ip_input.c revision 17072
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 1.44 1996/06/12 19:34:33 gpalmer Exp $ 35 */ 36 37#include "opt_ipfw.h" 38 39#include <sys/param.h> 40#include <sys/systm.h> 41#include <sys/malloc.h> 42#include <sys/mbuf.h> 43#include <sys/domain.h> 44#include <sys/protosw.h> 45#include <sys/socket.h> 46#include <sys/errno.h> 47#include <sys/time.h> 48#include <sys/kernel.h> 49#include <sys/syslog.h> 50#include <sys/sysctl.h> 51 52#include <net/if.h> 53#include <net/route.h> 54#include <net/netisr.h> 55 56#include <netinet/in.h> 57#include <netinet/in_systm.h> 58#include <netinet/in_var.h> 59#include <netinet/ip.h> 60#include <netinet/in_pcb.h> 61#include <netinet/in_var.h> 62#include <netinet/ip_var.h> 63#include <netinet/ip_icmp.h> 64 65#include <sys/socketvar.h> 66 67#ifdef IPFIREWALL 68#include <netinet/ip_fw.h> 69#endif 70 71int rsvp_on = 0; 72static int ip_rsvp_on; 73struct socket *ip_rsvpd; 74 75static int ipforwarding = 0; 76SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 77 &ipforwarding, 0, ""); 78 79static int ipsendredirects = 1; /* XXX */ 80SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 81 &ipsendredirects, 0, ""); 82 83int ip_defttl = IPDEFTTL; 84SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 85 &ip_defttl, 0, ""); 86 87static int ip_dosourceroute = 0; 88SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 89 &ip_dosourceroute, 0, ""); 90#ifdef DIAGNOSTIC 91static int ipprintfs = 0; 92#endif 93 94extern struct domain inetdomain; 95extern struct protosw inetsw[]; 96u_char ip_protox[IPPROTO_MAX]; 97static int ipqmaxlen = IFQ_MAXLEN; 98struct in_ifaddr *in_ifaddr; /* first inet address */ 99struct ifqueue ipintrq; 100SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RD, 101 &ipintrq.ifq_maxlen, 0, ""); 102SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 103 &ipintrq.ifq_drops, 0, ""); 104 105struct ipstat ipstat; 106static struct ipq ipq; 107 108#ifdef IPCTL_DEFMTU 109SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 110 &ip_mtu, 0, ""); 111#endif 112 113/* Firewall hooks */ 114ip_fw_chk_t *ip_fw_chk_ptr; 115ip_fw_ctl_t *ip_fw_ctl_ptr; 116 117/* 118 * We need to save the IP options in case a protocol wants to respond 119 * to an incoming packet over the same route if the packet got here 120 * using IP source routing. This allows connection establishment and 121 * maintenance when the remote end is on a network that is not known 122 * to us. 123 */ 124static int ip_nhops = 0; 125static struct ip_srcrt { 126 struct in_addr dst; /* final destination */ 127 char nop; /* one NOP to align */ 128 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 129 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 130} ip_srcrt; 131 132#ifdef IPDIVERT 133/* 134 * Shared variable between ip_input() and ip_reass() to communicate 135 * about which packets, once assembled from fragments, get diverted, 136 * and to which port. 137 */ 138static u_short frag_divert_port; 139#endif 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#ifdef IPFIREWALL 176 ip_fw_init(); 177#endif 178} 179 180static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 181static struct route ipforward_rt; 182 183/* 184 * Ip input routine. Checksum and byte swap header. If fragmented 185 * try to reassemble. Process options. Pass to next level. 186 */ 187void 188ip_input(struct mbuf *m) 189{ 190 struct ip *ip; 191 struct ipq *fp; 192 struct in_ifaddr *ia; 193 int hlen; 194 195#ifdef DIAGNOSTIC 196 if ((m->m_flags & M_PKTHDR) == 0) 197 panic("ipintr no HDR"); 198#endif 199 /* 200 * If no IP addresses have been set yet but the interfaces 201 * are receiving, can't do anything with incoming packets yet. 202 */ 203 if (in_ifaddr == NULL) 204 goto bad; 205 ipstat.ips_total++; 206 if (m->m_len < sizeof (struct ip) && 207 (m = m_pullup(m, sizeof (struct ip))) == 0) { 208 ipstat.ips_toosmall++; 209 return; 210 } 211 ip = mtod(m, struct ip *); 212 if (ip->ip_v != IPVERSION) { 213 ipstat.ips_badvers++; 214 goto bad; 215 } 216 hlen = ip->ip_hl << 2; 217 if (hlen < sizeof(struct ip)) { /* minimum header length */ 218 ipstat.ips_badhlen++; 219 goto bad; 220 } 221 if (hlen > m->m_len) { 222 if ((m = m_pullup(m, hlen)) == 0) { 223 ipstat.ips_badhlen++; 224 return; 225 } 226 ip = mtod(m, struct ip *); 227 } 228 ip->ip_sum = in_cksum(m, hlen); 229 if (ip->ip_sum) { 230 ipstat.ips_badsum++; 231 goto bad; 232 } 233 234 /* 235 * Convert fields to host representation. 236 */ 237 NTOHS(ip->ip_len); 238 if (ip->ip_len < hlen) { 239 ipstat.ips_badlen++; 240 goto bad; 241 } 242 NTOHS(ip->ip_id); 243 NTOHS(ip->ip_off); 244 245 /* 246 * Check that the amount of data in the buffers 247 * is as at least much as the IP header would have us expect. 248 * Trim mbufs if longer than we expect. 249 * Drop packet if shorter than we expect. 250 */ 251 if (m->m_pkthdr.len < ip->ip_len) { 252 ipstat.ips_tooshort++; 253 goto bad; 254 } 255 if (m->m_pkthdr.len > ip->ip_len) { 256 if (m->m_len == m->m_pkthdr.len) { 257 m->m_len = ip->ip_len; 258 m->m_pkthdr.len = ip->ip_len; 259 } else 260 m_adj(m, ip->ip_len - m->m_pkthdr.len); 261 } 262 /* 263 * IpHack's section. 264 * Right now when no processing on packet has done 265 * and it is still fresh out of network we do our black 266 * deals with it. 267 * - Firewall: deny/allow/divert 268 * - Wrap: fake packet's addr/port <unimpl.> 269 * - Encapsulate: put it in another IP and send out. <unimp.> 270 */ 271 272 if (ip_fw_chk_ptr) { 273 int action; 274 275#ifdef IPDIVERT 276 action = (*ip_fw_chk_ptr)(&ip, hlen, 277 m->m_pkthdr.rcvif, ip_divert_ignore, &m); 278#else 279 action = (*ip_fw_chk_ptr)(&ip, hlen, m->m_pkthdr.rcvif, 0, &m); 280#endif 281 if (action == -1) 282 return; 283 if (action != 0) { 284#ifdef IPDIVERT 285 frag_divert_port = action; 286 goto ours; 287#else 288 goto bad; /* ipfw said divert but we can't */ 289#endif 290 } 291 } 292 293 /* 294 * Process options and, if not destined for us, 295 * ship it on. ip_dooptions returns 1 when an 296 * error was detected (causing an icmp message 297 * to be sent and the original packet to be freed). 298 */ 299 ip_nhops = 0; /* for source routed packets */ 300 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 301 return; 302 303 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 304 * matter if it is destined to another node, or whether it is 305 * a multicast one, RSVP wants it! and prevents it from being forwarded 306 * anywhere else. Also checks if the rsvp daemon is running before 307 * grabbing the packet. 308 */ 309 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 310 goto ours; 311 312 /* 313 * Check our list of addresses, to see if the packet is for us. 314 */ 315 for (ia = in_ifaddr; ia; ia = ia->ia_next) { 316#define satosin(sa) ((struct sockaddr_in *)(sa)) 317 318 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 319 goto ours; 320 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 321#if 0 322 u_long t; 323#endif 324 325 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 326 ip->ip_dst.s_addr) 327 goto ours; 328 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 329 goto ours; 330#if 0 /* XXX - this should go away */ 331 /* 332 * Look for all-0's host part (old broadcast addr), 333 * either for subnet or net. 334 */ 335 t = ntohl(ip->ip_dst.s_addr); 336 if (t == ia->ia_subnet) 337 goto ours; 338 if (t == ia->ia_net) 339 goto ours; 340#endif /* compatibility cruft */ 341 } 342 } 343 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 344 struct in_multi *inm; 345 if (ip_mrouter) { 346 /* 347 * If we are acting as a multicast router, all 348 * incoming multicast packets are passed to the 349 * kernel-level multicast forwarding function. 350 * The packet is returned (relatively) intact; if 351 * ip_mforward() returns a non-zero value, the packet 352 * must be discarded, else it may be accepted below. 353 * 354 * (The IP ident field is put in the same byte order 355 * as expected when ip_mforward() is called from 356 * ip_output().) 357 */ 358 ip->ip_id = htons(ip->ip_id); 359 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 360 ipstat.ips_cantforward++; 361 m_freem(m); 362 return; 363 } 364 ip->ip_id = ntohs(ip->ip_id); 365 366 /* 367 * The process-level routing demon needs to receive 368 * all multicast IGMP packets, whether or not this 369 * host belongs to their destination groups. 370 */ 371 if (ip->ip_p == IPPROTO_IGMP) 372 goto ours; 373 ipstat.ips_forward++; 374 } 375 /* 376 * See if we belong to the destination multicast group on the 377 * arrival interface. 378 */ 379 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 380 if (inm == NULL) { 381 ipstat.ips_cantforward++; 382 m_freem(m); 383 return; 384 } 385 goto ours; 386 } 387 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 388 goto ours; 389 if (ip->ip_dst.s_addr == INADDR_ANY) 390 goto ours; 391 392 /* 393 * Not for us; forward if possible and desirable. 394 */ 395 if (ipforwarding == 0) { 396 ipstat.ips_cantforward++; 397 m_freem(m); 398 } else 399 ip_forward(m, 0); 400 return; 401 402ours: 403 404 /* 405 * If offset or IP_MF are set, must reassemble. 406 * Otherwise, nothing need be done. 407 * (We could look in the reassembly queue to see 408 * if the packet was previously fragmented, 409 * but it's not worth the time; just let them time out.) 410 */ 411 if (ip->ip_off &~ IP_DF) { 412 if (m->m_flags & M_EXT) { /* XXX */ 413 if ((m = m_pullup(m, sizeof (struct ip))) == 0) { 414 ipstat.ips_toosmall++; 415#ifdef IPDIVERT 416 frag_divert_port = 0; 417#endif 418 return; 419 } 420 ip = mtod(m, struct ip *); 421 } 422 /* 423 * Look for queue of fragments 424 * of this datagram. 425 */ 426 for (fp = ipq.next; fp != &ipq; fp = fp->next) 427 if (ip->ip_id == fp->ipq_id && 428 ip->ip_src.s_addr == fp->ipq_src.s_addr && 429 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 430 ip->ip_p == fp->ipq_p) 431 goto found; 432 fp = 0; 433found: 434 435 /* 436 * Adjust ip_len to not reflect header, 437 * set ip_mff if more fragments are expected, 438 * convert offset of this to bytes. 439 */ 440 ip->ip_len -= hlen; 441 ((struct ipasfrag *)ip)->ipf_mff &= ~1; 442 if (ip->ip_off & IP_MF) 443 ((struct ipasfrag *)ip)->ipf_mff |= 1; 444 ip->ip_off <<= 3; 445 446 /* 447 * If datagram marked as having more fragments 448 * or if this is not the first fragment, 449 * attempt reassembly; if it succeeds, proceed. 450 */ 451 if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) { 452 ipstat.ips_fragments++; 453 ip = ip_reass((struct ipasfrag *)ip, fp); 454 if (ip == 0) 455 return; 456 ipstat.ips_reassembled++; 457 m = dtom(ip); 458 } else 459 if (fp) 460 ip_freef(fp); 461 } else 462 ip->ip_len -= hlen; 463 464#ifdef IPDIVERT 465 /* 466 * Divert packets here to the divert protocol if required 467 */ 468 if (frag_divert_port) { 469 ip_divert_port = frag_divert_port; 470 frag_divert_port = 0; 471 (*inetsw[ip_protox[IPPROTO_DIVERT]].pr_input)(m, hlen); 472 return; 473 } 474#endif 475 476 /* 477 * Switch out to protocol's input routine. 478 */ 479 ipstat.ips_delivered++; 480 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 481 return; 482bad: 483 m_freem(m); 484} 485 486/* 487 * IP software interrupt routine - to go away sometime soon 488 */ 489static void 490ipintr(void) 491{ 492 int s; 493 struct mbuf *m; 494 495 while(1) { 496 s = splimp(); 497 IF_DEQUEUE(&ipintrq, m); 498 splx(s); 499 if (m == 0) 500 return; 501 ip_input(m); 502 } 503} 504 505NETISR_SET(NETISR_IP, ipintr); 506 507/* 508 * Take incoming datagram fragment and try to 509 * reassemble it into whole datagram. If a chain for 510 * reassembly of this datagram already exists, then it 511 * is given as fp; otherwise have to make a chain. 512 */ 513static struct ip * 514ip_reass(ip, fp) 515 register struct ipasfrag *ip; 516 register struct ipq *fp; 517{ 518 register struct mbuf *m = dtom(ip); 519 register struct ipasfrag *q; 520 struct mbuf *t; 521 int hlen = ip->ip_hl << 2; 522 int i, next; 523 524 /* 525 * Presence of header sizes in mbufs 526 * would confuse code below. 527 */ 528 m->m_data += hlen; 529 m->m_len -= hlen; 530 531 /* 532 * If first fragment to arrive, create a reassembly queue. 533 */ 534 if (fp == 0) { 535 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 536 goto dropfrag; 537 fp = mtod(t, struct ipq *); 538 insque(fp, &ipq); 539 fp->ipq_ttl = IPFRAGTTL; 540 fp->ipq_p = ip->ip_p; 541 fp->ipq_id = ip->ip_id; 542 fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp; 543 fp->ipq_src = ((struct ip *)ip)->ip_src; 544 fp->ipq_dst = ((struct ip *)ip)->ip_dst; 545#ifdef IPDIVERT 546 fp->ipq_divert = 0; 547#endif 548 q = (struct ipasfrag *)fp; 549 goto insert; 550 } 551 552 /* 553 * Find a segment which begins after this one does. 554 */ 555 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) 556 if (q->ip_off > ip->ip_off) 557 break; 558 559 /* 560 * If there is a preceding segment, it may provide some of 561 * our data already. If so, drop the data from the incoming 562 * segment. If it provides all of our data, drop us. 563 */ 564 if (q->ipf_prev != (struct ipasfrag *)fp) { 565 i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off; 566 if (i > 0) { 567 if (i >= ip->ip_len) 568 goto dropfrag; 569 m_adj(dtom(ip), i); 570 ip->ip_off += i; 571 ip->ip_len -= i; 572 } 573 } 574 575 /* 576 * While we overlap succeeding segments trim them or, 577 * if they are completely covered, dequeue them. 578 */ 579 while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) { 580 i = (ip->ip_off + ip->ip_len) - q->ip_off; 581 if (i < q->ip_len) { 582 q->ip_len -= i; 583 q->ip_off += i; 584 m_adj(dtom(q), i); 585 break; 586 } 587 q = q->ipf_next; 588 m_freem(dtom(q->ipf_prev)); 589 ip_deq(q->ipf_prev); 590 } 591 592insert: 593 594#ifdef IPDIVERT 595 /* 596 * Any fragment diverting causes the whole packet to divert 597 */ 598 if (frag_divert_port != 0) 599 fp->ipq_divert = frag_divert_port; 600 frag_divert_port = 0; 601#endif 602 603 /* 604 * Stick new segment in its place; 605 * check for complete reassembly. 606 */ 607 ip_enq(ip, q->ipf_prev); 608 next = 0; 609 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) { 610 if (q->ip_off != next) 611 return (0); 612 next += q->ip_len; 613 } 614 if (q->ipf_prev->ipf_mff & 1) 615 return (0); 616 617 /* 618 * Reassembly is complete; concatenate fragments. 619 */ 620 q = fp->ipq_next; 621 m = dtom(q); 622 t = m->m_next; 623 m->m_next = 0; 624 m_cat(m, t); 625 q = q->ipf_next; 626 while (q != (struct ipasfrag *)fp) { 627 t = dtom(q); 628 q = q->ipf_next; 629 m_cat(m, t); 630 } 631 632#ifdef IPDIVERT 633 /* 634 * Record divert port for packet, if any 635 */ 636 frag_divert_port = fp->ipq_divert; 637#endif 638 639 /* 640 * Create header for new ip packet by 641 * modifying header of first packet; 642 * dequeue and discard fragment reassembly header. 643 * Make header visible. 644 */ 645 ip = fp->ipq_next; 646 ip->ip_len = next; 647 ip->ipf_mff &= ~1; 648 ((struct ip *)ip)->ip_src = fp->ipq_src; 649 ((struct ip *)ip)->ip_dst = fp->ipq_dst; 650 remque(fp); 651 (void) m_free(dtom(fp)); 652 m = dtom(ip); 653 m->m_len += (ip->ip_hl << 2); 654 m->m_data -= (ip->ip_hl << 2); 655 /* some debugging cruft by sklower, below, will go away soon */ 656 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 657 register int plen = 0; 658 for (t = m; m; m = m->m_next) 659 plen += m->m_len; 660 t->m_pkthdr.len = plen; 661 } 662 return ((struct ip *)ip); 663 664dropfrag: 665 ipstat.ips_fragdropped++; 666 m_freem(m); 667 return (0); 668} 669 670/* 671 * Free a fragment reassembly header and all 672 * associated datagrams. 673 */ 674static void 675ip_freef(fp) 676 struct ipq *fp; 677{ 678 register struct ipasfrag *q, *p; 679 680 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) { 681 p = q->ipf_next; 682 ip_deq(q); 683 m_freem(dtom(q)); 684 } 685 remque(fp); 686 (void) m_free(dtom(fp)); 687} 688 689/* 690 * Put an ip fragment on a reassembly chain. 691 * Like insque, but pointers in middle of structure. 692 */ 693static void 694ip_enq(p, prev) 695 register struct ipasfrag *p, *prev; 696{ 697 698 p->ipf_prev = prev; 699 p->ipf_next = prev->ipf_next; 700 prev->ipf_next->ipf_prev = p; 701 prev->ipf_next = p; 702} 703 704/* 705 * To ip_enq as remque is to insque. 706 */ 707static void 708ip_deq(p) 709 register struct ipasfrag *p; 710{ 711 712 p->ipf_prev->ipf_next = p->ipf_next; 713 p->ipf_next->ipf_prev = p->ipf_prev; 714} 715 716/* 717 * IP timer processing; 718 * if a timer expires on a reassembly 719 * queue, discard it. 720 */ 721void 722ip_slowtimo() 723{ 724 register struct ipq *fp; 725 int s = splnet(); 726 727 fp = ipq.next; 728 if (fp == 0) { 729 splx(s); 730 return; 731 } 732 while (fp != &ipq) { 733 --fp->ipq_ttl; 734 fp = fp->next; 735 if (fp->prev->ipq_ttl == 0) { 736 ipstat.ips_fragtimeout++; 737 ip_freef(fp->prev); 738 } 739 } 740 splx(s); 741} 742 743/* 744 * Drain off all datagram fragments. 745 */ 746void 747ip_drain() 748{ 749 while (ipq.next != &ipq) { 750 ipstat.ips_fragdropped++; 751 ip_freef(ipq.next); 752 } 753 754 in_rtqdrain(); 755} 756 757/* 758 * Do option processing on a datagram, 759 * possibly discarding it if bad options are encountered, 760 * or forwarding it if source-routed. 761 * Returns 1 if packet has been forwarded/freed, 762 * 0 if the packet should be processed further. 763 */ 764static int 765ip_dooptions(m) 766 struct mbuf *m; 767{ 768 register struct ip *ip = mtod(m, struct ip *); 769 register u_char *cp; 770 register struct ip_timestamp *ipt; 771 register struct in_ifaddr *ia; 772 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 773 struct in_addr *sin, dst; 774 n_time ntime; 775 776 dst = ip->ip_dst; 777 cp = (u_char *)(ip + 1); 778 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 779 for (; cnt > 0; cnt -= optlen, cp += optlen) { 780 opt = cp[IPOPT_OPTVAL]; 781 if (opt == IPOPT_EOL) 782 break; 783 if (opt == IPOPT_NOP) 784 optlen = 1; 785 else { 786 optlen = cp[IPOPT_OLEN]; 787 if (optlen <= 0 || optlen > cnt) { 788 code = &cp[IPOPT_OLEN] - (u_char *)ip; 789 goto bad; 790 } 791 } 792 switch (opt) { 793 794 default: 795 break; 796 797 /* 798 * Source routing with record. 799 * Find interface with current destination address. 800 * If none on this machine then drop if strictly routed, 801 * or do nothing if loosely routed. 802 * Record interface address and bring up next address 803 * component. If strictly routed make sure next 804 * address is on directly accessible net. 805 */ 806 case IPOPT_LSRR: 807 case IPOPT_SSRR: 808 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 809 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 810 goto bad; 811 } 812 ipaddr.sin_addr = ip->ip_dst; 813 ia = (struct in_ifaddr *) 814 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 815 if (ia == 0) { 816 if (opt == IPOPT_SSRR) { 817 type = ICMP_UNREACH; 818 code = ICMP_UNREACH_SRCFAIL; 819 goto bad; 820 } 821 /* 822 * Loose routing, and not at next destination 823 * yet; nothing to do except forward. 824 */ 825 break; 826 } 827 off--; /* 0 origin */ 828 if (off > optlen - sizeof(struct in_addr)) { 829 /* 830 * End of source route. Should be for us. 831 */ 832 save_rte(cp, ip->ip_src); 833 break; 834 } 835 836 if (!ip_dosourceroute) { 837 char buf[4*sizeof "123"]; 838 strcpy(buf, inet_ntoa(ip->ip_dst)); 839 840 log(LOG_WARNING, 841 "attempted source route from %s to %s\n", 842 inet_ntoa(ip->ip_src), buf); 843 type = ICMP_UNREACH; 844 code = ICMP_UNREACH_SRCFAIL; 845 goto bad; 846 } 847 848 /* 849 * locate outgoing interface 850 */ 851 (void)memcpy(&ipaddr.sin_addr, cp + off, 852 sizeof(ipaddr.sin_addr)); 853 854 if (opt == IPOPT_SSRR) { 855#define INA struct in_ifaddr * 856#define SA struct sockaddr * 857 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 858 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 859 } else 860 ia = ip_rtaddr(ipaddr.sin_addr); 861 if (ia == 0) { 862 type = ICMP_UNREACH; 863 code = ICMP_UNREACH_SRCFAIL; 864 goto bad; 865 } 866 ip->ip_dst = ipaddr.sin_addr; 867 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 868 sizeof(struct in_addr)); 869 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 870 /* 871 * Let ip_intr's mcast routing check handle mcast pkts 872 */ 873 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 874 break; 875 876 case IPOPT_RR: 877 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 878 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 879 goto bad; 880 } 881 /* 882 * If no space remains, ignore. 883 */ 884 off--; /* 0 origin */ 885 if (off > optlen - sizeof(struct in_addr)) 886 break; 887 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 888 sizeof(ipaddr.sin_addr)); 889 /* 890 * locate outgoing interface; if we're the destination, 891 * use the incoming interface (should be same). 892 */ 893 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 894 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 895 type = ICMP_UNREACH; 896 code = ICMP_UNREACH_HOST; 897 goto bad; 898 } 899 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 900 sizeof(struct in_addr)); 901 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 902 break; 903 904 case IPOPT_TS: 905 code = cp - (u_char *)ip; 906 ipt = (struct ip_timestamp *)cp; 907 if (ipt->ipt_len < 5) 908 goto bad; 909 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) { 910 if (++ipt->ipt_oflw == 0) 911 goto bad; 912 break; 913 } 914 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 915 switch (ipt->ipt_flg) { 916 917 case IPOPT_TS_TSONLY: 918 break; 919 920 case IPOPT_TS_TSANDADDR: 921 if (ipt->ipt_ptr + sizeof(n_time) + 922 sizeof(struct in_addr) > ipt->ipt_len) 923 goto bad; 924 ipaddr.sin_addr = dst; 925 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 926 m->m_pkthdr.rcvif); 927 if (ia == 0) 928 continue; 929 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 930 sizeof(struct in_addr)); 931 ipt->ipt_ptr += sizeof(struct in_addr); 932 break; 933 934 case IPOPT_TS_PRESPEC: 935 if (ipt->ipt_ptr + sizeof(n_time) + 936 sizeof(struct in_addr) > ipt->ipt_len) 937 goto bad; 938 (void)memcpy(&ipaddr.sin_addr, sin, 939 sizeof(struct in_addr)); 940 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 941 continue; 942 ipt->ipt_ptr += sizeof(struct in_addr); 943 break; 944 945 default: 946 goto bad; 947 } 948 ntime = iptime(); 949 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 950 sizeof(n_time)); 951 ipt->ipt_ptr += sizeof(n_time); 952 } 953 } 954 if (forward) { 955 ip_forward(m, 1); 956 return (1); 957 } 958 return (0); 959bad: 960 ip->ip_len -= ip->ip_hl << 2; /* XXX icmp_error adds in hdr length */ 961 icmp_error(m, type, code, 0, 0); 962 ipstat.ips_badoptions++; 963 return (1); 964} 965 966/* 967 * Given address of next destination (final or next hop), 968 * return internet address info of interface to be used to get there. 969 */ 970static struct in_ifaddr * 971ip_rtaddr(dst) 972 struct in_addr dst; 973{ 974 register struct sockaddr_in *sin; 975 976 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 977 978 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 979 if (ipforward_rt.ro_rt) { 980 RTFREE(ipforward_rt.ro_rt); 981 ipforward_rt.ro_rt = 0; 982 } 983 sin->sin_family = AF_INET; 984 sin->sin_len = sizeof(*sin); 985 sin->sin_addr = dst; 986 987 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 988 } 989 if (ipforward_rt.ro_rt == 0) 990 return ((struct in_ifaddr *)0); 991 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 992} 993 994/* 995 * Save incoming source route for use in replies, 996 * to be picked up later by ip_srcroute if the receiver is interested. 997 */ 998void 999save_rte(option, dst) 1000 u_char *option; 1001 struct in_addr dst; 1002{ 1003 unsigned olen; 1004 1005 olen = option[IPOPT_OLEN]; 1006#ifdef DIAGNOSTIC 1007 if (ipprintfs) 1008 printf("save_rte: olen %d\n", olen); 1009#endif 1010 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1011 return; 1012 bcopy(option, ip_srcrt.srcopt, olen); 1013 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1014 ip_srcrt.dst = dst; 1015} 1016 1017/* 1018 * Retrieve incoming source route for use in replies, 1019 * in the same form used by setsockopt. 1020 * The first hop is placed before the options, will be removed later. 1021 */ 1022struct mbuf * 1023ip_srcroute() 1024{ 1025 register struct in_addr *p, *q; 1026 register struct mbuf *m; 1027 1028 if (ip_nhops == 0) 1029 return ((struct mbuf *)0); 1030 m = m_get(M_DONTWAIT, MT_SOOPTS); 1031 if (m == 0) 1032 return ((struct mbuf *)0); 1033 1034#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1035 1036 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1037 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1038 OPTSIZ; 1039#ifdef DIAGNOSTIC 1040 if (ipprintfs) 1041 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1042#endif 1043 1044 /* 1045 * First save first hop for return route 1046 */ 1047 p = &ip_srcrt.route[ip_nhops - 1]; 1048 *(mtod(m, struct in_addr *)) = *p--; 1049#ifdef DIAGNOSTIC 1050 if (ipprintfs) 1051 printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr)); 1052#endif 1053 1054 /* 1055 * Copy option fields and padding (nop) to mbuf. 1056 */ 1057 ip_srcrt.nop = IPOPT_NOP; 1058 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1059 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1060 &ip_srcrt.nop, OPTSIZ); 1061 q = (struct in_addr *)(mtod(m, caddr_t) + 1062 sizeof(struct in_addr) + OPTSIZ); 1063#undef OPTSIZ 1064 /* 1065 * Record return path as an IP source route, 1066 * reversing the path (pointers are now aligned). 1067 */ 1068 while (p >= ip_srcrt.route) { 1069#ifdef DIAGNOSTIC 1070 if (ipprintfs) 1071 printf(" %lx", ntohl(q->s_addr)); 1072#endif 1073 *q++ = *p--; 1074 } 1075 /* 1076 * Last hop goes to final destination. 1077 */ 1078 *q = ip_srcrt.dst; 1079#ifdef DIAGNOSTIC 1080 if (ipprintfs) 1081 printf(" %lx\n", ntohl(q->s_addr)); 1082#endif 1083 return (m); 1084} 1085 1086/* 1087 * Strip out IP options, at higher 1088 * level protocol in the kernel. 1089 * Second argument is buffer to which options 1090 * will be moved, and return value is their length. 1091 * XXX should be deleted; last arg currently ignored. 1092 */ 1093void 1094ip_stripoptions(m, mopt) 1095 register struct mbuf *m; 1096 struct mbuf *mopt; 1097{ 1098 register int i; 1099 struct ip *ip = mtod(m, struct ip *); 1100 register caddr_t opts; 1101 int olen; 1102 1103 olen = (ip->ip_hl<<2) - sizeof (struct ip); 1104 opts = (caddr_t)(ip + 1); 1105 i = m->m_len - (sizeof (struct ip) + olen); 1106 bcopy(opts + olen, opts, (unsigned)i); 1107 m->m_len -= olen; 1108 if (m->m_flags & M_PKTHDR) 1109 m->m_pkthdr.len -= olen; 1110 ip->ip_hl = sizeof(struct ip) >> 2; 1111} 1112 1113u_char inetctlerrmap[PRC_NCMDS] = { 1114 0, 0, 0, 0, 1115 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1116 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1117 EMSGSIZE, EHOSTUNREACH, 0, 0, 1118 0, 0, 0, 0, 1119 ENOPROTOOPT 1120}; 1121 1122/* 1123 * Forward a packet. If some error occurs return the sender 1124 * an icmp packet. Note we can't always generate a meaningful 1125 * icmp message because icmp doesn't have a large enough repertoire 1126 * of codes and types. 1127 * 1128 * If not forwarding, just drop the packet. This could be confusing 1129 * if ipforwarding was zero but some routing protocol was advancing 1130 * us as a gateway to somewhere. However, we must let the routing 1131 * protocol deal with that. 1132 * 1133 * The srcrt parameter indicates whether the packet is being forwarded 1134 * via a source route. 1135 */ 1136static void 1137ip_forward(m, srcrt) 1138 struct mbuf *m; 1139 int srcrt; 1140{ 1141 register struct ip *ip = mtod(m, struct ip *); 1142 register struct sockaddr_in *sin; 1143 register struct rtentry *rt; 1144 int error, type = 0, code = 0; 1145 struct mbuf *mcopy; 1146 n_long dest; 1147 struct ifnet *destifp; 1148 1149 dest = 0; 1150#ifdef DIAGNOSTIC 1151 if (ipprintfs) 1152 printf("forward: src %lx dst %lx ttl %x\n", 1153 ip->ip_src.s_addr, ip->ip_dst.s_addr, ip->ip_ttl); 1154#endif 1155 1156 1157 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) { 1158 ipstat.ips_cantforward++; 1159 m_freem(m); 1160 return; 1161 } 1162 HTONS(ip->ip_id); 1163 if (ip->ip_ttl <= IPTTLDEC) { 1164 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1165 return; 1166 } 1167 ip->ip_ttl -= IPTTLDEC; 1168 1169 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1170 if ((rt = ipforward_rt.ro_rt) == 0 || 1171 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1172 if (ipforward_rt.ro_rt) { 1173 RTFREE(ipforward_rt.ro_rt); 1174 ipforward_rt.ro_rt = 0; 1175 } 1176 sin->sin_family = AF_INET; 1177 sin->sin_len = sizeof(*sin); 1178 sin->sin_addr = ip->ip_dst; 1179 1180 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1181 if (ipforward_rt.ro_rt == 0) { 1182 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1183 return; 1184 } 1185 rt = ipforward_rt.ro_rt; 1186 } 1187 1188 /* 1189 * Save at most 64 bytes of the packet in case 1190 * we need to generate an ICMP message to the src. 1191 */ 1192 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1193 1194 /* 1195 * If forwarding packet using same interface that it came in on, 1196 * perhaps should send a redirect to sender to shortcut a hop. 1197 * Only send redirect if source is sending directly to us, 1198 * and if packet was not source routed (or has any options). 1199 * Also, don't send redirect if forwarding using a default route 1200 * or a route modified by a redirect. 1201 */ 1202#define satosin(sa) ((struct sockaddr_in *)(sa)) 1203 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1204 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1205 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1206 ipsendredirects && !srcrt) { 1207#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1208 u_long src = ntohl(ip->ip_src.s_addr); 1209 1210 if (RTA(rt) && 1211 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1212 if (rt->rt_flags & RTF_GATEWAY) 1213 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1214 else 1215 dest = ip->ip_dst.s_addr; 1216 /* Router requirements says to only send host redirects */ 1217 type = ICMP_REDIRECT; 1218 code = ICMP_REDIRECT_HOST; 1219#ifdef DIAGNOSTIC 1220 if (ipprintfs) 1221 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1222#endif 1223 } 1224 } 1225 1226 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1227 IP_FORWARDING, 0); 1228 if (error) 1229 ipstat.ips_cantforward++; 1230 else { 1231 ipstat.ips_forward++; 1232 if (type) 1233 ipstat.ips_redirectsent++; 1234 else { 1235 if (mcopy) 1236 m_freem(mcopy); 1237 return; 1238 } 1239 } 1240 if (mcopy == NULL) 1241 return; 1242 destifp = NULL; 1243 1244 switch (error) { 1245 1246 case 0: /* forwarded, but need redirect */ 1247 /* type, code set above */ 1248 break; 1249 1250 case ENETUNREACH: /* shouldn't happen, checked above */ 1251 case EHOSTUNREACH: 1252 case ENETDOWN: 1253 case EHOSTDOWN: 1254 default: 1255 type = ICMP_UNREACH; 1256 code = ICMP_UNREACH_HOST; 1257 break; 1258 1259 case EMSGSIZE: 1260 type = ICMP_UNREACH; 1261 code = ICMP_UNREACH_NEEDFRAG; 1262 if (ipforward_rt.ro_rt) 1263 destifp = ipforward_rt.ro_rt->rt_ifp; 1264 ipstat.ips_cantfrag++; 1265 break; 1266 1267 case ENOBUFS: 1268 type = ICMP_SOURCEQUENCH; 1269 code = 0; 1270 break; 1271 } 1272 icmp_error(mcopy, type, code, dest, destifp); 1273} 1274 1275int 1276ip_rsvp_init(struct socket *so) 1277{ 1278 if (so->so_type != SOCK_RAW || 1279 so->so_proto->pr_protocol != IPPROTO_RSVP) 1280 return EOPNOTSUPP; 1281 1282 if (ip_rsvpd != NULL) 1283 return EADDRINUSE; 1284 1285 ip_rsvpd = so; 1286 /* 1287 * This may seem silly, but we need to be sure we don't over-increment 1288 * the RSVP counter, in case something slips up. 1289 */ 1290 if (!ip_rsvp_on) { 1291 ip_rsvp_on = 1; 1292 rsvp_on++; 1293 } 1294 1295 return 0; 1296} 1297 1298int 1299ip_rsvp_done(void) 1300{ 1301 ip_rsvpd = NULL; 1302 /* 1303 * This may seem silly, but we need to be sure we don't over-decrement 1304 * the RSVP counter, in case something slips up. 1305 */ 1306 if (ip_rsvp_on) { 1307 ip_rsvp_on = 0; 1308 rsvp_on--; 1309 } 1310 return 0; 1311} 1312