ip_input.c revision 1.378
1/* $NetBSD: ip_input.c,v 1.378 2018/04/11 07:55:19 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 1.378 2018/04/11 07:55:19 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_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 408 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 409 ip_input(m); 410 } 411 SOFTNET_KERNEL_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 * Don't call hooks if the packet has already been processed by 574 * IPsec (encapsulated, tunnel mode). 575 */ 576#if defined(IPSEC) 577 if (!ipsec_used || !ipsec_indone(m)) 578#else 579 if (1) 580#endif 581 { 582 struct in_addr odst = ip->ip_dst; 583 bool freed; 584 585 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 586 if (freed || m == NULL) { 587 m = NULL; 588 goto out; 589 } 590 ip = mtod(m, struct ip *); 591 hlen = ip->ip_hl << 2; 592 593 /* 594 * XXX The setting of "srcrt" here is to prevent ip_forward() 595 * from generating ICMP redirects for packets that have 596 * been redirected by a hook back out on to the same LAN that 597 * they came from and is not an indication that the packet 598 * is being influenced by source routing options. This 599 * allows things like 600 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 601 * where tlp0 is both on the 1.1.1.0/24 network and is the 602 * default route for hosts on 1.1.1.0/24. Of course this 603 * also requires a "map tlp0 ..." to complete the story. 604 * One might argue whether or not this kind of network config. 605 * should be supported in this manner... 606 */ 607 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 608 } 609 610#ifdef ALTQ 611 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 612 if (altq_input) { 613 SOFTNET_LOCK(); 614 if ((*altq_input)(m, AF_INET) == 0) { 615 /* Packet dropped by traffic conditioner. */ 616 SOFTNET_UNLOCK(); 617 m = NULL; 618 goto out; 619 } 620 SOFTNET_UNLOCK(); 621 } 622#endif 623 624 /* 625 * Process options and, if not destined for us, 626 * ship it on. ip_dooptions returns 1 when an 627 * error was detected (causing an icmp message 628 * to be sent and the original packet to be freed). 629 */ 630 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 631 m = NULL; 632 goto out; 633 } 634 635 /* 636 * Check our list of addresses, to see if the packet is for us. 637 * 638 * Traditional 4.4BSD did not consult IFF_UP at all. 639 * The behavior here is to treat addresses on !IFF_UP interface 640 * or IN_IFF_NOTREADY addresses as not mine. 641 */ 642 downmatch = 0; 643 s = pserialize_read_enter(); 644 ia = ip_match_our_address(ifp, ip, &downmatch); 645 if (ia != NULL) { 646 pserialize_read_exit(s); 647 goto ours; 648 } 649 650 if (ifp->if_flags & IFF_BROADCAST) { 651 ia = ip_match_our_address_broadcast(ifp, ip); 652 if (ia != NULL) { 653 pserialize_read_exit(s); 654 goto ours; 655 } 656 } 657 pserialize_read_exit(s); 658 659 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 660#ifdef MROUTING 661 extern struct socket *ip_mrouter; 662 663 if (ip_mrouter) { 664 /* 665 * If we are acting as a multicast router, all 666 * incoming multicast packets are passed to the 667 * kernel-level multicast forwarding function. 668 * The packet is returned (relatively) intact; if 669 * ip_mforward() returns a non-zero value, the packet 670 * must be discarded, else it may be accepted below. 671 * 672 * (The IP ident field is put in the same byte order 673 * as expected when ip_mforward() is called from 674 * ip_output().) 675 */ 676 SOFTNET_LOCK(); 677 if (ip_mforward(m, ifp) != 0) { 678 SOFTNET_UNLOCK(); 679 IP_STATINC(IP_STAT_CANTFORWARD); 680 goto out; 681 } 682 SOFTNET_UNLOCK(); 683 684 /* 685 * The process-level routing demon needs to receive 686 * all multicast IGMP packets, whether or not this 687 * host belongs to their destination groups. 688 */ 689 if (ip->ip_p == IPPROTO_IGMP) { 690 goto ours; 691 } 692 IP_STATINC(IP_STAT_CANTFORWARD); 693 } 694#endif 695 /* 696 * See if we belong to the destination multicast group on the 697 * arrival interface. 698 */ 699 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 700 IP_STATINC(IP_STAT_CANTFORWARD); 701 goto out; 702 } 703 goto ours; 704 } 705 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 706 in_nullhost(ip->ip_dst)) 707 goto ours; 708 709 /* 710 * Not for us; forward if possible and desirable. 711 */ 712 if (ipforwarding == 0) { 713 m_put_rcvif_psref(ifp, &psref); 714 IP_STATINC(IP_STAT_CANTFORWARD); 715 m_freem(m); 716 } else { 717 /* 718 * If ip_dst matched any of my address on !IFF_UP interface, 719 * and there's no IFF_UP interface that matches ip_dst, 720 * send icmp unreach. Forwarding it will result in in-kernel 721 * forwarding loop till TTL goes to 0. 722 */ 723 if (downmatch) { 724 m_put_rcvif_psref(ifp, &psref); 725 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 726 IP_STATINC(IP_STAT_CANTFORWARD); 727 return; 728 } 729#ifdef IPSEC 730 /* Check the security policy (SP) for the packet */ 731 if (ipsec_used) { 732 if (ipsec4_input(m, IP_FORWARDING | 733 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 734 goto out; 735 } 736 } 737#endif 738 ip_forward(m, srcrt, ifp); 739 m_put_rcvif_psref(ifp, &psref); 740 } 741 return; 742 743ours: 744 m_put_rcvif_psref(ifp, &psref); 745 ifp = NULL; 746 747 /* 748 * If offset or IP_MF are set, must reassemble. 749 */ 750 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 751 /* 752 * Pass to IP reassembly mechanism. 753 */ 754 if (ip_reass_packet(&m, ip) != 0) { 755 /* Failed; invalid fragment(s) or packet. */ 756 goto out; 757 } 758 if (m == NULL) { 759 /* More fragments should come; silently return. */ 760 goto out; 761 } 762 /* 763 * Reassembly is done, we have the final packet. 764 * Update cached data in local variable(s). 765 */ 766 ip = mtod(m, struct ip *); 767 hlen = ip->ip_hl << 2; 768 } 769 770#ifdef IPSEC 771 /* 772 * Enforce IPsec policy checking if we are seeing last header. 773 * Note that we do not visit this with protocols with PCB layer 774 * code - like UDP/TCP/raw IP. 775 */ 776 if (ipsec_used && 777 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 778 if (ipsec4_input(m, 0) != 0) { 779 goto out; 780 } 781 } 782#endif 783 784 /* 785 * Switch out to protocol's input routine. 786 */ 787#if IFA_STATS 788 if (ia) { 789 struct in_ifaddr *_ia; 790 /* 791 * Keep a reference from ip_match_our_address with psref 792 * is expensive, so explore ia here again. 793 */ 794 s = pserialize_read_enter(); 795 _ia = in_get_ia(ip->ip_dst); 796 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 797 pserialize_read_exit(s); 798 } 799#endif 800 IP_STATINC(IP_STAT_DELIVERED); 801 802 const int off = hlen, nh = ip->ip_p; 803 804 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 805 return; 806 807out: 808 m_put_rcvif_psref(ifp, &psref); 809 if (m != NULL) 810 m_freem(m); 811} 812 813/* 814 * IP timer processing. 815 */ 816void 817ip_slowtimo(void) 818{ 819 820 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 821 822 ip_reass_slowtimo(); 823 824 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 825} 826 827/* 828 * IP drain processing. 829 */ 830void 831ip_drain(void) 832{ 833 834 KERNEL_LOCK(1, NULL); 835 ip_reass_drain(); 836 KERNEL_UNLOCK_ONE(NULL); 837} 838 839/* 840 * ip_dooptions: perform option processing on a datagram, possibly discarding 841 * it if bad options are encountered, or forwarding it if source-routed. 842 * 843 * => Returns true if packet has been forwarded/freed. 844 * => Returns false if the packet should be processed further. 845 */ 846static bool 847ip_dooptions(struct mbuf *m) 848{ 849 struct ip *ip = mtod(m, struct ip *); 850 u_char *cp, *cp0; 851 struct ip_timestamp *ipt; 852 struct in_ifaddr *ia; 853 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 854 int srr_present, rr_present, ts_present; 855 struct in_addr dst; 856 n_time ntime; 857 struct ifaddr *ifa = NULL; 858 int s; 859 860 srr_present = 0; 861 rr_present = 0; 862 ts_present = 0; 863 864 dst = ip->ip_dst; 865 cp = (u_char *)(ip + 1); 866 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 867 for (; cnt > 0; cnt -= optlen, cp += optlen) { 868 opt = cp[IPOPT_OPTVAL]; 869 if (opt == IPOPT_EOL) 870 break; 871 if (opt == IPOPT_NOP) 872 optlen = 1; 873 else { 874 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 875 code = &cp[IPOPT_OLEN] - (u_char *)ip; 876 goto bad; 877 } 878 optlen = cp[IPOPT_OLEN]; 879 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 880 code = &cp[IPOPT_OLEN] - (u_char *)ip; 881 goto bad; 882 } 883 } 884 switch (opt) { 885 886 default: 887 break; 888 889 /* 890 * Source routing with record. 891 * Find interface with current destination address. 892 * If none on this machine then drop if strictly routed, 893 * or do nothing if loosely routed. 894 * Record interface address and bring up next address 895 * component. If strictly routed make sure next 896 * address is on directly accessible net. 897 */ 898 case IPOPT_LSRR: 899 case IPOPT_SSRR: { 900 struct psref psref; 901 struct sockaddr_in ipaddr = { 902 .sin_len = sizeof(ipaddr), 903 .sin_family = AF_INET, 904 }; 905 906 if (ip_allowsrcrt == 0) { 907 type = ICMP_UNREACH; 908 code = ICMP_UNREACH_NET_PROHIB; 909 goto bad; 910 } 911 if (srr_present++) { 912 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 913 goto bad; 914 } 915 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 916 code = &cp[IPOPT_OLEN] - (u_char *)ip; 917 goto bad; 918 } 919 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 920 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 921 goto bad; 922 } 923 ipaddr.sin_addr = ip->ip_dst; 924 925 s = pserialize_read_enter(); 926 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 927 if (ifa == NULL) { 928 pserialize_read_exit(s); 929 if (opt == IPOPT_SSRR) { 930 type = ICMP_UNREACH; 931 code = ICMP_UNREACH_SRCFAIL; 932 goto bad; 933 } 934 /* 935 * Loose routing, and not at next destination 936 * yet; nothing to do except forward. 937 */ 938 break; 939 } 940 pserialize_read_exit(s); 941 942 off--; /* 0 origin */ 943 if ((off + sizeof(struct in_addr)) > optlen) { 944 /* 945 * End of source route. Should be for us. 946 */ 947 save_rte(m, cp, ip->ip_src); 948 break; 949 } 950 /* 951 * locate outgoing interface 952 */ 953 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 954 sizeof(ipaddr.sin_addr)); 955 if (opt == IPOPT_SSRR) { 956 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 957 &psref); 958 if (ifa != NULL) 959 ia = ifatoia(ifa); 960 else 961 ia = NULL; 962 } else { 963 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 964 } 965 if (ia == NULL) { 966 type = ICMP_UNREACH; 967 code = ICMP_UNREACH_SRCFAIL; 968 goto bad; 969 } 970 ip->ip_dst = ipaddr.sin_addr; 971 memcpy(cp + off, &ia->ia_addr.sin_addr, 972 sizeof(struct in_addr)); 973 ia4_release(ia, &psref); 974 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 975 /* 976 * Let ip_intr's mcast routing check handle mcast pkts 977 */ 978 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 979 break; 980 } 981 982 case IPOPT_RR: { 983 struct psref psref; 984 struct sockaddr_in ipaddr = { 985 .sin_len = sizeof(ipaddr), 986 .sin_family = AF_INET, 987 }; 988 989 if (rr_present++) { 990 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 991 goto bad; 992 } 993 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 994 code = &cp[IPOPT_OLEN] - (u_char *)ip; 995 goto bad; 996 } 997 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 998 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 999 goto bad; 1000 } 1001 /* 1002 * If no space remains, ignore. 1003 */ 1004 off--; /* 0 origin */ 1005 if ((off + sizeof(struct in_addr)) > optlen) 1006 break; 1007 memcpy((void *)&ipaddr.sin_addr, (void *)&ip->ip_dst, 1008 sizeof(ipaddr.sin_addr)); 1009 /* 1010 * locate outgoing interface; if we're the destination, 1011 * use the incoming interface (should be same). 1012 */ 1013 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1014 if (ifa == NULL) { 1015 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1016 if (ia == NULL) { 1017 type = ICMP_UNREACH; 1018 code = ICMP_UNREACH_HOST; 1019 goto bad; 1020 } 1021 } else { 1022 ia = ifatoia(ifa); 1023 } 1024 memcpy(cp + off, &ia->ia_addr.sin_addr, 1025 sizeof(struct in_addr)); 1026 ia4_release(ia, &psref); 1027 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1028 break; 1029 } 1030 1031 case IPOPT_TS: 1032 code = cp - (u_char *)ip; 1033 ipt = (struct ip_timestamp *)cp; 1034 if (ts_present++) { 1035 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1036 goto bad; 1037 } 1038 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1039 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1040 goto bad; 1041 } 1042 if (ipt->ipt_ptr < 5) { 1043 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1044 goto bad; 1045 } 1046 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1047 if (++ipt->ipt_oflw == 0) { 1048 code = (u_char *)&ipt->ipt_ptr - 1049 (u_char *)ip; 1050 goto bad; 1051 } 1052 break; 1053 } 1054 cp0 = (cp + ipt->ipt_ptr - 1); 1055 switch (ipt->ipt_flg) { 1056 1057 case IPOPT_TS_TSONLY: 1058 break; 1059 1060 case IPOPT_TS_TSANDADDR: { 1061 struct ifnet *rcvif; 1062 int _s, _ss; 1063 struct sockaddr_in ipaddr = { 1064 .sin_len = sizeof(ipaddr), 1065 .sin_family = AF_INET, 1066 }; 1067 1068 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1069 sizeof(struct in_addr) > ipt->ipt_len) { 1070 code = (u_char *)&ipt->ipt_ptr - 1071 (u_char *)ip; 1072 goto bad; 1073 } 1074 ipaddr.sin_addr = dst; 1075 _ss = pserialize_read_enter(); 1076 rcvif = m_get_rcvif(m, &_s); 1077 if (__predict_true(rcvif != NULL)) { 1078 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1079 rcvif); 1080 } 1081 m_put_rcvif(rcvif, &_s); 1082 if (ifa == NULL) { 1083 pserialize_read_exit(_ss); 1084 break; 1085 } 1086 ia = ifatoia(ifa); 1087 memcpy(cp0, &ia->ia_addr.sin_addr, 1088 sizeof(struct in_addr)); 1089 pserialize_read_exit(_ss); 1090 ipt->ipt_ptr += sizeof(struct in_addr); 1091 break; 1092 } 1093 1094 case IPOPT_TS_PRESPEC: { 1095 struct sockaddr_in ipaddr = { 1096 .sin_len = sizeof(ipaddr), 1097 .sin_family = AF_INET, 1098 }; 1099 1100 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1101 sizeof(struct in_addr) > ipt->ipt_len) { 1102 code = (u_char *)&ipt->ipt_ptr - 1103 (u_char *)ip; 1104 goto bad; 1105 } 1106 memcpy(&ipaddr.sin_addr, cp0, 1107 sizeof(struct in_addr)); 1108 s = pserialize_read_enter(); 1109 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1110 if (ifa == NULL) { 1111 pserialize_read_exit(s); 1112 continue; 1113 } 1114 pserialize_read_exit(s); 1115 ipt->ipt_ptr += sizeof(struct in_addr); 1116 break; 1117 } 1118 1119 default: 1120 /* XXX can't take &ipt->ipt_flg */ 1121 code = (u_char *)&ipt->ipt_ptr - 1122 (u_char *)ip + 1; 1123 goto bad; 1124 } 1125 ntime = iptime(); 1126 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1127 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1128 sizeof(n_time)); 1129 ipt->ipt_ptr += sizeof(n_time); 1130 } 1131 } 1132 if (forward) { 1133 struct ifnet *rcvif; 1134 struct psref _psref; 1135 1136 if (ip_forwsrcrt == 0) { 1137 type = ICMP_UNREACH; 1138 code = ICMP_UNREACH_SRCFAIL; 1139 goto bad; 1140 } 1141 1142 rcvif = m_get_rcvif_psref(m, &_psref); 1143 if (__predict_false(rcvif == NULL)) { 1144 type = ICMP_UNREACH; 1145 code = ICMP_UNREACH_HOST; 1146 goto bad; 1147 } 1148 ip_forward(m, 1, rcvif); 1149 m_put_rcvif_psref(rcvif, &_psref); 1150 return true; 1151 } 1152 return false; 1153bad: 1154 icmp_error(m, type, code, 0, 0); 1155 IP_STATINC(IP_STAT_BADOPTIONS); 1156 return true; 1157} 1158 1159/* 1160 * ip_rtaddr: given address of next destination (final or next hop), 1161 * return internet address info of interface to be used to get there. 1162 */ 1163static struct in_ifaddr * 1164ip_rtaddr(struct in_addr dst, struct psref *psref) 1165{ 1166 struct rtentry *rt; 1167 union { 1168 struct sockaddr dst; 1169 struct sockaddr_in dst4; 1170 } u; 1171 struct route *ro; 1172 1173 sockaddr_in_init(&u.dst4, &dst, 0); 1174 1175 ro = percpu_getref(ipforward_rt_percpu); 1176 rt = rtcache_lookup(ro, &u.dst); 1177 if (rt == NULL) { 1178 percpu_putref(ipforward_rt_percpu); 1179 return NULL; 1180 } 1181 1182 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1183 rtcache_unref(rt, ro); 1184 percpu_putref(ipforward_rt_percpu); 1185 1186 return ifatoia(rt->rt_ifa); 1187} 1188 1189/* 1190 * save_rte: save incoming source route for use in replies, to be picked 1191 * up later by ip_srcroute if the receiver is interested. 1192 */ 1193static void 1194save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1195{ 1196 struct ip_srcrt *isr; 1197 struct m_tag *mtag; 1198 unsigned olen; 1199 1200 olen = option[IPOPT_OLEN]; 1201 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1202 return; 1203 1204 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1205 if (mtag == NULL) 1206 return; 1207 isr = (struct ip_srcrt *)(mtag + 1); 1208 1209 memcpy(isr->isr_hdr, option, olen); 1210 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1211 isr->isr_dst = dst; 1212 m_tag_prepend(m, mtag); 1213} 1214 1215/* 1216 * Retrieve incoming source route for use in replies, 1217 * in the same form used by setsockopt. 1218 * The first hop is placed before the options, will be removed later. 1219 */ 1220struct mbuf * 1221ip_srcroute(struct mbuf *m0) 1222{ 1223 struct in_addr *p, *q; 1224 struct mbuf *m; 1225 struct ip_srcrt *isr; 1226 struct m_tag *mtag; 1227 1228 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL); 1229 if (mtag == NULL) 1230 return NULL; 1231 isr = (struct ip_srcrt *)(mtag + 1); 1232 1233 if (isr->isr_nhops == 0) 1234 return NULL; 1235 1236 m = m_get(M_DONTWAIT, MT_SOOPTS); 1237 if (m == NULL) 1238 return NULL; 1239 1240 MCLAIM(m, &inetdomain.dom_mowner); 1241#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1242 1243 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1244 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1245 1246 /* 1247 * First save first hop for return route 1248 */ 1249 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1250 *(mtod(m, struct in_addr *)) = *p--; 1251 1252 /* 1253 * Copy option fields and padding (nop) to mbuf. 1254 */ 1255 isr->isr_nop = IPOPT_NOP; 1256 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1257 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1258 OPTSIZ); 1259 q = (struct in_addr *)(mtod(m, char *) + 1260 sizeof(struct in_addr) + OPTSIZ); 1261#undef OPTSIZ 1262 /* 1263 * Record return path as an IP source route, 1264 * reversing the path (pointers are now aligned). 1265 */ 1266 while (p >= isr->isr_routes) { 1267 *q++ = *p--; 1268 } 1269 /* 1270 * Last hop goes to final destination. 1271 */ 1272 *q = isr->isr_dst; 1273 m_tag_delete(m0, mtag); 1274 return m; 1275} 1276 1277const int inetctlerrmap[PRC_NCMDS] = { 1278 [PRC_MSGSIZE] = EMSGSIZE, 1279 [PRC_HOSTDEAD] = EHOSTDOWN, 1280 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1281 [PRC_UNREACH_NET] = EHOSTUNREACH, 1282 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1283 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1284 [PRC_UNREACH_PORT] = ECONNREFUSED, 1285 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1286 [PRC_PARAMPROB] = ENOPROTOOPT, 1287}; 1288 1289void 1290ip_fasttimo(void) 1291{ 1292 if (ip_drainwanted) { 1293 ip_drain(); 1294 ip_drainwanted = 0; 1295 } 1296} 1297 1298void 1299ip_drainstub(void) 1300{ 1301 ip_drainwanted = 1; 1302} 1303 1304/* 1305 * Forward a packet. If some error occurs return the sender 1306 * an icmp packet. Note we can't always generate a meaningful 1307 * icmp message because icmp doesn't have a large enough repertoire 1308 * of codes and types. 1309 * 1310 * If not forwarding, just drop the packet. This could be confusing 1311 * if ipforwarding was zero but some routing protocol was advancing 1312 * us as a gateway to somewhere. However, we must let the routing 1313 * protocol deal with that. 1314 * 1315 * The srcrt parameter indicates whether the packet is being forwarded 1316 * via a source route. 1317 */ 1318static void 1319ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1320{ 1321 struct ip *ip = mtod(m, struct ip *); 1322 struct rtentry *rt; 1323 int error, type = 0, code = 0, destmtu = 0; 1324 struct mbuf *mcopy; 1325 n_long dest; 1326 union { 1327 struct sockaddr dst; 1328 struct sockaddr_in dst4; 1329 } u; 1330 uint64_t *ips; 1331 struct route *ro; 1332 1333 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1334 "interrupt handler; synchronization assumptions violated"); 1335 1336 /* 1337 * We are now in the output path. 1338 */ 1339 MCLAIM(m, &ip_tx_mowner); 1340 1341 /* 1342 * Clear any in-bound checksum flags for this packet. 1343 */ 1344 m->m_pkthdr.csum_flags = 0; 1345 1346 dest = 0; 1347 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1348 IP_STATINC(IP_STAT_CANTFORWARD); 1349 m_freem(m); 1350 return; 1351 } 1352 1353 if (ip->ip_ttl <= IPTTLDEC) { 1354 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1355 return; 1356 } 1357 1358 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1359 1360 ro = percpu_getref(ipforward_rt_percpu); 1361 rt = rtcache_lookup(ro, &u.dst); 1362 if (rt == NULL) { 1363 percpu_putref(ipforward_rt_percpu); 1364 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1365 return; 1366 } 1367 1368 /* 1369 * Save at most 68 bytes of the packet in case 1370 * we need to generate an ICMP message to the src. 1371 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1372 */ 1373 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1374 if (mcopy) 1375 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1376 1377 ip->ip_ttl -= IPTTLDEC; 1378 1379 /* 1380 * If forwarding packet using same interface that it came in on, 1381 * perhaps should send a redirect to sender to shortcut a hop. 1382 * Only send redirect if source is sending directly to us, 1383 * and if packet was not source routed (or has any options). 1384 * Also, don't send redirect if forwarding using a default route 1385 * or a route modified by a redirect. 1386 */ 1387 if (rt->rt_ifp == rcvif && 1388 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1389 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1390 ipsendredirects && !srcrt) { 1391 if (rt->rt_ifa && 1392 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1393 ifatoia(rt->rt_ifa)->ia_subnet) { 1394 if (rt->rt_flags & RTF_GATEWAY) 1395 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1396 else 1397 dest = ip->ip_dst.s_addr; 1398 /* 1399 * Router requirements says to only send host 1400 * redirects. 1401 */ 1402 type = ICMP_REDIRECT; 1403 code = ICMP_REDIRECT_HOST; 1404 } 1405 } 1406 rtcache_unref(rt, ro); 1407 1408 error = ip_output(m, NULL, ro, 1409 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1410 NULL, NULL); 1411 1412 if (error) { 1413 IP_STATINC(IP_STAT_CANTFORWARD); 1414 goto error; 1415 } 1416 1417 ips = IP_STAT_GETREF(); 1418 ips[IP_STAT_FORWARD]++; 1419 1420 if (type) { 1421 ips[IP_STAT_REDIRECTSENT]++; 1422 IP_STAT_PUTREF(); 1423 goto redirect; 1424 } 1425 1426 IP_STAT_PUTREF(); 1427 if (mcopy) { 1428#ifdef GATEWAY 1429 if (mcopy->m_flags & M_CANFASTFWD) 1430 ipflow_create(ro, mcopy); 1431#endif 1432 m_freem(mcopy); 1433 } 1434 1435 percpu_putref(ipforward_rt_percpu); 1436 return; 1437 1438redirect: 1439error: 1440 if (mcopy == NULL) { 1441 percpu_putref(ipforward_rt_percpu); 1442 return; 1443 } 1444 1445 switch (error) { 1446 1447 case 0: /* forwarded, but need redirect */ 1448 /* type, code set above */ 1449 break; 1450 1451 case ENETUNREACH: /* shouldn't happen, checked above */ 1452 case EHOSTUNREACH: 1453 case ENETDOWN: 1454 case EHOSTDOWN: 1455 default: 1456 type = ICMP_UNREACH; 1457 code = ICMP_UNREACH_HOST; 1458 break; 1459 1460 case EMSGSIZE: 1461 type = ICMP_UNREACH; 1462 code = ICMP_UNREACH_NEEDFRAG; 1463 1464 if ((rt = rtcache_validate(ro)) != NULL) { 1465 destmtu = rt->rt_ifp->if_mtu; 1466 rtcache_unref(rt, ro); 1467 } 1468#ifdef IPSEC 1469 if (ipsec_used) 1470 (void)ipsec4_forward(mcopy, &destmtu); 1471#endif 1472 IP_STATINC(IP_STAT_CANTFRAG); 1473 break; 1474 1475 case ENOBUFS: 1476 /* 1477 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1478 * Requirements for IP Version 4 Routers. Source quench can 1479 * be a big problem under DoS attacks or if the underlying 1480 * interface is rate-limited. 1481 */ 1482 if (mcopy) 1483 m_freem(mcopy); 1484 percpu_putref(ipforward_rt_percpu); 1485 return; 1486 } 1487 icmp_error(mcopy, type, code, dest, destmtu); 1488 percpu_putref(ipforward_rt_percpu); 1489} 1490 1491void 1492ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1493 struct mbuf *m) 1494{ 1495 struct socket *so = inp->inp_socket; 1496 int inpflags = inp->inp_flags; 1497 1498 if (SOOPT_TIMESTAMP(so->so_options)) 1499 mp = sbsavetimestamp(so->so_options, m, mp); 1500 1501 if (inpflags & INP_RECVDSTADDR) { 1502 *mp = sbcreatecontrol(&ip->ip_dst, 1503 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1504 if (*mp) 1505 mp = &(*mp)->m_next; 1506 } 1507 1508 if (inpflags & INP_RECVTTL) { 1509 *mp = sbcreatecontrol(&ip->ip_ttl, 1510 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1511 if (*mp) 1512 mp = &(*mp)->m_next; 1513 } 1514 1515 struct psref psref; 1516 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1517 if (__predict_false(ifp == NULL)) { 1518#ifdef DIAGNOSTIC 1519 printf("%s: missing receive interface\n", __func__); 1520#endif 1521 return; /* XXX should report error? */ 1522 } 1523 1524 if (inpflags & INP_RECVPKTINFO) { 1525 struct in_pktinfo ipi; 1526 ipi.ipi_addr = ip->ip_dst; 1527 ipi.ipi_ifindex = ifp->if_index; 1528 *mp = sbcreatecontrol(&ipi, 1529 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1530 if (*mp) 1531 mp = &(*mp)->m_next; 1532 } 1533 if (inpflags & INP_RECVIF) { 1534 struct sockaddr_dl sdl; 1535 1536 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1537 NULL, 0); 1538 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1539 if (*mp) 1540 mp = &(*mp)->m_next; 1541 } 1542 m_put_rcvif_psref(ifp, &psref); 1543} 1544 1545/* 1546 * sysctl helper routine for net.inet.ip.forwsrcrt. 1547 */ 1548static int 1549sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1550{ 1551 int error, tmp; 1552 struct sysctlnode node; 1553 1554 node = *rnode; 1555 tmp = ip_forwsrcrt; 1556 node.sysctl_data = &tmp; 1557 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1558 if (error || newp == NULL) 1559 return (error); 1560 1561 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1562 0, NULL, NULL, NULL); 1563 if (error) 1564 return (error); 1565 1566 ip_forwsrcrt = tmp; 1567 1568 return (0); 1569} 1570 1571/* 1572 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1573 * range of the new value and tweaks timers if it changes. 1574 */ 1575static int 1576sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1577{ 1578 int error, tmp; 1579 struct sysctlnode node; 1580 1581 icmp_mtudisc_lock(); 1582 1583 node = *rnode; 1584 tmp = ip_mtudisc_timeout; 1585 node.sysctl_data = &tmp; 1586 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1587 if (error || newp == NULL) 1588 goto out; 1589 if (tmp < 0) { 1590 error = EINVAL; 1591 goto out; 1592 } 1593 1594 ip_mtudisc_timeout = tmp; 1595 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1596 error = 0; 1597out: 1598 icmp_mtudisc_unlock(); 1599 return error; 1600} 1601 1602static int 1603sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1604{ 1605 1606 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1607} 1608 1609static void 1610sysctl_net_inet_ip_setup(struct sysctllog **clog) 1611{ 1612 sysctl_createv(clog, 0, NULL, NULL, 1613 CTLFLAG_PERMANENT, 1614 CTLTYPE_NODE, "inet", 1615 SYSCTL_DESCR("PF_INET related settings"), 1616 NULL, 0, NULL, 0, 1617 CTL_NET, PF_INET, CTL_EOL); 1618 sysctl_createv(clog, 0, NULL, NULL, 1619 CTLFLAG_PERMANENT, 1620 CTLTYPE_NODE, "ip", 1621 SYSCTL_DESCR("IPv4 related settings"), 1622 NULL, 0, NULL, 0, 1623 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1624 1625 sysctl_createv(clog, 0, NULL, NULL, 1626 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1627 CTLTYPE_INT, "forwarding", 1628 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1629 NULL, 0, &ipforwarding, 0, 1630 CTL_NET, PF_INET, IPPROTO_IP, 1631 IPCTL_FORWARDING, CTL_EOL); 1632 sysctl_createv(clog, 0, NULL, NULL, 1633 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1634 CTLTYPE_INT, "redirect", 1635 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1636 NULL, 0, &ipsendredirects, 0, 1637 CTL_NET, PF_INET, IPPROTO_IP, 1638 IPCTL_SENDREDIRECTS, CTL_EOL); 1639 sysctl_createv(clog, 0, NULL, NULL, 1640 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1641 CTLTYPE_INT, "ttl", 1642 SYSCTL_DESCR("Default TTL for an INET datagram"), 1643 NULL, 0, &ip_defttl, 0, 1644 CTL_NET, PF_INET, IPPROTO_IP, 1645 IPCTL_DEFTTL, CTL_EOL); 1646 sysctl_createv(clog, 0, NULL, NULL, 1647 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1648 CTLTYPE_INT, "forwsrcrt", 1649 SYSCTL_DESCR("Enable forwarding of source-routed " 1650 "datagrams"), 1651 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1652 CTL_NET, PF_INET, IPPROTO_IP, 1653 IPCTL_FORWSRCRT, CTL_EOL); 1654 sysctl_createv(clog, 0, NULL, NULL, 1655 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1656 CTLTYPE_INT, "directed-broadcast", 1657 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1658 NULL, 0, &ip_directedbcast, 0, 1659 CTL_NET, PF_INET, IPPROTO_IP, 1660 IPCTL_DIRECTEDBCAST, CTL_EOL); 1661 sysctl_createv(clog, 0, NULL, NULL, 1662 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1663 CTLTYPE_INT, "allowsrcrt", 1664 SYSCTL_DESCR("Accept source-routed datagrams"), 1665 NULL, 0, &ip_allowsrcrt, 0, 1666 CTL_NET, PF_INET, IPPROTO_IP, 1667 IPCTL_ALLOWSRCRT, CTL_EOL); 1668 1669 sysctl_createv(clog, 0, NULL, NULL, 1670 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1671 CTLTYPE_INT, "mtudisc", 1672 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1673 NULL, 0, &ip_mtudisc, 0, 1674 CTL_NET, PF_INET, IPPROTO_IP, 1675 IPCTL_MTUDISC, CTL_EOL); 1676 sysctl_createv(clog, 0, NULL, NULL, 1677 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1678 CTLTYPE_INT, "anonportmin", 1679 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1680 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1681 CTL_NET, PF_INET, IPPROTO_IP, 1682 IPCTL_ANONPORTMIN, CTL_EOL); 1683 sysctl_createv(clog, 0, NULL, NULL, 1684 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1685 CTLTYPE_INT, "anonportmax", 1686 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1687 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1688 CTL_NET, PF_INET, IPPROTO_IP, 1689 IPCTL_ANONPORTMAX, CTL_EOL); 1690 sysctl_createv(clog, 0, NULL, NULL, 1691 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1692 CTLTYPE_INT, "mtudisctimeout", 1693 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1694 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1695 CTL_NET, PF_INET, IPPROTO_IP, 1696 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1697#ifndef IPNOPRIVPORTS 1698 sysctl_createv(clog, 0, NULL, NULL, 1699 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1700 CTLTYPE_INT, "lowportmin", 1701 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1702 "to assign"), 1703 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1704 CTL_NET, PF_INET, IPPROTO_IP, 1705 IPCTL_LOWPORTMIN, CTL_EOL); 1706 sysctl_createv(clog, 0, NULL, NULL, 1707 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1708 CTLTYPE_INT, "lowportmax", 1709 SYSCTL_DESCR("Highest privileged ephemeral port number " 1710 "to assign"), 1711 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1712 CTL_NET, PF_INET, IPPROTO_IP, 1713 IPCTL_LOWPORTMAX, CTL_EOL); 1714#endif /* IPNOPRIVPORTS */ 1715#if NGRE > 0 1716 sysctl_createv(clog, 0, NULL, NULL, 1717 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1718 CTLTYPE_INT, "grettl", 1719 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1720 NULL, 0, &ip_gre_ttl, 0, 1721 CTL_NET, PF_INET, IPPROTO_IP, 1722 IPCTL_GRE_TTL, CTL_EOL); 1723#endif /* NGRE */ 1724 sysctl_createv(clog, 0, NULL, NULL, 1725 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1726 CTLTYPE_INT, "checkinterface", 1727 SYSCTL_DESCR("Enable receive side of Strong ES model " 1728 "from RFC1122"), 1729 NULL, 0, &ip_checkinterface, 0, 1730 CTL_NET, PF_INET, IPPROTO_IP, 1731 IPCTL_CHECKINTERFACE, CTL_EOL); 1732 sysctl_createv(clog, 0, NULL, NULL, 1733 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1734 CTLTYPE_INT, "random_id", 1735 SYSCTL_DESCR("Assign random ip_id values"), 1736 NULL, 0, &ip_do_randomid, 0, 1737 CTL_NET, PF_INET, IPPROTO_IP, 1738 IPCTL_RANDOMID, CTL_EOL); 1739 sysctl_createv(clog, 0, NULL, NULL, 1740 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1741 CTLTYPE_INT, "do_loopback_cksum", 1742 SYSCTL_DESCR("Perform IP checksum on loopback"), 1743 NULL, 0, &ip_do_loopback_cksum, 0, 1744 CTL_NET, PF_INET, IPPROTO_IP, 1745 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1746 sysctl_createv(clog, 0, NULL, NULL, 1747 CTLFLAG_PERMANENT, 1748 CTLTYPE_STRUCT, "stats", 1749 SYSCTL_DESCR("IP statistics"), 1750 sysctl_net_inet_ip_stats, 0, NULL, 0, 1751 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1752 CTL_EOL); 1753#if NARP 1754 sysctl_createv(clog, 0, NULL, NULL, 1755 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1756 CTLTYPE_INT, "dad_count", 1757 SYSCTL_DESCR("Number of Duplicate Address Detection " 1758 "probes to send"), 1759 NULL, 0, &ip_dad_count, 0, 1760 CTL_NET, PF_INET, IPPROTO_IP, 1761 IPCTL_DAD_COUNT, CTL_EOL); 1762#endif 1763 1764 /* anonportalgo RFC6056 subtree */ 1765 const struct sysctlnode *portalgo_node; 1766 sysctl_createv(clog, 0, NULL, &portalgo_node, 1767 CTLFLAG_PERMANENT, 1768 CTLTYPE_NODE, "anonportalgo", 1769 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1770 NULL, 0, NULL, 0, 1771 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1772 sysctl_createv(clog, 0, &portalgo_node, NULL, 1773 CTLFLAG_PERMANENT, 1774 CTLTYPE_STRING, "available", 1775 SYSCTL_DESCR("available algorithms"), 1776 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1777 CTL_CREATE, CTL_EOL); 1778 sysctl_createv(clog, 0, &portalgo_node, NULL, 1779 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1780 CTLTYPE_STRING, "selected", 1781 SYSCTL_DESCR("selected algorithm"), 1782 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1783 CTL_CREATE, CTL_EOL); 1784 sysctl_createv(clog, 0, &portalgo_node, NULL, 1785 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1786 CTLTYPE_STRUCT, "reserve", 1787 SYSCTL_DESCR("bitmap of reserved ports"), 1788 sysctl_portalgo_reserve4, 0, NULL, 0, 1789 CTL_CREATE, CTL_EOL); 1790} 1791 1792void 1793ip_statinc(u_int stat) 1794{ 1795 1796 KASSERT(stat < IP_NSTATS); 1797 IP_STATINC(stat); 1798} 1799