ip_input.c revision 1.362
1/* $NetBSD: ip_input.c,v 1.362 2017/11/17 07:37:12 ozaki-r 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 1.362 2017/11/17 07:37:12 ozaki-r 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 (ia->ia4_flags & IN_IFF_DETACHED) == 0) 376 break; 377 else 378 (*downmatch)++; 379 } 380 } 381 382 return ia; 383} 384 385static struct in_ifaddr * 386ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 387{ 388 struct in_ifaddr *ia = NULL; 389 struct ifaddr *ifa; 390 391 IFADDR_READER_FOREACH(ifa, ifp) { 392 if (ifa->ifa_addr->sa_family != AF_INET) 393 continue; 394 ia = ifatoia(ifa); 395 if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) 396 continue; 397 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 398 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 399 /* 400 * Look for all-0's host part (old broadcast addr), 401 * either for subnet or net. 402 */ 403 ip->ip_dst.s_addr == ia->ia_subnet || 404 ip->ip_dst.s_addr == ia->ia_net) 405 goto matched; 406 /* 407 * An interface with IP address zero accepts 408 * all packets that arrive on that interface. 409 */ 410 if (in_nullhost(ia->ia_addr.sin_addr)) 411 goto matched; 412 } 413 ia = NULL; 414 415matched: 416 return ia; 417} 418 419/* 420 * IP software interrupt routine. 421 */ 422static void 423ipintr(void *arg __unused) 424{ 425 struct mbuf *m; 426 427 KASSERT(cpu_softintr_p()); 428 429 SOFTNET_LOCK_UNLESS_NET_MPSAFE(); 430 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 431 ip_input(m); 432 } 433 SOFTNET_UNLOCK_UNLESS_NET_MPSAFE(); 434} 435 436/* 437 * IP input routine. Checksum and byte swap header. If fragmented 438 * try to reassemble. Process options. Pass to next level. 439 */ 440static void 441ip_input(struct mbuf *m) 442{ 443 struct ip *ip = NULL; 444 struct in_ifaddr *ia = NULL; 445 int hlen = 0, len; 446 int downmatch; 447 int srcrt = 0; 448 ifnet_t *ifp; 449 struct psref psref; 450 int s; 451 452 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 453 "interrupt handler; synchronization assumptions violated"); 454 455 MCLAIM(m, &ip_rx_mowner); 456 KASSERT((m->m_flags & M_PKTHDR) != 0); 457 458 ifp = m_get_rcvif_psref(m, &psref); 459 if (__predict_false(ifp == NULL)) 460 goto out; 461 462 /* 463 * If no IP addresses have been set yet but the interfaces 464 * are receiving, can't do anything with incoming packets yet. 465 * Note: we pre-check without locks held. 466 */ 467 if (IN_ADDRLIST_READER_EMPTY()) 468 goto out; 469 IP_STATINC(IP_STAT_TOTAL); 470 471 /* 472 * If the IP header is not aligned, slurp it up into a new 473 * mbuf with space for link headers, in the event we forward 474 * it. Otherwise, if it is aligned, make sure the entire 475 * base IP header is in the first mbuf of the chain. 476 */ 477 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 478 if ((m = m_copyup(m, sizeof(struct ip), 479 (max_linkhdr + 3) & ~3)) == NULL) { 480 /* XXXJRT new stat, please */ 481 IP_STATINC(IP_STAT_TOOSMALL); 482 goto out; 483 } 484 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 485 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 486 IP_STATINC(IP_STAT_TOOSMALL); 487 goto out; 488 } 489 } 490 ip = mtod(m, struct ip *); 491 if (ip->ip_v != IPVERSION) { 492 IP_STATINC(IP_STAT_BADVERS); 493 goto out; 494 } 495 hlen = ip->ip_hl << 2; 496 if (hlen < sizeof(struct ip)) { /* minimum header length */ 497 IP_STATINC(IP_STAT_BADHLEN); 498 goto out; 499 } 500 if (hlen > m->m_len) { 501 if ((m = m_pullup(m, hlen)) == NULL) { 502 IP_STATINC(IP_STAT_BADHLEN); 503 goto out; 504 } 505 ip = mtod(m, struct ip *); 506 } 507 508 /* 509 * RFC1122: packets with a multicast source address are 510 * not allowed. 511 */ 512 if (IN_MULTICAST(ip->ip_src.s_addr)) { 513 IP_STATINC(IP_STAT_BADADDR); 514 goto out; 515 } 516 517 /* 127/8 must not appear on wire - RFC1122 */ 518 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 519 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 520 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 521 IP_STATINC(IP_STAT_BADADDR); 522 goto out; 523 } 524 } 525 526 switch (m->m_pkthdr.csum_flags & 527 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | 528 M_CSUM_IPv4_BAD)) { 529 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 530 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 531 IP_STATINC(IP_STAT_BADSUM); 532 goto out; 533 534 case M_CSUM_IPv4: 535 /* Checksum was okay. */ 536 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 537 break; 538 539 default: 540 /* 541 * Must compute it ourselves. Maybe skip checksum on 542 * loopback interfaces. 543 */ 544 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 545 ip_do_loopback_cksum)) { 546 INET_CSUM_COUNTER_INCR(&ip_swcsum); 547 if (in_cksum(m, hlen) != 0) { 548 IP_STATINC(IP_STAT_BADSUM); 549 goto out; 550 } 551 } 552 break; 553 } 554 555 /* Retrieve the packet length. */ 556 len = ntohs(ip->ip_len); 557 558 /* 559 * Check for additional length bogosity 560 */ 561 if (len < hlen) { 562 IP_STATINC(IP_STAT_BADLEN); 563 goto out; 564 } 565 566 /* 567 * Check that the amount of data in the buffers 568 * is as at least much as the IP header would have us expect. 569 * Trim mbufs if longer than we expect. 570 * Drop packet if shorter than we expect. 571 */ 572 if (m->m_pkthdr.len < len) { 573 IP_STATINC(IP_STAT_TOOSHORT); 574 goto out; 575 } 576 if (m->m_pkthdr.len > len) { 577 if (m->m_len == m->m_pkthdr.len) { 578 m->m_len = len; 579 m->m_pkthdr.len = len; 580 } else 581 m_adj(m, len - m->m_pkthdr.len); 582 } 583 584 /* 585 * Assume that we can create a fast-forward IP flow entry 586 * based on this packet. 587 */ 588 m->m_flags |= M_CANFASTFWD; 589 590 /* 591 * Run through list of hooks for input packets. If there are any 592 * filters which require that additional packets in the flow are 593 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 594 * Note that filters must _never_ set this flag, as another filter 595 * in the list may have previously cleared it. 596 */ 597#if defined(IPSEC) 598 if (!ipsec_used || !ipsec_indone(m)) 599#else 600 if (1) 601#endif 602 { 603 struct in_addr odst = ip->ip_dst; 604 bool freed; 605 606 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 607 if (freed || m == NULL) { 608 m = NULL; 609 goto out; 610 } 611 ip = mtod(m, struct ip *); 612 hlen = ip->ip_hl << 2; 613 614 /* 615 * XXX The setting of "srcrt" here is to prevent ip_forward() 616 * from generating ICMP redirects for packets that have 617 * been redirected by a hook back out on to the same LAN that 618 * they came from and is not an indication that the packet 619 * is being inffluenced by source routing options. This 620 * allows things like 621 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 622 * where tlp0 is both on the 1.1.1.0/24 network and is the 623 * default route for hosts on 1.1.1.0/24. Of course this 624 * also requires a "map tlp0 ..." to complete the story. 625 * One might argue whether or not this kind of network config. 626 * should be supported in this manner... 627 */ 628 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 629 } 630 631#ifdef ALTQ 632 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 633 if (altq_input) { 634 SOFTNET_LOCK(); 635 if ((*altq_input)(m, AF_INET) == 0) { 636 /* Packet dropped by traffic conditioner. */ 637 SOFTNET_UNLOCK(); 638 m = NULL; 639 goto out; 640 } 641 SOFTNET_UNLOCK(); 642 } 643#endif 644 645 /* 646 * Process options and, if not destined for us, 647 * ship it on. ip_dooptions returns 1 when an 648 * error was detected (causing an icmp message 649 * to be sent and the original packet to be freed). 650 */ 651 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 652 m = NULL; 653 goto out; 654 } 655 656 /* 657 * Check our list of addresses, to see if the packet is for us. 658 * 659 * Traditional 4.4BSD did not consult IFF_UP at all. 660 * The behavior here is to treat addresses on !IFF_UP interface 661 * or IN_IFF_NOTREADY addresses as not mine. 662 */ 663 downmatch = 0; 664 s = pserialize_read_enter(); 665 ia = ip_match_our_address(ifp, ip, &downmatch); 666 if (ia != NULL) { 667 pserialize_read_exit(s); 668 goto ours; 669 } 670 671 if (ifp->if_flags & IFF_BROADCAST) { 672 ia = ip_match_our_address_broadcast(ifp, ip); 673 if (ia != NULL) { 674 pserialize_read_exit(s); 675 goto ours; 676 } 677 } 678 pserialize_read_exit(s); 679 680 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 681#ifdef MROUTING 682 extern struct socket *ip_mrouter; 683 684 if (ip_mrouter) { 685 /* 686 * If we are acting as a multicast router, all 687 * incoming multicast packets are passed to the 688 * kernel-level multicast forwarding function. 689 * The packet is returned (relatively) intact; if 690 * ip_mforward() returns a non-zero value, the packet 691 * must be discarded, else it may be accepted below. 692 * 693 * (The IP ident field is put in the same byte order 694 * as expected when ip_mforward() is called from 695 * ip_output().) 696 */ 697 SOFTNET_LOCK(); 698 if (ip_mforward(m, ifp) != 0) { 699 SOFTNET_UNLOCK(); 700 IP_STATINC(IP_STAT_CANTFORWARD); 701 goto out; 702 } 703 SOFTNET_UNLOCK(); 704 705 /* 706 * The process-level routing demon needs to receive 707 * all multicast IGMP packets, whether or not this 708 * host belongs to their destination groups. 709 */ 710 if (ip->ip_p == IPPROTO_IGMP) { 711 goto ours; 712 } 713 IP_STATINC(IP_STAT_CANTFORWARD); 714 } 715#endif 716 /* 717 * See if we belong to the destination multicast group on the 718 * arrival interface. 719 */ 720 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 721 IP_STATINC(IP_STAT_CANTFORWARD); 722 goto out; 723 } 724 goto ours; 725 } 726 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 727 in_nullhost(ip->ip_dst)) 728 goto ours; 729 730 /* 731 * Not for us; forward if possible and desirable. 732 */ 733 if (ipforwarding == 0) { 734 m_put_rcvif_psref(ifp, &psref); 735 IP_STATINC(IP_STAT_CANTFORWARD); 736 m_freem(m); 737 } else { 738 /* 739 * If ip_dst matched any of my address on !IFF_UP interface, 740 * and there's no IFF_UP interface that matches ip_dst, 741 * send icmp unreach. Forwarding it will result in in-kernel 742 * forwarding loop till TTL goes to 0. 743 */ 744 if (downmatch) { 745 m_put_rcvif_psref(ifp, &psref); 746 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 747 IP_STATINC(IP_STAT_CANTFORWARD); 748 return; 749 } 750#ifdef IPSEC 751 /* Check the security policy (SP) for the packet */ 752 if (ipsec_used) { 753 if (ipsec4_input(m, IP_FORWARDING | 754 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 755 goto out; 756 } 757 } 758#endif 759 ip_forward(m, srcrt, ifp); 760 m_put_rcvif_psref(ifp, &psref); 761 } 762 return; 763 764ours: 765 m_put_rcvif_psref(ifp, &psref); 766 ifp = NULL; 767 768 /* 769 * If offset or IP_MF are set, must reassemble. 770 */ 771 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 772 /* 773 * Pass to IP reassembly mechanism. 774 */ 775 if (ip_reass_packet(&m, ip) != 0) { 776 /* Failed; invalid fragment(s) or packet. */ 777 goto out; 778 } 779 if (m == NULL) { 780 /* More fragments should come; silently return. */ 781 goto out; 782 } 783 /* 784 * Reassembly is done, we have the final packet. 785 * Updated cached data in local variable(s). 786 */ 787 ip = mtod(m, struct ip *); 788 hlen = ip->ip_hl << 2; 789 } 790 791#ifdef IPSEC 792 /* 793 * Enforce IPsec policy checking if we are seeing last header. 794 * Note that we do not visit this with protocols with PCB layer 795 * code - like UDP/TCP/raw IP. 796 */ 797 if (ipsec_used && 798 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 799 if (ipsec4_input(m, 0) != 0) { 800 goto out; 801 } 802 } 803#endif 804 805 /* 806 * Switch out to protocol's input routine. 807 */ 808#if IFA_STATS 809 if (ia && ip) { 810 struct in_ifaddr *_ia; 811 /* 812 * Keep a reference from ip_match_our_address with psref 813 * is expensive, so explore ia here again. 814 */ 815 s = pserialize_read_enter(); 816 _ia = in_get_ia(ip->ip_dst); 817 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 818 pserialize_read_exit(s); 819 } 820#endif 821 IP_STATINC(IP_STAT_DELIVERED); 822 823 const int off = hlen, nh = ip->ip_p; 824 825 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 826 return; 827 828out: 829 m_put_rcvif_psref(ifp, &psref); 830 if (m != NULL) 831 m_freem(m); 832} 833 834/* 835 * IP timer processing. 836 */ 837void 838ip_slowtimo(void) 839{ 840 841 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 842 843 ip_reass_slowtimo(); 844 845 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 846} 847 848/* 849 * IP drain processing. 850 */ 851void 852ip_drain(void) 853{ 854 855 KERNEL_LOCK(1, NULL); 856 ip_reass_drain(); 857 KERNEL_UNLOCK_ONE(NULL); 858} 859 860/* 861 * ip_dooptions: perform option processing on a datagram, possibly discarding 862 * it if bad options are encountered, or forwarding it if source-routed. 863 * 864 * => Returns true if packet has been forwarded/freed. 865 * => Returns false if the packet should be processed further. 866 */ 867static bool 868ip_dooptions(struct mbuf *m) 869{ 870 struct ip *ip = mtod(m, struct ip *); 871 u_char *cp, *cp0; 872 struct ip_timestamp *ipt; 873 struct in_ifaddr *ia; 874 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 875 struct in_addr dst; 876 n_time ntime; 877 struct ifaddr *ifa = NULL; 878 int s; 879 880 dst = ip->ip_dst; 881 cp = (u_char *)(ip + 1); 882 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 883 for (; cnt > 0; cnt -= optlen, cp += optlen) { 884 opt = cp[IPOPT_OPTVAL]; 885 if (opt == IPOPT_EOL) 886 break; 887 if (opt == IPOPT_NOP) 888 optlen = 1; 889 else { 890 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 891 code = &cp[IPOPT_OLEN] - (u_char *)ip; 892 goto bad; 893 } 894 optlen = cp[IPOPT_OLEN]; 895 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 896 code = &cp[IPOPT_OLEN] - (u_char *)ip; 897 goto bad; 898 } 899 } 900 switch (opt) { 901 902 default: 903 break; 904 905 /* 906 * Source routing with record. 907 * Find interface with current destination address. 908 * If none on this machine then drop if strictly routed, 909 * or do nothing if loosely routed. 910 * Record interface address and bring up next address 911 * component. If strictly routed make sure next 912 * address is on directly accessible net. 913 */ 914 case IPOPT_LSRR: 915 case IPOPT_SSRR: { 916 struct psref psref; 917 struct sockaddr_in ipaddr = { 918 .sin_len = sizeof(ipaddr), 919 .sin_family = AF_INET, 920 }; 921 922 if (ip_allowsrcrt == 0) { 923 type = ICMP_UNREACH; 924 code = ICMP_UNREACH_NET_PROHIB; 925 goto bad; 926 } 927 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 928 code = &cp[IPOPT_OLEN] - (u_char *)ip; 929 goto bad; 930 } 931 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 932 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 933 goto bad; 934 } 935 ipaddr.sin_addr = ip->ip_dst; 936 937 s = pserialize_read_enter(); 938 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 939 if (ifa == NULL) { 940 pserialize_read_exit(s); 941 if (opt == IPOPT_SSRR) { 942 type = ICMP_UNREACH; 943 code = ICMP_UNREACH_SRCFAIL; 944 goto bad; 945 } 946 /* 947 * Loose routing, and not at next destination 948 * yet; nothing to do except forward. 949 */ 950 break; 951 } 952 pserialize_read_exit(s); 953 954 off--; /* 0 origin */ 955 if ((off + sizeof(struct in_addr)) > optlen) { 956 /* 957 * End of source route. Should be for us. 958 */ 959 save_rte(m, cp, ip->ip_src); 960 break; 961 } 962 /* 963 * locate outgoing interface 964 */ 965 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 966 sizeof(ipaddr.sin_addr)); 967 if (opt == IPOPT_SSRR) { 968 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 969 &psref); 970 if (ifa != NULL) 971 ia = ifatoia(ifa); 972 else 973 ia = NULL; 974 } else { 975 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 976 } 977 if (ia == NULL) { 978 type = ICMP_UNREACH; 979 code = ICMP_UNREACH_SRCFAIL; 980 goto bad; 981 } 982 ip->ip_dst = ipaddr.sin_addr; 983 bcopy((void *)&ia->ia_addr.sin_addr, 984 (void *)(cp + off), sizeof(struct in_addr)); 985 ia4_release(ia, &psref); 986 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 987 /* 988 * Let ip_intr's mcast routing check handle mcast pkts 989 */ 990 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 991 break; 992 } 993 994 case IPOPT_RR: { 995 struct psref psref; 996 struct sockaddr_in ipaddr = { 997 .sin_len = sizeof(ipaddr), 998 .sin_family = AF_INET, 999 }; 1000 1001 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1002 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1003 goto bad; 1004 } 1005 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1006 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1007 goto bad; 1008 } 1009 /* 1010 * If no space remains, ignore. 1011 */ 1012 off--; /* 0 origin */ 1013 if ((off + sizeof(struct in_addr)) > optlen) 1014 break; 1015 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1016 sizeof(ipaddr.sin_addr)); 1017 /* 1018 * locate outgoing interface; if we're the destination, 1019 * use the incoming interface (should be same). 1020 */ 1021 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1022 if (ifa == NULL) { 1023 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1024 if (ia == NULL) { 1025 type = ICMP_UNREACH; 1026 code = ICMP_UNREACH_HOST; 1027 goto bad; 1028 } 1029 } else { 1030 ia = ifatoia(ifa); 1031 } 1032 bcopy((void *)&ia->ia_addr.sin_addr, 1033 (void *)(cp + off), sizeof(struct in_addr)); 1034 ia4_release(ia, &psref); 1035 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1036 break; 1037 } 1038 1039 case IPOPT_TS: 1040 code = cp - (u_char *)ip; 1041 ipt = (struct ip_timestamp *)cp; 1042 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1043 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1044 goto bad; 1045 } 1046 if (ipt->ipt_ptr < 5) { 1047 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1048 goto bad; 1049 } 1050 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1051 if (++ipt->ipt_oflw == 0) { 1052 code = (u_char *)&ipt->ipt_ptr - 1053 (u_char *)ip; 1054 goto bad; 1055 } 1056 break; 1057 } 1058 cp0 = (cp + ipt->ipt_ptr - 1); 1059 switch (ipt->ipt_flg) { 1060 1061 case IPOPT_TS_TSONLY: 1062 break; 1063 1064 case IPOPT_TS_TSANDADDR: { 1065 struct ifnet *rcvif; 1066 int _s, _ss; 1067 struct sockaddr_in ipaddr = { 1068 .sin_len = sizeof(ipaddr), 1069 .sin_family = AF_INET, 1070 }; 1071 1072 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1073 sizeof(struct in_addr) > ipt->ipt_len) { 1074 code = (u_char *)&ipt->ipt_ptr - 1075 (u_char *)ip; 1076 goto bad; 1077 } 1078 ipaddr.sin_addr = dst; 1079 _ss = pserialize_read_enter(); 1080 rcvif = m_get_rcvif(m, &_s); 1081 if (__predict_true(rcvif != NULL)) { 1082 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1083 rcvif); 1084 } 1085 m_put_rcvif(rcvif, &_s); 1086 if (ifa == NULL) { 1087 pserialize_read_exit(_ss); 1088 break; 1089 } 1090 ia = ifatoia(ifa); 1091 bcopy(&ia->ia_addr.sin_addr, 1092 cp0, sizeof(struct in_addr)); 1093 pserialize_read_exit(_ss); 1094 ipt->ipt_ptr += sizeof(struct in_addr); 1095 break; 1096 } 1097 1098 case IPOPT_TS_PRESPEC: { 1099 struct sockaddr_in ipaddr = { 1100 .sin_len = sizeof(ipaddr), 1101 .sin_family = AF_INET, 1102 }; 1103 1104 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1105 sizeof(struct in_addr) > ipt->ipt_len) { 1106 code = (u_char *)&ipt->ipt_ptr - 1107 (u_char *)ip; 1108 goto bad; 1109 } 1110 memcpy(&ipaddr.sin_addr, cp0, 1111 sizeof(struct in_addr)); 1112 s = pserialize_read_enter(); 1113 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1114 if (ifa == NULL) { 1115 pserialize_read_exit(s); 1116 continue; 1117 } 1118 pserialize_read_exit(s); 1119 ipt->ipt_ptr += sizeof(struct in_addr); 1120 break; 1121 } 1122 1123 default: 1124 /* XXX can't take &ipt->ipt_flg */ 1125 code = (u_char *)&ipt->ipt_ptr - 1126 (u_char *)ip + 1; 1127 goto bad; 1128 } 1129 ntime = iptime(); 1130 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1131 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1132 sizeof(n_time)); 1133 ipt->ipt_ptr += sizeof(n_time); 1134 } 1135 } 1136 if (forward) { 1137 struct ifnet *rcvif; 1138 struct psref _psref; 1139 1140 if (ip_forwsrcrt == 0) { 1141 type = ICMP_UNREACH; 1142 code = ICMP_UNREACH_SRCFAIL; 1143 goto bad; 1144 } 1145 1146 rcvif = m_get_rcvif_psref(m, &_psref); 1147 if (__predict_false(rcvif == NULL)) { 1148 type = ICMP_UNREACH; 1149 code = ICMP_UNREACH_HOST; 1150 goto bad; 1151 } 1152 ip_forward(m, 1, rcvif); 1153 m_put_rcvif_psref(rcvif, &_psref); 1154 return true; 1155 } 1156 return false; 1157bad: 1158 icmp_error(m, type, code, 0, 0); 1159 IP_STATINC(IP_STAT_BADOPTIONS); 1160 return true; 1161} 1162 1163/* 1164 * ip_rtaddr: given address of next destination (final or next hop), 1165 * return internet address info of interface to be used to get there. 1166 */ 1167static struct in_ifaddr * 1168ip_rtaddr(struct in_addr dst, struct psref *psref) 1169{ 1170 struct rtentry *rt; 1171 union { 1172 struct sockaddr dst; 1173 struct sockaddr_in dst4; 1174 } u; 1175 struct route *ro; 1176 1177 sockaddr_in_init(&u.dst4, &dst, 0); 1178 1179 ro = percpu_getref(ipforward_rt_percpu); 1180 rt = rtcache_lookup(ro, &u.dst); 1181 if (rt == NULL) { 1182 percpu_putref(ipforward_rt_percpu); 1183 return NULL; 1184 } 1185 1186 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1187 rtcache_unref(rt, ro); 1188 percpu_putref(ipforward_rt_percpu); 1189 1190 return ifatoia(rt->rt_ifa); 1191} 1192 1193/* 1194 * save_rte: save incoming source route for use in replies, to be picked 1195 * up later by ip_srcroute if the receiver is interested. 1196 */ 1197static void 1198save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1199{ 1200 struct ip_srcrt *isr; 1201 struct m_tag *mtag; 1202 unsigned olen; 1203 1204 olen = option[IPOPT_OLEN]; 1205 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1206 return; 1207 1208 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1209 if (mtag == NULL) 1210 return; 1211 isr = (struct ip_srcrt *)(mtag + 1); 1212 1213 memcpy(isr->isr_hdr, option, olen); 1214 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1215 isr->isr_dst = dst; 1216 m_tag_prepend(m, mtag); 1217} 1218 1219/* 1220 * Retrieve incoming source route for use in replies, 1221 * in the same form used by setsockopt. 1222 * The first hop is placed before the options, will be removed later. 1223 */ 1224struct mbuf * 1225ip_srcroute(struct mbuf *m0) 1226{ 1227 struct in_addr *p, *q; 1228 struct mbuf *m; 1229 struct ip_srcrt *isr; 1230 struct m_tag *mtag; 1231 1232 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL); 1233 if (mtag == NULL) 1234 return NULL; 1235 isr = (struct ip_srcrt *)(mtag + 1); 1236 1237 if (isr->isr_nhops == 0) 1238 return NULL; 1239 1240 m = m_get(M_DONTWAIT, MT_SOOPTS); 1241 if (m == NULL) 1242 return NULL; 1243 1244 MCLAIM(m, &inetdomain.dom_mowner); 1245#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1246 1247 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1248 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1249 1250 /* 1251 * First save first hop for return route 1252 */ 1253 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1254 *(mtod(m, struct in_addr *)) = *p--; 1255 1256 /* 1257 * Copy option fields and padding (nop) to mbuf. 1258 */ 1259 isr->isr_nop = IPOPT_NOP; 1260 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1261 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1262 OPTSIZ); 1263 q = (struct in_addr *)(mtod(m, char *) + 1264 sizeof(struct in_addr) + OPTSIZ); 1265#undef OPTSIZ 1266 /* 1267 * Record return path as an IP source route, 1268 * reversing the path (pointers are now aligned). 1269 */ 1270 while (p >= isr->isr_routes) { 1271 *q++ = *p--; 1272 } 1273 /* 1274 * Last hop goes to final destination. 1275 */ 1276 *q = isr->isr_dst; 1277 m_tag_delete(m0, mtag); 1278 return m; 1279} 1280 1281const int inetctlerrmap[PRC_NCMDS] = { 1282 [PRC_MSGSIZE] = EMSGSIZE, 1283 [PRC_HOSTDEAD] = EHOSTDOWN, 1284 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1285 [PRC_UNREACH_NET] = EHOSTUNREACH, 1286 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1287 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1288 [PRC_UNREACH_PORT] = ECONNREFUSED, 1289 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1290 [PRC_PARAMPROB] = ENOPROTOOPT, 1291}; 1292 1293void 1294ip_fasttimo(void) 1295{ 1296 if (ip_drainwanted) { 1297 ip_drain(); 1298 ip_drainwanted = 0; 1299 } 1300} 1301 1302void 1303ip_drainstub(void) 1304{ 1305 ip_drainwanted = 1; 1306} 1307 1308/* 1309 * Forward a packet. If some error occurs return the sender 1310 * an icmp packet. Note we can't always generate a meaningful 1311 * icmp message because icmp doesn't have a large enough repertoire 1312 * of codes and types. 1313 * 1314 * If not forwarding, just drop the packet. This could be confusing 1315 * if ipforwarding was zero but some routing protocol was advancing 1316 * us as a gateway to somewhere. However, we must let the routing 1317 * protocol deal with that. 1318 * 1319 * The srcrt parameter indicates whether the packet is being forwarded 1320 * via a source route. 1321 */ 1322static void 1323ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1324{ 1325 struct ip *ip = mtod(m, struct ip *); 1326 struct rtentry *rt; 1327 int error, type = 0, code = 0, destmtu = 0; 1328 struct mbuf *mcopy; 1329 n_long dest; 1330 union { 1331 struct sockaddr dst; 1332 struct sockaddr_in dst4; 1333 } u; 1334 uint64_t *ips; 1335 struct route *ro; 1336 1337 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1338 "interrupt handler; synchronization assumptions violated"); 1339 1340 /* 1341 * We are now in the output path. 1342 */ 1343 MCLAIM(m, &ip_tx_mowner); 1344 1345 /* 1346 * Clear any in-bound checksum flags for this packet. 1347 */ 1348 m->m_pkthdr.csum_flags = 0; 1349 1350 dest = 0; 1351 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1352 IP_STATINC(IP_STAT_CANTFORWARD); 1353 m_freem(m); 1354 return; 1355 } 1356 1357 if (ip->ip_ttl <= IPTTLDEC) { 1358 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1359 return; 1360 } 1361 1362 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1363 1364 ro = percpu_getref(ipforward_rt_percpu); 1365 rt = rtcache_lookup(ro, &u.dst); 1366 if (rt == NULL) { 1367 percpu_putref(ipforward_rt_percpu); 1368 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1369 return; 1370 } 1371 1372 /* 1373 * Save at most 68 bytes of the packet in case 1374 * we need to generate an ICMP message to the src. 1375 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1376 */ 1377 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1378 if (mcopy) 1379 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1380 1381 ip->ip_ttl -= IPTTLDEC; 1382 1383 /* 1384 * If forwarding packet using same interface that it came in on, 1385 * perhaps should send a redirect to sender to shortcut a hop. 1386 * Only send redirect if source is sending directly to us, 1387 * and if packet was not source routed (or has any options). 1388 * Also, don't send redirect if forwarding using a default route 1389 * or a route modified by a redirect. 1390 */ 1391 if (rt->rt_ifp == rcvif && 1392 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1393 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1394 ipsendredirects && !srcrt) { 1395 if (rt->rt_ifa && 1396 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1397 ifatoia(rt->rt_ifa)->ia_subnet) { 1398 if (rt->rt_flags & RTF_GATEWAY) 1399 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1400 else 1401 dest = ip->ip_dst.s_addr; 1402 /* 1403 * Router requirements says to only send host 1404 * redirects. 1405 */ 1406 type = ICMP_REDIRECT; 1407 code = ICMP_REDIRECT_HOST; 1408 } 1409 } 1410 rtcache_unref(rt, ro); 1411 1412 error = ip_output(m, NULL, ro, 1413 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1414 NULL, NULL); 1415 1416 if (error) { 1417 IP_STATINC(IP_STAT_CANTFORWARD); 1418 goto error; 1419 } 1420 1421 ips = IP_STAT_GETREF(); 1422 ips[IP_STAT_FORWARD]++; 1423 1424 if (type) { 1425 ips[IP_STAT_REDIRECTSENT]++; 1426 IP_STAT_PUTREF(); 1427 goto redirect; 1428 } 1429 1430 IP_STAT_PUTREF(); 1431 if (mcopy) { 1432#ifdef GATEWAY 1433 if (mcopy->m_flags & M_CANFASTFWD) 1434 ipflow_create(ro, mcopy); 1435#endif 1436 m_freem(mcopy); 1437 } 1438 1439 percpu_putref(ipforward_rt_percpu); 1440 return; 1441 1442redirect: 1443error: 1444 if (mcopy == NULL) { 1445 percpu_putref(ipforward_rt_percpu); 1446 return; 1447 } 1448 1449 switch (error) { 1450 1451 case 0: /* forwarded, but need redirect */ 1452 /* type, code set above */ 1453 break; 1454 1455 case ENETUNREACH: /* shouldn't happen, checked above */ 1456 case EHOSTUNREACH: 1457 case ENETDOWN: 1458 case EHOSTDOWN: 1459 default: 1460 type = ICMP_UNREACH; 1461 code = ICMP_UNREACH_HOST; 1462 break; 1463 1464 case EMSGSIZE: 1465 type = ICMP_UNREACH; 1466 code = ICMP_UNREACH_NEEDFRAG; 1467 1468 if ((rt = rtcache_validate(ro)) != NULL) { 1469 destmtu = rt->rt_ifp->if_mtu; 1470 rtcache_unref(rt, ro); 1471 } 1472#ifdef IPSEC 1473 if (ipsec_used) 1474 (void)ipsec4_forward(mcopy, &destmtu); 1475#endif 1476 IP_STATINC(IP_STAT_CANTFRAG); 1477 break; 1478 1479 case ENOBUFS: 1480 /* 1481 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1482 * Requirements for IP Version 4 Routers. Source quench can 1483 * big problem under DoS attacks or if the underlying 1484 * interface is rate-limited. 1485 */ 1486 if (mcopy) 1487 m_freem(mcopy); 1488 percpu_putref(ipforward_rt_percpu); 1489 return; 1490 } 1491 icmp_error(mcopy, type, code, dest, destmtu); 1492 percpu_putref(ipforward_rt_percpu); 1493} 1494 1495void 1496ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1497 struct mbuf *m) 1498{ 1499 struct socket *so = inp->inp_socket; 1500 int inpflags = inp->inp_flags; 1501 1502 if (SOOPT_TIMESTAMP(so->so_options)) 1503 mp = sbsavetimestamp(so->so_options, m, mp); 1504 1505 if (inpflags & INP_RECVDSTADDR) { 1506 *mp = sbcreatecontrol(&ip->ip_dst, 1507 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1508 if (*mp) 1509 mp = &(*mp)->m_next; 1510 } 1511 1512 if (inpflags & INP_RECVTTL) { 1513 *mp = sbcreatecontrol(&ip->ip_ttl, 1514 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1515 if (*mp) 1516 mp = &(*mp)->m_next; 1517 } 1518 1519 struct psref psref; 1520 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1521 if (__predict_false(ifp == NULL)) { 1522#ifdef DIAGNOSTIC 1523 printf("%s: missing receive interface\n", __func__); 1524#endif 1525 return; /* XXX should report error? */ 1526 } 1527 1528 if (inpflags & INP_RECVPKTINFO) { 1529 struct in_pktinfo ipi; 1530 ipi.ipi_addr = ip->ip_src; 1531 ipi.ipi_ifindex = ifp->if_index; 1532 *mp = sbcreatecontrol(&ipi, 1533 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP); 1534 if (*mp) 1535 mp = &(*mp)->m_next; 1536 } 1537 if (inpflags & INP_PKTINFO) { 1538 struct in_pktinfo ipi; 1539 ipi.ipi_addr = ip->ip_dst; 1540 ipi.ipi_ifindex = ifp->if_index; 1541 *mp = sbcreatecontrol(&ipi, 1542 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1543 if (*mp) 1544 mp = &(*mp)->m_next; 1545 } 1546 if (inpflags & INP_RECVIF) { 1547 struct sockaddr_dl sdl; 1548 1549 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1550 NULL, 0); 1551 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1552 if (*mp) 1553 mp = &(*mp)->m_next; 1554 } 1555 m_put_rcvif_psref(ifp, &psref); 1556} 1557 1558/* 1559 * sysctl helper routine for net.inet.ip.forwsrcrt. 1560 */ 1561static int 1562sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1563{ 1564 int error, tmp; 1565 struct sysctlnode node; 1566 1567 node = *rnode; 1568 tmp = ip_forwsrcrt; 1569 node.sysctl_data = &tmp; 1570 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1571 if (error || newp == NULL) 1572 return (error); 1573 1574 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1575 0, NULL, NULL, NULL); 1576 if (error) 1577 return (error); 1578 1579 ip_forwsrcrt = tmp; 1580 1581 return (0); 1582} 1583 1584/* 1585 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1586 * range of the new value and tweaks timers if it changes. 1587 */ 1588static int 1589sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1590{ 1591 int error, tmp; 1592 struct sysctlnode node; 1593 1594 icmp_mtudisc_lock(); 1595 1596 node = *rnode; 1597 tmp = ip_mtudisc_timeout; 1598 node.sysctl_data = &tmp; 1599 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1600 if (error || newp == NULL) 1601 goto out; 1602 if (tmp < 0) { 1603 error = EINVAL; 1604 goto out; 1605 } 1606 1607 ip_mtudisc_timeout = tmp; 1608 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1609 error = 0; 1610out: 1611 icmp_mtudisc_unlock(); 1612 return error; 1613} 1614 1615static int 1616sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1617{ 1618 1619 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1620} 1621 1622static void 1623sysctl_net_inet_ip_setup(struct sysctllog **clog) 1624{ 1625 sysctl_createv(clog, 0, NULL, NULL, 1626 CTLFLAG_PERMANENT, 1627 CTLTYPE_NODE, "inet", 1628 SYSCTL_DESCR("PF_INET related settings"), 1629 NULL, 0, NULL, 0, 1630 CTL_NET, PF_INET, CTL_EOL); 1631 sysctl_createv(clog, 0, NULL, NULL, 1632 CTLFLAG_PERMANENT, 1633 CTLTYPE_NODE, "ip", 1634 SYSCTL_DESCR("IPv4 related settings"), 1635 NULL, 0, NULL, 0, 1636 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1637 1638 sysctl_createv(clog, 0, NULL, NULL, 1639 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1640 CTLTYPE_INT, "forwarding", 1641 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1642 NULL, 0, &ipforwarding, 0, 1643 CTL_NET, PF_INET, IPPROTO_IP, 1644 IPCTL_FORWARDING, CTL_EOL); 1645 sysctl_createv(clog, 0, NULL, NULL, 1646 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1647 CTLTYPE_INT, "redirect", 1648 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1649 NULL, 0, &ipsendredirects, 0, 1650 CTL_NET, PF_INET, IPPROTO_IP, 1651 IPCTL_SENDREDIRECTS, CTL_EOL); 1652 sysctl_createv(clog, 0, NULL, NULL, 1653 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1654 CTLTYPE_INT, "ttl", 1655 SYSCTL_DESCR("Default TTL for an INET datagram"), 1656 NULL, 0, &ip_defttl, 0, 1657 CTL_NET, PF_INET, IPPROTO_IP, 1658 IPCTL_DEFTTL, CTL_EOL); 1659#ifdef IPCTL_DEFMTU 1660 sysctl_createv(clog, 0, NULL, NULL, 1661 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1662 CTLTYPE_INT, "mtu", 1663 SYSCTL_DESCR("Default MTA for an INET route"), 1664 NULL, 0, &ip_mtu, 0, 1665 CTL_NET, PF_INET, IPPROTO_IP, 1666 IPCTL_DEFMTU, CTL_EOL); 1667#endif /* IPCTL_DEFMTU */ 1668 sysctl_createv(clog, 0, NULL, NULL, 1669 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1670 CTLTYPE_INT, "forwsrcrt", 1671 SYSCTL_DESCR("Enable forwarding of source-routed " 1672 "datagrams"), 1673 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1674 CTL_NET, PF_INET, IPPROTO_IP, 1675 IPCTL_FORWSRCRT, CTL_EOL); 1676 sysctl_createv(clog, 0, NULL, NULL, 1677 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1678 CTLTYPE_INT, "directed-broadcast", 1679 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1680 NULL, 0, &ip_directedbcast, 0, 1681 CTL_NET, PF_INET, IPPROTO_IP, 1682 IPCTL_DIRECTEDBCAST, CTL_EOL); 1683 sysctl_createv(clog, 0, NULL, NULL, 1684 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1685 CTLTYPE_INT, "allowsrcrt", 1686 SYSCTL_DESCR("Accept source-routed datagrams"), 1687 NULL, 0, &ip_allowsrcrt, 0, 1688 CTL_NET, PF_INET, IPPROTO_IP, 1689 IPCTL_ALLOWSRCRT, CTL_EOL); 1690 1691 sysctl_createv(clog, 0, NULL, NULL, 1692 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1693 CTLTYPE_INT, "mtudisc", 1694 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1695 NULL, 0, &ip_mtudisc, 0, 1696 CTL_NET, PF_INET, IPPROTO_IP, 1697 IPCTL_MTUDISC, CTL_EOL); 1698 sysctl_createv(clog, 0, NULL, NULL, 1699 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1700 CTLTYPE_INT, "anonportmin", 1701 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1702 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1703 CTL_NET, PF_INET, IPPROTO_IP, 1704 IPCTL_ANONPORTMIN, CTL_EOL); 1705 sysctl_createv(clog, 0, NULL, NULL, 1706 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1707 CTLTYPE_INT, "anonportmax", 1708 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1709 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1710 CTL_NET, PF_INET, IPPROTO_IP, 1711 IPCTL_ANONPORTMAX, CTL_EOL); 1712 sysctl_createv(clog, 0, NULL, NULL, 1713 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1714 CTLTYPE_INT, "mtudisctimeout", 1715 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1716 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1717 CTL_NET, PF_INET, IPPROTO_IP, 1718 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1719#ifndef IPNOPRIVPORTS 1720 sysctl_createv(clog, 0, NULL, NULL, 1721 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1722 CTLTYPE_INT, "lowportmin", 1723 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1724 "to assign"), 1725 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1726 CTL_NET, PF_INET, IPPROTO_IP, 1727 IPCTL_LOWPORTMIN, CTL_EOL); 1728 sysctl_createv(clog, 0, NULL, NULL, 1729 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1730 CTLTYPE_INT, "lowportmax", 1731 SYSCTL_DESCR("Highest privileged ephemeral port number " 1732 "to assign"), 1733 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1734 CTL_NET, PF_INET, IPPROTO_IP, 1735 IPCTL_LOWPORTMAX, CTL_EOL); 1736#endif /* IPNOPRIVPORTS */ 1737#if NGRE > 0 1738 sysctl_createv(clog, 0, NULL, NULL, 1739 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1740 CTLTYPE_INT, "grettl", 1741 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1742 NULL, 0, &ip_gre_ttl, 0, 1743 CTL_NET, PF_INET, IPPROTO_IP, 1744 IPCTL_GRE_TTL, CTL_EOL); 1745#endif /* NGRE */ 1746 sysctl_createv(clog, 0, NULL, NULL, 1747 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1748 CTLTYPE_INT, "checkinterface", 1749 SYSCTL_DESCR("Enable receive side of Strong ES model " 1750 "from RFC1122"), 1751 NULL, 0, &ip_checkinterface, 0, 1752 CTL_NET, PF_INET, IPPROTO_IP, 1753 IPCTL_CHECKINTERFACE, CTL_EOL); 1754 sysctl_createv(clog, 0, NULL, NULL, 1755 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1756 CTLTYPE_INT, "random_id", 1757 SYSCTL_DESCR("Assign random ip_id values"), 1758 NULL, 0, &ip_do_randomid, 0, 1759 CTL_NET, PF_INET, IPPROTO_IP, 1760 IPCTL_RANDOMID, CTL_EOL); 1761 sysctl_createv(clog, 0, NULL, NULL, 1762 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1763 CTLTYPE_INT, "do_loopback_cksum", 1764 SYSCTL_DESCR("Perform IP checksum on loopback"), 1765 NULL, 0, &ip_do_loopback_cksum, 0, 1766 CTL_NET, PF_INET, IPPROTO_IP, 1767 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1768 sysctl_createv(clog, 0, NULL, NULL, 1769 CTLFLAG_PERMANENT, 1770 CTLTYPE_STRUCT, "stats", 1771 SYSCTL_DESCR("IP statistics"), 1772 sysctl_net_inet_ip_stats, 0, NULL, 0, 1773 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1774 CTL_EOL); 1775#if NARP 1776 sysctl_createv(clog, 0, NULL, NULL, 1777 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1778 CTLTYPE_INT, "dad_count", 1779 SYSCTL_DESCR("Number of Duplicate Address Detection " 1780 "probes to send"), 1781 NULL, 0, &ip_dad_count, 0, 1782 CTL_NET, PF_INET, IPPROTO_IP, 1783 IPCTL_DAD_COUNT, CTL_EOL); 1784#endif 1785 1786 /* anonportalgo RFC6056 subtree */ 1787 const struct sysctlnode *portalgo_node; 1788 sysctl_createv(clog, 0, NULL, &portalgo_node, 1789 CTLFLAG_PERMANENT, 1790 CTLTYPE_NODE, "anonportalgo", 1791 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1792 NULL, 0, NULL, 0, 1793 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1794 sysctl_createv(clog, 0, &portalgo_node, NULL, 1795 CTLFLAG_PERMANENT, 1796 CTLTYPE_STRING, "available", 1797 SYSCTL_DESCR("available algorithms"), 1798 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1799 CTL_CREATE, CTL_EOL); 1800 sysctl_createv(clog, 0, &portalgo_node, NULL, 1801 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1802 CTLTYPE_STRING, "selected", 1803 SYSCTL_DESCR("selected algorithm"), 1804 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1805 CTL_CREATE, CTL_EOL); 1806 sysctl_createv(clog, 0, &portalgo_node, NULL, 1807 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1808 CTLTYPE_STRUCT, "reserve", 1809 SYSCTL_DESCR("bitmap of reserved ports"), 1810 sysctl_portalgo_reserve4, 0, NULL, 0, 1811 CTL_CREATE, CTL_EOL); 1812} 1813 1814void 1815ip_statinc(u_int stat) 1816{ 1817 1818 KASSERT(stat < IP_NSTATS); 1819 IP_STATINC(stat); 1820} 1821