ip_input.c revision 68169
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 68169 2000-11-01 11:21:45Z 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#ifdef DIAGNOSTIC 128static int ipprintfs = 0; 129#endif 130 131extern struct domain inetdomain; 132extern struct ipprotosw inetsw[]; 133u_char ip_protox[IPPROTO_MAX]; 134static int ipqmaxlen = IFQ_MAXLEN; 135struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 136SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 137 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 138SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 139 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 140 141struct ipstat ipstat; 142SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD, 143 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 144 145/* Packet reassembly stuff */ 146#define IPREASS_NHASH_LOG2 6 147#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 148#define IPREASS_HMASK (IPREASS_NHASH - 1) 149#define IPREASS_HASH(x,y) \ 150 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 151 152static struct ipq ipq[IPREASS_NHASH]; 153static int nipq = 0; /* total # of reass queues */ 154static int maxnipq; 155const int ipintrq_present = 1; 156 157#ifdef IPCTL_DEFMTU 158SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 159 &ip_mtu, 0, "Default MTU"); 160#endif 161 162#ifdef IPSTEALTH 163static int ipstealth = 0; 164SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 165 &ipstealth, 0, ""); 166#endif 167 168 169/* Firewall hooks */ 170ip_fw_chk_t *ip_fw_chk_ptr; 171ip_fw_ctl_t *ip_fw_ctl_ptr; 172int fw_enable = 1 ; 173 174#ifdef DUMMYNET 175ip_dn_ctl_t *ip_dn_ctl_ptr; 176#endif 177 178 179/* 180 * We need to save the IP options in case a protocol wants to respond 181 * to an incoming packet over the same route if the packet got here 182 * using IP source routing. This allows connection establishment and 183 * maintenance when the remote end is on a network that is not known 184 * to us. 185 */ 186static int ip_nhops = 0; 187static struct ip_srcrt { 188 struct in_addr dst; /* final destination */ 189 char nop; /* one NOP to align */ 190 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 191 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 192} ip_srcrt; 193 194struct sockaddr_in *ip_fw_fwd_addr; 195 196static void save_rte __P((u_char *, struct in_addr)); 197static int ip_dooptions __P((struct mbuf *)); 198static void ip_forward __P((struct mbuf *, int)); 199static void ip_freef __P((struct ipq *)); 200#ifdef IPDIVERT 201static struct mbuf *ip_reass __P((struct mbuf *, 202 struct ipq *, struct ipq *, u_int32_t *, u_int16_t *)); 203#else 204static struct mbuf *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *)); 205#endif 206static struct in_ifaddr *ip_rtaddr __P((struct in_addr)); 207static void ipintr __P((void)); 208 209/* 210 * IP initialization: fill in IP protocol switch table. 211 * All protocols not implemented in kernel go to raw IP protocol handler. 212 */ 213void 214ip_init() 215{ 216 register struct ipprotosw *pr; 217 register int i; 218 219 TAILQ_INIT(&in_ifaddrhead); 220 pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 221 if (pr == 0) 222 panic("ip_init"); 223 for (i = 0; i < IPPROTO_MAX; i++) 224 ip_protox[i] = pr - inetsw; 225 for (pr = (struct ipprotosw *)inetdomain.dom_protosw; 226 pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++) 227 if (pr->pr_domain->dom_family == PF_INET && 228 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 229 ip_protox[pr->pr_protocol] = pr - inetsw; 230 231 for (i = 0; i < IPREASS_NHASH; i++) 232 ipq[i].next = ipq[i].prev = &ipq[i]; 233 234 maxnipq = nmbclusters/4; 235 236 ip_id = time_second & 0xffff; 237 ipintrq.ifq_maxlen = ipqmaxlen; 238 239 register_netisr(NETISR_IP, ipintr); 240} 241 242static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 243static struct route ipforward_rt; 244 245/* 246 * Ip input routine. Checksum and byte swap header. If fragmented 247 * try to reassemble. Process options. Pass to next level. 248 */ 249void 250ip_input(struct mbuf *m) 251{ 252 struct ip *ip; 253 struct ipq *fp; 254 struct in_ifaddr *ia = NULL; 255 int i, hlen; 256 u_short sum; 257 u_int16_t divert_cookie; /* firewall cookie */ 258#ifdef IPDIVERT 259 u_int32_t divert_info = 0; /* packet divert/tee info */ 260#endif 261 struct ip_fw_chain *rule = NULL; 262#ifdef PFIL_HOOKS 263 struct packet_filter_hook *pfh; 264 struct mbuf *m0; 265 int rv; 266#endif /* PFIL_HOOKS */ 267 268#ifdef IPDIVERT 269 /* Get and reset firewall cookie */ 270 divert_cookie = ip_divert_cookie; 271 ip_divert_cookie = 0; 272#else 273 divert_cookie = 0; 274#endif 275 276#if defined(IPFIREWALL) && defined(DUMMYNET) 277 /* 278 * dummynet packet are prepended a vestigial mbuf with 279 * m_type = MT_DUMMYNET and m_data pointing to the matching 280 * rule. 281 */ 282 if (m->m_type == MT_DUMMYNET) { 283 rule = (struct ip_fw_chain *)(m->m_data) ; 284 m = m->m_next ; 285 ip = mtod(m, struct ip *); 286 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 287 goto iphack ; 288 } else 289 rule = NULL ; 290#endif 291 292#ifdef DIAGNOSTIC 293 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 294 panic("ip_input no HDR"); 295#endif 296 ipstat.ips_total++; 297 298 if (m->m_pkthdr.len < sizeof(struct ip)) 299 goto tooshort; 300 301 if (m->m_len < sizeof (struct ip) && 302 (m = m_pullup(m, sizeof (struct ip))) == 0) { 303 ipstat.ips_toosmall++; 304 return; 305 } 306 ip = mtod(m, struct ip *); 307 308 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 309 ipstat.ips_badvers++; 310 goto bad; 311 } 312 313 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 314 if (hlen < sizeof(struct ip)) { /* minimum header length */ 315 ipstat.ips_badhlen++; 316 goto bad; 317 } 318 if (hlen > m->m_len) { 319 if ((m = m_pullup(m, hlen)) == 0) { 320 ipstat.ips_badhlen++; 321 return; 322 } 323 ip = mtod(m, struct ip *); 324 } 325 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 326 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 327 } else { 328 if (hlen == sizeof(struct ip)) { 329 sum = in_cksum_hdr(ip); 330 } else { 331 sum = in_cksum(m, hlen); 332 } 333 } 334 if (sum) { 335 ipstat.ips_badsum++; 336 goto bad; 337 } 338 339 /* 340 * Convert fields to host representation. 341 */ 342 NTOHS(ip->ip_len); 343 if (ip->ip_len < hlen) { 344 ipstat.ips_badlen++; 345 goto bad; 346 } 347 NTOHS(ip->ip_off); 348 349 /* 350 * Check that the amount of data in the buffers 351 * is as at least much as the IP header would have us expect. 352 * Trim mbufs if longer than we expect. 353 * Drop packet if shorter than we expect. 354 */ 355 if (m->m_pkthdr.len < ip->ip_len) { 356tooshort: 357 ipstat.ips_tooshort++; 358 goto bad; 359 } 360 if (m->m_pkthdr.len > ip->ip_len) { 361 if (m->m_len == m->m_pkthdr.len) { 362 m->m_len = ip->ip_len; 363 m->m_pkthdr.len = ip->ip_len; 364 } else 365 m_adj(m, ip->ip_len - m->m_pkthdr.len); 366 } 367 /* 368 * IpHack's section. 369 * Right now when no processing on packet has done 370 * and it is still fresh out of network we do our black 371 * deals with it. 372 * - Firewall: deny/allow/divert 373 * - Xlate: translate packet's addr/port (NAT). 374 * - Pipe: pass pkt through dummynet. 375 * - Wrap: fake packet's addr/port <unimpl.> 376 * - Encapsulate: put it in another IP and send out. <unimp.> 377 */ 378 379#if defined(IPFIREWALL) && defined(DUMMYNET) 380iphack: 381#endif 382 383#ifdef PFIL_HOOKS 384 /* 385 * Run through list of hooks for input packets. If there are any 386 * filters which require that additional packets in the flow are 387 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 388 * Note that filters must _never_ set this flag, as another filter 389 * in the list may have previously cleared it. 390 */ 391 m0 = m; 392 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 393 for (; pfh; pfh = pfh->pfil_link.tqe_next) 394 if (pfh->pfil_func) { 395 rv = pfh->pfil_func(ip, hlen, 396 m->m_pkthdr.rcvif, 0, &m0); 397 if (rv) 398 return; 399 m = m0; 400 if (m == NULL) 401 return; 402 ip = mtod(m, struct ip *); 403 } 404#endif /* PFIL_HOOKS */ 405 406 if (fw_enable && ip_fw_chk_ptr) { 407#ifdef IPFIREWALL_FORWARD 408 /* 409 * If we've been forwarded from the output side, then 410 * skip the firewall a second time 411 */ 412 if (ip_fw_fwd_addr) 413 goto ours; 414#endif /* IPFIREWALL_FORWARD */ 415 /* 416 * See the comment in ip_output for the return values 417 * produced by the firewall. 418 */ 419 i = (*ip_fw_chk_ptr)(&ip, 420 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 421 if (m == NULL) /* Packet discarded by firewall */ 422 return; 423 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 424 goto pass; 425#ifdef DUMMYNET 426 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) { 427 /* Send packet to the appropriate pipe */ 428 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 429 0); 430 return; 431 } 432#endif 433#ifdef IPDIVERT 434 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 435 /* Divert or tee packet */ 436 divert_info = i; 437 goto ours; 438 } 439#endif 440#ifdef IPFIREWALL_FORWARD 441 if (i == 0 && ip_fw_fwd_addr != NULL) 442 goto pass; 443#endif 444 /* 445 * if we get here, the packet must be dropped 446 */ 447 m_freem(m); 448 return; 449 } 450pass: 451 452 /* 453 * Process options and, if not destined for us, 454 * ship it on. ip_dooptions returns 1 when an 455 * error was detected (causing an icmp message 456 * to be sent and the original packet to be freed). 457 */ 458 ip_nhops = 0; /* for source routed packets */ 459 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 460#ifdef IPFIREWALL_FORWARD 461 ip_fw_fwd_addr = NULL; 462#endif 463 return; 464 } 465 466 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 467 * matter if it is destined to another node, or whether it is 468 * a multicast one, RSVP wants it! and prevents it from being forwarded 469 * anywhere else. Also checks if the rsvp daemon is running before 470 * grabbing the packet. 471 */ 472 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 473 goto ours; 474 475 /* 476 * Check our list of addresses, to see if the packet is for us. 477 * If we don't have any addresses, assume any unicast packet 478 * we receive might be for us (and let the upper layers deal 479 * with it). 480 */ 481 if (TAILQ_EMPTY(&in_ifaddrhead) && 482 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 483 goto ours; 484 485 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 486 ia = TAILQ_NEXT(ia, ia_link)) { 487#define satosin(sa) ((struct sockaddr_in *)(sa)) 488 489#ifdef BOOTP_COMPAT 490 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 491 goto ours; 492#endif 493#ifdef IPFIREWALL_FORWARD 494 /* 495 * If the addr to forward to is one of ours, we pretend to 496 * be the destination for this packet. 497 */ 498 if (ip_fw_fwd_addr == NULL) { 499 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 500 goto ours; 501 } else if (IA_SIN(ia)->sin_addr.s_addr == 502 ip_fw_fwd_addr->sin_addr.s_addr) 503 goto ours; 504#else 505 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 506 goto ours; 507#endif 508 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 509 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 510 ip->ip_dst.s_addr) 511 goto ours; 512 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 513 goto ours; 514 } 515 } 516 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 517 struct in_multi *inm; 518 if (ip_mrouter) { 519 /* 520 * If we are acting as a multicast router, all 521 * incoming multicast packets are passed to the 522 * kernel-level multicast forwarding function. 523 * The packet is returned (relatively) intact; if 524 * ip_mforward() returns a non-zero value, the packet 525 * must be discarded, else it may be accepted below. 526 */ 527 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 528 ipstat.ips_cantforward++; 529 m_freem(m); 530 return; 531 } 532 533 /* 534 * The process-level routing demon needs to receive 535 * all multicast IGMP packets, whether or not this 536 * host belongs to their destination groups. 537 */ 538 if (ip->ip_p == IPPROTO_IGMP) 539 goto ours; 540 ipstat.ips_forward++; 541 } 542 /* 543 * See if we belong to the destination multicast group on the 544 * arrival interface. 545 */ 546 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 547 if (inm == NULL) { 548 ipstat.ips_notmember++; 549 m_freem(m); 550 return; 551 } 552 goto ours; 553 } 554 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 555 goto ours; 556 if (ip->ip_dst.s_addr == INADDR_ANY) 557 goto ours; 558 559#if defined(NFAITH) && 0 < NFAITH 560 /* 561 * FAITH(Firewall Aided Internet Translator) 562 */ 563 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 564 if (ip_keepfaith) { 565 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 566 goto ours; 567 } 568 m_freem(m); 569 return; 570 } 571#endif 572 /* 573 * Not for us; forward if possible and desirable. 574 */ 575 if (ipforwarding == 0) { 576 ipstat.ips_cantforward++; 577 m_freem(m); 578 } else 579 ip_forward(m, 0); 580#ifdef IPFIREWALL_FORWARD 581 ip_fw_fwd_addr = NULL; 582#endif 583 return; 584 585ours: 586 /* Count the packet in the ip address stats */ 587 if (ia != NULL) { 588 ia->ia_ifa.if_ipackets++; 589 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 590 } 591 592 /* 593 * If offset or IP_MF are set, must reassemble. 594 * Otherwise, nothing need be done. 595 * (We could look in the reassembly queue to see 596 * if the packet was previously fragmented, 597 * but it's not worth the time; just let them time out.) 598 */ 599 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 600 601 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 602 /* 603 * Look for queue of fragments 604 * of this datagram. 605 */ 606 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 607 if (ip->ip_id == fp->ipq_id && 608 ip->ip_src.s_addr == fp->ipq_src.s_addr && 609 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 610 ip->ip_p == fp->ipq_p) 611 goto found; 612 613 fp = 0; 614 615 /* check if there's a place for the new queue */ 616 if (nipq > maxnipq) { 617 /* 618 * drop something from the tail of the current queue 619 * before proceeding further 620 */ 621 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 622 for (i = 0; i < IPREASS_NHASH; i++) { 623 if (ipq[i].prev != &ipq[i]) { 624 ip_freef(ipq[i].prev); 625 break; 626 } 627 } 628 } else 629 ip_freef(ipq[sum].prev); 630 } 631found: 632 /* 633 * Adjust ip_len to not reflect header, 634 * convert offset of this to bytes. 635 */ 636 ip->ip_len -= hlen; 637 if (ip->ip_off & IP_MF) { 638 /* 639 * Make sure that fragments have a data length 640 * that's a non-zero multiple of 8 bytes. 641 */ 642 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 643 ipstat.ips_toosmall++; /* XXX */ 644 goto bad; 645 } 646 m->m_flags |= M_FRAG; 647 } 648 ip->ip_off <<= 3; 649 650 /* 651 * Attempt reassembly; if it succeeds, proceed. 652 */ 653 ipstat.ips_fragments++; 654 m->m_pkthdr.header = ip; 655#ifdef IPDIVERT 656 m = ip_reass(m, 657 fp, &ipq[sum], &divert_info, &divert_cookie); 658#else 659 m = ip_reass(m, fp, &ipq[sum]); 660#endif 661 if (m == 0) { 662#ifdef IPFIREWALL_FORWARD 663 ip_fw_fwd_addr = NULL; 664#endif 665 return; 666 } 667 ipstat.ips_reassembled++; 668 ip = mtod(m, struct ip *); 669 /* Get the header length of the reassembled packet */ 670 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 671#ifdef IPDIVERT 672 /* Restore original checksum before diverting packet */ 673 if (divert_info != 0) { 674 ip->ip_len += hlen; 675 HTONS(ip->ip_len); 676 HTONS(ip->ip_off); 677 ip->ip_sum = 0; 678 if (hlen == sizeof(struct ip)) 679 ip->ip_sum = in_cksum_hdr(ip); 680 else 681 ip->ip_sum = in_cksum(m, hlen); 682 NTOHS(ip->ip_off); 683 NTOHS(ip->ip_len); 684 ip->ip_len -= hlen; 685 } 686#endif 687 } else 688 ip->ip_len -= hlen; 689 690#ifdef IPDIVERT 691 /* 692 * Divert or tee packet to the divert protocol if required. 693 * 694 * If divert_info is zero then cookie should be too, so we shouldn't 695 * need to clear them here. Assume divert_packet() does so also. 696 */ 697 if (divert_info != 0) { 698 struct mbuf *clone = NULL; 699 700 /* Clone packet if we're doing a 'tee' */ 701 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 702 clone = m_dup(m, M_DONTWAIT); 703 704 /* Restore packet header fields to original values */ 705 ip->ip_len += hlen; 706 HTONS(ip->ip_len); 707 HTONS(ip->ip_off); 708 709 /* Deliver packet to divert input routine */ 710 ip_divert_cookie = divert_cookie; 711 divert_packet(m, 1, divert_info & 0xffff); 712 ipstat.ips_delivered++; 713 714 /* If 'tee', continue with original packet */ 715 if (clone == NULL) 716 return; 717 m = clone; 718 ip = mtod(m, struct ip *); 719 } 720#endif 721 722 /* 723 * Switch out to protocol's input routine. 724 */ 725 ipstat.ips_delivered++; 726 { 727 int off = hlen, nh = ip->ip_p; 728 729 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh); 730#ifdef IPFIREWALL_FORWARD 731 ip_fw_fwd_addr = NULL; /* tcp needed it */ 732#endif 733 return; 734 } 735bad: 736#ifdef IPFIREWALL_FORWARD 737 ip_fw_fwd_addr = NULL; 738#endif 739 m_freem(m); 740} 741 742/* 743 * IP software interrupt routine - to go away sometime soon 744 */ 745static void 746ipintr(void) 747{ 748 int s; 749 struct mbuf *m; 750 751 while(1) { 752 s = splimp(); 753 IF_DEQUEUE(&ipintrq, m); 754 splx(s); 755 if (m == 0) 756 return; 757 ip_input(m); 758 } 759} 760 761/* 762 * Take incoming datagram fragment and try to reassemble it into 763 * whole datagram. If a chain for reassembly of this datagram already 764 * exists, then it is given as fp; otherwise have to make a chain. 765 * 766 * When IPDIVERT enabled, keep additional state with each packet that 767 * tells us if we need to divert or tee the packet we're building. 768 */ 769 770static struct mbuf * 771#ifdef IPDIVERT 772ip_reass(m, fp, where, divinfo, divcookie) 773#else 774ip_reass(m, fp, where) 775#endif 776 register struct mbuf *m; 777 register struct ipq *fp; 778 struct ipq *where; 779#ifdef IPDIVERT 780 u_int32_t *divinfo; 781 u_int16_t *divcookie; 782#endif 783{ 784 struct ip *ip = mtod(m, struct ip *); 785 register struct mbuf *p, *q, *nq; 786 struct mbuf *t; 787 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 788 int i, next; 789 790 /* 791 * Presence of header sizes in mbufs 792 * would confuse code below. 793 */ 794 m->m_data += hlen; 795 m->m_len -= hlen; 796 797 /* 798 * If first fragment to arrive, create a reassembly queue. 799 */ 800 if (fp == 0) { 801 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 802 goto dropfrag; 803 fp = mtod(t, struct ipq *); 804 insque(fp, where); 805 nipq++; 806 fp->ipq_ttl = IPFRAGTTL; 807 fp->ipq_p = ip->ip_p; 808 fp->ipq_id = ip->ip_id; 809 fp->ipq_src = ip->ip_src; 810 fp->ipq_dst = ip->ip_dst; 811 fp->ipq_frags = m; 812 m->m_nextpkt = NULL; 813#ifdef IPDIVERT 814 fp->ipq_div_info = 0; 815 fp->ipq_div_cookie = 0; 816#endif 817 goto inserted; 818 } 819 820#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 821 822 /* 823 * Find a segment which begins after this one does. 824 */ 825 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 826 if (GETIP(q)->ip_off > ip->ip_off) 827 break; 828 829 /* 830 * If there is a preceding segment, it may provide some of 831 * our data already. If so, drop the data from the incoming 832 * segment. If it provides all of our data, drop us, otherwise 833 * stick new segment in the proper place. 834 * 835 * If some of the data is dropped from the the preceding 836 * segment, then it's checksum is invalidated. 837 */ 838 if (p) { 839 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 840 if (i > 0) { 841 if (i >= ip->ip_len) 842 goto dropfrag; 843 m_adj(m, i); 844 m->m_pkthdr.csum_flags = 0; 845 ip->ip_off += i; 846 ip->ip_len -= i; 847 } 848 m->m_nextpkt = p->m_nextpkt; 849 p->m_nextpkt = m; 850 } else { 851 m->m_nextpkt = fp->ipq_frags; 852 fp->ipq_frags = m; 853 } 854 855 /* 856 * While we overlap succeeding segments trim them or, 857 * if they are completely covered, dequeue them. 858 */ 859 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 860 q = nq) { 861 i = (ip->ip_off + ip->ip_len) - 862 GETIP(q)->ip_off; 863 if (i < GETIP(q)->ip_len) { 864 GETIP(q)->ip_len -= i; 865 GETIP(q)->ip_off += i; 866 m_adj(q, i); 867 q->m_pkthdr.csum_flags = 0; 868 break; 869 } 870 nq = q->m_nextpkt; 871 m->m_nextpkt = nq; 872 m_freem(q); 873 } 874 875inserted: 876 877#ifdef IPDIVERT 878 /* 879 * Transfer firewall instructions to the fragment structure. 880 * Any fragment diverting causes the whole packet to divert. 881 */ 882 fp->ipq_div_info = *divinfo; 883 fp->ipq_div_cookie = *divcookie; 884 *divinfo = 0; 885 *divcookie = 0; 886#endif 887 888 /* 889 * Check for complete reassembly. 890 */ 891 next = 0; 892 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 893 if (GETIP(q)->ip_off != next) 894 return (0); 895 next += GETIP(q)->ip_len; 896 } 897 /* Make sure the last packet didn't have the IP_MF flag */ 898 if (p->m_flags & M_FRAG) 899 return (0); 900 901 /* 902 * Reassembly is complete. Make sure the packet is a sane size. 903 */ 904 q = fp->ipq_frags; 905 ip = GETIP(q); 906 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 907 ipstat.ips_toolong++; 908 ip_freef(fp); 909 return (0); 910 } 911 912 /* 913 * Concatenate fragments. 914 */ 915 m = q; 916 t = m->m_next; 917 m->m_next = 0; 918 m_cat(m, t); 919 nq = q->m_nextpkt; 920 q->m_nextpkt = 0; 921 for (q = nq; q != NULL; q = nq) { 922 nq = q->m_nextpkt; 923 q->m_nextpkt = NULL; 924 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 925 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 926 m_cat(m, q); 927 } 928 929#ifdef IPDIVERT 930 /* 931 * Extract firewall instructions from the fragment structure. 932 */ 933 *divinfo = fp->ipq_div_info; 934 *divcookie = fp->ipq_div_cookie; 935#endif 936 937 /* 938 * Create header for new ip packet by 939 * modifying header of first packet; 940 * dequeue and discard fragment reassembly header. 941 * Make header visible. 942 */ 943 ip->ip_len = next; 944 ip->ip_src = fp->ipq_src; 945 ip->ip_dst = fp->ipq_dst; 946 remque(fp); 947 nipq--; 948 (void) m_free(dtom(fp)); 949 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 950 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 951 /* some debugging cruft by sklower, below, will go away soon */ 952 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 953 register int plen = 0; 954 for (t = m; t; t = t->m_next) 955 plen += t->m_len; 956 m->m_pkthdr.len = plen; 957 } 958 return (m); 959 960dropfrag: 961#ifdef IPDIVERT 962 *divinfo = 0; 963 *divcookie = 0; 964#endif 965 ipstat.ips_fragdropped++; 966 m_freem(m); 967 return (0); 968 969#undef GETIP 970} 971 972/* 973 * Free a fragment reassembly header and all 974 * associated datagrams. 975 */ 976static void 977ip_freef(fp) 978 struct ipq *fp; 979{ 980 register struct mbuf *q; 981 982 while (fp->ipq_frags) { 983 q = fp->ipq_frags; 984 fp->ipq_frags = q->m_nextpkt; 985 m_freem(q); 986 } 987 remque(fp); 988 (void) m_free(dtom(fp)); 989 nipq--; 990} 991 992/* 993 * IP timer processing; 994 * if a timer expires on a reassembly 995 * queue, discard it. 996 */ 997void 998ip_slowtimo() 999{ 1000 register struct ipq *fp; 1001 int s = splnet(); 1002 int i; 1003 1004 for (i = 0; i < IPREASS_NHASH; i++) { 1005 fp = ipq[i].next; 1006 if (fp == 0) 1007 continue; 1008 while (fp != &ipq[i]) { 1009 --fp->ipq_ttl; 1010 fp = fp->next; 1011 if (fp->prev->ipq_ttl == 0) { 1012 ipstat.ips_fragtimeout++; 1013 ip_freef(fp->prev); 1014 } 1015 } 1016 } 1017 ipflow_slowtimo(); 1018 splx(s); 1019} 1020 1021/* 1022 * Drain off all datagram fragments. 1023 */ 1024void 1025ip_drain() 1026{ 1027 int i; 1028 1029 for (i = 0; i < IPREASS_NHASH; i++) { 1030 while (ipq[i].next != &ipq[i]) { 1031 ipstat.ips_fragdropped++; 1032 ip_freef(ipq[i].next); 1033 } 1034 } 1035 in_rtqdrain(); 1036} 1037 1038/* 1039 * Do option processing on a datagram, 1040 * possibly discarding it if bad options are encountered, 1041 * or forwarding it if source-routed. 1042 * Returns 1 if packet has been forwarded/freed, 1043 * 0 if the packet should be processed further. 1044 */ 1045static int 1046ip_dooptions(m) 1047 struct mbuf *m; 1048{ 1049 register struct ip *ip = mtod(m, struct ip *); 1050 register u_char *cp; 1051 register struct ip_timestamp *ipt; 1052 register struct in_ifaddr *ia; 1053 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1054 struct in_addr *sin, dst; 1055 n_time ntime; 1056 1057 dst = ip->ip_dst; 1058 cp = (u_char *)(ip + 1); 1059 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1060 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1061 opt = cp[IPOPT_OPTVAL]; 1062 if (opt == IPOPT_EOL) 1063 break; 1064 if (opt == IPOPT_NOP) 1065 optlen = 1; 1066 else { 1067 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1068 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1069 goto bad; 1070 } 1071 optlen = cp[IPOPT_OLEN]; 1072 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1073 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1074 goto bad; 1075 } 1076 } 1077 switch (opt) { 1078 1079 default: 1080 break; 1081 1082 /* 1083 * Source routing with record. 1084 * Find interface with current destination address. 1085 * If none on this machine then drop if strictly routed, 1086 * or do nothing if loosely routed. 1087 * Record interface address and bring up next address 1088 * component. If strictly routed make sure next 1089 * address is on directly accessible net. 1090 */ 1091 case IPOPT_LSRR: 1092 case IPOPT_SSRR: 1093 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1094 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1095 goto bad; 1096 } 1097 ipaddr.sin_addr = ip->ip_dst; 1098 ia = (struct in_ifaddr *) 1099 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1100 if (ia == 0) { 1101 if (opt == IPOPT_SSRR) { 1102 type = ICMP_UNREACH; 1103 code = ICMP_UNREACH_SRCFAIL; 1104 goto bad; 1105 } 1106 if (!ip_dosourceroute) 1107 goto nosourcerouting; 1108 /* 1109 * Loose routing, and not at next destination 1110 * yet; nothing to do except forward. 1111 */ 1112 break; 1113 } 1114 off--; /* 0 origin */ 1115 if (off > optlen - (int)sizeof(struct in_addr)) { 1116 /* 1117 * End of source route. Should be for us. 1118 */ 1119 if (!ip_acceptsourceroute) 1120 goto nosourcerouting; 1121 save_rte(cp, ip->ip_src); 1122 break; 1123 } 1124 1125 if (!ip_dosourceroute) { 1126 if (ipforwarding) { 1127 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1128 /* 1129 * Acting as a router, so generate ICMP 1130 */ 1131nosourcerouting: 1132 strcpy(buf, inet_ntoa(ip->ip_dst)); 1133 log(LOG_WARNING, 1134 "attempted source route from %s to %s\n", 1135 inet_ntoa(ip->ip_src), buf); 1136 type = ICMP_UNREACH; 1137 code = ICMP_UNREACH_SRCFAIL; 1138 goto bad; 1139 } else { 1140 /* 1141 * Not acting as a router, so silently drop. 1142 */ 1143 ipstat.ips_cantforward++; 1144 m_freem(m); 1145 return (1); 1146 } 1147 } 1148 1149 /* 1150 * locate outgoing interface 1151 */ 1152 (void)memcpy(&ipaddr.sin_addr, cp + off, 1153 sizeof(ipaddr.sin_addr)); 1154 1155 if (opt == IPOPT_SSRR) { 1156#define INA struct in_ifaddr * 1157#define SA struct sockaddr * 1158 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1159 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1160 } else 1161 ia = ip_rtaddr(ipaddr.sin_addr); 1162 if (ia == 0) { 1163 type = ICMP_UNREACH; 1164 code = ICMP_UNREACH_SRCFAIL; 1165 goto bad; 1166 } 1167 ip->ip_dst = ipaddr.sin_addr; 1168 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1169 sizeof(struct in_addr)); 1170 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1171 /* 1172 * Let ip_intr's mcast routing check handle mcast pkts 1173 */ 1174 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1175 break; 1176 1177 case IPOPT_RR: 1178 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1179 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1180 goto bad; 1181 } 1182 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1183 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1184 goto bad; 1185 } 1186 /* 1187 * If no space remains, ignore. 1188 */ 1189 off--; /* 0 origin */ 1190 if (off > optlen - (int)sizeof(struct in_addr)) 1191 break; 1192 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1193 sizeof(ipaddr.sin_addr)); 1194 /* 1195 * locate outgoing interface; if we're the destination, 1196 * use the incoming interface (should be same). 1197 */ 1198 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1199 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1200 type = ICMP_UNREACH; 1201 code = ICMP_UNREACH_HOST; 1202 goto bad; 1203 } 1204 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1205 sizeof(struct in_addr)); 1206 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1207 break; 1208 1209 case IPOPT_TS: 1210 code = cp - (u_char *)ip; 1211 ipt = (struct ip_timestamp *)cp; 1212 if (ipt->ipt_len < 5) 1213 goto bad; 1214 if (ipt->ipt_ptr > 1215 ipt->ipt_len - (int)sizeof(int32_t)) { 1216 if (++ipt->ipt_oflw == 0) 1217 goto bad; 1218 break; 1219 } 1220 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1221 switch (ipt->ipt_flg) { 1222 1223 case IPOPT_TS_TSONLY: 1224 break; 1225 1226 case IPOPT_TS_TSANDADDR: 1227 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1228 sizeof(struct in_addr) > ipt->ipt_len) 1229 goto bad; 1230 ipaddr.sin_addr = dst; 1231 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1232 m->m_pkthdr.rcvif); 1233 if (ia == 0) 1234 continue; 1235 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1236 sizeof(struct in_addr)); 1237 ipt->ipt_ptr += sizeof(struct in_addr); 1238 break; 1239 1240 case IPOPT_TS_PRESPEC: 1241 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1242 sizeof(struct in_addr) > ipt->ipt_len) 1243 goto bad; 1244 (void)memcpy(&ipaddr.sin_addr, sin, 1245 sizeof(struct in_addr)); 1246 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1247 continue; 1248 ipt->ipt_ptr += sizeof(struct in_addr); 1249 break; 1250 1251 default: 1252 goto bad; 1253 } 1254 ntime = iptime(); 1255 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1256 sizeof(n_time)); 1257 ipt->ipt_ptr += sizeof(n_time); 1258 } 1259 } 1260 if (forward && ipforwarding) { 1261 ip_forward(m, 1); 1262 return (1); 1263 } 1264 return (0); 1265bad: 1266 icmp_error(m, type, code, 0, 0); 1267 ipstat.ips_badoptions++; 1268 return (1); 1269} 1270 1271/* 1272 * Given address of next destination (final or next hop), 1273 * return internet address info of interface to be used to get there. 1274 */ 1275static struct in_ifaddr * 1276ip_rtaddr(dst) 1277 struct in_addr dst; 1278{ 1279 register struct sockaddr_in *sin; 1280 1281 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1282 1283 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1284 if (ipforward_rt.ro_rt) { 1285 RTFREE(ipforward_rt.ro_rt); 1286 ipforward_rt.ro_rt = 0; 1287 } 1288 sin->sin_family = AF_INET; 1289 sin->sin_len = sizeof(*sin); 1290 sin->sin_addr = dst; 1291 1292 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1293 } 1294 if (ipforward_rt.ro_rt == 0) 1295 return ((struct in_ifaddr *)0); 1296 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1297} 1298 1299/* 1300 * Save incoming source route for use in replies, 1301 * to be picked up later by ip_srcroute if the receiver is interested. 1302 */ 1303void 1304save_rte(option, dst) 1305 u_char *option; 1306 struct in_addr dst; 1307{ 1308 unsigned olen; 1309 1310 olen = option[IPOPT_OLEN]; 1311#ifdef DIAGNOSTIC 1312 if (ipprintfs) 1313 printf("save_rte: olen %d\n", olen); 1314#endif 1315 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1316 return; 1317 bcopy(option, ip_srcrt.srcopt, olen); 1318 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1319 ip_srcrt.dst = dst; 1320} 1321 1322/* 1323 * Retrieve incoming source route for use in replies, 1324 * in the same form used by setsockopt. 1325 * The first hop is placed before the options, will be removed later. 1326 */ 1327struct mbuf * 1328ip_srcroute() 1329{ 1330 register struct in_addr *p, *q; 1331 register struct mbuf *m; 1332 1333 if (ip_nhops == 0) 1334 return ((struct mbuf *)0); 1335 m = m_get(M_DONTWAIT, MT_HEADER); 1336 if (m == 0) 1337 return ((struct mbuf *)0); 1338 1339#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1340 1341 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1342 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1343 OPTSIZ; 1344#ifdef DIAGNOSTIC 1345 if (ipprintfs) 1346 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1347#endif 1348 1349 /* 1350 * First save first hop for return route 1351 */ 1352 p = &ip_srcrt.route[ip_nhops - 1]; 1353 *(mtod(m, struct in_addr *)) = *p--; 1354#ifdef DIAGNOSTIC 1355 if (ipprintfs) 1356 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1357#endif 1358 1359 /* 1360 * Copy option fields and padding (nop) to mbuf. 1361 */ 1362 ip_srcrt.nop = IPOPT_NOP; 1363 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1364 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1365 &ip_srcrt.nop, OPTSIZ); 1366 q = (struct in_addr *)(mtod(m, caddr_t) + 1367 sizeof(struct in_addr) + OPTSIZ); 1368#undef OPTSIZ 1369 /* 1370 * Record return path as an IP source route, 1371 * reversing the path (pointers are now aligned). 1372 */ 1373 while (p >= ip_srcrt.route) { 1374#ifdef DIAGNOSTIC 1375 if (ipprintfs) 1376 printf(" %lx", (u_long)ntohl(q->s_addr)); 1377#endif 1378 *q++ = *p--; 1379 } 1380 /* 1381 * Last hop goes to final destination. 1382 */ 1383 *q = ip_srcrt.dst; 1384#ifdef DIAGNOSTIC 1385 if (ipprintfs) 1386 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1387#endif 1388 return (m); 1389} 1390 1391/* 1392 * Strip out IP options, at higher 1393 * level protocol in the kernel. 1394 * Second argument is buffer to which options 1395 * will be moved, and return value is their length. 1396 * XXX should be deleted; last arg currently ignored. 1397 */ 1398void 1399ip_stripoptions(m, mopt) 1400 register struct mbuf *m; 1401 struct mbuf *mopt; 1402{ 1403 register int i; 1404 struct ip *ip = mtod(m, struct ip *); 1405 register caddr_t opts; 1406 int olen; 1407 1408 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1409 opts = (caddr_t)(ip + 1); 1410 i = m->m_len - (sizeof (struct ip) + olen); 1411 bcopy(opts + olen, opts, (unsigned)i); 1412 m->m_len -= olen; 1413 if (m->m_flags & M_PKTHDR) 1414 m->m_pkthdr.len -= olen; 1415 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1416} 1417 1418u_char inetctlerrmap[PRC_NCMDS] = { 1419 0, 0, 0, 0, 1420 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1421 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1422 EMSGSIZE, EHOSTUNREACH, 0, 0, 1423 0, 0, 0, 0, 1424 ENOPROTOOPT 1425}; 1426 1427/* 1428 * Forward a packet. If some error occurs return the sender 1429 * an icmp packet. Note we can't always generate a meaningful 1430 * icmp message because icmp doesn't have a large enough repertoire 1431 * of codes and types. 1432 * 1433 * If not forwarding, just drop the packet. This could be confusing 1434 * if ipforwarding was zero but some routing protocol was advancing 1435 * us as a gateway to somewhere. However, we must let the routing 1436 * protocol deal with that. 1437 * 1438 * The srcrt parameter indicates whether the packet is being forwarded 1439 * via a source route. 1440 */ 1441static void 1442ip_forward(m, srcrt) 1443 struct mbuf *m; 1444 int srcrt; 1445{ 1446 register struct ip *ip = mtod(m, struct ip *); 1447 register struct sockaddr_in *sin; 1448 register struct rtentry *rt; 1449 int error, type = 0, code = 0; 1450 struct mbuf *mcopy; 1451 n_long dest; 1452 struct ifnet *destifp; 1453#ifdef IPSEC 1454 struct ifnet dummyifp; 1455#endif 1456 1457 dest = 0; 1458#ifdef DIAGNOSTIC 1459 if (ipprintfs) 1460 printf("forward: src %lx dst %lx ttl %x\n", 1461 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1462 ip->ip_ttl); 1463#endif 1464 1465 1466 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1467 ipstat.ips_cantforward++; 1468 m_freem(m); 1469 return; 1470 } 1471#ifdef IPSTEALTH 1472 if (!ipstealth) { 1473#endif 1474 if (ip->ip_ttl <= IPTTLDEC) { 1475 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1476 dest, 0); 1477 return; 1478 } 1479#ifdef IPSTEALTH 1480 } 1481#endif 1482 1483 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1484 if ((rt = ipforward_rt.ro_rt) == 0 || 1485 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1486 if (ipforward_rt.ro_rt) { 1487 RTFREE(ipforward_rt.ro_rt); 1488 ipforward_rt.ro_rt = 0; 1489 } 1490 sin->sin_family = AF_INET; 1491 sin->sin_len = sizeof(*sin); 1492 sin->sin_addr = ip->ip_dst; 1493 1494 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1495 if (ipforward_rt.ro_rt == 0) { 1496 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1497 return; 1498 } 1499 rt = ipforward_rt.ro_rt; 1500 } 1501 1502 /* 1503 * Save at most 64 bytes of the packet in case 1504 * we need to generate an ICMP message to the src. 1505 */ 1506 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1507 if (mcopy && (mcopy->m_flags & M_EXT)) 1508 m_copydata(mcopy, 0, sizeof(struct ip), mtod(mcopy, caddr_t)); 1509 1510#ifdef IPSTEALTH 1511 if (!ipstealth) { 1512#endif 1513 ip->ip_ttl -= IPTTLDEC; 1514#ifdef IPSTEALTH 1515 } 1516#endif 1517 1518 /* 1519 * If forwarding packet using same interface that it came in on, 1520 * perhaps should send a redirect to sender to shortcut a hop. 1521 * Only send redirect if source is sending directly to us, 1522 * and if packet was not source routed (or has any options). 1523 * Also, don't send redirect if forwarding using a default route 1524 * or a route modified by a redirect. 1525 */ 1526#define satosin(sa) ((struct sockaddr_in *)(sa)) 1527 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1528 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1529 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1530 ipsendredirects && !srcrt) { 1531#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1532 u_long src = ntohl(ip->ip_src.s_addr); 1533 1534 if (RTA(rt) && 1535 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1536 if (rt->rt_flags & RTF_GATEWAY) 1537 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1538 else 1539 dest = ip->ip_dst.s_addr; 1540 /* Router requirements says to only send host redirects */ 1541 type = ICMP_REDIRECT; 1542 code = ICMP_REDIRECT_HOST; 1543#ifdef DIAGNOSTIC 1544 if (ipprintfs) 1545 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1546#endif 1547 } 1548 } 1549 1550 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1551 IP_FORWARDING, 0); 1552 if (error) 1553 ipstat.ips_cantforward++; 1554 else { 1555 ipstat.ips_forward++; 1556 if (type) 1557 ipstat.ips_redirectsent++; 1558 else { 1559 if (mcopy) { 1560 ipflow_create(&ipforward_rt, mcopy); 1561 m_freem(mcopy); 1562 } 1563 return; 1564 } 1565 } 1566 if (mcopy == NULL) 1567 return; 1568 destifp = NULL; 1569 1570 switch (error) { 1571 1572 case 0: /* forwarded, but need redirect */ 1573 /* type, code set above */ 1574 break; 1575 1576 case ENETUNREACH: /* shouldn't happen, checked above */ 1577 case EHOSTUNREACH: 1578 case ENETDOWN: 1579 case EHOSTDOWN: 1580 default: 1581 type = ICMP_UNREACH; 1582 code = ICMP_UNREACH_HOST; 1583 break; 1584 1585 case EMSGSIZE: 1586 type = ICMP_UNREACH; 1587 code = ICMP_UNREACH_NEEDFRAG; 1588#ifndef IPSEC 1589 if (ipforward_rt.ro_rt) 1590 destifp = ipforward_rt.ro_rt->rt_ifp; 1591#else 1592 /* 1593 * If the packet is routed over IPsec tunnel, tell the 1594 * originator the tunnel MTU. 1595 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1596 * XXX quickhack!!! 1597 */ 1598 if (ipforward_rt.ro_rt) { 1599 struct secpolicy *sp = NULL; 1600 int ipsecerror; 1601 int ipsechdr; 1602 struct route *ro; 1603 1604 sp = ipsec4_getpolicybyaddr(mcopy, 1605 IPSEC_DIR_OUTBOUND, 1606 IP_FORWARDING, 1607 &ipsecerror); 1608 1609 if (sp == NULL) 1610 destifp = ipforward_rt.ro_rt->rt_ifp; 1611 else { 1612 /* count IPsec header size */ 1613 ipsechdr = ipsec4_hdrsiz(mcopy, 1614 IPSEC_DIR_OUTBOUND, 1615 NULL); 1616 1617 /* 1618 * find the correct route for outer IPv4 1619 * header, compute tunnel MTU. 1620 * 1621 * XXX BUG ALERT 1622 * The "dummyifp" code relies upon the fact 1623 * that icmp_error() touches only ifp->if_mtu. 1624 */ 1625 /*XXX*/ 1626 destifp = NULL; 1627 if (sp->req != NULL 1628 && sp->req->sav != NULL 1629 && sp->req->sav->sah != NULL) { 1630 ro = &sp->req->sav->sah->sa_route; 1631 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1632 dummyifp.if_mtu = 1633 ro->ro_rt->rt_ifp->if_mtu; 1634 dummyifp.if_mtu -= ipsechdr; 1635 destifp = &dummyifp; 1636 } 1637 } 1638 1639 key_freesp(sp); 1640 } 1641 } 1642#endif /*IPSEC*/ 1643 ipstat.ips_cantfrag++; 1644 break; 1645 1646 case ENOBUFS: 1647 type = ICMP_SOURCEQUENCH; 1648 code = 0; 1649 break; 1650 1651 case EACCES: /* ipfw denied packet */ 1652 m_freem(mcopy); 1653 return; 1654 } 1655 if (mcopy->m_flags & M_EXT) 1656 m_copyback(mcopy, 0, sizeof(struct ip), mtod(mcopy, caddr_t)); 1657 icmp_error(mcopy, type, code, dest, destifp); 1658} 1659 1660void 1661ip_savecontrol(inp, mp, ip, m) 1662 register struct inpcb *inp; 1663 register struct mbuf **mp; 1664 register struct ip *ip; 1665 register struct mbuf *m; 1666{ 1667 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1668 struct timeval tv; 1669 1670 microtime(&tv); 1671 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1672 SCM_TIMESTAMP, SOL_SOCKET); 1673 if (*mp) 1674 mp = &(*mp)->m_next; 1675 } 1676 if (inp->inp_flags & INP_RECVDSTADDR) { 1677 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1678 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1679 if (*mp) 1680 mp = &(*mp)->m_next; 1681 } 1682#ifdef notyet 1683 /* XXX 1684 * Moving these out of udp_input() made them even more broken 1685 * than they already were. 1686 */ 1687 /* options were tossed already */ 1688 if (inp->inp_flags & INP_RECVOPTS) { 1689 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1690 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1691 if (*mp) 1692 mp = &(*mp)->m_next; 1693 } 1694 /* ip_srcroute doesn't do what we want here, need to fix */ 1695 if (inp->inp_flags & INP_RECVRETOPTS) { 1696 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1697 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1698 if (*mp) 1699 mp = &(*mp)->m_next; 1700 } 1701#endif 1702 if (inp->inp_flags & INP_RECVIF) { 1703 struct ifnet *ifp; 1704 struct sdlbuf { 1705 struct sockaddr_dl sdl; 1706 u_char pad[32]; 1707 } sdlbuf; 1708 struct sockaddr_dl *sdp; 1709 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1710 1711 if (((ifp = m->m_pkthdr.rcvif)) 1712 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1713 sdp = (struct sockaddr_dl *)(ifnet_addrs 1714 [ifp->if_index - 1]->ifa_addr); 1715 /* 1716 * Change our mind and don't try copy. 1717 */ 1718 if ((sdp->sdl_family != AF_LINK) 1719 || (sdp->sdl_len > sizeof(sdlbuf))) { 1720 goto makedummy; 1721 } 1722 bcopy(sdp, sdl2, sdp->sdl_len); 1723 } else { 1724makedummy: 1725 sdl2->sdl_len 1726 = offsetof(struct sockaddr_dl, sdl_data[0]); 1727 sdl2->sdl_family = AF_LINK; 1728 sdl2->sdl_index = 0; 1729 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1730 } 1731 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1732 IP_RECVIF, IPPROTO_IP); 1733 if (*mp) 1734 mp = &(*mp)->m_next; 1735 } 1736} 1737 1738int 1739ip_rsvp_init(struct socket *so) 1740{ 1741 if (so->so_type != SOCK_RAW || 1742 so->so_proto->pr_protocol != IPPROTO_RSVP) 1743 return EOPNOTSUPP; 1744 1745 if (ip_rsvpd != NULL) 1746 return EADDRINUSE; 1747 1748 ip_rsvpd = so; 1749 /* 1750 * This may seem silly, but we need to be sure we don't over-increment 1751 * the RSVP counter, in case something slips up. 1752 */ 1753 if (!ip_rsvp_on) { 1754 ip_rsvp_on = 1; 1755 rsvp_on++; 1756 } 1757 1758 return 0; 1759} 1760 1761int 1762ip_rsvp_done(void) 1763{ 1764 ip_rsvpd = NULL; 1765 /* 1766 * This may seem silly, but we need to be sure we don't over-decrement 1767 * the RSVP counter, in case something slips up. 1768 */ 1769 if (ip_rsvp_on) { 1770 ip_rsvp_on = 0; 1771 rsvp_on--; 1772 } 1773 return 0; 1774} 1775