ip_input.c revision 41793
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 * $Id: ip_input.c,v 1.101 1998/09/10 08:56:40 dfr Exp $ 35 * $ANA: ip_input.c,v 1.5 1996/09/18 14:34:59 wollman Exp $ 36 */ 37 38#define _IP_VHL 39 40#include "opt_bootp.h" 41#include "opt_ipfw.h" 42#include "opt_ipdn.h" 43#include "opt_ipdivert.h" 44#include "opt_ipfilter.h" 45 46#include <stddef.h> 47 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/mbuf.h> 51#include <sys/malloc.h> 52#include <sys/domain.h> 53#include <sys/protosw.h> 54#include <sys/socket.h> 55#include <sys/time.h> 56#include <sys/kernel.h> 57#include <sys/syslog.h> 58#include <sys/sysctl.h> 59 60#include <net/if.h> 61#include <net/if_var.h> 62#include <net/if_dl.h> 63#include <net/route.h> 64#include <net/netisr.h> 65 66#include <netinet/in.h> 67#include <netinet/in_systm.h> 68#include <netinet/in_var.h> 69#include <netinet/ip.h> 70#include <netinet/in_pcb.h> 71#include <netinet/ip_var.h> 72#include <netinet/ip_icmp.h> 73#include <machine/in_cksum.h> 74 75#include <sys/socketvar.h> 76 77#ifdef IPFIREWALL 78#include <netinet/ip_fw.h> 79#endif 80 81#ifdef DUMMYNET 82#include <netinet/ip_dummynet.h> 83#endif 84 85int rsvp_on = 0; 86static int ip_rsvp_on; 87struct socket *ip_rsvpd; 88 89int ipforwarding = 0; 90SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 91 &ipforwarding, 0, ""); 92 93static int ipsendredirects = 1; /* XXX */ 94SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 95 &ipsendredirects, 0, ""); 96 97int ip_defttl = IPDEFTTL; 98SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 99 &ip_defttl, 0, ""); 100 101static int ip_dosourceroute = 0; 102SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 103 &ip_dosourceroute, 0, ""); 104 105static int ip_acceptsourceroute = 0; 106SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 107 CTLFLAG_RW, &ip_acceptsourceroute, 0, ""); 108#ifdef DIAGNOSTIC 109static int ipprintfs = 0; 110#endif 111 112extern struct domain inetdomain; 113extern struct protosw inetsw[]; 114u_char ip_protox[IPPROTO_MAX]; 115static int ipqmaxlen = IFQ_MAXLEN; 116struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 117struct ifqueue ipintrq; 118SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RD, 119 &ipintrq.ifq_maxlen, 0, ""); 120SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 121 &ipintrq.ifq_drops, 0, ""); 122 123struct ipstat ipstat; 124SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD, 125 &ipstat, ipstat, ""); 126 127/* Packet reassembly stuff */ 128#define IPREASS_NHASH_LOG2 6 129#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 130#define IPREASS_HMASK (IPREASS_NHASH - 1) 131#define IPREASS_HASH(x,y) \ 132 ((((x) & 0xF | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 133 134static struct ipq ipq[IPREASS_NHASH]; 135static int nipq = 0; /* total # of reass queues */ 136static int maxnipq; 137 138#ifdef IPCTL_DEFMTU 139SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 140 &ip_mtu, 0, ""); 141#endif 142 143#if !defined(COMPAT_IPFW) || COMPAT_IPFW == 1 144#undef COMPAT_IPFW 145#define COMPAT_IPFW 1 146#else 147#undef COMPAT_IPFW 148#endif 149 150#ifdef COMPAT_IPFW 151 152#include <netinet/ip_fw.h> 153 154/* Firewall hooks */ 155ip_fw_chk_t *ip_fw_chk_ptr; 156ip_fw_ctl_t *ip_fw_ctl_ptr; 157 158#ifdef DUMMYNET 159ip_dn_ctl_t *ip_dn_ctl_ptr; 160#endif 161 162/* IP Network Address Translation (NAT) hooks */ 163ip_nat_t *ip_nat_ptr; 164ip_nat_ctl_t *ip_nat_ctl_ptr; 165#endif 166 167#if defined(IPFILTER_LKM) || defined(IPFILTER) 168int iplattach __P((void)); 169int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL; 170#endif 171 172 173/* 174 * We need to save the IP options in case a protocol wants to respond 175 * to an incoming packet over the same route if the packet got here 176 * using IP source routing. This allows connection establishment and 177 * maintenance when the remote end is on a network that is not known 178 * to us. 179 */ 180static int ip_nhops = 0; 181static struct ip_srcrt { 182 struct in_addr dst; /* final destination */ 183 char nop; /* one NOP to align */ 184 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 185 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 186} ip_srcrt; 187 188#ifdef IPDIVERT 189/* 190 * Shared variable between ip_input() and ip_reass() to communicate 191 * about which packets, once assembled from fragments, get diverted, 192 * and to which port. 193 */ 194static u_short frag_divert_port; 195#endif 196 197struct sockaddr_in *ip_fw_fwd_addr; 198 199static void save_rte __P((u_char *, struct in_addr)); 200static int ip_dooptions __P((struct mbuf *)); 201static void ip_forward __P((struct mbuf *, int)); 202static void ip_freef __P((struct ipq *)); 203static struct ip * 204 ip_reass __P((struct mbuf *, struct ipq *, struct ipq *)); 205static struct in_ifaddr * 206 ip_rtaddr __P((struct in_addr)); 207static void ipintr __P((void)); 208/* 209 * IP initialization: fill in IP protocol switch table. 210 * All protocols not implemented in kernel go to raw IP protocol handler. 211 */ 212void 213ip_init() 214{ 215 register struct protosw *pr; 216 register int i; 217 218 TAILQ_INIT(&in_ifaddrhead); 219 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 220 if (pr == 0) 221 panic("ip_init"); 222 for (i = 0; i < IPPROTO_MAX; i++) 223 ip_protox[i] = pr - inetsw; 224 for (pr = inetdomain.dom_protosw; 225 pr < inetdomain.dom_protoswNPROTOSW; pr++) 226 if (pr->pr_domain->dom_family == PF_INET && 227 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 228 ip_protox[pr->pr_protocol] = pr - inetsw; 229 230 for (i = 0; i < IPREASS_NHASH; i++) 231 ipq[i].next = ipq[i].prev = &ipq[i]; 232 233 maxnipq = nmbclusters/4; 234 235 ip_id = time_second & 0xffff; 236 ipintrq.ifq_maxlen = ipqmaxlen; 237#ifdef IPFIREWALL 238 ip_fw_init(); 239#endif 240#ifdef DUMMYNET 241 ip_dn_init(); 242#endif 243#ifdef IPNAT 244 ip_nat_init(); 245#endif 246#ifdef IPFILTER 247 iplattach(); 248#endif 249 250} 251 252static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 253static struct route ipforward_rt; 254 255/* 256 * Ip input routine. Checksum and byte swap header. If fragmented 257 * try to reassemble. Process options. Pass to next level. 258 */ 259void 260ip_input(struct mbuf *m) 261{ 262 struct ip *ip; 263 struct ipq *fp; 264 struct ipqent *ipqe; 265 struct in_ifaddr *ia; 266 int i, hlen, mff; 267 u_short sum; 268#ifndef IPDIVERT /* dummy variable for the firewall code to play with */ 269 u_short ip_divert_cookie = 0 ; 270#endif 271#ifdef COMPAT_IPFW 272 struct ip_fw_chain *rule = NULL ; 273#endif 274 275#if defined(IPFIREWALL) && defined(DUMMYNET) 276 /* 277 * dummynet packet are prepended a vestigial mbuf with 278 * m_type = MT_DUMMYNET and m_data pointing to the matching 279 * rule. 280 */ 281 if (m->m_type == MT_DUMMYNET) { 282 struct mbuf *m0 = m ; 283 rule = (struct ip_fw_chain *)(m->m_data) ; 284 m = m->m_next ; 285 free(m0, M_IPFW); 286 ip = mtod(m, struct ip *); 287 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 288 goto iphack ; 289 } else 290 rule = NULL ; 291#endif 292 293#ifdef DIAGNOSTIC 294 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 295 panic("ip_input no HDR"); 296#endif 297 /* 298 * If no IP addresses have been set yet but the interfaces 299 * are receiving, can't do anything with incoming packets yet. 300 * XXX This is broken! We should be able to receive broadcasts 301 * and multicasts even without any local addresses configured. 302 */ 303 if (TAILQ_EMPTY(&in_ifaddrhead)) 304 goto bad; 305 ipstat.ips_total++; 306 307 if (m->m_pkthdr.len < sizeof(struct ip)) 308 goto tooshort; 309 310 if (m->m_len < sizeof (struct ip) && 311 (m = m_pullup(m, sizeof (struct ip))) == 0) { 312 ipstat.ips_toosmall++; 313 return; 314 } 315 ip = mtod(m, struct ip *); 316 317 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 318 ipstat.ips_badvers++; 319 goto bad; 320 } 321 322 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 323 if (hlen < sizeof(struct ip)) { /* minimum header length */ 324 ipstat.ips_badhlen++; 325 goto bad; 326 } 327 if (hlen > m->m_len) { 328 if ((m = m_pullup(m, hlen)) == 0) { 329 ipstat.ips_badhlen++; 330 return; 331 } 332 ip = mtod(m, struct ip *); 333 } 334 if (hlen == sizeof(struct ip)) { 335 sum = in_cksum_hdr(ip); 336 } else { 337 sum = in_cksum(m, hlen); 338 } 339 if (sum) { 340 ipstat.ips_badsum++; 341 goto bad; 342 } 343 344 /* 345 * Convert fields to host representation. 346 */ 347 NTOHS(ip->ip_len); 348 if (ip->ip_len < hlen) { 349 ipstat.ips_badlen++; 350 goto bad; 351 } 352 NTOHS(ip->ip_id); 353 NTOHS(ip->ip_off); 354 355 /* 356 * Check that the amount of data in the buffers 357 * is as at least much as the IP header would have us expect. 358 * Trim mbufs if longer than we expect. 359 * Drop packet if shorter than we expect. 360 */ 361 if (m->m_pkthdr.len < ip->ip_len) { 362tooshort: 363 ipstat.ips_tooshort++; 364 goto bad; 365 } 366 if (m->m_pkthdr.len > ip->ip_len) { 367 if (m->m_len == m->m_pkthdr.len) { 368 m->m_len = ip->ip_len; 369 m->m_pkthdr.len = ip->ip_len; 370 } else 371 m_adj(m, ip->ip_len - m->m_pkthdr.len); 372 } 373 /* 374 * IpHack's section. 375 * Right now when no processing on packet has done 376 * and it is still fresh out of network we do our black 377 * deals with it. 378 * - Firewall: deny/allow/divert 379 * - Xlate: translate packet's addr/port (NAT). 380 * - Pipe: pass pkt through dummynet. 381 * - Wrap: fake packet's addr/port <unimpl.> 382 * - Encapsulate: put it in another IP and send out. <unimp.> 383 */ 384 385iphack: 386#if defined(IPFILTER) || defined(IPFILTER_LKM) 387 /* 388 * Check if we want to allow this packet to be processed. 389 * Consider it to be bad if not. 390 */ 391 if (fr_checkp) { 392 struct mbuf *m1 = m; 393 394 if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1) 395 return; 396 ip = mtod(m = m1, struct ip *); 397 } 398#endif 399#ifdef COMPAT_IPFW 400 if (ip_fw_chk_ptr) { 401#ifdef IPFIREWALL_FORWARD 402 /* 403 * If we've been forwarded from the output side, then 404 * skip the firewall a second time 405 */ 406 if (ip_fw_fwd_addr) 407 goto ours; 408#endif /* IPFIREWALL_FORWARD */ 409 i = (*ip_fw_chk_ptr)(&ip, hlen, NULL, &ip_divert_cookie, 410 &m, &rule, &ip_fw_fwd_addr); 411 /* 412 * see the comment in ip_output for the return values 413 * produced by the firewall. 414 */ 415 if (!m) /* packet discarded by firewall */ 416 return ; 417 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 418 goto pass ; 419#ifdef DUMMYNET 420 if (i & 0x10000) { 421 /* send packet to the appropriate pipe */ 422 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule); 423 return ; 424 } 425#endif 426#ifdef IPDIVERT 427 if (i > 0 && i < 0x10000) { 428 /* Divert packet */ 429 frag_divert_port = i & 0xffff ; 430 goto ours; 431 } 432#endif 433#ifdef IPFIREWALL_FORWARD 434 if (i == 0 && ip_fw_fwd_addr != NULL) 435 goto pass ; 436#endif 437 /* 438 * if we get here, the packet must be dropped 439 */ 440 m_freem(m); 441 return; 442 } 443pass: 444 445 if (ip_nat_ptr && !(*ip_nat_ptr)(&ip, &m, m->m_pkthdr.rcvif, IP_NAT_IN)) 446 return; 447#endif /* !COMPAT_IPFW */ 448 449 /* 450 * Process options and, if not destined for us, 451 * ship it on. ip_dooptions returns 1 when an 452 * error was detected (causing an icmp message 453 * to be sent and the original packet to be freed). 454 */ 455 ip_nhops = 0; /* for source routed packets */ 456 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 457 return; 458 459 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 460 * matter if it is destined to another node, or whether it is 461 * a multicast one, RSVP wants it! and prevents it from being forwarded 462 * anywhere else. Also checks if the rsvp daemon is running before 463 * grabbing the packet. 464 */ 465 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 466 goto ours; 467 468 /* 469 * Check our list of addresses, to see if the packet is for us. 470 */ 471 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 472 ia = TAILQ_NEXT(ia, ia_link)) { 473#define satosin(sa) ((struct sockaddr_in *)(sa)) 474 475 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 476 goto ours; 477#ifdef BOOTP_COMPAT 478 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 479 goto ours; 480#endif 481#ifdef IPFIREWALL_FORWARD 482 /* 483 * If the addr to forward to is one of ours, we pretend to 484 * be the destination for this packet. 485 */ 486 if (ip_fw_fwd_addr != NULL && 487 IA_SIN(ia)->sin_addr.s_addr == 488 ip_fw_fwd_addr->sin_addr.s_addr) 489 goto ours; 490#endif 491 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 492 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 493 ip->ip_dst.s_addr) 494 goto ours; 495 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 496 goto ours; 497 } 498 } 499 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 500 struct in_multi *inm; 501 if (ip_mrouter) { 502 /* 503 * If we are acting as a multicast router, all 504 * incoming multicast packets are passed to the 505 * kernel-level multicast forwarding function. 506 * The packet is returned (relatively) intact; if 507 * ip_mforward() returns a non-zero value, the packet 508 * must be discarded, else it may be accepted below. 509 * 510 * (The IP ident field is put in the same byte order 511 * as expected when ip_mforward() is called from 512 * ip_output().) 513 */ 514 ip->ip_id = htons(ip->ip_id); 515 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 516 ipstat.ips_cantforward++; 517 m_freem(m); 518 return; 519 } 520 ip->ip_id = ntohs(ip->ip_id); 521 522 /* 523 * The process-level routing demon needs to receive 524 * all multicast IGMP packets, whether or not this 525 * host belongs to their destination groups. 526 */ 527 if (ip->ip_p == IPPROTO_IGMP) 528 goto ours; 529 ipstat.ips_forward++; 530 } 531 /* 532 * See if we belong to the destination multicast group on the 533 * arrival interface. 534 */ 535 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 536 if (inm == NULL) { 537 ipstat.ips_notmember++; 538 m_freem(m); 539 return; 540 } 541 goto ours; 542 } 543 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 544 goto ours; 545 if (ip->ip_dst.s_addr == INADDR_ANY) 546 goto ours; 547 548 /* 549 * Not for us; forward if possible and desirable. 550 */ 551 if (ipforwarding == 0) { 552 ipstat.ips_cantforward++; 553 m_freem(m); 554 } else 555 ip_forward(m, 0); 556 return; 557 558ours: 559 560 /* 561 * If offset or IP_MF are set, must reassemble. 562 * Otherwise, nothing need be done. 563 * (We could look in the reassembly queue to see 564 * if the packet was previously fragmented, 565 * but it's not worth the time; just let them time out.) 566 */ 567 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) { 568 if (m->m_flags & M_EXT) { /* XXX */ 569 if ((m = m_pullup(m, sizeof (struct ip))) == 0) { 570 ipstat.ips_toosmall++; 571#ifdef IPDIVERT 572 frag_divert_port = 0; 573 ip_divert_cookie = 0; 574#endif 575 return; 576 } 577 ip = mtod(m, struct ip *); 578 } 579 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 580 /* 581 * Look for queue of fragments 582 * of this datagram. 583 */ 584 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 585 if (ip->ip_id == fp->ipq_id && 586 ip->ip_src.s_addr == fp->ipq_src.s_addr && 587 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 588 ip->ip_p == fp->ipq_p) 589 goto found; 590 591 fp = 0; 592 593 /* check if there's a place for the new queue */ 594 if (nipq > maxnipq) { 595 /* 596 * drop something from the tail of the current queue 597 * before proceeding further 598 */ 599 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 600 for (i = 0; i < IPREASS_NHASH; i++) { 601 if (ipq[i].prev != &ipq[i]) { 602 ip_freef(ipq[i].prev); 603 break; 604 } 605 } 606 } else 607 ip_freef(ipq[sum].prev); 608 } 609found: 610 /* 611 * Adjust ip_len to not reflect header, 612 * set ip_mff if more fragments are expected, 613 * convert offset of this to bytes. 614 */ 615 ip->ip_len -= hlen; 616 mff = (ip->ip_off & IP_MF) != 0; 617 if (mff) { 618 /* 619 * Make sure that fragments have a data length 620 * that's a non-zero multiple of 8 bytes. 621 */ 622 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 623 ipstat.ips_toosmall++; /* XXX */ 624 goto bad; 625 } 626 m->m_flags |= M_FRAG; 627 } 628 ip->ip_off <<= 3; 629 630 /* 631 * If datagram marked as having more fragments 632 * or if this is not the first fragment, 633 * attempt reassembly; if it succeeds, proceed. 634 */ 635 if (mff || ip->ip_off) { 636 ipstat.ips_fragments++; 637 m->m_pkthdr.header = ip; 638 ip = ip_reass(m, fp, &ipq[sum]); 639 if (ip == 0) { 640#ifdef IPFIREWALL_FORWARD 641 ip_fw_fwd_addr = NULL; 642#endif 643 return; 644 } 645 /* Get the length of the reassembled packets header */ 646 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 647 ipstat.ips_reassembled++; 648 m = dtom(ip); 649#ifdef IPDIVERT 650 if (frag_divert_port) { 651 ip->ip_len += hlen; 652 HTONS(ip->ip_len); 653 HTONS(ip->ip_off); 654 HTONS(ip->ip_id); 655 ip->ip_sum = 0; 656 ip->ip_sum = in_cksum_hdr(ip); 657 NTOHS(ip->ip_id); 658 NTOHS(ip->ip_off); 659 NTOHS(ip->ip_len); 660 ip->ip_len -= hlen; 661 } 662#endif 663 } else 664 if (fp) 665 ip_freef(fp); 666 } else 667 ip->ip_len -= hlen; 668 669#ifdef IPDIVERT 670 /* 671 * Divert reassembled packets to the divert protocol if required 672 * If divert port is null then cookie should be too, 673 * so we shouldn't need to clear them here. Assume ip_divert does so. 674 */ 675 if (frag_divert_port) { 676 ipstat.ips_delivered++; 677 ip_divert_port = frag_divert_port; 678 frag_divert_port = 0; 679 (*inetsw[ip_protox[IPPROTO_DIVERT]].pr_input)(m, hlen); 680 return; 681 } 682 683 /* Don't let packets divert themselves */ 684 if (ip->ip_p == IPPROTO_DIVERT) { 685 ipstat.ips_noproto++; 686 goto bad; 687 } 688 689#endif 690 691 /* 692 * Switch out to protocol's input routine. 693 */ 694 ipstat.ips_delivered++; 695 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 696#ifdef IPFIREWALL_FORWARD 697 ip_fw_fwd_addr = NULL; /* tcp needed it */ 698#endif 699 return; 700bad: 701#ifdef IPFIREWALL_FORWARD 702 ip_fw_fwd_addr = NULL; 703#endif 704 m_freem(m); 705} 706 707/* 708 * IP software interrupt routine - to go away sometime soon 709 */ 710static void 711ipintr(void) 712{ 713 int s; 714 struct mbuf *m; 715 716 while(1) { 717 s = splimp(); 718 IF_DEQUEUE(&ipintrq, m); 719 splx(s); 720 if (m == 0) 721 return; 722 ip_input(m); 723 } 724} 725 726NETISR_SET(NETISR_IP, ipintr); 727 728/* 729 * Take incoming datagram fragment and try to 730 * reassemble it into whole datagram. If a chain for 731 * reassembly of this datagram already exists, then it 732 * is given as fp; otherwise have to make a chain. 733 */ 734static struct ip * 735ip_reass(m, fp, where) 736 register struct mbuf *m; 737 register struct ipq *fp; 738 struct ipq *where; 739{ 740 struct ip *ip = mtod(m, struct ip *); 741 register struct mbuf *p = 0, *q, *nq; 742 struct mbuf *t; 743 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 744 int i, next; 745 746 /* 747 * Presence of header sizes in mbufs 748 * would confuse code below. 749 */ 750 m->m_data += hlen; 751 m->m_len -= hlen; 752 753 /* 754 * If first fragment to arrive, create a reassembly queue. 755 */ 756 if (fp == 0) { 757 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 758 goto dropfrag; 759 fp = mtod(t, struct ipq *); 760 insque(fp, where); 761 nipq++; 762 fp->ipq_ttl = IPFRAGTTL; 763 fp->ipq_p = ip->ip_p; 764 fp->ipq_id = ip->ip_id; 765 fp->ipq_src = ip->ip_src; 766 fp->ipq_dst = ip->ip_dst; 767 fp->ipq_frags = 0; 768#ifdef IPDIVERT 769 fp->ipq_divert = 0; 770 fp->ipq_div_cookie = 0; 771#endif 772 q = 0; 773 goto insert; 774 } 775 776#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 777 778 /* 779 * Find a segment which begins after this one does. 780 */ 781 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 782 if (GETIP(q)->ip_off > ip->ip_off) 783 break; 784 785 /* 786 * If there is a preceding segment, it may provide some of 787 * our data already. If so, drop the data from the incoming 788 * segment. If it provides all of our data, drop us. 789 */ 790 if (p) { 791 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 792 if (i > 0) { 793 if (i >= ip->ip_len) 794 goto dropfrag; 795 m_adj(dtom(ip), i); 796 ip->ip_off += i; 797 ip->ip_len -= i; 798 } 799 } 800 801 /* 802 * While we overlap succeeding segments trim them or, 803 * if they are completely covered, dequeue them. 804 */ 805 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 806 p = q, q = nq) { 807 i = (ip->ip_off + ip->ip_len) - 808 GETIP(q)->ip_off; 809 if (i < GETIP(q)->ip_len) { 810 GETIP(q)->ip_len -= i; 811 GETIP(q)->ip_off += i; 812 m_adj(q, i); 813 break; 814 } 815 nq = q->m_nextpkt; 816 if (p) 817 p->m_nextpkt = nq; 818 else 819 fp->ipq_frags = nq; 820 m_freem(q); 821 } 822 823insert: 824 825#ifdef IPDIVERT 826 /* 827 * Any fragment diverting causes the whole packet to divert 828 */ 829 if (frag_divert_port) { 830 fp->ipq_divert = frag_divert_port; 831 fp->ipq_div_cookie = ip_divert_cookie; 832 } 833 frag_divert_port = 0; 834 ip_divert_cookie = 0; 835#endif 836 837 /* 838 * Stick new segment in its place; 839 * check for complete reassembly. 840 */ 841 if (p == NULL) { 842 m->m_nextpkt = fp->ipq_frags; 843 fp->ipq_frags = m; 844 } else { 845 m->m_nextpkt = p->m_nextpkt; 846 p->m_nextpkt = m; 847 } 848 next = 0; 849 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 850 if (GETIP(q)->ip_off != next) 851 return (0); 852 next += GETIP(q)->ip_len; 853 } 854 /* Make sure the last packet didn't have the IP_MF flag */ 855 if (p->m_flags & M_FRAG) 856 return (0); 857 858 /* 859 * Reassembly is complete. Make sure the packet is a sane size. 860 */ 861 q = fp->ipq_frags; 862 ip = GETIP(q); 863 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 864 ipstat.ips_toolong++; 865 ip_freef(fp); 866 return (0); 867 } 868 869 /* 870 * Concatenate fragments. 871 */ 872 m = q; 873 t = m->m_next; 874 m->m_next = 0; 875 m_cat(m, t); 876 nq = q->m_nextpkt; 877 q->m_nextpkt = 0; 878 for (q = nq; q != NULL; q = nq) { 879 nq = q->m_nextpkt; 880 q->m_nextpkt = NULL; 881 m_cat(m, q); 882 } 883 884#ifdef IPDIVERT 885 /* 886 * extract divert port for packet, if any 887 */ 888 frag_divert_port = fp->ipq_divert; 889 ip_divert_cookie = fp->ipq_div_cookie; 890#endif 891 892 /* 893 * Create header for new ip packet by 894 * modifying header of first packet; 895 * dequeue and discard fragment reassembly header. 896 * Make header visible. 897 */ 898 ip->ip_len = next; 899 ip->ip_src = fp->ipq_src; 900 ip->ip_dst = fp->ipq_dst; 901 remque(fp); 902 nipq--; 903 (void) m_free(dtom(fp)); 904 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 905 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 906 /* some debugging cruft by sklower, below, will go away soon */ 907 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 908 register int plen = 0; 909 for (t = m; m; m = m->m_next) 910 plen += m->m_len; 911 t->m_pkthdr.len = plen; 912 } 913 return (ip); 914 915dropfrag: 916#ifdef IPDIVERT 917 frag_divert_port = 0; 918 ip_divert_cookie = 0; 919#endif 920 ipstat.ips_fragdropped++; 921 m_freem(m); 922 return (0); 923 924#undef GETIP 925} 926 927/* 928 * Free a fragment reassembly header and all 929 * associated datagrams. 930 */ 931static void 932ip_freef(fp) 933 struct ipq *fp; 934{ 935 register struct mbuf *q; 936 937 while (fp->ipq_frags) { 938 q = fp->ipq_frags; 939 fp->ipq_frags = q->m_nextpkt; 940 m_freem(q); 941 } 942 remque(fp); 943 (void) m_free(dtom(fp)); 944 nipq--; 945} 946 947/* 948 * IP timer processing; 949 * if a timer expires on a reassembly 950 * queue, discard it. 951 */ 952void 953ip_slowtimo() 954{ 955 register struct ipq *fp; 956 int s = splnet(); 957 int i; 958 959 for (i = 0; i < IPREASS_NHASH; i++) { 960 fp = ipq[i].next; 961 if (fp == 0) 962 continue; 963 while (fp != &ipq[i]) { 964 --fp->ipq_ttl; 965 fp = fp->next; 966 if (fp->prev->ipq_ttl == 0) { 967 ipstat.ips_fragtimeout++; 968 ip_freef(fp->prev); 969 } 970 } 971 } 972 ipflow_slowtimo(); 973 splx(s); 974} 975 976/* 977 * Drain off all datagram fragments. 978 */ 979void 980ip_drain() 981{ 982 int i; 983 984 for (i = 0; i < IPREASS_NHASH; i++) { 985 while (ipq[i].next != &ipq[i]) { 986 ipstat.ips_fragdropped++; 987 ip_freef(ipq[i].next); 988 } 989 } 990 in_rtqdrain(); 991} 992 993/* 994 * Do option processing on a datagram, 995 * possibly discarding it if bad options are encountered, 996 * or forwarding it if source-routed. 997 * Returns 1 if packet has been forwarded/freed, 998 * 0 if the packet should be processed further. 999 */ 1000static int 1001ip_dooptions(m) 1002 struct mbuf *m; 1003{ 1004 register struct ip *ip = mtod(m, struct ip *); 1005 register u_char *cp; 1006 register struct ip_timestamp *ipt; 1007 register struct in_ifaddr *ia; 1008 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1009 struct in_addr *sin, dst; 1010 n_time ntime; 1011 1012 dst = ip->ip_dst; 1013 cp = (u_char *)(ip + 1); 1014 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1015 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1016 opt = cp[IPOPT_OPTVAL]; 1017 if (opt == IPOPT_EOL) 1018 break; 1019 if (opt == IPOPT_NOP) 1020 optlen = 1; 1021 else { 1022 optlen = cp[IPOPT_OLEN]; 1023 if (optlen <= 0 || optlen > cnt) { 1024 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1025 goto bad; 1026 } 1027 } 1028 switch (opt) { 1029 1030 default: 1031 break; 1032 1033 /* 1034 * Source routing with record. 1035 * Find interface with current destination address. 1036 * If none on this machine then drop if strictly routed, 1037 * or do nothing if loosely routed. 1038 * Record interface address and bring up next address 1039 * component. If strictly routed make sure next 1040 * address is on directly accessible net. 1041 */ 1042 case IPOPT_LSRR: 1043 case IPOPT_SSRR: 1044 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1045 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1046 goto bad; 1047 } 1048 ipaddr.sin_addr = ip->ip_dst; 1049 ia = (struct in_ifaddr *) 1050 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1051 if (ia == 0) { 1052 if (opt == IPOPT_SSRR) { 1053 type = ICMP_UNREACH; 1054 code = ICMP_UNREACH_SRCFAIL; 1055 goto bad; 1056 } 1057 if (!ip_dosourceroute) 1058 goto nosourcerouting; 1059 /* 1060 * Loose routing, and not at next destination 1061 * yet; nothing to do except forward. 1062 */ 1063 break; 1064 } 1065 off--; /* 0 origin */ 1066 if (off > optlen - sizeof(struct in_addr)) { 1067 /* 1068 * End of source route. Should be for us. 1069 */ 1070 if (!ip_acceptsourceroute) 1071 goto nosourcerouting; 1072 save_rte(cp, ip->ip_src); 1073 break; 1074 } 1075 1076 if (!ip_dosourceroute) { 1077 if (ipforwarding) { 1078 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1079 /* 1080 * Acting as a router, so generate ICMP 1081 */ 1082nosourcerouting: 1083 strcpy(buf, inet_ntoa(ip->ip_dst)); 1084 log(LOG_WARNING, 1085 "attempted source route from %s to %s\n", 1086 inet_ntoa(ip->ip_src), buf); 1087 type = ICMP_UNREACH; 1088 code = ICMP_UNREACH_SRCFAIL; 1089 goto bad; 1090 } else { 1091 /* 1092 * Not acting as a router, so silently drop. 1093 */ 1094 ipstat.ips_cantforward++; 1095 m_freem(m); 1096 return (1); 1097 } 1098 } 1099 1100 /* 1101 * locate outgoing interface 1102 */ 1103 (void)memcpy(&ipaddr.sin_addr, cp + off, 1104 sizeof(ipaddr.sin_addr)); 1105 1106 if (opt == IPOPT_SSRR) { 1107#define INA struct in_ifaddr * 1108#define SA struct sockaddr * 1109 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1110 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1111 } else 1112 ia = ip_rtaddr(ipaddr.sin_addr); 1113 if (ia == 0) { 1114 type = ICMP_UNREACH; 1115 code = ICMP_UNREACH_SRCFAIL; 1116 goto bad; 1117 } 1118 ip->ip_dst = ipaddr.sin_addr; 1119 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1120 sizeof(struct in_addr)); 1121 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1122 /* 1123 * Let ip_intr's mcast routing check handle mcast pkts 1124 */ 1125 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1126 break; 1127 1128 case IPOPT_RR: 1129 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1130 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1131 goto bad; 1132 } 1133 /* 1134 * If no space remains, ignore. 1135 */ 1136 off--; /* 0 origin */ 1137 if (off > optlen - sizeof(struct in_addr)) 1138 break; 1139 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1140 sizeof(ipaddr.sin_addr)); 1141 /* 1142 * locate outgoing interface; if we're the destination, 1143 * use the incoming interface (should be same). 1144 */ 1145 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1146 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1147 type = ICMP_UNREACH; 1148 code = ICMP_UNREACH_HOST; 1149 goto bad; 1150 } 1151 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1152 sizeof(struct in_addr)); 1153 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1154 break; 1155 1156 case IPOPT_TS: 1157 code = cp - (u_char *)ip; 1158 ipt = (struct ip_timestamp *)cp; 1159 if (ipt->ipt_len < 5) 1160 goto bad; 1161 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1162 if (++ipt->ipt_oflw == 0) 1163 goto bad; 1164 break; 1165 } 1166 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1167 switch (ipt->ipt_flg) { 1168 1169 case IPOPT_TS_TSONLY: 1170 break; 1171 1172 case IPOPT_TS_TSANDADDR: 1173 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1174 sizeof(struct in_addr) > ipt->ipt_len) 1175 goto bad; 1176 ipaddr.sin_addr = dst; 1177 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1178 m->m_pkthdr.rcvif); 1179 if (ia == 0) 1180 continue; 1181 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1182 sizeof(struct in_addr)); 1183 ipt->ipt_ptr += sizeof(struct in_addr); 1184 break; 1185 1186 case IPOPT_TS_PRESPEC: 1187 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1188 sizeof(struct in_addr) > ipt->ipt_len) 1189 goto bad; 1190 (void)memcpy(&ipaddr.sin_addr, sin, 1191 sizeof(struct in_addr)); 1192 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1193 continue; 1194 ipt->ipt_ptr += sizeof(struct in_addr); 1195 break; 1196 1197 default: 1198 goto bad; 1199 } 1200 ntime = iptime(); 1201 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1202 sizeof(n_time)); 1203 ipt->ipt_ptr += sizeof(n_time); 1204 } 1205 } 1206 if (forward && ipforwarding) { 1207 ip_forward(m, 1); 1208 return (1); 1209 } 1210 return (0); 1211bad: 1212 ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2; /* XXX icmp_error adds in hdr length */ 1213 icmp_error(m, type, code, 0, 0); 1214 ipstat.ips_badoptions++; 1215 return (1); 1216} 1217 1218/* 1219 * Given address of next destination (final or next hop), 1220 * return internet address info of interface to be used to get there. 1221 */ 1222static struct in_ifaddr * 1223ip_rtaddr(dst) 1224 struct in_addr dst; 1225{ 1226 register struct sockaddr_in *sin; 1227 1228 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1229 1230 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1231 if (ipforward_rt.ro_rt) { 1232 RTFREE(ipforward_rt.ro_rt); 1233 ipforward_rt.ro_rt = 0; 1234 } 1235 sin->sin_family = AF_INET; 1236 sin->sin_len = sizeof(*sin); 1237 sin->sin_addr = dst; 1238 1239 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1240 } 1241 if (ipforward_rt.ro_rt == 0) 1242 return ((struct in_ifaddr *)0); 1243 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1244} 1245 1246/* 1247 * Save incoming source route for use in replies, 1248 * to be picked up later by ip_srcroute if the receiver is interested. 1249 */ 1250void 1251save_rte(option, dst) 1252 u_char *option; 1253 struct in_addr dst; 1254{ 1255 unsigned olen; 1256 1257 olen = option[IPOPT_OLEN]; 1258#ifdef DIAGNOSTIC 1259 if (ipprintfs) 1260 printf("save_rte: olen %d\n", olen); 1261#endif 1262 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1263 return; 1264 bcopy(option, ip_srcrt.srcopt, olen); 1265 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1266 ip_srcrt.dst = dst; 1267} 1268 1269/* 1270 * Retrieve incoming source route for use in replies, 1271 * in the same form used by setsockopt. 1272 * The first hop is placed before the options, will be removed later. 1273 */ 1274struct mbuf * 1275ip_srcroute() 1276{ 1277 register struct in_addr *p, *q; 1278 register struct mbuf *m; 1279 1280 if (ip_nhops == 0) 1281 return ((struct mbuf *)0); 1282 m = m_get(M_DONTWAIT, MT_HEADER); 1283 if (m == 0) 1284 return ((struct mbuf *)0); 1285 1286#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1287 1288 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1289 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1290 OPTSIZ; 1291#ifdef DIAGNOSTIC 1292 if (ipprintfs) 1293 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1294#endif 1295 1296 /* 1297 * First save first hop for return route 1298 */ 1299 p = &ip_srcrt.route[ip_nhops - 1]; 1300 *(mtod(m, struct in_addr *)) = *p--; 1301#ifdef DIAGNOSTIC 1302 if (ipprintfs) 1303 printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr)); 1304#endif 1305 1306 /* 1307 * Copy option fields and padding (nop) to mbuf. 1308 */ 1309 ip_srcrt.nop = IPOPT_NOP; 1310 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1311 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1312 &ip_srcrt.nop, OPTSIZ); 1313 q = (struct in_addr *)(mtod(m, caddr_t) + 1314 sizeof(struct in_addr) + OPTSIZ); 1315#undef OPTSIZ 1316 /* 1317 * Record return path as an IP source route, 1318 * reversing the path (pointers are now aligned). 1319 */ 1320 while (p >= ip_srcrt.route) { 1321#ifdef DIAGNOSTIC 1322 if (ipprintfs) 1323 printf(" %lx", ntohl(q->s_addr)); 1324#endif 1325 *q++ = *p--; 1326 } 1327 /* 1328 * Last hop goes to final destination. 1329 */ 1330 *q = ip_srcrt.dst; 1331#ifdef DIAGNOSTIC 1332 if (ipprintfs) 1333 printf(" %lx\n", ntohl(q->s_addr)); 1334#endif 1335 return (m); 1336} 1337 1338/* 1339 * Strip out IP options, at higher 1340 * level protocol in the kernel. 1341 * Second argument is buffer to which options 1342 * will be moved, and return value is their length. 1343 * XXX should be deleted; last arg currently ignored. 1344 */ 1345void 1346ip_stripoptions(m, mopt) 1347 register struct mbuf *m; 1348 struct mbuf *mopt; 1349{ 1350 register int i; 1351 struct ip *ip = mtod(m, struct ip *); 1352 register caddr_t opts; 1353 int olen; 1354 1355 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1356 opts = (caddr_t)(ip + 1); 1357 i = m->m_len - (sizeof (struct ip) + olen); 1358 bcopy(opts + olen, opts, (unsigned)i); 1359 m->m_len -= olen; 1360 if (m->m_flags & M_PKTHDR) 1361 m->m_pkthdr.len -= olen; 1362 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1363} 1364 1365u_char inetctlerrmap[PRC_NCMDS] = { 1366 0, 0, 0, 0, 1367 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1368 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1369 EMSGSIZE, EHOSTUNREACH, 0, 0, 1370 0, 0, 0, 0, 1371 ENOPROTOOPT 1372}; 1373 1374/* 1375 * Forward a packet. If some error occurs return the sender 1376 * an icmp packet. Note we can't always generate a meaningful 1377 * icmp message because icmp doesn't have a large enough repertoire 1378 * of codes and types. 1379 * 1380 * If not forwarding, just drop the packet. This could be confusing 1381 * if ipforwarding was zero but some routing protocol was advancing 1382 * us as a gateway to somewhere. However, we must let the routing 1383 * protocol deal with that. 1384 * 1385 * The srcrt parameter indicates whether the packet is being forwarded 1386 * via a source route. 1387 */ 1388static void 1389ip_forward(m, srcrt) 1390 struct mbuf *m; 1391 int srcrt; 1392{ 1393 register struct ip *ip = mtod(m, struct ip *); 1394 register struct sockaddr_in *sin; 1395 register struct rtentry *rt; 1396 int error, type = 0, code = 0; 1397 struct mbuf *mcopy; 1398 n_long dest; 1399 struct ifnet *destifp; 1400 1401 dest = 0; 1402#ifdef DIAGNOSTIC 1403 if (ipprintfs) 1404 printf("forward: src %lx dst %lx ttl %x\n", 1405 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1406 ip->ip_ttl); 1407#endif 1408 1409 1410 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) { 1411 ipstat.ips_cantforward++; 1412 m_freem(m); 1413 return; 1414 } 1415 HTONS(ip->ip_id); 1416 if (ip->ip_ttl <= IPTTLDEC) { 1417 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1418 return; 1419 } 1420 ip->ip_ttl -= IPTTLDEC; 1421 1422 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1423 if ((rt = ipforward_rt.ro_rt) == 0 || 1424 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1425 if (ipforward_rt.ro_rt) { 1426 RTFREE(ipforward_rt.ro_rt); 1427 ipforward_rt.ro_rt = 0; 1428 } 1429 sin->sin_family = AF_INET; 1430 sin->sin_len = sizeof(*sin); 1431 sin->sin_addr = ip->ip_dst; 1432 1433 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1434 if (ipforward_rt.ro_rt == 0) { 1435 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1436 return; 1437 } 1438 rt = ipforward_rt.ro_rt; 1439 } 1440 1441 /* 1442 * Save at most 64 bytes of the packet in case 1443 * we need to generate an ICMP message to the src. 1444 */ 1445 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1446 1447 /* 1448 * If forwarding packet using same interface that it came in on, 1449 * perhaps should send a redirect to sender to shortcut a hop. 1450 * Only send redirect if source is sending directly to us, 1451 * and if packet was not source routed (or has any options). 1452 * Also, don't send redirect if forwarding using a default route 1453 * or a route modified by a redirect. 1454 */ 1455#define satosin(sa) ((struct sockaddr_in *)(sa)) 1456 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1457 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1458 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1459 ipsendredirects && !srcrt) { 1460#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1461 u_long src = ntohl(ip->ip_src.s_addr); 1462 1463 if (RTA(rt) && 1464 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1465 if (rt->rt_flags & RTF_GATEWAY) 1466 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1467 else 1468 dest = ip->ip_dst.s_addr; 1469 /* Router requirements says to only send host redirects */ 1470 type = ICMP_REDIRECT; 1471 code = ICMP_REDIRECT_HOST; 1472#ifdef DIAGNOSTIC 1473 if (ipprintfs) 1474 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1475#endif 1476 } 1477 } 1478 1479 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1480 IP_FORWARDING, 0); 1481 if (error) 1482 ipstat.ips_cantforward++; 1483 else { 1484 ipstat.ips_forward++; 1485 if (type) 1486 ipstat.ips_redirectsent++; 1487 else { 1488 if (mcopy) { 1489 ipflow_create(&ipforward_rt, mcopy); 1490 m_freem(mcopy); 1491 } 1492 return; 1493 } 1494 } 1495 if (mcopy == NULL) 1496 return; 1497 destifp = NULL; 1498 1499 switch (error) { 1500 1501 case 0: /* forwarded, but need redirect */ 1502 /* type, code set above */ 1503 break; 1504 1505 case ENETUNREACH: /* shouldn't happen, checked above */ 1506 case EHOSTUNREACH: 1507 case ENETDOWN: 1508 case EHOSTDOWN: 1509 default: 1510 type = ICMP_UNREACH; 1511 code = ICMP_UNREACH_HOST; 1512 break; 1513 1514 case EMSGSIZE: 1515 type = ICMP_UNREACH; 1516 code = ICMP_UNREACH_NEEDFRAG; 1517 if (ipforward_rt.ro_rt) 1518 destifp = ipforward_rt.ro_rt->rt_ifp; 1519 ipstat.ips_cantfrag++; 1520 break; 1521 1522 case ENOBUFS: 1523 type = ICMP_SOURCEQUENCH; 1524 code = 0; 1525 break; 1526 } 1527 icmp_error(mcopy, type, code, dest, destifp); 1528} 1529 1530void 1531ip_savecontrol(inp, mp, ip, m) 1532 register struct inpcb *inp; 1533 register struct mbuf **mp; 1534 register struct ip *ip; 1535 register struct mbuf *m; 1536{ 1537 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1538 struct timeval tv; 1539 1540 microtime(&tv); 1541 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1542 SCM_TIMESTAMP, SOL_SOCKET); 1543 if (*mp) 1544 mp = &(*mp)->m_next; 1545 } 1546 if (inp->inp_flags & INP_RECVDSTADDR) { 1547 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1548 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1549 if (*mp) 1550 mp = &(*mp)->m_next; 1551 } 1552#ifdef notyet 1553 /* XXX 1554 * Moving these out of udp_input() made them even more broken 1555 * than they already were. 1556 */ 1557 /* options were tossed already */ 1558 if (inp->inp_flags & INP_RECVOPTS) { 1559 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1560 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1561 if (*mp) 1562 mp = &(*mp)->m_next; 1563 } 1564 /* ip_srcroute doesn't do what we want here, need to fix */ 1565 if (inp->inp_flags & INP_RECVRETOPTS) { 1566 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1567 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1568 if (*mp) 1569 mp = &(*mp)->m_next; 1570 } 1571#endif 1572 if (inp->inp_flags & INP_RECVIF) { 1573 struct ifnet *ifp; 1574 struct sdlbuf { 1575 struct sockaddr_dl sdl; 1576 u_char pad[32]; 1577 } sdlbuf; 1578 struct sockaddr_dl *sdp; 1579 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1580 1581 if (((ifp = m->m_pkthdr.rcvif)) 1582 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1583 sdp = (struct sockaddr_dl *)(ifnet_addrs 1584 [ifp->if_index - 1]->ifa_addr); 1585 /* 1586 * Change our mind and don't try copy. 1587 */ 1588 if ((sdp->sdl_family != AF_LINK) 1589 || (sdp->sdl_len > sizeof(sdlbuf))) { 1590 goto makedummy; 1591 } 1592 bcopy(sdp, sdl2, sdp->sdl_len); 1593 } else { 1594makedummy: 1595 sdl2->sdl_len 1596 = offsetof(struct sockaddr_dl, sdl_data[0]); 1597 sdl2->sdl_family = AF_LINK; 1598 sdl2->sdl_index = 0; 1599 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1600 } 1601 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1602 IP_RECVIF, IPPROTO_IP); 1603 if (*mp) 1604 mp = &(*mp)->m_next; 1605 } 1606} 1607 1608int 1609ip_rsvp_init(struct socket *so) 1610{ 1611 if (so->so_type != SOCK_RAW || 1612 so->so_proto->pr_protocol != IPPROTO_RSVP) 1613 return EOPNOTSUPP; 1614 1615 if (ip_rsvpd != NULL) 1616 return EADDRINUSE; 1617 1618 ip_rsvpd = so; 1619 /* 1620 * This may seem silly, but we need to be sure we don't over-increment 1621 * the RSVP counter, in case something slips up. 1622 */ 1623 if (!ip_rsvp_on) { 1624 ip_rsvp_on = 1; 1625 rsvp_on++; 1626 } 1627 1628 return 0; 1629} 1630 1631int 1632ip_rsvp_done(void) 1633{ 1634 ip_rsvpd = NULL; 1635 /* 1636 * This may seem silly, but we need to be sure we don't over-decrement 1637 * the RSVP counter, in case something slips up. 1638 */ 1639 if (ip_rsvp_on) { 1640 ip_rsvp_on = 0; 1641 rsvp_on--; 1642 } 1643 return 0; 1644} 1645