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 1.40 1996/04/03 13:52:16 phk 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 <sys/socketvar.h> 64 int rsvp_on = 0; 65 static int ip_rsvp_on; 66 struct socket *ip_rsvpd; 67 68 static int ipforwarding = 0; 69 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 70 &ipforwarding, 0, ""); 71 72 static int ipsendredirects = 1; /* XXX */ 73 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 74 &ipsendredirects, 0, ""); 75 76 int ip_defttl = IPDEFTTL; 77 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 78 &ip_defttl, 0, ""); 79 80 static int ip_dosourceroute = 0; 81 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 82 &ip_dosourceroute, 0, ""); 83 #ifdef DIAGNOSTIC 84 static int ipprintfs = 0; 85 #endif 86 87 extern struct domain inetdomain; 88 extern struct protosw inetsw[]; 89 u_char ip_protox[IPPROTO_MAX]; 90 static int ipqmaxlen = IFQ_MAXLEN; 91 struct in_ifaddr *in_ifaddr; /* first inet address */ 92 struct ifqueue ipintrq; 93 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RD, 94 &ipintrq.ifq_maxlen, 0, ""); 95 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 96 &ipintrq.ifq_drops, 0, ""); 97 98 struct ipstat ipstat; 99 static struct ipq ipq; 100 101 #ifdef IPCTL_DEFMTU 102 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 103 &ip_mtu, 0, ""); 104 #endif 105 106 /* Firewall hooks */ 107 ip_fw_chk_t *ip_fw_chk_ptr; 108 ip_fw_ctl_t *ip_fw_ctl_ptr; 109 110 /* 111 * We need to save the IP options in case a protocol wants to respond 112 * to an incoming packet over the same route if the packet got here 113 * using IP source routing. This allows connection establishment and 114 * maintenance when the remote end is on a network that is not known 115 * to us. 116 */ 117 static int ip_nhops = 0; 118 static struct ip_srcrt { 119 struct in_addr dst; /* final destination */ 120 char nop; /* one NOP to align */ 121 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 122 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 123 } ip_srcrt; 124 125 static void save_rte __P((u_char *, struct in_addr)); 126 static void ip_deq __P((struct ipasfrag *)); 127 static int ip_dooptions __P((struct mbuf *)); 128 static void ip_enq __P((struct ipasfrag *, struct ipasfrag *)); 129 static void ip_forward __P((struct mbuf *, int)); 130 static void ip_freef __P((struct ipq *)); 131 static struct ip * 132 ip_reass __P((struct ipasfrag *, struct ipq *)); 133 static struct in_ifaddr * 134 ip_rtaddr __P((struct in_addr)); 135 static void ipintr __P((void)); 136 /* 137 * IP initialization: fill in IP protocol switch table. 138 * All protocols not implemented in kernel go to raw IP protocol handler. 139 */ 140 void 141 ip_init() 142 { 143 register struct protosw *pr; 144 register int i; 145 146 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 147 if (pr == 0) 148 panic("ip_init"); 149 for (i = 0; i < IPPROTO_MAX; i++) 150 ip_protox[i] = pr - inetsw; 151 for (pr = inetdomain.dom_protosw; 152 pr < inetdomain.dom_protoswNPROTOSW; pr++) 153 if (pr->pr_domain->dom_family == PF_INET && 154 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 155 ip_protox[pr->pr_protocol] = pr - inetsw; 156 ipq.next = ipq.prev = &ipq; 157 ip_id = time.tv_sec & 0xffff; 158 ipintrq.ifq_maxlen = ipqmaxlen; 159 #ifdef IPFIREWALL 160 ip_fw_init(); 161 #endif 162 } 163 164 static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 165 static struct route ipforward_rt; 166 167 /* 168 * Ip input routine. Checksum and byte swap header. If fragmented 169 * try to reassemble. Process options. Pass to next level. 170 */ 171 void 172 ip_input(struct mbuf *m) 173 { 174 struct ip *ip; 175 struct ipq *fp; 176 struct in_ifaddr *ia; 177 int hlen; 178 179 #ifdef DIAGNOSTIC 180 if ((m->m_flags & M_PKTHDR) == 0) 181 panic("ipintr no HDR"); 182 #endif 183 /* 184 * If no IP addresses have been set yet but the interfaces 185 * are receiving, can't do anything with incoming packets yet. 186 */ 187 if (in_ifaddr == NULL) 188 goto bad; 189 ipstat.ips_total++; 190 if (m->m_len < sizeof (struct ip) && 191 (m = m_pullup(m, sizeof (struct ip))) == 0) { 192 ipstat.ips_toosmall++; 193 return; 194 } 195 ip = mtod(m, struct ip *); 196 if (ip->ip_v != IPVERSION) { 197 ipstat.ips_badvers++; 198 goto bad; 199 } 200 hlen = ip->ip_hl << 2; 201 if (hlen < sizeof(struct ip)) { /* minimum header length */ 202 ipstat.ips_badhlen++; 203 goto bad; 204 } 205 if (hlen > m->m_len) { 206 if ((m = m_pullup(m, hlen)) == 0) { 207 ipstat.ips_badhlen++; 208 return; 209 } 210 ip = mtod(m, struct ip *); 211 } 212 ip->ip_sum = in_cksum(m, hlen); 213 if (ip->ip_sum) { 214 ipstat.ips_badsum++; 215 goto bad; 216 } 217 218 /* 219 * Convert fields to host representation. 220 */ 221 NTOHS(ip->ip_len); 222 if (ip->ip_len < hlen) { 223 ipstat.ips_badlen++; 224 goto bad; 225 } 226 NTOHS(ip->ip_id); 227 NTOHS(ip->ip_off); 228 229 /* 230 * Check that the amount of data in the buffers 231 * is as at least much as the IP header would have us expect. 232 * Trim mbufs if longer than we expect. 233 * Drop packet if shorter than we expect. 234 */ 235 if (m->m_pkthdr.len < ip->ip_len) { 236 ipstat.ips_tooshort++; 237 goto bad; 238 } 239 if (m->m_pkthdr.len > ip->ip_len) { 240 if (m->m_len == m->m_pkthdr.len) { 241 m->m_len = ip->ip_len; 242 m->m_pkthdr.len = ip->ip_len; 243 } else 244 m_adj(m, ip->ip_len - m->m_pkthdr.len); 245 } 246 /* 247 * IpHack's section. 248 * Right now when no processing on packet has done 249 * and it is still fresh out of network we do our black 250 * deals with it. 251 * - Firewall: deny/allow 252 * - Wrap: fake packet's addr/port <unimpl.> 253 * - Encapsulate: put it in another IP and send out. <unimp.> 254 */ 255 256 if (ip_fw_chk_ptr && 257 !(*ip_fw_chk_ptr)(&ip, hlen, m->m_pkthdr.rcvif, 0, &m)) 258 return; 259 260 /* 261 * Process options and, if not destined for us, 262 * ship it on. ip_dooptions returns 1 when an 263 * error was detected (causing an icmp message 264 * to be sent and the original packet to be freed). 265 */ 266 ip_nhops = 0; /* for source routed packets */ 267 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 268 return; 269 270 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 271 * matter if it is destined to another node, or whether it is 272 * a multicast one, RSVP wants it! and prevents it from being forwarded 273 * anywhere else. Also checks if the rsvp daemon is running before 274 * grabbing the packet. 275 */ 276 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 277 goto ours; 278 279 /* 280 * Check our list of addresses, to see if the packet is for us. 281 */ 282 for (ia = in_ifaddr; ia; ia = ia->ia_next) { 283 #define satosin(sa) ((struct sockaddr_in *)(sa)) 284 285 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 286 goto ours; 287 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 288 #if 0 289 u_long t; 290 #endif 291 292 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 293 ip->ip_dst.s_addr) 294 goto ours; 295 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 296 goto ours; 297 #if 0 /* XXX - this should go away */ 298 /* 299 * Look for all-0's host part (old broadcast addr), 300 * either for subnet or net. 301 */ 302 t = ntohl(ip->ip_dst.s_addr); 303 if (t == ia->ia_subnet) 304 goto ours; 305 if (t == ia->ia_net) 306 goto ours; 307 #endif /* compatibility cruft */ 308 } 309 } 310 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 311 struct in_multi *inm; 312 if (ip_mrouter) { 313 /* 314 * If we are acting as a multicast router, all 315 * incoming multicast packets are passed to the 316 * kernel-level multicast forwarding function. 317 * The packet is returned (relatively) intact; if 318 * ip_mforward() returns a non-zero value, the packet 319 * must be discarded, else it may be accepted below. 320 * 321 * (The IP ident field is put in the same byte order 322 * as expected when ip_mforward() is called from 323 * ip_output().) 324 */ 325 ip->ip_id = htons(ip->ip_id); 326 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 327 ipstat.ips_cantforward++; 328 m_freem(m); 329 return; 330 } 331 ip->ip_id = ntohs(ip->ip_id); 332 333 /* 334 * The process-level routing demon needs to receive 335 * all multicast IGMP packets, whether or not this 336 * host belongs to their destination groups. 337 */ 338 if (ip->ip_p == IPPROTO_IGMP) 339 goto ours; 340 ipstat.ips_forward++; 341 } 342 /* 343 * See if we belong to the destination multicast group on the 344 * arrival interface. 345 */ 346 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 347 if (inm == NULL) { 348 ipstat.ips_cantforward++; 349 m_freem(m); 350 return; 351 } 352 goto ours; 353 } 354 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 355 goto ours; 356 if (ip->ip_dst.s_addr == INADDR_ANY) 357 goto ours; 358 359 /* 360 * Not for us; forward if possible and desirable. 361 */ 362 if (ipforwarding == 0) { 363 ipstat.ips_cantforward++; 364 m_freem(m); 365 } else 366 ip_forward(m, 0); 367 return; 368 369 ours: 370 371 /* 372 * If offset or IP_MF are set, must reassemble. 373 * Otherwise, nothing need be done. 374 * (We could look in the reassembly queue to see 375 * if the packet was previously fragmented, 376 * but it's not worth the time; just let them time out.) 377 */ 378 if (ip->ip_off &~ IP_DF) { 379 if (m->m_flags & M_EXT) { /* XXX */ 380 if ((m = m_pullup(m, sizeof (struct ip))) == 0) { 381 ipstat.ips_toosmall++; 382 return; 383 } 384 ip = mtod(m, struct ip *); 385 } 386 /* 387 * Look for queue of fragments 388 * of this datagram. 389 */ 390 for (fp = ipq.next; fp != &ipq; fp = fp->next) 391 if (ip->ip_id == fp->ipq_id && 392 ip->ip_src.s_addr == fp->ipq_src.s_addr && 393 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 394 ip->ip_p == fp->ipq_p) 395 goto found; 396 fp = 0; 397 found: 398 399 /* 400 * Adjust ip_len to not reflect header, 401 * set ip_mff if more fragments are expected, 402 * convert offset of this to bytes. 403 */ 404 ip->ip_len -= hlen; 405 ((struct ipasfrag *)ip)->ipf_mff &= ~1; 406 if (ip->ip_off & IP_MF) 407 ((struct ipasfrag *)ip)->ipf_mff |= 1; 408 ip->ip_off <<= 3; 409 410 /* 411 * If datagram marked as having more fragments 412 * or if this is not the first fragment, 413 * attempt reassembly; if it succeeds, proceed. 414 */ 415 if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) { 416 ipstat.ips_fragments++; 417 ip = ip_reass((struct ipasfrag *)ip, fp); 418 if (ip == 0) 419 return; 420 ipstat.ips_reassembled++; 421 m = dtom(ip); 422 } else 423 if (fp) 424 ip_freef(fp); 425 } else 426 ip->ip_len -= hlen; 427 428 /* 429 * Switch out to protocol's input routine. 430 */ 431 ipstat.ips_delivered++; 432 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 433 return; 434 bad: 435 m_freem(m); 436 } 437 438 /* 439 * IP software interrupt routine - to go away sometime soon 440 */ 441 static void 442 ipintr(void) 443 { 444 int s; 445 struct mbuf *m; 446 447 while(1) { 448 s = splimp(); 449 IF_DEQUEUE(&ipintrq, m); 450 splx(s); 451 if (m == 0) 452 return; 453 ip_input(m); 454 } 455 } 456 457 NETISR_SET(NETISR_IP, ipintr); 458 459 /* 460 * Take incoming datagram fragment and try to 461 * reassemble it into whole datagram. If a chain for 462 * reassembly of this datagram already exists, then it 463 * is given as fp; otherwise have to make a chain. 464 */ 465 static struct ip * 466 ip_reass(ip, fp) 467 register struct ipasfrag *ip; 468 register struct ipq *fp; 469 { 470 register struct mbuf *m = dtom(ip); 471 register struct ipasfrag *q; 472 struct mbuf *t; 473 int hlen = ip->ip_hl << 2; 474 int i, next; 475 476 /* 477 * Presence of header sizes in mbufs 478 * would confuse code below. 479 */ 480 m->m_data += hlen; 481 m->m_len -= hlen; 482 483 /* 484 * If first fragment to arrive, create a reassembly queue. 485 */ 486 if (fp == 0) { 487 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 488 goto dropfrag; 489 fp = mtod(t, struct ipq *); 490 insque(fp, &ipq); 491 fp->ipq_ttl = IPFRAGTTL; 492 fp->ipq_p = ip->ip_p; 493 fp->ipq_id = ip->ip_id; 494 fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp; 495 fp->ipq_src = ((struct ip *)ip)->ip_src; 496 fp->ipq_dst = ((struct ip *)ip)->ip_dst; 497 q = (struct ipasfrag *)fp; 498 goto insert; 499 } 500 501 /* 502 * Find a segment which begins after this one does. 503 */ 504 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) 505 if (q->ip_off > ip->ip_off) 506 break; 507 508 /* 509 * If there is a preceding segment, it may provide some of 510 * our data already. If so, drop the data from the incoming 511 * segment. If it provides all of our data, drop us. 512 */ 513 if (q->ipf_prev != (struct ipasfrag *)fp) { 514 i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off; 515 if (i > 0) { 516 if (i >= ip->ip_len) 517 goto dropfrag; 518 m_adj(dtom(ip), i); 519 ip->ip_off += i; 520 ip->ip_len -= i; 521 } 522 } 523 524 /* 525 * While we overlap succeeding segments trim them or, 526 * if they are completely covered, dequeue them. 527 */ 528 while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) { 529 i = (ip->ip_off + ip->ip_len) - q->ip_off; 530 if (i < q->ip_len) { 531 q->ip_len -= i; 532 q->ip_off += i; 533 m_adj(dtom(q), i); 534 break; 535 } 536 q = q->ipf_next; 537 m_freem(dtom(q->ipf_prev)); 538 ip_deq(q->ipf_prev); 539 } 540 541 insert: 542 /* 543 * Stick new segment in its place; 544 * check for complete reassembly. 545 */ 546 ip_enq(ip, q->ipf_prev); 547 next = 0; 548 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) { 549 if (q->ip_off != next) 550 return (0); 551 next += q->ip_len; 552 } 553 if (q->ipf_prev->ipf_mff & 1) 554 return (0); 555 556 /* 557 * Reassembly is complete; concatenate fragments. 558 */ 559 q = fp->ipq_next; 560 m = dtom(q); 561 t = m->m_next; 562 m->m_next = 0; 563 m_cat(m, t); 564 q = q->ipf_next; 565 while (q != (struct ipasfrag *)fp) { 566 t = dtom(q); 567 q = q->ipf_next; 568 m_cat(m, t); 569 } 570 571 /* 572 * Create header for new ip packet by 573 * modifying header of first packet; 574 * dequeue and discard fragment reassembly header. 575 * Make header visible. 576 */ 577 ip = fp->ipq_next; 578 ip->ip_len = next; 579 ip->ipf_mff &= ~1; 580 ((struct ip *)ip)->ip_src = fp->ipq_src; 581 ((struct ip *)ip)->ip_dst = fp->ipq_dst; 582 remque(fp); 583 (void) m_free(dtom(fp)); 584 m = dtom(ip); 585 m->m_len += (ip->ip_hl << 2); 586 m->m_data -= (ip->ip_hl << 2); 587 /* some debugging cruft by sklower, below, will go away soon */ 588 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 589 register int plen = 0; 590 for (t = m; m; m = m->m_next) 591 plen += m->m_len; 592 t->m_pkthdr.len = plen; 593 } 594 return ((struct ip *)ip); 595 596 dropfrag: 597 ipstat.ips_fragdropped++; 598 m_freem(m); 599 return (0); 600 } 601 602 /* 603 * Free a fragment reassembly header and all 604 * associated datagrams. 605 */ 606 static void 607 ip_freef(fp) 608 struct ipq *fp; 609 { 610 register struct ipasfrag *q, *p; 611 612 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) { 613 p = q->ipf_next; 614 ip_deq(q); 615 m_freem(dtom(q)); 616 } 617 remque(fp); 618 (void) m_free(dtom(fp)); 619 } 620 621 /* 622 * Put an ip fragment on a reassembly chain. 623 * Like insque, but pointers in middle of structure. 624 */ 625 static void 626 ip_enq(p, prev) 627 register struct ipasfrag *p, *prev; 628 { 629 630 p->ipf_prev = prev; 631 p->ipf_next = prev->ipf_next; 632 prev->ipf_next->ipf_prev = p; 633 prev->ipf_next = p; 634 } 635 636 /* 637 * To ip_enq as remque is to insque. 638 */ 639 static void 640 ip_deq(p) 641 register struct ipasfrag *p; 642 { 643 644 p->ipf_prev->ipf_next = p->ipf_next; 645 p->ipf_next->ipf_prev = p->ipf_prev; 646 } 647 648 /* 649 * IP timer processing; 650 * if a timer expires on a reassembly 651 * queue, discard it. 652 */ 653 void 654 ip_slowtimo() 655 { 656 register struct ipq *fp; 657 int s = splnet(); 658 659 fp = ipq.next; 660 if (fp == 0) { 661 splx(s); 662 return; 663 } 664 while (fp != &ipq) { 665 --fp->ipq_ttl; 666 fp = fp->next; 667 if (fp->prev->ipq_ttl == 0) { 668 ipstat.ips_fragtimeout++; 669 ip_freef(fp->prev); 670 } 671 } 672 splx(s); 673 } 674 675 /* 676 * Drain off all datagram fragments. 677 */ 678 void 679 ip_drain() 680 { 681 while (ipq.next != &ipq) { 682 ipstat.ips_fragdropped++; 683 ip_freef(ipq.next); 684 } 685 686 in_rtqdrain(); 687 } 688 689 /* 690 * Do option processing on a datagram, 691 * possibly discarding it if bad options are encountered, 692 * or forwarding it if source-routed. 693 * Returns 1 if packet has been forwarded/freed, 694 * 0 if the packet should be processed further. 695 */ 696 static int 697 ip_dooptions(m) 698 struct mbuf *m; 699 { 700 register struct ip *ip = mtod(m, struct ip *); 701 register u_char *cp; 702 register struct ip_timestamp *ipt; 703 register struct in_ifaddr *ia; 704 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 705 struct in_addr *sin, dst; 706 n_time ntime; 707 708 dst = ip->ip_dst; 709 cp = (u_char *)(ip + 1); 710 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 711 for (; cnt > 0; cnt -= optlen, cp += optlen) { 712 opt = cp[IPOPT_OPTVAL]; 713 if (opt == IPOPT_EOL) 714 break; 715 if (opt == IPOPT_NOP) 716 optlen = 1; 717 else { 718 optlen = cp[IPOPT_OLEN]; 719 if (optlen <= 0 || optlen > cnt) { 720 code = &cp[IPOPT_OLEN] - (u_char *)ip; 721 goto bad; 722 } 723 } 724 switch (opt) { 725 726 default: 727 break; 728 729 /* 730 * Source routing with record. 731 * Find interface with current destination address. 732 * If none on this machine then drop if strictly routed, 733 * or do nothing if loosely routed. 734 * Record interface address and bring up next address 735 * component. If strictly routed make sure next 736 * address is on directly accessible net. 737 */ 738 case IPOPT_LSRR: 739 case IPOPT_SSRR: 740 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 741 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 742 goto bad; 743 } 744 ipaddr.sin_addr = ip->ip_dst; 745 ia = (struct in_ifaddr *) 746 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 747 if (ia == 0) { 748 if (opt == IPOPT_SSRR) { 749 type = ICMP_UNREACH; 750 code = ICMP_UNREACH_SRCFAIL; 751 goto bad; 752 } 753 /* 754 * Loose routing, and not at next destination 755 * yet; nothing to do except forward. 756 */ 757 break; 758 } 759 off--; /* 0 origin */ 760 if (off > optlen - sizeof(struct in_addr)) { 761 /* 762 * End of source route. Should be for us. 763 */ 764 save_rte(cp, ip->ip_src); 765 break; 766 } 767 768 if (!ip_dosourceroute) { 769 char buf[4*sizeof "123"]; 770 strcpy(buf, inet_ntoa(ip->ip_dst)); 771 772 log(LOG_WARNING, 773 "attempted source route from %s to %s\n", 774 inet_ntoa(ip->ip_src), buf); 775 type = ICMP_UNREACH; 776 code = ICMP_UNREACH_SRCFAIL; 777 goto bad; 778 } 779 780 /* 781 * locate outgoing interface 782 */ 783 (void)memcpy(&ipaddr.sin_addr, cp + off, 784 sizeof(ipaddr.sin_addr)); 785 786 if (opt == IPOPT_SSRR) { 787 #define INA struct in_ifaddr * 788 #define SA struct sockaddr * 789 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 790 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 791 } else 792 ia = ip_rtaddr(ipaddr.sin_addr); 793 if (ia == 0) { 794 type = ICMP_UNREACH; 795 code = ICMP_UNREACH_SRCFAIL; 796 goto bad; 797 } 798 ip->ip_dst = ipaddr.sin_addr; 799 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 800 sizeof(struct in_addr)); 801 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 802 /* 803 * Let ip_intr's mcast routing check handle mcast pkts 804 */ 805 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 806 break; 807 808 case IPOPT_RR: 809 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 810 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 811 goto bad; 812 } 813 /* 814 * If no space remains, ignore. 815 */ 816 off--; /* 0 origin */ 817 if (off > optlen - sizeof(struct in_addr)) 818 break; 819 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 820 sizeof(ipaddr.sin_addr)); 821 /* 822 * locate outgoing interface; if we're the destination, 823 * use the incoming interface (should be same). 824 */ 825 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 826 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 827 type = ICMP_UNREACH; 828 code = ICMP_UNREACH_HOST; 829 goto bad; 830 } 831 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 832 sizeof(struct in_addr)); 833 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 834 break; 835 836 case IPOPT_TS: 837 code = cp - (u_char *)ip; 838 ipt = (struct ip_timestamp *)cp; 839 if (ipt->ipt_len < 5) 840 goto bad; 841 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) { 842 if (++ipt->ipt_oflw == 0) 843 goto bad; 844 break; 845 } 846 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 847 switch (ipt->ipt_flg) { 848 849 case IPOPT_TS_TSONLY: 850 break; 851 852 case IPOPT_TS_TSANDADDR: 853 if (ipt->ipt_ptr + sizeof(n_time) + 854 sizeof(struct in_addr) > ipt->ipt_len) 855 goto bad; 856 ipaddr.sin_addr = dst; 857 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 858 m->m_pkthdr.rcvif); 859 if (ia == 0) 860 continue; 861 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 862 sizeof(struct in_addr)); 863 ipt->ipt_ptr += sizeof(struct in_addr); 864 break; 865 866 case IPOPT_TS_PRESPEC: 867 if (ipt->ipt_ptr + sizeof(n_time) + 868 sizeof(struct in_addr) > ipt->ipt_len) 869 goto bad; 870 (void)memcpy(&ipaddr.sin_addr, sin, 871 sizeof(struct in_addr)); 872 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 873 continue; 874 ipt->ipt_ptr += sizeof(struct in_addr); 875 break; 876 877 default: 878 goto bad; 879 } 880 ntime = iptime(); 881 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 882 sizeof(n_time)); 883 ipt->ipt_ptr += sizeof(n_time); 884 } 885 } 886 if (forward) { 887 ip_forward(m, 1); 888 return (1); 889 } 890 return (0); 891 bad: 892 ip->ip_len -= ip->ip_hl << 2; /* XXX icmp_error adds in hdr length */ 893 icmp_error(m, type, code, 0, 0); 894 ipstat.ips_badoptions++; 895 return (1); 896 } 897 898 /* 899 * Given address of next destination (final or next hop), 900 * return internet address info of interface to be used to get there. 901 */ 902 static struct in_ifaddr * 903 ip_rtaddr(dst) 904 struct in_addr dst; 905 { 906 register struct sockaddr_in *sin; 907 908 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 909 910 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 911 if (ipforward_rt.ro_rt) { 912 RTFREE(ipforward_rt.ro_rt); 913 ipforward_rt.ro_rt = 0; 914 } 915 sin->sin_family = AF_INET; 916 sin->sin_len = sizeof(*sin); 917 sin->sin_addr = dst; 918 919 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 920 } 921 if (ipforward_rt.ro_rt == 0) 922 return ((struct in_ifaddr *)0); 923 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 924 } 925 926 /* 927 * Save incoming source route for use in replies, 928 * to be picked up later by ip_srcroute if the receiver is interested. 929 */ 930 void 931 save_rte(option, dst) 932 u_char *option; 933 struct in_addr dst; 934 { 935 unsigned olen; 936 937 olen = option[IPOPT_OLEN]; 938 #ifdef DIAGNOSTIC 939 if (ipprintfs) 940 printf("save_rte: olen %d\n", olen); 941 #endif 942 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 943 return; 944 (void)memcpy(ip_srcrt.srcopt, option, olen); 945 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 946 ip_srcrt.dst = dst; 947 } 948 949 /* 950 * Retrieve incoming source route for use in replies, 951 * in the same form used by setsockopt. 952 * The first hop is placed before the options, will be removed later. 953 */ 954 struct mbuf * 955 ip_srcroute() 956 { 957 register struct in_addr *p, *q; 958 register struct mbuf *m; 959 960 if (ip_nhops == 0) 961 return ((struct mbuf *)0); 962 m = m_get(M_DONTWAIT, MT_SOOPTS); 963 if (m == 0) 964 return ((struct mbuf *)0); 965 966 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 967 968 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 969 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 970 OPTSIZ; 971 #ifdef DIAGNOSTIC 972 if (ipprintfs) 973 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 974 #endif 975 976 /* 977 * First save first hop for return route 978 */ 979 p = &ip_srcrt.route[ip_nhops - 1]; 980 *(mtod(m, struct in_addr *)) = *p--; 981 #ifdef DIAGNOSTIC 982 if (ipprintfs) 983 printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr)); 984 #endif 985 986 /* 987 * Copy option fields and padding (nop) to mbuf. 988 */ 989 ip_srcrt.nop = IPOPT_NOP; 990 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 991 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 992 &ip_srcrt.nop, OPTSIZ); 993 q = (struct in_addr *)(mtod(m, caddr_t) + 994 sizeof(struct in_addr) + OPTSIZ); 995 #undef OPTSIZ 996 /* 997 * Record return path as an IP source route, 998 * reversing the path (pointers are now aligned). 999 */ 1000 while (p >= ip_srcrt.route) { 1001 #ifdef DIAGNOSTIC 1002 if (ipprintfs) 1003 printf(" %lx", ntohl(q->s_addr)); 1004 #endif 1005 *q++ = *p--; 1006 } 1007 /* 1008 * Last hop goes to final destination. 1009 */ 1010 *q = ip_srcrt.dst; 1011 #ifdef DIAGNOSTIC 1012 if (ipprintfs) 1013 printf(" %lx\n", ntohl(q->s_addr)); 1014 #endif 1015 return (m); 1016 } 1017 1018 /* 1019 * Strip out IP options, at higher 1020 * level protocol in the kernel. 1021 * Second argument is buffer to which options 1022 * will be moved, and return value is their length. 1023 * XXX should be deleted; last arg currently ignored. 1024 */ 1025 void 1026 ip_stripoptions(m, mopt) 1027 register struct mbuf *m; 1028 struct mbuf *mopt; 1029 { 1030 register int i; 1031 struct ip *ip = mtod(m, struct ip *); 1032 register caddr_t opts; 1033 int olen; 1034 1035 olen = (ip->ip_hl<<2) - sizeof (struct ip); 1036 opts = (caddr_t)(ip + 1); 1037 i = m->m_len - (sizeof (struct ip) + olen); 1038 bcopy(opts + olen, opts, (unsigned)i); 1039 m->m_len -= olen; 1040 if (m->m_flags & M_PKTHDR) 1041 m->m_pkthdr.len -= olen; 1042 ip->ip_hl = sizeof(struct ip) >> 2; 1043 } 1044 1045 u_char inetctlerrmap[PRC_NCMDS] = { 1046 0, 0, 0, 0, 1047 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1048 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1049 EMSGSIZE, EHOSTUNREACH, 0, 0, 1050 0, 0, 0, 0, 1051 ENOPROTOOPT 1052 }; 1053 1054 /* 1055 * Forward a packet. If some error occurs return the sender 1056 * an icmp packet. Note we can't always generate a meaningful 1057 * icmp message because icmp doesn't have a large enough repertoire 1058 * of codes and types. 1059 * 1060 * If not forwarding, just drop the packet. This could be confusing 1061 * if ipforwarding was zero but some routing protocol was advancing 1062 * us as a gateway to somewhere. However, we must let the routing 1063 * protocol deal with that. 1064 * 1065 * The srcrt parameter indicates whether the packet is being forwarded 1066 * via a source route. 1067 */ 1068 static void 1069 ip_forward(m, srcrt) 1070 struct mbuf *m; 1071 int srcrt; 1072 { 1073 register struct ip *ip = mtod(m, struct ip *); 1074 register struct sockaddr_in *sin; 1075 register struct rtentry *rt; 1076 int error, type = 0, code = 0; 1077 struct mbuf *mcopy; 1078 n_long dest; 1079 struct ifnet *destifp; 1080 1081 dest = 0; 1082 #ifdef DIAGNOSTIC 1083 if (ipprintfs) 1084 printf("forward: src %lx dst %lx ttl %x\n", 1085 ip->ip_src.s_addr, ip->ip_dst.s_addr, ip->ip_ttl); 1086 #endif 1087 1088 1089 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) { 1090 ipstat.ips_cantforward++; 1091 m_freem(m); 1092 return; 1093 } 1094 HTONS(ip->ip_id); 1095 if (ip->ip_ttl <= IPTTLDEC) { 1096 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1097 return; 1098 } 1099 ip->ip_ttl -= IPTTLDEC; 1100 1101 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1102 if ((rt = ipforward_rt.ro_rt) == 0 || 1103 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1104 if (ipforward_rt.ro_rt) { 1105 RTFREE(ipforward_rt.ro_rt); 1106 ipforward_rt.ro_rt = 0; 1107 } 1108 sin->sin_family = AF_INET; 1109 sin->sin_len = sizeof(*sin); 1110 sin->sin_addr = ip->ip_dst; 1111 1112 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1113 if (ipforward_rt.ro_rt == 0) { 1114 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1115 return; 1116 } 1117 rt = ipforward_rt.ro_rt; 1118 } 1119 1120 /* 1121 * Save at most 64 bytes of the packet in case 1122 * we need to generate an ICMP message to the src. 1123 */ 1124 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1125 1126 /* 1127 * If forwarding packet using same interface that it came in on, 1128 * perhaps should send a redirect to sender to shortcut a hop. 1129 * Only send redirect if source is sending directly to us, 1130 * and if packet was not source routed (or has any options). 1131 * Also, don't send redirect if forwarding using a default route 1132 * or a route modified by a redirect. 1133 */ 1134 #define satosin(sa) ((struct sockaddr_in *)(sa)) 1135 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1136 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1137 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1138 ipsendredirects && !srcrt) { 1139 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1140 u_long src = ntohl(ip->ip_src.s_addr); 1141 1142 if (RTA(rt) && 1143 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1144 if (rt->rt_flags & RTF_GATEWAY) 1145 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1146 else 1147 dest = ip->ip_dst.s_addr; 1148 /* Router requirements says to only send host redirects */ 1149 type = ICMP_REDIRECT; 1150 code = ICMP_REDIRECT_HOST; 1151 #ifdef DIAGNOSTIC 1152 if (ipprintfs) 1153 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1154 #endif 1155 } 1156 } 1157 1158 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1159 IP_FORWARDING, 0); 1160 if (error) 1161 ipstat.ips_cantforward++; 1162 else { 1163 ipstat.ips_forward++; 1164 if (type) 1165 ipstat.ips_redirectsent++; 1166 else { 1167 if (mcopy) 1168 m_freem(mcopy); 1169 return; 1170 } 1171 } 1172 if (mcopy == NULL) 1173 return; 1174 destifp = NULL; 1175 1176 switch (error) { 1177 1178 case 0: /* forwarded, but need redirect */ 1179 /* type, code set above */ 1180 break; 1181 1182 case ENETUNREACH: /* shouldn't happen, checked above */ 1183 case EHOSTUNREACH: 1184 case ENETDOWN: 1185 case EHOSTDOWN: 1186 default: 1187 type = ICMP_UNREACH; 1188 code = ICMP_UNREACH_HOST; 1189 break; 1190 1191 case EMSGSIZE: 1192 type = ICMP_UNREACH; 1193 code = ICMP_UNREACH_NEEDFRAG; 1194 if (ipforward_rt.ro_rt) 1195 destifp = ipforward_rt.ro_rt->rt_ifp; 1196 ipstat.ips_cantfrag++; 1197 break; 1198 1199 case ENOBUFS: 1200 type = ICMP_SOURCEQUENCH; 1201 code = 0; 1202 break; 1203 } 1204 icmp_error(mcopy, type, code, dest, destifp); 1205 } 1206 1207 int 1208 ip_rsvp_init(struct socket *so) 1209 { 1210 if (so->so_type != SOCK_RAW || 1211 so->so_proto->pr_protocol != IPPROTO_RSVP) 1212 return EOPNOTSUPP; 1213 1214 if (ip_rsvpd != NULL) 1215 return EADDRINUSE; 1216 1217 ip_rsvpd = so; 1218 /* 1219 * This may seem silly, but we need to be sure we don't over-increment 1220 * the RSVP counter, in case something slips up. 1221 */ 1222 if (!ip_rsvp_on) { 1223 ip_rsvp_on = 1; 1224 rsvp_on++; 1225 } 1226 1227 return 0; 1228 } 1229 1230 int 1231 ip_rsvp_done(void) 1232 { 1233 ip_rsvpd = NULL; 1234 /* 1235 * This may seem silly, but we need to be sure we don't over-decrement 1236 * the RSVP counter, in case something slips up. 1237 */ 1238 if (ip_rsvp_on) { 1239 ip_rsvp_on = 0; 1240 rsvp_on--; 1241 } 1242 return 0; 1243 } 1244