ip_input.c revision 1.349
1/* $NetBSD: ip_input.c,v 1.349 2017/02/07 02:38:08 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 1.349 2017/02/07 02:38:08 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 358static struct in_ifaddr * 359ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 360{ 361 struct in_ifaddr *ia = NULL; 362 int checkif; 363 364 /* 365 * Enable a consistency check between the destination address 366 * and the arrival interface for a unicast packet (the RFC 1122 367 * strong ES model) if IP forwarding is disabled and the packet 368 * is not locally generated. 369 * 370 * XXX - Checking also should be disabled if the destination 371 * address is ipnat'ed to a different interface. 372 * 373 * XXX - Checking is incompatible with IP aliases added 374 * to the loopback interface instead of the interface where 375 * the packets are received. 376 * 377 * XXX - We need to add a per ifaddr flag for this so that 378 * we get finer grain control. 379 */ 380 checkif = ip_checkinterface && (ipforwarding == 0) && 381 (ifp->if_flags & IFF_LOOPBACK) == 0; 382 383 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 384 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 385 if (ia->ia4_flags & IN_IFF_NOTREADY) 386 continue; 387 if (checkif && ia->ia_ifp != ifp) 388 continue; 389 if ((ia->ia_ifp->if_flags & IFF_UP) != 0 && 390 (ia->ia4_flags & IN_IFF_DETACHED) == 0) 391 break; 392 else 393 (*downmatch)++; 394 } 395 } 396 397 return ia; 398} 399 400static struct in_ifaddr * 401ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 402{ 403 struct in_ifaddr *ia = NULL; 404 struct ifaddr *ifa; 405 406 IFADDR_READER_FOREACH(ifa, ifp) { 407 if (ifa->ifa_addr->sa_family != AF_INET) 408 continue; 409 ia = ifatoia(ifa); 410 if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) 411 continue; 412 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 413 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 414 /* 415 * Look for all-0's host part (old broadcast addr), 416 * either for subnet or net. 417 */ 418 ip->ip_dst.s_addr == ia->ia_subnet || 419 ip->ip_dst.s_addr == ia->ia_net) 420 goto matched; 421 /* 422 * An interface with IP address zero accepts 423 * all packets that arrive on that interface. 424 */ 425 if (in_nullhost(ia->ia_addr.sin_addr)) 426 goto matched; 427 } 428 ia = NULL; 429 430matched: 431 return ia; 432} 433 434/* 435 * IP software interrupt routine. 436 */ 437static void 438ipintr(void *arg __unused) 439{ 440 struct mbuf *m; 441 442 KASSERT(cpu_softintr_p()); 443 444#ifndef NET_MPSAFE 445 mutex_enter(softnet_lock); 446#endif 447 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 448 ip_input(m); 449 } 450#ifndef NET_MPSAFE 451 mutex_exit(softnet_lock); 452#endif 453} 454 455/* 456 * IP input routine. Checksum and byte swap header. If fragmented 457 * try to reassemble. Process options. Pass to next level. 458 */ 459static void 460ip_input(struct mbuf *m) 461{ 462 struct ip *ip = NULL; 463 struct in_ifaddr *ia = NULL; 464 int hlen = 0, len; 465 int downmatch; 466 int srcrt = 0; 467 ifnet_t *ifp; 468 struct psref psref; 469 int s; 470 471 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 472 "interrupt handler; synchronization assumptions violated"); 473 474 MCLAIM(m, &ip_rx_mowner); 475 KASSERT((m->m_flags & M_PKTHDR) != 0); 476 477 ifp = m_get_rcvif_psref(m, &psref); 478 if (__predict_false(ifp == NULL)) 479 goto out; 480 481 /* 482 * If no IP addresses have been set yet but the interfaces 483 * are receiving, can't do anything with incoming packets yet. 484 * Note: we pre-check without locks held. 485 */ 486 if (IN_ADDRLIST_READER_EMPTY()) 487 goto out; 488 IP_STATINC(IP_STAT_TOTAL); 489 490 /* 491 * If the IP header is not aligned, slurp it up into a new 492 * mbuf with space for link headers, in the event we forward 493 * it. Otherwise, if it is aligned, make sure the entire 494 * base IP header is in the first mbuf of the chain. 495 */ 496 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 497 if ((m = m_copyup(m, sizeof(struct ip), 498 (max_linkhdr + 3) & ~3)) == NULL) { 499 /* XXXJRT new stat, please */ 500 IP_STATINC(IP_STAT_TOOSMALL); 501 goto out; 502 } 503 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 504 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 505 IP_STATINC(IP_STAT_TOOSMALL); 506 goto out; 507 } 508 } 509 ip = mtod(m, struct ip *); 510 if (ip->ip_v != IPVERSION) { 511 IP_STATINC(IP_STAT_BADVERS); 512 goto out; 513 } 514 hlen = ip->ip_hl << 2; 515 if (hlen < sizeof(struct ip)) { /* minimum header length */ 516 IP_STATINC(IP_STAT_BADHLEN); 517 goto out; 518 } 519 if (hlen > m->m_len) { 520 if ((m = m_pullup(m, hlen)) == NULL) { 521 IP_STATINC(IP_STAT_BADHLEN); 522 goto out; 523 } 524 ip = mtod(m, struct ip *); 525 } 526 527 /* 528 * RFC1122: packets with a multicast source address are 529 * not allowed. 530 */ 531 if (IN_MULTICAST(ip->ip_src.s_addr)) { 532 IP_STATINC(IP_STAT_BADADDR); 533 goto out; 534 } 535 536 /* 127/8 must not appear on wire - RFC1122 */ 537 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 538 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 539 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 540 IP_STATINC(IP_STAT_BADADDR); 541 goto out; 542 } 543 } 544 545 switch (m->m_pkthdr.csum_flags & 546 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | 547 M_CSUM_IPv4_BAD)) { 548 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 549 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 550 IP_STATINC(IP_STAT_BADSUM); 551 goto out; 552 553 case M_CSUM_IPv4: 554 /* Checksum was okay. */ 555 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 556 break; 557 558 default: 559 /* 560 * Must compute it ourselves. Maybe skip checksum on 561 * loopback interfaces. 562 */ 563 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 564 ip_do_loopback_cksum)) { 565 INET_CSUM_COUNTER_INCR(&ip_swcsum); 566 if (in_cksum(m, hlen) != 0) { 567 IP_STATINC(IP_STAT_BADSUM); 568 goto out; 569 } 570 } 571 break; 572 } 573 574 /* Retrieve the packet length. */ 575 len = ntohs(ip->ip_len); 576 577 /* 578 * Check for additional length bogosity 579 */ 580 if (len < hlen) { 581 IP_STATINC(IP_STAT_BADLEN); 582 goto out; 583 } 584 585 /* 586 * Check that the amount of data in the buffers 587 * is as at least much as the IP header would have us expect. 588 * Trim mbufs if longer than we expect. 589 * Drop packet if shorter than we expect. 590 */ 591 if (m->m_pkthdr.len < len) { 592 IP_STATINC(IP_STAT_TOOSHORT); 593 goto out; 594 } 595 if (m->m_pkthdr.len > len) { 596 if (m->m_len == m->m_pkthdr.len) { 597 m->m_len = len; 598 m->m_pkthdr.len = len; 599 } else 600 m_adj(m, len - m->m_pkthdr.len); 601 } 602 603 /* 604 * Assume that we can create a fast-forward IP flow entry 605 * based on this packet. 606 */ 607 m->m_flags |= M_CANFASTFWD; 608 609 /* 610 * Run through list of hooks for input packets. If there are any 611 * filters which require that additional packets in the flow are 612 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 613 * Note that filters must _never_ set this flag, as another filter 614 * in the list may have previously cleared it. 615 */ 616#if defined(IPSEC) 617 if (!ipsec_used || !ipsec_indone(m)) 618#else 619 if (1) 620#endif 621 { 622 struct in_addr odst = ip->ip_dst; 623 bool freed; 624 625 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 626 if (freed || m == NULL) { 627 m = NULL; 628 goto out; 629 } 630 ip = mtod(m, struct ip *); 631 hlen = ip->ip_hl << 2; 632 633 /* 634 * XXX The setting of "srcrt" here is to prevent ip_forward() 635 * from generating ICMP redirects for packets that have 636 * been redirected by a hook back out on to the same LAN that 637 * they came from and is not an indication that the packet 638 * is being inffluenced by source routing options. This 639 * allows things like 640 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 641 * where tlp0 is both on the 1.1.1.0/24 network and is the 642 * default route for hosts on 1.1.1.0/24. Of course this 643 * also requires a "map tlp0 ..." to complete the story. 644 * One might argue whether or not this kind of network config. 645 * should be supported in this manner... 646 */ 647 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 648 } 649 650#ifdef ALTQ 651 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 652 if (altq_input) { 653 SOFTNET_LOCK(); 654 if ((*altq_input)(m, AF_INET) == 0) { 655 /* Packet dropped by traffic conditioner. */ 656 SOFTNET_UNLOCK(); 657 m = NULL; 658 goto out; 659 } 660 SOFTNET_UNLOCK(); 661 } 662#endif 663 664 /* 665 * Process options and, if not destined for us, 666 * ship it on. ip_dooptions returns 1 when an 667 * error was detected (causing an icmp message 668 * to be sent and the original packet to be freed). 669 */ 670 ip_nhops = 0; /* for source routed packets */ 671 if (hlen > sizeof (struct ip) && ip_dooptions(m)) { 672 m = NULL; 673 goto out; 674 } 675 676 /* 677 * Check our list of addresses, to see if the packet is for us. 678 * 679 * Traditional 4.4BSD did not consult IFF_UP at all. 680 * The behavior here is to treat addresses on !IFF_UP interface 681 * or IN_IFF_NOTREADY addresses as not mine. 682 */ 683 downmatch = 0; 684 s = pserialize_read_enter(); 685 ia = ip_match_our_address(ifp, ip, &downmatch); 686 if (ia != NULL) { 687 pserialize_read_exit(s); 688 goto ours; 689 } 690 691 if (ifp->if_flags & IFF_BROADCAST) { 692 ia = ip_match_our_address_broadcast(ifp, ip); 693 if (ia != NULL) { 694 pserialize_read_exit(s); 695 goto ours; 696 } 697 } 698 pserialize_read_exit(s); 699 700 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 701#ifdef MROUTING 702 extern struct socket *ip_mrouter; 703 704 if (ip_mrouter) { 705 /* 706 * If we are acting as a multicast router, all 707 * incoming multicast packets are passed to the 708 * kernel-level multicast forwarding function. 709 * The packet is returned (relatively) intact; if 710 * ip_mforward() returns a non-zero value, the packet 711 * must be discarded, else it may be accepted below. 712 * 713 * (The IP ident field is put in the same byte order 714 * as expected when ip_mforward() is called from 715 * ip_output().) 716 */ 717 SOFTNET_LOCK(); 718 if (ip_mforward(m, ifp) != 0) { 719 SOFTNET_UNLOCK(); 720 IP_STATINC(IP_STAT_CANTFORWARD); 721 goto out; 722 } 723 SOFTNET_UNLOCK(); 724 725 /* 726 * The process-level routing demon needs to receive 727 * all multicast IGMP packets, whether or not this 728 * host belongs to their destination groups. 729 */ 730 if (ip->ip_p == IPPROTO_IGMP) { 731 goto ours; 732 } 733 IP_STATINC(IP_STAT_CANTFORWARD); 734 } 735#endif 736 /* 737 * See if we belong to the destination multicast group on the 738 * arrival interface. 739 */ 740 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 741 IP_STATINC(IP_STAT_CANTFORWARD); 742 goto out; 743 } 744 goto ours; 745 } 746 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 747 in_nullhost(ip->ip_dst)) 748 goto ours; 749 750 /* 751 * Not for us; forward if possible and desirable. 752 */ 753 if (ipforwarding == 0) { 754 m_put_rcvif_psref(ifp, &psref); 755 IP_STATINC(IP_STAT_CANTFORWARD); 756 m_freem(m); 757 } else { 758 /* 759 * If ip_dst matched any of my address on !IFF_UP interface, 760 * and there's no IFF_UP interface that matches ip_dst, 761 * send icmp unreach. Forwarding it will result in in-kernel 762 * forwarding loop till TTL goes to 0. 763 */ 764 if (downmatch) { 765 m_put_rcvif_psref(ifp, &psref); 766 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 767 IP_STATINC(IP_STAT_CANTFORWARD); 768 return; 769 } 770#ifdef IPSEC 771 /* Perform IPsec, if any. */ 772 if (ipsec_used) { 773 SOFTNET_LOCK(); 774 if (ipsec4_input(m, IP_FORWARDING | 775 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 776 SOFTNET_UNLOCK(); 777 goto out; 778 } 779 SOFTNET_UNLOCK(); 780 } 781#endif 782 ip_forward(m, srcrt, ifp); 783 m_put_rcvif_psref(ifp, &psref); 784 } 785 return; 786 787ours: 788 m_put_rcvif_psref(ifp, &psref); 789 ifp = NULL; 790 791 /* 792 * If offset or IP_MF are set, must reassemble. 793 */ 794 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 795 /* 796 * Pass to IP reassembly mechanism. 797 */ 798 if (ip_reass_packet(&m, ip) != 0) { 799 /* Failed; invalid fragment(s) or packet. */ 800 goto out; 801 } 802 if (m == NULL) { 803 /* More fragments should come; silently return. */ 804 goto out; 805 } 806 /* 807 * Reassembly is done, we have the final packet. 808 * Updated cached data in local variable(s). 809 */ 810 ip = mtod(m, struct ip *); 811 hlen = ip->ip_hl << 2; 812 } 813 814#ifdef IPSEC 815 /* 816 * Enforce IPsec policy checking if we are seeing last header. 817 * Note that we do not visit this with protocols with PCB layer 818 * code - like UDP/TCP/raw IP. 819 */ 820 if (ipsec_used && 821 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 822 SOFTNET_LOCK(); 823 if (ipsec4_input(m, 0) != 0) { 824 SOFTNET_UNLOCK(); 825 goto out; 826 } 827 SOFTNET_UNLOCK(); 828 } 829#endif 830 831 /* 832 * Switch out to protocol's input routine. 833 */ 834#if IFA_STATS 835 if (ia && ip) { 836 struct in_ifaddr *_ia; 837 /* 838 * Keep a reference from ip_match_our_address with psref 839 * is expensive, so explore ia here again. 840 */ 841 s = pserialize_read_enter(); 842 _ia = in_get_ia(ip->ip_dst); 843 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 844 pserialize_read_exit(s); 845 } 846#endif 847 IP_STATINC(IP_STAT_DELIVERED); 848 849 const int off = hlen, nh = ip->ip_p; 850 851 SOFTNET_LOCK(); 852 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 853 SOFTNET_UNLOCK(); 854 return; 855 856out: 857 m_put_rcvif_psref(ifp, &psref); 858 if (m != NULL) 859 m_freem(m); 860} 861 862/* 863 * IP timer processing. 864 */ 865void 866ip_slowtimo(void) 867{ 868 869#ifndef NET_MPSAFE 870 mutex_enter(softnet_lock); 871 KERNEL_LOCK(1, NULL); 872#endif 873 874 ip_reass_slowtimo(); 875 876#ifndef NET_MPSAFE 877 KERNEL_UNLOCK_ONE(NULL); 878 mutex_exit(softnet_lock); 879#endif 880} 881 882/* 883 * IP drain processing. 884 */ 885void 886ip_drain(void) 887{ 888 889 KERNEL_LOCK(1, NULL); 890 ip_reass_drain(); 891 KERNEL_UNLOCK_ONE(NULL); 892} 893 894/* 895 * ip_dooptions: perform option processing on a datagram, possibly discarding 896 * it if bad options are encountered, or forwarding it if source-routed. 897 * 898 * => Returns true if packet has been forwarded/freed. 899 * => Returns false if the packet should be processed further. 900 */ 901static bool 902ip_dooptions(struct mbuf *m) 903{ 904 struct ip *ip = mtod(m, struct ip *); 905 u_char *cp, *cp0; 906 struct ip_timestamp *ipt; 907 struct in_ifaddr *ia; 908 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 909 struct in_addr dst; 910 n_time ntime; 911 struct ifaddr *ifa = NULL; 912 int s; 913 914 dst = ip->ip_dst; 915 cp = (u_char *)(ip + 1); 916 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 917 for (; cnt > 0; cnt -= optlen, cp += optlen) { 918 opt = cp[IPOPT_OPTVAL]; 919 if (opt == IPOPT_EOL) 920 break; 921 if (opt == IPOPT_NOP) 922 optlen = 1; 923 else { 924 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 925 code = &cp[IPOPT_OLEN] - (u_char *)ip; 926 goto bad; 927 } 928 optlen = cp[IPOPT_OLEN]; 929 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 930 code = &cp[IPOPT_OLEN] - (u_char *)ip; 931 goto bad; 932 } 933 } 934 switch (opt) { 935 936 default: 937 break; 938 939 /* 940 * Source routing with record. 941 * Find interface with current destination address. 942 * If none on this machine then drop if strictly routed, 943 * or do nothing if loosely routed. 944 * Record interface address and bring up next address 945 * component. If strictly routed make sure next 946 * address is on directly accessible net. 947 */ 948 case IPOPT_LSRR: 949 case IPOPT_SSRR: { 950 struct psref psref; 951 if (ip_allowsrcrt == 0) { 952 type = ICMP_UNREACH; 953 code = ICMP_UNREACH_NET_PROHIB; 954 goto bad; 955 } 956 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 957 code = &cp[IPOPT_OLEN] - (u_char *)ip; 958 goto bad; 959 } 960 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 961 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 962 goto bad; 963 } 964 ipaddr.sin_addr = ip->ip_dst; 965 966 s = pserialize_read_enter(); 967 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 968 if (ifa == NULL) { 969 pserialize_read_exit(s); 970 if (opt == IPOPT_SSRR) { 971 type = ICMP_UNREACH; 972 code = ICMP_UNREACH_SRCFAIL; 973 goto bad; 974 } 975 /* 976 * Loose routing, and not at next destination 977 * yet; nothing to do except forward. 978 */ 979 break; 980 } 981 pserialize_read_exit(s); 982 983 off--; /* 0 origin */ 984 if ((off + sizeof(struct in_addr)) > optlen) { 985 /* 986 * End of source route. Should be for us. 987 */ 988 save_rte(cp, ip->ip_src); 989 break; 990 } 991 /* 992 * locate outgoing interface 993 */ 994 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 995 sizeof(ipaddr.sin_addr)); 996 if (opt == IPOPT_SSRR) { 997 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 998 &psref); 999 if (ifa != NULL) 1000 ia = ifatoia(ifa); 1001 else 1002 ia = NULL; 1003 } else { 1004 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1005 } 1006 if (ia == NULL) { 1007 type = ICMP_UNREACH; 1008 code = ICMP_UNREACH_SRCFAIL; 1009 goto bad; 1010 } 1011 ip->ip_dst = ipaddr.sin_addr; 1012 bcopy((void *)&ia->ia_addr.sin_addr, 1013 (void *)(cp + off), sizeof(struct in_addr)); 1014 ia4_release(ia, &psref); 1015 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1016 /* 1017 * Let ip_intr's mcast routing check handle mcast pkts 1018 */ 1019 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1020 break; 1021 } 1022 1023 case IPOPT_RR: { 1024 struct psref psref; 1025 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1026 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1027 goto bad; 1028 } 1029 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1030 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1031 goto bad; 1032 } 1033 /* 1034 * If no space remains, ignore. 1035 */ 1036 off--; /* 0 origin */ 1037 if ((off + sizeof(struct in_addr)) > optlen) 1038 break; 1039 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1040 sizeof(ipaddr.sin_addr)); 1041 /* 1042 * locate outgoing interface; if we're the destination, 1043 * use the incoming interface (should be same). 1044 */ 1045 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1046 if (ifa == NULL) { 1047 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1048 if (ia == NULL) { 1049 type = ICMP_UNREACH; 1050 code = ICMP_UNREACH_HOST; 1051 goto bad; 1052 } 1053 } else { 1054 ia = ifatoia(ifa); 1055 } 1056 bcopy((void *)&ia->ia_addr.sin_addr, 1057 (void *)(cp + off), sizeof(struct in_addr)); 1058 ia4_release(ia, &psref); 1059 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1060 break; 1061 } 1062 1063 case IPOPT_TS: 1064 code = cp - (u_char *)ip; 1065 ipt = (struct ip_timestamp *)cp; 1066 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1067 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1068 goto bad; 1069 } 1070 if (ipt->ipt_ptr < 5) { 1071 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1072 goto bad; 1073 } 1074 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1075 if (++ipt->ipt_oflw == 0) { 1076 code = (u_char *)&ipt->ipt_ptr - 1077 (u_char *)ip; 1078 goto bad; 1079 } 1080 break; 1081 } 1082 cp0 = (cp + ipt->ipt_ptr - 1); 1083 switch (ipt->ipt_flg) { 1084 1085 case IPOPT_TS_TSONLY: 1086 break; 1087 1088 case IPOPT_TS_TSANDADDR: { 1089 struct ifnet *rcvif; 1090 int _s, _ss; 1091 1092 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1093 sizeof(struct in_addr) > ipt->ipt_len) { 1094 code = (u_char *)&ipt->ipt_ptr - 1095 (u_char *)ip; 1096 goto bad; 1097 } 1098 ipaddr.sin_addr = dst; 1099 _ss = pserialize_read_enter(); 1100 rcvif = m_get_rcvif(m, &_s); 1101 if (__predict_true(rcvif != NULL)) { 1102 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1103 rcvif); 1104 } 1105 m_put_rcvif(rcvif, &_s); 1106 if (ifa == NULL) { 1107 pserialize_read_exit(_ss); 1108 break; 1109 } 1110 ia = ifatoia(ifa); 1111 bcopy(&ia->ia_addr.sin_addr, 1112 cp0, sizeof(struct in_addr)); 1113 pserialize_read_exit(_ss); 1114 ipt->ipt_ptr += sizeof(struct in_addr); 1115 break; 1116 } 1117 1118 case IPOPT_TS_PRESPEC: 1119 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1120 sizeof(struct in_addr) > ipt->ipt_len) { 1121 code = (u_char *)&ipt->ipt_ptr - 1122 (u_char *)ip; 1123 goto bad; 1124 } 1125 memcpy(&ipaddr.sin_addr, cp0, 1126 sizeof(struct in_addr)); 1127 s = pserialize_read_enter(); 1128 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1129 if (ifa == NULL) { 1130 pserialize_read_exit(s); 1131 continue; 1132 } 1133 pserialize_read_exit(s); 1134 ipt->ipt_ptr += sizeof(struct in_addr); 1135 break; 1136 1137 default: 1138 /* XXX can't take &ipt->ipt_flg */ 1139 code = (u_char *)&ipt->ipt_ptr - 1140 (u_char *)ip + 1; 1141 goto bad; 1142 } 1143 ntime = iptime(); 1144 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1145 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1146 sizeof(n_time)); 1147 ipt->ipt_ptr += sizeof(n_time); 1148 } 1149 } 1150 if (forward) { 1151 struct ifnet *rcvif; 1152 struct psref _psref; 1153 1154 if (ip_forwsrcrt == 0) { 1155 type = ICMP_UNREACH; 1156 code = ICMP_UNREACH_SRCFAIL; 1157 goto bad; 1158 } 1159 1160 rcvif = m_get_rcvif_psref(m, &_psref); 1161 if (__predict_false(rcvif == NULL)) { 1162 type = ICMP_UNREACH; 1163 code = ICMP_UNREACH_HOST; 1164 goto bad; 1165 } 1166 ip_forward(m, 1, rcvif); 1167 m_put_rcvif_psref(rcvif, &_psref); 1168 return true; 1169 } 1170 return false; 1171bad: 1172 icmp_error(m, type, code, 0, 0); 1173 IP_STATINC(IP_STAT_BADOPTIONS); 1174 return true; 1175} 1176 1177/* 1178 * ip_rtaddr: given address of next destination (final or next hop), 1179 * return internet address info of interface to be used to get there. 1180 */ 1181static struct in_ifaddr * 1182ip_rtaddr(struct in_addr dst, struct psref *psref) 1183{ 1184 struct rtentry *rt; 1185 union { 1186 struct sockaddr dst; 1187 struct sockaddr_in dst4; 1188 } u; 1189 struct route *ro; 1190 1191 sockaddr_in_init(&u.dst4, &dst, 0); 1192 1193 ro = percpu_getref(ipforward_rt_percpu); 1194 rt = rtcache_lookup(ro, &u.dst); 1195 if (rt == NULL) { 1196 percpu_putref(ipforward_rt_percpu); 1197 return NULL; 1198 } 1199 1200 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1201 rtcache_unref(rt, ro); 1202 percpu_putref(ipforward_rt_percpu); 1203 1204 return ifatoia(rt->rt_ifa); 1205} 1206 1207/* 1208 * save_rte: save incoming source route for use in replies, to be picked 1209 * up later by ip_srcroute if the receiver is interested. 1210 */ 1211static void 1212save_rte(u_char *option, struct in_addr dst) 1213{ 1214 unsigned olen; 1215 1216 olen = option[IPOPT_OLEN]; 1217 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1218 return; 1219 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen); 1220 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1221 ip_srcrt.dst = dst; 1222} 1223 1224/* 1225 * Retrieve incoming source route for use in replies, 1226 * in the same form used by setsockopt. 1227 * The first hop is placed before the options, will be removed later. 1228 */ 1229struct mbuf * 1230ip_srcroute(void) 1231{ 1232 struct in_addr *p, *q; 1233 struct mbuf *m; 1234 1235 if (ip_nhops == 0) 1236 return NULL; 1237 m = m_get(M_DONTWAIT, MT_SOOPTS); 1238 if (m == 0) 1239 return NULL; 1240 1241 MCLAIM(m, &inetdomain.dom_mowner); 1242#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1243 1244 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1245 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1246 OPTSIZ; 1247 1248 /* 1249 * First save first hop for return route 1250 */ 1251 p = &ip_srcrt.route[ip_nhops - 1]; 1252 *(mtod(m, struct in_addr *)) = *p--; 1253 1254 /* 1255 * Copy option fields and padding (nop) to mbuf. 1256 */ 1257 ip_srcrt.nop = IPOPT_NOP; 1258 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1259 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1260 OPTSIZ); 1261 q = (struct in_addr *)(mtod(m, char *) + 1262 sizeof(struct in_addr) + OPTSIZ); 1263#undef OPTSIZ 1264 /* 1265 * Record return path as an IP source route, 1266 * reversing the path (pointers are now aligned). 1267 */ 1268 while (p >= ip_srcrt.route) { 1269 *q++ = *p--; 1270 } 1271 /* 1272 * Last hop goes to final destination. 1273 */ 1274 *q = ip_srcrt.dst; 1275 return (m); 1276} 1277 1278const int inetctlerrmap[PRC_NCMDS] = { 1279 [PRC_MSGSIZE] = EMSGSIZE, 1280 [PRC_HOSTDEAD] = EHOSTDOWN, 1281 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1282 [PRC_UNREACH_NET] = EHOSTUNREACH, 1283 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1284 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1285 [PRC_UNREACH_PORT] = ECONNREFUSED, 1286 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1287 [PRC_PARAMPROB] = ENOPROTOOPT, 1288}; 1289 1290void 1291ip_fasttimo(void) 1292{ 1293 if (ip_drainwanted) { 1294 ip_drain(); 1295 ip_drainwanted = 0; 1296 } 1297} 1298 1299void 1300ip_drainstub(void) 1301{ 1302 ip_drainwanted = 1; 1303} 1304 1305/* 1306 * Forward a packet. If some error occurs return the sender 1307 * an icmp packet. Note we can't always generate a meaningful 1308 * icmp message because icmp doesn't have a large enough repertoire 1309 * of codes and types. 1310 * 1311 * If not forwarding, just drop the packet. This could be confusing 1312 * if ipforwarding was zero but some routing protocol was advancing 1313 * us as a gateway to somewhere. However, we must let the routing 1314 * protocol deal with that. 1315 * 1316 * The srcrt parameter indicates whether the packet is being forwarded 1317 * via a source route. 1318 */ 1319static void 1320ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1321{ 1322 struct ip *ip = mtod(m, struct ip *); 1323 struct rtentry *rt; 1324 int error, type = 0, code = 0, destmtu = 0; 1325 struct mbuf *mcopy; 1326 n_long dest; 1327 union { 1328 struct sockaddr dst; 1329 struct sockaddr_in dst4; 1330 } u; 1331 uint64_t *ips; 1332 struct route *ro; 1333 1334 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1335 "interrupt handler; synchronization assumptions violated"); 1336 1337 /* 1338 * We are now in the output path. 1339 */ 1340 MCLAIM(m, &ip_tx_mowner); 1341 1342 /* 1343 * Clear any in-bound checksum flags for this packet. 1344 */ 1345 m->m_pkthdr.csum_flags = 0; 1346 1347 dest = 0; 1348 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1349 IP_STATINC(IP_STAT_CANTFORWARD); 1350 m_freem(m); 1351 return; 1352 } 1353 1354 if (ip->ip_ttl <= IPTTLDEC) { 1355 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1356 return; 1357 } 1358 1359 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1360 1361 ro = percpu_getref(ipforward_rt_percpu); 1362 rt = rtcache_lookup(ro, &u.dst); 1363 if (rt == NULL) { 1364 percpu_putref(ipforward_rt_percpu); 1365 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1366 return; 1367 } 1368 1369 /* 1370 * Save at most 68 bytes of the packet in case 1371 * we need to generate an ICMP message to the src. 1372 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1373 */ 1374 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1375 if (mcopy) 1376 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1377 1378 ip->ip_ttl -= IPTTLDEC; 1379 1380 /* 1381 * If forwarding packet using same interface that it came in on, 1382 * perhaps should send a redirect to sender to shortcut a hop. 1383 * Only send redirect if source is sending directly to us, 1384 * and if packet was not source routed (or has any options). 1385 * Also, don't send redirect if forwarding using a default route 1386 * or a route modified by a redirect. 1387 */ 1388 if (rt->rt_ifp == rcvif && 1389 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1390 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1391 ipsendredirects && !srcrt) { 1392 if (rt->rt_ifa && 1393 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1394 ifatoia(rt->rt_ifa)->ia_subnet) { 1395 if (rt->rt_flags & RTF_GATEWAY) 1396 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1397 else 1398 dest = ip->ip_dst.s_addr; 1399 /* 1400 * Router requirements says to only send host 1401 * redirects. 1402 */ 1403 type = ICMP_REDIRECT; 1404 code = ICMP_REDIRECT_HOST; 1405 } 1406 } 1407 rtcache_unref(rt, ro); 1408 1409 error = ip_output(m, NULL, ro, 1410 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1411 NULL, NULL); 1412 1413 if (error) { 1414 IP_STATINC(IP_STAT_CANTFORWARD); 1415 goto error; 1416 } 1417 1418 ips = IP_STAT_GETREF(); 1419 ips[IP_STAT_FORWARD]++; 1420 1421 if (type) { 1422 ips[IP_STAT_REDIRECTSENT]++; 1423 IP_STAT_PUTREF(); 1424 goto redirect; 1425 } 1426 1427 IP_STAT_PUTREF(); 1428 if (mcopy) { 1429#ifdef GATEWAY 1430 if (mcopy->m_flags & M_CANFASTFWD) 1431 ipflow_create(ro, mcopy); 1432#endif 1433 m_freem(mcopy); 1434 } 1435 1436 percpu_putref(ipforward_rt_percpu); 1437 return; 1438 1439redirect: 1440error: 1441 if (mcopy == NULL) { 1442 percpu_putref(ipforward_rt_percpu); 1443 return; 1444 } 1445 1446 switch (error) { 1447 1448 case 0: /* forwarded, but need redirect */ 1449 /* type, code set above */ 1450 break; 1451 1452 case ENETUNREACH: /* shouldn't happen, checked above */ 1453 case EHOSTUNREACH: 1454 case ENETDOWN: 1455 case EHOSTDOWN: 1456 default: 1457 type = ICMP_UNREACH; 1458 code = ICMP_UNREACH_HOST; 1459 break; 1460 1461 case EMSGSIZE: 1462 type = ICMP_UNREACH; 1463 code = ICMP_UNREACH_NEEDFRAG; 1464 1465 if ((rt = rtcache_validate(ro)) != NULL) { 1466 destmtu = rt->rt_ifp->if_mtu; 1467 rtcache_unref(rt, ro); 1468 } 1469#ifdef IPSEC 1470 if (ipsec_used) 1471 (void)ipsec4_forward(mcopy, &destmtu); 1472#endif 1473 IP_STATINC(IP_STAT_CANTFRAG); 1474 break; 1475 1476 case ENOBUFS: 1477 /* 1478 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1479 * Requirements for IP Version 4 Routers. Source quench can 1480 * big problem under DoS attacks or if the underlying 1481 * interface is rate-limited. 1482 */ 1483 if (mcopy) 1484 m_freem(mcopy); 1485 percpu_putref(ipforward_rt_percpu); 1486 return; 1487 } 1488 icmp_error(mcopy, type, code, dest, destmtu); 1489 percpu_putref(ipforward_rt_percpu); 1490} 1491 1492void 1493ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1494 struct mbuf *m) 1495{ 1496 struct socket *so = inp->inp_socket; 1497 ifnet_t *ifp; 1498 int inpflags = inp->inp_flags; 1499 struct psref psref; 1500 1501 ifp = m_get_rcvif_psref(m, &psref); 1502 if (__predict_false(ifp == NULL)) 1503 return; /* XXX should report error? */ 1504 1505 if (so->so_options & SO_TIMESTAMP 1506#ifdef SO_OTIMESTAMP 1507 || so->so_options & SO_OTIMESTAMP 1508#endif 1509 ) { 1510 struct timeval tv; 1511 1512 microtime(&tv); 1513#ifdef SO_OTIMESTAMP 1514 if (so->so_options & SO_OTIMESTAMP) { 1515 struct timeval50 tv50; 1516 timeval_to_timeval50(&tv, &tv50); 1517 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50), 1518 SCM_OTIMESTAMP, SOL_SOCKET); 1519 } else 1520#endif 1521 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 1522 SCM_TIMESTAMP, SOL_SOCKET); 1523 if (*mp) 1524 mp = &(*mp)->m_next; 1525 } 1526 if (inpflags & INP_RECVDSTADDR) { 1527 *mp = sbcreatecontrol((void *) &ip->ip_dst, 1528 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1529 if (*mp) 1530 mp = &(*mp)->m_next; 1531 } 1532 if (inpflags & INP_RECVPKTINFO) { 1533 struct in_pktinfo ipi; 1534 ipi.ipi_addr = ip->ip_src; 1535 ipi.ipi_ifindex = ifp->if_index; 1536 *mp = sbcreatecontrol((void *) &ipi, 1537 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP); 1538 if (*mp) 1539 mp = &(*mp)->m_next; 1540 } 1541 if (inpflags & INP_PKTINFO) { 1542 struct in_pktinfo ipi; 1543 ipi.ipi_addr = ip->ip_dst; 1544 ipi.ipi_ifindex = ifp->if_index; 1545 *mp = sbcreatecontrol((void *) &ipi, 1546 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1547 if (*mp) 1548 mp = &(*mp)->m_next; 1549 } 1550 if (inpflags & INP_RECVIF) { 1551 struct sockaddr_dl sdl; 1552 1553 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1554 NULL, 0); 1555 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1556 if (*mp) 1557 mp = &(*mp)->m_next; 1558 } 1559 if (inpflags & INP_RECVTTL) { 1560 *mp = sbcreatecontrol((void *) &ip->ip_ttl, 1561 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1562 if (*mp) 1563 mp = &(*mp)->m_next; 1564 } 1565 m_put_rcvif_psref(ifp, &psref); 1566} 1567 1568/* 1569 * sysctl helper routine for net.inet.ip.forwsrcrt. 1570 */ 1571static int 1572sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1573{ 1574 int error, tmp; 1575 struct sysctlnode node; 1576 1577 node = *rnode; 1578 tmp = ip_forwsrcrt; 1579 node.sysctl_data = &tmp; 1580 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1581 if (error || newp == NULL) 1582 return (error); 1583 1584 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1585 0, NULL, NULL, NULL); 1586 if (error) 1587 return (error); 1588 1589 ip_forwsrcrt = tmp; 1590 1591 return (0); 1592} 1593 1594/* 1595 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1596 * range of the new value and tweaks timers if it changes. 1597 */ 1598static int 1599sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1600{ 1601 int error, tmp; 1602 struct sysctlnode node; 1603 1604 node = *rnode; 1605 tmp = ip_mtudisc_timeout; 1606 node.sysctl_data = &tmp; 1607 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1608 if (error || newp == NULL) 1609 return (error); 1610 if (tmp < 0) 1611 return (EINVAL); 1612 1613 /* XXX NOMPSAFE still need softnet_lock */ 1614 mutex_enter(softnet_lock); 1615 1616 ip_mtudisc_timeout = tmp; 1617 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1618 1619 mutex_exit(softnet_lock); 1620 1621 return (0); 1622} 1623 1624static int 1625sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1626{ 1627 1628 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1629} 1630 1631static void 1632sysctl_net_inet_ip_setup(struct sysctllog **clog) 1633{ 1634 sysctl_createv(clog, 0, NULL, NULL, 1635 CTLFLAG_PERMANENT, 1636 CTLTYPE_NODE, "inet", 1637 SYSCTL_DESCR("PF_INET related settings"), 1638 NULL, 0, NULL, 0, 1639 CTL_NET, PF_INET, CTL_EOL); 1640 sysctl_createv(clog, 0, NULL, NULL, 1641 CTLFLAG_PERMANENT, 1642 CTLTYPE_NODE, "ip", 1643 SYSCTL_DESCR("IPv4 related settings"), 1644 NULL, 0, NULL, 0, 1645 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1646 1647 sysctl_createv(clog, 0, NULL, NULL, 1648 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1649 CTLTYPE_INT, "forwarding", 1650 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1651 NULL, 0, &ipforwarding, 0, 1652 CTL_NET, PF_INET, IPPROTO_IP, 1653 IPCTL_FORWARDING, CTL_EOL); 1654 sysctl_createv(clog, 0, NULL, NULL, 1655 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1656 CTLTYPE_INT, "redirect", 1657 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1658 NULL, 0, &ipsendredirects, 0, 1659 CTL_NET, PF_INET, IPPROTO_IP, 1660 IPCTL_SENDREDIRECTS, CTL_EOL); 1661 sysctl_createv(clog, 0, NULL, NULL, 1662 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1663 CTLTYPE_INT, "ttl", 1664 SYSCTL_DESCR("Default TTL for an INET datagram"), 1665 NULL, 0, &ip_defttl, 0, 1666 CTL_NET, PF_INET, IPPROTO_IP, 1667 IPCTL_DEFTTL, CTL_EOL); 1668#ifdef IPCTL_DEFMTU 1669 sysctl_createv(clog, 0, NULL, NULL, 1670 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1671 CTLTYPE_INT, "mtu", 1672 SYSCTL_DESCR("Default MTA for an INET route"), 1673 NULL, 0, &ip_mtu, 0, 1674 CTL_NET, PF_INET, IPPROTO_IP, 1675 IPCTL_DEFMTU, CTL_EOL); 1676#endif /* IPCTL_DEFMTU */ 1677 sysctl_createv(clog, 0, NULL, NULL, 1678 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1679 CTLTYPE_INT, "forwsrcrt", 1680 SYSCTL_DESCR("Enable forwarding of source-routed " 1681 "datagrams"), 1682 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1683 CTL_NET, PF_INET, IPPROTO_IP, 1684 IPCTL_FORWSRCRT, CTL_EOL); 1685 sysctl_createv(clog, 0, NULL, NULL, 1686 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1687 CTLTYPE_INT, "directed-broadcast", 1688 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1689 NULL, 0, &ip_directedbcast, 0, 1690 CTL_NET, PF_INET, IPPROTO_IP, 1691 IPCTL_DIRECTEDBCAST, CTL_EOL); 1692 sysctl_createv(clog, 0, NULL, NULL, 1693 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1694 CTLTYPE_INT, "allowsrcrt", 1695 SYSCTL_DESCR("Accept source-routed datagrams"), 1696 NULL, 0, &ip_allowsrcrt, 0, 1697 CTL_NET, PF_INET, IPPROTO_IP, 1698 IPCTL_ALLOWSRCRT, CTL_EOL); 1699 1700 sysctl_createv(clog, 0, NULL, NULL, 1701 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1702 CTLTYPE_INT, "mtudisc", 1703 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1704 NULL, 0, &ip_mtudisc, 0, 1705 CTL_NET, PF_INET, IPPROTO_IP, 1706 IPCTL_MTUDISC, CTL_EOL); 1707 sysctl_createv(clog, 0, NULL, NULL, 1708 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1709 CTLTYPE_INT, "anonportmin", 1710 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1711 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1712 CTL_NET, PF_INET, IPPROTO_IP, 1713 IPCTL_ANONPORTMIN, CTL_EOL); 1714 sysctl_createv(clog, 0, NULL, NULL, 1715 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1716 CTLTYPE_INT, "anonportmax", 1717 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1718 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1719 CTL_NET, PF_INET, IPPROTO_IP, 1720 IPCTL_ANONPORTMAX, CTL_EOL); 1721 sysctl_createv(clog, 0, NULL, NULL, 1722 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1723 CTLTYPE_INT, "mtudisctimeout", 1724 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1725 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1726 CTL_NET, PF_INET, IPPROTO_IP, 1727 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1728#ifndef IPNOPRIVPORTS 1729 sysctl_createv(clog, 0, NULL, NULL, 1730 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1731 CTLTYPE_INT, "lowportmin", 1732 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1733 "to assign"), 1734 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1735 CTL_NET, PF_INET, IPPROTO_IP, 1736 IPCTL_LOWPORTMIN, CTL_EOL); 1737 sysctl_createv(clog, 0, NULL, NULL, 1738 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1739 CTLTYPE_INT, "lowportmax", 1740 SYSCTL_DESCR("Highest privileged ephemeral port number " 1741 "to assign"), 1742 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1743 CTL_NET, PF_INET, IPPROTO_IP, 1744 IPCTL_LOWPORTMAX, CTL_EOL); 1745#endif /* IPNOPRIVPORTS */ 1746#if NGRE > 0 1747 sysctl_createv(clog, 0, NULL, NULL, 1748 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1749 CTLTYPE_INT, "grettl", 1750 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1751 NULL, 0, &ip_gre_ttl, 0, 1752 CTL_NET, PF_INET, IPPROTO_IP, 1753 IPCTL_GRE_TTL, CTL_EOL); 1754#endif /* NGRE */ 1755 sysctl_createv(clog, 0, NULL, NULL, 1756 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1757 CTLTYPE_INT, "checkinterface", 1758 SYSCTL_DESCR("Enable receive side of Strong ES model " 1759 "from RFC1122"), 1760 NULL, 0, &ip_checkinterface, 0, 1761 CTL_NET, PF_INET, IPPROTO_IP, 1762 IPCTL_CHECKINTERFACE, CTL_EOL); 1763 sysctl_createv(clog, 0, NULL, NULL, 1764 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1765 CTLTYPE_INT, "random_id", 1766 SYSCTL_DESCR("Assign random ip_id values"), 1767 NULL, 0, &ip_do_randomid, 0, 1768 CTL_NET, PF_INET, IPPROTO_IP, 1769 IPCTL_RANDOMID, CTL_EOL); 1770 sysctl_createv(clog, 0, NULL, NULL, 1771 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1772 CTLTYPE_INT, "do_loopback_cksum", 1773 SYSCTL_DESCR("Perform IP checksum on loopback"), 1774 NULL, 0, &ip_do_loopback_cksum, 0, 1775 CTL_NET, PF_INET, IPPROTO_IP, 1776 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1777 sysctl_createv(clog, 0, NULL, NULL, 1778 CTLFLAG_PERMANENT, 1779 CTLTYPE_STRUCT, "stats", 1780 SYSCTL_DESCR("IP statistics"), 1781 sysctl_net_inet_ip_stats, 0, NULL, 0, 1782 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1783 CTL_EOL); 1784#if NARP 1785 sysctl_createv(clog, 0, NULL, NULL, 1786 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1787 CTLTYPE_INT, "dad_count", 1788 SYSCTL_DESCR("Number of Duplicate Address Detection " 1789 "probes to send"), 1790 NULL, 0, &ip_dad_count, 0, 1791 CTL_NET, PF_INET, IPPROTO_IP, 1792 IPCTL_DAD_COUNT, CTL_EOL); 1793#endif 1794 1795 /* anonportalgo RFC6056 subtree */ 1796 const struct sysctlnode *portalgo_node; 1797 sysctl_createv(clog, 0, NULL, &portalgo_node, 1798 CTLFLAG_PERMANENT, 1799 CTLTYPE_NODE, "anonportalgo", 1800 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1801 NULL, 0, NULL, 0, 1802 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1803 sysctl_createv(clog, 0, &portalgo_node, NULL, 1804 CTLFLAG_PERMANENT, 1805 CTLTYPE_STRING, "available", 1806 SYSCTL_DESCR("available algorithms"), 1807 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1808 CTL_CREATE, CTL_EOL); 1809 sysctl_createv(clog, 0, &portalgo_node, NULL, 1810 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1811 CTLTYPE_STRING, "selected", 1812 SYSCTL_DESCR("selected algorithm"), 1813 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1814 CTL_CREATE, CTL_EOL); 1815 sysctl_createv(clog, 0, &portalgo_node, NULL, 1816 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1817 CTLTYPE_STRUCT, "reserve", 1818 SYSCTL_DESCR("bitmap of reserved ports"), 1819 sysctl_portalgo_reserve4, 0, NULL, 0, 1820 CTL_CREATE, CTL_EOL); 1821} 1822 1823void 1824ip_statinc(u_int stat) 1825{ 1826 1827 KASSERT(stat < IP_NSTATS); 1828 IP_STATINC(stat); 1829} 1830