ip_input.c revision 1.360
1/* $NetBSD: ip_input.c,v 1.360 2017/07/27 06:59:28 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 1.360 2017/07/27 06:59:28 ozaki-r Exp $"); 95 96#ifdef _KERNEL_OPT 97#include "opt_inet.h" 98#include "opt_gateway.h" 99#include "opt_ipsec.h" 100#include "opt_mrouting.h" 101#include "opt_mbuftrace.h" 102#include "opt_inet_csum.h" 103#include "opt_net_mpsafe.h" 104#endif 105 106#include "arp.h" 107 108#include <sys/param.h> 109#include <sys/systm.h> 110#include <sys/cpu.h> 111#include <sys/mbuf.h> 112#include <sys/domain.h> 113#include <sys/protosw.h> 114#include <sys/socket.h> 115#include <sys/socketvar.h> 116#include <sys/errno.h> 117#include <sys/time.h> 118#include <sys/kernel.h> 119#include <sys/pool.h> 120#include <sys/sysctl.h> 121#include <sys/kauth.h> 122 123#include <net/if.h> 124#include <net/if_dl.h> 125#include <net/route.h> 126#include <net/pktqueue.h> 127#include <net/pfil.h> 128 129#include <netinet/in.h> 130#include <netinet/in_systm.h> 131#include <netinet/ip.h> 132#include <netinet/in_pcb.h> 133#include <netinet/in_proto.h> 134#include <netinet/in_var.h> 135#include <netinet/ip_var.h> 136#include <netinet/ip_private.h> 137#include <netinet/ip_icmp.h> 138/* just for gif_ttl */ 139#include <netinet/in_gif.h> 140#include "gif.h" 141#include <net/if_gre.h> 142#include "gre.h" 143 144#ifdef MROUTING 145#include <netinet/ip_mroute.h> 146#endif 147#include <netinet/portalgo.h> 148 149#ifdef IPSEC 150#include <netipsec/ipsec.h> 151#endif 152 153#ifndef IPFORWARDING 154#ifdef GATEWAY 155#define IPFORWARDING 1 /* forward IP packets not for us */ 156#else /* GATEWAY */ 157#define IPFORWARDING 0 /* don't forward IP packets not for us */ 158#endif /* GATEWAY */ 159#endif /* IPFORWARDING */ 160#ifndef IPSENDREDIRECTS 161#define IPSENDREDIRECTS 1 162#endif 163#ifndef IPFORWSRCRT 164#define IPFORWSRCRT 1 /* forward source-routed packets */ 165#endif 166#ifndef IPALLOWSRCRT 167#define IPALLOWSRCRT 1 /* allow source-routed packets */ 168#endif 169#ifndef IPMTUDISC 170#define IPMTUDISC 1 171#endif 172#ifndef IPMTUDISCTIMEOUT 173#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 174#endif 175 176/* 177 * Note: DIRECTED_BROADCAST is handled this way so that previous 178 * configuration using this option will Just Work. 179 */ 180#ifndef IPDIRECTEDBCAST 181#ifdef DIRECTED_BROADCAST 182#define IPDIRECTEDBCAST 1 183#else 184#define IPDIRECTEDBCAST 0 185#endif /* DIRECTED_BROADCAST */ 186#endif /* IPDIRECTEDBCAST */ 187int ipforwarding = IPFORWARDING; 188int ipsendredirects = IPSENDREDIRECTS; 189int ip_defttl = IPDEFTTL; 190int ip_forwsrcrt = IPFORWSRCRT; 191int ip_directedbcast = IPDIRECTEDBCAST; 192int ip_allowsrcrt = IPALLOWSRCRT; 193int ip_mtudisc = IPMTUDISC; 194int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 195#ifdef DIAGNOSTIC 196int ipprintfs = 0; 197#endif 198 199int ip_do_randomid = 0; 200 201/* 202 * XXX - Setting ip_checkinterface mostly implements the receive side of 203 * the Strong ES model described in RFC 1122, but since the routing table 204 * and transmit implementation do not implement the Strong ES model, 205 * setting this to 1 results in an odd hybrid. 206 * 207 * XXX - ip_checkinterface currently must be disabled if you use ipnat 208 * to translate the destination address to another local interface. 209 * 210 * XXX - ip_checkinterface must be disabled if you add IP aliases 211 * to the loopback interface instead of the interface where the 212 * packets for those addresses are received. 213 */ 214static int ip_checkinterface __read_mostly = 0; 215 216struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 217 218pktqueue_t * ip_pktq __read_mostly; 219pfil_head_t * inet_pfil_hook __read_mostly; 220ipid_state_t * ip_ids __read_mostly; 221percpu_t * ipstat_percpu __read_mostly; 222 223static percpu_t *ipforward_rt_percpu __cacheline_aligned; 224 225uint16_t ip_id; 226 227#ifdef INET_CSUM_COUNTERS 228#include <sys/device.h> 229 230struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 231 NULL, "inet", "hwcsum bad"); 232struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 233 NULL, "inet", "hwcsum ok"); 234struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 235 NULL, "inet", "swcsum"); 236 237#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 238 239EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 240EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 241EVCNT_ATTACH_STATIC(ip_swcsum); 242 243#else 244 245#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 246 247#endif /* INET_CSUM_COUNTERS */ 248 249/* 250 * Used to save the IP options in case a protocol wants to respond 251 * to an incoming packet over the same route if the packet got here 252 * using IP source routing. This allows connection establishment and 253 * maintenance when the remote end is on a network that is not known 254 * to us. 255 */ 256struct ip_srcrt { 257 int isr_nhops; /* number of hops */ 258 struct in_addr isr_dst; /* final destination */ 259 char isr_nop; /* one NOP to align */ 260 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */ 261 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)]; 262}; 263 264static int ip_drainwanted; 265 266static void save_rte(struct mbuf *, u_char *, struct in_addr); 267 268#ifdef MBUFTRACE 269struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 270struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 271#endif 272 273static void ipintr(void *); 274static void ip_input(struct mbuf *); 275static void ip_forward(struct mbuf *, int, struct ifnet *); 276static bool ip_dooptions(struct mbuf *); 277static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *); 278static void sysctl_net_inet_ip_setup(struct sysctllog **); 279 280static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *, 281 int *); 282static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *, 283 struct ip *); 284 285#ifdef NET_MPSAFE 286#define SOFTNET_LOCK() mutex_enter(softnet_lock) 287#define SOFTNET_UNLOCK() mutex_exit(softnet_lock) 288#else 289#define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 290#define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 291#endif 292 293/* 294 * IP initialization: fill in IP protocol switch table. 295 * All protocols not implemented in kernel go to raw IP protocol handler. 296 */ 297void 298ip_init(void) 299{ 300 const struct protosw *pr; 301 302 in_init(); 303 sysctl_net_inet_ip_setup(NULL); 304 305 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 306 KASSERT(pr != NULL); 307 308 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 309 KASSERT(ip_pktq != NULL); 310 311 for (u_int i = 0; i < IPPROTO_MAX; i++) { 312 ip_protox[i] = pr - inetsw; 313 } 314 for (pr = inetdomain.dom_protosw; 315 pr < inetdomain.dom_protoswNPROTOSW; pr++) 316 if (pr->pr_domain->dom_family == PF_INET && 317 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 318 ip_protox[pr->pr_protocol] = pr - inetsw; 319 320 ip_reass_init(); 321 322 ip_ids = ip_id_init(); 323 ip_id = time_uptime & 0xfffff; 324 325#ifdef GATEWAY 326 ipflow_init(); 327#endif 328 329 /* Register our Packet Filter hook. */ 330 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 331 KASSERT(inet_pfil_hook != NULL); 332 333#ifdef MBUFTRACE 334 MOWNER_ATTACH(&ip_tx_mowner); 335 MOWNER_ATTACH(&ip_rx_mowner); 336#endif /* MBUFTRACE */ 337 338 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 339 ipforward_rt_percpu = percpu_alloc(sizeof(struct route)); 340 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 341} 342 343static struct in_ifaddr * 344ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 345{ 346 struct in_ifaddr *ia = NULL; 347 int checkif; 348 349 /* 350 * Enable a consistency check between the destination address 351 * and the arrival interface for a unicast packet (the RFC 1122 352 * strong ES model) if IP forwarding is disabled and the packet 353 * is not locally generated. 354 * 355 * XXX - Checking also should be disabled if the destination 356 * address is ipnat'ed to a different interface. 357 * 358 * XXX - Checking is incompatible with IP aliases added 359 * to the loopback interface instead of the interface where 360 * the packets are received. 361 * 362 * XXX - We need to add a per ifaddr flag for this so that 363 * we get finer grain control. 364 */ 365 checkif = ip_checkinterface && (ipforwarding == 0) && 366 (ifp->if_flags & IFF_LOOPBACK) == 0; 367 368 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 369 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 370 if (ia->ia4_flags & IN_IFF_NOTREADY) 371 continue; 372 if (checkif && ia->ia_ifp != ifp) 373 continue; 374 if ((ia->ia_ifp->if_flags & IFF_UP) != 0 && 375 (ia->ia4_flags & IN_IFF_DETACHED) == 0) 376 break; 377 else 378 (*downmatch)++; 379 } 380 } 381 382 return ia; 383} 384 385static struct in_ifaddr * 386ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 387{ 388 struct in_ifaddr *ia = NULL; 389 struct ifaddr *ifa; 390 391 IFADDR_READER_FOREACH(ifa, ifp) { 392 if (ifa->ifa_addr->sa_family != AF_INET) 393 continue; 394 ia = ifatoia(ifa); 395 if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) 396 continue; 397 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 398 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 399 /* 400 * Look for all-0's host part (old broadcast addr), 401 * either for subnet or net. 402 */ 403 ip->ip_dst.s_addr == ia->ia_subnet || 404 ip->ip_dst.s_addr == ia->ia_net) 405 goto matched; 406 /* 407 * An interface with IP address zero accepts 408 * all packets that arrive on that interface. 409 */ 410 if (in_nullhost(ia->ia_addr.sin_addr)) 411 goto matched; 412 } 413 ia = NULL; 414 415matched: 416 return ia; 417} 418 419/* 420 * IP software interrupt routine. 421 */ 422static void 423ipintr(void *arg __unused) 424{ 425 struct mbuf *m; 426 427 KASSERT(cpu_softintr_p()); 428 429#ifndef NET_MPSAFE 430 mutex_enter(softnet_lock); 431#endif 432 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 433 ip_input(m); 434 } 435#ifndef NET_MPSAFE 436 mutex_exit(softnet_lock); 437#endif 438} 439 440/* 441 * IP input routine. Checksum and byte swap header. If fragmented 442 * try to reassemble. Process options. Pass to next level. 443 */ 444static void 445ip_input(struct mbuf *m) 446{ 447 struct ip *ip = NULL; 448 struct in_ifaddr *ia = NULL; 449 int hlen = 0, len; 450 int downmatch; 451 int srcrt = 0; 452 ifnet_t *ifp; 453 struct psref psref; 454 int s; 455 456 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 457 "interrupt handler; synchronization assumptions violated"); 458 459 MCLAIM(m, &ip_rx_mowner); 460 KASSERT((m->m_flags & M_PKTHDR) != 0); 461 462 ifp = m_get_rcvif_psref(m, &psref); 463 if (__predict_false(ifp == NULL)) 464 goto out; 465 466 /* 467 * If no IP addresses have been set yet but the interfaces 468 * are receiving, can't do anything with incoming packets yet. 469 * Note: we pre-check without locks held. 470 */ 471 if (IN_ADDRLIST_READER_EMPTY()) 472 goto out; 473 IP_STATINC(IP_STAT_TOTAL); 474 475 /* 476 * If the IP header is not aligned, slurp it up into a new 477 * mbuf with space for link headers, in the event we forward 478 * it. Otherwise, if it is aligned, make sure the entire 479 * base IP header is in the first mbuf of the chain. 480 */ 481 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 482 if ((m = m_copyup(m, sizeof(struct ip), 483 (max_linkhdr + 3) & ~3)) == NULL) { 484 /* XXXJRT new stat, please */ 485 IP_STATINC(IP_STAT_TOOSMALL); 486 goto out; 487 } 488 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 489 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 490 IP_STATINC(IP_STAT_TOOSMALL); 491 goto out; 492 } 493 } 494 ip = mtod(m, struct ip *); 495 if (ip->ip_v != IPVERSION) { 496 IP_STATINC(IP_STAT_BADVERS); 497 goto out; 498 } 499 hlen = ip->ip_hl << 2; 500 if (hlen < sizeof(struct ip)) { /* minimum header length */ 501 IP_STATINC(IP_STAT_BADHLEN); 502 goto out; 503 } 504 if (hlen > m->m_len) { 505 if ((m = m_pullup(m, hlen)) == NULL) { 506 IP_STATINC(IP_STAT_BADHLEN); 507 goto out; 508 } 509 ip = mtod(m, struct ip *); 510 } 511 512 /* 513 * RFC1122: packets with a multicast source address are 514 * not allowed. 515 */ 516 if (IN_MULTICAST(ip->ip_src.s_addr)) { 517 IP_STATINC(IP_STAT_BADADDR); 518 goto out; 519 } 520 521 /* 127/8 must not appear on wire - RFC1122 */ 522 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 523 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 524 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 525 IP_STATINC(IP_STAT_BADADDR); 526 goto out; 527 } 528 } 529 530 switch (m->m_pkthdr.csum_flags & 531 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | 532 M_CSUM_IPv4_BAD)) { 533 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 534 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 535 IP_STATINC(IP_STAT_BADSUM); 536 goto out; 537 538 case M_CSUM_IPv4: 539 /* Checksum was okay. */ 540 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 541 break; 542 543 default: 544 /* 545 * Must compute it ourselves. Maybe skip checksum on 546 * loopback interfaces. 547 */ 548 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 549 ip_do_loopback_cksum)) { 550 INET_CSUM_COUNTER_INCR(&ip_swcsum); 551 if (in_cksum(m, hlen) != 0) { 552 IP_STATINC(IP_STAT_BADSUM); 553 goto out; 554 } 555 } 556 break; 557 } 558 559 /* Retrieve the packet length. */ 560 len = ntohs(ip->ip_len); 561 562 /* 563 * Check for additional length bogosity 564 */ 565 if (len < hlen) { 566 IP_STATINC(IP_STAT_BADLEN); 567 goto out; 568 } 569 570 /* 571 * Check that the amount of data in the buffers 572 * is as at least much as the IP header would have us expect. 573 * Trim mbufs if longer than we expect. 574 * Drop packet if shorter than we expect. 575 */ 576 if (m->m_pkthdr.len < len) { 577 IP_STATINC(IP_STAT_TOOSHORT); 578 goto out; 579 } 580 if (m->m_pkthdr.len > len) { 581 if (m->m_len == m->m_pkthdr.len) { 582 m->m_len = len; 583 m->m_pkthdr.len = len; 584 } else 585 m_adj(m, len - m->m_pkthdr.len); 586 } 587 588 /* 589 * Assume that we can create a fast-forward IP flow entry 590 * based on this packet. 591 */ 592 m->m_flags |= M_CANFASTFWD; 593 594 /* 595 * Run through list of hooks for input packets. If there are any 596 * filters which require that additional packets in the flow are 597 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 598 * Note that filters must _never_ set this flag, as another filter 599 * in the list may have previously cleared it. 600 */ 601#if defined(IPSEC) 602 if (!ipsec_used || !ipsec_indone(m)) 603#else 604 if (1) 605#endif 606 { 607 struct in_addr odst = ip->ip_dst; 608 bool freed; 609 610 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 611 if (freed || m == NULL) { 612 m = NULL; 613 goto out; 614 } 615 ip = mtod(m, struct ip *); 616 hlen = ip->ip_hl << 2; 617 618 /* 619 * XXX The setting of "srcrt" here is to prevent ip_forward() 620 * from generating ICMP redirects for packets that have 621 * been redirected by a hook back out on to the same LAN that 622 * they came from and is not an indication that the packet 623 * is being inffluenced by source routing options. This 624 * allows things like 625 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 626 * where tlp0 is both on the 1.1.1.0/24 network and is the 627 * default route for hosts on 1.1.1.0/24. Of course this 628 * also requires a "map tlp0 ..." to complete the story. 629 * One might argue whether or not this kind of network config. 630 * should be supported in this manner... 631 */ 632 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 633 } 634 635#ifdef ALTQ 636 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 637 if (altq_input) { 638 SOFTNET_LOCK(); 639 if ((*altq_input)(m, AF_INET) == 0) { 640 /* Packet dropped by traffic conditioner. */ 641 SOFTNET_UNLOCK(); 642 m = NULL; 643 goto out; 644 } 645 SOFTNET_UNLOCK(); 646 } 647#endif 648 649 /* 650 * Process options and, if not destined for us, 651 * ship it on. ip_dooptions returns 1 when an 652 * error was detected (causing an icmp message 653 * to be sent and the original packet to be freed). 654 */ 655 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 656 m = NULL; 657 goto out; 658 } 659 660 /* 661 * Check our list of addresses, to see if the packet is for us. 662 * 663 * Traditional 4.4BSD did not consult IFF_UP at all. 664 * The behavior here is to treat addresses on !IFF_UP interface 665 * or IN_IFF_NOTREADY addresses as not mine. 666 */ 667 downmatch = 0; 668 s = pserialize_read_enter(); 669 ia = ip_match_our_address(ifp, ip, &downmatch); 670 if (ia != NULL) { 671 pserialize_read_exit(s); 672 goto ours; 673 } 674 675 if (ifp->if_flags & IFF_BROADCAST) { 676 ia = ip_match_our_address_broadcast(ifp, ip); 677 if (ia != NULL) { 678 pserialize_read_exit(s); 679 goto ours; 680 } 681 } 682 pserialize_read_exit(s); 683 684 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 685#ifdef MROUTING 686 extern struct socket *ip_mrouter; 687 688 if (ip_mrouter) { 689 /* 690 * If we are acting as a multicast router, all 691 * incoming multicast packets are passed to the 692 * kernel-level multicast forwarding function. 693 * The packet is returned (relatively) intact; if 694 * ip_mforward() returns a non-zero value, the packet 695 * must be discarded, else it may be accepted below. 696 * 697 * (The IP ident field is put in the same byte order 698 * as expected when ip_mforward() is called from 699 * ip_output().) 700 */ 701 SOFTNET_LOCK(); 702 if (ip_mforward(m, ifp) != 0) { 703 SOFTNET_UNLOCK(); 704 IP_STATINC(IP_STAT_CANTFORWARD); 705 goto out; 706 } 707 SOFTNET_UNLOCK(); 708 709 /* 710 * The process-level routing demon needs to receive 711 * all multicast IGMP packets, whether or not this 712 * host belongs to their destination groups. 713 */ 714 if (ip->ip_p == IPPROTO_IGMP) { 715 goto ours; 716 } 717 IP_STATINC(IP_STAT_CANTFORWARD); 718 } 719#endif 720 /* 721 * See if we belong to the destination multicast group on the 722 * arrival interface. 723 */ 724 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 725 IP_STATINC(IP_STAT_CANTFORWARD); 726 goto out; 727 } 728 goto ours; 729 } 730 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 731 in_nullhost(ip->ip_dst)) 732 goto ours; 733 734 /* 735 * Not for us; forward if possible and desirable. 736 */ 737 if (ipforwarding == 0) { 738 m_put_rcvif_psref(ifp, &psref); 739 IP_STATINC(IP_STAT_CANTFORWARD); 740 m_freem(m); 741 } else { 742 /* 743 * If ip_dst matched any of my address on !IFF_UP interface, 744 * and there's no IFF_UP interface that matches ip_dst, 745 * send icmp unreach. Forwarding it will result in in-kernel 746 * forwarding loop till TTL goes to 0. 747 */ 748 if (downmatch) { 749 m_put_rcvif_psref(ifp, &psref); 750 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 751 IP_STATINC(IP_STAT_CANTFORWARD); 752 return; 753 } 754#ifdef IPSEC 755 /* Check the security policy (SP) for the packet */ 756 if (ipsec_used) { 757 if (ipsec4_input(m, IP_FORWARDING | 758 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 759 goto out; 760 } 761 } 762#endif 763 ip_forward(m, srcrt, ifp); 764 m_put_rcvif_psref(ifp, &psref); 765 } 766 return; 767 768ours: 769 m_put_rcvif_psref(ifp, &psref); 770 ifp = NULL; 771 772 /* 773 * If offset or IP_MF are set, must reassemble. 774 */ 775 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 776 /* 777 * Pass to IP reassembly mechanism. 778 */ 779 if (ip_reass_packet(&m, ip) != 0) { 780 /* Failed; invalid fragment(s) or packet. */ 781 goto out; 782 } 783 if (m == NULL) { 784 /* More fragments should come; silently return. */ 785 goto out; 786 } 787 /* 788 * Reassembly is done, we have the final packet. 789 * Updated cached data in local variable(s). 790 */ 791 ip = mtod(m, struct ip *); 792 hlen = ip->ip_hl << 2; 793 } 794 795#ifdef IPSEC 796 /* 797 * Enforce IPsec policy checking if we are seeing last header. 798 * Note that we do not visit this with protocols with PCB layer 799 * code - like UDP/TCP/raw IP. 800 */ 801 if (ipsec_used && 802 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 803 if (ipsec4_input(m, 0) != 0) { 804 goto out; 805 } 806 } 807#endif 808 809 /* 810 * Switch out to protocol's input routine. 811 */ 812#if IFA_STATS 813 if (ia && ip) { 814 struct in_ifaddr *_ia; 815 /* 816 * Keep a reference from ip_match_our_address with psref 817 * is expensive, so explore ia here again. 818 */ 819 s = pserialize_read_enter(); 820 _ia = in_get_ia(ip->ip_dst); 821 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 822 pserialize_read_exit(s); 823 } 824#endif 825 IP_STATINC(IP_STAT_DELIVERED); 826 827 const int off = hlen, nh = ip->ip_p; 828 829 SOFTNET_LOCK(); 830 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 831 SOFTNET_UNLOCK(); 832 return; 833 834out: 835 m_put_rcvif_psref(ifp, &psref); 836 if (m != NULL) 837 m_freem(m); 838} 839 840/* 841 * IP timer processing. 842 */ 843void 844ip_slowtimo(void) 845{ 846 847#ifndef NET_MPSAFE 848 mutex_enter(softnet_lock); 849 KERNEL_LOCK(1, NULL); 850#endif 851 852 ip_reass_slowtimo(); 853 854#ifndef NET_MPSAFE 855 KERNEL_UNLOCK_ONE(NULL); 856 mutex_exit(softnet_lock); 857#endif 858} 859 860/* 861 * IP drain processing. 862 */ 863void 864ip_drain(void) 865{ 866 867 KERNEL_LOCK(1, NULL); 868 ip_reass_drain(); 869 KERNEL_UNLOCK_ONE(NULL); 870} 871 872/* 873 * ip_dooptions: perform option processing on a datagram, possibly discarding 874 * it if bad options are encountered, or forwarding it if source-routed. 875 * 876 * => Returns true if packet has been forwarded/freed. 877 * => Returns false if the packet should be processed further. 878 */ 879static bool 880ip_dooptions(struct mbuf *m) 881{ 882 struct ip *ip = mtod(m, struct ip *); 883 u_char *cp, *cp0; 884 struct ip_timestamp *ipt; 885 struct in_ifaddr *ia; 886 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 887 struct in_addr dst; 888 n_time ntime; 889 struct ifaddr *ifa = NULL; 890 int s; 891 892 dst = ip->ip_dst; 893 cp = (u_char *)(ip + 1); 894 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 895 for (; cnt > 0; cnt -= optlen, cp += optlen) { 896 opt = cp[IPOPT_OPTVAL]; 897 if (opt == IPOPT_EOL) 898 break; 899 if (opt == IPOPT_NOP) 900 optlen = 1; 901 else { 902 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 903 code = &cp[IPOPT_OLEN] - (u_char *)ip; 904 goto bad; 905 } 906 optlen = cp[IPOPT_OLEN]; 907 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 908 code = &cp[IPOPT_OLEN] - (u_char *)ip; 909 goto bad; 910 } 911 } 912 switch (opt) { 913 914 default: 915 break; 916 917 /* 918 * Source routing with record. 919 * Find interface with current destination address. 920 * If none on this machine then drop if strictly routed, 921 * or do nothing if loosely routed. 922 * Record interface address and bring up next address 923 * component. If strictly routed make sure next 924 * address is on directly accessible net. 925 */ 926 case IPOPT_LSRR: 927 case IPOPT_SSRR: { 928 struct psref psref; 929 struct sockaddr_in ipaddr = { 930 .sin_len = sizeof(ipaddr), 931 .sin_family = AF_INET, 932 }; 933 934 if (ip_allowsrcrt == 0) { 935 type = ICMP_UNREACH; 936 code = ICMP_UNREACH_NET_PROHIB; 937 goto bad; 938 } 939 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 940 code = &cp[IPOPT_OLEN] - (u_char *)ip; 941 goto bad; 942 } 943 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 944 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 945 goto bad; 946 } 947 ipaddr.sin_addr = ip->ip_dst; 948 949 s = pserialize_read_enter(); 950 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 951 if (ifa == NULL) { 952 pserialize_read_exit(s); 953 if (opt == IPOPT_SSRR) { 954 type = ICMP_UNREACH; 955 code = ICMP_UNREACH_SRCFAIL; 956 goto bad; 957 } 958 /* 959 * Loose routing, and not at next destination 960 * yet; nothing to do except forward. 961 */ 962 break; 963 } 964 pserialize_read_exit(s); 965 966 off--; /* 0 origin */ 967 if ((off + sizeof(struct in_addr)) > optlen) { 968 /* 969 * End of source route. Should be for us. 970 */ 971 save_rte(m, cp, ip->ip_src); 972 break; 973 } 974 /* 975 * locate outgoing interface 976 */ 977 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 978 sizeof(ipaddr.sin_addr)); 979 if (opt == IPOPT_SSRR) { 980 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 981 &psref); 982 if (ifa != NULL) 983 ia = ifatoia(ifa); 984 else 985 ia = NULL; 986 } else { 987 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 988 } 989 if (ia == NULL) { 990 type = ICMP_UNREACH; 991 code = ICMP_UNREACH_SRCFAIL; 992 goto bad; 993 } 994 ip->ip_dst = ipaddr.sin_addr; 995 bcopy((void *)&ia->ia_addr.sin_addr, 996 (void *)(cp + off), sizeof(struct in_addr)); 997 ia4_release(ia, &psref); 998 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 999 /* 1000 * Let ip_intr's mcast routing check handle mcast pkts 1001 */ 1002 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1003 break; 1004 } 1005 1006 case IPOPT_RR: { 1007 struct psref psref; 1008 struct sockaddr_in ipaddr = { 1009 .sin_len = sizeof(ipaddr), 1010 .sin_family = AF_INET, 1011 }; 1012 1013 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1014 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1015 goto bad; 1016 } 1017 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1018 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1019 goto bad; 1020 } 1021 /* 1022 * If no space remains, ignore. 1023 */ 1024 off--; /* 0 origin */ 1025 if ((off + sizeof(struct in_addr)) > optlen) 1026 break; 1027 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1028 sizeof(ipaddr.sin_addr)); 1029 /* 1030 * locate outgoing interface; if we're the destination, 1031 * use the incoming interface (should be same). 1032 */ 1033 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1034 if (ifa == NULL) { 1035 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1036 if (ia == NULL) { 1037 type = ICMP_UNREACH; 1038 code = ICMP_UNREACH_HOST; 1039 goto bad; 1040 } 1041 } else { 1042 ia = ifatoia(ifa); 1043 } 1044 bcopy((void *)&ia->ia_addr.sin_addr, 1045 (void *)(cp + off), sizeof(struct in_addr)); 1046 ia4_release(ia, &psref); 1047 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1048 break; 1049 } 1050 1051 case IPOPT_TS: 1052 code = cp - (u_char *)ip; 1053 ipt = (struct ip_timestamp *)cp; 1054 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1055 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1056 goto bad; 1057 } 1058 if (ipt->ipt_ptr < 5) { 1059 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1060 goto bad; 1061 } 1062 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1063 if (++ipt->ipt_oflw == 0) { 1064 code = (u_char *)&ipt->ipt_ptr - 1065 (u_char *)ip; 1066 goto bad; 1067 } 1068 break; 1069 } 1070 cp0 = (cp + ipt->ipt_ptr - 1); 1071 switch (ipt->ipt_flg) { 1072 1073 case IPOPT_TS_TSONLY: 1074 break; 1075 1076 case IPOPT_TS_TSANDADDR: { 1077 struct ifnet *rcvif; 1078 int _s, _ss; 1079 struct sockaddr_in ipaddr = { 1080 .sin_len = sizeof(ipaddr), 1081 .sin_family = AF_INET, 1082 }; 1083 1084 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1085 sizeof(struct in_addr) > ipt->ipt_len) { 1086 code = (u_char *)&ipt->ipt_ptr - 1087 (u_char *)ip; 1088 goto bad; 1089 } 1090 ipaddr.sin_addr = dst; 1091 _ss = pserialize_read_enter(); 1092 rcvif = m_get_rcvif(m, &_s); 1093 if (__predict_true(rcvif != NULL)) { 1094 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1095 rcvif); 1096 } 1097 m_put_rcvif(rcvif, &_s); 1098 if (ifa == NULL) { 1099 pserialize_read_exit(_ss); 1100 break; 1101 } 1102 ia = ifatoia(ifa); 1103 bcopy(&ia->ia_addr.sin_addr, 1104 cp0, sizeof(struct in_addr)); 1105 pserialize_read_exit(_ss); 1106 ipt->ipt_ptr += sizeof(struct in_addr); 1107 break; 1108 } 1109 1110 case IPOPT_TS_PRESPEC: { 1111 struct sockaddr_in ipaddr = { 1112 .sin_len = sizeof(ipaddr), 1113 .sin_family = AF_INET, 1114 }; 1115 1116 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1117 sizeof(struct in_addr) > ipt->ipt_len) { 1118 code = (u_char *)&ipt->ipt_ptr - 1119 (u_char *)ip; 1120 goto bad; 1121 } 1122 memcpy(&ipaddr.sin_addr, cp0, 1123 sizeof(struct in_addr)); 1124 s = pserialize_read_enter(); 1125 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1126 if (ifa == NULL) { 1127 pserialize_read_exit(s); 1128 continue; 1129 } 1130 pserialize_read_exit(s); 1131 ipt->ipt_ptr += sizeof(struct in_addr); 1132 break; 1133 } 1134 1135 default: 1136 /* XXX can't take &ipt->ipt_flg */ 1137 code = (u_char *)&ipt->ipt_ptr - 1138 (u_char *)ip + 1; 1139 goto bad; 1140 } 1141 ntime = iptime(); 1142 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1143 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1144 sizeof(n_time)); 1145 ipt->ipt_ptr += sizeof(n_time); 1146 } 1147 } 1148 if (forward) { 1149 struct ifnet *rcvif; 1150 struct psref _psref; 1151 1152 if (ip_forwsrcrt == 0) { 1153 type = ICMP_UNREACH; 1154 code = ICMP_UNREACH_SRCFAIL; 1155 goto bad; 1156 } 1157 1158 rcvif = m_get_rcvif_psref(m, &_psref); 1159 if (__predict_false(rcvif == NULL)) { 1160 type = ICMP_UNREACH; 1161 code = ICMP_UNREACH_HOST; 1162 goto bad; 1163 } 1164 ip_forward(m, 1, rcvif); 1165 m_put_rcvif_psref(rcvif, &_psref); 1166 return true; 1167 } 1168 return false; 1169bad: 1170 icmp_error(m, type, code, 0, 0); 1171 IP_STATINC(IP_STAT_BADOPTIONS); 1172 return true; 1173} 1174 1175/* 1176 * ip_rtaddr: given address of next destination (final or next hop), 1177 * return internet address info of interface to be used to get there. 1178 */ 1179static struct in_ifaddr * 1180ip_rtaddr(struct in_addr dst, struct psref *psref) 1181{ 1182 struct rtentry *rt; 1183 union { 1184 struct sockaddr dst; 1185 struct sockaddr_in dst4; 1186 } u; 1187 struct route *ro; 1188 1189 sockaddr_in_init(&u.dst4, &dst, 0); 1190 1191 ro = percpu_getref(ipforward_rt_percpu); 1192 rt = rtcache_lookup(ro, &u.dst); 1193 if (rt == NULL) { 1194 percpu_putref(ipforward_rt_percpu); 1195 return NULL; 1196 } 1197 1198 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1199 rtcache_unref(rt, ro); 1200 percpu_putref(ipforward_rt_percpu); 1201 1202 return ifatoia(rt->rt_ifa); 1203} 1204 1205/* 1206 * save_rte: save incoming source route for use in replies, to be picked 1207 * up later by ip_srcroute if the receiver is interested. 1208 */ 1209static void 1210save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1211{ 1212 struct ip_srcrt *isr; 1213 struct m_tag *mtag; 1214 unsigned olen; 1215 1216 olen = option[IPOPT_OLEN]; 1217 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1218 return; 1219 1220 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1221 if (mtag == NULL) 1222 return; 1223 isr = (struct ip_srcrt *)(mtag + 1); 1224 1225 memcpy(isr->isr_hdr, option, olen); 1226 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1227 isr->isr_dst = dst; 1228 m_tag_prepend(m, mtag); 1229} 1230 1231/* 1232 * Retrieve incoming source route for use in replies, 1233 * in the same form used by setsockopt. 1234 * The first hop is placed before the options, will be removed later. 1235 */ 1236struct mbuf * 1237ip_srcroute(struct mbuf *m0) 1238{ 1239 struct in_addr *p, *q; 1240 struct mbuf *m; 1241 struct ip_srcrt *isr; 1242 struct m_tag *mtag; 1243 1244 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL); 1245 if (mtag == NULL) 1246 return NULL; 1247 isr = (struct ip_srcrt *)(mtag + 1); 1248 1249 if (isr->isr_nhops == 0) 1250 return NULL; 1251 1252 m = m_get(M_DONTWAIT, MT_SOOPTS); 1253 if (m == NULL) 1254 return NULL; 1255 1256 MCLAIM(m, &inetdomain.dom_mowner); 1257#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1258 1259 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1260 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1261 1262 /* 1263 * First save first hop for return route 1264 */ 1265 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1266 *(mtod(m, struct in_addr *)) = *p--; 1267 1268 /* 1269 * Copy option fields and padding (nop) to mbuf. 1270 */ 1271 isr->isr_nop = IPOPT_NOP; 1272 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1273 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1274 OPTSIZ); 1275 q = (struct in_addr *)(mtod(m, char *) + 1276 sizeof(struct in_addr) + OPTSIZ); 1277#undef OPTSIZ 1278 /* 1279 * Record return path as an IP source route, 1280 * reversing the path (pointers are now aligned). 1281 */ 1282 while (p >= isr->isr_routes) { 1283 *q++ = *p--; 1284 } 1285 /* 1286 * Last hop goes to final destination. 1287 */ 1288 *q = isr->isr_dst; 1289 m_tag_delete(m0, mtag); 1290 return m; 1291} 1292 1293const int inetctlerrmap[PRC_NCMDS] = { 1294 [PRC_MSGSIZE] = EMSGSIZE, 1295 [PRC_HOSTDEAD] = EHOSTDOWN, 1296 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1297 [PRC_UNREACH_NET] = EHOSTUNREACH, 1298 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1299 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1300 [PRC_UNREACH_PORT] = ECONNREFUSED, 1301 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1302 [PRC_PARAMPROB] = ENOPROTOOPT, 1303}; 1304 1305void 1306ip_fasttimo(void) 1307{ 1308 if (ip_drainwanted) { 1309 ip_drain(); 1310 ip_drainwanted = 0; 1311 } 1312} 1313 1314void 1315ip_drainstub(void) 1316{ 1317 ip_drainwanted = 1; 1318} 1319 1320/* 1321 * Forward a packet. If some error occurs return the sender 1322 * an icmp packet. Note we can't always generate a meaningful 1323 * icmp message because icmp doesn't have a large enough repertoire 1324 * of codes and types. 1325 * 1326 * If not forwarding, just drop the packet. This could be confusing 1327 * if ipforwarding was zero but some routing protocol was advancing 1328 * us as a gateway to somewhere. However, we must let the routing 1329 * protocol deal with that. 1330 * 1331 * The srcrt parameter indicates whether the packet is being forwarded 1332 * via a source route. 1333 */ 1334static void 1335ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1336{ 1337 struct ip *ip = mtod(m, struct ip *); 1338 struct rtentry *rt; 1339 int error, type = 0, code = 0, destmtu = 0; 1340 struct mbuf *mcopy; 1341 n_long dest; 1342 union { 1343 struct sockaddr dst; 1344 struct sockaddr_in dst4; 1345 } u; 1346 uint64_t *ips; 1347 struct route *ro; 1348 1349 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1350 "interrupt handler; synchronization assumptions violated"); 1351 1352 /* 1353 * We are now in the output path. 1354 */ 1355 MCLAIM(m, &ip_tx_mowner); 1356 1357 /* 1358 * Clear any in-bound checksum flags for this packet. 1359 */ 1360 m->m_pkthdr.csum_flags = 0; 1361 1362 dest = 0; 1363 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1364 IP_STATINC(IP_STAT_CANTFORWARD); 1365 m_freem(m); 1366 return; 1367 } 1368 1369 if (ip->ip_ttl <= IPTTLDEC) { 1370 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1371 return; 1372 } 1373 1374 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1375 1376 ro = percpu_getref(ipforward_rt_percpu); 1377 rt = rtcache_lookup(ro, &u.dst); 1378 if (rt == NULL) { 1379 percpu_putref(ipforward_rt_percpu); 1380 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1381 return; 1382 } 1383 1384 /* 1385 * Save at most 68 bytes of the packet in case 1386 * we need to generate an ICMP message to the src. 1387 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1388 */ 1389 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1390 if (mcopy) 1391 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1392 1393 ip->ip_ttl -= IPTTLDEC; 1394 1395 /* 1396 * If forwarding packet using same interface that it came in on, 1397 * perhaps should send a redirect to sender to shortcut a hop. 1398 * Only send redirect if source is sending directly to us, 1399 * and if packet was not source routed (or has any options). 1400 * Also, don't send redirect if forwarding using a default route 1401 * or a route modified by a redirect. 1402 */ 1403 if (rt->rt_ifp == rcvif && 1404 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1405 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1406 ipsendredirects && !srcrt) { 1407 if (rt->rt_ifa && 1408 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1409 ifatoia(rt->rt_ifa)->ia_subnet) { 1410 if (rt->rt_flags & RTF_GATEWAY) 1411 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1412 else 1413 dest = ip->ip_dst.s_addr; 1414 /* 1415 * Router requirements says to only send host 1416 * redirects. 1417 */ 1418 type = ICMP_REDIRECT; 1419 code = ICMP_REDIRECT_HOST; 1420 } 1421 } 1422 rtcache_unref(rt, ro); 1423 1424 error = ip_output(m, NULL, ro, 1425 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1426 NULL, NULL); 1427 1428 if (error) { 1429 IP_STATINC(IP_STAT_CANTFORWARD); 1430 goto error; 1431 } 1432 1433 ips = IP_STAT_GETREF(); 1434 ips[IP_STAT_FORWARD]++; 1435 1436 if (type) { 1437 ips[IP_STAT_REDIRECTSENT]++; 1438 IP_STAT_PUTREF(); 1439 goto redirect; 1440 } 1441 1442 IP_STAT_PUTREF(); 1443 if (mcopy) { 1444#ifdef GATEWAY 1445 if (mcopy->m_flags & M_CANFASTFWD) 1446 ipflow_create(ro, mcopy); 1447#endif 1448 m_freem(mcopy); 1449 } 1450 1451 percpu_putref(ipforward_rt_percpu); 1452 return; 1453 1454redirect: 1455error: 1456 if (mcopy == NULL) { 1457 percpu_putref(ipforward_rt_percpu); 1458 return; 1459 } 1460 1461 switch (error) { 1462 1463 case 0: /* forwarded, but need redirect */ 1464 /* type, code set above */ 1465 break; 1466 1467 case ENETUNREACH: /* shouldn't happen, checked above */ 1468 case EHOSTUNREACH: 1469 case ENETDOWN: 1470 case EHOSTDOWN: 1471 default: 1472 type = ICMP_UNREACH; 1473 code = ICMP_UNREACH_HOST; 1474 break; 1475 1476 case EMSGSIZE: 1477 type = ICMP_UNREACH; 1478 code = ICMP_UNREACH_NEEDFRAG; 1479 1480 if ((rt = rtcache_validate(ro)) != NULL) { 1481 destmtu = rt->rt_ifp->if_mtu; 1482 rtcache_unref(rt, ro); 1483 } 1484#ifdef IPSEC 1485 if (ipsec_used) 1486 (void)ipsec4_forward(mcopy, &destmtu); 1487#endif 1488 IP_STATINC(IP_STAT_CANTFRAG); 1489 break; 1490 1491 case ENOBUFS: 1492 /* 1493 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1494 * Requirements for IP Version 4 Routers. Source quench can 1495 * big problem under DoS attacks or if the underlying 1496 * interface is rate-limited. 1497 */ 1498 if (mcopy) 1499 m_freem(mcopy); 1500 percpu_putref(ipforward_rt_percpu); 1501 return; 1502 } 1503 icmp_error(mcopy, type, code, dest, destmtu); 1504 percpu_putref(ipforward_rt_percpu); 1505} 1506 1507void 1508ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1509 struct mbuf *m) 1510{ 1511 struct socket *so = inp->inp_socket; 1512 int inpflags = inp->inp_flags; 1513 1514 if (SOOPT_TIMESTAMP(so->so_options)) 1515 mp = sbsavetimestamp(so->so_options, m, mp); 1516 1517 if (inpflags & INP_RECVDSTADDR) { 1518 *mp = sbcreatecontrol(&ip->ip_dst, 1519 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1520 if (*mp) 1521 mp = &(*mp)->m_next; 1522 } 1523 1524 if (inpflags & INP_RECVTTL) { 1525 *mp = sbcreatecontrol(&ip->ip_ttl, 1526 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1527 if (*mp) 1528 mp = &(*mp)->m_next; 1529 } 1530 1531 struct psref psref; 1532 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1533 if (__predict_false(ifp == NULL)) { 1534#ifdef DIAGNOSTIC 1535 printf("%s: missing receive interface\n", __func__); 1536#endif 1537 return; /* XXX should report error? */ 1538 } 1539 1540 if (inpflags & INP_RECVPKTINFO) { 1541 struct in_pktinfo ipi; 1542 ipi.ipi_addr = ip->ip_src; 1543 ipi.ipi_ifindex = ifp->if_index; 1544 *mp = sbcreatecontrol(&ipi, 1545 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP); 1546 if (*mp) 1547 mp = &(*mp)->m_next; 1548 } 1549 if (inpflags & INP_PKTINFO) { 1550 struct in_pktinfo ipi; 1551 ipi.ipi_addr = ip->ip_dst; 1552 ipi.ipi_ifindex = ifp->if_index; 1553 *mp = sbcreatecontrol(&ipi, 1554 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1555 if (*mp) 1556 mp = &(*mp)->m_next; 1557 } 1558 if (inpflags & INP_RECVIF) { 1559 struct sockaddr_dl sdl; 1560 1561 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1562 NULL, 0); 1563 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, 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