ip_input.c revision 54175
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 * $FreeBSD: head/sys/netinet/ip_input.c 54175 1999-12-06 00:43:07Z archie $ 35 */ 36 37#define _IP_VHL 38 39#include "opt_bootp.h" 40#include "opt_ipfw.h" 41#include "opt_ipdn.h" 42#include "opt_ipdivert.h" 43#include "opt_ipfilter.h" 44#include "opt_ipstealth.h" 45 46#include <stddef.h> 47 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/mbuf.h> 51#include <sys/malloc.h> 52#include <sys/domain.h> 53#include <sys/protosw.h> 54#include <sys/socket.h> 55#include <sys/time.h> 56#include <sys/kernel.h> 57#include <sys/syslog.h> 58#include <sys/sysctl.h> 59 60#include <net/if.h> 61#include <net/if_var.h> 62#include <net/if_dl.h> 63#include <net/route.h> 64#include <net/netisr.h> 65 66#include <netinet/in.h> 67#include <netinet/in_systm.h> 68#include <netinet/in_var.h> 69#include <netinet/ip.h> 70#include <netinet/in_pcb.h> 71#include <netinet/ip_var.h> 72#include <netinet/ip_icmp.h> 73#include <machine/in_cksum.h> 74 75#include <sys/socketvar.h> 76 77#include <netinet/ip_fw.h> 78 79#ifdef DUMMYNET 80#include <netinet/ip_dummynet.h> 81#endif 82 83int rsvp_on = 0; 84static int ip_rsvp_on; 85struct socket *ip_rsvpd; 86 87int ipforwarding = 0; 88SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 89 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 90 91static int ipsendredirects = 1; /* XXX */ 92SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 93 &ipsendredirects, 0, "Enable sending IP redirects"); 94 95int ip_defttl = IPDEFTTL; 96SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 97 &ip_defttl, 0, "Maximum TTL on IP packets"); 98 99static int ip_dosourceroute = 0; 100SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 101 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 102 103static int ip_acceptsourceroute = 0; 104SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 105 CTLFLAG_RW, &ip_acceptsourceroute, 0, 106 "Enable accepting source routed IP packets"); 107#ifdef DIAGNOSTIC 108static int ipprintfs = 0; 109#endif 110 111extern struct domain inetdomain; 112extern struct protosw inetsw[]; 113u_char ip_protox[IPPROTO_MAX]; 114static int ipqmaxlen = IFQ_MAXLEN; 115struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 116struct ifqueue ipintrq; 117SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 118 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 119SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 120 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 121 122struct ipstat ipstat; 123SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD, 124 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 125 126/* Packet reassembly stuff */ 127#define IPREASS_NHASH_LOG2 6 128#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 129#define IPREASS_HMASK (IPREASS_NHASH - 1) 130#define IPREASS_HASH(x,y) \ 131 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 132 133static struct ipq ipq[IPREASS_NHASH]; 134static int nipq = 0; /* total # of reass queues */ 135static int maxnipq; 136 137#ifdef IPCTL_DEFMTU 138SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 139 &ip_mtu, 0, "Default MTU"); 140#endif 141 142#ifdef IPSTEALTH 143static int ipstealth = 0; 144SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 145 &ipstealth, 0, ""); 146#endif 147 148 149/* Firewall hooks */ 150ip_fw_chk_t *ip_fw_chk_ptr; 151ip_fw_ctl_t *ip_fw_ctl_ptr; 152 153#ifdef DUMMYNET 154ip_dn_ctl_t *ip_dn_ctl_ptr; 155#endif 156 157#if defined(IPFILTER_LKM) || defined(IPFILTER) 158int iplattach __P((void)); 159int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL; 160#endif 161 162 163/* 164 * We need to save the IP options in case a protocol wants to respond 165 * to an incoming packet over the same route if the packet got here 166 * using IP source routing. This allows connection establishment and 167 * maintenance when the remote end is on a network that is not known 168 * to us. 169 */ 170static int ip_nhops = 0; 171static struct ip_srcrt { 172 struct in_addr dst; /* final destination */ 173 char nop; /* one NOP to align */ 174 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 175 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 176} ip_srcrt; 177 178struct sockaddr_in *ip_fw_fwd_addr; 179 180static void save_rte __P((u_char *, struct in_addr)); 181static int ip_dooptions __P((struct mbuf *)); 182static void ip_forward __P((struct mbuf *, int)); 183static void ip_freef __P((struct ipq *)); 184#ifdef IPDIVERT 185static struct ip *ip_reass __P((struct mbuf *, 186 struct ipq *, struct ipq *, u_int32_t *, u_int16_t *)); 187#else 188static struct ip *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *)); 189#endif 190static struct in_ifaddr *ip_rtaddr __P((struct in_addr)); 191static void ipintr __P((void)); 192 193/* 194 * IP initialization: fill in IP protocol switch table. 195 * All protocols not implemented in kernel go to raw IP protocol handler. 196 */ 197void 198ip_init() 199{ 200 register struct protosw *pr; 201 register int i; 202 203 TAILQ_INIT(&in_ifaddrhead); 204 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 205 if (pr == 0) 206 panic("ip_init"); 207 for (i = 0; i < IPPROTO_MAX; i++) 208 ip_protox[i] = pr - inetsw; 209 for (pr = inetdomain.dom_protosw; 210 pr < inetdomain.dom_protoswNPROTOSW; pr++) 211 if (pr->pr_domain->dom_family == PF_INET && 212 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 213 ip_protox[pr->pr_protocol] = pr - inetsw; 214 215 for (i = 0; i < IPREASS_NHASH; i++) 216 ipq[i].next = ipq[i].prev = &ipq[i]; 217 218 maxnipq = nmbclusters/4; 219 220 ip_id = time_second & 0xffff; 221 ipintrq.ifq_maxlen = ipqmaxlen; 222#ifdef IPFILTER 223 iplattach(); 224#endif 225 226} 227 228static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 229static struct route ipforward_rt; 230 231/* 232 * Ip input routine. Checksum and byte swap header. If fragmented 233 * try to reassemble. Process options. Pass to next level. 234 */ 235void 236ip_input(struct mbuf *m) 237{ 238 struct ip *ip; 239 struct ipq *fp; 240 struct in_ifaddr *ia; 241 int i, hlen, mff; 242 u_short sum; 243 u_int16_t divert_cookie; /* firewall cookie */ 244#ifdef IPDIVERT 245 u_int32_t divert_info = 0; /* packet divert/tee info */ 246#endif 247 struct ip_fw_chain *rule = NULL; 248 249#ifdef IPDIVERT 250 /* Get and reset firewall cookie */ 251 divert_cookie = ip_divert_cookie; 252 ip_divert_cookie = 0; 253#else 254 divert_cookie = 0; 255#endif 256 257#if defined(IPFIREWALL) && defined(DUMMYNET) 258 /* 259 * dummynet packet are prepended a vestigial mbuf with 260 * m_type = MT_DUMMYNET and m_data pointing to the matching 261 * rule. 262 */ 263 if (m->m_type == MT_DUMMYNET) { 264 rule = (struct ip_fw_chain *)(m->m_data) ; 265 m = m->m_next ; 266 ip = mtod(m, struct ip *); 267 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 268 goto iphack ; 269 } else 270 rule = NULL ; 271#endif 272 273#ifdef DIAGNOSTIC 274 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 275 panic("ip_input no HDR"); 276#endif 277 ipstat.ips_total++; 278 279 if (m->m_pkthdr.len < sizeof(struct ip)) 280 goto tooshort; 281 282 if (m->m_len < sizeof (struct ip) && 283 (m = m_pullup(m, sizeof (struct ip))) == 0) { 284 ipstat.ips_toosmall++; 285 return; 286 } 287 ip = mtod(m, struct ip *); 288 289 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 290 ipstat.ips_badvers++; 291 goto bad; 292 } 293 294 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 295 if (hlen < sizeof(struct ip)) { /* minimum header length */ 296 ipstat.ips_badhlen++; 297 goto bad; 298 } 299 if (hlen > m->m_len) { 300 if ((m = m_pullup(m, hlen)) == 0) { 301 ipstat.ips_badhlen++; 302 return; 303 } 304 ip = mtod(m, struct ip *); 305 } 306 if (hlen == sizeof(struct ip)) { 307 sum = in_cksum_hdr(ip); 308 } else { 309 sum = in_cksum(m, hlen); 310 } 311 if (sum) { 312 ipstat.ips_badsum++; 313 goto bad; 314 } 315 316 /* 317 * Convert fields to host representation. 318 */ 319 NTOHS(ip->ip_len); 320 if (ip->ip_len < hlen) { 321 ipstat.ips_badlen++; 322 goto bad; 323 } 324 NTOHS(ip->ip_id); 325 NTOHS(ip->ip_off); 326 327 /* 328 * Check that the amount of data in the buffers 329 * is as at least much as the IP header would have us expect. 330 * Trim mbufs if longer than we expect. 331 * Drop packet if shorter than we expect. 332 */ 333 if (m->m_pkthdr.len < ip->ip_len) { 334tooshort: 335 ipstat.ips_tooshort++; 336 goto bad; 337 } 338 if (m->m_pkthdr.len > ip->ip_len) { 339 if (m->m_len == m->m_pkthdr.len) { 340 m->m_len = ip->ip_len; 341 m->m_pkthdr.len = ip->ip_len; 342 } else 343 m_adj(m, ip->ip_len - m->m_pkthdr.len); 344 } 345 /* 346 * IpHack's section. 347 * Right now when no processing on packet has done 348 * and it is still fresh out of network we do our black 349 * deals with it. 350 * - Firewall: deny/allow/divert 351 * - Xlate: translate packet's addr/port (NAT). 352 * - Pipe: pass pkt through dummynet. 353 * - Wrap: fake packet's addr/port <unimpl.> 354 * - Encapsulate: put it in another IP and send out. <unimp.> 355 */ 356 357#if defined(IPFIREWALL) && defined(DUMMYNET) 358iphack: 359#endif 360#if defined(IPFILTER) || defined(IPFILTER_LKM) 361 /* 362 * Check if we want to allow this packet to be processed. 363 * Consider it to be bad if not. 364 */ 365 if (fr_checkp) { 366 struct mbuf *m1 = m; 367 368 if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1) 369 return; 370 ip = mtod(m = m1, struct ip *); 371 } 372#endif 373 if (ip_fw_chk_ptr) { 374#ifdef IPFIREWALL_FORWARD 375 /* 376 * If we've been forwarded from the output side, then 377 * skip the firewall a second time 378 */ 379 if (ip_fw_fwd_addr) 380 goto ours; 381#endif /* IPFIREWALL_FORWARD */ 382 /* 383 * See the comment in ip_output for the return values 384 * produced by the firewall. 385 */ 386 i = (*ip_fw_chk_ptr)(&ip, 387 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 388 if (m == NULL) /* Packet discarded by firewall */ 389 return; 390 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 391 goto pass; 392#ifdef DUMMYNET 393 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) { 394 /* Send packet to the appropriate pipe */ 395 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule); 396 return; 397 } 398#endif 399#ifdef IPDIVERT 400 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 401 /* Divert or tee packet */ 402 divert_info = i; 403 goto ours; 404 } 405#endif 406#ifdef IPFIREWALL_FORWARD 407 if (i == 0 && ip_fw_fwd_addr != NULL) 408 goto pass; 409#endif 410 /* 411 * if we get here, the packet must be dropped 412 */ 413 m_freem(m); 414 return; 415 } 416pass: 417 418 /* 419 * Process options and, if not destined for us, 420 * ship it on. ip_dooptions returns 1 when an 421 * error was detected (causing an icmp message 422 * to be sent and the original packet to be freed). 423 */ 424 ip_nhops = 0; /* for source routed packets */ 425 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 426#ifdef IPFIREWALL_FORWARD 427 ip_fw_fwd_addr = NULL; 428#endif 429 return; 430 } 431 432 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 433 * matter if it is destined to another node, or whether it is 434 * a multicast one, RSVP wants it! and prevents it from being forwarded 435 * anywhere else. Also checks if the rsvp daemon is running before 436 * grabbing the packet. 437 */ 438 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 439 goto ours; 440 441 /* 442 * Check our list of addresses, to see if the packet is for us. 443 * If we don't have any addresses, assume any unicast packet 444 * we receive might be for us (and let the upper layers deal 445 * with it). 446 */ 447 if (TAILQ_EMPTY(&in_ifaddrhead) && 448 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 449 goto ours; 450 451 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 452 ia = TAILQ_NEXT(ia, ia_link)) { 453#define satosin(sa) ((struct sockaddr_in *)(sa)) 454 455#ifdef BOOTP_COMPAT 456 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 457 goto ours; 458#endif 459#ifdef IPFIREWALL_FORWARD 460 /* 461 * If the addr to forward to is one of ours, we pretend to 462 * be the destination for this packet. 463 */ 464 if (ip_fw_fwd_addr == NULL) { 465 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 466 goto ours; 467 } else if (IA_SIN(ia)->sin_addr.s_addr == 468 ip_fw_fwd_addr->sin_addr.s_addr) 469 goto ours; 470#else 471 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 472 goto ours; 473#endif 474 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 475 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 476 ip->ip_dst.s_addr) 477 goto ours; 478 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 479 goto ours; 480 } 481 } 482 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 483 struct in_multi *inm; 484 if (ip_mrouter) { 485 /* 486 * If we are acting as a multicast router, all 487 * incoming multicast packets are passed to the 488 * kernel-level multicast forwarding function. 489 * The packet is returned (relatively) intact; if 490 * ip_mforward() returns a non-zero value, the packet 491 * must be discarded, else it may be accepted below. 492 * 493 * (The IP ident field is put in the same byte order 494 * as expected when ip_mforward() is called from 495 * ip_output().) 496 */ 497 ip->ip_id = htons(ip->ip_id); 498 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 499 ipstat.ips_cantforward++; 500 m_freem(m); 501 return; 502 } 503 ip->ip_id = ntohs(ip->ip_id); 504 505 /* 506 * The process-level routing demon needs to receive 507 * all multicast IGMP packets, whether or not this 508 * host belongs to their destination groups. 509 */ 510 if (ip->ip_p == IPPROTO_IGMP) 511 goto ours; 512 ipstat.ips_forward++; 513 } 514 /* 515 * See if we belong to the destination multicast group on the 516 * arrival interface. 517 */ 518 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 519 if (inm == NULL) { 520 ipstat.ips_notmember++; 521 m_freem(m); 522 return; 523 } 524 goto ours; 525 } 526 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 527 goto ours; 528 if (ip->ip_dst.s_addr == INADDR_ANY) 529 goto ours; 530 531 /* 532 * Not for us; forward if possible and desirable. 533 */ 534 if (ipforwarding == 0) { 535 ipstat.ips_cantforward++; 536 m_freem(m); 537 } else 538 ip_forward(m, 0); 539#ifdef IPFIREWALL_FORWARD 540 ip_fw_fwd_addr = NULL; 541#endif 542 return; 543 544ours: 545 546 /* 547 * If offset or IP_MF are set, must reassemble. 548 * Otherwise, nothing need be done. 549 * (We could look in the reassembly queue to see 550 * if the packet was previously fragmented, 551 * but it's not worth the time; just let them time out.) 552 */ 553 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) { 554 if (m->m_flags & M_EXT) { /* XXX */ 555 if ((m = m_pullup(m, hlen)) == 0) { 556 ipstat.ips_toosmall++; 557#ifdef IPFIREWALL_FORWARD 558 ip_fw_fwd_addr = NULL; 559#endif 560 return; 561 } 562 ip = mtod(m, struct ip *); 563 } 564 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 565 /* 566 * Look for queue of fragments 567 * of this datagram. 568 */ 569 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 570 if (ip->ip_id == fp->ipq_id && 571 ip->ip_src.s_addr == fp->ipq_src.s_addr && 572 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 573 ip->ip_p == fp->ipq_p) 574 goto found; 575 576 fp = 0; 577 578 /* check if there's a place for the new queue */ 579 if (nipq > maxnipq) { 580 /* 581 * drop something from the tail of the current queue 582 * before proceeding further 583 */ 584 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 585 for (i = 0; i < IPREASS_NHASH; i++) { 586 if (ipq[i].prev != &ipq[i]) { 587 ip_freef(ipq[i].prev); 588 break; 589 } 590 } 591 } else 592 ip_freef(ipq[sum].prev); 593 } 594found: 595 /* 596 * Adjust ip_len to not reflect header, 597 * set ip_mff if more fragments are expected, 598 * convert offset of this to bytes. 599 */ 600 ip->ip_len -= hlen; 601 mff = (ip->ip_off & IP_MF) != 0; 602 if (mff) { 603 /* 604 * Make sure that fragments have a data length 605 * that's a non-zero multiple of 8 bytes. 606 */ 607 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 608 ipstat.ips_toosmall++; /* XXX */ 609 goto bad; 610 } 611 m->m_flags |= M_FRAG; 612 } 613 ip->ip_off <<= 3; 614 615 /* 616 * If datagram marked as having more fragments 617 * or if this is not the first fragment, 618 * attempt reassembly; if it succeeds, proceed. 619 */ 620 if (mff || ip->ip_off) { 621 ipstat.ips_fragments++; 622 m->m_pkthdr.header = ip; 623#ifdef IPDIVERT 624 ip = ip_reass(m, 625 fp, &ipq[sum], &divert_info, &divert_cookie); 626#else 627 ip = ip_reass(m, fp, &ipq[sum]); 628#endif 629 if (ip == 0) { 630#ifdef IPFIREWALL_FORWARD 631 ip_fw_fwd_addr = NULL; 632#endif 633 return; 634 } 635 /* Get the length of the reassembled packets header */ 636 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 637 ipstat.ips_reassembled++; 638 m = dtom(ip); 639#ifdef IPDIVERT 640 /* Restore original checksum before diverting packet */ 641 if (divert_info != 0) { 642 ip->ip_len += hlen; 643 HTONS(ip->ip_len); 644 HTONS(ip->ip_off); 645 HTONS(ip->ip_id); 646 ip->ip_sum = 0; 647 ip->ip_sum = in_cksum_hdr(ip); 648 NTOHS(ip->ip_id); 649 NTOHS(ip->ip_off); 650 NTOHS(ip->ip_len); 651 ip->ip_len -= hlen; 652 } 653#endif 654 } else 655 if (fp) 656 ip_freef(fp); 657 } else 658 ip->ip_len -= hlen; 659 660#ifdef IPDIVERT 661 /* 662 * Divert or tee packet to the divert protocol if required. 663 * 664 * If divert_info is zero then cookie should be too, so we shouldn't 665 * need to clear them here. Assume divert_packet() does so also. 666 */ 667 if (divert_info != 0) { 668 struct mbuf *clone = NULL; 669 670 /* Clone packet if we're doing a 'tee' */ 671 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 672 clone = m_dup(m, M_DONTWAIT); 673 674 /* Restore packet header fields to original values */ 675 ip->ip_len += hlen; 676 HTONS(ip->ip_len); 677 HTONS(ip->ip_off); 678 HTONS(ip->ip_id); 679 680 /* Deliver packet to divert input routine */ 681 ip_divert_cookie = divert_cookie; 682 divert_packet(m, 1, divert_info & 0xffff); 683 ipstat.ips_delivered++; 684 685 /* If 'tee', continue with original packet */ 686 if (clone == NULL) 687 return; 688 m = clone; 689 ip = mtod(m, struct ip *); 690 } 691#endif 692 693 /* 694 * Switch out to protocol's input routine. 695 */ 696 ipstat.ips_delivered++; 697 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen); 698#ifdef IPFIREWALL_FORWARD 699 ip_fw_fwd_addr = NULL; /* tcp needed it */ 700#endif 701 return; 702bad: 703#ifdef IPFIREWALL_FORWARD 704 ip_fw_fwd_addr = NULL; 705#endif 706 m_freem(m); 707} 708 709/* 710 * IP software interrupt routine - to go away sometime soon 711 */ 712static void 713ipintr(void) 714{ 715 int s; 716 struct mbuf *m; 717 718 while(1) { 719 s = splimp(); 720 IF_DEQUEUE(&ipintrq, m); 721 splx(s); 722 if (m == 0) 723 return; 724 ip_input(m); 725 } 726} 727 728NETISR_SET(NETISR_IP, ipintr); 729 730/* 731 * Take incoming datagram fragment and try to reassemble it into 732 * whole datagram. If a chain for reassembly of this datagram already 733 * exists, then it is given as fp; otherwise have to make a chain. 734 * 735 * When IPDIVERT enabled, keep additional state with each packet that 736 * tells us if we need to divert or tee the packet we're building. 737 */ 738 739static struct ip * 740#ifdef IPDIVERT 741ip_reass(m, fp, where, divinfo, divcookie) 742#else 743ip_reass(m, fp, where) 744#endif 745 register struct mbuf *m; 746 register struct ipq *fp; 747 struct ipq *where; 748#ifdef IPDIVERT 749 u_int32_t *divinfo; 750 u_int16_t *divcookie; 751#endif 752{ 753 struct ip *ip = mtod(m, struct ip *); 754 register struct mbuf *p = 0, *q, *nq; 755 struct mbuf *t; 756 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 757 int i, next; 758 759 /* 760 * Presence of header sizes in mbufs 761 * would confuse code below. 762 */ 763 m->m_data += hlen; 764 m->m_len -= hlen; 765 766 /* 767 * If first fragment to arrive, create a reassembly queue. 768 */ 769 if (fp == 0) { 770 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 771 goto dropfrag; 772 fp = mtod(t, struct ipq *); 773 insque(fp, where); 774 nipq++; 775 fp->ipq_ttl = IPFRAGTTL; 776 fp->ipq_p = ip->ip_p; 777 fp->ipq_id = ip->ip_id; 778 fp->ipq_src = ip->ip_src; 779 fp->ipq_dst = ip->ip_dst; 780 fp->ipq_frags = m; 781 m->m_nextpkt = NULL; 782#ifdef IPDIVERT 783 fp->ipq_div_info = 0; 784 fp->ipq_div_cookie = 0; 785#endif 786 goto inserted; 787 } 788 789#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 790 791 /* 792 * Find a segment which begins after this one does. 793 */ 794 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 795 if (GETIP(q)->ip_off > ip->ip_off) 796 break; 797 798 /* 799 * If there is a preceding segment, it may provide some of 800 * our data already. If so, drop the data from the incoming 801 * segment. If it provides all of our data, drop us, otherwise 802 * stick new segment in the proper place. 803 */ 804 if (p) { 805 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 806 if (i > 0) { 807 if (i >= ip->ip_len) 808 goto dropfrag; 809 m_adj(dtom(ip), i); 810 ip->ip_off += i; 811 ip->ip_len -= i; 812 } 813 m->m_nextpkt = p->m_nextpkt; 814 p->m_nextpkt = m; 815 } else { 816 m->m_nextpkt = fp->ipq_frags; 817 fp->ipq_frags = m; 818 } 819 820 /* 821 * While we overlap succeeding segments trim them or, 822 * if they are completely covered, dequeue them. 823 */ 824 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 825 q = nq) { 826 i = (ip->ip_off + ip->ip_len) - 827 GETIP(q)->ip_off; 828 if (i < GETIP(q)->ip_len) { 829 GETIP(q)->ip_len -= i; 830 GETIP(q)->ip_off += i; 831 m_adj(q, i); 832 break; 833 } 834 nq = q->m_nextpkt; 835 m->m_nextpkt = nq; 836 m_freem(q); 837 } 838 839inserted: 840 841#ifdef IPDIVERT 842 /* 843 * Transfer firewall instructions to the fragment structure. 844 * Any fragment diverting causes the whole packet to divert. 845 */ 846 fp->ipq_div_info = *divinfo; 847 fp->ipq_div_cookie = *divcookie; 848 *divinfo = 0; 849 *divcookie = 0; 850#endif 851 852 /* 853 * Check for complete reassembly. 854 */ 855 next = 0; 856 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 857 if (GETIP(q)->ip_off != next) 858 return (0); 859 next += GETIP(q)->ip_len; 860 } 861 /* Make sure the last packet didn't have the IP_MF flag */ 862 if (p->m_flags & M_FRAG) 863 return (0); 864 865 /* 866 * Reassembly is complete. Make sure the packet is a sane size. 867 */ 868 q = fp->ipq_frags; 869 ip = GETIP(q); 870 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 871 ipstat.ips_toolong++; 872 ip_freef(fp); 873 return (0); 874 } 875 876 /* 877 * Concatenate fragments. 878 */ 879 m = q; 880 t = m->m_next; 881 m->m_next = 0; 882 m_cat(m, t); 883 nq = q->m_nextpkt; 884 q->m_nextpkt = 0; 885 for (q = nq; q != NULL; q = nq) { 886 nq = q->m_nextpkt; 887 q->m_nextpkt = NULL; 888 m_cat(m, q); 889 } 890 891#ifdef IPDIVERT 892 /* 893 * Extract firewall instructions from the fragment structure. 894 */ 895 *divinfo = fp->ipq_div_info; 896 *divcookie = fp->ipq_div_cookie; 897#endif 898 899 /* 900 * Create header for new ip packet by 901 * modifying header of first packet; 902 * dequeue and discard fragment reassembly header. 903 * Make header visible. 904 */ 905 ip->ip_len = next; 906 ip->ip_src = fp->ipq_src; 907 ip->ip_dst = fp->ipq_dst; 908 remque(fp); 909 nipq--; 910 (void) m_free(dtom(fp)); 911 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 912 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 913 /* some debugging cruft by sklower, below, will go away soon */ 914 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 915 register int plen = 0; 916 for (t = m; m; m = m->m_next) 917 plen += m->m_len; 918 t->m_pkthdr.len = plen; 919 } 920 return (ip); 921 922dropfrag: 923#ifdef IPDIVERT 924 *divinfo = 0; 925 *divcookie = 0; 926#endif 927 ipstat.ips_fragdropped++; 928 m_freem(m); 929 return (0); 930 931#undef GETIP 932} 933 934/* 935 * Free a fragment reassembly header and all 936 * associated datagrams. 937 */ 938static void 939ip_freef(fp) 940 struct ipq *fp; 941{ 942 register struct mbuf *q; 943 944 while (fp->ipq_frags) { 945 q = fp->ipq_frags; 946 fp->ipq_frags = q->m_nextpkt; 947 m_freem(q); 948 } 949 remque(fp); 950 (void) m_free(dtom(fp)); 951 nipq--; 952} 953 954/* 955 * IP timer processing; 956 * if a timer expires on a reassembly 957 * queue, discard it. 958 */ 959void 960ip_slowtimo() 961{ 962 register struct ipq *fp; 963 int s = splnet(); 964 int i; 965 966 for (i = 0; i < IPREASS_NHASH; i++) { 967 fp = ipq[i].next; 968 if (fp == 0) 969 continue; 970 while (fp != &ipq[i]) { 971 --fp->ipq_ttl; 972 fp = fp->next; 973 if (fp->prev->ipq_ttl == 0) { 974 ipstat.ips_fragtimeout++; 975 ip_freef(fp->prev); 976 } 977 } 978 } 979 ipflow_slowtimo(); 980 splx(s); 981} 982 983/* 984 * Drain off all datagram fragments. 985 */ 986void 987ip_drain() 988{ 989 int i; 990 991 for (i = 0; i < IPREASS_NHASH; i++) { 992 while (ipq[i].next != &ipq[i]) { 993 ipstat.ips_fragdropped++; 994 ip_freef(ipq[i].next); 995 } 996 } 997 in_rtqdrain(); 998} 999 1000/* 1001 * Do option processing on a datagram, 1002 * possibly discarding it if bad options are encountered, 1003 * or forwarding it if source-routed. 1004 * Returns 1 if packet has been forwarded/freed, 1005 * 0 if the packet should be processed further. 1006 */ 1007static int 1008ip_dooptions(m) 1009 struct mbuf *m; 1010{ 1011 register struct ip *ip = mtod(m, struct ip *); 1012 register u_char *cp; 1013 register struct ip_timestamp *ipt; 1014 register struct in_ifaddr *ia; 1015 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1016 struct in_addr *sin, dst; 1017 n_time ntime; 1018 1019 dst = ip->ip_dst; 1020 cp = (u_char *)(ip + 1); 1021 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1022 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1023 opt = cp[IPOPT_OPTVAL]; 1024 if (opt == IPOPT_EOL) 1025 break; 1026 if (opt == IPOPT_NOP) 1027 optlen = 1; 1028 else { 1029 optlen = cp[IPOPT_OLEN]; 1030 if (optlen <= 0 || optlen > cnt) { 1031 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1032 goto bad; 1033 } 1034 } 1035 switch (opt) { 1036 1037 default: 1038 break; 1039 1040 /* 1041 * Source routing with record. 1042 * Find interface with current destination address. 1043 * If none on this machine then drop if strictly routed, 1044 * or do nothing if loosely routed. 1045 * Record interface address and bring up next address 1046 * component. If strictly routed make sure next 1047 * address is on directly accessible net. 1048 */ 1049 case IPOPT_LSRR: 1050 case IPOPT_SSRR: 1051 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1052 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1053 goto bad; 1054 } 1055 ipaddr.sin_addr = ip->ip_dst; 1056 ia = (struct in_ifaddr *) 1057 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1058 if (ia == 0) { 1059 if (opt == IPOPT_SSRR) { 1060 type = ICMP_UNREACH; 1061 code = ICMP_UNREACH_SRCFAIL; 1062 goto bad; 1063 } 1064 if (!ip_dosourceroute) 1065 goto nosourcerouting; 1066 /* 1067 * Loose routing, and not at next destination 1068 * yet; nothing to do except forward. 1069 */ 1070 break; 1071 } 1072 off--; /* 0 origin */ 1073 if (off > optlen - sizeof(struct in_addr)) { 1074 /* 1075 * End of source route. Should be for us. 1076 */ 1077 if (!ip_acceptsourceroute) 1078 goto nosourcerouting; 1079 save_rte(cp, ip->ip_src); 1080 break; 1081 } 1082 1083 if (!ip_dosourceroute) { 1084 if (ipforwarding) { 1085 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1086 /* 1087 * Acting as a router, so generate ICMP 1088 */ 1089nosourcerouting: 1090 strcpy(buf, inet_ntoa(ip->ip_dst)); 1091 log(LOG_WARNING, 1092 "attempted source route from %s to %s\n", 1093 inet_ntoa(ip->ip_src), buf); 1094 type = ICMP_UNREACH; 1095 code = ICMP_UNREACH_SRCFAIL; 1096 goto bad; 1097 } else { 1098 /* 1099 * Not acting as a router, so silently drop. 1100 */ 1101 ipstat.ips_cantforward++; 1102 m_freem(m); 1103 return (1); 1104 } 1105 } 1106 1107 /* 1108 * locate outgoing interface 1109 */ 1110 (void)memcpy(&ipaddr.sin_addr, cp + off, 1111 sizeof(ipaddr.sin_addr)); 1112 1113 if (opt == IPOPT_SSRR) { 1114#define INA struct in_ifaddr * 1115#define SA struct sockaddr * 1116 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1117 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1118 } else 1119 ia = ip_rtaddr(ipaddr.sin_addr); 1120 if (ia == 0) { 1121 type = ICMP_UNREACH; 1122 code = ICMP_UNREACH_SRCFAIL; 1123 goto bad; 1124 } 1125 ip->ip_dst = ipaddr.sin_addr; 1126 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1127 sizeof(struct in_addr)); 1128 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1129 /* 1130 * Let ip_intr's mcast routing check handle mcast pkts 1131 */ 1132 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1133 break; 1134 1135 case IPOPT_RR: 1136 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1137 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1138 goto bad; 1139 } 1140 /* 1141 * If no space remains, ignore. 1142 */ 1143 off--; /* 0 origin */ 1144 if (off > optlen - sizeof(struct in_addr)) 1145 break; 1146 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1147 sizeof(ipaddr.sin_addr)); 1148 /* 1149 * locate outgoing interface; if we're the destination, 1150 * use the incoming interface (should be same). 1151 */ 1152 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1153 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1154 type = ICMP_UNREACH; 1155 code = ICMP_UNREACH_HOST; 1156 goto bad; 1157 } 1158 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1159 sizeof(struct in_addr)); 1160 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1161 break; 1162 1163 case IPOPT_TS: 1164 code = cp - (u_char *)ip; 1165 ipt = (struct ip_timestamp *)cp; 1166 if (ipt->ipt_len < 5) 1167 goto bad; 1168 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1169 if (++ipt->ipt_oflw == 0) 1170 goto bad; 1171 break; 1172 } 1173 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1174 switch (ipt->ipt_flg) { 1175 1176 case IPOPT_TS_TSONLY: 1177 break; 1178 1179 case IPOPT_TS_TSANDADDR: 1180 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1181 sizeof(struct in_addr) > ipt->ipt_len) 1182 goto bad; 1183 ipaddr.sin_addr = dst; 1184 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1185 m->m_pkthdr.rcvif); 1186 if (ia == 0) 1187 continue; 1188 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1189 sizeof(struct in_addr)); 1190 ipt->ipt_ptr += sizeof(struct in_addr); 1191 break; 1192 1193 case IPOPT_TS_PRESPEC: 1194 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1195 sizeof(struct in_addr) > ipt->ipt_len) 1196 goto bad; 1197 (void)memcpy(&ipaddr.sin_addr, sin, 1198 sizeof(struct in_addr)); 1199 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1200 continue; 1201 ipt->ipt_ptr += sizeof(struct in_addr); 1202 break; 1203 1204 default: 1205 goto bad; 1206 } 1207 ntime = iptime(); 1208 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1209 sizeof(n_time)); 1210 ipt->ipt_ptr += sizeof(n_time); 1211 } 1212 } 1213 if (forward && ipforwarding) { 1214 ip_forward(m, 1); 1215 return (1); 1216 } 1217 return (0); 1218bad: 1219 ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2; /* XXX icmp_error adds in hdr length */ 1220 icmp_error(m, type, code, 0, 0); 1221 ipstat.ips_badoptions++; 1222 return (1); 1223} 1224 1225/* 1226 * Given address of next destination (final or next hop), 1227 * return internet address info of interface to be used to get there. 1228 */ 1229static struct in_ifaddr * 1230ip_rtaddr(dst) 1231 struct in_addr dst; 1232{ 1233 register struct sockaddr_in *sin; 1234 1235 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1236 1237 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1238 if (ipforward_rt.ro_rt) { 1239 RTFREE(ipforward_rt.ro_rt); 1240 ipforward_rt.ro_rt = 0; 1241 } 1242 sin->sin_family = AF_INET; 1243 sin->sin_len = sizeof(*sin); 1244 sin->sin_addr = dst; 1245 1246 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1247 } 1248 if (ipforward_rt.ro_rt == 0) 1249 return ((struct in_ifaddr *)0); 1250 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1251} 1252 1253/* 1254 * Save incoming source route for use in replies, 1255 * to be picked up later by ip_srcroute if the receiver is interested. 1256 */ 1257void 1258save_rte(option, dst) 1259 u_char *option; 1260 struct in_addr dst; 1261{ 1262 unsigned olen; 1263 1264 olen = option[IPOPT_OLEN]; 1265#ifdef DIAGNOSTIC 1266 if (ipprintfs) 1267 printf("save_rte: olen %d\n", olen); 1268#endif 1269 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1270 return; 1271 bcopy(option, ip_srcrt.srcopt, olen); 1272 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1273 ip_srcrt.dst = dst; 1274} 1275 1276/* 1277 * Retrieve incoming source route for use in replies, 1278 * in the same form used by setsockopt. 1279 * The first hop is placed before the options, will be removed later. 1280 */ 1281struct mbuf * 1282ip_srcroute() 1283{ 1284 register struct in_addr *p, *q; 1285 register struct mbuf *m; 1286 1287 if (ip_nhops == 0) 1288 return ((struct mbuf *)0); 1289 m = m_get(M_DONTWAIT, MT_HEADER); 1290 if (m == 0) 1291 return ((struct mbuf *)0); 1292 1293#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1294 1295 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1296 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1297 OPTSIZ; 1298#ifdef DIAGNOSTIC 1299 if (ipprintfs) 1300 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1301#endif 1302 1303 /* 1304 * First save first hop for return route 1305 */ 1306 p = &ip_srcrt.route[ip_nhops - 1]; 1307 *(mtod(m, struct in_addr *)) = *p--; 1308#ifdef DIAGNOSTIC 1309 if (ipprintfs) 1310 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1311#endif 1312 1313 /* 1314 * Copy option fields and padding (nop) to mbuf. 1315 */ 1316 ip_srcrt.nop = IPOPT_NOP; 1317 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1318 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1319 &ip_srcrt.nop, OPTSIZ); 1320 q = (struct in_addr *)(mtod(m, caddr_t) + 1321 sizeof(struct in_addr) + OPTSIZ); 1322#undef OPTSIZ 1323 /* 1324 * Record return path as an IP source route, 1325 * reversing the path (pointers are now aligned). 1326 */ 1327 while (p >= ip_srcrt.route) { 1328#ifdef DIAGNOSTIC 1329 if (ipprintfs) 1330 printf(" %lx", (u_long)ntohl(q->s_addr)); 1331#endif 1332 *q++ = *p--; 1333 } 1334 /* 1335 * Last hop goes to final destination. 1336 */ 1337 *q = ip_srcrt.dst; 1338#ifdef DIAGNOSTIC 1339 if (ipprintfs) 1340 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1341#endif 1342 return (m); 1343} 1344 1345/* 1346 * Strip out IP options, at higher 1347 * level protocol in the kernel. 1348 * Second argument is buffer to which options 1349 * will be moved, and return value is their length. 1350 * XXX should be deleted; last arg currently ignored. 1351 */ 1352void 1353ip_stripoptions(m, mopt) 1354 register struct mbuf *m; 1355 struct mbuf *mopt; 1356{ 1357 register int i; 1358 struct ip *ip = mtod(m, struct ip *); 1359 register caddr_t opts; 1360 int olen; 1361 1362 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1363 opts = (caddr_t)(ip + 1); 1364 i = m->m_len - (sizeof (struct ip) + olen); 1365 bcopy(opts + olen, opts, (unsigned)i); 1366 m->m_len -= olen; 1367 if (m->m_flags & M_PKTHDR) 1368 m->m_pkthdr.len -= olen; 1369 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1370} 1371 1372u_char inetctlerrmap[PRC_NCMDS] = { 1373 0, 0, 0, 0, 1374 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1375 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1376 EMSGSIZE, EHOSTUNREACH, 0, 0, 1377 0, 0, 0, 0, 1378 ENOPROTOOPT 1379}; 1380 1381/* 1382 * Forward a packet. If some error occurs return the sender 1383 * an icmp packet. Note we can't always generate a meaningful 1384 * icmp message because icmp doesn't have a large enough repertoire 1385 * of codes and types. 1386 * 1387 * If not forwarding, just drop the packet. This could be confusing 1388 * if ipforwarding was zero but some routing protocol was advancing 1389 * us as a gateway to somewhere. However, we must let the routing 1390 * protocol deal with that. 1391 * 1392 * The srcrt parameter indicates whether the packet is being forwarded 1393 * via a source route. 1394 */ 1395static void 1396ip_forward(m, srcrt) 1397 struct mbuf *m; 1398 int srcrt; 1399{ 1400 register struct ip *ip = mtod(m, struct ip *); 1401 register struct sockaddr_in *sin; 1402 register struct rtentry *rt; 1403 int error, type = 0, code = 0; 1404 struct mbuf *mcopy; 1405 n_long dest; 1406 struct ifnet *destifp; 1407 1408 dest = 0; 1409#ifdef DIAGNOSTIC 1410 if (ipprintfs) 1411 printf("forward: src %lx dst %lx ttl %x\n", 1412 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1413 ip->ip_ttl); 1414#endif 1415 1416 1417 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1418 ipstat.ips_cantforward++; 1419 m_freem(m); 1420 return; 1421 } 1422 HTONS(ip->ip_id); 1423#ifdef IPSTEALTH 1424 if (!ipstealth) { 1425#endif 1426 if (ip->ip_ttl <= IPTTLDEC) { 1427 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1428 dest, 0); 1429 return; 1430 } 1431 ip->ip_ttl -= IPTTLDEC; 1432#ifdef IPSTEALTH 1433 } 1434#endif 1435 1436 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1437 if ((rt = ipforward_rt.ro_rt) == 0 || 1438 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1439 if (ipforward_rt.ro_rt) { 1440 RTFREE(ipforward_rt.ro_rt); 1441 ipforward_rt.ro_rt = 0; 1442 } 1443 sin->sin_family = AF_INET; 1444 sin->sin_len = sizeof(*sin); 1445 sin->sin_addr = ip->ip_dst; 1446 1447 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1448 if (ipforward_rt.ro_rt == 0) { 1449 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1450 return; 1451 } 1452 rt = ipforward_rt.ro_rt; 1453 } 1454 1455 /* 1456 * Save at most 64 bytes of the packet in case 1457 * we need to generate an ICMP message to the src. 1458 */ 1459 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1460 1461 /* 1462 * If forwarding packet using same interface that it came in on, 1463 * perhaps should send a redirect to sender to shortcut a hop. 1464 * Only send redirect if source is sending directly to us, 1465 * and if packet was not source routed (or has any options). 1466 * Also, don't send redirect if forwarding using a default route 1467 * or a route modified by a redirect. 1468 */ 1469#define satosin(sa) ((struct sockaddr_in *)(sa)) 1470 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1471 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1472 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1473 ipsendredirects && !srcrt) { 1474#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1475 u_long src = ntohl(ip->ip_src.s_addr); 1476 1477 if (RTA(rt) && 1478 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1479 if (rt->rt_flags & RTF_GATEWAY) 1480 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1481 else 1482 dest = ip->ip_dst.s_addr; 1483 /* Router requirements says to only send host redirects */ 1484 type = ICMP_REDIRECT; 1485 code = ICMP_REDIRECT_HOST; 1486#ifdef DIAGNOSTIC 1487 if (ipprintfs) 1488 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1489#endif 1490 } 1491 } 1492 1493 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1494 IP_FORWARDING, 0); 1495 if (error) 1496 ipstat.ips_cantforward++; 1497 else { 1498 ipstat.ips_forward++; 1499 if (type) 1500 ipstat.ips_redirectsent++; 1501 else { 1502 if (mcopy) { 1503 ipflow_create(&ipforward_rt, mcopy); 1504 m_freem(mcopy); 1505 } 1506 return; 1507 } 1508 } 1509 if (mcopy == NULL) 1510 return; 1511 destifp = NULL; 1512 1513 switch (error) { 1514 1515 case 0: /* forwarded, but need redirect */ 1516 /* type, code set above */ 1517 break; 1518 1519 case ENETUNREACH: /* shouldn't happen, checked above */ 1520 case EHOSTUNREACH: 1521 case ENETDOWN: 1522 case EHOSTDOWN: 1523 default: 1524 type = ICMP_UNREACH; 1525 code = ICMP_UNREACH_HOST; 1526 break; 1527 1528 case EMSGSIZE: 1529 type = ICMP_UNREACH; 1530 code = ICMP_UNREACH_NEEDFRAG; 1531 if (ipforward_rt.ro_rt) 1532 destifp = ipforward_rt.ro_rt->rt_ifp; 1533 ipstat.ips_cantfrag++; 1534 break; 1535 1536 case ENOBUFS: 1537 type = ICMP_SOURCEQUENCH; 1538 code = 0; 1539 break; 1540 } 1541 icmp_error(mcopy, type, code, dest, destifp); 1542} 1543 1544void 1545ip_savecontrol(inp, mp, ip, m) 1546 register struct inpcb *inp; 1547 register struct mbuf **mp; 1548 register struct ip *ip; 1549 register struct mbuf *m; 1550{ 1551 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1552 struct timeval tv; 1553 1554 microtime(&tv); 1555 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1556 SCM_TIMESTAMP, SOL_SOCKET); 1557 if (*mp) 1558 mp = &(*mp)->m_next; 1559 } 1560 if (inp->inp_flags & INP_RECVDSTADDR) { 1561 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1562 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1563 if (*mp) 1564 mp = &(*mp)->m_next; 1565 } 1566#ifdef notyet 1567 /* XXX 1568 * Moving these out of udp_input() made them even more broken 1569 * than they already were. 1570 */ 1571 /* options were tossed already */ 1572 if (inp->inp_flags & INP_RECVOPTS) { 1573 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1574 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1575 if (*mp) 1576 mp = &(*mp)->m_next; 1577 } 1578 /* ip_srcroute doesn't do what we want here, need to fix */ 1579 if (inp->inp_flags & INP_RECVRETOPTS) { 1580 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1581 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1582 if (*mp) 1583 mp = &(*mp)->m_next; 1584 } 1585#endif 1586 if (inp->inp_flags & INP_RECVIF) { 1587 struct ifnet *ifp; 1588 struct sdlbuf { 1589 struct sockaddr_dl sdl; 1590 u_char pad[32]; 1591 } sdlbuf; 1592 struct sockaddr_dl *sdp; 1593 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1594 1595 if (((ifp = m->m_pkthdr.rcvif)) 1596 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1597 sdp = (struct sockaddr_dl *)(ifnet_addrs 1598 [ifp->if_index - 1]->ifa_addr); 1599 /* 1600 * Change our mind and don't try copy. 1601 */ 1602 if ((sdp->sdl_family != AF_LINK) 1603 || (sdp->sdl_len > sizeof(sdlbuf))) { 1604 goto makedummy; 1605 } 1606 bcopy(sdp, sdl2, sdp->sdl_len); 1607 } else { 1608makedummy: 1609 sdl2->sdl_len 1610 = offsetof(struct sockaddr_dl, sdl_data[0]); 1611 sdl2->sdl_family = AF_LINK; 1612 sdl2->sdl_index = 0; 1613 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1614 } 1615 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1616 IP_RECVIF, IPPROTO_IP); 1617 if (*mp) 1618 mp = &(*mp)->m_next; 1619 } 1620} 1621 1622int 1623ip_rsvp_init(struct socket *so) 1624{ 1625 if (so->so_type != SOCK_RAW || 1626 so->so_proto->pr_protocol != IPPROTO_RSVP) 1627 return EOPNOTSUPP; 1628 1629 if (ip_rsvpd != NULL) 1630 return EADDRINUSE; 1631 1632 ip_rsvpd = so; 1633 /* 1634 * This may seem silly, but we need to be sure we don't over-increment 1635 * the RSVP counter, in case something slips up. 1636 */ 1637 if (!ip_rsvp_on) { 1638 ip_rsvp_on = 1; 1639 rsvp_on++; 1640 } 1641 1642 return 0; 1643} 1644 1645int 1646ip_rsvp_done(void) 1647{ 1648 ip_rsvpd = NULL; 1649 /* 1650 * This may seem silly, but we need to be sure we don't over-decrement 1651 * the RSVP counter, in case something slips up. 1652 */ 1653 if (ip_rsvp_on) { 1654 ip_rsvp_on = 0; 1655 rsvp_on--; 1656 } 1657 return 0; 1658} 1659