ip_input.c revision 1.332
1/* $NetBSD: ip_input.c,v 1.332 2016/06/30 06:56:27 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 1.332 2016/06/30 06:56:27 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#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#ifdef COMPAT_50 177#include <compat/sys/time.h> 178#include <compat/sys/socket.h> 179#endif 180 181/* 182 * Note: DIRECTED_BROADCAST is handled this way so that previous 183 * configuration using this option will Just Work. 184 */ 185#ifndef IPDIRECTEDBCAST 186#ifdef DIRECTED_BROADCAST 187#define IPDIRECTEDBCAST 1 188#else 189#define IPDIRECTEDBCAST 0 190#endif /* DIRECTED_BROADCAST */ 191#endif /* IPDIRECTEDBCAST */ 192int ipforwarding = IPFORWARDING; 193int ipsendredirects = IPSENDREDIRECTS; 194int ip_defttl = IPDEFTTL; 195int ip_forwsrcrt = IPFORWSRCRT; 196int ip_directedbcast = IPDIRECTEDBCAST; 197int ip_allowsrcrt = IPALLOWSRCRT; 198int ip_mtudisc = IPMTUDISC; 199int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 200#ifdef DIAGNOSTIC 201int ipprintfs = 0; 202#endif 203 204int ip_do_randomid = 0; 205 206/* 207 * XXX - Setting ip_checkinterface mostly implements the receive side of 208 * the Strong ES model described in RFC 1122, but since the routing table 209 * and transmit implementation do not implement the Strong ES model, 210 * setting this to 1 results in an odd hybrid. 211 * 212 * XXX - ip_checkinterface currently must be disabled if you use ipnat 213 * to translate the destination address to another local interface. 214 * 215 * XXX - ip_checkinterface must be disabled if you add IP aliases 216 * to the loopback interface instead of the interface where the 217 * packets for those addresses are received. 218 */ 219static int ip_checkinterface __read_mostly = 0; 220 221struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 222 223pktqueue_t * ip_pktq __read_mostly; 224pfil_head_t * inet_pfil_hook __read_mostly; 225ipid_state_t * ip_ids __read_mostly; 226percpu_t * ipstat_percpu __read_mostly; 227 228static struct route ipforward_rt __cacheline_aligned; 229 230uint16_t ip_id; 231 232#ifdef INET_CSUM_COUNTERS 233#include <sys/device.h> 234 235struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 236 NULL, "inet", "hwcsum bad"); 237struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 238 NULL, "inet", "hwcsum ok"); 239struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 240 NULL, "inet", "swcsum"); 241 242#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 243 244EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 245EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 246EVCNT_ATTACH_STATIC(ip_swcsum); 247 248#else 249 250#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 251 252#endif /* INET_CSUM_COUNTERS */ 253 254/* 255 * We need to save the IP options in case a protocol wants to respond 256 * to an incoming packet over the same route if the packet got here 257 * using IP source routing. This allows connection establishment and 258 * maintenance when the remote end is on a network that is not known 259 * to us. 260 */ 261 262static int ip_nhops = 0; 263 264static struct ip_srcrt { 265 struct in_addr dst; /* final destination */ 266 char nop; /* one NOP to align */ 267 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 268 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 269} ip_srcrt; 270 271static int ip_drainwanted; 272 273struct sockaddr_in ipaddr = { 274 .sin_len = sizeof(ipaddr), 275 .sin_family = AF_INET, 276}; 277 278static void save_rte(u_char *, struct in_addr); 279 280#ifdef MBUFTRACE 281struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 282struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 283#endif 284 285static void ipintr(void *); 286static void ip_input(struct mbuf *); 287static void ip_forward(struct mbuf *, int, struct ifnet *); 288static bool ip_dooptions(struct mbuf *); 289static struct in_ifaddr *ip_rtaddr(struct in_addr); 290static void sysctl_net_inet_ip_setup(struct sysctllog **); 291 292/* XXX: Not yet enabled. */ 293#define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 294#define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 295 296/* 297 * IP initialization: fill in IP protocol switch table. 298 * All protocols not implemented in kernel go to raw IP protocol handler. 299 */ 300void 301ip_init(void) 302{ 303 const struct protosw *pr; 304 305 in_init(); 306 sysctl_net_inet_ip_setup(NULL); 307 308 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 309 KASSERT(pr != NULL); 310 311 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 312 KASSERT(ip_pktq != NULL); 313 314 for (u_int i = 0; i < IPPROTO_MAX; i++) { 315 ip_protox[i] = pr - inetsw; 316 } 317 for (pr = inetdomain.dom_protosw; 318 pr < inetdomain.dom_protoswNPROTOSW; pr++) 319 if (pr->pr_domain->dom_family == PF_INET && 320 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 321 ip_protox[pr->pr_protocol] = pr - inetsw; 322 323 ip_reass_init(); 324 325 ip_ids = ip_id_init(); 326 ip_id = time_uptime & 0xfffff; 327 328 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 329#ifdef GATEWAY 330 ipflow_init(); 331#endif 332 333 /* Register our Packet Filter hook. */ 334 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 335 KASSERT(inet_pfil_hook != NULL); 336 337#ifdef MBUFTRACE 338 MOWNER_ATTACH(&ip_tx_mowner); 339 MOWNER_ATTACH(&ip_rx_mowner); 340#endif /* MBUFTRACE */ 341 342 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 343} 344 345/* 346 * IP software interrupt routine. 347 */ 348static void 349ipintr(void *arg __unused) 350{ 351 struct mbuf *m; 352 353 KASSERT(cpu_softintr_p()); 354 355 mutex_enter(softnet_lock); 356 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 357 ip_input(m); 358 } 359 mutex_exit(softnet_lock); 360} 361 362/* 363 * IP input routine. Checksum and byte swap header. If fragmented 364 * try to reassemble. Process options. Pass to next level. 365 */ 366static void 367ip_input(struct mbuf *m) 368{ 369 struct ip *ip = NULL; 370 struct in_ifaddr *ia; 371 struct ifaddr *ifa; 372 int hlen = 0, len; 373 int downmatch; 374 int checkif; 375 int srcrt = 0; 376 ifnet_t *ifp; 377 struct psref psref; 378 379 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 380 "interrupt handler; synchronization assumptions violated"); 381 382 MCLAIM(m, &ip_rx_mowner); 383 KASSERT((m->m_flags & M_PKTHDR) != 0); 384 385 ifp = m_get_rcvif_psref(m, &psref); 386 if (__predict_false(ifp == NULL)) 387 goto out; 388 389 /* 390 * If no IP addresses have been set yet but the interfaces 391 * are receiving, can't do anything with incoming packets yet. 392 * Note: we pre-check without locks held. 393 */ 394 if (!TAILQ_FIRST(&in_ifaddrhead)) { 395 goto out; 396 } 397 IP_STATINC(IP_STAT_TOTAL); 398 399 /* 400 * If the IP header is not aligned, slurp it up into a new 401 * mbuf with space for link headers, in the event we forward 402 * it. Otherwise, if it is aligned, make sure the entire 403 * base IP header is in the first mbuf of the chain. 404 */ 405 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 406 if ((m = m_copyup(m, sizeof(struct ip), 407 (max_linkhdr + 3) & ~3)) == NULL) { 408 /* XXXJRT new stat, please */ 409 IP_STATINC(IP_STAT_TOOSMALL); 410 goto out; 411 } 412 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 413 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 414 IP_STATINC(IP_STAT_TOOSMALL); 415 goto out; 416 } 417 } 418 ip = mtod(m, struct ip *); 419 if (ip->ip_v != IPVERSION) { 420 IP_STATINC(IP_STAT_BADVERS); 421 goto out; 422 } 423 hlen = ip->ip_hl << 2; 424 if (hlen < sizeof(struct ip)) { /* minimum header length */ 425 IP_STATINC(IP_STAT_BADHLEN); 426 goto out; 427 } 428 if (hlen > m->m_len) { 429 if ((m = m_pullup(m, hlen)) == NULL) { 430 IP_STATINC(IP_STAT_BADHLEN); 431 goto out; 432 } 433 ip = mtod(m, struct ip *); 434 } 435 436 /* 437 * RFC1122: packets with a multicast source address are 438 * not allowed. 439 */ 440 if (IN_MULTICAST(ip->ip_src.s_addr)) { 441 IP_STATINC(IP_STAT_BADADDR); 442 goto out; 443 } 444 445 /* 127/8 must not appear on wire - RFC1122 */ 446 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 447 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 448 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 449 IP_STATINC(IP_STAT_BADADDR); 450 goto out; 451 } 452 } 453 454 switch (m->m_pkthdr.csum_flags & 455 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | 456 M_CSUM_IPv4_BAD)) { 457 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 458 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 459 IP_STATINC(IP_STAT_BADSUM); 460 goto out; 461 462 case M_CSUM_IPv4: 463 /* Checksum was okay. */ 464 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 465 break; 466 467 default: 468 /* 469 * Must compute it ourselves. Maybe skip checksum on 470 * loopback interfaces. 471 */ 472 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 473 ip_do_loopback_cksum)) { 474 INET_CSUM_COUNTER_INCR(&ip_swcsum); 475 if (in_cksum(m, hlen) != 0) { 476 IP_STATINC(IP_STAT_BADSUM); 477 goto out; 478 } 479 } 480 break; 481 } 482 483 /* Retrieve the packet length. */ 484 len = ntohs(ip->ip_len); 485 486 /* 487 * Check for additional length bogosity 488 */ 489 if (len < hlen) { 490 IP_STATINC(IP_STAT_BADLEN); 491 goto out; 492 } 493 494 /* 495 * Check that the amount of data in the buffers 496 * is as at least much as the IP header would have us expect. 497 * Trim mbufs if longer than we expect. 498 * Drop packet if shorter than we expect. 499 */ 500 if (m->m_pkthdr.len < len) { 501 IP_STATINC(IP_STAT_TOOSHORT); 502 goto out; 503 } 504 if (m->m_pkthdr.len > len) { 505 if (m->m_len == m->m_pkthdr.len) { 506 m->m_len = len; 507 m->m_pkthdr.len = len; 508 } else 509 m_adj(m, len - m->m_pkthdr.len); 510 } 511 512 /* 513 * Assume that we can create a fast-forward IP flow entry 514 * based on this packet. 515 */ 516 m->m_flags |= M_CANFASTFWD; 517 518 /* 519 * Run through list of hooks for input packets. If there are any 520 * filters which require that additional packets in the flow are 521 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 522 * Note that filters must _never_ set this flag, as another filter 523 * in the list may have previously cleared it. 524 */ 525#if defined(IPSEC) 526 if (!ipsec_used || !ipsec_indone(m)) 527#else 528 if (1) 529#endif 530 { 531 struct in_addr odst = ip->ip_dst; 532 bool freed; 533 534 SOFTNET_LOCK(); 535 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 536 SOFTNET_UNLOCK(); 537 if (freed || m == NULL) { 538 goto out; 539 } 540 ip = mtod(m, struct ip *); 541 hlen = ip->ip_hl << 2; 542 543 /* 544 * XXX The setting of "srcrt" here is to prevent ip_forward() 545 * from generating ICMP redirects for packets that have 546 * been redirected by a hook back out on to the same LAN that 547 * they came from and is not an indication that the packet 548 * is being inffluenced by source routing options. This 549 * allows things like 550 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 551 * where tlp0 is both on the 1.1.1.0/24 network and is the 552 * default route for hosts on 1.1.1.0/24. Of course this 553 * also requires a "map tlp0 ..." to complete the story. 554 * One might argue whether or not this kind of network config. 555 * should be supported in this manner... 556 */ 557 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 558 } 559 560#ifdef ALTQ 561 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 562 if (altq_input) { 563 SOFTNET_LOCK(); 564 if ((*altq_input)(m, AF_INET) == 0) { 565 /* Packet dropped by traffic conditioner. */ 566 SOFTNET_UNLOCK(); 567 goto out; 568 } 569 SOFTNET_UNLOCK(); 570 } 571#endif 572 573 /* 574 * Process options and, if not destined for us, 575 * ship it on. ip_dooptions returns 1 when an 576 * error was detected (causing an icmp message 577 * to be sent and the original packet to be freed). 578 */ 579 ip_nhops = 0; /* for source routed packets */ 580 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 581 goto out; 582 583 /* 584 * Enable a consistency check between the destination address 585 * and the arrival interface for a unicast packet (the RFC 1122 586 * strong ES model) if IP forwarding is disabled and the packet 587 * is not locally generated. 588 * 589 * XXX - Checking also should be disabled if the destination 590 * address is ipnat'ed to a different interface. 591 * 592 * XXX - Checking is incompatible with IP aliases added 593 * to the loopback interface instead of the interface where 594 * the packets are received. 595 * 596 * XXX - We need to add a per ifaddr flag for this so that 597 * we get finer grain control. 598 */ 599 checkif = ip_checkinterface && (ipforwarding == 0) && 600 (ifp->if_flags & IFF_LOOPBACK) == 0; 601 602 /* 603 * Check our list of addresses, to see if the packet is for us. 604 * 605 * Traditional 4.4BSD did not consult IFF_UP at all. 606 * The behavior here is to treat addresses on !IFF_UP interface 607 * or IN_IFF_NOTREADY addresses as not mine. 608 */ 609 downmatch = 0; 610 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 611 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 612 if (ia->ia4_flags & IN_IFF_NOTREADY) 613 continue; 614 if (checkif && ia->ia_ifp != ifp) 615 continue; 616 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 617 break; 618 else 619 downmatch++; 620 } 621 } 622 if (ia != NULL) 623 goto ours; 624 if (ifp->if_flags & IFF_BROADCAST) { 625 IFADDR_FOREACH(ifa, ifp) { 626 if (ifa->ifa_addr->sa_family != AF_INET) 627 continue; 628 ia = ifatoia(ifa); 629 if (ia->ia4_flags & IN_IFF_NOTREADY) 630 continue; 631 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 632 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 633 /* 634 * Look for all-0's host part (old broadcast addr), 635 * either for subnet or net. 636 */ 637 ip->ip_dst.s_addr == ia->ia_subnet || 638 ip->ip_dst.s_addr == ia->ia_net) 639 goto ours; 640 /* 641 * An interface with IP address zero accepts 642 * all packets that arrive on that interface. 643 */ 644 if (in_nullhost(ia->ia_addr.sin_addr)) 645 goto ours; 646 } 647 } 648 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 649#ifdef MROUTING 650 extern struct socket *ip_mrouter; 651 652 if (ip_mrouter) { 653 /* 654 * If we are acting as a multicast router, all 655 * incoming multicast packets are passed to the 656 * kernel-level multicast forwarding function. 657 * The packet is returned (relatively) intact; if 658 * ip_mforward() returns a non-zero value, the packet 659 * must be discarded, else it may be accepted below. 660 * 661 * (The IP ident field is put in the same byte order 662 * as expected when ip_mforward() is called from 663 * ip_output().) 664 */ 665 SOFTNET_LOCK(); 666 if (ip_mforward(m, ifp) != 0) { 667 SOFTNET_UNLOCK(); 668 IP_STATINC(IP_STAT_CANTFORWARD); 669 goto out; 670 } 671 SOFTNET_UNLOCK(); 672 673 /* 674 * The process-level routing demon needs to receive 675 * all multicast IGMP packets, whether or not this 676 * host belongs to their destination groups. 677 */ 678 if (ip->ip_p == IPPROTO_IGMP) { 679 goto ours; 680 } 681 IP_STATINC(IP_STAT_CANTFORWARD); 682 } 683#endif 684 /* 685 * See if we belong to the destination multicast group on the 686 * arrival interface. 687 */ 688 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 689 IP_STATINC(IP_STAT_CANTFORWARD); 690 goto out; 691 } 692 goto ours; 693 } 694 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 695 in_nullhost(ip->ip_dst)) 696 goto ours; 697 698 /* 699 * Not for us; forward if possible and desirable. 700 */ 701 if (ipforwarding == 0) { 702 m_put_rcvif_psref(ifp, &psref); 703 IP_STATINC(IP_STAT_CANTFORWARD); 704 m_freem(m); 705 } else { 706 /* 707 * If ip_dst matched any of my address on !IFF_UP interface, 708 * and there's no IFF_UP interface that matches ip_dst, 709 * send icmp unreach. Forwarding it will result in in-kernel 710 * forwarding loop till TTL goes to 0. 711 */ 712 if (downmatch) { 713 m_put_rcvif_psref(ifp, &psref); 714 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 715 IP_STATINC(IP_STAT_CANTFORWARD); 716 return; 717 } 718#ifdef IPSEC 719 /* Perform IPsec, if any. */ 720 if (ipsec_used) { 721 SOFTNET_LOCK(); 722 if (ipsec4_input(m, IP_FORWARDING | 723 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 724 SOFTNET_UNLOCK(); 725 goto out; 726 } 727 SOFTNET_UNLOCK(); 728 } 729#endif 730 ip_forward(m, srcrt, ifp); 731 m_put_rcvif_psref(ifp, &psref); 732 } 733 return; 734 735ours: 736 m_put_rcvif_psref(ifp, &psref); 737 ifp = NULL; 738 739 /* 740 * If offset or IP_MF are set, must reassemble. 741 */ 742 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 743 /* 744 * Pass to IP reassembly mechanism. 745 */ 746 if (ip_reass_packet(&m, ip) != 0) { 747 /* Failed; invalid fragment(s) or packet. */ 748 goto out; 749 } 750 if (m == NULL) { 751 /* More fragments should come; silently return. */ 752 goto out; 753 } 754 /* 755 * Reassembly is done, we have the final packet. 756 * Updated cached data in local variable(s). 757 */ 758 ip = mtod(m, struct ip *); 759 hlen = ip->ip_hl << 2; 760 } 761 762#ifdef IPSEC 763 /* 764 * Enforce IPsec policy checking if we are seeing last header. 765 * Note that we do not visit this with protocols with PCB layer 766 * code - like UDP/TCP/raw IP. 767 */ 768 if (ipsec_used && 769 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 770 SOFTNET_LOCK(); 771 if (ipsec4_input(m, 0) != 0) { 772 SOFTNET_UNLOCK(); 773 goto out; 774 } 775 SOFTNET_UNLOCK(); 776 } 777#endif 778 779 /* 780 * Switch out to protocol's input routine. 781 */ 782#if IFA_STATS 783 if (ia && ip) 784 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 785#endif 786 IP_STATINC(IP_STAT_DELIVERED); 787 788 const int off = hlen, nh = ip->ip_p; 789 790 SOFTNET_LOCK(); 791 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 792 SOFTNET_UNLOCK(); 793 return; 794 795out: 796 m_put_rcvif_psref(ifp, &psref); 797 if (m != NULL) 798 m_freem(m); 799} 800 801/* 802 * IP timer processing. 803 */ 804void 805ip_slowtimo(void) 806{ 807 808 mutex_enter(softnet_lock); 809 KERNEL_LOCK(1, NULL); 810 811 ip_reass_slowtimo(); 812 813 KERNEL_UNLOCK_ONE(NULL); 814 mutex_exit(softnet_lock); 815} 816 817/* 818 * IP drain processing. 819 */ 820void 821ip_drain(void) 822{ 823 824 KERNEL_LOCK(1, NULL); 825 ip_reass_drain(); 826 KERNEL_UNLOCK_ONE(NULL); 827} 828 829/* 830 * ip_dooptions: perform option processing on a datagram, possibly discarding 831 * it if bad options are encountered, or forwarding it if source-routed. 832 * 833 * => Returns true if packet has been forwarded/freed. 834 * => Returns false if the packet should be processed further. 835 */ 836static bool 837ip_dooptions(struct mbuf *m) 838{ 839 struct ip *ip = mtod(m, struct ip *); 840 u_char *cp, *cp0; 841 struct ip_timestamp *ipt; 842 struct in_ifaddr *ia; 843 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 844 struct in_addr dst; 845 n_time ntime; 846 847 dst = ip->ip_dst; 848 cp = (u_char *)(ip + 1); 849 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 850 for (; cnt > 0; cnt -= optlen, cp += optlen) { 851 opt = cp[IPOPT_OPTVAL]; 852 if (opt == IPOPT_EOL) 853 break; 854 if (opt == IPOPT_NOP) 855 optlen = 1; 856 else { 857 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 858 code = &cp[IPOPT_OLEN] - (u_char *)ip; 859 goto bad; 860 } 861 optlen = cp[IPOPT_OLEN]; 862 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 863 code = &cp[IPOPT_OLEN] - (u_char *)ip; 864 goto bad; 865 } 866 } 867 switch (opt) { 868 869 default: 870 break; 871 872 /* 873 * Source routing with record. 874 * Find interface with current destination address. 875 * If none on this machine then drop if strictly routed, 876 * or do nothing if loosely routed. 877 * Record interface address and bring up next address 878 * component. If strictly routed make sure next 879 * address is on directly accessible net. 880 */ 881 case IPOPT_LSRR: 882 case IPOPT_SSRR: 883 if (ip_allowsrcrt == 0) { 884 type = ICMP_UNREACH; 885 code = ICMP_UNREACH_NET_PROHIB; 886 goto bad; 887 } 888 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 889 code = &cp[IPOPT_OLEN] - (u_char *)ip; 890 goto bad; 891 } 892 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 893 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 894 goto bad; 895 } 896 ipaddr.sin_addr = ip->ip_dst; 897 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 898 if (ia == 0) { 899 if (opt == IPOPT_SSRR) { 900 type = ICMP_UNREACH; 901 code = ICMP_UNREACH_SRCFAIL; 902 goto bad; 903 } 904 /* 905 * Loose routing, and not at next destination 906 * yet; nothing to do except forward. 907 */ 908 break; 909 } 910 off--; /* 0 origin */ 911 if ((off + sizeof(struct in_addr)) > optlen) { 912 /* 913 * End of source route. Should be for us. 914 */ 915 save_rte(cp, ip->ip_src); 916 break; 917 } 918 /* 919 * locate outgoing interface 920 */ 921 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 922 sizeof(ipaddr.sin_addr)); 923 if (opt == IPOPT_SSRR) 924 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 925 else 926 ia = ip_rtaddr(ipaddr.sin_addr); 927 if (ia == 0) { 928 type = ICMP_UNREACH; 929 code = ICMP_UNREACH_SRCFAIL; 930 goto bad; 931 } 932 ip->ip_dst = ipaddr.sin_addr; 933 bcopy((void *)&ia->ia_addr.sin_addr, 934 (void *)(cp + off), sizeof(struct in_addr)); 935 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 936 /* 937 * Let ip_intr's mcast routing check handle mcast pkts 938 */ 939 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 940 break; 941 942 case IPOPT_RR: 943 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 944 code = &cp[IPOPT_OLEN] - (u_char *)ip; 945 goto bad; 946 } 947 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 948 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 949 goto bad; 950 } 951 /* 952 * If no space remains, ignore. 953 */ 954 off--; /* 0 origin */ 955 if ((off + sizeof(struct in_addr)) > optlen) 956 break; 957 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 958 sizeof(ipaddr.sin_addr)); 959 /* 960 * locate outgoing interface; if we're the destination, 961 * use the incoming interface (should be same). 962 */ 963 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 964 == NULL && 965 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 966 type = ICMP_UNREACH; 967 code = ICMP_UNREACH_HOST; 968 goto bad; 969 } 970 bcopy((void *)&ia->ia_addr.sin_addr, 971 (void *)(cp + off), sizeof(struct in_addr)); 972 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 973 break; 974 975 case IPOPT_TS: 976 code = cp - (u_char *)ip; 977 ipt = (struct ip_timestamp *)cp; 978 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 979 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 980 goto bad; 981 } 982 if (ipt->ipt_ptr < 5) { 983 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 984 goto bad; 985 } 986 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 987 if (++ipt->ipt_oflw == 0) { 988 code = (u_char *)&ipt->ipt_ptr - 989 (u_char *)ip; 990 goto bad; 991 } 992 break; 993 } 994 cp0 = (cp + ipt->ipt_ptr - 1); 995 switch (ipt->ipt_flg) { 996 997 case IPOPT_TS_TSONLY: 998 break; 999 1000 case IPOPT_TS_TSANDADDR: { 1001 struct ifnet *rcvif; 1002 int s; 1003 1004 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1005 sizeof(struct in_addr) > ipt->ipt_len) { 1006 code = (u_char *)&ipt->ipt_ptr - 1007 (u_char *)ip; 1008 goto bad; 1009 } 1010 ipaddr.sin_addr = dst; 1011 rcvif = m_get_rcvif(m, &s); 1012 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1013 rcvif)); 1014 m_put_rcvif(rcvif, &s); 1015 if (ia == 0) 1016 continue; 1017 bcopy(&ia->ia_addr.sin_addr, 1018 cp0, sizeof(struct in_addr)); 1019 ipt->ipt_ptr += sizeof(struct in_addr); 1020 break; 1021 } 1022 1023 case IPOPT_TS_PRESPEC: 1024 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1025 sizeof(struct in_addr) > ipt->ipt_len) { 1026 code = (u_char *)&ipt->ipt_ptr - 1027 (u_char *)ip; 1028 goto bad; 1029 } 1030 memcpy(&ipaddr.sin_addr, cp0, 1031 sizeof(struct in_addr)); 1032 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1033 == NULL) 1034 continue; 1035 ipt->ipt_ptr += sizeof(struct in_addr); 1036 break; 1037 1038 default: 1039 /* XXX can't take &ipt->ipt_flg */ 1040 code = (u_char *)&ipt->ipt_ptr - 1041 (u_char *)ip + 1; 1042 goto bad; 1043 } 1044 ntime = iptime(); 1045 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1046 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1047 sizeof(n_time)); 1048 ipt->ipt_ptr += sizeof(n_time); 1049 } 1050 } 1051 if (forward) { 1052 struct ifnet *rcvif; 1053 struct psref psref; 1054 1055 if (ip_forwsrcrt == 0) { 1056 type = ICMP_UNREACH; 1057 code = ICMP_UNREACH_SRCFAIL; 1058 goto bad; 1059 } 1060 1061 rcvif = m_get_rcvif_psref(m, &psref); 1062 if (__predict_false(rcvif == NULL)) { 1063 type = ICMP_UNREACH; 1064 code = ICMP_UNREACH_HOST; 1065 goto bad; 1066 } 1067 ip_forward(m, 1, rcvif); 1068 m_put_rcvif_psref(rcvif, &psref); 1069 return true; 1070 } 1071 return false; 1072bad: 1073 icmp_error(m, type, code, 0, 0); 1074 IP_STATINC(IP_STAT_BADOPTIONS); 1075 return true; 1076} 1077 1078/* 1079 * ip_rtaddr: given address of next destination (final or next hop), 1080 * return internet address info of interface to be used to get there. 1081 */ 1082static struct in_ifaddr * 1083ip_rtaddr(struct in_addr dst) 1084{ 1085 struct rtentry *rt; 1086 union { 1087 struct sockaddr dst; 1088 struct sockaddr_in dst4; 1089 } u; 1090 1091 sockaddr_in_init(&u.dst4, &dst, 0); 1092 1093 SOFTNET_LOCK(); 1094 rt = rtcache_lookup(&ipforward_rt, &u.dst); 1095 SOFTNET_UNLOCK(); 1096 if (rt == NULL) 1097 return NULL; 1098 1099 return ifatoia(rt->rt_ifa); 1100} 1101 1102/* 1103 * save_rte: save incoming source route for use in replies, to be picked 1104 * up later by ip_srcroute if the receiver is interested. 1105 */ 1106static void 1107save_rte(u_char *option, struct in_addr dst) 1108{ 1109 unsigned olen; 1110 1111 olen = option[IPOPT_OLEN]; 1112 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1113 return; 1114 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen); 1115 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1116 ip_srcrt.dst = dst; 1117} 1118 1119/* 1120 * Retrieve incoming source route for use in replies, 1121 * in the same form used by setsockopt. 1122 * The first hop is placed before the options, will be removed later. 1123 */ 1124struct mbuf * 1125ip_srcroute(void) 1126{ 1127 struct in_addr *p, *q; 1128 struct mbuf *m; 1129 1130 if (ip_nhops == 0) 1131 return NULL; 1132 m = m_get(M_DONTWAIT, MT_SOOPTS); 1133 if (m == 0) 1134 return NULL; 1135 1136 MCLAIM(m, &inetdomain.dom_mowner); 1137#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1138 1139 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1140 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1141 OPTSIZ; 1142 1143 /* 1144 * First save first hop for return route 1145 */ 1146 p = &ip_srcrt.route[ip_nhops - 1]; 1147 *(mtod(m, struct in_addr *)) = *p--; 1148 1149 /* 1150 * Copy option fields and padding (nop) to mbuf. 1151 */ 1152 ip_srcrt.nop = IPOPT_NOP; 1153 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1154 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1155 OPTSIZ); 1156 q = (struct in_addr *)(mtod(m, char *) + 1157 sizeof(struct in_addr) + OPTSIZ); 1158#undef OPTSIZ 1159 /* 1160 * Record return path as an IP source route, 1161 * reversing the path (pointers are now aligned). 1162 */ 1163 while (p >= ip_srcrt.route) { 1164 *q++ = *p--; 1165 } 1166 /* 1167 * Last hop goes to final destination. 1168 */ 1169 *q = ip_srcrt.dst; 1170 return (m); 1171} 1172 1173const int inetctlerrmap[PRC_NCMDS] = { 1174 [PRC_MSGSIZE] = EMSGSIZE, 1175 [PRC_HOSTDEAD] = EHOSTDOWN, 1176 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1177 [PRC_UNREACH_NET] = EHOSTUNREACH, 1178 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1179 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1180 [PRC_UNREACH_PORT] = ECONNREFUSED, 1181 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1182 [PRC_PARAMPROB] = ENOPROTOOPT, 1183}; 1184 1185void 1186ip_fasttimo(void) 1187{ 1188 if (ip_drainwanted) { 1189 ip_drain(); 1190 ip_drainwanted = 0; 1191 } 1192} 1193 1194void 1195ip_drainstub(void) 1196{ 1197 ip_drainwanted = 1; 1198} 1199 1200/* 1201 * Forward a packet. If some error occurs return the sender 1202 * an icmp packet. Note we can't always generate a meaningful 1203 * icmp message because icmp doesn't have a large enough repertoire 1204 * of codes and types. 1205 * 1206 * If not forwarding, just drop the packet. This could be confusing 1207 * if ipforwarding was zero but some routing protocol was advancing 1208 * us as a gateway to somewhere. However, we must let the routing 1209 * protocol deal with that. 1210 * 1211 * The srcrt parameter indicates whether the packet is being forwarded 1212 * via a source route. 1213 */ 1214static void 1215ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1216{ 1217 struct ip *ip = mtod(m, struct ip *); 1218 struct rtentry *rt; 1219 int error, type = 0, code = 0, destmtu = 0; 1220 struct mbuf *mcopy; 1221 n_long dest; 1222 union { 1223 struct sockaddr dst; 1224 struct sockaddr_in dst4; 1225 } u; 1226 uint64_t *ips; 1227 1228 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1229 "interrupt handler; synchronization assumptions violated"); 1230 1231 /* 1232 * We are now in the output path. 1233 */ 1234 MCLAIM(m, &ip_tx_mowner); 1235 1236 /* 1237 * Clear any in-bound checksum flags for this packet. 1238 */ 1239 m->m_pkthdr.csum_flags = 0; 1240 1241 dest = 0; 1242 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1243 IP_STATINC(IP_STAT_CANTFORWARD); 1244 m_freem(m); 1245 return; 1246 } 1247 1248 SOFTNET_LOCK(); 1249 1250 if (ip->ip_ttl <= IPTTLDEC) { 1251 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1252 SOFTNET_UNLOCK(); 1253 return; 1254 } 1255 1256 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1257 1258 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1259 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1260 SOFTNET_UNLOCK(); 1261 return; 1262 } 1263 1264 /* 1265 * Save at most 68 bytes of the packet in case 1266 * we need to generate an ICMP message to the src. 1267 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1268 */ 1269 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1270 if (mcopy) 1271 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1272 1273 ip->ip_ttl -= IPTTLDEC; 1274 1275 /* 1276 * If forwarding packet using same interface that it came in on, 1277 * perhaps should send a redirect to sender to shortcut a hop. 1278 * Only send redirect if source is sending directly to us, 1279 * and if packet was not source routed (or has any options). 1280 * Also, don't send redirect if forwarding using a default route 1281 * or a route modified by a redirect. 1282 */ 1283 if (rt->rt_ifp == rcvif && 1284 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1285 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1286 ipsendredirects && !srcrt) { 1287 if (rt->rt_ifa && 1288 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1289 ifatoia(rt->rt_ifa)->ia_subnet) { 1290 if (rt->rt_flags & RTF_GATEWAY) 1291 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1292 else 1293 dest = ip->ip_dst.s_addr; 1294 /* 1295 * Router requirements says to only send host 1296 * redirects. 1297 */ 1298 type = ICMP_REDIRECT; 1299 code = ICMP_REDIRECT_HOST; 1300 } 1301 } 1302 1303 error = ip_output(m, NULL, &ipforward_rt, 1304 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1305 NULL, NULL); 1306 1307 if (error) { 1308 IP_STATINC(IP_STAT_CANTFORWARD); 1309 goto error; 1310 } 1311 1312 ips = IP_STAT_GETREF(); 1313 ips[IP_STAT_FORWARD]++; 1314 1315 if (type) { 1316 ips[IP_STAT_REDIRECTSENT]++; 1317 IP_STAT_PUTREF(); 1318 goto redirect; 1319 } 1320 1321 IP_STAT_PUTREF(); 1322 if (mcopy) { 1323#ifdef GATEWAY 1324 if (mcopy->m_flags & M_CANFASTFWD) 1325 ipflow_create(&ipforward_rt, mcopy); 1326#endif 1327 m_freem(mcopy); 1328 } 1329 1330 SOFTNET_UNLOCK(); 1331 return; 1332 1333redirect: 1334error: 1335 if (mcopy == NULL) { 1336 SOFTNET_UNLOCK(); 1337 return; 1338 } 1339 1340 switch (error) { 1341 1342 case 0: /* forwarded, but need redirect */ 1343 /* type, code set above */ 1344 break; 1345 1346 case ENETUNREACH: /* shouldn't happen, checked above */ 1347 case EHOSTUNREACH: 1348 case ENETDOWN: 1349 case EHOSTDOWN: 1350 default: 1351 type = ICMP_UNREACH; 1352 code = ICMP_UNREACH_HOST; 1353 break; 1354 1355 case EMSGSIZE: 1356 type = ICMP_UNREACH; 1357 code = ICMP_UNREACH_NEEDFRAG; 1358 1359 if ((rt = rtcache_validate(&ipforward_rt)) != NULL) 1360 destmtu = rt->rt_ifp->if_mtu; 1361#ifdef IPSEC 1362 if (ipsec_used) 1363 (void)ipsec4_forward(mcopy, &destmtu); 1364#endif 1365 IP_STATINC(IP_STAT_CANTFRAG); 1366 break; 1367 1368 case ENOBUFS: 1369 /* 1370 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1371 * Requirements for IP Version 4 Routers. Source quench can 1372 * big problem under DoS attacks or if the underlying 1373 * interface is rate-limited. 1374 */ 1375 if (mcopy) 1376 m_freem(mcopy); 1377 SOFTNET_UNLOCK(); 1378 return; 1379 } 1380 icmp_error(mcopy, type, code, dest, destmtu); 1381 SOFTNET_UNLOCK(); 1382} 1383 1384void 1385ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1386 struct mbuf *m) 1387{ 1388 struct socket *so = inp->inp_socket; 1389 ifnet_t *ifp; 1390 int inpflags = inp->inp_flags; 1391 struct psref psref; 1392 1393 ifp = m_get_rcvif_psref(m, &psref); 1394 if (__predict_false(ifp == NULL)) 1395 return; /* XXX should report error? */ 1396 1397 if (so->so_options & SO_TIMESTAMP 1398#ifdef SO_OTIMESTAMP 1399 || so->so_options & SO_OTIMESTAMP 1400#endif 1401 ) { 1402 struct timeval tv; 1403 1404 microtime(&tv); 1405#ifdef SO_OTIMESTAMP 1406 if (so->so_options & SO_OTIMESTAMP) { 1407 struct timeval50 tv50; 1408 timeval_to_timeval50(&tv, &tv50); 1409 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50), 1410 SCM_OTIMESTAMP, SOL_SOCKET); 1411 } else 1412#endif 1413 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 1414 SCM_TIMESTAMP, SOL_SOCKET); 1415 if (*mp) 1416 mp = &(*mp)->m_next; 1417 } 1418 if (inpflags & INP_RECVDSTADDR) { 1419 *mp = sbcreatecontrol((void *) &ip->ip_dst, 1420 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1421 if (*mp) 1422 mp = &(*mp)->m_next; 1423 } 1424 if (inpflags & INP_RECVPKTINFO) { 1425 struct in_pktinfo ipi; 1426 ipi.ipi_addr = ip->ip_src; 1427 ipi.ipi_ifindex = ifp->if_index; 1428 *mp = sbcreatecontrol((void *) &ipi, 1429 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP); 1430 if (*mp) 1431 mp = &(*mp)->m_next; 1432 } 1433 if (inpflags & INP_PKTINFO) { 1434 struct in_pktinfo ipi; 1435 ipi.ipi_addr = ip->ip_dst; 1436 ipi.ipi_ifindex = ifp->if_index; 1437 *mp = sbcreatecontrol((void *) &ipi, 1438 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1439 if (*mp) 1440 mp = &(*mp)->m_next; 1441 } 1442 if (inpflags & INP_RECVIF) { 1443 struct sockaddr_dl sdl; 1444 1445 sockaddr_dl_init(&sdl, sizeof(sdl), ifp ? 1446 ifp->if_index : 0, 0, NULL, 0, NULL, 0); 1447 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1448 if (*mp) 1449 mp = &(*mp)->m_next; 1450 } 1451 if (inpflags & INP_RECVTTL) { 1452 *mp = sbcreatecontrol((void *) &ip->ip_ttl, 1453 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1454 if (*mp) 1455 mp = &(*mp)->m_next; 1456 } 1457 m_put_rcvif_psref(ifp, &psref); 1458} 1459 1460/* 1461 * sysctl helper routine for net.inet.ip.forwsrcrt. 1462 */ 1463static int 1464sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1465{ 1466 int error, tmp; 1467 struct sysctlnode node; 1468 1469 node = *rnode; 1470 tmp = ip_forwsrcrt; 1471 node.sysctl_data = &tmp; 1472 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1473 if (error || newp == NULL) 1474 return (error); 1475 1476 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1477 0, NULL, NULL, NULL); 1478 if (error) 1479 return (error); 1480 1481 ip_forwsrcrt = tmp; 1482 1483 return (0); 1484} 1485 1486/* 1487 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1488 * range of the new value and tweaks timers if it changes. 1489 */ 1490static int 1491sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1492{ 1493 int error, tmp; 1494 struct sysctlnode node; 1495 1496 node = *rnode; 1497 tmp = ip_mtudisc_timeout; 1498 node.sysctl_data = &tmp; 1499 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1500 if (error || newp == NULL) 1501 return (error); 1502 if (tmp < 0) 1503 return (EINVAL); 1504 1505 mutex_enter(softnet_lock); 1506 1507 ip_mtudisc_timeout = tmp; 1508 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1509 1510 mutex_exit(softnet_lock); 1511 1512 return (0); 1513} 1514 1515static int 1516sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1517{ 1518 1519 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1520} 1521 1522static void 1523sysctl_net_inet_ip_setup(struct sysctllog **clog) 1524{ 1525 sysctl_createv(clog, 0, NULL, NULL, 1526 CTLFLAG_PERMANENT, 1527 CTLTYPE_NODE, "inet", 1528 SYSCTL_DESCR("PF_INET related settings"), 1529 NULL, 0, NULL, 0, 1530 CTL_NET, PF_INET, CTL_EOL); 1531 sysctl_createv(clog, 0, NULL, NULL, 1532 CTLFLAG_PERMANENT, 1533 CTLTYPE_NODE, "ip", 1534 SYSCTL_DESCR("IPv4 related settings"), 1535 NULL, 0, NULL, 0, 1536 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1537 1538 sysctl_createv(clog, 0, NULL, NULL, 1539 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1540 CTLTYPE_INT, "forwarding", 1541 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1542 NULL, 0, &ipforwarding, 0, 1543 CTL_NET, PF_INET, IPPROTO_IP, 1544 IPCTL_FORWARDING, CTL_EOL); 1545 sysctl_createv(clog, 0, NULL, NULL, 1546 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1547 CTLTYPE_INT, "redirect", 1548 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1549 NULL, 0, &ipsendredirects, 0, 1550 CTL_NET, PF_INET, IPPROTO_IP, 1551 IPCTL_SENDREDIRECTS, CTL_EOL); 1552 sysctl_createv(clog, 0, NULL, NULL, 1553 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1554 CTLTYPE_INT, "ttl", 1555 SYSCTL_DESCR("Default TTL for an INET datagram"), 1556 NULL, 0, &ip_defttl, 0, 1557 CTL_NET, PF_INET, IPPROTO_IP, 1558 IPCTL_DEFTTL, CTL_EOL); 1559#ifdef IPCTL_DEFMTU 1560 sysctl_createv(clog, 0, NULL, NULL, 1561 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1562 CTLTYPE_INT, "mtu", 1563 SYSCTL_DESCR("Default MTA for an INET route"), 1564 NULL, 0, &ip_mtu, 0, 1565 CTL_NET, PF_INET, IPPROTO_IP, 1566 IPCTL_DEFMTU, CTL_EOL); 1567#endif /* IPCTL_DEFMTU */ 1568 sysctl_createv(clog, 0, NULL, NULL, 1569 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1570 CTLTYPE_INT, "forwsrcrt", 1571 SYSCTL_DESCR("Enable forwarding of source-routed " 1572 "datagrams"), 1573 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1574 CTL_NET, PF_INET, IPPROTO_IP, 1575 IPCTL_FORWSRCRT, CTL_EOL); 1576 sysctl_createv(clog, 0, NULL, NULL, 1577 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1578 CTLTYPE_INT, "directed-broadcast", 1579 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1580 NULL, 0, &ip_directedbcast, 0, 1581 CTL_NET, PF_INET, IPPROTO_IP, 1582 IPCTL_DIRECTEDBCAST, CTL_EOL); 1583 sysctl_createv(clog, 0, NULL, NULL, 1584 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1585 CTLTYPE_INT, "allowsrcrt", 1586 SYSCTL_DESCR("Accept source-routed datagrams"), 1587 NULL, 0, &ip_allowsrcrt, 0, 1588 CTL_NET, PF_INET, IPPROTO_IP, 1589 IPCTL_ALLOWSRCRT, CTL_EOL); 1590 1591 sysctl_createv(clog, 0, NULL, NULL, 1592 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1593 CTLTYPE_INT, "mtudisc", 1594 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1595 NULL, 0, &ip_mtudisc, 0, 1596 CTL_NET, PF_INET, IPPROTO_IP, 1597 IPCTL_MTUDISC, CTL_EOL); 1598 sysctl_createv(clog, 0, NULL, NULL, 1599 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1600 CTLTYPE_INT, "anonportmin", 1601 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1602 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1603 CTL_NET, PF_INET, IPPROTO_IP, 1604 IPCTL_ANONPORTMIN, CTL_EOL); 1605 sysctl_createv(clog, 0, NULL, NULL, 1606 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1607 CTLTYPE_INT, "anonportmax", 1608 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1609 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1610 CTL_NET, PF_INET, IPPROTO_IP, 1611 IPCTL_ANONPORTMAX, CTL_EOL); 1612 sysctl_createv(clog, 0, NULL, NULL, 1613 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1614 CTLTYPE_INT, "mtudisctimeout", 1615 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1616 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1617 CTL_NET, PF_INET, IPPROTO_IP, 1618 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1619#ifndef IPNOPRIVPORTS 1620 sysctl_createv(clog, 0, NULL, NULL, 1621 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1622 CTLTYPE_INT, "lowportmin", 1623 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1624 "to assign"), 1625 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1626 CTL_NET, PF_INET, IPPROTO_IP, 1627 IPCTL_LOWPORTMIN, CTL_EOL); 1628 sysctl_createv(clog, 0, NULL, NULL, 1629 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1630 CTLTYPE_INT, "lowportmax", 1631 SYSCTL_DESCR("Highest privileged ephemeral port number " 1632 "to assign"), 1633 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1634 CTL_NET, PF_INET, IPPROTO_IP, 1635 IPCTL_LOWPORTMAX, CTL_EOL); 1636#endif /* IPNOPRIVPORTS */ 1637#if NGRE > 0 1638 sysctl_createv(clog, 0, NULL, NULL, 1639 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1640 CTLTYPE_INT, "grettl", 1641 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1642 NULL, 0, &ip_gre_ttl, 0, 1643 CTL_NET, PF_INET, IPPROTO_IP, 1644 IPCTL_GRE_TTL, CTL_EOL); 1645#endif /* NGRE */ 1646 sysctl_createv(clog, 0, NULL, NULL, 1647 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1648 CTLTYPE_INT, "checkinterface", 1649 SYSCTL_DESCR("Enable receive side of Strong ES model " 1650 "from RFC1122"), 1651 NULL, 0, &ip_checkinterface, 0, 1652 CTL_NET, PF_INET, IPPROTO_IP, 1653 IPCTL_CHECKINTERFACE, CTL_EOL); 1654 sysctl_createv(clog, 0, NULL, NULL, 1655 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1656 CTLTYPE_INT, "random_id", 1657 SYSCTL_DESCR("Assign random ip_id values"), 1658 NULL, 0, &ip_do_randomid, 0, 1659 CTL_NET, PF_INET, IPPROTO_IP, 1660 IPCTL_RANDOMID, CTL_EOL); 1661 sysctl_createv(clog, 0, NULL, NULL, 1662 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1663 CTLTYPE_INT, "do_loopback_cksum", 1664 SYSCTL_DESCR("Perform IP checksum on loopback"), 1665 NULL, 0, &ip_do_loopback_cksum, 0, 1666 CTL_NET, PF_INET, IPPROTO_IP, 1667 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1668 sysctl_createv(clog, 0, NULL, NULL, 1669 CTLFLAG_PERMANENT, 1670 CTLTYPE_STRUCT, "stats", 1671 SYSCTL_DESCR("IP statistics"), 1672 sysctl_net_inet_ip_stats, 0, NULL, 0, 1673 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1674 CTL_EOL); 1675#if NARP 1676 sysctl_createv(clog, 0, NULL, NULL, 1677 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1678 CTLTYPE_INT, "dad_count", 1679 SYSCTL_DESCR("Number of Duplicate Address Detection " 1680 "probes to send"), 1681 NULL, 0, &ip_dad_count, 0, 1682 CTL_NET, PF_INET, IPPROTO_IP, 1683 IPCTL_DAD_COUNT, CTL_EOL); 1684#endif 1685 1686 /* anonportalgo RFC6056 subtree */ 1687 const struct sysctlnode *portalgo_node; 1688 sysctl_createv(clog, 0, NULL, &portalgo_node, 1689 CTLFLAG_PERMANENT, 1690 CTLTYPE_NODE, "anonportalgo", 1691 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1692 NULL, 0, NULL, 0, 1693 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1694 sysctl_createv(clog, 0, &portalgo_node, NULL, 1695 CTLFLAG_PERMANENT, 1696 CTLTYPE_STRING, "available", 1697 SYSCTL_DESCR("available algorithms"), 1698 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1699 CTL_CREATE, CTL_EOL); 1700 sysctl_createv(clog, 0, &portalgo_node, NULL, 1701 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1702 CTLTYPE_STRING, "selected", 1703 SYSCTL_DESCR("selected algorithm"), 1704 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1705 CTL_CREATE, CTL_EOL); 1706 sysctl_createv(clog, 0, &portalgo_node, NULL, 1707 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1708 CTLTYPE_STRUCT, "reserve", 1709 SYSCTL_DESCR("bitmap of reserved ports"), 1710 sysctl_portalgo_reserve4, 0, NULL, 0, 1711 CTL_CREATE, CTL_EOL); 1712} 1713 1714void 1715ip_statinc(u_int stat) 1716{ 1717 1718 KASSERT(stat < IP_NSTATS); 1719 IP_STATINC(stat); 1720} 1721