ip_reass.c revision 78667
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 78667 2001-06-23 17:17:59Z ru $ 35 */ 36 37#define _IP_VHL 38 39#include "opt_bootp.h" 40#include "opt_ipfw.h" 41#include "opt_ipdn.h" 42#include "opt_ipdivert.h" 43#include "opt_ipfilter.h" 44#include "opt_ipstealth.h" 45#include "opt_ipsec.h" 46#include "opt_pfil_hooks.h" 47#include "opt_random_ip_id.h" 48 49#include <sys/param.h> 50#include <sys/systm.h> 51#include <sys/mbuf.h> 52#include <sys/malloc.h> 53#include <sys/domain.h> 54#include <sys/protosw.h> 55#include <sys/socket.h> 56#include <sys/time.h> 57#include <sys/kernel.h> 58#include <sys/syslog.h> 59#include <sys/sysctl.h> 60 61#include <net/pfil.h> 62#include <net/if.h> 63#include <net/if_var.h> 64#include <net/if_dl.h> 65#include <net/route.h> 66#include <net/netisr.h> 67#include <net/intrq.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 <netinet/ipprotosw.h> 79 80#include <sys/socketvar.h> 81 82#include <netinet/ip_fw.h> 83 84#ifdef IPSEC 85#include <netinet6/ipsec.h> 86#include <netkey/key.h> 87#endif 88 89#include "faith.h" 90#if defined(NFAITH) && NFAITH > 0 91#include <net/if_types.h> 92#endif 93 94#ifdef DUMMYNET 95#include <netinet/ip_dummynet.h> 96#endif 97 98int rsvp_on = 0; 99static int ip_rsvp_on; 100struct socket *ip_rsvpd; 101 102int ipforwarding = 0; 103SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 104 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 105 106static int ipsendredirects = 1; /* XXX */ 107SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 108 &ipsendredirects, 0, "Enable sending IP redirects"); 109 110int ip_defttl = IPDEFTTL; 111SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 112 &ip_defttl, 0, "Maximum TTL on IP packets"); 113 114static int ip_dosourceroute = 0; 115SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 116 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 117 118static int ip_acceptsourceroute = 0; 119SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 120 CTLFLAG_RW, &ip_acceptsourceroute, 0, 121 "Enable accepting source routed IP packets"); 122 123static int ip_keepfaith = 0; 124SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 125 &ip_keepfaith, 0, 126 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 127 128static int ip_nfragpackets = 0; 129static int ip_maxfragpackets; /* initialized in ip_init() */ 130SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 131 &ip_maxfragpackets, 0, 132 "Maximum number of IPv4 fragment reassembly queue entries"); 133 134/* 135 * XXX - Setting ip_checkinterface mostly implements the receive side of 136 * the Strong ES model described in RFC 1122, but since the routing table 137 * and transmit implementation do not implement the Strong ES model, 138 * setting this to 1 results in an odd hybrid. 139 * 140 * XXX - ip_checkinterface currently must be disabled if you use ipnat 141 * to translate the destination address to another local interface. 142 * 143 * XXX - ip_checkinterface must be disabled if you add IP aliases 144 * to the loopback interface instead of the interface where the 145 * packets for those addresses are received. 146 */ 147static int ip_checkinterface = 1; 148SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 149 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 150 151#ifdef DIAGNOSTIC 152static int ipprintfs = 0; 153#endif 154 155extern struct domain inetdomain; 156extern struct ipprotosw inetsw[]; 157u_char ip_protox[IPPROTO_MAX]; 158static int ipqmaxlen = IFQ_MAXLEN; 159struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 160SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 161 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 162SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 163 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 164 165struct ipstat ipstat; 166SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW, 167 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 168 169/* Packet reassembly stuff */ 170#define IPREASS_NHASH_LOG2 6 171#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 172#define IPREASS_HMASK (IPREASS_NHASH - 1) 173#define IPREASS_HASH(x,y) \ 174 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 175 176static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 177static int nipq = 0; /* total # of reass queues */ 178static int maxnipq; 179const int ipintrq_present = 1; 180 181#ifdef IPCTL_DEFMTU 182SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 183 &ip_mtu, 0, "Default MTU"); 184#endif 185 186#ifdef IPSTEALTH 187static int ipstealth = 0; 188SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 189 &ipstealth, 0, ""); 190#endif 191 192 193/* Firewall hooks */ 194ip_fw_chk_t *ip_fw_chk_ptr; 195ip_fw_ctl_t *ip_fw_ctl_ptr; 196int fw_enable = 1 ; 197 198#ifdef DUMMYNET 199ip_dn_ctl_t *ip_dn_ctl_ptr; 200#endif 201 202 203/* 204 * We need to save the IP options in case a protocol wants to respond 205 * to an incoming packet over the same route if the packet got here 206 * using IP source routing. This allows connection establishment and 207 * maintenance when the remote end is on a network that is not known 208 * to us. 209 */ 210static int ip_nhops = 0; 211static struct ip_srcrt { 212 struct in_addr dst; /* final destination */ 213 char nop; /* one NOP to align */ 214 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 215 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 216} ip_srcrt; 217 218struct sockaddr_in *ip_fw_fwd_addr; 219 220static void save_rte __P((u_char *, struct in_addr)); 221static int ip_dooptions __P((struct mbuf *)); 222static void ip_forward __P((struct mbuf *, int)); 223static void ip_freef __P((struct ipqhead *, struct ipq *)); 224#ifdef IPDIVERT 225static struct mbuf *ip_reass __P((struct mbuf *, struct ipqhead *, struct ipq *, u_int32_t *, u_int16_t *)); 226#else 227static struct mbuf *ip_reass __P((struct mbuf *, struct ipqhead *, struct ipq *)); 228#endif 229static struct in_ifaddr *ip_rtaddr __P((struct in_addr)); 230static void ipintr __P((void)); 231 232/* 233 * IP initialization: fill in IP protocol switch table. 234 * All protocols not implemented in kernel go to raw IP protocol handler. 235 */ 236void 237ip_init() 238{ 239 register struct ipprotosw *pr; 240 register int i; 241 242 TAILQ_INIT(&in_ifaddrhead); 243 pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 244 if (pr == 0) 245 panic("ip_init"); 246 for (i = 0; i < IPPROTO_MAX; i++) 247 ip_protox[i] = pr - inetsw; 248 for (pr = (struct ipprotosw *)inetdomain.dom_protosw; 249 pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++) 250 if (pr->pr_domain->dom_family == PF_INET && 251 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 252 ip_protox[pr->pr_protocol] = pr - inetsw; 253 254 for (i = 0; i < IPREASS_NHASH; i++) 255 TAILQ_INIT(&ipq[i]); 256 257 maxnipq = nmbclusters / 4; 258 ip_maxfragpackets = nmbclusters / 4; 259 260#ifndef RANDOM_IP_ID 261 ip_id = time_second & 0xffff; 262#endif 263 ipintrq.ifq_maxlen = ipqmaxlen; 264 mtx_init(&ipintrq.ifq_mtx, "ip_inq", MTX_DEF); 265 266 register_netisr(NETISR_IP, ipintr); 267} 268 269static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 270struct route ipforward_rt; 271 272/* 273 * Ip input routine. Checksum and byte swap header. If fragmented 274 * try to reassemble. Process options. Pass to next level. 275 */ 276void 277ip_input(struct mbuf *m) 278{ 279 struct ip *ip; 280 struct ipq *fp; 281 struct in_ifaddr *ia = NULL; 282 int i, hlen, checkif; 283 u_short sum; 284 u_int16_t divert_cookie; /* firewall cookie */ 285 struct in_addr pkt_dst; 286#ifdef IPDIVERT 287 u_int32_t divert_info = 0; /* packet divert/tee info */ 288#endif 289 struct ip_fw_chain *rule = NULL; 290#ifdef PFIL_HOOKS 291 struct packet_filter_hook *pfh; 292 struct mbuf *m0; 293 int rv; 294#endif /* PFIL_HOOKS */ 295 296#ifdef IPDIVERT 297 /* Get and reset firewall cookie */ 298 divert_cookie = ip_divert_cookie; 299 ip_divert_cookie = 0; 300#else 301 divert_cookie = 0; 302#endif 303 304#if defined(IPFIREWALL) && defined(DUMMYNET) 305 /* 306 * dummynet packet are prepended a vestigial mbuf with 307 * m_type = MT_DUMMYNET and m_data pointing to the matching 308 * rule. 309 */ 310 if (m->m_type == MT_DUMMYNET) { 311 rule = (struct ip_fw_chain *)(m->m_data) ; 312 m = m->m_next ; 313 ip = mtod(m, struct ip *); 314 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 315 goto iphack ; 316 } else 317 rule = NULL ; 318#endif 319 320#ifdef DIAGNOSTIC 321 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 322 panic("ip_input no HDR"); 323#endif 324 ipstat.ips_total++; 325 326 if (m->m_pkthdr.len < sizeof(struct ip)) 327 goto tooshort; 328 329 if (m->m_len < sizeof (struct ip) && 330 (m = m_pullup(m, sizeof (struct ip))) == 0) { 331 ipstat.ips_toosmall++; 332 return; 333 } 334 ip = mtod(m, struct ip *); 335 336 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 337 ipstat.ips_badvers++; 338 goto bad; 339 } 340 341 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 342 if (hlen < sizeof(struct ip)) { /* minimum header length */ 343 ipstat.ips_badhlen++; 344 goto bad; 345 } 346 if (hlen > m->m_len) { 347 if ((m = m_pullup(m, hlen)) == 0) { 348 ipstat.ips_badhlen++; 349 return; 350 } 351 ip = mtod(m, struct ip *); 352 } 353 354 /* 127/8 must not appear on wire - RFC1122 */ 355 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 356 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 357 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 358 ipstat.ips_badaddr++; 359 goto bad; 360 } 361 } 362 363 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 364 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 365 } else { 366 if (hlen == sizeof(struct ip)) { 367 sum = in_cksum_hdr(ip); 368 } else { 369 sum = in_cksum(m, hlen); 370 } 371 } 372 if (sum) { 373 ipstat.ips_badsum++; 374 goto bad; 375 } 376 377 /* 378 * Convert fields to host representation. 379 */ 380 NTOHS(ip->ip_len); 381 if (ip->ip_len < hlen) { 382 ipstat.ips_badlen++; 383 goto bad; 384 } 385 NTOHS(ip->ip_off); 386 387 /* 388 * Check that the amount of data in the buffers 389 * is as at least much as the IP header would have us expect. 390 * Trim mbufs if longer than we expect. 391 * Drop packet if shorter than we expect. 392 */ 393 if (m->m_pkthdr.len < ip->ip_len) { 394tooshort: 395 ipstat.ips_tooshort++; 396 goto bad; 397 } 398 if (m->m_pkthdr.len > ip->ip_len) { 399 if (m->m_len == m->m_pkthdr.len) { 400 m->m_len = ip->ip_len; 401 m->m_pkthdr.len = ip->ip_len; 402 } else 403 m_adj(m, ip->ip_len - m->m_pkthdr.len); 404 } 405 406#ifdef IPSEC 407 if (ipsec_gethist(m, NULL)) 408 goto pass; 409#endif 410 411 /* 412 * IpHack's section. 413 * Right now when no processing on packet has done 414 * and it is still fresh out of network we do our black 415 * deals with it. 416 * - Firewall: deny/allow/divert 417 * - Xlate: translate packet's addr/port (NAT). 418 * - Pipe: pass pkt through dummynet. 419 * - Wrap: fake packet's addr/port <unimpl.> 420 * - Encapsulate: put it in another IP and send out. <unimp.> 421 */ 422 423#if defined(IPFIREWALL) && defined(DUMMYNET) 424iphack: 425#endif 426 427#ifdef PFIL_HOOKS 428 /* 429 * Run through list of hooks for input packets. If there are any 430 * filters which require that additional packets in the flow are 431 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 432 * Note that filters must _never_ set this flag, as another filter 433 * in the list may have previously cleared it. 434 */ 435 m0 = m; 436 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 437 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 438 if (pfh->pfil_func) { 439 rv = pfh->pfil_func(ip, hlen, 440 m->m_pkthdr.rcvif, 0, &m0); 441 if (rv) 442 return; 443 m = m0; 444 if (m == NULL) 445 return; 446 ip = mtod(m, struct ip *); 447 } 448#endif /* PFIL_HOOKS */ 449 450 if (fw_enable && ip_fw_chk_ptr) { 451#ifdef IPFIREWALL_FORWARD 452 /* 453 * If we've been forwarded from the output side, then 454 * skip the firewall a second time 455 */ 456 if (ip_fw_fwd_addr) 457 goto ours; 458#endif /* IPFIREWALL_FORWARD */ 459 /* 460 * See the comment in ip_output for the return values 461 * produced by the firewall. 462 */ 463 i = (*ip_fw_chk_ptr)(&ip, 464 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 465 if (i & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */ 466 if (m) 467 m_freem(m); 468 return ; 469 } 470 if (m == NULL) { /* Packet discarded by firewall */ 471 static int __debug=10; 472 if (__debug >0) { 473 printf("firewall returns NULL, please update!\n"); 474 __debug-- ; 475 } 476 return; 477 } 478 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 479 goto pass; 480#ifdef DUMMYNET 481 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) { 482 /* Send packet to the appropriate pipe */ 483 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 484 0); 485 return; 486 } 487#endif 488#ifdef IPDIVERT 489 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 490 /* Divert or tee packet */ 491 divert_info = i; 492 goto ours; 493 } 494#endif 495#ifdef IPFIREWALL_FORWARD 496 if (i == 0 && ip_fw_fwd_addr != NULL) 497 goto pass; 498#endif 499 /* 500 * if we get here, the packet must be dropped 501 */ 502 m_freem(m); 503 return; 504 } 505pass: 506 507 /* 508 * Process options and, if not destined for us, 509 * ship it on. ip_dooptions returns 1 when an 510 * error was detected (causing an icmp message 511 * to be sent and the original packet to be freed). 512 */ 513 ip_nhops = 0; /* for source routed packets */ 514 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 515#ifdef IPFIREWALL_FORWARD 516 ip_fw_fwd_addr = NULL; 517#endif 518 return; 519 } 520 521 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 522 * matter if it is destined to another node, or whether it is 523 * a multicast one, RSVP wants it! and prevents it from being forwarded 524 * anywhere else. Also checks if the rsvp daemon is running before 525 * grabbing the packet. 526 */ 527 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 528 goto ours; 529 530 /* 531 * Check our list of addresses, to see if the packet is for us. 532 * If we don't have any addresses, assume any unicast packet 533 * we receive might be for us (and let the upper layers deal 534 * with it). 535 */ 536 if (TAILQ_EMPTY(&in_ifaddrhead) && 537 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 538 goto ours; 539 540 /* 541 * Cache the destination address of the packet; this may be 542 * changed by use of 'ipfw fwd'. 543 */ 544 pkt_dst = ip_fw_fwd_addr == NULL ? 545 ip->ip_dst : ip_fw_fwd_addr->sin_addr; 546 547 /* 548 * Enable a consistency check between the destination address 549 * and the arrival interface for a unicast packet (the RFC 1122 550 * strong ES model) if IP forwarding is disabled and the packet 551 * is not locally generated and the packet is not subject to 552 * 'ipfw fwd'. 553 * 554 * XXX - Checking also should be disabled if the destination 555 * address is ipnat'ed to a different interface. 556 * 557 * XXX - Checking is incompatible with IP aliases added 558 * to the loopback interface instead of the interface where 559 * the packets are received. 560 */ 561 checkif = ip_checkinterface && (ipforwarding == 0) && 562 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 563 (ip_fw_fwd_addr == NULL); 564 565 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 566#define satosin(sa) ((struct sockaddr_in *)(sa)) 567 568#ifdef BOOTP_COMPAT 569 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 570 goto ours; 571#endif 572 /* 573 * If the address matches, verify that the packet 574 * arrived via the correct interface if checking is 575 * enabled. 576 */ 577 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 578 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 579 goto ours; 580 /* 581 * Only accept broadcast packets that arrive via the 582 * matching interface. Reception of forwarded directed 583 * broadcasts would be handled via ip_forward() and 584 * ether_output() with the loopback into the stack for 585 * SIMPLEX interfaces handled by ether_output(). 586 */ 587 if (ia->ia_ifp == m->m_pkthdr.rcvif && 588 ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 589 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 590 pkt_dst.s_addr) 591 goto ours; 592 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 593 goto ours; 594 } 595 } 596 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 597 struct in_multi *inm; 598 if (ip_mrouter) { 599 /* 600 * If we are acting as a multicast router, all 601 * incoming multicast packets are passed to the 602 * kernel-level multicast forwarding function. 603 * The packet is returned (relatively) intact; if 604 * ip_mforward() returns a non-zero value, the packet 605 * must be discarded, else it may be accepted below. 606 */ 607 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 608 ipstat.ips_cantforward++; 609 m_freem(m); 610 return; 611 } 612 613 /* 614 * The process-level routing demon needs to receive 615 * all multicast IGMP packets, whether or not this 616 * host belongs to their destination groups. 617 */ 618 if (ip->ip_p == IPPROTO_IGMP) 619 goto ours; 620 ipstat.ips_forward++; 621 } 622 /* 623 * See if we belong to the destination multicast group on the 624 * arrival interface. 625 */ 626 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 627 if (inm == NULL) { 628 ipstat.ips_notmember++; 629 m_freem(m); 630 return; 631 } 632 goto ours; 633 } 634 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 635 goto ours; 636 if (ip->ip_dst.s_addr == INADDR_ANY) 637 goto ours; 638 639#if defined(NFAITH) && 0 < NFAITH 640 /* 641 * FAITH(Firewall Aided Internet Translator) 642 */ 643 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 644 if (ip_keepfaith) { 645 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 646 goto ours; 647 } 648 m_freem(m); 649 return; 650 } 651#endif 652 /* 653 * Not for us; forward if possible and desirable. 654 */ 655 if (ipforwarding == 0) { 656 ipstat.ips_cantforward++; 657 m_freem(m); 658 } else 659 ip_forward(m, 0); 660#ifdef IPFIREWALL_FORWARD 661 ip_fw_fwd_addr = NULL; 662#endif 663 return; 664 665ours: 666 /* Count the packet in the ip address stats */ 667 if (ia != NULL) { 668 ia->ia_ifa.if_ipackets++; 669 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 670 } 671 672 /* 673 * If offset or IP_MF are set, must reassemble. 674 * Otherwise, nothing need be done. 675 * (We could look in the reassembly queue to see 676 * if the packet was previously fragmented, 677 * but it's not worth the time; just let them time out.) 678 */ 679 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 680 681 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 682 /* 683 * Look for queue of fragments 684 * of this datagram. 685 */ 686 TAILQ_FOREACH(fp, &ipq[sum], ipq_list) 687 if (ip->ip_id == fp->ipq_id && 688 ip->ip_src.s_addr == fp->ipq_src.s_addr && 689 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 690 ip->ip_p == fp->ipq_p) 691 goto found; 692 693 fp = 0; 694 695 /* check if there's a place for the new queue */ 696 if (nipq > maxnipq) { 697 /* 698 * drop something from the tail of the current queue 699 * before proceeding further 700 */ 701 struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead); 702 if (q == NULL) { /* gak */ 703 for (i = 0; i < IPREASS_NHASH; i++) { 704 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 705 if (r) { 706 ip_freef(&ipq[i], r); 707 break; 708 } 709 } 710 } else 711 ip_freef(&ipq[sum], q); 712 } 713found: 714 /* 715 * Adjust ip_len to not reflect header, 716 * convert offset of this to bytes. 717 */ 718 ip->ip_len -= hlen; 719 if (ip->ip_off & IP_MF) { 720 /* 721 * Make sure that fragments have a data length 722 * that's a non-zero multiple of 8 bytes. 723 */ 724 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 725 ipstat.ips_toosmall++; /* XXX */ 726 goto bad; 727 } 728 m->m_flags |= M_FRAG; 729 } 730 ip->ip_off <<= 3; 731 732 /* 733 * Attempt reassembly; if it succeeds, proceed. 734 */ 735 ipstat.ips_fragments++; 736 m->m_pkthdr.header = ip; 737#ifdef IPDIVERT 738 m = ip_reass(m, 739 &ipq[sum], fp, &divert_info, &divert_cookie); 740#else 741 m = ip_reass(m, &ipq[sum], fp); 742#endif 743 if (m == 0) { 744#ifdef IPFIREWALL_FORWARD 745 ip_fw_fwd_addr = NULL; 746#endif 747 return; 748 } 749 ipstat.ips_reassembled++; 750 ip = mtod(m, struct ip *); 751 /* Get the header length of the reassembled packet */ 752 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 753#ifdef IPDIVERT 754 /* Restore original checksum before diverting packet */ 755 if (divert_info != 0) { 756 ip->ip_len += hlen; 757 HTONS(ip->ip_len); 758 HTONS(ip->ip_off); 759 ip->ip_sum = 0; 760 if (hlen == sizeof(struct ip)) 761 ip->ip_sum = in_cksum_hdr(ip); 762 else 763 ip->ip_sum = in_cksum(m, hlen); 764 NTOHS(ip->ip_off); 765 NTOHS(ip->ip_len); 766 ip->ip_len -= hlen; 767 } 768#endif 769 } else 770 ip->ip_len -= hlen; 771 772#ifdef IPDIVERT 773 /* 774 * Divert or tee packet to the divert protocol if required. 775 * 776 * If divert_info is zero then cookie should be too, so we shouldn't 777 * need to clear them here. Assume divert_packet() does so also. 778 */ 779 if (divert_info != 0) { 780 struct mbuf *clone = NULL; 781 782 /* Clone packet if we're doing a 'tee' */ 783 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 784 clone = m_dup(m, M_DONTWAIT); 785 786 /* Restore packet header fields to original values */ 787 ip->ip_len += hlen; 788 HTONS(ip->ip_len); 789 HTONS(ip->ip_off); 790 791 /* Deliver packet to divert input routine */ 792 ip_divert_cookie = divert_cookie; 793 divert_packet(m, 1, divert_info & 0xffff); 794 ipstat.ips_delivered++; 795 796 /* If 'tee', continue with original packet */ 797 if (clone == NULL) 798 return; 799 m = clone; 800 ip = mtod(m, struct ip *); 801 } 802#endif 803 804#ifdef IPSEC 805 /* 806 * enforce IPsec policy checking if we are seeing last header. 807 * note that we do not visit this with protocols with pcb layer 808 * code - like udp/tcp/raw ip. 809 */ 810 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 && 811 ipsec4_in_reject(m, NULL)) { 812 ipsecstat.in_polvio++; 813 goto bad; 814 } 815#endif 816 817 /* 818 * Switch out to protocol's input routine. 819 */ 820 ipstat.ips_delivered++; 821 { 822 int off = hlen, nh = ip->ip_p; 823 824 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh); 825#ifdef IPFIREWALL_FORWARD 826 ip_fw_fwd_addr = NULL; /* tcp needed it */ 827#endif 828 return; 829 } 830bad: 831#ifdef IPFIREWALL_FORWARD 832 ip_fw_fwd_addr = NULL; 833#endif 834 m_freem(m); 835} 836 837/* 838 * IP software interrupt routine - to go away sometime soon 839 */ 840static void 841ipintr(void) 842{ 843 struct mbuf *m; 844 845 while (1) { 846 IF_DEQUEUE(&ipintrq, m); 847 if (m == 0) 848 return; 849 ip_input(m); 850 } 851} 852 853/* 854 * Take incoming datagram fragment and try to reassemble it into 855 * whole datagram. If a chain for reassembly of this datagram already 856 * exists, then it is given as fp; otherwise have to make a chain. 857 * 858 * When IPDIVERT enabled, keep additional state with each packet that 859 * tells us if we need to divert or tee the packet we're building. 860 */ 861 862static struct mbuf * 863#ifdef IPDIVERT 864ip_reass(m, head, fp, divinfo, divcookie) 865#else 866ip_reass(m, head, fp) 867#endif 868 struct mbuf *m; 869 struct ipqhead *head; 870 struct ipq *fp; 871#ifdef IPDIVERT 872 u_int32_t *divinfo; 873 u_int16_t *divcookie; 874#endif 875{ 876 struct ip *ip = mtod(m, struct ip *); 877 register struct mbuf *p, *q, *nq; 878 struct mbuf *t; 879 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 880 int i, next; 881 882 /* 883 * Presence of header sizes in mbufs 884 * would confuse code below. 885 */ 886 m->m_data += hlen; 887 m->m_len -= hlen; 888 889 /* 890 * If first fragment to arrive, create a reassembly queue. 891 */ 892 if (fp == 0) { 893 /* 894 * Enforce upper bound on number of fragmented packets 895 * for which we attempt reassembly; 896 * If maxfrag is 0, never accept fragments. 897 * If maxfrag is -1, accept all fragments without limitation. 898 */ 899 if ((ip_maxfragpackets >= 0) && (ip_nfragpackets >= ip_maxfragpackets)) 900 goto dropfrag; 901 ip_nfragpackets++; 902 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 903 goto dropfrag; 904 fp = mtod(t, struct ipq *); 905 TAILQ_INSERT_HEAD(head, fp, ipq_list); 906 nipq++; 907 fp->ipq_ttl = IPFRAGTTL; 908 fp->ipq_p = ip->ip_p; 909 fp->ipq_id = ip->ip_id; 910 fp->ipq_src = ip->ip_src; 911 fp->ipq_dst = ip->ip_dst; 912 fp->ipq_frags = m; 913 m->m_nextpkt = NULL; 914#ifdef IPDIVERT 915 fp->ipq_div_info = 0; 916 fp->ipq_div_cookie = 0; 917#endif 918 goto inserted; 919 } 920 921#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 922 923 /* 924 * Find a segment which begins after this one does. 925 */ 926 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 927 if (GETIP(q)->ip_off > ip->ip_off) 928 break; 929 930 /* 931 * If there is a preceding segment, it may provide some of 932 * our data already. If so, drop the data from the incoming 933 * segment. If it provides all of our data, drop us, otherwise 934 * stick new segment in the proper place. 935 * 936 * If some of the data is dropped from the the preceding 937 * segment, then it's checksum is invalidated. 938 */ 939 if (p) { 940 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 941 if (i > 0) { 942 if (i >= ip->ip_len) 943 goto dropfrag; 944 m_adj(m, i); 945 m->m_pkthdr.csum_flags = 0; 946 ip->ip_off += i; 947 ip->ip_len -= i; 948 } 949 m->m_nextpkt = p->m_nextpkt; 950 p->m_nextpkt = m; 951 } else { 952 m->m_nextpkt = fp->ipq_frags; 953 fp->ipq_frags = m; 954 } 955 956 /* 957 * While we overlap succeeding segments trim them or, 958 * if they are completely covered, dequeue them. 959 */ 960 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 961 q = nq) { 962 i = (ip->ip_off + ip->ip_len) - 963 GETIP(q)->ip_off; 964 if (i < GETIP(q)->ip_len) { 965 GETIP(q)->ip_len -= i; 966 GETIP(q)->ip_off += i; 967 m_adj(q, i); 968 q->m_pkthdr.csum_flags = 0; 969 break; 970 } 971 nq = q->m_nextpkt; 972 m->m_nextpkt = nq; 973 m_freem(q); 974 } 975 976inserted: 977 978#ifdef IPDIVERT 979 /* 980 * Transfer firewall instructions to the fragment structure. 981 * Any fragment diverting causes the whole packet to divert. 982 */ 983 fp->ipq_div_info = *divinfo; 984 fp->ipq_div_cookie = *divcookie; 985 *divinfo = 0; 986 *divcookie = 0; 987#endif 988 989 /* 990 * Check for complete reassembly. 991 */ 992 next = 0; 993 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 994 if (GETIP(q)->ip_off != next) 995 return (0); 996 next += GETIP(q)->ip_len; 997 } 998 /* Make sure the last packet didn't have the IP_MF flag */ 999 if (p->m_flags & M_FRAG) 1000 return (0); 1001 1002 /* 1003 * Reassembly is complete. Make sure the packet is a sane size. 1004 */ 1005 q = fp->ipq_frags; 1006 ip = GETIP(q); 1007 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1008 ipstat.ips_toolong++; 1009 ip_freef(head, fp); 1010 return (0); 1011 } 1012 1013 /* 1014 * Concatenate fragments. 1015 */ 1016 m = q; 1017 t = m->m_next; 1018 m->m_next = 0; 1019 m_cat(m, t); 1020 nq = q->m_nextpkt; 1021 q->m_nextpkt = 0; 1022 for (q = nq; q != NULL; q = nq) { 1023 nq = q->m_nextpkt; 1024 q->m_nextpkt = NULL; 1025 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1026 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1027 m_cat(m, q); 1028 } 1029 1030#ifdef IPDIVERT 1031 /* 1032 * Extract firewall instructions from the fragment structure. 1033 */ 1034 *divinfo = fp->ipq_div_info; 1035 *divcookie = fp->ipq_div_cookie; 1036#endif 1037 1038 /* 1039 * Create header for new ip packet by 1040 * modifying header of first packet; 1041 * dequeue and discard fragment reassembly header. 1042 * Make header visible. 1043 */ 1044 ip->ip_len = next; 1045 ip->ip_src = fp->ipq_src; 1046 ip->ip_dst = fp->ipq_dst; 1047 TAILQ_REMOVE(head, fp, ipq_list); 1048 nipq--; 1049 (void) m_free(dtom(fp)); 1050 ip_nfragpackets--; 1051 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1052 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1053 /* some debugging cruft by sklower, below, will go away soon */ 1054 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1055 register int plen = 0; 1056 for (t = m; t; t = t->m_next) 1057 plen += t->m_len; 1058 m->m_pkthdr.len = plen; 1059 } 1060 return (m); 1061 1062dropfrag: 1063#ifdef IPDIVERT 1064 *divinfo = 0; 1065 *divcookie = 0; 1066#endif 1067 ipstat.ips_fragdropped++; 1068 m_freem(m); 1069 return (0); 1070 1071#undef GETIP 1072} 1073 1074/* 1075 * Free a fragment reassembly header and all 1076 * associated datagrams. 1077 */ 1078static void 1079ip_freef(fhp, fp) 1080 struct ipqhead *fhp; 1081 struct ipq *fp; 1082{ 1083 register struct mbuf *q; 1084 1085 while (fp->ipq_frags) { 1086 q = fp->ipq_frags; 1087 fp->ipq_frags = q->m_nextpkt; 1088 m_freem(q); 1089 } 1090 TAILQ_REMOVE(fhp, fp, ipq_list); 1091 (void) m_free(dtom(fp)); 1092 ip_nfragpackets--; 1093 nipq--; 1094} 1095 1096/* 1097 * IP timer processing; 1098 * if a timer expires on a reassembly 1099 * queue, discard it. 1100 */ 1101void 1102ip_slowtimo() 1103{ 1104 register struct ipq *fp; 1105 int s = splnet(); 1106 int i; 1107 1108 for (i = 0; i < IPREASS_NHASH; i++) { 1109 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1110 struct ipq *fpp; 1111 1112 fpp = fp; 1113 fp = TAILQ_NEXT(fp, ipq_list); 1114 if(--fpp->ipq_ttl == 0) { 1115 ipstat.ips_fragtimeout++; 1116 ip_freef(&ipq[i], fpp); 1117 } 1118 } 1119 } 1120 /* 1121 * If we are over the maximum number of fragments 1122 * (due to the limit being lowered), drain off 1123 * enough to get down to the new limit. 1124 */ 1125 for (i = 0; i < IPREASS_NHASH; i++) { 1126 if (ip_maxfragpackets >= 0) { 1127 while (ip_nfragpackets > ip_maxfragpackets && 1128 !TAILQ_EMPTY(&ipq[i])) { 1129 ipstat.ips_fragdropped++; 1130 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1131 } 1132 } 1133 } 1134 ipflow_slowtimo(); 1135 splx(s); 1136} 1137 1138/* 1139 * Drain off all datagram fragments. 1140 */ 1141void 1142ip_drain() 1143{ 1144 int i; 1145 1146 for (i = 0; i < IPREASS_NHASH; i++) { 1147 while(!TAILQ_EMPTY(&ipq[i])) { 1148 ipstat.ips_fragdropped++; 1149 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1150 } 1151 } 1152 in_rtqdrain(); 1153} 1154 1155/* 1156 * Do option processing on a datagram, 1157 * possibly discarding it if bad options are encountered, 1158 * or forwarding it if source-routed. 1159 * Returns 1 if packet has been forwarded/freed, 1160 * 0 if the packet should be processed further. 1161 */ 1162static int 1163ip_dooptions(m) 1164 struct mbuf *m; 1165{ 1166 register struct ip *ip = mtod(m, struct ip *); 1167 register u_char *cp; 1168 register struct ip_timestamp *ipt; 1169 register struct in_ifaddr *ia; 1170 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1171 struct in_addr *sin, dst; 1172 n_time ntime; 1173 1174 dst = ip->ip_dst; 1175 cp = (u_char *)(ip + 1); 1176 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1177 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1178 opt = cp[IPOPT_OPTVAL]; 1179 if (opt == IPOPT_EOL) 1180 break; 1181 if (opt == IPOPT_NOP) 1182 optlen = 1; 1183 else { 1184 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1185 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1186 goto bad; 1187 } 1188 optlen = cp[IPOPT_OLEN]; 1189 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1190 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1191 goto bad; 1192 } 1193 } 1194 switch (opt) { 1195 1196 default: 1197 break; 1198 1199 /* 1200 * Source routing with record. 1201 * Find interface with current destination address. 1202 * If none on this machine then drop if strictly routed, 1203 * or do nothing if loosely routed. 1204 * Record interface address and bring up next address 1205 * component. If strictly routed make sure next 1206 * address is on directly accessible net. 1207 */ 1208 case IPOPT_LSRR: 1209 case IPOPT_SSRR: 1210 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1211 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1212 goto bad; 1213 } 1214 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1215 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1216 goto bad; 1217 } 1218 ipaddr.sin_addr = ip->ip_dst; 1219 ia = (struct in_ifaddr *) 1220 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1221 if (ia == 0) { 1222 if (opt == IPOPT_SSRR) { 1223 type = ICMP_UNREACH; 1224 code = ICMP_UNREACH_SRCFAIL; 1225 goto bad; 1226 } 1227 if (!ip_dosourceroute) 1228 goto nosourcerouting; 1229 /* 1230 * Loose routing, and not at next destination 1231 * yet; nothing to do except forward. 1232 */ 1233 break; 1234 } 1235 off--; /* 0 origin */ 1236 if (off > optlen - (int)sizeof(struct in_addr)) { 1237 /* 1238 * End of source route. Should be for us. 1239 */ 1240 if (!ip_acceptsourceroute) 1241 goto nosourcerouting; 1242 save_rte(cp, ip->ip_src); 1243 break; 1244 } 1245 1246 if (!ip_dosourceroute) { 1247 if (ipforwarding) { 1248 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1249 /* 1250 * Acting as a router, so generate ICMP 1251 */ 1252nosourcerouting: 1253 strcpy(buf, inet_ntoa(ip->ip_dst)); 1254 log(LOG_WARNING, 1255 "attempted source route from %s to %s\n", 1256 inet_ntoa(ip->ip_src), buf); 1257 type = ICMP_UNREACH; 1258 code = ICMP_UNREACH_SRCFAIL; 1259 goto bad; 1260 } else { 1261 /* 1262 * Not acting as a router, so silently drop. 1263 */ 1264 ipstat.ips_cantforward++; 1265 m_freem(m); 1266 return (1); 1267 } 1268 } 1269 1270 /* 1271 * locate outgoing interface 1272 */ 1273 (void)memcpy(&ipaddr.sin_addr, cp + off, 1274 sizeof(ipaddr.sin_addr)); 1275 1276 if (opt == IPOPT_SSRR) { 1277#define INA struct in_ifaddr * 1278#define SA struct sockaddr * 1279 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1280 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1281 } else 1282 ia = ip_rtaddr(ipaddr.sin_addr); 1283 if (ia == 0) { 1284 type = ICMP_UNREACH; 1285 code = ICMP_UNREACH_SRCFAIL; 1286 goto bad; 1287 } 1288 ip->ip_dst = ipaddr.sin_addr; 1289 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1290 sizeof(struct in_addr)); 1291 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1292 /* 1293 * Let ip_intr's mcast routing check handle mcast pkts 1294 */ 1295 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1296 break; 1297 1298 case IPOPT_RR: 1299 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1300 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1301 goto bad; 1302 } 1303 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1304 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1305 goto bad; 1306 } 1307 /* 1308 * If no space remains, ignore. 1309 */ 1310 off--; /* 0 origin */ 1311 if (off > optlen - (int)sizeof(struct in_addr)) 1312 break; 1313 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1314 sizeof(ipaddr.sin_addr)); 1315 /* 1316 * locate outgoing interface; if we're the destination, 1317 * use the incoming interface (should be same). 1318 */ 1319 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1320 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1321 type = ICMP_UNREACH; 1322 code = ICMP_UNREACH_HOST; 1323 goto bad; 1324 } 1325 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1326 sizeof(struct in_addr)); 1327 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1328 break; 1329 1330 case IPOPT_TS: 1331 code = cp - (u_char *)ip; 1332 ipt = (struct ip_timestamp *)cp; 1333 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1334 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1335 goto bad; 1336 } 1337 if (ipt->ipt_ptr < 5) { 1338 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1339 goto bad; 1340 } 1341 if (ipt->ipt_ptr > 1342 ipt->ipt_len - (int)sizeof(int32_t)) { 1343 if (++ipt->ipt_oflw == 0) { 1344 code = (u_char *)&ipt->ipt_ptr - 1345 (u_char *)ip; 1346 goto bad; 1347 } 1348 break; 1349 } 1350 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1351 switch (ipt->ipt_flg) { 1352 1353 case IPOPT_TS_TSONLY: 1354 break; 1355 1356 case IPOPT_TS_TSANDADDR: 1357 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1358 sizeof(struct in_addr) > ipt->ipt_len) { 1359 code = (u_char *)&ipt->ipt_ptr - 1360 (u_char *)ip; 1361 goto bad; 1362 } 1363 ipaddr.sin_addr = dst; 1364 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1365 m->m_pkthdr.rcvif); 1366 if (ia == 0) 1367 continue; 1368 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1369 sizeof(struct in_addr)); 1370 ipt->ipt_ptr += sizeof(struct in_addr); 1371 break; 1372 1373 case IPOPT_TS_PRESPEC: 1374 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1375 sizeof(struct in_addr) > ipt->ipt_len) { 1376 code = (u_char *)&ipt->ipt_ptr - 1377 (u_char *)ip; 1378 goto bad; 1379 } 1380 (void)memcpy(&ipaddr.sin_addr, sin, 1381 sizeof(struct in_addr)); 1382 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1383 continue; 1384 ipt->ipt_ptr += sizeof(struct in_addr); 1385 break; 1386 1387 default: 1388 /* XXX can't take &ipt->ipt_flg */ 1389 code = (u_char *)&ipt->ipt_ptr - 1390 (u_char *)ip + 1; 1391 goto bad; 1392 } 1393 ntime = iptime(); 1394 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1395 sizeof(n_time)); 1396 ipt->ipt_ptr += sizeof(n_time); 1397 } 1398 } 1399 if (forward && ipforwarding) { 1400 ip_forward(m, 1); 1401 return (1); 1402 } 1403 return (0); 1404bad: 1405 icmp_error(m, type, code, 0, 0); 1406 ipstat.ips_badoptions++; 1407 return (1); 1408} 1409 1410/* 1411 * Given address of next destination (final or next hop), 1412 * return internet address info of interface to be used to get there. 1413 */ 1414static struct in_ifaddr * 1415ip_rtaddr(dst) 1416 struct in_addr dst; 1417{ 1418 register struct sockaddr_in *sin; 1419 1420 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1421 1422 if (ipforward_rt.ro_rt == 0 || 1423 !(ipforward_rt.ro_rt->rt_flags & RTF_UP) || 1424 dst.s_addr != sin->sin_addr.s_addr) { 1425 if (ipforward_rt.ro_rt) { 1426 RTFREE(ipforward_rt.ro_rt); 1427 ipforward_rt.ro_rt = 0; 1428 } 1429 sin->sin_family = AF_INET; 1430 sin->sin_len = sizeof(*sin); 1431 sin->sin_addr = dst; 1432 1433 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1434 } 1435 if (ipforward_rt.ro_rt == 0) 1436 return ((struct in_ifaddr *)0); 1437 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1438} 1439 1440/* 1441 * Save incoming source route for use in replies, 1442 * to be picked up later by ip_srcroute if the receiver is interested. 1443 */ 1444void 1445save_rte(option, dst) 1446 u_char *option; 1447 struct in_addr dst; 1448{ 1449 unsigned olen; 1450 1451 olen = option[IPOPT_OLEN]; 1452#ifdef DIAGNOSTIC 1453 if (ipprintfs) 1454 printf("save_rte: olen %d\n", olen); 1455#endif 1456 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1457 return; 1458 bcopy(option, ip_srcrt.srcopt, olen); 1459 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1460 ip_srcrt.dst = dst; 1461} 1462 1463/* 1464 * Retrieve incoming source route for use in replies, 1465 * in the same form used by setsockopt. 1466 * The first hop is placed before the options, will be removed later. 1467 */ 1468struct mbuf * 1469ip_srcroute() 1470{ 1471 register struct in_addr *p, *q; 1472 register struct mbuf *m; 1473 1474 if (ip_nhops == 0) 1475 return ((struct mbuf *)0); 1476 m = m_get(M_DONTWAIT, MT_HEADER); 1477 if (m == 0) 1478 return ((struct mbuf *)0); 1479 1480#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1481 1482 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1483 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1484 OPTSIZ; 1485#ifdef DIAGNOSTIC 1486 if (ipprintfs) 1487 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1488#endif 1489 1490 /* 1491 * First save first hop for return route 1492 */ 1493 p = &ip_srcrt.route[ip_nhops - 1]; 1494 *(mtod(m, struct in_addr *)) = *p--; 1495#ifdef DIAGNOSTIC 1496 if (ipprintfs) 1497 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1498#endif 1499 1500 /* 1501 * Copy option fields and padding (nop) to mbuf. 1502 */ 1503 ip_srcrt.nop = IPOPT_NOP; 1504 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1505 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1506 &ip_srcrt.nop, OPTSIZ); 1507 q = (struct in_addr *)(mtod(m, caddr_t) + 1508 sizeof(struct in_addr) + OPTSIZ); 1509#undef OPTSIZ 1510 /* 1511 * Record return path as an IP source route, 1512 * reversing the path (pointers are now aligned). 1513 */ 1514 while (p >= ip_srcrt.route) { 1515#ifdef DIAGNOSTIC 1516 if (ipprintfs) 1517 printf(" %lx", (u_long)ntohl(q->s_addr)); 1518#endif 1519 *q++ = *p--; 1520 } 1521 /* 1522 * Last hop goes to final destination. 1523 */ 1524 *q = ip_srcrt.dst; 1525#ifdef DIAGNOSTIC 1526 if (ipprintfs) 1527 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1528#endif 1529 return (m); 1530} 1531 1532/* 1533 * Strip out IP options, at higher 1534 * level protocol in the kernel. 1535 * Second argument is buffer to which options 1536 * will be moved, and return value is their length. 1537 * XXX should be deleted; last arg currently ignored. 1538 */ 1539void 1540ip_stripoptions(m, mopt) 1541 register struct mbuf *m; 1542 struct mbuf *mopt; 1543{ 1544 register int i; 1545 struct ip *ip = mtod(m, struct ip *); 1546 register caddr_t opts; 1547 int olen; 1548 1549 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1550 opts = (caddr_t)(ip + 1); 1551 i = m->m_len - (sizeof (struct ip) + olen); 1552 bcopy(opts + olen, opts, (unsigned)i); 1553 m->m_len -= olen; 1554 if (m->m_flags & M_PKTHDR) 1555 m->m_pkthdr.len -= olen; 1556 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1557} 1558 1559u_char inetctlerrmap[PRC_NCMDS] = { 1560 0, 0, 0, 0, 1561 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1562 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1563 EMSGSIZE, EHOSTUNREACH, 0, 0, 1564 0, 0, 0, 0, 1565 ENOPROTOOPT, ENETRESET 1566}; 1567 1568/* 1569 * Forward a packet. If some error occurs return the sender 1570 * an icmp packet. Note we can't always generate a meaningful 1571 * icmp message because icmp doesn't have a large enough repertoire 1572 * of codes and types. 1573 * 1574 * If not forwarding, just drop the packet. This could be confusing 1575 * if ipforwarding was zero but some routing protocol was advancing 1576 * us as a gateway to somewhere. However, we must let the routing 1577 * protocol deal with that. 1578 * 1579 * The srcrt parameter indicates whether the packet is being forwarded 1580 * via a source route. 1581 */ 1582static void 1583ip_forward(m, srcrt) 1584 struct mbuf *m; 1585 int srcrt; 1586{ 1587 register struct ip *ip = mtod(m, struct ip *); 1588 register struct rtentry *rt; 1589 int error, type = 0, code = 0; 1590 struct mbuf *mcopy; 1591 n_long dest; 1592 struct ifnet *destifp; 1593#ifdef IPSEC 1594 struct ifnet dummyifp; 1595#endif 1596 1597 dest = 0; 1598#ifdef DIAGNOSTIC 1599 if (ipprintfs) 1600 printf("forward: src %lx dst %lx ttl %x\n", 1601 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1602 ip->ip_ttl); 1603#endif 1604 1605 1606 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1607 ipstat.ips_cantforward++; 1608 m_freem(m); 1609 return; 1610 } 1611#ifdef IPSTEALTH 1612 if (!ipstealth) { 1613#endif 1614 if (ip->ip_ttl <= IPTTLDEC) { 1615 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1616 dest, 0); 1617 return; 1618 } 1619#ifdef IPSTEALTH 1620 } 1621#endif 1622 1623 if (ip_rtaddr(ip->ip_dst) == 0) { 1624 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1625 return; 1626 } else 1627 rt = ipforward_rt.ro_rt; 1628 1629 /* 1630 * Save the IP header and at most 8 bytes of the payload, 1631 * in case we need to generate an ICMP message to the src. 1632 * 1633 * We don't use m_copy() because it might return a reference 1634 * to a shared cluster. Both this function and ip_output() 1635 * assume exclusive access to the IP header in `m', so any 1636 * data in a cluster may change before we reach icmp_error(). 1637 */ 1638 MGET(mcopy, M_DONTWAIT, m->m_type); 1639 if (mcopy != NULL) { 1640 M_COPY_PKTHDR(mcopy, m); 1641 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1642 (int)ip->ip_len); 1643 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1644 } 1645 1646#ifdef IPSTEALTH 1647 if (!ipstealth) { 1648#endif 1649 ip->ip_ttl -= IPTTLDEC; 1650#ifdef IPSTEALTH 1651 } 1652#endif 1653 1654 /* 1655 * If forwarding packet using same interface that it came in on, 1656 * perhaps should send a redirect to sender to shortcut a hop. 1657 * Only send redirect if source is sending directly to us, 1658 * and if packet was not source routed (or has any options). 1659 * Also, don't send redirect if forwarding using a default route 1660 * or a route modified by a redirect. 1661 */ 1662#define satosin(sa) ((struct sockaddr_in *)(sa)) 1663 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1664 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1665 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1666 ipsendredirects && !srcrt) { 1667#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1668 u_long src = ntohl(ip->ip_src.s_addr); 1669 1670 if (RTA(rt) && 1671 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1672 if (rt->rt_flags & RTF_GATEWAY) 1673 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1674 else 1675 dest = ip->ip_dst.s_addr; 1676 /* Router requirements says to only send host redirects */ 1677 type = ICMP_REDIRECT; 1678 code = ICMP_REDIRECT_HOST; 1679#ifdef DIAGNOSTIC 1680 if (ipprintfs) 1681 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1682#endif 1683 } 1684 } 1685 1686 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1687 IP_FORWARDING, 0); 1688 if (error) 1689 ipstat.ips_cantforward++; 1690 else { 1691 ipstat.ips_forward++; 1692 if (type) 1693 ipstat.ips_redirectsent++; 1694 else { 1695 if (mcopy) { 1696 ipflow_create(&ipforward_rt, mcopy); 1697 m_freem(mcopy); 1698 } 1699 return; 1700 } 1701 } 1702 if (mcopy == NULL) 1703 return; 1704 destifp = NULL; 1705 1706 switch (error) { 1707 1708 case 0: /* forwarded, but need redirect */ 1709 /* type, code set above */ 1710 break; 1711 1712 case ENETUNREACH: /* shouldn't happen, checked above */ 1713 case EHOSTUNREACH: 1714 case ENETDOWN: 1715 case EHOSTDOWN: 1716 default: 1717 type = ICMP_UNREACH; 1718 code = ICMP_UNREACH_HOST; 1719 break; 1720 1721 case EMSGSIZE: 1722 type = ICMP_UNREACH; 1723 code = ICMP_UNREACH_NEEDFRAG; 1724#ifndef IPSEC 1725 if (ipforward_rt.ro_rt) 1726 destifp = ipforward_rt.ro_rt->rt_ifp; 1727#else 1728 /* 1729 * If the packet is routed over IPsec tunnel, tell the 1730 * originator the tunnel MTU. 1731 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1732 * XXX quickhack!!! 1733 */ 1734 if (ipforward_rt.ro_rt) { 1735 struct secpolicy *sp = NULL; 1736 int ipsecerror; 1737 int ipsechdr; 1738 struct route *ro; 1739 1740 sp = ipsec4_getpolicybyaddr(mcopy, 1741 IPSEC_DIR_OUTBOUND, 1742 IP_FORWARDING, 1743 &ipsecerror); 1744 1745 if (sp == NULL) 1746 destifp = ipforward_rt.ro_rt->rt_ifp; 1747 else { 1748 /* count IPsec header size */ 1749 ipsechdr = ipsec4_hdrsiz(mcopy, 1750 IPSEC_DIR_OUTBOUND, 1751 NULL); 1752 1753 /* 1754 * find the correct route for outer IPv4 1755 * header, compute tunnel MTU. 1756 * 1757 * XXX BUG ALERT 1758 * The "dummyifp" code relies upon the fact 1759 * that icmp_error() touches only ifp->if_mtu. 1760 */ 1761 /*XXX*/ 1762 destifp = NULL; 1763 if (sp->req != NULL 1764 && sp->req->sav != NULL 1765 && sp->req->sav->sah != NULL) { 1766 ro = &sp->req->sav->sah->sa_route; 1767 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1768 dummyifp.if_mtu = 1769 ro->ro_rt->rt_ifp->if_mtu; 1770 dummyifp.if_mtu -= ipsechdr; 1771 destifp = &dummyifp; 1772 } 1773 } 1774 1775 key_freesp(sp); 1776 } 1777 } 1778#endif /*IPSEC*/ 1779 ipstat.ips_cantfrag++; 1780 break; 1781 1782 case ENOBUFS: 1783 type = ICMP_SOURCEQUENCH; 1784 code = 0; 1785 break; 1786 1787 case EACCES: /* ipfw denied packet */ 1788 m_freem(mcopy); 1789 return; 1790 } 1791 icmp_error(mcopy, type, code, dest, destifp); 1792} 1793 1794void 1795ip_savecontrol(inp, mp, ip, m) 1796 register struct inpcb *inp; 1797 register struct mbuf **mp; 1798 register struct ip *ip; 1799 register struct mbuf *m; 1800{ 1801 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1802 struct timeval tv; 1803 1804 microtime(&tv); 1805 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1806 SCM_TIMESTAMP, SOL_SOCKET); 1807 if (*mp) 1808 mp = &(*mp)->m_next; 1809 } 1810 if (inp->inp_flags & INP_RECVDSTADDR) { 1811 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1812 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1813 if (*mp) 1814 mp = &(*mp)->m_next; 1815 } 1816#ifdef notyet 1817 /* XXX 1818 * Moving these out of udp_input() made them even more broken 1819 * than they already were. 1820 */ 1821 /* options were tossed already */ 1822 if (inp->inp_flags & INP_RECVOPTS) { 1823 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1824 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1825 if (*mp) 1826 mp = &(*mp)->m_next; 1827 } 1828 /* ip_srcroute doesn't do what we want here, need to fix */ 1829 if (inp->inp_flags & INP_RECVRETOPTS) { 1830 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1831 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1832 if (*mp) 1833 mp = &(*mp)->m_next; 1834 } 1835#endif 1836 if (inp->inp_flags & INP_RECVIF) { 1837 struct ifnet *ifp; 1838 struct sdlbuf { 1839 struct sockaddr_dl sdl; 1840 u_char pad[32]; 1841 } sdlbuf; 1842 struct sockaddr_dl *sdp; 1843 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1844 1845 if (((ifp = m->m_pkthdr.rcvif)) 1846 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1847 sdp = (struct sockaddr_dl *)(ifnet_addrs 1848 [ifp->if_index - 1]->ifa_addr); 1849 /* 1850 * Change our mind and don't try copy. 1851 */ 1852 if ((sdp->sdl_family != AF_LINK) 1853 || (sdp->sdl_len > sizeof(sdlbuf))) { 1854 goto makedummy; 1855 } 1856 bcopy(sdp, sdl2, sdp->sdl_len); 1857 } else { 1858makedummy: 1859 sdl2->sdl_len 1860 = offsetof(struct sockaddr_dl, sdl_data[0]); 1861 sdl2->sdl_family = AF_LINK; 1862 sdl2->sdl_index = 0; 1863 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1864 } 1865 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1866 IP_RECVIF, IPPROTO_IP); 1867 if (*mp) 1868 mp = &(*mp)->m_next; 1869 } 1870} 1871 1872int 1873ip_rsvp_init(struct socket *so) 1874{ 1875 if (so->so_type != SOCK_RAW || 1876 so->so_proto->pr_protocol != IPPROTO_RSVP) 1877 return EOPNOTSUPP; 1878 1879 if (ip_rsvpd != NULL) 1880 return EADDRINUSE; 1881 1882 ip_rsvpd = so; 1883 /* 1884 * This may seem silly, but we need to be sure we don't over-increment 1885 * the RSVP counter, in case something slips up. 1886 */ 1887 if (!ip_rsvp_on) { 1888 ip_rsvp_on = 1; 1889 rsvp_on++; 1890 } 1891 1892 return 0; 1893} 1894 1895int 1896ip_rsvp_done(void) 1897{ 1898 ip_rsvpd = NULL; 1899 /* 1900 * This may seem silly, but we need to be sure we don't over-decrement 1901 * the RSVP counter, in case something slips up. 1902 */ 1903 if (ip_rsvp_on) { 1904 ip_rsvp_on = 0; 1905 rsvp_on--; 1906 } 1907 return 0; 1908} 1909