 0; 96SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 97 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 98 99static int ipsendredirects = 1; /* XXX */ 100SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 101 &ipsendredirects, 0, "Enable sending IP redirects"); 102 103int ip_defttl = IPDEFTTL; 104SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 105 &ip_defttl, 0, "Maximum TTL on IP packets"); 106 107static int ip_dosourceroute = 0; 108SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 109 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 110 111static int ip_acceptsourceroute = 0; 112SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 113 CTLFLAG_RW, &ip_acceptsourceroute, 0, 114 "Enable accepting source routed IP packets"); 115 116static int ip_keepfaith = 0; 117SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 118 &ip_keepfaith, 0, 119 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 120 121static int nipq = 0; /* total # of reass queues */ 122static int maxnipq; 123SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 124 &maxnipq, 0, 125 "Maximum number of IPv4 fragment reassembly queue entries"); 126 127static int maxfragsperpacket; 128SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 129 &maxfragsperpacket, 0, 130 "Maximum number of IPv4 fragments allowed per packet"); 131 132static int ip_sendsourcequench = 0; 133SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 134 &ip_sendsourcequench, 0, 135 "Enable the transmission of source quench packets"); 136 137static int hear_no_evil = 0; 138SYSCTL_INT(_net_inet_ip, OID_AUTO, hear_no_evil, CTLFLAG_RW, 139 &hear_no_evil, 0, 140 "Drop all received EVIL packets."); 141 142/* 143 * XXX - Setting ip_checkinterface mostly implements the receive side of 144 * the Strong ES model described in RFC 1122, but since the routing table 145 * and transmit implementation do not implement the Strong ES model, 146 * setting this to 1 results in an odd hybrid. 147 * 148 * XXX - ip_checkinterface currently must be disabled if you use ipnat 149 * to translate the destination address to another local interface. 150 * 151 * XXX - ip_checkinterface must be disabled if you add IP aliases 152 * to the loopback interface instead of the interface where the 153 * packets for those addresses are received. 154 */ 155static int ip_checkinterface = 1; 156SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 157 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 158 159#ifdef DIAGNOSTIC 160static int ipprintfs = 0; 161#endif 162 163static struct ifqueue ipintrq; 164static int ipqmaxlen = IFQ_MAXLEN; 165 166extern struct domain inetdomain; 167extern struct protosw inetsw[]; 168u_char ip_protox[IPPROTO_MAX]; 169struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 170struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 171u_long in_ifaddrhmask; /* mask for hash table */ 172 173SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 174 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 175SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 176 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 177 178struct ipstat ipstat; 179SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 180 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 181 182/* Packet reassembly stuff */ 183#define IPREASS_NHASH_LOG2 6 184#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 185#define IPREASS_HMASK (IPREASS_NHASH - 1) 186#define IPREASS_HASH(x,y) \ 187 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 188 189static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 190 191#ifdef IPCTL_DEFMTU 192SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 193 &ip_mtu, 0, "Default MTU"); 194#endif 195 196#ifdef IPSTEALTH 197static int ipstealth = 0; 198SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 199 &ipstealth, 0, ""); 200#endif 201 202 203/* Firewall hooks */ 204ip_fw_chk_t *ip_fw_chk_ptr; 205int fw_enable = 1 ; 206int fw_one_pass = 1; 207 208/* Dummynet hooks */ 209ip_dn_io_t *ip_dn_io_ptr; 210 211 212/* 213 * XXX this is ugly -- the following two global variables are 214 * used to store packet state while it travels through the stack. 215 * Note that the code even makes assumptions on the size and 216 * alignment of fields inside struct ip_srcrt so e.g. adding some 217 * fields will break the code. This needs to be fixed. 218 * 219 * We need to save the IP options in case a protocol wants to respond 220 * to an incoming packet over the same route if the packet got here 221 * using IP source routing. This allows connection establishment and 222 * maintenance when the remote end is on a network that is not known 223 * to us. 224 */ 225static int ip_nhops = 0; 226static struct ip_srcrt { 227 struct in_addr dst; /* final destination */ 228 char nop; /* one NOP to align */ 229 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 230 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 231} ip_srcrt; 232 233static void save_rte(u_char *, struct in_addr); 234static int ip_dooptions(struct mbuf *m, int, 235 struct sockaddr_in *next_hop); 236static void ip_forward(struct mbuf *m, int srcrt, 237 struct sockaddr_in *next_hop); 238static void ip_freef(struct ipqhead *, struct ipq *); 239static struct mbuf *ip_reass(struct mbuf *, struct ipqhead *, 240 struct ipq *, u_int32_t *, u_int16_t *); 241 242/* 243 * IP initialization: fill in IP protocol switch table. 244 * All protocols not implemented in kernel go to raw IP protocol handler. 245 */ 246void 247ip_init() 248{ 249 register struct protosw *pr; 250 register int i; 251 252 TAILQ_INIT(&in_ifaddrhead); 253 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 254 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 255 if (pr == 0) 256 panic("ip_init"); 257 for (i = 0; i < IPPROTO_MAX; i++) 258 ip_protox[i] = pr - inetsw; 259 for (pr = inetdomain.dom_protosw; 260 pr < inetdomain.dom_protoswNPROTOSW; pr++) 261 if (pr->pr_domain->dom_family == PF_INET && 262 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 263 ip_protox[pr->pr_protocol] = pr - inetsw; 264 265 for (i = 0; i < IPREASS_NHASH; i++) 266 TAILQ_INIT(&ipq[i]); 267 268 maxnipq = nmbclusters / 32; 269 maxfragsperpacket = 16; 270 271#ifndef RANDOM_IP_ID 272 ip_id = time_second & 0xffff; 273#endif 274 ipintrq.ifq_maxlen = ipqmaxlen; 275 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 276 netisr_register(NETISR_IP, ip_input, &ipintrq); 277} 278 279/* 280 * XXX watch out this one. It is perhaps used as a cache for 281 * the most recently used route ? it is cleared in in_addroute() 282 * when a new route is successfully created. 283 */ 284struct route ipforward_rt; 285 286/* 287 * Ip input routine. Checksum and byte swap header. If fragmented 288 * try to reassemble. Process options. Pass to next level. 289 */ 290void 291ip_input(struct mbuf *m) 292{ 293 struct ip *ip; 294 struct ipq *fp; 295 struct in_ifaddr *ia = NULL; 296 struct ifaddr *ifa; 297 int i, hlen, checkif; 298 u_short sum; 299 struct in_addr pkt_dst; 300 u_int32_t divert_info = 0; /* packet divert/tee info */ 301 struct ip_fw_args args; 302#ifdef PFIL_HOOKS 303 struct packet_filter_hook *pfh; 304 struct mbuf *m0; 305 int rv; 306#endif /* PFIL_HOOKS */ 307#ifdef FAST_IPSEC 308 struct m_tag *mtag; 309 struct tdb_ident *tdbi; 310 struct secpolicy *sp; 311 int s, error; 312#endif /* FAST_IPSEC */ 313 314 args.eh = NULL; 315 args.oif = NULL; 316 args.rule = NULL; 317 args.divert_rule = 0; /* divert cookie */ 318 args.next_hop = NULL; 319 320 /* Grab info from MT_TAG mbufs prepended to the chain. */ 321 for (; m && m->m_type == MT_TAG; m = m->m_next) { 322 switch(m->_m_tag_id) { 323 default: 324 printf("ip_input: unrecognised MT_TAG tag %d\n", 325 m->_m_tag_id); 326 break; 327 328 case PACKET_TAG_DUMMYNET: 329 args.rule = ((struct dn_pkt *)m)->rule; 330 break; 331 332 case PACKET_TAG_DIVERT: 333 args.divert_rule = (intptr_t)m->m_hdr.mh_data & 0xffff; 334 break; 335 336 case PACKET_TAG_IPFORWARD: 337 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data; 338 break; 339 } 340 } 341 342 KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0, 343 ("ip_input: no HDR")); 344 345 if (args.rule) { /* dummynet already filtered us */ 346 ip = mtod(m, struct ip *); 347 hlen = ip->ip_hl << 2; 348 goto iphack ; 349 } 350 351 ipstat.ips_total++; 352 353 if (m->m_pkthdr.len < sizeof(struct ip)) 354 goto tooshort; 355 356 if (m->m_len < sizeof (struct ip) && 357 (m = m_pullup(m, sizeof (struct ip))) == 0) { 358 ipstat.ips_toosmall++; 359 return; 360 } 361 ip = mtod(m, struct ip *); 362 363 if (ip->ip_v != IPVERSION) { 364 ipstat.ips_badvers++; 365 goto bad; 366 } 367 368 hlen = ip->ip_hl << 2; 369 if (hlen < sizeof(struct ip)) { /* minimum header length */ 370 ipstat.ips_badhlen++; 371 goto bad; 372 } 373 if (hlen > m->m_len) { 374 if ((m = m_pullup(m, hlen)) == 0) { 375 ipstat.ips_badhlen++; 376 return; 377 } 378 ip = mtod(m, struct ip *); 379 } 380 381 /* 127/8 must not appear on wire - RFC1122 */ 382 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 383 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 384 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 385 ipstat.ips_badaddr++; 386 goto bad; 387 } 388 } 389 390 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 391 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 392 } else { 393 if (hlen == sizeof(struct ip)) { 394 sum = in_cksum_hdr(ip); 395 } else { 396 sum = in_cksum(m, hlen); 397 } 398 } 399 if (sum) { 400 ipstat.ips_badsum++; 401 goto bad; 402 } 403 404 /* 405 * Convert fields to host representation. 406 */ 407 ip->ip_len = ntohs(ip->ip_len); 408 if (ip->ip_len < hlen) { 409 ipstat.ips_badlen++; 410 goto bad; 411 } 412 ip->ip_off = ntohs(ip->ip_off); 413 414 /* 415 * Check for RFC3514 (EVIL) packets. 416 */ 417 if (ip->ip_off & IP_EF) { 418 ipstat.ips_evil++; 419 if (hear_no_evil) 420 goto bad; 421 } 422 423 /* 424 * Check that the amount of data in the buffers 425 * is as at least much as the IP header would have us expect. 426 * Trim mbufs if longer than we expect. 427 * Drop packet if shorter than we expect. 428 */ 429 if (m->m_pkthdr.len < ip->ip_len) { 430tooshort: 431 ipstat.ips_tooshort++; 432 goto bad; 433 } 434 if (m->m_pkthdr.len > ip->ip_len) { 435 if (m->m_len == m->m_pkthdr.len) { 436 m->m_len = ip->ip_len; 437 m->m_pkthdr.len = ip->ip_len; 438 } else 439 m_adj(m, ip->ip_len - m->m_pkthdr.len); 440 } 441#if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 442 /* 443 * Bypass packet filtering for packets from a tunnel (gif). 444 */ 445 if (ipsec_gethist(m, NULL)) 446 goto pass; 447#endif 448 449 /* 450 * IpHack's section. 451 * Right now when no processing on packet has done 452 * and it is still fresh out of network we do our black 453 * deals with it. 454 * - Firewall: deny/allow/divert 455 * - Xlate: translate packet's addr/port (NAT). 456 * - Pipe: pass pkt through dummynet. 457 * - Wrap: fake packet's addr/port <unimpl.> 458 * - Encapsulate: put it in another IP and send out. <unimp.> 459 */ 460 461iphack: 462 463#ifdef PFIL_HOOKS 464 /* 465 * Run through list of hooks for input packets. If there are any 466 * filters which require that additional packets in the flow are 467 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 468 * Note that filters must _never_ set this flag, as another filter 469 * in the list may have previously cleared it. 470 */ 471 m0 = m; 472 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 473 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 474 if (pfh->pfil_func) { 475 rv = pfh->pfil_func(ip, hlen, 476 m->m_pkthdr.rcvif, 0, &m0); 477 if (rv) 478 return; 479 m = m0; 480 if (m == NULL) 481 return; 482 ip = mtod(m, struct ip *); 483 } 484#endif /* PFIL_HOOKS */ 485 486 if (fw_enable && IPFW_LOADED) { 487 /* 488 * If we've been forwarded from the output side, then 489 * skip the firewall a second time 490 */ 491 if (args.next_hop) 492 goto ours; 493 494 args.m = m; 495 i = ip_fw_chk_ptr(&args); 496 m = args.m; 497 498 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */ 499 if (m) 500 m_freem(m); 501 return; 502 } 503 ip = mtod(m, struct ip *); /* just in case m changed */ 504 if (i == 0 && args.next_hop == NULL) /* common case */ 505 goto pass; 506 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) { 507 /* Send packet to the appropriate pipe */ 508 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args); 509 return; 510 } 511#ifdef IPDIVERT 512 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 513 /* Divert or tee packet */ 514 divert_info = i; 515 goto ours; 516 } 517#endif 518 if (i == 0 && args.next_hop != NULL) 519 goto pass; 520 /* 521 * if we get here, the packet must be dropped 522 */ 523 m_freem(m); 524 return; 525 } 526pass: 527 528 /* 529 * Process options and, if not destined for us, 530 * ship it on. ip_dooptions returns 1 when an 531 * error was detected (causing an icmp message 532 * to be sent and the original packet to be freed). 533 */ 534 ip_nhops = 0; /* for source routed packets */ 535 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop)) 536 return; 537 538 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 539 * matter if it is destined to another node, or whether it is 540 * a multicast one, RSVP wants it! and prevents it from being forwarded 541 * anywhere else. Also checks if the rsvp daemon is running before 542 * grabbing the packet. 543 */ 544 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 545 goto ours; 546 547 /* 548 * Check our list of addresses, to see if the packet is for us. 549 * If we don't have any addresses, assume any unicast packet 550 * we receive might be for us (and let the upper layers deal 551 * with it). 552 */ 553 if (TAILQ_EMPTY(&in_ifaddrhead) && 554 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 555 goto ours; 556 557 /* 558 * Cache the destination address of the packet; this may be 559 * changed by use of 'ipfw fwd'. 560 */ 561 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 562 563 /* 564 * Enable a consistency check between the destination address 565 * and the arrival interface for a unicast packet (the RFC 1122 566 * strong ES model) if IP forwarding is disabled and the packet 567 * is not locally generated and the packet is not subject to 568 * 'ipfw fwd'. 569 * 570 * XXX - Checking also should be disabled if the destination 571 * address is ipnat'ed to a different interface. 572 * 573 * XXX - Checking is incompatible with IP aliases added 574 * to the loopback interface instead of the interface where 575 * the packets are received. 576 */ 577 checkif = ip_checkinterface && (ipforwarding == 0) && 578 m->m_pkthdr.rcvif != NULL && 579 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 580 (args.next_hop == NULL); 581 582 /* 583 * Check for exact addresses in the hash bucket. 584 */ 585 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 586 /* 587 * If the address matches, verify that the packet 588 * arrived via the correct interface if checking is 589 * enabled. 590 */ 591 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 592 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 593 goto ours; 594 } 595 /* 596 * Check for broadcast addresses. 597 * 598 * Only accept broadcast packets that arrive via the matching 599 * interface. Reception of forwarded directed broadcasts would 600 * be handled via ip_forward() and ether_output() with the loopback 601 * into the stack for SIMPLEX interfaces handled by ether_output(). 602 */ 603 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 604 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 605 if (ifa->ifa_addr->sa_family != AF_INET) 606 continue; 607 ia = ifatoia(ifa); 608 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 609 pkt_dst.s_addr) 610 goto ours; 611 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 612 goto ours; 613#ifdef BOOTP_COMPAT 614 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 615 goto ours; 616#endif 617 } 618 } 619 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 620 struct in_multi *inm; 621 if (ip_mrouter) { 622 /* 623 * If we are acting as a multicast router, all 624 * incoming multicast packets are passed to the 625 * kernel-level multicast forwarding function. 626 * The packet is returned (relatively) intact; if 627 * ip_mforward() returns a non-zero value, the packet 628 * must be discarded, else it may be accepted below. 629 */ 630 if (ip_mforward && 631 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 632 ipstat.ips_cantforward++; 633 m_freem(m); 634 return; 635 } 636 637 /* 638 * The process-level routing daemon needs to receive 639 * all multicast IGMP packets, whether or not this 640 * host belongs to their destination groups. 641 */ 642 if (ip->ip_p == IPPROTO_IGMP) 643 goto ours; 644 ipstat.ips_forward++; 645 } 646 /* 647 * See if we belong to the destination multicast group on the 648 * arrival interface. 649 */ 650 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 651 if (inm == NULL) { 652 ipstat.ips_notmember++; 653 m_freem(m); 654 return; 655 } 656 goto ours; 657 } 658 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 659 goto ours; 660 if (ip->ip_dst.s_addr == INADDR_ANY) 661 goto ours; 662 663 /* 664 * FAITH(Firewall Aided Internet Translator) 665 */ 666 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 667 if (ip_keepfaith) { 668 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 669 goto ours; 670 } 671 m_freem(m); 672 return; 673 } 674 675 /* 676 * Not for us; forward if possible and desirable. 677 */ 678 if (ipforwarding == 0) { 679 ipstat.ips_cantforward++; 680 m_freem(m); 681 } else { 682#ifdef IPSEC 683 /* 684 * Enforce inbound IPsec SPD. 685 */ 686 if (ipsec4_in_reject(m, NULL)) { 687 ipsecstat.in_polvio++; 688 goto bad; 689 } 690#endif /* IPSEC */ 691#ifdef FAST_IPSEC 692 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 693 s = splnet(); 694 if (mtag != NULL) { 695 tdbi = (struct tdb_ident *)(mtag + 1); 696 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 697 } else { 698 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 699 IP_FORWARDING, &error); 700 } 701 if (sp == NULL) { /* NB: can happen if error */ 702 splx(s); 703 /*XXX error stat???*/ 704 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 705 goto bad; 706 } 707 708 /* 709 * Check security policy against packet attributes. 710 */ 711 error = ipsec_in_reject(sp, m); 712 KEY_FREESP(&sp); 713 splx(s); 714 if (error) { 715 ipstat.ips_cantforward++; 716 goto bad; 717 } 718#endif /* FAST_IPSEC */ 719 ip_forward(m, 0, args.next_hop); 720 } 721 return; 722 723ours: 724#ifdef IPSTEALTH 725 /* 726 * IPSTEALTH: Process non-routing options only 727 * if the packet is destined for us. 728 */ 729 if (ipstealth && hlen > sizeof (struct ip) && 730 ip_dooptions(m, 1, args.next_hop)) 731 return; 732#endif /* IPSTEALTH */ 733 734 /* Count the packet in the ip address stats */ 735 if (ia != NULL) { 736 ia->ia_ifa.if_ipackets++; 737 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 738 } 739 740 /* 741 * If offset or IP_MF are set, must reassemble. 742 * Otherwise, nothing need be done. 743 * (We could look in the reassembly queue to see 744 * if the packet was previously fragmented, 745 * but it's not worth the time; just let them time out.) 746 */ 747 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 748 749 /* If maxnipq is 0, never accept fragments. */ 750 if (maxnipq == 0) { 751 ipstat.ips_fragments++; 752 ipstat.ips_fragdropped++; 753 goto bad; 754 } 755 756 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 757 /* 758 * Look for queue of fragments 759 * of this datagram. 760 */ 761 TAILQ_FOREACH(fp, &ipq[sum], ipq_list) 762 if (ip->ip_id == fp->ipq_id && 763 ip->ip_src.s_addr == fp->ipq_src.s_addr && 764 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 765#ifdef MAC 766 mac_fragment_match(m, fp) && 767#endif 768 ip->ip_p == fp->ipq_p) 769 goto found; 770 771 fp = 0; 772 773 /* 774 * Enforce upper bound on number of fragmented packets 775 * for which we attempt reassembly; 776 * If maxnipq is -1, accept all fragments without limitation. 777 */ 778 if ((nipq > maxnipq) && (maxnipq > 0)) { 779 /* 780 * drop something from the tail of the current queue 781 * before proceeding further 782 */ 783 struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead); 784 if (q == NULL) { /* gak */ 785 for (i = 0; i < IPREASS_NHASH; i++) { 786 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 787 if (r) { 788 ipstat.ips_fragtimeout += r->ipq_nfrags; 789 ip_freef(&ipq[i], r); 790 break; 791 } 792 } 793 } else { 794 ipstat.ips_fragtimeout += q->ipq_nfrags; 795 ip_freef(&ipq[sum], q); 796 } 797 } 798found: 799 /* 800 * Adjust ip_len to not reflect header, 801 * convert offset of this to bytes. 802 */ 803 ip->ip_len -= hlen; 804 if (ip->ip_off & IP_MF) { 805 /* 806 * Make sure that fragments have a data length 807 * that's a non-zero multiple of 8 bytes. 808 */ 809 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 810 ipstat.ips_toosmall++; /* XXX */ 811 goto bad; 812 } 813 m->m_flags |= M_FRAG; 814 } else 815 m->m_flags &= ~M_FRAG; 816 ip->ip_off <<= 3; 817 818 /* 819 * Attempt reassembly; if it succeeds, proceed. 820 * ip_reass() will return a different mbuf, and update 821 * the divert info in divert_info and args.divert_rule. 822 */ 823 ipstat.ips_fragments++; 824 m->m_pkthdr.header = ip; 825 m = ip_reass(m, 826 &ipq[sum], fp, &divert_info, &args.divert_rule); 827 if (m == 0) 828 return; 829 ipstat.ips_reassembled++; 830 ip = mtod(m, struct ip *); 831 /* Get the header length of the reassembled packet */ 832 hlen = ip->ip_hl << 2; 833#ifdef IPDIVERT 834 /* Restore original checksum before diverting packet */ 835 if (divert_info != 0) { 836 ip->ip_len += hlen; 837 ip->ip_len = htons(ip->ip_len); 838 ip->ip_off = htons(ip->ip_off); 839 ip->ip_sum = 0; 840 if (hlen == sizeof(struct ip)) 841 ip->ip_sum = in_cksum_hdr(ip); 842 else 843 ip->ip_sum = in_cksum(m, hlen); 844 ip->ip_off = ntohs(ip->ip_off); 845 ip->ip_len = ntohs(ip->ip_len); 846 ip->ip_len -= hlen; 847 } 848#endif 849 } else 850 ip->ip_len -= hlen; 851 852#ifdef IPDIVERT 853 /* 854 * Divert or tee packet to the divert protocol if required. 855 */ 856 if (divert_info != 0) { 857 struct mbuf *clone = NULL; 858 859 /* Clone packet if we're doing a 'tee' */ 860 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 861 clone = m_dup(m, M_DONTWAIT); 862 863 /* Restore packet header fields to original values */ 864 ip->ip_len += hlen; 865 ip->ip_len = htons(ip->ip_len); 866 ip->ip_off = htons(ip->ip_off); 867 868 /* Deliver packet to divert input routine */ 869 divert_packet(m, 1, divert_info & 0xffff, args.divert_rule); 870 ipstat.ips_delivered++; 871 872 /* If 'tee', continue with original packet */ 873 if (clone == NULL) 874 return; 875 m = clone; 876 ip = mtod(m, struct ip *); 877 ip->ip_len += hlen; 878 /* 879 * Jump backwards to complete processing of the 880 * packet. But first clear divert_info to avoid 881 * entering this block again. 882 * We do not need to clear args.divert_rule 883 * or args.next_hop as they will not be used. 884 */ 885 divert_info = 0; 886 goto pass; 887 } 888#endif 889 890#ifdef IPSEC 891 /* 892 * enforce IPsec policy checking if we are seeing last header. 893 * note that we do not visit this with protocols with pcb layer 894 * code - like udp/tcp/raw ip. 895 */ 896 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 897 ipsec4_in_reject(m, NULL)) { 898 ipsecstat.in_polvio++; 899 goto bad; 900 } 901#endif 902#if FAST_IPSEC 903 /* 904 * enforce IPsec policy checking if we are seeing last header. 905 * note that we do not visit this with protocols with pcb layer 906 * code - like udp/tcp/raw ip. 907 */ 908 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 909 /* 910 * Check if the packet has already had IPsec processing 911 * done. If so, then just pass it along. This tag gets 912 * set during AH, ESP, etc. input handling, before the 913 * packet is returned to the ip input queue for delivery. 914 */ 915 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 916 s = splnet(); 917 if (mtag != NULL) { 918 tdbi = (struct tdb_ident *)(mtag + 1); 919 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 920 } else { 921 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 922 IP_FORWARDING, &error); 923 } 924 if (sp != NULL) { 925 /* 926 * Check security policy against packet attributes. 927 */ 928 error = ipsec_in_reject(sp, m); 929 KEY_FREESP(&sp); 930 } else { 931 /* XXX error stat??? */ 932 error = EINVAL; 933DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 934 goto bad; 935 } 936 splx(s); 937 if (error) 938 goto bad; 939 } 940#endif /* FAST_IPSEC */ 941 942 /* 943 * Switch out to protocol's input routine. 944 */ 945 ipstat.ips_delivered++; 946 if (args.next_hop && ip->ip_p == IPPROTO_TCP) { 947 /* TCP needs IPFORWARD info if available */ 948 struct m_hdr tag; 949 950 tag.mh_type = MT_TAG; 951 tag.mh_flags = PACKET_TAG_IPFORWARD; 952 tag.mh_data = (caddr_t)args.next_hop; 953 tag.mh_next = m; 954 955 (*inetsw[ip_protox[ip->ip_p]].pr_input)( 956 (struct mbuf *)&tag, hlen); 957 } else 958 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 959 return; 960bad: 961 m_freem(m); 962} 963 964/* 965 * Take incoming datagram fragment and try to reassemble it into 966 * whole datagram. If a chain for reassembly of this datagram already 967 * exists, then it is given as fp; otherwise have to make a chain. 968 * 969 * When IPDIVERT enabled, keep additional state with each packet that 970 * tells us if we need to divert or tee the packet we're building. 971 * In particular, *divinfo includes the port and TEE flag, 972 * *divert_rule is the number of the matching rule. 973 */ 974 975static struct mbuf * 976ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp, 977 u_int32_t *divinfo, u_int16_t *divert_rule) 978{ 979 struct ip *ip = mtod(m, struct ip *); 980 register struct mbuf *p, *q, *nq; 981 struct mbuf *t; 982 int hlen = ip->ip_hl << 2; 983 int i, next; 984 985 /* 986 * Presence of header sizes in mbufs 987 * would confuse code below. 988 */ 989 m->m_data += hlen; 990 m->m_len -= hlen; 991 992 /* 993 * If first fragment to arrive, create a reassembly queue. 994 */ 995 if (fp == 0) { 996 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 997 goto dropfrag; 998 fp = mtod(t, struct ipq *); 999#ifdef MAC 1000 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 1001 m_free(t); 1002 goto dropfrag; 1003 } 1004 mac_create_ipq(m, fp); 1005#endif 1006 TAILQ_INSERT_HEAD(head, fp, ipq_list); 1007 nipq++; 1008 fp->ipq_nfrags = 1; 1009 fp->ipq_ttl = IPFRAGTTL; 1010 fp->ipq_p = ip->ip_p; 1011 fp->ipq_id = ip->ip_id; 1012 fp->ipq_src = ip->ip_src; 1013 fp->ipq_dst = ip->ip_dst; 1014 fp->ipq_frags = m; 1015 m->m_nextpkt = NULL; 1016#ifdef IPDIVERT 1017 fp->ipq_div_info = 0; 1018 fp->ipq_div_cookie = 0; 1019#endif 1020 goto inserted; 1021 } else { 1022 fp->ipq_nfrags++; 1023#ifdef MAC 1024 mac_update_ipq(m, fp); 1025#endif 1026 } 1027 1028#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1029 1030 /* 1031 * Find a segment which begins after this one does. 1032 */ 1033 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1034 if (GETIP(q)->ip_off > ip->ip_off) 1035 break; 1036 1037 /* 1038 * If there is a preceding segment, it may provide some of 1039 * our data already. If so, drop the data from the incoming 1040 * segment. If it provides all of our data, drop us, otherwise 1041 * stick new segment in the proper place. 1042 * 1043 * If some of the data is dropped from the the preceding 1044 * segment, then it's checksum is invalidated. 1045 */ 1046 if (p) { 1047 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1048 if (i > 0) { 1049 if (i >= ip->ip_len) 1050 goto dropfrag; 1051 m_adj(m, i); 1052 m->m_pkthdr.csum_flags = 0; 1053 ip->ip_off += i; 1054 ip->ip_len -= i; 1055 } 1056 m->m_nextpkt = p->m_nextpkt; 1057 p->m_nextpkt = m; 1058 } else { 1059 m->m_nextpkt = fp->ipq_frags; 1060 fp->ipq_frags = m; 1061 } 1062 1063 /* 1064 * While we overlap succeeding segments trim them or, 1065 * if they are completely covered, dequeue them. 1066 */ 1067 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1068 q = nq) { 1069 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1070 if (i < GETIP(q)->ip_len) { 1071 GETIP(q)->ip_len -= i; 1072 GETIP(q)->ip_off += i; 1073 m_adj(q, i); 1074 q->m_pkthdr.csum_flags = 0; 1075 break; 1076 } 1077 nq = q->m_nextpkt; 1078 m->m_nextpkt = nq; 1079 ipstat.ips_fragdropped++; 1080 fp->ipq_nfrags--; 1081 m_freem(q); 1082 } 1083 1084inserted: 1085 1086#ifdef IPDIVERT 1087 /* 1088 * Transfer firewall instructions to the fragment structure. 1089 * Only trust info in the fragment at offset 0. 1090 */ 1091 if (ip->ip_off == 0) { 1092 fp->ipq_div_info = *divinfo; 1093 fp->ipq_div_cookie = *divert_rule; 1094 } 1095 *divinfo = 0; 1096 *divert_rule = 0; 1097#endif 1098 1099 /* 1100 * Check for complete reassembly and perform frag per packet 1101 * limiting. 1102 * 1103 * Frag limiting is performed here so that the nth frag has 1104 * a chance to complete the packet before we drop the packet. 1105 * As a result, n+1 frags are actually allowed per packet, but 1106 * only n will ever be stored. (n = maxfragsperpacket.) 1107 * 1108 */ 1109 next = 0; 1110 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1111 if (GETIP(q)->ip_off != next) { 1112 if (fp->ipq_nfrags > maxfragsperpacket) { 1113 ipstat.ips_fragdropped += fp->ipq_nfrags; 1114 ip_freef(head, fp); 1115 } 1116 return (0); 1117 } 1118 next += GETIP(q)->ip_len; 1119 } 1120 /* Make sure the last packet didn't have the IP_MF flag */ 1121 if (p->m_flags & M_FRAG) { 1122 if (fp->ipq_nfrags > maxfragsperpacket) { 1123 ipstat.ips_fragdropped += fp->ipq_nfrags; 1124 ip_freef(head, fp); 1125 } 1126 return (0); 1127 } 1128 1129 /* 1130 * Reassembly is complete. Make sure the packet is a sane size. 1131 */ 1132 q = fp->ipq_frags; 1133 ip = GETIP(q); 1134 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1135 ipstat.ips_toolong++; 1136 ipstat.ips_fragdropped += fp->ipq_nfrags; 1137 ip_freef(head, fp); 1138 return (0); 1139 } 1140 1141 /* 1142 * Concatenate fragments. 1143 */ 1144 m = q; 1145 t = m->m_next; 1146 m->m_next = 0; 1147 m_cat(m, t); 1148 nq = q->m_nextpkt; 1149 q->m_nextpkt = 0; 1150 for (q = nq; q != NULL; q = nq) { 1151 nq = q->m_nextpkt; 1152 q->m_nextpkt = NULL; 1153 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1154 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1155 m_cat(m, q); 1156 } 1157#ifdef MAC 1158 mac_create_datagram_from_ipq(fp, m); 1159 mac_destroy_ipq(fp); 1160#endif 1161 1162#ifdef IPDIVERT 1163 /* 1164 * Extract firewall instructions from the fragment structure. 1165 */ 1166 *divinfo = fp->ipq_div_info; 1167 *divert_rule = fp->ipq_div_cookie; 1168#endif 1169 1170 /* 1171 * Create header for new ip packet by 1172 * modifying header of first packet; 1173 * dequeue and discard fragment reassembly header. 1174 * Make header visible. 1175 */ 1176 ip->ip_len = next; 1177 ip->ip_src = fp->ipq_src; 1178 ip->ip_dst = fp->ipq_dst; 1179 TAILQ_REMOVE(head, fp, ipq_list); 1180 nipq--; 1181 (void) m_free(dtom(fp)); 1182 m->m_len += (ip->ip_hl << 2); 1183 m->m_data -= (ip->ip_hl << 2); 1184 /* some debugging cruft by sklower, below, will go away soon */ 1185 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1186 m_fixhdr(m); 1187 return (m); 1188 1189dropfrag: 1190#ifdef IPDIVERT 1191 *divinfo = 0; 1192 *divert_rule = 0; 1193#endif 1194 ipstat.ips_fragdropped++; 1195 if (fp != 0) 1196 fp->ipq_nfrags--; 1197 m_freem(m); 1198 return (0); 1199 1200#undef GETIP 1201} 1202 1203/* 1204 * Free a fragment reassembly header and all 1205 * associated datagrams. 1206 */ 1207static void 1208ip_freef(fhp, fp) 1209 struct ipqhead *fhp; 1210 struct ipq *fp; 1211{ 1212 register struct mbuf *q; 1213 1214 while (fp->ipq_frags) { 1215 q = fp->ipq_frags; 1216 fp->ipq_frags = q->m_nextpkt; 1217 m_freem(q); 1218 } 1219 TAILQ_REMOVE(fhp, fp, ipq_list); 1220 (void) m_free(dtom(fp)); 1221 nipq--; 1222} 1223 1224/* 1225 * IP timer processing; 1226 * if a timer expires on a reassembly 1227 * queue, discard it. 1228 */ 1229void 1230ip_slowtimo() 1231{ 1232 register struct ipq *fp; 1233 int s = splnet(); 1234 int i; 1235 1236 for (i = 0; i < IPREASS_NHASH; i++) { 1237 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1238 struct ipq *fpp; 1239 1240 fpp = fp; 1241 fp = TAILQ_NEXT(fp, ipq_list); 1242 if(--fpp->ipq_ttl == 0) { 1243 ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1244 ip_freef(&ipq[i], fpp); 1245 } 1246 } 1247 } 1248 /* 1249 * If we are over the maximum number of fragments 1250 * (due to the limit being lowered), drain off 1251 * enough to get down to the new limit. 1252 */ 1253 if (maxnipq >= 0 && nipq > maxnipq) { 1254 for (i = 0; i < IPREASS_NHASH; i++) { 1255 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) { 1256 ipstat.ips_fragdropped += 1257 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1258 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1259 } 1260 } 1261 } 1262 ipflow_slowtimo(); 1263 splx(s); 1264} 1265 1266/* 1267 * Drain off all datagram fragments. 1268 */ 1269void 1270ip_drain() 1271{ 1272 int i; 1273 1274 for (i = 0; i < IPREASS_NHASH; i++) { 1275 while(!TAILQ_EMPTY(&ipq[i])) { 1276 ipstat.ips_fragdropped += 1277 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1278 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1279 } 1280 } 1281 in_rtqdrain(); 1282} 1283 1284/* 1285 * Do option processing on a datagram, 1286 * possibly discarding it if bad options are encountered, 1287 * or forwarding it if source-routed. 1288 * The pass argument is used when operating in the IPSTEALTH 1289 * mode to tell what options to process: 1290 * [LS]SRR (pass 0) or the others (pass 1). 1291 * The reason for as many as two passes is that when doing IPSTEALTH, 1292 * non-routing options should be processed only if the packet is for us. 1293 * Returns 1 if packet has been forwarded/freed, 1294 * 0 if the packet should be processed further. 1295 */ 1296static int 1297ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1298{ 1299 struct ip *ip = mtod(m, struct ip *); 1300 u_char *cp; 1301 struct in_ifaddr *ia; 1302 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1303 struct in_addr *sin, dst; 1304 n_time ntime; 1305 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 1306 1307 dst = ip->ip_dst; 1308 cp = (u_char *)(ip + 1); 1309 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1310 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1311 opt = cp[IPOPT_OPTVAL]; 1312 if (opt == IPOPT_EOL) 1313 break; 1314 if (opt == IPOPT_NOP) 1315 optlen = 1; 1316 else { 1317 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1318 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1319 goto bad; 1320 } 1321 optlen = cp[IPOPT_OLEN]; 1322 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1323 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1324 goto bad; 1325 } 1326 } 1327 switch (opt) { 1328 1329 default: 1330 break; 1331 1332 /* 1333 * Source routing with record. 1334 * Find interface with current destination address. 1335 * If none on this machine then drop if strictly routed, 1336 * or do nothing if loosely routed. 1337 * Record interface address and bring up next address 1338 * component. If strictly routed make sure next 1339 * address is on directly accessible net. 1340 */ 1341 case IPOPT_LSRR: 1342 case IPOPT_SSRR: 1343#ifdef IPSTEALTH 1344 if (ipstealth && pass > 0) 1345 break; 1346#endif 1347 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1348 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1349 goto bad; 1350 } 1351 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1352 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1353 goto bad; 1354 } 1355 ipaddr.sin_addr = ip->ip_dst; 1356 ia = (struct in_ifaddr *) 1357 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1358 if (ia == 0) { 1359 if (opt == IPOPT_SSRR) { 1360 type = ICMP_UNREACH; 1361 code = ICMP_UNREACH_SRCFAIL; 1362 goto bad; 1363 } 1364 if (!ip_dosourceroute) 1365 goto nosourcerouting; 1366 /* 1367 * Loose routing, and not at next destination 1368 * yet; nothing to do except forward. 1369 */ 1370 break; 1371 } 1372 off--; /* 0 origin */ 1373 if (off > optlen - (int)sizeof(struct in_addr)) { 1374 /* 1375 * End of source route. Should be for us. 1376 */ 1377 if (!ip_acceptsourceroute) 1378 goto nosourcerouting; 1379 save_rte(cp, ip->ip_src); 1380 break; 1381 } 1382#ifdef IPSTEALTH 1383 if (ipstealth) 1384 goto dropit; 1385#endif 1386 if (!ip_dosourceroute) { 1387 if (ipforwarding) { 1388 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1389 /* 1390 * Acting as a router, so generate ICMP 1391 */ 1392nosourcerouting: 1393 strcpy(buf, inet_ntoa(ip->ip_dst)); 1394 log(LOG_WARNING, 1395 "attempted source route from %s to %s\n", 1396 inet_ntoa(ip->ip_src), buf); 1397 type = ICMP_UNREACH; 1398 code = ICMP_UNREACH_SRCFAIL; 1399 goto bad; 1400 } else { 1401 /* 1402 * Not acting as a router, so silently drop. 1403 */ 1404#ifdef IPSTEALTH 1405dropit: 1406#endif 1407 ipstat.ips_cantforward++; 1408 m_freem(m); 1409 return (1); 1410 } 1411 } 1412 1413 /* 1414 * locate outgoing interface 1415 */ 1416 (void)memcpy(&ipaddr.sin_addr, cp + off, 1417 sizeof(ipaddr.sin_addr)); 1418 1419 if (opt == IPOPT_SSRR) { 1420#define INA struct in_ifaddr * 1421#define SA struct sockaddr * 1422 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1423 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1424 } else 1425 ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt); 1426 if (ia == 0) { 1427 type = ICMP_UNREACH; 1428 code = ICMP_UNREACH_SRCFAIL; 1429 goto bad; 1430 } 1431 ip->ip_dst = ipaddr.sin_addr; 1432 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1433 sizeof(struct in_addr)); 1434 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1435 /* 1436 * Let ip_intr's mcast routing check handle mcast pkts 1437 */ 1438 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1439 break; 1440 1441 case IPOPT_RR: 1442#ifdef IPSTEALTH 1443 if (ipstealth && pass == 0) 1444 break; 1445#endif 1446 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1447 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1448 goto bad; 1449 } 1450 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1451 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1452 goto bad; 1453 } 1454 /* 1455 * If no space remains, ignore. 1456 */ 1457 off--; /* 0 origin */ 1458 if (off > optlen - (int)sizeof(struct in_addr)) 1459 break; 1460 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1461 sizeof(ipaddr.sin_addr)); 1462 /* 1463 * locate outgoing interface; if we're the destination, 1464 * use the incoming interface (should be same). 1465 */ 1466 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1467 (ia = ip_rtaddr(ipaddr.sin_addr, 1468 &ipforward_rt)) == 0) { 1469 type = ICMP_UNREACH; 1470 code = ICMP_UNREACH_HOST; 1471 goto bad; 1472 } 1473 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1474 sizeof(struct in_addr)); 1475 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1476 break; 1477 1478 case IPOPT_TS: 1479#ifdef IPSTEALTH 1480 if (ipstealth && pass == 0) 1481 break; 1482#endif 1483 code = cp - (u_char *)ip; 1484 if (optlen < 4 || optlen > 40) { 1485 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1486 goto bad; 1487 } 1488 if ((off = cp[IPOPT_OFFSET]) < 5) { 1489 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1490 goto bad; 1491 } 1492 if (off > optlen - (int)sizeof(int32_t)) { 1493 cp[IPOPT_OFFSET + 1] += (1 << 4); 1494 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1495 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1496 goto bad; 1497 } 1498 break; 1499 } 1500 off--; /* 0 origin */ 1501 sin = (struct in_addr *)(cp + off); 1502 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1503 1504 case IPOPT_TS_TSONLY: 1505 break; 1506 1507 case IPOPT_TS_TSANDADDR: 1508 if (off + sizeof(n_time) + 1509 sizeof(struct in_addr) > optlen) { 1510 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1511 goto bad; 1512 } 1513 ipaddr.sin_addr = dst; 1514 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1515 m->m_pkthdr.rcvif); 1516 if (ia == 0) 1517 continue; 1518 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1519 sizeof(struct in_addr)); 1520 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1521 off += sizeof(struct in_addr); 1522 break; 1523 1524 case IPOPT_TS_PRESPEC: 1525 if (off + sizeof(n_time) + 1526 sizeof(struct in_addr) > optlen) { 1527 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1528 goto bad; 1529 } 1530 (void)memcpy(&ipaddr.sin_addr, sin, 1531 sizeof(struct in_addr)); 1532 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1533 continue; 1534 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1535 off += sizeof(struct in_addr); 1536 break; 1537 1538 default: 1539 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1540 goto bad; 1541 } 1542 ntime = iptime(); 1543 (void)memcpy(cp + off, &ntime, sizeof(n_time)); 1544 cp[IPOPT_OFFSET] += sizeof(n_time); 1545 } 1546 } 1547 if (forward && ipforwarding) { 1548 ip_forward(m, 1, next_hop); 1549 return (1); 1550 } 1551 return (0); 1552bad: 1553 icmp_error(m, type, code, 0, 0); 1554 ipstat.ips_badoptions++; 1555 return (1); 1556} 1557 1558/* 1559 * Given address of next destination (final or next hop), 1560 * return internet address info of interface to be used to get there. 1561 */ 1562struct in_ifaddr * 1563ip_rtaddr(dst, rt) 1564 struct in_addr dst; 1565 struct route *rt; 1566{ 1567 register struct sockaddr_in *sin; 1568 1569 sin = (struct sockaddr_in *)&rt->ro_dst; 1570 1571 if (rt->ro_rt == 0 || 1572 !(rt->ro_rt->rt_flags & RTF_UP) || 1573 dst.s_addr != sin->sin_addr.s_addr) { 1574 if (rt->ro_rt) { 1575 RTFREE(rt->ro_rt); 1576 rt->ro_rt = 0; 1577 } 1578 sin->sin_family = AF_INET; 1579 sin->sin_len = sizeof(*sin); 1580 sin->sin_addr = dst; 1581 1582 rtalloc_ign(rt, RTF_PRCLONING); 1583 } 1584 if (rt->ro_rt == 0) 1585 return ((struct in_ifaddr *)0); 1586 return (ifatoia(rt->ro_rt->rt_ifa)); 1587} 1588 1589/* 1590 * Save incoming source route for use in replies, 1591 * to be picked up later by ip_srcroute if the receiver is interested. 1592 */ 1593static void 1594save_rte(option, dst) 1595 u_char *option; 1596 struct in_addr dst; 1597{ 1598 unsigned olen; 1599 1600 olen = option[IPOPT_OLEN]; 1601#ifdef DIAGNOSTIC 1602 if (ipprintfs) 1603 printf("save_rte: olen %d\n", olen); 1604#endif 1605 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1606 return; 1607 bcopy(option, ip_srcrt.srcopt, olen); 1608 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1609 ip_srcrt.dst = dst; 1610} 1611 1612/* 1613 * Retrieve incoming source route for use in replies, 1614 * in the same form used by setsockopt. 1615 * The first hop is placed before the options, will be removed later. 1616 */ 1617struct mbuf * 1618ip_srcroute() 1619{ 1620 register struct in_addr *p, *q; 1621 register struct mbuf *m; 1622 1623 if (ip_nhops == 0) 1624 return ((struct mbuf *)0); 1625 m = m_get(M_DONTWAIT, MT_HEADER); 1626 if (m == 0) 1627 return ((struct mbuf *)0); 1628 1629#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1630 1631 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1632 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1633 OPTSIZ; 1634#ifdef DIAGNOSTIC 1635 if (ipprintfs) 1636 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1637#endif 1638 1639 /* 1640 * First save first hop for return route 1641 */ 1642 p = &ip_srcrt.route[ip_nhops - 1]; 1643 *(mtod(m, struct in_addr *)) = *p--; 1644#ifdef DIAGNOSTIC 1645 if (ipprintfs) 1646 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1647#endif 1648 1649 /* 1650 * Copy option fields and padding (nop) to mbuf. 1651 */ 1652 ip_srcrt.nop = IPOPT_NOP; 1653 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1654 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1655 &ip_srcrt.nop, OPTSIZ); 1656 q = (struct in_addr *)(mtod(m, caddr_t) + 1657 sizeof(struct in_addr) + OPTSIZ); 1658#undef OPTSIZ 1659 /* 1660 * Record return path as an IP source route, 1661 * reversing the path (pointers are now aligned). 1662 */ 1663 while (p >= ip_srcrt.route) { 1664#ifdef DIAGNOSTIC 1665 if (ipprintfs) 1666 printf(" %lx", (u_long)ntohl(q->s_addr)); 1667#endif 1668 *q++ = *p--; 1669 } 1670 /* 1671 * Last hop goes to final destination. 1672 */ 1673 *q = ip_srcrt.dst; 1674#ifdef DIAGNOSTIC 1675 if (ipprintfs) 1676 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1677#endif 1678 return (m); 1679} 1680 1681/* 1682 * Strip out IP options, at higher 1683 * level protocol in the kernel. 1684 * Second argument is buffer to which options 1685 * will be moved, and return value is their length. 1686 * XXX should be deleted; last arg currently ignored. 1687 */ 1688void 1689ip_stripoptions(m, mopt) 1690 register struct mbuf *m; 1691 struct mbuf *mopt; 1692{ 1693 register int i; 1694 struct ip *ip = mtod(m, struct ip *); 1695 register caddr_t opts; 1696 int olen; 1697 1698 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1699 opts = (caddr_t)(ip + 1); 1700 i = m->m_len - (sizeof (struct ip) + olen); 1701 bcopy(opts + olen, opts, (unsigned)i); 1702 m->m_len -= olen; 1703 if (m->m_flags & M_PKTHDR) 1704 m->m_pkthdr.len -= olen; 1705 ip->ip_v = IPVERSION; 1706 ip->ip_hl = sizeof(struct ip) >> 2; 1707} 1708 1709u_char inetctlerrmap[PRC_NCMDS] = { 1710 0, 0, 0, 0, 1711 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1712 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1713 EMSGSIZE, EHOSTUNREACH, 0, 0, 1714 0, 0, 0, 0, 1715 ENOPROTOOPT, ECONNREFUSED 1716}; 1717 1718/* 1719 * Forward a packet. If some error occurs return the sender 1720 * an icmp packet. Note we can't always generate a meaningful 1721 * icmp message because icmp doesn't have a large enough repertoire 1722 * of codes and types. 1723 * 1724 * If not forwarding, just drop the packet. This could be confusing 1725 * if ipforwarding was zero but some routing protocol was advancing 1726 * us as a gateway to somewhere. However, we must let the routing 1727 * protocol deal with that. 1728 * 1729 * The srcrt parameter indicates whether the packet is being forwarded 1730 * via a source route. 1731 */ 1732static void 1733ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop) 1734{ 1735 struct ip *ip = mtod(m, struct ip *); 1736 struct rtentry *rt; 1737 int error, type = 0, code = 0; 1738 struct mbuf *mcopy; 1739 n_long dest; 1740 struct in_addr pkt_dst; 1741 struct ifnet *destifp; 1742#if defined(IPSEC) || defined(FAST_IPSEC) 1743 struct ifnet dummyifp; 1744#endif 1745 1746 dest = 0; 1747 /* 1748 * Cache the destination address of the packet; this may be 1749 * changed by use of 'ipfw fwd'. 1750 */ 1751 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 1752 1753#ifdef DIAGNOSTIC 1754 if (ipprintfs) 1755 printf("forward: src %lx dst %lx ttl %x\n", 1756 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr, 1757 ip->ip_ttl); 1758#endif 1759 1760 1761 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) { 1762 ipstat.ips_cantforward++; 1763 m_freem(m); 1764 return; 1765 } 1766#ifdef IPSTEALTH 1767 if (!ipstealth) { 1768#endif 1769 if (ip->ip_ttl <= IPTTLDEC) { 1770 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1771 dest, 0); 1772 return; 1773 } 1774#ifdef IPSTEALTH 1775 } 1776#endif 1777 1778 if (ip_rtaddr(pkt_dst, &ipforward_rt) == 0) { 1779 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1780 return; 1781 } else 1782 rt = ipforward_rt.ro_rt; 1783 1784 /* 1785 * Save the IP header and at most 8 bytes of the payload, 1786 * in case we need to generate an ICMP message to the src. 1787 * 1788 * XXX this can be optimized a lot by saving the data in a local 1789 * buffer on the stack (72 bytes at most), and only allocating the 1790 * mbuf if really necessary. The vast majority of the packets 1791 * are forwarded without having to send an ICMP back (either 1792 * because unnecessary, or because rate limited), so we are 1793 * really we are wasting a lot of work here. 1794 * 1795 * We don't use m_copy() because it might return a reference 1796 * to a shared cluster. Both this function and ip_output() 1797 * assume exclusive access to the IP header in `m', so any 1798 * data in a cluster may change before we reach icmp_error(). 1799 */ 1800 MGET(mcopy, M_DONTWAIT, m->m_type); 1801 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1802 /* 1803 * It's probably ok if the pkthdr dup fails (because 1804 * the deep copy of the tag chain failed), but for now 1805 * be conservative and just discard the copy since 1806 * code below may some day want the tags. 1807 */ 1808 m_free(mcopy); 1809 mcopy = NULL; 1810 } 1811 if (mcopy != NULL) { 1812 mcopy->m_len = imin((ip->ip_hl << 2) + 8, 1813 (int)ip->ip_len); 1814 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1815#ifdef MAC 1816 /* 1817 * XXXMAC: This will eventually become an explicit 1818 * labeling point. 1819 */ 1820 mac_create_mbuf_from_mbuf(m, mcopy); 1821#endif 1822 } 1823 1824#ifdef IPSTEALTH 1825 if (!ipstealth) { 1826#endif 1827 ip->ip_ttl -= IPTTLDEC; 1828#ifdef IPSTEALTH 1829 } 1830#endif 1831 1832 /* 1833 * If forwarding packet using same interface that it came in on, 1834 * perhaps should send a redirect to sender to shortcut a hop. 1835 * Only send redirect if source is sending directly to us, 1836 * and if packet was not source routed (or has any options). 1837 * Also, don't send redirect if forwarding using a default route 1838 * or a route modified by a redirect. 1839 */ 1840 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1841 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1842 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1843 ipsendredirects && !srcrt && !next_hop) { 1844#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1845 u_long src = ntohl(ip->ip_src.s_addr); 1846 1847 if (RTA(rt) && 1848 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1849 if (rt->rt_flags & RTF_GATEWAY) 1850 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1851 else 1852 dest = pkt_dst.s_addr; 1853 /* Router requirements says to only send host redirects */ 1854 type = ICMP_REDIRECT; 1855 code = ICMP_REDIRECT_HOST; 1856#ifdef DIAGNOSTIC 1857 if (ipprintfs) 1858 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1859#endif 1860 } 1861 } 1862 1863 { 1864 struct m_hdr tag; 1865 1866 if (next_hop) { 1867 /* Pass IPFORWARD info if available */ 1868 1869 tag.mh_type = MT_TAG; 1870 tag.mh_flags = PACKET_TAG_IPFORWARD; 1871 tag.mh_data = (caddr_t)next_hop; 1872 tag.mh_next = m; 1873 m = (struct mbuf *)&tag; 1874 } 1875 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1876 IP_FORWARDING, 0, NULL); 1877 } 1878 if (error) 1879 ipstat.ips_cantforward++; 1880 else { 1881 ipstat.ips_forward++; 1882 if (type) 1883 ipstat.ips_redirectsent++; 1884 else { 1885 if (mcopy) { 1886 ipflow_create(&ipforward_rt, mcopy); 1887 m_freem(mcopy); 1888 } 1889 return; 1890 } 1891 } 1892 if (mcopy == NULL) 1893 return; 1894 destifp = NULL; 1895 1896 switch (error) { 1897 1898 case 0: /* forwarded, but need redirect */ 1899 /* type, code set above */ 1900 break; 1901 1902 case ENETUNREACH: /* shouldn't happen, checked above */ 1903 case EHOSTUNREACH: 1904 case ENETDOWN: 1905 case EHOSTDOWN: 1906 default: 1907 type = ICMP_UNREACH; 1908 code = ICMP_UNREACH_HOST; 1909 break; 1910 1911 case EMSGSIZE: 1912 type = ICMP_UNREACH; 1913 code = ICMP_UNREACH_NEEDFRAG; 1914#ifdef IPSEC 1915 /* 1916 * If the packet is routed over IPsec tunnel, tell the 1917 * originator the tunnel MTU. 1918 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1919 * XXX quickhack!!! 1920 */ 1921 if (ipforward_rt.ro_rt) { 1922 struct secpolicy *sp = NULL; 1923 int ipsecerror; 1924 int ipsechdr; 1925 struct route *ro; 1926 1927 sp = ipsec4_getpolicybyaddr(mcopy, 1928 IPSEC_DIR_OUTBOUND, 1929 IP_FORWARDING, 1930 &ipsecerror); 1931 1932 if (sp == NULL) 1933 destifp = ipforward_rt.ro_rt->rt_ifp; 1934 else { 1935 /* count IPsec header size */ 1936 ipsechdr = ipsec4_hdrsiz(mcopy, 1937 IPSEC_DIR_OUTBOUND, 1938 NULL); 1939 1940 /* 1941 * find the correct route for outer IPv4 1942 * header, compute tunnel MTU. 1943 * 1944 * XXX BUG ALERT 1945 * The "dummyifp" code relies upon the fact 1946 * that icmp_error() touches only ifp->if_mtu. 1947 */ 1948 /*XXX*/ 1949 destifp = NULL; 1950 if (sp->req != NULL 1951 && sp->req->sav != NULL 1952 && sp->req->sav->sah != NULL) { 1953 ro = &sp->req->sav->sah->sa_route; 1954 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1955 dummyifp.if_mtu = 1956 ro->ro_rt->rt_ifp->if_mtu; 1957 dummyifp.if_mtu -= ipsechdr; 1958 destifp = &dummyifp; 1959 } 1960 } 1961 1962 key_freesp(sp); 1963 } 1964 } 1965#elif FAST_IPSEC 1966 /* 1967 * If the packet is routed over IPsec tunnel, tell the 1968 * originator the tunnel MTU. 1969 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1970 * XXX quickhack!!! 1971 */ 1972 if (ipforward_rt.ro_rt) { 1973 struct secpolicy *sp = NULL; 1974 int ipsecerror; 1975 int ipsechdr; 1976 struct route *ro; 1977 1978 sp = ipsec_getpolicybyaddr(mcopy, 1979 IPSEC_DIR_OUTBOUND, 1980 IP_FORWARDING, 1981 &ipsecerror); 1982 1983 if (sp == NULL) 1984 destifp = ipforward_rt.ro_rt->rt_ifp; 1985 else { 1986 /* count IPsec header size */ 1987 ipsechdr = ipsec4_hdrsiz(mcopy, 1988 IPSEC_DIR_OUTBOUND, 1989 NULL); 1990 1991 /* 1992 * find the correct route for outer IPv4 1993 * header, compute tunnel MTU. 1994 * 1995 * XXX BUG ALERT 1996 * The "dummyifp" code relies upon the fact 1997 * that icmp_error() touches only ifp->if_mtu. 1998 */ 1999 /*XXX*/ 2000 destifp = NULL; 2001 if (sp->req != NULL 2002 && sp->req->sav != NULL 2003 && sp->req->sav->sah != NULL) { 2004 ro = &sp->req->sav->sah->sa_route; 2005 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 2006 dummyifp.if_mtu = 2007 ro->ro_rt->rt_ifp->if_mtu; 2008 dummyifp.if_mtu -= ipsechdr; 2009 destifp = &dummyifp; 2010 } 2011 } 2012 2013 KEY_FREESP(&sp); 2014 } 2015 } 2016#else /* !IPSEC && !FAST_IPSEC */ 2017 if (ipforward_rt.ro_rt) 2018 destifp = ipforward_rt.ro_rt->rt_ifp; 2019#endif /*IPSEC*/ 2020 ipstat.ips_cantfrag++; 2021 break; 2022 2023 case ENOBUFS: 2024 /* 2025 * A router should not generate ICMP_SOURCEQUENCH as 2026 * required in RFC1812 Requirements for IP Version 4 Routers. 2027 * Source quench could be a big problem under DoS attacks, 2028 * or if the underlying interface is rate-limited. 2029 * Those who need source quench packets may re-enable them 2030 * via the net.inet.ip.sendsourcequench sysctl. 2031 */ 2032 if (ip_sendsourcequench == 0) { 2033 m_freem(mcopy); 2034 return; 2035 } else { 2036 type = ICMP_SOURCEQUENCH; 2037 code = 0; 2038 } 2039 break; 2040 2041 case EACCES: /* ipfw denied packet */ 2042 m_freem(mcopy); 2043 return; 2044 } 2045 icmp_error(mcopy, type, code, dest, destifp); 2046} 2047 2048void 2049ip_savecontrol(inp, mp, ip, m) 2050 register struct inpcb *inp; 2051 register struct mbuf **mp; 2052 register struct ip *ip; 2053 register struct mbuf *m; 2054{ 2055 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2056 struct timeval tv; 2057 2058 microtime(&tv); 2059 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2060 SCM_TIMESTAMP, SOL_SOCKET); 2061 if (*mp) 2062 mp = &(*mp)->m_next; 2063 } 2064 if (inp->inp_flags & INP_RECVDSTADDR) { 2065 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2066 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2067 if (*mp) 2068 mp = &(*mp)->m_next; 2069 } 2070#ifdef notyet 2071 /* XXX 2072 * Moving these out of udp_input() made them even more broken 2073 * than they already were. 2074 */ 2075 /* options were tossed already */ 2076 if (inp->inp_flags & INP_RECVOPTS) { 2077 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2078 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2079 if (*mp) 2080 mp = &(*mp)->m_next; 2081 } 2082 /* ip_srcroute doesn't do what we want here, need to fix */ 2083 if (inp->inp_flags & INP_RECVRETOPTS) { 2084 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2085 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2086 if (*mp) 2087 mp = &(*mp)->m_next; 2088 } 2089#endif 2090 if (inp->inp_flags & INP_RECVIF) { 2091 struct ifnet *ifp; 2092 struct sdlbuf { 2093 struct sockaddr_dl sdl; 2094 u_char pad[32]; 2095 } sdlbuf; 2096 struct sockaddr_dl *sdp; 2097 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2098 2099 if (((ifp = m->m_pkthdr.rcvif)) 2100 && ( ifp->if_index && (ifp->if_index <= if_index))) { 2101 sdp = (struct sockaddr_dl *) 2102 (ifaddr_byindex(ifp->if_index)->ifa_addr); 2103 /* 2104 * Change our mind and don't try copy. 2105 */ 2106 if ((sdp->sdl_family != AF_LINK) 2107 || (sdp->sdl_len > sizeof(sdlbuf))) { 2108 goto makedummy; 2109 } 2110 bcopy(sdp, sdl2, sdp->sdl_len); 2111 } else { 2112makedummy: 2113 sdl2->sdl_len 2114 = offsetof(struct sockaddr_dl, sdl_data[0]); 2115 sdl2->sdl_family = AF_LINK; 2116 sdl2->sdl_index = 0; 2117 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2118 } 2119 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2120 IP_RECVIF, IPPROTO_IP); 2121 if (*mp) 2122 mp = &(*mp)->m_next; 2123 } 2124} 2125 2126/* 2127 * XXX these routines are called from the upper part of the kernel. 2128 * They need to be locked when we remove Giant. 2129 * 2130 * They could also be moved to ip_mroute.c, since all the RSVP 2131 * handling is done there already. 2132 */ 2133static int ip_rsvp_on; 2134struct socket *ip_rsvpd; 2135int 2136ip_rsvp_init(struct socket *so) 2137{ 2138 if (so->so_type != SOCK_RAW || 2139 so->so_proto->pr_protocol != IPPROTO_RSVP) 2140 return EOPNOTSUPP; 2141 2142 if (ip_rsvpd != NULL) 2143 return EADDRINUSE; 2144 2145 ip_rsvpd = so; 2146 /* 2147 * This may seem silly, but we need to be sure we don't over-increment 2148 * the RSVP counter, in case something slips up. 2149 */ 2150 if (!ip_rsvp_on) { 2151 ip_rsvp_on = 1; 2152 rsvp_on++; 2153 } 2154 2155 return 0; 2156} 2157 2158int 2159ip_rsvp_done(void) 2160{ 2161 ip_rsvpd = NULL; 2162 /* 2163 * This may seem silly, but we need to be sure we don't over-decrement 2164 * the RSVP counter, in case something slips up. 2165 */ 2166 if (ip_rsvp_on) { 2167 ip_rsvp_on = 0; 2168 rsvp_on--; 2169 } 2170 return 0; 2171} 2172 2173void 2174rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 2175{ 2176 if (rsvp_input_p) { /* call the real one if loaded */ 2177 rsvp_input_p(m, off); 2178 return; 2179 } 2180 2181 /* Can still get packets with rsvp_on = 0 if there is a local member 2182 * of the group to which the RSVP packet is addressed. But in this 2183 * case we want to throw the packet away. 2184 */ 2185 2186 if (!rsvp_on) { 2187 m_freem(m); 2188 return; 2189 } 2190 2191 if (ip_rsvpd != NULL) { 2192 rip_input(m, off); 2193 return; 2194 } 2195 /* Drop the packet */ 2196 m_freem(m); 2197} 2198