ip_input.c revision 31163
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 1.71 1997/11/05 20:17:21 joerg Exp $ 35 * $ANA: ip_input.c,v 1.5 1996/09/18 14:34:59 wollman Exp $ 36 */ 37 38#define _IP_VHL 39 40#include "opt_ipfw.h" 41#include "opt_ipdivert.h" 42 43#include <stddef.h> 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/mbuf.h> 48#include <sys/domain.h> 49#include <sys/protosw.h> 50#include <sys/socket.h> 51#include <sys/time.h> 52#include <sys/kernel.h> 53#include <sys/syslog.h> 54#include <sys/sysctl.h> 55 56#include <net/if.h> 57#include <net/if_var.h> 58#include <net/if_dl.h> 59#include <net/route.h> 60#include <net/netisr.h> 61 62#include <netinet/in.h> 63#include <netinet/in_systm.h> 64#include <netinet/in_var.h> 65#include <netinet/ip.h> 66#include <netinet/in_pcb.h> 67#include <netinet/ip_var.h> 68#include <netinet/ip_icmp.h> 69#include <machine/in_cksum.h> 70 71#include <sys/socketvar.h> 72 73#ifdef IPFIREWALL 74#include <netinet/ip_fw.h> 75#endif 76 77int rsvp_on = 0; 78static int ip_rsvp_on; 79struct socket *ip_rsvpd; 80 81static int ipforwarding = 0; 82SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 83 &ipforwarding, 0, ""); 84 85static int ipsendredirects = 1; /* XXX */ 86SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 87 &ipsendredirects, 0, ""); 88 89int ip_defttl = IPDEFTTL; 90SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 91 &ip_defttl, 0, ""); 92 93static int ip_dosourceroute = 0; 94SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 95 &ip_dosourceroute, 0, ""); 96#ifdef DIAGNOSTIC 97static int ipprintfs = 0; 98#endif 99 100extern struct domain inetdomain; 101extern struct protosw inetsw[]; 102u_char ip_protox[IPPROTO_MAX]; 103static int ipqmaxlen = IFQ_MAXLEN; 104struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 105struct ifqueue ipintrq; 106SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RD, 107 &ipintrq.ifq_maxlen, 0, ""); 108SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 109 &ipintrq.ifq_drops, 0, ""); 110 111struct ipstat ipstat; 112SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD, 113 &ipstat, ipstat, ""); 114 115/* Packet reassembly stuff */ 116#define IPREASS_NHASH_LOG2 6 117#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 118#define IPREASS_HMASK (IPREASS_NHASH - 1) 119#define IPREASS_HASH(x,y) \ 120 ((((x) & 0xF | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 121 122static struct ipq ipq[IPREASS_NHASH]; 123static int nipq = 0; /* total # of reass queues */ 124static int maxnipq; 125 126#ifdef IPCTL_DEFMTU 127SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 128 &ip_mtu, 0, ""); 129#endif 130 131#if !defined(COMPAT_IPFW) || COMPAT_IPFW == 1 132#undef COMPAT_IPFW 133#define COMPAT_IPFW 1 134#else 135#undef COMPAT_IPFW 136#endif 137 138#ifdef COMPAT_IPFW 139/* Firewall hooks */ 140ip_fw_chk_t *ip_fw_chk_ptr; 141ip_fw_ctl_t *ip_fw_ctl_ptr; 142 143/* IP Network Address Translation (NAT) hooks */ 144ip_nat_t *ip_nat_ptr; 145ip_nat_ctl_t *ip_nat_ctl_ptr; 146#endif 147 148#if defined(IPFILTER_LKM) || defined(IPFILTER) 149int fr_check __P((struct ip *, int, struct ifnet *, int, struct mbuf **)); 150int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL; 151#endif 152 153 154/* 155 * We need to save the IP options in case a protocol wants to respond 156 * to an incoming packet over the same route if the packet got here 157 * using IP source routing. This allows connection establishment and 158 * maintenance when the remote end is on a network that is not known 159 * to us. 160 */ 161static int ip_nhops = 0; 162static struct ip_srcrt { 163 struct in_addr dst; /* final destination */ 164 char nop; /* one NOP to align */ 165 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 166 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 167} ip_srcrt; 168 169#ifdef IPDIVERT 170/* 171 * Shared variable between ip_input() and ip_reass() to communicate 172 * about which packets, once assembled from fragments, get diverted, 173 * and to which port. 174 */ 175static u_short frag_divert_port; 176#endif 177 178static void save_rte __P((u_char *, struct in_addr)); 179static void ip_deq __P((struct ipasfrag *)); 180static int ip_dooptions __P((struct mbuf *)); 181static void ip_enq __P((struct ipasfrag *, struct ipasfrag *)); 182static void ip_forward __P((struct mbuf *, int)); 183static void ip_freef __P((struct ipq *)); 184static struct ip * 185 ip_reass __P((struct ipasfrag *, struct ipq *, struct ipq *)); 186static struct in_ifaddr * 187 ip_rtaddr __P((struct in_addr)); 188static void ipintr __P((void)); 189/* 190 * IP initialization: fill in IP protocol switch table. 191 * All protocols not implemented in kernel go to raw IP protocol handler. 192 */ 193void 194ip_init() 195{ 196 register struct protosw *pr; 197 register int i; 198 199 TAILQ_INIT(&in_ifaddrhead); 200 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 201 if (pr == 0) 202 panic("ip_init"); 203 for (i = 0; i < IPPROTO_MAX; i++) 204 ip_protox[i] = pr - inetsw; 205 for (pr = inetdomain.dom_protosw; 206 pr < inetdomain.dom_protoswNPROTOSW; pr++) 207 if (pr->pr_domain->dom_family == PF_INET && 208 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 209 ip_protox[pr->pr_protocol] = pr - inetsw; 210 211 for (i = 0; i < IPREASS_NHASH; i++) 212 ipq[i].next = ipq[i].prev = &ipq[i]; 213 214 maxnipq = nmbclusters/4; 215 216 ip_id = time.tv_sec & 0xffff; 217 ipintrq.ifq_maxlen = ipqmaxlen; 218#ifdef IPFIREWALL 219 ip_fw_init(); 220#endif 221#ifdef IPNAT 222 ip_nat_init(); 223#endif 224 225} 226 227static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 228static struct route ipforward_rt; 229 230/* 231 * Ip input routine. Checksum and byte swap header. If fragmented 232 * try to reassemble. Process options. Pass to next level. 233 */ 234void 235ip_input(struct mbuf *m) 236{ 237 struct ip *ip; 238 struct ipq *fp; 239 struct in_ifaddr *ia; 240 int i, hlen; 241 u_short sum; 242 243#ifdef DIAGNOSTIC 244 if ((m->m_flags & M_PKTHDR) == 0) 245 panic("ip_input no HDR"); 246#endif 247 /* 248 * If no IP addresses have been set yet but the interfaces 249 * are receiving, can't do anything with incoming packets yet. 250 * XXX This is broken! We should be able to receive broadcasts 251 * and multicasts even without any local addresses configured. 252 */ 253 if (TAILQ_EMPTY(&in_ifaddrhead)) 254 goto bad; 255 ipstat.ips_total++; 256 257 if (m->m_pkthdr.len < sizeof(struct ip)) 258 goto tooshort; 259 260#ifdef DIAGNOSTIC 261 if (m->m_len < sizeof(struct ip)) 262 panic("ipintr mbuf too short"); 263#endif 264 265 if (m->m_len < sizeof (struct ip) && 266 (m = m_pullup(m, sizeof (struct ip))) == 0) { 267 ipstat.ips_toosmall++; 268 return; 269 } 270 ip = mtod(m, struct ip *); 271 272 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 273 ipstat.ips_badvers++; 274 goto bad; 275 } 276 277 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 278 if (hlen < sizeof(struct ip)) { /* minimum header length */ 279 ipstat.ips_badhlen++; 280 goto bad; 281 } 282 if (hlen > m->m_len) { 283 if ((m = m_pullup(m, hlen)) == 0) { 284 ipstat.ips_badhlen++; 285 return; 286 } 287 ip = mtod(m, struct ip *); 288 } 289 if (hlen == sizeof(struct ip)) { 290 sum = in_cksum_hdr(ip); 291 } else { 292 sum = in_cksum(m, hlen); 293 } 294 if (sum) { 295 ipstat.ips_badsum++; 296 goto bad; 297 } 298 299 /* 300 * Convert fields to host representation. 301 */ 302 NTOHS(ip->ip_len); 303 if (ip->ip_len < hlen) { 304 ipstat.ips_badlen++; 305 goto bad; 306 } 307 NTOHS(ip->ip_id); 308 NTOHS(ip->ip_off); 309 310 /* 311 * Check that the amount of data in the buffers 312 * is as at least much as the IP header would have us expect. 313 * Trim mbufs if longer than we expect. 314 * Drop packet if shorter than we expect. 315 */ 316 if (m->m_pkthdr.len < ip->ip_len) { 317tooshort: 318 ipstat.ips_tooshort++; 319 goto bad; 320 } 321 if (m->m_pkthdr.len > ip->ip_len) { 322 if (m->m_len == m->m_pkthdr.len) { 323 m->m_len = ip->ip_len; 324 m->m_pkthdr.len = ip->ip_len; 325 } else 326 m_adj(m, ip->ip_len - m->m_pkthdr.len); 327 } 328 /* 329 * IpHack's section. 330 * Right now when no processing on packet has done 331 * and it is still fresh out of network we do our black 332 * deals with it. 333 * - Firewall: deny/allow/divert 334 * - Xlate: translate packet's addr/port (NAT). 335 * - Wrap: fake packet's addr/port <unimpl.> 336 * - Encapsulate: put it in another IP and send out. <unimp.> 337 */ 338#if defined(IPFILTER) || defined(IPFILTER_LKM) 339 /* 340 * Check if we want to allow this packet to be processed. 341 * Consider it to be bad if not. 342 */ 343 if (fr_check) { 344 struct mbuf *m1 = m; 345 346 if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1) 347 return; 348 ip = mtod(m = m1, struct ip *); 349 } 350#endif 351#ifdef COMPAT_IPFW 352 if (ip_fw_chk_ptr) { 353#ifdef IPDIVERT 354 u_short port; 355 356 port = (*ip_fw_chk_ptr)(&ip, hlen, NULL, ip_divert_ignore, &m); 357 ip_divert_ignore = 0; 358 if (port) { /* Divert packet */ 359 frag_divert_port = port; 360 goto ours; 361 } 362#else 363 /* If ipfw says divert, we have to just drop packet */ 364 if ((*ip_fw_chk_ptr)(&ip, hlen, NULL, 0, &m)) { 365 m_freem(m); 366 m = NULL; 367 } 368#endif 369 if (!m) 370 return; 371 } 372 373 if (ip_nat_ptr && !(*ip_nat_ptr)(&ip, &m, m->m_pkthdr.rcvif, IP_NAT_IN)) 374 return; 375#endif 376 377 /* 378 * Process options and, if not destined for us, 379 * ship it on. ip_dooptions returns 1 when an 380 * error was detected (causing an icmp message 381 * to be sent and the original packet to be freed). 382 */ 383 ip_nhops = 0; /* for source routed packets */ 384 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 385 return; 386 387 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 388 * matter if it is destined to another node, or whether it is 389 * a multicast one, RSVP wants it! and prevents it from being forwarded 390 * anywhere else. Also checks if the rsvp daemon is running before 391 * grabbing the packet. 392 */ 393 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 394 goto ours; 395 396 /* 397 * Check our list of addresses, to see if the packet is for us. 398 */ 399 for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) { 400#define satosin(sa) ((struct sockaddr_in *)(sa)) 401 402 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 403 goto ours; 404#ifdef BOOTP_COMPAT 405 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 406 goto ours; 407#endif 408 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 409 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 410 ip->ip_dst.s_addr) 411 goto ours; 412 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 413 goto ours; 414 } 415 } 416 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 417 struct in_multi *inm; 418 if (ip_mrouter) { 419 /* 420 * If we are acting as a multicast router, all 421 * incoming multicast packets are passed to the 422 * kernel-level multicast forwarding function. 423 * The packet is returned (relatively) intact; if 424 * ip_mforward() returns a non-zero value, the packet 425 * must be discarded, else it may be accepted below. 426 * 427 * (The IP ident field is put in the same byte order 428 * as expected when ip_mforward() is called from 429 * ip_output().) 430 */ 431 ip->ip_id = htons(ip->ip_id); 432 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 433 ipstat.ips_cantforward++; 434 m_freem(m); 435 return; 436 } 437 ip->ip_id = ntohs(ip->ip_id); 438 439 /* 440 * The process-level routing demon needs to receive 441 * all multicast IGMP packets, whether or not this 442 * host belongs to their destination groups. 443 */ 444 if (ip->ip_p == IPPROTO_IGMP) 445 goto ours; 446 ipstat.ips_forward++; 447 } 448 /* 449 * See if we belong to the destination multicast group on the 450 * arrival interface. 451 */ 452 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 453 if (inm == NULL) { 454 ipstat.ips_notmember++; 455 m_freem(m); 456 return; 457 } 458 goto ours; 459 } 460 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 461 goto ours; 462 if (ip->ip_dst.s_addr == INADDR_ANY) 463 goto ours; 464 465 /* 466 * Not for us; forward if possible and desirable. 467 */ 468 if (ipforwarding == 0) { 469 ipstat.ips_cantforward++; 470 m_freem(m); 471 } else 472 ip_forward(m, 0); 473 return; 474 475ours: 476 477 /* 478 * If offset or IP_MF are set, must reassemble. 479 * Otherwise, nothing need be done. 480 * (We could look in the reassembly queue to see 481 * if the packet was previously fragmented, 482 * but it's not worth the time; just let them time out.) 483 */ 484 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 485 if (m->m_flags & M_EXT) { /* XXX */ 486 if ((m = m_pullup(m, sizeof (struct ip))) == 0) { 487 ipstat.ips_toosmall++; 488#ifdef IPDIVERT 489 frag_divert_port = 0; 490#endif 491 return; 492 } 493 ip = mtod(m, struct ip *); 494 } 495 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 496 /* 497 * Look for queue of fragments 498 * of this datagram. 499 */ 500 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 501 if (ip->ip_id == fp->ipq_id && 502 ip->ip_src.s_addr == fp->ipq_src.s_addr && 503 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 504 ip->ip_p == fp->ipq_p) 505 goto found; 506 507 fp = 0; 508 509 /* check if there's a place for the new queue */ 510 if (nipq > maxnipq) { 511 /* 512 * drop something from the tail of the current queue 513 * before proceeding further 514 */ 515 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 516 for (i = 0; i < IPREASS_NHASH; i++) { 517 if (ipq[i].prev != &ipq[i]) { 518 ip_freef(ipq[i].prev); 519 break; 520 } 521 } 522 } else 523 ip_freef(ipq[sum].prev); 524 } 525found: 526 /* 527 * Adjust ip_len to not reflect header, 528 * set ip_mff if more fragments are expected, 529 * convert offset of this to bytes. 530 */ 531 ip->ip_len -= hlen; 532 ((struct ipasfrag *)ip)->ipf_mff &= ~1; 533 if (ip->ip_off & IP_MF) 534 ((struct ipasfrag *)ip)->ipf_mff |= 1; 535 ip->ip_off <<= 3; 536 537 /* 538 * If datagram marked as having more fragments 539 * or if this is not the first fragment, 540 * attempt reassembly; if it succeeds, proceed. 541 */ 542 if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) { 543 ipstat.ips_fragments++; 544 ip = ip_reass((struct ipasfrag *)ip, fp, &ipq[sum]); 545 if (ip == 0) 546 return; 547 ipstat.ips_reassembled++; 548 m = dtom(ip); 549#ifdef IPDIVERT 550 if (frag_divert_port) { 551 ip->ip_len += hlen; 552 HTONS(ip->ip_len); 553 HTONS(ip->ip_off); 554 HTONS(ip->ip_id); 555 ip->ip_sum = 0; 556 ip->ip_sum = in_cksum_hdr(ip); 557 NTOHS(ip->ip_id); 558 NTOHS(ip->ip_off); 559 NTOHS(ip->ip_len); 560 ip->ip_len -= hlen; 561 } 562#endif 563 } else 564 if (fp) 565 ip_freef(fp); 566 } else 567 ip->ip_len -= hlen; 568 569#ifdef IPDIVERT 570 /* 571 * Divert reassembled packets to the divert protocol if required 572 */ 573 if (frag_divert_port) { 574 ipstat.ips_delivered++; 575 ip_divert_port = frag_divert_port; 576 frag_divert_port = 0; 577 (*inetsw[ip_protox[IPPROTO_DIVERT]].pr_input)(m, hlen); 578 return; 579 } 580 581 /* Don't let packets divert themselves */ 582 if (ip->ip_p == IPPROTO_DIVERT) { 583 ipstat.ips_noproto++; 584 goto bad; 585 } 586#endif 587 588 /* 589 * Switch out to protocol's input routine. 590 */ 591 ipstat.ips_delivered++; 592 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 593 return; 594bad: 595 m_freem(m); 596} 597 598/* 599 * IP software interrupt routine - to go away sometime soon 600 */ 601static void 602ipintr(void) 603{ 604 int s; 605 struct mbuf *m; 606 607 while(1) { 608 s = splimp(); 609 IF_DEQUEUE(&ipintrq, m); 610 splx(s); 611 if (m == 0) 612 return; 613 ip_input(m); 614 } 615} 616 617NETISR_SET(NETISR_IP, ipintr); 618 619/* 620 * Take incoming datagram fragment and try to 621 * reassemble it into whole datagram. If a chain for 622 * reassembly of this datagram already exists, then it 623 * is given as fp; otherwise have to make a chain. 624 */ 625static struct ip * 626ip_reass(ip, fp, where) 627 register struct ipasfrag *ip; 628 register struct ipq *fp; 629 struct ipq *where; 630{ 631 register struct mbuf *m = dtom(ip); 632 register struct ipasfrag *q; 633 struct mbuf *t; 634 int hlen = ip->ip_hl << 2; 635 int i, next; 636 637 /* 638 * Presence of header sizes in mbufs 639 * would confuse code below. 640 */ 641 m->m_data += hlen; 642 m->m_len -= hlen; 643 644 /* 645 * If first fragment to arrive, create a reassembly queue. 646 */ 647 if (fp == 0) { 648 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 649 goto dropfrag; 650 fp = mtod(t, struct ipq *); 651 insque(fp, where); 652 nipq++; 653 fp->ipq_ttl = IPFRAGTTL; 654 fp->ipq_p = ip->ip_p; 655 fp->ipq_id = ip->ip_id; 656 fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp; 657 fp->ipq_src = ((struct ip *)ip)->ip_src; 658 fp->ipq_dst = ((struct ip *)ip)->ip_dst; 659#ifdef IPDIVERT 660 fp->ipq_divert = 0; 661#endif 662 q = (struct ipasfrag *)fp; 663 goto insert; 664 } 665 666 /* 667 * Find a segment which begins after this one does. 668 */ 669 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) 670 if (q->ip_off > ip->ip_off) 671 break; 672 673 /* 674 * If there is a preceding segment, it may provide some of 675 * our data already. If so, drop the data from the incoming 676 * segment. If it provides all of our data, drop us. 677 */ 678 if (q->ipf_prev != (struct ipasfrag *)fp) { 679 i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off; 680 if (i > 0) { 681 if (i >= ip->ip_len) 682 goto dropfrag; 683 m_adj(dtom(ip), i); 684 ip->ip_off += i; 685 ip->ip_len -= i; 686 } 687 } 688 689 /* 690 * While we overlap succeeding segments trim them or, 691 * if they are completely covered, dequeue them. 692 */ 693 while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) { 694 struct mbuf *m0; 695 696 i = (ip->ip_off + ip->ip_len) - q->ip_off; 697 if (i < q->ip_len) { 698 q->ip_len -= i; 699 q->ip_off += i; 700 m_adj(dtom(q), i); 701 break; 702 } 703 m0 = dtom(q); 704 q = q->ipf_next; 705 ip_deq(q->ipf_prev); 706 m_freem(m0); 707 } 708 709insert: 710 711#ifdef IPDIVERT 712 /* 713 * Any fragment diverting causes the whole packet to divert 714 */ 715 if (frag_divert_port != 0) 716 fp->ipq_divert = frag_divert_port; 717 frag_divert_port = 0; 718#endif 719 720 /* 721 * Stick new segment in its place; 722 * check for complete reassembly. 723 */ 724 ip_enq(ip, q->ipf_prev); 725 next = 0; 726 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) { 727 if (q->ip_off != next) 728 return (0); 729 next += q->ip_len; 730 } 731 if (q->ipf_prev->ipf_mff & 1) 732 return (0); 733 734 /* 735 * Reassembly is complete. Make sure the packet is a sane size. 736 */ 737 if (next + (IP_VHL_HL(((struct ip *)fp->ipq_next)->ip_vhl) << 2) 738 > IP_MAXPACKET) { 739 ipstat.ips_toolong++; 740 ip_freef(fp); 741 return (0); 742 } 743 744 /* 745 * Concatenate fragments. 746 */ 747 q = fp->ipq_next; 748 m = dtom(q); 749 t = m->m_next; 750 m->m_next = 0; 751 m_cat(m, t); 752 q = q->ipf_next; 753 while (q != (struct ipasfrag *)fp) { 754 t = dtom(q); 755 q = q->ipf_next; 756 m_cat(m, t); 757 } 758 759#ifdef IPDIVERT 760 /* 761 * Record divert port for packet, if any 762 */ 763 frag_divert_port = fp->ipq_divert; 764#endif 765 766 /* 767 * Create header for new ip packet by 768 * modifying header of first packet; 769 * dequeue and discard fragment reassembly header. 770 * Make header visible. 771 */ 772 ip = fp->ipq_next; 773 ip->ip_len = next; 774 ip->ipf_mff &= ~1; 775 ((struct ip *)ip)->ip_src = fp->ipq_src; 776 ((struct ip *)ip)->ip_dst = fp->ipq_dst; 777 remque(fp); 778 nipq--; 779 (void) m_free(dtom(fp)); 780 m = dtom(ip); 781 m->m_len += (ip->ip_hl << 2); 782 m->m_data -= (ip->ip_hl << 2); 783 /* some debugging cruft by sklower, below, will go away soon */ 784 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 785 register int plen = 0; 786 for (t = m; m; m = m->m_next) 787 plen += m->m_len; 788 t->m_pkthdr.len = plen; 789 } 790 return ((struct ip *)ip); 791 792dropfrag: 793 ipstat.ips_fragdropped++; 794 m_freem(m); 795 return (0); 796} 797 798/* 799 * Free a fragment reassembly header and all 800 * associated datagrams. 801 */ 802static void 803ip_freef(fp) 804 struct ipq *fp; 805{ 806 register struct ipasfrag *q, *p; 807 808 for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) { 809 p = q->ipf_next; 810 ip_deq(q); 811 m_freem(dtom(q)); 812 } 813 remque(fp); 814 (void) m_free(dtom(fp)); 815 nipq--; 816} 817 818/* 819 * Put an ip fragment on a reassembly chain. 820 * Like insque, but pointers in middle of structure. 821 */ 822static void 823ip_enq(p, prev) 824 register struct ipasfrag *p, *prev; 825{ 826 827 p->ipf_prev = prev; 828 p->ipf_next = prev->ipf_next; 829 prev->ipf_next->ipf_prev = p; 830 prev->ipf_next = p; 831} 832 833/* 834 * To ip_enq as remque is to insque. 835 */ 836static void 837ip_deq(p) 838 register struct ipasfrag *p; 839{ 840 841 p->ipf_prev->ipf_next = p->ipf_next; 842 p->ipf_next->ipf_prev = p->ipf_prev; 843} 844 845/* 846 * IP timer processing; 847 * if a timer expires on a reassembly 848 * queue, discard it. 849 */ 850void 851ip_slowtimo() 852{ 853 register struct ipq *fp; 854 int s = splnet(); 855 int i; 856 857 for (i = 0; i < IPREASS_NHASH; i++) { 858 fp = ipq[i].next; 859 if (fp == 0) 860 continue; 861 while (fp != &ipq[i]) { 862 --fp->ipq_ttl; 863 fp = fp->next; 864 if (fp->prev->ipq_ttl == 0) { 865 ipstat.ips_fragtimeout++; 866 ip_freef(fp->prev); 867 } 868 } 869 } 870 splx(s); 871} 872 873/* 874 * Drain off all datagram fragments. 875 */ 876void 877ip_drain() 878{ 879 int i; 880 881 for (i = 0; i < IPREASS_NHASH; i++) { 882 while (ipq[i].next != &ipq[i]) { 883 ipstat.ips_fragdropped++; 884 ip_freef(ipq[i].next); 885 } 886 } 887 in_rtqdrain(); 888} 889 890/* 891 * Do option processing on a datagram, 892 * possibly discarding it if bad options are encountered, 893 * or forwarding it if source-routed. 894 * Returns 1 if packet has been forwarded/freed, 895 * 0 if the packet should be processed further. 896 */ 897static int 898ip_dooptions(m) 899 struct mbuf *m; 900{ 901 register struct ip *ip = mtod(m, struct ip *); 902 register u_char *cp; 903 register struct ip_timestamp *ipt; 904 register struct in_ifaddr *ia; 905 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 906 struct in_addr *sin, dst; 907 n_time ntime; 908 909 dst = ip->ip_dst; 910 cp = (u_char *)(ip + 1); 911 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 912 for (; cnt > 0; cnt -= optlen, cp += optlen) { 913 opt = cp[IPOPT_OPTVAL]; 914 if (opt == IPOPT_EOL) 915 break; 916 if (opt == IPOPT_NOP) 917 optlen = 1; 918 else { 919 optlen = cp[IPOPT_OLEN]; 920 if (optlen <= 0 || optlen > cnt) { 921 code = &cp[IPOPT_OLEN] - (u_char *)ip; 922 goto bad; 923 } 924 } 925 switch (opt) { 926 927 default: 928 break; 929 930 /* 931 * Source routing with record. 932 * Find interface with current destination address. 933 * If none on this machine then drop if strictly routed, 934 * or do nothing if loosely routed. 935 * Record interface address and bring up next address 936 * component. If strictly routed make sure next 937 * address is on directly accessible net. 938 */ 939 case IPOPT_LSRR: 940 case IPOPT_SSRR: 941 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 942 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 943 goto bad; 944 } 945 ipaddr.sin_addr = ip->ip_dst; 946 ia = (struct in_ifaddr *) 947 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 948 if (ia == 0) { 949 if (opt == IPOPT_SSRR) { 950 type = ICMP_UNREACH; 951 code = ICMP_UNREACH_SRCFAIL; 952 goto bad; 953 } 954 if (!ip_dosourceroute) 955 goto nosourcerouting; 956 /* 957 * Loose routing, and not at next destination 958 * yet; nothing to do except forward. 959 */ 960 break; 961 } 962 off--; /* 0 origin */ 963 if (off > optlen - sizeof(struct in_addr)) { 964 /* 965 * End of source route. Should be for us. 966 */ 967 save_rte(cp, ip->ip_src); 968 break; 969 } 970 971 if (!ip_dosourceroute) { 972 char buf[4*sizeof "123"]; 973 974nosourcerouting: 975 strcpy(buf, inet_ntoa(ip->ip_dst)); 976 log(LOG_WARNING, 977 "attempted source route from %s to %s\n", 978 inet_ntoa(ip->ip_src), buf); 979 type = ICMP_UNREACH; 980 code = ICMP_UNREACH_SRCFAIL; 981 goto bad; 982 } 983 984 /* 985 * locate outgoing interface 986 */ 987 (void)memcpy(&ipaddr.sin_addr, cp + off, 988 sizeof(ipaddr.sin_addr)); 989 990 if (opt == IPOPT_SSRR) { 991#define INA struct in_ifaddr * 992#define SA struct sockaddr * 993 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 994 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 995 } else 996 ia = ip_rtaddr(ipaddr.sin_addr); 997 if (ia == 0) { 998 type = ICMP_UNREACH; 999 code = ICMP_UNREACH_SRCFAIL; 1000 goto bad; 1001 } 1002 ip->ip_dst = ipaddr.sin_addr; 1003 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1004 sizeof(struct in_addr)); 1005 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1006 /* 1007 * Let ip_intr's mcast routing check handle mcast pkts 1008 */ 1009 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1010 break; 1011 1012 case IPOPT_RR: 1013 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1014 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1015 goto bad; 1016 } 1017 /* 1018 * If no space remains, ignore. 1019 */ 1020 off--; /* 0 origin */ 1021 if (off > optlen - sizeof(struct in_addr)) 1022 break; 1023 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1024 sizeof(ipaddr.sin_addr)); 1025 /* 1026 * locate outgoing interface; if we're the destination, 1027 * use the incoming interface (should be same). 1028 */ 1029 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1030 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1031 type = ICMP_UNREACH; 1032 code = ICMP_UNREACH_HOST; 1033 goto bad; 1034 } 1035 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1036 sizeof(struct in_addr)); 1037 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1038 break; 1039 1040 case IPOPT_TS: 1041 code = cp - (u_char *)ip; 1042 ipt = (struct ip_timestamp *)cp; 1043 if (ipt->ipt_len < 5) 1044 goto bad; 1045 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) { 1046 if (++ipt->ipt_oflw == 0) 1047 goto bad; 1048 break; 1049 } 1050 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1051 switch (ipt->ipt_flg) { 1052 1053 case IPOPT_TS_TSONLY: 1054 break; 1055 1056 case IPOPT_TS_TSANDADDR: 1057 if (ipt->ipt_ptr + sizeof(n_time) + 1058 sizeof(struct in_addr) > ipt->ipt_len) 1059 goto bad; 1060 ipaddr.sin_addr = dst; 1061 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1062 m->m_pkthdr.rcvif); 1063 if (ia == 0) 1064 continue; 1065 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1066 sizeof(struct in_addr)); 1067 ipt->ipt_ptr += sizeof(struct in_addr); 1068 break; 1069 1070 case IPOPT_TS_PRESPEC: 1071 if (ipt->ipt_ptr + sizeof(n_time) + 1072 sizeof(struct in_addr) > ipt->ipt_len) 1073 goto bad; 1074 (void)memcpy(&ipaddr.sin_addr, sin, 1075 sizeof(struct in_addr)); 1076 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1077 continue; 1078 ipt->ipt_ptr += sizeof(struct in_addr); 1079 break; 1080 1081 default: 1082 goto bad; 1083 } 1084 ntime = iptime(); 1085 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1086 sizeof(n_time)); 1087 ipt->ipt_ptr += sizeof(n_time); 1088 } 1089 } 1090 if (forward) { 1091 ip_forward(m, 1); 1092 return (1); 1093 } 1094 return (0); 1095bad: 1096 ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2; /* XXX icmp_error adds in hdr length */ 1097 icmp_error(m, type, code, 0, 0); 1098 ipstat.ips_badoptions++; 1099 return (1); 1100} 1101 1102/* 1103 * Given address of next destination (final or next hop), 1104 * return internet address info of interface to be used to get there. 1105 */ 1106static struct in_ifaddr * 1107ip_rtaddr(dst) 1108 struct in_addr dst; 1109{ 1110 register struct sockaddr_in *sin; 1111 1112 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1113 1114 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1115 if (ipforward_rt.ro_rt) { 1116 RTFREE(ipforward_rt.ro_rt); 1117 ipforward_rt.ro_rt = 0; 1118 } 1119 sin->sin_family = AF_INET; 1120 sin->sin_len = sizeof(*sin); 1121 sin->sin_addr = dst; 1122 1123 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1124 } 1125 if (ipforward_rt.ro_rt == 0) 1126 return ((struct in_ifaddr *)0); 1127 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1128} 1129 1130/* 1131 * Save incoming source route for use in replies, 1132 * to be picked up later by ip_srcroute if the receiver is interested. 1133 */ 1134void 1135save_rte(option, dst) 1136 u_char *option; 1137 struct in_addr dst; 1138{ 1139 unsigned olen; 1140 1141 olen = option[IPOPT_OLEN]; 1142#ifdef DIAGNOSTIC 1143 if (ipprintfs) 1144 printf("save_rte: olen %d\n", olen); 1145#endif 1146 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1147 return; 1148 bcopy(option, ip_srcrt.srcopt, olen); 1149 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1150 ip_srcrt.dst = dst; 1151} 1152 1153/* 1154 * Retrieve incoming source route for use in replies, 1155 * in the same form used by setsockopt. 1156 * The first hop is placed before the options, will be removed later. 1157 */ 1158struct mbuf * 1159ip_srcroute() 1160{ 1161 register struct in_addr *p, *q; 1162 register struct mbuf *m; 1163 1164 if (ip_nhops == 0) 1165 return ((struct mbuf *)0); 1166 m = m_get(M_DONTWAIT, MT_SOOPTS); 1167 if (m == 0) 1168 return ((struct mbuf *)0); 1169 1170#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1171 1172 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1173 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1174 OPTSIZ; 1175#ifdef DIAGNOSTIC 1176 if (ipprintfs) 1177 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1178#endif 1179 1180 /* 1181 * First save first hop for return route 1182 */ 1183 p = &ip_srcrt.route[ip_nhops - 1]; 1184 *(mtod(m, struct in_addr *)) = *p--; 1185#ifdef DIAGNOSTIC 1186 if (ipprintfs) 1187 printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr)); 1188#endif 1189 1190 /* 1191 * Copy option fields and padding (nop) to mbuf. 1192 */ 1193 ip_srcrt.nop = IPOPT_NOP; 1194 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1195 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1196 &ip_srcrt.nop, OPTSIZ); 1197 q = (struct in_addr *)(mtod(m, caddr_t) + 1198 sizeof(struct in_addr) + OPTSIZ); 1199#undef OPTSIZ 1200 /* 1201 * Record return path as an IP source route, 1202 * reversing the path (pointers are now aligned). 1203 */ 1204 while (p >= ip_srcrt.route) { 1205#ifdef DIAGNOSTIC 1206 if (ipprintfs) 1207 printf(" %lx", ntohl(q->s_addr)); 1208#endif 1209 *q++ = *p--; 1210 } 1211 /* 1212 * Last hop goes to final destination. 1213 */ 1214 *q = ip_srcrt.dst; 1215#ifdef DIAGNOSTIC 1216 if (ipprintfs) 1217 printf(" %lx\n", ntohl(q->s_addr)); 1218#endif 1219 return (m); 1220} 1221 1222/* 1223 * Strip out IP options, at higher 1224 * level protocol in the kernel. 1225 * Second argument is buffer to which options 1226 * will be moved, and return value is their length. 1227 * XXX should be deleted; last arg currently ignored. 1228 */ 1229void 1230ip_stripoptions(m, mopt) 1231 register struct mbuf *m; 1232 struct mbuf *mopt; 1233{ 1234 register int i; 1235 struct ip *ip = mtod(m, struct ip *); 1236 register caddr_t opts; 1237 int olen; 1238 1239 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1240 opts = (caddr_t)(ip + 1); 1241 i = m->m_len - (sizeof (struct ip) + olen); 1242 bcopy(opts + olen, opts, (unsigned)i); 1243 m->m_len -= olen; 1244 if (m->m_flags & M_PKTHDR) 1245 m->m_pkthdr.len -= olen; 1246 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1247} 1248 1249u_char inetctlerrmap[PRC_NCMDS] = { 1250 0, 0, 0, 0, 1251 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1252 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1253 EMSGSIZE, EHOSTUNREACH, 0, 0, 1254 0, 0, 0, 0, 1255 ENOPROTOOPT 1256}; 1257 1258/* 1259 * Forward a packet. If some error occurs return the sender 1260 * an icmp packet. Note we can't always generate a meaningful 1261 * icmp message because icmp doesn't have a large enough repertoire 1262 * of codes and types. 1263 * 1264 * If not forwarding, just drop the packet. This could be confusing 1265 * if ipforwarding was zero but some routing protocol was advancing 1266 * us as a gateway to somewhere. However, we must let the routing 1267 * protocol deal with that. 1268 * 1269 * The srcrt parameter indicates whether the packet is being forwarded 1270 * via a source route. 1271 */ 1272static void 1273ip_forward(m, srcrt) 1274 struct mbuf *m; 1275 int srcrt; 1276{ 1277 register struct ip *ip = mtod(m, struct ip *); 1278 register struct sockaddr_in *sin; 1279 register struct rtentry *rt; 1280 int error, type = 0, code = 0; 1281 struct mbuf *mcopy; 1282 n_long dest; 1283 struct ifnet *destifp; 1284 1285 dest = 0; 1286#ifdef DIAGNOSTIC 1287 if (ipprintfs) 1288 printf("forward: src %lx dst %lx ttl %x\n", 1289 ip->ip_src.s_addr, ip->ip_dst.s_addr, ip->ip_ttl); 1290#endif 1291 1292 1293 if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) { 1294 ipstat.ips_cantforward++; 1295 m_freem(m); 1296 return; 1297 } 1298 HTONS(ip->ip_id); 1299 if (ip->ip_ttl <= IPTTLDEC) { 1300 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1301 return; 1302 } 1303 ip->ip_ttl -= IPTTLDEC; 1304 1305 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1306 if ((rt = ipforward_rt.ro_rt) == 0 || 1307 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1308 if (ipforward_rt.ro_rt) { 1309 RTFREE(ipforward_rt.ro_rt); 1310 ipforward_rt.ro_rt = 0; 1311 } 1312 sin->sin_family = AF_INET; 1313 sin->sin_len = sizeof(*sin); 1314 sin->sin_addr = ip->ip_dst; 1315 1316 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1317 if (ipforward_rt.ro_rt == 0) { 1318 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1319 return; 1320 } 1321 rt = ipforward_rt.ro_rt; 1322 } 1323 1324 /* 1325 * Save at most 64 bytes of the packet in case 1326 * we need to generate an ICMP message to the src. 1327 */ 1328 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1329 1330 /* 1331 * If forwarding packet using same interface that it came in on, 1332 * perhaps should send a redirect to sender to shortcut a hop. 1333 * Only send redirect if source is sending directly to us, 1334 * and if packet was not source routed (or has any options). 1335 * Also, don't send redirect if forwarding using a default route 1336 * or a route modified by a redirect. 1337 */ 1338#define satosin(sa) ((struct sockaddr_in *)(sa)) 1339 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1340 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1341 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1342 ipsendredirects && !srcrt) { 1343#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1344 u_long src = ntohl(ip->ip_src.s_addr); 1345 1346 if (RTA(rt) && 1347 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1348 if (rt->rt_flags & RTF_GATEWAY) 1349 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1350 else 1351 dest = ip->ip_dst.s_addr; 1352 /* Router requirements says to only send host redirects */ 1353 type = ICMP_REDIRECT; 1354 code = ICMP_REDIRECT_HOST; 1355#ifdef DIAGNOSTIC 1356 if (ipprintfs) 1357 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1358#endif 1359 } 1360 } 1361 1362 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1363 IP_FORWARDING, 0); 1364 if (error) 1365 ipstat.ips_cantforward++; 1366 else { 1367 ipstat.ips_forward++; 1368 if (type) 1369 ipstat.ips_redirectsent++; 1370 else { 1371 if (mcopy) 1372 m_freem(mcopy); 1373 return; 1374 } 1375 } 1376 if (mcopy == NULL) 1377 return; 1378 destifp = NULL; 1379 1380 switch (error) { 1381 1382 case 0: /* forwarded, but need redirect */ 1383 /* type, code set above */ 1384 break; 1385 1386 case ENETUNREACH: /* shouldn't happen, checked above */ 1387 case EHOSTUNREACH: 1388 case ENETDOWN: 1389 case EHOSTDOWN: 1390 default: 1391 type = ICMP_UNREACH; 1392 code = ICMP_UNREACH_HOST; 1393 break; 1394 1395 case EMSGSIZE: 1396 type = ICMP_UNREACH; 1397 code = ICMP_UNREACH_NEEDFRAG; 1398 if (ipforward_rt.ro_rt) 1399 destifp = ipforward_rt.ro_rt->rt_ifp; 1400 ipstat.ips_cantfrag++; 1401 break; 1402 1403 case ENOBUFS: 1404 type = ICMP_SOURCEQUENCH; 1405 code = 0; 1406 break; 1407 } 1408 icmp_error(mcopy, type, code, dest, destifp); 1409} 1410 1411void 1412ip_savecontrol(inp, mp, ip, m) 1413 register struct inpcb *inp; 1414 register struct mbuf **mp; 1415 register struct ip *ip; 1416 register struct mbuf *m; 1417{ 1418 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1419 struct timeval tv; 1420 1421 microtime(&tv); 1422 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1423 SCM_TIMESTAMP, SOL_SOCKET); 1424 if (*mp) 1425 mp = &(*mp)->m_next; 1426 } 1427 if (inp->inp_flags & INP_RECVDSTADDR) { 1428 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1429 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1430 if (*mp) 1431 mp = &(*mp)->m_next; 1432 } 1433#ifdef notyet 1434 /* XXX 1435 * Moving these out of udp_input() made them even more broken 1436 * than they already were. 1437 */ 1438 /* options were tossed already */ 1439 if (inp->inp_flags & INP_RECVOPTS) { 1440 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1441 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1442 if (*mp) 1443 mp = &(*mp)->m_next; 1444 } 1445 /* ip_srcroute doesn't do what we want here, need to fix */ 1446 if (inp->inp_flags & INP_RECVRETOPTS) { 1447 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1448 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1449 if (*mp) 1450 mp = &(*mp)->m_next; 1451 } 1452#endif 1453 if (inp->inp_flags & INP_RECVIF) { 1454 struct ifnet *ifp; 1455 struct sdlbuf { 1456 struct sockaddr_dl sdl; 1457 u_char pad[32]; 1458 } sdlbuf; 1459 struct sockaddr_dl *sdp; 1460 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1461 1462 if (((ifp = m->m_pkthdr.rcvif)) 1463 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1464 sdp = (struct sockaddr_dl *)(ifnet_addrs 1465 [ifp->if_index - 1]->ifa_addr); 1466 /* 1467 * Change our mind and don't try copy. 1468 */ 1469 if ((sdp->sdl_family != AF_LINK) 1470 || (sdp->sdl_len > sizeof(sdlbuf))) { 1471 goto makedummy; 1472 } 1473 bcopy(sdp, sdl2, sdp->sdl_len); 1474 } else { 1475makedummy: 1476 sdl2->sdl_len 1477 = offsetof(struct sockaddr_dl, sdl_data[0]); 1478 sdl2->sdl_family = AF_LINK; 1479 sdl2->sdl_index = 0; 1480 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1481 } 1482 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1483 IP_RECVIF, IPPROTO_IP); 1484 if (*mp) 1485 mp = &(*mp)->m_next; 1486 } 1487} 1488 1489int 1490ip_rsvp_init(struct socket *so) 1491{ 1492 if (so->so_type != SOCK_RAW || 1493 so->so_proto->pr_protocol != IPPROTO_RSVP) 1494 return EOPNOTSUPP; 1495 1496 if (ip_rsvpd != NULL) 1497 return EADDRINUSE; 1498 1499 ip_rsvpd = so; 1500 /* 1501 * This may seem silly, but we need to be sure we don't over-increment 1502 * the RSVP counter, in case something slips up. 1503 */ 1504 if (!ip_rsvp_on) { 1505 ip_rsvp_on = 1; 1506 rsvp_on++; 1507 } 1508 1509 return 0; 1510} 1511 1512int 1513ip_rsvp_done(void) 1514{ 1515 ip_rsvpd = NULL; 1516 /* 1517 * This may seem silly, but we need to be sure we don't over-decrement 1518 * the RSVP counter, in case something slips up. 1519 */ 1520 if (ip_rsvp_on) { 1521 ip_rsvp_on = 0; 1522 rsvp_on--; 1523 } 1524 return 0; 1525} 1526