ip_input.c revision 64075
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 64075 2000-07-31 23:41:47Z ache $ 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 * (The IP ident field is put in the same byte order 531 * as expected when ip_mforward() is called from 532 * ip_output().) 533 */ 534 ip->ip_id = htons(ip->ip_id); 535 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 536 ipstat.ips_cantforward++; 537 m_freem(m); 538 return; 539 } 540 ip->ip_id = ntohs(ip->ip_id); 541 542 /* 543 * The process-level routing demon needs to receive 544 * all multicast IGMP packets, whether or not this 545 * host belongs to their destination groups. 546 */ 547 if (ip->ip_p == IPPROTO_IGMP) 548 goto ours; 549 ipstat.ips_forward++; 550 } 551 /* 552 * See if we belong to the destination multicast group on the 553 * arrival interface. 554 */ 555 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 556 if (inm == NULL) { 557 ipstat.ips_notmember++; 558 m_freem(m); 559 return; 560 } 561 goto ours; 562 } 563 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 564 goto ours; 565 if (ip->ip_dst.s_addr == INADDR_ANY) 566 goto ours; 567 568#if defined(NFAITH) && 0 < NFAITH 569 /* 570 * FAITH(Firewall Aided Internet Translator) 571 */ 572 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 573 if (ip_keepfaith) { 574 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 575 goto ours; 576 } 577 m_freem(m); 578 return; 579 } 580#endif 581 /* 582 * Not for us; forward if possible and desirable. 583 */ 584 if (ipforwarding == 0) { 585 ipstat.ips_cantforward++; 586 m_freem(m); 587 } else 588 ip_forward(m, 0); 589#ifdef IPFIREWALL_FORWARD 590 ip_fw_fwd_addr = NULL; 591#endif 592 return; 593 594ours: 595 596 /* 597 * If offset or IP_MF are set, must reassemble. 598 * Otherwise, nothing need be done. 599 * (We could look in the reassembly queue to see 600 * if the packet was previously fragmented, 601 * but it's not worth the time; just let them time out.) 602 */ 603 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) { 604 605#if 0 /* 606 * Reassembly should be able to treat a mbuf cluster, for later 607 * operation of contiguous protocol headers on the cluster. (KAME) 608 */ 609 if (m->m_flags & M_EXT) { /* XXX */ 610 if ((m = m_pullup(m, hlen)) == 0) { 611 ipstat.ips_toosmall++; 612#ifdef IPFIREWALL_FORWARD 613 ip_fw_fwd_addr = NULL; 614#endif 615 return; 616 } 617 ip = mtod(m, struct ip *); 618 } 619#endif 620 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 621 /* 622 * Look for queue of fragments 623 * of this datagram. 624 */ 625 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 626 if (ip->ip_id == fp->ipq_id && 627 ip->ip_src.s_addr == fp->ipq_src.s_addr && 628 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 629 ip->ip_p == fp->ipq_p) 630 goto found; 631 632 fp = 0; 633 634 /* check if there's a place for the new queue */ 635 if (nipq > maxnipq) { 636 /* 637 * drop something from the tail of the current queue 638 * before proceeding further 639 */ 640 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 641 for (i = 0; i < IPREASS_NHASH; i++) { 642 if (ipq[i].prev != &ipq[i]) { 643 ip_freef(ipq[i].prev); 644 break; 645 } 646 } 647 } else 648 ip_freef(ipq[sum].prev); 649 } 650found: 651 /* 652 * Adjust ip_len to not reflect header, 653 * set ip_mff if more fragments are expected, 654 * convert offset of this to bytes. 655 */ 656 ip->ip_len -= hlen; 657 mff = (ip->ip_off & IP_MF) != 0; 658 if (mff) { 659 /* 660 * Make sure that fragments have a data length 661 * that's a non-zero multiple of 8 bytes. 662 */ 663 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 664 ipstat.ips_toosmall++; /* XXX */ 665 goto bad; 666 } 667 m->m_flags |= M_FRAG; 668 } 669 ip->ip_off <<= 3; 670 671 /* 672 * If datagram marked as having more fragments 673 * or if this is not the first fragment, 674 * attempt reassembly; if it succeeds, proceed. 675 */ 676 if (mff || ip->ip_off) { 677 ipstat.ips_fragments++; 678 m->m_pkthdr.header = ip; 679#ifdef IPDIVERT 680 m = ip_reass(m, 681 fp, &ipq[sum], &divert_info, &divert_cookie); 682#else 683 m = ip_reass(m, fp, &ipq[sum]); 684#endif 685 if (m == 0) { 686#ifdef IPFIREWALL_FORWARD 687 ip_fw_fwd_addr = NULL; 688#endif 689 return; 690 } 691 /* Get the length of the reassembled packets header */ 692 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 693 ipstat.ips_reassembled++; 694 ip = mtod(m, struct ip *); 695#ifdef IPDIVERT 696 /* Restore original checksum before diverting packet */ 697 if (divert_info != 0) { 698 ip->ip_len += hlen; 699 HTONS(ip->ip_len); 700 HTONS(ip->ip_off); 701 HTONS(ip->ip_id); 702 ip->ip_sum = 0; 703 ip->ip_sum = in_cksum_hdr(ip); 704 NTOHS(ip->ip_id); 705 NTOHS(ip->ip_off); 706 NTOHS(ip->ip_len); 707 ip->ip_len -= hlen; 708 } 709#endif 710 } else 711 if (fp) 712 ip_freef(fp); 713 } else 714 ip->ip_len -= hlen; 715 716#ifdef IPDIVERT 717 /* 718 * Divert or tee packet to the divert protocol if required. 719 * 720 * If divert_info is zero then cookie should be too, so we shouldn't 721 * need to clear them here. Assume divert_packet() does so also. 722 */ 723 if (divert_info != 0) { 724 struct mbuf *clone = NULL; 725 726 /* Clone packet if we're doing a 'tee' */ 727 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 728 clone = m_dup(m, M_DONTWAIT); 729 730 /* Restore packet header fields to original values */ 731 ip->ip_len += hlen; 732 HTONS(ip->ip_len); 733 HTONS(ip->ip_off); 734 HTONS(ip->ip_id); 735 736 /* Deliver packet to divert input routine */ 737 ip_divert_cookie = divert_cookie; 738 divert_packet(m, 1, divert_info & 0xffff); 739 ipstat.ips_delivered++; 740 741 /* If 'tee', continue with original packet */ 742 if (clone == NULL) 743 return; 744 m = clone; 745 ip = mtod(m, struct ip *); 746 } 747#endif 748 749 /* 750 * Switch out to protocol's input routine. 751 */ 752 ipstat.ips_delivered++; 753 { 754 int off = hlen, nh = ip->ip_p; 755 756 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh); 757#ifdef IPFIREWALL_FORWARD 758 ip_fw_fwd_addr = NULL; /* tcp needed it */ 759#endif 760 return; 761 } 762bad: 763#ifdef IPFIREWALL_FORWARD 764 ip_fw_fwd_addr = NULL; 765#endif 766 m_freem(m); 767} 768 769/* 770 * IP software interrupt routine - to go away sometime soon 771 */ 772static void 773ipintr(void) 774{ 775 int s; 776 struct mbuf *m; 777 778 while(1) { 779 s = splimp(); 780 IF_DEQUEUE(&ipintrq, m); 781 splx(s); 782 if (m == 0) 783 return; 784 ip_input(m); 785 } 786} 787 788/* 789 * Take incoming datagram fragment and try to reassemble it into 790 * whole datagram. If a chain for reassembly of this datagram already 791 * exists, then it is given as fp; otherwise have to make a chain. 792 * 793 * When IPDIVERT enabled, keep additional state with each packet that 794 * tells us if we need to divert or tee the packet we're building. 795 */ 796 797static struct mbuf * 798#ifdef IPDIVERT 799ip_reass(m, fp, where, divinfo, divcookie) 800#else 801ip_reass(m, fp, where) 802#endif 803 register struct mbuf *m; 804 register struct ipq *fp; 805 struct ipq *where; 806#ifdef IPDIVERT 807 u_int32_t *divinfo; 808 u_int16_t *divcookie; 809#endif 810{ 811 struct ip *ip = mtod(m, struct ip *); 812 register struct mbuf *p = 0, *q, *nq; 813 struct mbuf *t; 814 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 815 int i, next; 816 817 /* 818 * Presence of header sizes in mbufs 819 * would confuse code below. 820 */ 821 m->m_data += hlen; 822 m->m_len -= hlen; 823 824 /* 825 * If first fragment to arrive, create a reassembly queue. 826 */ 827 if (fp == 0) { 828 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 829 goto dropfrag; 830 fp = mtod(t, struct ipq *); 831 insque(fp, where); 832 nipq++; 833 fp->ipq_ttl = IPFRAGTTL; 834 fp->ipq_p = ip->ip_p; 835 fp->ipq_id = ip->ip_id; 836 fp->ipq_src = ip->ip_src; 837 fp->ipq_dst = ip->ip_dst; 838 fp->ipq_frags = m; 839 m->m_nextpkt = NULL; 840#ifdef IPDIVERT 841 fp->ipq_div_info = 0; 842 fp->ipq_div_cookie = 0; 843#endif 844 goto inserted; 845 } 846 847#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 848 849 /* 850 * Find a segment which begins after this one does. 851 */ 852 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 853 if (GETIP(q)->ip_off > ip->ip_off) 854 break; 855 856 /* 857 * If there is a preceding segment, it may provide some of 858 * our data already. If so, drop the data from the incoming 859 * segment. If it provides all of our data, drop us, otherwise 860 * stick new segment in the proper place. 861 * 862 * If some of the data is dropped from the the preceding 863 * segment, then it's checksum is invalidated. 864 */ 865 if (p) { 866 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 867 if (i > 0) { 868 if (i >= ip->ip_len) 869 goto dropfrag; 870 m_adj(m, i); 871 m->m_pkthdr.csum_flags = 0; 872 ip->ip_off += i; 873 ip->ip_len -= i; 874 } 875 m->m_nextpkt = p->m_nextpkt; 876 p->m_nextpkt = m; 877 } else { 878 m->m_nextpkt = fp->ipq_frags; 879 fp->ipq_frags = m; 880 } 881 882 /* 883 * While we overlap succeeding segments trim them or, 884 * if they are completely covered, dequeue them. 885 */ 886 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 887 q = nq) { 888 i = (ip->ip_off + ip->ip_len) - 889 GETIP(q)->ip_off; 890 if (i < GETIP(q)->ip_len) { 891 GETIP(q)->ip_len -= i; 892 GETIP(q)->ip_off += i; 893 m_adj(q, i); 894 q->m_pkthdr.csum_flags = 0; 895 break; 896 } 897 nq = q->m_nextpkt; 898 m->m_nextpkt = nq; 899 m_freem(q); 900 } 901 902inserted: 903 904#ifdef IPDIVERT 905 /* 906 * Transfer firewall instructions to the fragment structure. 907 * Any fragment diverting causes the whole packet to divert. 908 */ 909 fp->ipq_div_info = *divinfo; 910 fp->ipq_div_cookie = *divcookie; 911 *divinfo = 0; 912 *divcookie = 0; 913#endif 914 915 /* 916 * Check for complete reassembly. 917 */ 918 next = 0; 919 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 920 if (GETIP(q)->ip_off != next) 921 return (0); 922 next += GETIP(q)->ip_len; 923 } 924 /* Make sure the last packet didn't have the IP_MF flag */ 925 if (p->m_flags & M_FRAG) 926 return (0); 927 928 /* 929 * Reassembly is complete. Make sure the packet is a sane size. 930 */ 931 q = fp->ipq_frags; 932 ip = GETIP(q); 933 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 934 ipstat.ips_toolong++; 935 ip_freef(fp); 936 return (0); 937 } 938 939 /* 940 * Concatenate fragments. 941 */ 942 m = q; 943 t = m->m_next; 944 m->m_next = 0; 945 m_cat(m, t); 946 nq = q->m_nextpkt; 947 q->m_nextpkt = 0; 948 for (q = nq; q != NULL; q = nq) { 949 nq = q->m_nextpkt; 950 q->m_nextpkt = NULL; 951 m_cat(m, q); 952 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 953 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 954 } 955 956#ifdef IPDIVERT 957 /* 958 * Extract firewall instructions from the fragment structure. 959 */ 960 *divinfo = fp->ipq_div_info; 961 *divcookie = fp->ipq_div_cookie; 962#endif 963 964 /* 965 * Create header for new ip packet by 966 * modifying header of first packet; 967 * dequeue and discard fragment reassembly header. 968 * Make header visible. 969 */ 970 ip->ip_len = next; 971 ip->ip_src = fp->ipq_src; 972 ip->ip_dst = fp->ipq_dst; 973 remque(fp); 974 nipq--; 975 (void) m_free(dtom(fp)); 976 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 977 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 978 /* some debugging cruft by sklower, below, will go away soon */ 979 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 980 register int plen = 0; 981 for (t = m; t; t = t->m_next) 982 plen += t->m_len; 983 m->m_pkthdr.len = plen; 984 } 985 return (m); 986 987dropfrag: 988#ifdef IPDIVERT 989 *divinfo = 0; 990 *divcookie = 0; 991#endif 992 ipstat.ips_fragdropped++; 993 m_freem(m); 994 return (0); 995 996#undef GETIP 997} 998 999/* 1000 * Free a fragment reassembly header and all 1001 * associated datagrams. 1002 */ 1003static void 1004ip_freef(fp) 1005 struct ipq *fp; 1006{ 1007 register struct mbuf *q; 1008 1009 while (fp->ipq_frags) { 1010 q = fp->ipq_frags; 1011 fp->ipq_frags = q->m_nextpkt; 1012 m_freem(q); 1013 } 1014 remque(fp); 1015 (void) m_free(dtom(fp)); 1016 nipq--; 1017} 1018 1019/* 1020 * IP timer processing; 1021 * if a timer expires on a reassembly 1022 * queue, discard it. 1023 */ 1024void 1025ip_slowtimo() 1026{ 1027 register struct ipq *fp; 1028 int s = splnet(); 1029 int i; 1030 1031 for (i = 0; i < IPREASS_NHASH; i++) { 1032 fp = ipq[i].next; 1033 if (fp == 0) 1034 continue; 1035 while (fp != &ipq[i]) { 1036 --fp->ipq_ttl; 1037 fp = fp->next; 1038 if (fp->prev->ipq_ttl == 0) { 1039 ipstat.ips_fragtimeout++; 1040 ip_freef(fp->prev); 1041 } 1042 } 1043 } 1044 ipflow_slowtimo(); 1045 splx(s); 1046} 1047 1048/* 1049 * Drain off all datagram fragments. 1050 */ 1051void 1052ip_drain() 1053{ 1054 int i; 1055 1056 for (i = 0; i < IPREASS_NHASH; i++) { 1057 while (ipq[i].next != &ipq[i]) { 1058 ipstat.ips_fragdropped++; 1059 ip_freef(ipq[i].next); 1060 } 1061 } 1062 in_rtqdrain(); 1063} 1064 1065/* 1066 * Do option processing on a datagram, 1067 * possibly discarding it if bad options are encountered, 1068 * or forwarding it if source-routed. 1069 * Returns 1 if packet has been forwarded/freed, 1070 * 0 if the packet should be processed further. 1071 */ 1072static int 1073ip_dooptions(m) 1074 struct mbuf *m; 1075{ 1076 register struct ip *ip = mtod(m, struct ip *); 1077 register u_char *cp; 1078 register struct ip_timestamp *ipt; 1079 register struct in_ifaddr *ia; 1080 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1081 struct in_addr *sin, dst; 1082 n_time ntime; 1083 1084 dst = ip->ip_dst; 1085 cp = (u_char *)(ip + 1); 1086 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1087 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1088 opt = cp[IPOPT_OPTVAL]; 1089 if (opt == IPOPT_EOL) 1090 break; 1091 if (opt == IPOPT_NOP) 1092 optlen = 1; 1093 else { 1094 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1095 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1096 goto bad; 1097 } 1098 optlen = cp[IPOPT_OLEN]; 1099 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1100 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1101 goto bad; 1102 } 1103 } 1104 switch (opt) { 1105 1106 default: 1107 break; 1108 1109 /* 1110 * Source routing with record. 1111 * Find interface with current destination address. 1112 * If none on this machine then drop if strictly routed, 1113 * or do nothing if loosely routed. 1114 * Record interface address and bring up next address 1115 * component. If strictly routed make sure next 1116 * address is on directly accessible net. 1117 */ 1118 case IPOPT_LSRR: 1119 case IPOPT_SSRR: 1120 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1121 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1122 goto bad; 1123 } 1124 ipaddr.sin_addr = ip->ip_dst; 1125 ia = (struct in_ifaddr *) 1126 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1127 if (ia == 0) { 1128 if (opt == IPOPT_SSRR) { 1129 type = ICMP_UNREACH; 1130 code = ICMP_UNREACH_SRCFAIL; 1131 goto bad; 1132 } 1133 if (!ip_dosourceroute) 1134 goto nosourcerouting; 1135 /* 1136 * Loose routing, and not at next destination 1137 * yet; nothing to do except forward. 1138 */ 1139 break; 1140 } 1141 off--; /* 0 origin */ 1142 if (off > optlen - (int)sizeof(struct in_addr)) { 1143 /* 1144 * End of source route. Should be for us. 1145 */ 1146 if (!ip_acceptsourceroute) 1147 goto nosourcerouting; 1148 save_rte(cp, ip->ip_src); 1149 break; 1150 } 1151 1152 if (!ip_dosourceroute) { 1153 if (ipforwarding) { 1154 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1155 /* 1156 * Acting as a router, so generate ICMP 1157 */ 1158nosourcerouting: 1159 strcpy(buf, inet_ntoa(ip->ip_dst)); 1160 log(LOG_WARNING, 1161 "attempted source route from %s to %s\n", 1162 inet_ntoa(ip->ip_src), buf); 1163 type = ICMP_UNREACH; 1164 code = ICMP_UNREACH_SRCFAIL; 1165 goto bad; 1166 } else { 1167 /* 1168 * Not acting as a router, so silently drop. 1169 */ 1170 ipstat.ips_cantforward++; 1171 m_freem(m); 1172 return (1); 1173 } 1174 } 1175 1176 /* 1177 * locate outgoing interface 1178 */ 1179 (void)memcpy(&ipaddr.sin_addr, cp + off, 1180 sizeof(ipaddr.sin_addr)); 1181 1182 if (opt == IPOPT_SSRR) { 1183#define INA struct in_ifaddr * 1184#define SA struct sockaddr * 1185 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1186 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1187 } else 1188 ia = ip_rtaddr(ipaddr.sin_addr); 1189 if (ia == 0) { 1190 type = ICMP_UNREACH; 1191 code = ICMP_UNREACH_SRCFAIL; 1192 goto bad; 1193 } 1194 ip->ip_dst = ipaddr.sin_addr; 1195 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1196 sizeof(struct in_addr)); 1197 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1198 /* 1199 * Let ip_intr's mcast routing check handle mcast pkts 1200 */ 1201 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1202 break; 1203 1204 case IPOPT_RR: 1205 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1206 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1207 goto bad; 1208 } 1209 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1210 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1211 goto bad; 1212 } 1213 /* 1214 * If no space remains, ignore. 1215 */ 1216 off--; /* 0 origin */ 1217 if (off > optlen - (int)sizeof(struct in_addr)) 1218 break; 1219 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1220 sizeof(ipaddr.sin_addr)); 1221 /* 1222 * locate outgoing interface; if we're the destination, 1223 * use the incoming interface (should be same). 1224 */ 1225 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1226 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1227 type = ICMP_UNREACH; 1228 code = ICMP_UNREACH_HOST; 1229 goto bad; 1230 } 1231 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1232 sizeof(struct in_addr)); 1233 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1234 break; 1235 1236 case IPOPT_TS: 1237 code = cp - (u_char *)ip; 1238 ipt = (struct ip_timestamp *)cp; 1239 if (ipt->ipt_len < 5) 1240 goto bad; 1241 if (ipt->ipt_ptr > 1242 ipt->ipt_len - (int)sizeof(int32_t)) { 1243 if (++ipt->ipt_oflw == 0) 1244 goto bad; 1245 break; 1246 } 1247 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1248 switch (ipt->ipt_flg) { 1249 1250 case IPOPT_TS_TSONLY: 1251 break; 1252 1253 case IPOPT_TS_TSANDADDR: 1254 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1255 sizeof(struct in_addr) > ipt->ipt_len) 1256 goto bad; 1257 ipaddr.sin_addr = dst; 1258 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1259 m->m_pkthdr.rcvif); 1260 if (ia == 0) 1261 continue; 1262 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1263 sizeof(struct in_addr)); 1264 ipt->ipt_ptr += sizeof(struct in_addr); 1265 break; 1266 1267 case IPOPT_TS_PRESPEC: 1268 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1269 sizeof(struct in_addr) > ipt->ipt_len) 1270 goto bad; 1271 (void)memcpy(&ipaddr.sin_addr, sin, 1272 sizeof(struct in_addr)); 1273 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1274 continue; 1275 ipt->ipt_ptr += sizeof(struct in_addr); 1276 break; 1277 1278 default: 1279 goto bad; 1280 } 1281 ntime = iptime(); 1282 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1283 sizeof(n_time)); 1284 ipt->ipt_ptr += sizeof(n_time); 1285 } 1286 } 1287 if (forward && ipforwarding) { 1288 ip_forward(m, 1); 1289 return (1); 1290 } 1291 return (0); 1292bad: 1293 ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2; /* XXX icmp_error adds in hdr length */ 1294 icmp_error(m, type, code, 0, 0); 1295 ipstat.ips_badoptions++; 1296 return (1); 1297} 1298 1299/* 1300 * Given address of next destination (final or next hop), 1301 * return internet address info of interface to be used to get there. 1302 */ 1303static struct in_ifaddr * 1304ip_rtaddr(dst) 1305 struct in_addr dst; 1306{ 1307 register struct sockaddr_in *sin; 1308 1309 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1310 1311 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1312 if (ipforward_rt.ro_rt) { 1313 RTFREE(ipforward_rt.ro_rt); 1314 ipforward_rt.ro_rt = 0; 1315 } 1316 sin->sin_family = AF_INET; 1317 sin->sin_len = sizeof(*sin); 1318 sin->sin_addr = dst; 1319 1320 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1321 } 1322 if (ipforward_rt.ro_rt == 0) 1323 return ((struct in_ifaddr *)0); 1324 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1325} 1326 1327/* 1328 * Save incoming source route for use in replies, 1329 * to be picked up later by ip_srcroute if the receiver is interested. 1330 */ 1331void 1332save_rte(option, dst) 1333 u_char *option; 1334 struct in_addr dst; 1335{ 1336 unsigned olen; 1337 1338 olen = option[IPOPT_OLEN]; 1339#ifdef DIAGNOSTIC 1340 if (ipprintfs) 1341 printf("save_rte: olen %d\n", olen); 1342#endif 1343 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1344 return; 1345 bcopy(option, ip_srcrt.srcopt, olen); 1346 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1347 ip_srcrt.dst = dst; 1348} 1349 1350/* 1351 * Retrieve incoming source route for use in replies, 1352 * in the same form used by setsockopt. 1353 * The first hop is placed before the options, will be removed later. 1354 */ 1355struct mbuf * 1356ip_srcroute() 1357{ 1358 register struct in_addr *p, *q; 1359 register struct mbuf *m; 1360 1361 if (ip_nhops == 0) 1362 return ((struct mbuf *)0); 1363 m = m_get(M_DONTWAIT, MT_HEADER); 1364 if (m == 0) 1365 return ((struct mbuf *)0); 1366 1367#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1368 1369 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1370 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1371 OPTSIZ; 1372#ifdef DIAGNOSTIC 1373 if (ipprintfs) 1374 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1375#endif 1376 1377 /* 1378 * First save first hop for return route 1379 */ 1380 p = &ip_srcrt.route[ip_nhops - 1]; 1381 *(mtod(m, struct in_addr *)) = *p--; 1382#ifdef DIAGNOSTIC 1383 if (ipprintfs) 1384 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1385#endif 1386 1387 /* 1388 * Copy option fields and padding (nop) to mbuf. 1389 */ 1390 ip_srcrt.nop = IPOPT_NOP; 1391 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1392 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1393 &ip_srcrt.nop, OPTSIZ); 1394 q = (struct in_addr *)(mtod(m, caddr_t) + 1395 sizeof(struct in_addr) + OPTSIZ); 1396#undef OPTSIZ 1397 /* 1398 * Record return path as an IP source route, 1399 * reversing the path (pointers are now aligned). 1400 */ 1401 while (p >= ip_srcrt.route) { 1402#ifdef DIAGNOSTIC 1403 if (ipprintfs) 1404 printf(" %lx", (u_long)ntohl(q->s_addr)); 1405#endif 1406 *q++ = *p--; 1407 } 1408 /* 1409 * Last hop goes to final destination. 1410 */ 1411 *q = ip_srcrt.dst; 1412#ifdef DIAGNOSTIC 1413 if (ipprintfs) 1414 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1415#endif 1416 return (m); 1417} 1418 1419/* 1420 * Strip out IP options, at higher 1421 * level protocol in the kernel. 1422 * Second argument is buffer to which options 1423 * will be moved, and return value is their length. 1424 * XXX should be deleted; last arg currently ignored. 1425 */ 1426void 1427ip_stripoptions(m, mopt) 1428 register struct mbuf *m; 1429 struct mbuf *mopt; 1430{ 1431 register int i; 1432 struct ip *ip = mtod(m, struct ip *); 1433 register caddr_t opts; 1434 int olen; 1435 1436 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1437 opts = (caddr_t)(ip + 1); 1438 i = m->m_len - (sizeof (struct ip) + olen); 1439 bcopy(opts + olen, opts, (unsigned)i); 1440 m->m_len -= olen; 1441 if (m->m_flags & M_PKTHDR) 1442 m->m_pkthdr.len -= olen; 1443 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1444} 1445 1446u_char inetctlerrmap[PRC_NCMDS] = { 1447 0, 0, 0, 0, 1448 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1449 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1450 EMSGSIZE, EHOSTUNREACH, 0, 0, 1451 0, 0, 0, 0, 1452 ENOPROTOOPT 1453}; 1454 1455/* 1456 * Forward a packet. If some error occurs return the sender 1457 * an icmp packet. Note we can't always generate a meaningful 1458 * icmp message because icmp doesn't have a large enough repertoire 1459 * of codes and types. 1460 * 1461 * If not forwarding, just drop the packet. This could be confusing 1462 * if ipforwarding was zero but some routing protocol was advancing 1463 * us as a gateway to somewhere. However, we must let the routing 1464 * protocol deal with that. 1465 * 1466 * The srcrt parameter indicates whether the packet is being forwarded 1467 * via a source route. 1468 */ 1469static void 1470ip_forward(m, srcrt) 1471 struct mbuf *m; 1472 int srcrt; 1473{ 1474 register struct ip *ip = mtod(m, struct ip *); 1475 register struct sockaddr_in *sin; 1476 register struct rtentry *rt; 1477 int error, type = 0, code = 0; 1478 struct mbuf *mcopy; 1479 n_long dest; 1480 struct ifnet *destifp; 1481#ifdef IPSEC 1482 struct ifnet dummyifp; 1483#endif 1484 1485 dest = 0; 1486#ifdef DIAGNOSTIC 1487 if (ipprintfs) 1488 printf("forward: src %lx dst %lx ttl %x\n", 1489 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1490 ip->ip_ttl); 1491#endif 1492 1493 1494 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1495 ipstat.ips_cantforward++; 1496 m_freem(m); 1497 return; 1498 } 1499 HTONS(ip->ip_id); 1500#ifdef IPSTEALTH 1501 if (!ipstealth) { 1502#endif 1503 if (ip->ip_ttl <= IPTTLDEC) { 1504 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1505 dest, 0); 1506 return; 1507 } 1508 ip->ip_ttl -= IPTTLDEC; 1509#ifdef IPSTEALTH 1510 } 1511#endif 1512 1513 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1514 if ((rt = ipforward_rt.ro_rt) == 0 || 1515 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1516 if (ipforward_rt.ro_rt) { 1517 RTFREE(ipforward_rt.ro_rt); 1518 ipforward_rt.ro_rt = 0; 1519 } 1520 sin->sin_family = AF_INET; 1521 sin->sin_len = sizeof(*sin); 1522 sin->sin_addr = ip->ip_dst; 1523 1524 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1525 if (ipforward_rt.ro_rt == 0) { 1526 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1527 return; 1528 } 1529 rt = ipforward_rt.ro_rt; 1530 } 1531 1532 /* 1533 * Save at most 64 bytes of the packet in case 1534 * we need to generate an ICMP message to the src. 1535 */ 1536 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1537 1538 /* 1539 * If forwarding packet using same interface that it came in on, 1540 * perhaps should send a redirect to sender to shortcut a hop. 1541 * Only send redirect if source is sending directly to us, 1542 * and if packet was not source routed (or has any options). 1543 * Also, don't send redirect if forwarding using a default route 1544 * or a route modified by a redirect. 1545 */ 1546#define satosin(sa) ((struct sockaddr_in *)(sa)) 1547 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1548 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1549 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1550 ipsendredirects && !srcrt) { 1551#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1552 u_long src = ntohl(ip->ip_src.s_addr); 1553 1554 if (RTA(rt) && 1555 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1556 if (rt->rt_flags & RTF_GATEWAY) 1557 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1558 else 1559 dest = ip->ip_dst.s_addr; 1560 /* Router requirements says to only send host redirects */ 1561 type = ICMP_REDIRECT; 1562 code = ICMP_REDIRECT_HOST; 1563#ifdef DIAGNOSTIC 1564 if (ipprintfs) 1565 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1566#endif 1567 } 1568 } 1569 1570 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1571 IP_FORWARDING, 0); 1572 if (error) 1573 ipstat.ips_cantforward++; 1574 else { 1575 ipstat.ips_forward++; 1576 if (type) 1577 ipstat.ips_redirectsent++; 1578 else { 1579 if (mcopy) { 1580 ipflow_create(&ipforward_rt, mcopy); 1581 m_freem(mcopy); 1582 } 1583 return; 1584 } 1585 } 1586 if (mcopy == NULL) 1587 return; 1588 destifp = NULL; 1589 1590 switch (error) { 1591 1592 case 0: /* forwarded, but need redirect */ 1593 /* type, code set above */ 1594 break; 1595 1596 case ENETUNREACH: /* shouldn't happen, checked above */ 1597 case EHOSTUNREACH: 1598 case ENETDOWN: 1599 case EHOSTDOWN: 1600 default: 1601 type = ICMP_UNREACH; 1602 code = ICMP_UNREACH_HOST; 1603 break; 1604 1605 case EMSGSIZE: 1606 type = ICMP_UNREACH; 1607 code = ICMP_UNREACH_NEEDFRAG; 1608#ifndef IPSEC 1609 if (ipforward_rt.ro_rt) 1610 destifp = ipforward_rt.ro_rt->rt_ifp; 1611#else 1612 /* 1613 * If the packet is routed over IPsec tunnel, tell the 1614 * originator the tunnel MTU. 1615 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1616 * XXX quickhack!!! 1617 */ 1618 if (ipforward_rt.ro_rt) { 1619 struct secpolicy *sp = NULL; 1620 int ipsecerror; 1621 int ipsechdr; 1622 struct route *ro; 1623 1624 sp = ipsec4_getpolicybyaddr(mcopy, 1625 IPSEC_DIR_OUTBOUND, 1626 IP_FORWARDING, 1627 &ipsecerror); 1628 1629 if (sp == NULL) 1630 destifp = ipforward_rt.ro_rt->rt_ifp; 1631 else { 1632 /* count IPsec header size */ 1633 ipsechdr = ipsec4_hdrsiz(mcopy, 1634 IPSEC_DIR_OUTBOUND, 1635 NULL); 1636 1637 /* 1638 * find the correct route for outer IPv4 1639 * header, compute tunnel MTU. 1640 * 1641 * XXX BUG ALERT 1642 * The "dummyifp" code relies upon the fact 1643 * that icmp_error() touches only ifp->if_mtu. 1644 */ 1645 /*XXX*/ 1646 destifp = NULL; 1647 if (sp->req != NULL 1648 && sp->req->sav != NULL 1649 && sp->req->sav->sah != NULL) { 1650 ro = &sp->req->sav->sah->sa_route; 1651 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1652 dummyifp.if_mtu = 1653 ro->ro_rt->rt_ifp->if_mtu; 1654 dummyifp.if_mtu -= ipsechdr; 1655 destifp = &dummyifp; 1656 } 1657 } 1658 1659 key_freesp(sp); 1660 } 1661 } 1662#endif /*IPSEC*/ 1663 ipstat.ips_cantfrag++; 1664 break; 1665 1666 case ENOBUFS: 1667 type = ICMP_SOURCEQUENCH; 1668 code = 0; 1669 break; 1670 1671 case EACCES: /* ipfw denied packet */ 1672 m_freem(mcopy); 1673 return; 1674 } 1675 icmp_error(mcopy, type, code, dest, destifp); 1676} 1677 1678void 1679ip_savecontrol(inp, mp, ip, m) 1680 register struct inpcb *inp; 1681 register struct mbuf **mp; 1682 register struct ip *ip; 1683 register struct mbuf *m; 1684{ 1685 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1686 struct timeval tv; 1687 1688 microtime(&tv); 1689 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1690 SCM_TIMESTAMP, SOL_SOCKET); 1691 if (*mp) 1692 mp = &(*mp)->m_next; 1693 } 1694 if (inp->inp_flags & INP_RECVDSTADDR) { 1695 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1696 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1697 if (*mp) 1698 mp = &(*mp)->m_next; 1699 } 1700#ifdef notyet 1701 /* XXX 1702 * Moving these out of udp_input() made them even more broken 1703 * than they already were. 1704 */ 1705 /* options were tossed already */ 1706 if (inp->inp_flags & INP_RECVOPTS) { 1707 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1708 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1709 if (*mp) 1710 mp = &(*mp)->m_next; 1711 } 1712 /* ip_srcroute doesn't do what we want here, need to fix */ 1713 if (inp->inp_flags & INP_RECVRETOPTS) { 1714 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1715 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1716 if (*mp) 1717 mp = &(*mp)->m_next; 1718 } 1719#endif 1720 if (inp->inp_flags & INP_RECVIF) { 1721 struct ifnet *ifp; 1722 struct sdlbuf { 1723 struct sockaddr_dl sdl; 1724 u_char pad[32]; 1725 } sdlbuf; 1726 struct sockaddr_dl *sdp; 1727 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1728 1729 if (((ifp = m->m_pkthdr.rcvif)) 1730 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1731 sdp = (struct sockaddr_dl *)(ifnet_addrs 1732 [ifp->if_index - 1]->ifa_addr); 1733 /* 1734 * Change our mind and don't try copy. 1735 */ 1736 if ((sdp->sdl_family != AF_LINK) 1737 || (sdp->sdl_len > sizeof(sdlbuf))) { 1738 goto makedummy; 1739 } 1740 bcopy(sdp, sdl2, sdp->sdl_len); 1741 } else { 1742makedummy: 1743 sdl2->sdl_len 1744 = offsetof(struct sockaddr_dl, sdl_data[0]); 1745 sdl2->sdl_family = AF_LINK; 1746 sdl2->sdl_index = 0; 1747 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1748 } 1749 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1750 IP_RECVIF, IPPROTO_IP); 1751 if (*mp) 1752 mp = &(*mp)->m_next; 1753 } 1754} 1755 1756int 1757ip_rsvp_init(struct socket *so) 1758{ 1759 if (so->so_type != SOCK_RAW || 1760 so->so_proto->pr_protocol != IPPROTO_RSVP) 1761 return EOPNOTSUPP; 1762 1763 if (ip_rsvpd != NULL) 1764 return EADDRINUSE; 1765 1766 ip_rsvpd = so; 1767 /* 1768 * This may seem silly, but we need to be sure we don't over-increment 1769 * the RSVP counter, in case something slips up. 1770 */ 1771 if (!ip_rsvp_on) { 1772 ip_rsvp_on = 1; 1773 rsvp_on++; 1774 } 1775 1776 return 0; 1777} 1778 1779int 1780ip_rsvp_done(void) 1781{ 1782 ip_rsvpd = NULL; 1783 /* 1784 * This may seem silly, but we need to be sure we don't over-decrement 1785 * the RSVP counter, in case something slips up. 1786 */ 1787 if (ip_rsvp_on) { 1788 ip_rsvp_on = 0; 1789 rsvp_on--; 1790 } 1791 return 0; 1792} 1793