ip_input.c revision 1.331
1/* $NetBSD: ip_input.c,v 1.331 2016/06/30 06:48:58 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 1.331 2016/06/30 06:48:58 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 bad; 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 bad; 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 bad; 422 } 423 hlen = ip->ip_hl << 2; 424 if (hlen < sizeof(struct ip)) { /* minimum header length */ 425 IP_STATINC(IP_STAT_BADHLEN); 426 goto bad; 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 bad; 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 bad; 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 goto badcsum; 460 461 case M_CSUM_IPv4: 462 /* Checksum was okay. */ 463 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 464 break; 465 466 default: 467 /* 468 * Must compute it ourselves. Maybe skip checksum on 469 * loopback interfaces. 470 */ 471 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 472 ip_do_loopback_cksum)) { 473 INET_CSUM_COUNTER_INCR(&ip_swcsum); 474 if (in_cksum(m, hlen) != 0) 475 goto badcsum; 476 } 477 break; 478 } 479 480 /* Retrieve the packet length. */ 481 len = ntohs(ip->ip_len); 482 483 /* 484 * Check for additional length bogosity 485 */ 486 if (len < hlen) { 487 IP_STATINC(IP_STAT_BADLEN); 488 goto bad; 489 } 490 491 /* 492 * Check that the amount of data in the buffers 493 * is as at least much as the IP header would have us expect. 494 * Trim mbufs if longer than we expect. 495 * Drop packet if shorter than we expect. 496 */ 497 if (m->m_pkthdr.len < len) { 498 IP_STATINC(IP_STAT_TOOSHORT); 499 goto bad; 500 } 501 if (m->m_pkthdr.len > len) { 502 if (m->m_len == m->m_pkthdr.len) { 503 m->m_len = len; 504 m->m_pkthdr.len = len; 505 } else 506 m_adj(m, len - m->m_pkthdr.len); 507 } 508 509 /* 510 * Assume that we can create a fast-forward IP flow entry 511 * based on this packet. 512 */ 513 m->m_flags |= M_CANFASTFWD; 514 515 /* 516 * Run through list of hooks for input packets. If there are any 517 * filters which require that additional packets in the flow are 518 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 519 * Note that filters must _never_ set this flag, as another filter 520 * in the list may have previously cleared it. 521 */ 522#if defined(IPSEC) 523 if (!ipsec_used || !ipsec_indone(m)) 524#else 525 if (1) 526#endif 527 { 528 struct in_addr odst = ip->ip_dst; 529 bool freed; 530 531 SOFTNET_LOCK(); 532 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 533 SOFTNET_UNLOCK(); 534 if (freed || m == NULL) { 535 goto out; 536 } 537 ip = mtod(m, struct ip *); 538 hlen = ip->ip_hl << 2; 539 540 /* 541 * XXX The setting of "srcrt" here is to prevent ip_forward() 542 * from generating ICMP redirects for packets that have 543 * been redirected by a hook back out on to the same LAN that 544 * they came from and is not an indication that the packet 545 * is being inffluenced by source routing options. This 546 * allows things like 547 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 548 * where tlp0 is both on the 1.1.1.0/24 network and is the 549 * default route for hosts on 1.1.1.0/24. Of course this 550 * also requires a "map tlp0 ..." to complete the story. 551 * One might argue whether or not this kind of network config. 552 * should be supported in this manner... 553 */ 554 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 555 } 556 557#ifdef ALTQ 558 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 559 if (altq_input) { 560 SOFTNET_LOCK(); 561 if ((*altq_input)(m, AF_INET) == 0) { 562 /* Packet dropped by traffic conditioner. */ 563 SOFTNET_UNLOCK(); 564 goto out; 565 } 566 SOFTNET_UNLOCK(); 567 } 568#endif 569 570 /* 571 * Process options and, if not destined for us, 572 * ship it on. ip_dooptions returns 1 when an 573 * error was detected (causing an icmp message 574 * to be sent and the original packet to be freed). 575 */ 576 ip_nhops = 0; /* for source routed packets */ 577 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 578 goto out; 579 580 /* 581 * Enable a consistency check between the destination address 582 * and the arrival interface for a unicast packet (the RFC 1122 583 * strong ES model) if IP forwarding is disabled and the packet 584 * is not locally generated. 585 * 586 * XXX - Checking also should be disabled if the destination 587 * address is ipnat'ed to a different interface. 588 * 589 * XXX - Checking is incompatible with IP aliases added 590 * to the loopback interface instead of the interface where 591 * the packets are received. 592 * 593 * XXX - We need to add a per ifaddr flag for this so that 594 * we get finer grain control. 595 */ 596 checkif = ip_checkinterface && (ipforwarding == 0) && 597 (ifp->if_flags & IFF_LOOPBACK) == 0; 598 599 /* 600 * Check our list of addresses, to see if the packet is for us. 601 * 602 * Traditional 4.4BSD did not consult IFF_UP at all. 603 * The behavior here is to treat addresses on !IFF_UP interface 604 * or IN_IFF_NOTREADY addresses as not mine. 605 */ 606 downmatch = 0; 607 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 608 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 609 if (ia->ia4_flags & IN_IFF_NOTREADY) 610 continue; 611 if (checkif && ia->ia_ifp != ifp) 612 continue; 613 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 614 break; 615 else 616 downmatch++; 617 } 618 } 619 if (ia != NULL) 620 goto ours; 621 if (ifp->if_flags & IFF_BROADCAST) { 622 IFADDR_FOREACH(ifa, ifp) { 623 if (ifa->ifa_addr->sa_family != AF_INET) 624 continue; 625 ia = ifatoia(ifa); 626 if (ia->ia4_flags & IN_IFF_NOTREADY) 627 continue; 628 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 629 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 630 /* 631 * Look for all-0's host part (old broadcast addr), 632 * either for subnet or net. 633 */ 634 ip->ip_dst.s_addr == ia->ia_subnet || 635 ip->ip_dst.s_addr == ia->ia_net) 636 goto ours; 637 /* 638 * An interface with IP address zero accepts 639 * all packets that arrive on that interface. 640 */ 641 if (in_nullhost(ia->ia_addr.sin_addr)) 642 goto ours; 643 } 644 } 645 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 646#ifdef MROUTING 647 extern struct socket *ip_mrouter; 648 649 if (ip_mrouter) { 650 /* 651 * If we are acting as a multicast router, all 652 * incoming multicast packets are passed to the 653 * kernel-level multicast forwarding function. 654 * The packet is returned (relatively) intact; if 655 * ip_mforward() returns a non-zero value, the packet 656 * must be discarded, else it may be accepted below. 657 * 658 * (The IP ident field is put in the same byte order 659 * as expected when ip_mforward() is called from 660 * ip_output().) 661 */ 662 SOFTNET_LOCK(); 663 if (ip_mforward(m, ifp) != 0) { 664 SOFTNET_UNLOCK(); 665 IP_STATINC(IP_STAT_CANTFORWARD); 666 goto bad; 667 } 668 SOFTNET_UNLOCK(); 669 670 /* 671 * The process-level routing demon needs to receive 672 * all multicast IGMP packets, whether or not this 673 * host belongs to their destination groups. 674 */ 675 if (ip->ip_p == IPPROTO_IGMP) { 676 goto ours; 677 } 678 IP_STATINC(IP_STAT_CANTFORWARD); 679 } 680#endif 681 /* 682 * See if we belong to the destination multicast group on the 683 * arrival interface. 684 */ 685 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 686 IP_STATINC(IP_STAT_CANTFORWARD); 687 goto bad; 688 } 689 goto ours; 690 } 691 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 692 in_nullhost(ip->ip_dst)) 693 goto ours; 694 695 /* 696 * Not for us; forward if possible and desirable. 697 */ 698 if (ipforwarding == 0) { 699 m_put_rcvif_psref(ifp, &psref); 700 IP_STATINC(IP_STAT_CANTFORWARD); 701 m_freem(m); 702 } else { 703 /* 704 * If ip_dst matched any of my address on !IFF_UP interface, 705 * and there's no IFF_UP interface that matches ip_dst, 706 * send icmp unreach. Forwarding it will result in in-kernel 707 * forwarding loop till TTL goes to 0. 708 */ 709 if (downmatch) { 710 m_put_rcvif_psref(ifp, &psref); 711 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 712 IP_STATINC(IP_STAT_CANTFORWARD); 713 return; 714 } 715#ifdef IPSEC 716 /* Perform IPsec, if any. */ 717 if (ipsec_used) { 718 SOFTNET_LOCK(); 719 if (ipsec4_input(m, IP_FORWARDING | 720 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 721 SOFTNET_UNLOCK(); 722 goto bad; 723 } 724 SOFTNET_UNLOCK(); 725 } 726#endif 727 ip_forward(m, srcrt, ifp); 728 m_put_rcvif_psref(ifp, &psref); 729 } 730 return; 731 732ours: 733 m_put_rcvif_psref(ifp, &psref); 734 ifp = NULL; 735 736 /* 737 * If offset or IP_MF are set, must reassemble. 738 */ 739 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 740 /* 741 * Pass to IP reassembly mechanism. 742 */ 743 if (ip_reass_packet(&m, ip) != 0) { 744 /* Failed; invalid fragment(s) or packet. */ 745 goto bad; 746 } 747 if (m == NULL) { 748 /* More fragments should come; silently return. */ 749 return; 750 } 751 /* 752 * Reassembly is done, we have the final packet. 753 * Updated cached data in local variable(s). 754 */ 755 ip = mtod(m, struct ip *); 756 hlen = ip->ip_hl << 2; 757 } 758 759#ifdef IPSEC 760 /* 761 * Enforce IPsec policy checking if we are seeing last header. 762 * Note that we do not visit this with protocols with PCB layer 763 * code - like UDP/TCP/raw IP. 764 */ 765 if (ipsec_used && 766 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 767 SOFTNET_LOCK(); 768 if (ipsec4_input(m, 0) != 0) { 769 SOFTNET_UNLOCK(); 770 goto bad; 771 } 772 SOFTNET_UNLOCK(); 773 } 774#endif 775 776 /* 777 * Switch out to protocol's input routine. 778 */ 779#if IFA_STATS 780 if (ia && ip) 781 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 782#endif 783 IP_STATINC(IP_STAT_DELIVERED); 784 785 const int off = hlen, nh = ip->ip_p; 786 787 SOFTNET_LOCK(); 788 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 789 SOFTNET_UNLOCK(); 790 return; 791bad: 792 m_put_rcvif_psref(ifp, &psref); 793 m_freem(m); 794 return; 795 796badcsum: 797 m_put_rcvif_psref(ifp, &psref); 798 IP_STATINC(IP_STAT_BADSUM); 799 m_freem(m); 800 return; 801out: 802 m_put_rcvif_psref(ifp, &psref); 803} 804 805/* 806 * IP timer processing. 807 */ 808void 809ip_slowtimo(void) 810{ 811 812 mutex_enter(softnet_lock); 813 KERNEL_LOCK(1, NULL); 814 815 ip_reass_slowtimo(); 816 817 KERNEL_UNLOCK_ONE(NULL); 818 mutex_exit(softnet_lock); 819} 820 821/* 822 * IP drain processing. 823 */ 824void 825ip_drain(void) 826{ 827 828 KERNEL_LOCK(1, NULL); 829 ip_reass_drain(); 830 KERNEL_UNLOCK_ONE(NULL); 831} 832 833/* 834 * ip_dooptions: perform option processing on a datagram, possibly discarding 835 * it if bad options are encountered, or forwarding it if source-routed. 836 * 837 * => Returns true if packet has been forwarded/freed. 838 * => Returns false if the packet should be processed further. 839 */ 840static bool 841ip_dooptions(struct mbuf *m) 842{ 843 struct ip *ip = mtod(m, struct ip *); 844 u_char *cp, *cp0; 845 struct ip_timestamp *ipt; 846 struct in_ifaddr *ia; 847 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 848 struct in_addr dst; 849 n_time ntime; 850 851 dst = ip->ip_dst; 852 cp = (u_char *)(ip + 1); 853 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 854 for (; cnt > 0; cnt -= optlen, cp += optlen) { 855 opt = cp[IPOPT_OPTVAL]; 856 if (opt == IPOPT_EOL) 857 break; 858 if (opt == IPOPT_NOP) 859 optlen = 1; 860 else { 861 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 862 code = &cp[IPOPT_OLEN] - (u_char *)ip; 863 goto bad; 864 } 865 optlen = cp[IPOPT_OLEN]; 866 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 867 code = &cp[IPOPT_OLEN] - (u_char *)ip; 868 goto bad; 869 } 870 } 871 switch (opt) { 872 873 default: 874 break; 875 876 /* 877 * Source routing with record. 878 * Find interface with current destination address. 879 * If none on this machine then drop if strictly routed, 880 * or do nothing if loosely routed. 881 * Record interface address and bring up next address 882 * component. If strictly routed make sure next 883 * address is on directly accessible net. 884 */ 885 case IPOPT_LSRR: 886 case IPOPT_SSRR: 887 if (ip_allowsrcrt == 0) { 888 type = ICMP_UNREACH; 889 code = ICMP_UNREACH_NET_PROHIB; 890 goto bad; 891 } 892 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 893 code = &cp[IPOPT_OLEN] - (u_char *)ip; 894 goto bad; 895 } 896 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 897 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 898 goto bad; 899 } 900 ipaddr.sin_addr = ip->ip_dst; 901 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 902 if (ia == 0) { 903 if (opt == IPOPT_SSRR) { 904 type = ICMP_UNREACH; 905 code = ICMP_UNREACH_SRCFAIL; 906 goto bad; 907 } 908 /* 909 * Loose routing, and not at next destination 910 * yet; nothing to do except forward. 911 */ 912 break; 913 } 914 off--; /* 0 origin */ 915 if ((off + sizeof(struct in_addr)) > optlen) { 916 /* 917 * End of source route. Should be for us. 918 */ 919 save_rte(cp, ip->ip_src); 920 break; 921 } 922 /* 923 * locate outgoing interface 924 */ 925 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 926 sizeof(ipaddr.sin_addr)); 927 if (opt == IPOPT_SSRR) 928 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 929 else 930 ia = ip_rtaddr(ipaddr.sin_addr); 931 if (ia == 0) { 932 type = ICMP_UNREACH; 933 code = ICMP_UNREACH_SRCFAIL; 934 goto bad; 935 } 936 ip->ip_dst = ipaddr.sin_addr; 937 bcopy((void *)&ia->ia_addr.sin_addr, 938 (void *)(cp + off), sizeof(struct in_addr)); 939 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 940 /* 941 * Let ip_intr's mcast routing check handle mcast pkts 942 */ 943 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 944 break; 945 946 case IPOPT_RR: 947 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 948 code = &cp[IPOPT_OLEN] - (u_char *)ip; 949 goto bad; 950 } 951 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 952 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 953 goto bad; 954 } 955 /* 956 * If no space remains, ignore. 957 */ 958 off--; /* 0 origin */ 959 if ((off + sizeof(struct in_addr)) > optlen) 960 break; 961 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 962 sizeof(ipaddr.sin_addr)); 963 /* 964 * locate outgoing interface; if we're the destination, 965 * use the incoming interface (should be same). 966 */ 967 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 968 == NULL && 969 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 970 type = ICMP_UNREACH; 971 code = ICMP_UNREACH_HOST; 972 goto bad; 973 } 974 bcopy((void *)&ia->ia_addr.sin_addr, 975 (void *)(cp + off), sizeof(struct in_addr)); 976 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 977 break; 978 979 case IPOPT_TS: 980 code = cp - (u_char *)ip; 981 ipt = (struct ip_timestamp *)cp; 982 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 983 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 984 goto bad; 985 } 986 if (ipt->ipt_ptr < 5) { 987 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 988 goto bad; 989 } 990 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 991 if (++ipt->ipt_oflw == 0) { 992 code = (u_char *)&ipt->ipt_ptr - 993 (u_char *)ip; 994 goto bad; 995 } 996 break; 997 } 998 cp0 = (cp + ipt->ipt_ptr - 1); 999 switch (ipt->ipt_flg) { 1000 1001 case IPOPT_TS_TSONLY: 1002 break; 1003 1004 case IPOPT_TS_TSANDADDR: { 1005 struct ifnet *rcvif; 1006 int s; 1007 1008 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1009 sizeof(struct in_addr) > ipt->ipt_len) { 1010 code = (u_char *)&ipt->ipt_ptr - 1011 (u_char *)ip; 1012 goto bad; 1013 } 1014 ipaddr.sin_addr = dst; 1015 rcvif = m_get_rcvif(m, &s); 1016 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1017 rcvif)); 1018 m_put_rcvif(rcvif, &s); 1019 if (ia == 0) 1020 continue; 1021 bcopy(&ia->ia_addr.sin_addr, 1022 cp0, sizeof(struct in_addr)); 1023 ipt->ipt_ptr += sizeof(struct in_addr); 1024 break; 1025 } 1026 1027 case IPOPT_TS_PRESPEC: 1028 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1029 sizeof(struct in_addr) > ipt->ipt_len) { 1030 code = (u_char *)&ipt->ipt_ptr - 1031 (u_char *)ip; 1032 goto bad; 1033 } 1034 memcpy(&ipaddr.sin_addr, cp0, 1035 sizeof(struct in_addr)); 1036 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1037 == NULL) 1038 continue; 1039 ipt->ipt_ptr += sizeof(struct in_addr); 1040 break; 1041 1042 default: 1043 /* XXX can't take &ipt->ipt_flg */ 1044 code = (u_char *)&ipt->ipt_ptr - 1045 (u_char *)ip + 1; 1046 goto bad; 1047 } 1048 ntime = iptime(); 1049 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1050 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1051 sizeof(n_time)); 1052 ipt->ipt_ptr += sizeof(n_time); 1053 } 1054 } 1055 if (forward) { 1056 struct ifnet *rcvif; 1057 struct psref psref; 1058 1059 if (ip_forwsrcrt == 0) { 1060 type = ICMP_UNREACH; 1061 code = ICMP_UNREACH_SRCFAIL; 1062 goto bad; 1063 } 1064 1065 rcvif = m_get_rcvif_psref(m, &psref); 1066 if (__predict_false(rcvif == NULL)) { 1067 type = ICMP_UNREACH; 1068 code = ICMP_UNREACH_HOST; 1069 goto bad; 1070 } 1071 ip_forward(m, 1, rcvif); 1072 m_put_rcvif_psref(rcvif, &psref); 1073 return true; 1074 } 1075 return false; 1076bad: 1077 icmp_error(m, type, code, 0, 0); 1078 IP_STATINC(IP_STAT_BADOPTIONS); 1079 return true; 1080} 1081 1082/* 1083 * ip_rtaddr: given address of next destination (final or next hop), 1084 * return internet address info of interface to be used to get there. 1085 */ 1086static struct in_ifaddr * 1087ip_rtaddr(struct in_addr dst) 1088{ 1089 struct rtentry *rt; 1090 union { 1091 struct sockaddr dst; 1092 struct sockaddr_in dst4; 1093 } u; 1094 1095 sockaddr_in_init(&u.dst4, &dst, 0); 1096 1097 SOFTNET_LOCK(); 1098 rt = rtcache_lookup(&ipforward_rt, &u.dst); 1099 SOFTNET_UNLOCK(); 1100 if (rt == NULL) 1101 return NULL; 1102 1103 return ifatoia(rt->rt_ifa); 1104} 1105 1106/* 1107 * save_rte: save incoming source route for use in replies, to be picked 1108 * up later by ip_srcroute if the receiver is interested. 1109 */ 1110static void 1111save_rte(u_char *option, struct in_addr dst) 1112{ 1113 unsigned olen; 1114 1115 olen = option[IPOPT_OLEN]; 1116 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1117 return; 1118 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen); 1119 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1120 ip_srcrt.dst = dst; 1121} 1122 1123/* 1124 * Retrieve incoming source route for use in replies, 1125 * in the same form used by setsockopt. 1126 * The first hop is placed before the options, will be removed later. 1127 */ 1128struct mbuf * 1129ip_srcroute(void) 1130{ 1131 struct in_addr *p, *q; 1132 struct mbuf *m; 1133 1134 if (ip_nhops == 0) 1135 return NULL; 1136 m = m_get(M_DONTWAIT, MT_SOOPTS); 1137 if (m == 0) 1138 return NULL; 1139 1140 MCLAIM(m, &inetdomain.dom_mowner); 1141#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1142 1143 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1144 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1145 OPTSIZ; 1146 1147 /* 1148 * First save first hop for return route 1149 */ 1150 p = &ip_srcrt.route[ip_nhops - 1]; 1151 *(mtod(m, struct in_addr *)) = *p--; 1152 1153 /* 1154 * Copy option fields and padding (nop) to mbuf. 1155 */ 1156 ip_srcrt.nop = IPOPT_NOP; 1157 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1158 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1159 OPTSIZ); 1160 q = (struct in_addr *)(mtod(m, char *) + 1161 sizeof(struct in_addr) + OPTSIZ); 1162#undef OPTSIZ 1163 /* 1164 * Record return path as an IP source route, 1165 * reversing the path (pointers are now aligned). 1166 */ 1167 while (p >= ip_srcrt.route) { 1168 *q++ = *p--; 1169 } 1170 /* 1171 * Last hop goes to final destination. 1172 */ 1173 *q = ip_srcrt.dst; 1174 return (m); 1175} 1176 1177const int inetctlerrmap[PRC_NCMDS] = { 1178 [PRC_MSGSIZE] = EMSGSIZE, 1179 [PRC_HOSTDEAD] = EHOSTDOWN, 1180 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1181 [PRC_UNREACH_NET] = EHOSTUNREACH, 1182 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1183 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1184 [PRC_UNREACH_PORT] = ECONNREFUSED, 1185 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1186 [PRC_PARAMPROB] = ENOPROTOOPT, 1187}; 1188 1189void 1190ip_fasttimo(void) 1191{ 1192 if (ip_drainwanted) { 1193 ip_drain(); 1194 ip_drainwanted = 0; 1195 } 1196} 1197 1198void 1199ip_drainstub(void) 1200{ 1201 ip_drainwanted = 1; 1202} 1203 1204/* 1205 * Forward a packet. If some error occurs return the sender 1206 * an icmp packet. Note we can't always generate a meaningful 1207 * icmp message because icmp doesn't have a large enough repertoire 1208 * of codes and types. 1209 * 1210 * If not forwarding, just drop the packet. This could be confusing 1211 * if ipforwarding was zero but some routing protocol was advancing 1212 * us as a gateway to somewhere. However, we must let the routing 1213 * protocol deal with that. 1214 * 1215 * The srcrt parameter indicates whether the packet is being forwarded 1216 * via a source route. 1217 */ 1218static void 1219ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1220{ 1221 struct ip *ip = mtod(m, struct ip *); 1222 struct rtentry *rt; 1223 int error, type = 0, code = 0, destmtu = 0; 1224 struct mbuf *mcopy; 1225 n_long dest; 1226 union { 1227 struct sockaddr dst; 1228 struct sockaddr_in dst4; 1229 } u; 1230 uint64_t *ips; 1231 1232 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1233 "interrupt handler; synchronization assumptions violated"); 1234 1235 /* 1236 * We are now in the output path. 1237 */ 1238 MCLAIM(m, &ip_tx_mowner); 1239 1240 /* 1241 * Clear any in-bound checksum flags for this packet. 1242 */ 1243 m->m_pkthdr.csum_flags = 0; 1244 1245 dest = 0; 1246 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1247 IP_STATINC(IP_STAT_CANTFORWARD); 1248 m_freem(m); 1249 return; 1250 } 1251 1252 SOFTNET_LOCK(); 1253 1254 if (ip->ip_ttl <= IPTTLDEC) { 1255 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1256 SOFTNET_UNLOCK(); 1257 return; 1258 } 1259 1260 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1261 1262 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1263 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1264 SOFTNET_UNLOCK(); 1265 return; 1266 } 1267 1268 /* 1269 * Save at most 68 bytes of the packet in case 1270 * we need to generate an ICMP message to the src. 1271 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1272 */ 1273 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1274 if (mcopy) 1275 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1276 1277 ip->ip_ttl -= IPTTLDEC; 1278 1279 /* 1280 * If forwarding packet using same interface that it came in on, 1281 * perhaps should send a redirect to sender to shortcut a hop. 1282 * Only send redirect if source is sending directly to us, 1283 * and if packet was not source routed (or has any options). 1284 * Also, don't send redirect if forwarding using a default route 1285 * or a route modified by a redirect. 1286 */ 1287 if (rt->rt_ifp == rcvif && 1288 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1289 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1290 ipsendredirects && !srcrt) { 1291 if (rt->rt_ifa && 1292 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1293 ifatoia(rt->rt_ifa)->ia_subnet) { 1294 if (rt->rt_flags & RTF_GATEWAY) 1295 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1296 else 1297 dest = ip->ip_dst.s_addr; 1298 /* 1299 * Router requirements says to only send host 1300 * redirects. 1301 */ 1302 type = ICMP_REDIRECT; 1303 code = ICMP_REDIRECT_HOST; 1304 } 1305 } 1306 1307 error = ip_output(m, NULL, &ipforward_rt, 1308 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1309 NULL, NULL); 1310 1311 if (error) { 1312 IP_STATINC(IP_STAT_CANTFORWARD); 1313 goto error; 1314 } 1315 1316 ips = IP_STAT_GETREF(); 1317 ips[IP_STAT_FORWARD]++; 1318 1319 if (type) { 1320 ips[IP_STAT_REDIRECTSENT]++; 1321 IP_STAT_PUTREF(); 1322 goto redirect; 1323 } 1324 1325 IP_STAT_PUTREF(); 1326 if (mcopy) { 1327#ifdef GATEWAY 1328 if (mcopy->m_flags & M_CANFASTFWD) 1329 ipflow_create(&ipforward_rt, mcopy); 1330#endif 1331 m_freem(mcopy); 1332 } 1333 1334 SOFTNET_UNLOCK(); 1335 return; 1336 1337redirect: 1338error: 1339 if (mcopy == NULL) { 1340 SOFTNET_UNLOCK(); 1341 return; 1342 } 1343 1344 switch (error) { 1345 1346 case 0: /* forwarded, but need redirect */ 1347 /* type, code set above */ 1348 break; 1349 1350 case ENETUNREACH: /* shouldn't happen, checked above */ 1351 case EHOSTUNREACH: 1352 case ENETDOWN: 1353 case EHOSTDOWN: 1354 default: 1355 type = ICMP_UNREACH; 1356 code = ICMP_UNREACH_HOST; 1357 break; 1358 1359 case EMSGSIZE: 1360 type = ICMP_UNREACH; 1361 code = ICMP_UNREACH_NEEDFRAG; 1362 1363 if ((rt = rtcache_validate(&ipforward_rt)) != NULL) 1364 destmtu = rt->rt_ifp->if_mtu; 1365#ifdef IPSEC 1366 if (ipsec_used) 1367 (void)ipsec4_forward(mcopy, &destmtu); 1368#endif 1369 IP_STATINC(IP_STAT_CANTFRAG); 1370 break; 1371 1372 case ENOBUFS: 1373 /* 1374 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1375 * Requirements for IP Version 4 Routers. Source quench can 1376 * big problem under DoS attacks or if the underlying 1377 * interface is rate-limited. 1378 */ 1379 if (mcopy) 1380 m_freem(mcopy); 1381 SOFTNET_UNLOCK(); 1382 return; 1383 } 1384 icmp_error(mcopy, type, code, dest, destmtu); 1385 SOFTNET_UNLOCK(); 1386} 1387 1388void 1389ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1390 struct mbuf *m) 1391{ 1392 struct socket *so = inp->inp_socket; 1393 ifnet_t *ifp; 1394 int inpflags = inp->inp_flags; 1395 struct psref psref; 1396 1397 ifp = m_get_rcvif_psref(m, &psref); 1398 if (__predict_false(ifp == NULL)) 1399 return; /* XXX should report error? */ 1400 1401 if (so->so_options & SO_TIMESTAMP 1402#ifdef SO_OTIMESTAMP 1403 || so->so_options & SO_OTIMESTAMP 1404#endif 1405 ) { 1406 struct timeval tv; 1407 1408 microtime(&tv); 1409#ifdef SO_OTIMESTAMP 1410 if (so->so_options & SO_OTIMESTAMP) { 1411 struct timeval50 tv50; 1412 timeval_to_timeval50(&tv, &tv50); 1413 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50), 1414 SCM_OTIMESTAMP, SOL_SOCKET); 1415 } else 1416#endif 1417 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 1418 SCM_TIMESTAMP, SOL_SOCKET); 1419 if (*mp) 1420 mp = &(*mp)->m_next; 1421 } 1422 if (inpflags & INP_RECVDSTADDR) { 1423 *mp = sbcreatecontrol((void *) &ip->ip_dst, 1424 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1425 if (*mp) 1426 mp = &(*mp)->m_next; 1427 } 1428 if (inpflags & INP_RECVPKTINFO) { 1429 struct in_pktinfo ipi; 1430 ipi.ipi_addr = ip->ip_src; 1431 ipi.ipi_ifindex = ifp->if_index; 1432 *mp = sbcreatecontrol((void *) &ipi, 1433 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP); 1434 if (*mp) 1435 mp = &(*mp)->m_next; 1436 } 1437 if (inpflags & INP_PKTINFO) { 1438 struct in_pktinfo ipi; 1439 ipi.ipi_addr = ip->ip_dst; 1440 ipi.ipi_ifindex = ifp->if_index; 1441 *mp = sbcreatecontrol((void *) &ipi, 1442 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1443 if (*mp) 1444 mp = &(*mp)->m_next; 1445 } 1446 if (inpflags & INP_RECVIF) { 1447 struct sockaddr_dl sdl; 1448 1449 sockaddr_dl_init(&sdl, sizeof(sdl), ifp ? 1450 ifp->if_index : 0, 0, NULL, 0, NULL, 0); 1451 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1452 if (*mp) 1453 mp = &(*mp)->m_next; 1454 } 1455 if (inpflags & INP_RECVTTL) { 1456 *mp = sbcreatecontrol((void *) &ip->ip_ttl, 1457 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1458 if (*mp) 1459 mp = &(*mp)->m_next; 1460 } 1461 m_put_rcvif_psref(ifp, &psref); 1462} 1463 1464/* 1465 * sysctl helper routine for net.inet.ip.forwsrcrt. 1466 */ 1467static int 1468sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1469{ 1470 int error, tmp; 1471 struct sysctlnode node; 1472 1473 node = *rnode; 1474 tmp = ip_forwsrcrt; 1475 node.sysctl_data = &tmp; 1476 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1477 if (error || newp == NULL) 1478 return (error); 1479 1480 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1481 0, NULL, NULL, NULL); 1482 if (error) 1483 return (error); 1484 1485 ip_forwsrcrt = tmp; 1486 1487 return (0); 1488} 1489 1490/* 1491 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1492 * range of the new value and tweaks timers if it changes. 1493 */ 1494static int 1495sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1496{ 1497 int error, tmp; 1498 struct sysctlnode node; 1499 1500 node = *rnode; 1501 tmp = ip_mtudisc_timeout; 1502 node.sysctl_data = &tmp; 1503 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1504 if (error || newp == NULL) 1505 return (error); 1506 if (tmp < 0) 1507 return (EINVAL); 1508 1509 mutex_enter(softnet_lock); 1510 1511 ip_mtudisc_timeout = tmp; 1512 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1513 1514 mutex_exit(softnet_lock); 1515 1516 return (0); 1517} 1518 1519static int 1520sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1521{ 1522 1523 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1524} 1525 1526static void 1527sysctl_net_inet_ip_setup(struct sysctllog **clog) 1528{ 1529 sysctl_createv(clog, 0, NULL, NULL, 1530 CTLFLAG_PERMANENT, 1531 CTLTYPE_NODE, "inet", 1532 SYSCTL_DESCR("PF_INET related settings"), 1533 NULL, 0, NULL, 0, 1534 CTL_NET, PF_INET, CTL_EOL); 1535 sysctl_createv(clog, 0, NULL, NULL, 1536 CTLFLAG_PERMANENT, 1537 CTLTYPE_NODE, "ip", 1538 SYSCTL_DESCR("IPv4 related settings"), 1539 NULL, 0, NULL, 0, 1540 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1541 1542 sysctl_createv(clog, 0, NULL, NULL, 1543 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1544 CTLTYPE_INT, "forwarding", 1545 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1546 NULL, 0, &ipforwarding, 0, 1547 CTL_NET, PF_INET, IPPROTO_IP, 1548 IPCTL_FORWARDING, CTL_EOL); 1549 sysctl_createv(clog, 0, NULL, NULL, 1550 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1551 CTLTYPE_INT, "redirect", 1552 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1553 NULL, 0, &ipsendredirects, 0, 1554 CTL_NET, PF_INET, IPPROTO_IP, 1555 IPCTL_SENDREDIRECTS, CTL_EOL); 1556 sysctl_createv(clog, 0, NULL, NULL, 1557 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1558 CTLTYPE_INT, "ttl", 1559 SYSCTL_DESCR("Default TTL for an INET datagram"), 1560 NULL, 0, &ip_defttl, 0, 1561 CTL_NET, PF_INET, IPPROTO_IP, 1562 IPCTL_DEFTTL, CTL_EOL); 1563#ifdef IPCTL_DEFMTU 1564 sysctl_createv(clog, 0, NULL, NULL, 1565 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1566 CTLTYPE_INT, "mtu", 1567 SYSCTL_DESCR("Default MTA for an INET route"), 1568 NULL, 0, &ip_mtu, 0, 1569 CTL_NET, PF_INET, IPPROTO_IP, 1570 IPCTL_DEFMTU, CTL_EOL); 1571#endif /* IPCTL_DEFMTU */ 1572 sysctl_createv(clog, 0, NULL, NULL, 1573 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1574 CTLTYPE_INT, "forwsrcrt", 1575 SYSCTL_DESCR("Enable forwarding of source-routed " 1576 "datagrams"), 1577 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1578 CTL_NET, PF_INET, IPPROTO_IP, 1579 IPCTL_FORWSRCRT, CTL_EOL); 1580 sysctl_createv(clog, 0, NULL, NULL, 1581 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1582 CTLTYPE_INT, "directed-broadcast", 1583 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1584 NULL, 0, &ip_directedbcast, 0, 1585 CTL_NET, PF_INET, IPPROTO_IP, 1586 IPCTL_DIRECTEDBCAST, CTL_EOL); 1587 sysctl_createv(clog, 0, NULL, NULL, 1588 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1589 CTLTYPE_INT, "allowsrcrt", 1590 SYSCTL_DESCR("Accept source-routed datagrams"), 1591 NULL, 0, &ip_allowsrcrt, 0, 1592 CTL_NET, PF_INET, IPPROTO_IP, 1593 IPCTL_ALLOWSRCRT, CTL_EOL); 1594 1595 sysctl_createv(clog, 0, NULL, NULL, 1596 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1597 CTLTYPE_INT, "mtudisc", 1598 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1599 NULL, 0, &ip_mtudisc, 0, 1600 CTL_NET, PF_INET, IPPROTO_IP, 1601 IPCTL_MTUDISC, CTL_EOL); 1602 sysctl_createv(clog, 0, NULL, NULL, 1603 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1604 CTLTYPE_INT, "anonportmin", 1605 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1606 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1607 CTL_NET, PF_INET, IPPROTO_IP, 1608 IPCTL_ANONPORTMIN, CTL_EOL); 1609 sysctl_createv(clog, 0, NULL, NULL, 1610 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1611 CTLTYPE_INT, "anonportmax", 1612 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1613 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1614 CTL_NET, PF_INET, IPPROTO_IP, 1615 IPCTL_ANONPORTMAX, CTL_EOL); 1616 sysctl_createv(clog, 0, NULL, NULL, 1617 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1618 CTLTYPE_INT, "mtudisctimeout", 1619 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1620 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1621 CTL_NET, PF_INET, IPPROTO_IP, 1622 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1623#ifndef IPNOPRIVPORTS 1624 sysctl_createv(clog, 0, NULL, NULL, 1625 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1626 CTLTYPE_INT, "lowportmin", 1627 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1628 "to assign"), 1629 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1630 CTL_NET, PF_INET, IPPROTO_IP, 1631 IPCTL_LOWPORTMIN, CTL_EOL); 1632 sysctl_createv(clog, 0, NULL, NULL, 1633 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1634 CTLTYPE_INT, "lowportmax", 1635 SYSCTL_DESCR("Highest privileged ephemeral port number " 1636 "to assign"), 1637 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1638 CTL_NET, PF_INET, IPPROTO_IP, 1639 IPCTL_LOWPORTMAX, CTL_EOL); 1640#endif /* IPNOPRIVPORTS */ 1641#if NGRE > 0 1642 sysctl_createv(clog, 0, NULL, NULL, 1643 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1644 CTLTYPE_INT, "grettl", 1645 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1646 NULL, 0, &ip_gre_ttl, 0, 1647 CTL_NET, PF_INET, IPPROTO_IP, 1648 IPCTL_GRE_TTL, CTL_EOL); 1649#endif /* NGRE */ 1650 sysctl_createv(clog, 0, NULL, NULL, 1651 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1652 CTLTYPE_INT, "checkinterface", 1653 SYSCTL_DESCR("Enable receive side of Strong ES model " 1654 "from RFC1122"), 1655 NULL, 0, &ip_checkinterface, 0, 1656 CTL_NET, PF_INET, IPPROTO_IP, 1657 IPCTL_CHECKINTERFACE, CTL_EOL); 1658 sysctl_createv(clog, 0, NULL, NULL, 1659 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1660 CTLTYPE_INT, "random_id", 1661 SYSCTL_DESCR("Assign random ip_id values"), 1662 NULL, 0, &ip_do_randomid, 0, 1663 CTL_NET, PF_INET, IPPROTO_IP, 1664 IPCTL_RANDOMID, CTL_EOL); 1665 sysctl_createv(clog, 0, NULL, NULL, 1666 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1667 CTLTYPE_INT, "do_loopback_cksum", 1668 SYSCTL_DESCR("Perform IP checksum on loopback"), 1669 NULL, 0, &ip_do_loopback_cksum, 0, 1670 CTL_NET, PF_INET, IPPROTO_IP, 1671 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1672 sysctl_createv(clog, 0, NULL, NULL, 1673 CTLFLAG_PERMANENT, 1674 CTLTYPE_STRUCT, "stats", 1675 SYSCTL_DESCR("IP statistics"), 1676 sysctl_net_inet_ip_stats, 0, NULL, 0, 1677 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1678 CTL_EOL); 1679#if NARP 1680 sysctl_createv(clog, 0, NULL, NULL, 1681 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1682 CTLTYPE_INT, "dad_count", 1683 SYSCTL_DESCR("Number of Duplicate Address Detection " 1684 "probes to send"), 1685 NULL, 0, &ip_dad_count, 0, 1686 CTL_NET, PF_INET, IPPROTO_IP, 1687 IPCTL_DAD_COUNT, CTL_EOL); 1688#endif 1689 1690 /* anonportalgo RFC6056 subtree */ 1691 const struct sysctlnode *portalgo_node; 1692 sysctl_createv(clog, 0, NULL, &portalgo_node, 1693 CTLFLAG_PERMANENT, 1694 CTLTYPE_NODE, "anonportalgo", 1695 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1696 NULL, 0, NULL, 0, 1697 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1698 sysctl_createv(clog, 0, &portalgo_node, NULL, 1699 CTLFLAG_PERMANENT, 1700 CTLTYPE_STRING, "available", 1701 SYSCTL_DESCR("available algorithms"), 1702 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1703 CTL_CREATE, CTL_EOL); 1704 sysctl_createv(clog, 0, &portalgo_node, NULL, 1705 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1706 CTLTYPE_STRING, "selected", 1707 SYSCTL_DESCR("selected algorithm"), 1708 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1709 CTL_CREATE, CTL_EOL); 1710 sysctl_createv(clog, 0, &portalgo_node, NULL, 1711 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1712 CTLTYPE_STRUCT, "reserve", 1713 SYSCTL_DESCR("bitmap of reserved ports"), 1714 sysctl_portalgo_reserve4, 0, NULL, 0, 1715 CTL_CREATE, CTL_EOL); 1716} 1717 1718void 1719ip_statinc(u_int stat) 1720{ 1721 1722 KASSERT(stat < IP_NSTATS); 1723 IP_STATINC(stat); 1724} 1725