ip_reass.c revision 57401
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 57401 2000-02-23 20:11:57Z guido $ 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#include "opt_ipsec.h" 46 47#include <stddef.h> 48 49#include <sys/param.h> 50#include <sys/systm.h> 51#include <sys/mbuf.h> 52#include <sys/malloc.h> 53#include <sys/domain.h> 54#include <sys/protosw.h> 55#include <sys/socket.h> 56#include <sys/time.h> 57#include <sys/kernel.h> 58#include <sys/syslog.h> 59#include <sys/sysctl.h> 60 61#include <net/if.h> 62#include <net/if_var.h> 63#include <net/if_dl.h> 64#include <net/route.h> 65#include <net/netisr.h> 66#include <net/intrq.h> 67 68#include <netinet/in.h> 69#include <netinet/in_systm.h> 70#include <netinet/in_var.h> 71#include <netinet/ip.h> 72#include <netinet/in_pcb.h> 73#include <netinet/ip_var.h> 74#include <netinet/ip_icmp.h> 75#include <machine/in_cksum.h> 76 77#include <netinet/ipprotosw.h> 78 79#include <sys/socketvar.h> 80 81#include <netinet/ip_fw.h> 82 83#ifdef IPSEC 84#include <netinet6/ipsec.h> 85#include <netkey/key.h> 86#ifdef IPSEC_DEBUG 87#include <netkey/key_debug.h> 88#else 89#define KEYDEBUG(lev,arg) 90#endif 91#endif 92 93#include "faith.h" 94#if defined(NFAITH) && NFAITH > 0 95#include <net/if_types.h> 96#endif 97 98#ifdef DUMMYNET 99#include <netinet/ip_dummynet.h> 100#endif 101 102int rsvp_on = 0; 103static int ip_rsvp_on; 104struct socket *ip_rsvpd; 105 106int ipforwarding = 0; 107SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 108 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 109 110static int ipsendredirects = 1; /* XXX */ 111SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 112 &ipsendredirects, 0, "Enable sending IP redirects"); 113 114int ip_defttl = IPDEFTTL; 115SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 116 &ip_defttl, 0, "Maximum TTL on IP packets"); 117 118static int ip_dosourceroute = 0; 119SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 120 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 121 122static int ip_acceptsourceroute = 0; 123SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 124 CTLFLAG_RW, &ip_acceptsourceroute, 0, 125 "Enable accepting source routed IP packets"); 126 127static int ip_keepfaith = 0; 128SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 129 &ip_keepfaith, 0, 130 "Enable packet capture for FAITH IPv4->IPv6 translater daemon"); 131 132#ifdef DIAGNOSTIC 133static int ipprintfs = 0; 134#endif 135 136extern struct domain inetdomain; 137extern struct ipprotosw inetsw[]; 138u_char ip_protox[IPPROTO_MAX]; 139static int ipqmaxlen = IFQ_MAXLEN; 140struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 141SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW, 142 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 143SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD, 144 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 145 146struct ipstat ipstat; 147SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD, 148 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 149 150/* Packet reassembly stuff */ 151#define IPREASS_NHASH_LOG2 6 152#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 153#define IPREASS_HMASK (IPREASS_NHASH - 1) 154#define IPREASS_HASH(x,y) \ 155 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 156 157static struct ipq ipq[IPREASS_NHASH]; 158static int nipq = 0; /* total # of reass queues */ 159static int maxnipq; 160const int ipintrq_present = 1; 161 162#ifdef IPCTL_DEFMTU 163SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 164 &ip_mtu, 0, "Default MTU"); 165#endif 166 167#ifdef IPSTEALTH 168static int ipstealth = 0; 169SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, 170 &ipstealth, 0, ""); 171#endif 172 173 174/* Firewall hooks */ 175ip_fw_chk_t *ip_fw_chk_ptr; 176ip_fw_ctl_t *ip_fw_ctl_ptr; 177int fw_enable = 1 ; 178 179#ifdef DUMMYNET 180ip_dn_ctl_t *ip_dn_ctl_ptr; 181#endif 182 183int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL; 184 185 186/* 187 * We need to save the IP options in case a protocol wants to respond 188 * to an incoming packet over the same route if the packet got here 189 * using IP source routing. This allows connection establishment and 190 * maintenance when the remote end is on a network that is not known 191 * to us. 192 */ 193static int ip_nhops = 0; 194static struct ip_srcrt { 195 struct in_addr dst; /* final destination */ 196 char nop; /* one NOP to align */ 197 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 198 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 199} ip_srcrt; 200 201struct sockaddr_in *ip_fw_fwd_addr; 202 203static void save_rte __P((u_char *, struct in_addr)); 204static int ip_dooptions __P((struct mbuf *)); 205static void ip_forward __P((struct mbuf *, int)); 206static void ip_freef __P((struct ipq *)); 207#ifdef IPDIVERT 208static struct mbuf *ip_reass __P((struct mbuf *, 209 struct ipq *, struct ipq *, u_int32_t *, u_int16_t *)); 210#else 211static struct mbuf *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *)); 212#endif 213static struct in_ifaddr *ip_rtaddr __P((struct in_addr)); 214static void ipintr __P((void)); 215 216/* 217 * IP initialization: fill in IP protocol switch table. 218 * All protocols not implemented in kernel go to raw IP protocol handler. 219 */ 220void 221ip_init() 222{ 223 register struct ipprotosw *pr; 224 register int i; 225 226 TAILQ_INIT(&in_ifaddrhead); 227 pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 228 if (pr == 0) 229 panic("ip_init"); 230 for (i = 0; i < IPPROTO_MAX; i++) 231 ip_protox[i] = pr - inetsw; 232 for (pr = (struct ipprotosw *)inetdomain.dom_protosw; 233 pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++) 234 if (pr->pr_domain->dom_family == PF_INET && 235 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 236 ip_protox[pr->pr_protocol] = pr - inetsw; 237 238 for (i = 0; i < IPREASS_NHASH; i++) 239 ipq[i].next = ipq[i].prev = &ipq[i]; 240 241 maxnipq = nmbclusters/4; 242 243 ip_id = time_second & 0xffff; 244 ipintrq.ifq_maxlen = ipqmaxlen; 245 246 register_netisr(NETISR_IP, ipintr); 247} 248 249static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 250static struct route ipforward_rt; 251 252/* 253 * Ip input routine. Checksum and byte swap header. If fragmented 254 * try to reassemble. Process options. Pass to next level. 255 */ 256void 257ip_input(struct mbuf *m) 258{ 259 struct ip *ip; 260 struct ipq *fp; 261 struct in_ifaddr *ia; 262 int i, hlen, mff; 263 u_short sum; 264 u_int16_t divert_cookie; /* firewall cookie */ 265#ifdef IPDIVERT 266 u_int32_t divert_info = 0; /* packet divert/tee info */ 267#endif 268 struct ip_fw_chain *rule = NULL; 269 270#ifdef IPDIVERT 271 /* Get and reset firewall cookie */ 272 divert_cookie = ip_divert_cookie; 273 ip_divert_cookie = 0; 274#else 275 divert_cookie = 0; 276#endif 277 278#if defined(IPFIREWALL) && defined(DUMMYNET) 279 /* 280 * dummynet packet are prepended a vestigial mbuf with 281 * m_type = MT_DUMMYNET and m_data pointing to the matching 282 * rule. 283 */ 284 if (m->m_type == MT_DUMMYNET) { 285 rule = (struct ip_fw_chain *)(m->m_data) ; 286 m = m->m_next ; 287 ip = mtod(m, struct ip *); 288 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 289 goto iphack ; 290 } else 291 rule = NULL ; 292#endif 293 294#ifdef DIAGNOSTIC 295 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 296 panic("ip_input no HDR"); 297#endif 298 ipstat.ips_total++; 299 300 if (m->m_pkthdr.len < sizeof(struct ip)) 301 goto tooshort; 302 303 if (m->m_len < sizeof (struct ip) && 304 (m = m_pullup(m, sizeof (struct ip))) == 0) { 305 ipstat.ips_toosmall++; 306 return; 307 } 308 ip = mtod(m, struct ip *); 309 310 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 311 ipstat.ips_badvers++; 312 goto bad; 313 } 314 315 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 316 if (hlen < sizeof(struct ip)) { /* minimum header length */ 317 ipstat.ips_badhlen++; 318 goto bad; 319 } 320 if (hlen > m->m_len) { 321 if ((m = m_pullup(m, hlen)) == 0) { 322 ipstat.ips_badhlen++; 323 return; 324 } 325 ip = mtod(m, struct ip *); 326 } 327 if (hlen == sizeof(struct ip)) { 328 sum = in_cksum_hdr(ip); 329 } else { 330 sum = in_cksum(m, hlen); 331 } 332 if (sum) { 333 ipstat.ips_badsum++; 334 goto bad; 335 } 336 337 /* 338 * Convert fields to host representation. 339 */ 340 NTOHS(ip->ip_len); 341 if (ip->ip_len < hlen) { 342 ipstat.ips_badlen++; 343 goto bad; 344 } 345 NTOHS(ip->ip_id); 346 NTOHS(ip->ip_off); 347 348 /* 349 * Check that the amount of data in the buffers 350 * is as at least much as the IP header would have us expect. 351 * Trim mbufs if longer than we expect. 352 * Drop packet if shorter than we expect. 353 */ 354 if (m->m_pkthdr.len < ip->ip_len) { 355tooshort: 356 ipstat.ips_tooshort++; 357 goto bad; 358 } 359 if (m->m_pkthdr.len > ip->ip_len) { 360 if (m->m_len == m->m_pkthdr.len) { 361 m->m_len = ip->ip_len; 362 m->m_pkthdr.len = ip->ip_len; 363 } else 364 m_adj(m, ip->ip_len - m->m_pkthdr.len); 365 } 366 /* 367 * IpHack's section. 368 * Right now when no processing on packet has done 369 * and it is still fresh out of network we do our black 370 * deals with it. 371 * - Firewall: deny/allow/divert 372 * - Xlate: translate packet's addr/port (NAT). 373 * - Pipe: pass pkt through dummynet. 374 * - Wrap: fake packet's addr/port <unimpl.> 375 * - Encapsulate: put it in another IP and send out. <unimp.> 376 */ 377 378#if defined(IPFIREWALL) && defined(DUMMYNET) 379iphack: 380#endif 381 /* 382 * Check if we want to allow this packet to be processed. 383 * Consider it to be bad if not. 384 */ 385 if (fr_checkp) { 386 struct mbuf *m1 = m; 387 388 if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1) 389 return; 390 ip = mtod(m = m1, struct ip *); 391 } 392 if (fw_enable && ip_fw_chk_ptr) { 393#ifdef IPFIREWALL_FORWARD 394 /* 395 * If we've been forwarded from the output side, then 396 * skip the firewall a second time 397 */ 398 if (ip_fw_fwd_addr) 399 goto ours; 400#endif /* IPFIREWALL_FORWARD */ 401 /* 402 * See the comment in ip_output for the return values 403 * produced by the firewall. 404 */ 405 i = (*ip_fw_chk_ptr)(&ip, 406 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 407 if (m == NULL) /* Packet discarded by firewall */ 408 return; 409 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 410 goto pass; 411#ifdef DUMMYNET 412 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) { 413 /* Send packet to the appropriate pipe */ 414 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 415 0); 416 return; 417 } 418#endif 419#ifdef IPDIVERT 420 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 421 /* Divert or tee packet */ 422 divert_info = i; 423 goto ours; 424 } 425#endif 426#ifdef IPFIREWALL_FORWARD 427 if (i == 0 && ip_fw_fwd_addr != NULL) 428 goto pass; 429#endif 430 /* 431 * if we get here, the packet must be dropped 432 */ 433 m_freem(m); 434 return; 435 } 436pass: 437 438 /* 439 * Process options and, if not destined for us, 440 * ship it on. ip_dooptions returns 1 when an 441 * error was detected (causing an icmp message 442 * to be sent and the original packet to be freed). 443 */ 444 ip_nhops = 0; /* for source routed packets */ 445 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 446#ifdef IPFIREWALL_FORWARD 447 ip_fw_fwd_addr = NULL; 448#endif 449 return; 450 } 451 452 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 453 * matter if it is destined to another node, or whether it is 454 * a multicast one, RSVP wants it! and prevents it from being forwarded 455 * anywhere else. Also checks if the rsvp daemon is running before 456 * grabbing the packet. 457 */ 458 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 459 goto ours; 460 461 /* 462 * Check our list of addresses, to see if the packet is for us. 463 * If we don't have any addresses, assume any unicast packet 464 * we receive might be for us (and let the upper layers deal 465 * with it). 466 */ 467 if (TAILQ_EMPTY(&in_ifaddrhead) && 468 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 469 goto ours; 470 471 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 472 ia = TAILQ_NEXT(ia, ia_link)) { 473#define satosin(sa) ((struct sockaddr_in *)(sa)) 474 475#ifdef BOOTP_COMPAT 476 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 477 goto ours; 478#endif 479#ifdef IPFIREWALL_FORWARD 480 /* 481 * If the addr to forward to is one of ours, we pretend to 482 * be the destination for this packet. 483 */ 484 if (ip_fw_fwd_addr == NULL) { 485 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 486 goto ours; 487 } else if (IA_SIN(ia)->sin_addr.s_addr == 488 ip_fw_fwd_addr->sin_addr.s_addr) 489 goto ours; 490#else 491 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 492 goto ours; 493#endif 494 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 495 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 496 ip->ip_dst.s_addr) 497 goto ours; 498 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 499 goto ours; 500 } 501 } 502 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 503 struct in_multi *inm; 504 if (ip_mrouter) { 505 /* 506 * If we are acting as a multicast router, all 507 * incoming multicast packets are passed to the 508 * kernel-level multicast forwarding function. 509 * The packet is returned (relatively) intact; if 510 * ip_mforward() returns a non-zero value, the packet 511 * must be discarded, else it may be accepted below. 512 * 513 * (The IP ident field is put in the same byte order 514 * as expected when ip_mforward() is called from 515 * ip_output().) 516 */ 517 ip->ip_id = htons(ip->ip_id); 518 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 519 ipstat.ips_cantforward++; 520 m_freem(m); 521 return; 522 } 523 ip->ip_id = ntohs(ip->ip_id); 524 525 /* 526 * The process-level routing demon needs to receive 527 * all multicast IGMP packets, whether or not this 528 * host belongs to their destination groups. 529 */ 530 if (ip->ip_p == IPPROTO_IGMP) 531 goto ours; 532 ipstat.ips_forward++; 533 } 534 /* 535 * See if we belong to the destination multicast group on the 536 * arrival interface. 537 */ 538 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 539 if (inm == NULL) { 540 ipstat.ips_notmember++; 541 m_freem(m); 542 return; 543 } 544 goto ours; 545 } 546 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 547 goto ours; 548 if (ip->ip_dst.s_addr == INADDR_ANY) 549 goto ours; 550 551#if defined(NFAITH) && 0 < NFAITH 552 /* 553 * FAITH(Firewall Aided Internet Translator) 554 */ 555 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 556 if (ip_keepfaith) { 557 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 558 goto ours; 559 } 560 m_freem(m); 561 return; 562 } 563#endif 564 /* 565 * Not for us; forward if possible and desirable. 566 */ 567 if (ipforwarding == 0) { 568 ipstat.ips_cantforward++; 569 m_freem(m); 570 } else 571 ip_forward(m, 0); 572#ifdef IPFIREWALL_FORWARD 573 ip_fw_fwd_addr = NULL; 574#endif 575 return; 576 577ours: 578 579 /* 580 * If offset or IP_MF are set, must reassemble. 581 * Otherwise, nothing need be done. 582 * (We could look in the reassembly queue to see 583 * if the packet was previously fragmented, 584 * but it's not worth the time; just let them time out.) 585 */ 586 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) { 587 588#if 0 /* 589 * Reassembly should be able to treat a mbuf cluster, for later 590 * operation of contiguous protocol headers on the cluster. (KAME) 591 */ 592 if (m->m_flags & M_EXT) { /* XXX */ 593 if ((m = m_pullup(m, hlen)) == 0) { 594 ipstat.ips_toosmall++; 595#ifdef IPFIREWALL_FORWARD 596 ip_fw_fwd_addr = NULL; 597#endif 598 return; 599 } 600 ip = mtod(m, struct ip *); 601 } 602#endif 603 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 604 /* 605 * Look for queue of fragments 606 * of this datagram. 607 */ 608 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 609 if (ip->ip_id == fp->ipq_id && 610 ip->ip_src.s_addr == fp->ipq_src.s_addr && 611 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 612 ip->ip_p == fp->ipq_p) 613 goto found; 614 615 fp = 0; 616 617 /* check if there's a place for the new queue */ 618 if (nipq > maxnipq) { 619 /* 620 * drop something from the tail of the current queue 621 * before proceeding further 622 */ 623 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 624 for (i = 0; i < IPREASS_NHASH; i++) { 625 if (ipq[i].prev != &ipq[i]) { 626 ip_freef(ipq[i].prev); 627 break; 628 } 629 } 630 } else 631 ip_freef(ipq[sum].prev); 632 } 633found: 634 /* 635 * Adjust ip_len to not reflect header, 636 * set ip_mff if more fragments are expected, 637 * convert offset of this to bytes. 638 */ 639 ip->ip_len -= hlen; 640 mff = (ip->ip_off & IP_MF) != 0; 641 if (mff) { 642 /* 643 * Make sure that fragments have a data length 644 * that's a non-zero multiple of 8 bytes. 645 */ 646 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 647 ipstat.ips_toosmall++; /* XXX */ 648 goto bad; 649 } 650 m->m_flags |= M_FRAG; 651 } 652 ip->ip_off <<= 3; 653 654 /* 655 * If datagram marked as having more fragments 656 * or if this is not the first fragment, 657 * attempt reassembly; if it succeeds, proceed. 658 */ 659 if (mff || ip->ip_off) { 660 ipstat.ips_fragments++; 661 m->m_pkthdr.header = ip; 662#ifdef IPDIVERT 663 m = ip_reass(m, 664 fp, &ipq[sum], &divert_info, &divert_cookie); 665#else 666 m = ip_reass(m, fp, &ipq[sum]); 667#endif 668 if (m == 0) { 669#ifdef IPFIREWALL_FORWARD 670 ip_fw_fwd_addr = NULL; 671#endif 672 return; 673 } 674 /* Get the length of the reassembled packets header */ 675 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 676 ipstat.ips_reassembled++; 677 ip = mtod(m, struct ip *); 678#ifdef IPDIVERT 679 /* Restore original checksum before diverting packet */ 680 if (divert_info != 0) { 681 ip->ip_len += hlen; 682 HTONS(ip->ip_len); 683 HTONS(ip->ip_off); 684 HTONS(ip->ip_id); 685 ip->ip_sum = 0; 686 ip->ip_sum = in_cksum_hdr(ip); 687 NTOHS(ip->ip_id); 688 NTOHS(ip->ip_off); 689 NTOHS(ip->ip_len); 690 ip->ip_len -= hlen; 691 } 692#endif 693 } else 694 if (fp) 695 ip_freef(fp); 696 } else 697 ip->ip_len -= hlen; 698 699#ifdef IPDIVERT 700 /* 701 * Divert or tee packet to the divert protocol if required. 702 * 703 * If divert_info is zero then cookie should be too, so we shouldn't 704 * need to clear them here. Assume divert_packet() does so also. 705 */ 706 if (divert_info != 0) { 707 struct mbuf *clone = NULL; 708 709 /* Clone packet if we're doing a 'tee' */ 710 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 711 clone = m_dup(m, M_DONTWAIT); 712 713 /* Restore packet header fields to original values */ 714 ip->ip_len += hlen; 715 HTONS(ip->ip_len); 716 HTONS(ip->ip_off); 717 HTONS(ip->ip_id); 718 719 /* Deliver packet to divert input routine */ 720 ip_divert_cookie = divert_cookie; 721 divert_packet(m, 1, divert_info & 0xffff); 722 ipstat.ips_delivered++; 723 724 /* If 'tee', continue with original packet */ 725 if (clone == NULL) 726 return; 727 m = clone; 728 ip = mtod(m, struct ip *); 729 } 730#endif 731 732 /* 733 * Switch out to protocol's input routine. 734 */ 735 ipstat.ips_delivered++; 736 { 737 int off = hlen, nh = ip->ip_p; 738 739 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh); 740#ifdef IPFIREWALL_FORWARD 741 ip_fw_fwd_addr = NULL; /* tcp needed it */ 742#endif 743 return; 744 } 745bad: 746#ifdef IPFIREWALL_FORWARD 747 ip_fw_fwd_addr = NULL; 748#endif 749 m_freem(m); 750} 751 752/* 753 * IP software interrupt routine - to go away sometime soon 754 */ 755static void 756ipintr(void) 757{ 758 int s; 759 struct mbuf *m; 760 761 while(1) { 762 s = splimp(); 763 IF_DEQUEUE(&ipintrq, m); 764 splx(s); 765 if (m == 0) 766 return; 767 ip_input(m); 768 } 769} 770 771/* 772 * Take incoming datagram fragment and try to reassemble it into 773 * whole datagram. If a chain for reassembly of this datagram already 774 * exists, then it is given as fp; otherwise have to make a chain. 775 * 776 * When IPDIVERT enabled, keep additional state with each packet that 777 * tells us if we need to divert or tee the packet we're building. 778 */ 779 780static struct mbuf * 781#ifdef IPDIVERT 782ip_reass(m, fp, where, divinfo, divcookie) 783#else 784ip_reass(m, fp, where) 785#endif 786 register struct mbuf *m; 787 register struct ipq *fp; 788 struct ipq *where; 789#ifdef IPDIVERT 790 u_int32_t *divinfo; 791 u_int16_t *divcookie; 792#endif 793{ 794 struct ip *ip = mtod(m, struct ip *); 795 register struct mbuf *p = 0, *q, *nq; 796 struct mbuf *t; 797 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 798 int i, next; 799 800 /* 801 * Presence of header sizes in mbufs 802 * would confuse code below. 803 */ 804 m->m_data += hlen; 805 m->m_len -= hlen; 806 807 /* 808 * If first fragment to arrive, create a reassembly queue. 809 */ 810 if (fp == 0) { 811 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 812 goto dropfrag; 813 fp = mtod(t, struct ipq *); 814 insque(fp, where); 815 nipq++; 816 fp->ipq_ttl = IPFRAGTTL; 817 fp->ipq_p = ip->ip_p; 818 fp->ipq_id = ip->ip_id; 819 fp->ipq_src = ip->ip_src; 820 fp->ipq_dst = ip->ip_dst; 821 fp->ipq_frags = m; 822 m->m_nextpkt = NULL; 823#ifdef IPDIVERT 824 fp->ipq_div_info = 0; 825 fp->ipq_div_cookie = 0; 826#endif 827 goto inserted; 828 } 829 830#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 831 832 /* 833 * Find a segment which begins after this one does. 834 */ 835 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 836 if (GETIP(q)->ip_off > ip->ip_off) 837 break; 838 839 /* 840 * If there is a preceding segment, it may provide some of 841 * our data already. If so, drop the data from the incoming 842 * segment. If it provides all of our data, drop us, otherwise 843 * stick new segment in the proper place. 844 */ 845 if (p) { 846 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 847 if (i > 0) { 848 if (i >= ip->ip_len) 849 goto dropfrag; 850 m_adj(m, i); 851 ip->ip_off += i; 852 ip->ip_len -= i; 853 } 854 m->m_nextpkt = p->m_nextpkt; 855 p->m_nextpkt = m; 856 } else { 857 m->m_nextpkt = fp->ipq_frags; 858 fp->ipq_frags = m; 859 } 860 861 /* 862 * While we overlap succeeding segments trim them or, 863 * if they are completely covered, dequeue them. 864 */ 865 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 866 q = nq) { 867 i = (ip->ip_off + ip->ip_len) - 868 GETIP(q)->ip_off; 869 if (i < GETIP(q)->ip_len) { 870 GETIP(q)->ip_len -= i; 871 GETIP(q)->ip_off += i; 872 m_adj(q, i); 873 break; 874 } 875 nq = q->m_nextpkt; 876 m->m_nextpkt = nq; 877 m_freem(q); 878 } 879 880inserted: 881 882#ifdef IPDIVERT 883 /* 884 * Transfer firewall instructions to the fragment structure. 885 * Any fragment diverting causes the whole packet to divert. 886 */ 887 fp->ipq_div_info = *divinfo; 888 fp->ipq_div_cookie = *divcookie; 889 *divinfo = 0; 890 *divcookie = 0; 891#endif 892 893 /* 894 * Check for complete reassembly. 895 */ 896 next = 0; 897 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 898 if (GETIP(q)->ip_off != next) 899 return (0); 900 next += GETIP(q)->ip_len; 901 } 902 /* Make sure the last packet didn't have the IP_MF flag */ 903 if (p->m_flags & M_FRAG) 904 return (0); 905 906 /* 907 * Reassembly is complete. Make sure the packet is a sane size. 908 */ 909 q = fp->ipq_frags; 910 ip = GETIP(q); 911 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 912 ipstat.ips_toolong++; 913 ip_freef(fp); 914 return (0); 915 } 916 917 /* 918 * Concatenate fragments. 919 */ 920 m = q; 921 t = m->m_next; 922 m->m_next = 0; 923 m_cat(m, t); 924 nq = q->m_nextpkt; 925 q->m_nextpkt = 0; 926 for (q = nq; q != NULL; q = nq) { 927 nq = q->m_nextpkt; 928 q->m_nextpkt = NULL; 929 m_cat(m, q); 930 } 931 932#ifdef IPDIVERT 933 /* 934 * Extract firewall instructions from the fragment structure. 935 */ 936 *divinfo = fp->ipq_div_info; 937 *divcookie = fp->ipq_div_cookie; 938#endif 939 940 /* 941 * Create header for new ip packet by 942 * modifying header of first packet; 943 * dequeue and discard fragment reassembly header. 944 * Make header visible. 945 */ 946 ip->ip_len = next; 947 ip->ip_src = fp->ipq_src; 948 ip->ip_dst = fp->ipq_dst; 949 remque(fp); 950 nipq--; 951 (void) m_free(dtom(fp)); 952 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 953 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 954 /* some debugging cruft by sklower, below, will go away soon */ 955 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 956 register int plen = 0; 957 for (t = m; t; t = t->m_next) 958 plen += t->m_len; 959 m->m_pkthdr.len = plen; 960 } 961 return (m); 962 963dropfrag: 964#ifdef IPDIVERT 965 *divinfo = 0; 966 *divcookie = 0; 967#endif 968 ipstat.ips_fragdropped++; 969 m_freem(m); 970 return (0); 971 972#undef GETIP 973} 974 975/* 976 * Free a fragment reassembly header and all 977 * associated datagrams. 978 */ 979static void 980ip_freef(fp) 981 struct ipq *fp; 982{ 983 register struct mbuf *q; 984 985 while (fp->ipq_frags) { 986 q = fp->ipq_frags; 987 fp->ipq_frags = q->m_nextpkt; 988 m_freem(q); 989 } 990 remque(fp); 991 (void) m_free(dtom(fp)); 992 nipq--; 993} 994 995/* 996 * IP timer processing; 997 * if a timer expires on a reassembly 998 * queue, discard it. 999 */ 1000void 1001ip_slowtimo() 1002{ 1003 register struct ipq *fp; 1004 int s = splnet(); 1005 int i; 1006 1007 for (i = 0; i < IPREASS_NHASH; i++) { 1008 fp = ipq[i].next; 1009 if (fp == 0) 1010 continue; 1011 while (fp != &ipq[i]) { 1012 --fp->ipq_ttl; 1013 fp = fp->next; 1014 if (fp->prev->ipq_ttl == 0) { 1015 ipstat.ips_fragtimeout++; 1016 ip_freef(fp->prev); 1017 } 1018 } 1019 } 1020 ipflow_slowtimo(); 1021 splx(s); 1022} 1023 1024/* 1025 * Drain off all datagram fragments. 1026 */ 1027void 1028ip_drain() 1029{ 1030 int i; 1031 1032 for (i = 0; i < IPREASS_NHASH; i++) { 1033 while (ipq[i].next != &ipq[i]) { 1034 ipstat.ips_fragdropped++; 1035 ip_freef(ipq[i].next); 1036 } 1037 } 1038 in_rtqdrain(); 1039} 1040 1041/* 1042 * Do option processing on a datagram, 1043 * possibly discarding it if bad options are encountered, 1044 * or forwarding it if source-routed. 1045 * Returns 1 if packet has been forwarded/freed, 1046 * 0 if the packet should be processed further. 1047 */ 1048static int 1049ip_dooptions(m) 1050 struct mbuf *m; 1051{ 1052 register struct ip *ip = mtod(m, struct ip *); 1053 register u_char *cp; 1054 register struct ip_timestamp *ipt; 1055 register struct in_ifaddr *ia; 1056 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1057 struct in_addr *sin, dst; 1058 n_time ntime; 1059 1060 dst = ip->ip_dst; 1061 cp = (u_char *)(ip + 1); 1062 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1063 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1064 opt = cp[IPOPT_OPTVAL]; 1065 if (opt == IPOPT_EOL) 1066 break; 1067 if (opt == IPOPT_NOP) 1068 optlen = 1; 1069 else { 1070 optlen = cp[IPOPT_OLEN]; 1071 if (optlen <= 0 || optlen > cnt) { 1072 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1073 goto bad; 1074 } 1075 } 1076 switch (opt) { 1077 1078 default: 1079 break; 1080 1081 /* 1082 * Source routing with record. 1083 * Find interface with current destination address. 1084 * If none on this machine then drop if strictly routed, 1085 * or do nothing if loosely routed. 1086 * Record interface address and bring up next address 1087 * component. If strictly routed make sure next 1088 * address is on directly accessible net. 1089 */ 1090 case IPOPT_LSRR: 1091 case IPOPT_SSRR: 1092 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1093 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1094 goto bad; 1095 } 1096 ipaddr.sin_addr = ip->ip_dst; 1097 ia = (struct in_ifaddr *) 1098 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1099 if (ia == 0) { 1100 if (opt == IPOPT_SSRR) { 1101 type = ICMP_UNREACH; 1102 code = ICMP_UNREACH_SRCFAIL; 1103 goto bad; 1104 } 1105 if (!ip_dosourceroute) 1106 goto nosourcerouting; 1107 /* 1108 * Loose routing, and not at next destination 1109 * yet; nothing to do except forward. 1110 */ 1111 break; 1112 } 1113 off--; /* 0 origin */ 1114 if (off > optlen - sizeof(struct in_addr)) { 1115 /* 1116 * End of source route. Should be for us. 1117 */ 1118 if (!ip_acceptsourceroute) 1119 goto nosourcerouting; 1120 save_rte(cp, ip->ip_src); 1121 break; 1122 } 1123 1124 if (!ip_dosourceroute) { 1125 if (ipforwarding) { 1126 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1127 /* 1128 * Acting as a router, so generate ICMP 1129 */ 1130nosourcerouting: 1131 strcpy(buf, inet_ntoa(ip->ip_dst)); 1132 log(LOG_WARNING, 1133 "attempted source route from %s to %s\n", 1134 inet_ntoa(ip->ip_src), buf); 1135 type = ICMP_UNREACH; 1136 code = ICMP_UNREACH_SRCFAIL; 1137 goto bad; 1138 } else { 1139 /* 1140 * Not acting as a router, so silently drop. 1141 */ 1142 ipstat.ips_cantforward++; 1143 m_freem(m); 1144 return (1); 1145 } 1146 } 1147 1148 /* 1149 * locate outgoing interface 1150 */ 1151 (void)memcpy(&ipaddr.sin_addr, cp + off, 1152 sizeof(ipaddr.sin_addr)); 1153 1154 if (opt == IPOPT_SSRR) { 1155#define INA struct in_ifaddr * 1156#define SA struct sockaddr * 1157 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1158 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1159 } else 1160 ia = ip_rtaddr(ipaddr.sin_addr); 1161 if (ia == 0) { 1162 type = ICMP_UNREACH; 1163 code = ICMP_UNREACH_SRCFAIL; 1164 goto bad; 1165 } 1166 ip->ip_dst = ipaddr.sin_addr; 1167 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1168 sizeof(struct in_addr)); 1169 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1170 /* 1171 * Let ip_intr's mcast routing check handle mcast pkts 1172 */ 1173 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1174 break; 1175 1176 case IPOPT_RR: 1177 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1178 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1179 goto bad; 1180 } 1181 /* 1182 * If no space remains, ignore. 1183 */ 1184 off--; /* 0 origin */ 1185 if (off > optlen - sizeof(struct in_addr)) 1186 break; 1187 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1188 sizeof(ipaddr.sin_addr)); 1189 /* 1190 * locate outgoing interface; if we're the destination, 1191 * use the incoming interface (should be same). 1192 */ 1193 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1194 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1195 type = ICMP_UNREACH; 1196 code = ICMP_UNREACH_HOST; 1197 goto bad; 1198 } 1199 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1200 sizeof(struct in_addr)); 1201 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1202 break; 1203 1204 case IPOPT_TS: 1205 code = cp - (u_char *)ip; 1206 ipt = (struct ip_timestamp *)cp; 1207 if (ipt->ipt_len < 5) 1208 goto bad; 1209 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1210 if (++ipt->ipt_oflw == 0) 1211 goto bad; 1212 break; 1213 } 1214 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1215 switch (ipt->ipt_flg) { 1216 1217 case IPOPT_TS_TSONLY: 1218 break; 1219 1220 case IPOPT_TS_TSANDADDR: 1221 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1222 sizeof(struct in_addr) > ipt->ipt_len) 1223 goto bad; 1224 ipaddr.sin_addr = dst; 1225 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1226 m->m_pkthdr.rcvif); 1227 if (ia == 0) 1228 continue; 1229 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1230 sizeof(struct in_addr)); 1231 ipt->ipt_ptr += sizeof(struct in_addr); 1232 break; 1233 1234 case IPOPT_TS_PRESPEC: 1235 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1236 sizeof(struct in_addr) > ipt->ipt_len) 1237 goto bad; 1238 (void)memcpy(&ipaddr.sin_addr, sin, 1239 sizeof(struct in_addr)); 1240 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1241 continue; 1242 ipt->ipt_ptr += sizeof(struct in_addr); 1243 break; 1244 1245 default: 1246 goto bad; 1247 } 1248 ntime = iptime(); 1249 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1250 sizeof(n_time)); 1251 ipt->ipt_ptr += sizeof(n_time); 1252 } 1253 } 1254 if (forward && ipforwarding) { 1255 ip_forward(m, 1); 1256 return (1); 1257 } 1258 return (0); 1259bad: 1260 ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2; /* XXX icmp_error adds in hdr length */ 1261 icmp_error(m, type, code, 0, 0); 1262 ipstat.ips_badoptions++; 1263 return (1); 1264} 1265 1266/* 1267 * Given address of next destination (final or next hop), 1268 * return internet address info of interface to be used to get there. 1269 */ 1270static struct in_ifaddr * 1271ip_rtaddr(dst) 1272 struct in_addr dst; 1273{ 1274 register struct sockaddr_in *sin; 1275 1276 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1277 1278 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1279 if (ipforward_rt.ro_rt) { 1280 RTFREE(ipforward_rt.ro_rt); 1281 ipforward_rt.ro_rt = 0; 1282 } 1283 sin->sin_family = AF_INET; 1284 sin->sin_len = sizeof(*sin); 1285 sin->sin_addr = dst; 1286 1287 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1288 } 1289 if (ipforward_rt.ro_rt == 0) 1290 return ((struct in_ifaddr *)0); 1291 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1292} 1293 1294/* 1295 * Save incoming source route for use in replies, 1296 * to be picked up later by ip_srcroute if the receiver is interested. 1297 */ 1298void 1299save_rte(option, dst) 1300 u_char *option; 1301 struct in_addr dst; 1302{ 1303 unsigned olen; 1304 1305 olen = option[IPOPT_OLEN]; 1306#ifdef DIAGNOSTIC 1307 if (ipprintfs) 1308 printf("save_rte: olen %d\n", olen); 1309#endif 1310 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1311 return; 1312 bcopy(option, ip_srcrt.srcopt, olen); 1313 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1314 ip_srcrt.dst = dst; 1315} 1316 1317/* 1318 * Retrieve incoming source route for use in replies, 1319 * in the same form used by setsockopt. 1320 * The first hop is placed before the options, will be removed later. 1321 */ 1322struct mbuf * 1323ip_srcroute() 1324{ 1325 register struct in_addr *p, *q; 1326 register struct mbuf *m; 1327 1328 if (ip_nhops == 0) 1329 return ((struct mbuf *)0); 1330 m = m_get(M_DONTWAIT, MT_HEADER); 1331 if (m == 0) 1332 return ((struct mbuf *)0); 1333 1334#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1335 1336 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1337 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1338 OPTSIZ; 1339#ifdef DIAGNOSTIC 1340 if (ipprintfs) 1341 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1342#endif 1343 1344 /* 1345 * First save first hop for return route 1346 */ 1347 p = &ip_srcrt.route[ip_nhops - 1]; 1348 *(mtod(m, struct in_addr *)) = *p--; 1349#ifdef DIAGNOSTIC 1350 if (ipprintfs) 1351 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1352#endif 1353 1354 /* 1355 * Copy option fields and padding (nop) to mbuf. 1356 */ 1357 ip_srcrt.nop = IPOPT_NOP; 1358 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1359 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1360 &ip_srcrt.nop, OPTSIZ); 1361 q = (struct in_addr *)(mtod(m, caddr_t) + 1362 sizeof(struct in_addr) + OPTSIZ); 1363#undef OPTSIZ 1364 /* 1365 * Record return path as an IP source route, 1366 * reversing the path (pointers are now aligned). 1367 */ 1368 while (p >= ip_srcrt.route) { 1369#ifdef DIAGNOSTIC 1370 if (ipprintfs) 1371 printf(" %lx", (u_long)ntohl(q->s_addr)); 1372#endif 1373 *q++ = *p--; 1374 } 1375 /* 1376 * Last hop goes to final destination. 1377 */ 1378 *q = ip_srcrt.dst; 1379#ifdef DIAGNOSTIC 1380 if (ipprintfs) 1381 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1382#endif 1383 return (m); 1384} 1385 1386/* 1387 * Strip out IP options, at higher 1388 * level protocol in the kernel. 1389 * Second argument is buffer to which options 1390 * will be moved, and return value is their length. 1391 * XXX should be deleted; last arg currently ignored. 1392 */ 1393void 1394ip_stripoptions(m, mopt) 1395 register struct mbuf *m; 1396 struct mbuf *mopt; 1397{ 1398 register int i; 1399 struct ip *ip = mtod(m, struct ip *); 1400 register caddr_t opts; 1401 int olen; 1402 1403 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1404 opts = (caddr_t)(ip + 1); 1405 i = m->m_len - (sizeof (struct ip) + olen); 1406 bcopy(opts + olen, opts, (unsigned)i); 1407 m->m_len -= olen; 1408 if (m->m_flags & M_PKTHDR) 1409 m->m_pkthdr.len -= olen; 1410 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1411} 1412 1413u_char inetctlerrmap[PRC_NCMDS] = { 1414 0, 0, 0, 0, 1415 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1416 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1417 EMSGSIZE, EHOSTUNREACH, 0, 0, 1418 0, 0, 0, 0, 1419 ENOPROTOOPT 1420}; 1421 1422/* 1423 * Forward a packet. If some error occurs return the sender 1424 * an icmp packet. Note we can't always generate a meaningful 1425 * icmp message because icmp doesn't have a large enough repertoire 1426 * of codes and types. 1427 * 1428 * If not forwarding, just drop the packet. This could be confusing 1429 * if ipforwarding was zero but some routing protocol was advancing 1430 * us as a gateway to somewhere. However, we must let the routing 1431 * protocol deal with that. 1432 * 1433 * The srcrt parameter indicates whether the packet is being forwarded 1434 * via a source route. 1435 */ 1436static void 1437ip_forward(m, srcrt) 1438 struct mbuf *m; 1439 int srcrt; 1440{ 1441 register struct ip *ip = mtod(m, struct ip *); 1442 register struct sockaddr_in *sin; 1443 register struct rtentry *rt; 1444 int error, type = 0, code = 0; 1445 struct mbuf *mcopy; 1446 n_long dest; 1447 struct ifnet *destifp; 1448#ifdef IPSEC 1449 struct ifnet dummyifp; 1450#endif 1451 1452 dest = 0; 1453#ifdef DIAGNOSTIC 1454 if (ipprintfs) 1455 printf("forward: src %lx dst %lx ttl %x\n", 1456 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1457 ip->ip_ttl); 1458#endif 1459 1460 1461 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1462 ipstat.ips_cantforward++; 1463 m_freem(m); 1464 return; 1465 } 1466 HTONS(ip->ip_id); 1467#ifdef IPSTEALTH 1468 if (!ipstealth) { 1469#endif 1470 if (ip->ip_ttl <= IPTTLDEC) { 1471 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1472 dest, 0); 1473 return; 1474 } 1475 ip->ip_ttl -= IPTTLDEC; 1476#ifdef IPSTEALTH 1477 } 1478#endif 1479 1480 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1481 if ((rt = ipforward_rt.ro_rt) == 0 || 1482 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1483 if (ipforward_rt.ro_rt) { 1484 RTFREE(ipforward_rt.ro_rt); 1485 ipforward_rt.ro_rt = 0; 1486 } 1487 sin->sin_family = AF_INET; 1488 sin->sin_len = sizeof(*sin); 1489 sin->sin_addr = ip->ip_dst; 1490 1491 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1492 if (ipforward_rt.ro_rt == 0) { 1493 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1494 return; 1495 } 1496 rt = ipforward_rt.ro_rt; 1497 } 1498 1499 /* 1500 * Save at most 64 bytes of the packet in case 1501 * we need to generate an ICMP message to the src. 1502 */ 1503 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1504 1505 /* 1506 * If forwarding packet using same interface that it came in on, 1507 * perhaps should send a redirect to sender to shortcut a hop. 1508 * Only send redirect if source is sending directly to us, 1509 * and if packet was not source routed (or has any options). 1510 * Also, don't send redirect if forwarding using a default route 1511 * or a route modified by a redirect. 1512 */ 1513#define satosin(sa) ((struct sockaddr_in *)(sa)) 1514 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1515 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1516 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1517 ipsendredirects && !srcrt) { 1518#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1519 u_long src = ntohl(ip->ip_src.s_addr); 1520 1521 if (RTA(rt) && 1522 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1523 if (rt->rt_flags & RTF_GATEWAY) 1524 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1525 else 1526 dest = ip->ip_dst.s_addr; 1527 /* Router requirements says to only send host redirects */ 1528 type = ICMP_REDIRECT; 1529 code = ICMP_REDIRECT_HOST; 1530#ifdef DIAGNOSTIC 1531 if (ipprintfs) 1532 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1533#endif 1534 } 1535 } 1536 1537 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1538 IP_FORWARDING, 0); 1539 if (error) 1540 ipstat.ips_cantforward++; 1541 else { 1542 ipstat.ips_forward++; 1543 if (type) 1544 ipstat.ips_redirectsent++; 1545 else { 1546 if (mcopy) { 1547 ipflow_create(&ipforward_rt, mcopy); 1548 m_freem(mcopy); 1549 } 1550 return; 1551 } 1552 } 1553 if (mcopy == NULL) 1554 return; 1555 destifp = NULL; 1556 1557 switch (error) { 1558 1559 case 0: /* forwarded, but need redirect */ 1560 /* type, code set above */ 1561 break; 1562 1563 case ENETUNREACH: /* shouldn't happen, checked above */ 1564 case EHOSTUNREACH: 1565 case ENETDOWN: 1566 case EHOSTDOWN: 1567 default: 1568 type = ICMP_UNREACH; 1569 code = ICMP_UNREACH_HOST; 1570 break; 1571 1572 case EMSGSIZE: 1573 type = ICMP_UNREACH; 1574 code = ICMP_UNREACH_NEEDFRAG; 1575#ifndef IPSEC 1576 if (ipforward_rt.ro_rt) 1577 destifp = ipforward_rt.ro_rt->rt_ifp; 1578#else 1579 /* 1580 * If the packet is routed over IPsec tunnel, tell the 1581 * originator the tunnel MTU. 1582 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1583 * XXX quickhack!!! 1584 */ 1585 if (ipforward_rt.ro_rt) { 1586 struct secpolicy *sp = NULL; 1587 int ipsecerror; 1588 int ipsechdr; 1589 struct route *ro; 1590 1591 sp = ipsec4_getpolicybyaddr(mcopy, 1592 IPSEC_DIR_OUTBOUND, 1593 IP_FORWARDING, 1594 &ipsecerror); 1595 1596 if (sp == NULL) 1597 destifp = ipforward_rt.ro_rt->rt_ifp; 1598 else { 1599 /* count IPsec header size */ 1600 ipsechdr = ipsec4_hdrsiz(mcopy, 1601 IPSEC_DIR_OUTBOUND, 1602 NULL); 1603 1604 /* 1605 * find the correct route for outer IPv4 1606 * header, compute tunnel MTU. 1607 * 1608 * XXX BUG ALERT 1609 * The "dummyifp" code relies upon the fact 1610 * that icmp_error() touches only ifp->if_mtu. 1611 */ 1612 /*XXX*/ 1613 destifp = NULL; 1614 if (sp->req != NULL 1615 && sp->req->sav != NULL 1616 && sp->req->sav->sah != NULL) { 1617 ro = &sp->req->sav->sah->sa_route; 1618 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1619 dummyifp.if_mtu = 1620 ro->ro_rt->rt_ifp->if_mtu; 1621 dummyifp.if_mtu -= ipsechdr; 1622 destifp = &dummyifp; 1623 } 1624 } 1625 1626 key_freesp(sp); 1627 } 1628 } 1629#endif /*IPSEC*/ 1630 ipstat.ips_cantfrag++; 1631 break; 1632 1633 case ENOBUFS: 1634 type = ICMP_SOURCEQUENCH; 1635 code = 0; 1636 break; 1637 } 1638 icmp_error(mcopy, type, code, dest, destifp); 1639} 1640 1641void 1642ip_savecontrol(inp, mp, ip, m) 1643 register struct inpcb *inp; 1644 register struct mbuf **mp; 1645 register struct ip *ip; 1646 register struct mbuf *m; 1647{ 1648 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1649 struct timeval tv; 1650 1651 microtime(&tv); 1652 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1653 SCM_TIMESTAMP, SOL_SOCKET); 1654 if (*mp) 1655 mp = &(*mp)->m_next; 1656 } 1657 if (inp->inp_flags & INP_RECVDSTADDR) { 1658 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1659 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1660 if (*mp) 1661 mp = &(*mp)->m_next; 1662 } 1663#ifdef notyet 1664 /* XXX 1665 * Moving these out of udp_input() made them even more broken 1666 * than they already were. 1667 */ 1668 /* options were tossed already */ 1669 if (inp->inp_flags & INP_RECVOPTS) { 1670 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1671 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1672 if (*mp) 1673 mp = &(*mp)->m_next; 1674 } 1675 /* ip_srcroute doesn't do what we want here, need to fix */ 1676 if (inp->inp_flags & INP_RECVRETOPTS) { 1677 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1678 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1679 if (*mp) 1680 mp = &(*mp)->m_next; 1681 } 1682#endif 1683 if (inp->inp_flags & INP_RECVIF) { 1684 struct ifnet *ifp; 1685 struct sdlbuf { 1686 struct sockaddr_dl sdl; 1687 u_char pad[32]; 1688 } sdlbuf; 1689 struct sockaddr_dl *sdp; 1690 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1691 1692 if (((ifp = m->m_pkthdr.rcvif)) 1693 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1694 sdp = (struct sockaddr_dl *)(ifnet_addrs 1695 [ifp->if_index - 1]->ifa_addr); 1696 /* 1697 * Change our mind and don't try copy. 1698 */ 1699 if ((sdp->sdl_family != AF_LINK) 1700 || (sdp->sdl_len > sizeof(sdlbuf))) { 1701 goto makedummy; 1702 } 1703 bcopy(sdp, sdl2, sdp->sdl_len); 1704 } else { 1705makedummy: 1706 sdl2->sdl_len 1707 = offsetof(struct sockaddr_dl, sdl_data[0]); 1708 sdl2->sdl_family = AF_LINK; 1709 sdl2->sdl_index = 0; 1710 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1711 } 1712 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1713 IP_RECVIF, IPPROTO_IP); 1714 if (*mp) 1715 mp = &(*mp)->m_next; 1716 } 1717} 1718 1719int 1720ip_rsvp_init(struct socket *so) 1721{ 1722 if (so->so_type != SOCK_RAW || 1723 so->so_proto->pr_protocol != IPPROTO_RSVP) 1724 return EOPNOTSUPP; 1725 1726 if (ip_rsvpd != NULL) 1727 return EADDRINUSE; 1728 1729 ip_rsvpd = so; 1730 /* 1731 * This may seem silly, but we need to be sure we don't over-increment 1732 * the RSVP counter, in case something slips up. 1733 */ 1734 if (!ip_rsvp_on) { 1735 ip_rsvp_on = 1; 1736 rsvp_on++; 1737 } 1738 1739 return 0; 1740} 1741 1742int 1743ip_rsvp_done(void) 1744{ 1745 ip_rsvpd = NULL; 1746 /* 1747 * This may seem silly, but we need to be sure we don't over-decrement 1748 * the RSVP counter, in case something slips up. 1749 */ 1750 if (ip_rsvp_on) { 1751 ip_rsvp_on = 0; 1752 rsvp_on--; 1753 } 1754 return 0; 1755} 1756