ip_input.c revision 74454
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 74454 2001-03-19 09:16:16Z 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 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/mbuf.h> 51#include <sys/malloc.h> 52#include <sys/domain.h> 53#include <sys/protosw.h> 54#include <sys/socket.h> 55#include <sys/time.h> 56#include <sys/kernel.h> 57#include <sys/syslog.h> 58#include <sys/sysctl.h> 59 60#include <net/pfil.h> 61#include <net/if.h> 62#include <net/if_var.h> 63#include <net/if_dl.h> 64#include <net/route.h> 65#include <net/netisr.h> 66#include <net/intrq.h> 67 68#include <netinet/in.h> 69#include <netinet/in_systm.h> 70#include <netinet/in_var.h> 71#include <netinet/ip.h> 72#include <netinet/in_pcb.h> 73#include <netinet/ip_var.h> 74#include <netinet/ip_icmp.h> 75#include <machine/in_cksum.h> 76 77#include <netinet/ipprotosw.h> 78 79#include <sys/socketvar.h> 80 81#include <netinet/ip_fw.h> 82 83#ifdef IPSEC 84#include <netinet6/ipsec.h> 85#include <netkey/key.h> 86#endif 87 88#include "faith.h" 89#if defined(NFAITH) && NFAITH > 0 90#include <net/if_types.h> 91#endif 92 93#ifdef DUMMYNET 94#include <netinet/ip_dummynet.h> 95#endif 96 97int rsvp_on = 0; 98static int ip_rsvp_on; 99struct socket *ip_rsvpd; 100 101int ipforwarding = 0; 102SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 103 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 104 105static int ipsendredirects = 1; /* XXX */ 106SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 107 &ipsendredirects, 0, "Enable sending IP redirects"); 108 109int ip_defttl = IPDEFTTL; 110SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 111 &ip_defttl, 0, "Maximum TTL on IP packets"); 112 113static int ip_dosourceroute = 0; 114SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 115 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 116 117static int ip_acceptsourceroute = 0; 118SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 119 CTLFLAG_RW, &ip_acceptsourceroute, 0, 120 "Enable accepting source routed IP packets"); 121 122static int ip_keepfaith = 0; 123SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 124 &ip_keepfaith, 0, 125 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 126 127/* 128 * XXX - Setting ip_checkinterface mostly implements the receive side of 129 * the Strong ES model described in RFC 1122, but since the routing table 130 * and transmit implementation do not implement the Strong ES model, 131 * setting this to 1 results in an odd hybrid. 132 * 133 * XXX - ip_checkinterface currently must be disabled if you use ipnat 134 * to translate the destination address to another local interface. 135 * 136 * XXX - ip_checkinterface must be disabled if you add IP aliases 137 * to the loopback interface instead of the interface where the 138 * packets for those addresses are received. 139 */ 140static int ip_checkinterface = 1; 141SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 142 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 143 144#ifdef DIAGNOSTIC 145static int ipprintfs = 0; 146#endif 147 148extern struct domain inetdomain; 149extern struct ipprotosw inetsw[]; 150u_char ip_protox[IPPROTO_MAX]; 151static int ipqmaxlen = IFQ_MAXLEN; 152struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 153SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 154 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 155SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 156 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 157 158struct ipstat ipstat; 159SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD, 160 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 161 162/* Packet reassembly stuff */ 163#define IPREASS_NHASH_LOG2 6 164#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 165#define IPREASS_HMASK (IPREASS_NHASH - 1) 166#define IPREASS_HASH(x,y) \ 167 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 168 169static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH]; 170static int nipq = 0; /* total # of reass queues */ 171static int maxnipq; 172const int ipintrq_present = 1; 173 174#ifdef IPCTL_DEFMTU 175SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 176 &ip_mtu, 0, "Default MTU"); 177#endif 178 179#ifdef IPSTEALTH 180static int ipstealth = 0; 181SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 182 &ipstealth, 0, ""); 183#endif 184 185 186/* Firewall hooks */ 187ip_fw_chk_t *ip_fw_chk_ptr; 188ip_fw_ctl_t *ip_fw_ctl_ptr; 189int fw_enable = 1 ; 190 191#ifdef DUMMYNET 192ip_dn_ctl_t *ip_dn_ctl_ptr; 193#endif 194 195 196/* 197 * We need to save the IP options in case a protocol wants to respond 198 * to an incoming packet over the same route if the packet got here 199 * using IP source routing. This allows connection establishment and 200 * maintenance when the remote end is on a network that is not known 201 * to us. 202 */ 203static int ip_nhops = 0; 204static struct ip_srcrt { 205 struct in_addr dst; /* final destination */ 206 char nop; /* one NOP to align */ 207 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 208 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 209} ip_srcrt; 210 211struct sockaddr_in *ip_fw_fwd_addr; 212 213static void save_rte __P((u_char *, struct in_addr)); 214static int ip_dooptions __P((struct mbuf *)); 215static void ip_forward __P((struct mbuf *, int)); 216static void ip_freef __P((struct ipqhead *, struct ipq *)); 217#ifdef IPDIVERT 218static struct mbuf *ip_reass __P((struct mbuf *, struct ipqhead *, struct ipq *, u_int32_t *, u_int16_t *)); 219#else 220static struct mbuf *ip_reass __P((struct mbuf *, struct ipqhead *, struct ipq *)); 221#endif 222static struct in_ifaddr *ip_rtaddr __P((struct in_addr)); 223static void ipintr __P((void)); 224 225/* 226 * IP initialization: fill in IP protocol switch table. 227 * All protocols not implemented in kernel go to raw IP protocol handler. 228 */ 229void 230ip_init() 231{ 232 register struct ipprotosw *pr; 233 register int i; 234 235 TAILQ_INIT(&in_ifaddrhead); 236 pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 237 if (pr == 0) 238 panic("ip_init"); 239 for (i = 0; i < IPPROTO_MAX; i++) 240 ip_protox[i] = pr - inetsw; 241 for (pr = (struct ipprotosw *)inetdomain.dom_protosw; 242 pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++) 243 if (pr->pr_domain->dom_family == PF_INET && 244 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 245 ip_protox[pr->pr_protocol] = pr - inetsw; 246 247 for (i = 0; i < IPREASS_NHASH; i++) 248 TAILQ_INIT(&ipq[i]); 249 250 maxnipq = nmbclusters/4; 251 252 ip_id = time_second & 0xffff; 253 ipintrq.ifq_maxlen = ipqmaxlen; 254 mtx_init(&ipintrq.ifq_mtx, "ip_inq", MTX_DEF); 255 256 register_netisr(NETISR_IP, ipintr); 257} 258 259static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 260struct route ipforward_rt; 261 262/* 263 * Ip input routine. Checksum and byte swap header. If fragmented 264 * try to reassemble. Process options. Pass to next level. 265 */ 266void 267ip_input(struct mbuf *m) 268{ 269 struct ip *ip; 270 struct ipq *fp; 271 struct in_ifaddr *ia = NULL; 272 int i, hlen, checkif; 273 u_short sum; 274 u_int16_t divert_cookie; /* firewall cookie */ 275 struct in_addr pkt_dst; 276#ifdef IPDIVERT 277 u_int32_t divert_info = 0; /* packet divert/tee info */ 278#endif 279 struct ip_fw_chain *rule = NULL; 280#ifdef PFIL_HOOKS 281 struct packet_filter_hook *pfh; 282 struct mbuf *m0; 283 int rv; 284#endif /* PFIL_HOOKS */ 285 286#ifdef IPDIVERT 287 /* Get and reset firewall cookie */ 288 divert_cookie = ip_divert_cookie; 289 ip_divert_cookie = 0; 290#else 291 divert_cookie = 0; 292#endif 293 294#if defined(IPFIREWALL) && defined(DUMMYNET) 295 /* 296 * dummynet packet are prepended a vestigial mbuf with 297 * m_type = MT_DUMMYNET and m_data pointing to the matching 298 * rule. 299 */ 300 if (m->m_type == MT_DUMMYNET) { 301 rule = (struct ip_fw_chain *)(m->m_data) ; 302 m = m->m_next ; 303 ip = mtod(m, struct ip *); 304 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 305 goto iphack ; 306 } else 307 rule = NULL ; 308#endif 309 310#ifdef DIAGNOSTIC 311 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 312 panic("ip_input no HDR"); 313#endif 314 ipstat.ips_total++; 315 316 if (m->m_pkthdr.len < sizeof(struct ip)) 317 goto tooshort; 318 319 if (m->m_len < sizeof (struct ip) && 320 (m = m_pullup(m, sizeof (struct ip))) == 0) { 321 ipstat.ips_toosmall++; 322 return; 323 } 324 ip = mtod(m, struct ip *); 325 326 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 327 ipstat.ips_badvers++; 328 goto bad; 329 } 330 331 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 332 if (hlen < sizeof(struct ip)) { /* minimum header length */ 333 ipstat.ips_badhlen++; 334 goto bad; 335 } 336 if (hlen > m->m_len) { 337 if ((m = m_pullup(m, hlen)) == 0) { 338 ipstat.ips_badhlen++; 339 return; 340 } 341 ip = mtod(m, struct ip *); 342 } 343 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 344 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 345 } else { 346 if (hlen == sizeof(struct ip)) { 347 sum = in_cksum_hdr(ip); 348 } else { 349 sum = in_cksum(m, hlen); 350 } 351 } 352 if (sum) { 353 ipstat.ips_badsum++; 354 goto bad; 355 } 356 357 /* 358 * Convert fields to host representation. 359 */ 360 NTOHS(ip->ip_len); 361 if (ip->ip_len < hlen) { 362 ipstat.ips_badlen++; 363 goto bad; 364 } 365 NTOHS(ip->ip_off); 366 367 /* 368 * Check that the amount of data in the buffers 369 * is as at least much as the IP header would have us expect. 370 * Trim mbufs if longer than we expect. 371 * Drop packet if shorter than we expect. 372 */ 373 if (m->m_pkthdr.len < ip->ip_len) { 374tooshort: 375 ipstat.ips_tooshort++; 376 goto bad; 377 } 378 if (m->m_pkthdr.len > ip->ip_len) { 379 if (m->m_len == m->m_pkthdr.len) { 380 m->m_len = ip->ip_len; 381 m->m_pkthdr.len = ip->ip_len; 382 } else 383 m_adj(m, ip->ip_len - m->m_pkthdr.len); 384 } 385 386 /* 387 * Don't accept packets with a loopback destination address 388 * unless they arrived via the loopback interface. 389 */ 390 if ((ntohl(ip->ip_dst.s_addr) & IN_CLASSA_NET) == 391 (IN_LOOPBACKNET << IN_CLASSA_NSHIFT) && 392 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 393 goto bad; 394 } 395 396 /* 397 * IpHack's section. 398 * Right now when no processing on packet has done 399 * and it is still fresh out of network we do our black 400 * deals with it. 401 * - Firewall: deny/allow/divert 402 * - Xlate: translate packet's addr/port (NAT). 403 * - Pipe: pass pkt through dummynet. 404 * - Wrap: fake packet's addr/port <unimpl.> 405 * - Encapsulate: put it in another IP and send out. <unimp.> 406 */ 407 408#if defined(IPFIREWALL) && defined(DUMMYNET) 409iphack: 410#endif 411 412#ifdef PFIL_HOOKS 413 /* 414 * Run through list of hooks for input packets. If there are any 415 * filters which require that additional packets in the flow are 416 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 417 * Note that filters must _never_ set this flag, as another filter 418 * in the list may have previously cleared it. 419 */ 420 m0 = m; 421 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 422 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 423 if (pfh->pfil_func) { 424 rv = pfh->pfil_func(ip, hlen, 425 m->m_pkthdr.rcvif, 0, &m0); 426 if (rv) 427 return; 428 m = m0; 429 if (m == NULL) 430 return; 431 ip = mtod(m, struct ip *); 432 } 433#endif /* PFIL_HOOKS */ 434 435 if (fw_enable && ip_fw_chk_ptr) { 436#ifdef IPFIREWALL_FORWARD 437 /* 438 * If we've been forwarded from the output side, then 439 * skip the firewall a second time 440 */ 441 if (ip_fw_fwd_addr) 442 goto ours; 443#endif /* IPFIREWALL_FORWARD */ 444 /* 445 * See the comment in ip_output for the return values 446 * produced by the firewall. 447 */ 448 i = (*ip_fw_chk_ptr)(&ip, 449 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 450 if (i & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */ 451 if (m) 452 m_freem(m); 453 return ; 454 } 455 if (m == NULL) { /* Packet discarded by firewall */ 456 static int __debug=10; 457 if (__debug >0) { 458 printf("firewall returns NULL, please update!\n"); 459 __debug-- ; 460 } 461 return; 462 } 463 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 464 goto pass; 465#ifdef DUMMYNET 466 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) { 467 /* Send packet to the appropriate pipe */ 468 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 469 0); 470 return; 471 } 472#endif 473#ifdef IPDIVERT 474 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 475 /* Divert or tee packet */ 476 divert_info = i; 477 goto ours; 478 } 479#endif 480#ifdef IPFIREWALL_FORWARD 481 if (i == 0 && ip_fw_fwd_addr != NULL) 482 goto pass; 483#endif 484 /* 485 * if we get here, the packet must be dropped 486 */ 487 m_freem(m); 488 return; 489 } 490pass: 491 492 /* 493 * Process options and, if not destined for us, 494 * ship it on. ip_dooptions returns 1 when an 495 * error was detected (causing an icmp message 496 * to be sent and the original packet to be freed). 497 */ 498 ip_nhops = 0; /* for source routed packets */ 499 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 500#ifdef IPFIREWALL_FORWARD 501 ip_fw_fwd_addr = NULL; 502#endif 503 return; 504 } 505 506 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 507 * matter if it is destined to another node, or whether it is 508 * a multicast one, RSVP wants it! and prevents it from being forwarded 509 * anywhere else. Also checks if the rsvp daemon is running before 510 * grabbing the packet. 511 */ 512 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 513 goto ours; 514 515 /* 516 * Check our list of addresses, to see if the packet is for us. 517 * If we don't have any addresses, assume any unicast packet 518 * we receive might be for us (and let the upper layers deal 519 * with it). 520 */ 521 if (TAILQ_EMPTY(&in_ifaddrhead) && 522 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 523 goto ours; 524 525 /* 526 * Cache the destination address of the packet; this may be 527 * changed by use of 'ipfw fwd'. 528 */ 529 pkt_dst = ip_fw_fwd_addr == NULL ? 530 ip->ip_dst : ip_fw_fwd_addr->sin_addr; 531 532 /* 533 * Enable a consistency check between the destination address 534 * and the arrival interface for a unicast packet (the RFC 1122 535 * strong ES model) if IP forwarding is disabled and the packet 536 * is not locally generated and the packet is not subject to 537 * 'ipfw fwd'. 538 * 539 * XXX - Checking also should be disabled if the destination 540 * address is ipnat'ed to a different interface. 541 * 542 * XXX - Checking is incompatible with IP aliases added 543 * to the loopback interface instead of the interface where 544 * the packets are received. 545 */ 546 checkif = ip_checkinterface && (ipforwarding == 0) && 547 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 548 (ip_fw_fwd_addr == NULL); 549 550 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 551#define satosin(sa) ((struct sockaddr_in *)(sa)) 552 553#ifdef BOOTP_COMPAT 554 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 555 goto ours; 556#endif 557 /* 558 * If the address matches, verify that the packet 559 * arrived via the correct interface if checking is 560 * enabled. 561 */ 562 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 563 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 564 goto ours; 565 /* 566 * Only accept broadcast packets that arrive via the 567 * matching interface. Reception of forwarded directed 568 * broadcasts would be handled via ip_forward() and 569 * ether_output() with the loopback into the stack for 570 * SIMPLEX interfaces handled by ether_output(). 571 */ 572 if (ia->ia_ifp == m->m_pkthdr.rcvif && 573 ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 574 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 575 pkt_dst.s_addr) 576 goto ours; 577 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 578 goto ours; 579 } 580 } 581 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 582 struct in_multi *inm; 583 if (ip_mrouter) { 584 /* 585 * If we are acting as a multicast router, all 586 * incoming multicast packets are passed to the 587 * kernel-level multicast forwarding function. 588 * The packet is returned (relatively) intact; if 589 * ip_mforward() returns a non-zero value, the packet 590 * must be discarded, else it may be accepted below. 591 */ 592 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 593 ipstat.ips_cantforward++; 594 m_freem(m); 595 return; 596 } 597 598 /* 599 * The process-level routing demon needs to receive 600 * all multicast IGMP packets, whether or not this 601 * host belongs to their destination groups. 602 */ 603 if (ip->ip_p == IPPROTO_IGMP) 604 goto ours; 605 ipstat.ips_forward++; 606 } 607 /* 608 * See if we belong to the destination multicast group on the 609 * arrival interface. 610 */ 611 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 612 if (inm == NULL) { 613 ipstat.ips_notmember++; 614 m_freem(m); 615 return; 616 } 617 goto ours; 618 } 619 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 620 goto ours; 621 if (ip->ip_dst.s_addr == INADDR_ANY) 622 goto ours; 623 624#if defined(NFAITH) && 0 < NFAITH 625 /* 626 * FAITH(Firewall Aided Internet Translator) 627 */ 628 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 629 if (ip_keepfaith) { 630 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 631 goto ours; 632 } 633 m_freem(m); 634 return; 635 } 636#endif 637 /* 638 * Not for us; forward if possible and desirable. 639 */ 640 if (ipforwarding == 0) { 641 ipstat.ips_cantforward++; 642 m_freem(m); 643 } else 644 ip_forward(m, 0); 645#ifdef IPFIREWALL_FORWARD 646 ip_fw_fwd_addr = NULL; 647#endif 648 return; 649 650ours: 651 /* Count the packet in the ip address stats */ 652 if (ia != NULL) { 653 ia->ia_ifa.if_ipackets++; 654 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 655 } 656 657 /* 658 * If offset or IP_MF are set, must reassemble. 659 * Otherwise, nothing need be done. 660 * (We could look in the reassembly queue to see 661 * if the packet was previously fragmented, 662 * but it's not worth the time; just let them time out.) 663 */ 664 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 665 666 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 667 /* 668 * Look for queue of fragments 669 * of this datagram. 670 */ 671 TAILQ_FOREACH(fp, &ipq[sum], ipq_list) 672 if (ip->ip_id == fp->ipq_id && 673 ip->ip_src.s_addr == fp->ipq_src.s_addr && 674 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 675 ip->ip_p == fp->ipq_p) 676 goto found; 677 678 fp = 0; 679 680 /* check if there's a place for the new queue */ 681 if (nipq > maxnipq) { 682 /* 683 * drop something from the tail of the current queue 684 * before proceeding further 685 */ 686 struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead); 687 if (q == NULL) { /* gak */ 688 for (i = 0; i < IPREASS_NHASH; i++) { 689 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead); 690 if (r) { 691 ip_freef(&ipq[i], r); 692 break; 693 } 694 } 695 } else 696 ip_freef(&ipq[sum], q); 697 } 698found: 699 /* 700 * Adjust ip_len to not reflect header, 701 * convert offset of this to bytes. 702 */ 703 ip->ip_len -= hlen; 704 if (ip->ip_off & IP_MF) { 705 /* 706 * Make sure that fragments have a data length 707 * that's a non-zero multiple of 8 bytes. 708 */ 709 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 710 ipstat.ips_toosmall++; /* XXX */ 711 goto bad; 712 } 713 m->m_flags |= M_FRAG; 714 } 715 ip->ip_off <<= 3; 716 717 /* 718 * Attempt reassembly; if it succeeds, proceed. 719 */ 720 ipstat.ips_fragments++; 721 m->m_pkthdr.header = ip; 722#ifdef IPDIVERT 723 m = ip_reass(m, 724 &ipq[sum], fp, &divert_info, &divert_cookie); 725#else 726 m = ip_reass(m, &ipq[sum], fp); 727#endif 728 if (m == 0) { 729#ifdef IPFIREWALL_FORWARD 730 ip_fw_fwd_addr = NULL; 731#endif 732 return; 733 } 734 ipstat.ips_reassembled++; 735 ip = mtod(m, struct ip *); 736 /* Get the header length of the reassembled packet */ 737 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 738#ifdef IPDIVERT 739 /* Restore original checksum before diverting packet */ 740 if (divert_info != 0) { 741 ip->ip_len += hlen; 742 HTONS(ip->ip_len); 743 HTONS(ip->ip_off); 744 ip->ip_sum = 0; 745 if (hlen == sizeof(struct ip)) 746 ip->ip_sum = in_cksum_hdr(ip); 747 else 748 ip->ip_sum = in_cksum(m, hlen); 749 NTOHS(ip->ip_off); 750 NTOHS(ip->ip_len); 751 ip->ip_len -= hlen; 752 } 753#endif 754 } else 755 ip->ip_len -= hlen; 756 757#ifdef IPDIVERT 758 /* 759 * Divert or tee packet to the divert protocol if required. 760 * 761 * If divert_info is zero then cookie should be too, so we shouldn't 762 * need to clear them here. Assume divert_packet() does so also. 763 */ 764 if (divert_info != 0) { 765 struct mbuf *clone = NULL; 766 767 /* Clone packet if we're doing a 'tee' */ 768 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 769 clone = m_dup(m, M_DONTWAIT); 770 771 /* Restore packet header fields to original values */ 772 ip->ip_len += hlen; 773 HTONS(ip->ip_len); 774 HTONS(ip->ip_off); 775 776 /* Deliver packet to divert input routine */ 777 ip_divert_cookie = divert_cookie; 778 divert_packet(m, 1, divert_info & 0xffff); 779 ipstat.ips_delivered++; 780 781 /* If 'tee', continue with original packet */ 782 if (clone == NULL) 783 return; 784 m = clone; 785 ip = mtod(m, struct ip *); 786 } 787#endif 788 789 /* 790 * Switch out to protocol's input routine. 791 */ 792 ipstat.ips_delivered++; 793 { 794 int off = hlen, nh = ip->ip_p; 795 796 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh); 797#ifdef IPFIREWALL_FORWARD 798 ip_fw_fwd_addr = NULL; /* tcp needed it */ 799#endif 800 return; 801 } 802bad: 803#ifdef IPFIREWALL_FORWARD 804 ip_fw_fwd_addr = NULL; 805#endif 806 m_freem(m); 807} 808 809/* 810 * IP software interrupt routine - to go away sometime soon 811 */ 812static void 813ipintr(void) 814{ 815 struct mbuf *m; 816 817 while (1) { 818 IF_DEQUEUE(&ipintrq, m); 819 if (m == 0) 820 return; 821 ip_input(m); 822 } 823} 824 825/* 826 * Take incoming datagram fragment and try to reassemble it into 827 * whole datagram. If a chain for reassembly of this datagram already 828 * exists, then it is given as fp; otherwise have to make a chain. 829 * 830 * When IPDIVERT enabled, keep additional state with each packet that 831 * tells us if we need to divert or tee the packet we're building. 832 */ 833 834static struct mbuf * 835#ifdef IPDIVERT 836ip_reass(m, head, fp, divinfo, divcookie) 837#else 838ip_reass(m, head, fp) 839#endif 840 struct mbuf *m; 841 struct ipqhead *head; 842 struct ipq *fp; 843#ifdef IPDIVERT 844 u_int32_t *divinfo; 845 u_int16_t *divcookie; 846#endif 847{ 848 struct ip *ip = mtod(m, struct ip *); 849 register struct mbuf *p, *q, *nq; 850 struct mbuf *t; 851 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 852 int i, next; 853 854 /* 855 * Presence of header sizes in mbufs 856 * would confuse code below. 857 */ 858 m->m_data += hlen; 859 m->m_len -= hlen; 860 861 /* 862 * If first fragment to arrive, create a reassembly queue. 863 */ 864 if (fp == 0) { 865 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 866 goto dropfrag; 867 fp = mtod(t, struct ipq *); 868 TAILQ_INSERT_HEAD(head, fp, ipq_list); 869 nipq++; 870 fp->ipq_ttl = IPFRAGTTL; 871 fp->ipq_p = ip->ip_p; 872 fp->ipq_id = ip->ip_id; 873 fp->ipq_src = ip->ip_src; 874 fp->ipq_dst = ip->ip_dst; 875 fp->ipq_frags = m; 876 m->m_nextpkt = NULL; 877#ifdef IPDIVERT 878 fp->ipq_div_info = 0; 879 fp->ipq_div_cookie = 0; 880#endif 881 goto inserted; 882 } 883 884#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 885 886 /* 887 * Find a segment which begins after this one does. 888 */ 889 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 890 if (GETIP(q)->ip_off > ip->ip_off) 891 break; 892 893 /* 894 * If there is a preceding segment, it may provide some of 895 * our data already. If so, drop the data from the incoming 896 * segment. If it provides all of our data, drop us, otherwise 897 * stick new segment in the proper place. 898 * 899 * If some of the data is dropped from the the preceding 900 * segment, then it's checksum is invalidated. 901 */ 902 if (p) { 903 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 904 if (i > 0) { 905 if (i >= ip->ip_len) 906 goto dropfrag; 907 m_adj(m, i); 908 m->m_pkthdr.csum_flags = 0; 909 ip->ip_off += i; 910 ip->ip_len -= i; 911 } 912 m->m_nextpkt = p->m_nextpkt; 913 p->m_nextpkt = m; 914 } else { 915 m->m_nextpkt = fp->ipq_frags; 916 fp->ipq_frags = m; 917 } 918 919 /* 920 * While we overlap succeeding segments trim them or, 921 * if they are completely covered, dequeue them. 922 */ 923 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 924 q = nq) { 925 i = (ip->ip_off + ip->ip_len) - 926 GETIP(q)->ip_off; 927 if (i < GETIP(q)->ip_len) { 928 GETIP(q)->ip_len -= i; 929 GETIP(q)->ip_off += i; 930 m_adj(q, i); 931 q->m_pkthdr.csum_flags = 0; 932 break; 933 } 934 nq = q->m_nextpkt; 935 m->m_nextpkt = nq; 936 m_freem(q); 937 } 938 939inserted: 940 941#ifdef IPDIVERT 942 /* 943 * Transfer firewall instructions to the fragment structure. 944 * Any fragment diverting causes the whole packet to divert. 945 */ 946 fp->ipq_div_info = *divinfo; 947 fp->ipq_div_cookie = *divcookie; 948 *divinfo = 0; 949 *divcookie = 0; 950#endif 951 952 /* 953 * Check for complete reassembly. 954 */ 955 next = 0; 956 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 957 if (GETIP(q)->ip_off != next) 958 return (0); 959 next += GETIP(q)->ip_len; 960 } 961 /* Make sure the last packet didn't have the IP_MF flag */ 962 if (p->m_flags & M_FRAG) 963 return (0); 964 965 /* 966 * Reassembly is complete. Make sure the packet is a sane size. 967 */ 968 q = fp->ipq_frags; 969 ip = GETIP(q); 970 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 971 ipstat.ips_toolong++; 972 ip_freef(head, fp); 973 return (0); 974 } 975 976 /* 977 * Concatenate fragments. 978 */ 979 m = q; 980 t = m->m_next; 981 m->m_next = 0; 982 m_cat(m, t); 983 nq = q->m_nextpkt; 984 q->m_nextpkt = 0; 985 for (q = nq; q != NULL; q = nq) { 986 nq = q->m_nextpkt; 987 q->m_nextpkt = NULL; 988 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 989 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 990 m_cat(m, q); 991 } 992 993#ifdef IPDIVERT 994 /* 995 * Extract firewall instructions from the fragment structure. 996 */ 997 *divinfo = fp->ipq_div_info; 998 *divcookie = fp->ipq_div_cookie; 999#endif 1000 1001 /* 1002 * Create header for new ip packet by 1003 * modifying header of first packet; 1004 * dequeue and discard fragment reassembly header. 1005 * Make header visible. 1006 */ 1007 ip->ip_len = next; 1008 ip->ip_src = fp->ipq_src; 1009 ip->ip_dst = fp->ipq_dst; 1010 TAILQ_REMOVE(head, fp, ipq_list); 1011 nipq--; 1012 (void) m_free(dtom(fp)); 1013 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1014 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1015 /* some debugging cruft by sklower, below, will go away soon */ 1016 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1017 register int plen = 0; 1018 for (t = m; t; t = t->m_next) 1019 plen += t->m_len; 1020 m->m_pkthdr.len = plen; 1021 } 1022 return (m); 1023 1024dropfrag: 1025#ifdef IPDIVERT 1026 *divinfo = 0; 1027 *divcookie = 0; 1028#endif 1029 ipstat.ips_fragdropped++; 1030 m_freem(m); 1031 return (0); 1032 1033#undef GETIP 1034} 1035 1036/* 1037 * Free a fragment reassembly header and all 1038 * associated datagrams. 1039 */ 1040static void 1041ip_freef(fhp, fp) 1042 struct ipqhead *fhp; 1043 struct ipq *fp; 1044{ 1045 register struct mbuf *q; 1046 1047 while (fp->ipq_frags) { 1048 q = fp->ipq_frags; 1049 fp->ipq_frags = q->m_nextpkt; 1050 m_freem(q); 1051 } 1052 TAILQ_REMOVE(fhp, fp, ipq_list); 1053 (void) m_free(dtom(fp)); 1054 nipq--; 1055} 1056 1057/* 1058 * IP timer processing; 1059 * if a timer expires on a reassembly 1060 * queue, discard it. 1061 */ 1062void 1063ip_slowtimo() 1064{ 1065 register struct ipq *fp; 1066 int s = splnet(); 1067 int i; 1068 1069 for (i = 0; i < IPREASS_NHASH; i++) { 1070 for(fp = TAILQ_FIRST(&ipq[i]); fp;) { 1071 struct ipq *fpp; 1072 1073 fpp = fp; 1074 fp = TAILQ_NEXT(fp, ipq_list); 1075 if(--fpp->ipq_ttl == 0) { 1076 ipstat.ips_fragtimeout++; 1077 ip_freef(&ipq[i], fpp); 1078 } 1079 } 1080 } 1081 ipflow_slowtimo(); 1082 splx(s); 1083} 1084 1085/* 1086 * Drain off all datagram fragments. 1087 */ 1088void 1089ip_drain() 1090{ 1091 int i; 1092 1093 for (i = 0; i < IPREASS_NHASH; i++) { 1094 while(!TAILQ_EMPTY(&ipq[i])) { 1095 ipstat.ips_fragdropped++; 1096 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i])); 1097 } 1098 } 1099 in_rtqdrain(); 1100} 1101 1102/* 1103 * Do option processing on a datagram, 1104 * possibly discarding it if bad options are encountered, 1105 * or forwarding it if source-routed. 1106 * Returns 1 if packet has been forwarded/freed, 1107 * 0 if the packet should be processed further. 1108 */ 1109static int 1110ip_dooptions(m) 1111 struct mbuf *m; 1112{ 1113 register struct ip *ip = mtod(m, struct ip *); 1114 register u_char *cp; 1115 register struct ip_timestamp *ipt; 1116 register struct in_ifaddr *ia; 1117 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1118 struct in_addr *sin, dst; 1119 n_time ntime; 1120 1121 dst = ip->ip_dst; 1122 cp = (u_char *)(ip + 1); 1123 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1124 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1125 opt = cp[IPOPT_OPTVAL]; 1126 if (opt == IPOPT_EOL) 1127 break; 1128 if (opt == IPOPT_NOP) 1129 optlen = 1; 1130 else { 1131 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1132 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1133 goto bad; 1134 } 1135 optlen = cp[IPOPT_OLEN]; 1136 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1137 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1138 goto bad; 1139 } 1140 } 1141 switch (opt) { 1142 1143 default: 1144 break; 1145 1146 /* 1147 * Source routing with record. 1148 * Find interface with current destination address. 1149 * If none on this machine then drop if strictly routed, 1150 * or do nothing if loosely routed. 1151 * Record interface address and bring up next address 1152 * component. If strictly routed make sure next 1153 * address is on directly accessible net. 1154 */ 1155 case IPOPT_LSRR: 1156 case IPOPT_SSRR: 1157 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1158 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1159 goto bad; 1160 } 1161 ipaddr.sin_addr = ip->ip_dst; 1162 ia = (struct in_ifaddr *) 1163 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1164 if (ia == 0) { 1165 if (opt == IPOPT_SSRR) { 1166 type = ICMP_UNREACH; 1167 code = ICMP_UNREACH_SRCFAIL; 1168 goto bad; 1169 } 1170 if (!ip_dosourceroute) 1171 goto nosourcerouting; 1172 /* 1173 * Loose routing, and not at next destination 1174 * yet; nothing to do except forward. 1175 */ 1176 break; 1177 } 1178 off--; /* 0 origin */ 1179 if (off > optlen - (int)sizeof(struct in_addr)) { 1180 /* 1181 * End of source route. Should be for us. 1182 */ 1183 if (!ip_acceptsourceroute) 1184 goto nosourcerouting; 1185 save_rte(cp, ip->ip_src); 1186 break; 1187 } 1188 1189 if (!ip_dosourceroute) { 1190 if (ipforwarding) { 1191 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1192 /* 1193 * Acting as a router, so generate ICMP 1194 */ 1195nosourcerouting: 1196 strcpy(buf, inet_ntoa(ip->ip_dst)); 1197 log(LOG_WARNING, 1198 "attempted source route from %s to %s\n", 1199 inet_ntoa(ip->ip_src), buf); 1200 type = ICMP_UNREACH; 1201 code = ICMP_UNREACH_SRCFAIL; 1202 goto bad; 1203 } else { 1204 /* 1205 * Not acting as a router, so silently drop. 1206 */ 1207 ipstat.ips_cantforward++; 1208 m_freem(m); 1209 return (1); 1210 } 1211 } 1212 1213 /* 1214 * locate outgoing interface 1215 */ 1216 (void)memcpy(&ipaddr.sin_addr, cp + off, 1217 sizeof(ipaddr.sin_addr)); 1218 1219 if (opt == IPOPT_SSRR) { 1220#define INA struct in_ifaddr * 1221#define SA struct sockaddr * 1222 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1223 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1224 } else 1225 ia = ip_rtaddr(ipaddr.sin_addr); 1226 if (ia == 0) { 1227 type = ICMP_UNREACH; 1228 code = ICMP_UNREACH_SRCFAIL; 1229 goto bad; 1230 } 1231 ip->ip_dst = ipaddr.sin_addr; 1232 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1233 sizeof(struct in_addr)); 1234 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1235 /* 1236 * Let ip_intr's mcast routing check handle mcast pkts 1237 */ 1238 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1239 break; 1240 1241 case IPOPT_RR: 1242 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1243 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1244 goto bad; 1245 } 1246 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1247 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1248 goto bad; 1249 } 1250 /* 1251 * If no space remains, ignore. 1252 */ 1253 off--; /* 0 origin */ 1254 if (off > optlen - (int)sizeof(struct in_addr)) 1255 break; 1256 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1257 sizeof(ipaddr.sin_addr)); 1258 /* 1259 * locate outgoing interface; if we're the destination, 1260 * use the incoming interface (should be same). 1261 */ 1262 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1263 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1264 type = ICMP_UNREACH; 1265 code = ICMP_UNREACH_HOST; 1266 goto bad; 1267 } 1268 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1269 sizeof(struct in_addr)); 1270 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1271 break; 1272 1273 case IPOPT_TS: 1274 code = cp - (u_char *)ip; 1275 ipt = (struct ip_timestamp *)cp; 1276 if (ipt->ipt_len < 5) 1277 goto bad; 1278 if (ipt->ipt_ptr > 1279 ipt->ipt_len - (int)sizeof(int32_t)) { 1280 if (++ipt->ipt_oflw == 0) 1281 goto bad; 1282 break; 1283 } 1284 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1285 switch (ipt->ipt_flg) { 1286 1287 case IPOPT_TS_TSONLY: 1288 break; 1289 1290 case IPOPT_TS_TSANDADDR: 1291 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1292 sizeof(struct in_addr) > ipt->ipt_len) 1293 goto bad; 1294 ipaddr.sin_addr = dst; 1295 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1296 m->m_pkthdr.rcvif); 1297 if (ia == 0) 1298 continue; 1299 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1300 sizeof(struct in_addr)); 1301 ipt->ipt_ptr += sizeof(struct in_addr); 1302 break; 1303 1304 case IPOPT_TS_PRESPEC: 1305 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1306 sizeof(struct in_addr) > ipt->ipt_len) 1307 goto bad; 1308 (void)memcpy(&ipaddr.sin_addr, sin, 1309 sizeof(struct in_addr)); 1310 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1311 continue; 1312 ipt->ipt_ptr += sizeof(struct in_addr); 1313 break; 1314 1315 default: 1316 goto bad; 1317 } 1318 ntime = iptime(); 1319 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1320 sizeof(n_time)); 1321 ipt->ipt_ptr += sizeof(n_time); 1322 } 1323 } 1324 if (forward && ipforwarding) { 1325 ip_forward(m, 1); 1326 return (1); 1327 } 1328 return (0); 1329bad: 1330 icmp_error(m, type, code, 0, 0); 1331 ipstat.ips_badoptions++; 1332 return (1); 1333} 1334 1335/* 1336 * Given address of next destination (final or next hop), 1337 * return internet address info of interface to be used to get there. 1338 */ 1339static struct in_ifaddr * 1340ip_rtaddr(dst) 1341 struct in_addr dst; 1342{ 1343 register struct sockaddr_in *sin; 1344 1345 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1346 1347 if (ipforward_rt.ro_rt == 0 || 1348 !(ipforward_rt.ro_rt->rt_flags & RTF_UP) || 1349 dst.s_addr != sin->sin_addr.s_addr) { 1350 if (ipforward_rt.ro_rt) { 1351 RTFREE(ipforward_rt.ro_rt); 1352 ipforward_rt.ro_rt = 0; 1353 } 1354 sin->sin_family = AF_INET; 1355 sin->sin_len = sizeof(*sin); 1356 sin->sin_addr = dst; 1357 1358 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1359 } 1360 if (ipforward_rt.ro_rt == 0) 1361 return ((struct in_ifaddr *)0); 1362 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1363} 1364 1365/* 1366 * Save incoming source route for use in replies, 1367 * to be picked up later by ip_srcroute if the receiver is interested. 1368 */ 1369void 1370save_rte(option, dst) 1371 u_char *option; 1372 struct in_addr dst; 1373{ 1374 unsigned olen; 1375 1376 olen = option[IPOPT_OLEN]; 1377#ifdef DIAGNOSTIC 1378 if (ipprintfs) 1379 printf("save_rte: olen %d\n", olen); 1380#endif 1381 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1382 return; 1383 bcopy(option, ip_srcrt.srcopt, olen); 1384 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1385 ip_srcrt.dst = dst; 1386} 1387 1388/* 1389 * Retrieve incoming source route for use in replies, 1390 * in the same form used by setsockopt. 1391 * The first hop is placed before the options, will be removed later. 1392 */ 1393struct mbuf * 1394ip_srcroute() 1395{ 1396 register struct in_addr *p, *q; 1397 register struct mbuf *m; 1398 1399 if (ip_nhops == 0) 1400 return ((struct mbuf *)0); 1401 m = m_get(M_DONTWAIT, MT_HEADER); 1402 if (m == 0) 1403 return ((struct mbuf *)0); 1404 1405#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1406 1407 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1408 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1409 OPTSIZ; 1410#ifdef DIAGNOSTIC 1411 if (ipprintfs) 1412 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1413#endif 1414 1415 /* 1416 * First save first hop for return route 1417 */ 1418 p = &ip_srcrt.route[ip_nhops - 1]; 1419 *(mtod(m, struct in_addr *)) = *p--; 1420#ifdef DIAGNOSTIC 1421 if (ipprintfs) 1422 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1423#endif 1424 1425 /* 1426 * Copy option fields and padding (nop) to mbuf. 1427 */ 1428 ip_srcrt.nop = IPOPT_NOP; 1429 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1430 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1431 &ip_srcrt.nop, OPTSIZ); 1432 q = (struct in_addr *)(mtod(m, caddr_t) + 1433 sizeof(struct in_addr) + OPTSIZ); 1434#undef OPTSIZ 1435 /* 1436 * Record return path as an IP source route, 1437 * reversing the path (pointers are now aligned). 1438 */ 1439 while (p >= ip_srcrt.route) { 1440#ifdef DIAGNOSTIC 1441 if (ipprintfs) 1442 printf(" %lx", (u_long)ntohl(q->s_addr)); 1443#endif 1444 *q++ = *p--; 1445 } 1446 /* 1447 * Last hop goes to final destination. 1448 */ 1449 *q = ip_srcrt.dst; 1450#ifdef DIAGNOSTIC 1451 if (ipprintfs) 1452 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1453#endif 1454 return (m); 1455} 1456 1457/* 1458 * Strip out IP options, at higher 1459 * level protocol in the kernel. 1460 * Second argument is buffer to which options 1461 * will be moved, and return value is their length. 1462 * XXX should be deleted; last arg currently ignored. 1463 */ 1464void 1465ip_stripoptions(m, mopt) 1466 register struct mbuf *m; 1467 struct mbuf *mopt; 1468{ 1469 register int i; 1470 struct ip *ip = mtod(m, struct ip *); 1471 register caddr_t opts; 1472 int olen; 1473 1474 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1475 opts = (caddr_t)(ip + 1); 1476 i = m->m_len - (sizeof (struct ip) + olen); 1477 bcopy(opts + olen, opts, (unsigned)i); 1478 m->m_len -= olen; 1479 if (m->m_flags & M_PKTHDR) 1480 m->m_pkthdr.len -= olen; 1481 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1482} 1483 1484u_char inetctlerrmap[PRC_NCMDS] = { 1485 0, 0, 0, 0, 1486 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1487 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1488 EMSGSIZE, EHOSTUNREACH, 0, 0, 1489 0, 0, 0, 0, 1490 ENOPROTOOPT, ENETRESET 1491}; 1492 1493/* 1494 * Forward a packet. If some error occurs return the sender 1495 * an icmp packet. Note we can't always generate a meaningful 1496 * icmp message because icmp doesn't have a large enough repertoire 1497 * of codes and types. 1498 * 1499 * If not forwarding, just drop the packet. This could be confusing 1500 * if ipforwarding was zero but some routing protocol was advancing 1501 * us as a gateway to somewhere. However, we must let the routing 1502 * protocol deal with that. 1503 * 1504 * The srcrt parameter indicates whether the packet is being forwarded 1505 * via a source route. 1506 */ 1507static void 1508ip_forward(m, srcrt) 1509 struct mbuf *m; 1510 int srcrt; 1511{ 1512 register struct ip *ip = mtod(m, struct ip *); 1513 register struct rtentry *rt; 1514 int error, type = 0, code = 0; 1515 struct mbuf *mcopy; 1516 n_long dest; 1517 struct ifnet *destifp; 1518#ifdef IPSEC 1519 struct ifnet dummyifp; 1520#endif 1521 1522 dest = 0; 1523#ifdef DIAGNOSTIC 1524 if (ipprintfs) 1525 printf("forward: src %lx dst %lx ttl %x\n", 1526 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1527 ip->ip_ttl); 1528#endif 1529 1530 1531 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1532 ipstat.ips_cantforward++; 1533 m_freem(m); 1534 return; 1535 } 1536#ifdef IPSTEALTH 1537 if (!ipstealth) { 1538#endif 1539 if (ip->ip_ttl <= IPTTLDEC) { 1540 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1541 dest, 0); 1542 return; 1543 } 1544#ifdef IPSTEALTH 1545 } 1546#endif 1547 1548 if (ip_rtaddr(ip->ip_dst) == 0) { 1549 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1550 return; 1551 } else 1552 rt = ipforward_rt.ro_rt; 1553 1554 /* 1555 * Save the IP header and at most 8 bytes of the payload, 1556 * in case we need to generate an ICMP message to the src. 1557 * 1558 * We don't use m_copy() because it might return a reference 1559 * to a shared cluster. Both this function and ip_output() 1560 * assume exclusive access to the IP header in `m', so any 1561 * data in a cluster may change before we reach icmp_error(). 1562 */ 1563 MGET(mcopy, M_DONTWAIT, m->m_type); 1564 if (mcopy != NULL) { 1565 M_COPY_PKTHDR(mcopy, m); 1566 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1567 (int)ip->ip_len); 1568 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1569 } 1570 1571#ifdef IPSTEALTH 1572 if (!ipstealth) { 1573#endif 1574 ip->ip_ttl -= IPTTLDEC; 1575#ifdef IPSTEALTH 1576 } 1577#endif 1578 1579 /* 1580 * If forwarding packet using same interface that it came in on, 1581 * perhaps should send a redirect to sender to shortcut a hop. 1582 * Only send redirect if source is sending directly to us, 1583 * and if packet was not source routed (or has any options). 1584 * Also, don't send redirect if forwarding using a default route 1585 * or a route modified by a redirect. 1586 */ 1587#define satosin(sa) ((struct sockaddr_in *)(sa)) 1588 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1589 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1590 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1591 ipsendredirects && !srcrt) { 1592#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1593 u_long src = ntohl(ip->ip_src.s_addr); 1594 1595 if (RTA(rt) && 1596 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1597 if (rt->rt_flags & RTF_GATEWAY) 1598 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1599 else 1600 dest = ip->ip_dst.s_addr; 1601 /* Router requirements says to only send host redirects */ 1602 type = ICMP_REDIRECT; 1603 code = ICMP_REDIRECT_HOST; 1604#ifdef DIAGNOSTIC 1605 if (ipprintfs) 1606 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1607#endif 1608 } 1609 } 1610 1611 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1612 IP_FORWARDING, 0); 1613 if (error) 1614 ipstat.ips_cantforward++; 1615 else { 1616 ipstat.ips_forward++; 1617 if (type) 1618 ipstat.ips_redirectsent++; 1619 else { 1620 if (mcopy) { 1621 ipflow_create(&ipforward_rt, mcopy); 1622 m_freem(mcopy); 1623 } 1624 return; 1625 } 1626 } 1627 if (mcopy == NULL) 1628 return; 1629 destifp = NULL; 1630 1631 switch (error) { 1632 1633 case 0: /* forwarded, but need redirect */ 1634 /* type, code set above */ 1635 break; 1636 1637 case ENETUNREACH: /* shouldn't happen, checked above */ 1638 case EHOSTUNREACH: 1639 case ENETDOWN: 1640 case EHOSTDOWN: 1641 default: 1642 type = ICMP_UNREACH; 1643 code = ICMP_UNREACH_HOST; 1644 break; 1645 1646 case EMSGSIZE: 1647 type = ICMP_UNREACH; 1648 code = ICMP_UNREACH_NEEDFRAG; 1649#ifndef IPSEC 1650 if (ipforward_rt.ro_rt) 1651 destifp = ipforward_rt.ro_rt->rt_ifp; 1652#else 1653 /* 1654 * If the packet is routed over IPsec tunnel, tell the 1655 * originator the tunnel MTU. 1656 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1657 * XXX quickhack!!! 1658 */ 1659 if (ipforward_rt.ro_rt) { 1660 struct secpolicy *sp = NULL; 1661 int ipsecerror; 1662 int ipsechdr; 1663 struct route *ro; 1664 1665 sp = ipsec4_getpolicybyaddr(mcopy, 1666 IPSEC_DIR_OUTBOUND, 1667 IP_FORWARDING, 1668 &ipsecerror); 1669 1670 if (sp == NULL) 1671 destifp = ipforward_rt.ro_rt->rt_ifp; 1672 else { 1673 /* count IPsec header size */ 1674 ipsechdr = ipsec4_hdrsiz(mcopy, 1675 IPSEC_DIR_OUTBOUND, 1676 NULL); 1677 1678 /* 1679 * find the correct route for outer IPv4 1680 * header, compute tunnel MTU. 1681 * 1682 * XXX BUG ALERT 1683 * The "dummyifp" code relies upon the fact 1684 * that icmp_error() touches only ifp->if_mtu. 1685 */ 1686 /*XXX*/ 1687 destifp = NULL; 1688 if (sp->req != NULL 1689 && sp->req->sav != NULL 1690 && sp->req->sav->sah != NULL) { 1691 ro = &sp->req->sav->sah->sa_route; 1692 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1693 dummyifp.if_mtu = 1694 ro->ro_rt->rt_ifp->if_mtu; 1695 dummyifp.if_mtu -= ipsechdr; 1696 destifp = &dummyifp; 1697 } 1698 } 1699 1700 key_freesp(sp); 1701 } 1702 } 1703#endif /*IPSEC*/ 1704 ipstat.ips_cantfrag++; 1705 break; 1706 1707 case ENOBUFS: 1708 type = ICMP_SOURCEQUENCH; 1709 code = 0; 1710 break; 1711 1712 case EACCES: /* ipfw denied packet */ 1713 m_freem(mcopy); 1714 return; 1715 } 1716 icmp_error(mcopy, type, code, dest, destifp); 1717} 1718 1719void 1720ip_savecontrol(inp, mp, ip, m) 1721 register struct inpcb *inp; 1722 register struct mbuf **mp; 1723 register struct ip *ip; 1724 register struct mbuf *m; 1725{ 1726 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1727 struct timeval tv; 1728 1729 microtime(&tv); 1730 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1731 SCM_TIMESTAMP, SOL_SOCKET); 1732 if (*mp) 1733 mp = &(*mp)->m_next; 1734 } 1735 if (inp->inp_flags & INP_RECVDSTADDR) { 1736 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1737 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1738 if (*mp) 1739 mp = &(*mp)->m_next; 1740 } 1741#ifdef notyet 1742 /* XXX 1743 * Moving these out of udp_input() made them even more broken 1744 * than they already were. 1745 */ 1746 /* options were tossed already */ 1747 if (inp->inp_flags & INP_RECVOPTS) { 1748 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1749 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1750 if (*mp) 1751 mp = &(*mp)->m_next; 1752 } 1753 /* ip_srcroute doesn't do what we want here, need to fix */ 1754 if (inp->inp_flags & INP_RECVRETOPTS) { 1755 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1756 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1757 if (*mp) 1758 mp = &(*mp)->m_next; 1759 } 1760#endif 1761 if (inp->inp_flags & INP_RECVIF) { 1762 struct ifnet *ifp; 1763 struct sdlbuf { 1764 struct sockaddr_dl sdl; 1765 u_char pad[32]; 1766 } sdlbuf; 1767 struct sockaddr_dl *sdp; 1768 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1769 1770 if (((ifp = m->m_pkthdr.rcvif)) 1771 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1772 sdp = (struct sockaddr_dl *)(ifnet_addrs 1773 [ifp->if_index - 1]->ifa_addr); 1774 /* 1775 * Change our mind and don't try copy. 1776 */ 1777 if ((sdp->sdl_family != AF_LINK) 1778 || (sdp->sdl_len > sizeof(sdlbuf))) { 1779 goto makedummy; 1780 } 1781 bcopy(sdp, sdl2, sdp->sdl_len); 1782 } else { 1783makedummy: 1784 sdl2->sdl_len 1785 = offsetof(struct sockaddr_dl, sdl_data[0]); 1786 sdl2->sdl_family = AF_LINK; 1787 sdl2->sdl_index = 0; 1788 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1789 } 1790 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1791 IP_RECVIF, IPPROTO_IP); 1792 if (*mp) 1793 mp = &(*mp)->m_next; 1794 } 1795} 1796 1797int 1798ip_rsvp_init(struct socket *so) 1799{ 1800 if (so->so_type != SOCK_RAW || 1801 so->so_proto->pr_protocol != IPPROTO_RSVP) 1802 return EOPNOTSUPP; 1803 1804 if (ip_rsvpd != NULL) 1805 return EADDRINUSE; 1806 1807 ip_rsvpd = so; 1808 /* 1809 * This may seem silly, but we need to be sure we don't over-increment 1810 * the RSVP counter, in case something slips up. 1811 */ 1812 if (!ip_rsvp_on) { 1813 ip_rsvp_on = 1; 1814 rsvp_on++; 1815 } 1816 1817 return 0; 1818} 1819 1820int 1821ip_rsvp_done(void) 1822{ 1823 ip_rsvpd = NULL; 1824 /* 1825 * This may seem silly, but we need to be sure we don't over-decrement 1826 * the RSVP counter, in case something slips up. 1827 */ 1828 if (ip_rsvp_on) { 1829 ip_rsvp_on = 0; 1830 rsvp_on--; 1831 } 1832 return 0; 1833} 1834