ip_input.c revision 1.369
1/* $NetBSD: ip_input.c,v 1.369 2018/02/05 14:23:38 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 1.369 2018/02/05 14:23:38 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 196int ip_do_randomid = 0; 197 198/* 199 * XXX - Setting ip_checkinterface mostly implements the receive side of 200 * the Strong ES model described in RFC 1122, but since the routing table 201 * and transmit implementation do not implement the Strong ES model, 202 * setting this to 1 results in an odd hybrid. 203 * 204 * XXX - ip_checkinterface currently must be disabled if you use ipnat 205 * to translate the destination address to another local interface. 206 * 207 * XXX - ip_checkinterface must be disabled if you add IP aliases 208 * to the loopback interface instead of the interface where the 209 * packets for those addresses are received. 210 */ 211static int ip_checkinterface __read_mostly = 0; 212 213struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 214 215pktqueue_t * ip_pktq __read_mostly; 216pfil_head_t * inet_pfil_hook __read_mostly; 217ipid_state_t * ip_ids __read_mostly; 218percpu_t * ipstat_percpu __read_mostly; 219 220static percpu_t *ipforward_rt_percpu __cacheline_aligned; 221 222uint16_t ip_id; 223 224#ifdef INET_CSUM_COUNTERS 225#include <sys/device.h> 226 227struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 228 NULL, "inet", "hwcsum bad"); 229struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 230 NULL, "inet", "hwcsum ok"); 231struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 232 NULL, "inet", "swcsum"); 233 234#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 235 236EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 237EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 238EVCNT_ATTACH_STATIC(ip_swcsum); 239 240#else 241 242#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 243 244#endif /* INET_CSUM_COUNTERS */ 245 246/* 247 * Used to save the IP options in case a protocol wants to respond 248 * to an incoming packet over the same route if the packet got here 249 * using IP source routing. This allows connection establishment and 250 * maintenance when the remote end is on a network that is not known 251 * to us. 252 */ 253struct ip_srcrt { 254 int isr_nhops; /* number of hops */ 255 struct in_addr isr_dst; /* final destination */ 256 char isr_nop; /* one NOP to align */ 257 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */ 258 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)]; 259}; 260 261static int ip_drainwanted; 262 263static void save_rte(struct mbuf *, u_char *, struct in_addr); 264 265#ifdef MBUFTRACE 266struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 267struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 268#endif 269 270static void ipintr(void *); 271static void ip_input(struct mbuf *); 272static void ip_forward(struct mbuf *, int, struct ifnet *); 273static bool ip_dooptions(struct mbuf *); 274static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *); 275static void sysctl_net_inet_ip_setup(struct sysctllog **); 276 277static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *, 278 int *); 279static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *, 280 struct ip *); 281 282#ifdef NET_MPSAFE 283#define SOFTNET_LOCK() mutex_enter(softnet_lock) 284#define SOFTNET_UNLOCK() mutex_exit(softnet_lock) 285#else 286#define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 287#define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 288#endif 289 290/* 291 * IP initialization: fill in IP protocol switch table. 292 * All protocols not implemented in kernel go to raw IP protocol handler. 293 */ 294void 295ip_init(void) 296{ 297 const struct protosw *pr; 298 299 in_init(); 300 sysctl_net_inet_ip_setup(NULL); 301 302 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 303 KASSERT(pr != NULL); 304 305 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 306 KASSERT(ip_pktq != NULL); 307 308 for (u_int i = 0; i < IPPROTO_MAX; i++) { 309 ip_protox[i] = pr - inetsw; 310 } 311 for (pr = inetdomain.dom_protosw; 312 pr < inetdomain.dom_protoswNPROTOSW; pr++) 313 if (pr->pr_domain->dom_family == PF_INET && 314 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 315 ip_protox[pr->pr_protocol] = pr - inetsw; 316 317 ip_reass_init(); 318 319 ip_ids = ip_id_init(); 320 ip_id = time_uptime & 0xfffff; 321 322#ifdef GATEWAY 323 ipflow_init(); 324#endif 325 326 /* Register our Packet Filter hook. */ 327 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 328 KASSERT(inet_pfil_hook != NULL); 329 330#ifdef MBUFTRACE 331 MOWNER_ATTACH(&ip_tx_mowner); 332 MOWNER_ATTACH(&ip_rx_mowner); 333#endif /* MBUFTRACE */ 334 335 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 336 ipforward_rt_percpu = percpu_alloc(sizeof(struct route)); 337 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 338} 339 340static struct in_ifaddr * 341ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 342{ 343 struct in_ifaddr *ia = NULL; 344 int checkif; 345 346 /* 347 * Enable a consistency check between the destination address 348 * and the arrival interface for a unicast packet (the RFC 1122 349 * strong ES model) if IP forwarding is disabled and the packet 350 * is not locally generated. 351 * 352 * XXX - Checking also should be disabled if the destination 353 * address is ipnat'ed to a different interface. 354 * 355 * XXX - Checking is incompatible with IP aliases added 356 * to the loopback interface instead of the interface where 357 * the packets are received. 358 * 359 * XXX - We need to add a per ifaddr flag for this so that 360 * we get finer grain control. 361 */ 362 checkif = ip_checkinterface && (ipforwarding == 0) && 363 (ifp->if_flags & IFF_LOOPBACK) == 0; 364 365 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 366 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 367 if (ia->ia4_flags & IN_IFF_NOTREADY) 368 continue; 369 if (checkif && ia->ia_ifp != ifp) 370 continue; 371 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) { 372 (*downmatch)++; 373 continue; 374 } 375 if (ia->ia4_flags & IN_IFF_DETACHED && 376 (ifp->if_flags & IFF_LOOPBACK) == 0) 377 continue; 378 break; 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) 396 continue; 397 if (ia->ia4_flags & IN_IFF_DETACHED && 398 (ifp->if_flags & IFF_LOOPBACK) == 0) 399 continue; 400 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 401 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 402 /* 403 * Look for all-0's host part (old broadcast addr), 404 * either for subnet or net. 405 */ 406 ip->ip_dst.s_addr == ia->ia_subnet || 407 ip->ip_dst.s_addr == ia->ia_net) 408 goto matched; 409 /* 410 * An interface with IP address zero accepts 411 * all packets that arrive on that interface. 412 */ 413 if (in_nullhost(ia->ia_addr.sin_addr)) 414 goto matched; 415 } 416 ia = NULL; 417 418matched: 419 return ia; 420} 421 422/* 423 * IP software interrupt routine. 424 */ 425static void 426ipintr(void *arg __unused) 427{ 428 struct mbuf *m; 429 430 KASSERT(cpu_softintr_p()); 431 432 SOFTNET_LOCK_UNLESS_NET_MPSAFE(); 433 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 434 ip_input(m); 435 } 436 SOFTNET_UNLOCK_UNLESS_NET_MPSAFE(); 437} 438 439/* 440 * IP input routine. Checksum and byte swap header. If fragmented 441 * try to reassemble. Process options. Pass to next level. 442 */ 443static void 444ip_input(struct mbuf *m) 445{ 446 struct ip *ip = NULL; 447 struct in_ifaddr *ia = NULL; 448 int hlen = 0, len; 449 int downmatch; 450 int srcrt = 0; 451 ifnet_t *ifp; 452 struct psref psref; 453 int s; 454 455 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 456 "interrupt handler; synchronization assumptions violated"); 457 458 MCLAIM(m, &ip_rx_mowner); 459 KASSERT((m->m_flags & M_PKTHDR) != 0); 460 461 ifp = m_get_rcvif_psref(m, &psref); 462 if (__predict_false(ifp == NULL)) 463 goto out; 464 465 /* 466 * If no IP addresses have been set yet but the interfaces 467 * are receiving, can't do anything with incoming packets yet. 468 * Note: we pre-check without locks held. 469 */ 470 if (IN_ADDRLIST_READER_EMPTY()) 471 goto out; 472 IP_STATINC(IP_STAT_TOTAL); 473 474 /* 475 * If the IP header is not aligned, slurp it up into a new 476 * mbuf with space for link headers, in the event we forward 477 * it. Otherwise, if it is aligned, make sure the entire 478 * base IP header is in the first mbuf of the chain. 479 */ 480 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 481 if ((m = m_copyup(m, sizeof(struct ip), 482 (max_linkhdr + 3) & ~3)) == NULL) { 483 /* XXXJRT new stat, please */ 484 IP_STATINC(IP_STAT_TOOSMALL); 485 goto out; 486 } 487 } else if (__predict_false(m->m_len < sizeof(struct ip))) { 488 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) { 489 IP_STATINC(IP_STAT_TOOSMALL); 490 goto out; 491 } 492 } 493 ip = mtod(m, struct ip *); 494 if (ip->ip_v != IPVERSION) { 495 IP_STATINC(IP_STAT_BADVERS); 496 goto out; 497 } 498 hlen = ip->ip_hl << 2; 499 if (hlen < sizeof(struct ip)) { /* minimum header length */ 500 IP_STATINC(IP_STAT_BADHLEN); 501 goto out; 502 } 503 if (hlen > m->m_len) { 504 if ((m = m_pullup(m, hlen)) == NULL) { 505 IP_STATINC(IP_STAT_BADHLEN); 506 goto out; 507 } 508 ip = mtod(m, struct ip *); 509 } 510 511 /* 512 * RFC1122: packets with a multicast source address are 513 * not allowed. 514 */ 515 if (IN_MULTICAST(ip->ip_src.s_addr)) { 516 IP_STATINC(IP_STAT_BADADDR); 517 goto out; 518 } 519 520 /* 127/8 must not appear on wire - RFC1122 */ 521 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 522 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 523 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 524 IP_STATINC(IP_STAT_BADADDR); 525 goto out; 526 } 527 } 528 529 switch (m->m_pkthdr.csum_flags & 530 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) { 531 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 532 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 533 IP_STATINC(IP_STAT_BADSUM); 534 goto out; 535 536 case M_CSUM_IPv4: 537 /* Checksum was okay. */ 538 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 539 break; 540 541 default: 542 /* 543 * Must compute it ourselves. Maybe skip checksum on 544 * loopback interfaces. 545 */ 546 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 547 ip_do_loopback_cksum)) { 548 INET_CSUM_COUNTER_INCR(&ip_swcsum); 549 if (in_cksum(m, hlen) != 0) { 550 IP_STATINC(IP_STAT_BADSUM); 551 goto out; 552 } 553 } 554 break; 555 } 556 557 /* Retrieve the packet length. */ 558 len = ntohs(ip->ip_len); 559 560 /* 561 * Check for additional length bogosity 562 */ 563 if (len < hlen) { 564 IP_STATINC(IP_STAT_BADLEN); 565 goto out; 566 } 567 568 /* 569 * Check that the amount of data in the buffers is at least as much 570 * as the IP header would have us expect. Trim mbufs if longer than 571 * we expect. Drop packet if shorter than we expect. 572 */ 573 if (m->m_pkthdr.len < len) { 574 IP_STATINC(IP_STAT_TOOSHORT); 575 goto out; 576 } 577 if (m->m_pkthdr.len > len) { 578 if (m->m_len == m->m_pkthdr.len) { 579 m->m_len = len; 580 m->m_pkthdr.len = len; 581 } else 582 m_adj(m, len - m->m_pkthdr.len); 583 } 584 585 /* 586 * Assume that we can create a fast-forward IP flow entry 587 * based on this packet. 588 */ 589 m->m_flags |= M_CANFASTFWD; 590 591 /* 592 * Run through list of hooks for input packets. If there are any 593 * filters which require that additional packets in the flow are 594 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 595 * Note that filters must _never_ set this flag, as another filter 596 * in the list may have previously cleared it. 597 */ 598#if defined(IPSEC) 599 if (!ipsec_used || !ipsec_indone(m)) 600#else 601 if (1) 602#endif 603 { 604 struct in_addr odst = ip->ip_dst; 605 bool freed; 606 607 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 608 if (freed || m == NULL) { 609 m = NULL; 610 goto out; 611 } 612 ip = mtod(m, struct ip *); 613 hlen = ip->ip_hl << 2; 614 615 /* 616 * XXX The setting of "srcrt" here is to prevent ip_forward() 617 * from generating ICMP redirects for packets that have 618 * been redirected by a hook back out on to the same LAN that 619 * they came from and is not an indication that the packet 620 * is being inffluenced by source routing options. This 621 * allows things like 622 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 623 * where tlp0 is both on the 1.1.1.0/24 network and is the 624 * default route for hosts on 1.1.1.0/24. Of course this 625 * also requires a "map tlp0 ..." to complete the story. 626 * One might argue whether or not this kind of network config. 627 * should be supported in this manner... 628 */ 629 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 630 } 631 632#ifdef ALTQ 633 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 634 if (altq_input) { 635 SOFTNET_LOCK(); 636 if ((*altq_input)(m, AF_INET) == 0) { 637 /* Packet dropped by traffic conditioner. */ 638 SOFTNET_UNLOCK(); 639 m = NULL; 640 goto out; 641 } 642 SOFTNET_UNLOCK(); 643 } 644#endif 645 646 /* 647 * Process options and, if not destined for us, 648 * ship it on. ip_dooptions returns 1 when an 649 * error was detected (causing an icmp message 650 * to be sent and the original packet to be freed). 651 */ 652 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 653 m = NULL; 654 goto out; 655 } 656 657 /* 658 * Check our list of addresses, to see if the packet is for us. 659 * 660 * Traditional 4.4BSD did not consult IFF_UP at all. 661 * The behavior here is to treat addresses on !IFF_UP interface 662 * or IN_IFF_NOTREADY addresses as not mine. 663 */ 664 downmatch = 0; 665 s = pserialize_read_enter(); 666 ia = ip_match_our_address(ifp, ip, &downmatch); 667 if (ia != NULL) { 668 pserialize_read_exit(s); 669 goto ours; 670 } 671 672 if (ifp->if_flags & IFF_BROADCAST) { 673 ia = ip_match_our_address_broadcast(ifp, ip); 674 if (ia != NULL) { 675 pserialize_read_exit(s); 676 goto ours; 677 } 678 } 679 pserialize_read_exit(s); 680 681 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 682#ifdef MROUTING 683 extern struct socket *ip_mrouter; 684 685 if (ip_mrouter) { 686 /* 687 * If we are acting as a multicast router, all 688 * incoming multicast packets are passed to the 689 * kernel-level multicast forwarding function. 690 * The packet is returned (relatively) intact; if 691 * ip_mforward() returns a non-zero value, the packet 692 * must be discarded, else it may be accepted below. 693 * 694 * (The IP ident field is put in the same byte order 695 * as expected when ip_mforward() is called from 696 * ip_output().) 697 */ 698 SOFTNET_LOCK(); 699 if (ip_mforward(m, ifp) != 0) { 700 SOFTNET_UNLOCK(); 701 IP_STATINC(IP_STAT_CANTFORWARD); 702 goto out; 703 } 704 SOFTNET_UNLOCK(); 705 706 /* 707 * The process-level routing demon needs to receive 708 * all multicast IGMP packets, whether or not this 709 * host belongs to their destination groups. 710 */ 711 if (ip->ip_p == IPPROTO_IGMP) { 712 goto ours; 713 } 714 IP_STATINC(IP_STAT_CANTFORWARD); 715 } 716#endif 717 /* 718 * See if we belong to the destination multicast group on the 719 * arrival interface. 720 */ 721 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 722 IP_STATINC(IP_STAT_CANTFORWARD); 723 goto out; 724 } 725 goto ours; 726 } 727 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 728 in_nullhost(ip->ip_dst)) 729 goto ours; 730 731 /* 732 * Not for us; forward if possible and desirable. 733 */ 734 if (ipforwarding == 0) { 735 m_put_rcvif_psref(ifp, &psref); 736 IP_STATINC(IP_STAT_CANTFORWARD); 737 m_freem(m); 738 } else { 739 /* 740 * If ip_dst matched any of my address on !IFF_UP interface, 741 * and there's no IFF_UP interface that matches ip_dst, 742 * send icmp unreach. Forwarding it will result in in-kernel 743 * forwarding loop till TTL goes to 0. 744 */ 745 if (downmatch) { 746 m_put_rcvif_psref(ifp, &psref); 747 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 748 IP_STATINC(IP_STAT_CANTFORWARD); 749 return; 750 } 751#ifdef IPSEC 752 /* Check the security policy (SP) for the packet */ 753 if (ipsec_used) { 754 if (ipsec4_input(m, IP_FORWARDING | 755 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 756 goto out; 757 } 758 } 759#endif 760 ip_forward(m, srcrt, ifp); 761 m_put_rcvif_psref(ifp, &psref); 762 } 763 return; 764 765ours: 766 m_put_rcvif_psref(ifp, &psref); 767 ifp = NULL; 768 769 /* 770 * If offset or IP_MF are set, must reassemble. 771 */ 772 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 773 /* 774 * Pass to IP reassembly mechanism. 775 */ 776 if (ip_reass_packet(&m, ip) != 0) { 777 /* Failed; invalid fragment(s) or packet. */ 778 goto out; 779 } 780 if (m == NULL) { 781 /* More fragments should come; silently return. */ 782 goto out; 783 } 784 /* 785 * Reassembly is done, we have the final packet. 786 * Update cached data in local variable(s). 787 */ 788 ip = mtod(m, struct ip *); 789 hlen = ip->ip_hl << 2; 790 } 791 792#ifdef IPSEC 793 /* 794 * Enforce IPsec policy checking if we are seeing last header. 795 * Note that we do not visit this with protocols with PCB layer 796 * code - like UDP/TCP/raw IP. 797 */ 798 if (ipsec_used && 799 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 800 if (ipsec4_input(m, 0) != 0) { 801 goto out; 802 } 803 } 804#endif 805 806 /* 807 * Switch out to protocol's input routine. 808 */ 809#if IFA_STATS 810 if (ia && ip) { 811 struct in_ifaddr *_ia; 812 /* 813 * Keep a reference from ip_match_our_address with psref 814 * is expensive, so explore ia here again. 815 */ 816 s = pserialize_read_enter(); 817 _ia = in_get_ia(ip->ip_dst); 818 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 819 pserialize_read_exit(s); 820 } 821#endif 822 IP_STATINC(IP_STAT_DELIVERED); 823 824 const int off = hlen, nh = ip->ip_p; 825 826 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 827 return; 828 829out: 830 m_put_rcvif_psref(ifp, &psref); 831 if (m != NULL) 832 m_freem(m); 833} 834 835/* 836 * IP timer processing. 837 */ 838void 839ip_slowtimo(void) 840{ 841 842 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 843 844 ip_reass_slowtimo(); 845 846 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 847} 848 849/* 850 * IP drain processing. 851 */ 852void 853ip_drain(void) 854{ 855 856 KERNEL_LOCK(1, NULL); 857 ip_reass_drain(); 858 KERNEL_UNLOCK_ONE(NULL); 859} 860 861/* 862 * ip_dooptions: perform option processing on a datagram, possibly discarding 863 * it if bad options are encountered, or forwarding it if source-routed. 864 * 865 * => Returns true if packet has been forwarded/freed. 866 * => Returns false if the packet should be processed further. 867 */ 868static bool 869ip_dooptions(struct mbuf *m) 870{ 871 struct ip *ip = mtod(m, struct ip *); 872 u_char *cp, *cp0; 873 struct ip_timestamp *ipt; 874 struct in_ifaddr *ia; 875 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 876 int srr_present, rr_present, ts_present; 877 struct in_addr dst; 878 n_time ntime; 879 struct ifaddr *ifa = NULL; 880 int s; 881 882 srr_present = 0; 883 rr_present = 0; 884 ts_present = 0; 885 886 dst = ip->ip_dst; 887 cp = (u_char *)(ip + 1); 888 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 889 for (; cnt > 0; cnt -= optlen, cp += optlen) { 890 opt = cp[IPOPT_OPTVAL]; 891 if (opt == IPOPT_EOL) 892 break; 893 if (opt == IPOPT_NOP) 894 optlen = 1; 895 else { 896 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 897 code = &cp[IPOPT_OLEN] - (u_char *)ip; 898 goto bad; 899 } 900 optlen = cp[IPOPT_OLEN]; 901 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 902 code = &cp[IPOPT_OLEN] - (u_char *)ip; 903 goto bad; 904 } 905 } 906 switch (opt) { 907 908 default: 909 break; 910 911 /* 912 * Source routing with record. 913 * Find interface with current destination address. 914 * If none on this machine then drop if strictly routed, 915 * or do nothing if loosely routed. 916 * Record interface address and bring up next address 917 * component. If strictly routed make sure next 918 * address is on directly accessible net. 919 */ 920 case IPOPT_LSRR: 921 case IPOPT_SSRR: { 922 struct psref psref; 923 struct sockaddr_in ipaddr = { 924 .sin_len = sizeof(ipaddr), 925 .sin_family = AF_INET, 926 }; 927 928 if (ip_allowsrcrt == 0) { 929 type = ICMP_UNREACH; 930 code = ICMP_UNREACH_NET_PROHIB; 931 goto bad; 932 } 933 if (srr_present++) { 934 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 935 goto bad; 936 } 937 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 938 code = &cp[IPOPT_OLEN] - (u_char *)ip; 939 goto bad; 940 } 941 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 942 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 943 goto bad; 944 } 945 ipaddr.sin_addr = ip->ip_dst; 946 947 s = pserialize_read_enter(); 948 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 949 if (ifa == NULL) { 950 pserialize_read_exit(s); 951 if (opt == IPOPT_SSRR) { 952 type = ICMP_UNREACH; 953 code = ICMP_UNREACH_SRCFAIL; 954 goto bad; 955 } 956 /* 957 * Loose routing, and not at next destination 958 * yet; nothing to do except forward. 959 */ 960 break; 961 } 962 pserialize_read_exit(s); 963 964 off--; /* 0 origin */ 965 if ((off + sizeof(struct in_addr)) > optlen) { 966 /* 967 * End of source route. Should be for us. 968 */ 969 save_rte(m, cp, ip->ip_src); 970 break; 971 } 972 /* 973 * locate outgoing interface 974 */ 975 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 976 sizeof(ipaddr.sin_addr)); 977 if (opt == IPOPT_SSRR) { 978 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 979 &psref); 980 if (ifa != NULL) 981 ia = ifatoia(ifa); 982 else 983 ia = NULL; 984 } else { 985 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 986 } 987 if (ia == NULL) { 988 type = ICMP_UNREACH; 989 code = ICMP_UNREACH_SRCFAIL; 990 goto bad; 991 } 992 ip->ip_dst = ipaddr.sin_addr; 993 bcopy((void *)&ia->ia_addr.sin_addr, 994 (void *)(cp + off), sizeof(struct in_addr)); 995 ia4_release(ia, &psref); 996 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 997 /* 998 * Let ip_intr's mcast routing check handle mcast pkts 999 */ 1000 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1001 break; 1002 } 1003 1004 case IPOPT_RR: { 1005 struct psref psref; 1006 struct sockaddr_in ipaddr = { 1007 .sin_len = sizeof(ipaddr), 1008 .sin_family = AF_INET, 1009 }; 1010 1011 if (rr_present++) { 1012 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1013 goto bad; 1014 } 1015 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1016 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1017 goto bad; 1018 } 1019 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1020 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1021 goto bad; 1022 } 1023 /* 1024 * If no space remains, ignore. 1025 */ 1026 off--; /* 0 origin */ 1027 if ((off + sizeof(struct in_addr)) > optlen) 1028 break; 1029 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1030 sizeof(ipaddr.sin_addr)); 1031 /* 1032 * locate outgoing interface; if we're the destination, 1033 * use the incoming interface (should be same). 1034 */ 1035 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1036 if (ifa == NULL) { 1037 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1038 if (ia == NULL) { 1039 type = ICMP_UNREACH; 1040 code = ICMP_UNREACH_HOST; 1041 goto bad; 1042 } 1043 } else { 1044 ia = ifatoia(ifa); 1045 } 1046 bcopy((void *)&ia->ia_addr.sin_addr, 1047 (void *)(cp + off), sizeof(struct in_addr)); 1048 ia4_release(ia, &psref); 1049 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1050 break; 1051 } 1052 1053 case IPOPT_TS: 1054 code = cp - (u_char *)ip; 1055 ipt = (struct ip_timestamp *)cp; 1056 if (ts_present++) { 1057 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1058 goto bad; 1059 } 1060 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1061 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1062 goto bad; 1063 } 1064 if (ipt->ipt_ptr < 5) { 1065 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1066 goto bad; 1067 } 1068 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1069 if (++ipt->ipt_oflw == 0) { 1070 code = (u_char *)&ipt->ipt_ptr - 1071 (u_char *)ip; 1072 goto bad; 1073 } 1074 break; 1075 } 1076 cp0 = (cp + ipt->ipt_ptr - 1); 1077 switch (ipt->ipt_flg) { 1078 1079 case IPOPT_TS_TSONLY: 1080 break; 1081 1082 case IPOPT_TS_TSANDADDR: { 1083 struct ifnet *rcvif; 1084 int _s, _ss; 1085 struct sockaddr_in ipaddr = { 1086 .sin_len = sizeof(ipaddr), 1087 .sin_family = AF_INET, 1088 }; 1089 1090 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1091 sizeof(struct in_addr) > ipt->ipt_len) { 1092 code = (u_char *)&ipt->ipt_ptr - 1093 (u_char *)ip; 1094 goto bad; 1095 } 1096 ipaddr.sin_addr = dst; 1097 _ss = pserialize_read_enter(); 1098 rcvif = m_get_rcvif(m, &_s); 1099 if (__predict_true(rcvif != NULL)) { 1100 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1101 rcvif); 1102 } 1103 m_put_rcvif(rcvif, &_s); 1104 if (ifa == NULL) { 1105 pserialize_read_exit(_ss); 1106 break; 1107 } 1108 ia = ifatoia(ifa); 1109 bcopy(&ia->ia_addr.sin_addr, 1110 cp0, sizeof(struct in_addr)); 1111 pserialize_read_exit(_ss); 1112 ipt->ipt_ptr += sizeof(struct in_addr); 1113 break; 1114 } 1115 1116 case IPOPT_TS_PRESPEC: { 1117 struct sockaddr_in ipaddr = { 1118 .sin_len = sizeof(ipaddr), 1119 .sin_family = AF_INET, 1120 }; 1121 1122 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1123 sizeof(struct in_addr) > ipt->ipt_len) { 1124 code = (u_char *)&ipt->ipt_ptr - 1125 (u_char *)ip; 1126 goto bad; 1127 } 1128 memcpy(&ipaddr.sin_addr, cp0, 1129 sizeof(struct in_addr)); 1130 s = pserialize_read_enter(); 1131 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1132 if (ifa == NULL) { 1133 pserialize_read_exit(s); 1134 continue; 1135 } 1136 pserialize_read_exit(s); 1137 ipt->ipt_ptr += sizeof(struct in_addr); 1138 break; 1139 } 1140 1141 default: 1142 /* XXX can't take &ipt->ipt_flg */ 1143 code = (u_char *)&ipt->ipt_ptr - 1144 (u_char *)ip + 1; 1145 goto bad; 1146 } 1147 ntime = iptime(); 1148 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1149 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1150 sizeof(n_time)); 1151 ipt->ipt_ptr += sizeof(n_time); 1152 } 1153 } 1154 if (forward) { 1155 struct ifnet *rcvif; 1156 struct psref _psref; 1157 1158 if (ip_forwsrcrt == 0) { 1159 type = ICMP_UNREACH; 1160 code = ICMP_UNREACH_SRCFAIL; 1161 goto bad; 1162 } 1163 1164 rcvif = m_get_rcvif_psref(m, &_psref); 1165 if (__predict_false(rcvif == NULL)) { 1166 type = ICMP_UNREACH; 1167 code = ICMP_UNREACH_HOST; 1168 goto bad; 1169 } 1170 ip_forward(m, 1, rcvif); 1171 m_put_rcvif_psref(rcvif, &_psref); 1172 return true; 1173 } 1174 return false; 1175bad: 1176 icmp_error(m, type, code, 0, 0); 1177 IP_STATINC(IP_STAT_BADOPTIONS); 1178 return true; 1179} 1180 1181/* 1182 * ip_rtaddr: given address of next destination (final or next hop), 1183 * return internet address info of interface to be used to get there. 1184 */ 1185static struct in_ifaddr * 1186ip_rtaddr(struct in_addr dst, struct psref *psref) 1187{ 1188 struct rtentry *rt; 1189 union { 1190 struct sockaddr dst; 1191 struct sockaddr_in dst4; 1192 } u; 1193 struct route *ro; 1194 1195 sockaddr_in_init(&u.dst4, &dst, 0); 1196 1197 ro = percpu_getref(ipforward_rt_percpu); 1198 rt = rtcache_lookup(ro, &u.dst); 1199 if (rt == NULL) { 1200 percpu_putref(ipforward_rt_percpu); 1201 return NULL; 1202 } 1203 1204 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1205 rtcache_unref(rt, ro); 1206 percpu_putref(ipforward_rt_percpu); 1207 1208 return ifatoia(rt->rt_ifa); 1209} 1210 1211/* 1212 * save_rte: save incoming source route for use in replies, to be picked 1213 * up later by ip_srcroute if the receiver is interested. 1214 */ 1215static void 1216save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1217{ 1218 struct ip_srcrt *isr; 1219 struct m_tag *mtag; 1220 unsigned olen; 1221 1222 olen = option[IPOPT_OLEN]; 1223 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1224 return; 1225 1226 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1227 if (mtag == NULL) 1228 return; 1229 isr = (struct ip_srcrt *)(mtag + 1); 1230 1231 memcpy(isr->isr_hdr, option, olen); 1232 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1233 isr->isr_dst = dst; 1234 m_tag_prepend(m, mtag); 1235} 1236 1237/* 1238 * Retrieve incoming source route for use in replies, 1239 * in the same form used by setsockopt. 1240 * The first hop is placed before the options, will be removed later. 1241 */ 1242struct mbuf * 1243ip_srcroute(struct mbuf *m0) 1244{ 1245 struct in_addr *p, *q; 1246 struct mbuf *m; 1247 struct ip_srcrt *isr; 1248 struct m_tag *mtag; 1249 1250 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL); 1251 if (mtag == NULL) 1252 return NULL; 1253 isr = (struct ip_srcrt *)(mtag + 1); 1254 1255 if (isr->isr_nhops == 0) 1256 return NULL; 1257 1258 m = m_get(M_DONTWAIT, MT_SOOPTS); 1259 if (m == NULL) 1260 return NULL; 1261 1262 MCLAIM(m, &inetdomain.dom_mowner); 1263#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1264 1265 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1266 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1267 1268 /* 1269 * First save first hop for return route 1270 */ 1271 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1272 *(mtod(m, struct in_addr *)) = *p--; 1273 1274 /* 1275 * Copy option fields and padding (nop) to mbuf. 1276 */ 1277 isr->isr_nop = IPOPT_NOP; 1278 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1279 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1280 OPTSIZ); 1281 q = (struct in_addr *)(mtod(m, char *) + 1282 sizeof(struct in_addr) + OPTSIZ); 1283#undef OPTSIZ 1284 /* 1285 * Record return path as an IP source route, 1286 * reversing the path (pointers are now aligned). 1287 */ 1288 while (p >= isr->isr_routes) { 1289 *q++ = *p--; 1290 } 1291 /* 1292 * Last hop goes to final destination. 1293 */ 1294 *q = isr->isr_dst; 1295 m_tag_delete(m0, mtag); 1296 return m; 1297} 1298 1299const int inetctlerrmap[PRC_NCMDS] = { 1300 [PRC_MSGSIZE] = EMSGSIZE, 1301 [PRC_HOSTDEAD] = EHOSTDOWN, 1302 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1303 [PRC_UNREACH_NET] = EHOSTUNREACH, 1304 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1305 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1306 [PRC_UNREACH_PORT] = ECONNREFUSED, 1307 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1308 [PRC_PARAMPROB] = ENOPROTOOPT, 1309}; 1310 1311void 1312ip_fasttimo(void) 1313{ 1314 if (ip_drainwanted) { 1315 ip_drain(); 1316 ip_drainwanted = 0; 1317 } 1318} 1319 1320void 1321ip_drainstub(void) 1322{ 1323 ip_drainwanted = 1; 1324} 1325 1326/* 1327 * Forward a packet. If some error occurs return the sender 1328 * an icmp packet. Note we can't always generate a meaningful 1329 * icmp message because icmp doesn't have a large enough repertoire 1330 * of codes and types. 1331 * 1332 * If not forwarding, just drop the packet. This could be confusing 1333 * if ipforwarding was zero but some routing protocol was advancing 1334 * us as a gateway to somewhere. However, we must let the routing 1335 * protocol deal with that. 1336 * 1337 * The srcrt parameter indicates whether the packet is being forwarded 1338 * via a source route. 1339 */ 1340static void 1341ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1342{ 1343 struct ip *ip = mtod(m, struct ip *); 1344 struct rtentry *rt; 1345 int error, type = 0, code = 0, destmtu = 0; 1346 struct mbuf *mcopy; 1347 n_long dest; 1348 union { 1349 struct sockaddr dst; 1350 struct sockaddr_in dst4; 1351 } u; 1352 uint64_t *ips; 1353 struct route *ro; 1354 1355 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1356 "interrupt handler; synchronization assumptions violated"); 1357 1358 /* 1359 * We are now in the output path. 1360 */ 1361 MCLAIM(m, &ip_tx_mowner); 1362 1363 /* 1364 * Clear any in-bound checksum flags for this packet. 1365 */ 1366 m->m_pkthdr.csum_flags = 0; 1367 1368 dest = 0; 1369 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1370 IP_STATINC(IP_STAT_CANTFORWARD); 1371 m_freem(m); 1372 return; 1373 } 1374 1375 if (ip->ip_ttl <= IPTTLDEC) { 1376 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1377 return; 1378 } 1379 1380 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1381 1382 ro = percpu_getref(ipforward_rt_percpu); 1383 rt = rtcache_lookup(ro, &u.dst); 1384 if (rt == NULL) { 1385 percpu_putref(ipforward_rt_percpu); 1386 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1387 return; 1388 } 1389 1390 /* 1391 * Save at most 68 bytes of the packet in case 1392 * we need to generate an ICMP message to the src. 1393 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1394 */ 1395 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1396 if (mcopy) 1397 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1398 1399 ip->ip_ttl -= IPTTLDEC; 1400 1401 /* 1402 * If forwarding packet using same interface that it came in on, 1403 * perhaps should send a redirect to sender to shortcut a hop. 1404 * Only send redirect if source is sending directly to us, 1405 * and if packet was not source routed (or has any options). 1406 * Also, don't send redirect if forwarding using a default route 1407 * or a route modified by a redirect. 1408 */ 1409 if (rt->rt_ifp == rcvif && 1410 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1411 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1412 ipsendredirects && !srcrt) { 1413 if (rt->rt_ifa && 1414 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1415 ifatoia(rt->rt_ifa)->ia_subnet) { 1416 if (rt->rt_flags & RTF_GATEWAY) 1417 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1418 else 1419 dest = ip->ip_dst.s_addr; 1420 /* 1421 * Router requirements says to only send host 1422 * redirects. 1423 */ 1424 type = ICMP_REDIRECT; 1425 code = ICMP_REDIRECT_HOST; 1426 } 1427 } 1428 rtcache_unref(rt, ro); 1429 1430 error = ip_output(m, NULL, ro, 1431 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1432 NULL, NULL); 1433 1434 if (error) { 1435 IP_STATINC(IP_STAT_CANTFORWARD); 1436 goto error; 1437 } 1438 1439 ips = IP_STAT_GETREF(); 1440 ips[IP_STAT_FORWARD]++; 1441 1442 if (type) { 1443 ips[IP_STAT_REDIRECTSENT]++; 1444 IP_STAT_PUTREF(); 1445 goto redirect; 1446 } 1447 1448 IP_STAT_PUTREF(); 1449 if (mcopy) { 1450#ifdef GATEWAY 1451 if (mcopy->m_flags & M_CANFASTFWD) 1452 ipflow_create(ro, mcopy); 1453#endif 1454 m_freem(mcopy); 1455 } 1456 1457 percpu_putref(ipforward_rt_percpu); 1458 return; 1459 1460redirect: 1461error: 1462 if (mcopy == NULL) { 1463 percpu_putref(ipforward_rt_percpu); 1464 return; 1465 } 1466 1467 switch (error) { 1468 1469 case 0: /* forwarded, but need redirect */ 1470 /* type, code set above */ 1471 break; 1472 1473 case ENETUNREACH: /* shouldn't happen, checked above */ 1474 case EHOSTUNREACH: 1475 case ENETDOWN: 1476 case EHOSTDOWN: 1477 default: 1478 type = ICMP_UNREACH; 1479 code = ICMP_UNREACH_HOST; 1480 break; 1481 1482 case EMSGSIZE: 1483 type = ICMP_UNREACH; 1484 code = ICMP_UNREACH_NEEDFRAG; 1485 1486 if ((rt = rtcache_validate(ro)) != NULL) { 1487 destmtu = rt->rt_ifp->if_mtu; 1488 rtcache_unref(rt, ro); 1489 } 1490#ifdef IPSEC 1491 if (ipsec_used) 1492 (void)ipsec4_forward(mcopy, &destmtu); 1493#endif 1494 IP_STATINC(IP_STAT_CANTFRAG); 1495 break; 1496 1497 case ENOBUFS: 1498 /* 1499 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1500 * Requirements for IP Version 4 Routers. Source quench can 1501 * big problem under DoS attacks or if the underlying 1502 * interface is rate-limited. 1503 */ 1504 if (mcopy) 1505 m_freem(mcopy); 1506 percpu_putref(ipforward_rt_percpu); 1507 return; 1508 } 1509 icmp_error(mcopy, type, code, dest, destmtu); 1510 percpu_putref(ipforward_rt_percpu); 1511} 1512 1513void 1514ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1515 struct mbuf *m) 1516{ 1517 struct socket *so = inp->inp_socket; 1518 int inpflags = inp->inp_flags; 1519 1520 if (SOOPT_TIMESTAMP(so->so_options)) 1521 mp = sbsavetimestamp(so->so_options, m, mp); 1522 1523 if (inpflags & INP_RECVDSTADDR) { 1524 *mp = sbcreatecontrol(&ip->ip_dst, 1525 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1526 if (*mp) 1527 mp = &(*mp)->m_next; 1528 } 1529 1530 if (inpflags & INP_RECVTTL) { 1531 *mp = sbcreatecontrol(&ip->ip_ttl, 1532 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1533 if (*mp) 1534 mp = &(*mp)->m_next; 1535 } 1536 1537 struct psref psref; 1538 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1539 if (__predict_false(ifp == NULL)) { 1540#ifdef DIAGNOSTIC 1541 printf("%s: missing receive interface\n", __func__); 1542#endif 1543 return; /* XXX should report error? */ 1544 } 1545 1546 if (inpflags & INP_RECVPKTINFO) { 1547 struct in_pktinfo ipi; 1548 ipi.ipi_addr = ip->ip_dst; 1549 ipi.ipi_ifindex = ifp->if_index; 1550 *mp = sbcreatecontrol(&ipi, 1551 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1552 if (*mp) 1553 mp = &(*mp)->m_next; 1554 } 1555 if (inpflags & INP_RECVIF) { 1556 struct sockaddr_dl sdl; 1557 1558 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1559 NULL, 0); 1560 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1561 if (*mp) 1562 mp = &(*mp)->m_next; 1563 } 1564 m_put_rcvif_psref(ifp, &psref); 1565} 1566 1567/* 1568 * sysctl helper routine for net.inet.ip.forwsrcrt. 1569 */ 1570static int 1571sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1572{ 1573 int error, tmp; 1574 struct sysctlnode node; 1575 1576 node = *rnode; 1577 tmp = ip_forwsrcrt; 1578 node.sysctl_data = &tmp; 1579 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1580 if (error || newp == NULL) 1581 return (error); 1582 1583 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1584 0, NULL, NULL, NULL); 1585 if (error) 1586 return (error); 1587 1588 ip_forwsrcrt = tmp; 1589 1590 return (0); 1591} 1592 1593/* 1594 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1595 * range of the new value and tweaks timers if it changes. 1596 */ 1597static int 1598sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1599{ 1600 int error, tmp; 1601 struct sysctlnode node; 1602 1603 icmp_mtudisc_lock(); 1604 1605 node = *rnode; 1606 tmp = ip_mtudisc_timeout; 1607 node.sysctl_data = &tmp; 1608 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1609 if (error || newp == NULL) 1610 goto out; 1611 if (tmp < 0) { 1612 error = EINVAL; 1613 goto out; 1614 } 1615 1616 ip_mtudisc_timeout = tmp; 1617 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1618 error = 0; 1619out: 1620 icmp_mtudisc_unlock(); 1621 return error; 1622} 1623 1624static int 1625sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1626{ 1627 1628 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1629} 1630 1631static void 1632sysctl_net_inet_ip_setup(struct sysctllog **clog) 1633{ 1634 sysctl_createv(clog, 0, NULL, NULL, 1635 CTLFLAG_PERMANENT, 1636 CTLTYPE_NODE, "inet", 1637 SYSCTL_DESCR("PF_INET related settings"), 1638 NULL, 0, NULL, 0, 1639 CTL_NET, PF_INET, CTL_EOL); 1640 sysctl_createv(clog, 0, NULL, NULL, 1641 CTLFLAG_PERMANENT, 1642 CTLTYPE_NODE, "ip", 1643 SYSCTL_DESCR("IPv4 related settings"), 1644 NULL, 0, NULL, 0, 1645 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1646 1647 sysctl_createv(clog, 0, NULL, NULL, 1648 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1649 CTLTYPE_INT, "forwarding", 1650 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1651 NULL, 0, &ipforwarding, 0, 1652 CTL_NET, PF_INET, IPPROTO_IP, 1653 IPCTL_FORWARDING, CTL_EOL); 1654 sysctl_createv(clog, 0, NULL, NULL, 1655 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1656 CTLTYPE_INT, "redirect", 1657 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1658 NULL, 0, &ipsendredirects, 0, 1659 CTL_NET, PF_INET, IPPROTO_IP, 1660 IPCTL_SENDREDIRECTS, CTL_EOL); 1661 sysctl_createv(clog, 0, NULL, NULL, 1662 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1663 CTLTYPE_INT, "ttl", 1664 SYSCTL_DESCR("Default TTL for an INET datagram"), 1665 NULL, 0, &ip_defttl, 0, 1666 CTL_NET, PF_INET, IPPROTO_IP, 1667 IPCTL_DEFTTL, CTL_EOL); 1668#ifdef IPCTL_DEFMTU 1669 sysctl_createv(clog, 0, NULL, NULL, 1670 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1671 CTLTYPE_INT, "mtu", 1672 SYSCTL_DESCR("Default MTA for an INET route"), 1673 NULL, 0, &ip_mtu, 0, 1674 CTL_NET, PF_INET, IPPROTO_IP, 1675 IPCTL_DEFMTU, CTL_EOL); 1676#endif /* IPCTL_DEFMTU */ 1677 sysctl_createv(clog, 0, NULL, NULL, 1678 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1679 CTLTYPE_INT, "forwsrcrt", 1680 SYSCTL_DESCR("Enable forwarding of source-routed " 1681 "datagrams"), 1682 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1683 CTL_NET, PF_INET, IPPROTO_IP, 1684 IPCTL_FORWSRCRT, CTL_EOL); 1685 sysctl_createv(clog, 0, NULL, NULL, 1686 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1687 CTLTYPE_INT, "directed-broadcast", 1688 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1689 NULL, 0, &ip_directedbcast, 0, 1690 CTL_NET, PF_INET, IPPROTO_IP, 1691 IPCTL_DIRECTEDBCAST, CTL_EOL); 1692 sysctl_createv(clog, 0, NULL, NULL, 1693 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1694 CTLTYPE_INT, "allowsrcrt", 1695 SYSCTL_DESCR("Accept source-routed datagrams"), 1696 NULL, 0, &ip_allowsrcrt, 0, 1697 CTL_NET, PF_INET, IPPROTO_IP, 1698 IPCTL_ALLOWSRCRT, CTL_EOL); 1699 1700 sysctl_createv(clog, 0, NULL, NULL, 1701 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1702 CTLTYPE_INT, "mtudisc", 1703 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1704 NULL, 0, &ip_mtudisc, 0, 1705 CTL_NET, PF_INET, IPPROTO_IP, 1706 IPCTL_MTUDISC, CTL_EOL); 1707 sysctl_createv(clog, 0, NULL, NULL, 1708 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1709 CTLTYPE_INT, "anonportmin", 1710 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1711 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1712 CTL_NET, PF_INET, IPPROTO_IP, 1713 IPCTL_ANONPORTMIN, CTL_EOL); 1714 sysctl_createv(clog, 0, NULL, NULL, 1715 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1716 CTLTYPE_INT, "anonportmax", 1717 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1718 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1719 CTL_NET, PF_INET, IPPROTO_IP, 1720 IPCTL_ANONPORTMAX, CTL_EOL); 1721 sysctl_createv(clog, 0, NULL, NULL, 1722 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1723 CTLTYPE_INT, "mtudisctimeout", 1724 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1725 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1726 CTL_NET, PF_INET, IPPROTO_IP, 1727 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1728#ifndef IPNOPRIVPORTS 1729 sysctl_createv(clog, 0, NULL, NULL, 1730 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1731 CTLTYPE_INT, "lowportmin", 1732 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1733 "to assign"), 1734 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1735 CTL_NET, PF_INET, IPPROTO_IP, 1736 IPCTL_LOWPORTMIN, CTL_EOL); 1737 sysctl_createv(clog, 0, NULL, NULL, 1738 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1739 CTLTYPE_INT, "lowportmax", 1740 SYSCTL_DESCR("Highest privileged ephemeral port number " 1741 "to assign"), 1742 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1743 CTL_NET, PF_INET, IPPROTO_IP, 1744 IPCTL_LOWPORTMAX, CTL_EOL); 1745#endif /* IPNOPRIVPORTS */ 1746#if NGRE > 0 1747 sysctl_createv(clog, 0, NULL, NULL, 1748 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1749 CTLTYPE_INT, "grettl", 1750 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1751 NULL, 0, &ip_gre_ttl, 0, 1752 CTL_NET, PF_INET, IPPROTO_IP, 1753 IPCTL_GRE_TTL, CTL_EOL); 1754#endif /* NGRE */ 1755 sysctl_createv(clog, 0, NULL, NULL, 1756 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1757 CTLTYPE_INT, "checkinterface", 1758 SYSCTL_DESCR("Enable receive side of Strong ES model " 1759 "from RFC1122"), 1760 NULL, 0, &ip_checkinterface, 0, 1761 CTL_NET, PF_INET, IPPROTO_IP, 1762 IPCTL_CHECKINTERFACE, CTL_EOL); 1763 sysctl_createv(clog, 0, NULL, NULL, 1764 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1765 CTLTYPE_INT, "random_id", 1766 SYSCTL_DESCR("Assign random ip_id values"), 1767 NULL, 0, &ip_do_randomid, 0, 1768 CTL_NET, PF_INET, IPPROTO_IP, 1769 IPCTL_RANDOMID, CTL_EOL); 1770 sysctl_createv(clog, 0, NULL, NULL, 1771 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1772 CTLTYPE_INT, "do_loopback_cksum", 1773 SYSCTL_DESCR("Perform IP checksum on loopback"), 1774 NULL, 0, &ip_do_loopback_cksum, 0, 1775 CTL_NET, PF_INET, IPPROTO_IP, 1776 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1777 sysctl_createv(clog, 0, NULL, NULL, 1778 CTLFLAG_PERMANENT, 1779 CTLTYPE_STRUCT, "stats", 1780 SYSCTL_DESCR("IP statistics"), 1781 sysctl_net_inet_ip_stats, 0, NULL, 0, 1782 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1783 CTL_EOL); 1784#if NARP 1785 sysctl_createv(clog, 0, NULL, NULL, 1786 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1787 CTLTYPE_INT, "dad_count", 1788 SYSCTL_DESCR("Number of Duplicate Address Detection " 1789 "probes to send"), 1790 NULL, 0, &ip_dad_count, 0, 1791 CTL_NET, PF_INET, IPPROTO_IP, 1792 IPCTL_DAD_COUNT, CTL_EOL); 1793#endif 1794 1795 /* anonportalgo RFC6056 subtree */ 1796 const struct sysctlnode *portalgo_node; 1797 sysctl_createv(clog, 0, NULL, &portalgo_node, 1798 CTLFLAG_PERMANENT, 1799 CTLTYPE_NODE, "anonportalgo", 1800 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1801 NULL, 0, NULL, 0, 1802 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1803 sysctl_createv(clog, 0, &portalgo_node, NULL, 1804 CTLFLAG_PERMANENT, 1805 CTLTYPE_STRING, "available", 1806 SYSCTL_DESCR("available algorithms"), 1807 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1808 CTL_CREATE, CTL_EOL); 1809 sysctl_createv(clog, 0, &portalgo_node, NULL, 1810 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1811 CTLTYPE_STRING, "selected", 1812 SYSCTL_DESCR("selected algorithm"), 1813 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1814 CTL_CREATE, CTL_EOL); 1815 sysctl_createv(clog, 0, &portalgo_node, NULL, 1816 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1817 CTLTYPE_STRUCT, "reserve", 1818 SYSCTL_DESCR("bitmap of reserved ports"), 1819 sysctl_portalgo_reserve4, 0, NULL, 0, 1820 CTL_CREATE, CTL_EOL); 1821} 1822 1823void 1824ip_statinc(u_int stat) 1825{ 1826 1827 KASSERT(stat < IP_NSTATS); 1828 IP_STATINC(stat); 1829} 1830