ip_input.c revision 1.400
1/* $NetBSD: ip_input.c,v 1.400 2021/03/07 14:58:27 christos 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 1.400 2021/03/07 14:58:27 christos Exp $"); 95 96#ifdef _KERNEL_OPT 97#include "opt_inet.h" 98#include "opt_gateway.h" 99#include "opt_ipsec.h" 100#include "opt_mrouting.h" 101#include "opt_mbuftrace.h" 102#include "opt_inet_csum.h" 103#include "opt_net_mpsafe.h" 104#endif 105 106#include "arp.h" 107 108#include <sys/param.h> 109#include <sys/systm.h> 110#include <sys/cpu.h> 111#include <sys/mbuf.h> 112#include <sys/domain.h> 113#include <sys/protosw.h> 114#include <sys/socket.h> 115#include <sys/socketvar.h> 116#include <sys/errno.h> 117#include <sys/time.h> 118#include <sys/kernel.h> 119#include <sys/pool.h> 120#include <sys/sysctl.h> 121#include <sys/kauth.h> 122 123#include <net/if.h> 124#include <net/if_dl.h> 125#include <net/route.h> 126#include <net/pktqueue.h> 127#include <net/pfil.h> 128 129#include <netinet/in.h> 130#include <netinet/in_systm.h> 131#include <netinet/ip.h> 132#include <netinet/in_pcb.h> 133#include <netinet/in_proto.h> 134#include <netinet/in_var.h> 135#include <netinet/ip_var.h> 136#include <netinet/ip_private.h> 137#include <netinet/ip_icmp.h> 138/* just for gif_ttl */ 139#include <netinet/in_gif.h> 140#include "gif.h" 141#include <net/if_gre.h> 142#include "gre.h" 143 144#ifdef MROUTING 145#include <netinet/ip_mroute.h> 146#endif 147#include <netinet/portalgo.h> 148 149#ifdef IPSEC 150#include <netipsec/ipsec.h> 151#endif 152 153#ifndef IPFORWARDING 154#ifdef GATEWAY 155#define IPFORWARDING 1 /* forward IP packets not for us */ 156#else 157#define IPFORWARDING 0 /* don't forward IP packets not for us */ 158#endif 159#endif 160 161#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 162 163int ipforwarding = IPFORWARDING; 164int ipsendredirects = 1; 165int ip_defttl = IPDEFTTL; 166int ip_forwsrcrt = 0; 167int ip_directedbcast = 0; 168int ip_allowsrcrt = 0; 169int ip_mtudisc = 1; 170int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 171int ip_do_randomid = 1; 172 173/* 174 * XXX - Setting ip_checkinterface mostly implements the receive side of 175 * the Strong ES model described in RFC 1122, but since the routing table 176 * and transmit implementation do not implement the Strong ES model, 177 * setting this to 1 results in an odd hybrid. 178 * 179 * XXX - ip_checkinterface currently must be disabled if you use NAT 180 * to translate the destination address to another local interface. 181 * 182 * XXX - ip_checkinterface must be disabled if you add IP aliases 183 * to the loopback interface instead of the interface where the 184 * packets for those addresses are received. 185 */ 186static int ip_checkinterface __read_mostly = 0; 187 188struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 189 190pktqueue_t * ip_pktq __read_mostly; 191pfil_head_t * inet_pfil_hook __read_mostly; 192ipid_state_t * ip_ids __read_mostly; 193percpu_t * ipstat_percpu __read_mostly; 194 195static percpu_t *ipforward_rt_percpu __cacheline_aligned; 196 197uint16_t ip_id; 198 199#ifdef INET_CSUM_COUNTERS 200#include <sys/device.h> 201 202struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 203 NULL, "inet", "hwcsum bad"); 204struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 205 NULL, "inet", "hwcsum ok"); 206struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 207 NULL, "inet", "swcsum"); 208 209#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 210 211EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 212EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 213EVCNT_ATTACH_STATIC(ip_swcsum); 214 215#else 216 217#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 218 219#endif /* INET_CSUM_COUNTERS */ 220 221/* 222 * Used to save the IP options in case a protocol wants to respond 223 * to an incoming packet over the same route if the packet got here 224 * using IP source routing. This allows connection establishment and 225 * maintenance when the remote end is on a network that is not known 226 * to us. 227 */ 228struct ip_srcrt { 229 int isr_nhops; /* number of hops */ 230 struct in_addr isr_dst; /* final destination */ 231 char isr_nop; /* one NOP to align */ 232 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */ 233 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)]; 234}; 235 236static int ip_drainwanted; 237 238static void save_rte(struct mbuf *, u_char *, struct in_addr); 239 240#ifdef MBUFTRACE 241struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 242struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 243#endif 244 245static void ipintr(void *); 246static void ip_input(struct mbuf *, struct ifnet *); 247static void ip_forward(struct mbuf *, int, struct ifnet *); 248static bool ip_dooptions(struct mbuf *); 249static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *); 250static void sysctl_net_inet_ip_setup(struct sysctllog **); 251 252static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *, 253 int *); 254static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *, 255 struct ip *); 256 257#ifdef NET_MPSAFE 258#define SOFTNET_LOCK() mutex_enter(softnet_lock) 259#define SOFTNET_UNLOCK() mutex_exit(softnet_lock) 260#else 261#define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 262#define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 263#endif 264 265/* 266 * IP initialization: fill in IP protocol switch table. 267 * All protocols not implemented in kernel go to raw IP protocol handler. 268 */ 269void 270ip_init(void) 271{ 272 const struct protosw *pr; 273 274 in_init(); 275 sysctl_net_inet_ip_setup(NULL); 276 277 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 278 KASSERT(pr != NULL); 279 280 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 281 KASSERT(ip_pktq != NULL); 282 283 for (u_int i = 0; i < IPPROTO_MAX; i++) { 284 ip_protox[i] = pr - inetsw; 285 } 286 for (pr = inetdomain.dom_protosw; 287 pr < inetdomain.dom_protoswNPROTOSW; pr++) 288 if (pr->pr_domain->dom_family == PF_INET && 289 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 290 ip_protox[pr->pr_protocol] = pr - inetsw; 291 292 ip_reass_init(); 293 294 ip_ids = ip_id_init(); 295 ip_id = time_uptime & 0xfffff; 296 297#ifdef GATEWAY 298 ipflow_init(); 299#endif 300 301 /* Register our Packet Filter hook. */ 302 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 303 KASSERT(inet_pfil_hook != NULL); 304 305#ifdef MBUFTRACE 306 MOWNER_ATTACH(&ip_tx_mowner); 307 MOWNER_ATTACH(&ip_rx_mowner); 308#endif 309 310 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 311 ipforward_rt_percpu = rtcache_percpu_alloc(); 312 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 313} 314 315static struct in_ifaddr * 316ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 317{ 318 struct in_ifaddr *ia = NULL; 319 int checkif; 320 321 /* 322 * Enable a consistency check between the destination address 323 * and the arrival interface for a unicast packet (the RFC 1122 324 * strong ES model) if IP forwarding is disabled and the packet 325 * is not locally generated. 326 * 327 * XXX - We need to add a per ifaddr flag for this so that 328 * we get finer grain control. 329 */ 330 checkif = ip_checkinterface && (ipforwarding == 0) && 331 (ifp->if_flags & IFF_LOOPBACK) == 0; 332 333 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 334 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 335 if (ia->ia4_flags & IN_IFF_NOTREADY) 336 continue; 337 if (checkif && ia->ia_ifp != ifp) 338 continue; 339 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) { 340 (*downmatch)++; 341 continue; 342 } 343 if (ia->ia4_flags & IN_IFF_DETACHED && 344 (ifp->if_flags & IFF_LOOPBACK) == 0) 345 continue; 346 break; 347 } 348 } 349 350 return ia; 351} 352 353static struct in_ifaddr * 354ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 355{ 356 struct in_ifaddr *ia = NULL; 357 struct ifaddr *ifa; 358 359 IFADDR_READER_FOREACH(ifa, ifp) { 360 if (ifa->ifa_addr->sa_family != AF_INET) 361 continue; 362 ia = ifatoia(ifa); 363 if (ia->ia4_flags & IN_IFF_NOTREADY) 364 continue; 365 if (ia->ia4_flags & IN_IFF_DETACHED && 366 (ifp->if_flags & IFF_LOOPBACK) == 0) 367 continue; 368 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 369 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 370 /* 371 * Look for all-0's host part (old broadcast addr), 372 * either for subnet or net. 373 */ 374 ip->ip_dst.s_addr == ia->ia_subnet || 375 ip->ip_dst.s_addr == ia->ia_net) 376 goto matched; 377 /* 378 * An interface with IP address zero accepts 379 * all packets that arrive on that interface. 380 */ 381 if (in_nullhost(ia->ia_addr.sin_addr)) 382 goto matched; 383 } 384 ia = NULL; 385 386matched: 387 return ia; 388} 389 390/* 391 * IP software interrupt routine. 392 */ 393static void 394ipintr(void *arg __unused) 395{ 396 struct mbuf *m; 397 398 KASSERT(cpu_softintr_p()); 399 400 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 401 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 402 struct ifnet *ifp; 403 struct psref psref; 404 405 ifp = m_get_rcvif_psref(m, &psref); 406 if (__predict_false(ifp == NULL)) { 407 IP_STATINC(IP_STAT_IFDROP); 408 m_freem(m); 409 continue; 410 } 411 412 ip_input(m, ifp); 413 414 m_put_rcvif_psref(ifp, &psref); 415 } 416 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 417} 418 419/* 420 * IP input routine. Checksum and byte swap header. If fragmented 421 * try to reassemble. Process options. Pass to next level. 422 */ 423static void 424ip_input(struct mbuf *m, struct ifnet *ifp) 425{ 426 struct ip *ip = NULL; 427 struct in_ifaddr *ia = NULL; 428 int hlen = 0, len; 429 int downmatch; 430 int srcrt = 0; 431 int s; 432 433 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 434 "interrupt handler; synchronization assumptions violated"); 435 436 MCLAIM(m, &ip_rx_mowner); 437 KASSERT((m->m_flags & M_PKTHDR) != 0); 438 439 /* 440 * If no IP addresses have been set yet but the interfaces 441 * are receiving, can't do anything with incoming packets yet. 442 * Note: we pre-check without locks held. 443 */ 444 if (IN_ADDRLIST_READER_EMPTY()) { 445 IP_STATINC(IP_STAT_IFDROP); 446 goto out; 447 } 448 449 IP_STATINC(IP_STAT_TOTAL); 450 451 /* 452 * If the IP header is not aligned, slurp it up into a new 453 * mbuf with space for link headers, in the event we forward 454 * it. Otherwise, if it is aligned, make sure the entire 455 * base IP header is in the first mbuf of the chain. 456 */ 457 if (M_GET_ALIGNED_HDR(&m, struct ip, true) != 0) { 458 /* XXXJRT new stat, please */ 459 IP_STATINC(IP_STAT_TOOSMALL); 460 goto out; 461 } 462 ip = mtod(m, struct ip *); 463 if (ip->ip_v != IPVERSION) { 464 IP_STATINC(IP_STAT_BADVERS); 465 goto out; 466 } 467 hlen = ip->ip_hl << 2; 468 if (hlen < sizeof(struct ip)) { /* minimum header length */ 469 IP_STATINC(IP_STAT_BADHLEN); 470 goto out; 471 } 472 if (hlen > m->m_len) { 473 if ((m = m_pullup(m, hlen)) == NULL) { 474 IP_STATINC(IP_STAT_BADHLEN); 475 goto out; 476 } 477 ip = mtod(m, struct ip *); 478 } 479 480 /* 481 * RFC1122: packets with a multicast source address are 482 * not allowed. 483 */ 484 if (IN_MULTICAST(ip->ip_src.s_addr)) { 485 IP_STATINC(IP_STAT_BADADDR); 486 goto out; 487 } 488 489 /* 127/8 must not appear on wire - RFC1122 */ 490 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 491 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 492 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 493 IP_STATINC(IP_STAT_BADADDR); 494 goto out; 495 } 496 } 497 498 switch (m->m_pkthdr.csum_flags & 499 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) { 500 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 501 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 502 IP_STATINC(IP_STAT_BADSUM); 503 goto out; 504 505 case M_CSUM_IPv4: 506 /* Checksum was okay. */ 507 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 508 break; 509 510 default: 511 /* 512 * Must compute it ourselves. Maybe skip checksum on 513 * loopback interfaces. 514 */ 515 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 516 ip_do_loopback_cksum)) { 517 INET_CSUM_COUNTER_INCR(&ip_swcsum); 518 if (in_cksum(m, hlen) != 0) { 519 IP_STATINC(IP_STAT_BADSUM); 520 goto out; 521 } 522 } 523 break; 524 } 525 526 /* Retrieve the packet length. */ 527 len = ntohs(ip->ip_len); 528 529 /* 530 * Check for additional length bogosity 531 */ 532 if (len < hlen) { 533 IP_STATINC(IP_STAT_BADLEN); 534 goto out; 535 } 536 537 /* 538 * Check that the amount of data in the buffers is at least as much 539 * as the IP header would have us expect. Trim mbufs if longer than 540 * we expect. Drop packet if shorter than we expect. 541 */ 542 if (m->m_pkthdr.len < len) { 543 IP_STATINC(IP_STAT_TOOSHORT); 544 goto out; 545 } 546 if (m->m_pkthdr.len > len) { 547 if (m->m_len == m->m_pkthdr.len) { 548 m->m_len = len; 549 m->m_pkthdr.len = len; 550 } else 551 m_adj(m, len - m->m_pkthdr.len); 552 } 553 554 /* 555 * Assume that we can create a fast-forward IP flow entry 556 * based on this packet. 557 */ 558 m->m_flags |= M_CANFASTFWD; 559 560 /* 561 * Run through list of hooks for input packets. If there are any 562 * filters which require that additional packets in the flow are 563 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 564 * Note that filters must _never_ set this flag, as another filter 565 * in the list may have previously cleared it. 566 * 567 * Don't call hooks if the packet has already been processed by 568 * IPsec (encapsulated, tunnel mode). 569 */ 570#if defined(IPSEC) 571 if (!ipsec_used || !ipsec_skip_pfil(m)) 572#else 573 if (1) 574#endif 575 { 576 struct in_addr odst = ip->ip_dst; 577 bool freed; 578 579 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 580 if (freed || m == NULL) { 581 m = NULL; 582 IP_STATINC(IP_STAT_PFILDROP_IN); 583 goto out; 584 } 585 if (__predict_false(m->m_len < sizeof(struct ip))) { 586 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) { 587 IP_STATINC(IP_STAT_TOOSMALL); 588 goto out; 589 } 590 } 591 ip = mtod(m, struct ip *); 592 hlen = ip->ip_hl << 2; 593 if (hlen < sizeof(struct ip)) { /* minimum header length */ 594 IP_STATINC(IP_STAT_BADHLEN); 595 goto out; 596 } 597 if (hlen > m->m_len) { 598 if ((m = m_pullup(m, hlen)) == NULL) { 599 IP_STATINC(IP_STAT_BADHLEN); 600 goto out; 601 } 602 ip = mtod(m, struct ip *); 603 } 604 605 /* 606 * XXX The setting of "srcrt" here is to prevent ip_forward() 607 * from generating ICMP redirects for packets that have 608 * been redirected by a hook back out on to the same LAN that 609 * they came from and is not an indication that the packet 610 * is being influenced by source routing options. This 611 * allows things like 612 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 613 * where tlp0 is both on the 1.1.1.0/24 network and is the 614 * default route for hosts on 1.1.1.0/24. Of course this 615 * also requires a "map tlp0 ..." to complete the story. 616 * One might argue whether or not this kind of network config. 617 * should be supported in this manner... 618 */ 619 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 620 } 621 622#ifdef ALTQ 623 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 624 if (altq_input) { 625 SOFTNET_LOCK(); 626 if ((*altq_input)(m, AF_INET) == 0) { 627 /* Packet dropped by traffic conditioner. */ 628 SOFTNET_UNLOCK(); 629 m = NULL; 630 goto out; 631 } 632 SOFTNET_UNLOCK(); 633 } 634#endif 635 636 /* 637 * Process options and, if not destined for us, 638 * ship it on. ip_dooptions returns 1 when an 639 * error was detected (causing an icmp message 640 * to be sent and the original packet to be freed). 641 */ 642 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 643 m = NULL; 644 goto out; 645 } 646 647 /* 648 * Check our list of addresses, to see if the packet is for us. 649 * 650 * Traditional 4.4BSD did not consult IFF_UP at all. 651 * The behavior here is to treat addresses on !IFF_UP interface 652 * or IN_IFF_NOTREADY addresses as not mine. 653 */ 654 downmatch = 0; 655 s = pserialize_read_enter(); 656 ia = ip_match_our_address(ifp, ip, &downmatch); 657 if (ia != NULL) { 658 pserialize_read_exit(s); 659 goto ours; 660 } 661 662 if (ifp->if_flags & IFF_BROADCAST) { 663 ia = ip_match_our_address_broadcast(ifp, ip); 664 if (ia != NULL) { 665 pserialize_read_exit(s); 666 goto ours; 667 } 668 } 669 pserialize_read_exit(s); 670 671 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 672#ifdef MROUTING 673 extern struct socket *ip_mrouter; 674 675 if (ip_mrouter) { 676 /* 677 * If we are acting as a multicast router, all 678 * incoming multicast packets are passed to the 679 * kernel-level multicast forwarding function. 680 * The packet is returned (relatively) intact; if 681 * ip_mforward() returns a non-zero value, the packet 682 * must be discarded, else it may be accepted below. 683 * 684 * (The IP ident field is put in the same byte order 685 * as expected when ip_mforward() is called from 686 * ip_output().) 687 */ 688 SOFTNET_LOCK(); 689 if (ip_mforward(m, ifp) != 0) { 690 SOFTNET_UNLOCK(); 691 IP_STATINC(IP_STAT_CANTFORWARD); 692 goto out; 693 } 694 SOFTNET_UNLOCK(); 695 696 /* 697 * The process-level routing demon needs to receive 698 * all multicast IGMP packets, whether or not this 699 * host belongs to their destination groups. 700 */ 701 if (ip->ip_p == IPPROTO_IGMP) { 702 goto ours; 703 } 704 IP_STATINC(IP_STAT_CANTFORWARD); 705 } 706#endif 707 /* 708 * See if we belong to the destination multicast group on the 709 * arrival interface. 710 */ 711 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 712 IP_STATINC(IP_STAT_CANTFORWARD); 713 goto out; 714 } 715 goto ours; 716 } 717 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 718 in_nullhost(ip->ip_dst)) 719 goto ours; 720 721 /* 722 * Not for us; forward if possible and desirable. 723 */ 724 if (ipforwarding == 0) { 725 IP_STATINC(IP_STAT_CANTFORWARD); 726 m_freem(m); 727 } else { 728 /* 729 * If ip_dst matched any of my address on !IFF_UP interface, 730 * and there's no IFF_UP interface that matches ip_dst, 731 * send icmp unreach. Forwarding it will result in in-kernel 732 * forwarding loop till TTL goes to 0. 733 */ 734 if (downmatch) { 735 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 736 IP_STATINC(IP_STAT_CANTFORWARD); 737 return; 738 } 739#ifdef IPSEC 740 /* Check the security policy (SP) for the packet */ 741 if (ipsec_used) { 742 if (ipsec_ip_input_checkpolicy(m, true) != 0) { 743 IP_STATINC(IP_STAT_IPSECDROP_IN); 744 goto out; 745 } 746 } 747#endif 748 ip_forward(m, srcrt, ifp); 749 } 750 return; 751 752ours: 753 /* 754 * If offset or IP_MF are set, must reassemble. 755 */ 756 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 757 /* 758 * Pass to IP reassembly mechanism. 759 */ 760 if (ip_reass_packet(&m) != 0) { 761 /* Failed; invalid fragment(s) or packet. */ 762 goto out; 763 } 764 if (m == NULL) { 765 /* More fragments should come; silently return. */ 766 goto out; 767 } 768 /* 769 * Reassembly is done, we have the final packet. 770 * Update cached data in local variable(s). 771 */ 772 ip = mtod(m, struct ip *); 773 hlen = ip->ip_hl << 2; 774 } 775 776 M_VERIFY_PACKET(m); 777 778#ifdef IPSEC 779 /* 780 * Enforce IPsec policy checking if we are seeing last header. 781 * Note that we do not visit this with protocols with PCB layer 782 * code - like UDP/TCP/raw IP. 783 */ 784 if (ipsec_used && 785 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 786 if (ipsec_ip_input_checkpolicy(m, false) != 0) { 787 IP_STATINC(IP_STAT_IPSECDROP_IN); 788 goto out; 789 } 790 } 791#endif 792 793 /* 794 * Switch out to protocol's input routine. 795 */ 796#if IFA_STATS 797 if (ia) { 798 struct in_ifaddr *_ia; 799 /* 800 * Keep a reference from ip_match_our_address with psref 801 * is expensive, so explore ia here again. 802 */ 803 s = pserialize_read_enter(); 804 _ia = in_get_ia(ip->ip_dst); 805 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 806 pserialize_read_exit(s); 807 } 808#endif 809 IP_STATINC(IP_STAT_DELIVERED); 810 811 const int off = hlen, nh = ip->ip_p; 812 813 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 814 return; 815 816out: 817 if (m != NULL) 818 m_freem(m); 819} 820 821/* 822 * IP timer processing. 823 */ 824void 825ip_slowtimo(void) 826{ 827 828 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 829 830 ip_reass_slowtimo(); 831 832 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 833} 834 835/* 836 * IP drain processing. 837 */ 838void 839ip_drain(void) 840{ 841 842 KERNEL_LOCK(1, NULL); 843 ip_reass_drain(); 844 KERNEL_UNLOCK_ONE(NULL); 845} 846 847/* 848 * ip_dooptions: perform option processing on a datagram, possibly discarding 849 * it if bad options are encountered, or forwarding it if source-routed. 850 * 851 * => Returns true if packet has been forwarded/freed. 852 * => Returns false if the packet should be processed further. 853 */ 854static bool 855ip_dooptions(struct mbuf *m) 856{ 857 struct ip *ip = mtod(m, struct ip *); 858 u_char *cp, *cp0; 859 struct ip_timestamp *ipt; 860 struct in_ifaddr *ia; 861 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 862 int srr_present, rr_present, ts_present; 863 struct in_addr dst; 864 n_time ntime; 865 struct ifaddr *ifa = NULL; 866 int s; 867 868 srr_present = 0; 869 rr_present = 0; 870 ts_present = 0; 871 872 dst = ip->ip_dst; 873 cp = (u_char *)(ip + 1); 874 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 875 for (; cnt > 0; cnt -= optlen, cp += optlen) { 876 opt = cp[IPOPT_OPTVAL]; 877 if (opt == IPOPT_EOL) 878 break; 879 if (opt == IPOPT_NOP) 880 optlen = 1; 881 else { 882 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 883 code = &cp[IPOPT_OLEN] - (u_char *)ip; 884 goto bad; 885 } 886 optlen = cp[IPOPT_OLEN]; 887 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 888 code = &cp[IPOPT_OLEN] - (u_char *)ip; 889 goto bad; 890 } 891 } 892 switch (opt) { 893 894 default: 895 break; 896 897 /* 898 * Source routing with record. 899 * Find interface with current destination address. 900 * If none on this machine then drop if strictly routed, 901 * or do nothing if loosely routed. 902 * Record interface address and bring up next address 903 * component. If strictly routed make sure next 904 * address is on directly accessible net. 905 */ 906 case IPOPT_LSRR: 907 case IPOPT_SSRR: { 908 struct psref psref; 909 struct sockaddr_in ipaddr = { 910 .sin_len = sizeof(ipaddr), 911 .sin_family = AF_INET, 912 }; 913 914 if (ip_allowsrcrt == 0) { 915 type = ICMP_UNREACH; 916 code = ICMP_UNREACH_NET_PROHIB; 917 goto bad; 918 } 919 if (srr_present++) { 920 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 921 goto bad; 922 } 923 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 924 code = &cp[IPOPT_OLEN] - (u_char *)ip; 925 goto bad; 926 } 927 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 928 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 929 goto bad; 930 } 931 ipaddr.sin_addr = ip->ip_dst; 932 933 s = pserialize_read_enter(); 934 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 935 if (ifa == NULL) { 936 pserialize_read_exit(s); 937 if (opt == IPOPT_SSRR) { 938 type = ICMP_UNREACH; 939 code = ICMP_UNREACH_SRCFAIL; 940 goto bad; 941 } 942 /* 943 * Loose routing, and not at next destination 944 * yet; nothing to do except forward. 945 */ 946 break; 947 } 948 pserialize_read_exit(s); 949 950 off--; /* 0 origin */ 951 if ((off + sizeof(struct in_addr)) > optlen) { 952 /* 953 * End of source route. Should be for us. 954 */ 955 save_rte(m, cp, ip->ip_src); 956 break; 957 } 958 /* 959 * locate outgoing interface 960 */ 961 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 962 sizeof(ipaddr.sin_addr)); 963 if (opt == IPOPT_SSRR) { 964 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 965 &psref); 966 if (ifa != NULL) 967 ia = ifatoia(ifa); 968 else 969 ia = NULL; 970 } else { 971 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 972 } 973 if (ia == NULL) { 974 type = ICMP_UNREACH; 975 code = ICMP_UNREACH_SRCFAIL; 976 goto bad; 977 } 978 ip->ip_dst = ipaddr.sin_addr; 979 memcpy(cp + off, &ia->ia_addr.sin_addr, 980 sizeof(struct in_addr)); 981 ia4_release(ia, &psref); 982 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 983 /* 984 * Let ip_intr's mcast routing check handle mcast pkts 985 */ 986 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 987 break; 988 } 989 990 case IPOPT_RR: { 991 struct psref psref; 992 struct sockaddr_in ipaddr = { 993 .sin_len = sizeof(ipaddr), 994 .sin_family = AF_INET, 995 }; 996 997 if (rr_present++) { 998 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 999 goto bad; 1000 } 1001 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1002 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1003 goto bad; 1004 } 1005 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1006 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1007 goto bad; 1008 } 1009 /* 1010 * If no space remains, ignore. 1011 */ 1012 off--; /* 0 origin */ 1013 if ((off + sizeof(struct in_addr)) > optlen) 1014 break; 1015 memcpy((void *)&ipaddr.sin_addr, (void *)&ip->ip_dst, 1016 sizeof(ipaddr.sin_addr)); 1017 /* 1018 * locate outgoing interface; if we're the destination, 1019 * use the incoming interface (should be same). 1020 */ 1021 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1022 if (ifa == NULL) { 1023 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1024 if (ia == NULL) { 1025 type = ICMP_UNREACH; 1026 code = ICMP_UNREACH_HOST; 1027 goto bad; 1028 } 1029 } else { 1030 ia = ifatoia(ifa); 1031 } 1032 memcpy(cp + off, &ia->ia_addr.sin_addr, 1033 sizeof(struct in_addr)); 1034 ia4_release(ia, &psref); 1035 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1036 break; 1037 } 1038 1039 case IPOPT_TS: 1040 code = cp - (u_char *)ip; 1041 ipt = (struct ip_timestamp *)cp; 1042 if (ts_present++) { 1043 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1044 goto bad; 1045 } 1046 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1047 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1048 goto bad; 1049 } 1050 if (ipt->ipt_ptr < 5) { 1051 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1052 goto bad; 1053 } 1054 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1055 if (++ipt->ipt_oflw == 0) { 1056 code = (u_char *)&ipt->ipt_ptr - 1057 (u_char *)ip; 1058 goto bad; 1059 } 1060 break; 1061 } 1062 cp0 = (cp + ipt->ipt_ptr - 1); 1063 switch (ipt->ipt_flg) { 1064 1065 case IPOPT_TS_TSONLY: 1066 break; 1067 1068 case IPOPT_TS_TSANDADDR: { 1069 struct ifnet *rcvif; 1070 int _s, _ss; 1071 struct sockaddr_in ipaddr = { 1072 .sin_len = sizeof(ipaddr), 1073 .sin_family = AF_INET, 1074 }; 1075 1076 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1077 sizeof(struct in_addr) > ipt->ipt_len) { 1078 code = (u_char *)&ipt->ipt_ptr - 1079 (u_char *)ip; 1080 goto bad; 1081 } 1082 ipaddr.sin_addr = dst; 1083 _ss = pserialize_read_enter(); 1084 rcvif = m_get_rcvif(m, &_s); 1085 if (__predict_true(rcvif != NULL)) { 1086 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1087 rcvif); 1088 } 1089 m_put_rcvif(rcvif, &_s); 1090 if (ifa == NULL) { 1091 pserialize_read_exit(_ss); 1092 break; 1093 } 1094 ia = ifatoia(ifa); 1095 memcpy(cp0, &ia->ia_addr.sin_addr, 1096 sizeof(struct in_addr)); 1097 pserialize_read_exit(_ss); 1098 ipt->ipt_ptr += sizeof(struct in_addr); 1099 break; 1100 } 1101 1102 case IPOPT_TS_PRESPEC: { 1103 struct sockaddr_in ipaddr = { 1104 .sin_len = sizeof(ipaddr), 1105 .sin_family = AF_INET, 1106 }; 1107 1108 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1109 sizeof(struct in_addr) > ipt->ipt_len) { 1110 code = (u_char *)&ipt->ipt_ptr - 1111 (u_char *)ip; 1112 goto bad; 1113 } 1114 memcpy(&ipaddr.sin_addr, cp0, 1115 sizeof(struct in_addr)); 1116 s = pserialize_read_enter(); 1117 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1118 if (ifa == NULL) { 1119 pserialize_read_exit(s); 1120 continue; 1121 } 1122 pserialize_read_exit(s); 1123 ipt->ipt_ptr += sizeof(struct in_addr); 1124 break; 1125 } 1126 1127 default: 1128 /* XXX can't take &ipt->ipt_flg */ 1129 code = (u_char *)&ipt->ipt_ptr - 1130 (u_char *)ip + 1; 1131 goto bad; 1132 } 1133 ntime = iptime(); 1134 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1135 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1136 sizeof(n_time)); 1137 ipt->ipt_ptr += sizeof(n_time); 1138 } 1139 } 1140 if (forward) { 1141 struct ifnet *rcvif; 1142 struct psref _psref; 1143 1144 if (ip_forwsrcrt == 0) { 1145 type = ICMP_UNREACH; 1146 code = ICMP_UNREACH_SRCFAIL; 1147 goto bad; 1148 } 1149 1150 rcvif = m_get_rcvif_psref(m, &_psref); 1151 if (__predict_false(rcvif == NULL)) { 1152 type = ICMP_UNREACH; 1153 code = ICMP_UNREACH_HOST; 1154 goto bad; 1155 } 1156 ip_forward(m, 1, rcvif); 1157 m_put_rcvif_psref(rcvif, &_psref); 1158 return true; 1159 } 1160 return false; 1161bad: 1162 icmp_error(m, type, code, 0, 0); 1163 IP_STATINC(IP_STAT_BADOPTIONS); 1164 return true; 1165} 1166 1167/* 1168 * ip_rtaddr: given address of next destination (final or next hop), 1169 * return internet address info of interface to be used to get there. 1170 */ 1171static struct in_ifaddr * 1172ip_rtaddr(struct in_addr dst, struct psref *psref) 1173{ 1174 struct rtentry *rt; 1175 union { 1176 struct sockaddr dst; 1177 struct sockaddr_in dst4; 1178 } u; 1179 struct route *ro; 1180 1181 sockaddr_in_init(&u.dst4, &dst, 0); 1182 1183 ro = rtcache_percpu_getref(ipforward_rt_percpu); 1184 rt = rtcache_lookup(ro, &u.dst); 1185 if (rt == NULL) { 1186 rtcache_percpu_putref(ipforward_rt_percpu); 1187 return NULL; 1188 } 1189 1190 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1191 rtcache_unref(rt, ro); 1192 rtcache_percpu_putref(ipforward_rt_percpu); 1193 1194 return ifatoia(rt->rt_ifa); 1195} 1196 1197/* 1198 * save_rte: save incoming source route for use in replies, to be picked 1199 * up later by ip_srcroute if the receiver is interested. 1200 */ 1201static void 1202save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1203{ 1204 struct ip_srcrt *isr; 1205 struct m_tag *mtag; 1206 unsigned olen; 1207 1208 olen = option[IPOPT_OLEN]; 1209 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1210 return; 1211 1212 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1213 if (mtag == NULL) 1214 return; 1215 isr = (struct ip_srcrt *)(mtag + 1); 1216 1217 memcpy(isr->isr_hdr, option, olen); 1218 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1219 isr->isr_dst = dst; 1220 m_tag_prepend(m, mtag); 1221} 1222 1223/* 1224 * Retrieve incoming source route for use in replies, 1225 * in the same form used by setsockopt. 1226 * The first hop is placed before the options, will be removed later. 1227 */ 1228struct mbuf * 1229ip_srcroute(struct mbuf *m0) 1230{ 1231 struct in_addr *p, *q; 1232 struct mbuf *m; 1233 struct ip_srcrt *isr; 1234 struct m_tag *mtag; 1235 1236 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE); 1237 if (mtag == NULL) 1238 return NULL; 1239 isr = (struct ip_srcrt *)(mtag + 1); 1240 1241 if (isr->isr_nhops == 0) 1242 return NULL; 1243 1244 m = m_get(M_DONTWAIT, MT_SOOPTS); 1245 if (m == NULL) 1246 return NULL; 1247 1248 MCLAIM(m, &inetdomain.dom_mowner); 1249#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1250 1251 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1252 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1253 1254 /* 1255 * First save first hop for return route 1256 */ 1257 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1258 *(mtod(m, struct in_addr *)) = *p--; 1259 1260 /* 1261 * Copy option fields and padding (nop) to mbuf. 1262 */ 1263 isr->isr_nop = IPOPT_NOP; 1264 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1265 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1266 OPTSIZ); 1267 q = (struct in_addr *)(mtod(m, char *) + 1268 sizeof(struct in_addr) + OPTSIZ); 1269#undef OPTSIZ 1270 /* 1271 * Record return path as an IP source route, 1272 * reversing the path (pointers are now aligned). 1273 */ 1274 while (p >= isr->isr_routes) { 1275 *q++ = *p--; 1276 } 1277 /* 1278 * Last hop goes to final destination. 1279 */ 1280 *q = isr->isr_dst; 1281 m_tag_delete(m0, mtag); 1282 return m; 1283} 1284 1285const int inetctlerrmap[PRC_NCMDS] = { 1286 [PRC_MSGSIZE] = EMSGSIZE, 1287 [PRC_HOSTDEAD] = EHOSTDOWN, 1288 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1289 [PRC_UNREACH_NET] = EHOSTUNREACH, 1290 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1291 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1292 [PRC_UNREACH_PORT] = ECONNREFUSED, 1293 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1294 [PRC_PARAMPROB] = ENOPROTOOPT, 1295}; 1296 1297void 1298ip_fasttimo(void) 1299{ 1300 if (ip_drainwanted) { 1301 ip_drain(); 1302 ip_drainwanted = 0; 1303 } 1304} 1305 1306void 1307ip_drainstub(void) 1308{ 1309 ip_drainwanted = 1; 1310} 1311 1312/* 1313 * Forward a packet. If some error occurs return the sender 1314 * an icmp packet. Note we can't always generate a meaningful 1315 * icmp message because icmp doesn't have a large enough repertoire 1316 * of codes and types. 1317 * 1318 * If not forwarding, just drop the packet. This could be confusing 1319 * if ipforwarding was zero but some routing protocol was advancing 1320 * us as a gateway to somewhere. However, we must let the routing 1321 * protocol deal with that. 1322 * 1323 * The srcrt parameter indicates whether the packet is being forwarded 1324 * via a source route. 1325 */ 1326static void 1327ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1328{ 1329 struct ip *ip = mtod(m, struct ip *); 1330 struct rtentry *rt; 1331 int error, type = 0, code = 0, destmtu = 0; 1332 struct mbuf *mcopy; 1333 n_long dest; 1334 union { 1335 struct sockaddr dst; 1336 struct sockaddr_in dst4; 1337 } u; 1338 uint64_t *ips; 1339 struct route *ro; 1340 1341 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1342 "interrupt handler; synchronization assumptions violated"); 1343 1344 /* 1345 * We are now in the output path. 1346 */ 1347 MCLAIM(m, &ip_tx_mowner); 1348 1349 /* 1350 * Clear any in-bound checksum flags for this packet. 1351 */ 1352 m->m_pkthdr.csum_flags = 0; 1353 1354 dest = 0; 1355 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1356 IP_STATINC(IP_STAT_CANTFORWARD); 1357 m_freem(m); 1358 return; 1359 } 1360 1361 if (ip->ip_ttl <= IPTTLDEC) { 1362 IP_STATINC(IP_STAT_TIMXCEED); 1363 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1364 return; 1365 } 1366 1367 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1368 1369 ro = rtcache_percpu_getref(ipforward_rt_percpu); 1370 rt = rtcache_lookup(ro, &u.dst); 1371 if (rt == NULL) { 1372 rtcache_percpu_putref(ipforward_rt_percpu); 1373 IP_STATINC(IP_STAT_NOROUTE); 1374 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1375 return; 1376 } 1377 1378 /* 1379 * Save at most 68 bytes of the packet in case 1380 * we need to generate an ICMP message to the src. 1381 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1382 */ 1383 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1384 if (mcopy) 1385 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1386 1387 ip->ip_ttl -= IPTTLDEC; 1388 1389 /* 1390 * If forwarding packet using same interface that it came in on, 1391 * perhaps should send a redirect to sender to shortcut a hop. 1392 * Only send redirect if source is sending directly to us, 1393 * and if packet was not source routed (or has any options). 1394 * Also, don't send redirect if forwarding using a default route 1395 * or a route modified by a redirect. 1396 */ 1397 if (rt->rt_ifp == rcvif && 1398 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1399 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1400 ipsendredirects && !srcrt) { 1401 if ((ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1402 ifatoia(rt->rt_ifa)->ia_subnet) { 1403 if (rt->rt_flags & RTF_GATEWAY) 1404 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1405 else 1406 dest = ip->ip_dst.s_addr; 1407 /* 1408 * Router requirements says to only send host 1409 * redirects. 1410 */ 1411 type = ICMP_REDIRECT; 1412 code = ICMP_REDIRECT_HOST; 1413 } 1414 } 1415 rtcache_unref(rt, ro); 1416 1417 error = ip_output(m, NULL, ro, 1418 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1419 NULL, NULL); 1420 1421 if (error) { 1422 IP_STATINC(IP_STAT_CANTFORWARD); 1423 goto error; 1424 } 1425 1426 ips = IP_STAT_GETREF(); 1427 ips[IP_STAT_FORWARD]++; 1428 1429 if (type) { 1430 ips[IP_STAT_REDIRECTSENT]++; 1431 IP_STAT_PUTREF(); 1432 goto redirect; 1433 } 1434 1435 IP_STAT_PUTREF(); 1436 if (mcopy) { 1437#ifdef GATEWAY 1438 if (mcopy->m_flags & M_CANFASTFWD) 1439 ipflow_create(ro, mcopy); 1440#endif 1441 m_freem(mcopy); 1442 } 1443 1444 rtcache_percpu_putref(ipforward_rt_percpu); 1445 return; 1446 1447redirect: 1448error: 1449 if (mcopy == NULL) { 1450 rtcache_percpu_putref(ipforward_rt_percpu); 1451 return; 1452 } 1453 1454 switch (error) { 1455 1456 case 0: /* forwarded, but need redirect */ 1457 /* type, code set above */ 1458 break; 1459 1460 case ENETUNREACH: /* shouldn't happen, checked above */ 1461 case EHOSTUNREACH: 1462 case ENETDOWN: 1463 case EHOSTDOWN: 1464 default: 1465 type = ICMP_UNREACH; 1466 code = ICMP_UNREACH_HOST; 1467 break; 1468 1469 case EMSGSIZE: 1470 type = ICMP_UNREACH; 1471 code = ICMP_UNREACH_NEEDFRAG; 1472 1473 if ((rt = rtcache_validate(ro)) != NULL) { 1474 destmtu = rt->rt_ifp->if_mtu; 1475 rtcache_unref(rt, ro); 1476 } 1477#ifdef IPSEC 1478 if (ipsec_used) 1479 ipsec_mtu(mcopy, &destmtu); 1480#endif 1481 IP_STATINC(IP_STAT_CANTFRAG); 1482 break; 1483 1484 case ENOBUFS: 1485 /* 1486 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1487 * Requirements for IP Version 4 Routers. Source quench can 1488 * be a big problem under DoS attacks or if the underlying 1489 * interface is rate-limited. 1490 */ 1491 if (mcopy) 1492 m_freem(mcopy); 1493 rtcache_percpu_putref(ipforward_rt_percpu); 1494 return; 1495 } 1496 icmp_error(mcopy, type, code, dest, destmtu); 1497 rtcache_percpu_putref(ipforward_rt_percpu); 1498} 1499 1500void 1501ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1502 struct mbuf *m) 1503{ 1504 struct socket *so = inp->inp_socket; 1505 int inpflags = inp->inp_flags; 1506 1507 if (SOOPT_TIMESTAMP(so->so_options)) 1508 mp = sbsavetimestamp(so->so_options, mp); 1509 1510 if (inpflags & INP_RECVDSTADDR) { 1511 *mp = sbcreatecontrol(&ip->ip_dst, 1512 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1513 if (*mp) 1514 mp = &(*mp)->m_next; 1515 } 1516 1517 if (inpflags & INP_RECVTTL) { 1518 *mp = sbcreatecontrol(&ip->ip_ttl, 1519 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1520 if (*mp) 1521 mp = &(*mp)->m_next; 1522 } 1523 1524 struct psref psref; 1525 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1526 if (__predict_false(ifp == NULL)) { 1527#ifdef DIAGNOSTIC 1528 printf("%s: missing receive interface\n", __func__); 1529#endif 1530 return; /* XXX should report error? */ 1531 } 1532 1533 if (inpflags & INP_RECVPKTINFO) { 1534 struct in_pktinfo ipi; 1535 ipi.ipi_addr = ip->ip_dst; 1536 ipi.ipi_ifindex = ifp->if_index; 1537 *mp = sbcreatecontrol(&ipi, 1538 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1539 if (*mp) 1540 mp = &(*mp)->m_next; 1541 } 1542 if (inpflags & INP_RECVIF) { 1543 struct sockaddr_dl sdl; 1544 1545 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1546 NULL, 0); 1547 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1548 if (*mp) 1549 mp = &(*mp)->m_next; 1550 } 1551 m_put_rcvif_psref(ifp, &psref); 1552} 1553 1554/* 1555 * sysctl helper routine for net.inet.ip.forwsrcrt. 1556 */ 1557static int 1558sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1559{ 1560 int error, tmp; 1561 struct sysctlnode node; 1562 1563 node = *rnode; 1564 tmp = ip_forwsrcrt; 1565 node.sysctl_data = &tmp; 1566 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1567 if (error || newp == NULL) 1568 return (error); 1569 1570 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1571 0, NULL, NULL, NULL); 1572 if (error) 1573 return (error); 1574 1575 ip_forwsrcrt = tmp; 1576 1577 return (0); 1578} 1579 1580/* 1581 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1582 * range of the new value and tweaks timers if it changes. 1583 */ 1584static int 1585sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1586{ 1587 int error, tmp; 1588 struct sysctlnode node; 1589 1590 icmp_mtudisc_lock(); 1591 1592 node = *rnode; 1593 tmp = ip_mtudisc_timeout; 1594 node.sysctl_data = &tmp; 1595 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1596 if (error || newp == NULL) 1597 goto out; 1598 if (tmp < 0) { 1599 error = EINVAL; 1600 goto out; 1601 } 1602 1603 ip_mtudisc_timeout = tmp; 1604 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1605 error = 0; 1606out: 1607 icmp_mtudisc_unlock(); 1608 return error; 1609} 1610 1611static int 1612sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1613{ 1614 1615 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1616} 1617 1618static void 1619sysctl_net_inet_ip_setup(struct sysctllog **clog) 1620{ 1621 sysctl_createv(clog, 0, NULL, NULL, 1622 CTLFLAG_PERMANENT, 1623 CTLTYPE_NODE, "inet", 1624 SYSCTL_DESCR("PF_INET related settings"), 1625 NULL, 0, NULL, 0, 1626 CTL_NET, PF_INET, CTL_EOL); 1627 sysctl_createv(clog, 0, NULL, NULL, 1628 CTLFLAG_PERMANENT, 1629 CTLTYPE_NODE, "ip", 1630 SYSCTL_DESCR("IPv4 related settings"), 1631 NULL, 0, NULL, 0, 1632 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1633 1634 sysctl_createv(clog, 0, NULL, NULL, 1635 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1636 CTLTYPE_INT, "forwarding", 1637 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1638 NULL, 0, &ipforwarding, 0, 1639 CTL_NET, PF_INET, IPPROTO_IP, 1640 IPCTL_FORWARDING, CTL_EOL); 1641 sysctl_createv(clog, 0, NULL, NULL, 1642 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1643 CTLTYPE_INT, "redirect", 1644 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1645 NULL, 0, &ipsendredirects, 0, 1646 CTL_NET, PF_INET, IPPROTO_IP, 1647 IPCTL_SENDREDIRECTS, CTL_EOL); 1648 sysctl_createv(clog, 0, NULL, NULL, 1649 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1650 CTLTYPE_INT, "ttl", 1651 SYSCTL_DESCR("Default TTL for an INET datagram"), 1652 NULL, 0, &ip_defttl, 0, 1653 CTL_NET, PF_INET, IPPROTO_IP, 1654 IPCTL_DEFTTL, CTL_EOL); 1655 sysctl_createv(clog, 0, NULL, NULL, 1656 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1657 CTLTYPE_INT, "forwsrcrt", 1658 SYSCTL_DESCR("Enable forwarding of source-routed " 1659 "datagrams"), 1660 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1661 CTL_NET, PF_INET, IPPROTO_IP, 1662 IPCTL_FORWSRCRT, CTL_EOL); 1663 sysctl_createv(clog, 0, NULL, NULL, 1664 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1665 CTLTYPE_INT, "directed-broadcast", 1666 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1667 NULL, 0, &ip_directedbcast, 0, 1668 CTL_NET, PF_INET, IPPROTO_IP, 1669 IPCTL_DIRECTEDBCAST, CTL_EOL); 1670 sysctl_createv(clog, 0, NULL, NULL, 1671 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1672 CTLTYPE_INT, "allowsrcrt", 1673 SYSCTL_DESCR("Accept source-routed datagrams"), 1674 NULL, 0, &ip_allowsrcrt, 0, 1675 CTL_NET, PF_INET, IPPROTO_IP, 1676 IPCTL_ALLOWSRCRT, CTL_EOL); 1677 1678 sysctl_createv(clog, 0, NULL, NULL, 1679 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1680 CTLTYPE_INT, "mtudisc", 1681 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1682 NULL, 0, &ip_mtudisc, 0, 1683 CTL_NET, PF_INET, IPPROTO_IP, 1684 IPCTL_MTUDISC, CTL_EOL); 1685 sysctl_createv(clog, 0, NULL, NULL, 1686 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1687 CTLTYPE_INT, "anonportmin", 1688 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1689 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1690 CTL_NET, PF_INET, IPPROTO_IP, 1691 IPCTL_ANONPORTMIN, CTL_EOL); 1692 sysctl_createv(clog, 0, NULL, NULL, 1693 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1694 CTLTYPE_INT, "anonportmax", 1695 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1696 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1697 CTL_NET, PF_INET, IPPROTO_IP, 1698 IPCTL_ANONPORTMAX, CTL_EOL); 1699 sysctl_createv(clog, 0, NULL, NULL, 1700 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1701 CTLTYPE_INT, "mtudisctimeout", 1702 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1703 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1704 CTL_NET, PF_INET, IPPROTO_IP, 1705 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1706#ifndef IPNOPRIVPORTS 1707 sysctl_createv(clog, 0, NULL, NULL, 1708 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1709 CTLTYPE_INT, "lowportmin", 1710 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1711 "to assign"), 1712 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1713 CTL_NET, PF_INET, IPPROTO_IP, 1714 IPCTL_LOWPORTMIN, CTL_EOL); 1715 sysctl_createv(clog, 0, NULL, NULL, 1716 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1717 CTLTYPE_INT, "lowportmax", 1718 SYSCTL_DESCR("Highest privileged ephemeral port number " 1719 "to assign"), 1720 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1721 CTL_NET, PF_INET, IPPROTO_IP, 1722 IPCTL_LOWPORTMAX, CTL_EOL); 1723#endif /* IPNOPRIVPORTS */ 1724#if NGRE > 0 1725 sysctl_createv(clog, 0, NULL, NULL, 1726 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1727 CTLTYPE_INT, "grettl", 1728 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1729 NULL, 0, &ip_gre_ttl, 0, 1730 CTL_NET, PF_INET, IPPROTO_IP, 1731 IPCTL_GRE_TTL, CTL_EOL); 1732#endif /* NGRE */ 1733 sysctl_createv(clog, 0, NULL, NULL, 1734 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1735 CTLTYPE_INT, "checkinterface", 1736 SYSCTL_DESCR("Enable receive side of Strong ES model " 1737 "from RFC1122"), 1738 NULL, 0, &ip_checkinterface, 0, 1739 CTL_NET, PF_INET, IPPROTO_IP, 1740 IPCTL_CHECKINTERFACE, CTL_EOL); 1741 sysctl_createv(clog, 0, NULL, NULL, 1742 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1743 CTLTYPE_INT, "random_id", 1744 SYSCTL_DESCR("Assign random ip_id values"), 1745 NULL, 0, &ip_do_randomid, 0, 1746 CTL_NET, PF_INET, IPPROTO_IP, 1747 IPCTL_RANDOMID, CTL_EOL); 1748 sysctl_createv(clog, 0, NULL, NULL, 1749 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1750 CTLTYPE_INT, "do_loopback_cksum", 1751 SYSCTL_DESCR("Perform IP checksum on loopback"), 1752 NULL, 0, &ip_do_loopback_cksum, 0, 1753 CTL_NET, PF_INET, IPPROTO_IP, 1754 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1755 sysctl_createv(clog, 0, NULL, NULL, 1756 CTLFLAG_PERMANENT, 1757 CTLTYPE_STRUCT, "stats", 1758 SYSCTL_DESCR("IP statistics"), 1759 sysctl_net_inet_ip_stats, 0, NULL, 0, 1760 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1761 CTL_EOL); 1762#if NARP 1763 sysctl_createv(clog, 0, NULL, NULL, 1764 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1765 CTLTYPE_INT, "dad_count", 1766 SYSCTL_DESCR("Number of Duplicate Address Detection " 1767 "probes to send"), 1768 NULL, 0, &ip_dad_count, 0, 1769 CTL_NET, PF_INET, IPPROTO_IP, 1770 IPCTL_DAD_COUNT, CTL_EOL); 1771#endif 1772 1773 /* anonportalgo RFC6056 subtree */ 1774 const struct sysctlnode *portalgo_node; 1775 sysctl_createv(clog, 0, NULL, &portalgo_node, 1776 CTLFLAG_PERMANENT, 1777 CTLTYPE_NODE, "anonportalgo", 1778 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1779 NULL, 0, NULL, 0, 1780 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1781 sysctl_createv(clog, 0, &portalgo_node, NULL, 1782 CTLFLAG_PERMANENT, 1783 CTLTYPE_STRING, "available", 1784 SYSCTL_DESCR("available algorithms"), 1785 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1786 CTL_CREATE, CTL_EOL); 1787 sysctl_createv(clog, 0, &portalgo_node, NULL, 1788 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1789 CTLTYPE_STRING, "selected", 1790 SYSCTL_DESCR("selected algorithm"), 1791 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1792 CTL_CREATE, CTL_EOL); 1793 sysctl_createv(clog, 0, &portalgo_node, NULL, 1794 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1795 CTLTYPE_STRUCT, "reserve", 1796 SYSCTL_DESCR("bitmap of reserved ports"), 1797 sysctl_portalgo_reserve4, 0, NULL, 0, 1798 CTL_CREATE, CTL_EOL); 1799} 1800 1801void 1802ip_statinc(u_int stat) 1803{ 1804 1805 KASSERT(stat < IP_NSTATS); 1806 IP_STATINC(stat); 1807} 1808