ip_input.c revision 1.366
1/* $NetBSD: ip_input.c,v 1.366 2018/02/05 13:23:11 maxv 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 1.366 2018/02/05 13:23:11 maxv 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 0 /* forward source-routed packets */ 165#endif 166#ifndef IPALLOWSRCRT 167#define IPALLOWSRCRT 0 /* 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) | M_CSUM_IPv4_BAD)) { 534 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 535 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 536 IP_STATINC(IP_STAT_BADSUM); 537 goto out; 538 539 case M_CSUM_IPv4: 540 /* Checksum was okay. */ 541 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 542 break; 543 544 default: 545 /* 546 * Must compute it ourselves. Maybe skip checksum on 547 * loopback interfaces. 548 */ 549 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 550 ip_do_loopback_cksum)) { 551 INET_CSUM_COUNTER_INCR(&ip_swcsum); 552 if (in_cksum(m, hlen) != 0) { 553 IP_STATINC(IP_STAT_BADSUM); 554 goto out; 555 } 556 } 557 break; 558 } 559 560 /* Retrieve the packet length. */ 561 len = ntohs(ip->ip_len); 562 563 /* 564 * Check for additional length bogosity 565 */ 566 if (len < hlen) { 567 IP_STATINC(IP_STAT_BADLEN); 568 goto out; 569 } 570 571 /* 572 * Check that the amount of data in the buffers is at least as much 573 * as the IP header would have us expect. Trim mbufs if longer than 574 * we expect. Drop packet if shorter than we expect. 575 */ 576 if (m->m_pkthdr.len < len) { 577 IP_STATINC(IP_STAT_TOOSHORT); 578 goto out; 579 } 580 if (m->m_pkthdr.len > len) { 581 if (m->m_len == m->m_pkthdr.len) { 582 m->m_len = len; 583 m->m_pkthdr.len = len; 584 } else 585 m_adj(m, len - m->m_pkthdr.len); 586 } 587 588 /* 589 * Assume that we can create a fast-forward IP flow entry 590 * based on this packet. 591 */ 592 m->m_flags |= M_CANFASTFWD; 593 594 /* 595 * Run through list of hooks for input packets. If there are any 596 * filters which require that additional packets in the flow are 597 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 598 * Note that filters must _never_ set this flag, as another filter 599 * in the list may have previously cleared it. 600 */ 601#if defined(IPSEC) 602 if (!ipsec_used || !ipsec_indone(m)) 603#else 604 if (1) 605#endif 606 { 607 struct in_addr odst = ip->ip_dst; 608 bool freed; 609 610 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 611 if (freed || m == NULL) { 612 m = NULL; 613 goto out; 614 } 615 ip = mtod(m, struct ip *); 616 hlen = ip->ip_hl << 2; 617 618 /* 619 * XXX The setting of "srcrt" here is to prevent ip_forward() 620 * from generating ICMP redirects for packets that have 621 * been redirected by a hook back out on to the same LAN that 622 * they came from and is not an indication that the packet 623 * is being inffluenced by source routing options. This 624 * allows things like 625 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 626 * where tlp0 is both on the 1.1.1.0/24 network and is the 627 * default route for hosts on 1.1.1.0/24. Of course this 628 * also requires a "map tlp0 ..." to complete the story. 629 * One might argue whether or not this kind of network config. 630 * should be supported in this manner... 631 */ 632 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 633 } 634 635#ifdef ALTQ 636 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 637 if (altq_input) { 638 SOFTNET_LOCK(); 639 if ((*altq_input)(m, AF_INET) == 0) { 640 /* Packet dropped by traffic conditioner. */ 641 SOFTNET_UNLOCK(); 642 m = NULL; 643 goto out; 644 } 645 SOFTNET_UNLOCK(); 646 } 647#endif 648 649 /* 650 * Process options and, if not destined for us, 651 * ship it on. ip_dooptions returns 1 when an 652 * error was detected (causing an icmp message 653 * to be sent and the original packet to be freed). 654 */ 655 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 656 m = NULL; 657 goto out; 658 } 659 660 /* 661 * Check our list of addresses, to see if the packet is for us. 662 * 663 * Traditional 4.4BSD did not consult IFF_UP at all. 664 * The behavior here is to treat addresses on !IFF_UP interface 665 * or IN_IFF_NOTREADY addresses as not mine. 666 */ 667 downmatch = 0; 668 s = pserialize_read_enter(); 669 ia = ip_match_our_address(ifp, ip, &downmatch); 670 if (ia != NULL) { 671 pserialize_read_exit(s); 672 goto ours; 673 } 674 675 if (ifp->if_flags & IFF_BROADCAST) { 676 ia = ip_match_our_address_broadcast(ifp, ip); 677 if (ia != NULL) { 678 pserialize_read_exit(s); 679 goto ours; 680 } 681 } 682 pserialize_read_exit(s); 683 684 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 685#ifdef MROUTING 686 extern struct socket *ip_mrouter; 687 688 if (ip_mrouter) { 689 /* 690 * If we are acting as a multicast router, all 691 * incoming multicast packets are passed to the 692 * kernel-level multicast forwarding function. 693 * The packet is returned (relatively) intact; if 694 * ip_mforward() returns a non-zero value, the packet 695 * must be discarded, else it may be accepted below. 696 * 697 * (The IP ident field is put in the same byte order 698 * as expected when ip_mforward() is called from 699 * ip_output().) 700 */ 701 SOFTNET_LOCK(); 702 if (ip_mforward(m, ifp) != 0) { 703 SOFTNET_UNLOCK(); 704 IP_STATINC(IP_STAT_CANTFORWARD); 705 goto out; 706 } 707 SOFTNET_UNLOCK(); 708 709 /* 710 * The process-level routing demon needs to receive 711 * all multicast IGMP packets, whether or not this 712 * host belongs to their destination groups. 713 */ 714 if (ip->ip_p == IPPROTO_IGMP) { 715 goto ours; 716 } 717 IP_STATINC(IP_STAT_CANTFORWARD); 718 } 719#endif 720 /* 721 * See if we belong to the destination multicast group on the 722 * arrival interface. 723 */ 724 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 725 IP_STATINC(IP_STAT_CANTFORWARD); 726 goto out; 727 } 728 goto ours; 729 } 730 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 731 in_nullhost(ip->ip_dst)) 732 goto ours; 733 734 /* 735 * Not for us; forward if possible and desirable. 736 */ 737 if (ipforwarding == 0) { 738 m_put_rcvif_psref(ifp, &psref); 739 IP_STATINC(IP_STAT_CANTFORWARD); 740 m_freem(m); 741 } else { 742 /* 743 * If ip_dst matched any of my address on !IFF_UP interface, 744 * and there's no IFF_UP interface that matches ip_dst, 745 * send icmp unreach. Forwarding it will result in in-kernel 746 * forwarding loop till TTL goes to 0. 747 */ 748 if (downmatch) { 749 m_put_rcvif_psref(ifp, &psref); 750 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 751 IP_STATINC(IP_STAT_CANTFORWARD); 752 return; 753 } 754#ifdef IPSEC 755 /* Check the security policy (SP) for the packet */ 756 if (ipsec_used) { 757 if (ipsec4_input(m, IP_FORWARDING | 758 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 759 goto out; 760 } 761 } 762#endif 763 ip_forward(m, srcrt, ifp); 764 m_put_rcvif_psref(ifp, &psref); 765 } 766 return; 767 768ours: 769 m_put_rcvif_psref(ifp, &psref); 770 ifp = NULL; 771 772 /* 773 * If offset or IP_MF are set, must reassemble. 774 */ 775 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 776 /* 777 * Pass to IP reassembly mechanism. 778 */ 779 if (ip_reass_packet(&m, ip) != 0) { 780 /* Failed; invalid fragment(s) or packet. */ 781 goto out; 782 } 783 if (m == NULL) { 784 /* More fragments should come; silently return. */ 785 goto out; 786 } 787 /* 788 * Reassembly is done, we have the final packet. 789 * Update cached data in local variable(s). 790 */ 791 ip = mtod(m, struct ip *); 792 hlen = ip->ip_hl << 2; 793 } 794 795#ifdef IPSEC 796 /* 797 * Enforce IPsec policy checking if we are seeing last header. 798 * Note that we do not visit this with protocols with PCB layer 799 * code - like UDP/TCP/raw IP. 800 */ 801 if (ipsec_used && 802 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 803 if (ipsec4_input(m, 0) != 0) { 804 goto out; 805 } 806 } 807#endif 808 809 /* 810 * Switch out to protocol's input routine. 811 */ 812#if IFA_STATS 813 if (ia && ip) { 814 struct in_ifaddr *_ia; 815 /* 816 * Keep a reference from ip_match_our_address with psref 817 * is expensive, so explore ia here again. 818 */ 819 s = pserialize_read_enter(); 820 _ia = in_get_ia(ip->ip_dst); 821 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 822 pserialize_read_exit(s); 823 } 824#endif 825 IP_STATINC(IP_STAT_DELIVERED); 826 827 const int off = hlen, nh = ip->ip_p; 828 829 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 830 return; 831 832out: 833 m_put_rcvif_psref(ifp, &psref); 834 if (m != NULL) 835 m_freem(m); 836} 837 838/* 839 * IP timer processing. 840 */ 841void 842ip_slowtimo(void) 843{ 844 845 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 846 847 ip_reass_slowtimo(); 848 849 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 850} 851 852/* 853 * IP drain processing. 854 */ 855void 856ip_drain(void) 857{ 858 859 KERNEL_LOCK(1, NULL); 860 ip_reass_drain(); 861 KERNEL_UNLOCK_ONE(NULL); 862} 863 864/* 865 * ip_dooptions: perform option processing on a datagram, possibly discarding 866 * it if bad options are encountered, or forwarding it if source-routed. 867 * 868 * => Returns true if packet has been forwarded/freed. 869 * => Returns false if the packet should be processed further. 870 */ 871static bool 872ip_dooptions(struct mbuf *m) 873{ 874 struct ip *ip = mtod(m, struct ip *); 875 u_char *cp, *cp0; 876 struct ip_timestamp *ipt; 877 struct in_ifaddr *ia; 878 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 879 struct in_addr dst; 880 n_time ntime; 881 struct ifaddr *ifa = NULL; 882 int s; 883 884 dst = ip->ip_dst; 885 cp = (u_char *)(ip + 1); 886 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 887 for (; cnt > 0; cnt -= optlen, cp += optlen) { 888 opt = cp[IPOPT_OPTVAL]; 889 if (opt == IPOPT_EOL) 890 break; 891 if (opt == IPOPT_NOP) 892 optlen = 1; 893 else { 894 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 895 code = &cp[IPOPT_OLEN] - (u_char *)ip; 896 goto bad; 897 } 898 optlen = cp[IPOPT_OLEN]; 899 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 900 code = &cp[IPOPT_OLEN] - (u_char *)ip; 901 goto bad; 902 } 903 } 904 switch (opt) { 905 906 default: 907 break; 908 909 /* 910 * Source routing with record. 911 * Find interface with current destination address. 912 * If none on this machine then drop if strictly routed, 913 * or do nothing if loosely routed. 914 * Record interface address and bring up next address 915 * component. If strictly routed make sure next 916 * address is on directly accessible net. 917 */ 918 case IPOPT_LSRR: 919 case IPOPT_SSRR: { 920 struct psref psref; 921 struct sockaddr_in ipaddr = { 922 .sin_len = sizeof(ipaddr), 923 .sin_family = AF_INET, 924 }; 925 926 if (ip_allowsrcrt == 0) { 927 type = ICMP_UNREACH; 928 code = ICMP_UNREACH_NET_PROHIB; 929 goto bad; 930 } 931 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 932 code = &cp[IPOPT_OLEN] - (u_char *)ip; 933 goto bad; 934 } 935 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 936 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 937 goto bad; 938 } 939 ipaddr.sin_addr = ip->ip_dst; 940 941 s = pserialize_read_enter(); 942 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 943 if (ifa == NULL) { 944 pserialize_read_exit(s); 945 if (opt == IPOPT_SSRR) { 946 type = ICMP_UNREACH; 947 code = ICMP_UNREACH_SRCFAIL; 948 goto bad; 949 } 950 /* 951 * Loose routing, and not at next destination 952 * yet; nothing to do except forward. 953 */ 954 break; 955 } 956 pserialize_read_exit(s); 957 958 off--; /* 0 origin */ 959 if ((off + sizeof(struct in_addr)) > optlen) { 960 /* 961 * End of source route. Should be for us. 962 */ 963 save_rte(m, cp, ip->ip_src); 964 break; 965 } 966 /* 967 * locate outgoing interface 968 */ 969 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 970 sizeof(ipaddr.sin_addr)); 971 if (opt == IPOPT_SSRR) { 972 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 973 &psref); 974 if (ifa != NULL) 975 ia = ifatoia(ifa); 976 else 977 ia = NULL; 978 } else { 979 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 980 } 981 if (ia == NULL) { 982 type = ICMP_UNREACH; 983 code = ICMP_UNREACH_SRCFAIL; 984 goto bad; 985 } 986 ip->ip_dst = ipaddr.sin_addr; 987 bcopy((void *)&ia->ia_addr.sin_addr, 988 (void *)(cp + off), sizeof(struct in_addr)); 989 ia4_release(ia, &psref); 990 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 991 /* 992 * Let ip_intr's mcast routing check handle mcast pkts 993 */ 994 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 995 break; 996 } 997 998 case IPOPT_RR: { 999 struct psref psref; 1000 struct sockaddr_in ipaddr = { 1001 .sin_len = sizeof(ipaddr), 1002 .sin_family = AF_INET, 1003 }; 1004 1005 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1006 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1007 goto bad; 1008 } 1009 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1010 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1011 goto bad; 1012 } 1013 /* 1014 * If no space remains, ignore. 1015 */ 1016 off--; /* 0 origin */ 1017 if ((off + sizeof(struct in_addr)) > optlen) 1018 break; 1019 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1020 sizeof(ipaddr.sin_addr)); 1021 /* 1022 * locate outgoing interface; if we're the destination, 1023 * use the incoming interface (should be same). 1024 */ 1025 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1026 if (ifa == NULL) { 1027 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1028 if (ia == NULL) { 1029 type = ICMP_UNREACH; 1030 code = ICMP_UNREACH_HOST; 1031 goto bad; 1032 } 1033 } else { 1034 ia = ifatoia(ifa); 1035 } 1036 bcopy((void *)&ia->ia_addr.sin_addr, 1037 (void *)(cp + off), sizeof(struct in_addr)); 1038 ia4_release(ia, &psref); 1039 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1040 break; 1041 } 1042 1043 case IPOPT_TS: 1044 code = cp - (u_char *)ip; 1045 ipt = (struct ip_timestamp *)cp; 1046 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1047 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1048 goto bad; 1049 } 1050 if (ipt->ipt_ptr < 5) { 1051 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1052 goto bad; 1053 } 1054 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1055 if (++ipt->ipt_oflw == 0) { 1056 code = (u_char *)&ipt->ipt_ptr - 1057 (u_char *)ip; 1058 goto bad; 1059 } 1060 break; 1061 } 1062 cp0 = (cp + ipt->ipt_ptr - 1); 1063 switch (ipt->ipt_flg) { 1064 1065 case IPOPT_TS_TSONLY: 1066 break; 1067 1068 case IPOPT_TS_TSANDADDR: { 1069 struct ifnet *rcvif; 1070 int _s, _ss; 1071 struct sockaddr_in ipaddr = { 1072 .sin_len = sizeof(ipaddr), 1073 .sin_family = AF_INET, 1074 }; 1075 1076 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1077 sizeof(struct in_addr) > ipt->ipt_len) { 1078 code = (u_char *)&ipt->ipt_ptr - 1079 (u_char *)ip; 1080 goto bad; 1081 } 1082 ipaddr.sin_addr = dst; 1083 _ss = pserialize_read_enter(); 1084 rcvif = m_get_rcvif(m, &_s); 1085 if (__predict_true(rcvif != NULL)) { 1086 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1087 rcvif); 1088 } 1089 m_put_rcvif(rcvif, &_s); 1090 if (ifa == NULL) { 1091 pserialize_read_exit(_ss); 1092 break; 1093 } 1094 ia = ifatoia(ifa); 1095 bcopy(&ia->ia_addr.sin_addr, 1096 cp0, sizeof(struct in_addr)); 1097 pserialize_read_exit(_ss); 1098 ipt->ipt_ptr += sizeof(struct in_addr); 1099 break; 1100 } 1101 1102 case IPOPT_TS_PRESPEC: { 1103 struct sockaddr_in ipaddr = { 1104 .sin_len = sizeof(ipaddr), 1105 .sin_family = AF_INET, 1106 }; 1107 1108 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1109 sizeof(struct in_addr) > ipt->ipt_len) { 1110 code = (u_char *)&ipt->ipt_ptr - 1111 (u_char *)ip; 1112 goto bad; 1113 } 1114 memcpy(&ipaddr.sin_addr, cp0, 1115 sizeof(struct in_addr)); 1116 s = pserialize_read_enter(); 1117 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1118 if (ifa == NULL) { 1119 pserialize_read_exit(s); 1120 continue; 1121 } 1122 pserialize_read_exit(s); 1123 ipt->ipt_ptr += sizeof(struct in_addr); 1124 break; 1125 } 1126 1127 default: 1128 /* XXX can't take &ipt->ipt_flg */ 1129 code = (u_char *)&ipt->ipt_ptr - 1130 (u_char *)ip + 1; 1131 goto bad; 1132 } 1133 ntime = iptime(); 1134 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1135 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1136 sizeof(n_time)); 1137 ipt->ipt_ptr += sizeof(n_time); 1138 } 1139 } 1140 if (forward) { 1141 struct ifnet *rcvif; 1142 struct psref _psref; 1143 1144 if (ip_forwsrcrt == 0) { 1145 type = ICMP_UNREACH; 1146 code = ICMP_UNREACH_SRCFAIL; 1147 goto bad; 1148 } 1149 1150 rcvif = m_get_rcvif_psref(m, &_psref); 1151 if (__predict_false(rcvif == NULL)) { 1152 type = ICMP_UNREACH; 1153 code = ICMP_UNREACH_HOST; 1154 goto bad; 1155 } 1156 ip_forward(m, 1, rcvif); 1157 m_put_rcvif_psref(rcvif, &_psref); 1158 return true; 1159 } 1160 return false; 1161bad: 1162 icmp_error(m, type, code, 0, 0); 1163 IP_STATINC(IP_STAT_BADOPTIONS); 1164 return true; 1165} 1166 1167/* 1168 * ip_rtaddr: given address of next destination (final or next hop), 1169 * return internet address info of interface to be used to get there. 1170 */ 1171static struct in_ifaddr * 1172ip_rtaddr(struct in_addr dst, struct psref *psref) 1173{ 1174 struct rtentry *rt; 1175 union { 1176 struct sockaddr dst; 1177 struct sockaddr_in dst4; 1178 } u; 1179 struct route *ro; 1180 1181 sockaddr_in_init(&u.dst4, &dst, 0); 1182 1183 ro = percpu_getref(ipforward_rt_percpu); 1184 rt = rtcache_lookup(ro, &u.dst); 1185 if (rt == NULL) { 1186 percpu_putref(ipforward_rt_percpu); 1187 return NULL; 1188 } 1189 1190 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1191 rtcache_unref(rt, ro); 1192 percpu_putref(ipforward_rt_percpu); 1193 1194 return ifatoia(rt->rt_ifa); 1195} 1196 1197/* 1198 * save_rte: save incoming source route for use in replies, to be picked 1199 * up later by ip_srcroute if the receiver is interested. 1200 */ 1201static void 1202save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1203{ 1204 struct ip_srcrt *isr; 1205 struct m_tag *mtag; 1206 unsigned olen; 1207 1208 olen = option[IPOPT_OLEN]; 1209 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1210 return; 1211 1212 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1213 if (mtag == NULL) 1214 return; 1215 isr = (struct ip_srcrt *)(mtag + 1); 1216 1217 memcpy(isr->isr_hdr, option, olen); 1218 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1219 isr->isr_dst = dst; 1220 m_tag_prepend(m, mtag); 1221} 1222 1223/* 1224 * Retrieve incoming source route for use in replies, 1225 * in the same form used by setsockopt. 1226 * The first hop is placed before the options, will be removed later. 1227 */ 1228struct mbuf * 1229ip_srcroute(struct mbuf *m0) 1230{ 1231 struct in_addr *p, *q; 1232 struct mbuf *m; 1233 struct ip_srcrt *isr; 1234 struct m_tag *mtag; 1235 1236 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL); 1237 if (mtag == NULL) 1238 return NULL; 1239 isr = (struct ip_srcrt *)(mtag + 1); 1240 1241 if (isr->isr_nhops == 0) 1242 return NULL; 1243 1244 m = m_get(M_DONTWAIT, MT_SOOPTS); 1245 if (m == NULL) 1246 return NULL; 1247 1248 MCLAIM(m, &inetdomain.dom_mowner); 1249#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1250 1251 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1252 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1253 1254 /* 1255 * First save first hop for return route 1256 */ 1257 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1258 *(mtod(m, struct in_addr *)) = *p--; 1259 1260 /* 1261 * Copy option fields and padding (nop) to mbuf. 1262 */ 1263 isr->isr_nop = IPOPT_NOP; 1264 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1265 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1266 OPTSIZ); 1267 q = (struct in_addr *)(mtod(m, char *) + 1268 sizeof(struct in_addr) + OPTSIZ); 1269#undef OPTSIZ 1270 /* 1271 * Record return path as an IP source route, 1272 * reversing the path (pointers are now aligned). 1273 */ 1274 while (p >= isr->isr_routes) { 1275 *q++ = *p--; 1276 } 1277 /* 1278 * Last hop goes to final destination. 1279 */ 1280 *q = isr->isr_dst; 1281 m_tag_delete(m0, mtag); 1282 return m; 1283} 1284 1285const int inetctlerrmap[PRC_NCMDS] = { 1286 [PRC_MSGSIZE] = EMSGSIZE, 1287 [PRC_HOSTDEAD] = EHOSTDOWN, 1288 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1289 [PRC_UNREACH_NET] = EHOSTUNREACH, 1290 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1291 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1292 [PRC_UNREACH_PORT] = ECONNREFUSED, 1293 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1294 [PRC_PARAMPROB] = ENOPROTOOPT, 1295}; 1296 1297void 1298ip_fasttimo(void) 1299{ 1300 if (ip_drainwanted) { 1301 ip_drain(); 1302 ip_drainwanted = 0; 1303 } 1304} 1305 1306void 1307ip_drainstub(void) 1308{ 1309 ip_drainwanted = 1; 1310} 1311 1312/* 1313 * Forward a packet. If some error occurs return the sender 1314 * an icmp packet. Note we can't always generate a meaningful 1315 * icmp message because icmp doesn't have a large enough repertoire 1316 * of codes and types. 1317 * 1318 * If not forwarding, just drop the packet. This could be confusing 1319 * if ipforwarding was zero but some routing protocol was advancing 1320 * us as a gateway to somewhere. However, we must let the routing 1321 * protocol deal with that. 1322 * 1323 * The srcrt parameter indicates whether the packet is being forwarded 1324 * via a source route. 1325 */ 1326static void 1327ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1328{ 1329 struct ip *ip = mtod(m, struct ip *); 1330 struct rtentry *rt; 1331 int error, type = 0, code = 0, destmtu = 0; 1332 struct mbuf *mcopy; 1333 n_long dest; 1334 union { 1335 struct sockaddr dst; 1336 struct sockaddr_in dst4; 1337 } u; 1338 uint64_t *ips; 1339 struct route *ro; 1340 1341 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1342 "interrupt handler; synchronization assumptions violated"); 1343 1344 /* 1345 * We are now in the output path. 1346 */ 1347 MCLAIM(m, &ip_tx_mowner); 1348 1349 /* 1350 * Clear any in-bound checksum flags for this packet. 1351 */ 1352 m->m_pkthdr.csum_flags = 0; 1353 1354 dest = 0; 1355 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1356 IP_STATINC(IP_STAT_CANTFORWARD); 1357 m_freem(m); 1358 return; 1359 } 1360 1361 if (ip->ip_ttl <= IPTTLDEC) { 1362 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1363 return; 1364 } 1365 1366 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1367 1368 ro = percpu_getref(ipforward_rt_percpu); 1369 rt = rtcache_lookup(ro, &u.dst); 1370 if (rt == NULL) { 1371 percpu_putref(ipforward_rt_percpu); 1372 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1373 return; 1374 } 1375 1376 /* 1377 * Save at most 68 bytes of the packet in case 1378 * we need to generate an ICMP message to the src. 1379 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1380 */ 1381 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1382 if (mcopy) 1383 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1384 1385 ip->ip_ttl -= IPTTLDEC; 1386 1387 /* 1388 * If forwarding packet using same interface that it came in on, 1389 * perhaps should send a redirect to sender to shortcut a hop. 1390 * Only send redirect if source is sending directly to us, 1391 * and if packet was not source routed (or has any options). 1392 * Also, don't send redirect if forwarding using a default route 1393 * or a route modified by a redirect. 1394 */ 1395 if (rt->rt_ifp == rcvif && 1396 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1397 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1398 ipsendredirects && !srcrt) { 1399 if (rt->rt_ifa && 1400 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1401 ifatoia(rt->rt_ifa)->ia_subnet) { 1402 if (rt->rt_flags & RTF_GATEWAY) 1403 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1404 else 1405 dest = ip->ip_dst.s_addr; 1406 /* 1407 * Router requirements says to only send host 1408 * redirects. 1409 */ 1410 type = ICMP_REDIRECT; 1411 code = ICMP_REDIRECT_HOST; 1412 } 1413 } 1414 rtcache_unref(rt, ro); 1415 1416 error = ip_output(m, NULL, ro, 1417 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1418 NULL, NULL); 1419 1420 if (error) { 1421 IP_STATINC(IP_STAT_CANTFORWARD); 1422 goto error; 1423 } 1424 1425 ips = IP_STAT_GETREF(); 1426 ips[IP_STAT_FORWARD]++; 1427 1428 if (type) { 1429 ips[IP_STAT_REDIRECTSENT]++; 1430 IP_STAT_PUTREF(); 1431 goto redirect; 1432 } 1433 1434 IP_STAT_PUTREF(); 1435 if (mcopy) { 1436#ifdef GATEWAY 1437 if (mcopy->m_flags & M_CANFASTFWD) 1438 ipflow_create(ro, mcopy); 1439#endif 1440 m_freem(mcopy); 1441 } 1442 1443 percpu_putref(ipforward_rt_percpu); 1444 return; 1445 1446redirect: 1447error: 1448 if (mcopy == NULL) { 1449 percpu_putref(ipforward_rt_percpu); 1450 return; 1451 } 1452 1453 switch (error) { 1454 1455 case 0: /* forwarded, but need redirect */ 1456 /* type, code set above */ 1457 break; 1458 1459 case ENETUNREACH: /* shouldn't happen, checked above */ 1460 case EHOSTUNREACH: 1461 case ENETDOWN: 1462 case EHOSTDOWN: 1463 default: 1464 type = ICMP_UNREACH; 1465 code = ICMP_UNREACH_HOST; 1466 break; 1467 1468 case EMSGSIZE: 1469 type = ICMP_UNREACH; 1470 code = ICMP_UNREACH_NEEDFRAG; 1471 1472 if ((rt = rtcache_validate(ro)) != NULL) { 1473 destmtu = rt->rt_ifp->if_mtu; 1474 rtcache_unref(rt, ro); 1475 } 1476#ifdef IPSEC 1477 if (ipsec_used) 1478 (void)ipsec4_forward(mcopy, &destmtu); 1479#endif 1480 IP_STATINC(IP_STAT_CANTFRAG); 1481 break; 1482 1483 case ENOBUFS: 1484 /* 1485 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1486 * Requirements for IP Version 4 Routers. Source quench can 1487 * big problem under DoS attacks or if the underlying 1488 * interface is rate-limited. 1489 */ 1490 if (mcopy) 1491 m_freem(mcopy); 1492 percpu_putref(ipforward_rt_percpu); 1493 return; 1494 } 1495 icmp_error(mcopy, type, code, dest, destmtu); 1496 percpu_putref(ipforward_rt_percpu); 1497} 1498 1499void 1500ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1501 struct mbuf *m) 1502{ 1503 struct socket *so = inp->inp_socket; 1504 int inpflags = inp->inp_flags; 1505 1506 if (SOOPT_TIMESTAMP(so->so_options)) 1507 mp = sbsavetimestamp(so->so_options, m, mp); 1508 1509 if (inpflags & INP_RECVDSTADDR) { 1510 *mp = sbcreatecontrol(&ip->ip_dst, 1511 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1512 if (*mp) 1513 mp = &(*mp)->m_next; 1514 } 1515 1516 if (inpflags & INP_RECVTTL) { 1517 *mp = sbcreatecontrol(&ip->ip_ttl, 1518 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1519 if (*mp) 1520 mp = &(*mp)->m_next; 1521 } 1522 1523 struct psref psref; 1524 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1525 if (__predict_false(ifp == NULL)) { 1526#ifdef DIAGNOSTIC 1527 printf("%s: missing receive interface\n", __func__); 1528#endif 1529 return; /* XXX should report error? */ 1530 } 1531 1532 if (inpflags & INP_RECVPKTINFO) { 1533 struct in_pktinfo ipi; 1534 ipi.ipi_addr = ip->ip_dst; 1535 ipi.ipi_ifindex = ifp->if_index; 1536 *mp = sbcreatecontrol(&ipi, 1537 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1538 if (*mp) 1539 mp = &(*mp)->m_next; 1540 } 1541 if (inpflags & INP_RECVIF) { 1542 struct sockaddr_dl sdl; 1543 1544 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1545 NULL, 0); 1546 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1547 if (*mp) 1548 mp = &(*mp)->m_next; 1549 } 1550 m_put_rcvif_psref(ifp, &psref); 1551} 1552 1553/* 1554 * sysctl helper routine for net.inet.ip.forwsrcrt. 1555 */ 1556static int 1557sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1558{ 1559 int error, tmp; 1560 struct sysctlnode node; 1561 1562 node = *rnode; 1563 tmp = ip_forwsrcrt; 1564 node.sysctl_data = &tmp; 1565 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1566 if (error || newp == NULL) 1567 return (error); 1568 1569 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1570 0, NULL, NULL, NULL); 1571 if (error) 1572 return (error); 1573 1574 ip_forwsrcrt = tmp; 1575 1576 return (0); 1577} 1578 1579/* 1580 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1581 * range of the new value and tweaks timers if it changes. 1582 */ 1583static int 1584sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1585{ 1586 int error, tmp; 1587 struct sysctlnode node; 1588 1589 icmp_mtudisc_lock(); 1590 1591 node = *rnode; 1592 tmp = ip_mtudisc_timeout; 1593 node.sysctl_data = &tmp; 1594 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1595 if (error || newp == NULL) 1596 goto out; 1597 if (tmp < 0) { 1598 error = EINVAL; 1599 goto out; 1600 } 1601 1602 ip_mtudisc_timeout = tmp; 1603 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1604 error = 0; 1605out: 1606 icmp_mtudisc_unlock(); 1607 return error; 1608} 1609 1610static int 1611sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1612{ 1613 1614 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1615} 1616 1617static void 1618sysctl_net_inet_ip_setup(struct sysctllog **clog) 1619{ 1620 sysctl_createv(clog, 0, NULL, NULL, 1621 CTLFLAG_PERMANENT, 1622 CTLTYPE_NODE, "inet", 1623 SYSCTL_DESCR("PF_INET related settings"), 1624 NULL, 0, NULL, 0, 1625 CTL_NET, PF_INET, CTL_EOL); 1626 sysctl_createv(clog, 0, NULL, NULL, 1627 CTLFLAG_PERMANENT, 1628 CTLTYPE_NODE, "ip", 1629 SYSCTL_DESCR("IPv4 related settings"), 1630 NULL, 0, NULL, 0, 1631 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1632 1633 sysctl_createv(clog, 0, NULL, NULL, 1634 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1635 CTLTYPE_INT, "forwarding", 1636 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1637 NULL, 0, &ipforwarding, 0, 1638 CTL_NET, PF_INET, IPPROTO_IP, 1639 IPCTL_FORWARDING, CTL_EOL); 1640 sysctl_createv(clog, 0, NULL, NULL, 1641 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1642 CTLTYPE_INT, "redirect", 1643 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1644 NULL, 0, &ipsendredirects, 0, 1645 CTL_NET, PF_INET, IPPROTO_IP, 1646 IPCTL_SENDREDIRECTS, CTL_EOL); 1647 sysctl_createv(clog, 0, NULL, NULL, 1648 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1649 CTLTYPE_INT, "ttl", 1650 SYSCTL_DESCR("Default TTL for an INET datagram"), 1651 NULL, 0, &ip_defttl, 0, 1652 CTL_NET, PF_INET, IPPROTO_IP, 1653 IPCTL_DEFTTL, CTL_EOL); 1654#ifdef IPCTL_DEFMTU 1655 sysctl_createv(clog, 0, NULL, NULL, 1656 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1657 CTLTYPE_INT, "mtu", 1658 SYSCTL_DESCR("Default MTA for an INET route"), 1659 NULL, 0, &ip_mtu, 0, 1660 CTL_NET, PF_INET, IPPROTO_IP, 1661 IPCTL_DEFMTU, CTL_EOL); 1662#endif /* IPCTL_DEFMTU */ 1663 sysctl_createv(clog, 0, NULL, NULL, 1664 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1665 CTLTYPE_INT, "forwsrcrt", 1666 SYSCTL_DESCR("Enable forwarding of source-routed " 1667 "datagrams"), 1668 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1669 CTL_NET, PF_INET, IPPROTO_IP, 1670 IPCTL_FORWSRCRT, CTL_EOL); 1671 sysctl_createv(clog, 0, NULL, NULL, 1672 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1673 CTLTYPE_INT, "directed-broadcast", 1674 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1675 NULL, 0, &ip_directedbcast, 0, 1676 CTL_NET, PF_INET, IPPROTO_IP, 1677 IPCTL_DIRECTEDBCAST, CTL_EOL); 1678 sysctl_createv(clog, 0, NULL, NULL, 1679 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1680 CTLTYPE_INT, "allowsrcrt", 1681 SYSCTL_DESCR("Accept source-routed datagrams"), 1682 NULL, 0, &ip_allowsrcrt, 0, 1683 CTL_NET, PF_INET, IPPROTO_IP, 1684 IPCTL_ALLOWSRCRT, CTL_EOL); 1685 1686 sysctl_createv(clog, 0, NULL, NULL, 1687 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1688 CTLTYPE_INT, "mtudisc", 1689 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1690 NULL, 0, &ip_mtudisc, 0, 1691 CTL_NET, PF_INET, IPPROTO_IP, 1692 IPCTL_MTUDISC, CTL_EOL); 1693 sysctl_createv(clog, 0, NULL, NULL, 1694 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1695 CTLTYPE_INT, "anonportmin", 1696 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1697 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1698 CTL_NET, PF_INET, IPPROTO_IP, 1699 IPCTL_ANONPORTMIN, CTL_EOL); 1700 sysctl_createv(clog, 0, NULL, NULL, 1701 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1702 CTLTYPE_INT, "anonportmax", 1703 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1704 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1705 CTL_NET, PF_INET, IPPROTO_IP, 1706 IPCTL_ANONPORTMAX, CTL_EOL); 1707 sysctl_createv(clog, 0, NULL, NULL, 1708 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1709 CTLTYPE_INT, "mtudisctimeout", 1710 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1711 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1712 CTL_NET, PF_INET, IPPROTO_IP, 1713 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1714#ifndef IPNOPRIVPORTS 1715 sysctl_createv(clog, 0, NULL, NULL, 1716 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1717 CTLTYPE_INT, "lowportmin", 1718 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1719 "to assign"), 1720 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1721 CTL_NET, PF_INET, IPPROTO_IP, 1722 IPCTL_LOWPORTMIN, CTL_EOL); 1723 sysctl_createv(clog, 0, NULL, NULL, 1724 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1725 CTLTYPE_INT, "lowportmax", 1726 SYSCTL_DESCR("Highest privileged ephemeral port number " 1727 "to assign"), 1728 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1729 CTL_NET, PF_INET, IPPROTO_IP, 1730 IPCTL_LOWPORTMAX, CTL_EOL); 1731#endif /* IPNOPRIVPORTS */ 1732#if NGRE > 0 1733 sysctl_createv(clog, 0, NULL, NULL, 1734 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1735 CTLTYPE_INT, "grettl", 1736 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1737 NULL, 0, &ip_gre_ttl, 0, 1738 CTL_NET, PF_INET, IPPROTO_IP, 1739 IPCTL_GRE_TTL, CTL_EOL); 1740#endif /* NGRE */ 1741 sysctl_createv(clog, 0, NULL, NULL, 1742 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1743 CTLTYPE_INT, "checkinterface", 1744 SYSCTL_DESCR("Enable receive side of Strong ES model " 1745 "from RFC1122"), 1746 NULL, 0, &ip_checkinterface, 0, 1747 CTL_NET, PF_INET, IPPROTO_IP, 1748 IPCTL_CHECKINTERFACE, CTL_EOL); 1749 sysctl_createv(clog, 0, NULL, NULL, 1750 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1751 CTLTYPE_INT, "random_id", 1752 SYSCTL_DESCR("Assign random ip_id values"), 1753 NULL, 0, &ip_do_randomid, 0, 1754 CTL_NET, PF_INET, IPPROTO_IP, 1755 IPCTL_RANDOMID, CTL_EOL); 1756 sysctl_createv(clog, 0, NULL, NULL, 1757 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1758 CTLTYPE_INT, "do_loopback_cksum", 1759 SYSCTL_DESCR("Perform IP checksum on loopback"), 1760 NULL, 0, &ip_do_loopback_cksum, 0, 1761 CTL_NET, PF_INET, IPPROTO_IP, 1762 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1763 sysctl_createv(clog, 0, NULL, NULL, 1764 CTLFLAG_PERMANENT, 1765 CTLTYPE_STRUCT, "stats", 1766 SYSCTL_DESCR("IP statistics"), 1767 sysctl_net_inet_ip_stats, 0, NULL, 0, 1768 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1769 CTL_EOL); 1770#if NARP 1771 sysctl_createv(clog, 0, NULL, NULL, 1772 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1773 CTLTYPE_INT, "dad_count", 1774 SYSCTL_DESCR("Number of Duplicate Address Detection " 1775 "probes to send"), 1776 NULL, 0, &ip_dad_count, 0, 1777 CTL_NET, PF_INET, IPPROTO_IP, 1778 IPCTL_DAD_COUNT, CTL_EOL); 1779#endif 1780 1781 /* anonportalgo RFC6056 subtree */ 1782 const struct sysctlnode *portalgo_node; 1783 sysctl_createv(clog, 0, NULL, &portalgo_node, 1784 CTLFLAG_PERMANENT, 1785 CTLTYPE_NODE, "anonportalgo", 1786 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1787 NULL, 0, NULL, 0, 1788 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1789 sysctl_createv(clog, 0, &portalgo_node, NULL, 1790 CTLFLAG_PERMANENT, 1791 CTLTYPE_STRING, "available", 1792 SYSCTL_DESCR("available algorithms"), 1793 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1794 CTL_CREATE, CTL_EOL); 1795 sysctl_createv(clog, 0, &portalgo_node, NULL, 1796 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1797 CTLTYPE_STRING, "selected", 1798 SYSCTL_DESCR("selected algorithm"), 1799 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1800 CTL_CREATE, CTL_EOL); 1801 sysctl_createv(clog, 0, &portalgo_node, NULL, 1802 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1803 CTLTYPE_STRUCT, "reserve", 1804 SYSCTL_DESCR("bitmap of reserved ports"), 1805 sysctl_portalgo_reserve4, 0, NULL, 0, 1806 CTL_CREATE, CTL_EOL); 1807} 1808 1809void 1810ip_statinc(u_int stat) 1811{ 1812 1813 KASSERT(stat < IP_NSTATS); 1814 IP_STATINC(stat); 1815} 1816