ip_input.c revision 65327
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 65327 2000-09-01 12:33:03Z 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 <stddef.h> 49 50#include <sys/param.h> 51#include <sys/systm.h> 52#include <sys/mbuf.h> 53#include <sys/malloc.h> 54#include <sys/domain.h> 55#include <sys/protosw.h> 56#include <sys/socket.h> 57#include <sys/time.h> 58#include <sys/kernel.h> 59#include <sys/syslog.h> 60#include <sys/sysctl.h> 61 62#include <net/pfil.h> 63#include <net/if.h> 64#include <net/if_var.h> 65#include <net/if_dl.h> 66#include <net/route.h> 67#include <net/netisr.h> 68#include <net/intrq.h> 69 70#include <netinet/in.h> 71#include <netinet/in_systm.h> 72#include <netinet/in_var.h> 73#include <netinet/ip.h> 74#include <netinet/in_pcb.h> 75#include <netinet/ip_var.h> 76#include <netinet/ip_icmp.h> 77#include <machine/in_cksum.h> 78 79#include <netinet/ipprotosw.h> 80 81#include <sys/socketvar.h> 82 83#include <netinet/ip_fw.h> 84 85#ifdef IPSEC 86#include <netinet6/ipsec.h> 87#include <netkey/key.h> 88#endif 89 90#include "faith.h" 91#if defined(NFAITH) && NFAITH > 0 92#include <net/if_types.h> 93#endif 94 95#ifdef DUMMYNET 96#include <netinet/ip_dummynet.h> 97#endif 98 99int rsvp_on = 0; 100static int ip_rsvp_on; 101struct socket *ip_rsvpd; 102 103int ipforwarding = 0; 104SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 105 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 106 107static int ipsendredirects = 1; /* XXX */ 108SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 109 &ipsendredirects, 0, "Enable sending IP redirects"); 110 111int ip_defttl = IPDEFTTL; 112SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 113 &ip_defttl, 0, "Maximum TTL on IP packets"); 114 115static int ip_dosourceroute = 0; 116SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 117 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 118 119static int ip_acceptsourceroute = 0; 120SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 121 CTLFLAG_RW, &ip_acceptsourceroute, 0, 122 "Enable accepting source routed IP packets"); 123 124static int ip_keepfaith = 0; 125SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 126 &ip_keepfaith, 0, 127 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 128 129#ifdef DIAGNOSTIC 130static int ipprintfs = 0; 131#endif 132 133extern struct domain inetdomain; 134extern struct ipprotosw inetsw[]; 135u_char ip_protox[IPPROTO_MAX]; 136static int ipqmaxlen = IFQ_MAXLEN; 137struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 138SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 139 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 140SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 141 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 142 143struct ipstat ipstat; 144SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD, 145 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 146 147/* Packet reassembly stuff */ 148#define IPREASS_NHASH_LOG2 6 149#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 150#define IPREASS_HMASK (IPREASS_NHASH - 1) 151#define IPREASS_HASH(x,y) \ 152 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 153 154static struct ipq ipq[IPREASS_NHASH]; 155static int nipq = 0; /* total # of reass queues */ 156static int maxnipq; 157const int ipintrq_present = 1; 158 159#ifdef IPCTL_DEFMTU 160SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 161 &ip_mtu, 0, "Default MTU"); 162#endif 163 164#ifdef IPSTEALTH 165static int ipstealth = 0; 166SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 167 &ipstealth, 0, ""); 168#endif 169 170 171/* Firewall hooks */ 172ip_fw_chk_t *ip_fw_chk_ptr; 173ip_fw_ctl_t *ip_fw_ctl_ptr; 174int fw_enable = 1 ; 175 176#ifdef DUMMYNET 177ip_dn_ctl_t *ip_dn_ctl_ptr; 178#endif 179 180 181/* 182 * We need to save the IP options in case a protocol wants to respond 183 * to an incoming packet over the same route if the packet got here 184 * using IP source routing. This allows connection establishment and 185 * maintenance when the remote end is on a network that is not known 186 * to us. 187 */ 188static int ip_nhops = 0; 189static struct ip_srcrt { 190 struct in_addr dst; /* final destination */ 191 char nop; /* one NOP to align */ 192 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 193 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 194} ip_srcrt; 195 196struct sockaddr_in *ip_fw_fwd_addr; 197 198static void save_rte __P((u_char *, struct in_addr)); 199static int ip_dooptions __P((struct mbuf *)); 200static void ip_forward __P((struct mbuf *, int)); 201static void ip_freef __P((struct ipq *)); 202#ifdef IPDIVERT 203static struct mbuf *ip_reass __P((struct mbuf *, 204 struct ipq *, struct ipq *, u_int32_t *, u_int16_t *)); 205#else 206static struct mbuf *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *)); 207#endif 208static struct in_ifaddr *ip_rtaddr __P((struct in_addr)); 209static void ipintr __P((void)); 210 211/* 212 * IP initialization: fill in IP protocol switch table. 213 * All protocols not implemented in kernel go to raw IP protocol handler. 214 */ 215void 216ip_init() 217{ 218 register struct ipprotosw *pr; 219 register int i; 220 221 TAILQ_INIT(&in_ifaddrhead); 222 pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 223 if (pr == 0) 224 panic("ip_init"); 225 for (i = 0; i < IPPROTO_MAX; i++) 226 ip_protox[i] = pr - inetsw; 227 for (pr = (struct ipprotosw *)inetdomain.dom_protosw; 228 pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++) 229 if (pr->pr_domain->dom_family == PF_INET && 230 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 231 ip_protox[pr->pr_protocol] = pr - inetsw; 232 233 for (i = 0; i < IPREASS_NHASH; i++) 234 ipq[i].next = ipq[i].prev = &ipq[i]; 235 236 maxnipq = nmbclusters/4; 237 238 ip_id = time_second & 0xffff; 239 ipintrq.ifq_maxlen = ipqmaxlen; 240 241 register_netisr(NETISR_IP, ipintr); 242} 243 244static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 245static struct route ipforward_rt; 246 247/* 248 * Ip input routine. Checksum and byte swap header. If fragmented 249 * try to reassemble. Process options. Pass to next level. 250 */ 251void 252ip_input(struct mbuf *m) 253{ 254 struct ip *ip; 255 struct ipq *fp; 256 struct in_ifaddr *ia; 257 int i, hlen, mff; 258 u_short sum; 259 u_int16_t divert_cookie; /* firewall cookie */ 260#ifdef IPDIVERT 261 u_int32_t divert_info = 0; /* packet divert/tee info */ 262#endif 263 struct ip_fw_chain *rule = NULL; 264#ifdef PFIL_HOOKS 265 struct packet_filter_hook *pfh; 266 struct mbuf *m0; 267 int rv; 268#endif /* PFIL_HOOKS */ 269 270#ifdef IPDIVERT 271 /* Get and reset firewall cookie */ 272 divert_cookie = ip_divert_cookie; 273 ip_divert_cookie = 0; 274#else 275 divert_cookie = 0; 276#endif 277 278#if defined(IPFIREWALL) && defined(DUMMYNET) 279 /* 280 * dummynet packet are prepended a vestigial mbuf with 281 * m_type = MT_DUMMYNET and m_data pointing to the matching 282 * rule. 283 */ 284 if (m->m_type == MT_DUMMYNET) { 285 rule = (struct ip_fw_chain *)(m->m_data) ; 286 m = m->m_next ; 287 ip = mtod(m, struct ip *); 288 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 289 goto iphack ; 290 } else 291 rule = NULL ; 292#endif 293 294#ifdef DIAGNOSTIC 295 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 296 panic("ip_input no HDR"); 297#endif 298 ipstat.ips_total++; 299 300 if (m->m_pkthdr.len < sizeof(struct ip)) 301 goto tooshort; 302 303 if (m->m_len < sizeof (struct ip) && 304 (m = m_pullup(m, sizeof (struct ip))) == 0) { 305 ipstat.ips_toosmall++; 306 return; 307 } 308 ip = mtod(m, struct ip *); 309 310 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 311 ipstat.ips_badvers++; 312 goto bad; 313 } 314 315 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 316 if (hlen < sizeof(struct ip)) { /* minimum header length */ 317 ipstat.ips_badhlen++; 318 goto bad; 319 } 320 if (hlen > m->m_len) { 321 if ((m = m_pullup(m, hlen)) == 0) { 322 ipstat.ips_badhlen++; 323 return; 324 } 325 ip = mtod(m, struct ip *); 326 } 327 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 328 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 329 } else { 330 if (hlen == sizeof(struct ip)) { 331 sum = in_cksum_hdr(ip); 332 } else { 333 sum = in_cksum(m, hlen); 334 } 335 } 336 if (sum) { 337 ipstat.ips_badsum++; 338 goto bad; 339 } 340 341 /* 342 * Convert fields to host representation. 343 */ 344 NTOHS(ip->ip_len); 345 if (ip->ip_len < hlen) { 346 ipstat.ips_badlen++; 347 goto bad; 348 } 349 NTOHS(ip->ip_id); 350 NTOHS(ip->ip_off); 351 352 /* 353 * Check that the amount of data in the buffers 354 * is as at least much as the IP header would have us expect. 355 * Trim mbufs if longer than we expect. 356 * Drop packet if shorter than we expect. 357 */ 358 if (m->m_pkthdr.len < ip->ip_len) { 359tooshort: 360 ipstat.ips_tooshort++; 361 goto bad; 362 } 363 if (m->m_pkthdr.len > ip->ip_len) { 364 if (m->m_len == m->m_pkthdr.len) { 365 m->m_len = ip->ip_len; 366 m->m_pkthdr.len = ip->ip_len; 367 } else 368 m_adj(m, ip->ip_len - m->m_pkthdr.len); 369 } 370 /* 371 * IpHack's section. 372 * Right now when no processing on packet has done 373 * and it is still fresh out of network we do our black 374 * deals with it. 375 * - Firewall: deny/allow/divert 376 * - Xlate: translate packet's addr/port (NAT). 377 * - Pipe: pass pkt through dummynet. 378 * - Wrap: fake packet's addr/port <unimpl.> 379 * - Encapsulate: put it in another IP and send out. <unimp.> 380 */ 381 382#if defined(IPFIREWALL) && defined(DUMMYNET) 383iphack: 384#endif 385 386#ifdef PFIL_HOOKS 387 /* 388 * Run through list of hooks for input packets. If there are any 389 * filters which require that additional packets in the flow are 390 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 391 * Note that filters must _never_ set this flag, as another filter 392 * in the list may have previously cleared it. 393 */ 394 m0 = m; 395 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 396 for (; pfh; pfh = pfh->pfil_link.tqe_next) 397 if (pfh->pfil_func) { 398 rv = pfh->pfil_func(ip, hlen, 399 m->m_pkthdr.rcvif, 0, &m0); 400 if (rv) 401 return; 402 m = m0; 403 if (m == NULL) 404 return; 405 ip = mtod(m, struct ip *); 406 } 407#endif /* PFIL_HOOKS */ 408 409 if (fw_enable && ip_fw_chk_ptr) { 410#ifdef IPFIREWALL_FORWARD 411 /* 412 * If we've been forwarded from the output side, then 413 * skip the firewall a second time 414 */ 415 if (ip_fw_fwd_addr) 416 goto ours; 417#endif /* IPFIREWALL_FORWARD */ 418 /* 419 * See the comment in ip_output for the return values 420 * produced by the firewall. 421 */ 422 i = (*ip_fw_chk_ptr)(&ip, 423 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 424 if (m == NULL) /* Packet discarded by firewall */ 425 return; 426 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 427 goto pass; 428#ifdef DUMMYNET 429 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) { 430 /* Send packet to the appropriate pipe */ 431 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 432 0); 433 return; 434 } 435#endif 436#ifdef IPDIVERT 437 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 438 /* Divert or tee packet */ 439 divert_info = i; 440 goto ours; 441 } 442#endif 443#ifdef IPFIREWALL_FORWARD 444 if (i == 0 && ip_fw_fwd_addr != NULL) 445 goto pass; 446#endif 447 /* 448 * if we get here, the packet must be dropped 449 */ 450 m_freem(m); 451 return; 452 } 453pass: 454 455 /* 456 * Process options and, if not destined for us, 457 * ship it on. ip_dooptions returns 1 when an 458 * error was detected (causing an icmp message 459 * to be sent and the original packet to be freed). 460 */ 461 ip_nhops = 0; /* for source routed packets */ 462 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 463#ifdef IPFIREWALL_FORWARD 464 ip_fw_fwd_addr = NULL; 465#endif 466 return; 467 } 468 469 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 470 * matter if it is destined to another node, or whether it is 471 * a multicast one, RSVP wants it! and prevents it from being forwarded 472 * anywhere else. Also checks if the rsvp daemon is running before 473 * grabbing the packet. 474 */ 475 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 476 goto ours; 477 478 /* 479 * Check our list of addresses, to see if the packet is for us. 480 * If we don't have any addresses, assume any unicast packet 481 * we receive might be for us (and let the upper layers deal 482 * with it). 483 */ 484 if (TAILQ_EMPTY(&in_ifaddrhead) && 485 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 486 goto ours; 487 488 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 489 ia = TAILQ_NEXT(ia, ia_link)) { 490#define satosin(sa) ((struct sockaddr_in *)(sa)) 491 492#ifdef BOOTP_COMPAT 493 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 494 goto ours; 495#endif 496#ifdef IPFIREWALL_FORWARD 497 /* 498 * If the addr to forward to is one of ours, we pretend to 499 * be the destination for this packet. 500 */ 501 if (ip_fw_fwd_addr == NULL) { 502 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 503 goto ours; 504 } else if (IA_SIN(ia)->sin_addr.s_addr == 505 ip_fw_fwd_addr->sin_addr.s_addr) 506 goto ours; 507#else 508 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 509 goto ours; 510#endif 511 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 512 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 513 ip->ip_dst.s_addr) 514 goto ours; 515 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 516 goto ours; 517 } 518 } 519 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 520 struct in_multi *inm; 521 if (ip_mrouter) { 522 /* 523 * If we are acting as a multicast router, all 524 * incoming multicast packets are passed to the 525 * kernel-level multicast forwarding function. 526 * The packet is returned (relatively) intact; if 527 * ip_mforward() returns a non-zero value, the packet 528 * must be discarded, else it may be accepted below. 529 */ 530 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 531 ipstat.ips_cantforward++; 532 m_freem(m); 533 return; 534 } 535 536 /* 537 * The process-level routing demon needs to receive 538 * all multicast IGMP packets, whether or not this 539 * host belongs to their destination groups. 540 */ 541 if (ip->ip_p == IPPROTO_IGMP) 542 goto ours; 543 ipstat.ips_forward++; 544 } 545 /* 546 * See if we belong to the destination multicast group on the 547 * arrival interface. 548 */ 549 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 550 if (inm == NULL) { 551 ipstat.ips_notmember++; 552 m_freem(m); 553 return; 554 } 555 goto ours; 556 } 557 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 558 goto ours; 559 if (ip->ip_dst.s_addr == INADDR_ANY) 560 goto ours; 561 562#if defined(NFAITH) && 0 < NFAITH 563 /* 564 * FAITH(Firewall Aided Internet Translator) 565 */ 566 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 567 if (ip_keepfaith) { 568 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 569 goto ours; 570 } 571 m_freem(m); 572 return; 573 } 574#endif 575 /* 576 * Not for us; forward if possible and desirable. 577 */ 578 if (ipforwarding == 0) { 579 ipstat.ips_cantforward++; 580 m_freem(m); 581 } else 582 ip_forward(m, 0); 583#ifdef IPFIREWALL_FORWARD 584 ip_fw_fwd_addr = NULL; 585#endif 586 return; 587 588ours: 589 590 /* 591 * If offset or IP_MF are set, must reassemble. 592 * Otherwise, nothing need be done. 593 * (We could look in the reassembly queue to see 594 * if the packet was previously fragmented, 595 * but it's not worth the time; just let them time out.) 596 */ 597 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) { 598 599#if 0 /* 600 * Reassembly should be able to treat a mbuf cluster, for later 601 * operation of contiguous protocol headers on the cluster. (KAME) 602 */ 603 if (m->m_flags & M_EXT) { /* XXX */ 604 if ((m = m_pullup(m, hlen)) == 0) { 605 ipstat.ips_toosmall++; 606#ifdef IPFIREWALL_FORWARD 607 ip_fw_fwd_addr = NULL; 608#endif 609 return; 610 } 611 ip = mtod(m, struct ip *); 612 } 613#endif 614 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 615 /* 616 * Look for queue of fragments 617 * of this datagram. 618 */ 619 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 620 if (ip->ip_id == fp->ipq_id && 621 ip->ip_src.s_addr == fp->ipq_src.s_addr && 622 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 623 ip->ip_p == fp->ipq_p) 624 goto found; 625 626 fp = 0; 627 628 /* check if there's a place for the new queue */ 629 if (nipq > maxnipq) { 630 /* 631 * drop something from the tail of the current queue 632 * before proceeding further 633 */ 634 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 635 for (i = 0; i < IPREASS_NHASH; i++) { 636 if (ipq[i].prev != &ipq[i]) { 637 ip_freef(ipq[i].prev); 638 break; 639 } 640 } 641 } else 642 ip_freef(ipq[sum].prev); 643 } 644found: 645 /* 646 * Adjust ip_len to not reflect header, 647 * set ip_mff if more fragments are expected, 648 * convert offset of this to bytes. 649 */ 650 ip->ip_len -= hlen; 651 mff = (ip->ip_off & IP_MF) != 0; 652 if (mff) { 653 /* 654 * Make sure that fragments have a data length 655 * that's a non-zero multiple of 8 bytes. 656 */ 657 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 658 ipstat.ips_toosmall++; /* XXX */ 659 goto bad; 660 } 661 m->m_flags |= M_FRAG; 662 } 663 ip->ip_off <<= 3; 664 665 /* 666 * If datagram marked as having more fragments 667 * or if this is not the first fragment, 668 * attempt reassembly; if it succeeds, proceed. 669 */ 670 if (mff || ip->ip_off) { 671 ipstat.ips_fragments++; 672 m->m_pkthdr.header = ip; 673#ifdef IPDIVERT 674 m = ip_reass(m, 675 fp, &ipq[sum], &divert_info, &divert_cookie); 676#else 677 m = ip_reass(m, fp, &ipq[sum]); 678#endif 679 if (m == 0) { 680#ifdef IPFIREWALL_FORWARD 681 ip_fw_fwd_addr = NULL; 682#endif 683 return; 684 } 685 /* Get the length of the reassembled packets header */ 686 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 687 ipstat.ips_reassembled++; 688 ip = mtod(m, struct ip *); 689#ifdef IPDIVERT 690 /* Restore original checksum before diverting packet */ 691 if (divert_info != 0) { 692 ip->ip_len += hlen; 693 HTONS(ip->ip_len); 694 HTONS(ip->ip_off); 695 HTONS(ip->ip_id); 696 ip->ip_sum = 0; 697 ip->ip_sum = in_cksum_hdr(ip); 698 NTOHS(ip->ip_id); 699 NTOHS(ip->ip_off); 700 NTOHS(ip->ip_len); 701 ip->ip_len -= hlen; 702 } 703#endif 704 } else 705 if (fp) 706 ip_freef(fp); 707 } else 708 ip->ip_len -= hlen; 709 710#ifdef IPDIVERT 711 /* 712 * Divert or tee packet to the divert protocol if required. 713 * 714 * If divert_info is zero then cookie should be too, so we shouldn't 715 * need to clear them here. Assume divert_packet() does so also. 716 */ 717 if (divert_info != 0) { 718 struct mbuf *clone = NULL; 719 720 /* Clone packet if we're doing a 'tee' */ 721 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 722 clone = m_dup(m, M_DONTWAIT); 723 724 /* Restore packet header fields to original values */ 725 ip->ip_len += hlen; 726 HTONS(ip->ip_len); 727 HTONS(ip->ip_off); 728 HTONS(ip->ip_id); 729 730 /* Deliver packet to divert input routine */ 731 ip_divert_cookie = divert_cookie; 732 divert_packet(m, 1, divert_info & 0xffff); 733 ipstat.ips_delivered++; 734 735 /* If 'tee', continue with original packet */ 736 if (clone == NULL) 737 return; 738 m = clone; 739 ip = mtod(m, struct ip *); 740 } 741#endif 742 743 /* 744 * Switch out to protocol's input routine. 745 */ 746 ipstat.ips_delivered++; 747 { 748 int off = hlen, nh = ip->ip_p; 749 750 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh); 751#ifdef IPFIREWALL_FORWARD 752 ip_fw_fwd_addr = NULL; /* tcp needed it */ 753#endif 754 return; 755 } 756bad: 757#ifdef IPFIREWALL_FORWARD 758 ip_fw_fwd_addr = NULL; 759#endif 760 m_freem(m); 761} 762 763/* 764 * IP software interrupt routine - to go away sometime soon 765 */ 766static void 767ipintr(void) 768{ 769 int s; 770 struct mbuf *m; 771 772 while(1) { 773 s = splimp(); 774 IF_DEQUEUE(&ipintrq, m); 775 splx(s); 776 if (m == 0) 777 return; 778 ip_input(m); 779 } 780} 781 782/* 783 * Take incoming datagram fragment and try to reassemble it into 784 * whole datagram. If a chain for reassembly of this datagram already 785 * exists, then it is given as fp; otherwise have to make a chain. 786 * 787 * When IPDIVERT enabled, keep additional state with each packet that 788 * tells us if we need to divert or tee the packet we're building. 789 */ 790 791static struct mbuf * 792#ifdef IPDIVERT 793ip_reass(m, fp, where, divinfo, divcookie) 794#else 795ip_reass(m, fp, where) 796#endif 797 register struct mbuf *m; 798 register struct ipq *fp; 799 struct ipq *where; 800#ifdef IPDIVERT 801 u_int32_t *divinfo; 802 u_int16_t *divcookie; 803#endif 804{ 805 struct ip *ip = mtod(m, struct ip *); 806 register struct mbuf *p = 0, *q, *nq; 807 struct mbuf *t; 808 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 809 int i, next; 810 811 /* 812 * Presence of header sizes in mbufs 813 * would confuse code below. 814 */ 815 m->m_data += hlen; 816 m->m_len -= hlen; 817 818 /* 819 * If first fragment to arrive, create a reassembly queue. 820 */ 821 if (fp == 0) { 822 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 823 goto dropfrag; 824 fp = mtod(t, struct ipq *); 825 insque(fp, where); 826 nipq++; 827 fp->ipq_ttl = IPFRAGTTL; 828 fp->ipq_p = ip->ip_p; 829 fp->ipq_id = ip->ip_id; 830 fp->ipq_src = ip->ip_src; 831 fp->ipq_dst = ip->ip_dst; 832 fp->ipq_frags = m; 833 m->m_nextpkt = NULL; 834#ifdef IPDIVERT 835 fp->ipq_div_info = 0; 836 fp->ipq_div_cookie = 0; 837#endif 838 goto inserted; 839 } 840 841#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 842 843 /* 844 * Find a segment which begins after this one does. 845 */ 846 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 847 if (GETIP(q)->ip_off > ip->ip_off) 848 break; 849 850 /* 851 * If there is a preceding segment, it may provide some of 852 * our data already. If so, drop the data from the incoming 853 * segment. If it provides all of our data, drop us, otherwise 854 * stick new segment in the proper place. 855 * 856 * If some of the data is dropped from the the preceding 857 * segment, then it's checksum is invalidated. 858 */ 859 if (p) { 860 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 861 if (i > 0) { 862 if (i >= ip->ip_len) 863 goto dropfrag; 864 m_adj(m, i); 865 m->m_pkthdr.csum_flags = 0; 866 ip->ip_off += i; 867 ip->ip_len -= i; 868 } 869 m->m_nextpkt = p->m_nextpkt; 870 p->m_nextpkt = m; 871 } else { 872 m->m_nextpkt = fp->ipq_frags; 873 fp->ipq_frags = m; 874 } 875 876 /* 877 * While we overlap succeeding segments trim them or, 878 * if they are completely covered, dequeue them. 879 */ 880 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 881 q = nq) { 882 i = (ip->ip_off + ip->ip_len) - 883 GETIP(q)->ip_off; 884 if (i < GETIP(q)->ip_len) { 885 GETIP(q)->ip_len -= i; 886 GETIP(q)->ip_off += i; 887 m_adj(q, i); 888 q->m_pkthdr.csum_flags = 0; 889 break; 890 } 891 nq = q->m_nextpkt; 892 m->m_nextpkt = nq; 893 m_freem(q); 894 } 895 896inserted: 897 898#ifdef IPDIVERT 899 /* 900 * Transfer firewall instructions to the fragment structure. 901 * Any fragment diverting causes the whole packet to divert. 902 */ 903 fp->ipq_div_info = *divinfo; 904 fp->ipq_div_cookie = *divcookie; 905 *divinfo = 0; 906 *divcookie = 0; 907#endif 908 909 /* 910 * Check for complete reassembly. 911 */ 912 next = 0; 913 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 914 if (GETIP(q)->ip_off != next) 915 return (0); 916 next += GETIP(q)->ip_len; 917 } 918 /* Make sure the last packet didn't have the IP_MF flag */ 919 if (p->m_flags & M_FRAG) 920 return (0); 921 922 /* 923 * Reassembly is complete. Make sure the packet is a sane size. 924 */ 925 q = fp->ipq_frags; 926 ip = GETIP(q); 927 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 928 ipstat.ips_toolong++; 929 ip_freef(fp); 930 return (0); 931 } 932 933 /* 934 * Concatenate fragments. 935 */ 936 m = q; 937 t = m->m_next; 938 m->m_next = 0; 939 m_cat(m, t); 940 nq = q->m_nextpkt; 941 q->m_nextpkt = 0; 942 for (q = nq; q != NULL; q = nq) { 943 nq = q->m_nextpkt; 944 q->m_nextpkt = NULL; 945 m_cat(m, q); 946 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 947 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 948 } 949 950#ifdef IPDIVERT 951 /* 952 * Extract firewall instructions from the fragment structure. 953 */ 954 *divinfo = fp->ipq_div_info; 955 *divcookie = fp->ipq_div_cookie; 956#endif 957 958 /* 959 * Create header for new ip packet by 960 * modifying header of first packet; 961 * dequeue and discard fragment reassembly header. 962 * Make header visible. 963 */ 964 ip->ip_len = next; 965 ip->ip_src = fp->ipq_src; 966 ip->ip_dst = fp->ipq_dst; 967 remque(fp); 968 nipq--; 969 (void) m_free(dtom(fp)); 970 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 971 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 972 /* some debugging cruft by sklower, below, will go away soon */ 973 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 974 register int plen = 0; 975 for (t = m; t; t = t->m_next) 976 plen += t->m_len; 977 m->m_pkthdr.len = plen; 978 } 979 return (m); 980 981dropfrag: 982#ifdef IPDIVERT 983 *divinfo = 0; 984 *divcookie = 0; 985#endif 986 ipstat.ips_fragdropped++; 987 m_freem(m); 988 return (0); 989 990#undef GETIP 991} 992 993/* 994 * Free a fragment reassembly header and all 995 * associated datagrams. 996 */ 997static void 998ip_freef(fp) 999 struct ipq *fp; 1000{ 1001 register struct mbuf *q; 1002 1003 while (fp->ipq_frags) { 1004 q = fp->ipq_frags; 1005 fp->ipq_frags = q->m_nextpkt; 1006 m_freem(q); 1007 } 1008 remque(fp); 1009 (void) m_free(dtom(fp)); 1010 nipq--; 1011} 1012 1013/* 1014 * IP timer processing; 1015 * if a timer expires on a reassembly 1016 * queue, discard it. 1017 */ 1018void 1019ip_slowtimo() 1020{ 1021 register struct ipq *fp; 1022 int s = splnet(); 1023 int i; 1024 1025 for (i = 0; i < IPREASS_NHASH; i++) { 1026 fp = ipq[i].next; 1027 if (fp == 0) 1028 continue; 1029 while (fp != &ipq[i]) { 1030 --fp->ipq_ttl; 1031 fp = fp->next; 1032 if (fp->prev->ipq_ttl == 0) { 1033 ipstat.ips_fragtimeout++; 1034 ip_freef(fp->prev); 1035 } 1036 } 1037 } 1038 ipflow_slowtimo(); 1039 splx(s); 1040} 1041 1042/* 1043 * Drain off all datagram fragments. 1044 */ 1045void 1046ip_drain() 1047{ 1048 int i; 1049 1050 for (i = 0; i < IPREASS_NHASH; i++) { 1051 while (ipq[i].next != &ipq[i]) { 1052 ipstat.ips_fragdropped++; 1053 ip_freef(ipq[i].next); 1054 } 1055 } 1056 in_rtqdrain(); 1057} 1058 1059/* 1060 * Do option processing on a datagram, 1061 * possibly discarding it if bad options are encountered, 1062 * or forwarding it if source-routed. 1063 * Returns 1 if packet has been forwarded/freed, 1064 * 0 if the packet should be processed further. 1065 */ 1066static int 1067ip_dooptions(m) 1068 struct mbuf *m; 1069{ 1070 register struct ip *ip = mtod(m, struct ip *); 1071 register u_char *cp; 1072 register struct ip_timestamp *ipt; 1073 register struct in_ifaddr *ia; 1074 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1075 struct in_addr *sin, dst; 1076 n_time ntime; 1077 1078 dst = ip->ip_dst; 1079 cp = (u_char *)(ip + 1); 1080 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1081 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1082 opt = cp[IPOPT_OPTVAL]; 1083 if (opt == IPOPT_EOL) 1084 break; 1085 if (opt == IPOPT_NOP) 1086 optlen = 1; 1087 else { 1088 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1089 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1090 goto bad; 1091 } 1092 optlen = cp[IPOPT_OLEN]; 1093 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1094 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1095 goto bad; 1096 } 1097 } 1098 switch (opt) { 1099 1100 default: 1101 break; 1102 1103 /* 1104 * Source routing with record. 1105 * Find interface with current destination address. 1106 * If none on this machine then drop if strictly routed, 1107 * or do nothing if loosely routed. 1108 * Record interface address and bring up next address 1109 * component. If strictly routed make sure next 1110 * address is on directly accessible net. 1111 */ 1112 case IPOPT_LSRR: 1113 case IPOPT_SSRR: 1114 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1115 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1116 goto bad; 1117 } 1118 ipaddr.sin_addr = ip->ip_dst; 1119 ia = (struct in_ifaddr *) 1120 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1121 if (ia == 0) { 1122 if (opt == IPOPT_SSRR) { 1123 type = ICMP_UNREACH; 1124 code = ICMP_UNREACH_SRCFAIL; 1125 goto bad; 1126 } 1127 if (!ip_dosourceroute) 1128 goto nosourcerouting; 1129 /* 1130 * Loose routing, and not at next destination 1131 * yet; nothing to do except forward. 1132 */ 1133 break; 1134 } 1135 off--; /* 0 origin */ 1136 if (off > optlen - (int)sizeof(struct in_addr)) { 1137 /* 1138 * End of source route. Should be for us. 1139 */ 1140 if (!ip_acceptsourceroute) 1141 goto nosourcerouting; 1142 save_rte(cp, ip->ip_src); 1143 break; 1144 } 1145 1146 if (!ip_dosourceroute) { 1147 if (ipforwarding) { 1148 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1149 /* 1150 * Acting as a router, so generate ICMP 1151 */ 1152nosourcerouting: 1153 strcpy(buf, inet_ntoa(ip->ip_dst)); 1154 log(LOG_WARNING, 1155 "attempted source route from %s to %s\n", 1156 inet_ntoa(ip->ip_src), buf); 1157 type = ICMP_UNREACH; 1158 code = ICMP_UNREACH_SRCFAIL; 1159 goto bad; 1160 } else { 1161 /* 1162 * Not acting as a router, so silently drop. 1163 */ 1164 ipstat.ips_cantforward++; 1165 m_freem(m); 1166 return (1); 1167 } 1168 } 1169 1170 /* 1171 * locate outgoing interface 1172 */ 1173 (void)memcpy(&ipaddr.sin_addr, cp + off, 1174 sizeof(ipaddr.sin_addr)); 1175 1176 if (opt == IPOPT_SSRR) { 1177#define INA struct in_ifaddr * 1178#define SA struct sockaddr * 1179 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1180 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1181 } else 1182 ia = ip_rtaddr(ipaddr.sin_addr); 1183 if (ia == 0) { 1184 type = ICMP_UNREACH; 1185 code = ICMP_UNREACH_SRCFAIL; 1186 goto bad; 1187 } 1188 ip->ip_dst = ipaddr.sin_addr; 1189 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1190 sizeof(struct in_addr)); 1191 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1192 /* 1193 * Let ip_intr's mcast routing check handle mcast pkts 1194 */ 1195 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1196 break; 1197 1198 case IPOPT_RR: 1199 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1200 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1201 goto bad; 1202 } 1203 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1204 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1205 goto bad; 1206 } 1207 /* 1208 * If no space remains, ignore. 1209 */ 1210 off--; /* 0 origin */ 1211 if (off > optlen - (int)sizeof(struct in_addr)) 1212 break; 1213 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1214 sizeof(ipaddr.sin_addr)); 1215 /* 1216 * locate outgoing interface; if we're the destination, 1217 * use the incoming interface (should be same). 1218 */ 1219 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1220 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1221 type = ICMP_UNREACH; 1222 code = ICMP_UNREACH_HOST; 1223 goto bad; 1224 } 1225 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1226 sizeof(struct in_addr)); 1227 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1228 break; 1229 1230 case IPOPT_TS: 1231 code = cp - (u_char *)ip; 1232 ipt = (struct ip_timestamp *)cp; 1233 if (ipt->ipt_len < 5) 1234 goto bad; 1235 if (ipt->ipt_ptr > 1236 ipt->ipt_len - (int)sizeof(int32_t)) { 1237 if (++ipt->ipt_oflw == 0) 1238 goto bad; 1239 break; 1240 } 1241 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1242 switch (ipt->ipt_flg) { 1243 1244 case IPOPT_TS_TSONLY: 1245 break; 1246 1247 case IPOPT_TS_TSANDADDR: 1248 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1249 sizeof(struct in_addr) > ipt->ipt_len) 1250 goto bad; 1251 ipaddr.sin_addr = dst; 1252 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1253 m->m_pkthdr.rcvif); 1254 if (ia == 0) 1255 continue; 1256 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1257 sizeof(struct in_addr)); 1258 ipt->ipt_ptr += sizeof(struct in_addr); 1259 break; 1260 1261 case IPOPT_TS_PRESPEC: 1262 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1263 sizeof(struct in_addr) > ipt->ipt_len) 1264 goto bad; 1265 (void)memcpy(&ipaddr.sin_addr, sin, 1266 sizeof(struct in_addr)); 1267 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1268 continue; 1269 ipt->ipt_ptr += sizeof(struct in_addr); 1270 break; 1271 1272 default: 1273 goto bad; 1274 } 1275 ntime = iptime(); 1276 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1277 sizeof(n_time)); 1278 ipt->ipt_ptr += sizeof(n_time); 1279 } 1280 } 1281 if (forward && ipforwarding) { 1282 ip_forward(m, 1); 1283 return (1); 1284 } 1285 return (0); 1286bad: 1287 icmp_error(m, type, code, 0, 0); 1288 ipstat.ips_badoptions++; 1289 return (1); 1290} 1291 1292/* 1293 * Given address of next destination (final or next hop), 1294 * return internet address info of interface to be used to get there. 1295 */ 1296static struct in_ifaddr * 1297ip_rtaddr(dst) 1298 struct in_addr dst; 1299{ 1300 register struct sockaddr_in *sin; 1301 1302 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1303 1304 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1305 if (ipforward_rt.ro_rt) { 1306 RTFREE(ipforward_rt.ro_rt); 1307 ipforward_rt.ro_rt = 0; 1308 } 1309 sin->sin_family = AF_INET; 1310 sin->sin_len = sizeof(*sin); 1311 sin->sin_addr = dst; 1312 1313 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1314 } 1315 if (ipforward_rt.ro_rt == 0) 1316 return ((struct in_ifaddr *)0); 1317 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1318} 1319 1320/* 1321 * Save incoming source route for use in replies, 1322 * to be picked up later by ip_srcroute if the receiver is interested. 1323 */ 1324void 1325save_rte(option, dst) 1326 u_char *option; 1327 struct in_addr dst; 1328{ 1329 unsigned olen; 1330 1331 olen = option[IPOPT_OLEN]; 1332#ifdef DIAGNOSTIC 1333 if (ipprintfs) 1334 printf("save_rte: olen %d\n", olen); 1335#endif 1336 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1337 return; 1338 bcopy(option, ip_srcrt.srcopt, olen); 1339 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1340 ip_srcrt.dst = dst; 1341} 1342 1343/* 1344 * Retrieve incoming source route for use in replies, 1345 * in the same form used by setsockopt. 1346 * The first hop is placed before the options, will be removed later. 1347 */ 1348struct mbuf * 1349ip_srcroute() 1350{ 1351 register struct in_addr *p, *q; 1352 register struct mbuf *m; 1353 1354 if (ip_nhops == 0) 1355 return ((struct mbuf *)0); 1356 m = m_get(M_DONTWAIT, MT_HEADER); 1357 if (m == 0) 1358 return ((struct mbuf *)0); 1359 1360#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1361 1362 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1363 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1364 OPTSIZ; 1365#ifdef DIAGNOSTIC 1366 if (ipprintfs) 1367 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1368#endif 1369 1370 /* 1371 * First save first hop for return route 1372 */ 1373 p = &ip_srcrt.route[ip_nhops - 1]; 1374 *(mtod(m, struct in_addr *)) = *p--; 1375#ifdef DIAGNOSTIC 1376 if (ipprintfs) 1377 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1378#endif 1379 1380 /* 1381 * Copy option fields and padding (nop) to mbuf. 1382 */ 1383 ip_srcrt.nop = IPOPT_NOP; 1384 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1385 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1386 &ip_srcrt.nop, OPTSIZ); 1387 q = (struct in_addr *)(mtod(m, caddr_t) + 1388 sizeof(struct in_addr) + OPTSIZ); 1389#undef OPTSIZ 1390 /* 1391 * Record return path as an IP source route, 1392 * reversing the path (pointers are now aligned). 1393 */ 1394 while (p >= ip_srcrt.route) { 1395#ifdef DIAGNOSTIC 1396 if (ipprintfs) 1397 printf(" %lx", (u_long)ntohl(q->s_addr)); 1398#endif 1399 *q++ = *p--; 1400 } 1401 /* 1402 * Last hop goes to final destination. 1403 */ 1404 *q = ip_srcrt.dst; 1405#ifdef DIAGNOSTIC 1406 if (ipprintfs) 1407 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1408#endif 1409 return (m); 1410} 1411 1412/* 1413 * Strip out IP options, at higher 1414 * level protocol in the kernel. 1415 * Second argument is buffer to which options 1416 * will be moved, and return value is their length. 1417 * XXX should be deleted; last arg currently ignored. 1418 */ 1419void 1420ip_stripoptions(m, mopt) 1421 register struct mbuf *m; 1422 struct mbuf *mopt; 1423{ 1424 register int i; 1425 struct ip *ip = mtod(m, struct ip *); 1426 register caddr_t opts; 1427 int olen; 1428 1429 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1430 opts = (caddr_t)(ip + 1); 1431 i = m->m_len - (sizeof (struct ip) + olen); 1432 bcopy(opts + olen, opts, (unsigned)i); 1433 m->m_len -= olen; 1434 if (m->m_flags & M_PKTHDR) 1435 m->m_pkthdr.len -= olen; 1436 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1437} 1438 1439u_char inetctlerrmap[PRC_NCMDS] = { 1440 0, 0, 0, 0, 1441 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1442 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1443 EMSGSIZE, EHOSTUNREACH, 0, 0, 1444 0, 0, 0, 0, 1445 ENOPROTOOPT 1446}; 1447 1448/* 1449 * Forward a packet. If some error occurs return the sender 1450 * an icmp packet. Note we can't always generate a meaningful 1451 * icmp message because icmp doesn't have a large enough repertoire 1452 * of codes and types. 1453 * 1454 * If not forwarding, just drop the packet. This could be confusing 1455 * if ipforwarding was zero but some routing protocol was advancing 1456 * us as a gateway to somewhere. However, we must let the routing 1457 * protocol deal with that. 1458 * 1459 * The srcrt parameter indicates whether the packet is being forwarded 1460 * via a source route. 1461 */ 1462static void 1463ip_forward(m, srcrt) 1464 struct mbuf *m; 1465 int srcrt; 1466{ 1467 register struct ip *ip = mtod(m, struct ip *); 1468 register struct sockaddr_in *sin; 1469 register struct rtentry *rt; 1470 int error, type = 0, code = 0; 1471 struct mbuf *mcopy; 1472 n_long dest; 1473 struct ifnet *destifp; 1474#ifdef IPSEC 1475 struct ifnet dummyifp; 1476#endif 1477 1478 dest = 0; 1479#ifdef DIAGNOSTIC 1480 if (ipprintfs) 1481 printf("forward: src %lx dst %lx ttl %x\n", 1482 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1483 ip->ip_ttl); 1484#endif 1485 1486 1487 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1488 ipstat.ips_cantforward++; 1489 m_freem(m); 1490 return; 1491 } 1492#ifdef IPSTEALTH 1493 if (!ipstealth) { 1494#endif 1495 if (ip->ip_ttl <= IPTTLDEC) { 1496 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1497 dest, 0); 1498 return; 1499 } 1500#ifdef IPSTEALTH 1501 } 1502#endif 1503 1504 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1505 if ((rt = ipforward_rt.ro_rt) == 0 || 1506 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1507 if (ipforward_rt.ro_rt) { 1508 RTFREE(ipforward_rt.ro_rt); 1509 ipforward_rt.ro_rt = 0; 1510 } 1511 sin->sin_family = AF_INET; 1512 sin->sin_len = sizeof(*sin); 1513 sin->sin_addr = ip->ip_dst; 1514 1515 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1516 if (ipforward_rt.ro_rt == 0) { 1517 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1518 return; 1519 } 1520 rt = ipforward_rt.ro_rt; 1521 } 1522 1523 /* 1524 * Save at most 64 bytes of the packet in case 1525 * we need to generate an ICMP message to the src. 1526 */ 1527 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1528 if (mcopy && (mcopy->m_flags & M_EXT)) 1529 m_copydata(mcopy, 0, sizeof(struct ip), mtod(mcopy, caddr_t)); 1530 1531#ifdef IPSTEALTH 1532 if (!ipstealth) { 1533#endif 1534 ip->ip_ttl -= IPTTLDEC; 1535#ifdef IPSTEALTH 1536 } 1537#endif 1538 1539 /* 1540 * If forwarding packet using same interface that it came in on, 1541 * perhaps should send a redirect to sender to shortcut a hop. 1542 * Only send redirect if source is sending directly to us, 1543 * and if packet was not source routed (or has any options). 1544 * Also, don't send redirect if forwarding using a default route 1545 * or a route modified by a redirect. 1546 */ 1547#define satosin(sa) ((struct sockaddr_in *)(sa)) 1548 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1549 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1550 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1551 ipsendredirects && !srcrt) { 1552#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1553 u_long src = ntohl(ip->ip_src.s_addr); 1554 1555 if (RTA(rt) && 1556 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1557 if (rt->rt_flags & RTF_GATEWAY) 1558 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1559 else 1560 dest = ip->ip_dst.s_addr; 1561 /* Router requirements says to only send host redirects */ 1562 type = ICMP_REDIRECT; 1563 code = ICMP_REDIRECT_HOST; 1564#ifdef DIAGNOSTIC 1565 if (ipprintfs) 1566 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1567#endif 1568 } 1569 } 1570 1571 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1572 IP_FORWARDING, 0); 1573 if (error) 1574 ipstat.ips_cantforward++; 1575 else { 1576 ipstat.ips_forward++; 1577 if (type) 1578 ipstat.ips_redirectsent++; 1579 else { 1580 if (mcopy) { 1581 ipflow_create(&ipforward_rt, mcopy); 1582 m_freem(mcopy); 1583 } 1584 return; 1585 } 1586 } 1587 if (mcopy == NULL) 1588 return; 1589 destifp = NULL; 1590 1591 switch (error) { 1592 1593 case 0: /* forwarded, but need redirect */ 1594 /* type, code set above */ 1595 break; 1596 1597 case ENETUNREACH: /* shouldn't happen, checked above */ 1598 case EHOSTUNREACH: 1599 case ENETDOWN: 1600 case EHOSTDOWN: 1601 default: 1602 type = ICMP_UNREACH; 1603 code = ICMP_UNREACH_HOST; 1604 break; 1605 1606 case EMSGSIZE: 1607 type = ICMP_UNREACH; 1608 code = ICMP_UNREACH_NEEDFRAG; 1609#ifndef IPSEC 1610 if (ipforward_rt.ro_rt) 1611 destifp = ipforward_rt.ro_rt->rt_ifp; 1612#else 1613 /* 1614 * If the packet is routed over IPsec tunnel, tell the 1615 * originator the tunnel MTU. 1616 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1617 * XXX quickhack!!! 1618 */ 1619 if (ipforward_rt.ro_rt) { 1620 struct secpolicy *sp = NULL; 1621 int ipsecerror; 1622 int ipsechdr; 1623 struct route *ro; 1624 1625 sp = ipsec4_getpolicybyaddr(mcopy, 1626 IPSEC_DIR_OUTBOUND, 1627 IP_FORWARDING, 1628 &ipsecerror); 1629 1630 if (sp == NULL) 1631 destifp = ipforward_rt.ro_rt->rt_ifp; 1632 else { 1633 /* count IPsec header size */ 1634 ipsechdr = ipsec4_hdrsiz(mcopy, 1635 IPSEC_DIR_OUTBOUND, 1636 NULL); 1637 1638 /* 1639 * find the correct route for outer IPv4 1640 * header, compute tunnel MTU. 1641 * 1642 * XXX BUG ALERT 1643 * The "dummyifp" code relies upon the fact 1644 * that icmp_error() touches only ifp->if_mtu. 1645 */ 1646 /*XXX*/ 1647 destifp = NULL; 1648 if (sp->req != NULL 1649 && sp->req->sav != NULL 1650 && sp->req->sav->sah != NULL) { 1651 ro = &sp->req->sav->sah->sa_route; 1652 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1653 dummyifp.if_mtu = 1654 ro->ro_rt->rt_ifp->if_mtu; 1655 dummyifp.if_mtu -= ipsechdr; 1656 destifp = &dummyifp; 1657 } 1658 } 1659 1660 key_freesp(sp); 1661 } 1662 } 1663#endif /*IPSEC*/ 1664 ipstat.ips_cantfrag++; 1665 break; 1666 1667 case ENOBUFS: 1668 type = ICMP_SOURCEQUENCH; 1669 code = 0; 1670 break; 1671 1672 case EACCES: /* ipfw denied packet */ 1673 m_freem(mcopy); 1674 return; 1675 } 1676 if (mcopy->m_flags & M_EXT) 1677 m_copyback(mcopy, 0, sizeof(struct ip), mtod(mcopy, caddr_t)); 1678 icmp_error(mcopy, type, code, dest, destifp); 1679} 1680 1681void 1682ip_savecontrol(inp, mp, ip, m) 1683 register struct inpcb *inp; 1684 register struct mbuf **mp; 1685 register struct ip *ip; 1686 register struct mbuf *m; 1687{ 1688 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1689 struct timeval tv; 1690 1691 microtime(&tv); 1692 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1693 SCM_TIMESTAMP, SOL_SOCKET); 1694 if (*mp) 1695 mp = &(*mp)->m_next; 1696 } 1697 if (inp->inp_flags & INP_RECVDSTADDR) { 1698 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1699 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1700 if (*mp) 1701 mp = &(*mp)->m_next; 1702 } 1703#ifdef notyet 1704 /* XXX 1705 * Moving these out of udp_input() made them even more broken 1706 * than they already were. 1707 */ 1708 /* options were tossed already */ 1709 if (inp->inp_flags & INP_RECVOPTS) { 1710 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1711 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1712 if (*mp) 1713 mp = &(*mp)->m_next; 1714 } 1715 /* ip_srcroute doesn't do what we want here, need to fix */ 1716 if (inp->inp_flags & INP_RECVRETOPTS) { 1717 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1718 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1719 if (*mp) 1720 mp = &(*mp)->m_next; 1721 } 1722#endif 1723 if (inp->inp_flags & INP_RECVIF) { 1724 struct ifnet *ifp; 1725 struct sdlbuf { 1726 struct sockaddr_dl sdl; 1727 u_char pad[32]; 1728 } sdlbuf; 1729 struct sockaddr_dl *sdp; 1730 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1731 1732 if (((ifp = m->m_pkthdr.rcvif)) 1733 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1734 sdp = (struct sockaddr_dl *)(ifnet_addrs 1735 [ifp->if_index - 1]->ifa_addr); 1736 /* 1737 * Change our mind and don't try copy. 1738 */ 1739 if ((sdp->sdl_family != AF_LINK) 1740 || (sdp->sdl_len > sizeof(sdlbuf))) { 1741 goto makedummy; 1742 } 1743 bcopy(sdp, sdl2, sdp->sdl_len); 1744 } else { 1745makedummy: 1746 sdl2->sdl_len 1747 = offsetof(struct sockaddr_dl, sdl_data[0]); 1748 sdl2->sdl_family = AF_LINK; 1749 sdl2->sdl_index = 0; 1750 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1751 } 1752 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1753 IP_RECVIF, IPPROTO_IP); 1754 if (*mp) 1755 mp = &(*mp)->m_next; 1756 } 1757} 1758 1759int 1760ip_rsvp_init(struct socket *so) 1761{ 1762 if (so->so_type != SOCK_RAW || 1763 so->so_proto->pr_protocol != IPPROTO_RSVP) 1764 return EOPNOTSUPP; 1765 1766 if (ip_rsvpd != NULL) 1767 return EADDRINUSE; 1768 1769 ip_rsvpd = so; 1770 /* 1771 * This may seem silly, but we need to be sure we don't over-increment 1772 * the RSVP counter, in case something slips up. 1773 */ 1774 if (!ip_rsvp_on) { 1775 ip_rsvp_on = 1; 1776 rsvp_on++; 1777 } 1778 1779 return 0; 1780} 1781 1782int 1783ip_rsvp_done(void) 1784{ 1785 ip_rsvpd = NULL; 1786 /* 1787 * This may seem silly, but we need to be sure we don't over-decrement 1788 * the RSVP counter, in case something slips up. 1789 */ 1790 if (ip_rsvp_on) { 1791 ip_rsvp_on = 0; 1792 rsvp_on--; 1793 } 1794 return 0; 1795} 1796