ip_input.c revision 133720
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 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 30 * $FreeBSD: head/sys/netinet/ip_input.c 133720 2004-08-14 15:32:40Z dwmalone $ 31 */ 32 33#include "opt_bootp.h" 34#include "opt_ipfw.h" 35#include "opt_ipdn.h" 36#include "opt_ipdivert.h" 37#include "opt_ipfilter.h" 38#include "opt_ipstealth.h" 39#include "opt_ipsec.h" 40#include "opt_mac.h" 41#include "opt_pfil_hooks.h" 42 43#include <sys/param.h> 44#include <sys/systm.h> 45#include <sys/mac.h> 46#include <sys/mbuf.h> 47#include <sys/malloc.h> 48#include <sys/domain.h> 49#include <sys/protosw.h> 50#include <sys/socket.h> 51#include <sys/time.h> 52#include <sys/kernel.h> 53#include <sys/syslog.h> 54#include <sys/sysctl.h> 55 56#include <net/pfil.h> 57#include <net/if.h> 58#include <net/if_types.h> 59#include <net/if_var.h> 60#include <net/if_dl.h> 61#include <net/route.h> 62#include <net/netisr.h> 63 64#include <netinet/in.h> 65#include <netinet/in_systm.h> 66#include <netinet/in_var.h> 67#include <netinet/ip.h> 68#include <netinet/in_pcb.h> 69#include <netinet/ip_var.h> 70#include <netinet/ip_icmp.h> 71#include <machine/in_cksum.h> 72 73#include <sys/socketvar.h> 74 75#include <netinet/ip_fw.h> 76#include <netinet/ip_divert.h> 77#include <netinet/ip_dummynet.h> 78 79#ifdef IPSEC 80#include <netinet6/ipsec.h> 81#include <netkey/key.h> 82#endif 83 84#ifdef FAST_IPSEC 85#include <netipsec/ipsec.h> 86#include <netipsec/key.h> 87#endif 88 89int rsvp_on = 0; 90 91int ipforwarding = 0; 92SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 93 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 94 95static int ipsendredirects = 1; /* XXX */ 96SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 97 &ipsendredirects, 0, "Enable sending IP redirects"); 98 99int ip_defttl = IPDEFTTL; 100SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 101 &ip_defttl, 0, "Maximum TTL on IP packets"); 102 103static int ip_dosourceroute = 0; 104SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 105 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 106 107static int ip_acceptsourceroute = 0; 108SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 109 CTLFLAG_RW, &ip_acceptsourceroute, 0, 110 "Enable accepting source routed IP packets"); 111 112int ip_doopts = 1; /* 0 = ignore, 1 = process, 2 = reject */ 113SYSCTL_INT(_net_inet_ip, OID_AUTO, process_options, CTLFLAG_RW, 114 &ip_doopts, 0, "Enable IP options processing ([LS]SRR, RR, TS)"); 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 137int ip_do_randomid = 0; 138SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 139 &ip_do_randomid, 0, 140 "Assign random ip_id values"); 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#ifdef PFIL_HOOKS 163struct pfil_head inet_pfil_hook; 164#endif 165 166static struct ifqueue ipintrq; 167static int ipqmaxlen = IFQ_MAXLEN; 168 169extern struct domain inetdomain; 170extern struct protosw inetsw[]; 171u_char ip_protox[IPPROTO_MAX]; 172struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 173struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 174u_long in_ifaddrhmask; /* mask for hash table */ 175 176SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 177 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 178SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 179 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 180 181struct ipstat ipstat; 182SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 183 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 184 185/* Packet reassembly stuff */ 186#define IPREASS_NHASH_LOG2 6 187#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 188#define IPREASS_HMASK (IPREASS_NHASH - 1) 189#define IPREASS_HASH(x,y) \ 190 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 191 192static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 193struct mtx ipqlock; 194 195#define IPQ_LOCK() mtx_lock(&ipqlock) 196#define IPQ_UNLOCK() mtx_unlock(&ipqlock) 197#define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF) 198#define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED) 199 200#ifdef IPCTL_DEFMTU 201SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 202 &ip_mtu, 0, "Default MTU"); 203#endif 204 205#ifdef IPSTEALTH 206int ipstealth = 0; 207SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 208 &ipstealth, 0, ""); 209#endif 210 211 212/* Firewall hooks */ 213ip_fw_chk_t *ip_fw_chk_ptr; 214int fw_enable = 1 ; 215int fw_one_pass = 1; 216 217/* Dummynet hooks */ 218ip_dn_io_t *ip_dn_io_ptr; 219 220/* 221 * XXX this is ugly -- the following two global variables are 222 * used to store packet state while it travels through the stack. 223 * Note that the code even makes assumptions on the size and 224 * alignment of fields inside struct ip_srcrt so e.g. adding some 225 * fields will break the code. This needs to be fixed. 226 * 227 * We need to save the IP options in case a protocol wants to respond 228 * to an incoming packet over the same route if the packet got here 229 * using IP source routing. This allows connection establishment and 230 * maintenance when the remote end is on a network that is not known 231 * to us. 232 * XXX: Broken on SMP and possibly preemption! 233 */ 234static int ip_nhops = 0; 235static struct ip_srcrt { 236 struct in_addr dst; /* final destination */ 237 char nop; /* one NOP to align */ 238 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 239 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 240} ip_srcrt; 241 242static void save_rte(u_char *, struct in_addr); 243static int ip_dooptions(struct mbuf *m, int, 244 struct sockaddr_in *next_hop); 245static void ip_forward(struct mbuf *m, int srcrt, 246 struct sockaddr_in *next_hop); 247static void ip_freef(struct ipqhead *, struct ipq *); 248static struct mbuf *ip_reass(struct mbuf *); 249 250/* 251 * IP initialization: fill in IP protocol switch table. 252 * All protocols not implemented in kernel go to raw IP protocol handler. 253 */ 254void 255ip_init() 256{ 257 register struct protosw *pr; 258 register int i; 259 260 TAILQ_INIT(&in_ifaddrhead); 261 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 262 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 263 if (pr == 0) 264 panic("ip_init"); 265 for (i = 0; i < IPPROTO_MAX; i++) 266 ip_protox[i] = pr - inetsw; 267 for (pr = inetdomain.dom_protosw; 268 pr < inetdomain.dom_protoswNPROTOSW; pr++) 269 if (pr->pr_domain->dom_family == PF_INET && 270 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 271 ip_protox[pr->pr_protocol] = pr - inetsw; 272 273#ifdef PFIL_HOOKS 274 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 275 inet_pfil_hook.ph_af = AF_INET; 276 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) 277 printf("%s: WARNING: unable to register pfil hook, " 278 "error %d\n", __func__, i); 279#endif /* PFIL_HOOKS */ 280 281 IPQ_LOCK_INIT(); 282 for (i = 0; i < IPREASS_NHASH; i++) 283 TAILQ_INIT(&ipq[i]); 284 285 maxnipq = nmbclusters / 32; 286 maxfragsperpacket = 16; 287 288 ip_id = time_second & 0xffff; 289 ipintrq.ifq_maxlen = ipqmaxlen; 290 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF); 291 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE); 292} 293 294/* 295 * Ip input routine. Checksum and byte swap header. If fragmented 296 * try to reassemble. Process options. Pass to next level. 297 */ 298void 299ip_input(struct mbuf *m) 300{ 301 struct ip *ip = NULL; 302 struct in_ifaddr *ia = NULL; 303 struct ifaddr *ifa; 304 int i, checkif, hlen = 0; 305 u_short sum; 306 struct in_addr pkt_dst; 307#ifdef IPDIVERT 308 u_int32_t divert_info; /* packet divert/tee info */ 309#endif 310 struct ip_fw_args args; 311 int dchg = 0; /* dest changed after fw */ 312#ifdef PFIL_HOOKS 313 struct in_addr odst; /* original dst address */ 314#endif 315#ifdef FAST_IPSEC 316 struct m_tag *mtag; 317 struct tdb_ident *tdbi; 318 struct secpolicy *sp; 319 int s, error; 320#endif /* FAST_IPSEC */ 321 322 args.eh = NULL; 323 args.oif = NULL; 324 325 M_ASSERTPKTHDR(m); 326 327 args.next_hop = m_claim_next(m, PACKET_TAG_IPFORWARD); 328 args.rule = ip_dn_claim_rule(m); 329 330 if (m->m_flags & M_FASTFWD_OURS) { 331 /* ip_fastforward firewall changed dest to local */ 332 m->m_flags &= ~M_FASTFWD_OURS; /* for reflected mbufs */ 333 goto ours; 334 } 335 336 if (args.rule) { /* dummynet already filtered us */ 337 ip = mtod(m, struct ip *); 338 hlen = ip->ip_hl << 2; 339 goto iphack ; 340 } 341 342 ipstat.ips_total++; 343 344 if (m->m_pkthdr.len < sizeof(struct ip)) 345 goto tooshort; 346 347 if (m->m_len < sizeof (struct ip) && 348 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 349 ipstat.ips_toosmall++; 350 return; 351 } 352 ip = mtod(m, struct ip *); 353 354 if (ip->ip_v != IPVERSION) { 355 ipstat.ips_badvers++; 356 goto bad; 357 } 358 359 hlen = ip->ip_hl << 2; 360 if (hlen < sizeof(struct ip)) { /* minimum header length */ 361 ipstat.ips_badhlen++; 362 goto bad; 363 } 364 if (hlen > m->m_len) { 365 if ((m = m_pullup(m, hlen)) == NULL) { 366 ipstat.ips_badhlen++; 367 return; 368 } 369 ip = mtod(m, struct ip *); 370 } 371 372 /* 127/8 must not appear on wire - RFC1122 */ 373 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 374 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 375 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 376 ipstat.ips_badaddr++; 377 goto bad; 378 } 379 } 380 381 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 382 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 383 } else { 384 if (hlen == sizeof(struct ip)) { 385 sum = in_cksum_hdr(ip); 386 } else { 387 sum = in_cksum(m, hlen); 388 } 389 } 390 if (sum) { 391 ipstat.ips_badsum++; 392 goto bad; 393 } 394 395#ifdef ALTQ 396 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 397 /* packet is dropped by traffic conditioner */ 398 return; 399#endif 400 401 /* 402 * Convert fields to host representation. 403 */ 404 ip->ip_len = ntohs(ip->ip_len); 405 if (ip->ip_len < hlen) { 406 ipstat.ips_badlen++; 407 goto bad; 408 } 409 ip->ip_off = ntohs(ip->ip_off); 410 411 /* 412 * Check that the amount of data in the buffers 413 * is as at least much as the IP header would have us expect. 414 * Trim mbufs if longer than we expect. 415 * Drop packet if shorter than we expect. 416 */ 417 if (m->m_pkthdr.len < ip->ip_len) { 418tooshort: 419 ipstat.ips_tooshort++; 420 goto bad; 421 } 422 if (m->m_pkthdr.len > ip->ip_len) { 423 if (m->m_len == m->m_pkthdr.len) { 424 m->m_len = ip->ip_len; 425 m->m_pkthdr.len = ip->ip_len; 426 } else 427 m_adj(m, ip->ip_len - m->m_pkthdr.len); 428 } 429#if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 430 /* 431 * Bypass packet filtering for packets from a tunnel (gif). 432 */ 433 if (ipsec_getnhist(m)) 434 goto pass; 435#endif 436#if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF) 437 /* 438 * Bypass packet filtering for packets from a tunnel (gif). 439 */ 440 if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL) 441 goto pass; 442#endif 443 444 /* 445 * IpHack's section. 446 * Right now when no processing on packet has done 447 * and it is still fresh out of network we do our black 448 * deals with it. 449 * - Firewall: deny/allow/divert 450 * - Xlate: translate packet's addr/port (NAT). 451 * - Pipe: pass pkt through dummynet. 452 * - Wrap: fake packet's addr/port <unimpl.> 453 * - Encapsulate: put it in another IP and send out. <unimp.> 454 */ 455 456iphack: 457 458#ifdef PFIL_HOOKS 459 /* 460 * Run through list of hooks for input packets. 461 * 462 * NB: Beware of the destination address changing (e.g. 463 * by NAT rewriting). When this happens, tell 464 * ip_forward to do the right thing. 465 */ 466 odst = ip->ip_dst; 467 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, 468 PFIL_IN) != 0) 469 return; 470 if (m == NULL) /* consumed by filter */ 471 return; 472 ip = mtod(m, struct ip *); 473 dchg = (odst.s_addr != ip->ip_dst.s_addr); 474#endif /* PFIL_HOOKS */ 475 476 if (fw_enable && IPFW_LOADED) { 477 /* 478 * If we've been forwarded from the output side, then 479 * skip the firewall a second time 480 */ 481 if (args.next_hop) 482 goto ours; 483 484 args.m = m; 485 i = ip_fw_chk_ptr(&args); 486 m = args.m; 487 488 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */ 489 if (m) 490 m_freem(m); 491 return; 492 } 493 ip = mtod(m, struct ip *); /* just in case m changed */ 494 if (i == 0 && args.next_hop == NULL) /* common case */ 495 goto pass; 496 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) { 497 /* Send packet to the appropriate pipe */ 498 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args); 499 return; 500 } 501#ifdef IPDIVERT 502 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 503 /* Divert or tee packet */ 504 goto ours; 505 } 506#endif 507 if (i == 0 && args.next_hop != NULL) 508 goto pass; 509 /* 510 * if we get here, the packet must be dropped 511 */ 512 m_freem(m); 513 return; 514 } 515pass: 516 517 /* 518 * Process options and, if not destined for us, 519 * ship it on. ip_dooptions returns 1 when an 520 * error was detected (causing an icmp message 521 * to be sent and the original packet to be freed). 522 */ 523 ip_nhops = 0; /* for source routed packets */ 524 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop)) 525 return; 526 527 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 528 * matter if it is destined to another node, or whether it is 529 * a multicast one, RSVP wants it! and prevents it from being forwarded 530 * anywhere else. Also checks if the rsvp daemon is running before 531 * grabbing the packet. 532 */ 533 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 534 goto ours; 535 536 /* 537 * Check our list of addresses, to see if the packet is for us. 538 * If we don't have any addresses, assume any unicast packet 539 * we receive might be for us (and let the upper layers deal 540 * with it). 541 */ 542 if (TAILQ_EMPTY(&in_ifaddrhead) && 543 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 544 goto ours; 545 546 /* 547 * Cache the destination address of the packet; this may be 548 * changed by use of 'ipfw fwd'. 549 */ 550 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 551 552 /* 553 * Enable a consistency check between the destination address 554 * and the arrival interface for a unicast packet (the RFC 1122 555 * strong ES model) if IP forwarding is disabled and the packet 556 * is not locally generated and the packet is not subject to 557 * 'ipfw fwd'. 558 * 559 * XXX - Checking also should be disabled if the destination 560 * address is ipnat'ed to a different interface. 561 * 562 * XXX - Checking is incompatible with IP aliases added 563 * to the loopback interface instead of the interface where 564 * the packets are received. 565 */ 566 checkif = ip_checkinterface && (ipforwarding == 0) && 567 m->m_pkthdr.rcvif != NULL && 568 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 569 (args.next_hop == NULL) && (dchg == 0); 570 571 /* 572 * Check for exact addresses in the hash bucket. 573 */ 574 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 575 /* 576 * If the address matches, verify that the packet 577 * arrived via the correct interface if checking is 578 * enabled. 579 */ 580 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 581 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 582 goto ours; 583 } 584 /* 585 * Check for broadcast addresses. 586 * 587 * Only accept broadcast packets that arrive via the matching 588 * interface. Reception of forwarded directed broadcasts would 589 * be handled via ip_forward() and ether_output() with the loopback 590 * into the stack for SIMPLEX interfaces handled by ether_output(). 591 */ 592 if (m->m_pkthdr.rcvif != NULL && 593 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 594 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 595 if (ifa->ifa_addr->sa_family != AF_INET) 596 continue; 597 ia = ifatoia(ifa); 598 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 599 pkt_dst.s_addr) 600 goto ours; 601 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 602 goto ours; 603#ifdef BOOTP_COMPAT 604 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 605 goto ours; 606#endif 607 } 608 } 609 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 610 struct in_multi *inm; 611 if (ip_mrouter) { 612 /* 613 * If we are acting as a multicast router, all 614 * incoming multicast packets are passed to the 615 * kernel-level multicast forwarding function. 616 * The packet is returned (relatively) intact; if 617 * ip_mforward() returns a non-zero value, the packet 618 * must be discarded, else it may be accepted below. 619 */ 620 if (ip_mforward && 621 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 622 ipstat.ips_cantforward++; 623 m_freem(m); 624 return; 625 } 626 627 /* 628 * The process-level routing daemon needs to receive 629 * all multicast IGMP packets, whether or not this 630 * host belongs to their destination groups. 631 */ 632 if (ip->ip_p == IPPROTO_IGMP) 633 goto ours; 634 ipstat.ips_forward++; 635 } 636 /* 637 * See if we belong to the destination multicast group on the 638 * arrival interface. 639 */ 640 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 641 if (inm == NULL) { 642 ipstat.ips_notmember++; 643 m_freem(m); 644 return; 645 } 646 goto ours; 647 } 648 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 649 goto ours; 650 if (ip->ip_dst.s_addr == INADDR_ANY) 651 goto ours; 652 653 /* 654 * FAITH(Firewall Aided Internet Translator) 655 */ 656 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 657 if (ip_keepfaith) { 658 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 659 goto ours; 660 } 661 m_freem(m); 662 return; 663 } 664 665 /* 666 * Not for us; forward if possible and desirable. 667 */ 668 if (ipforwarding == 0) { 669 ipstat.ips_cantforward++; 670 m_freem(m); 671 } else { 672#ifdef IPSEC 673 /* 674 * Enforce inbound IPsec SPD. 675 */ 676 if (ipsec4_in_reject(m, NULL)) { 677 ipsecstat.in_polvio++; 678 goto bad; 679 } 680#endif /* IPSEC */ 681#ifdef FAST_IPSEC 682 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 683 s = splnet(); 684 if (mtag != NULL) { 685 tdbi = (struct tdb_ident *)(mtag + 1); 686 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 687 } else { 688 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 689 IP_FORWARDING, &error); 690 } 691 if (sp == NULL) { /* NB: can happen if error */ 692 splx(s); 693 /*XXX error stat???*/ 694 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 695 goto bad; 696 } 697 698 /* 699 * Check security policy against packet attributes. 700 */ 701 error = ipsec_in_reject(sp, m); 702 KEY_FREESP(&sp); 703 splx(s); 704 if (error) { 705 ipstat.ips_cantforward++; 706 goto bad; 707 } 708#endif /* FAST_IPSEC */ 709 ip_forward(m, dchg, args.next_hop); 710 } 711 return; 712 713ours: 714#ifdef IPSTEALTH 715 /* 716 * IPSTEALTH: Process non-routing options only 717 * if the packet is destined for us. 718 */ 719 if (ipstealth && hlen > sizeof (struct ip) && 720 ip_dooptions(m, 1, args.next_hop)) 721 return; 722#endif /* IPSTEALTH */ 723 724 /* Count the packet in the ip address stats */ 725 if (ia != NULL) { 726 ia->ia_ifa.if_ipackets++; 727 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 728 } 729 730 /* 731 * Attempt reassembly; if it succeeds, proceed. 732 * ip_reass() will return a different mbuf. 733 */ 734 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 735 m = ip_reass(m); 736 if (m == NULL) 737 return; 738 ip = mtod(m, struct ip *); 739 /* Get the header length of the reassembled packet */ 740 hlen = ip->ip_hl << 2; 741#ifdef IPDIVERT 742 /* Restore original checksum before diverting packet */ 743 if (divert_find_info(m) != 0) { 744 ip->ip_len = htons(ip->ip_len); 745 ip->ip_off = htons(ip->ip_off); 746 ip->ip_sum = 0; 747 if (hlen == sizeof(struct ip)) 748 ip->ip_sum = in_cksum_hdr(ip); 749 else 750 ip->ip_sum = in_cksum(m, hlen); 751 ip->ip_off = ntohs(ip->ip_off); 752 ip->ip_len = ntohs(ip->ip_len); 753 } 754#endif 755 } 756 757 /* 758 * Further protocols expect the packet length to be w/o the 759 * IP header. 760 */ 761 ip->ip_len -= hlen; 762 763#ifdef IPDIVERT 764 /* 765 * Divert or tee packet to the divert protocol if required. 766 */ 767 divert_info = divert_find_info(m); 768 if (divert_info != 0) { 769 struct mbuf *clone; 770 771 /* Clone packet if we're doing a 'tee' */ 772 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 773 clone = divert_clone(m); 774 else 775 clone = NULL; 776 777 /* Restore packet header fields to original values */ 778 ip->ip_len += hlen; 779 ip->ip_len = htons(ip->ip_len); 780 ip->ip_off = htons(ip->ip_off); 781 782 /* Deliver packet to divert input routine */ 783 divert_packet(m, 1); 784 ipstat.ips_delivered++; 785 786 /* If 'tee', continue with original packet */ 787 if (clone == NULL) 788 return; 789 m = clone; 790 ip = mtod(m, struct ip *); 791 ip->ip_len += hlen; 792 /* 793 * Jump backwards to complete processing of the 794 * packet. We do not need to clear args.next_hop 795 * as that will not be used again and the cloned packet 796 * doesn't contain a divert packet tag so we won't 797 * re-entry this block. 798 */ 799 goto pass; 800 } 801#endif 802 803#ifdef IPSEC 804 /* 805 * enforce IPsec policy checking if we are seeing last header. 806 * note that we do not visit this with protocols with pcb layer 807 * code - like udp/tcp/raw ip. 808 */ 809 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 810 ipsec4_in_reject(m, NULL)) { 811 ipsecstat.in_polvio++; 812 goto bad; 813 } 814#endif 815#if FAST_IPSEC 816 /* 817 * enforce IPsec policy checking if we are seeing last header. 818 * note that we do not visit this with protocols with pcb layer 819 * code - like udp/tcp/raw ip. 820 */ 821 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 822 /* 823 * Check if the packet has already had IPsec processing 824 * done. If so, then just pass it along. This tag gets 825 * set during AH, ESP, etc. input handling, before the 826 * packet is returned to the ip input queue for delivery. 827 */ 828 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 829 s = splnet(); 830 if (mtag != NULL) { 831 tdbi = (struct tdb_ident *)(mtag + 1); 832 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 833 } else { 834 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 835 IP_FORWARDING, &error); 836 } 837 if (sp != NULL) { 838 /* 839 * Check security policy against packet attributes. 840 */ 841 error = ipsec_in_reject(sp, m); 842 KEY_FREESP(&sp); 843 } else { 844 /* XXX error stat??? */ 845 error = EINVAL; 846DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 847 goto bad; 848 } 849 splx(s); 850 if (error) 851 goto bad; 852 } 853#endif /* FAST_IPSEC */ 854 855 /* 856 * Switch out to protocol's input routine. 857 */ 858 ipstat.ips_delivered++; 859 if (args.next_hop && ip->ip_p == IPPROTO_TCP) { 860 /* attach next hop info for TCP */ 861 struct m_tag *mtag = m_tag_get(PACKET_TAG_IPFORWARD, 862 sizeof(struct sockaddr_in *), M_NOWAIT); 863 if (mtag == NULL) 864 goto bad; 865 *(struct sockaddr_in **)(mtag+1) = args.next_hop; 866 m_tag_prepend(m, mtag); 867 } 868 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 869 return; 870bad: 871 m_freem(m); 872} 873 874/* 875 * Take incoming datagram fragment and try to reassemble it into 876 * whole datagram. If the argument is the first fragment or one 877 * in between the function will return NULL and store the mbuf 878 * in the fragment chain. If the argument is the last fragment 879 * the packet will be reassembled and the pointer to the new 880 * mbuf returned for further processing. Only m_tags attached 881 * to the first packet/fragment are preserved. 882 * The IP header is *NOT* adjusted out of iplen. 883 */ 884 885struct mbuf * 886ip_reass(struct mbuf *m) 887{ 888 struct ip *ip; 889 struct mbuf *p, *q, *nq, *t; 890 struct ipq *fp = NULL; 891 struct ipqhead *head; 892 int i, hlen, next; 893 u_int8_t ecn, ecn0; 894 u_short hash; 895 896 /* If maxnipq is 0, never accept fragments. */ 897 if (maxnipq == 0) { 898 ipstat.ips_fragments++; 899 ipstat.ips_fragdropped++; 900 m_freem(m); 901 return (NULL); 902 } 903 904 ip = mtod(m, struct ip *); 905 hlen = ip->ip_hl << 2; 906 907 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 908 head = &ipq[hash]; 909 IPQ_LOCK(); 910 911 /* 912 * Look for queue of fragments 913 * of this datagram. 914 */ 915 TAILQ_FOREACH(fp, head, ipq_list) 916 if (ip->ip_id == fp->ipq_id && 917 ip->ip_src.s_addr == fp->ipq_src.s_addr && 918 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 919#ifdef MAC 920 mac_fragment_match(m, fp) && 921#endif 922 ip->ip_p == fp->ipq_p) 923 goto found; 924 925 fp = NULL; 926 927 /* 928 * Enforce upper bound on number of fragmented packets 929 * for which we attempt reassembly; 930 * If maxnipq is -1, accept all fragments without limitation. 931 */ 932 if ((nipq > maxnipq) && (maxnipq > 0)) { 933 /* 934 * drop something from the tail of the current queue 935 * before proceeding further 936 */ 937 struct ipq *q = TAILQ_LAST(head, ipqhead); 938 if (q == NULL) { /* gak */ 939 for (i = 0; i < IPREASS_NHASH; i++) { 940 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 941 if (r) { 942 ipstat.ips_fragtimeout += r->ipq_nfrags; 943 ip_freef(&ipq[i], r); 944 break; 945 } 946 } 947 } else { 948 ipstat.ips_fragtimeout += q->ipq_nfrags; 949 ip_freef(head, q); 950 } 951 } 952 953found: 954 /* 955 * Adjust ip_len to not reflect header, 956 * convert offset of this to bytes. 957 */ 958 ip->ip_len -= hlen; 959 if (ip->ip_off & IP_MF) { 960 /* 961 * Make sure that fragments have a data length 962 * that's a non-zero multiple of 8 bytes. 963 */ 964 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 965 ipstat.ips_toosmall++; /* XXX */ 966 goto dropfrag; 967 } 968 m->m_flags |= M_FRAG; 969 } else 970 m->m_flags &= ~M_FRAG; 971 ip->ip_off <<= 3; 972 973 974 /* 975 * Attempt reassembly; if it succeeds, proceed. 976 * ip_reass() will return a different mbuf. 977 */ 978 ipstat.ips_fragments++; 979 m->m_pkthdr.header = ip; 980 981 /* Previous ip_reass() started here. */ 982 /* 983 * Presence of header sizes in mbufs 984 * would confuse code below. 985 */ 986 m->m_data += hlen; 987 m->m_len -= hlen; 988 989 /* 990 * If first fragment to arrive, create a reassembly queue. 991 */ 992 if (fp == NULL) { 993 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 994 goto dropfrag; 995 fp = mtod(t, struct ipq *); 996#ifdef MAC 997 if (mac_init_ipq(fp, M_NOWAIT) != 0) { 998 m_free(t); 999 goto dropfrag; 1000 } 1001 mac_create_ipq(m, fp); 1002#endif 1003 TAILQ_INSERT_HEAD(head, fp, ipq_list); 1004 nipq++; 1005 fp->ipq_nfrags = 1; 1006 fp->ipq_ttl = IPFRAGTTL; 1007 fp->ipq_p = ip->ip_p; 1008 fp->ipq_id = ip->ip_id; 1009 fp->ipq_src = ip->ip_src; 1010 fp->ipq_dst = ip->ip_dst; 1011 fp->ipq_frags = m; 1012 m->m_nextpkt = NULL; 1013 goto inserted; 1014 } else { 1015 fp->ipq_nfrags++; 1016#ifdef MAC 1017 mac_update_ipq(m, fp); 1018#endif 1019 } 1020 1021#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1022 1023 /* 1024 * Handle ECN by comparing this segment with the first one; 1025 * if CE is set, do not lose CE. 1026 * drop if CE and not-ECT are mixed for the same packet. 1027 */ 1028 ecn = ip->ip_tos & IPTOS_ECN_MASK; 1029 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 1030 if (ecn == IPTOS_ECN_CE) { 1031 if (ecn0 == IPTOS_ECN_NOTECT) 1032 goto dropfrag; 1033 if (ecn0 != IPTOS_ECN_CE) 1034 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 1035 } 1036 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 1037 goto dropfrag; 1038 1039 /* 1040 * Find a segment which begins after this one does. 1041 */ 1042 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1043 if (GETIP(q)->ip_off > ip->ip_off) 1044 break; 1045 1046 /* 1047 * If there is a preceding segment, it may provide some of 1048 * our data already. If so, drop the data from the incoming 1049 * segment. If it provides all of our data, drop us, otherwise 1050 * stick new segment in the proper place. 1051 * 1052 * If some of the data is dropped from the the preceding 1053 * segment, then it's checksum is invalidated. 1054 */ 1055 if (p) { 1056 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1057 if (i > 0) { 1058 if (i >= ip->ip_len) 1059 goto dropfrag; 1060 m_adj(m, i); 1061 m->m_pkthdr.csum_flags = 0; 1062 ip->ip_off += i; 1063 ip->ip_len -= i; 1064 } 1065 m->m_nextpkt = p->m_nextpkt; 1066 p->m_nextpkt = m; 1067 } else { 1068 m->m_nextpkt = fp->ipq_frags; 1069 fp->ipq_frags = m; 1070 } 1071 1072 /* 1073 * While we overlap succeeding segments trim them or, 1074 * if they are completely covered, dequeue them. 1075 */ 1076 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1077 q = nq) { 1078 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1079 if (i < GETIP(q)->ip_len) { 1080 GETIP(q)->ip_len -= i; 1081 GETIP(q)->ip_off += i; 1082 m_adj(q, i); 1083 q->m_pkthdr.csum_flags = 0; 1084 break; 1085 } 1086 nq = q->m_nextpkt; 1087 m->m_nextpkt = nq; 1088 ipstat.ips_fragdropped++; 1089 fp->ipq_nfrags--; 1090 m_freem(q); 1091 } 1092 1093inserted: 1094 1095#ifdef IPDIVERT 1096 if (ip->ip_off != 0) { 1097 /* 1098 * Strip any divert information; only the info 1099 * on the first fragment is used/kept. 1100 */ 1101 struct m_tag *mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 1102 if (mtag) 1103 m_tag_delete(m, mtag); 1104 } 1105#endif 1106 1107 /* 1108 * Check for complete reassembly and perform frag per packet 1109 * limiting. 1110 * 1111 * Frag limiting is performed here so that the nth frag has 1112 * a chance to complete the packet before we drop the packet. 1113 * As a result, n+1 frags are actually allowed per packet, but 1114 * only n will ever be stored. (n = maxfragsperpacket.) 1115 * 1116 */ 1117 next = 0; 1118 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1119 if (GETIP(q)->ip_off != next) { 1120 if (fp->ipq_nfrags > maxfragsperpacket) { 1121 ipstat.ips_fragdropped += fp->ipq_nfrags; 1122 ip_freef(head, fp); 1123 } 1124 goto done; 1125 } 1126 next += GETIP(q)->ip_len; 1127 } 1128 /* Make sure the last packet didn't have the IP_MF flag */ 1129 if (p->m_flags & M_FRAG) { 1130 if (fp->ipq_nfrags > maxfragsperpacket) { 1131 ipstat.ips_fragdropped += fp->ipq_nfrags; 1132 ip_freef(head, fp); 1133 } 1134 goto done; 1135 } 1136 1137 /* 1138 * Reassembly is complete. Make sure the packet is a sane size. 1139 */ 1140 q = fp->ipq_frags; 1141 ip = GETIP(q); 1142 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) { 1143 ipstat.ips_toolong++; 1144 ipstat.ips_fragdropped += fp->ipq_nfrags; 1145 ip_freef(head, fp); 1146 goto done; 1147 } 1148 1149 /* 1150 * Concatenate fragments. 1151 */ 1152 m = q; 1153 t = m->m_next; 1154 m->m_next = 0; 1155 m_cat(m, t); 1156 nq = q->m_nextpkt; 1157 q->m_nextpkt = 0; 1158 for (q = nq; q != NULL; q = nq) { 1159 nq = q->m_nextpkt; 1160 q->m_nextpkt = NULL; 1161 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1162 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1163 m_cat(m, q); 1164 } 1165#ifdef MAC 1166 mac_create_datagram_from_ipq(fp, m); 1167 mac_destroy_ipq(fp); 1168#endif 1169 1170 /* 1171 * Create header for new ip packet by modifying header of first 1172 * packet; dequeue and discard fragment reassembly header. 1173 * Make header visible. 1174 */ 1175 ip->ip_len = (ip->ip_hl << 2) + next; 1176 ip->ip_src = fp->ipq_src; 1177 ip->ip_dst = fp->ipq_dst; 1178 TAILQ_REMOVE(head, fp, ipq_list); 1179 nipq--; 1180 (void) m_free(dtom(fp)); 1181 m->m_len += (ip->ip_hl << 2); 1182 m->m_data -= (ip->ip_hl << 2); 1183 /* some debugging cruft by sklower, below, will go away soon */ 1184 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */ 1185 m_fixhdr(m); 1186 ipstat.ips_reassembled++; 1187 IPQ_UNLOCK(); 1188 return (m); 1189 1190dropfrag: 1191 ipstat.ips_fragdropped++; 1192 if (fp != NULL) 1193 fp->ipq_nfrags--; 1194 m_freem(m); 1195done: 1196 IPQ_UNLOCK(); 1197 return (NULL); 1198 1199#undef GETIP 1200} 1201 1202/* 1203 * Free a fragment reassembly header and all 1204 * associated datagrams. 1205 */ 1206static void 1207ip_freef(fhp, fp) 1208 struct ipqhead *fhp; 1209 struct ipq *fp; 1210{ 1211 register struct mbuf *q; 1212 1213 IPQ_LOCK_ASSERT(); 1214 1215 while (fp->ipq_frags) { 1216 q = fp->ipq_frags; 1217 fp->ipq_frags = q->m_nextpkt; 1218 m_freem(q); 1219 } 1220 TAILQ_REMOVE(fhp, fp, ipq_list); 1221 (void) m_free(dtom(fp)); 1222 nipq--; 1223} 1224 1225/* 1226 * IP timer processing; 1227 * if a timer expires on a reassembly 1228 * queue, discard it. 1229 */ 1230void 1231ip_slowtimo() 1232{ 1233 register struct ipq *fp; 1234 int s = splnet(); 1235 int i; 1236 1237 IPQ_LOCK(); 1238 for (i = 0; i < IPREASS_NHASH; i++) { 1239 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1240 struct ipq *fpp; 1241 1242 fpp = fp; 1243 fp = TAILQ_NEXT(fp, ipq_list); 1244 if(--fpp->ipq_ttl == 0) { 1245 ipstat.ips_fragtimeout += fpp->ipq_nfrags; 1246 ip_freef(&ipq[i], fpp); 1247 } 1248 } 1249 } 1250 /* 1251 * If we are over the maximum number of fragments 1252 * (due to the limit being lowered), drain off 1253 * enough to get down to the new limit. 1254 */ 1255 if (maxnipq >= 0 && nipq > maxnipq) { 1256 for (i = 0; i < IPREASS_NHASH; i++) { 1257 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) { 1258 ipstat.ips_fragdropped += 1259 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1260 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1261 } 1262 } 1263 } 1264 IPQ_UNLOCK(); 1265 splx(s); 1266} 1267 1268/* 1269 * Drain off all datagram fragments. 1270 */ 1271void 1272ip_drain() 1273{ 1274 int i; 1275 1276 IPQ_LOCK(); 1277 for (i = 0; i < IPREASS_NHASH; i++) { 1278 while(!TAILQ_EMPTY(&ipq[i])) { 1279 ipstat.ips_fragdropped += 1280 TAILQ_FIRST(&ipq[i])->ipq_nfrags; 1281 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1282 } 1283 } 1284 IPQ_UNLOCK(); 1285 in_rtqdrain(); 1286} 1287 1288/* 1289 * Do option processing on a datagram, 1290 * possibly discarding it if bad options are encountered, 1291 * or forwarding it if source-routed. 1292 * The pass argument is used when operating in the IPSTEALTH 1293 * mode to tell what options to process: 1294 * [LS]SRR (pass 0) or the others (pass 1). 1295 * The reason for as many as two passes is that when doing IPSTEALTH, 1296 * non-routing options should be processed only if the packet is for us. 1297 * Returns 1 if packet has been forwarded/freed, 1298 * 0 if the packet should be processed further. 1299 */ 1300static int 1301ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1302{ 1303 struct ip *ip = mtod(m, struct ip *); 1304 u_char *cp; 1305 struct in_ifaddr *ia; 1306 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1307 struct in_addr *sin, dst; 1308 n_time ntime; 1309 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 1310 1311 /* ignore or reject packets with IP options */ 1312 if (ip_doopts == 0) 1313 return 0; 1314 else if (ip_doopts == 2) { 1315 type = ICMP_UNREACH; 1316 code = ICMP_UNREACH_FILTER_PROHIB; 1317 goto bad; 1318 } 1319 1320 dst = ip->ip_dst; 1321 cp = (u_char *)(ip + 1); 1322 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1323 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1324 opt = cp[IPOPT_OPTVAL]; 1325 if (opt == IPOPT_EOL) 1326 break; 1327 if (opt == IPOPT_NOP) 1328 optlen = 1; 1329 else { 1330 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1331 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1332 goto bad; 1333 } 1334 optlen = cp[IPOPT_OLEN]; 1335 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1336 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1337 goto bad; 1338 } 1339 } 1340 switch (opt) { 1341 1342 default: 1343 break; 1344 1345 /* 1346 * Source routing with record. 1347 * Find interface with current destination address. 1348 * If none on this machine then drop if strictly routed, 1349 * or do nothing if loosely routed. 1350 * Record interface address and bring up next address 1351 * component. If strictly routed make sure next 1352 * address is on directly accessible net. 1353 */ 1354 case IPOPT_LSRR: 1355 case IPOPT_SSRR: 1356#ifdef IPSTEALTH 1357 if (ipstealth && pass > 0) 1358 break; 1359#endif 1360 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1361 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1362 goto bad; 1363 } 1364 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1365 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1366 goto bad; 1367 } 1368 ipaddr.sin_addr = ip->ip_dst; 1369 ia = (struct in_ifaddr *) 1370 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1371 if (ia == NULL) { 1372 if (opt == IPOPT_SSRR) { 1373 type = ICMP_UNREACH; 1374 code = ICMP_UNREACH_SRCFAIL; 1375 goto bad; 1376 } 1377 if (!ip_dosourceroute) 1378 goto nosourcerouting; 1379 /* 1380 * Loose routing, and not at next destination 1381 * yet; nothing to do except forward. 1382 */ 1383 break; 1384 } 1385 off--; /* 0 origin */ 1386 if (off > optlen - (int)sizeof(struct in_addr)) { 1387 /* 1388 * End of source route. Should be for us. 1389 */ 1390 if (!ip_acceptsourceroute) 1391 goto nosourcerouting; 1392 save_rte(cp, ip->ip_src); 1393 break; 1394 } 1395#ifdef IPSTEALTH 1396 if (ipstealth) 1397 goto dropit; 1398#endif 1399 if (!ip_dosourceroute) { 1400 if (ipforwarding) { 1401 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1402 /* 1403 * Acting as a router, so generate ICMP 1404 */ 1405nosourcerouting: 1406 strcpy(buf, inet_ntoa(ip->ip_dst)); 1407 log(LOG_WARNING, 1408 "attempted source route from %s to %s\n", 1409 inet_ntoa(ip->ip_src), buf); 1410 type = ICMP_UNREACH; 1411 code = ICMP_UNREACH_SRCFAIL; 1412 goto bad; 1413 } else { 1414 /* 1415 * Not acting as a router, so silently drop. 1416 */ 1417#ifdef IPSTEALTH 1418dropit: 1419#endif 1420 ipstat.ips_cantforward++; 1421 m_freem(m); 1422 return (1); 1423 } 1424 } 1425 1426 /* 1427 * locate outgoing interface 1428 */ 1429 (void)memcpy(&ipaddr.sin_addr, cp + off, 1430 sizeof(ipaddr.sin_addr)); 1431 1432 if (opt == IPOPT_SSRR) { 1433#define INA struct in_ifaddr * 1434#define SA struct sockaddr * 1435 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == NULL) 1436 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1437 } else 1438 ia = ip_rtaddr(ipaddr.sin_addr); 1439 if (ia == NULL) { 1440 type = ICMP_UNREACH; 1441 code = ICMP_UNREACH_SRCFAIL; 1442 goto bad; 1443 } 1444 ip->ip_dst = ipaddr.sin_addr; 1445 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1446 sizeof(struct in_addr)); 1447 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1448 /* 1449 * Let ip_intr's mcast routing check handle mcast pkts 1450 */ 1451 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1452 break; 1453 1454 case IPOPT_RR: 1455#ifdef IPSTEALTH 1456 if (ipstealth && pass == 0) 1457 break; 1458#endif 1459 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1460 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1461 goto bad; 1462 } 1463 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1464 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1465 goto bad; 1466 } 1467 /* 1468 * If no space remains, ignore. 1469 */ 1470 off--; /* 0 origin */ 1471 if (off > optlen - (int)sizeof(struct in_addr)) 1472 break; 1473 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1474 sizeof(ipaddr.sin_addr)); 1475 /* 1476 * locate outgoing interface; if we're the destination, 1477 * use the incoming interface (should be same). 1478 */ 1479 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1480 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 1481 type = ICMP_UNREACH; 1482 code = ICMP_UNREACH_HOST; 1483 goto bad; 1484 } 1485 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1486 sizeof(struct in_addr)); 1487 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1488 break; 1489 1490 case IPOPT_TS: 1491#ifdef IPSTEALTH 1492 if (ipstealth && pass == 0) 1493 break; 1494#endif 1495 code = cp - (u_char *)ip; 1496 if (optlen < 4 || optlen > 40) { 1497 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1498 goto bad; 1499 } 1500 if ((off = cp[IPOPT_OFFSET]) < 5) { 1501 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1502 goto bad; 1503 } 1504 if (off > optlen - (int)sizeof(int32_t)) { 1505 cp[IPOPT_OFFSET + 1] += (1 << 4); 1506 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1507 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1508 goto bad; 1509 } 1510 break; 1511 } 1512 off--; /* 0 origin */ 1513 sin = (struct in_addr *)(cp + off); 1514 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1515 1516 case IPOPT_TS_TSONLY: 1517 break; 1518 1519 case IPOPT_TS_TSANDADDR: 1520 if (off + sizeof(n_time) + 1521 sizeof(struct in_addr) > optlen) { 1522 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1523 goto bad; 1524 } 1525 ipaddr.sin_addr = dst; 1526 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1527 m->m_pkthdr.rcvif); 1528 if (ia == NULL) 1529 continue; 1530 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1531 sizeof(struct in_addr)); 1532 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1533 off += sizeof(struct in_addr); 1534 break; 1535 1536 case IPOPT_TS_PRESPEC: 1537 if (off + sizeof(n_time) + 1538 sizeof(struct in_addr) > optlen) { 1539 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1540 goto bad; 1541 } 1542 (void)memcpy(&ipaddr.sin_addr, sin, 1543 sizeof(struct in_addr)); 1544 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1545 continue; 1546 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1547 off += sizeof(struct in_addr); 1548 break; 1549 1550 default: 1551 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1552 goto bad; 1553 } 1554 ntime = iptime(); 1555 (void)memcpy(cp + off, &ntime, sizeof(n_time)); 1556 cp[IPOPT_OFFSET] += sizeof(n_time); 1557 } 1558 } 1559 if (forward && ipforwarding) { 1560 ip_forward(m, 1, next_hop); 1561 return (1); 1562 } 1563 return (0); 1564bad: 1565 icmp_error(m, type, code, 0, 0); 1566 ipstat.ips_badoptions++; 1567 return (1); 1568} 1569 1570/* 1571 * Given address of next destination (final or next hop), 1572 * return internet address info of interface to be used to get there. 1573 */ 1574struct in_ifaddr * 1575ip_rtaddr(dst) 1576 struct in_addr dst; 1577{ 1578 struct route sro; 1579 struct sockaddr_in *sin; 1580 struct in_ifaddr *ifa; 1581 1582 bzero(&sro, sizeof(sro)); 1583 sin = (struct sockaddr_in *)&sro.ro_dst; 1584 sin->sin_family = AF_INET; 1585 sin->sin_len = sizeof(*sin); 1586 sin->sin_addr = dst; 1587 rtalloc_ign(&sro, RTF_CLONING); 1588 1589 if (sro.ro_rt == NULL) 1590 return ((struct in_ifaddr *)0); 1591 1592 ifa = ifatoia(sro.ro_rt->rt_ifa); 1593 RTFREE(sro.ro_rt); 1594 return ifa; 1595} 1596 1597/* 1598 * Save incoming source route for use in replies, 1599 * to be picked up later by ip_srcroute if the receiver is interested. 1600 */ 1601static void 1602save_rte(option, dst) 1603 u_char *option; 1604 struct in_addr dst; 1605{ 1606 unsigned olen; 1607 1608 olen = option[IPOPT_OLEN]; 1609#ifdef DIAGNOSTIC 1610 if (ipprintfs) 1611 printf("save_rte: olen %d\n", olen); 1612#endif 1613 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1614 return; 1615 bcopy(option, ip_srcrt.srcopt, olen); 1616 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1617 ip_srcrt.dst = dst; 1618} 1619 1620/* 1621 * Retrieve incoming source route for use in replies, 1622 * in the same form used by setsockopt. 1623 * The first hop is placed before the options, will be removed later. 1624 */ 1625struct mbuf * 1626ip_srcroute() 1627{ 1628 register struct in_addr *p, *q; 1629 register struct mbuf *m; 1630 1631 if (ip_nhops == 0) 1632 return ((struct mbuf *)0); 1633 m = m_get(M_DONTWAIT, MT_HEADER); 1634 if (m == NULL) 1635 return ((struct mbuf *)0); 1636 1637#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1638 1639 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1640 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1641 OPTSIZ; 1642#ifdef DIAGNOSTIC 1643 if (ipprintfs) 1644 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1645#endif 1646 1647 /* 1648 * First save first hop for return route 1649 */ 1650 p = &ip_srcrt.route[ip_nhops - 1]; 1651 *(mtod(m, struct in_addr *)) = *p--; 1652#ifdef DIAGNOSTIC 1653 if (ipprintfs) 1654 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1655#endif 1656 1657 /* 1658 * Copy option fields and padding (nop) to mbuf. 1659 */ 1660 ip_srcrt.nop = IPOPT_NOP; 1661 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1662 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1663 &ip_srcrt.nop, OPTSIZ); 1664 q = (struct in_addr *)(mtod(m, caddr_t) + 1665 sizeof(struct in_addr) + OPTSIZ); 1666#undef OPTSIZ 1667 /* 1668 * Record return path as an IP source route, 1669 * reversing the path (pointers are now aligned). 1670 */ 1671 while (p >= ip_srcrt.route) { 1672#ifdef DIAGNOSTIC 1673 if (ipprintfs) 1674 printf(" %lx", (u_long)ntohl(q->s_addr)); 1675#endif 1676 *q++ = *p--; 1677 } 1678 /* 1679 * Last hop goes to final destination. 1680 */ 1681 *q = ip_srcrt.dst; 1682#ifdef DIAGNOSTIC 1683 if (ipprintfs) 1684 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1685#endif 1686 return (m); 1687} 1688 1689/* 1690 * Strip out IP options, at higher 1691 * level protocol in the kernel. 1692 * Second argument is buffer to which options 1693 * will be moved, and return value is their length. 1694 * XXX should be deleted; last arg currently ignored. 1695 */ 1696void 1697ip_stripoptions(m, mopt) 1698 register struct mbuf *m; 1699 struct mbuf *mopt; 1700{ 1701 register int i; 1702 struct ip *ip = mtod(m, struct ip *); 1703 register caddr_t opts; 1704 int olen; 1705 1706 olen = (ip->ip_hl << 2) - sizeof (struct ip); 1707 opts = (caddr_t)(ip + 1); 1708 i = m->m_len - (sizeof (struct ip) + olen); 1709 bcopy(opts + olen, opts, (unsigned)i); 1710 m->m_len -= olen; 1711 if (m->m_flags & M_PKTHDR) 1712 m->m_pkthdr.len -= olen; 1713 ip->ip_v = IPVERSION; 1714 ip->ip_hl = sizeof(struct ip) >> 2; 1715} 1716 1717u_char inetctlerrmap[PRC_NCMDS] = { 1718 0, 0, 0, 0, 1719 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1720 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1721 EMSGSIZE, EHOSTUNREACH, 0, 0, 1722 0, 0, EHOSTUNREACH, 0, 1723 ENOPROTOOPT, ECONNREFUSED 1724}; 1725 1726/* 1727 * Forward a packet. If some error occurs return the sender 1728 * an icmp packet. Note we can't always generate a meaningful 1729 * icmp message because icmp doesn't have a large enough repertoire 1730 * of codes and types. 1731 * 1732 * If not forwarding, just drop the packet. This could be confusing 1733 * if ipforwarding was zero but some routing protocol was advancing 1734 * us as a gateway to somewhere. However, we must let the routing 1735 * protocol deal with that. 1736 * 1737 * The srcrt parameter indicates whether the packet is being forwarded 1738 * via a source route. 1739 */ 1740static void 1741ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop) 1742{ 1743 struct ip *ip = mtod(m, struct ip *); 1744 struct in_ifaddr *ia; 1745 int error, type = 0, code = 0; 1746 struct mbuf *mcopy; 1747 n_long dest; 1748 struct in_addr pkt_dst; 1749 struct ifnet *destifp; 1750#if defined(IPSEC) || defined(FAST_IPSEC) 1751 struct ifnet dummyifp; 1752#endif 1753 1754 /* 1755 * Cache the destination address of the packet; this may be 1756 * changed by use of 'ipfw fwd'. 1757 */ 1758 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 1759 1760#ifdef DIAGNOSTIC 1761 if (ipprintfs) 1762 printf("forward: src %lx dst %lx ttl %x\n", 1763 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr, 1764 ip->ip_ttl); 1765#endif 1766 1767 1768 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) { 1769 ipstat.ips_cantforward++; 1770 m_freem(m); 1771 return; 1772 } 1773#ifdef IPSTEALTH 1774 if (!ipstealth) { 1775#endif 1776 if (ip->ip_ttl <= IPTTLDEC) { 1777 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1778 0, 0); 1779 return; 1780 } 1781#ifdef IPSTEALTH 1782 } 1783#endif 1784 1785 if ((ia = ip_rtaddr(pkt_dst)) == NULL) { 1786 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 1787 return; 1788 } 1789 1790 /* 1791 * Save the IP header and at most 8 bytes of the payload, 1792 * in case we need to generate an ICMP message to the src. 1793 * 1794 * XXX this can be optimized a lot by saving the data in a local 1795 * buffer on the stack (72 bytes at most), and only allocating the 1796 * mbuf if really necessary. The vast majority of the packets 1797 * are forwarded without having to send an ICMP back (either 1798 * because unnecessary, or because rate limited), so we are 1799 * really we are wasting a lot of work here. 1800 * 1801 * We don't use m_copy() because it might return a reference 1802 * to a shared cluster. Both this function and ip_output() 1803 * assume exclusive access to the IP header in `m', so any 1804 * data in a cluster may change before we reach icmp_error(). 1805 */ 1806 MGET(mcopy, M_DONTWAIT, m->m_type); 1807 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) { 1808 /* 1809 * It's probably ok if the pkthdr dup fails (because 1810 * the deep copy of the tag chain failed), but for now 1811 * be conservative and just discard the copy since 1812 * code below may some day want the tags. 1813 */ 1814 m_free(mcopy); 1815 mcopy = NULL; 1816 } 1817 if (mcopy != NULL) { 1818 mcopy->m_len = imin((ip->ip_hl << 2) + 8, 1819 (int)ip->ip_len); 1820 mcopy->m_pkthdr.len = mcopy->m_len; 1821 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 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 dest = 0; 1841 if (ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) { 1842 struct sockaddr_in *sin; 1843 struct route ro; 1844 struct rtentry *rt; 1845 1846 bzero(&ro, sizeof(ro)); 1847 sin = (struct sockaddr_in *)&ro.ro_dst; 1848 sin->sin_family = AF_INET; 1849 sin->sin_len = sizeof(*sin); 1850 sin->sin_addr = pkt_dst; 1851 rtalloc_ign(&ro, RTF_CLONING); 1852 1853 rt = ro.ro_rt; 1854 1855 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1856 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1857 ipsendredirects && !srcrt && !next_hop) { 1858#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1859 u_long src = ntohl(ip->ip_src.s_addr); 1860 1861 if (RTA(rt) && 1862 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1863 if (rt->rt_flags & RTF_GATEWAY) 1864 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1865 else 1866 dest = pkt_dst.s_addr; 1867 /* Router requirements says to only send host redirects */ 1868 type = ICMP_REDIRECT; 1869 code = ICMP_REDIRECT_HOST; 1870#ifdef DIAGNOSTIC 1871 if (ipprintfs) 1872 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1873#endif 1874 } 1875 } 1876 if (rt) 1877 RTFREE(rt); 1878 } 1879 1880 if (next_hop) { 1881 struct m_tag *mtag = m_tag_get(PACKET_TAG_IPFORWARD, 1882 sizeof(struct sockaddr_in *), M_NOWAIT); 1883 if (mtag == NULL) { 1884 m_freem(m); 1885 return; 1886 } 1887 *(struct sockaddr_in **)(mtag+1) = next_hop; 1888 m_tag_prepend(m, mtag); 1889 } 1890 error = ip_output(m, (struct mbuf *)0, NULL, IP_FORWARDING, 0, NULL); 1891 if (error) 1892 ipstat.ips_cantforward++; 1893 else { 1894 ipstat.ips_forward++; 1895 if (type) 1896 ipstat.ips_redirectsent++; 1897 else { 1898 if (mcopy) 1899 m_freem(mcopy); 1900 return; 1901 } 1902 } 1903 if (mcopy == NULL) 1904 return; 1905 destifp = NULL; 1906 1907 switch (error) { 1908 1909 case 0: /* forwarded, but need redirect */ 1910 /* type, code set above */ 1911 break; 1912 1913 case ENETUNREACH: /* shouldn't happen, checked above */ 1914 case EHOSTUNREACH: 1915 case ENETDOWN: 1916 case EHOSTDOWN: 1917 default: 1918 type = ICMP_UNREACH; 1919 code = ICMP_UNREACH_HOST; 1920 break; 1921 1922 case EMSGSIZE: 1923 type = ICMP_UNREACH; 1924 code = ICMP_UNREACH_NEEDFRAG; 1925#if defined(IPSEC) || defined(FAST_IPSEC) 1926 /* 1927 * If the packet is routed over IPsec tunnel, tell the 1928 * originator the tunnel MTU. 1929 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1930 * XXX quickhack!!! 1931 */ 1932 { 1933 struct secpolicy *sp = NULL; 1934 int ipsecerror; 1935 int ipsechdr; 1936 struct route *ro; 1937 1938#ifdef IPSEC 1939 sp = ipsec4_getpolicybyaddr(mcopy, 1940 IPSEC_DIR_OUTBOUND, 1941 IP_FORWARDING, 1942 &ipsecerror); 1943#else /* FAST_IPSEC */ 1944 sp = ipsec_getpolicybyaddr(mcopy, 1945 IPSEC_DIR_OUTBOUND, 1946 IP_FORWARDING, 1947 &ipsecerror); 1948#endif 1949 if (sp != NULL) { 1950 /* count IPsec header size */ 1951 ipsechdr = ipsec4_hdrsiz(mcopy, 1952 IPSEC_DIR_OUTBOUND, 1953 NULL); 1954 1955 /* 1956 * find the correct route for outer IPv4 1957 * header, compute tunnel MTU. 1958 * 1959 * XXX BUG ALERT 1960 * The "dummyifp" code relies upon the fact 1961 * that icmp_error() touches only ifp->if_mtu. 1962 */ 1963 /*XXX*/ 1964 destifp = NULL; 1965 if (sp->req != NULL 1966 && sp->req->sav != NULL 1967 && sp->req->sav->sah != NULL) { 1968 ro = &sp->req->sav->sah->sa_route; 1969 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1970 dummyifp.if_mtu = 1971 ro->ro_rt->rt_rmx.rmx_mtu ? 1972 ro->ro_rt->rt_rmx.rmx_mtu : 1973 ro->ro_rt->rt_ifp->if_mtu; 1974 dummyifp.if_mtu -= ipsechdr; 1975 destifp = &dummyifp; 1976 } 1977 } 1978 1979#ifdef IPSEC 1980 key_freesp(sp); 1981#else /* FAST_IPSEC */ 1982 KEY_FREESP(&sp); 1983#endif 1984 ipstat.ips_cantfrag++; 1985 break; 1986 } else 1987#endif /*IPSEC || FAST_IPSEC*/ 1988 destifp = ia->ia_ifp; 1989#if defined(IPSEC) || defined(FAST_IPSEC) 1990 } 1991#endif /*IPSEC || FAST_IPSEC*/ 1992 ipstat.ips_cantfrag++; 1993 break; 1994 1995 case ENOBUFS: 1996 /* 1997 * A router should not generate ICMP_SOURCEQUENCH as 1998 * required in RFC1812 Requirements for IP Version 4 Routers. 1999 * Source quench could be a big problem under DoS attacks, 2000 * or if the underlying interface is rate-limited. 2001 * Those who need source quench packets may re-enable them 2002 * via the net.inet.ip.sendsourcequench sysctl. 2003 */ 2004 if (ip_sendsourcequench == 0) { 2005 m_freem(mcopy); 2006 return; 2007 } else { 2008 type = ICMP_SOURCEQUENCH; 2009 code = 0; 2010 } 2011 break; 2012 2013 case EACCES: /* ipfw denied packet */ 2014 m_freem(mcopy); 2015 return; 2016 } 2017 icmp_error(mcopy, type, code, dest, destifp); 2018} 2019 2020void 2021ip_savecontrol(inp, mp, ip, m) 2022 register struct inpcb *inp; 2023 register struct mbuf **mp; 2024 register struct ip *ip; 2025 register struct mbuf *m; 2026{ 2027 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 2028 struct bintime bt; 2029 2030 bintime(&bt); 2031 if (inp->inp_socket->so_options & SO_BINTIME) { 2032 *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt), 2033 SCM_BINTIME, SOL_SOCKET); 2034 if (*mp) 2035 mp = &(*mp)->m_next; 2036 } 2037 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2038 struct timeval tv; 2039 2040 bintime2timeval(&bt, &tv); 2041 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2042 SCM_TIMESTAMP, SOL_SOCKET); 2043 if (*mp) 2044 mp = &(*mp)->m_next; 2045 } 2046 } 2047 if (inp->inp_flags & INP_RECVDSTADDR) { 2048 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2049 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2050 if (*mp) 2051 mp = &(*mp)->m_next; 2052 } 2053 if (inp->inp_flags & INP_RECVTTL) { 2054 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2055 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2056 if (*mp) 2057 mp = &(*mp)->m_next; 2058 } 2059#ifdef notyet 2060 /* XXX 2061 * Moving these out of udp_input() made them even more broken 2062 * than they already were. 2063 */ 2064 /* options were tossed already */ 2065 if (inp->inp_flags & INP_RECVOPTS) { 2066 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2067 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2068 if (*mp) 2069 mp = &(*mp)->m_next; 2070 } 2071 /* ip_srcroute doesn't do what we want here, need to fix */ 2072 if (inp->inp_flags & INP_RECVRETOPTS) { 2073 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 2074 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2075 if (*mp) 2076 mp = &(*mp)->m_next; 2077 } 2078#endif 2079 if (inp->inp_flags & INP_RECVIF) { 2080 struct ifnet *ifp; 2081 struct sdlbuf { 2082 struct sockaddr_dl sdl; 2083 u_char pad[32]; 2084 } sdlbuf; 2085 struct sockaddr_dl *sdp; 2086 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2087 2088 if (((ifp = m->m_pkthdr.rcvif)) 2089 && ( ifp->if_index && (ifp->if_index <= if_index))) { 2090 sdp = (struct sockaddr_dl *) 2091 (ifaddr_byindex(ifp->if_index)->ifa_addr); 2092 /* 2093 * Change our mind and don't try copy. 2094 */ 2095 if ((sdp->sdl_family != AF_LINK) 2096 || (sdp->sdl_len > sizeof(sdlbuf))) { 2097 goto makedummy; 2098 } 2099 bcopy(sdp, sdl2, sdp->sdl_len); 2100 } else { 2101makedummy: 2102 sdl2->sdl_len 2103 = offsetof(struct sockaddr_dl, sdl_data[0]); 2104 sdl2->sdl_family = AF_LINK; 2105 sdl2->sdl_index = 0; 2106 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2107 } 2108 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2109 IP_RECVIF, IPPROTO_IP); 2110 if (*mp) 2111 mp = &(*mp)->m_next; 2112 } 2113} 2114 2115/* 2116 * XXX these routines are called from the upper part of the kernel. 2117 * They need to be locked when we remove Giant. 2118 * 2119 * They could also be moved to ip_mroute.c, since all the RSVP 2120 * handling is done there already. 2121 */ 2122static int ip_rsvp_on; 2123struct socket *ip_rsvpd; 2124int 2125ip_rsvp_init(struct socket *so) 2126{ 2127 if (so->so_type != SOCK_RAW || 2128 so->so_proto->pr_protocol != IPPROTO_RSVP) 2129 return EOPNOTSUPP; 2130 2131 if (ip_rsvpd != NULL) 2132 return EADDRINUSE; 2133 2134 ip_rsvpd = so; 2135 /* 2136 * This may seem silly, but we need to be sure we don't over-increment 2137 * the RSVP counter, in case something slips up. 2138 */ 2139 if (!ip_rsvp_on) { 2140 ip_rsvp_on = 1; 2141 rsvp_on++; 2142 } 2143 2144 return 0; 2145} 2146 2147int 2148ip_rsvp_done(void) 2149{ 2150 ip_rsvpd = NULL; 2151 /* 2152 * This may seem silly, but we need to be sure we don't over-decrement 2153 * the RSVP counter, in case something slips up. 2154 */ 2155 if (ip_rsvp_on) { 2156 ip_rsvp_on = 0; 2157 rsvp_on--; 2158 } 2159 return 0; 2160} 2161 2162void 2163rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */ 2164{ 2165 if (rsvp_input_p) { /* call the real one if loaded */ 2166 rsvp_input_p(m, off); 2167 return; 2168 } 2169 2170 /* Can still get packets with rsvp_on = 0 if there is a local member 2171 * of the group to which the RSVP packet is addressed. But in this 2172 * case we want to throw the packet away. 2173 */ 2174 2175 if (!rsvp_on) { 2176 m_freem(m); 2177 return; 2178 } 2179 2180 if (ip_rsvpd != NULL) { 2181 rip_input(m, off); 2182 return; 2183 } 2184 /* Drop the packet */ 2185 m_freem(m); 2186} 2187