ip_input.c revision 38482
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 1.98 1998/08/17 01:05:24 bde 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_ipdivert.h" 43#include "opt_ipfilter.h" 44 45#include <stddef.h> 46 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/mbuf.h> 50#include <sys/domain.h> 51#include <sys/protosw.h> 52#include <sys/socket.h> 53#include <sys/time.h> 54#include <sys/kernel.h> 55#include <sys/syslog.h> 56#include <sys/sysctl.h> 57 58#include <net/if.h> 59#include <net/if_var.h> 60#include <net/if_dl.h> 61#include <net/route.h> 62#include <net/netisr.h> 63 64#include <netinet/in.h> 65#include <netinet/in_systm.h> 66#include <netinet/in_var.h> 67#include <netinet/ip.h> 68#include <netinet/in_pcb.h> 69#include <netinet/ip_var.h> 70#include <netinet/ip_icmp.h> 71#include <machine/in_cksum.h> 72 73#include <sys/socketvar.h> 74 75#ifdef IPFIREWALL 76#include <netinet/ip_fw.h> 77#endif 78 79int rsvp_on = 0; 80static int ip_rsvp_on; 81struct socket *ip_rsvpd; 82 83int ipforwarding = 0; 84SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 85 &ipforwarding, 0, ""); 86 87static int ipsendredirects = 1; /* XXX */ 88SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 89 &ipsendredirects, 0, ""); 90 91int ip_defttl = IPDEFTTL; 92SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 93 &ip_defttl, 0, ""); 94 95static int ip_dosourceroute = 0; 96SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 97 &ip_dosourceroute, 0, ""); 98 99static int ip_acceptsourceroute = 0; 100SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 101 CTLFLAG_RW, &ip_acceptsourceroute, 0, ""); 102#ifdef DIAGNOSTIC 103static int ipprintfs = 0; 104#endif 105 106extern struct domain inetdomain; 107extern struct protosw inetsw[]; 108u_char ip_protox[IPPROTO_MAX]; 109static int ipqmaxlen = IFQ_MAXLEN; 110struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 111struct ifqueue ipintrq; 112SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RD, 113 &ipintrq.ifq_maxlen, 0, ""); 114SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 115 &ipintrq.ifq_drops, 0, ""); 116 117struct ipstat ipstat; 118SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD, 119 &ipstat, ipstat, ""); 120 121/* Packet reassembly stuff */ 122#define IPREASS_NHASH_LOG2 6 123#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 124#define IPREASS_HMASK (IPREASS_NHASH - 1) 125#define IPREASS_HASH(x,y) \ 126 ((((x) & 0xF | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 127 128static struct ipq ipq[IPREASS_NHASH]; 129static int nipq = 0; /* total # of reass queues */ 130static int maxnipq; 131 132#ifdef IPCTL_DEFMTU 133SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 134 &ip_mtu, 0, ""); 135#endif 136 137#if !defined(COMPAT_IPFW) || COMPAT_IPFW == 1 138#undef COMPAT_IPFW 139#define COMPAT_IPFW 1 140#else 141#undef COMPAT_IPFW 142#endif 143 144#ifdef COMPAT_IPFW 145 146#include <netinet/ip_fw.h> 147 148/* Firewall hooks */ 149ip_fw_chk_t *ip_fw_chk_ptr; 150ip_fw_ctl_t *ip_fw_ctl_ptr; 151 152/* IP Network Address Translation (NAT) hooks */ 153ip_nat_t *ip_nat_ptr; 154ip_nat_ctl_t *ip_nat_ctl_ptr; 155#endif 156 157#if defined(IPFILTER_LKM) || defined(IPFILTER) 158int iplattach __P((void)); 159int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL; 160#endif 161 162 163/* 164 * We need to save the IP options in case a protocol wants to respond 165 * to an incoming packet over the same route if the packet got here 166 * using IP source routing. This allows connection establishment and 167 * maintenance when the remote end is on a network that is not known 168 * to us. 169 */ 170static int ip_nhops = 0; 171static struct ip_srcrt { 172 struct in_addr dst; /* final destination */ 173 char nop; /* one NOP to align */ 174 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 175 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 176} ip_srcrt; 177 178#ifdef IPDIVERT 179/* 180 * Shared variable between ip_input() and ip_reass() to communicate 181 * about which packets, once assembled from fragments, get diverted, 182 * and to which port. 183 */ 184static u_short frag_divert_port; 185#endif 186 187struct sockaddr_in *ip_fw_fwd_addr; 188 189static void save_rte __P((u_char *, struct in_addr)); 190static void ip_deq __P((struct ipasfrag *)); 191static int ip_dooptions __P((struct mbuf *)); 192static void ip_enq __P((struct ipasfrag *, struct ipasfrag *)); 193static void ip_forward __P((struct mbuf *, int)); 194static void ip_freef __P((struct ipq *)); 195static struct ip * 196 ip_reass __P((struct ipasfrag *, struct ipq *, struct ipq *)); 197static struct in_ifaddr * 198 ip_rtaddr __P((struct in_addr)); 199static void ipintr __P((void)); 200/* 201 * IP initialization: fill in IP protocol switch table. 202 * All protocols not implemented in kernel go to raw IP protocol handler. 203 */ 204void 205ip_init() 206{ 207 register struct protosw *pr; 208 register int i; 209 210 TAILQ_INIT(&in_ifaddrhead); 211 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 212 if (pr == 0) 213 panic("ip_init"); 214 for (i = 0; i < IPPROTO_MAX; i++) 215 ip_protox[i] = pr - inetsw; 216 for (pr = inetdomain.dom_protosw; 217 pr < inetdomain.dom_protoswNPROTOSW; pr++) 218 if (pr->pr_domain->dom_family == PF_INET && 219 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 220 ip_protox[pr->pr_protocol] = pr - inetsw; 221 222 for (i = 0; i < IPREASS_NHASH; i++) 223 ipq[i].next = ipq[i].prev = &ipq[i]; 224 225 maxnipq = nmbclusters/4; 226 227 ip_id = time_second & 0xffff; 228 ipintrq.ifq_maxlen = ipqmaxlen; 229#ifdef IPFIREWALL 230 ip_fw_init(); 231#endif 232#ifdef IPNAT 233 ip_nat_init(); 234#endif 235#ifdef IPFILTER 236 iplattach(); 237#endif 238 239} 240 241static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 242static struct route ipforward_rt; 243 244/* 245 * Ip input routine. Checksum and byte swap header. If fragmented 246 * try to reassemble. Process options. Pass to next level. 247 */ 248void 249ip_input(struct mbuf *m) 250{ 251 struct ip *ip; 252 struct ipq *fp; 253 struct in_ifaddr *ia; 254 int i, hlen; 255 u_short sum; 256 257#ifdef DIAGNOSTIC 258 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 259 panic("ip_input no HDR"); 260#endif 261 /* 262 * If no IP addresses have been set yet but the interfaces 263 * are receiving, can't do anything with incoming packets yet. 264 * XXX This is broken! We should be able to receive broadcasts 265 * and multicasts even without any local addresses configured. 266 */ 267 if (TAILQ_EMPTY(&in_ifaddrhead)) 268 goto bad; 269 ipstat.ips_total++; 270 271 if (m->m_pkthdr.len < sizeof(struct ip)) 272 goto tooshort; 273 274 if (m->m_len < sizeof (struct ip) && 275 (m = m_pullup(m, sizeof (struct ip))) == 0) { 276 ipstat.ips_toosmall++; 277 return; 278 } 279 ip = mtod(m, struct ip *); 280 281 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 282 ipstat.ips_badvers++; 283 goto bad; 284 } 285 286 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 287 if (hlen < sizeof(struct ip)) { /* minimum header length */ 288 ipstat.ips_badhlen++; 289 goto bad; 290 } 291 if (hlen > m->m_len) { 292 if ((m = m_pullup(m, hlen)) == 0) { 293 ipstat.ips_badhlen++; 294 return; 295 } 296 ip = mtod(m, struct ip *); 297 } 298 if (hlen == sizeof(struct ip)) { 299 sum = in_cksum_hdr(ip); 300 } else { 301 sum = in_cksum(m, hlen); 302 } 303 if (sum) { 304 ipstat.ips_badsum++; 305 goto bad; 306 } 307 308 /* 309 * Convert fields to host representation. 310 */ 311 NTOHS(ip->ip_len); 312 if (ip->ip_len < hlen) { 313 ipstat.ips_badlen++; 314 goto bad; 315 } 316 NTOHS(ip->ip_id); 317 NTOHS(ip->ip_off); 318 319 /* 320 * Check that the amount of data in the buffers 321 * is as at least much as the IP header would have us expect. 322 * Trim mbufs if longer than we expect. 323 * Drop packet if shorter than we expect. 324 */ 325 if (m->m_pkthdr.len < ip->ip_len) { 326tooshort: 327 ipstat.ips_tooshort++; 328 goto bad; 329 } 330 if (m->m_pkthdr.len > ip->ip_len) { 331 if (m->m_len == m->m_pkthdr.len) { 332 m->m_len = ip->ip_len; 333 m->m_pkthdr.len = ip->ip_len; 334 } else 335 m_adj(m, ip->ip_len - m->m_pkthdr.len); 336 } 337 /* 338 * IpHack's section. 339 * Right now when no processing on packet has done 340 * and it is still fresh out of network we do our black 341 * deals with it. 342 * - Firewall: deny/allow/divert 343 * - Xlate: translate packet's addr/port (NAT). 344 * - Wrap: fake packet's addr/port <unimpl.> 345 * - Encapsulate: put it in another IP and send out. <unimp.> 346 */ 347#if defined(IPFILTER) || defined(IPFILTER_LKM) 348 /* 349 * Check if we want to allow this packet to be processed. 350 * Consider it to be bad if not. 351 */ 352 if (fr_checkp) { 353 struct mbuf *m1 = m; 354 355 if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1) 356 return; 357 ip = mtod(m = m1, struct ip *); 358 } 359#endif 360#ifdef COMPAT_IPFW 361 if (ip_fw_chk_ptr) { 362 u_int16_t port; 363 364#ifdef IPFIREWALL_FORWARD 365 /* 366 * If we've been forwarded from the output side, then 367 * skip the firewall a second time 368 */ 369 if (ip_fw_fwd_addr) 370 goto ours; 371#endif /* IPFIREWALL_FORWARD */ 372#ifdef IPDIVERT 373 port = (*ip_fw_chk_ptr)(&ip, hlen, NULL, &ip_divert_cookie, 374 &m, &ip_fw_fwd_addr); 375 if (port) { 376 /* Divert packet */ 377 frag_divert_port = port; 378 goto ours; 379 } 380#else /* !DIVERT */ 381 /* 382 * If ipfw says divert, we have to just drop packet 383 * Use port as a dummy argument. 384 */ 385 port = 0; 386 if ((*ip_fw_chk_ptr)(&ip, hlen, NULL, &port, 387 &m, &ip_fw_fwd_addr)) { 388 m_freem(m); 389 m = NULL; 390 } 391#endif /* !DIVERT */ 392 if (!m) 393 return; 394 } 395 396 if (ip_nat_ptr && !(*ip_nat_ptr)(&ip, &m, m->m_pkthdr.rcvif, IP_NAT_IN)) 397 return; 398#endif /* !COMPAT_IPFW */ 399 400 /* 401 * Process options and, if not destined for us, 402 * ship it on. ip_dooptions returns 1 when an 403 * error was detected (causing an icmp message 404 * to be sent and the original packet to be freed). 405 */ 406 ip_nhops = 0; /* for source routed packets */ 407 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 408 return; 409 410 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 411 * matter if it is destined to another node, or whether it is 412 * a multicast one, RSVP wants it! and prevents it from being forwarded 413 * anywhere else. Also checks if the rsvp daemon is running before 414 * grabbing the packet. 415 */ 416 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 417 goto ours; 418 419 /* 420 * Check our list of addresses, to see if the packet is for us. 421 */ 422 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 423 ia = TAILQ_NEXT(ia, ia_link)) { 424#define satosin(sa) ((struct sockaddr_in *)(sa)) 425 426 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 427 goto ours; 428#ifdef BOOTP_COMPAT 429 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 430 goto ours; 431#endif 432#ifdef IPFIREWALL_FORWARD 433 /* 434 * If the addr to forward to is one of ours, we pretend to 435 * be the destination for this packet. 436 */ 437 if (ip_fw_fwd_addr != NULL && 438 IA_SIN(ia)->sin_addr.s_addr == 439 ip_fw_fwd_addr->sin_addr.s_addr) 440 goto ours; 441#endif 442 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 443 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 444 ip->ip_dst.s_addr) 445 goto ours; 446 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 447 goto ours; 448 } 449 } 450 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 451 struct in_multi *inm; 452 if (ip_mrouter) { 453 /* 454 * If we are acting as a multicast router, all 455 * incoming multicast packets are passed to the 456 * kernel-level multicast forwarding function. 457 * The packet is returned (relatively) intact; if 458 * ip_mforward() returns a non-zero value, the packet 459 * must be discarded, else it may be accepted below. 460 * 461 * (The IP ident field is put in the same byte order 462 * as expected when ip_mforward() is called from 463 * ip_output().) 464 */ 465 ip->ip_id = htons(ip->ip_id); 466 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 467 ipstat.ips_cantforward++; 468 m_freem(m); 469 return; 470 } 471 ip->ip_id = ntohs(ip->ip_id); 472 473 /* 474 * The process-level routing demon needs to receive 475 * all multicast IGMP packets, whether or not this 476 * host belongs to their destination groups. 477 */ 478 if (ip->ip_p == IPPROTO_IGMP) 479 goto ours; 480 ipstat.ips_forward++; 481 } 482 /* 483 * See if we belong to the destination multicast group on the 484 * arrival interface. 485 */ 486 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 487 if (inm == NULL) { 488 ipstat.ips_notmember++; 489 m_freem(m); 490 return; 491 } 492 goto ours; 493 } 494 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 495 goto ours; 496 if (ip->ip_dst.s_addr == INADDR_ANY) 497 goto ours; 498 499 /* 500 * Not for us; forward if possible and desirable. 501 */ 502 if (ipforwarding == 0) { 503 ipstat.ips_cantforward++; 504 m_freem(m); 505 } else 506 ip_forward(m, 0); 507 return; 508 509ours: 510 511 /* 512 * If offset or IP_MF are set, must reassemble. 513 * Otherwise, nothing need be done. 514 * (We could look in the reassembly queue to see 515 * if the packet was previously fragmented, 516 * but it's not worth the time; just let them time out.) 517 */ 518 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) { 519 if (m->m_flags & M_EXT) { /* XXX */ 520 if ((m = m_pullup(m, sizeof (struct ip))) == 0) { 521 ipstat.ips_toosmall++; 522#ifdef IPDIVERT 523 frag_divert_port = 0; 524 ip_divert_cookie = 0; 525#endif 526 return; 527 } 528 ip = mtod(m, struct ip *); 529 } 530 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 531 /* 532 * Look for queue of fragments 533 * of this datagram. 534 */ 535 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 536 if (ip->ip_id == fp->ipq_id && 537 ip->ip_src.s_addr == fp->ipq_src.s_addr && 538 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 539 ip->ip_p == fp->ipq_p) 540 goto found; 541 542 fp = 0; 543 544 /* check if there's a place for the new queue */ 545 if (nipq > maxnipq) { 546 /* 547 * drop something from the tail of the current queue 548 * before proceeding further 549 */ 550 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 551 for (i = 0; i < IPREASS_NHASH; i++) { 552 if (ipq[i].prev != &ipq[i]) { 553 ip_freef(ipq[i].prev); 554 break; 555 } 556 } 557 } else 558 ip_freef(ipq[sum].prev); 559 } 560found: 561 /* 562 * Adjust ip_len to not reflect header, 563 * set ip_mff if more fragments are expected, 564 * convert offset of this to bytes. 565 */ 566 ip->ip_len -= hlen; 567 ((struct ipasfrag *)ip)->ipf_mff &= ~1; 568 if (ip->ip_off & IP_MF) 569 ((struct ipasfrag *)ip)->ipf_mff |= 1; 570 ip->ip_off <<= 3; 571 572 /* 573 * If datagram marked as having more fragments 574 * or if this is not the first fragment, 575 * attempt reassembly; if it succeeds, proceed. 576 */ 577 if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) { 578 ipstat.ips_fragments++; 579 ip = ip_reass((struct ipasfrag *)ip, fp, &ipq[sum]); 580 if (ip == 0) { 581#ifdef IPFIREWALL_FORWARD 582 ip_fw_fwd_addr = NULL; 583#endif 584 return; 585 } 586 /* Get the length of the reassembled packets header */ 587 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 588 ipstat.ips_reassembled++; 589 m = dtom(ip); 590#ifdef IPDIVERT 591 if (frag_divert_port) { 592 ip->ip_len += hlen; 593 HTONS(ip->ip_len); 594 HTONS(ip->ip_off); 595 HTONS(ip->ip_id); 596 ip->ip_sum = 0; 597 ip->ip_sum = in_cksum_hdr(ip); 598 NTOHS(ip->ip_id); 599 NTOHS(ip->ip_off); 600 NTOHS(ip->ip_len); 601 ip->ip_len -= hlen; 602 } 603#endif 604 } else 605 if (fp) 606 ip_freef(fp); 607 } else 608 ip->ip_len -= hlen; 609 610#ifdef IPDIVERT 611 /* 612 * Divert reassembled packets to the divert protocol if required 613 * If divert port is null then cookie should be too, 614 * so we shouldn't need to clear them here. Assume ip_divert does so. 615 */ 616 if (frag_divert_port) { 617 ipstat.ips_delivered++; 618 ip_divert_port = frag_divert_port; 619 frag_divert_port = 0; 620 (*inetsw[ip_protox[IPPROTO_DIVERT]].pr_input)(m, hlen); 621 return; 622 } 623 624 /* Don't let packets divert themselves */ 625 if (ip->ip_p == IPPROTO_DIVERT) { 626 ipstat.ips_noproto++; 627 goto bad; 628 } 629 630#endif 631 632 /* 633 * Switch out to protocol's input routine. 634 */ 635 ipstat.ips_delivered++; 636 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 637#ifdef IPFIREWALL_FORWARD 638 ip_fw_fwd_addr = NULL; /* tcp needed it */ 639#endif 640 return; 641bad: 642#ifdef IPFIREWALL_FORWARD 643 ip_fw_fwd_addr = NULL; 644#endif 645 m_freem(m); 646} 647 648/* 649 * IP software interrupt routine - to go away sometime soon 650 */ 651static void 652ipintr(void) 653{ 654 int s; 655 struct mbuf *m; 656 657 while(1) { 658 s = splimp(); 659 IF_DEQUEUE(&ipintrq, m); 660 splx(s); 661 if (m == 0) 662 return; 663 ip_input(m); 664 } 665} 666 667NETISR_SET(NETISR_IP, ipintr); 668 669/* 670 * Take incoming datagram fragment and try to 671 * reassemble it into whole datagram. If a chain for 672 * reassembly of this datagram already exists, then it 673 * is given as fp; otherwise have to make a chain. 674 */ 675static struct ip * 676ip_reass(ip, fp, where) 677 register struct ipasfrag *ip; 678 register struct ipq *fp; 679 struct ipq *where; 680{ 681 register struct mbuf *m = dtom(ip); 682 register struct ipasfrag *q; 683 struct mbuf *t; 684 int hlen = ip->ip_hl << 2; 685 int i, next; 686 687 /* 688 * Presence of header sizes in mbufs 689 * would confuse code below. 690 */ 691 m->m_data += hlen; 692 m->m_len -= hlen; 693 694 /* 695 * If first fragment to arrive, create a reassembly queue. 696 */ 697 if (fp == 0) { 698 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 699 goto dropfrag; 700 fp = mtod(t, struct ipq *); 701 insque(fp, where); 702 nipq++; 703 fp->ipq_ttl = IPFRAGTTL; 704 fp->ipq_p = ip->ip_p; 705 fp->ipq_id = ip->ip_id; 706 fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp; 707 fp->ipq_src = ((struct ip *)ip)->ip_src; 708 fp->ipq_dst = ((struct ip *)ip)->ip_dst; 709#ifdef IPDIVERT 710 fp->ipq_divert = 0; 711 fp->ipq_div_cookie = 0; 712#endif 713 q = (struct ipasfrag *)fp; 714 goto insert; 715 } 716 717 /* 718 * Find a segment which begins after this one does. 719 */ 720 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) 721 if (q->ip_off > ip->ip_off) 722 break; 723 724 /* 725 * If there is a preceding segment, it may provide some of 726 * our data already. If so, drop the data from the incoming 727 * segment. If it provides all of our data, drop us. 728 */ 729 if (q->ipf_prev != (struct ipasfrag *)fp) { 730 i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off; 731 if (i > 0) { 732 if (i >= ip->ip_len) 733 goto dropfrag; 734 m_adj(dtom(ip), i); 735 ip->ip_off += i; 736 ip->ip_len -= i; 737 } 738 } 739 740 /* 741 * While we overlap succeeding segments trim them or, 742 * if they are completely covered, dequeue them. 743 */ 744 while (q != (struct ipasfrag *)fp && 745 ip->ip_off + ip->ip_len > q->ip_off) { 746 struct mbuf *m0; 747 748 i = (ip->ip_off + ip->ip_len) - q->ip_off; 749 if (i < q->ip_len) { 750 q->ip_len -= i; 751 q->ip_off += i; 752 m_adj(dtom(q), i); 753 break; 754 } 755 m0 = dtom(q); 756 q = q->ipf_next; 757 ip_deq(q->ipf_prev); 758 m_freem(m0); 759 } 760 761insert: 762 763#ifdef IPDIVERT 764 /* 765 * Any fragment diverting causes the whole packet to divert 766 */ 767 if (frag_divert_port) { 768 fp->ipq_divert = frag_divert_port; 769 fp->ipq_div_cookie = ip_divert_cookie; 770 } 771 frag_divert_port = 0; 772 ip_divert_cookie = 0; 773#endif 774 775 /* 776 * Stick new segment in its place; 777 * check for complete reassembly. 778 */ 779 ip_enq(ip, q->ipf_prev); 780 next = 0; 781 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) { 782 if (q->ip_off != next) 783 return (0); 784 next += q->ip_len; 785 } 786 if (q->ipf_prev->ipf_mff & 1) 787 return (0); 788 789 /* 790 * Reassembly is complete. Make sure the packet is a sane size. 791 */ 792 if (next + (IP_VHL_HL(((struct ip *)fp->ipq_next)->ip_vhl) << 2) 793 > IP_MAXPACKET) { 794 ipstat.ips_toolong++; 795 ip_freef(fp); 796 return (0); 797 } 798 799 /* 800 * Concatenate fragments. 801 */ 802 q = fp->ipq_next; 803 m = dtom(q); 804 t = m->m_next; 805 m->m_next = 0; 806 m_cat(m, t); 807 q = q->ipf_next; 808 while (q != (struct ipasfrag *)fp) { 809 t = dtom(q); 810 q = q->ipf_next; 811 m_cat(m, t); 812 } 813 814#ifdef IPDIVERT 815 /* 816 * extract divert port for packet, if any 817 */ 818 frag_divert_port = fp->ipq_divert; 819 ip_divert_cookie = fp->ipq_div_cookie; 820#endif 821 822 /* 823 * Create header for new ip packet by 824 * modifying header of first packet; 825 * dequeue and discard fragment reassembly header. 826 * Make header visible. 827 */ 828 ip = fp->ipq_next; 829 ip->ip_len = next; 830 ip->ipf_mff &= ~1; 831 ((struct ip *)ip)->ip_src = fp->ipq_src; 832 ((struct ip *)ip)->ip_dst = fp->ipq_dst; 833 remque(fp); 834 nipq--; 835 (void) m_free(dtom(fp)); 836 m = dtom(ip); 837 m->m_len += (ip->ip_hl << 2); 838 m->m_data -= (ip->ip_hl << 2); 839 /* some debugging cruft by sklower, below, will go away soon */ 840 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 841 register int plen = 0; 842 for (t = m; m; m = m->m_next) 843 plen += m->m_len; 844 t->m_pkthdr.len = plen; 845 } 846 return ((struct ip *)ip); 847 848dropfrag: 849#ifdef IPDIVERT 850 frag_divert_port = 0; 851 ip_divert_cookie = 0; 852#endif 853 ipstat.ips_fragdropped++; 854 m_freem(m); 855 return (0); 856} 857 858/* 859 * Free a fragment reassembly header and all 860 * associated datagrams. 861 */ 862static void 863ip_freef(fp) 864 struct ipq *fp; 865{ 866 register struct ipasfrag *q, *p; 867 868 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) { 869 p = q->ipf_next; 870 ip_deq(q); 871 m_freem(dtom(q)); 872 } 873 remque(fp); 874 (void) m_free(dtom(fp)); 875 nipq--; 876} 877 878/* 879 * Put an ip fragment on a reassembly chain. 880 * Like insque, but pointers in middle of structure. 881 */ 882static void 883ip_enq(p, prev) 884 register struct ipasfrag *p, *prev; 885{ 886 887 p->ipf_prev = prev; 888 p->ipf_next = prev->ipf_next; 889 prev->ipf_next->ipf_prev = p; 890 prev->ipf_next = p; 891} 892 893/* 894 * To ip_enq as remque is to insque. 895 */ 896static void 897ip_deq(p) 898 register struct ipasfrag *p; 899{ 900 901 p->ipf_prev->ipf_next = p->ipf_next; 902 p->ipf_next->ipf_prev = p->ipf_prev; 903} 904 905/* 906 * IP timer processing; 907 * if a timer expires on a reassembly 908 * queue, discard it. 909 */ 910void 911ip_slowtimo() 912{ 913 register struct ipq *fp; 914 int s = splnet(); 915 int i; 916 917 for (i = 0; i < IPREASS_NHASH; i++) { 918 fp = ipq[i].next; 919 if (fp == 0) 920 continue; 921 while (fp != &ipq[i]) { 922 --fp->ipq_ttl; 923 fp = fp->next; 924 if (fp->prev->ipq_ttl == 0) { 925 ipstat.ips_fragtimeout++; 926 ip_freef(fp->prev); 927 } 928 } 929 } 930 ipflow_slowtimo(); 931 splx(s); 932} 933 934/* 935 * Drain off all datagram fragments. 936 */ 937void 938ip_drain() 939{ 940 int i; 941 942 for (i = 0; i < IPREASS_NHASH; i++) { 943 while (ipq[i].next != &ipq[i]) { 944 ipstat.ips_fragdropped++; 945 ip_freef(ipq[i].next); 946 } 947 } 948 in_rtqdrain(); 949} 950 951/* 952 * Do option processing on a datagram, 953 * possibly discarding it if bad options are encountered, 954 * or forwarding it if source-routed. 955 * Returns 1 if packet has been forwarded/freed, 956 * 0 if the packet should be processed further. 957 */ 958static int 959ip_dooptions(m) 960 struct mbuf *m; 961{ 962 register struct ip *ip = mtod(m, struct ip *); 963 register u_char *cp; 964 register struct ip_timestamp *ipt; 965 register struct in_ifaddr *ia; 966 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 967 struct in_addr *sin, dst; 968 n_time ntime; 969 970 dst = ip->ip_dst; 971 cp = (u_char *)(ip + 1); 972 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 973 for (; cnt > 0; cnt -= optlen, cp += optlen) { 974 opt = cp[IPOPT_OPTVAL]; 975 if (opt == IPOPT_EOL) 976 break; 977 if (opt == IPOPT_NOP) 978 optlen = 1; 979 else { 980 optlen = cp[IPOPT_OLEN]; 981 if (optlen <= 0 || optlen > cnt) { 982 code = &cp[IPOPT_OLEN] - (u_char *)ip; 983 goto bad; 984 } 985 } 986 switch (opt) { 987 988 default: 989 break; 990 991 /* 992 * Source routing with record. 993 * Find interface with current destination address. 994 * If none on this machine then drop if strictly routed, 995 * or do nothing if loosely routed. 996 * Record interface address and bring up next address 997 * component. If strictly routed make sure next 998 * address is on directly accessible net. 999 */ 1000 case IPOPT_LSRR: 1001 case IPOPT_SSRR: 1002 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1003 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1004 goto bad; 1005 } 1006 ipaddr.sin_addr = ip->ip_dst; 1007 ia = (struct in_ifaddr *) 1008 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1009 if (ia == 0) { 1010 if (opt == IPOPT_SSRR) { 1011 type = ICMP_UNREACH; 1012 code = ICMP_UNREACH_SRCFAIL; 1013 goto bad; 1014 } 1015 if (!ip_dosourceroute) 1016 goto nosourcerouting; 1017 /* 1018 * Loose routing, and not at next destination 1019 * yet; nothing to do except forward. 1020 */ 1021 break; 1022 } 1023 off--; /* 0 origin */ 1024 if (off > optlen - sizeof(struct in_addr)) { 1025 /* 1026 * End of source route. Should be for us. 1027 */ 1028 if (!ip_acceptsourceroute) 1029 goto nosourcerouting; 1030 save_rte(cp, ip->ip_src); 1031 break; 1032 } 1033 1034 if (!ip_dosourceroute) { 1035 if (ipforwarding) { 1036 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1037 /* 1038 * Acting as a router, so generate ICMP 1039 */ 1040nosourcerouting: 1041 strcpy(buf, inet_ntoa(ip->ip_dst)); 1042 log(LOG_WARNING, 1043 "attempted source route from %s to %s\n", 1044 inet_ntoa(ip->ip_src), buf); 1045 type = ICMP_UNREACH; 1046 code = ICMP_UNREACH_SRCFAIL; 1047 goto bad; 1048 } else { 1049 /* 1050 * Not acting as a router, so silently drop. 1051 */ 1052 ipstat.ips_cantforward++; 1053 m_freem(m); 1054 return (1); 1055 } 1056 } 1057 1058 /* 1059 * locate outgoing interface 1060 */ 1061 (void)memcpy(&ipaddr.sin_addr, cp + off, 1062 sizeof(ipaddr.sin_addr)); 1063 1064 if (opt == IPOPT_SSRR) { 1065#define INA struct in_ifaddr * 1066#define SA struct sockaddr * 1067 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1068 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1069 } else 1070 ia = ip_rtaddr(ipaddr.sin_addr); 1071 if (ia == 0) { 1072 type = ICMP_UNREACH; 1073 code = ICMP_UNREACH_SRCFAIL; 1074 goto bad; 1075 } 1076 ip->ip_dst = ipaddr.sin_addr; 1077 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1078 sizeof(struct in_addr)); 1079 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1080 /* 1081 * Let ip_intr's mcast routing check handle mcast pkts 1082 */ 1083 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1084 break; 1085 1086 case IPOPT_RR: 1087 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1088 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1089 goto bad; 1090 } 1091 /* 1092 * If no space remains, ignore. 1093 */ 1094 off--; /* 0 origin */ 1095 if (off > optlen - sizeof(struct in_addr)) 1096 break; 1097 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1098 sizeof(ipaddr.sin_addr)); 1099 /* 1100 * locate outgoing interface; if we're the destination, 1101 * use the incoming interface (should be same). 1102 */ 1103 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1104 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1105 type = ICMP_UNREACH; 1106 code = ICMP_UNREACH_HOST; 1107 goto bad; 1108 } 1109 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1110 sizeof(struct in_addr)); 1111 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1112 break; 1113 1114 case IPOPT_TS: 1115 code = cp - (u_char *)ip; 1116 ipt = (struct ip_timestamp *)cp; 1117 if (ipt->ipt_len < 5) 1118 goto bad; 1119 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1120 if (++ipt->ipt_oflw == 0) 1121 goto bad; 1122 break; 1123 } 1124 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1125 switch (ipt->ipt_flg) { 1126 1127 case IPOPT_TS_TSONLY: 1128 break; 1129 1130 case IPOPT_TS_TSANDADDR: 1131 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1132 sizeof(struct in_addr) > ipt->ipt_len) 1133 goto bad; 1134 ipaddr.sin_addr = dst; 1135 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1136 m->m_pkthdr.rcvif); 1137 if (ia == 0) 1138 continue; 1139 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1140 sizeof(struct in_addr)); 1141 ipt->ipt_ptr += sizeof(struct in_addr); 1142 break; 1143 1144 case IPOPT_TS_PRESPEC: 1145 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1146 sizeof(struct in_addr) > ipt->ipt_len) 1147 goto bad; 1148 (void)memcpy(&ipaddr.sin_addr, sin, 1149 sizeof(struct in_addr)); 1150 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1151 continue; 1152 ipt->ipt_ptr += sizeof(struct in_addr); 1153 break; 1154 1155 default: 1156 goto bad; 1157 } 1158 ntime = iptime(); 1159 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1160 sizeof(n_time)); 1161 ipt->ipt_ptr += sizeof(n_time); 1162 } 1163 } 1164 if (forward && ipforwarding) { 1165 ip_forward(m, 1); 1166 return (1); 1167 } 1168 return (0); 1169bad: 1170 ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2; /* XXX icmp_error adds in hdr length */ 1171 icmp_error(m, type, code, 0, 0); 1172 ipstat.ips_badoptions++; 1173 return (1); 1174} 1175 1176/* 1177 * Given address of next destination (final or next hop), 1178 * return internet address info of interface to be used to get there. 1179 */ 1180static struct in_ifaddr * 1181ip_rtaddr(dst) 1182 struct in_addr dst; 1183{ 1184 register struct sockaddr_in *sin; 1185 1186 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1187 1188 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1189 if (ipforward_rt.ro_rt) { 1190 RTFREE(ipforward_rt.ro_rt); 1191 ipforward_rt.ro_rt = 0; 1192 } 1193 sin->sin_family = AF_INET; 1194 sin->sin_len = sizeof(*sin); 1195 sin->sin_addr = dst; 1196 1197 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1198 } 1199 if (ipforward_rt.ro_rt == 0) 1200 return ((struct in_ifaddr *)0); 1201 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1202} 1203 1204/* 1205 * Save incoming source route for use in replies, 1206 * to be picked up later by ip_srcroute if the receiver is interested. 1207 */ 1208void 1209save_rte(option, dst) 1210 u_char *option; 1211 struct in_addr dst; 1212{ 1213 unsigned olen; 1214 1215 olen = option[IPOPT_OLEN]; 1216#ifdef DIAGNOSTIC 1217 if (ipprintfs) 1218 printf("save_rte: olen %d\n", olen); 1219#endif 1220 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1221 return; 1222 bcopy(option, ip_srcrt.srcopt, olen); 1223 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1224 ip_srcrt.dst = dst; 1225} 1226 1227/* 1228 * Retrieve incoming source route for use in replies, 1229 * in the same form used by setsockopt. 1230 * The first hop is placed before the options, will be removed later. 1231 */ 1232struct mbuf * 1233ip_srcroute() 1234{ 1235 register struct in_addr *p, *q; 1236 register struct mbuf *m; 1237 1238 if (ip_nhops == 0) 1239 return ((struct mbuf *)0); 1240 m = m_get(M_DONTWAIT, MT_HEADER); 1241 if (m == 0) 1242 return ((struct mbuf *)0); 1243 1244#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1245 1246 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1247 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1248 OPTSIZ; 1249#ifdef DIAGNOSTIC 1250 if (ipprintfs) 1251 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1252#endif 1253 1254 /* 1255 * First save first hop for return route 1256 */ 1257 p = &ip_srcrt.route[ip_nhops - 1]; 1258 *(mtod(m, struct in_addr *)) = *p--; 1259#ifdef DIAGNOSTIC 1260 if (ipprintfs) 1261 printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr)); 1262#endif 1263 1264 /* 1265 * Copy option fields and padding (nop) to mbuf. 1266 */ 1267 ip_srcrt.nop = IPOPT_NOP; 1268 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1269 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1270 &ip_srcrt.nop, OPTSIZ); 1271 q = (struct in_addr *)(mtod(m, caddr_t) + 1272 sizeof(struct in_addr) + OPTSIZ); 1273#undef OPTSIZ 1274 /* 1275 * Record return path as an IP source route, 1276 * reversing the path (pointers are now aligned). 1277 */ 1278 while (p >= ip_srcrt.route) { 1279#ifdef DIAGNOSTIC 1280 if (ipprintfs) 1281 printf(" %lx", ntohl(q->s_addr)); 1282#endif 1283 *q++ = *p--; 1284 } 1285 /* 1286 * Last hop goes to final destination. 1287 */ 1288 *q = ip_srcrt.dst; 1289#ifdef DIAGNOSTIC 1290 if (ipprintfs) 1291 printf(" %lx\n", ntohl(q->s_addr)); 1292#endif 1293 return (m); 1294} 1295 1296/* 1297 * Strip out IP options, at higher 1298 * level protocol in the kernel. 1299 * Second argument is buffer to which options 1300 * will be moved, and return value is their length. 1301 * XXX should be deleted; last arg currently ignored. 1302 */ 1303void 1304ip_stripoptions(m, mopt) 1305 register struct mbuf *m; 1306 struct mbuf *mopt; 1307{ 1308 register int i; 1309 struct ip *ip = mtod(m, struct ip *); 1310 register caddr_t opts; 1311 int olen; 1312 1313 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1314 opts = (caddr_t)(ip + 1); 1315 i = m->m_len - (sizeof (struct ip) + olen); 1316 bcopy(opts + olen, opts, (unsigned)i); 1317 m->m_len -= olen; 1318 if (m->m_flags & M_PKTHDR) 1319 m->m_pkthdr.len -= olen; 1320 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1321} 1322 1323u_char inetctlerrmap[PRC_NCMDS] = { 1324 0, 0, 0, 0, 1325 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1326 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1327 EMSGSIZE, EHOSTUNREACH, 0, 0, 1328 0, 0, 0, 0, 1329 ENOPROTOOPT 1330}; 1331 1332/* 1333 * Forward a packet. If some error occurs return the sender 1334 * an icmp packet. Note we can't always generate a meaningful 1335 * icmp message because icmp doesn't have a large enough repertoire 1336 * of codes and types. 1337 * 1338 * If not forwarding, just drop the packet. This could be confusing 1339 * if ipforwarding was zero but some routing protocol was advancing 1340 * us as a gateway to somewhere. However, we must let the routing 1341 * protocol deal with that. 1342 * 1343 * The srcrt parameter indicates whether the packet is being forwarded 1344 * via a source route. 1345 */ 1346static void 1347ip_forward(m, srcrt) 1348 struct mbuf *m; 1349 int srcrt; 1350{ 1351 register struct ip *ip = mtod(m, struct ip *); 1352 register struct sockaddr_in *sin; 1353 register struct rtentry *rt; 1354 int error, type = 0, code = 0; 1355 struct mbuf *mcopy; 1356 n_long dest; 1357 struct ifnet *destifp; 1358 1359 dest = 0; 1360#ifdef DIAGNOSTIC 1361 if (ipprintfs) 1362 printf("forward: src %lx dst %lx ttl %x\n", 1363 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1364 ip->ip_ttl); 1365#endif 1366 1367 1368 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) { 1369 ipstat.ips_cantforward++; 1370 m_freem(m); 1371 return; 1372 } 1373 HTONS(ip->ip_id); 1374 if (ip->ip_ttl <= IPTTLDEC) { 1375 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1376 return; 1377 } 1378 ip->ip_ttl -= IPTTLDEC; 1379 1380 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1381 if ((rt = ipforward_rt.ro_rt) == 0 || 1382 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1383 if (ipforward_rt.ro_rt) { 1384 RTFREE(ipforward_rt.ro_rt); 1385 ipforward_rt.ro_rt = 0; 1386 } 1387 sin->sin_family = AF_INET; 1388 sin->sin_len = sizeof(*sin); 1389 sin->sin_addr = ip->ip_dst; 1390 1391 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1392 if (ipforward_rt.ro_rt == 0) { 1393 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1394 return; 1395 } 1396 rt = ipforward_rt.ro_rt; 1397 } 1398 1399 /* 1400 * Save at most 64 bytes of the packet in case 1401 * we need to generate an ICMP message to the src. 1402 */ 1403 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1404 1405 /* 1406 * If forwarding packet using same interface that it came in on, 1407 * perhaps should send a redirect to sender to shortcut a hop. 1408 * Only send redirect if source is sending directly to us, 1409 * and if packet was not source routed (or has any options). 1410 * Also, don't send redirect if forwarding using a default route 1411 * or a route modified by a redirect. 1412 */ 1413#define satosin(sa) ((struct sockaddr_in *)(sa)) 1414 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1415 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1416 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1417 ipsendredirects && !srcrt) { 1418#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1419 u_long src = ntohl(ip->ip_src.s_addr); 1420 1421 if (RTA(rt) && 1422 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1423 if (rt->rt_flags & RTF_GATEWAY) 1424 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1425 else 1426 dest = ip->ip_dst.s_addr; 1427 /* Router requirements says to only send host redirects */ 1428 type = ICMP_REDIRECT; 1429 code = ICMP_REDIRECT_HOST; 1430#ifdef DIAGNOSTIC 1431 if (ipprintfs) 1432 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1433#endif 1434 } 1435 } 1436 1437 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1438 IP_FORWARDING, 0); 1439 if (error) 1440 ipstat.ips_cantforward++; 1441 else { 1442 ipstat.ips_forward++; 1443 if (type) 1444 ipstat.ips_redirectsent++; 1445 else { 1446 if (mcopy) { 1447 ipflow_create(&ipforward_rt, mcopy); 1448 m_freem(mcopy); 1449 } 1450 return; 1451 } 1452 } 1453 if (mcopy == NULL) 1454 return; 1455 destifp = NULL; 1456 1457 switch (error) { 1458 1459 case 0: /* forwarded, but need redirect */ 1460 /* type, code set above */ 1461 break; 1462 1463 case ENETUNREACH: /* shouldn't happen, checked above */ 1464 case EHOSTUNREACH: 1465 case ENETDOWN: 1466 case EHOSTDOWN: 1467 default: 1468 type = ICMP_UNREACH; 1469 code = ICMP_UNREACH_HOST; 1470 break; 1471 1472 case EMSGSIZE: 1473 type = ICMP_UNREACH; 1474 code = ICMP_UNREACH_NEEDFRAG; 1475 if (ipforward_rt.ro_rt) 1476 destifp = ipforward_rt.ro_rt->rt_ifp; 1477 ipstat.ips_cantfrag++; 1478 break; 1479 1480 case ENOBUFS: 1481 type = ICMP_SOURCEQUENCH; 1482 code = 0; 1483 break; 1484 } 1485 icmp_error(mcopy, type, code, dest, destifp); 1486} 1487 1488void 1489ip_savecontrol(inp, mp, ip, m) 1490 register struct inpcb *inp; 1491 register struct mbuf **mp; 1492 register struct ip *ip; 1493 register struct mbuf *m; 1494{ 1495 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1496 struct timeval tv; 1497 1498 microtime(&tv); 1499 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1500 SCM_TIMESTAMP, SOL_SOCKET); 1501 if (*mp) 1502 mp = &(*mp)->m_next; 1503 } 1504 if (inp->inp_flags & INP_RECVDSTADDR) { 1505 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1506 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1507 if (*mp) 1508 mp = &(*mp)->m_next; 1509 } 1510#ifdef notyet 1511 /* XXX 1512 * Moving these out of udp_input() made them even more broken 1513 * than they already were. 1514 */ 1515 /* options were tossed already */ 1516 if (inp->inp_flags & INP_RECVOPTS) { 1517 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1518 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1519 if (*mp) 1520 mp = &(*mp)->m_next; 1521 } 1522 /* ip_srcroute doesn't do what we want here, need to fix */ 1523 if (inp->inp_flags & INP_RECVRETOPTS) { 1524 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1525 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1526 if (*mp) 1527 mp = &(*mp)->m_next; 1528 } 1529#endif 1530 if (inp->inp_flags & INP_RECVIF) { 1531 struct ifnet *ifp; 1532 struct sdlbuf { 1533 struct sockaddr_dl sdl; 1534 u_char pad[32]; 1535 } sdlbuf; 1536 struct sockaddr_dl *sdp; 1537 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1538 1539 if (((ifp = m->m_pkthdr.rcvif)) 1540 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1541 sdp = (struct sockaddr_dl *)(ifnet_addrs 1542 [ifp->if_index - 1]->ifa_addr); 1543 /* 1544 * Change our mind and don't try copy. 1545 */ 1546 if ((sdp->sdl_family != AF_LINK) 1547 || (sdp->sdl_len > sizeof(sdlbuf))) { 1548 goto makedummy; 1549 } 1550 bcopy(sdp, sdl2, sdp->sdl_len); 1551 } else { 1552makedummy: 1553 sdl2->sdl_len 1554 = offsetof(struct sockaddr_dl, sdl_data[0]); 1555 sdl2->sdl_family = AF_LINK; 1556 sdl2->sdl_index = 0; 1557 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1558 } 1559 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1560 IP_RECVIF, IPPROTO_IP); 1561 if (*mp) 1562 mp = &(*mp)->m_next; 1563 } 1564} 1565 1566int 1567ip_rsvp_init(struct socket *so) 1568{ 1569 if (so->so_type != SOCK_RAW || 1570 so->so_proto->pr_protocol != IPPROTO_RSVP) 1571 return EOPNOTSUPP; 1572 1573 if (ip_rsvpd != NULL) 1574 return EADDRINUSE; 1575 1576 ip_rsvpd = so; 1577 /* 1578 * This may seem silly, but we need to be sure we don't over-increment 1579 * the RSVP counter, in case something slips up. 1580 */ 1581 if (!ip_rsvp_on) { 1582 ip_rsvp_on = 1; 1583 rsvp_on++; 1584 } 1585 1586 return 0; 1587} 1588 1589int 1590ip_rsvp_done(void) 1591{ 1592 ip_rsvpd = NULL; 1593 /* 1594 * This may seem silly, but we need to be sure we don't over-decrement 1595 * the RSVP counter, in case something slips up. 1596 */ 1597 if (ip_rsvp_on) { 1598 ip_rsvp_on = 0; 1599 rsvp_on--; 1600 } 1601 return 0; 1602} 1603