ip_input.c revision 1.364
1/* $NetBSD: ip_input.c,v 1.364 2018/01/01 00:51:36 christos Exp $ */ 2 3/* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32/*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 59 * POSSIBILITY OF SUCH DAMAGE. 60 */ 61 62/* 63 * Copyright (c) 1982, 1986, 1988, 1993 64 * The Regents of the University of California. All rights reserved. 65 * 66 * Redistribution and use in source and binary forms, with or without 67 * modification, are permitted provided that the following conditions 68 * are met: 69 * 1. Redistributions of source code must retain the above copyright 70 * notice, this list of conditions and the following disclaimer. 71 * 2. Redistributions in binary form must reproduce the above copyright 72 * notice, this list of conditions and the following disclaimer in the 73 * documentation and/or other materials provided with the distribution. 74 * 3. Neither the name of the University nor the names of its contributors 75 * may be used to endorse or promote products derived from this software 76 * without specific prior written permission. 77 * 78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 88 * SUCH DAMAGE. 89 * 90 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 91 */ 92 93#include <sys/cdefs.h> 94__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.364 2018/01/01 00:51:36 christos Exp $"); 95 96#ifdef _KERNEL_OPT 97#include "opt_inet.h" 98#include "opt_gateway.h" 99#include "opt_ipsec.h" 100#include "opt_mrouting.h" 101#include "opt_mbuftrace.h" 102#include "opt_inet_csum.h" 103#include "opt_net_mpsafe.h" 104#endif 105 106#include "arp.h" 107 108#include <sys/param.h> 109#include <sys/systm.h> 110#include <sys/cpu.h> 111#include <sys/mbuf.h> 112#include <sys/domain.h> 113#include <sys/protosw.h> 114#include <sys/socket.h> 115#include <sys/socketvar.h> 116#include <sys/errno.h> 117#include <sys/time.h> 118#include <sys/kernel.h> 119#include <sys/pool.h> 120#include <sys/sysctl.h> 121#include <sys/kauth.h> 122 123#include <net/if.h> 124#include <net/if_dl.h> 125#include <net/route.h> 126#include <net/pktqueue.h> 127#include <net/pfil.h> 128 129#include <netinet/in.h> 130#include <netinet/in_systm.h> 131#include <netinet/ip.h> 132#include <netinet/in_pcb.h> 133#include <netinet/in_proto.h> 134#include <netinet/in_var.h> 135#include <netinet/ip_var.h> 136#include <netinet/ip_private.h> 137#include <netinet/ip_icmp.h> 138/* just for gif_ttl */ 139#include <netinet/in_gif.h> 140#include "gif.h" 141#include <net/if_gre.h> 142#include "gre.h" 143 144#ifdef MROUTING 145#include <netinet/ip_mroute.h> 146#endif 147#include <netinet/portalgo.h> 148 149#ifdef IPSEC 150#include <netipsec/ipsec.h> 151#endif 152 153#ifndef IPFORWARDING 154#ifdef GATEWAY 155#define IPFORWARDING 1 /* forward IP packets not for us */ 156#else /* GATEWAY */ 157#define IPFORWARDING 0 /* don't forward IP packets not for us */ 158#endif /* GATEWAY */ 159#endif /* IPFORWARDING */ 160#ifndef IPSENDREDIRECTS 161#define IPSENDREDIRECTS 1 162#endif 163#ifndef IPFORWSRCRT 164#define IPFORWSRCRT 1 /* forward source-routed packets */ 165#endif 166#ifndef IPALLOWSRCRT 167#define IPALLOWSRCRT 1 /* allow source-routed packets */ 168#endif 169#ifndef IPMTUDISC 170#define IPMTUDISC 1 171#endif 172#ifndef IPMTUDISCTIMEOUT 173#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 174#endif 175 176/* 177 * Note: DIRECTED_BROADCAST is handled this way so that previous 178 * configuration using this option will Just Work. 179 */ 180#ifndef IPDIRECTEDBCAST 181#ifdef DIRECTED_BROADCAST 182#define IPDIRECTEDBCAST 1 183#else 184#define IPDIRECTEDBCAST 0 185#endif /* DIRECTED_BROADCAST */ 186#endif /* IPDIRECTEDBCAST */ 187int ipforwarding = IPFORWARDING; 188int ipsendredirects = IPSENDREDIRECTS; 189int ip_defttl = IPDEFTTL; 190int ip_forwsrcrt = IPFORWSRCRT; 191int ip_directedbcast = IPDIRECTEDBCAST; 192int ip_allowsrcrt = IPALLOWSRCRT; 193int ip_mtudisc = IPMTUDISC; 194int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 195#ifdef DIAGNOSTIC 196int ipprintfs = 0; 197#endif 198 199int ip_do_randomid = 0; 200 201/* 202 * XXX - Setting ip_checkinterface mostly implements the receive side of 203 * the Strong ES model described in RFC 1122, but since the routing table 204 * and transmit implementation do not implement the Strong ES model, 205 * setting this to 1 results in an odd hybrid. 206 * 207 * XXX - ip_checkinterface currently must be disabled if you use ipnat 208 * to translate the destination address to another local interface. 209 * 210 * XXX - ip_checkinterface must be disabled if you add IP aliases 211 * to the loopback interface instead of the interface where the 212 * packets for those addresses are received. 213 */ 214static int ip_checkinterface __read_mostly = 0; 215 216struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 217 218pktqueue_t * ip_pktq __read_mostly; 219pfil_head_t * inet_pfil_hook __read_mostly; 220ipid_state_t * ip_ids __read_mostly; 221percpu_t * ipstat_percpu __read_mostly; 222 223static percpu_t *ipforward_rt_percpu __cacheline_aligned; 224 225uint16_t ip_id; 226 227#ifdef INET_CSUM_COUNTERS 228#include <sys/device.h> 229 230struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 231 NULL, "inet", "hwcsum bad"); 232struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 233 NULL, "inet", "hwcsum ok"); 234struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 235 NULL, "inet", "swcsum"); 236 237#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 238 239EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 240EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 241EVCNT_ATTACH_STATIC(ip_swcsum); 242 243#else 244 245#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 246 247#endif /* INET_CSUM_COUNTERS */ 248 249/* 250 * Used to save the IP options in case a protocol wants to respond 251 * to an incoming packet over the same route if the packet got here 252 * using IP source routing. This allows connection establishment and 253 * maintenance when the remote end is on a network that is not known 254 * to us. 255 */ 256struct ip_srcrt { 257 int isr_nhops; /* number of hops */ 258 struct in_addr isr_dst; /* final destination */ 259 char isr_nop; /* one NOP to align */ 260 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */ 261 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)]; 262}; 263 264static int ip_drainwanted; 265 266static void save_rte(struct mbuf *, u_char *, struct in_addr); 267 268#ifdef MBUFTRACE 269struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 270struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 271#endif 272 273static void ipintr(void *); 274static void ip_input(struct mbuf *); 275static void ip_forward(struct mbuf *, int, struct ifnet *); 276static bool ip_dooptions(struct mbuf *); 277static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *); 278static void sysctl_net_inet_ip_setup(struct sysctllog **); 279 280static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *, 281 int *); 282static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *, 283 struct ip *); 284 285#ifdef NET_MPSAFE 286#define SOFTNET_LOCK() mutex_enter(softnet_lock) 287#define SOFTNET_UNLOCK() mutex_exit(softnet_lock) 288#else 289#define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 290#define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 291#endif 292 293/* 294 * IP initialization: fill in IP protocol switch table. 295 * All protocols not implemented in kernel go to raw IP protocol handler. 296 */ 297void 298ip_init(void) 299{ 300 const struct protosw *pr; 301 302 in_init(); 303 sysctl_net_inet_ip_setup(NULL); 304 305 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 306 KASSERT(pr != NULL); 307 308 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 309 KASSERT(ip_pktq != NULL); 310 311 for (u_int i = 0; i < IPPROTO_MAX; i++) { 312 ip_protox[i] = pr - inetsw; 313 } 314 for (pr = inetdomain.dom_protosw; 315 pr < inetdomain.dom_protoswNPROTOSW; pr++) 316 if (pr->pr_domain->dom_family == PF_INET && 317 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 318 ip_protox[pr->pr_protocol] = pr - inetsw; 319 320 ip_reass_init(); 321 322 ip_ids = ip_id_init(); 323 ip_id = time_uptime & 0xfffff; 324 325#ifdef GATEWAY 326 ipflow_init(); 327#endif 328 329 /* Register our Packet Filter hook. */ 330 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 331 KASSERT(inet_pfil_hook != NULL); 332 333#ifdef MBUFTRACE 334 MOWNER_ATTACH(&ip_tx_mowner); 335 MOWNER_ATTACH(&ip_rx_mowner); 336#endif /* MBUFTRACE */ 337 338 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 339 ipforward_rt_percpu = percpu_alloc(sizeof(struct route)); 340 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 341} 342 343static struct in_ifaddr * 344ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 345{ 346 struct in_ifaddr *ia = NULL; 347 int checkif; 348 349 /* 350 * Enable a consistency check between the destination address 351 * and the arrival interface for a unicast packet (the RFC 1122 352 * strong ES model) if IP forwarding is disabled and the packet 353 * is not locally generated. 354 * 355 * XXX - Checking also should be disabled if the destination 356 * address is ipnat'ed to a different interface. 357 * 358 * XXX - Checking is incompatible with IP aliases added 359 * to the loopback interface instead of the interface where 360 * the packets are received. 361 * 362 * XXX - We need to add a per ifaddr flag for this so that 363 * we get finer grain control. 364 */ 365 checkif = ip_checkinterface && (ipforwarding == 0) && 366 (ifp->if_flags & IFF_LOOPBACK) == 0; 367 368 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 369 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 370 if (ia->ia4_flags & IN_IFF_NOTREADY) 371 continue; 372 if (checkif && ia->ia_ifp != ifp) 373 continue; 374 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) { 375 (*downmatch)++; 376 continue; 377 } 378 if (ia->ia4_flags & IN_IFF_DETACHED && 379 (ifp->if_flags & IFF_LOOPBACK) == 0) 380 continue; 381 break; 382 } 383 } 384 385 return ia; 386} 387 388static struct in_ifaddr * 389ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 390{ 391 struct in_ifaddr *ia = NULL; 392 struct ifaddr *ifa; 393 394 IFADDR_READER_FOREACH(ifa, ifp) { 395 if (ifa->ifa_addr->sa_family != AF_INET) 396 continue; 397 ia = ifatoia(ifa); 398 if (ia->ia4_flags & IN_IFF_NOTREADY) 399 continue; 400 if (ia->ia4_flags & IN_IFF_DETACHED && 401 (ifp->if_flags & IFF_LOOPBACK) == 0) 402 continue; 403 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 404 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 405 /* 406 * Look for all-0's host part (old broadcast addr), 407 * either for subnet or net. 408 */ 409 ip->ip_dst.s_addr == ia->ia_subnet || 410 ip->ip_dst.s_addr == ia->ia_net) 411 goto matched; 412 /* 413 * An interface with IP address zero accepts 414 * all packets that arrive on that interface. 415 */ 416 if (in_nullhost(ia->ia_addr.sin_addr)) 417 goto matched; 418 } 419 ia = NULL; 420 421matched: 422 return ia; 423} 424 425/* 426 * IP software interrupt routine. 427 */ 428static void 429ipintr(void *arg __unused) 430{ 431 struct mbuf *m; 432 433 KASSERT(cpu_softintr_p()); 434 435 SOFTNET_LOCK_UNLESS_NET_MPSAFE(); 436 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 437 ip_input(m); 438 } 439 SOFTNET_UNLOCK_UNLESS_NET_MPSAFE(); 440} 441 442/* 443 * IP input routine. Checksum and byte swap header. If fragmented 444 * try to reassemble. Process options. Pass to next level. 445 */ 446static void 447ip_input(struct mbuf *m) 448{ 449 struct ip *ip = NULL; 450 struct in_ifaddr *ia = NULL; 451 int hlen = 0, len; 452 int downmatch; 453 int srcrt = 0; 454 ifnet_t *ifp; 455 struct psref psref; 456 int s; 457 458 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 459 "interrupt handler; synchronization assumptions violated"); 460 461 MCLAIM(m, &ip_rx_mowner); 462 KASSERT((m->m_flags & M_PKTHDR) != 0); 463 464 ifp = m_get_rcvif_psref(m, &psref); 465 if (__predict_false(ifp == NULL)) 466 goto out; 467 468 /* 469 * If no IP addresses have been set yet but the interfaces 470 * are receiving, can't do anything with incoming packets yet. 471 * Note: we pre-check without locks held. 472 */ 473 if (IN_ADDRLIST_READER_EMPTY()) 474 goto out; 475 IP_STATINC(IP_STAT_TOTAL); 476 477 /* 478 * If the IP header is not aligned, slurp it up into a new 479 * mbuf with space for link headers, in the event we forward 480 * it. Otherwise, if it is aligned, make sure the entire 481 * base IP header is in the first mbuf of the chain. 482 */ 483 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 484 if ((m = m_copyup(m, sizeof(struct ip), 485 (max_linkhdr + 3) & ~3)) == NULL) { 486 /* XXXJRT new stat, please */ 487 IP_STATINC(IP_STAT_TOOSMALL); 488 goto out; 489 } 490 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 491 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 492 IP_STATINC(IP_STAT_TOOSMALL); 493 goto out; 494 } 495 } 496 ip = mtod(m, struct ip *); 497 if (ip->ip_v != IPVERSION) { 498 IP_STATINC(IP_STAT_BADVERS); 499 goto out; 500 } 501 hlen = ip->ip_hl << 2; 502 if (hlen < sizeof(struct ip)) { /* minimum header length */ 503 IP_STATINC(IP_STAT_BADHLEN); 504 goto out; 505 } 506 if (hlen > m->m_len) { 507 if ((m = m_pullup(m, hlen)) == NULL) { 508 IP_STATINC(IP_STAT_BADHLEN); 509 goto out; 510 } 511 ip = mtod(m, struct ip *); 512 } 513 514 /* 515 * RFC1122: packets with a multicast source address are 516 * not allowed. 517 */ 518 if (IN_MULTICAST(ip->ip_src.s_addr)) { 519 IP_STATINC(IP_STAT_BADADDR); 520 goto out; 521 } 522 523 /* 127/8 must not appear on wire - RFC1122 */ 524 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 525 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 526 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 527 IP_STATINC(IP_STAT_BADADDR); 528 goto out; 529 } 530 } 531 532 switch (m->m_pkthdr.csum_flags & 533 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | 534 M_CSUM_IPv4_BAD)) { 535 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 536 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 537 IP_STATINC(IP_STAT_BADSUM); 538 goto out; 539 540 case M_CSUM_IPv4: 541 /* Checksum was okay. */ 542 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 543 break; 544 545 default: 546 /* 547 * Must compute it ourselves. Maybe skip checksum on 548 * loopback interfaces. 549 */ 550 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 551 ip_do_loopback_cksum)) { 552 INET_CSUM_COUNTER_INCR(&ip_swcsum); 553 if (in_cksum(m, hlen) != 0) { 554 IP_STATINC(IP_STAT_BADSUM); 555 goto out; 556 } 557 } 558 break; 559 } 560 561 /* Retrieve the packet length. */ 562 len = ntohs(ip->ip_len); 563 564 /* 565 * Check for additional length bogosity 566 */ 567 if (len < hlen) { 568 IP_STATINC(IP_STAT_BADLEN); 569 goto out; 570 } 571 572 /* 573 * Check that the amount of data in the buffers 574 * is as at least much as the IP header would have us expect. 575 * Trim mbufs if longer than we expect. 576 * Drop packet if shorter than we expect. 577 */ 578 if (m->m_pkthdr.len < len) { 579 IP_STATINC(IP_STAT_TOOSHORT); 580 goto out; 581 } 582 if (m->m_pkthdr.len > len) { 583 if (m->m_len == m->m_pkthdr.len) { 584 m->m_len = len; 585 m->m_pkthdr.len = len; 586 } else 587 m_adj(m, len - m->m_pkthdr.len); 588 } 589 590 /* 591 * Assume that we can create a fast-forward IP flow entry 592 * based on this packet. 593 */ 594 m->m_flags |= M_CANFASTFWD; 595 596 /* 597 * Run through list of hooks for input packets. If there are any 598 * filters which require that additional packets in the flow are 599 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 600 * Note that filters must _never_ set this flag, as another filter 601 * in the list may have previously cleared it. 602 */ 603#if defined(IPSEC) 604 if (!ipsec_used || !ipsec_indone(m)) 605#else 606 if (1) 607#endif 608 { 609 struct in_addr odst = ip->ip_dst; 610 bool freed; 611 612 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 613 if (freed || m == NULL) { 614 m = NULL; 615 goto out; 616 } 617 ip = mtod(m, struct ip *); 618 hlen = ip->ip_hl << 2; 619 620 /* 621 * XXX The setting of "srcrt" here is to prevent ip_forward() 622 * from generating ICMP redirects for packets that have 623 * been redirected by a hook back out on to the same LAN that 624 * they came from and is not an indication that the packet 625 * is being inffluenced by source routing options. This 626 * allows things like 627 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 628 * where tlp0 is both on the 1.1.1.0/24 network and is the 629 * default route for hosts on 1.1.1.0/24. Of course this 630 * also requires a "map tlp0 ..." to complete the story. 631 * One might argue whether or not this kind of network config. 632 * should be supported in this manner... 633 */ 634 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 635 } 636 637#ifdef ALTQ 638 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 639 if (altq_input) { 640 SOFTNET_LOCK(); 641 if ((*altq_input)(m, AF_INET) == 0) { 642 /* Packet dropped by traffic conditioner. */ 643 SOFTNET_UNLOCK(); 644 m = NULL; 645 goto out; 646 } 647 SOFTNET_UNLOCK(); 648 } 649#endif 650 651 /* 652 * Process options and, if not destined for us, 653 * ship it on. ip_dooptions returns 1 when an 654 * error was detected (causing an icmp message 655 * to be sent and the original packet to be freed). 656 */ 657 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 658 m = NULL; 659 goto out; 660 } 661 662 /* 663 * Check our list of addresses, to see if the packet is for us. 664 * 665 * Traditional 4.4BSD did not consult IFF_UP at all. 666 * The behavior here is to treat addresses on !IFF_UP interface 667 * or IN_IFF_NOTREADY addresses as not mine. 668 */ 669 downmatch = 0; 670 s = pserialize_read_enter(); 671 ia = ip_match_our_address(ifp, ip, &downmatch); 672 if (ia != NULL) { 673 pserialize_read_exit(s); 674 goto ours; 675 } 676 677 if (ifp->if_flags & IFF_BROADCAST) { 678 ia = ip_match_our_address_broadcast(ifp, ip); 679 if (ia != NULL) { 680 pserialize_read_exit(s); 681 goto ours; 682 } 683 } 684 pserialize_read_exit(s); 685 686 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 687#ifdef MROUTING 688 extern struct socket *ip_mrouter; 689 690 if (ip_mrouter) { 691 /* 692 * If we are acting as a multicast router, all 693 * incoming multicast packets are passed to the 694 * kernel-level multicast forwarding function. 695 * The packet is returned (relatively) intact; if 696 * ip_mforward() returns a non-zero value, the packet 697 * must be discarded, else it may be accepted below. 698 * 699 * (The IP ident field is put in the same byte order 700 * as expected when ip_mforward() is called from 701 * ip_output().) 702 */ 703 SOFTNET_LOCK(); 704 if (ip_mforward(m, ifp) != 0) { 705 SOFTNET_UNLOCK(); 706 IP_STATINC(IP_STAT_CANTFORWARD); 707 goto out; 708 } 709 SOFTNET_UNLOCK(); 710 711 /* 712 * The process-level routing demon needs to receive 713 * all multicast IGMP packets, whether or not this 714 * host belongs to their destination groups. 715 */ 716 if (ip->ip_p == IPPROTO_IGMP) { 717 goto ours; 718 } 719 IP_STATINC(IP_STAT_CANTFORWARD); 720 } 721#endif 722 /* 723 * See if we belong to the destination multicast group on the 724 * arrival interface. 725 */ 726 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 727 IP_STATINC(IP_STAT_CANTFORWARD); 728 goto out; 729 } 730 goto ours; 731 } 732 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 733 in_nullhost(ip->ip_dst)) 734 goto ours; 735 736 /* 737 * Not for us; forward if possible and desirable. 738 */ 739 if (ipforwarding == 0) { 740 m_put_rcvif_psref(ifp, &psref); 741 IP_STATINC(IP_STAT_CANTFORWARD); 742 m_freem(m); 743 } else { 744 /* 745 * If ip_dst matched any of my address on !IFF_UP interface, 746 * and there's no IFF_UP interface that matches ip_dst, 747 * send icmp unreach. Forwarding it will result in in-kernel 748 * forwarding loop till TTL goes to 0. 749 */ 750 if (downmatch) { 751 m_put_rcvif_psref(ifp, &psref); 752 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 753 IP_STATINC(IP_STAT_CANTFORWARD); 754 return; 755 } 756#ifdef IPSEC 757 /* Check the security policy (SP) for the packet */ 758 if (ipsec_used) { 759 if (ipsec4_input(m, IP_FORWARDING | 760 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 761 goto out; 762 } 763 } 764#endif 765 ip_forward(m, srcrt, ifp); 766 m_put_rcvif_psref(ifp, &psref); 767 } 768 return; 769 770ours: 771 m_put_rcvif_psref(ifp, &psref); 772 ifp = NULL; 773 774 /* 775 * If offset or IP_MF are set, must reassemble. 776 */ 777 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 778 /* 779 * Pass to IP reassembly mechanism. 780 */ 781 if (ip_reass_packet(&m, ip) != 0) { 782 /* Failed; invalid fragment(s) or packet. */ 783 goto out; 784 } 785 if (m == NULL) { 786 /* More fragments should come; silently return. */ 787 goto out; 788 } 789 /* 790 * Reassembly is done, we have the final packet. 791 * Updated cached data in local variable(s). 792 */ 793 ip = mtod(m, struct ip *); 794 hlen = ip->ip_hl << 2; 795 } 796 797#ifdef IPSEC 798 /* 799 * Enforce IPsec policy checking if we are seeing last header. 800 * Note that we do not visit this with protocols with PCB layer 801 * code - like UDP/TCP/raw IP. 802 */ 803 if (ipsec_used && 804 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 805 if (ipsec4_input(m, 0) != 0) { 806 goto out; 807 } 808 } 809#endif 810 811 /* 812 * Switch out to protocol's input routine. 813 */ 814#if IFA_STATS 815 if (ia && ip) { 816 struct in_ifaddr *_ia; 817 /* 818 * Keep a reference from ip_match_our_address with psref 819 * is expensive, so explore ia here again. 820 */ 821 s = pserialize_read_enter(); 822 _ia = in_get_ia(ip->ip_dst); 823 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 824 pserialize_read_exit(s); 825 } 826#endif 827 IP_STATINC(IP_STAT_DELIVERED); 828 829 const int off = hlen, nh = ip->ip_p; 830 831 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 832 return; 833 834out: 835 m_put_rcvif_psref(ifp, &psref); 836 if (m != NULL) 837 m_freem(m); 838} 839 840/* 841 * IP timer processing. 842 */ 843void 844ip_slowtimo(void) 845{ 846 847 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 848 849 ip_reass_slowtimo(); 850 851 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 852} 853 854/* 855 * IP drain processing. 856 */ 857void 858ip_drain(void) 859{ 860 861 KERNEL_LOCK(1, NULL); 862 ip_reass_drain(); 863 KERNEL_UNLOCK_ONE(NULL); 864} 865 866/* 867 * ip_dooptions: perform option processing on a datagram, possibly discarding 868 * it if bad options are encountered, or forwarding it if source-routed. 869 * 870 * => Returns true if packet has been forwarded/freed. 871 * => Returns false if the packet should be processed further. 872 */ 873static bool 874ip_dooptions(struct mbuf *m) 875{ 876 struct ip *ip = mtod(m, struct ip *); 877 u_char *cp, *cp0; 878 struct ip_timestamp *ipt; 879 struct in_ifaddr *ia; 880 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 881 struct in_addr dst; 882 n_time ntime; 883 struct ifaddr *ifa = NULL; 884 int s; 885 886 dst = ip->ip_dst; 887 cp = (u_char *)(ip + 1); 888 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 889 for (; cnt > 0; cnt -= optlen, cp += optlen) { 890 opt = cp[IPOPT_OPTVAL]; 891 if (opt == IPOPT_EOL) 892 break; 893 if (opt == IPOPT_NOP) 894 optlen = 1; 895 else { 896 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 897 code = &cp[IPOPT_OLEN] - (u_char *)ip; 898 goto bad; 899 } 900 optlen = cp[IPOPT_OLEN]; 901 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 902 code = &cp[IPOPT_OLEN] - (u_char *)ip; 903 goto bad; 904 } 905 } 906 switch (opt) { 907 908 default: 909 break; 910 911 /* 912 * Source routing with record. 913 * Find interface with current destination address. 914 * If none on this machine then drop if strictly routed, 915 * or do nothing if loosely routed. 916 * Record interface address and bring up next address 917 * component. If strictly routed make sure next 918 * address is on directly accessible net. 919 */ 920 case IPOPT_LSRR: 921 case IPOPT_SSRR: { 922 struct psref psref; 923 struct sockaddr_in ipaddr = { 924 .sin_len = sizeof(ipaddr), 925 .sin_family = AF_INET, 926 }; 927 928 if (ip_allowsrcrt == 0) { 929 type = ICMP_UNREACH; 930 code = ICMP_UNREACH_NET_PROHIB; 931 goto bad; 932 } 933 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 934 code = &cp[IPOPT_OLEN] - (u_char *)ip; 935 goto bad; 936 } 937 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 938 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 939 goto bad; 940 } 941 ipaddr.sin_addr = ip->ip_dst; 942 943 s = pserialize_read_enter(); 944 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 945 if (ifa == NULL) { 946 pserialize_read_exit(s); 947 if (opt == IPOPT_SSRR) { 948 type = ICMP_UNREACH; 949 code = ICMP_UNREACH_SRCFAIL; 950 goto bad; 951 } 952 /* 953 * Loose routing, and not at next destination 954 * yet; nothing to do except forward. 955 */ 956 break; 957 } 958 pserialize_read_exit(s); 959 960 off--; /* 0 origin */ 961 if ((off + sizeof(struct in_addr)) > optlen) { 962 /* 963 * End of source route. Should be for us. 964 */ 965 save_rte(m, cp, ip->ip_src); 966 break; 967 } 968 /* 969 * locate outgoing interface 970 */ 971 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 972 sizeof(ipaddr.sin_addr)); 973 if (opt == IPOPT_SSRR) { 974 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 975 &psref); 976 if (ifa != NULL) 977 ia = ifatoia(ifa); 978 else 979 ia = NULL; 980 } else { 981 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 982 } 983 if (ia == NULL) { 984 type = ICMP_UNREACH; 985 code = ICMP_UNREACH_SRCFAIL; 986 goto bad; 987 } 988 ip->ip_dst = ipaddr.sin_addr; 989 bcopy((void *)&ia->ia_addr.sin_addr, 990 (void *)(cp + off), sizeof(struct in_addr)); 991 ia4_release(ia, &psref); 992 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 993 /* 994 * Let ip_intr's mcast routing check handle mcast pkts 995 */ 996 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 997 break; 998 } 999 1000 case IPOPT_RR: { 1001 struct psref psref; 1002 struct sockaddr_in ipaddr = { 1003 .sin_len = sizeof(ipaddr), 1004 .sin_family = AF_INET, 1005 }; 1006 1007 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1008 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1009 goto bad; 1010 } 1011 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1012 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1013 goto bad; 1014 } 1015 /* 1016 * If no space remains, ignore. 1017 */ 1018 off--; /* 0 origin */ 1019 if ((off + sizeof(struct in_addr)) > optlen) 1020 break; 1021 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1022 sizeof(ipaddr.sin_addr)); 1023 /* 1024 * locate outgoing interface; if we're the destination, 1025 * use the incoming interface (should be same). 1026 */ 1027 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1028 if (ifa == NULL) { 1029 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1030 if (ia == NULL) { 1031 type = ICMP_UNREACH; 1032 code = ICMP_UNREACH_HOST; 1033 goto bad; 1034 } 1035 } else { 1036 ia = ifatoia(ifa); 1037 } 1038 bcopy((void *)&ia->ia_addr.sin_addr, 1039 (void *)(cp + off), sizeof(struct in_addr)); 1040 ia4_release(ia, &psref); 1041 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1042 break; 1043 } 1044 1045 case IPOPT_TS: 1046 code = cp - (u_char *)ip; 1047 ipt = (struct ip_timestamp *)cp; 1048 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1049 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1050 goto bad; 1051 } 1052 if (ipt->ipt_ptr < 5) { 1053 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1054 goto bad; 1055 } 1056 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1057 if (++ipt->ipt_oflw == 0) { 1058 code = (u_char *)&ipt->ipt_ptr - 1059 (u_char *)ip; 1060 goto bad; 1061 } 1062 break; 1063 } 1064 cp0 = (cp + ipt->ipt_ptr - 1); 1065 switch (ipt->ipt_flg) { 1066 1067 case IPOPT_TS_TSONLY: 1068 break; 1069 1070 case IPOPT_TS_TSANDADDR: { 1071 struct ifnet *rcvif; 1072 int _s, _ss; 1073 struct sockaddr_in ipaddr = { 1074 .sin_len = sizeof(ipaddr), 1075 .sin_family = AF_INET, 1076 }; 1077 1078 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1079 sizeof(struct in_addr) > ipt->ipt_len) { 1080 code = (u_char *)&ipt->ipt_ptr - 1081 (u_char *)ip; 1082 goto bad; 1083 } 1084 ipaddr.sin_addr = dst; 1085 _ss = pserialize_read_enter(); 1086 rcvif = m_get_rcvif(m, &_s); 1087 if (__predict_true(rcvif != NULL)) { 1088 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1089 rcvif); 1090 } 1091 m_put_rcvif(rcvif, &_s); 1092 if (ifa == NULL) { 1093 pserialize_read_exit(_ss); 1094 break; 1095 } 1096 ia = ifatoia(ifa); 1097 bcopy(&ia->ia_addr.sin_addr, 1098 cp0, sizeof(struct in_addr)); 1099 pserialize_read_exit(_ss); 1100 ipt->ipt_ptr += sizeof(struct in_addr); 1101 break; 1102 } 1103 1104 case IPOPT_TS_PRESPEC: { 1105 struct sockaddr_in ipaddr = { 1106 .sin_len = sizeof(ipaddr), 1107 .sin_family = AF_INET, 1108 }; 1109 1110 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1111 sizeof(struct in_addr) > ipt->ipt_len) { 1112 code = (u_char *)&ipt->ipt_ptr - 1113 (u_char *)ip; 1114 goto bad; 1115 } 1116 memcpy(&ipaddr.sin_addr, cp0, 1117 sizeof(struct in_addr)); 1118 s = pserialize_read_enter(); 1119 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1120 if (ifa == NULL) { 1121 pserialize_read_exit(s); 1122 continue; 1123 } 1124 pserialize_read_exit(s); 1125 ipt->ipt_ptr += sizeof(struct in_addr); 1126 break; 1127 } 1128 1129 default: 1130 /* XXX can't take &ipt->ipt_flg */ 1131 code = (u_char *)&ipt->ipt_ptr - 1132 (u_char *)ip + 1; 1133 goto bad; 1134 } 1135 ntime = iptime(); 1136 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1137 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1138 sizeof(n_time)); 1139 ipt->ipt_ptr += sizeof(n_time); 1140 } 1141 } 1142 if (forward) { 1143 struct ifnet *rcvif; 1144 struct psref _psref; 1145 1146 if (ip_forwsrcrt == 0) { 1147 type = ICMP_UNREACH; 1148 code = ICMP_UNREACH_SRCFAIL; 1149 goto bad; 1150 } 1151 1152 rcvif = m_get_rcvif_psref(m, &_psref); 1153 if (__predict_false(rcvif == NULL)) { 1154 type = ICMP_UNREACH; 1155 code = ICMP_UNREACH_HOST; 1156 goto bad; 1157 } 1158 ip_forward(m, 1, rcvif); 1159 m_put_rcvif_psref(rcvif, &_psref); 1160 return true; 1161 } 1162 return false; 1163bad: 1164 icmp_error(m, type, code, 0, 0); 1165 IP_STATINC(IP_STAT_BADOPTIONS); 1166 return true; 1167} 1168 1169/* 1170 * ip_rtaddr: given address of next destination (final or next hop), 1171 * return internet address info of interface to be used to get there. 1172 */ 1173static struct in_ifaddr * 1174ip_rtaddr(struct in_addr dst, struct psref *psref) 1175{ 1176 struct rtentry *rt; 1177 union { 1178 struct sockaddr dst; 1179 struct sockaddr_in dst4; 1180 } u; 1181 struct route *ro; 1182 1183 sockaddr_in_init(&u.dst4, &dst, 0); 1184 1185 ro = percpu_getref(ipforward_rt_percpu); 1186 rt = rtcache_lookup(ro, &u.dst); 1187 if (rt == NULL) { 1188 percpu_putref(ipforward_rt_percpu); 1189 return NULL; 1190 } 1191 1192 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1193 rtcache_unref(rt, ro); 1194 percpu_putref(ipforward_rt_percpu); 1195 1196 return ifatoia(rt->rt_ifa); 1197} 1198 1199/* 1200 * save_rte: save incoming source route for use in replies, to be picked 1201 * up later by ip_srcroute if the receiver is interested. 1202 */ 1203static void 1204save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1205{ 1206 struct ip_srcrt *isr; 1207 struct m_tag *mtag; 1208 unsigned olen; 1209 1210 olen = option[IPOPT_OLEN]; 1211 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1212 return; 1213 1214 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1215 if (mtag == NULL) 1216 return; 1217 isr = (struct ip_srcrt *)(mtag + 1); 1218 1219 memcpy(isr->isr_hdr, option, olen); 1220 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1221 isr->isr_dst = dst; 1222 m_tag_prepend(m, mtag); 1223} 1224 1225/* 1226 * Retrieve incoming source route for use in replies, 1227 * in the same form used by setsockopt. 1228 * The first hop is placed before the options, will be removed later. 1229 */ 1230struct mbuf * 1231ip_srcroute(struct mbuf *m0) 1232{ 1233 struct in_addr *p, *q; 1234 struct mbuf *m; 1235 struct ip_srcrt *isr; 1236 struct m_tag *mtag; 1237 1238 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL); 1239 if (mtag == NULL) 1240 return NULL; 1241 isr = (struct ip_srcrt *)(mtag + 1); 1242 1243 if (isr->isr_nhops == 0) 1244 return NULL; 1245 1246 m = m_get(M_DONTWAIT, MT_SOOPTS); 1247 if (m == NULL) 1248 return NULL; 1249 1250 MCLAIM(m, &inetdomain.dom_mowner); 1251#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1252 1253 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1254 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1255 1256 /* 1257 * First save first hop for return route 1258 */ 1259 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1260 *(mtod(m, struct in_addr *)) = *p--; 1261 1262 /* 1263 * Copy option fields and padding (nop) to mbuf. 1264 */ 1265 isr->isr_nop = IPOPT_NOP; 1266 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1267 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1268 OPTSIZ); 1269 q = (struct in_addr *)(mtod(m, char *) + 1270 sizeof(struct in_addr) + OPTSIZ); 1271#undef OPTSIZ 1272 /* 1273 * Record return path as an IP source route, 1274 * reversing the path (pointers are now aligned). 1275 */ 1276 while (p >= isr->isr_routes) { 1277 *q++ = *p--; 1278 } 1279 /* 1280 * Last hop goes to final destination. 1281 */ 1282 *q = isr->isr_dst; 1283 m_tag_delete(m0, mtag); 1284 return m; 1285} 1286 1287const int inetctlerrmap[PRC_NCMDS] = { 1288 [PRC_MSGSIZE] = EMSGSIZE, 1289 [PRC_HOSTDEAD] = EHOSTDOWN, 1290 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1291 [PRC_UNREACH_NET] = EHOSTUNREACH, 1292 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1293 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1294 [PRC_UNREACH_PORT] = ECONNREFUSED, 1295 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1296 [PRC_PARAMPROB] = ENOPROTOOPT, 1297}; 1298 1299void 1300ip_fasttimo(void) 1301{ 1302 if (ip_drainwanted) { 1303 ip_drain(); 1304 ip_drainwanted = 0; 1305 } 1306} 1307 1308void 1309ip_drainstub(void) 1310{ 1311 ip_drainwanted = 1; 1312} 1313 1314/* 1315 * Forward a packet. If some error occurs return the sender 1316 * an icmp packet. Note we can't always generate a meaningful 1317 * icmp message because icmp doesn't have a large enough repertoire 1318 * of codes and types. 1319 * 1320 * If not forwarding, just drop the packet. This could be confusing 1321 * if ipforwarding was zero but some routing protocol was advancing 1322 * us as a gateway to somewhere. However, we must let the routing 1323 * protocol deal with that. 1324 * 1325 * The srcrt parameter indicates whether the packet is being forwarded 1326 * via a source route. 1327 */ 1328static void 1329ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1330{ 1331 struct ip *ip = mtod(m, struct ip *); 1332 struct rtentry *rt; 1333 int error, type = 0, code = 0, destmtu = 0; 1334 struct mbuf *mcopy; 1335 n_long dest; 1336 union { 1337 struct sockaddr dst; 1338 struct sockaddr_in dst4; 1339 } u; 1340 uint64_t *ips; 1341 struct route *ro; 1342 1343 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1344 "interrupt handler; synchronization assumptions violated"); 1345 1346 /* 1347 * We are now in the output path. 1348 */ 1349 MCLAIM(m, &ip_tx_mowner); 1350 1351 /* 1352 * Clear any in-bound checksum flags for this packet. 1353 */ 1354 m->m_pkthdr.csum_flags = 0; 1355 1356 dest = 0; 1357 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1358 IP_STATINC(IP_STAT_CANTFORWARD); 1359 m_freem(m); 1360 return; 1361 } 1362 1363 if (ip->ip_ttl <= IPTTLDEC) { 1364 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1365 return; 1366 } 1367 1368 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1369 1370 ro = percpu_getref(ipforward_rt_percpu); 1371 rt = rtcache_lookup(ro, &u.dst); 1372 if (rt == NULL) { 1373 percpu_putref(ipforward_rt_percpu); 1374 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1375 return; 1376 } 1377 1378 /* 1379 * Save at most 68 bytes of the packet in case 1380 * we need to generate an ICMP message to the src. 1381 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1382 */ 1383 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1384 if (mcopy) 1385 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1386 1387 ip->ip_ttl -= IPTTLDEC; 1388 1389 /* 1390 * If forwarding packet using same interface that it came in on, 1391 * perhaps should send a redirect to sender to shortcut a hop. 1392 * Only send redirect if source is sending directly to us, 1393 * and if packet was not source routed (or has any options). 1394 * Also, don't send redirect if forwarding using a default route 1395 * or a route modified by a redirect. 1396 */ 1397 if (rt->rt_ifp == rcvif && 1398 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1399 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1400 ipsendredirects && !srcrt) { 1401 if (rt->rt_ifa && 1402 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1403 ifatoia(rt->rt_ifa)->ia_subnet) { 1404 if (rt->rt_flags & RTF_GATEWAY) 1405 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1406 else 1407 dest = ip->ip_dst.s_addr; 1408 /* 1409 * Router requirements says to only send host 1410 * redirects. 1411 */ 1412 type = ICMP_REDIRECT; 1413 code = ICMP_REDIRECT_HOST; 1414 } 1415 } 1416 rtcache_unref(rt, ro); 1417 1418 error = ip_output(m, NULL, ro, 1419 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1420 NULL, NULL); 1421 1422 if (error) { 1423 IP_STATINC(IP_STAT_CANTFORWARD); 1424 goto error; 1425 } 1426 1427 ips = IP_STAT_GETREF(); 1428 ips[IP_STAT_FORWARD]++; 1429 1430 if (type) { 1431 ips[IP_STAT_REDIRECTSENT]++; 1432 IP_STAT_PUTREF(); 1433 goto redirect; 1434 } 1435 1436 IP_STAT_PUTREF(); 1437 if (mcopy) { 1438#ifdef GATEWAY 1439 if (mcopy->m_flags & M_CANFASTFWD) 1440 ipflow_create(ro, mcopy); 1441#endif 1442 m_freem(mcopy); 1443 } 1444 1445 percpu_putref(ipforward_rt_percpu); 1446 return; 1447 1448redirect: 1449error: 1450 if (mcopy == NULL) { 1451 percpu_putref(ipforward_rt_percpu); 1452 return; 1453 } 1454 1455 switch (error) { 1456 1457 case 0: /* forwarded, but need redirect */ 1458 /* type, code set above */ 1459 break; 1460 1461 case ENETUNREACH: /* shouldn't happen, checked above */ 1462 case EHOSTUNREACH: 1463 case ENETDOWN: 1464 case EHOSTDOWN: 1465 default: 1466 type = ICMP_UNREACH; 1467 code = ICMP_UNREACH_HOST; 1468 break; 1469 1470 case EMSGSIZE: 1471 type = ICMP_UNREACH; 1472 code = ICMP_UNREACH_NEEDFRAG; 1473 1474 if ((rt = rtcache_validate(ro)) != NULL) { 1475 destmtu = rt->rt_ifp->if_mtu; 1476 rtcache_unref(rt, ro); 1477 } 1478#ifdef IPSEC 1479 if (ipsec_used) 1480 (void)ipsec4_forward(mcopy, &destmtu); 1481#endif 1482 IP_STATINC(IP_STAT_CANTFRAG); 1483 break; 1484 1485 case ENOBUFS: 1486 /* 1487 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1488 * Requirements for IP Version 4 Routers. Source quench can 1489 * big problem under DoS attacks or if the underlying 1490 * interface is rate-limited. 1491 */ 1492 if (mcopy) 1493 m_freem(mcopy); 1494 percpu_putref(ipforward_rt_percpu); 1495 return; 1496 } 1497 icmp_error(mcopy, type, code, dest, destmtu); 1498 percpu_putref(ipforward_rt_percpu); 1499} 1500 1501void 1502ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1503 struct mbuf *m) 1504{ 1505 struct socket *so = inp->inp_socket; 1506 int inpflags = inp->inp_flags; 1507 1508 if (SOOPT_TIMESTAMP(so->so_options)) 1509 mp = sbsavetimestamp(so->so_options, m, mp); 1510 1511 if (inpflags & INP_RECVDSTADDR) { 1512 *mp = sbcreatecontrol(&ip->ip_dst, 1513 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1514 if (*mp) 1515 mp = &(*mp)->m_next; 1516 } 1517 1518 if (inpflags & INP_RECVTTL) { 1519 *mp = sbcreatecontrol(&ip->ip_ttl, 1520 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1521 if (*mp) 1522 mp = &(*mp)->m_next; 1523 } 1524 1525 struct psref psref; 1526 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1527 if (__predict_false(ifp == NULL)) { 1528#ifdef DIAGNOSTIC 1529 printf("%s: missing receive interface\n", __func__); 1530#endif 1531 return; /* XXX should report error? */ 1532 } 1533 1534 if (inpflags & INP_RECVPKTINFO) { 1535 struct in_pktinfo ipi; 1536 ipi.ipi_addr = ip->ip_dst; 1537 ipi.ipi_ifindex = ifp->if_index; 1538 *mp = sbcreatecontrol(&ipi, 1539 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1540 if (*mp) 1541 mp = &(*mp)->m_next; 1542 } 1543 if (inpflags & INP_RECVIF) { 1544 struct sockaddr_dl sdl; 1545 1546 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1547 NULL, 0); 1548 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1549 if (*mp) 1550 mp = &(*mp)->m_next; 1551 } 1552 m_put_rcvif_psref(ifp, &psref); 1553} 1554 1555/* 1556 * sysctl helper routine for net.inet.ip.forwsrcrt. 1557 */ 1558static int 1559sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1560{ 1561 int error, tmp; 1562 struct sysctlnode node; 1563 1564 node = *rnode; 1565 tmp = ip_forwsrcrt; 1566 node.sysctl_data = &tmp; 1567 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1568 if (error || newp == NULL) 1569 return (error); 1570 1571 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1572 0, NULL, NULL, NULL); 1573 if (error) 1574 return (error); 1575 1576 ip_forwsrcrt = tmp; 1577 1578 return (0); 1579} 1580 1581/* 1582 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1583 * range of the new value and tweaks timers if it changes. 1584 */ 1585static int 1586sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1587{ 1588 int error, tmp; 1589 struct sysctlnode node; 1590 1591 icmp_mtudisc_lock(); 1592 1593 node = *rnode; 1594 tmp = ip_mtudisc_timeout; 1595 node.sysctl_data = &tmp; 1596 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1597 if (error || newp == NULL) 1598 goto out; 1599 if (tmp < 0) { 1600 error = EINVAL; 1601 goto out; 1602 } 1603 1604 ip_mtudisc_timeout = tmp; 1605 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1606 error = 0; 1607out: 1608 icmp_mtudisc_unlock(); 1609 return error; 1610} 1611 1612static int 1613sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1614{ 1615 1616 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1617} 1618 1619static void 1620sysctl_net_inet_ip_setup(struct sysctllog **clog) 1621{ 1622 sysctl_createv(clog, 0, NULL, NULL, 1623 CTLFLAG_PERMANENT, 1624 CTLTYPE_NODE, "inet", 1625 SYSCTL_DESCR("PF_INET related settings"), 1626 NULL, 0, NULL, 0, 1627 CTL_NET, PF_INET, CTL_EOL); 1628 sysctl_createv(clog, 0, NULL, NULL, 1629 CTLFLAG_PERMANENT, 1630 CTLTYPE_NODE, "ip", 1631 SYSCTL_DESCR("IPv4 related settings"), 1632 NULL, 0, NULL, 0, 1633 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1634 1635 sysctl_createv(clog, 0, NULL, NULL, 1636 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1637 CTLTYPE_INT, "forwarding", 1638 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1639 NULL, 0, &ipforwarding, 0, 1640 CTL_NET, PF_INET, IPPROTO_IP, 1641 IPCTL_FORWARDING, CTL_EOL); 1642 sysctl_createv(clog, 0, NULL, NULL, 1643 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1644 CTLTYPE_INT, "redirect", 1645 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1646 NULL, 0, &ipsendredirects, 0, 1647 CTL_NET, PF_INET, IPPROTO_IP, 1648 IPCTL_SENDREDIRECTS, CTL_EOL); 1649 sysctl_createv(clog, 0, NULL, NULL, 1650 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1651 CTLTYPE_INT, "ttl", 1652 SYSCTL_DESCR("Default TTL for an INET datagram"), 1653 NULL, 0, &ip_defttl, 0, 1654 CTL_NET, PF_INET, IPPROTO_IP, 1655 IPCTL_DEFTTL, CTL_EOL); 1656#ifdef IPCTL_DEFMTU 1657 sysctl_createv(clog, 0, NULL, NULL, 1658 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1659 CTLTYPE_INT, "mtu", 1660 SYSCTL_DESCR("Default MTA for an INET route"), 1661 NULL, 0, &ip_mtu, 0, 1662 CTL_NET, PF_INET, IPPROTO_IP, 1663 IPCTL_DEFMTU, CTL_EOL); 1664#endif /* IPCTL_DEFMTU */ 1665 sysctl_createv(clog, 0, NULL, NULL, 1666 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1667 CTLTYPE_INT, "forwsrcrt", 1668 SYSCTL_DESCR("Enable forwarding of source-routed " 1669 "datagrams"), 1670 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1671 CTL_NET, PF_INET, IPPROTO_IP, 1672 IPCTL_FORWSRCRT, CTL_EOL); 1673 sysctl_createv(clog, 0, NULL, NULL, 1674 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1675 CTLTYPE_INT, "directed-broadcast", 1676 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1677 NULL, 0, &ip_directedbcast, 0, 1678 CTL_NET, PF_INET, IPPROTO_IP, 1679 IPCTL_DIRECTEDBCAST, CTL_EOL); 1680 sysctl_createv(clog, 0, NULL, NULL, 1681 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1682 CTLTYPE_INT, "allowsrcrt", 1683 SYSCTL_DESCR("Accept source-routed datagrams"), 1684 NULL, 0, &ip_allowsrcrt, 0, 1685 CTL_NET, PF_INET, IPPROTO_IP, 1686 IPCTL_ALLOWSRCRT, CTL_EOL); 1687 1688 sysctl_createv(clog, 0, NULL, NULL, 1689 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1690 CTLTYPE_INT, "mtudisc", 1691 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1692 NULL, 0, &ip_mtudisc, 0, 1693 CTL_NET, PF_INET, IPPROTO_IP, 1694 IPCTL_MTUDISC, CTL_EOL); 1695 sysctl_createv(clog, 0, NULL, NULL, 1696 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1697 CTLTYPE_INT, "anonportmin", 1698 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1699 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1700 CTL_NET, PF_INET, IPPROTO_IP, 1701 IPCTL_ANONPORTMIN, CTL_EOL); 1702 sysctl_createv(clog, 0, NULL, NULL, 1703 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1704 CTLTYPE_INT, "anonportmax", 1705 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1706 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1707 CTL_NET, PF_INET, IPPROTO_IP, 1708 IPCTL_ANONPORTMAX, CTL_EOL); 1709 sysctl_createv(clog, 0, NULL, NULL, 1710 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1711 CTLTYPE_INT, "mtudisctimeout", 1712 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1713 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1714 CTL_NET, PF_INET, IPPROTO_IP, 1715 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1716#ifndef IPNOPRIVPORTS 1717 sysctl_createv(clog, 0, NULL, NULL, 1718 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1719 CTLTYPE_INT, "lowportmin", 1720 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1721 "to assign"), 1722 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1723 CTL_NET, PF_INET, IPPROTO_IP, 1724 IPCTL_LOWPORTMIN, CTL_EOL); 1725 sysctl_createv(clog, 0, NULL, NULL, 1726 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1727 CTLTYPE_INT, "lowportmax", 1728 SYSCTL_DESCR("Highest privileged ephemeral port number " 1729 "to assign"), 1730 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1731 CTL_NET, PF_INET, IPPROTO_IP, 1732 IPCTL_LOWPORTMAX, CTL_EOL); 1733#endif /* IPNOPRIVPORTS */ 1734#if NGRE > 0 1735 sysctl_createv(clog, 0, NULL, NULL, 1736 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1737 CTLTYPE_INT, "grettl", 1738 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1739 NULL, 0, &ip_gre_ttl, 0, 1740 CTL_NET, PF_INET, IPPROTO_IP, 1741 IPCTL_GRE_TTL, CTL_EOL); 1742#endif /* NGRE */ 1743 sysctl_createv(clog, 0, NULL, NULL, 1744 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1745 CTLTYPE_INT, "checkinterface", 1746 SYSCTL_DESCR("Enable receive side of Strong ES model " 1747 "from RFC1122"), 1748 NULL, 0, &ip_checkinterface, 0, 1749 CTL_NET, PF_INET, IPPROTO_IP, 1750 IPCTL_CHECKINTERFACE, CTL_EOL); 1751 sysctl_createv(clog, 0, NULL, NULL, 1752 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1753 CTLTYPE_INT, "random_id", 1754 SYSCTL_DESCR("Assign random ip_id values"), 1755 NULL, 0, &ip_do_randomid, 0, 1756 CTL_NET, PF_INET, IPPROTO_IP, 1757 IPCTL_RANDOMID, CTL_EOL); 1758 sysctl_createv(clog, 0, NULL, NULL, 1759 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1760 CTLTYPE_INT, "do_loopback_cksum", 1761 SYSCTL_DESCR("Perform IP checksum on loopback"), 1762 NULL, 0, &ip_do_loopback_cksum, 0, 1763 CTL_NET, PF_INET, IPPROTO_IP, 1764 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1765 sysctl_createv(clog, 0, NULL, NULL, 1766 CTLFLAG_PERMANENT, 1767 CTLTYPE_STRUCT, "stats", 1768 SYSCTL_DESCR("IP statistics"), 1769 sysctl_net_inet_ip_stats, 0, NULL, 0, 1770 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1771 CTL_EOL); 1772#if NARP 1773 sysctl_createv(clog, 0, NULL, NULL, 1774 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1775 CTLTYPE_INT, "dad_count", 1776 SYSCTL_DESCR("Number of Duplicate Address Detection " 1777 "probes to send"), 1778 NULL, 0, &ip_dad_count, 0, 1779 CTL_NET, PF_INET, IPPROTO_IP, 1780 IPCTL_DAD_COUNT, CTL_EOL); 1781#endif 1782 1783 /* anonportalgo RFC6056 subtree */ 1784 const struct sysctlnode *portalgo_node; 1785 sysctl_createv(clog, 0, NULL, &portalgo_node, 1786 CTLFLAG_PERMANENT, 1787 CTLTYPE_NODE, "anonportalgo", 1788 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1789 NULL, 0, NULL, 0, 1790 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1791 sysctl_createv(clog, 0, &portalgo_node, NULL, 1792 CTLFLAG_PERMANENT, 1793 CTLTYPE_STRING, "available", 1794 SYSCTL_DESCR("available algorithms"), 1795 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1796 CTL_CREATE, CTL_EOL); 1797 sysctl_createv(clog, 0, &portalgo_node, NULL, 1798 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1799 CTLTYPE_STRING, "selected", 1800 SYSCTL_DESCR("selected algorithm"), 1801 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1802 CTL_CREATE, CTL_EOL); 1803 sysctl_createv(clog, 0, &portalgo_node, NULL, 1804 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1805 CTLTYPE_STRUCT, "reserve", 1806 SYSCTL_DESCR("bitmap of reserved ports"), 1807 sysctl_portalgo_reserve4, 0, NULL, 0, 1808 CTL_CREATE, CTL_EOL); 1809} 1810 1811void 1812ip_statinc(u_int stat) 1813{ 1814 1815 KASSERT(stat < IP_NSTATS); 1816 IP_STATINC(stat); 1817} 1818