ip_input.c revision 12820
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 1.29 1995/12/06 23:37:34 bde Exp $ 35 */ 36 37#include <sys/param.h> 38#include <sys/systm.h> 39#include <sys/malloc.h> 40#include <sys/mbuf.h> 41#include <sys/domain.h> 42#include <sys/protosw.h> 43#include <sys/socket.h> 44#include <sys/errno.h> 45#include <sys/time.h> 46#include <sys/kernel.h> 47#include <sys/syslog.h> 48#include <sys/sysctl.h> 49 50#include <net/if.h> 51#include <net/route.h> 52#include <net/netisr.h> 53 54#include <netinet/in.h> 55#include <netinet/in_systm.h> 56#include <netinet/in_var.h> 57#include <netinet/ip.h> 58#include <netinet/in_pcb.h> 59#include <netinet/in_var.h> 60#include <netinet/ip_var.h> 61#include <netinet/ip_icmp.h> 62 63#include <netinet/ip_fw.h> 64 65#include <sys/socketvar.h> 66int rsvp_on = 0; 67static int ip_rsvp_on; 68struct socket *ip_rsvpd; 69 70#ifndef IPFORWARDING 71#ifdef GATEWAY 72#define IPFORWARDING 1 /* forward IP packets not for us */ 73#else /* GATEWAY */ 74#define IPFORWARDING 0 /* don't forward IP packets not for us */ 75#endif /* GATEWAY */ 76#endif /* IPFORWARDING */ 77#ifndef IPSENDREDIRECTS 78#define IPSENDREDIRECTS 1 79#endif 80#ifndef DIRECTED_BROADCAST 81#define DIRECTED_BROADCAST 0 82#endif 83 84static int ipforwarding = IPFORWARDING; 85SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 86 &ipforwarding, 0, ""); 87 88static int ipsendredirects = IPSENDREDIRECTS; 89SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 90 &ipsendredirects, 0, ""); 91 92static int ipdirbroadcast = DIRECTED_BROADCAST; 93SYSCTL_INT(_net_inet_ip, IPCTL_DIRECTEDBROADCAST, directed_broadcast, 94 CTLFLAG_RW, &ipdirbroadcast, 0, ""); 95 96int ip_defttl = IPDEFTTL; 97SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 98 &ip_defttl, 0, ""); 99 100static int ip_dosourceroute = 0; 101SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 102 &ip_dosourceroute, 0, ""); 103#ifdef DIAGNOSTIC 104static int ipprintfs = 0; 105#endif 106 107extern struct domain inetdomain; 108extern struct protosw inetsw[]; 109u_char ip_protox[IPPROTO_MAX]; 110static int ipqmaxlen = IFQ_MAXLEN; 111struct in_ifaddr *in_ifaddr; /* first inet address */ 112struct ifqueue ipintrq; 113SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RD, 114 &ipintrq.ifq_maxlen, 0, ""); 115SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 116 &ipintrq.ifq_drops, 0, ""); 117 118struct ipstat ipstat; 119static struct ipq ipq; 120 121#ifdef IPCTL_DEFMTU 122SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 123 &ip_mtu, 0, ""); 124#endif 125 126/* 127 * We need to save the IP options in case a protocol wants to respond 128 * to an incoming packet over the same route if the packet got here 129 * using IP source routing. This allows connection establishment and 130 * maintenance when the remote end is on a network that is not known 131 * to us. 132 */ 133static int ip_nhops = 0; 134static struct ip_srcrt { 135 struct in_addr dst; /* final destination */ 136 char nop; /* one NOP to align */ 137 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 138 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 139} ip_srcrt; 140 141static void save_rte __P((u_char *, struct in_addr)); 142static void ip_deq __P((struct ipasfrag *)); 143static int ip_dooptions __P((struct mbuf *)); 144static void ip_enq __P((struct ipasfrag *, struct ipasfrag *)); 145static void ip_forward __P((struct mbuf *, int)); 146static void ip_freef __P((struct ipq *)); 147static struct ip * 148 ip_reass __P((struct ipasfrag *, struct ipq *)); 149static struct in_ifaddr * 150 ip_rtaddr __P((struct in_addr)); 151static void ipintr __P((void)); 152/* 153 * IP initialization: fill in IP protocol switch table. 154 * All protocols not implemented in kernel go to raw IP protocol handler. 155 */ 156void 157ip_init() 158{ 159 register struct protosw *pr; 160 register int i; 161 162 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 163 if (pr == 0) 164 panic("ip_init"); 165 for (i = 0; i < IPPROTO_MAX; i++) 166 ip_protox[i] = pr - inetsw; 167 for (pr = inetdomain.dom_protosw; 168 pr < inetdomain.dom_protoswNPROTOSW; pr++) 169 if (pr->pr_domain->dom_family == PF_INET && 170 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 171 ip_protox[pr->pr_protocol] = pr - inetsw; 172 ipq.next = ipq.prev = &ipq; 173 ip_id = time.tv_sec & 0xffff; 174 ipintrq.ifq_maxlen = ipqmaxlen; 175} 176 177static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 178static struct route ipforward_rt; 179 180/* 181 * Ip input routine. Checksum and byte swap header. If fragmented 182 * try to reassemble. Process options. Pass to next level. 183 */ 184static void 185ipintr(void) 186{ 187 register struct ip *ip; 188 register struct mbuf *m; 189 register struct ipq *fp; 190 register struct in_ifaddr *ia; 191 int hlen, s; 192 193next: 194 /* 195 * Get next datagram off input queue and get IP header 196 * in first mbuf. 197 */ 198 s = splimp(); 199 IF_DEQUEUE(&ipintrq, m); 200 splx(s); 201 if (m == 0) 202 return; 203#ifdef DIAGNOSTIC 204 if ((m->m_flags & M_PKTHDR) == 0) 205 panic("ipintr no HDR"); 206#endif 207 /* 208 * If no IP addresses have been set yet but the interfaces 209 * are receiving, can't do anything with incoming packets yet. 210 */ 211 if (in_ifaddr == NULL) 212 goto bad; 213 ipstat.ips_total++; 214 if (m->m_len < sizeof (struct ip) && 215 (m = m_pullup(m, sizeof (struct ip))) == 0) { 216 ipstat.ips_toosmall++; 217 goto next; 218 } 219 ip = mtod(m, struct ip *); 220 if (ip->ip_v != IPVERSION) { 221 ipstat.ips_badvers++; 222 goto bad; 223 } 224 hlen = ip->ip_hl << 2; 225 if (hlen < sizeof(struct ip)) { /* minimum header length */ 226 ipstat.ips_badhlen++; 227 goto bad; 228 } 229 if (hlen > m->m_len) { 230 if ((m = m_pullup(m, hlen)) == 0) { 231 ipstat.ips_badhlen++; 232 goto next; 233 } 234 ip = mtod(m, struct ip *); 235 } 236 ip->ip_sum = in_cksum(m, hlen); 237 if (ip->ip_sum) { 238 ipstat.ips_badsum++; 239 goto bad; 240 } 241 242 /* 243 * Convert fields to host representation. 244 */ 245 NTOHS(ip->ip_len); 246 if (ip->ip_len < hlen) { 247 ipstat.ips_badlen++; 248 goto bad; 249 } 250 NTOHS(ip->ip_id); 251 NTOHS(ip->ip_off); 252 253 /* 254 * Check that the amount of data in the buffers 255 * is as at least much as the IP header would have us expect. 256 * Trim mbufs if longer than we expect. 257 * Drop packet if shorter than we expect. 258 */ 259 if (m->m_pkthdr.len < ip->ip_len) { 260 ipstat.ips_tooshort++; 261 goto bad; 262 } 263 if (m->m_pkthdr.len > ip->ip_len) { 264 if (m->m_len == m->m_pkthdr.len) { 265 m->m_len = ip->ip_len; 266 m->m_pkthdr.len = ip->ip_len; 267 } else 268 m_adj(m, ip->ip_len - m->m_pkthdr.len); 269 } 270 /* 271 * IpHack's section. 272 * Right now when no processing on packet has done 273 * and it is still fresh out of network we do our black 274 * deals with it. 275 * - Firewall: deny/allow 276 * - Wrap: fake packet's addr/port <unimpl.> 277 * - Encapsulate: put it in another IP and send out. <unimp.> 278 */ 279 280 if (ip_fw_chk_ptr!=NULL) 281 if (!(*ip_fw_chk_ptr)(m,ip,m->m_pkthdr.rcvif,ip_fw_chain) ) { 282 goto next; 283 } 284 285 /* 286 * Process options and, if not destined for us, 287 * ship it on. ip_dooptions returns 1 when an 288 * error was detected (causing an icmp message 289 * to be sent and the original packet to be freed). 290 */ 291 ip_nhops = 0; /* for source routed packets */ 292 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 293 goto next; 294 295 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 296 * matter if it is destined to another node, or whether it is 297 * a multicast one, RSVP wants it! and prevents it from being forwarded 298 * anywhere else. Also checks if the rsvp daemon is running before 299 * grabbing the packet. 300 */ 301 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 302 goto ours; 303 304 /* 305 * Check our list of addresses, to see if the packet is for us. 306 */ 307 for (ia = in_ifaddr; ia; ia = ia->ia_next) { 308#define satosin(sa) ((struct sockaddr_in *)(sa)) 309 310 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 311 goto ours; 312 if ((!ipdirbroadcast || ia->ia_ifp == m->m_pkthdr.rcvif) && 313 (ia->ia_ifp->if_flags & IFF_BROADCAST)) { 314 u_long t; 315 316 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 317 ip->ip_dst.s_addr) 318 goto ours; 319 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 320 goto ours; 321 /* 322 * Look for all-0's host part (old broadcast addr), 323 * either for subnet or net. 324 */ 325 t = ntohl(ip->ip_dst.s_addr); 326 if (t == ia->ia_subnet) 327 goto ours; 328 if (t == ia->ia_net) 329 goto ours; 330 } 331 } 332 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 333 struct in_multi *inm; 334 if (ip_mrouter) { 335 /* 336 * If we are acting as a multicast router, all 337 * incoming multicast packets are passed to the 338 * kernel-level multicast forwarding function. 339 * The packet is returned (relatively) intact; if 340 * ip_mforward() returns a non-zero value, the packet 341 * must be discarded, else it may be accepted below. 342 * 343 * (The IP ident field is put in the same byte order 344 * as expected when ip_mforward() is called from 345 * ip_output().) 346 */ 347 ip->ip_id = htons(ip->ip_id); 348 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 349 ipstat.ips_cantforward++; 350 m_freem(m); 351 goto next; 352 } 353 ip->ip_id = ntohs(ip->ip_id); 354 355 /* 356 * The process-level routing demon needs to receive 357 * all multicast IGMP packets, whether or not this 358 * host belongs to their destination groups. 359 */ 360 if (ip->ip_p == IPPROTO_IGMP) 361 goto ours; 362 ipstat.ips_forward++; 363 } 364 /* 365 * See if we belong to the destination multicast group on the 366 * arrival interface. 367 */ 368 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 369 if (inm == NULL) { 370 ipstat.ips_cantforward++; 371 m_freem(m); 372 goto next; 373 } 374 goto ours; 375 } 376 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 377 goto ours; 378 if (ip->ip_dst.s_addr == INADDR_ANY) 379 goto ours; 380 381 /* 382 * Not for us; forward if possible and desirable. 383 */ 384 if (ipforwarding == 0) { 385 ipstat.ips_cantforward++; 386 m_freem(m); 387 } else 388 ip_forward(m, 0); 389 goto next; 390 391ours: 392 393 /* 394 * If packet came to us we count it... 395 * This way we count all incoming packets which has 396 * not been forwarded... 397 * Do not convert ip_len to host byte order when 398 * counting,ppl already made it for us before.. 399 */ 400 if (ip_acct_cnt_ptr!=NULL) 401 (*ip_acct_cnt_ptr)(ip,m->m_pkthdr.rcvif,ip_acct_chain,0); 402 403 /* 404 * If offset or IP_MF are set, must reassemble. 405 * Otherwise, nothing need be done. 406 * (We could look in the reassembly queue to see 407 * if the packet was previously fragmented, 408 * but it's not worth the time; just let them time out.) 409 */ 410 if (ip->ip_off &~ IP_DF) { 411 if (m->m_flags & M_EXT) { /* XXX */ 412 if ((m = m_pullup(m, sizeof (struct ip))) == 0) { 413 ipstat.ips_toosmall++; 414 goto next; 415 } 416 ip = mtod(m, struct ip *); 417 } 418 /* 419 * Look for queue of fragments 420 * of this datagram. 421 */ 422 for (fp = ipq.next; fp != &ipq; fp = fp->next) 423 if (ip->ip_id == fp->ipq_id && 424 ip->ip_src.s_addr == fp->ipq_src.s_addr && 425 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 426 ip->ip_p == fp->ipq_p) 427 goto found; 428 fp = 0; 429found: 430 431 /* 432 * Adjust ip_len to not reflect header, 433 * set ip_mff if more fragments are expected, 434 * convert offset of this to bytes. 435 */ 436 ip->ip_len -= hlen; 437 ((struct ipasfrag *)ip)->ipf_mff &= ~1; 438 if (ip->ip_off & IP_MF) 439 ((struct ipasfrag *)ip)->ipf_mff |= 1; 440 ip->ip_off <<= 3; 441 442 /* 443 * If datagram marked as having more fragments 444 * or if this is not the first fragment, 445 * attempt reassembly; if it succeeds, proceed. 446 */ 447 if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) { 448 ipstat.ips_fragments++; 449 ip = ip_reass((struct ipasfrag *)ip, fp); 450 if (ip == 0) 451 goto next; 452 ipstat.ips_reassembled++; 453 m = dtom(ip); 454 } else 455 if (fp) 456 ip_freef(fp); 457 } else 458 ip->ip_len -= hlen; 459 460 /* 461 * Switch out to protocol's input routine. 462 */ 463 ipstat.ips_delivered++; 464 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 465 goto next; 466bad: 467 m_freem(m); 468 goto next; 469} 470 471NETISR_SET(NETISR_IP, ipintr); 472 473/* 474 * Take incoming datagram fragment and try to 475 * reassemble it into whole datagram. If a chain for 476 * reassembly of this datagram already exists, then it 477 * is given as fp; otherwise have to make a chain. 478 */ 479static struct ip * 480ip_reass(ip, fp) 481 register struct ipasfrag *ip; 482 register struct ipq *fp; 483{ 484 register struct mbuf *m = dtom(ip); 485 register struct ipasfrag *q; 486 struct mbuf *t; 487 int hlen = ip->ip_hl << 2; 488 int i, next; 489 490 /* 491 * Presence of header sizes in mbufs 492 * would confuse code below. 493 */ 494 m->m_data += hlen; 495 m->m_len -= hlen; 496 497 /* 498 * If first fragment to arrive, create a reassembly queue. 499 */ 500 if (fp == 0) { 501 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 502 goto dropfrag; 503 fp = mtod(t, struct ipq *); 504 insque(fp, &ipq); 505 fp->ipq_ttl = IPFRAGTTL; 506 fp->ipq_p = ip->ip_p; 507 fp->ipq_id = ip->ip_id; 508 fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp; 509 fp->ipq_src = ((struct ip *)ip)->ip_src; 510 fp->ipq_dst = ((struct ip *)ip)->ip_dst; 511 q = (struct ipasfrag *)fp; 512 goto insert; 513 } 514 515 /* 516 * Find a segment which begins after this one does. 517 */ 518 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) 519 if (q->ip_off > ip->ip_off) 520 break; 521 522 /* 523 * If there is a preceding segment, it may provide some of 524 * our data already. If so, drop the data from the incoming 525 * segment. If it provides all of our data, drop us. 526 */ 527 if (q->ipf_prev != (struct ipasfrag *)fp) { 528 i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off; 529 if (i > 0) { 530 if (i >= ip->ip_len) 531 goto dropfrag; 532 m_adj(dtom(ip), i); 533 ip->ip_off += i; 534 ip->ip_len -= i; 535 } 536 } 537 538 /* 539 * While we overlap succeeding segments trim them or, 540 * if they are completely covered, dequeue them. 541 */ 542 while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) { 543 i = (ip->ip_off + ip->ip_len) - q->ip_off; 544 if (i < q->ip_len) { 545 q->ip_len -= i; 546 q->ip_off += i; 547 m_adj(dtom(q), i); 548 break; 549 } 550 q = q->ipf_next; 551 m_freem(dtom(q->ipf_prev)); 552 ip_deq(q->ipf_prev); 553 } 554 555insert: 556 /* 557 * Stick new segment in its place; 558 * check for complete reassembly. 559 */ 560 ip_enq(ip, q->ipf_prev); 561 next = 0; 562 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) { 563 if (q->ip_off != next) 564 return (0); 565 next += q->ip_len; 566 } 567 if (q->ipf_prev->ipf_mff & 1) 568 return (0); 569 570 /* 571 * Reassembly is complete; concatenate fragments. 572 */ 573 q = fp->ipq_next; 574 m = dtom(q); 575 t = m->m_next; 576 m->m_next = 0; 577 m_cat(m, t); 578 q = q->ipf_next; 579 while (q != (struct ipasfrag *)fp) { 580 t = dtom(q); 581 q = q->ipf_next; 582 m_cat(m, t); 583 } 584 585 /* 586 * Create header for new ip packet by 587 * modifying header of first packet; 588 * dequeue and discard fragment reassembly header. 589 * Make header visible. 590 */ 591 ip = fp->ipq_next; 592 ip->ip_len = next; 593 ip->ipf_mff &= ~1; 594 ((struct ip *)ip)->ip_src = fp->ipq_src; 595 ((struct ip *)ip)->ip_dst = fp->ipq_dst; 596 remque(fp); 597 (void) m_free(dtom(fp)); 598 m = dtom(ip); 599 m->m_len += (ip->ip_hl << 2); 600 m->m_data -= (ip->ip_hl << 2); 601 /* some debugging cruft by sklower, below, will go away soon */ 602 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 603 register int plen = 0; 604 for (t = m; m; m = m->m_next) 605 plen += m->m_len; 606 t->m_pkthdr.len = plen; 607 } 608 return ((struct ip *)ip); 609 610dropfrag: 611 ipstat.ips_fragdropped++; 612 m_freem(m); 613 return (0); 614} 615 616/* 617 * Free a fragment reassembly header and all 618 * associated datagrams. 619 */ 620static void 621ip_freef(fp) 622 struct ipq *fp; 623{ 624 register struct ipasfrag *q, *p; 625 626 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) { 627 p = q->ipf_next; 628 ip_deq(q); 629 m_freem(dtom(q)); 630 } 631 remque(fp); 632 (void) m_free(dtom(fp)); 633} 634 635/* 636 * Put an ip fragment on a reassembly chain. 637 * Like insque, but pointers in middle of structure. 638 */ 639static void 640ip_enq(p, prev) 641 register struct ipasfrag *p, *prev; 642{ 643 644 p->ipf_prev = prev; 645 p->ipf_next = prev->ipf_next; 646 prev->ipf_next->ipf_prev = p; 647 prev->ipf_next = p; 648} 649 650/* 651 * To ip_enq as remque is to insque. 652 */ 653static void 654ip_deq(p) 655 register struct ipasfrag *p; 656{ 657 658 p->ipf_prev->ipf_next = p->ipf_next; 659 p->ipf_next->ipf_prev = p->ipf_prev; 660} 661 662/* 663 * IP timer processing; 664 * if a timer expires on a reassembly 665 * queue, discard it. 666 */ 667void 668ip_slowtimo() 669{ 670 register struct ipq *fp; 671 int s = splnet(); 672 673 fp = ipq.next; 674 if (fp == 0) { 675 splx(s); 676 return; 677 } 678 while (fp != &ipq) { 679 --fp->ipq_ttl; 680 fp = fp->next; 681 if (fp->prev->ipq_ttl == 0) { 682 ipstat.ips_fragtimeout++; 683 ip_freef(fp->prev); 684 } 685 } 686 splx(s); 687} 688 689/* 690 * Drain off all datagram fragments. 691 */ 692void 693ip_drain() 694{ 695 696 while (ipq.next != &ipq) { 697 ipstat.ips_fragdropped++; 698 ip_freef(ipq.next); 699 } 700} 701 702/* 703 * Do option processing on a datagram, 704 * possibly discarding it if bad options are encountered, 705 * or forwarding it if source-routed. 706 * Returns 1 if packet has been forwarded/freed, 707 * 0 if the packet should be processed further. 708 */ 709static int 710ip_dooptions(m) 711 struct mbuf *m; 712{ 713 register struct ip *ip = mtod(m, struct ip *); 714 register u_char *cp; 715 register struct ip_timestamp *ipt; 716 register struct in_ifaddr *ia; 717 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 718 struct in_addr *sin, dst; 719 n_time ntime; 720 721 dst = ip->ip_dst; 722 cp = (u_char *)(ip + 1); 723 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 724 for (; cnt > 0; cnt -= optlen, cp += optlen) { 725 opt = cp[IPOPT_OPTVAL]; 726 if (opt == IPOPT_EOL) 727 break; 728 if (opt == IPOPT_NOP) 729 optlen = 1; 730 else { 731 optlen = cp[IPOPT_OLEN]; 732 if (optlen <= 0 || optlen > cnt) { 733 code = &cp[IPOPT_OLEN] - (u_char *)ip; 734 goto bad; 735 } 736 } 737 switch (opt) { 738 739 default: 740 break; 741 742 /* 743 * Source routing with record. 744 * Find interface with current destination address. 745 * If none on this machine then drop if strictly routed, 746 * or do nothing if loosely routed. 747 * Record interface address and bring up next address 748 * component. If strictly routed make sure next 749 * address is on directly accessible net. 750 */ 751 case IPOPT_LSRR: 752 case IPOPT_SSRR: 753 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 754 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 755 goto bad; 756 } 757 ipaddr.sin_addr = ip->ip_dst; 758 ia = (struct in_ifaddr *) 759 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 760 if (ia == 0) { 761 if (opt == IPOPT_SSRR) { 762 type = ICMP_UNREACH; 763 code = ICMP_UNREACH_SRCFAIL; 764 goto bad; 765 } 766 /* 767 * Loose routing, and not at next destination 768 * yet; nothing to do except forward. 769 */ 770 break; 771 } 772 off--; /* 0 origin */ 773 if (off > optlen - sizeof(struct in_addr)) { 774 /* 775 * End of source route. Should be for us. 776 */ 777 save_rte(cp, ip->ip_src); 778 break; 779 } 780 781 if (!ip_dosourceroute) { 782 char buf[4*sizeof "123"]; 783 strcpy(buf, inet_ntoa(ip->ip_dst)); 784 785 log(LOG_WARNING, 786 "attempted source route from %s to %s\n", 787 inet_ntoa(ip->ip_src), buf); 788 type = ICMP_UNREACH; 789 code = ICMP_UNREACH_SRCFAIL; 790 goto bad; 791 } 792 793 /* 794 * locate outgoing interface 795 */ 796 (void)memcpy(&ipaddr.sin_addr, cp + off, 797 sizeof(ipaddr.sin_addr)); 798 799 if (opt == IPOPT_SSRR) { 800#define INA struct in_ifaddr * 801#define SA struct sockaddr * 802 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 803 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 804 } else 805 ia = ip_rtaddr(ipaddr.sin_addr); 806 if (ia == 0) { 807 type = ICMP_UNREACH; 808 code = ICMP_UNREACH_SRCFAIL; 809 goto bad; 810 } 811 ip->ip_dst = ipaddr.sin_addr; 812 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 813 sizeof(struct in_addr)); 814 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 815 /* 816 * Let ip_intr's mcast routing check handle mcast pkts 817 */ 818 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 819 break; 820 821 case IPOPT_RR: 822 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 823 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 824 goto bad; 825 } 826 /* 827 * If no space remains, ignore. 828 */ 829 off--; /* 0 origin */ 830 if (off > optlen - sizeof(struct in_addr)) 831 break; 832 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 833 sizeof(ipaddr.sin_addr)); 834 /* 835 * locate outgoing interface; if we're the destination, 836 * use the incoming interface (should be same). 837 */ 838 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 839 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 840 type = ICMP_UNREACH; 841 code = ICMP_UNREACH_HOST; 842 goto bad; 843 } 844 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 845 sizeof(struct in_addr)); 846 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 847 break; 848 849 case IPOPT_TS: 850 code = cp - (u_char *)ip; 851 ipt = (struct ip_timestamp *)cp; 852 if (ipt->ipt_len < 5) 853 goto bad; 854 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) { 855 if (++ipt->ipt_oflw == 0) 856 goto bad; 857 break; 858 } 859 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 860 switch (ipt->ipt_flg) { 861 862 case IPOPT_TS_TSONLY: 863 break; 864 865 case IPOPT_TS_TSANDADDR: 866 if (ipt->ipt_ptr + sizeof(n_time) + 867 sizeof(struct in_addr) > ipt->ipt_len) 868 goto bad; 869 ipaddr.sin_addr = dst; 870 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 871 m->m_pkthdr.rcvif); 872 if (ia == 0) 873 continue; 874 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 875 sizeof(struct in_addr)); 876 ipt->ipt_ptr += sizeof(struct in_addr); 877 break; 878 879 case IPOPT_TS_PRESPEC: 880 if (ipt->ipt_ptr + sizeof(n_time) + 881 sizeof(struct in_addr) > ipt->ipt_len) 882 goto bad; 883 (void)memcpy(&ipaddr.sin_addr, sin, 884 sizeof(struct in_addr)); 885 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 886 continue; 887 ipt->ipt_ptr += sizeof(struct in_addr); 888 break; 889 890 default: 891 goto bad; 892 } 893 ntime = iptime(); 894 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 895 sizeof(n_time)); 896 ipt->ipt_ptr += sizeof(n_time); 897 } 898 } 899 if (forward) { 900 ip_forward(m, 1); 901 return (1); 902 } 903 return (0); 904bad: 905 ip->ip_len -= ip->ip_hl << 2; /* XXX icmp_error adds in hdr length */ 906 icmp_error(m, type, code, 0, 0); 907 ipstat.ips_badoptions++; 908 return (1); 909} 910 911/* 912 * Given address of next destination (final or next hop), 913 * return internet address info of interface to be used to get there. 914 */ 915static struct in_ifaddr * 916ip_rtaddr(dst) 917 struct in_addr dst; 918{ 919 register struct sockaddr_in *sin; 920 921 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 922 923 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 924 if (ipforward_rt.ro_rt) { 925 RTFREE(ipforward_rt.ro_rt); 926 ipforward_rt.ro_rt = 0; 927 } 928 sin->sin_family = AF_INET; 929 sin->sin_len = sizeof(*sin); 930 sin->sin_addr = dst; 931 932 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 933 } 934 if (ipforward_rt.ro_rt == 0) 935 return ((struct in_ifaddr *)0); 936 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 937} 938 939/* 940 * Save incoming source route for use in replies, 941 * to be picked up later by ip_srcroute if the receiver is interested. 942 */ 943void 944save_rte(option, dst) 945 u_char *option; 946 struct in_addr dst; 947{ 948 unsigned olen; 949 950 olen = option[IPOPT_OLEN]; 951#ifdef DIAGNOSTIC 952 if (ipprintfs) 953 printf("save_rte: olen %d\n", olen); 954#endif 955 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 956 return; 957 (void)memcpy(ip_srcrt.srcopt, option, olen); 958 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 959 ip_srcrt.dst = dst; 960} 961 962/* 963 * Retrieve incoming source route for use in replies, 964 * in the same form used by setsockopt. 965 * The first hop is placed before the options, will be removed later. 966 */ 967struct mbuf * 968ip_srcroute() 969{ 970 register struct in_addr *p, *q; 971 register struct mbuf *m; 972 973 if (ip_nhops == 0) 974 return ((struct mbuf *)0); 975 m = m_get(M_DONTWAIT, MT_SOOPTS); 976 if (m == 0) 977 return ((struct mbuf *)0); 978 979#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 980 981 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 982 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 983 OPTSIZ; 984#ifdef DIAGNOSTIC 985 if (ipprintfs) 986 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 987#endif 988 989 /* 990 * First save first hop for return route 991 */ 992 p = &ip_srcrt.route[ip_nhops - 1]; 993 *(mtod(m, struct in_addr *)) = *p--; 994#ifdef DIAGNOSTIC 995 if (ipprintfs) 996 printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr)); 997#endif 998 999 /* 1000 * Copy option fields and padding (nop) to mbuf. 1001 */ 1002 ip_srcrt.nop = IPOPT_NOP; 1003 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1004 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1005 &ip_srcrt.nop, OPTSIZ); 1006 q = (struct in_addr *)(mtod(m, caddr_t) + 1007 sizeof(struct in_addr) + OPTSIZ); 1008#undef OPTSIZ 1009 /* 1010 * Record return path as an IP source route, 1011 * reversing the path (pointers are now aligned). 1012 */ 1013 while (p >= ip_srcrt.route) { 1014#ifdef DIAGNOSTIC 1015 if (ipprintfs) 1016 printf(" %lx", ntohl(q->s_addr)); 1017#endif 1018 *q++ = *p--; 1019 } 1020 /* 1021 * Last hop goes to final destination. 1022 */ 1023 *q = ip_srcrt.dst; 1024#ifdef DIAGNOSTIC 1025 if (ipprintfs) 1026 printf(" %lx\n", ntohl(q->s_addr)); 1027#endif 1028 return (m); 1029} 1030 1031/* 1032 * Strip out IP options, at higher 1033 * level protocol in the kernel. 1034 * Second argument is buffer to which options 1035 * will be moved, and return value is their length. 1036 * XXX should be deleted; last arg currently ignored. 1037 */ 1038void 1039ip_stripoptions(m, mopt) 1040 register struct mbuf *m; 1041 struct mbuf *mopt; 1042{ 1043 register int i; 1044 struct ip *ip = mtod(m, struct ip *); 1045 register caddr_t opts; 1046 int olen; 1047 1048 olen = (ip->ip_hl<<2) - sizeof (struct ip); 1049 opts = (caddr_t)(ip + 1); 1050 i = m->m_len - (sizeof (struct ip) + olen); 1051 bcopy(opts + olen, opts, (unsigned)i); 1052 m->m_len -= olen; 1053 if (m->m_flags & M_PKTHDR) 1054 m->m_pkthdr.len -= olen; 1055 ip->ip_hl = sizeof(struct ip) >> 2; 1056} 1057 1058u_char inetctlerrmap[PRC_NCMDS] = { 1059 0, 0, 0, 0, 1060 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1061 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1062 EMSGSIZE, EHOSTUNREACH, 0, 0, 1063 0, 0, 0, 0, 1064 ENOPROTOOPT 1065}; 1066 1067/* 1068 * Forward a packet. If some error occurs return the sender 1069 * an icmp packet. Note we can't always generate a meaningful 1070 * icmp message because icmp doesn't have a large enough repertoire 1071 * of codes and types. 1072 * 1073 * If not forwarding, just drop the packet. This could be confusing 1074 * if ipforwarding was zero but some routing protocol was advancing 1075 * us as a gateway to somewhere. However, we must let the routing 1076 * protocol deal with that. 1077 * 1078 * The srcrt parameter indicates whether the packet is being forwarded 1079 * via a source route. 1080 */ 1081static void 1082ip_forward(m, srcrt) 1083 struct mbuf *m; 1084 int srcrt; 1085{ 1086 register struct ip *ip = mtod(m, struct ip *); 1087 register struct sockaddr_in *sin; 1088 register struct rtentry *rt; 1089 int error, type = 0, code = 0; 1090 struct mbuf *mcopy; 1091 n_long dest; 1092 struct ifnet *destifp; 1093 1094 dest = 0; 1095#ifdef DIAGNOSTIC 1096 if (ipprintfs) 1097 printf("forward: src %lx dst %lx ttl %x\n", 1098 ip->ip_src.s_addr, ip->ip_dst.s_addr, ip->ip_ttl); 1099#endif 1100 1101 1102 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) { 1103 ipstat.ips_cantforward++; 1104 m_freem(m); 1105 return; 1106 } 1107 HTONS(ip->ip_id); 1108 if (ip->ip_ttl <= IPTTLDEC) { 1109 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1110 return; 1111 } 1112 ip->ip_ttl -= IPTTLDEC; 1113 1114 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1115 if ((rt = ipforward_rt.ro_rt) == 0 || 1116 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1117 if (ipforward_rt.ro_rt) { 1118 RTFREE(ipforward_rt.ro_rt); 1119 ipforward_rt.ro_rt = 0; 1120 } 1121 sin->sin_family = AF_INET; 1122 sin->sin_len = sizeof(*sin); 1123 sin->sin_addr = ip->ip_dst; 1124 1125 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1126 if (ipforward_rt.ro_rt == 0) { 1127 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1128 return; 1129 } 1130 rt = ipforward_rt.ro_rt; 1131 } 1132 1133 /* 1134 * Save at most 64 bytes of the packet in case 1135 * we need to generate an ICMP message to the src. 1136 */ 1137 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1138 1139 /* 1140 * If forwarding packet using same interface that it came in on, 1141 * perhaps should send a redirect to sender to shortcut a hop. 1142 * Only send redirect if source is sending directly to us, 1143 * and if packet was not source routed (or has any options). 1144 * Also, don't send redirect if forwarding using a default route 1145 * or a route modified by a redirect. 1146 */ 1147#define satosin(sa) ((struct sockaddr_in *)(sa)) 1148 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1149 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1150 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1151 ipsendredirects && !srcrt) { 1152#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1153 u_long src = ntohl(ip->ip_src.s_addr); 1154 1155 if (RTA(rt) && 1156 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1157 if (rt->rt_flags & RTF_GATEWAY) 1158 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1159 else 1160 dest = ip->ip_dst.s_addr; 1161 /* Router requirements says to only send host redirects */ 1162 type = ICMP_REDIRECT; 1163 code = ICMP_REDIRECT_HOST; 1164#ifdef DIAGNOSTIC 1165 if (ipprintfs) 1166 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1167#endif 1168 } 1169 } 1170 1171 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, IP_FORWARDING 1172#ifdef DIRECTED_BROADCAST 1173 | IP_ALLOWBROADCAST 1174#endif 1175 , 0); 1176 if (error) 1177 ipstat.ips_cantforward++; 1178 else { 1179 ipstat.ips_forward++; 1180 if (type) 1181 ipstat.ips_redirectsent++; 1182 else { 1183 if (mcopy) 1184 m_freem(mcopy); 1185 return; 1186 } 1187 } 1188 if (mcopy == NULL) 1189 return; 1190 destifp = NULL; 1191 1192 switch (error) { 1193 1194 case 0: /* forwarded, but need redirect */ 1195 /* type, code set above */ 1196 break; 1197 1198 case ENETUNREACH: /* shouldn't happen, checked above */ 1199 case EHOSTUNREACH: 1200 case ENETDOWN: 1201 case EHOSTDOWN: 1202 default: 1203 type = ICMP_UNREACH; 1204 code = ICMP_UNREACH_HOST; 1205 break; 1206 1207 case EMSGSIZE: 1208 type = ICMP_UNREACH; 1209 code = ICMP_UNREACH_NEEDFRAG; 1210 if (ipforward_rt.ro_rt) 1211 destifp = ipforward_rt.ro_rt->rt_ifp; 1212 ipstat.ips_cantfrag++; 1213 break; 1214 1215 case ENOBUFS: 1216 type = ICMP_SOURCEQUENCH; 1217 code = 0; 1218 break; 1219 } 1220 icmp_error(mcopy, type, code, dest, destifp); 1221} 1222 1223int 1224ip_rsvp_init(struct socket *so) 1225{ 1226 if (so->so_type != SOCK_RAW || 1227 so->so_proto->pr_protocol != IPPROTO_RSVP) 1228 return EOPNOTSUPP; 1229 1230 if (ip_rsvpd != NULL) 1231 return EADDRINUSE; 1232 1233 ip_rsvpd = so; 1234 /* 1235 * This may seem silly, but we need to be sure we don't over-increment 1236 * the RSVP counter, in case something slips up. 1237 */ 1238 if (!ip_rsvp_on) { 1239 ip_rsvp_on = 1; 1240 rsvp_on++; 1241 } 1242 1243 return 0; 1244} 1245 1246int 1247ip_rsvp_done(void) 1248{ 1249 ip_rsvpd = NULL; 1250 /* 1251 * This may seem silly, but we need to be sure we don't over-decrement 1252 * the RSVP counter, in case something slips up. 1253 */ 1254 if (ip_rsvp_on) { 1255 ip_rsvp_on = 0; 1256 rsvp_on--; 1257 } 1258 return 0; 1259} 1260