ip_input.c revision 1.373
1/* $NetBSD: ip_input.c,v 1.373 2018/02/06 06:36:40 maxv 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 1.373 2018/02/06 06:36:40 maxv 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 = 0; 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 ipnat 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 *); 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 = percpu_alloc(sizeof(struct route)); 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 - Checking also should be disabled if the destination 328 * address is ipnat'ed to a different interface. 329 * 330 * XXX - Checking is incompatible with IP aliases added 331 * to the loopback interface instead of the interface where 332 * the packets are received. 333 * 334 * XXX - We need to add a per ifaddr flag for this so that 335 * we get finer grain control. 336 */ 337 checkif = ip_checkinterface && (ipforwarding == 0) && 338 (ifp->if_flags & IFF_LOOPBACK) == 0; 339 340 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 341 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 342 if (ia->ia4_flags & IN_IFF_NOTREADY) 343 continue; 344 if (checkif && ia->ia_ifp != ifp) 345 continue; 346 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) { 347 (*downmatch)++; 348 continue; 349 } 350 if (ia->ia4_flags & IN_IFF_DETACHED && 351 (ifp->if_flags & IFF_LOOPBACK) == 0) 352 continue; 353 break; 354 } 355 } 356 357 return ia; 358} 359 360static struct in_ifaddr * 361ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 362{ 363 struct in_ifaddr *ia = NULL; 364 struct ifaddr *ifa; 365 366 IFADDR_READER_FOREACH(ifa, ifp) { 367 if (ifa->ifa_addr->sa_family != AF_INET) 368 continue; 369 ia = ifatoia(ifa); 370 if (ia->ia4_flags & IN_IFF_NOTREADY) 371 continue; 372 if (ia->ia4_flags & IN_IFF_DETACHED && 373 (ifp->if_flags & IFF_LOOPBACK) == 0) 374 continue; 375 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 376 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 377 /* 378 * Look for all-0's host part (old broadcast addr), 379 * either for subnet or net. 380 */ 381 ip->ip_dst.s_addr == ia->ia_subnet || 382 ip->ip_dst.s_addr == ia->ia_net) 383 goto matched; 384 /* 385 * An interface with IP address zero accepts 386 * all packets that arrive on that interface. 387 */ 388 if (in_nullhost(ia->ia_addr.sin_addr)) 389 goto matched; 390 } 391 ia = NULL; 392 393matched: 394 return ia; 395} 396 397/* 398 * IP software interrupt routine. 399 */ 400static void 401ipintr(void *arg __unused) 402{ 403 struct mbuf *m; 404 405 KASSERT(cpu_softintr_p()); 406 407 SOFTNET_LOCK_UNLESS_NET_MPSAFE(); 408 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 409 ip_input(m); 410 } 411 SOFTNET_UNLOCK_UNLESS_NET_MPSAFE(); 412} 413 414/* 415 * IP input routine. Checksum and byte swap header. If fragmented 416 * try to reassemble. Process options. Pass to next level. 417 */ 418static void 419ip_input(struct mbuf *m) 420{ 421 struct ip *ip = NULL; 422 struct in_ifaddr *ia = NULL; 423 int hlen = 0, len; 424 int downmatch; 425 int srcrt = 0; 426 ifnet_t *ifp; 427 struct psref psref; 428 int s; 429 430 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 431 "interrupt handler; synchronization assumptions violated"); 432 433 MCLAIM(m, &ip_rx_mowner); 434 KASSERT((m->m_flags & M_PKTHDR) != 0); 435 436 ifp = m_get_rcvif_psref(m, &psref); 437 if (__predict_false(ifp == NULL)) 438 goto out; 439 440 /* 441 * If no IP addresses have been set yet but the interfaces 442 * are receiving, can't do anything with incoming packets yet. 443 * Note: we pre-check without locks held. 444 */ 445 if (IN_ADDRLIST_READER_EMPTY()) 446 goto out; 447 IP_STATINC(IP_STAT_TOTAL); 448 449 /* 450 * If the IP header is not aligned, slurp it up into a new 451 * mbuf with space for link headers, in the event we forward 452 * it. Otherwise, if it is aligned, make sure the entire 453 * base IP header is in the first mbuf of the chain. 454 */ 455 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 456 if ((m = m_copyup(m, sizeof(struct ip), 457 (max_linkhdr + 3) & ~3)) == NULL) { 458 /* XXXJRT new stat, please */ 459 IP_STATINC(IP_STAT_TOOSMALL); 460 goto out; 461 } 462 } else if (__predict_false(m->m_len < sizeof(struct ip))) { 463 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) { 464 IP_STATINC(IP_STAT_TOOSMALL); 465 goto out; 466 } 467 } 468 ip = mtod(m, struct ip *); 469 if (ip->ip_v != IPVERSION) { 470 IP_STATINC(IP_STAT_BADVERS); 471 goto out; 472 } 473 hlen = ip->ip_hl << 2; 474 if (hlen < sizeof(struct ip)) { /* minimum header length */ 475 IP_STATINC(IP_STAT_BADHLEN); 476 goto out; 477 } 478 if (hlen > m->m_len) { 479 if ((m = m_pullup(m, hlen)) == NULL) { 480 IP_STATINC(IP_STAT_BADHLEN); 481 goto out; 482 } 483 ip = mtod(m, struct ip *); 484 } 485 486 /* 487 * RFC1122: packets with a multicast source address are 488 * not allowed. 489 */ 490 if (IN_MULTICAST(ip->ip_src.s_addr)) { 491 IP_STATINC(IP_STAT_BADADDR); 492 goto out; 493 } 494 495 /* 127/8 must not appear on wire - RFC1122 */ 496 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 497 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 498 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 499 IP_STATINC(IP_STAT_BADADDR); 500 goto out; 501 } 502 } 503 504 switch (m->m_pkthdr.csum_flags & 505 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) { 506 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 507 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 508 IP_STATINC(IP_STAT_BADSUM); 509 goto out; 510 511 case M_CSUM_IPv4: 512 /* Checksum was okay. */ 513 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 514 break; 515 516 default: 517 /* 518 * Must compute it ourselves. Maybe skip checksum on 519 * loopback interfaces. 520 */ 521 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 522 ip_do_loopback_cksum)) { 523 INET_CSUM_COUNTER_INCR(&ip_swcsum); 524 if (in_cksum(m, hlen) != 0) { 525 IP_STATINC(IP_STAT_BADSUM); 526 goto out; 527 } 528 } 529 break; 530 } 531 532 /* Retrieve the packet length. */ 533 len = ntohs(ip->ip_len); 534 535 /* 536 * Check for additional length bogosity 537 */ 538 if (len < hlen) { 539 IP_STATINC(IP_STAT_BADLEN); 540 goto out; 541 } 542 543 /* 544 * Check that the amount of data in the buffers is at least as much 545 * as the IP header would have us expect. Trim mbufs if longer than 546 * we expect. Drop packet if shorter than we expect. 547 */ 548 if (m->m_pkthdr.len < len) { 549 IP_STATINC(IP_STAT_TOOSHORT); 550 goto out; 551 } 552 if (m->m_pkthdr.len > len) { 553 if (m->m_len == m->m_pkthdr.len) { 554 m->m_len = len; 555 m->m_pkthdr.len = len; 556 } else 557 m_adj(m, len - m->m_pkthdr.len); 558 } 559 560 /* 561 * Assume that we can create a fast-forward IP flow entry 562 * based on this packet. 563 */ 564 m->m_flags |= M_CANFASTFWD; 565 566 /* 567 * Run through list of hooks for input packets. If there are any 568 * filters which require that additional packets in the flow are 569 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 570 * Note that filters must _never_ set this flag, as another filter 571 * in the list may have previously cleared it. 572 */ 573#if defined(IPSEC) 574 if (!ipsec_used || !ipsec_indone(m)) 575#else 576 if (1) 577#endif 578 { 579 struct in_addr odst = ip->ip_dst; 580 bool freed; 581 582 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 583 if (freed || m == NULL) { 584 m = NULL; 585 goto out; 586 } 587 ip = mtod(m, struct ip *); 588 hlen = ip->ip_hl << 2; 589 590 /* 591 * XXX The setting of "srcrt" here is to prevent ip_forward() 592 * from generating ICMP redirects for packets that have 593 * been redirected by a hook back out on to the same LAN that 594 * they came from and is not an indication that the packet 595 * is being influenced by source routing options. This 596 * allows things like 597 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 598 * where tlp0 is both on the 1.1.1.0/24 network and is the 599 * default route for hosts on 1.1.1.0/24. Of course this 600 * also requires a "map tlp0 ..." to complete the story. 601 * One might argue whether or not this kind of network config. 602 * should be supported in this manner... 603 */ 604 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 605 } 606 607#ifdef ALTQ 608 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 609 if (altq_input) { 610 SOFTNET_LOCK(); 611 if ((*altq_input)(m, AF_INET) == 0) { 612 /* Packet dropped by traffic conditioner. */ 613 SOFTNET_UNLOCK(); 614 m = NULL; 615 goto out; 616 } 617 SOFTNET_UNLOCK(); 618 } 619#endif 620 621 /* 622 * Process options and, if not destined for us, 623 * ship it on. ip_dooptions returns 1 when an 624 * error was detected (causing an icmp message 625 * to be sent and the original packet to be freed). 626 */ 627 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 628 m = NULL; 629 goto out; 630 } 631 632 /* 633 * Check our list of addresses, to see if the packet is for us. 634 * 635 * Traditional 4.4BSD did not consult IFF_UP at all. 636 * The behavior here is to treat addresses on !IFF_UP interface 637 * or IN_IFF_NOTREADY addresses as not mine. 638 */ 639 downmatch = 0; 640 s = pserialize_read_enter(); 641 ia = ip_match_our_address(ifp, ip, &downmatch); 642 if (ia != NULL) { 643 pserialize_read_exit(s); 644 goto ours; 645 } 646 647 if (ifp->if_flags & IFF_BROADCAST) { 648 ia = ip_match_our_address_broadcast(ifp, ip); 649 if (ia != NULL) { 650 pserialize_read_exit(s); 651 goto ours; 652 } 653 } 654 pserialize_read_exit(s); 655 656 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 657#ifdef MROUTING 658 extern struct socket *ip_mrouter; 659 660 if (ip_mrouter) { 661 /* 662 * If we are acting as a multicast router, all 663 * incoming multicast packets are passed to the 664 * kernel-level multicast forwarding function. 665 * The packet is returned (relatively) intact; if 666 * ip_mforward() returns a non-zero value, the packet 667 * must be discarded, else it may be accepted below. 668 * 669 * (The IP ident field is put in the same byte order 670 * as expected when ip_mforward() is called from 671 * ip_output().) 672 */ 673 SOFTNET_LOCK(); 674 if (ip_mforward(m, ifp) != 0) { 675 SOFTNET_UNLOCK(); 676 IP_STATINC(IP_STAT_CANTFORWARD); 677 goto out; 678 } 679 SOFTNET_UNLOCK(); 680 681 /* 682 * The process-level routing demon needs to receive 683 * all multicast IGMP packets, whether or not this 684 * host belongs to their destination groups. 685 */ 686 if (ip->ip_p == IPPROTO_IGMP) { 687 goto ours; 688 } 689 IP_STATINC(IP_STAT_CANTFORWARD); 690 } 691#endif 692 /* 693 * See if we belong to the destination multicast group on the 694 * arrival interface. 695 */ 696 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 697 IP_STATINC(IP_STAT_CANTFORWARD); 698 goto out; 699 } 700 goto ours; 701 } 702 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 703 in_nullhost(ip->ip_dst)) 704 goto ours; 705 706 /* 707 * Not for us; forward if possible and desirable. 708 */ 709 if (ipforwarding == 0) { 710 m_put_rcvif_psref(ifp, &psref); 711 IP_STATINC(IP_STAT_CANTFORWARD); 712 m_freem(m); 713 } else { 714 /* 715 * If ip_dst matched any of my address on !IFF_UP interface, 716 * and there's no IFF_UP interface that matches ip_dst, 717 * send icmp unreach. Forwarding it will result in in-kernel 718 * forwarding loop till TTL goes to 0. 719 */ 720 if (downmatch) { 721 m_put_rcvif_psref(ifp, &psref); 722 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 723 IP_STATINC(IP_STAT_CANTFORWARD); 724 return; 725 } 726#ifdef IPSEC 727 /* Check the security policy (SP) for the packet */ 728 if (ipsec_used) { 729 if (ipsec4_input(m, IP_FORWARDING | 730 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 731 goto out; 732 } 733 } 734#endif 735 ip_forward(m, srcrt, ifp); 736 m_put_rcvif_psref(ifp, &psref); 737 } 738 return; 739 740ours: 741 m_put_rcvif_psref(ifp, &psref); 742 ifp = NULL; 743 744 /* 745 * If offset or IP_MF are set, must reassemble. 746 */ 747 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 748 /* 749 * Pass to IP reassembly mechanism. 750 */ 751 if (ip_reass_packet(&m, ip) != 0) { 752 /* Failed; invalid fragment(s) or packet. */ 753 goto out; 754 } 755 if (m == NULL) { 756 /* More fragments should come; silently return. */ 757 goto out; 758 } 759 /* 760 * Reassembly is done, we have the final packet. 761 * Update cached data in local variable(s). 762 */ 763 ip = mtod(m, struct ip *); 764 hlen = ip->ip_hl << 2; 765 } 766 767#ifdef IPSEC 768 /* 769 * Enforce IPsec policy checking if we are seeing last header. 770 * Note that we do not visit this with protocols with PCB layer 771 * code - like UDP/TCP/raw IP. 772 */ 773 if (ipsec_used && 774 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 775 if (ipsec4_input(m, 0) != 0) { 776 goto out; 777 } 778 } 779#endif 780 781 /* 782 * Switch out to protocol's input routine. 783 */ 784#if IFA_STATS 785 if (ia && ip) { 786 struct in_ifaddr *_ia; 787 /* 788 * Keep a reference from ip_match_our_address with psref 789 * is expensive, so explore ia here again. 790 */ 791 s = pserialize_read_enter(); 792 _ia = in_get_ia(ip->ip_dst); 793 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 794 pserialize_read_exit(s); 795 } 796#endif 797 IP_STATINC(IP_STAT_DELIVERED); 798 799 const int off = hlen, nh = ip->ip_p; 800 801 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 802 return; 803 804out: 805 m_put_rcvif_psref(ifp, &psref); 806 if (m != NULL) 807 m_freem(m); 808} 809 810/* 811 * IP timer processing. 812 */ 813void 814ip_slowtimo(void) 815{ 816 817 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 818 819 ip_reass_slowtimo(); 820 821 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 822} 823 824/* 825 * IP drain processing. 826 */ 827void 828ip_drain(void) 829{ 830 831 KERNEL_LOCK(1, NULL); 832 ip_reass_drain(); 833 KERNEL_UNLOCK_ONE(NULL); 834} 835 836/* 837 * ip_dooptions: perform option processing on a datagram, possibly discarding 838 * it if bad options are encountered, or forwarding it if source-routed. 839 * 840 * => Returns true if packet has been forwarded/freed. 841 * => Returns false if the packet should be processed further. 842 */ 843static bool 844ip_dooptions(struct mbuf *m) 845{ 846 struct ip *ip = mtod(m, struct ip *); 847 u_char *cp, *cp0; 848 struct ip_timestamp *ipt; 849 struct in_ifaddr *ia; 850 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 851 int srr_present, rr_present, ts_present; 852 struct in_addr dst; 853 n_time ntime; 854 struct ifaddr *ifa = NULL; 855 int s; 856 857 srr_present = 0; 858 rr_present = 0; 859 ts_present = 0; 860 861 dst = ip->ip_dst; 862 cp = (u_char *)(ip + 1); 863 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 864 for (; cnt > 0; cnt -= optlen, cp += optlen) { 865 opt = cp[IPOPT_OPTVAL]; 866 if (opt == IPOPT_EOL) 867 break; 868 if (opt == IPOPT_NOP) 869 optlen = 1; 870 else { 871 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 872 code = &cp[IPOPT_OLEN] - (u_char *)ip; 873 goto bad; 874 } 875 optlen = cp[IPOPT_OLEN]; 876 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 877 code = &cp[IPOPT_OLEN] - (u_char *)ip; 878 goto bad; 879 } 880 } 881 switch (opt) { 882 883 default: 884 break; 885 886 /* 887 * Source routing with record. 888 * Find interface with current destination address. 889 * If none on this machine then drop if strictly routed, 890 * or do nothing if loosely routed. 891 * Record interface address and bring up next address 892 * component. If strictly routed make sure next 893 * address is on directly accessible net. 894 */ 895 case IPOPT_LSRR: 896 case IPOPT_SSRR: { 897 struct psref psref; 898 struct sockaddr_in ipaddr = { 899 .sin_len = sizeof(ipaddr), 900 .sin_family = AF_INET, 901 }; 902 903 if (ip_allowsrcrt == 0) { 904 type = ICMP_UNREACH; 905 code = ICMP_UNREACH_NET_PROHIB; 906 goto bad; 907 } 908 if (srr_present++) { 909 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 910 goto bad; 911 } 912 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 913 code = &cp[IPOPT_OLEN] - (u_char *)ip; 914 goto bad; 915 } 916 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 917 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 918 goto bad; 919 } 920 ipaddr.sin_addr = ip->ip_dst; 921 922 s = pserialize_read_enter(); 923 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 924 if (ifa == NULL) { 925 pserialize_read_exit(s); 926 if (opt == IPOPT_SSRR) { 927 type = ICMP_UNREACH; 928 code = ICMP_UNREACH_SRCFAIL; 929 goto bad; 930 } 931 /* 932 * Loose routing, and not at next destination 933 * yet; nothing to do except forward. 934 */ 935 break; 936 } 937 pserialize_read_exit(s); 938 939 off--; /* 0 origin */ 940 if ((off + sizeof(struct in_addr)) > optlen) { 941 /* 942 * End of source route. Should be for us. 943 */ 944 save_rte(m, cp, ip->ip_src); 945 break; 946 } 947 /* 948 * locate outgoing interface 949 */ 950 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 951 sizeof(ipaddr.sin_addr)); 952 if (opt == IPOPT_SSRR) { 953 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 954 &psref); 955 if (ifa != NULL) 956 ia = ifatoia(ifa); 957 else 958 ia = NULL; 959 } else { 960 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 961 } 962 if (ia == NULL) { 963 type = ICMP_UNREACH; 964 code = ICMP_UNREACH_SRCFAIL; 965 goto bad; 966 } 967 ip->ip_dst = ipaddr.sin_addr; 968 bcopy((void *)&ia->ia_addr.sin_addr, 969 (void *)(cp + off), sizeof(struct in_addr)); 970 ia4_release(ia, &psref); 971 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 972 /* 973 * Let ip_intr's mcast routing check handle mcast pkts 974 */ 975 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 976 break; 977 } 978 979 case IPOPT_RR: { 980 struct psref psref; 981 struct sockaddr_in ipaddr = { 982 .sin_len = sizeof(ipaddr), 983 .sin_family = AF_INET, 984 }; 985 986 if (rr_present++) { 987 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 988 goto bad; 989 } 990 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 991 code = &cp[IPOPT_OLEN] - (u_char *)ip; 992 goto bad; 993 } 994 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 995 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 996 goto bad; 997 } 998 /* 999 * If no space remains, ignore. 1000 */ 1001 off--; /* 0 origin */ 1002 if ((off + sizeof(struct in_addr)) > optlen) 1003 break; 1004 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1005 sizeof(ipaddr.sin_addr)); 1006 /* 1007 * locate outgoing interface; if we're the destination, 1008 * use the incoming interface (should be same). 1009 */ 1010 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1011 if (ifa == NULL) { 1012 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1013 if (ia == NULL) { 1014 type = ICMP_UNREACH; 1015 code = ICMP_UNREACH_HOST; 1016 goto bad; 1017 } 1018 } else { 1019 ia = ifatoia(ifa); 1020 } 1021 bcopy((void *)&ia->ia_addr.sin_addr, 1022 (void *)(cp + off), sizeof(struct in_addr)); 1023 ia4_release(ia, &psref); 1024 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1025 break; 1026 } 1027 1028 case IPOPT_TS: 1029 code = cp - (u_char *)ip; 1030 ipt = (struct ip_timestamp *)cp; 1031 if (ts_present++) { 1032 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1033 goto bad; 1034 } 1035 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1036 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1037 goto bad; 1038 } 1039 if (ipt->ipt_ptr < 5) { 1040 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1041 goto bad; 1042 } 1043 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1044 if (++ipt->ipt_oflw == 0) { 1045 code = (u_char *)&ipt->ipt_ptr - 1046 (u_char *)ip; 1047 goto bad; 1048 } 1049 break; 1050 } 1051 cp0 = (cp + ipt->ipt_ptr - 1); 1052 switch (ipt->ipt_flg) { 1053 1054 case IPOPT_TS_TSONLY: 1055 break; 1056 1057 case IPOPT_TS_TSANDADDR: { 1058 struct ifnet *rcvif; 1059 int _s, _ss; 1060 struct sockaddr_in ipaddr = { 1061 .sin_len = sizeof(ipaddr), 1062 .sin_family = AF_INET, 1063 }; 1064 1065 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1066 sizeof(struct in_addr) > ipt->ipt_len) { 1067 code = (u_char *)&ipt->ipt_ptr - 1068 (u_char *)ip; 1069 goto bad; 1070 } 1071 ipaddr.sin_addr = dst; 1072 _ss = pserialize_read_enter(); 1073 rcvif = m_get_rcvif(m, &_s); 1074 if (__predict_true(rcvif != NULL)) { 1075 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1076 rcvif); 1077 } 1078 m_put_rcvif(rcvif, &_s); 1079 if (ifa == NULL) { 1080 pserialize_read_exit(_ss); 1081 break; 1082 } 1083 ia = ifatoia(ifa); 1084 bcopy(&ia->ia_addr.sin_addr, 1085 cp0, sizeof(struct in_addr)); 1086 pserialize_read_exit(_ss); 1087 ipt->ipt_ptr += sizeof(struct in_addr); 1088 break; 1089 } 1090 1091 case IPOPT_TS_PRESPEC: { 1092 struct sockaddr_in ipaddr = { 1093 .sin_len = sizeof(ipaddr), 1094 .sin_family = AF_INET, 1095 }; 1096 1097 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1098 sizeof(struct in_addr) > ipt->ipt_len) { 1099 code = (u_char *)&ipt->ipt_ptr - 1100 (u_char *)ip; 1101 goto bad; 1102 } 1103 memcpy(&ipaddr.sin_addr, cp0, 1104 sizeof(struct in_addr)); 1105 s = pserialize_read_enter(); 1106 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1107 if (ifa == NULL) { 1108 pserialize_read_exit(s); 1109 continue; 1110 } 1111 pserialize_read_exit(s); 1112 ipt->ipt_ptr += sizeof(struct in_addr); 1113 break; 1114 } 1115 1116 default: 1117 /* XXX can't take &ipt->ipt_flg */ 1118 code = (u_char *)&ipt->ipt_ptr - 1119 (u_char *)ip + 1; 1120 goto bad; 1121 } 1122 ntime = iptime(); 1123 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1124 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1125 sizeof(n_time)); 1126 ipt->ipt_ptr += sizeof(n_time); 1127 } 1128 } 1129 if (forward) { 1130 struct ifnet *rcvif; 1131 struct psref _psref; 1132 1133 if (ip_forwsrcrt == 0) { 1134 type = ICMP_UNREACH; 1135 code = ICMP_UNREACH_SRCFAIL; 1136 goto bad; 1137 } 1138 1139 rcvif = m_get_rcvif_psref(m, &_psref); 1140 if (__predict_false(rcvif == NULL)) { 1141 type = ICMP_UNREACH; 1142 code = ICMP_UNREACH_HOST; 1143 goto bad; 1144 } 1145 ip_forward(m, 1, rcvif); 1146 m_put_rcvif_psref(rcvif, &_psref); 1147 return true; 1148 } 1149 return false; 1150bad: 1151 icmp_error(m, type, code, 0, 0); 1152 IP_STATINC(IP_STAT_BADOPTIONS); 1153 return true; 1154} 1155 1156/* 1157 * ip_rtaddr: given address of next destination (final or next hop), 1158 * return internet address info of interface to be used to get there. 1159 */ 1160static struct in_ifaddr * 1161ip_rtaddr(struct in_addr dst, struct psref *psref) 1162{ 1163 struct rtentry *rt; 1164 union { 1165 struct sockaddr dst; 1166 struct sockaddr_in dst4; 1167 } u; 1168 struct route *ro; 1169 1170 sockaddr_in_init(&u.dst4, &dst, 0); 1171 1172 ro = percpu_getref(ipforward_rt_percpu); 1173 rt = rtcache_lookup(ro, &u.dst); 1174 if (rt == NULL) { 1175 percpu_putref(ipforward_rt_percpu); 1176 return NULL; 1177 } 1178 1179 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1180 rtcache_unref(rt, ro); 1181 percpu_putref(ipforward_rt_percpu); 1182 1183 return ifatoia(rt->rt_ifa); 1184} 1185 1186/* 1187 * save_rte: save incoming source route for use in replies, to be picked 1188 * up later by ip_srcroute if the receiver is interested. 1189 */ 1190static void 1191save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1192{ 1193 struct ip_srcrt *isr; 1194 struct m_tag *mtag; 1195 unsigned olen; 1196 1197 olen = option[IPOPT_OLEN]; 1198 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1199 return; 1200 1201 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1202 if (mtag == NULL) 1203 return; 1204 isr = (struct ip_srcrt *)(mtag + 1); 1205 1206 memcpy(isr->isr_hdr, option, olen); 1207 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1208 isr->isr_dst = dst; 1209 m_tag_prepend(m, mtag); 1210} 1211 1212/* 1213 * Retrieve incoming source route for use in replies, 1214 * in the same form used by setsockopt. 1215 * The first hop is placed before the options, will be removed later. 1216 */ 1217struct mbuf * 1218ip_srcroute(struct mbuf *m0) 1219{ 1220 struct in_addr *p, *q; 1221 struct mbuf *m; 1222 struct ip_srcrt *isr; 1223 struct m_tag *mtag; 1224 1225 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL); 1226 if (mtag == NULL) 1227 return NULL; 1228 isr = (struct ip_srcrt *)(mtag + 1); 1229 1230 if (isr->isr_nhops == 0) 1231 return NULL; 1232 1233 m = m_get(M_DONTWAIT, MT_SOOPTS); 1234 if (m == NULL) 1235 return NULL; 1236 1237 MCLAIM(m, &inetdomain.dom_mowner); 1238#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1239 1240 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1241 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1242 1243 /* 1244 * First save first hop for return route 1245 */ 1246 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1247 *(mtod(m, struct in_addr *)) = *p--; 1248 1249 /* 1250 * Copy option fields and padding (nop) to mbuf. 1251 */ 1252 isr->isr_nop = IPOPT_NOP; 1253 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1254 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1255 OPTSIZ); 1256 q = (struct in_addr *)(mtod(m, char *) + 1257 sizeof(struct in_addr) + OPTSIZ); 1258#undef OPTSIZ 1259 /* 1260 * Record return path as an IP source route, 1261 * reversing the path (pointers are now aligned). 1262 */ 1263 while (p >= isr->isr_routes) { 1264 *q++ = *p--; 1265 } 1266 /* 1267 * Last hop goes to final destination. 1268 */ 1269 *q = isr->isr_dst; 1270 m_tag_delete(m0, mtag); 1271 return m; 1272} 1273 1274const int inetctlerrmap[PRC_NCMDS] = { 1275 [PRC_MSGSIZE] = EMSGSIZE, 1276 [PRC_HOSTDEAD] = EHOSTDOWN, 1277 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1278 [PRC_UNREACH_NET] = EHOSTUNREACH, 1279 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1280 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1281 [PRC_UNREACH_PORT] = ECONNREFUSED, 1282 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1283 [PRC_PARAMPROB] = ENOPROTOOPT, 1284}; 1285 1286void 1287ip_fasttimo(void) 1288{ 1289 if (ip_drainwanted) { 1290 ip_drain(); 1291 ip_drainwanted = 0; 1292 } 1293} 1294 1295void 1296ip_drainstub(void) 1297{ 1298 ip_drainwanted = 1; 1299} 1300 1301/* 1302 * Forward a packet. If some error occurs return the sender 1303 * an icmp packet. Note we can't always generate a meaningful 1304 * icmp message because icmp doesn't have a large enough repertoire 1305 * of codes and types. 1306 * 1307 * If not forwarding, just drop the packet. This could be confusing 1308 * if ipforwarding was zero but some routing protocol was advancing 1309 * us as a gateway to somewhere. However, we must let the routing 1310 * protocol deal with that. 1311 * 1312 * The srcrt parameter indicates whether the packet is being forwarded 1313 * via a source route. 1314 */ 1315static void 1316ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1317{ 1318 struct ip *ip = mtod(m, struct ip *); 1319 struct rtentry *rt; 1320 int error, type = 0, code = 0, destmtu = 0; 1321 struct mbuf *mcopy; 1322 n_long dest; 1323 union { 1324 struct sockaddr dst; 1325 struct sockaddr_in dst4; 1326 } u; 1327 uint64_t *ips; 1328 struct route *ro; 1329 1330 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1331 "interrupt handler; synchronization assumptions violated"); 1332 1333 /* 1334 * We are now in the output path. 1335 */ 1336 MCLAIM(m, &ip_tx_mowner); 1337 1338 /* 1339 * Clear any in-bound checksum flags for this packet. 1340 */ 1341 m->m_pkthdr.csum_flags = 0; 1342 1343 dest = 0; 1344 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1345 IP_STATINC(IP_STAT_CANTFORWARD); 1346 m_freem(m); 1347 return; 1348 } 1349 1350 if (ip->ip_ttl <= IPTTLDEC) { 1351 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1352 return; 1353 } 1354 1355 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1356 1357 ro = percpu_getref(ipforward_rt_percpu); 1358 rt = rtcache_lookup(ro, &u.dst); 1359 if (rt == NULL) { 1360 percpu_putref(ipforward_rt_percpu); 1361 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1362 return; 1363 } 1364 1365 /* 1366 * Save at most 68 bytes of the packet in case 1367 * we need to generate an ICMP message to the src. 1368 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1369 */ 1370 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1371 if (mcopy) 1372 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1373 1374 ip->ip_ttl -= IPTTLDEC; 1375 1376 /* 1377 * If forwarding packet using same interface that it came in on, 1378 * perhaps should send a redirect to sender to shortcut a hop. 1379 * Only send redirect if source is sending directly to us, 1380 * and if packet was not source routed (or has any options). 1381 * Also, don't send redirect if forwarding using a default route 1382 * or a route modified by a redirect. 1383 */ 1384 if (rt->rt_ifp == rcvif && 1385 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1386 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1387 ipsendredirects && !srcrt) { 1388 if (rt->rt_ifa && 1389 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1390 ifatoia(rt->rt_ifa)->ia_subnet) { 1391 if (rt->rt_flags & RTF_GATEWAY) 1392 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1393 else 1394 dest = ip->ip_dst.s_addr; 1395 /* 1396 * Router requirements says to only send host 1397 * redirects. 1398 */ 1399 type = ICMP_REDIRECT; 1400 code = ICMP_REDIRECT_HOST; 1401 } 1402 } 1403 rtcache_unref(rt, ro); 1404 1405 error = ip_output(m, NULL, ro, 1406 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1407 NULL, NULL); 1408 1409 if (error) { 1410 IP_STATINC(IP_STAT_CANTFORWARD); 1411 goto error; 1412 } 1413 1414 ips = IP_STAT_GETREF(); 1415 ips[IP_STAT_FORWARD]++; 1416 1417 if (type) { 1418 ips[IP_STAT_REDIRECTSENT]++; 1419 IP_STAT_PUTREF(); 1420 goto redirect; 1421 } 1422 1423 IP_STAT_PUTREF(); 1424 if (mcopy) { 1425#ifdef GATEWAY 1426 if (mcopy->m_flags & M_CANFASTFWD) 1427 ipflow_create(ro, mcopy); 1428#endif 1429 m_freem(mcopy); 1430 } 1431 1432 percpu_putref(ipforward_rt_percpu); 1433 return; 1434 1435redirect: 1436error: 1437 if (mcopy == NULL) { 1438 percpu_putref(ipforward_rt_percpu); 1439 return; 1440 } 1441 1442 switch (error) { 1443 1444 case 0: /* forwarded, but need redirect */ 1445 /* type, code set above */ 1446 break; 1447 1448 case ENETUNREACH: /* shouldn't happen, checked above */ 1449 case EHOSTUNREACH: 1450 case ENETDOWN: 1451 case EHOSTDOWN: 1452 default: 1453 type = ICMP_UNREACH; 1454 code = ICMP_UNREACH_HOST; 1455 break; 1456 1457 case EMSGSIZE: 1458 type = ICMP_UNREACH; 1459 code = ICMP_UNREACH_NEEDFRAG; 1460 1461 if ((rt = rtcache_validate(ro)) != NULL) { 1462 destmtu = rt->rt_ifp->if_mtu; 1463 rtcache_unref(rt, ro); 1464 } 1465#ifdef IPSEC 1466 if (ipsec_used) 1467 (void)ipsec4_forward(mcopy, &destmtu); 1468#endif 1469 IP_STATINC(IP_STAT_CANTFRAG); 1470 break; 1471 1472 case ENOBUFS: 1473 /* 1474 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1475 * Requirements for IP Version 4 Routers. Source quench can 1476 * be a big problem under DoS attacks or if the underlying 1477 * interface is rate-limited. 1478 */ 1479 if (mcopy) 1480 m_freem(mcopy); 1481 percpu_putref(ipforward_rt_percpu); 1482 return; 1483 } 1484 icmp_error(mcopy, type, code, dest, destmtu); 1485 percpu_putref(ipforward_rt_percpu); 1486} 1487 1488void 1489ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1490 struct mbuf *m) 1491{ 1492 struct socket *so = inp->inp_socket; 1493 int inpflags = inp->inp_flags; 1494 1495 if (SOOPT_TIMESTAMP(so->so_options)) 1496 mp = sbsavetimestamp(so->so_options, m, mp); 1497 1498 if (inpflags & INP_RECVDSTADDR) { 1499 *mp = sbcreatecontrol(&ip->ip_dst, 1500 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1501 if (*mp) 1502 mp = &(*mp)->m_next; 1503 } 1504 1505 if (inpflags & INP_RECVTTL) { 1506 *mp = sbcreatecontrol(&ip->ip_ttl, 1507 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1508 if (*mp) 1509 mp = &(*mp)->m_next; 1510 } 1511 1512 struct psref psref; 1513 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1514 if (__predict_false(ifp == NULL)) { 1515#ifdef DIAGNOSTIC 1516 printf("%s: missing receive interface\n", __func__); 1517#endif 1518 return; /* XXX should report error? */ 1519 } 1520 1521 if (inpflags & INP_RECVPKTINFO) { 1522 struct in_pktinfo ipi; 1523 ipi.ipi_addr = ip->ip_dst; 1524 ipi.ipi_ifindex = ifp->if_index; 1525 *mp = sbcreatecontrol(&ipi, 1526 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1527 if (*mp) 1528 mp = &(*mp)->m_next; 1529 } 1530 if (inpflags & INP_RECVIF) { 1531 struct sockaddr_dl sdl; 1532 1533 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1534 NULL, 0); 1535 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1536 if (*mp) 1537 mp = &(*mp)->m_next; 1538 } 1539 m_put_rcvif_psref(ifp, &psref); 1540} 1541 1542/* 1543 * sysctl helper routine for net.inet.ip.forwsrcrt. 1544 */ 1545static int 1546sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1547{ 1548 int error, tmp; 1549 struct sysctlnode node; 1550 1551 node = *rnode; 1552 tmp = ip_forwsrcrt; 1553 node.sysctl_data = &tmp; 1554 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1555 if (error || newp == NULL) 1556 return (error); 1557 1558 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1559 0, NULL, NULL, NULL); 1560 if (error) 1561 return (error); 1562 1563 ip_forwsrcrt = tmp; 1564 1565 return (0); 1566} 1567 1568/* 1569 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1570 * range of the new value and tweaks timers if it changes. 1571 */ 1572static int 1573sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1574{ 1575 int error, tmp; 1576 struct sysctlnode node; 1577 1578 icmp_mtudisc_lock(); 1579 1580 node = *rnode; 1581 tmp = ip_mtudisc_timeout; 1582 node.sysctl_data = &tmp; 1583 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1584 if (error || newp == NULL) 1585 goto out; 1586 if (tmp < 0) { 1587 error = EINVAL; 1588 goto out; 1589 } 1590 1591 ip_mtudisc_timeout = tmp; 1592 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1593 error = 0; 1594out: 1595 icmp_mtudisc_unlock(); 1596 return error; 1597} 1598 1599static int 1600sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1601{ 1602 1603 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1604} 1605 1606static void 1607sysctl_net_inet_ip_setup(struct sysctllog **clog) 1608{ 1609 sysctl_createv(clog, 0, NULL, NULL, 1610 CTLFLAG_PERMANENT, 1611 CTLTYPE_NODE, "inet", 1612 SYSCTL_DESCR("PF_INET related settings"), 1613 NULL, 0, NULL, 0, 1614 CTL_NET, PF_INET, CTL_EOL); 1615 sysctl_createv(clog, 0, NULL, NULL, 1616 CTLFLAG_PERMANENT, 1617 CTLTYPE_NODE, "ip", 1618 SYSCTL_DESCR("IPv4 related settings"), 1619 NULL, 0, NULL, 0, 1620 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1621 1622 sysctl_createv(clog, 0, NULL, NULL, 1623 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1624 CTLTYPE_INT, "forwarding", 1625 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1626 NULL, 0, &ipforwarding, 0, 1627 CTL_NET, PF_INET, IPPROTO_IP, 1628 IPCTL_FORWARDING, CTL_EOL); 1629 sysctl_createv(clog, 0, NULL, NULL, 1630 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1631 CTLTYPE_INT, "redirect", 1632 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1633 NULL, 0, &ipsendredirects, 0, 1634 CTL_NET, PF_INET, IPPROTO_IP, 1635 IPCTL_SENDREDIRECTS, CTL_EOL); 1636 sysctl_createv(clog, 0, NULL, NULL, 1637 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1638 CTLTYPE_INT, "ttl", 1639 SYSCTL_DESCR("Default TTL for an INET datagram"), 1640 NULL, 0, &ip_defttl, 0, 1641 CTL_NET, PF_INET, IPPROTO_IP, 1642 IPCTL_DEFTTL, CTL_EOL); 1643#ifdef IPCTL_DEFMTU 1644 sysctl_createv(clog, 0, NULL, NULL, 1645 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1646 CTLTYPE_INT, "mtu", 1647 SYSCTL_DESCR("Default MTA for an INET route"), 1648 NULL, 0, &ip_mtu, 0, 1649 CTL_NET, PF_INET, IPPROTO_IP, 1650 IPCTL_DEFMTU, CTL_EOL); 1651#endif /* IPCTL_DEFMTU */ 1652 sysctl_createv(clog, 0, NULL, NULL, 1653 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1654 CTLTYPE_INT, "forwsrcrt", 1655 SYSCTL_DESCR("Enable forwarding of source-routed " 1656 "datagrams"), 1657 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1658 CTL_NET, PF_INET, IPPROTO_IP, 1659 IPCTL_FORWSRCRT, CTL_EOL); 1660 sysctl_createv(clog, 0, NULL, NULL, 1661 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1662 CTLTYPE_INT, "directed-broadcast", 1663 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1664 NULL, 0, &ip_directedbcast, 0, 1665 CTL_NET, PF_INET, IPPROTO_IP, 1666 IPCTL_DIRECTEDBCAST, CTL_EOL); 1667 sysctl_createv(clog, 0, NULL, NULL, 1668 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1669 CTLTYPE_INT, "allowsrcrt", 1670 SYSCTL_DESCR("Accept source-routed datagrams"), 1671 NULL, 0, &ip_allowsrcrt, 0, 1672 CTL_NET, PF_INET, IPPROTO_IP, 1673 IPCTL_ALLOWSRCRT, CTL_EOL); 1674 1675 sysctl_createv(clog, 0, NULL, NULL, 1676 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1677 CTLTYPE_INT, "mtudisc", 1678 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1679 NULL, 0, &ip_mtudisc, 0, 1680 CTL_NET, PF_INET, IPPROTO_IP, 1681 IPCTL_MTUDISC, CTL_EOL); 1682 sysctl_createv(clog, 0, NULL, NULL, 1683 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1684 CTLTYPE_INT, "anonportmin", 1685 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1686 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1687 CTL_NET, PF_INET, IPPROTO_IP, 1688 IPCTL_ANONPORTMIN, CTL_EOL); 1689 sysctl_createv(clog, 0, NULL, NULL, 1690 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1691 CTLTYPE_INT, "anonportmax", 1692 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1693 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1694 CTL_NET, PF_INET, IPPROTO_IP, 1695 IPCTL_ANONPORTMAX, CTL_EOL); 1696 sysctl_createv(clog, 0, NULL, NULL, 1697 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1698 CTLTYPE_INT, "mtudisctimeout", 1699 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1700 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1701 CTL_NET, PF_INET, IPPROTO_IP, 1702 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1703#ifndef IPNOPRIVPORTS 1704 sysctl_createv(clog, 0, NULL, NULL, 1705 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1706 CTLTYPE_INT, "lowportmin", 1707 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1708 "to assign"), 1709 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1710 CTL_NET, PF_INET, IPPROTO_IP, 1711 IPCTL_LOWPORTMIN, CTL_EOL); 1712 sysctl_createv(clog, 0, NULL, NULL, 1713 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1714 CTLTYPE_INT, "lowportmax", 1715 SYSCTL_DESCR("Highest privileged ephemeral port number " 1716 "to assign"), 1717 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1718 CTL_NET, PF_INET, IPPROTO_IP, 1719 IPCTL_LOWPORTMAX, CTL_EOL); 1720#endif /* IPNOPRIVPORTS */ 1721#if NGRE > 0 1722 sysctl_createv(clog, 0, NULL, NULL, 1723 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1724 CTLTYPE_INT, "grettl", 1725 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1726 NULL, 0, &ip_gre_ttl, 0, 1727 CTL_NET, PF_INET, IPPROTO_IP, 1728 IPCTL_GRE_TTL, CTL_EOL); 1729#endif /* NGRE */ 1730 sysctl_createv(clog, 0, NULL, NULL, 1731 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1732 CTLTYPE_INT, "checkinterface", 1733 SYSCTL_DESCR("Enable receive side of Strong ES model " 1734 "from RFC1122"), 1735 NULL, 0, &ip_checkinterface, 0, 1736 CTL_NET, PF_INET, IPPROTO_IP, 1737 IPCTL_CHECKINTERFACE, CTL_EOL); 1738 sysctl_createv(clog, 0, NULL, NULL, 1739 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1740 CTLTYPE_INT, "random_id", 1741 SYSCTL_DESCR("Assign random ip_id values"), 1742 NULL, 0, &ip_do_randomid, 0, 1743 CTL_NET, PF_INET, IPPROTO_IP, 1744 IPCTL_RANDOMID, CTL_EOL); 1745 sysctl_createv(clog, 0, NULL, NULL, 1746 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1747 CTLTYPE_INT, "do_loopback_cksum", 1748 SYSCTL_DESCR("Perform IP checksum on loopback"), 1749 NULL, 0, &ip_do_loopback_cksum, 0, 1750 CTL_NET, PF_INET, IPPROTO_IP, 1751 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1752 sysctl_createv(clog, 0, NULL, NULL, 1753 CTLFLAG_PERMANENT, 1754 CTLTYPE_STRUCT, "stats", 1755 SYSCTL_DESCR("IP statistics"), 1756 sysctl_net_inet_ip_stats, 0, NULL, 0, 1757 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1758 CTL_EOL); 1759#if NARP 1760 sysctl_createv(clog, 0, NULL, NULL, 1761 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1762 CTLTYPE_INT, "dad_count", 1763 SYSCTL_DESCR("Number of Duplicate Address Detection " 1764 "probes to send"), 1765 NULL, 0, &ip_dad_count, 0, 1766 CTL_NET, PF_INET, IPPROTO_IP, 1767 IPCTL_DAD_COUNT, CTL_EOL); 1768#endif 1769 1770 /* anonportalgo RFC6056 subtree */ 1771 const struct sysctlnode *portalgo_node; 1772 sysctl_createv(clog, 0, NULL, &portalgo_node, 1773 CTLFLAG_PERMANENT, 1774 CTLTYPE_NODE, "anonportalgo", 1775 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1776 NULL, 0, NULL, 0, 1777 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1778 sysctl_createv(clog, 0, &portalgo_node, NULL, 1779 CTLFLAG_PERMANENT, 1780 CTLTYPE_STRING, "available", 1781 SYSCTL_DESCR("available algorithms"), 1782 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1783 CTL_CREATE, CTL_EOL); 1784 sysctl_createv(clog, 0, &portalgo_node, NULL, 1785 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1786 CTLTYPE_STRING, "selected", 1787 SYSCTL_DESCR("selected algorithm"), 1788 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1789 CTL_CREATE, CTL_EOL); 1790 sysctl_createv(clog, 0, &portalgo_node, NULL, 1791 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1792 CTLTYPE_STRUCT, "reserve", 1793 SYSCTL_DESCR("bitmap of reserved ports"), 1794 sysctl_portalgo_reserve4, 0, NULL, 0, 1795 CTL_CREATE, CTL_EOL); 1796} 1797 1798void 1799ip_statinc(u_int stat) 1800{ 1801 1802 KASSERT(stat < IP_NSTATS); 1803 IP_STATINC(stat); 1804} 1805