ip_input.c revision 56555
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 56555 2000-01-24 20:39:02Z brian $ 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; 177 178#ifdef DUMMYNET 179ip_dn_ctl_t *ip_dn_ctl_ptr; 180#endif 181 182#if defined(IPFILTER_LKM) || defined(IPFILTER) 183int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL; 184#endif 185 186 187/* 188 * We need to save the IP options in case a protocol wants to respond 189 * to an incoming packet over the same route if the packet got here 190 * using IP source routing. This allows connection establishment and 191 * maintenance when the remote end is on a network that is not known 192 * to us. 193 */ 194static int ip_nhops = 0; 195static struct ip_srcrt { 196 struct in_addr dst; /* final destination */ 197 char nop; /* one NOP to align */ 198 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 199 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 200} ip_srcrt; 201 202struct sockaddr_in *ip_fw_fwd_addr; 203 204static void save_rte __P((u_char *, struct in_addr)); 205static int ip_dooptions __P((struct mbuf *)); 206static void ip_forward __P((struct mbuf *, int)); 207static void ip_freef __P((struct ipq *)); 208#ifdef IPDIVERT 209static struct mbuf *ip_reass __P((struct mbuf *, 210 struct ipq *, struct ipq *, u_int32_t *, u_int16_t *)); 211#else 212static struct mbuf *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *)); 213#endif 214static struct in_ifaddr *ip_rtaddr __P((struct in_addr)); 215static void ipintr __P((void)); 216 217/* 218 * IP initialization: fill in IP protocol switch table. 219 * All protocols not implemented in kernel go to raw IP protocol handler. 220 */ 221void 222ip_init() 223{ 224 register struct ipprotosw *pr; 225 register int i; 226 227 TAILQ_INIT(&in_ifaddrhead); 228 pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 229 if (pr == 0) 230 panic("ip_init"); 231 for (i = 0; i < IPPROTO_MAX; i++) 232 ip_protox[i] = pr - inetsw; 233 for (pr = (struct ipprotosw *)inetdomain.dom_protosw; 234 pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++) 235 if (pr->pr_domain->dom_family == PF_INET && 236 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 237 ip_protox[pr->pr_protocol] = pr - inetsw; 238 239 for (i = 0; i < IPREASS_NHASH; i++) 240 ipq[i].next = ipq[i].prev = &ipq[i]; 241 242 maxnipq = nmbclusters/4; 243 244 ip_id = time_second & 0xffff; 245 ipintrq.ifq_maxlen = ipqmaxlen; 246} 247 248static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; 249static struct route ipforward_rt; 250 251/* 252 * Ip input routine. Checksum and byte swap header. If fragmented 253 * try to reassemble. Process options. Pass to next level. 254 */ 255void 256ip_input(struct mbuf *m) 257{ 258 struct ip *ip; 259 struct ipq *fp; 260 struct in_ifaddr *ia; 261 int i, hlen, mff; 262 u_short sum; 263 u_int16_t divert_cookie; /* firewall cookie */ 264#ifdef IPDIVERT 265 u_int32_t divert_info = 0; /* packet divert/tee info */ 266#endif 267 struct ip_fw_chain *rule = NULL; 268 269#ifdef IPDIVERT 270 /* Get and reset firewall cookie */ 271 divert_cookie = ip_divert_cookie; 272 ip_divert_cookie = 0; 273#else 274 divert_cookie = 0; 275#endif 276 277#if defined(IPFIREWALL) && defined(DUMMYNET) 278 /* 279 * dummynet packet are prepended a vestigial mbuf with 280 * m_type = MT_DUMMYNET and m_data pointing to the matching 281 * rule. 282 */ 283 if (m->m_type == MT_DUMMYNET) { 284 rule = (struct ip_fw_chain *)(m->m_data) ; 285 m = m->m_next ; 286 ip = mtod(m, struct ip *); 287 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 288 goto iphack ; 289 } else 290 rule = NULL ; 291#endif 292 293#ifdef DIAGNOSTIC 294 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 295 panic("ip_input no HDR"); 296#endif 297 ipstat.ips_total++; 298 299 if (m->m_pkthdr.len < sizeof(struct ip)) 300 goto tooshort; 301 302 if (m->m_len < sizeof (struct ip) && 303 (m = m_pullup(m, sizeof (struct ip))) == 0) { 304 ipstat.ips_toosmall++; 305 return; 306 } 307 ip = mtod(m, struct ip *); 308 309 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 310 ipstat.ips_badvers++; 311 goto bad; 312 } 313 314 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 315 if (hlen < sizeof(struct ip)) { /* minimum header length */ 316 ipstat.ips_badhlen++; 317 goto bad; 318 } 319 if (hlen > m->m_len) { 320 if ((m = m_pullup(m, hlen)) == 0) { 321 ipstat.ips_badhlen++; 322 return; 323 } 324 ip = mtod(m, struct ip *); 325 } 326 if (hlen == sizeof(struct ip)) { 327 sum = in_cksum_hdr(ip); 328 } else { 329 sum = in_cksum(m, hlen); 330 } 331 if (sum) { 332 ipstat.ips_badsum++; 333 goto bad; 334 } 335 336 /* 337 * Convert fields to host representation. 338 */ 339 NTOHS(ip->ip_len); 340 if (ip->ip_len < hlen) { 341 ipstat.ips_badlen++; 342 goto bad; 343 } 344 NTOHS(ip->ip_id); 345 NTOHS(ip->ip_off); 346 347 /* 348 * Check that the amount of data in the buffers 349 * is as at least much as the IP header would have us expect. 350 * Trim mbufs if longer than we expect. 351 * Drop packet if shorter than we expect. 352 */ 353 if (m->m_pkthdr.len < ip->ip_len) { 354tooshort: 355 ipstat.ips_tooshort++; 356 goto bad; 357 } 358 if (m->m_pkthdr.len > ip->ip_len) { 359 if (m->m_len == m->m_pkthdr.len) { 360 m->m_len = ip->ip_len; 361 m->m_pkthdr.len = ip->ip_len; 362 } else 363 m_adj(m, ip->ip_len - m->m_pkthdr.len); 364 } 365 /* 366 * IpHack's section. 367 * Right now when no processing on packet has done 368 * and it is still fresh out of network we do our black 369 * deals with it. 370 * - Firewall: deny/allow/divert 371 * - Xlate: translate packet's addr/port (NAT). 372 * - Pipe: pass pkt through dummynet. 373 * - Wrap: fake packet's addr/port <unimpl.> 374 * - Encapsulate: put it in another IP and send out. <unimp.> 375 */ 376 377#if defined(IPFIREWALL) && defined(DUMMYNET) 378iphack: 379#endif 380#if defined(IPFILTER) || defined(IPFILTER_LKM) 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#endif 393 if (ip_fw_chk_ptr) { 394#ifdef IPFIREWALL_FORWARD 395 /* 396 * If we've been forwarded from the output side, then 397 * skip the firewall a second time 398 */ 399 if (ip_fw_fwd_addr) 400 goto ours; 401#endif /* IPFIREWALL_FORWARD */ 402 /* 403 * See the comment in ip_output for the return values 404 * produced by the firewall. 405 */ 406 i = (*ip_fw_chk_ptr)(&ip, 407 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr); 408 if (m == NULL) /* Packet discarded by firewall */ 409 return; 410 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */ 411 goto pass; 412#ifdef DUMMYNET 413 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) { 414 /* Send packet to the appropriate pipe */ 415 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule, 416 0); 417 return; 418 } 419#endif 420#ifdef IPDIVERT 421 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 422 /* Divert or tee packet */ 423 divert_info = i; 424 goto ours; 425 } 426#endif 427#ifdef IPFIREWALL_FORWARD 428 if (i == 0 && ip_fw_fwd_addr != NULL) 429 goto pass; 430#endif 431 /* 432 * if we get here, the packet must be dropped 433 */ 434 m_freem(m); 435 return; 436 } 437pass: 438 439 /* 440 * Process options and, if not destined for us, 441 * ship it on. ip_dooptions returns 1 when an 442 * error was detected (causing an icmp message 443 * to be sent and the original packet to be freed). 444 */ 445 ip_nhops = 0; /* for source routed packets */ 446 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 447#ifdef IPFIREWALL_FORWARD 448 ip_fw_fwd_addr = NULL; 449#endif 450 return; 451 } 452 453 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 454 * matter if it is destined to another node, or whether it is 455 * a multicast one, RSVP wants it! and prevents it from being forwarded 456 * anywhere else. Also checks if the rsvp daemon is running before 457 * grabbing the packet. 458 */ 459 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 460 goto ours; 461 462 /* 463 * Check our list of addresses, to see if the packet is for us. 464 * If we don't have any addresses, assume any unicast packet 465 * we receive might be for us (and let the upper layers deal 466 * with it). 467 */ 468 if (TAILQ_EMPTY(&in_ifaddrhead) && 469 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 470 goto ours; 471 472 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 473 ia = TAILQ_NEXT(ia, ia_link)) { 474#define satosin(sa) ((struct sockaddr_in *)(sa)) 475 476#ifdef BOOTP_COMPAT 477 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 478 goto ours; 479#endif 480#ifdef IPFIREWALL_FORWARD 481 /* 482 * If the addr to forward to is one of ours, we pretend to 483 * be the destination for this packet. 484 */ 485 if (ip_fw_fwd_addr == NULL) { 486 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 487 goto ours; 488 } else if (IA_SIN(ia)->sin_addr.s_addr == 489 ip_fw_fwd_addr->sin_addr.s_addr) 490 goto ours; 491#else 492 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr) 493 goto ours; 494#endif 495 if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) { 496 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 497 ip->ip_dst.s_addr) 498 goto ours; 499 if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr) 500 goto ours; 501 } 502 } 503 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 504 struct in_multi *inm; 505 if (ip_mrouter) { 506 /* 507 * If we are acting as a multicast router, all 508 * incoming multicast packets are passed to the 509 * kernel-level multicast forwarding function. 510 * The packet is returned (relatively) intact; if 511 * ip_mforward() returns a non-zero value, the packet 512 * must be discarded, else it may be accepted below. 513 * 514 * (The IP ident field is put in the same byte order 515 * as expected when ip_mforward() is called from 516 * ip_output().) 517 */ 518 ip->ip_id = htons(ip->ip_id); 519 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) { 520 ipstat.ips_cantforward++; 521 m_freem(m); 522 return; 523 } 524 ip->ip_id = ntohs(ip->ip_id); 525 526 /* 527 * The process-level routing demon needs to receive 528 * all multicast IGMP packets, whether or not this 529 * host belongs to their destination groups. 530 */ 531 if (ip->ip_p == IPPROTO_IGMP) 532 goto ours; 533 ipstat.ips_forward++; 534 } 535 /* 536 * See if we belong to the destination multicast group on the 537 * arrival interface. 538 */ 539 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm); 540 if (inm == NULL) { 541 ipstat.ips_notmember++; 542 m_freem(m); 543 return; 544 } 545 goto ours; 546 } 547 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 548 goto ours; 549 if (ip->ip_dst.s_addr == INADDR_ANY) 550 goto ours; 551 552#if defined(NFAITH) && 0 < NFAITH 553 /* 554 * FAITH(Firewall Aided Internet Translator) 555 */ 556 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 557 if (ip_keepfaith) { 558 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 559 goto ours; 560 } 561 m_freem(m); 562 return; 563 } 564#endif 565 /* 566 * Not for us; forward if possible and desirable. 567 */ 568 if (ipforwarding == 0) { 569 ipstat.ips_cantforward++; 570 m_freem(m); 571 } else 572 ip_forward(m, 0); 573#ifdef IPFIREWALL_FORWARD 574 ip_fw_fwd_addr = NULL; 575#endif 576 return; 577 578ours: 579 580 /* 581 * If offset or IP_MF are set, must reassemble. 582 * Otherwise, nothing need be done. 583 * (We could look in the reassembly queue to see 584 * if the packet was previously fragmented, 585 * but it's not worth the time; just let them time out.) 586 */ 587 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) { 588 589#if 0 /* 590 * Reassembly should be able to treat a mbuf cluster, for later 591 * operation of contiguous protocol headers on the cluster. (KAME) 592 */ 593 if (m->m_flags & M_EXT) { /* XXX */ 594 if ((m = m_pullup(m, hlen)) == 0) { 595 ipstat.ips_toosmall++; 596#ifdef IPFIREWALL_FORWARD 597 ip_fw_fwd_addr = NULL; 598#endif 599 return; 600 } 601 ip = mtod(m, struct ip *); 602 } 603#endif 604 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 605 /* 606 * Look for queue of fragments 607 * of this datagram. 608 */ 609 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 610 if (ip->ip_id == fp->ipq_id && 611 ip->ip_src.s_addr == fp->ipq_src.s_addr && 612 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 613 ip->ip_p == fp->ipq_p) 614 goto found; 615 616 fp = 0; 617 618 /* check if there's a place for the new queue */ 619 if (nipq > maxnipq) { 620 /* 621 * drop something from the tail of the current queue 622 * before proceeding further 623 */ 624 if (ipq[sum].prev == &ipq[sum]) { /* gak */ 625 for (i = 0; i < IPREASS_NHASH; i++) { 626 if (ipq[i].prev != &ipq[i]) { 627 ip_freef(ipq[i].prev); 628 break; 629 } 630 } 631 } else 632 ip_freef(ipq[sum].prev); 633 } 634found: 635 /* 636 * Adjust ip_len to not reflect header, 637 * set ip_mff if more fragments are expected, 638 * convert offset of this to bytes. 639 */ 640 ip->ip_len -= hlen; 641 mff = (ip->ip_off & IP_MF) != 0; 642 if (mff) { 643 /* 644 * Make sure that fragments have a data length 645 * that's a non-zero multiple of 8 bytes. 646 */ 647 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 648 ipstat.ips_toosmall++; /* XXX */ 649 goto bad; 650 } 651 m->m_flags |= M_FRAG; 652 } 653 ip->ip_off <<= 3; 654 655 /* 656 * If datagram marked as having more fragments 657 * or if this is not the first fragment, 658 * attempt reassembly; if it succeeds, proceed. 659 */ 660 if (mff || ip->ip_off) { 661 ipstat.ips_fragments++; 662 m->m_pkthdr.header = ip; 663#ifdef IPDIVERT 664 m = ip_reass(m, 665 fp, &ipq[sum], &divert_info, &divert_cookie); 666#else 667 m = ip_reass(m, fp, &ipq[sum]); 668#endif 669 if (m == 0) { 670#ifdef IPFIREWALL_FORWARD 671 ip_fw_fwd_addr = NULL; 672#endif 673 return; 674 } 675 /* Get the length of the reassembled packets header */ 676 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 677 ipstat.ips_reassembled++; 678 ip = mtod(m, struct ip *); 679#ifdef IPDIVERT 680 /* Restore original checksum before diverting packet */ 681 if (divert_info != 0) { 682 ip->ip_len += hlen; 683 HTONS(ip->ip_len); 684 HTONS(ip->ip_off); 685 HTONS(ip->ip_id); 686 ip->ip_sum = 0; 687 ip->ip_sum = in_cksum_hdr(ip); 688 NTOHS(ip->ip_id); 689 NTOHS(ip->ip_off); 690 NTOHS(ip->ip_len); 691 ip->ip_len -= hlen; 692 } 693#endif 694 } else 695 if (fp) 696 ip_freef(fp); 697 } else 698 ip->ip_len -= hlen; 699 700#ifdef IPDIVERT 701 /* 702 * Divert or tee packet to the divert protocol if required. 703 * 704 * If divert_info is zero then cookie should be too, so we shouldn't 705 * need to clear them here. Assume divert_packet() does so also. 706 */ 707 if (divert_info != 0) { 708 struct mbuf *clone = NULL; 709 710 /* Clone packet if we're doing a 'tee' */ 711 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0) 712 clone = m_dup(m, M_DONTWAIT); 713 714 /* Restore packet header fields to original values */ 715 ip->ip_len += hlen; 716 HTONS(ip->ip_len); 717 HTONS(ip->ip_off); 718 HTONS(ip->ip_id); 719 720 /* Deliver packet to divert input routine */ 721 ip_divert_cookie = divert_cookie; 722 divert_packet(m, 1, divert_info & 0xffff); 723 ipstat.ips_delivered++; 724 725 /* If 'tee', continue with original packet */ 726 if (clone == NULL) 727 return; 728 m = clone; 729 ip = mtod(m, struct ip *); 730 } 731#endif 732 733 /* 734 * Switch out to protocol's input routine. 735 */ 736 ipstat.ips_delivered++; 737 { 738 int off = hlen, nh = ip->ip_p; 739 740 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh); 741#ifdef IPFIREWALL_FORWARD 742 ip_fw_fwd_addr = NULL; /* tcp needed it */ 743#endif 744 return; 745 } 746bad: 747#ifdef IPFIREWALL_FORWARD 748 ip_fw_fwd_addr = NULL; 749#endif 750 m_freem(m); 751} 752 753/* 754 * IP software interrupt routine - to go away sometime soon 755 */ 756static void 757ipintr(void) 758{ 759 int s; 760 struct mbuf *m; 761 762 while(1) { 763 s = splimp(); 764 IF_DEQUEUE(&ipintrq, m); 765 splx(s); 766 if (m == 0) 767 return; 768 ip_input(m); 769 } 770} 771 772NETISR_SET(NETISR_IP, ipintr); 773 774/* 775 * Take incoming datagram fragment and try to reassemble it into 776 * whole datagram. If a chain for reassembly of this datagram already 777 * exists, then it is given as fp; otherwise have to make a chain. 778 * 779 * When IPDIVERT enabled, keep additional state with each packet that 780 * tells us if we need to divert or tee the packet we're building. 781 */ 782 783static struct mbuf * 784#ifdef IPDIVERT 785ip_reass(m, fp, where, divinfo, divcookie) 786#else 787ip_reass(m, fp, where) 788#endif 789 register struct mbuf *m; 790 register struct ipq *fp; 791 struct ipq *where; 792#ifdef IPDIVERT 793 u_int32_t *divinfo; 794 u_int16_t *divcookie; 795#endif 796{ 797 struct ip *ip = mtod(m, struct ip *); 798 register struct mbuf *p = 0, *q, *nq; 799 struct mbuf *t; 800 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 801 int i, next; 802 803 /* 804 * Presence of header sizes in mbufs 805 * would confuse code below. 806 */ 807 m->m_data += hlen; 808 m->m_len -= hlen; 809 810 /* 811 * If first fragment to arrive, create a reassembly queue. 812 */ 813 if (fp == 0) { 814 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 815 goto dropfrag; 816 fp = mtod(t, struct ipq *); 817 insque(fp, where); 818 nipq++; 819 fp->ipq_ttl = IPFRAGTTL; 820 fp->ipq_p = ip->ip_p; 821 fp->ipq_id = ip->ip_id; 822 fp->ipq_src = ip->ip_src; 823 fp->ipq_dst = ip->ip_dst; 824 fp->ipq_frags = m; 825 m->m_nextpkt = NULL; 826#ifdef IPDIVERT 827 fp->ipq_div_info = 0; 828 fp->ipq_div_cookie = 0; 829#endif 830 goto inserted; 831 } 832 833#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 834 835 /* 836 * Find a segment which begins after this one does. 837 */ 838 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 839 if (GETIP(q)->ip_off > ip->ip_off) 840 break; 841 842 /* 843 * If there is a preceding segment, it may provide some of 844 * our data already. If so, drop the data from the incoming 845 * segment. If it provides all of our data, drop us, otherwise 846 * stick new segment in the proper place. 847 */ 848 if (p) { 849 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 850 if (i > 0) { 851 if (i >= ip->ip_len) 852 goto dropfrag; 853 m_adj(m, i); 854 ip->ip_off += i; 855 ip->ip_len -= i; 856 } 857 m->m_nextpkt = p->m_nextpkt; 858 p->m_nextpkt = m; 859 } else { 860 m->m_nextpkt = fp->ipq_frags; 861 fp->ipq_frags = m; 862 } 863 864 /* 865 * While we overlap succeeding segments trim them or, 866 * if they are completely covered, dequeue them. 867 */ 868 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 869 q = nq) { 870 i = (ip->ip_off + ip->ip_len) - 871 GETIP(q)->ip_off; 872 if (i < GETIP(q)->ip_len) { 873 GETIP(q)->ip_len -= i; 874 GETIP(q)->ip_off += i; 875 m_adj(q, i); 876 break; 877 } 878 nq = q->m_nextpkt; 879 m->m_nextpkt = nq; 880 m_freem(q); 881 } 882 883inserted: 884 885#ifdef IPDIVERT 886 /* 887 * Transfer firewall instructions to the fragment structure. 888 * Any fragment diverting causes the whole packet to divert. 889 */ 890 fp->ipq_div_info = *divinfo; 891 fp->ipq_div_cookie = *divcookie; 892 *divinfo = 0; 893 *divcookie = 0; 894#endif 895 896 /* 897 * Check for complete reassembly. 898 */ 899 next = 0; 900 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 901 if (GETIP(q)->ip_off != next) 902 return (0); 903 next += GETIP(q)->ip_len; 904 } 905 /* Make sure the last packet didn't have the IP_MF flag */ 906 if (p->m_flags & M_FRAG) 907 return (0); 908 909 /* 910 * Reassembly is complete. Make sure the packet is a sane size. 911 */ 912 q = fp->ipq_frags; 913 ip = GETIP(q); 914 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 915 ipstat.ips_toolong++; 916 ip_freef(fp); 917 return (0); 918 } 919 920 /* 921 * Concatenate fragments. 922 */ 923 m = q; 924 t = m->m_next; 925 m->m_next = 0; 926 m_cat(m, t); 927 nq = q->m_nextpkt; 928 q->m_nextpkt = 0; 929 for (q = nq; q != NULL; q = nq) { 930 nq = q->m_nextpkt; 931 q->m_nextpkt = NULL; 932 m_cat(m, q); 933 } 934 935#ifdef IPDIVERT 936 /* 937 * Extract firewall instructions from the fragment structure. 938 */ 939 *divinfo = fp->ipq_div_info; 940 *divcookie = fp->ipq_div_cookie; 941#endif 942 943 /* 944 * Create header for new ip packet by 945 * modifying header of first packet; 946 * dequeue and discard fragment reassembly header. 947 * Make header visible. 948 */ 949 ip->ip_len = next; 950 ip->ip_src = fp->ipq_src; 951 ip->ip_dst = fp->ipq_dst; 952 remque(fp); 953 nipq--; 954 (void) m_free(dtom(fp)); 955 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 956 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 957 /* some debugging cruft by sklower, below, will go away soon */ 958 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 959 register int plen = 0; 960 for (t = m; t; t = t->m_next) 961 plen += t->m_len; 962 m->m_pkthdr.len = plen; 963 } 964 return (m); 965 966dropfrag: 967#ifdef IPDIVERT 968 *divinfo = 0; 969 *divcookie = 0; 970#endif 971 ipstat.ips_fragdropped++; 972 m_freem(m); 973 return (0); 974 975#undef GETIP 976} 977 978/* 979 * Free a fragment reassembly header and all 980 * associated datagrams. 981 */ 982static void 983ip_freef(fp) 984 struct ipq *fp; 985{ 986 register struct mbuf *q; 987 988 while (fp->ipq_frags) { 989 q = fp->ipq_frags; 990 fp->ipq_frags = q->m_nextpkt; 991 m_freem(q); 992 } 993 remque(fp); 994 (void) m_free(dtom(fp)); 995 nipq--; 996} 997 998/* 999 * IP timer processing; 1000 * if a timer expires on a reassembly 1001 * queue, discard it. 1002 */ 1003void 1004ip_slowtimo() 1005{ 1006 register struct ipq *fp; 1007 int s = splnet(); 1008 int i; 1009 1010 for (i = 0; i < IPREASS_NHASH; i++) { 1011 fp = ipq[i].next; 1012 if (fp == 0) 1013 continue; 1014 while (fp != &ipq[i]) { 1015 --fp->ipq_ttl; 1016 fp = fp->next; 1017 if (fp->prev->ipq_ttl == 0) { 1018 ipstat.ips_fragtimeout++; 1019 ip_freef(fp->prev); 1020 } 1021 } 1022 } 1023 ipflow_slowtimo(); 1024 splx(s); 1025} 1026 1027/* 1028 * Drain off all datagram fragments. 1029 */ 1030void 1031ip_drain() 1032{ 1033 int i; 1034 1035 for (i = 0; i < IPREASS_NHASH; i++) { 1036 while (ipq[i].next != &ipq[i]) { 1037 ipstat.ips_fragdropped++; 1038 ip_freef(ipq[i].next); 1039 } 1040 } 1041 in_rtqdrain(); 1042} 1043 1044/* 1045 * Do option processing on a datagram, 1046 * possibly discarding it if bad options are encountered, 1047 * or forwarding it if source-routed. 1048 * Returns 1 if packet has been forwarded/freed, 1049 * 0 if the packet should be processed further. 1050 */ 1051static int 1052ip_dooptions(m) 1053 struct mbuf *m; 1054{ 1055 register struct ip *ip = mtod(m, struct ip *); 1056 register u_char *cp; 1057 register struct ip_timestamp *ipt; 1058 register struct in_ifaddr *ia; 1059 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1060 struct in_addr *sin, dst; 1061 n_time ntime; 1062 1063 dst = ip->ip_dst; 1064 cp = (u_char *)(ip + 1); 1065 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1066 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1067 opt = cp[IPOPT_OPTVAL]; 1068 if (opt == IPOPT_EOL) 1069 break; 1070 if (opt == IPOPT_NOP) 1071 optlen = 1; 1072 else { 1073 optlen = cp[IPOPT_OLEN]; 1074 if (optlen <= 0 || optlen > cnt) { 1075 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1076 goto bad; 1077 } 1078 } 1079 switch (opt) { 1080 1081 default: 1082 break; 1083 1084 /* 1085 * Source routing with record. 1086 * Find interface with current destination address. 1087 * If none on this machine then drop if strictly routed, 1088 * or do nothing if loosely routed. 1089 * Record interface address and bring up next address 1090 * component. If strictly routed make sure next 1091 * address is on directly accessible net. 1092 */ 1093 case IPOPT_LSRR: 1094 case IPOPT_SSRR: 1095 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1096 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1097 goto bad; 1098 } 1099 ipaddr.sin_addr = ip->ip_dst; 1100 ia = (struct in_ifaddr *) 1101 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1102 if (ia == 0) { 1103 if (opt == IPOPT_SSRR) { 1104 type = ICMP_UNREACH; 1105 code = ICMP_UNREACH_SRCFAIL; 1106 goto bad; 1107 } 1108 if (!ip_dosourceroute) 1109 goto nosourcerouting; 1110 /* 1111 * Loose routing, and not at next destination 1112 * yet; nothing to do except forward. 1113 */ 1114 break; 1115 } 1116 off--; /* 0 origin */ 1117 if (off > optlen - sizeof(struct in_addr)) { 1118 /* 1119 * End of source route. Should be for us. 1120 */ 1121 if (!ip_acceptsourceroute) 1122 goto nosourcerouting; 1123 save_rte(cp, ip->ip_src); 1124 break; 1125 } 1126 1127 if (!ip_dosourceroute) { 1128 if (ipforwarding) { 1129 char buf[16]; /* aaa.bbb.ccc.ddd\0 */ 1130 /* 1131 * Acting as a router, so generate ICMP 1132 */ 1133nosourcerouting: 1134 strcpy(buf, inet_ntoa(ip->ip_dst)); 1135 log(LOG_WARNING, 1136 "attempted source route from %s to %s\n", 1137 inet_ntoa(ip->ip_src), buf); 1138 type = ICMP_UNREACH; 1139 code = ICMP_UNREACH_SRCFAIL; 1140 goto bad; 1141 } else { 1142 /* 1143 * Not acting as a router, so silently drop. 1144 */ 1145 ipstat.ips_cantforward++; 1146 m_freem(m); 1147 return (1); 1148 } 1149 } 1150 1151 /* 1152 * locate outgoing interface 1153 */ 1154 (void)memcpy(&ipaddr.sin_addr, cp + off, 1155 sizeof(ipaddr.sin_addr)); 1156 1157 if (opt == IPOPT_SSRR) { 1158#define INA struct in_ifaddr * 1159#define SA struct sockaddr * 1160 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0) 1161 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1162 } else 1163 ia = ip_rtaddr(ipaddr.sin_addr); 1164 if (ia == 0) { 1165 type = ICMP_UNREACH; 1166 code = ICMP_UNREACH_SRCFAIL; 1167 goto bad; 1168 } 1169 ip->ip_dst = ipaddr.sin_addr; 1170 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1171 sizeof(struct in_addr)); 1172 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1173 /* 1174 * Let ip_intr's mcast routing check handle mcast pkts 1175 */ 1176 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1177 break; 1178 1179 case IPOPT_RR: 1180 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1181 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1182 goto bad; 1183 } 1184 /* 1185 * If no space remains, ignore. 1186 */ 1187 off--; /* 0 origin */ 1188 if (off > optlen - sizeof(struct in_addr)) 1189 break; 1190 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1191 sizeof(ipaddr.sin_addr)); 1192 /* 1193 * locate outgoing interface; if we're the destination, 1194 * use the incoming interface (should be same). 1195 */ 1196 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 && 1197 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 1198 type = ICMP_UNREACH; 1199 code = ICMP_UNREACH_HOST; 1200 goto bad; 1201 } 1202 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1203 sizeof(struct in_addr)); 1204 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1205 break; 1206 1207 case IPOPT_TS: 1208 code = cp - (u_char *)ip; 1209 ipt = (struct ip_timestamp *)cp; 1210 if (ipt->ipt_len < 5) 1211 goto bad; 1212 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1213 if (++ipt->ipt_oflw == 0) 1214 goto bad; 1215 break; 1216 } 1217 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1); 1218 switch (ipt->ipt_flg) { 1219 1220 case IPOPT_TS_TSONLY: 1221 break; 1222 1223 case IPOPT_TS_TSANDADDR: 1224 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1225 sizeof(struct in_addr) > ipt->ipt_len) 1226 goto bad; 1227 ipaddr.sin_addr = dst; 1228 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1229 m->m_pkthdr.rcvif); 1230 if (ia == 0) 1231 continue; 1232 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 1233 sizeof(struct in_addr)); 1234 ipt->ipt_ptr += sizeof(struct in_addr); 1235 break; 1236 1237 case IPOPT_TS_PRESPEC: 1238 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1239 sizeof(struct in_addr) > ipt->ipt_len) 1240 goto bad; 1241 (void)memcpy(&ipaddr.sin_addr, sin, 1242 sizeof(struct in_addr)); 1243 if (ifa_ifwithaddr((SA)&ipaddr) == 0) 1244 continue; 1245 ipt->ipt_ptr += sizeof(struct in_addr); 1246 break; 1247 1248 default: 1249 goto bad; 1250 } 1251 ntime = iptime(); 1252 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 1253 sizeof(n_time)); 1254 ipt->ipt_ptr += sizeof(n_time); 1255 } 1256 } 1257 if (forward && ipforwarding) { 1258 ip_forward(m, 1); 1259 return (1); 1260 } 1261 return (0); 1262bad: 1263 ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2; /* XXX icmp_error adds in hdr length */ 1264 icmp_error(m, type, code, 0, 0); 1265 ipstat.ips_badoptions++; 1266 return (1); 1267} 1268 1269/* 1270 * Given address of next destination (final or next hop), 1271 * return internet address info of interface to be used to get there. 1272 */ 1273static struct in_ifaddr * 1274ip_rtaddr(dst) 1275 struct in_addr dst; 1276{ 1277 register struct sockaddr_in *sin; 1278 1279 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst; 1280 1281 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) { 1282 if (ipforward_rt.ro_rt) { 1283 RTFREE(ipforward_rt.ro_rt); 1284 ipforward_rt.ro_rt = 0; 1285 } 1286 sin->sin_family = AF_INET; 1287 sin->sin_len = sizeof(*sin); 1288 sin->sin_addr = dst; 1289 1290 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1291 } 1292 if (ipforward_rt.ro_rt == 0) 1293 return ((struct in_ifaddr *)0); 1294 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa); 1295} 1296 1297/* 1298 * Save incoming source route for use in replies, 1299 * to be picked up later by ip_srcroute if the receiver is interested. 1300 */ 1301void 1302save_rte(option, dst) 1303 u_char *option; 1304 struct in_addr dst; 1305{ 1306 unsigned olen; 1307 1308 olen = option[IPOPT_OLEN]; 1309#ifdef DIAGNOSTIC 1310 if (ipprintfs) 1311 printf("save_rte: olen %d\n", olen); 1312#endif 1313 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1314 return; 1315 bcopy(option, ip_srcrt.srcopt, olen); 1316 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1317 ip_srcrt.dst = dst; 1318} 1319 1320/* 1321 * Retrieve incoming source route for use in replies, 1322 * in the same form used by setsockopt. 1323 * The first hop is placed before the options, will be removed later. 1324 */ 1325struct mbuf * 1326ip_srcroute() 1327{ 1328 register struct in_addr *p, *q; 1329 register struct mbuf *m; 1330 1331 if (ip_nhops == 0) 1332 return ((struct mbuf *)0); 1333 m = m_get(M_DONTWAIT, MT_HEADER); 1334 if (m == 0) 1335 return ((struct mbuf *)0); 1336 1337#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1338 1339 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1340 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1341 OPTSIZ; 1342#ifdef DIAGNOSTIC 1343 if (ipprintfs) 1344 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 1345#endif 1346 1347 /* 1348 * First save first hop for return route 1349 */ 1350 p = &ip_srcrt.route[ip_nhops - 1]; 1351 *(mtod(m, struct in_addr *)) = *p--; 1352#ifdef DIAGNOSTIC 1353 if (ipprintfs) 1354 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr)); 1355#endif 1356 1357 /* 1358 * Copy option fields and padding (nop) to mbuf. 1359 */ 1360 ip_srcrt.nop = IPOPT_NOP; 1361 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1362 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 1363 &ip_srcrt.nop, OPTSIZ); 1364 q = (struct in_addr *)(mtod(m, caddr_t) + 1365 sizeof(struct in_addr) + OPTSIZ); 1366#undef OPTSIZ 1367 /* 1368 * Record return path as an IP source route, 1369 * reversing the path (pointers are now aligned). 1370 */ 1371 while (p >= ip_srcrt.route) { 1372#ifdef DIAGNOSTIC 1373 if (ipprintfs) 1374 printf(" %lx", (u_long)ntohl(q->s_addr)); 1375#endif 1376 *q++ = *p--; 1377 } 1378 /* 1379 * Last hop goes to final destination. 1380 */ 1381 *q = ip_srcrt.dst; 1382#ifdef DIAGNOSTIC 1383 if (ipprintfs) 1384 printf(" %lx\n", (u_long)ntohl(q->s_addr)); 1385#endif 1386 return (m); 1387} 1388 1389/* 1390 * Strip out IP options, at higher 1391 * level protocol in the kernel. 1392 * Second argument is buffer to which options 1393 * will be moved, and return value is their length. 1394 * XXX should be deleted; last arg currently ignored. 1395 */ 1396void 1397ip_stripoptions(m, mopt) 1398 register struct mbuf *m; 1399 struct mbuf *mopt; 1400{ 1401 register int i; 1402 struct ip *ip = mtod(m, struct ip *); 1403 register caddr_t opts; 1404 int olen; 1405 1406 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1407 opts = (caddr_t)(ip + 1); 1408 i = m->m_len - (sizeof (struct ip) + olen); 1409 bcopy(opts + olen, opts, (unsigned)i); 1410 m->m_len -= olen; 1411 if (m->m_flags & M_PKTHDR) 1412 m->m_pkthdr.len -= olen; 1413 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1414} 1415 1416u_char inetctlerrmap[PRC_NCMDS] = { 1417 0, 0, 0, 0, 1418 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1419 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1420 EMSGSIZE, EHOSTUNREACH, 0, 0, 1421 0, 0, 0, 0, 1422 ENOPROTOOPT 1423}; 1424 1425/* 1426 * Forward a packet. If some error occurs return the sender 1427 * an icmp packet. Note we can't always generate a meaningful 1428 * icmp message because icmp doesn't have a large enough repertoire 1429 * of codes and types. 1430 * 1431 * If not forwarding, just drop the packet. This could be confusing 1432 * if ipforwarding was zero but some routing protocol was advancing 1433 * us as a gateway to somewhere. However, we must let the routing 1434 * protocol deal with that. 1435 * 1436 * The srcrt parameter indicates whether the packet is being forwarded 1437 * via a source route. 1438 */ 1439static void 1440ip_forward(m, srcrt) 1441 struct mbuf *m; 1442 int srcrt; 1443{ 1444 register struct ip *ip = mtod(m, struct ip *); 1445 register struct sockaddr_in *sin; 1446 register struct rtentry *rt; 1447 int error, type = 0, code = 0; 1448 struct mbuf *mcopy; 1449 n_long dest; 1450 struct ifnet *destifp; 1451#ifdef IPSEC 1452 struct ifnet dummyifp; 1453#endif 1454 1455 dest = 0; 1456#ifdef DIAGNOSTIC 1457 if (ipprintfs) 1458 printf("forward: src %lx dst %lx ttl %x\n", 1459 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr, 1460 ip->ip_ttl); 1461#endif 1462 1463 1464 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1465 ipstat.ips_cantforward++; 1466 m_freem(m); 1467 return; 1468 } 1469 HTONS(ip->ip_id); 1470#ifdef IPSTEALTH 1471 if (!ipstealth) { 1472#endif 1473 if (ip->ip_ttl <= IPTTLDEC) { 1474 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 1475 dest, 0); 1476 return; 1477 } 1478 ip->ip_ttl -= IPTTLDEC; 1479#ifdef IPSTEALTH 1480 } 1481#endif 1482 1483 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst; 1484 if ((rt = ipforward_rt.ro_rt) == 0 || 1485 ip->ip_dst.s_addr != sin->sin_addr.s_addr) { 1486 if (ipforward_rt.ro_rt) { 1487 RTFREE(ipforward_rt.ro_rt); 1488 ipforward_rt.ro_rt = 0; 1489 } 1490 sin->sin_family = AF_INET; 1491 sin->sin_len = sizeof(*sin); 1492 sin->sin_addr = ip->ip_dst; 1493 1494 rtalloc_ign(&ipforward_rt, RTF_PRCLONING); 1495 if (ipforward_rt.ro_rt == 0) { 1496 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1497 return; 1498 } 1499 rt = ipforward_rt.ro_rt; 1500 } 1501 1502 /* 1503 * Save at most 64 bytes of the packet in case 1504 * we need to generate an ICMP message to the src. 1505 */ 1506 mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64)); 1507 1508 /* 1509 * If forwarding packet using same interface that it came in on, 1510 * perhaps should send a redirect to sender to shortcut a hop. 1511 * Only send redirect if source is sending directly to us, 1512 * and if packet was not source routed (or has any options). 1513 * Also, don't send redirect if forwarding using a default route 1514 * or a route modified by a redirect. 1515 */ 1516#define satosin(sa) ((struct sockaddr_in *)(sa)) 1517 if (rt->rt_ifp == m->m_pkthdr.rcvif && 1518 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1519 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1520 ipsendredirects && !srcrt) { 1521#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1522 u_long src = ntohl(ip->ip_src.s_addr); 1523 1524 if (RTA(rt) && 1525 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1526 if (rt->rt_flags & RTF_GATEWAY) 1527 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1528 else 1529 dest = ip->ip_dst.s_addr; 1530 /* Router requirements says to only send host redirects */ 1531 type = ICMP_REDIRECT; 1532 code = ICMP_REDIRECT_HOST; 1533#ifdef DIAGNOSTIC 1534 if (ipprintfs) 1535 printf("redirect (%d) to %lx\n", code, (u_long)dest); 1536#endif 1537 } 1538 } 1539 1540 error = ip_output(m, (struct mbuf *)0, &ipforward_rt, 1541 IP_FORWARDING, 0); 1542 if (error) 1543 ipstat.ips_cantforward++; 1544 else { 1545 ipstat.ips_forward++; 1546 if (type) 1547 ipstat.ips_redirectsent++; 1548 else { 1549 if (mcopy) { 1550 ipflow_create(&ipforward_rt, mcopy); 1551 m_freem(mcopy); 1552 } 1553 return; 1554 } 1555 } 1556 if (mcopy == NULL) 1557 return; 1558 destifp = NULL; 1559 1560 switch (error) { 1561 1562 case 0: /* forwarded, but need redirect */ 1563 /* type, code set above */ 1564 break; 1565 1566 case ENETUNREACH: /* shouldn't happen, checked above */ 1567 case EHOSTUNREACH: 1568 case ENETDOWN: 1569 case EHOSTDOWN: 1570 default: 1571 type = ICMP_UNREACH; 1572 code = ICMP_UNREACH_HOST; 1573 break; 1574 1575 case EMSGSIZE: 1576 type = ICMP_UNREACH; 1577 code = ICMP_UNREACH_NEEDFRAG; 1578#ifndef IPSEC 1579 if (ipforward_rt.ro_rt) 1580 destifp = ipforward_rt.ro_rt->rt_ifp; 1581#else 1582 /* 1583 * If the packet is routed over IPsec tunnel, tell the 1584 * originator the tunnel MTU. 1585 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 1586 * XXX quickhack!!! 1587 */ 1588 if (ipforward_rt.ro_rt) { 1589 struct secpolicy *sp = NULL; 1590 int ipsecerror; 1591 int ipsechdr; 1592 struct route *ro; 1593 1594 sp = ipsec4_getpolicybyaddr(mcopy, 1595 IPSEC_DIR_OUTBOUND, 1596 IP_FORWARDING, 1597 &ipsecerror); 1598 1599 if (sp == NULL) 1600 destifp = ipforward_rt.ro_rt->rt_ifp; 1601 else { 1602 /* count IPsec header size */ 1603 ipsechdr = ipsec4_hdrsiz(mcopy, 1604 IPSEC_DIR_OUTBOUND, 1605 NULL); 1606 1607 /* 1608 * find the correct route for outer IPv4 1609 * header, compute tunnel MTU. 1610 * 1611 * XXX BUG ALERT 1612 * The "dummyifp" code relies upon the fact 1613 * that icmp_error() touches only ifp->if_mtu. 1614 */ 1615 /*XXX*/ 1616 destifp = NULL; 1617 if (sp->req != NULL 1618 && sp->req->sav != NULL 1619 && sp->req->sav->sah != NULL) { 1620 ro = &sp->req->sav->sah->sa_route; 1621 if (ro->ro_rt && ro->ro_rt->rt_ifp) { 1622 dummyifp.if_mtu = 1623 ro->ro_rt->rt_ifp->if_mtu; 1624 dummyifp.if_mtu -= ipsechdr; 1625 destifp = &dummyifp; 1626 } 1627 } 1628 1629 key_freesp(sp); 1630 } 1631 } 1632#endif /*IPSEC*/ 1633 ipstat.ips_cantfrag++; 1634 break; 1635 1636 case ENOBUFS: 1637 type = ICMP_SOURCEQUENCH; 1638 code = 0; 1639 break; 1640 } 1641 icmp_error(mcopy, type, code, dest, destifp); 1642} 1643 1644void 1645ip_savecontrol(inp, mp, ip, m) 1646 register struct inpcb *inp; 1647 register struct mbuf **mp; 1648 register struct ip *ip; 1649 register struct mbuf *m; 1650{ 1651 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1652 struct timeval tv; 1653 1654 microtime(&tv); 1655 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1656 SCM_TIMESTAMP, SOL_SOCKET); 1657 if (*mp) 1658 mp = &(*mp)->m_next; 1659 } 1660 if (inp->inp_flags & INP_RECVDSTADDR) { 1661 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1662 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1663 if (*mp) 1664 mp = &(*mp)->m_next; 1665 } 1666#ifdef notyet 1667 /* XXX 1668 * Moving these out of udp_input() made them even more broken 1669 * than they already were. 1670 */ 1671 /* options were tossed already */ 1672 if (inp->inp_flags & INP_RECVOPTS) { 1673 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1674 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1675 if (*mp) 1676 mp = &(*mp)->m_next; 1677 } 1678 /* ip_srcroute doesn't do what we want here, need to fix */ 1679 if (inp->inp_flags & INP_RECVRETOPTS) { 1680 *mp = sbcreatecontrol((caddr_t) ip_srcroute(), 1681 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1682 if (*mp) 1683 mp = &(*mp)->m_next; 1684 } 1685#endif 1686 if (inp->inp_flags & INP_RECVIF) { 1687 struct ifnet *ifp; 1688 struct sdlbuf { 1689 struct sockaddr_dl sdl; 1690 u_char pad[32]; 1691 } sdlbuf; 1692 struct sockaddr_dl *sdp; 1693 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1694 1695 if (((ifp = m->m_pkthdr.rcvif)) 1696 && ( ifp->if_index && (ifp->if_index <= if_index))) { 1697 sdp = (struct sockaddr_dl *)(ifnet_addrs 1698 [ifp->if_index - 1]->ifa_addr); 1699 /* 1700 * Change our mind and don't try copy. 1701 */ 1702 if ((sdp->sdl_family != AF_LINK) 1703 || (sdp->sdl_len > sizeof(sdlbuf))) { 1704 goto makedummy; 1705 } 1706 bcopy(sdp, sdl2, sdp->sdl_len); 1707 } else { 1708makedummy: 1709 sdl2->sdl_len 1710 = offsetof(struct sockaddr_dl, sdl_data[0]); 1711 sdl2->sdl_family = AF_LINK; 1712 sdl2->sdl_index = 0; 1713 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1714 } 1715 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 1716 IP_RECVIF, IPPROTO_IP); 1717 if (*mp) 1718 mp = &(*mp)->m_next; 1719 } 1720} 1721 1722int 1723ip_rsvp_init(struct socket *so) 1724{ 1725 if (so->so_type != SOCK_RAW || 1726 so->so_proto->pr_protocol != IPPROTO_RSVP) 1727 return EOPNOTSUPP; 1728 1729 if (ip_rsvpd != NULL) 1730 return EADDRINUSE; 1731 1732 ip_rsvpd = so; 1733 /* 1734 * This may seem silly, but we need to be sure we don't over-increment 1735 * the RSVP counter, in case something slips up. 1736 */ 1737 if (!ip_rsvp_on) { 1738 ip_rsvp_on = 1; 1739 rsvp_on++; 1740 } 1741 1742 return 0; 1743} 1744 1745int 1746ip_rsvp_done(void) 1747{ 1748 ip_rsvpd = NULL; 1749 /* 1750 * This may seem silly, but we need to be sure we don't over-decrement 1751 * the RSVP counter, in case something slips up. 1752 */ 1753 if (ip_rsvp_on) { 1754 ip_rsvp_on = 0; 1755 rsvp_on--; 1756 } 1757 return 0; 1758} 1759