ip_reass.c revision 73996
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 73996 2001-03-08 19:03:26Z iedowse $ 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 struct 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 ipq *)); 217#ifdef IPDIVERT 218static struct mbuf *ip_reass __P((struct mbuf *, 219 struct ipq *, struct ipq *, u_int32_t *, u_int16_t *)); 220#else 221static struct mbuf *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *)); 222#endif 223static struct in_ifaddr *ip_rtaddr __P((struct in_addr)); 224static void ipintr __P((void)); 225 226/* 227 * IP initialization: fill in IP protocol switch table. 228 * All protocols not implemented in kernel go to raw IP protocol handler. 229 */ 230void 231ip_init() 232{ 233 register struct ipprotosw *pr; 234 register int i; 235 236 TAILQ_INIT(&in_ifaddrhead); 237 pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 238 if (pr == 0) 239 panic("ip_init"); 240 for (i = 0; i < IPPROTO_MAX; i++) 241 ip_protox[i] = pr - inetsw; 242 for (pr = (struct ipprotosw *)inetdomain.dom_protosw; 243 pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++) 244 if (pr->pr_domain->dom_family == PF_INET && 245 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 246 ip_protox[pr->pr_protocol] = pr - inetsw; 247 248 for (i = 0; i < IPREASS_NHASH; i++) 249 ipq[i].next = ipq[i].prev = &ipq[i]; 250 251 maxnipq = nmbclusters/4; 252 253 ip_id = time_second & 0xffff; 254 ipintrq.ifq_maxlen = ipqmaxlen; 255 mtx_init(&ipintrq.ifq_mtx, "ip_inq", MTX_DEF); 256 257 register_netisr(NETISR_IP, ipintr); 258} 259 260static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 261static struct route ipforward_rt; 262 263/* 264 * Ip input routine. Checksum and byte swap header. If fragmented 265 * try to reassemble. Process options. Pass to next level. 266 */ 267void 268ip_input(struct mbuf *m) 269{ 270 struct ip *ip; 271 struct ipq *fp; 272 struct in_ifaddr *ia = NULL; 273 int i, hlen, checkif; 274 u_short sum; 275 u_int16_t divert_cookie; /* firewall cookie */ 276 struct in_addr pkt_dst; 277#ifdef IPDIVERT 278 u_int32_t divert_info = 0; /* packet divert/tee info */ 279#endif 280 struct ip_fw_chain *rule = NULL; 281#ifdef PFIL_HOOKS 282 struct packet_filter_hook *pfh; 283 struct mbuf *m0; 284 int rv; 285#endif /* PFIL_HOOKS */ 286 287#ifdef IPDIVERT 288 /* Get and reset firewall cookie */ 289 divert_cookie = ip_divert_cookie; 290 ip_divert_cookie = 0; 291#else 292 divert_cookie = 0; 293#endif 294 295#if defined(IPFIREWALL) && defined(DUMMYNET) 296 /* 297 * dummynet packet are prepended a vestigial mbuf with 298 * m_type = MT_DUMMYNET and m_data pointing to the matching 299 * rule. 300 */ 301 if (m->m_type == MT_DUMMYNET) { 302 rule = (struct ip_fw_chain *)(m->m_data) ; 303 m = m->m_next ; 304 ip = mtod(m, struct ip *); 305 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 306 goto iphack ; 307 } else 308 rule = NULL ; 309#endif 310 311#ifdef DIAGNOSTIC 312 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 313 panic("ip_input no HDR"); 314#endif 315 ipstat.ips_total++; 316 317 if (m->m_pkthdr.len < sizeof(struct ip)) 318 goto tooshort; 319 320 if (m->m_len < sizeof (struct ip) && 321 (m = m_pullup(m, sizeof (struct ip))) == 0) { 322 ipstat.ips_toosmall++; 323 return; 324 } 325 ip = mtod(m, struct ip *); 326 327 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 328 ipstat.ips_badvers++; 329 goto bad; 330 } 331 332 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 333 if (hlen < sizeof(struct ip)) { /* minimum header length */ 334 ipstat.ips_badhlen++; 335 goto bad; 336 } 337 if (hlen > m->m_len) { 338 if ((m = m_pullup(m, hlen)) == 0) { 339 ipstat.ips_badhlen++; 340 return; 341 } 342 ip = mtod(m, struct ip *); 343 } 344 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 345 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 346 } else { 347 if (hlen == sizeof(struct ip)) { 348 sum = in_cksum_hdr(ip); 349 } else { 350 sum = in_cksum(m, hlen); 351 } 352 } 353 if (sum) { 354 ipstat.ips_badsum++; 355 goto bad; 356 } 357 358 /* 359 * Convert fields to host representation. 360 */ 361 NTOHS(ip->ip_len); 362 if (ip->ip_len < hlen) { 363 ipstat.ips_badlen++; 364 goto bad; 365 } 366 NTOHS(ip->ip_off); 367 368 /* 369 * Check that the amount of data in the buffers 370 * is as at least much as the IP header would have us expect. 371 * Trim mbufs if longer than we expect. 372 * Drop packet if shorter than we expect. 373 */ 374 if (m->m_pkthdr.len < ip->ip_len) { 375tooshort: 376 ipstat.ips_tooshort++; 377 goto bad; 378 } 379 if (m->m_pkthdr.len > ip->ip_len) { 380 if (m->m_len == m->m_pkthdr.len) { 381 m->m_len = ip->ip_len; 382 m->m_pkthdr.len = ip->ip_len; 383 } else 384 m_adj(m, ip->ip_len - m->m_pkthdr.len); 385 } 386 387 /* 388 * Don't accept packets with a loopback destination address 389 * unless they arrived via the loopback interface. 390 */ 391 if ((ntohl(ip->ip_dst.s_addr) & IN_CLASSA_NET) == 392 (IN_LOOPBACKNET << IN_CLASSA_NSHIFT) && 393 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 394 goto bad; 395 } 396 397 /* 398 * IpHack's section. 399 * Right now when no processing on packet has done 400 * and it is still fresh out of network we do our black 401 * deals with it. 402 * - Firewall: deny/allow/divert 403 * - Xlate: translate packet's addr/port (NAT). 404 * - Pipe: pass pkt through dummynet. 405 * - Wrap: fake packet's addr/port <unimpl.> 406 * - Encapsulate: put it in another IP and send out. <unimp.> 407 */ 408 409#if defined(IPFIREWALL) && defined(DUMMYNET) 410iphack: 411#endif 412 413#ifdef PFIL_HOOKS 414 /* 415 * Run through list of hooks for input packets. If there are any 416 * filters which require that additional packets in the flow are 417 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 418 * Note that filters must _never_ set this flag, as another filter 419 * in the list may have previously cleared it. 420 */ 421 m0 = m; 422 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 423 for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 424 if (pfh->pfil_func) { 425 rv = pfh->pfil_func(ip, hlen, 426 m->m_pkthdr.rcvif, 0, &m0); 427 if (rv) 428 return; 429 m = m0; 430 if (m == NULL) 431 return; 432 ip = mtod(m, struct ip *); 433 } 434#endif /* PFIL_HOOKS */ 435 436 if (fw_enable && ip_fw_chk_ptr) { 437#ifdef IPFIREWALL_FORWARD 438 /* 439 * If we've been forwarded from the output side, then 440 * skip the firewall a second time 441 */ 442 if (ip_fw_fwd_addr) 443 goto ours; 444#endif /* IPFIREWALL_FORWARD */ 445 /* 446 * See the comment in ip_output for the return values 447 * produced by the firewall. 448 */ 449 i = (*ip_fw_chk_ptr)(&ip, 450 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 451 if (i & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */ 452 if (m) 453 m_freem(m); 454 return ; 455 } 456 if (m == NULL) { /* Packet discarded by firewall */ 457 static int __debug=10; 458 if (__debug >0) { 459 printf("firewall returns NULL, please update!\n"); 460 __debug-- ; 461 } 462 return; 463 } 464 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 465 goto pass; 466#ifdef DUMMYNET 467 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) { 468 /* Send packet to the appropriate pipe */ 469 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 470 0); 471 return; 472 } 473#endif 474#ifdef IPDIVERT 475 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 476 /* Divert or tee packet */ 477 divert_info = i; 478 goto ours; 479 } 480#endif 481#ifdef IPFIREWALL_FORWARD 482 if (i == 0 && ip_fw_fwd_addr != NULL) 483 goto pass; 484#endif 485 /* 486 * if we get here, the packet must be dropped 487 */ 488 m_freem(m); 489 return; 490 } 491pass: 492 493 /* 494 * Process options and, if not destined for us, 495 * ship it on. ip_dooptions returns 1 when an 496 * error was detected (causing an icmp message 497 * to be sent and the original packet to be freed). 498 */ 499 ip_nhops = 0; /* for source routed packets */ 500 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 501#ifdef IPFIREWALL_FORWARD 502 ip_fw_fwd_addr = NULL; 503#endif 504 return; 505 } 506 507 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 508 * matter if it is destined to another node, or whether it is 509 * a multicast one, RSVP wants it! and prevents it from being forwarded 510 * anywhere else. Also checks if the rsvp daemon is running before 511 * grabbing the packet. 512 */ 513 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 514 goto ours; 515 516 /* 517 * Check our list of addresses, to see if the packet is for us. 518 * If we don't have any addresses, assume any unicast packet 519 * we receive might be for us (and let the upper layers deal 520 * with it). 521 */ 522 if (TAILQ_EMPTY(&in_ifaddrhead) && 523 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 524 goto ours; 525 526 /* 527 * Cache the destination address of the packet; this may be 528 * changed by use of 'ipfw fwd'. 529 */ 530 pkt_dst = ip_fw_fwd_addr == NULL ? 531 ip->ip_dst : ip_fw_fwd_addr->sin_addr; 532 533 /* 534 * Enable a consistency check between the destination address 535 * and the arrival interface for a unicast packet (the RFC 1122 536 * strong ES model) if IP forwarding is disabled and the packet 537 * is not locally generated and the packet is not subject to 538 * 'ipfw fwd'. 539 * 540 * XXX - Checking also should be disabled if the destination 541 * address is ipnat'ed to a different interface. 542 * 543 * XXX - Checking is incompatible with IP aliases added 544 * to the loopback interface instead of the interface where 545 * the packets are received. 546 */ 547 checkif = ip_checkinterface && (ipforwarding == 0) && 548 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) && 549 (ip_fw_fwd_addr == NULL); 550 551 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 552#define satosin(sa) ((struct sockaddr_in *)(sa)) 553 554#ifdef BOOTP_COMPAT 555 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 556 goto ours; 557#endif 558 /* 559 * If the address matches, verify that the packet 560 * arrived via the correct interface if checking is 561 * enabled. 562 */ 563 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 564 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 565 goto ours; 566 /* 567 * Only accept broadcast packets that arrive via the 568 * matching interface. Reception of forwarded directed 569 * broadcasts would be handled via ip_forward() and 570 * ether_output() with the loopback into the stack for 571 * SIMPLEX interfaces handled by ether_output(). 572 */ 573 if (ia->ia_ifp == m->m_pkthdr.rcvif && 574 ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 575 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 576 pkt_dst.s_addr) 577 goto ours; 578 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 579 goto ours; 580 } 581 } 582 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 583 struct in_multi *inm; 584 if (ip_mrouter) { 585 /* 586 * If we are acting as a multicast router, all 587 * incoming multicast packets are passed to the 588 * kernel-level multicast forwarding function. 589 * The packet is returned (relatively) intact; if 590 * ip_mforward() returns a non-zero value, the packet 591 * must be discarded, else it may be accepted below. 592 */ 593 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 594 ipstat.ips_cantforward++; 595 m_freem(m); 596 return; 597 } 598 599 /* 600 * The process-level routing demon needs to receive 601 * all multicast IGMP packets, whether or not this 602 * host belongs to their destination groups. 603 */ 604 if (ip->ip_p == IPPROTO_IGMP) 605 goto ours; 606 ipstat.ips_forward++; 607 } 608 /* 609 * See if we belong to the destination multicast group on the 610 * arrival interface. 611 */ 612 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 613 if (inm == NULL) { 614 ipstat.ips_notmember++; 615 m_freem(m); 616 return; 617 } 618 goto ours; 619 } 620 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 621 goto ours; 622 if (ip->ip_dst.s_addr == INADDR_ANY) 623 goto ours; 624 625#if defined(NFAITH) && 0 < NFAITH 626 /* 627 * FAITH(Firewall Aided Internet Translator) 628 */ 629 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 630 if (ip_keepfaith) { 631 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 632 goto ours; 633 } 634 m_freem(m); 635 return; 636 } 637#endif 638 /* 639 * Not for us; forward if possible and desirable. 640 */ 641 if (ipforwarding == 0) { 642 ipstat.ips_cantforward++; 643 m_freem(m); 644 } else 645 ip_forward(m, 0); 646#ifdef IPFIREWALL_FORWARD 647 ip_fw_fwd_addr = NULL; 648#endif 649 return; 650 651ours: 652 /* Count the packet in the ip address stats */ 653 if (ia != NULL) { 654 ia->ia_ifa.if_ipackets++; 655 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 656 } 657 658 /* 659 * If offset or IP_MF are set, must reassemble. 660 * Otherwise, nothing need be done. 661 * (We could look in the reassembly queue to see 662 * if the packet was previously fragmented, 663 * but it's not worth the time; just let them time out.) 664 */ 665 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 666 667 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 668 /* 669 * Look for queue of fragments 670 * of this datagram. 671 */ 672 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 673 if (ip->ip_id == fp->ipq_id && 674 ip->ip_src.s_addr == fp->ipq_src.s_addr && 675 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 676 ip->ip_p == fp->ipq_p) 677 goto found; 678 679 fp = 0; 680 681 /* check if there's a place for the new queue */ 682 if (nipq > maxnipq) { 683 /* 684 * drop something from the tail of the current queue 685 * before proceeding further 686 */ 687 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 688 for (i = 0; i < IPREASS_NHASH; i++) { 689 if (ipq[i].prev != &ipq[i]) { 690 ip_freef(ipq[i].prev); 691 break; 692 } 693 } 694 } else 695 ip_freef(ipq[sum].prev); 696 } 697found: 698 /* 699 * Adjust ip_len to not reflect header, 700 * convert offset of this to bytes. 701 */ 702 ip->ip_len -= hlen; 703 if (ip->ip_off & IP_MF) { 704 /* 705 * Make sure that fragments have a data length 706 * that's a non-zero multiple of 8 bytes. 707 */ 708 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 709 ipstat.ips_toosmall++; /* XXX */ 710 goto bad; 711 } 712 m->m_flags |= M_FRAG; 713 } 714 ip->ip_off <<= 3; 715 716 /* 717 * Attempt reassembly; if it succeeds, proceed. 718 */ 719 ipstat.ips_fragments++; 720 m->m_pkthdr.header = ip; 721#ifdef IPDIVERT 722 m = ip_reass(m, 723 fp, &ipq[sum], &divert_info, &divert_cookie); 724#else 725 m = ip_reass(m, fp, &ipq[sum]); 726#endif 727 if (m == 0) { 728#ifdef IPFIREWALL_FORWARD 729 ip_fw_fwd_addr = NULL; 730#endif 731 return; 732 } 733 ipstat.ips_reassembled++; 734 ip = mtod(m, struct ip *); 735 /* Get the header length of the reassembled packet */ 736 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 737#ifdef IPDIVERT 738 /* Restore original checksum before diverting packet */ 739 if (divert_info != 0) { 740 ip->ip_len += hlen; 741 HTONS(ip->ip_len); 742 HTONS(ip->ip_off); 743 ip->ip_sum = 0; 744 if (hlen == sizeof(struct ip)) 745 ip->ip_sum = in_cksum_hdr(ip); 746 else 747 ip->ip_sum = in_cksum(m, hlen); 748 NTOHS(ip->ip_off); 749 NTOHS(ip->ip_len); 750 ip->ip_len -= hlen; 751 } 752#endif 753 } else 754 ip->ip_len -= hlen; 755 756#ifdef IPDIVERT 757 /* 758 * Divert or tee packet to the divert protocol if required. 759 * 760 * If divert_info is zero then cookie should be too, so we shouldn't 761 * need to clear them here. Assume divert_packet() does so also. 762 */ 763 if (divert_info != 0) { 764 struct mbuf *clone = NULL; 765 766 /* Clone packet if we're doing a 'tee' */ 767 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 768 clone = m_dup(m, M_DONTWAIT); 769 770 /* Restore packet header fields to original values */ 771 ip->ip_len += hlen; 772 HTONS(ip->ip_len); 773 HTONS(ip->ip_off); 774 775 /* Deliver packet to divert input routine */ 776 ip_divert_cookie = divert_cookie; 777 divert_packet(m, 1, divert_info & 0xffff); 778 ipstat.ips_delivered++; 779 780 /* If 'tee', continue with original packet */ 781 if (clone == NULL) 782 return; 783 m = clone; 784 ip = mtod(m, struct ip *); 785 } 786#endif 787 788 /* 789 * Switch out to protocol's input routine. 790 */ 791 ipstat.ips_delivered++; 792 { 793 int off = hlen, nh = ip->ip_p; 794 795 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh); 796#ifdef IPFIREWALL_FORWARD 797 ip_fw_fwd_addr = NULL; /* tcp needed it */ 798#endif 799 return; 800 } 801bad: 802#ifdef IPFIREWALL_FORWARD 803 ip_fw_fwd_addr = NULL; 804#endif 805 m_freem(m); 806} 807 808/* 809 * IP software interrupt routine - to go away sometime soon 810 */ 811static void 812ipintr(void) 813{ 814 struct mbuf *m; 815 816 while (1) { 817 IF_DEQUEUE(&ipintrq, m); 818 if (m == 0) 819 return; 820 ip_input(m); 821 } 822} 823 824/* 825 * Take incoming datagram fragment and try to reassemble it into 826 * whole datagram. If a chain for reassembly of this datagram already 827 * exists, then it is given as fp; otherwise have to make a chain. 828 * 829 * When IPDIVERT enabled, keep additional state with each packet that 830 * tells us if we need to divert or tee the packet we're building. 831 */ 832 833static struct mbuf * 834#ifdef IPDIVERT 835ip_reass(m, fp, where, divinfo, divcookie) 836#else 837ip_reass(m, fp, where) 838#endif 839 register struct mbuf *m; 840 register struct ipq *fp; 841 struct ipq *where; 842#ifdef IPDIVERT 843 u_int32_t *divinfo; 844 u_int16_t *divcookie; 845#endif 846{ 847 struct ip *ip = mtod(m, struct ip *); 848 register struct mbuf *p, *q, *nq; 849 struct mbuf *t; 850 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 851 int i, next; 852 853 /* 854 * Presence of header sizes in mbufs 855 * would confuse code below. 856 */ 857 m->m_data += hlen; 858 m->m_len -= hlen; 859 860 /* 861 * If first fragment to arrive, create a reassembly queue. 862 */ 863 if (fp == 0) { 864 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 865 goto dropfrag; 866 fp = mtod(t, struct ipq *); 867 insque(fp, where); 868 nipq++; 869 fp->ipq_ttl = IPFRAGTTL; 870 fp->ipq_p = ip->ip_p; 871 fp->ipq_id = ip->ip_id; 872 fp->ipq_src = ip->ip_src; 873 fp->ipq_dst = ip->ip_dst; 874 fp->ipq_frags = m; 875 m->m_nextpkt = NULL; 876#ifdef IPDIVERT 877 fp->ipq_div_info = 0; 878 fp->ipq_div_cookie = 0; 879#endif 880 goto inserted; 881 } 882 883#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 884 885 /* 886 * Find a segment which begins after this one does. 887 */ 888 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 889 if (GETIP(q)->ip_off > ip->ip_off) 890 break; 891 892 /* 893 * If there is a preceding segment, it may provide some of 894 * our data already. If so, drop the data from the incoming 895 * segment. If it provides all of our data, drop us, otherwise 896 * stick new segment in the proper place. 897 * 898 * If some of the data is dropped from the the preceding 899 * segment, then it's checksum is invalidated. 900 */ 901 if (p) { 902 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 903 if (i > 0) { 904 if (i >= ip->ip_len) 905 goto dropfrag; 906 m_adj(m, i); 907 m->m_pkthdr.csum_flags = 0; 908 ip->ip_off += i; 909 ip->ip_len -= i; 910 } 911 m->m_nextpkt = p->m_nextpkt; 912 p->m_nextpkt = m; 913 } else { 914 m->m_nextpkt = fp->ipq_frags; 915 fp->ipq_frags = m; 916 } 917 918 /* 919 * While we overlap succeeding segments trim them or, 920 * if they are completely covered, dequeue them. 921 */ 922 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 923 q = nq) { 924 i = (ip->ip_off + ip->ip_len) - 925 GETIP(q)->ip_off; 926 if (i < GETIP(q)->ip_len) { 927 GETIP(q)->ip_len -= i; 928 GETIP(q)->ip_off += i; 929 m_adj(q, i); 930 q->m_pkthdr.csum_flags = 0; 931 break; 932 } 933 nq = q->m_nextpkt; 934 m->m_nextpkt = nq; 935 m_freem(q); 936 } 937 938inserted: 939 940#ifdef IPDIVERT 941 /* 942 * Transfer firewall instructions to the fragment structure. 943 * Any fragment diverting causes the whole packet to divert. 944 */ 945 fp->ipq_div_info = *divinfo; 946 fp->ipq_div_cookie = *divcookie; 947 *divinfo = 0; 948 *divcookie = 0; 949#endif 950 951 /* 952 * Check for complete reassembly. 953 */ 954 next = 0; 955 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 956 if (GETIP(q)->ip_off != next) 957 return (0); 958 next += GETIP(q)->ip_len; 959 } 960 /* Make sure the last packet didn't have the IP_MF flag */ 961 if (p->m_flags & M_FRAG) 962 return (0); 963 964 /* 965 * Reassembly is complete. Make sure the packet is a sane size. 966 */ 967 q = fp->ipq_frags; 968 ip = GETIP(q); 969 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 970 ipstat.ips_toolong++; 971 ip_freef(fp); 972 return (0); 973 } 974 975 /* 976 * Concatenate fragments. 977 */ 978 m = q; 979 t = m->m_next; 980 m->m_next = 0; 981 m_cat(m, t); 982 nq = q->m_nextpkt; 983 q->m_nextpkt = 0; 984 for (q = nq; q != NULL; q = nq) { 985 nq = q->m_nextpkt; 986 q->m_nextpkt = NULL; 987 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 988 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 989 m_cat(m, q); 990 } 991 992#ifdef IPDIVERT 993 /* 994 * Extract firewall instructions from the fragment structure. 995 */ 996 *divinfo = fp->ipq_div_info; 997 *divcookie = fp->ipq_div_cookie; 998#endif 999 1000 /* 1001 * Create header for new ip packet by 1002 * modifying header of first packet; 1003 * dequeue and discard fragment reassembly header. 1004 * Make header visible. 1005 */ 1006 ip->ip_len = next; 1007 ip->ip_src = fp->ipq_src; 1008 ip->ip_dst = fp->ipq_dst; 1009 remque(fp); 1010 nipq--; 1011 (void) m_free(dtom(fp)); 1012 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1013 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1014 /* some debugging cruft by sklower, below, will go away soon */ 1015 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1016 register int plen = 0; 1017 for (t = m; t; t = t->m_next) 1018 plen += t->m_len; 1019 m->m_pkthdr.len = plen; 1020 } 1021 return (m); 1022 1023dropfrag: 1024#ifdef IPDIVERT 1025 *divinfo = 0; 1026 *divcookie = 0; 1027#endif 1028 ipstat.ips_fragdropped++; 1029 m_freem(m); 1030 return (0); 1031 1032#undef GETIP 1033} 1034 1035/* 1036 * Free a fragment reassembly header and all 1037 * associated datagrams. 1038 */ 1039static void 1040ip_freef(fp) 1041 struct ipq *fp; 1042{ 1043 register struct mbuf *q; 1044 1045 while (fp->ipq_frags) { 1046 q = fp->ipq_frags; 1047 fp->ipq_frags = q->m_nextpkt; 1048 m_freem(q); 1049 } 1050 remque(fp); 1051 (void) m_free(dtom(fp)); 1052 nipq--; 1053} 1054 1055/* 1056 * IP timer processing; 1057 * if a timer expires on a reassembly 1058 * queue, discard it. 1059 */ 1060void 1061ip_slowtimo() 1062{ 1063 register struct ipq *fp; 1064 int s = splnet(); 1065 int i; 1066 1067 for (i = 0; i < IPREASS_NHASH; i++) { 1068 fp = ipq[i].next; 1069 if (fp == 0) 1070 continue; 1071 while (fp != &ipq[i]) { 1072 --fp->ipq_ttl; 1073 fp = fp->next; 1074 if (fp->prev->ipq_ttl == 0) { 1075 ipstat.ips_fragtimeout++; 1076 ip_freef(fp->prev); 1077 } 1078 } 1079 } 1080 ipflow_slowtimo(); 1081 splx(s); 1082} 1083 1084/* 1085 * Drain off all datagram fragments. 1086 */ 1087void 1088ip_drain() 1089{ 1090 int i; 1091 1092 for (i = 0; i < IPREASS_NHASH; i++) { 1093 while (ipq[i].next != &ipq[i]) { 1094 ipstat.ips_fragdropped++; 1095 ip_freef(ipq[i].next); 1096 } 1097 } 1098 in_rtqdrain(); 1099} 1100 1101/* 1102 * Do option processing on a datagram, 1103 * possibly discarding it if bad options are encountered, 1104 * or forwarding it if source-routed. 1105 * Returns 1 if packet has been forwarded/freed, 1106 * 0 if the packet should be processed further. 1107 */ 1108static int 1109ip_dooptions(m) 1110 struct mbuf *m; 1111{ 1112 register struct ip *ip = mtod(m, struct ip *); 1113 register u_char *cp; 1114 register struct ip_timestamp *ipt; 1115 register struct in_ifaddr *ia; 1116 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1117 struct in_addr *sin, dst; 1118 n_time ntime; 1119 1120 dst = ip->ip_dst; 1121 cp = (u_char *)(ip + 1); 1122 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1123 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1124 opt = cp[IPOPT_OPTVAL]; 1125 if (opt == IPOPT_EOL) 1126 break; 1127 if (opt == IPOPT_NOP) 1128 optlen = 1; 1129 else { 1130 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1131 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1132 goto bad; 1133 } 1134 optlen = cp[IPOPT_OLEN]; 1135 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1136 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1137 goto bad; 1138 } 1139 } 1140 switch (opt) { 1141 1142 default: 1143 break; 1144 1145 /* 1146 * Source routing with record. 1147 * Find interface with current destination address. 1148 * If none on this machine then drop if strictly routed, 1149 * or do nothing if loosely routed. 1150 * Record interface address and bring up next address 1151 * component. If strictly routed make sure next 1152 * address is on directly accessible net. 1153 */ 1154 case IPOPT_LSRR: 1155 case IPOPT_SSRR: 1156 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1157 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1158 goto bad; 1159 } 1160 ipaddr.sin_addr = ip->ip_dst; 1161 ia = (struct in_ifaddr *) 1162 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1163 if (ia == 0) { 1164 if (opt == IPOPT_SSRR) { 1165 type = ICMP_UNREACH; 1166 code = ICMP_UNREACH_SRCFAIL; 1167 goto bad; 1168 } 1169 if (!ip_dosourceroute) 1170 goto nosourcerouting; 1171 /* 1172 * Loose routing, and not at next destination 1173 * yet; nothing to do except forward. 1174 */ 1175 break; 1176 } 1177 off--; /* 0 origin */ 1178 if (off > optlen - (int)sizeof(struct in_addr)) { 1179 /* 1180 * End of source route. Should be for us. 1181 */ 1182 if (!ip_acceptsourceroute) 1183 goto nosourcerouting; 1184 save_rte(cp, ip->ip_src); 1185 break; 1186 } 1187 1188 if (!ip_dosourceroute) { 1189 if (ipforwarding) { 1190 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1191 /* 1192 * Acting as a router, so generate ICMP 1193 */ 1194nosourcerouting: 1195 strcpy(buf, inet_ntoa(ip->ip_dst)); 1196 log(LOG_WARNING, 1197 "attempted source route from %s to %s\n", 1198 inet_ntoa(ip->ip_src), buf); 1199 type = ICMP_UNREACH; 1200 code = ICMP_UNREACH_SRCFAIL; 1201 goto bad; 1202 } else { 1203 /* 1204 * Not acting as a router, so silently drop. 1205 */ 1206 ipstat.ips_cantforward++; 1207 m_freem(m); 1208 return (1); 1209 } 1210 } 1211 1212 /* 1213 * locate outgoing interface 1214 */ 1215 (void)memcpy(&ipaddr.sin_addr, cp + off, 1216 sizeof(ipaddr.sin_addr)); 1217 1218 if (opt == IPOPT_SSRR) { 1219#define INA struct in_ifaddr * 1220#define SA struct sockaddr * 1221 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1222 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1223 } else 1224 ia = ip_rtaddr(ipaddr.sin_addr); 1225 if (ia == 0) { 1226 type = ICMP_UNREACH; 1227 code = ICMP_UNREACH_SRCFAIL; 1228 goto bad; 1229 } 1230 ip->ip_dst = ipaddr.sin_addr; 1231 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1232 sizeof(struct in_addr)); 1233 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1234 /* 1235 * Let ip_intr's mcast routing check handle mcast pkts 1236 */ 1237 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1238 break; 1239 1240 case IPOPT_RR: 1241 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1242 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1243 goto bad; 1244 } 1245 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1246 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1247 goto bad; 1248 } 1249 /* 1250 * If no space remains, ignore. 1251 */ 1252 off--; /* 0 origin */ 1253 if (off > optlen - (int)sizeof(struct in_addr)) 1254 break; 1255 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1256 sizeof(ipaddr.sin_addr)); 1257 /* 1258 * locate outgoing interface; if we're the destination, 1259 * use the incoming interface (should be same). 1260 */ 1261 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1262 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1263 type = ICMP_UNREACH; 1264 code = ICMP_UNREACH_HOST; 1265 goto bad; 1266 } 1267 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1268 sizeof(struct in_addr)); 1269 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1270 break; 1271 1272 case IPOPT_TS: 1273 code = cp - (u_char *)ip; 1274 ipt = (struct ip_timestamp *)cp; 1275 if (ipt->ipt_len < 5) 1276 goto bad; 1277 if (ipt->ipt_ptr > 1278 ipt->ipt_len - (int)sizeof(int32_t)) { 1279 if (++ipt->ipt_oflw == 0) 1280 goto bad; 1281 break; 1282 } 1283 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1284 switch (ipt->ipt_flg) { 1285 1286 case IPOPT_TS_TSONLY: 1287 break; 1288 1289 case IPOPT_TS_TSANDADDR: 1290 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1291 sizeof(struct in_addr) > ipt->ipt_len) 1292 goto bad; 1293 ipaddr.sin_addr = dst; 1294 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1295 m->m_pkthdr.rcvif); 1296 if (ia == 0) 1297 continue; 1298 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1299 sizeof(struct in_addr)); 1300 ipt->ipt_ptr += sizeof(struct in_addr); 1301 break; 1302 1303 case IPOPT_TS_PRESPEC: 1304 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1305 sizeof(struct in_addr) > ipt->ipt_len) 1306 goto bad; 1307 (void)memcpy(&ipaddr.sin_addr, sin, 1308 sizeof(struct in_addr)); 1309 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1310 continue; 1311 ipt->ipt_ptr += sizeof(struct in_addr); 1312 break; 1313 1314 default: 1315 goto bad; 1316 } 1317 ntime = iptime(); 1318 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1319 sizeof(n_time)); 1320 ipt->ipt_ptr += sizeof(n_time); 1321 } 1322 } 1323 if (forward && ipforwarding) { 1324 ip_forward(m, 1); 1325 return (1); 1326 } 1327 return (0); 1328bad: 1329 icmp_error(m, type, code, 0, 0); 1330 ipstat.ips_badoptions++; 1331 return (1); 1332} 1333 1334/* 1335 * Given address of next destination (final or next hop), 1336 * return internet address info of interface to be used to get there. 1337 */ 1338static struct in_ifaddr * 1339ip_rtaddr(dst) 1340 struct in_addr dst; 1341{ 1342 register struct sockaddr_in *sin; 1343 1344 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1345 1346 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1347 if (ipforward_rt.ro_rt) { 1348 RTFREE(ipforward_rt.ro_rt); 1349 ipforward_rt.ro_rt = 0; 1350 } 1351 sin->sin_family = AF_INET; 1352 sin->sin_len = sizeof(*sin); 1353 sin->sin_addr = dst; 1354 1355 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1356 } 1357 if (ipforward_rt.ro_rt == 0) 1358 return ((struct in_ifaddr *)0); 1359 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1360} 1361 1362/* 1363 * Save incoming source route for use in replies, 1364 * to be picked up later by ip_srcroute if the receiver is interested. 1365 */ 1366void 1367save_rte(option, dst) 1368 u_char *option; 1369 struct in_addr dst; 1370{ 1371 unsigned olen; 1372 1373 olen = option[IPOPT_OLEN]; 1374#ifdef DIAGNOSTIC 1375 if (ipprintfs) 1376 printf("save_rte: olen %d\n", olen); 1377#endif 1378 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1379 return; 1380 bcopy(option, ip_srcrt.srcopt, olen); 1381 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1382 ip_srcrt.dst = dst; 1383} 1384 1385/* 1386 * Retrieve incoming source route for use in replies, 1387 * in the same form used by setsockopt. 1388 * The first hop is placed before the options, will be removed later. 1389 */ 1390struct mbuf * 1391ip_srcroute() 1392{ 1393 register struct in_addr *p, *q; 1394 register struct mbuf *m; 1395 1396 if (ip_nhops == 0) 1397 return ((struct mbuf *)0); 1398 m = m_get(M_DONTWAIT, MT_HEADER); 1399 if (m == 0) 1400 return ((struct mbuf *)0); 1401 1402#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1403 1404 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1405 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1406 OPTSIZ; 1407#ifdef DIAGNOSTIC 1408 if (ipprintfs) 1409 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1410#endif 1411 1412 /* 1413 * First save first hop for return route 1414 */ 1415 p = &ip_srcrt.route[ip_nhops - 1]; 1416 *(mtod(m, struct in_addr *)) = *p--; 1417#ifdef DIAGNOSTIC 1418 if (ipprintfs) 1419 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1420#endif 1421 1422 /* 1423 * Copy option fields and padding (nop) to mbuf. 1424 */ 1425 ip_srcrt.nop = IPOPT_NOP; 1426 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1427 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1428 &ip_srcrt.nop, OPTSIZ); 1429 q = (struct in_addr *)(mtod(m, caddr_t) + 1430 sizeof(struct in_addr) + OPTSIZ); 1431#undef OPTSIZ 1432 /* 1433 * Record return path as an IP source route, 1434 * reversing the path (pointers are now aligned). 1435 */ 1436 while (p >= ip_srcrt.route) { 1437#ifdef DIAGNOSTIC 1438 if (ipprintfs) 1439 printf(" %lx", (u_long)ntohl(q->s_addr)); 1440#endif 1441 *q++ = *p--; 1442 } 1443 /* 1444 * Last hop goes to final destination. 1445 */ 1446 *q = ip_srcrt.dst; 1447#ifdef DIAGNOSTIC 1448 if (ipprintfs) 1449 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1450#endif 1451 return (m); 1452} 1453 1454/* 1455 * Strip out IP options, at higher 1456 * level protocol in the kernel. 1457 * Second argument is buffer to which options 1458 * will be moved, and return value is their length. 1459 * XXX should be deleted; last arg currently ignored. 1460 */ 1461void 1462ip_stripoptions(m, mopt) 1463 register struct mbuf *m; 1464 struct mbuf *mopt; 1465{ 1466 register int i; 1467 struct ip *ip = mtod(m, struct ip *); 1468 register caddr_t opts; 1469 int olen; 1470 1471 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1472 opts = (caddr_t)(ip + 1); 1473 i = m->m_len - (sizeof (struct ip) + olen); 1474 bcopy(opts + olen, opts, (unsigned)i); 1475 m->m_len -= olen; 1476 if (m->m_flags & M_PKTHDR) 1477 m->m_pkthdr.len -= olen; 1478 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1479} 1480 1481u_char inetctlerrmap[PRC_NCMDS] = { 1482 0, 0, 0, 0, 1483 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1484 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1485 EMSGSIZE, EHOSTUNREACH, 0, 0, 1486 0, 0, 0, 0, 1487 ENOPROTOOPT, ENETRESET 1488}; 1489 1490/* 1491 * Forward a packet. If some error occurs return the sender 1492 * an icmp packet. Note we can't always generate a meaningful 1493 * icmp message because icmp doesn't have a large enough repertoire 1494 * of codes and types. 1495 * 1496 * If not forwarding, just drop the packet. This could be confusing 1497 * if ipforwarding was zero but some routing protocol was advancing 1498 * us as a gateway to somewhere. However, we must let the routing 1499 * protocol deal with that. 1500 * 1501 * The srcrt parameter indicates whether the packet is being forwarded 1502 * via a source route. 1503 */ 1504static void 1505ip_forward(m, srcrt) 1506 struct mbuf *m; 1507 int srcrt; 1508{ 1509 register struct ip *ip = mtod(m, struct ip *); 1510 register struct sockaddr_in *sin; 1511 register struct rtentry *rt; 1512 int error, type = 0, code = 0; 1513 struct mbuf *mcopy; 1514 n_long dest; 1515 struct ifnet *destifp; 1516#ifdef IPSEC 1517 struct ifnet dummyifp; 1518#endif 1519 1520 dest = 0; 1521#ifdef DIAGNOSTIC 1522 if (ipprintfs) 1523 printf("forward: src %lx dst %lx ttl %x\n", 1524 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1525 ip->ip_ttl); 1526#endif 1527 1528 1529 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1530 ipstat.ips_cantforward++; 1531 m_freem(m); 1532 return; 1533 } 1534#ifdef IPSTEALTH 1535 if (!ipstealth) { 1536#endif 1537 if (ip->ip_ttl <= IPTTLDEC) { 1538 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1539 dest, 0); 1540 return; 1541 } 1542#ifdef IPSTEALTH 1543 } 1544#endif 1545 1546 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1547 if ((rt = ipforward_rt.ro_rt) == 0 || 1548 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1549 if (ipforward_rt.ro_rt) { 1550 RTFREE(ipforward_rt.ro_rt); 1551 ipforward_rt.ro_rt = 0; 1552 } 1553 sin->sin_family = AF_INET; 1554 sin->sin_len = sizeof(*sin); 1555 sin->sin_addr = ip->ip_dst; 1556 1557 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1558 if (ipforward_rt.ro_rt == 0) { 1559 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1560 return; 1561 } 1562 rt = ipforward_rt.ro_rt; 1563 } 1564 1565 /* 1566 * Save the IP header and at most 8 bytes of the payload, 1567 * in case we need to generate an ICMP message to the src. 1568 * 1569 * We don't use m_copy() because it might return a reference 1570 * to a shared cluster. Both this function and ip_output() 1571 * assume exclusive access to the IP header in `m', so any 1572 * data in a cluster may change before we reach icmp_error(). 1573 */ 1574 MGET(mcopy, M_DONTWAIT, m->m_type); 1575 if (mcopy != NULL) { 1576 M_COPY_PKTHDR(mcopy, m); 1577 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 1578 (int)ip->ip_len); 1579 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1580 } 1581 1582#ifdef IPSTEALTH 1583 if (!ipstealth) { 1584#endif 1585 ip->ip_ttl -= IPTTLDEC; 1586#ifdef IPSTEALTH 1587 } 1588#endif 1589 1590 /* 1591 * If forwarding packet using same interface that it came in on, 1592 * perhaps should send a redirect to sender to shortcut a hop. 1593 * Only send redirect if source is sending directly to us, 1594 * and if packet was not source routed (or has any options). 1595 * Also, don't send redirect if forwarding using a default route 1596 * or a route modified by a redirect. 1597 */ 1598#define satosin(sa) ((struct sockaddr_in *)(sa)) 1599 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1600 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1601 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1602 ipsendredirects && !srcrt) { 1603#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1604 u_long src = ntohl(ip->ip_src.s_addr); 1605 1606 if (RTA(rt) && 1607 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1608 if (rt->rt_flags & RTF_GATEWAY) 1609 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1610 else 1611 dest = ip->ip_dst.s_addr; 1612 /* Router requirements says to only send host redirects */ 1613 type = ICMP_REDIRECT; 1614 code = ICMP_REDIRECT_HOST; 1615#ifdef DIAGNOSTIC 1616 if (ipprintfs) 1617 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1618#endif 1619 } 1620 } 1621 1622 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1623 IP_FORWARDING, 0); 1624 if (error) 1625 ipstat.ips_cantforward++; 1626 else { 1627 ipstat.ips_forward++; 1628 if (type) 1629 ipstat.ips_redirectsent++; 1630 else { 1631 if (mcopy) { 1632 ipflow_create(&ipforward_rt, mcopy); 1633 m_freem(mcopy); 1634 } 1635 return; 1636 } 1637 } 1638 if (mcopy == NULL) 1639 return; 1640 destifp = NULL; 1641 1642 switch (error) { 1643 1644 case 0: /* forwarded, but need redirect */ 1645 /* type, code set above */ 1646 break; 1647 1648 case ENETUNREACH: /* shouldn't happen, checked above */ 1649 case EHOSTUNREACH: 1650 case ENETDOWN: 1651 case EHOSTDOWN: 1652 default: 1653 type = ICMP_UNREACH; 1654 code = ICMP_UNREACH_HOST; 1655 break; 1656 1657 case EMSGSIZE: 1658 type = ICMP_UNREACH; 1659 code = ICMP_UNREACH_NEEDFRAG; 1660#ifndef IPSEC 1661 if (ipforward_rt.ro_rt) 1662 destifp = ipforward_rt.ro_rt->rt_ifp; 1663#else 1664 /* 1665 * If the packet is routed over IPsec tunnel, tell the 1666 * originator the tunnel MTU. 1667 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1668 * XXX quickhack!!! 1669 */ 1670 if (ipforward_rt.ro_rt) { 1671 struct secpolicy *sp = NULL; 1672 int ipsecerror; 1673 int ipsechdr; 1674 struct route *ro; 1675 1676 sp = ipsec4_getpolicybyaddr(mcopy, 1677 IPSEC_DIR_OUTBOUND, 1678 IP_FORWARDING, 1679 &ipsecerror); 1680 1681 if (sp == NULL) 1682 destifp = ipforward_rt.ro_rt->rt_ifp; 1683 else { 1684 /* count IPsec header size */ 1685 ipsechdr = ipsec4_hdrsiz(mcopy, 1686 IPSEC_DIR_OUTBOUND, 1687 NULL); 1688 1689 /* 1690 * find the correct route for outer IPv4 1691 * header, compute tunnel MTU. 1692 * 1693 * XXX BUG ALERT 1694 * The "dummyifp" code relies upon the fact 1695 * that icmp_error() touches only ifp->if_mtu. 1696 */ 1697 /*XXX*/ 1698 destifp = NULL; 1699 if (sp->req != NULL 1700 && sp->req->sav != NULL 1701 && sp->req->sav->sah != NULL) { 1702 ro = &sp->req->sav->sah->sa_route; 1703 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1704 dummyifp.if_mtu = 1705 ro->ro_rt->rt_ifp->if_mtu; 1706 dummyifp.if_mtu -= ipsechdr; 1707 destifp = &dummyifp; 1708 } 1709 } 1710 1711 key_freesp(sp); 1712 } 1713 } 1714#endif /*IPSEC*/ 1715 ipstat.ips_cantfrag++; 1716 break; 1717 1718 case ENOBUFS: 1719 type = ICMP_SOURCEQUENCH; 1720 code = 0; 1721 break; 1722 1723 case EACCES: /* ipfw denied packet */ 1724 m_freem(mcopy); 1725 return; 1726 } 1727 icmp_error(mcopy, type, code, dest, destifp); 1728} 1729 1730void 1731ip_savecontrol(inp, mp, ip, m) 1732 register struct inpcb *inp; 1733 register struct mbuf **mp; 1734 register struct ip *ip; 1735 register struct mbuf *m; 1736{ 1737 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1738 struct timeval tv; 1739 1740 microtime(&tv); 1741 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1742 SCM_TIMESTAMP, SOL_SOCKET); 1743 if (*mp) 1744 mp = &(*mp)->m_next; 1745 } 1746 if (inp->inp_flags & INP_RECVDSTADDR) { 1747 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1748 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1749 if (*mp) 1750 mp = &(*mp)->m_next; 1751 } 1752#ifdef notyet 1753 /* XXX 1754 * Moving these out of udp_input() made them even more broken 1755 * than they already were. 1756 */ 1757 /* options were tossed already */ 1758 if (inp->inp_flags & INP_RECVOPTS) { 1759 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1760 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1761 if (*mp) 1762 mp = &(*mp)->m_next; 1763 } 1764 /* ip_srcroute doesn't do what we want here, need to fix */ 1765 if (inp->inp_flags & INP_RECVRETOPTS) { 1766 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1767 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1768 if (*mp) 1769 mp = &(*mp)->m_next; 1770 } 1771#endif 1772 if (inp->inp_flags & INP_RECVIF) { 1773 struct ifnet *ifp; 1774 struct sdlbuf { 1775 struct sockaddr_dl sdl; 1776 u_char pad[32]; 1777 } sdlbuf; 1778 struct sockaddr_dl *sdp; 1779 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1780 1781 if (((ifp = m->m_pkthdr.rcvif)) 1782 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1783 sdp = (struct sockaddr_dl *)(ifnet_addrs 1784 [ifp->if_index - 1]->ifa_addr); 1785 /* 1786 * Change our mind and don't try copy. 1787 */ 1788 if ((sdp->sdl_family != AF_LINK) 1789 || (sdp->sdl_len > sizeof(sdlbuf))) { 1790 goto makedummy; 1791 } 1792 bcopy(sdp, sdl2, sdp->sdl_len); 1793 } else { 1794makedummy: 1795 sdl2->sdl_len 1796 = offsetof(struct sockaddr_dl, sdl_data[0]); 1797 sdl2->sdl_family = AF_LINK; 1798 sdl2->sdl_index = 0; 1799 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1800 } 1801 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1802 IP_RECVIF, IPPROTO_IP); 1803 if (*mp) 1804 mp = &(*mp)->m_next; 1805 } 1806} 1807 1808int 1809ip_rsvp_init(struct socket *so) 1810{ 1811 if (so->so_type != SOCK_RAW || 1812 so->so_proto->pr_protocol != IPPROTO_RSVP) 1813 return EOPNOTSUPP; 1814 1815 if (ip_rsvpd != NULL) 1816 return EADDRINUSE; 1817 1818 ip_rsvpd = so; 1819 /* 1820 * This may seem silly, but we need to be sure we don't over-increment 1821 * the RSVP counter, in case something slips up. 1822 */ 1823 if (!ip_rsvp_on) { 1824 ip_rsvp_on = 1; 1825 rsvp_on++; 1826 } 1827 1828 return 0; 1829} 1830 1831int 1832ip_rsvp_done(void) 1833{ 1834 ip_rsvpd = NULL; 1835 /* 1836 * This may seem silly, but we need to be sure we don't over-decrement 1837 * the RSVP counter, in case something slips up. 1838 */ 1839 if (ip_rsvp_on) { 1840 ip_rsvp_on = 0; 1841 rsvp_on--; 1842 } 1843 return 0; 1844} 1845