ip_input.c revision 1.352
1/* $NetBSD: ip_input.c,v 1.352 2017/03/06 07:31:15 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 1.352 2017/03/06 07:31:15 ozaki-r Exp $"); 95 96#ifdef _KERNEL_OPT 97#include "opt_inet.h" 98#include "opt_compat_netbsd.h" 99#include "opt_gateway.h" 100#include "opt_ipsec.h" 101#include "opt_mrouting.h" 102#include "opt_mbuftrace.h" 103#include "opt_inet_csum.h" 104#include "opt_net_mpsafe.h" 105#endif 106 107#include "arp.h" 108 109#include <sys/param.h> 110#include <sys/systm.h> 111#include <sys/cpu.h> 112#include <sys/mbuf.h> 113#include <sys/domain.h> 114#include <sys/protosw.h> 115#include <sys/socket.h> 116#include <sys/socketvar.h> 117#include <sys/errno.h> 118#include <sys/time.h> 119#include <sys/kernel.h> 120#include <sys/pool.h> 121#include <sys/sysctl.h> 122#include <sys/kauth.h> 123 124#include <net/if.h> 125#include <net/if_dl.h> 126#include <net/route.h> 127#include <net/pktqueue.h> 128#include <net/pfil.h> 129 130#include <netinet/in.h> 131#include <netinet/in_systm.h> 132#include <netinet/ip.h> 133#include <netinet/in_pcb.h> 134#include <netinet/in_proto.h> 135#include <netinet/in_var.h> 136#include <netinet/ip_var.h> 137#include <netinet/ip_private.h> 138#include <netinet/ip_icmp.h> 139/* just for gif_ttl */ 140#include <netinet/in_gif.h> 141#include "gif.h" 142#include <net/if_gre.h> 143#include "gre.h" 144 145#ifdef MROUTING 146#include <netinet/ip_mroute.h> 147#endif 148#include <netinet/portalgo.h> 149 150#ifdef IPSEC 151#include <netipsec/ipsec.h> 152#endif 153 154#ifndef IPFORWARDING 155#ifdef GATEWAY 156#define IPFORWARDING 1 /* forward IP packets not for us */ 157#else /* GATEWAY */ 158#define IPFORWARDING 0 /* don't forward IP packets not for us */ 159#endif /* GATEWAY */ 160#endif /* IPFORWARDING */ 161#ifndef IPSENDREDIRECTS 162#define IPSENDREDIRECTS 1 163#endif 164#ifndef IPFORWSRCRT 165#define IPFORWSRCRT 1 /* forward source-routed packets */ 166#endif 167#ifndef IPALLOWSRCRT 168#define IPALLOWSRCRT 1 /* allow source-routed packets */ 169#endif 170#ifndef IPMTUDISC 171#define IPMTUDISC 1 172#endif 173#ifndef IPMTUDISCTIMEOUT 174#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 175#endif 176 177#ifdef COMPAT_50 178#include <compat/sys/time.h> 179#include <compat/sys/socket.h> 180#endif 181 182/* 183 * Note: DIRECTED_BROADCAST is handled this way so that previous 184 * configuration using this option will Just Work. 185 */ 186#ifndef IPDIRECTEDBCAST 187#ifdef DIRECTED_BROADCAST 188#define IPDIRECTEDBCAST 1 189#else 190#define IPDIRECTEDBCAST 0 191#endif /* DIRECTED_BROADCAST */ 192#endif /* IPDIRECTEDBCAST */ 193int ipforwarding = IPFORWARDING; 194int ipsendredirects = IPSENDREDIRECTS; 195int ip_defttl = IPDEFTTL; 196int ip_forwsrcrt = IPFORWSRCRT; 197int ip_directedbcast = IPDIRECTEDBCAST; 198int ip_allowsrcrt = IPALLOWSRCRT; 199int ip_mtudisc = IPMTUDISC; 200int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 201#ifdef DIAGNOSTIC 202int ipprintfs = 0; 203#endif 204 205int ip_do_randomid = 0; 206 207/* 208 * XXX - Setting ip_checkinterface mostly implements the receive side of 209 * the Strong ES model described in RFC 1122, but since the routing table 210 * and transmit implementation do not implement the Strong ES model, 211 * setting this to 1 results in an odd hybrid. 212 * 213 * XXX - ip_checkinterface currently must be disabled if you use ipnat 214 * to translate the destination address to another local interface. 215 * 216 * XXX - ip_checkinterface must be disabled if you add IP aliases 217 * to the loopback interface instead of the interface where the 218 * packets for those addresses are received. 219 */ 220static int ip_checkinterface __read_mostly = 0; 221 222struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 223 224pktqueue_t * ip_pktq __read_mostly; 225pfil_head_t * inet_pfil_hook __read_mostly; 226ipid_state_t * ip_ids __read_mostly; 227percpu_t * ipstat_percpu __read_mostly; 228 229static percpu_t *ipforward_rt_percpu __cacheline_aligned; 230 231uint16_t ip_id; 232 233#ifdef INET_CSUM_COUNTERS 234#include <sys/device.h> 235 236struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 237 NULL, "inet", "hwcsum bad"); 238struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 239 NULL, "inet", "hwcsum ok"); 240struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 241 NULL, "inet", "swcsum"); 242 243#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 244 245EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 246EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 247EVCNT_ATTACH_STATIC(ip_swcsum); 248 249#else 250 251#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 252 253#endif /* INET_CSUM_COUNTERS */ 254 255/* 256 * We need to save the IP options in case a protocol wants to respond 257 * to an incoming packet over the same route if the packet got here 258 * using IP source routing. This allows connection establishment and 259 * maintenance when the remote end is on a network that is not known 260 * to us. 261 */ 262 263static int ip_nhops = 0; 264 265static struct ip_srcrt { 266 struct in_addr dst; /* final destination */ 267 char nop; /* one NOP to align */ 268 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 269 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 270} ip_srcrt; 271 272static int ip_drainwanted; 273 274struct sockaddr_in ipaddr = { 275 .sin_len = sizeof(ipaddr), 276 .sin_family = AF_INET, 277}; 278 279static void save_rte(u_char *, struct in_addr); 280 281#ifdef MBUFTRACE 282struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 283struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 284#endif 285 286static void ipintr(void *); 287static void ip_input(struct mbuf *); 288static void ip_forward(struct mbuf *, int, struct ifnet *); 289static bool ip_dooptions(struct mbuf *); 290static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *); 291static void sysctl_net_inet_ip_setup(struct sysctllog **); 292 293static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *, 294 int *); 295static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *, 296 struct ip *); 297 298#ifdef NET_MPSAFE 299#define SOFTNET_LOCK() mutex_enter(softnet_lock) 300#define SOFTNET_UNLOCK() mutex_exit(softnet_lock) 301#else 302#define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 303#define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 304#endif 305 306/* 307 * IP initialization: fill in IP protocol switch table. 308 * All protocols not implemented in kernel go to raw IP protocol handler. 309 */ 310void 311ip_init(void) 312{ 313 const struct protosw *pr; 314 315 in_init(); 316 sysctl_net_inet_ip_setup(NULL); 317 318 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 319 KASSERT(pr != NULL); 320 321 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 322 KASSERT(ip_pktq != NULL); 323 324 for (u_int i = 0; i < IPPROTO_MAX; i++) { 325 ip_protox[i] = pr - inetsw; 326 } 327 for (pr = inetdomain.dom_protosw; 328 pr < inetdomain.dom_protoswNPROTOSW; pr++) 329 if (pr->pr_domain->dom_family == PF_INET && 330 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 331 ip_protox[pr->pr_protocol] = pr - inetsw; 332 333 ip_reass_init(); 334 335 ip_ids = ip_id_init(); 336 ip_id = time_uptime & 0xfffff; 337 338#ifdef GATEWAY 339 ipflow_init(); 340#endif 341 342 /* Register our Packet Filter hook. */ 343 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 344 KASSERT(inet_pfil_hook != NULL); 345 346#ifdef MBUFTRACE 347 MOWNER_ATTACH(&ip_tx_mowner); 348 MOWNER_ATTACH(&ip_rx_mowner); 349#endif /* MBUFTRACE */ 350 351 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 352 353 ipforward_rt_percpu = percpu_alloc(sizeof(struct route)); 354 if (ipforward_rt_percpu == NULL) 355 panic("failed to allocate ipforward_rt_percpu"); 356 357 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 358} 359 360static struct in_ifaddr * 361ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 362{ 363 struct in_ifaddr *ia = NULL; 364 int checkif; 365 366 /* 367 * Enable a consistency check between the destination address 368 * and the arrival interface for a unicast packet (the RFC 1122 369 * strong ES model) if IP forwarding is disabled and the packet 370 * is not locally generated. 371 * 372 * XXX - Checking also should be disabled if the destination 373 * address is ipnat'ed to a different interface. 374 * 375 * XXX - Checking is incompatible with IP aliases added 376 * to the loopback interface instead of the interface where 377 * the packets are received. 378 * 379 * XXX - We need to add a per ifaddr flag for this so that 380 * we get finer grain control. 381 */ 382 checkif = ip_checkinterface && (ipforwarding == 0) && 383 (ifp->if_flags & IFF_LOOPBACK) == 0; 384 385 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 386 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 387 if (ia->ia4_flags & IN_IFF_NOTREADY) 388 continue; 389 if (checkif && ia->ia_ifp != ifp) 390 continue; 391 if ((ia->ia_ifp->if_flags & IFF_UP) != 0 && 392 (ia->ia4_flags & IN_IFF_DETACHED) == 0) 393 break; 394 else 395 (*downmatch)++; 396 } 397 } 398 399 return ia; 400} 401 402static struct in_ifaddr * 403ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 404{ 405 struct in_ifaddr *ia = NULL; 406 struct ifaddr *ifa; 407 408 IFADDR_READER_FOREACH(ifa, ifp) { 409 if (ifa->ifa_addr->sa_family != AF_INET) 410 continue; 411 ia = ifatoia(ifa); 412 if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) 413 continue; 414 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 415 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 416 /* 417 * Look for all-0's host part (old broadcast addr), 418 * either for subnet or net. 419 */ 420 ip->ip_dst.s_addr == ia->ia_subnet || 421 ip->ip_dst.s_addr == ia->ia_net) 422 goto matched; 423 /* 424 * An interface with IP address zero accepts 425 * all packets that arrive on that interface. 426 */ 427 if (in_nullhost(ia->ia_addr.sin_addr)) 428 goto matched; 429 } 430 ia = NULL; 431 432matched: 433 return ia; 434} 435 436/* 437 * IP software interrupt routine. 438 */ 439static void 440ipintr(void *arg __unused) 441{ 442 struct mbuf *m; 443 444 KASSERT(cpu_softintr_p()); 445 446#ifndef NET_MPSAFE 447 mutex_enter(softnet_lock); 448#endif 449 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 450 ip_input(m); 451 } 452#ifndef NET_MPSAFE 453 mutex_exit(softnet_lock); 454#endif 455} 456 457/* 458 * IP input routine. Checksum and byte swap header. If fragmented 459 * try to reassemble. Process options. Pass to next level. 460 */ 461static void 462ip_input(struct mbuf *m) 463{ 464 struct ip *ip = NULL; 465 struct in_ifaddr *ia = NULL; 466 int hlen = 0, len; 467 int downmatch; 468 int srcrt = 0; 469 ifnet_t *ifp; 470 struct psref psref; 471 int s; 472 473 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 474 "interrupt handler; synchronization assumptions violated"); 475 476 MCLAIM(m, &ip_rx_mowner); 477 KASSERT((m->m_flags & M_PKTHDR) != 0); 478 479 ifp = m_get_rcvif_psref(m, &psref); 480 if (__predict_false(ifp == NULL)) 481 goto out; 482 483 /* 484 * If no IP addresses have been set yet but the interfaces 485 * are receiving, can't do anything with incoming packets yet. 486 * Note: we pre-check without locks held. 487 */ 488 if (IN_ADDRLIST_READER_EMPTY()) 489 goto out; 490 IP_STATINC(IP_STAT_TOTAL); 491 492 /* 493 * If the IP header is not aligned, slurp it up into a new 494 * mbuf with space for link headers, in the event we forward 495 * it. Otherwise, if it is aligned, make sure the entire 496 * base IP header is in the first mbuf of the chain. 497 */ 498 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 499 if ((m = m_copyup(m, sizeof(struct ip), 500 (max_linkhdr + 3) & ~3)) == NULL) { 501 /* XXXJRT new stat, please */ 502 IP_STATINC(IP_STAT_TOOSMALL); 503 goto out; 504 } 505 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 506 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 507 IP_STATINC(IP_STAT_TOOSMALL); 508 goto out; 509 } 510 } 511 ip = mtod(m, struct ip *); 512 if (ip->ip_v != IPVERSION) { 513 IP_STATINC(IP_STAT_BADVERS); 514 goto out; 515 } 516 hlen = ip->ip_hl << 2; 517 if (hlen < sizeof(struct ip)) { /* minimum header length */ 518 IP_STATINC(IP_STAT_BADHLEN); 519 goto out; 520 } 521 if (hlen > m->m_len) { 522 if ((m = m_pullup(m, hlen)) == NULL) { 523 IP_STATINC(IP_STAT_BADHLEN); 524 goto out; 525 } 526 ip = mtod(m, struct ip *); 527 } 528 529 /* 530 * RFC1122: packets with a multicast source address are 531 * not allowed. 532 */ 533 if (IN_MULTICAST(ip->ip_src.s_addr)) { 534 IP_STATINC(IP_STAT_BADADDR); 535 goto out; 536 } 537 538 /* 127/8 must not appear on wire - RFC1122 */ 539 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 540 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 541 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 542 IP_STATINC(IP_STAT_BADADDR); 543 goto out; 544 } 545 } 546 547 switch (m->m_pkthdr.csum_flags & 548 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | 549 M_CSUM_IPv4_BAD)) { 550 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 551 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 552 IP_STATINC(IP_STAT_BADSUM); 553 goto out; 554 555 case M_CSUM_IPv4: 556 /* Checksum was okay. */ 557 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 558 break; 559 560 default: 561 /* 562 * Must compute it ourselves. Maybe skip checksum on 563 * loopback interfaces. 564 */ 565 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 566 ip_do_loopback_cksum)) { 567 INET_CSUM_COUNTER_INCR(&ip_swcsum); 568 if (in_cksum(m, hlen) != 0) { 569 IP_STATINC(IP_STAT_BADSUM); 570 goto out; 571 } 572 } 573 break; 574 } 575 576 /* Retrieve the packet length. */ 577 len = ntohs(ip->ip_len); 578 579 /* 580 * Check for additional length bogosity 581 */ 582 if (len < hlen) { 583 IP_STATINC(IP_STAT_BADLEN); 584 goto out; 585 } 586 587 /* 588 * Check that the amount of data in the buffers 589 * is as at least much as the IP header would have us expect. 590 * Trim mbufs if longer than we expect. 591 * Drop packet if shorter than we expect. 592 */ 593 if (m->m_pkthdr.len < len) { 594 IP_STATINC(IP_STAT_TOOSHORT); 595 goto out; 596 } 597 if (m->m_pkthdr.len > len) { 598 if (m->m_len == m->m_pkthdr.len) { 599 m->m_len = len; 600 m->m_pkthdr.len = len; 601 } else 602 m_adj(m, len - m->m_pkthdr.len); 603 } 604 605 /* 606 * Assume that we can create a fast-forward IP flow entry 607 * based on this packet. 608 */ 609 m->m_flags |= M_CANFASTFWD; 610 611 /* 612 * Run through list of hooks for input packets. If there are any 613 * filters which require that additional packets in the flow are 614 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 615 * Note that filters must _never_ set this flag, as another filter 616 * in the list may have previously cleared it. 617 */ 618#if defined(IPSEC) 619 if (!ipsec_used || !ipsec_indone(m)) 620#else 621 if (1) 622#endif 623 { 624 struct in_addr odst = ip->ip_dst; 625 bool freed; 626 627 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 628 if (freed || m == NULL) { 629 m = NULL; 630 goto out; 631 } 632 ip = mtod(m, struct ip *); 633 hlen = ip->ip_hl << 2; 634 635 /* 636 * XXX The setting of "srcrt" here is to prevent ip_forward() 637 * from generating ICMP redirects for packets that have 638 * been redirected by a hook back out on to the same LAN that 639 * they came from and is not an indication that the packet 640 * is being inffluenced by source routing options. This 641 * allows things like 642 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 643 * where tlp0 is both on the 1.1.1.0/24 network and is the 644 * default route for hosts on 1.1.1.0/24. Of course this 645 * also requires a "map tlp0 ..." to complete the story. 646 * One might argue whether or not this kind of network config. 647 * should be supported in this manner... 648 */ 649 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 650 } 651 652#ifdef ALTQ 653 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 654 if (altq_input) { 655 SOFTNET_LOCK(); 656 if ((*altq_input)(m, AF_INET) == 0) { 657 /* Packet dropped by traffic conditioner. */ 658 SOFTNET_UNLOCK(); 659 m = NULL; 660 goto out; 661 } 662 SOFTNET_UNLOCK(); 663 } 664#endif 665 666 /* 667 * Process options and, if not destined for us, 668 * ship it on. ip_dooptions returns 1 when an 669 * error was detected (causing an icmp message 670 * to be sent and the original packet to be freed). 671 */ 672 ip_nhops = 0; /* for source routed packets */ 673 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 674 m = NULL; 675 goto out; 676 } 677 678 /* 679 * Check our list of addresses, to see if the packet is for us. 680 * 681 * Traditional 4.4BSD did not consult IFF_UP at all. 682 * The behavior here is to treat addresses on !IFF_UP interface 683 * or IN_IFF_NOTREADY addresses as not mine. 684 */ 685 downmatch = 0; 686 s = pserialize_read_enter(); 687 ia = ip_match_our_address(ifp, ip, &downmatch); 688 if (ia != NULL) { 689 pserialize_read_exit(s); 690 goto ours; 691 } 692 693 if (ifp->if_flags & IFF_BROADCAST) { 694 ia = ip_match_our_address_broadcast(ifp, ip); 695 if (ia != NULL) { 696 pserialize_read_exit(s); 697 goto ours; 698 } 699 } 700 pserialize_read_exit(s); 701 702 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 703#ifdef MROUTING 704 extern struct socket *ip_mrouter; 705 706 if (ip_mrouter) { 707 /* 708 * If we are acting as a multicast router, all 709 * incoming multicast packets are passed to the 710 * kernel-level multicast forwarding function. 711 * The packet is returned (relatively) intact; if 712 * ip_mforward() returns a non-zero value, the packet 713 * must be discarded, else it may be accepted below. 714 * 715 * (The IP ident field is put in the same byte order 716 * as expected when ip_mforward() is called from 717 * ip_output().) 718 */ 719 SOFTNET_LOCK(); 720 if (ip_mforward(m, ifp) != 0) { 721 SOFTNET_UNLOCK(); 722 IP_STATINC(IP_STAT_CANTFORWARD); 723 goto out; 724 } 725 SOFTNET_UNLOCK(); 726 727 /* 728 * The process-level routing demon needs to receive 729 * all multicast IGMP packets, whether or not this 730 * host belongs to their destination groups. 731 */ 732 if (ip->ip_p == IPPROTO_IGMP) { 733 goto ours; 734 } 735 IP_STATINC(IP_STAT_CANTFORWARD); 736 } 737#endif 738 /* 739 * See if we belong to the destination multicast group on the 740 * arrival interface. 741 */ 742 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 743 IP_STATINC(IP_STAT_CANTFORWARD); 744 goto out; 745 } 746 goto ours; 747 } 748 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 749 in_nullhost(ip->ip_dst)) 750 goto ours; 751 752 /* 753 * Not for us; forward if possible and desirable. 754 */ 755 if (ipforwarding == 0) { 756 m_put_rcvif_psref(ifp, &psref); 757 IP_STATINC(IP_STAT_CANTFORWARD); 758 m_freem(m); 759 } else { 760 /* 761 * If ip_dst matched any of my address on !IFF_UP interface, 762 * and there's no IFF_UP interface that matches ip_dst, 763 * send icmp unreach. Forwarding it will result in in-kernel 764 * forwarding loop till TTL goes to 0. 765 */ 766 if (downmatch) { 767 m_put_rcvif_psref(ifp, &psref); 768 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 769 IP_STATINC(IP_STAT_CANTFORWARD); 770 return; 771 } 772#ifdef IPSEC 773 /* Perform IPsec, if any. */ 774 if (ipsec_used) { 775 SOFTNET_LOCK(); 776 if (ipsec4_input(m, IP_FORWARDING | 777 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 778 SOFTNET_UNLOCK(); 779 goto out; 780 } 781 SOFTNET_UNLOCK(); 782 } 783#endif 784 ip_forward(m, srcrt, ifp); 785 m_put_rcvif_psref(ifp, &psref); 786 } 787 return; 788 789ours: 790 m_put_rcvif_psref(ifp, &psref); 791 ifp = NULL; 792 793 /* 794 * If offset or IP_MF are set, must reassemble. 795 */ 796 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 797 /* 798 * Pass to IP reassembly mechanism. 799 */ 800 if (ip_reass_packet(&m, ip) != 0) { 801 /* Failed; invalid fragment(s) or packet. */ 802 goto out; 803 } 804 if (m == NULL) { 805 /* More fragments should come; silently return. */ 806 goto out; 807 } 808 /* 809 * Reassembly is done, we have the final packet. 810 * Updated cached data in local variable(s). 811 */ 812 ip = mtod(m, struct ip *); 813 hlen = ip->ip_hl << 2; 814 } 815 816#ifdef IPSEC 817 /* 818 * Enforce IPsec policy checking if we are seeing last header. 819 * Note that we do not visit this with protocols with PCB layer 820 * code - like UDP/TCP/raw IP. 821 */ 822 if (ipsec_used && 823 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 824 SOFTNET_LOCK(); 825 if (ipsec4_input(m, 0) != 0) { 826 SOFTNET_UNLOCK(); 827 goto out; 828 } 829 SOFTNET_UNLOCK(); 830 } 831#endif 832 833 /* 834 * Switch out to protocol's input routine. 835 */ 836#if IFA_STATS 837 if (ia && ip) { 838 struct in_ifaddr *_ia; 839 /* 840 * Keep a reference from ip_match_our_address with psref 841 * is expensive, so explore ia here again. 842 */ 843 s = pserialize_read_enter(); 844 _ia = in_get_ia(ip->ip_dst); 845 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 846 pserialize_read_exit(s); 847 } 848#endif 849 IP_STATINC(IP_STAT_DELIVERED); 850 851 const int off = hlen, nh = ip->ip_p; 852 853 SOFTNET_LOCK(); 854 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 855 SOFTNET_UNLOCK(); 856 return; 857 858out: 859 m_put_rcvif_psref(ifp, &psref); 860 if (m != NULL) 861 m_freem(m); 862} 863 864/* 865 * IP timer processing. 866 */ 867void 868ip_slowtimo(void) 869{ 870 871#ifndef NET_MPSAFE 872 mutex_enter(softnet_lock); 873 KERNEL_LOCK(1, NULL); 874#endif 875 876 ip_reass_slowtimo(); 877 878#ifndef NET_MPSAFE 879 KERNEL_UNLOCK_ONE(NULL); 880 mutex_exit(softnet_lock); 881#endif 882} 883 884/* 885 * IP drain processing. 886 */ 887void 888ip_drain(void) 889{ 890 891 KERNEL_LOCK(1, NULL); 892 ip_reass_drain(); 893 KERNEL_UNLOCK_ONE(NULL); 894} 895 896/* 897 * ip_dooptions: perform option processing on a datagram, possibly discarding 898 * it if bad options are encountered, or forwarding it if source-routed. 899 * 900 * => Returns true if packet has been forwarded/freed. 901 * => Returns false if the packet should be processed further. 902 */ 903static bool 904ip_dooptions(struct mbuf *m) 905{ 906 struct ip *ip = mtod(m, struct ip *); 907 u_char *cp, *cp0; 908 struct ip_timestamp *ipt; 909 struct in_ifaddr *ia; 910 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 911 struct in_addr dst; 912 n_time ntime; 913 struct ifaddr *ifa = NULL; 914 int s; 915 916 dst = ip->ip_dst; 917 cp = (u_char *)(ip + 1); 918 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 919 for (; cnt > 0; cnt -= optlen, cp += optlen) { 920 opt = cp[IPOPT_OPTVAL]; 921 if (opt == IPOPT_EOL) 922 break; 923 if (opt == IPOPT_NOP) 924 optlen = 1; 925 else { 926 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 927 code = &cp[IPOPT_OLEN] - (u_char *)ip; 928 goto bad; 929 } 930 optlen = cp[IPOPT_OLEN]; 931 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 932 code = &cp[IPOPT_OLEN] - (u_char *)ip; 933 goto bad; 934 } 935 } 936 switch (opt) { 937 938 default: 939 break; 940 941 /* 942 * Source routing with record. 943 * Find interface with current destination address. 944 * If none on this machine then drop if strictly routed, 945 * or do nothing if loosely routed. 946 * Record interface address and bring up next address 947 * component. If strictly routed make sure next 948 * address is on directly accessible net. 949 */ 950 case IPOPT_LSRR: 951 case IPOPT_SSRR: { 952 struct psref psref; 953 if (ip_allowsrcrt == 0) { 954 type = ICMP_UNREACH; 955 code = ICMP_UNREACH_NET_PROHIB; 956 goto bad; 957 } 958 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 959 code = &cp[IPOPT_OLEN] - (u_char *)ip; 960 goto bad; 961 } 962 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 963 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 964 goto bad; 965 } 966 ipaddr.sin_addr = ip->ip_dst; 967 968 s = pserialize_read_enter(); 969 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 970 if (ifa == NULL) { 971 pserialize_read_exit(s); 972 if (opt == IPOPT_SSRR) { 973 type = ICMP_UNREACH; 974 code = ICMP_UNREACH_SRCFAIL; 975 goto bad; 976 } 977 /* 978 * Loose routing, and not at next destination 979 * yet; nothing to do except forward. 980 */ 981 break; 982 } 983 pserialize_read_exit(s); 984 985 off--; /* 0 origin */ 986 if ((off + sizeof(struct in_addr)) > optlen) { 987 /* 988 * End of source route. Should be for us. 989 */ 990 save_rte(cp, ip->ip_src); 991 break; 992 } 993 /* 994 * locate outgoing interface 995 */ 996 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 997 sizeof(ipaddr.sin_addr)); 998 if (opt == IPOPT_SSRR) { 999 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 1000 &psref); 1001 if (ifa != NULL) 1002 ia = ifatoia(ifa); 1003 else 1004 ia = NULL; 1005 } else { 1006 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1007 } 1008 if (ia == NULL) { 1009 type = ICMP_UNREACH; 1010 code = ICMP_UNREACH_SRCFAIL; 1011 goto bad; 1012 } 1013 ip->ip_dst = ipaddr.sin_addr; 1014 bcopy((void *)&ia->ia_addr.sin_addr, 1015 (void *)(cp + off), sizeof(struct in_addr)); 1016 ia4_release(ia, &psref); 1017 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1018 /* 1019 * Let ip_intr's mcast routing check handle mcast pkts 1020 */ 1021 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1022 break; 1023 } 1024 1025 case IPOPT_RR: { 1026 struct psref psref; 1027 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1028 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1029 goto bad; 1030 } 1031 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1032 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1033 goto bad; 1034 } 1035 /* 1036 * If no space remains, ignore. 1037 */ 1038 off--; /* 0 origin */ 1039 if ((off + sizeof(struct in_addr)) > optlen) 1040 break; 1041 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1042 sizeof(ipaddr.sin_addr)); 1043 /* 1044 * locate outgoing interface; if we're the destination, 1045 * use the incoming interface (should be same). 1046 */ 1047 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1048 if (ifa == NULL) { 1049 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1050 if (ia == NULL) { 1051 type = ICMP_UNREACH; 1052 code = ICMP_UNREACH_HOST; 1053 goto bad; 1054 } 1055 } else { 1056 ia = ifatoia(ifa); 1057 } 1058 bcopy((void *)&ia->ia_addr.sin_addr, 1059 (void *)(cp + off), sizeof(struct in_addr)); 1060 ia4_release(ia, &psref); 1061 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1062 break; 1063 } 1064 1065 case IPOPT_TS: 1066 code = cp - (u_char *)ip; 1067 ipt = (struct ip_timestamp *)cp; 1068 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1069 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1070 goto bad; 1071 } 1072 if (ipt->ipt_ptr < 5) { 1073 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1074 goto bad; 1075 } 1076 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1077 if (++ipt->ipt_oflw == 0) { 1078 code = (u_char *)&ipt->ipt_ptr - 1079 (u_char *)ip; 1080 goto bad; 1081 } 1082 break; 1083 } 1084 cp0 = (cp + ipt->ipt_ptr - 1); 1085 switch (ipt->ipt_flg) { 1086 1087 case IPOPT_TS_TSONLY: 1088 break; 1089 1090 case IPOPT_TS_TSANDADDR: { 1091 struct ifnet *rcvif; 1092 int _s, _ss; 1093 1094 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1095 sizeof(struct in_addr) > ipt->ipt_len) { 1096 code = (u_char *)&ipt->ipt_ptr - 1097 (u_char *)ip; 1098 goto bad; 1099 } 1100 ipaddr.sin_addr = dst; 1101 _ss = pserialize_read_enter(); 1102 rcvif = m_get_rcvif(m, &_s); 1103 if (__predict_true(rcvif != NULL)) { 1104 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1105 rcvif); 1106 } 1107 m_put_rcvif(rcvif, &_s); 1108 if (ifa == NULL) { 1109 pserialize_read_exit(_ss); 1110 break; 1111 } 1112 ia = ifatoia(ifa); 1113 bcopy(&ia->ia_addr.sin_addr, 1114 cp0, sizeof(struct in_addr)); 1115 pserialize_read_exit(_ss); 1116 ipt->ipt_ptr += sizeof(struct in_addr); 1117 break; 1118 } 1119 1120 case IPOPT_TS_PRESPEC: 1121 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1122 sizeof(struct in_addr) > ipt->ipt_len) { 1123 code = (u_char *)&ipt->ipt_ptr - 1124 (u_char *)ip; 1125 goto bad; 1126 } 1127 memcpy(&ipaddr.sin_addr, cp0, 1128 sizeof(struct in_addr)); 1129 s = pserialize_read_enter(); 1130 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1131 if (ifa == NULL) { 1132 pserialize_read_exit(s); 1133 continue; 1134 } 1135 pserialize_read_exit(s); 1136 ipt->ipt_ptr += sizeof(struct in_addr); 1137 break; 1138 1139 default: 1140 /* XXX can't take &ipt->ipt_flg */ 1141 code = (u_char *)&ipt->ipt_ptr - 1142 (u_char *)ip + 1; 1143 goto bad; 1144 } 1145 ntime = iptime(); 1146 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1147 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1148 sizeof(n_time)); 1149 ipt->ipt_ptr += sizeof(n_time); 1150 } 1151 } 1152 if (forward) { 1153 struct ifnet *rcvif; 1154 struct psref _psref; 1155 1156 if (ip_forwsrcrt == 0) { 1157 type = ICMP_UNREACH; 1158 code = ICMP_UNREACH_SRCFAIL; 1159 goto bad; 1160 } 1161 1162 rcvif = m_get_rcvif_psref(m, &_psref); 1163 if (__predict_false(rcvif == NULL)) { 1164 type = ICMP_UNREACH; 1165 code = ICMP_UNREACH_HOST; 1166 goto bad; 1167 } 1168 ip_forward(m, 1, rcvif); 1169 m_put_rcvif_psref(rcvif, &_psref); 1170 return true; 1171 } 1172 return false; 1173bad: 1174 icmp_error(m, type, code, 0, 0); 1175 IP_STATINC(IP_STAT_BADOPTIONS); 1176 return true; 1177} 1178 1179/* 1180 * ip_rtaddr: given address of next destination (final or next hop), 1181 * return internet address info of interface to be used to get there. 1182 */ 1183static struct in_ifaddr * 1184ip_rtaddr(struct in_addr dst, struct psref *psref) 1185{ 1186 struct rtentry *rt; 1187 union { 1188 struct sockaddr dst; 1189 struct sockaddr_in dst4; 1190 } u; 1191 struct route *ro; 1192 1193 sockaddr_in_init(&u.dst4, &dst, 0); 1194 1195 ro = percpu_getref(ipforward_rt_percpu); 1196 rt = rtcache_lookup(ro, &u.dst); 1197 if (rt == NULL) { 1198 percpu_putref(ipforward_rt_percpu); 1199 return NULL; 1200 } 1201 1202 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1203 rtcache_unref(rt, ro); 1204 percpu_putref(ipforward_rt_percpu); 1205 1206 return ifatoia(rt->rt_ifa); 1207} 1208 1209/* 1210 * save_rte: save incoming source route for use in replies, to be picked 1211 * up later by ip_srcroute if the receiver is interested. 1212 */ 1213static void 1214save_rte(u_char *option, struct in_addr dst) 1215{ 1216 unsigned olen; 1217 1218 olen = option[IPOPT_OLEN]; 1219 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1220 return; 1221 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen); 1222 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1223 ip_srcrt.dst = dst; 1224} 1225 1226/* 1227 * Retrieve incoming source route for use in replies, 1228 * in the same form used by setsockopt. 1229 * The first hop is placed before the options, will be removed later. 1230 */ 1231struct mbuf * 1232ip_srcroute(void) 1233{ 1234 struct in_addr *p, *q; 1235 struct mbuf *m; 1236 1237 if (ip_nhops == 0) 1238 return NULL; 1239 m = m_get(M_DONTWAIT, MT_SOOPTS); 1240 if (m == 0) 1241 return NULL; 1242 1243 MCLAIM(m, &inetdomain.dom_mowner); 1244#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1245 1246 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1247 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1248 OPTSIZ; 1249 1250 /* 1251 * First save first hop for return route 1252 */ 1253 p = &ip_srcrt.route[ip_nhops - 1]; 1254 *(mtod(m, struct in_addr *)) = *p--; 1255 1256 /* 1257 * Copy option fields and padding (nop) to mbuf. 1258 */ 1259 ip_srcrt.nop = IPOPT_NOP; 1260 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1261 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.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 >= ip_srcrt.route) { 1271 *q++ = *p--; 1272 } 1273 /* 1274 * Last hop goes to final destination. 1275 */ 1276 *q = ip_srcrt.dst; 1277 return (m); 1278} 1279 1280const int inetctlerrmap[PRC_NCMDS] = { 1281 [PRC_MSGSIZE] = EMSGSIZE, 1282 [PRC_HOSTDEAD] = EHOSTDOWN, 1283 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1284 [PRC_UNREACH_NET] = EHOSTUNREACH, 1285 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1286 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1287 [PRC_UNREACH_PORT] = ECONNREFUSED, 1288 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1289 [PRC_PARAMPROB] = ENOPROTOOPT, 1290}; 1291 1292void 1293ip_fasttimo(void) 1294{ 1295 if (ip_drainwanted) { 1296 ip_drain(); 1297 ip_drainwanted = 0; 1298 } 1299} 1300 1301void 1302ip_drainstub(void) 1303{ 1304 ip_drainwanted = 1; 1305} 1306 1307/* 1308 * Forward a packet. If some error occurs return the sender 1309 * an icmp packet. Note we can't always generate a meaningful 1310 * icmp message because icmp doesn't have a large enough repertoire 1311 * of codes and types. 1312 * 1313 * If not forwarding, just drop the packet. This could be confusing 1314 * if ipforwarding was zero but some routing protocol was advancing 1315 * us as a gateway to somewhere. However, we must let the routing 1316 * protocol deal with that. 1317 * 1318 * The srcrt parameter indicates whether the packet is being forwarded 1319 * via a source route. 1320 */ 1321static void 1322ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1323{ 1324 struct ip *ip = mtod(m, struct ip *); 1325 struct rtentry *rt; 1326 int error, type = 0, code = 0, destmtu = 0; 1327 struct mbuf *mcopy; 1328 n_long dest; 1329 union { 1330 struct sockaddr dst; 1331 struct sockaddr_in dst4; 1332 } u; 1333 uint64_t *ips; 1334 struct route *ro; 1335 1336 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1337 "interrupt handler; synchronization assumptions violated"); 1338 1339 /* 1340 * We are now in the output path. 1341 */ 1342 MCLAIM(m, &ip_tx_mowner); 1343 1344 /* 1345 * Clear any in-bound checksum flags for this packet. 1346 */ 1347 m->m_pkthdr.csum_flags = 0; 1348 1349 dest = 0; 1350 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1351 IP_STATINC(IP_STAT_CANTFORWARD); 1352 m_freem(m); 1353 return; 1354 } 1355 1356 if (ip->ip_ttl <= IPTTLDEC) { 1357 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1358 return; 1359 } 1360 1361 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1362 1363 ro = percpu_getref(ipforward_rt_percpu); 1364 rt = rtcache_lookup(ro, &u.dst); 1365 if (rt == NULL) { 1366 percpu_putref(ipforward_rt_percpu); 1367 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1368 return; 1369 } 1370 1371 /* 1372 * Save at most 68 bytes of the packet in case 1373 * we need to generate an ICMP message to the src. 1374 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1375 */ 1376 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1377 if (mcopy) 1378 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1379 1380 ip->ip_ttl -= IPTTLDEC; 1381 1382 /* 1383 * If forwarding packet using same interface that it came in on, 1384 * perhaps should send a redirect to sender to shortcut a hop. 1385 * Only send redirect if source is sending directly to us, 1386 * and if packet was not source routed (or has any options). 1387 * Also, don't send redirect if forwarding using a default route 1388 * or a route modified by a redirect. 1389 */ 1390 if (rt->rt_ifp == rcvif && 1391 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1392 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1393 ipsendredirects && !srcrt) { 1394 if (rt->rt_ifa && 1395 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1396 ifatoia(rt->rt_ifa)->ia_subnet) { 1397 if (rt->rt_flags & RTF_GATEWAY) 1398 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1399 else 1400 dest = ip->ip_dst.s_addr; 1401 /* 1402 * Router requirements says to only send host 1403 * redirects. 1404 */ 1405 type = ICMP_REDIRECT; 1406 code = ICMP_REDIRECT_HOST; 1407 } 1408 } 1409 rtcache_unref(rt, ro); 1410 1411 error = ip_output(m, NULL, ro, 1412 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1413 NULL, NULL); 1414 1415 if (error) { 1416 IP_STATINC(IP_STAT_CANTFORWARD); 1417 goto error; 1418 } 1419 1420 ips = IP_STAT_GETREF(); 1421 ips[IP_STAT_FORWARD]++; 1422 1423 if (type) { 1424 ips[IP_STAT_REDIRECTSENT]++; 1425 IP_STAT_PUTREF(); 1426 goto redirect; 1427 } 1428 1429 IP_STAT_PUTREF(); 1430 if (mcopy) { 1431#ifdef GATEWAY 1432 if (mcopy->m_flags & M_CANFASTFWD) 1433 ipflow_create(ro, mcopy); 1434#endif 1435 m_freem(mcopy); 1436 } 1437 1438 percpu_putref(ipforward_rt_percpu); 1439 return; 1440 1441redirect: 1442error: 1443 if (mcopy == NULL) { 1444 percpu_putref(ipforward_rt_percpu); 1445 return; 1446 } 1447 1448 switch (error) { 1449 1450 case 0: /* forwarded, but need redirect */ 1451 /* type, code set above */ 1452 break; 1453 1454 case ENETUNREACH: /* shouldn't happen, checked above */ 1455 case EHOSTUNREACH: 1456 case ENETDOWN: 1457 case EHOSTDOWN: 1458 default: 1459 type = ICMP_UNREACH; 1460 code = ICMP_UNREACH_HOST; 1461 break; 1462 1463 case EMSGSIZE: 1464 type = ICMP_UNREACH; 1465 code = ICMP_UNREACH_NEEDFRAG; 1466 1467 if ((rt = rtcache_validate(ro)) != NULL) { 1468 destmtu = rt->rt_ifp->if_mtu; 1469 rtcache_unref(rt, ro); 1470 } 1471#ifdef IPSEC 1472 if (ipsec_used) 1473 (void)ipsec4_forward(mcopy, &destmtu); 1474#endif 1475 IP_STATINC(IP_STAT_CANTFRAG); 1476 break; 1477 1478 case ENOBUFS: 1479 /* 1480 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1481 * Requirements for IP Version 4 Routers. Source quench can 1482 * big problem under DoS attacks or if the underlying 1483 * interface is rate-limited. 1484 */ 1485 if (mcopy) 1486 m_freem(mcopy); 1487 percpu_putref(ipforward_rt_percpu); 1488 return; 1489 } 1490 icmp_error(mcopy, type, code, dest, destmtu); 1491 percpu_putref(ipforward_rt_percpu); 1492} 1493 1494void 1495ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1496 struct mbuf *m) 1497{ 1498 struct socket *so = inp->inp_socket; 1499 ifnet_t *ifp; 1500 int inpflags = inp->inp_flags; 1501 struct psref psref; 1502 1503 ifp = m_get_rcvif_psref(m, &psref); 1504 if (__predict_false(ifp == NULL)) 1505 return; /* XXX should report error? */ 1506 1507 if (so->so_options & SO_TIMESTAMP 1508#ifdef SO_OTIMESTAMP 1509 || so->so_options & SO_OTIMESTAMP 1510#endif 1511 ) { 1512 struct timeval tv; 1513 1514 microtime(&tv); 1515#ifdef SO_OTIMESTAMP 1516 if (so->so_options & SO_OTIMESTAMP) { 1517 struct timeval50 tv50; 1518 timeval_to_timeval50(&tv, &tv50); 1519 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50), 1520 SCM_OTIMESTAMP, SOL_SOCKET); 1521 } else 1522#endif 1523 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 1524 SCM_TIMESTAMP, SOL_SOCKET); 1525 if (*mp) 1526 mp = &(*mp)->m_next; 1527 } 1528 if (inpflags & INP_RECVDSTADDR) { 1529 *mp = sbcreatecontrol((void *) &ip->ip_dst, 1530 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1531 if (*mp) 1532 mp = &(*mp)->m_next; 1533 } 1534 if (inpflags & INP_RECVPKTINFO) { 1535 struct in_pktinfo ipi; 1536 ipi.ipi_addr = ip->ip_src; 1537 ipi.ipi_ifindex = ifp->if_index; 1538 *mp = sbcreatecontrol((void *) &ipi, 1539 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP); 1540 if (*mp) 1541 mp = &(*mp)->m_next; 1542 } 1543 if (inpflags & INP_PKTINFO) { 1544 struct in_pktinfo ipi; 1545 ipi.ipi_addr = ip->ip_dst; 1546 ipi.ipi_ifindex = ifp->if_index; 1547 *mp = sbcreatecontrol((void *) &ipi, 1548 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1549 if (*mp) 1550 mp = &(*mp)->m_next; 1551 } 1552 if (inpflags & INP_RECVIF) { 1553 struct sockaddr_dl sdl; 1554 1555 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1556 NULL, 0); 1557 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1558 if (*mp) 1559 mp = &(*mp)->m_next; 1560 } 1561 if (inpflags & INP_RECVTTL) { 1562 *mp = sbcreatecontrol((void *) &ip->ip_ttl, 1563 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1564 if (*mp) 1565 mp = &(*mp)->m_next; 1566 } 1567 m_put_rcvif_psref(ifp, &psref); 1568} 1569 1570/* 1571 * sysctl helper routine for net.inet.ip.forwsrcrt. 1572 */ 1573static int 1574sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1575{ 1576 int error, tmp; 1577 struct sysctlnode node; 1578 1579 node = *rnode; 1580 tmp = ip_forwsrcrt; 1581 node.sysctl_data = &tmp; 1582 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1583 if (error || newp == NULL) 1584 return (error); 1585 1586 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1587 0, NULL, NULL, NULL); 1588 if (error) 1589 return (error); 1590 1591 ip_forwsrcrt = tmp; 1592 1593 return (0); 1594} 1595 1596/* 1597 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1598 * range of the new value and tweaks timers if it changes. 1599 */ 1600static int 1601sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1602{ 1603 int error, tmp; 1604 struct sysctlnode node; 1605 1606 icmp_mtudisc_lock(); 1607 1608 node = *rnode; 1609 tmp = ip_mtudisc_timeout; 1610 node.sysctl_data = &tmp; 1611 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1612 if (error || newp == NULL) 1613 goto out; 1614 if (tmp < 0) { 1615 error = EINVAL; 1616 goto out; 1617 } 1618 1619 ip_mtudisc_timeout = tmp; 1620 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1621 error = 0; 1622out: 1623 icmp_mtudisc_unlock(); 1624 return error; 1625} 1626 1627static int 1628sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1629{ 1630 1631 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1632} 1633 1634static void 1635sysctl_net_inet_ip_setup(struct sysctllog **clog) 1636{ 1637 sysctl_createv(clog, 0, NULL, NULL, 1638 CTLFLAG_PERMANENT, 1639 CTLTYPE_NODE, "inet", 1640 SYSCTL_DESCR("PF_INET related settings"), 1641 NULL, 0, NULL, 0, 1642 CTL_NET, PF_INET, CTL_EOL); 1643 sysctl_createv(clog, 0, NULL, NULL, 1644 CTLFLAG_PERMANENT, 1645 CTLTYPE_NODE, "ip", 1646 SYSCTL_DESCR("IPv4 related settings"), 1647 NULL, 0, NULL, 0, 1648 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1649 1650 sysctl_createv(clog, 0, NULL, NULL, 1651 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1652 CTLTYPE_INT, "forwarding", 1653 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1654 NULL, 0, &ipforwarding, 0, 1655 CTL_NET, PF_INET, IPPROTO_IP, 1656 IPCTL_FORWARDING, CTL_EOL); 1657 sysctl_createv(clog, 0, NULL, NULL, 1658 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1659 CTLTYPE_INT, "redirect", 1660 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1661 NULL, 0, &ipsendredirects, 0, 1662 CTL_NET, PF_INET, IPPROTO_IP, 1663 IPCTL_SENDREDIRECTS, CTL_EOL); 1664 sysctl_createv(clog, 0, NULL, NULL, 1665 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1666 CTLTYPE_INT, "ttl", 1667 SYSCTL_DESCR("Default TTL for an INET datagram"), 1668 NULL, 0, &ip_defttl, 0, 1669 CTL_NET, PF_INET, IPPROTO_IP, 1670 IPCTL_DEFTTL, CTL_EOL); 1671#ifdef IPCTL_DEFMTU 1672 sysctl_createv(clog, 0, NULL, NULL, 1673 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1674 CTLTYPE_INT, "mtu", 1675 SYSCTL_DESCR("Default MTA for an INET route"), 1676 NULL, 0, &ip_mtu, 0, 1677 CTL_NET, PF_INET, IPPROTO_IP, 1678 IPCTL_DEFMTU, CTL_EOL); 1679#endif /* IPCTL_DEFMTU */ 1680 sysctl_createv(clog, 0, NULL, NULL, 1681 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1682 CTLTYPE_INT, "forwsrcrt", 1683 SYSCTL_DESCR("Enable forwarding of source-routed " 1684 "datagrams"), 1685 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1686 CTL_NET, PF_INET, IPPROTO_IP, 1687 IPCTL_FORWSRCRT, CTL_EOL); 1688 sysctl_createv(clog, 0, NULL, NULL, 1689 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1690 CTLTYPE_INT, "directed-broadcast", 1691 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1692 NULL, 0, &ip_directedbcast, 0, 1693 CTL_NET, PF_INET, IPPROTO_IP, 1694 IPCTL_DIRECTEDBCAST, CTL_EOL); 1695 sysctl_createv(clog, 0, NULL, NULL, 1696 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1697 CTLTYPE_INT, "allowsrcrt", 1698 SYSCTL_DESCR("Accept source-routed datagrams"), 1699 NULL, 0, &ip_allowsrcrt, 0, 1700 CTL_NET, PF_INET, IPPROTO_IP, 1701 IPCTL_ALLOWSRCRT, CTL_EOL); 1702 1703 sysctl_createv(clog, 0, NULL, NULL, 1704 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1705 CTLTYPE_INT, "mtudisc", 1706 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1707 NULL, 0, &ip_mtudisc, 0, 1708 CTL_NET, PF_INET, IPPROTO_IP, 1709 IPCTL_MTUDISC, CTL_EOL); 1710 sysctl_createv(clog, 0, NULL, NULL, 1711 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1712 CTLTYPE_INT, "anonportmin", 1713 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1714 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1715 CTL_NET, PF_INET, IPPROTO_IP, 1716 IPCTL_ANONPORTMIN, CTL_EOL); 1717 sysctl_createv(clog, 0, NULL, NULL, 1718 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1719 CTLTYPE_INT, "anonportmax", 1720 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1721 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1722 CTL_NET, PF_INET, IPPROTO_IP, 1723 IPCTL_ANONPORTMAX, CTL_EOL); 1724 sysctl_createv(clog, 0, NULL, NULL, 1725 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1726 CTLTYPE_INT, "mtudisctimeout", 1727 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1728 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1729 CTL_NET, PF_INET, IPPROTO_IP, 1730 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1731#ifndef IPNOPRIVPORTS 1732 sysctl_createv(clog, 0, NULL, NULL, 1733 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1734 CTLTYPE_INT, "lowportmin", 1735 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1736 "to assign"), 1737 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1738 CTL_NET, PF_INET, IPPROTO_IP, 1739 IPCTL_LOWPORTMIN, CTL_EOL); 1740 sysctl_createv(clog, 0, NULL, NULL, 1741 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1742 CTLTYPE_INT, "lowportmax", 1743 SYSCTL_DESCR("Highest privileged ephemeral port number " 1744 "to assign"), 1745 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1746 CTL_NET, PF_INET, IPPROTO_IP, 1747 IPCTL_LOWPORTMAX, CTL_EOL); 1748#endif /* IPNOPRIVPORTS */ 1749#if NGRE > 0 1750 sysctl_createv(clog, 0, NULL, NULL, 1751 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1752 CTLTYPE_INT, "grettl", 1753 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1754 NULL, 0, &ip_gre_ttl, 0, 1755 CTL_NET, PF_INET, IPPROTO_IP, 1756 IPCTL_GRE_TTL, CTL_EOL); 1757#endif /* NGRE */ 1758 sysctl_createv(clog, 0, NULL, NULL, 1759 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1760 CTLTYPE_INT, "checkinterface", 1761 SYSCTL_DESCR("Enable receive side of Strong ES model " 1762 "from RFC1122"), 1763 NULL, 0, &ip_checkinterface, 0, 1764 CTL_NET, PF_INET, IPPROTO_IP, 1765 IPCTL_CHECKINTERFACE, CTL_EOL); 1766 sysctl_createv(clog, 0, NULL, NULL, 1767 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1768 CTLTYPE_INT, "random_id", 1769 SYSCTL_DESCR("Assign random ip_id values"), 1770 NULL, 0, &ip_do_randomid, 0, 1771 CTL_NET, PF_INET, IPPROTO_IP, 1772 IPCTL_RANDOMID, CTL_EOL); 1773 sysctl_createv(clog, 0, NULL, NULL, 1774 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1775 CTLTYPE_INT, "do_loopback_cksum", 1776 SYSCTL_DESCR("Perform IP checksum on loopback"), 1777 NULL, 0, &ip_do_loopback_cksum, 0, 1778 CTL_NET, PF_INET, IPPROTO_IP, 1779 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1780 sysctl_createv(clog, 0, NULL, NULL, 1781 CTLFLAG_PERMANENT, 1782 CTLTYPE_STRUCT, "stats", 1783 SYSCTL_DESCR("IP statistics"), 1784 sysctl_net_inet_ip_stats, 0, NULL, 0, 1785 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1786 CTL_EOL); 1787#if NARP 1788 sysctl_createv(clog, 0, NULL, NULL, 1789 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1790 CTLTYPE_INT, "dad_count", 1791 SYSCTL_DESCR("Number of Duplicate Address Detection " 1792 "probes to send"), 1793 NULL, 0, &ip_dad_count, 0, 1794 CTL_NET, PF_INET, IPPROTO_IP, 1795 IPCTL_DAD_COUNT, CTL_EOL); 1796#endif 1797 1798 /* anonportalgo RFC6056 subtree */ 1799 const struct sysctlnode *portalgo_node; 1800 sysctl_createv(clog, 0, NULL, &portalgo_node, 1801 CTLFLAG_PERMANENT, 1802 CTLTYPE_NODE, "anonportalgo", 1803 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1804 NULL, 0, NULL, 0, 1805 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1806 sysctl_createv(clog, 0, &portalgo_node, NULL, 1807 CTLFLAG_PERMANENT, 1808 CTLTYPE_STRING, "available", 1809 SYSCTL_DESCR("available algorithms"), 1810 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1811 CTL_CREATE, CTL_EOL); 1812 sysctl_createv(clog, 0, &portalgo_node, NULL, 1813 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1814 CTLTYPE_STRING, "selected", 1815 SYSCTL_DESCR("selected algorithm"), 1816 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1817 CTL_CREATE, CTL_EOL); 1818 sysctl_createv(clog, 0, &portalgo_node, NULL, 1819 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1820 CTLTYPE_STRUCT, "reserve", 1821 SYSCTL_DESCR("bitmap of reserved ports"), 1822 sysctl_portalgo_reserve4, 0, NULL, 0, 1823 CTL_CREATE, CTL_EOL); 1824} 1825 1826void 1827ip_statinc(u_int stat) 1828{ 1829 1830 KASSERT(stat < IP_NSTATS); 1831 IP_STATINC(stat); 1832} 1833