1 /* 2 * Copyright (c) 1982, 1986, 1988 Regents of the University of California. 3 * All rights reserved. 4 * 5 * %sccs.include.redist.c% 6 * 7 * @(#)ip_input.c 7.23 (Berkeley) 10/11/92 8 */ 9 10 #include <sys/param.h> 11 #include <sys/systm.h> 12 #include <sys/malloc.h> 13 #include <sys/mbuf.h> 14 #include <sys/domain.h> 15 #include <sys/protosw.h> 16 #include <sys/socket.h> 17 #include <sys/errno.h> 18 #include <sys/time.h> 19 #include <sys/kernel.h> 20 21 #include <net/if.h> 22 #include <net/route.h> 23 24 #include <netinet/in.h> 25 #include <netinet/in_systm.h> 26 #include <netinet/ip.h> 27 #include <netinet/in_pcb.h> 28 #include <netinet/in_var.h> 29 #include <netinet/ip_var.h> 30 #include <netinet/ip_icmp.h> 31 32 #ifndef IPFORWARDING 33 #ifdef GATEWAY 34 #define IPFORWARDING 1 /* forward IP packets not for us */ 35 #else /* GATEWAY */ 36 #define IPFORWARDING 0 /* don't forward IP packets not for us */ 37 #endif /* GATEWAY */ 38 #endif /* IPFORWARDING */ 39 #ifndef IPSENDREDIRECTS 40 #define IPSENDREDIRECTS 1 41 #endif 42 int ipforwarding = IPFORWARDING; 43 int ipsendredirects = IPSENDREDIRECTS; 44 #ifdef DIAGNOSTIC 45 int ipprintfs = 0; 46 #endif 47 48 extern struct domain inetdomain; 49 extern struct protosw inetsw[]; 50 u_char ip_protox[IPPROTO_MAX]; 51 int ipqmaxlen = IFQ_MAXLEN; 52 struct in_ifaddr *in_ifaddr; /* first inet address */ 53 54 /* 55 * We need to save the IP options in case a protocol wants to respond 56 * to an incoming packet over the same route if the packet got here 57 * using IP source routing. This allows connection establishment and 58 * maintenance when the remote end is on a network that is not known 59 * to us. 60 */ 61 int ip_nhops = 0; 62 static struct ip_srcrt { 63 struct in_addr dst; /* final destination */ 64 char nop; /* one NOP to align */ 65 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 66 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 67 } ip_srcrt; 68 69 #ifdef GATEWAY 70 extern int if_index; 71 u_long *ip_ifmatrix; 72 #endif 73 74 /* 75 * IP initialization: fill in IP protocol switch table. 76 * All protocols not implemented in kernel go to raw IP protocol handler. 77 */ 78 ip_init() 79 { 80 register struct protosw *pr; 81 register int i; 82 83 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 84 if (pr == 0) 85 panic("ip_init"); 86 for (i = 0; i < IPPROTO_MAX; i++) 87 ip_protox[i] = pr - inetsw; 88 for (pr = inetdomain.dom_protosw; 89 pr < inetdomain.dom_protoswNPROTOSW; pr++) 90 if (pr->pr_domain->dom_family == PF_INET && 91 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 92 ip_protox[pr->pr_protocol] = pr - inetsw; 93 ipq.next = ipq.prev = &ipq; 94 ip_id = time.tv_sec & 0xffff; 95 ipintrq.ifq_maxlen = ipqmaxlen; 96 #ifdef GATEWAY 97 i = (if_index + 1) * (if_index + 1) * sizeof (u_long); 98 ip_ifmatrix = (u_long *) malloc(i, M_RTABLE, M_WAITOK); 99 bzero((char *)ip_ifmatrix, i); 100 #endif 101 } 102 103 struct ip *ip_reass(); 104 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 105 struct route ipforward_rt; 106 107 /* 108 * Ip input routine. Checksum and byte swap header. If fragmented 109 * try to reassemble. Process options. Pass to next level. 110 */ 111 ipintr() 112 { 113 register struct ip *ip; 114 register struct mbuf *m; 115 register struct ipq *fp; 116 register struct in_ifaddr *ia; 117 int hlen, s; 118 119 next: 120 /* 121 * Get next datagram off input queue and get IP header 122 * in first mbuf. 123 */ 124 s = splimp(); 125 IF_DEQUEUE(&ipintrq, m); 126 splx(s); 127 if (m == 0) 128 return; 129 #ifdef DIAGNOSTIC 130 if ((m->m_flags & M_PKTHDR) == 0) 131 panic("ipintr no HDR"); 132 #endif 133 /* 134 * If no IP addresses have been set yet but the interfaces 135 * are receiving, can't do anything with incoming packets yet. 136 */ 137 if (in_ifaddr == NULL) 138 goto bad; 139 ipstat.ips_total++; 140 if (m->m_len < sizeof (struct ip) && 141 (m = m_pullup(m, sizeof (struct ip))) == 0) { 142 ipstat.ips_toosmall++; 143 goto next; 144 } 145 ip = mtod(m, struct ip *); 146 hlen = ip->ip_hl << 2; 147 if (hlen < sizeof(struct ip)) { /* minimum header length */ 148 ipstat.ips_badhlen++; 149 goto bad; 150 } 151 if (hlen > m->m_len) { 152 if ((m = m_pullup(m, hlen)) == 0) { 153 ipstat.ips_badhlen++; 154 goto next; 155 } 156 ip = mtod(m, struct ip *); 157 } 158 if (ip->ip_sum = in_cksum(m, hlen)) { 159 ipstat.ips_badsum++; 160 goto bad; 161 } 162 163 /* 164 * Convert fields to host representation. 165 */ 166 NTOHS(ip->ip_len); 167 if (ip->ip_len < hlen) { 168 ipstat.ips_badlen++; 169 goto bad; 170 } 171 NTOHS(ip->ip_id); 172 NTOHS(ip->ip_off); 173 174 /* 175 * Check that the amount of data in the buffers 176 * is as at least much as the IP header would have us expect. 177 * Trim mbufs if longer than we expect. 178 * Drop packet if shorter than we expect. 179 */ 180 if (m->m_pkthdr.len < ip->ip_len) { 181 ipstat.ips_tooshort++; 182 goto bad; 183 } 184 if (m->m_pkthdr.len > ip->ip_len) { 185 if (m->m_len == m->m_pkthdr.len) { 186 m->m_len = ip->ip_len; 187 m->m_pkthdr.len = ip->ip_len; 188 } else 189 m_adj(m, ip->ip_len - m->m_pkthdr.len); 190 } 191 192 /* 193 * Process options and, if not destined for us, 194 * ship it on. ip_dooptions returns 1 when an 195 * error was detected (causing an icmp message 196 * to be sent and the original packet to be freed). 197 */ 198 ip_nhops = 0; /* for source routed packets */ 199 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 200 goto next; 201 202 /* 203 * Check our list of addresses, to see if the packet is for us. 204 */ 205 for (ia = in_ifaddr; ia; ia = ia->ia_next) { 206 #define satosin(sa) ((struct sockaddr_in *)(sa)) 207 208 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 209 goto ours; 210 if ( 211 #ifdef DIRECTED_BROADCAST 212 ia->ia_ifp == m->m_pkthdr.rcvif && 213 #endif 214 (ia->ia_ifp->if_flags & IFF_BROADCAST)) { 215 u_long t; 216 217 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 218 ip->ip_dst.s_addr) 219 goto ours; 220 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 221 goto ours; 222 /* 223 * Look for all-0's host part (old broadcast addr), 224 * either for subnet or net. 225 */ 226 t = ntohl(ip->ip_dst.s_addr); 227 if (t == ia->ia_subnet) 228 goto ours; 229 if (t == ia->ia_net) 230 goto ours; 231 } 232 } 233 #ifdef MULTICAST 234 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 235 struct in_multi *inm; 236 #ifdef MROUTING 237 extern struct socket *ip_mrouter; 238 239 if (ip_mrouter) { 240 /* 241 * If we are acting as a multicast router, all 242 * incoming multicast packets are passed to the 243 * kernel-level multicast forwarding function. 244 * The packet is returned (relatively) intact; if 245 * ip_mforward() returns a non-zero value, the packet 246 * must be discarded, else it may be accepted below. 247 * 248 * (The IP ident field is put in the same byte order 249 * as expected when ip_mforward() is called from 250 * ip_output().) 251 */ 252 ip->ip_id = htons(ip->ip_id); 253 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) { 254 m_freem(m); 255 goto next; 256 } 257 ip->ip_id = ntohs(ip->ip_id); 258 259 /* 260 * The process-level routing demon needs to receive 261 * all multicast IGMP packets, whether or not this 262 * host belongs to their destination groups. 263 */ 264 if (ip->ip_p == IPPROTO_IGMP) 265 goto ours; 266 } 267 #endif 268 /* 269 * See if we belong to the destination multicast group on the 270 * arrival interface. 271 */ 272 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 273 if (inm == NULL) { 274 m_freem(m); 275 goto next; 276 } 277 goto ours; 278 } 279 #endif 280 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 281 goto ours; 282 if (ip->ip_dst.s_addr == INADDR_ANY) 283 goto ours; 284 285 /* 286 * Not for us; forward if possible and desirable. 287 */ 288 if (ipforwarding == 0) { 289 ipstat.ips_cantforward++; 290 m_freem(m); 291 } else 292 ip_forward(m, 0); 293 goto next; 294 295 ours: 296 /* 297 * If offset or IP_MF are set, must reassemble. 298 * Otherwise, nothing need be done. 299 * (We could look in the reassembly queue to see 300 * if the packet was previously fragmented, 301 * but it's not worth the time; just let them time out.) 302 */ 303 if (ip->ip_off &~ IP_DF) { 304 if (m->m_flags & M_EXT) { /* XXX */ 305 if ((m = m_pullup(m, sizeof (struct ip))) == 0) { 306 ipstat.ips_toosmall++; 307 goto next; 308 } 309 ip = mtod(m, struct ip *); 310 } 311 /* 312 * Look for queue of fragments 313 * of this datagram. 314 */ 315 for (fp = ipq.next; fp != &ipq; fp = fp->next) 316 if (ip->ip_id == fp->ipq_id && 317 ip->ip_src.s_addr == fp->ipq_src.s_addr && 318 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 319 ip->ip_p == fp->ipq_p) 320 goto found; 321 fp = 0; 322 found: 323 324 /* 325 * Adjust ip_len to not reflect header, 326 * set ip_mff if more fragments are expected, 327 * convert offset of this to bytes. 328 */ 329 ip->ip_len -= hlen; 330 ((struct ipasfrag *)ip)->ipf_mff = 0; 331 if (ip->ip_off & IP_MF) 332 ((struct ipasfrag *)ip)->ipf_mff = 1; 333 ip->ip_off <<= 3; 334 335 /* 336 * If datagram marked as having more fragments 337 * or if this is not the first fragment, 338 * attempt reassembly; if it succeeds, proceed. 339 */ 340 if (((struct ipasfrag *)ip)->ipf_mff || ip->ip_off) { 341 ipstat.ips_fragments++; 342 ip = ip_reass((struct ipasfrag *)ip, fp); 343 if (ip == 0) 344 goto next; 345 else 346 ipstat.ips_reassembled++; 347 m = dtom(ip); 348 } else 349 if (fp) 350 ip_freef(fp); 351 } else 352 ip->ip_len -= hlen; 353 354 /* 355 * Switch out to protocol's input routine. 356 */ 357 ipstat.ips_delivered++; 358 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 359 goto next; 360 bad: 361 m_freem(m); 362 goto next; 363 } 364 365 /* 366 * Take incoming datagram fragment and try to 367 * reassemble it into whole datagram. If a chain for 368 * reassembly of this datagram already exists, then it 369 * is given as fp; otherwise have to make a chain. 370 */ 371 struct ip * 372 ip_reass(ip, fp) 373 register struct ipasfrag *ip; 374 register struct ipq *fp; 375 { 376 register struct mbuf *m = dtom(ip); 377 register struct ipasfrag *q; 378 struct mbuf *t; 379 int hlen = ip->ip_hl << 2; 380 int i, next; 381 382 /* 383 * Presence of header sizes in mbufs 384 * would confuse code below. 385 */ 386 m->m_data += hlen; 387 m->m_len -= hlen; 388 389 /* 390 * If first fragment to arrive, create a reassembly queue. 391 */ 392 if (fp == 0) { 393 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 394 goto dropfrag; 395 fp = mtod(t, struct ipq *); 396 insque(fp, &ipq); 397 fp->ipq_ttl = IPFRAGTTL; 398 fp->ipq_p = ip->ip_p; 399 fp->ipq_id = ip->ip_id; 400 fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp; 401 fp->ipq_src = ((struct ip *)ip)->ip_src; 402 fp->ipq_dst = ((struct ip *)ip)->ip_dst; 403 q = (struct ipasfrag *)fp; 404 goto insert; 405 } 406 407 /* 408 * Find a segment which begins after this one does. 409 */ 410 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) 411 if (q->ip_off > ip->ip_off) 412 break; 413 414 /* 415 * If there is a preceding segment, it may provide some of 416 * our data already. If so, drop the data from the incoming 417 * segment. If it provides all of our data, drop us. 418 */ 419 if (q->ipf_prev != (struct ipasfrag *)fp) { 420 i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off; 421 if (i > 0) { 422 if (i >= ip->ip_len) 423 goto dropfrag; 424 m_adj(dtom(ip), i); 425 ip->ip_off += i; 426 ip->ip_len -= i; 427 } 428 } 429 430 /* 431 * While we overlap succeeding segments trim them or, 432 * if they are completely covered, dequeue them. 433 */ 434 while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) { 435 i = (ip->ip_off + ip->ip_len) - q->ip_off; 436 if (i < q->ip_len) { 437 q->ip_len -= i; 438 q->ip_off += i; 439 m_adj(dtom(q), i); 440 break; 441 } 442 q = q->ipf_next; 443 m_freem(dtom(q->ipf_prev)); 444 ip_deq(q->ipf_prev); 445 } 446 447 insert: 448 /* 449 * Stick new segment in its place; 450 * check for complete reassembly. 451 */ 452 ip_enq(ip, q->ipf_prev); 453 next = 0; 454 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) { 455 if (q->ip_off != next) 456 return (0); 457 next += q->ip_len; 458 } 459 if (q->ipf_prev->ipf_mff) 460 return (0); 461 462 /* 463 * Reassembly is complete; concatenate fragments. 464 */ 465 q = fp->ipq_next; 466 m = dtom(q); 467 t = m->m_next; 468 m->m_next = 0; 469 m_cat(m, t); 470 q = q->ipf_next; 471 while (q != (struct ipasfrag *)fp) { 472 t = dtom(q); 473 q = q->ipf_next; 474 m_cat(m, t); 475 } 476 477 /* 478 * Create header for new ip packet by 479 * modifying header of first packet; 480 * dequeue and discard fragment reassembly header. 481 * Make header visible. 482 */ 483 ip = fp->ipq_next; 484 ip->ip_len = next; 485 ((struct ip *)ip)->ip_src = fp->ipq_src; 486 ((struct ip *)ip)->ip_dst = fp->ipq_dst; 487 remque(fp); 488 (void) m_free(dtom(fp)); 489 m = dtom(ip); 490 m->m_len += (ip->ip_hl << 2); 491 m->m_data -= (ip->ip_hl << 2); 492 /* some debugging cruft by sklower, below, will go away soon */ 493 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 494 register int plen = 0; 495 for (t = m; m; m = m->m_next) 496 plen += m->m_len; 497 t->m_pkthdr.len = plen; 498 } 499 return ((struct ip *)ip); 500 501 dropfrag: 502 ipstat.ips_fragdropped++; 503 m_freem(m); 504 return (0); 505 } 506 507 /* 508 * Free a fragment reassembly header and all 509 * associated datagrams. 510 */ 511 ip_freef(fp) 512 struct ipq *fp; 513 { 514 register struct ipasfrag *q, *p; 515 516 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) { 517 p = q->ipf_next; 518 ip_deq(q); 519 m_freem(dtom(q)); 520 } 521 remque(fp); 522 (void) m_free(dtom(fp)); 523 } 524 525 /* 526 * Put an ip fragment on a reassembly chain. 527 * Like insque, but pointers in middle of structure. 528 */ 529 ip_enq(p, prev) 530 register struct ipasfrag *p, *prev; 531 { 532 533 p->ipf_prev = prev; 534 p->ipf_next = prev->ipf_next; 535 prev->ipf_next->ipf_prev = p; 536 prev->ipf_next = p; 537 } 538 539 /* 540 * To ip_enq as remque is to insque. 541 */ 542 ip_deq(p) 543 register struct ipasfrag *p; 544 { 545 546 p->ipf_prev->ipf_next = p->ipf_next; 547 p->ipf_next->ipf_prev = p->ipf_prev; 548 } 549 550 /* 551 * IP timer processing; 552 * if a timer expires on a reassembly 553 * queue, discard it. 554 */ 555 ip_slowtimo() 556 { 557 register struct ipq *fp; 558 int s = splnet(); 559 560 fp = ipq.next; 561 if (fp == 0) { 562 splx(s); 563 return; 564 } 565 while (fp != &ipq) { 566 --fp->ipq_ttl; 567 fp = fp->next; 568 if (fp->prev->ipq_ttl == 0) { 569 ipstat.ips_fragtimeout++; 570 ip_freef(fp->prev); 571 } 572 } 573 splx(s); 574 } 575 576 /* 577 * Drain off all datagram fragments. 578 */ 579 ip_drain() 580 { 581 582 while (ipq.next != &ipq) { 583 ipstat.ips_fragdropped++; 584 ip_freef(ipq.next); 585 } 586 } 587 588 extern struct in_ifaddr *ifptoia(); 589 struct in_ifaddr *ip_rtaddr(); 590 591 /* 592 * Do option processing on a datagram, 593 * possibly discarding it if bad options are encountered, 594 * or forwarding it if source-routed. 595 * Returns 1 if packet has been forwarded/freed, 596 * 0 if the packet should be processed further. 597 */ 598 ip_dooptions(m) 599 struct mbuf *m; 600 { 601 register struct ip *ip = mtod(m, struct ip *); 602 register u_char *cp; 603 register struct ip_timestamp *ipt; 604 register struct in_ifaddr *ia; 605 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 606 struct in_addr *sin; 607 n_time ntime; 608 609 cp = (u_char *)(ip + 1); 610 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 611 for (; cnt > 0; cnt -= optlen, cp += optlen) { 612 opt = cp[IPOPT_OPTVAL]; 613 if (opt == IPOPT_EOL) 614 break; 615 if (opt == IPOPT_NOP) 616 optlen = 1; 617 else { 618 optlen = cp[IPOPT_OLEN]; 619 if (optlen <= 0 || optlen > cnt) { 620 code = &cp[IPOPT_OLEN] - (u_char *)ip; 621 goto bad; 622 } 623 } 624 switch (opt) { 625 626 default: 627 break; 628 629 /* 630 * Source routing with record. 631 * Find interface with current destination address. 632 * If none on this machine then drop if strictly routed, 633 * or do nothing if loosely routed. 634 * Record interface address and bring up next address 635 * component. If strictly routed make sure next 636 * address is on directly accessible net. 637 */ 638 case IPOPT_LSRR: 639 case IPOPT_SSRR: 640 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 641 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 642 goto bad; 643 } 644 ipaddr.sin_addr = ip->ip_dst; 645 ia = (struct in_ifaddr *) 646 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 647 if (ia == 0) { 648 if (opt == IPOPT_SSRR) { 649 type = ICMP_UNREACH; 650 code = ICMP_UNREACH_SRCFAIL; 651 goto bad; 652 } 653 /* 654 * Loose routing, and not at next destination 655 * yet; nothing to do except forward. 656 */ 657 break; 658 } 659 off--; /* 0 origin */ 660 if (off > optlen - sizeof(struct in_addr)) { 661 /* 662 * End of source route. Should be for us. 663 */ 664 save_rte(cp, ip->ip_src); 665 break; 666 } 667 /* 668 * locate outgoing interface 669 */ 670 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, 671 sizeof(ipaddr.sin_addr)); 672 if (opt == IPOPT_SSRR) { 673 #define INA struct in_ifaddr * 674 #define SA struct sockaddr * 675 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 676 ia = in_iaonnetof(in_netof(ipaddr.sin_addr)); 677 } else 678 ia = ip_rtaddr(ipaddr.sin_addr); 679 if (ia == 0) { 680 type = ICMP_UNREACH; 681 code = ICMP_UNREACH_SRCFAIL; 682 goto bad; 683 } 684 ip->ip_dst = ipaddr.sin_addr; 685 bcopy((caddr_t)&(IA_SIN(ia)->sin_addr), 686 (caddr_t)(cp + off), sizeof(struct in_addr)); 687 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 688 forward = 1; 689 break; 690 691 case IPOPT_RR: 692 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 693 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 694 goto bad; 695 } 696 /* 697 * If no space remains, ignore. 698 */ 699 off--; /* 0 origin */ 700 if (off > optlen - sizeof(struct in_addr)) 701 break; 702 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, 703 sizeof(ipaddr.sin_addr)); 704 /* 705 * locate outgoing interface; if we're the destination, 706 * use the incoming interface (should be same). 707 */ 708 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 709 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 710 type = ICMP_UNREACH; 711 code = ICMP_UNREACH_HOST; 712 goto bad; 713 } 714 bcopy((caddr_t)&(IA_SIN(ia)->sin_addr), 715 (caddr_t)(cp + off), sizeof(struct in_addr)); 716 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 717 break; 718 719 case IPOPT_TS: 720 code = cp - (u_char *)ip; 721 ipt = (struct ip_timestamp *)cp; 722 if (ipt->ipt_len < 5) 723 goto bad; 724 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) { 725 if (++ipt->ipt_oflw == 0) 726 goto bad; 727 break; 728 } 729 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 730 switch (ipt->ipt_flg) { 731 732 case IPOPT_TS_TSONLY: 733 break; 734 735 case IPOPT_TS_TSANDADDR: 736 if (ipt->ipt_ptr + sizeof(n_time) + 737 sizeof(struct in_addr) > ipt->ipt_len) 738 goto bad; 739 ia = ifptoia(m->m_pkthdr.rcvif); 740 bcopy((caddr_t)&IA_SIN(ia)->sin_addr, 741 (caddr_t)sin, sizeof(struct in_addr)); 742 ipt->ipt_ptr += sizeof(struct in_addr); 743 break; 744 745 case IPOPT_TS_PRESPEC: 746 if (ipt->ipt_ptr + sizeof(n_time) + 747 sizeof(struct in_addr) > ipt->ipt_len) 748 goto bad; 749 bcopy((caddr_t)sin, (caddr_t)&ipaddr.sin_addr, 750 sizeof(struct in_addr)); 751 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 752 continue; 753 ipt->ipt_ptr += sizeof(struct in_addr); 754 break; 755 756 default: 757 goto bad; 758 } 759 ntime = iptime(); 760 bcopy((caddr_t)&ntime, (caddr_t)cp + ipt->ipt_ptr - 1, 761 sizeof(n_time)); 762 ipt->ipt_ptr += sizeof(n_time); 763 } 764 } 765 if (forward) { 766 ip_forward(m, 1); 767 return (1); 768 } else 769 return (0); 770 bad: 771 icmp_error(m, type, code); 772 return (1); 773 } 774 775 /* 776 * Given address of next destination (final or next hop), 777 * return internet address info of interface to be used to get there. 778 */ 779 struct in_ifaddr * 780 ip_rtaddr(dst) 781 struct in_addr dst; 782 { 783 register struct sockaddr_in *sin; 784 785 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 786 787 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 788 if (ipforward_rt.ro_rt) { 789 RTFREE(ipforward_rt.ro_rt); 790 ipforward_rt.ro_rt = 0; 791 } 792 sin->sin_family = AF_INET; 793 sin->sin_len = sizeof(*sin); 794 sin->sin_addr = dst; 795 796 rtalloc(&ipforward_rt); 797 } 798 if (ipforward_rt.ro_rt == 0) 799 return ((struct in_ifaddr *)0); 800 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 801 } 802 803 /* 804 * Save incoming source route for use in replies, 805 * to be picked up later by ip_srcroute if the receiver is interested. 806 */ 807 save_rte(option, dst) 808 u_char *option; 809 struct in_addr dst; 810 { 811 unsigned olen; 812 813 olen = option[IPOPT_OLEN]; 814 #ifdef DIAGNOSTIC 815 if (ipprintfs) 816 printf("save_rte: olen %d\n", olen); 817 #endif 818 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 819 return; 820 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); 821 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 822 ip_srcrt.dst = dst; 823 } 824 825 /* 826 * Retrieve incoming source route for use in replies, 827 * in the same form used by setsockopt. 828 * The first hop is placed before the options, will be removed later. 829 */ 830 struct mbuf * 831 ip_srcroute() 832 { 833 register struct in_addr *p, *q; 834 register struct mbuf *m; 835 836 if (ip_nhops == 0) 837 return ((struct mbuf *)0); 838 m = m_get(M_DONTWAIT, MT_SOOPTS); 839 if (m == 0) 840 return ((struct mbuf *)0); 841 842 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 843 844 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 845 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 846 OPTSIZ; 847 #ifdef DIAGNOSTIC 848 if (ipprintfs) 849 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 850 #endif 851 852 /* 853 * First save first hop for return route 854 */ 855 p = &ip_srcrt.route[ip_nhops - 1]; 856 *(mtod(m, struct in_addr *)) = *p--; 857 #ifdef DIAGNOSTIC 858 if (ipprintfs) 859 printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr)); 860 #endif 861 862 /* 863 * Copy option fields and padding (nop) to mbuf. 864 */ 865 ip_srcrt.nop = IPOPT_NOP; 866 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 867 bcopy((caddr_t)&ip_srcrt.nop, 868 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); 869 q = (struct in_addr *)(mtod(m, caddr_t) + 870 sizeof(struct in_addr) + OPTSIZ); 871 #undef OPTSIZ 872 /* 873 * Record return path as an IP source route, 874 * reversing the path (pointers are now aligned). 875 */ 876 while (p >= ip_srcrt.route) { 877 #ifdef DIAGNOSTIC 878 if (ipprintfs) 879 printf(" %lx", ntohl(q->s_addr)); 880 #endif 881 *q++ = *p--; 882 } 883 /* 884 * Last hop goes to final destination. 885 */ 886 *q = ip_srcrt.dst; 887 #ifdef DIAGNOSTIC 888 if (ipprintfs) 889 printf(" %lx\n", ntohl(q->s_addr)); 890 #endif 891 return (m); 892 } 893 894 /* 895 * Strip out IP options, at higher 896 * level protocol in the kernel. 897 * Second argument is buffer to which options 898 * will be moved, and return value is their length. 899 * XXX should be deleted; last arg currently ignored. 900 */ 901 ip_stripoptions(m, mopt) 902 register struct mbuf *m; 903 struct mbuf *mopt; 904 { 905 register int i; 906 struct ip *ip = mtod(m, struct ip *); 907 register caddr_t opts; 908 int olen; 909 910 olen = (ip->ip_hl<<2) - sizeof (struct ip); 911 opts = (caddr_t)(ip + 1); 912 i = m->m_len - (sizeof (struct ip) + olen); 913 bcopy(opts + olen, opts, (unsigned)i); 914 m->m_len -= olen; 915 if (m->m_flags & M_PKTHDR) 916 m->m_pkthdr.len -= olen; 917 ip->ip_hl = sizeof(struct ip) >> 2; 918 } 919 920 u_char inetctlerrmap[PRC_NCMDS] = { 921 0, 0, 0, 0, 922 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 923 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 924 EMSGSIZE, EHOSTUNREACH, 0, 0, 925 0, 0, 0, 0, 926 ENOPROTOOPT 927 }; 928 929 /* 930 * Forward a packet. If some error occurs return the sender 931 * an icmp packet. Note we can't always generate a meaningful 932 * icmp message because icmp doesn't have a large enough repertoire 933 * of codes and types. 934 * 935 * If not forwarding, just drop the packet. This could be confusing 936 * if ipforwarding was zero but some routing protocol was advancing 937 * us as a gateway to somewhere. However, we must let the routing 938 * protocol deal with that. 939 * 940 * The srcrt parameter indicates whether the packet is being forwarded 941 * via a source route. 942 */ 943 ip_forward(m, srcrt) 944 struct mbuf *m; 945 int srcrt; 946 { 947 register struct ip *ip = mtod(m, struct ip *); 948 register struct sockaddr_in *sin; 949 register struct rtentry *rt; 950 int error, type = 0, code; 951 struct mbuf *mcopy; 952 struct in_addr dest; 953 954 dest.s_addr = 0; 955 #ifdef DIAGNOSTIC 956 if (ipprintfs) 957 printf("forward: src %x dst %x ttl %x\n", ip->ip_src, 958 ip->ip_dst, ip->ip_ttl); 959 #endif 960 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) { 961 ipstat.ips_cantforward++; 962 m_freem(m); 963 return; 964 } 965 HTONS(ip->ip_id); 966 if (ip->ip_ttl <= IPTTLDEC) { 967 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest); 968 return; 969 } 970 ip->ip_ttl -= IPTTLDEC; 971 972 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 973 if ((rt = ipforward_rt.ro_rt) == 0 || 974 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 975 if (ipforward_rt.ro_rt) { 976 RTFREE(ipforward_rt.ro_rt); 977 ipforward_rt.ro_rt = 0; 978 } 979 sin->sin_family = AF_INET; 980 sin->sin_len = sizeof(*sin); 981 sin->sin_addr = ip->ip_dst; 982 983 rtalloc(&ipforward_rt); 984 if (ipforward_rt.ro_rt == 0) { 985 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest); 986 return; 987 } 988 rt = ipforward_rt.ro_rt; 989 } 990 991 /* 992 * Save at most 64 bytes of the packet in case 993 * we need to generate an ICMP message to the src. 994 */ 995 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 996 997 #ifdef GATEWAY 998 ip_ifmatrix[rt->rt_ifp->if_index + 999 if_index * m->m_pkthdr.rcvif->if_index]++; 1000 #endif 1001 /* 1002 * If forwarding packet using same interface that it came in on, 1003 * perhaps should send a redirect to sender to shortcut a hop. 1004 * Only send redirect if source is sending directly to us, 1005 * and if packet was not source routed (or has any options). 1006 * Also, don't send redirect if forwarding using a default route 1007 * or a route modified by a redirect. 1008 */ 1009 #define satosin(sa) ((struct sockaddr_in *)(sa)) 1010 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1011 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1012 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1013 ipsendredirects && !srcrt) { 1014 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1015 u_long src = ntohl(ip->ip_src.s_addr); 1016 u_long dst = ntohl(ip->ip_dst.s_addr); 1017 1018 if (RTA(rt) && 1019 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1020 if (rt->rt_flags & RTF_GATEWAY) 1021 dest = satosin(rt->rt_gateway)->sin_addr; 1022 else 1023 dest = ip->ip_dst; 1024 /* 1025 * If the destination is reached by a route to host, 1026 * is on a subnet of a local net, or is directly 1027 * on the attached net (!), use host redirect. 1028 * (We may be the correct first hop for other subnets.) 1029 */ 1030 type = ICMP_REDIRECT; 1031 if ((rt->rt_flags & RTF_HOST) || 1032 (rt->rt_flags & RTF_GATEWAY) == 0) 1033 code = ICMP_REDIRECT_HOST; 1034 else if (RTA(rt)->ia_subnetmask != RTA(rt)->ia_netmask && 1035 (dst & RTA(rt)->ia_netmask) == RTA(rt)->ia_net) 1036 code = ICMP_REDIRECT_HOST; 1037 else 1038 code = ICMP_REDIRECT_NET; 1039 #ifdef DIAGNOSTIC 1040 if (ipprintfs) 1041 printf("redirect (%d) to %x\n", code, dest.s_addr); 1042 #endif 1043 } 1044 } 1045 1046 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, IP_FORWARDING); 1047 if (error) 1048 ipstat.ips_cantforward++; 1049 else { 1050 ipstat.ips_forward++; 1051 if (type) 1052 ipstat.ips_redirectsent++; 1053 else { 1054 if (mcopy) 1055 m_freem(mcopy); 1056 return; 1057 } 1058 } 1059 if (mcopy == NULL) 1060 return; 1061 switch (error) { 1062 1063 case 0: /* forwarded, but need redirect */ 1064 /* type, code set above */ 1065 break; 1066 1067 case ENETUNREACH: /* shouldn't happen, checked above */ 1068 case EHOSTUNREACH: 1069 case ENETDOWN: 1070 case EHOSTDOWN: 1071 default: 1072 type = ICMP_UNREACH; 1073 code = ICMP_UNREACH_HOST; 1074 break; 1075 1076 case EMSGSIZE: 1077 type = ICMP_UNREACH; 1078 code = ICMP_UNREACH_NEEDFRAG; 1079 ipstat.ips_cantfrag++; 1080 break; 1081 1082 case ENOBUFS: 1083 type = ICMP_SOURCEQUENCH; 1084 code = 0; 1085 break; 1086 } 1087 icmp_error(mcopy, type, code, dest); 1088 } 1089