ip_input.c revision 67334
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 67334 2000-10-19 23:15:54Z joe $ 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 = NULL; 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_off); 350 351 /* 352 * Check that the amount of data in the buffers 353 * is as at least much as the IP header would have us expect. 354 * Trim mbufs if longer than we expect. 355 * Drop packet if shorter than we expect. 356 */ 357 if (m->m_pkthdr.len < ip->ip_len) { 358tooshort: 359 ipstat.ips_tooshort++; 360 goto bad; 361 } 362 if (m->m_pkthdr.len > ip->ip_len) { 363 if (m->m_len == m->m_pkthdr.len) { 364 m->m_len = ip->ip_len; 365 m->m_pkthdr.len = ip->ip_len; 366 } else 367 m_adj(m, ip->ip_len - m->m_pkthdr.len); 368 } 369 /* 370 * IpHack's section. 371 * Right now when no processing on packet has done 372 * and it is still fresh out of network we do our black 373 * deals with it. 374 * - Firewall: deny/allow/divert 375 * - Xlate: translate packet's addr/port (NAT). 376 * - Pipe: pass pkt through dummynet. 377 * - Wrap: fake packet's addr/port <unimpl.> 378 * - Encapsulate: put it in another IP and send out. <unimp.> 379 */ 380 381#if defined(IPFIREWALL) && defined(DUMMYNET) 382iphack: 383#endif 384 385#ifdef PFIL_HOOKS 386 /* 387 * Run through list of hooks for input packets. If there are any 388 * filters which require that additional packets in the flow are 389 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 390 * Note that filters must _never_ set this flag, as another filter 391 * in the list may have previously cleared it. 392 */ 393 m0 = m; 394 pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 395 for (; pfh; pfh = pfh->pfil_link.tqe_next) 396 if (pfh->pfil_func) { 397 rv = pfh->pfil_func(ip, hlen, 398 m->m_pkthdr.rcvif, 0, &m0); 399 if (rv) 400 return; 401 m = m0; 402 if (m == NULL) 403 return; 404 ip = mtod(m, struct ip *); 405 } 406#endif /* PFIL_HOOKS */ 407 408 if (fw_enable && ip_fw_chk_ptr) { 409#ifdef IPFIREWALL_FORWARD 410 /* 411 * If we've been forwarded from the output side, then 412 * skip the firewall a second time 413 */ 414 if (ip_fw_fwd_addr) 415 goto ours; 416#endif /* IPFIREWALL_FORWARD */ 417 /* 418 * See the comment in ip_output for the return values 419 * produced by the firewall. 420 */ 421 i = (*ip_fw_chk_ptr)(&ip, 422 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 423 if (m == NULL) /* Packet discarded by firewall */ 424 return; 425 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 426 goto pass; 427#ifdef DUMMYNET 428 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) { 429 /* Send packet to the appropriate pipe */ 430 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 431 0); 432 return; 433 } 434#endif 435#ifdef IPDIVERT 436 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 437 /* Divert or tee packet */ 438 divert_info = i; 439 goto ours; 440 } 441#endif 442#ifdef IPFIREWALL_FORWARD 443 if (i == 0 && ip_fw_fwd_addr != NULL) 444 goto pass; 445#endif 446 /* 447 * if we get here, the packet must be dropped 448 */ 449 m_freem(m); 450 return; 451 } 452pass: 453 454 /* 455 * Process options and, if not destined for us, 456 * ship it on. ip_dooptions returns 1 when an 457 * error was detected (causing an icmp message 458 * to be sent and the original packet to be freed). 459 */ 460 ip_nhops = 0; /* for source routed packets */ 461 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 462#ifdef IPFIREWALL_FORWARD 463 ip_fw_fwd_addr = NULL; 464#endif 465 return; 466 } 467 468 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 469 * matter if it is destined to another node, or whether it is 470 * a multicast one, RSVP wants it! and prevents it from being forwarded 471 * anywhere else. Also checks if the rsvp daemon is running before 472 * grabbing the packet. 473 */ 474 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 475 goto ours; 476 477 /* 478 * Check our list of addresses, to see if the packet is for us. 479 * If we don't have any addresses, assume any unicast packet 480 * we receive might be for us (and let the upper layers deal 481 * with it). 482 */ 483 if (TAILQ_EMPTY(&in_ifaddrhead) && 484 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 485 goto ours; 486 487 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 488 ia = TAILQ_NEXT(ia, ia_link)) { 489#define satosin(sa) ((struct sockaddr_in *)(sa)) 490 491#ifdef BOOTP_COMPAT 492 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 493 goto ours; 494#endif 495#ifdef IPFIREWALL_FORWARD 496 /* 497 * If the addr to forward to is one of ours, we pretend to 498 * be the destination for this packet. 499 */ 500 if (ip_fw_fwd_addr == NULL) { 501 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 502 goto ours; 503 } else if (IA_SIN(ia)->sin_addr.s_addr == 504 ip_fw_fwd_addr->sin_addr.s_addr) 505 goto ours; 506#else 507 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 508 goto ours; 509#endif 510 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 511 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 512 ip->ip_dst.s_addr) 513 goto ours; 514 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 515 goto ours; 516 } 517 } 518 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 519 struct in_multi *inm; 520 if (ip_mrouter) { 521 /* 522 * If we are acting as a multicast router, all 523 * incoming multicast packets are passed to the 524 * kernel-level multicast forwarding function. 525 * The packet is returned (relatively) intact; if 526 * ip_mforward() returns a non-zero value, the packet 527 * must be discarded, else it may be accepted below. 528 */ 529 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 530 ipstat.ips_cantforward++; 531 m_freem(m); 532 return; 533 } 534 535 /* 536 * The process-level routing demon needs to receive 537 * all multicast IGMP packets, whether or not this 538 * host belongs to their destination groups. 539 */ 540 if (ip->ip_p == IPPROTO_IGMP) 541 goto ours; 542 ipstat.ips_forward++; 543 } 544 /* 545 * See if we belong to the destination multicast group on the 546 * arrival interface. 547 */ 548 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 549 if (inm == NULL) { 550 ipstat.ips_notmember++; 551 m_freem(m); 552 return; 553 } 554 goto ours; 555 } 556 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 557 goto ours; 558 if (ip->ip_dst.s_addr == INADDR_ANY) 559 goto ours; 560 561#if defined(NFAITH) && 0 < NFAITH 562 /* 563 * FAITH(Firewall Aided Internet Translator) 564 */ 565 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 566 if (ip_keepfaith) { 567 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 568 goto ours; 569 } 570 m_freem(m); 571 return; 572 } 573#endif 574 /* 575 * Not for us; forward if possible and desirable. 576 */ 577 if (ipforwarding == 0) { 578 ipstat.ips_cantforward++; 579 m_freem(m); 580 } else 581 ip_forward(m, 0); 582#ifdef IPFIREWALL_FORWARD 583 ip_fw_fwd_addr = NULL; 584#endif 585 return; 586 587ours: 588 /* Count the packet in the ip address stats */ 589 if (ia != NULL) { 590 ia->ia_ifa.if_ipackets++; 591 ia->ia_ifa.if_ibytes += m->m_pkthdr.len; 592 } 593 594 /* 595 * If offset or IP_MF are set, must reassemble. 596 * Otherwise, nothing need be done. 597 * (We could look in the reassembly queue to see 598 * if the packet was previously fragmented, 599 * but it's not worth the time; just let them time out.) 600 */ 601 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) { 602 603#if 0 /* 604 * Reassembly should be able to treat a mbuf cluster, for later 605 * operation of contiguous protocol headers on the cluster. (KAME) 606 */ 607 if (m->m_flags & M_EXT) { /* XXX */ 608 if ((m = m_pullup(m, hlen)) == 0) { 609 ipstat.ips_toosmall++; 610#ifdef IPFIREWALL_FORWARD 611 ip_fw_fwd_addr = NULL; 612#endif 613 return; 614 } 615 ip = mtod(m, struct ip *); 616 } 617#endif 618 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 619 /* 620 * Look for queue of fragments 621 * of this datagram. 622 */ 623 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 624 if (ip->ip_id == fp->ipq_id && 625 ip->ip_src.s_addr == fp->ipq_src.s_addr && 626 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 627 ip->ip_p == fp->ipq_p) 628 goto found; 629 630 fp = 0; 631 632 /* check if there's a place for the new queue */ 633 if (nipq > maxnipq) { 634 /* 635 * drop something from the tail of the current queue 636 * before proceeding further 637 */ 638 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 639 for (i = 0; i < IPREASS_NHASH; i++) { 640 if (ipq[i].prev != &ipq[i]) { 641 ip_freef(ipq[i].prev); 642 break; 643 } 644 } 645 } else 646 ip_freef(ipq[sum].prev); 647 } 648found: 649 /* 650 * Adjust ip_len to not reflect header, 651 * set ip_mff if more fragments are expected, 652 * convert offset of this to bytes. 653 */ 654 ip->ip_len -= hlen; 655 mff = (ip->ip_off & IP_MF) != 0; 656 if (mff) { 657 /* 658 * Make sure that fragments have a data length 659 * that's a non-zero multiple of 8 bytes. 660 */ 661 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 662 ipstat.ips_toosmall++; /* XXX */ 663 goto bad; 664 } 665 m->m_flags |= M_FRAG; 666 } 667 ip->ip_off <<= 3; 668 669 /* 670 * If datagram marked as having more fragments 671 * or if this is not the first fragment, 672 * attempt reassembly; if it succeeds, proceed. 673 */ 674 if (mff || ip->ip_off) { 675 ipstat.ips_fragments++; 676 m->m_pkthdr.header = ip; 677#ifdef IPDIVERT 678 m = ip_reass(m, 679 fp, &ipq[sum], &divert_info, &divert_cookie); 680#else 681 m = ip_reass(m, fp, &ipq[sum]); 682#endif 683 if (m == 0) { 684#ifdef IPFIREWALL_FORWARD 685 ip_fw_fwd_addr = NULL; 686#endif 687 return; 688 } 689 /* Get the length of the reassembled packets header */ 690 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 691 ipstat.ips_reassembled++; 692 ip = mtod(m, struct ip *); 693#ifdef IPDIVERT 694 /* Restore original checksum before diverting packet */ 695 if (divert_info != 0) { 696 ip->ip_len += hlen; 697 HTONS(ip->ip_len); 698 HTONS(ip->ip_off); 699 ip->ip_sum = 0; 700 ip->ip_sum = in_cksum_hdr(ip); 701 NTOHS(ip->ip_off); 702 NTOHS(ip->ip_len); 703 ip->ip_len -= hlen; 704 } 705#endif 706 } else 707 if (fp) 708 ip_freef(fp); 709 } else 710 ip->ip_len -= hlen; 711 712#ifdef IPDIVERT 713 /* 714 * Divert or tee packet to the divert protocol if required. 715 * 716 * If divert_info is zero then cookie should be too, so we shouldn't 717 * need to clear them here. Assume divert_packet() does so also. 718 */ 719 if (divert_info != 0) { 720 struct mbuf *clone = NULL; 721 722 /* Clone packet if we're doing a 'tee' */ 723 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 724 clone = m_dup(m, M_DONTWAIT); 725 726 /* Restore packet header fields to original values */ 727 ip->ip_len += hlen; 728 HTONS(ip->ip_len); 729 HTONS(ip->ip_off); 730 731 /* Deliver packet to divert input routine */ 732 ip_divert_cookie = divert_cookie; 733 divert_packet(m, 1, divert_info & 0xffff); 734 ipstat.ips_delivered++; 735 736 /* If 'tee', continue with original packet */ 737 if (clone == NULL) 738 return; 739 m = clone; 740 ip = mtod(m, struct ip *); 741 } 742#endif 743 744 /* 745 * Switch out to protocol's input routine. 746 */ 747 ipstat.ips_delivered++; 748 { 749 int off = hlen, nh = ip->ip_p; 750 751 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh); 752#ifdef IPFIREWALL_FORWARD 753 ip_fw_fwd_addr = NULL; /* tcp needed it */ 754#endif 755 return; 756 } 757bad: 758#ifdef IPFIREWALL_FORWARD 759 ip_fw_fwd_addr = NULL; 760#endif 761 m_freem(m); 762} 763 764/* 765 * IP software interrupt routine - to go away sometime soon 766 */ 767static void 768ipintr(void) 769{ 770 int s; 771 struct mbuf *m; 772 773 while(1) { 774 s = splimp(); 775 IF_DEQUEUE(&ipintrq, m); 776 splx(s); 777 if (m == 0) 778 return; 779 ip_input(m); 780 } 781} 782 783/* 784 * Take incoming datagram fragment and try to reassemble it into 785 * whole datagram. If a chain for reassembly of this datagram already 786 * exists, then it is given as fp; otherwise have to make a chain. 787 * 788 * When IPDIVERT enabled, keep additional state with each packet that 789 * tells us if we need to divert or tee the packet we're building. 790 */ 791 792static struct mbuf * 793#ifdef IPDIVERT 794ip_reass(m, fp, where, divinfo, divcookie) 795#else 796ip_reass(m, fp, where) 797#endif 798 register struct mbuf *m; 799 register struct ipq *fp; 800 struct ipq *where; 801#ifdef IPDIVERT 802 u_int32_t *divinfo; 803 u_int16_t *divcookie; 804#endif 805{ 806 struct ip *ip = mtod(m, struct ip *); 807 register struct mbuf *p = 0, *q, *nq; 808 struct mbuf *t; 809 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 810 int i, next; 811 812 /* 813 * Presence of header sizes in mbufs 814 * would confuse code below. 815 */ 816 m->m_data += hlen; 817 m->m_len -= hlen; 818 819 /* 820 * If first fragment to arrive, create a reassembly queue. 821 */ 822 if (fp == 0) { 823 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 824 goto dropfrag; 825 fp = mtod(t, struct ipq *); 826 insque(fp, where); 827 nipq++; 828 fp->ipq_ttl = IPFRAGTTL; 829 fp->ipq_p = ip->ip_p; 830 fp->ipq_id = ip->ip_id; 831 fp->ipq_src = ip->ip_src; 832 fp->ipq_dst = ip->ip_dst; 833 fp->ipq_frags = m; 834 m->m_nextpkt = NULL; 835#ifdef IPDIVERT 836 fp->ipq_div_info = 0; 837 fp->ipq_div_cookie = 0; 838#endif 839 goto inserted; 840 } 841 842#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 843 844 /* 845 * Find a segment which begins after this one does. 846 */ 847 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 848 if (GETIP(q)->ip_off > ip->ip_off) 849 break; 850 851 /* 852 * If there is a preceding segment, it may provide some of 853 * our data already. If so, drop the data from the incoming 854 * segment. If it provides all of our data, drop us, otherwise 855 * stick new segment in the proper place. 856 * 857 * If some of the data is dropped from the the preceding 858 * segment, then it's checksum is invalidated. 859 */ 860 if (p) { 861 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 862 if (i > 0) { 863 if (i >= ip->ip_len) 864 goto dropfrag; 865 m_adj(m, i); 866 m->m_pkthdr.csum_flags = 0; 867 ip->ip_off += i; 868 ip->ip_len -= i; 869 } 870 m->m_nextpkt = p->m_nextpkt; 871 p->m_nextpkt = m; 872 } else { 873 m->m_nextpkt = fp->ipq_frags; 874 fp->ipq_frags = m; 875 } 876 877 /* 878 * While we overlap succeeding segments trim them or, 879 * if they are completely covered, dequeue them. 880 */ 881 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 882 q = nq) { 883 i = (ip->ip_off + ip->ip_len) - 884 GETIP(q)->ip_off; 885 if (i < GETIP(q)->ip_len) { 886 GETIP(q)->ip_len -= i; 887 GETIP(q)->ip_off += i; 888 m_adj(q, i); 889 q->m_pkthdr.csum_flags = 0; 890 break; 891 } 892 nq = q->m_nextpkt; 893 m->m_nextpkt = nq; 894 m_freem(q); 895 } 896 897inserted: 898 899#ifdef IPDIVERT 900 /* 901 * Transfer firewall instructions to the fragment structure. 902 * Any fragment diverting causes the whole packet to divert. 903 */ 904 fp->ipq_div_info = *divinfo; 905 fp->ipq_div_cookie = *divcookie; 906 *divinfo = 0; 907 *divcookie = 0; 908#endif 909 910 /* 911 * Check for complete reassembly. 912 */ 913 next = 0; 914 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 915 if (GETIP(q)->ip_off != next) 916 return (0); 917 next += GETIP(q)->ip_len; 918 } 919 /* Make sure the last packet didn't have the IP_MF flag */ 920 if (p->m_flags & M_FRAG) 921 return (0); 922 923 /* 924 * Reassembly is complete. Make sure the packet is a sane size. 925 */ 926 q = fp->ipq_frags; 927 ip = GETIP(q); 928 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 929 ipstat.ips_toolong++; 930 ip_freef(fp); 931 return (0); 932 } 933 934 /* 935 * Concatenate fragments. 936 */ 937 m = q; 938 t = m->m_next; 939 m->m_next = 0; 940 m_cat(m, t); 941 nq = q->m_nextpkt; 942 q->m_nextpkt = 0; 943 for (q = nq; q != NULL; q = nq) { 944 nq = q->m_nextpkt; 945 q->m_nextpkt = NULL; 946 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 947 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 948 m_cat(m, q); 949 } 950 951#ifdef IPDIVERT 952 /* 953 * Extract firewall instructions from the fragment structure. 954 */ 955 *divinfo = fp->ipq_div_info; 956 *divcookie = fp->ipq_div_cookie; 957#endif 958 959 /* 960 * Create header for new ip packet by 961 * modifying header of first packet; 962 * dequeue and discard fragment reassembly header. 963 * Make header visible. 964 */ 965 ip->ip_len = next; 966 ip->ip_src = fp->ipq_src; 967 ip->ip_dst = fp->ipq_dst; 968 remque(fp); 969 nipq--; 970 (void) m_free(dtom(fp)); 971 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 972 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 973 /* some debugging cruft by sklower, below, will go away soon */ 974 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 975 register int plen = 0; 976 for (t = m; t; t = t->m_next) 977 plen += t->m_len; 978 m->m_pkthdr.len = plen; 979 } 980 return (m); 981 982dropfrag: 983#ifdef IPDIVERT 984 *divinfo = 0; 985 *divcookie = 0; 986#endif 987 ipstat.ips_fragdropped++; 988 m_freem(m); 989 return (0); 990 991#undef GETIP 992} 993 994/* 995 * Free a fragment reassembly header and all 996 * associated datagrams. 997 */ 998static void 999ip_freef(fp) 1000 struct ipq *fp; 1001{ 1002 register struct mbuf *q; 1003 1004 while (fp->ipq_frags) { 1005 q = fp->ipq_frags; 1006 fp->ipq_frags = q->m_nextpkt; 1007 m_freem(q); 1008 } 1009 remque(fp); 1010 (void) m_free(dtom(fp)); 1011 nipq--; 1012} 1013 1014/* 1015 * IP timer processing; 1016 * if a timer expires on a reassembly 1017 * queue, discard it. 1018 */ 1019void 1020ip_slowtimo() 1021{ 1022 register struct ipq *fp; 1023 int s = splnet(); 1024 int i; 1025 1026 for (i = 0; i < IPREASS_NHASH; i++) { 1027 fp = ipq[i].next; 1028 if (fp == 0) 1029 continue; 1030 while (fp != &ipq[i]) { 1031 --fp->ipq_ttl; 1032 fp = fp->next; 1033 if (fp->prev->ipq_ttl == 0) { 1034 ipstat.ips_fragtimeout++; 1035 ip_freef(fp->prev); 1036 } 1037 } 1038 } 1039 ipflow_slowtimo(); 1040 splx(s); 1041} 1042 1043/* 1044 * Drain off all datagram fragments. 1045 */ 1046void 1047ip_drain() 1048{ 1049 int i; 1050 1051 for (i = 0; i < IPREASS_NHASH; i++) { 1052 while (ipq[i].next != &ipq[i]) { 1053 ipstat.ips_fragdropped++; 1054 ip_freef(ipq[i].next); 1055 } 1056 } 1057 in_rtqdrain(); 1058} 1059 1060/* 1061 * Do option processing on a datagram, 1062 * possibly discarding it if bad options are encountered, 1063 * or forwarding it if source-routed. 1064 * Returns 1 if packet has been forwarded/freed, 1065 * 0 if the packet should be processed further. 1066 */ 1067static int 1068ip_dooptions(m) 1069 struct mbuf *m; 1070{ 1071 register struct ip *ip = mtod(m, struct ip *); 1072 register u_char *cp; 1073 register struct ip_timestamp *ipt; 1074 register struct in_ifaddr *ia; 1075 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1076 struct in_addr *sin, dst; 1077 n_time ntime; 1078 1079 dst = ip->ip_dst; 1080 cp = (u_char *)(ip + 1); 1081 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1082 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1083 opt = cp[IPOPT_OPTVAL]; 1084 if (opt == IPOPT_EOL) 1085 break; 1086 if (opt == IPOPT_NOP) 1087 optlen = 1; 1088 else { 1089 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1090 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1091 goto bad; 1092 } 1093 optlen = cp[IPOPT_OLEN]; 1094 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1095 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1096 goto bad; 1097 } 1098 } 1099 switch (opt) { 1100 1101 default: 1102 break; 1103 1104 /* 1105 * Source routing with record. 1106 * Find interface with current destination address. 1107 * If none on this machine then drop if strictly routed, 1108 * or do nothing if loosely routed. 1109 * Record interface address and bring up next address 1110 * component. If strictly routed make sure next 1111 * address is on directly accessible net. 1112 */ 1113 case IPOPT_LSRR: 1114 case IPOPT_SSRR: 1115 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1116 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1117 goto bad; 1118 } 1119 ipaddr.sin_addr = ip->ip_dst; 1120 ia = (struct in_ifaddr *) 1121 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1122 if (ia == 0) { 1123 if (opt == IPOPT_SSRR) { 1124 type = ICMP_UNREACH; 1125 code = ICMP_UNREACH_SRCFAIL; 1126 goto bad; 1127 } 1128 if (!ip_dosourceroute) 1129 goto nosourcerouting; 1130 /* 1131 * Loose routing, and not at next destination 1132 * yet; nothing to do except forward. 1133 */ 1134 break; 1135 } 1136 off--; /* 0 origin */ 1137 if (off > optlen - (int)sizeof(struct in_addr)) { 1138 /* 1139 * End of source route. Should be for us. 1140 */ 1141 if (!ip_acceptsourceroute) 1142 goto nosourcerouting; 1143 save_rte(cp, ip->ip_src); 1144 break; 1145 } 1146 1147 if (!ip_dosourceroute) { 1148 if (ipforwarding) { 1149 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1150 /* 1151 * Acting as a router, so generate ICMP 1152 */ 1153nosourcerouting: 1154 strcpy(buf, inet_ntoa(ip->ip_dst)); 1155 log(LOG_WARNING, 1156 "attempted source route from %s to %s\n", 1157 inet_ntoa(ip->ip_src), buf); 1158 type = ICMP_UNREACH; 1159 code = ICMP_UNREACH_SRCFAIL; 1160 goto bad; 1161 } else { 1162 /* 1163 * Not acting as a router, so silently drop. 1164 */ 1165 ipstat.ips_cantforward++; 1166 m_freem(m); 1167 return (1); 1168 } 1169 } 1170 1171 /* 1172 * locate outgoing interface 1173 */ 1174 (void)memcpy(&ipaddr.sin_addr, cp + off, 1175 sizeof(ipaddr.sin_addr)); 1176 1177 if (opt == IPOPT_SSRR) { 1178#define INA struct in_ifaddr * 1179#define SA struct sockaddr * 1180 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1181 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1182 } else 1183 ia = ip_rtaddr(ipaddr.sin_addr); 1184 if (ia == 0) { 1185 type = ICMP_UNREACH; 1186 code = ICMP_UNREACH_SRCFAIL; 1187 goto bad; 1188 } 1189 ip->ip_dst = ipaddr.sin_addr; 1190 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1191 sizeof(struct in_addr)); 1192 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1193 /* 1194 * Let ip_intr's mcast routing check handle mcast pkts 1195 */ 1196 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1197 break; 1198 1199 case IPOPT_RR: 1200 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1201 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1202 goto bad; 1203 } 1204 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1205 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1206 goto bad; 1207 } 1208 /* 1209 * If no space remains, ignore. 1210 */ 1211 off--; /* 0 origin */ 1212 if (off > optlen - (int)sizeof(struct in_addr)) 1213 break; 1214 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1215 sizeof(ipaddr.sin_addr)); 1216 /* 1217 * locate outgoing interface; if we're the destination, 1218 * use the incoming interface (should be same). 1219 */ 1220 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1221 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1222 type = ICMP_UNREACH; 1223 code = ICMP_UNREACH_HOST; 1224 goto bad; 1225 } 1226 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1227 sizeof(struct in_addr)); 1228 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1229 break; 1230 1231 case IPOPT_TS: 1232 code = cp - (u_char *)ip; 1233 ipt = (struct ip_timestamp *)cp; 1234 if (ipt->ipt_len < 5) 1235 goto bad; 1236 if (ipt->ipt_ptr > 1237 ipt->ipt_len - (int)sizeof(int32_t)) { 1238 if (++ipt->ipt_oflw == 0) 1239 goto bad; 1240 break; 1241 } 1242 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1243 switch (ipt->ipt_flg) { 1244 1245 case IPOPT_TS_TSONLY: 1246 break; 1247 1248 case IPOPT_TS_TSANDADDR: 1249 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1250 sizeof(struct in_addr) > ipt->ipt_len) 1251 goto bad; 1252 ipaddr.sin_addr = dst; 1253 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1254 m->m_pkthdr.rcvif); 1255 if (ia == 0) 1256 continue; 1257 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1258 sizeof(struct in_addr)); 1259 ipt->ipt_ptr += sizeof(struct in_addr); 1260 break; 1261 1262 case IPOPT_TS_PRESPEC: 1263 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1264 sizeof(struct in_addr) > ipt->ipt_len) 1265 goto bad; 1266 (void)memcpy(&ipaddr.sin_addr, sin, 1267 sizeof(struct in_addr)); 1268 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1269 continue; 1270 ipt->ipt_ptr += sizeof(struct in_addr); 1271 break; 1272 1273 default: 1274 goto bad; 1275 } 1276 ntime = iptime(); 1277 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1278 sizeof(n_time)); 1279 ipt->ipt_ptr += sizeof(n_time); 1280 } 1281 } 1282 if (forward && ipforwarding) { 1283 ip_forward(m, 1); 1284 return (1); 1285 } 1286 return (0); 1287bad: 1288 icmp_error(m, type, code, 0, 0); 1289 ipstat.ips_badoptions++; 1290 return (1); 1291} 1292 1293/* 1294 * Given address of next destination (final or next hop), 1295 * return internet address info of interface to be used to get there. 1296 */ 1297static struct in_ifaddr * 1298ip_rtaddr(dst) 1299 struct in_addr dst; 1300{ 1301 register struct sockaddr_in *sin; 1302 1303 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1304 1305 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1306 if (ipforward_rt.ro_rt) { 1307 RTFREE(ipforward_rt.ro_rt); 1308 ipforward_rt.ro_rt = 0; 1309 } 1310 sin->sin_family = AF_INET; 1311 sin->sin_len = sizeof(*sin); 1312 sin->sin_addr = dst; 1313 1314 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1315 } 1316 if (ipforward_rt.ro_rt == 0) 1317 return ((struct in_ifaddr *)0); 1318 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1319} 1320 1321/* 1322 * Save incoming source route for use in replies, 1323 * to be picked up later by ip_srcroute if the receiver is interested. 1324 */ 1325void 1326save_rte(option, dst) 1327 u_char *option; 1328 struct in_addr dst; 1329{ 1330 unsigned olen; 1331 1332 olen = option[IPOPT_OLEN]; 1333#ifdef DIAGNOSTIC 1334 if (ipprintfs) 1335 printf("save_rte: olen %d\n", olen); 1336#endif 1337 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1338 return; 1339 bcopy(option, ip_srcrt.srcopt, olen); 1340 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1341 ip_srcrt.dst = dst; 1342} 1343 1344/* 1345 * Retrieve incoming source route for use in replies, 1346 * in the same form used by setsockopt. 1347 * The first hop is placed before the options, will be removed later. 1348 */ 1349struct mbuf * 1350ip_srcroute() 1351{ 1352 register struct in_addr *p, *q; 1353 register struct mbuf *m; 1354 1355 if (ip_nhops == 0) 1356 return ((struct mbuf *)0); 1357 m = m_get(M_DONTWAIT, MT_HEADER); 1358 if (m == 0) 1359 return ((struct mbuf *)0); 1360 1361#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1362 1363 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1364 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1365 OPTSIZ; 1366#ifdef DIAGNOSTIC 1367 if (ipprintfs) 1368 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1369#endif 1370 1371 /* 1372 * First save first hop for return route 1373 */ 1374 p = &ip_srcrt.route[ip_nhops - 1]; 1375 *(mtod(m, struct in_addr *)) = *p--; 1376#ifdef DIAGNOSTIC 1377 if (ipprintfs) 1378 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1379#endif 1380 1381 /* 1382 * Copy option fields and padding (nop) to mbuf. 1383 */ 1384 ip_srcrt.nop = IPOPT_NOP; 1385 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1386 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1387 &ip_srcrt.nop, OPTSIZ); 1388 q = (struct in_addr *)(mtod(m, caddr_t) + 1389 sizeof(struct in_addr) + OPTSIZ); 1390#undef OPTSIZ 1391 /* 1392 * Record return path as an IP source route, 1393 * reversing the path (pointers are now aligned). 1394 */ 1395 while (p >= ip_srcrt.route) { 1396#ifdef DIAGNOSTIC 1397 if (ipprintfs) 1398 printf(" %lx", (u_long)ntohl(q->s_addr)); 1399#endif 1400 *q++ = *p--; 1401 } 1402 /* 1403 * Last hop goes to final destination. 1404 */ 1405 *q = ip_srcrt.dst; 1406#ifdef DIAGNOSTIC 1407 if (ipprintfs) 1408 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1409#endif 1410 return (m); 1411} 1412 1413/* 1414 * Strip out IP options, at higher 1415 * level protocol in the kernel. 1416 * Second argument is buffer to which options 1417 * will be moved, and return value is their length. 1418 * XXX should be deleted; last arg currently ignored. 1419 */ 1420void 1421ip_stripoptions(m, mopt) 1422 register struct mbuf *m; 1423 struct mbuf *mopt; 1424{ 1425 register int i; 1426 struct ip *ip = mtod(m, struct ip *); 1427 register caddr_t opts; 1428 int olen; 1429 1430 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1431 opts = (caddr_t)(ip + 1); 1432 i = m->m_len - (sizeof (struct ip) + olen); 1433 bcopy(opts + olen, opts, (unsigned)i); 1434 m->m_len -= olen; 1435 if (m->m_flags & M_PKTHDR) 1436 m->m_pkthdr.len -= olen; 1437 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1438} 1439 1440u_char inetctlerrmap[PRC_NCMDS] = { 1441 0, 0, 0, 0, 1442 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1443 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1444 EMSGSIZE, EHOSTUNREACH, 0, 0, 1445 0, 0, 0, 0, 1446 ENOPROTOOPT 1447}; 1448 1449/* 1450 * Forward a packet. If some error occurs return the sender 1451 * an icmp packet. Note we can't always generate a meaningful 1452 * icmp message because icmp doesn't have a large enough repertoire 1453 * of codes and types. 1454 * 1455 * If not forwarding, just drop the packet. This could be confusing 1456 * if ipforwarding was zero but some routing protocol was advancing 1457 * us as a gateway to somewhere. However, we must let the routing 1458 * protocol deal with that. 1459 * 1460 * The srcrt parameter indicates whether the packet is being forwarded 1461 * via a source route. 1462 */ 1463static void 1464ip_forward(m, srcrt) 1465 struct mbuf *m; 1466 int srcrt; 1467{ 1468 register struct ip *ip = mtod(m, struct ip *); 1469 register struct sockaddr_in *sin; 1470 register struct rtentry *rt; 1471 int error, type = 0, code = 0; 1472 struct mbuf *mcopy; 1473 n_long dest; 1474 struct ifnet *destifp; 1475#ifdef IPSEC 1476 struct ifnet dummyifp; 1477#endif 1478 1479 dest = 0; 1480#ifdef DIAGNOSTIC 1481 if (ipprintfs) 1482 printf("forward: src %lx dst %lx ttl %x\n", 1483 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1484 ip->ip_ttl); 1485#endif 1486 1487 1488 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1489 ipstat.ips_cantforward++; 1490 m_freem(m); 1491 return; 1492 } 1493#ifdef IPSTEALTH 1494 if (!ipstealth) { 1495#endif 1496 if (ip->ip_ttl <= IPTTLDEC) { 1497 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1498 dest, 0); 1499 return; 1500 } 1501#ifdef IPSTEALTH 1502 } 1503#endif 1504 1505 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1506 if ((rt = ipforward_rt.ro_rt) == 0 || 1507 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1508 if (ipforward_rt.ro_rt) { 1509 RTFREE(ipforward_rt.ro_rt); 1510 ipforward_rt.ro_rt = 0; 1511 } 1512 sin->sin_family = AF_INET; 1513 sin->sin_len = sizeof(*sin); 1514 sin->sin_addr = ip->ip_dst; 1515 1516 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1517 if (ipforward_rt.ro_rt == 0) { 1518 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1519 return; 1520 } 1521 rt = ipforward_rt.ro_rt; 1522 } 1523 1524 /* 1525 * Save at most 64 bytes of the packet in case 1526 * we need to generate an ICMP message to the src. 1527 */ 1528 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1529 if (mcopy && (mcopy->m_flags & M_EXT)) 1530 m_copydata(mcopy, 0, sizeof(struct ip), mtod(mcopy, caddr_t)); 1531 1532#ifdef IPSTEALTH 1533 if (!ipstealth) { 1534#endif 1535 ip->ip_ttl -= IPTTLDEC; 1536#ifdef IPSTEALTH 1537 } 1538#endif 1539 1540 /* 1541 * If forwarding packet using same interface that it came in on, 1542 * perhaps should send a redirect to sender to shortcut a hop. 1543 * Only send redirect if source is sending directly to us, 1544 * and if packet was not source routed (or has any options). 1545 * Also, don't send redirect if forwarding using a default route 1546 * or a route modified by a redirect. 1547 */ 1548#define satosin(sa) ((struct sockaddr_in *)(sa)) 1549 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1550 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1551 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1552 ipsendredirects && !srcrt) { 1553#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1554 u_long src = ntohl(ip->ip_src.s_addr); 1555 1556 if (RTA(rt) && 1557 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1558 if (rt->rt_flags & RTF_GATEWAY) 1559 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1560 else 1561 dest = ip->ip_dst.s_addr; 1562 /* Router requirements says to only send host redirects */ 1563 type = ICMP_REDIRECT; 1564 code = ICMP_REDIRECT_HOST; 1565#ifdef DIAGNOSTIC 1566 if (ipprintfs) 1567 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1568#endif 1569 } 1570 } 1571 1572 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1573 IP_FORWARDING, 0); 1574 if (error) 1575 ipstat.ips_cantforward++; 1576 else { 1577 ipstat.ips_forward++; 1578 if (type) 1579 ipstat.ips_redirectsent++; 1580 else { 1581 if (mcopy) { 1582 ipflow_create(&ipforward_rt, mcopy); 1583 m_freem(mcopy); 1584 } 1585 return; 1586 } 1587 } 1588 if (mcopy == NULL) 1589 return; 1590 destifp = NULL; 1591 1592 switch (error) { 1593 1594 case 0: /* forwarded, but need redirect */ 1595 /* type, code set above */ 1596 break; 1597 1598 case ENETUNREACH: /* shouldn't happen, checked above */ 1599 case EHOSTUNREACH: 1600 case ENETDOWN: 1601 case EHOSTDOWN: 1602 default: 1603 type = ICMP_UNREACH; 1604 code = ICMP_UNREACH_HOST; 1605 break; 1606 1607 case EMSGSIZE: 1608 type = ICMP_UNREACH; 1609 code = ICMP_UNREACH_NEEDFRAG; 1610#ifndef IPSEC 1611 if (ipforward_rt.ro_rt) 1612 destifp = ipforward_rt.ro_rt->rt_ifp; 1613#else 1614 /* 1615 * If the packet is routed over IPsec tunnel, tell the 1616 * originator the tunnel MTU. 1617 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1618 * XXX quickhack!!! 1619 */ 1620 if (ipforward_rt.ro_rt) { 1621 struct secpolicy *sp = NULL; 1622 int ipsecerror; 1623 int ipsechdr; 1624 struct route *ro; 1625 1626 sp = ipsec4_getpolicybyaddr(mcopy, 1627 IPSEC_DIR_OUTBOUND, 1628 IP_FORWARDING, 1629 &ipsecerror); 1630 1631 if (sp == NULL) 1632 destifp = ipforward_rt.ro_rt->rt_ifp; 1633 else { 1634 /* count IPsec header size */ 1635 ipsechdr = ipsec4_hdrsiz(mcopy, 1636 IPSEC_DIR_OUTBOUND, 1637 NULL); 1638 1639 /* 1640 * find the correct route for outer IPv4 1641 * header, compute tunnel MTU. 1642 * 1643 * XXX BUG ALERT 1644 * The "dummyifp" code relies upon the fact 1645 * that icmp_error() touches only ifp->if_mtu. 1646 */ 1647 /*XXX*/ 1648 destifp = NULL; 1649 if (sp->req != NULL 1650 && sp->req->sav != NULL 1651 && sp->req->sav->sah != NULL) { 1652 ro = &sp->req->sav->sah->sa_route; 1653 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1654 dummyifp.if_mtu = 1655 ro->ro_rt->rt_ifp->if_mtu; 1656 dummyifp.if_mtu -= ipsechdr; 1657 destifp = &dummyifp; 1658 } 1659 } 1660 1661 key_freesp(sp); 1662 } 1663 } 1664#endif /*IPSEC*/ 1665 ipstat.ips_cantfrag++; 1666 break; 1667 1668 case ENOBUFS: 1669 type = ICMP_SOURCEQUENCH; 1670 code = 0; 1671 break; 1672 1673 case EACCES: /* ipfw denied packet */ 1674 m_freem(mcopy); 1675 return; 1676 } 1677 if (mcopy->m_flags & M_EXT) 1678 m_copyback(mcopy, 0, sizeof(struct ip), mtod(mcopy, caddr_t)); 1679 icmp_error(mcopy, type, code, dest, destifp); 1680} 1681 1682void 1683ip_savecontrol(inp, mp, ip, m) 1684 register struct inpcb *inp; 1685 register struct mbuf **mp; 1686 register struct ip *ip; 1687 register struct mbuf *m; 1688{ 1689 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1690 struct timeval tv; 1691 1692 microtime(&tv); 1693 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1694 SCM_TIMESTAMP, SOL_SOCKET); 1695 if (*mp) 1696 mp = &(*mp)->m_next; 1697 } 1698 if (inp->inp_flags & INP_RECVDSTADDR) { 1699 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1700 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1701 if (*mp) 1702 mp = &(*mp)->m_next; 1703 } 1704#ifdef notyet 1705 /* XXX 1706 * Moving these out of udp_input() made them even more broken 1707 * than they already were. 1708 */ 1709 /* options were tossed already */ 1710 if (inp->inp_flags & INP_RECVOPTS) { 1711 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1712 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1713 if (*mp) 1714 mp = &(*mp)->m_next; 1715 } 1716 /* ip_srcroute doesn't do what we want here, need to fix */ 1717 if (inp->inp_flags & INP_RECVRETOPTS) { 1718 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1719 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1720 if (*mp) 1721 mp = &(*mp)->m_next; 1722 } 1723#endif 1724 if (inp->inp_flags & INP_RECVIF) { 1725 struct ifnet *ifp; 1726 struct sdlbuf { 1727 struct sockaddr_dl sdl; 1728 u_char pad[32]; 1729 } sdlbuf; 1730 struct sockaddr_dl *sdp; 1731 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1732 1733 if (((ifp = m->m_pkthdr.rcvif)) 1734 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1735 sdp = (struct sockaddr_dl *)(ifnet_addrs 1736 [ifp->if_index - 1]->ifa_addr); 1737 /* 1738 * Change our mind and don't try copy. 1739 */ 1740 if ((sdp->sdl_family != AF_LINK) 1741 || (sdp->sdl_len > sizeof(sdlbuf))) { 1742 goto makedummy; 1743 } 1744 bcopy(sdp, sdl2, sdp->sdl_len); 1745 } else { 1746makedummy: 1747 sdl2->sdl_len 1748 = offsetof(struct sockaddr_dl, sdl_data[0]); 1749 sdl2->sdl_family = AF_LINK; 1750 sdl2->sdl_index = 0; 1751 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1752 } 1753 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1754 IP_RECVIF, IPPROTO_IP); 1755 if (*mp) 1756 mp = &(*mp)->m_next; 1757 } 1758} 1759 1760int 1761ip_rsvp_init(struct socket *so) 1762{ 1763 if (so->so_type != SOCK_RAW || 1764 so->so_proto->pr_protocol != IPPROTO_RSVP) 1765 return EOPNOTSUPP; 1766 1767 if (ip_rsvpd != NULL) 1768 return EADDRINUSE; 1769 1770 ip_rsvpd = so; 1771 /* 1772 * This may seem silly, but we need to be sure we don't over-increment 1773 * the RSVP counter, in case something slips up. 1774 */ 1775 if (!ip_rsvp_on) { 1776 ip_rsvp_on = 1; 1777 rsvp_on++; 1778 } 1779 1780 return 0; 1781} 1782 1783int 1784ip_rsvp_done(void) 1785{ 1786 ip_rsvpd = NULL; 1787 /* 1788 * This may seem silly, but we need to be sure we don't over-decrement 1789 * the RSVP counter, in case something slips up. 1790 */ 1791 if (ip_rsvp_on) { 1792 ip_rsvp_on = 0; 1793 rsvp_on--; 1794 } 1795 return 0; 1796} 1797