ip_input.c revision 1.388
1/* $NetBSD: ip_input.c,v 1.388 2019/01/17 02:47:15 knakahara 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 1.388 2019/01/17 02:47:15 knakahara 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 NAT 180 * to translate the destination address to another local interface. 181 * 182 * XXX - ip_checkinterface must be disabled if you add IP aliases 183 * to the loopback interface instead of the interface where the 184 * packets for those addresses are received. 185 */ 186static int ip_checkinterface __read_mostly = 0; 187 188struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 189 190pktqueue_t * ip_pktq __read_mostly; 191pfil_head_t * inet_pfil_hook __read_mostly; 192ipid_state_t * ip_ids __read_mostly; 193percpu_t * ipstat_percpu __read_mostly; 194 195static percpu_t *ipforward_rt_percpu __cacheline_aligned; 196 197uint16_t ip_id; 198 199#ifdef INET_CSUM_COUNTERS 200#include <sys/device.h> 201 202struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 203 NULL, "inet", "hwcsum bad"); 204struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 205 NULL, "inet", "hwcsum ok"); 206struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 207 NULL, "inet", "swcsum"); 208 209#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 210 211EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 212EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 213EVCNT_ATTACH_STATIC(ip_swcsum); 214 215#else 216 217#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 218 219#endif /* INET_CSUM_COUNTERS */ 220 221/* 222 * Used to save the IP options in case a protocol wants to respond 223 * to an incoming packet over the same route if the packet got here 224 * using IP source routing. This allows connection establishment and 225 * maintenance when the remote end is on a network that is not known 226 * to us. 227 */ 228struct ip_srcrt { 229 int isr_nhops; /* number of hops */ 230 struct in_addr isr_dst; /* final destination */ 231 char isr_nop; /* one NOP to align */ 232 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */ 233 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)]; 234}; 235 236static int ip_drainwanted; 237 238static void save_rte(struct mbuf *, u_char *, struct in_addr); 239 240#ifdef MBUFTRACE 241struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 242struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 243#endif 244 245static void ipintr(void *); 246static void ip_input(struct mbuf *); 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 - We need to add a per ifaddr flag for this so that 328 * we get finer grain control. 329 */ 330 checkif = ip_checkinterface && (ipforwarding == 0) && 331 (ifp->if_flags & IFF_LOOPBACK) == 0; 332 333 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 334 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 335 if (ia->ia4_flags & IN_IFF_NOTREADY) 336 continue; 337 if (checkif && ia->ia_ifp != ifp) 338 continue; 339 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) { 340 (*downmatch)++; 341 continue; 342 } 343 if (ia->ia4_flags & IN_IFF_DETACHED && 344 (ifp->if_flags & IFF_LOOPBACK) == 0) 345 continue; 346 break; 347 } 348 } 349 350 return ia; 351} 352 353static struct in_ifaddr * 354ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 355{ 356 struct in_ifaddr *ia = NULL; 357 struct ifaddr *ifa; 358 359 IFADDR_READER_FOREACH(ifa, ifp) { 360 if (ifa->ifa_addr->sa_family != AF_INET) 361 continue; 362 ia = ifatoia(ifa); 363 if (ia->ia4_flags & IN_IFF_NOTREADY) 364 continue; 365 if (ia->ia4_flags & IN_IFF_DETACHED && 366 (ifp->if_flags & IFF_LOOPBACK) == 0) 367 continue; 368 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 369 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 370 /* 371 * Look for all-0's host part (old broadcast addr), 372 * either for subnet or net. 373 */ 374 ip->ip_dst.s_addr == ia->ia_subnet || 375 ip->ip_dst.s_addr == ia->ia_net) 376 goto matched; 377 /* 378 * An interface with IP address zero accepts 379 * all packets that arrive on that interface. 380 */ 381 if (in_nullhost(ia->ia_addr.sin_addr)) 382 goto matched; 383 } 384 ia = NULL; 385 386matched: 387 return ia; 388} 389 390/* 391 * IP software interrupt routine. 392 */ 393static void 394ipintr(void *arg __unused) 395{ 396 struct mbuf *m; 397 398 KASSERT(cpu_softintr_p()); 399 400 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 401 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 402 ip_input(m); 403 } 404 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 405} 406 407/* 408 * IP input routine. Checksum and byte swap header. If fragmented 409 * try to reassemble. Process options. Pass to next level. 410 */ 411static void 412ip_input(struct mbuf *m) 413{ 414 struct ip *ip = NULL; 415 struct in_ifaddr *ia = NULL; 416 int hlen = 0, len; 417 int downmatch; 418 int srcrt = 0; 419 ifnet_t *ifp; 420 struct psref psref; 421 int s; 422 423 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 424 "interrupt handler; synchronization assumptions violated"); 425 426 MCLAIM(m, &ip_rx_mowner); 427 KASSERT((m->m_flags & M_PKTHDR) != 0); 428 429 ifp = m_get_rcvif_psref(m, &psref); 430 if (__predict_false(ifp == NULL)) 431 goto out; 432 433 /* 434 * If no IP addresses have been set yet but the interfaces 435 * are receiving, can't do anything with incoming packets yet. 436 * Note: we pre-check without locks held. 437 */ 438 if (IN_ADDRLIST_READER_EMPTY()) 439 goto out; 440 IP_STATINC(IP_STAT_TOTAL); 441 442 /* 443 * If the IP header is not aligned, slurp it up into a new 444 * mbuf with space for link headers, in the event we forward 445 * it. Otherwise, if it is aligned, make sure the entire 446 * base IP header is in the first mbuf of the chain. 447 */ 448 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 449 if ((m = m_copyup(m, sizeof(struct ip), 450 (max_linkhdr + 3) & ~3)) == NULL) { 451 /* XXXJRT new stat, please */ 452 IP_STATINC(IP_STAT_TOOSMALL); 453 goto out; 454 } 455 } else if (__predict_false(m->m_len < sizeof(struct ip))) { 456 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) { 457 IP_STATINC(IP_STAT_TOOSMALL); 458 goto out; 459 } 460 } 461 ip = mtod(m, struct ip *); 462 if (ip->ip_v != IPVERSION) { 463 IP_STATINC(IP_STAT_BADVERS); 464 goto out; 465 } 466 hlen = ip->ip_hl << 2; 467 if (hlen < sizeof(struct ip)) { /* minimum header length */ 468 IP_STATINC(IP_STAT_BADHLEN); 469 goto out; 470 } 471 if (hlen > m->m_len) { 472 if ((m = m_pullup(m, hlen)) == NULL) { 473 IP_STATINC(IP_STAT_BADHLEN); 474 goto out; 475 } 476 ip = mtod(m, struct ip *); 477 } 478 479 /* 480 * RFC1122: packets with a multicast source address are 481 * not allowed. 482 */ 483 if (IN_MULTICAST(ip->ip_src.s_addr)) { 484 IP_STATINC(IP_STAT_BADADDR); 485 goto out; 486 } 487 488 /* 127/8 must not appear on wire - RFC1122 */ 489 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 490 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 491 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 492 IP_STATINC(IP_STAT_BADADDR); 493 goto out; 494 } 495 } 496 497 switch (m->m_pkthdr.csum_flags & 498 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) { 499 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 500 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 501 IP_STATINC(IP_STAT_BADSUM); 502 goto out; 503 504 case M_CSUM_IPv4: 505 /* Checksum was okay. */ 506 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 507 break; 508 509 default: 510 /* 511 * Must compute it ourselves. Maybe skip checksum on 512 * loopback interfaces. 513 */ 514 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 515 ip_do_loopback_cksum)) { 516 INET_CSUM_COUNTER_INCR(&ip_swcsum); 517 if (in_cksum(m, hlen) != 0) { 518 IP_STATINC(IP_STAT_BADSUM); 519 goto out; 520 } 521 } 522 break; 523 } 524 525 /* Retrieve the packet length. */ 526 len = ntohs(ip->ip_len); 527 528 /* 529 * Check for additional length bogosity 530 */ 531 if (len < hlen) { 532 IP_STATINC(IP_STAT_BADLEN); 533 goto out; 534 } 535 536 /* 537 * Check that the amount of data in the buffers is at least as much 538 * as the IP header would have us expect. Trim mbufs if longer than 539 * we expect. Drop packet if shorter than we expect. 540 */ 541 if (m->m_pkthdr.len < len) { 542 IP_STATINC(IP_STAT_TOOSHORT); 543 goto out; 544 } 545 if (m->m_pkthdr.len > len) { 546 if (m->m_len == m->m_pkthdr.len) { 547 m->m_len = len; 548 m->m_pkthdr.len = len; 549 } else 550 m_adj(m, len - m->m_pkthdr.len); 551 } 552 553 /* 554 * Assume that we can create a fast-forward IP flow entry 555 * based on this packet. 556 */ 557 m->m_flags |= M_CANFASTFWD; 558 559 /* 560 * Run through list of hooks for input packets. If there are any 561 * filters which require that additional packets in the flow are 562 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 563 * Note that filters must _never_ set this flag, as another filter 564 * in the list may have previously cleared it. 565 * 566 * Don't call hooks if the packet has already been processed by 567 * IPsec (encapsulated, tunnel mode). 568 */ 569#if defined(IPSEC) 570 if (!ipsec_used || !ipsec_skip_pfil(m)) 571#else 572 if (1) 573#endif 574 { 575 struct in_addr odst = ip->ip_dst; 576 bool freed; 577 578 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 579 if (freed || m == NULL) { 580 m = NULL; 581 goto out; 582 } 583 KASSERT(m->m_len >= sizeof(struct ip)); 584 ip = mtod(m, struct ip *); 585 hlen = ip->ip_hl << 2; 586 KASSERT(m->m_len >= hlen); 587 588 /* 589 * XXX The setting of "srcrt" here is to prevent ip_forward() 590 * from generating ICMP redirects for packets that have 591 * been redirected by a hook back out on to the same LAN that 592 * they came from and is not an indication that the packet 593 * is being influenced by source routing options. This 594 * allows things like 595 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 596 * where tlp0 is both on the 1.1.1.0/24 network and is the 597 * default route for hosts on 1.1.1.0/24. Of course this 598 * also requires a "map tlp0 ..." to complete the story. 599 * One might argue whether or not this kind of network config. 600 * should be supported in this manner... 601 */ 602 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 603 } 604 605#ifdef ALTQ 606 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 607 if (altq_input) { 608 SOFTNET_LOCK(); 609 if ((*altq_input)(m, AF_INET) == 0) { 610 /* Packet dropped by traffic conditioner. */ 611 SOFTNET_UNLOCK(); 612 m = NULL; 613 goto out; 614 } 615 SOFTNET_UNLOCK(); 616 } 617#endif 618 619 /* 620 * Process options and, if not destined for us, 621 * ship it on. ip_dooptions returns 1 when an 622 * error was detected (causing an icmp message 623 * to be sent and the original packet to be freed). 624 */ 625 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 626 m = NULL; 627 goto out; 628 } 629 630 /* 631 * Check our list of addresses, to see if the packet is for us. 632 * 633 * Traditional 4.4BSD did not consult IFF_UP at all. 634 * The behavior here is to treat addresses on !IFF_UP interface 635 * or IN_IFF_NOTREADY addresses as not mine. 636 */ 637 downmatch = 0; 638 s = pserialize_read_enter(); 639 ia = ip_match_our_address(ifp, ip, &downmatch); 640 if (ia != NULL) { 641 pserialize_read_exit(s); 642 goto ours; 643 } 644 645 if (ifp->if_flags & IFF_BROADCAST) { 646 ia = ip_match_our_address_broadcast(ifp, ip); 647 if (ia != NULL) { 648 pserialize_read_exit(s); 649 goto ours; 650 } 651 } 652 pserialize_read_exit(s); 653 654 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 655#ifdef MROUTING 656 extern struct socket *ip_mrouter; 657 658 if (ip_mrouter) { 659 /* 660 * If we are acting as a multicast router, all 661 * incoming multicast packets are passed to the 662 * kernel-level multicast forwarding function. 663 * The packet is returned (relatively) intact; if 664 * ip_mforward() returns a non-zero value, the packet 665 * must be discarded, else it may be accepted below. 666 * 667 * (The IP ident field is put in the same byte order 668 * as expected when ip_mforward() is called from 669 * ip_output().) 670 */ 671 SOFTNET_LOCK(); 672 if (ip_mforward(m, ifp) != 0) { 673 SOFTNET_UNLOCK(); 674 IP_STATINC(IP_STAT_CANTFORWARD); 675 goto out; 676 } 677 SOFTNET_UNLOCK(); 678 679 /* 680 * The process-level routing demon needs to receive 681 * all multicast IGMP packets, whether or not this 682 * host belongs to their destination groups. 683 */ 684 if (ip->ip_p == IPPROTO_IGMP) { 685 goto ours; 686 } 687 IP_STATINC(IP_STAT_CANTFORWARD); 688 } 689#endif 690 /* 691 * See if we belong to the destination multicast group on the 692 * arrival interface. 693 */ 694 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 695 IP_STATINC(IP_STAT_CANTFORWARD); 696 goto out; 697 } 698 goto ours; 699 } 700 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 701 in_nullhost(ip->ip_dst)) 702 goto ours; 703 704 /* 705 * Not for us; forward if possible and desirable. 706 */ 707 if (ipforwarding == 0) { 708 m_put_rcvif_psref(ifp, &psref); 709 IP_STATINC(IP_STAT_CANTFORWARD); 710 m_freem(m); 711 } else { 712 /* 713 * If ip_dst matched any of my address on !IFF_UP interface, 714 * and there's no IFF_UP interface that matches ip_dst, 715 * send icmp unreach. Forwarding it will result in in-kernel 716 * forwarding loop till TTL goes to 0. 717 */ 718 if (downmatch) { 719 m_put_rcvif_psref(ifp, &psref); 720 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 721 IP_STATINC(IP_STAT_CANTFORWARD); 722 return; 723 } 724#ifdef IPSEC 725 /* Check the security policy (SP) for the packet */ 726 if (ipsec_used) { 727 if (ipsec_ip_input(m, true) != 0) { 728 goto out; 729 } 730 } 731#endif 732 ip_forward(m, srcrt, ifp); 733 m_put_rcvif_psref(ifp, &psref); 734 } 735 return; 736 737ours: 738 m_put_rcvif_psref(ifp, &psref); 739 ifp = NULL; 740 741 /* 742 * If offset or IP_MF are set, must reassemble. 743 */ 744 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 745 /* 746 * Pass to IP reassembly mechanism. 747 */ 748 if (ip_reass_packet(&m) != 0) { 749 /* Failed; invalid fragment(s) or packet. */ 750 goto out; 751 } 752 if (m == NULL) { 753 /* More fragments should come; silently return. */ 754 goto out; 755 } 756 /* 757 * Reassembly is done, we have the final packet. 758 * Update cached data in local variable(s). 759 */ 760 ip = mtod(m, struct ip *); 761 hlen = ip->ip_hl << 2; 762 } 763 764 M_VERIFY_PACKET(m); 765 766#ifdef IPSEC 767 /* 768 * Enforce IPsec policy checking if we are seeing last header. 769 * Note that we do not visit this with protocols with PCB layer 770 * code - like UDP/TCP/raw IP. 771 */ 772 if (ipsec_used && 773 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 774 if (ipsec_ip_input(m, false) != 0) { 775 goto out; 776 } 777 } 778#endif 779 780 /* 781 * Switch out to protocol's input routine. 782 */ 783#if IFA_STATS 784 if (ia) { 785 struct in_ifaddr *_ia; 786 /* 787 * Keep a reference from ip_match_our_address with psref 788 * is expensive, so explore ia here again. 789 */ 790 s = pserialize_read_enter(); 791 _ia = in_get_ia(ip->ip_dst); 792 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 793 pserialize_read_exit(s); 794 } 795#endif 796 IP_STATINC(IP_STAT_DELIVERED); 797 798 const int off = hlen, nh = ip->ip_p; 799 800 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 801 return; 802 803out: 804 m_put_rcvif_psref(ifp, &psref); 805 if (m != NULL) 806 m_freem(m); 807} 808 809/* 810 * IP timer processing. 811 */ 812void 813ip_slowtimo(void) 814{ 815 816 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 817 818 ip_reass_slowtimo(); 819 820 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 821} 822 823/* 824 * IP drain processing. 825 */ 826void 827ip_drain(void) 828{ 829 830 KERNEL_LOCK(1, NULL); 831 ip_reass_drain(); 832 KERNEL_UNLOCK_ONE(NULL); 833} 834 835/* 836 * ip_dooptions: perform option processing on a datagram, possibly discarding 837 * it if bad options are encountered, or forwarding it if source-routed. 838 * 839 * => Returns true if packet has been forwarded/freed. 840 * => Returns false if the packet should be processed further. 841 */ 842static bool 843ip_dooptions(struct mbuf *m) 844{ 845 struct ip *ip = mtod(m, struct ip *); 846 u_char *cp, *cp0; 847 struct ip_timestamp *ipt; 848 struct in_ifaddr *ia; 849 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 850 int srr_present, rr_present, ts_present; 851 struct in_addr dst; 852 n_time ntime; 853 struct ifaddr *ifa = NULL; 854 int s; 855 856 srr_present = 0; 857 rr_present = 0; 858 ts_present = 0; 859 860 dst = ip->ip_dst; 861 cp = (u_char *)(ip + 1); 862 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 863 for (; cnt > 0; cnt -= optlen, cp += optlen) { 864 opt = cp[IPOPT_OPTVAL]; 865 if (opt == IPOPT_EOL) 866 break; 867 if (opt == IPOPT_NOP) 868 optlen = 1; 869 else { 870 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 871 code = &cp[IPOPT_OLEN] - (u_char *)ip; 872 goto bad; 873 } 874 optlen = cp[IPOPT_OLEN]; 875 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 876 code = &cp[IPOPT_OLEN] - (u_char *)ip; 877 goto bad; 878 } 879 } 880 switch (opt) { 881 882 default: 883 break; 884 885 /* 886 * Source routing with record. 887 * Find interface with current destination address. 888 * If none on this machine then drop if strictly routed, 889 * or do nothing if loosely routed. 890 * Record interface address and bring up next address 891 * component. If strictly routed make sure next 892 * address is on directly accessible net. 893 */ 894 case IPOPT_LSRR: 895 case IPOPT_SSRR: { 896 struct psref psref; 897 struct sockaddr_in ipaddr = { 898 .sin_len = sizeof(ipaddr), 899 .sin_family = AF_INET, 900 }; 901 902 if (ip_allowsrcrt == 0) { 903 type = ICMP_UNREACH; 904 code = ICMP_UNREACH_NET_PROHIB; 905 goto bad; 906 } 907 if (srr_present++) { 908 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 909 goto bad; 910 } 911 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 912 code = &cp[IPOPT_OLEN] - (u_char *)ip; 913 goto bad; 914 } 915 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 916 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 917 goto bad; 918 } 919 ipaddr.sin_addr = ip->ip_dst; 920 921 s = pserialize_read_enter(); 922 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 923 if (ifa == NULL) { 924 pserialize_read_exit(s); 925 if (opt == IPOPT_SSRR) { 926 type = ICMP_UNREACH; 927 code = ICMP_UNREACH_SRCFAIL; 928 goto bad; 929 } 930 /* 931 * Loose routing, and not at next destination 932 * yet; nothing to do except forward. 933 */ 934 break; 935 } 936 pserialize_read_exit(s); 937 938 off--; /* 0 origin */ 939 if ((off + sizeof(struct in_addr)) > optlen) { 940 /* 941 * End of source route. Should be for us. 942 */ 943 save_rte(m, cp, ip->ip_src); 944 break; 945 } 946 /* 947 * locate outgoing interface 948 */ 949 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 950 sizeof(ipaddr.sin_addr)); 951 if (opt == IPOPT_SSRR) { 952 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 953 &psref); 954 if (ifa != NULL) 955 ia = ifatoia(ifa); 956 else 957 ia = NULL; 958 } else { 959 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 960 } 961 if (ia == NULL) { 962 type = ICMP_UNREACH; 963 code = ICMP_UNREACH_SRCFAIL; 964 goto bad; 965 } 966 ip->ip_dst = ipaddr.sin_addr; 967 memcpy(cp + off, &ia->ia_addr.sin_addr, 968 sizeof(struct in_addr)); 969 ia4_release(ia, &psref); 970 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 971 /* 972 * Let ip_intr's mcast routing check handle mcast pkts 973 */ 974 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 975 break; 976 } 977 978 case IPOPT_RR: { 979 struct psref psref; 980 struct sockaddr_in ipaddr = { 981 .sin_len = sizeof(ipaddr), 982 .sin_family = AF_INET, 983 }; 984 985 if (rr_present++) { 986 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 987 goto bad; 988 } 989 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 990 code = &cp[IPOPT_OLEN] - (u_char *)ip; 991 goto bad; 992 } 993 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 994 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 995 goto bad; 996 } 997 /* 998 * If no space remains, ignore. 999 */ 1000 off--; /* 0 origin */ 1001 if ((off + sizeof(struct in_addr)) > optlen) 1002 break; 1003 memcpy((void *)&ipaddr.sin_addr, (void *)&ip->ip_dst, 1004 sizeof(ipaddr.sin_addr)); 1005 /* 1006 * locate outgoing interface; if we're the destination, 1007 * use the incoming interface (should be same). 1008 */ 1009 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1010 if (ifa == NULL) { 1011 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1012 if (ia == NULL) { 1013 type = ICMP_UNREACH; 1014 code = ICMP_UNREACH_HOST; 1015 goto bad; 1016 } 1017 } else { 1018 ia = ifatoia(ifa); 1019 } 1020 memcpy(cp + off, &ia->ia_addr.sin_addr, 1021 sizeof(struct in_addr)); 1022 ia4_release(ia, &psref); 1023 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1024 break; 1025 } 1026 1027 case IPOPT_TS: 1028 code = cp - (u_char *)ip; 1029 ipt = (struct ip_timestamp *)cp; 1030 if (ts_present++) { 1031 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1032 goto bad; 1033 } 1034 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1035 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1036 goto bad; 1037 } 1038 if (ipt->ipt_ptr < 5) { 1039 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1040 goto bad; 1041 } 1042 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1043 if (++ipt->ipt_oflw == 0) { 1044 code = (u_char *)&ipt->ipt_ptr - 1045 (u_char *)ip; 1046 goto bad; 1047 } 1048 break; 1049 } 1050 cp0 = (cp + ipt->ipt_ptr - 1); 1051 switch (ipt->ipt_flg) { 1052 1053 case IPOPT_TS_TSONLY: 1054 break; 1055 1056 case IPOPT_TS_TSANDADDR: { 1057 struct ifnet *rcvif; 1058 int _s, _ss; 1059 struct sockaddr_in ipaddr = { 1060 .sin_len = sizeof(ipaddr), 1061 .sin_family = AF_INET, 1062 }; 1063 1064 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1065 sizeof(struct in_addr) > ipt->ipt_len) { 1066 code = (u_char *)&ipt->ipt_ptr - 1067 (u_char *)ip; 1068 goto bad; 1069 } 1070 ipaddr.sin_addr = dst; 1071 _ss = pserialize_read_enter(); 1072 rcvif = m_get_rcvif(m, &_s); 1073 if (__predict_true(rcvif != NULL)) { 1074 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1075 rcvif); 1076 } 1077 m_put_rcvif(rcvif, &_s); 1078 if (ifa == NULL) { 1079 pserialize_read_exit(_ss); 1080 break; 1081 } 1082 ia = ifatoia(ifa); 1083 memcpy(cp0, &ia->ia_addr.sin_addr, 1084 sizeof(struct in_addr)); 1085 pserialize_read_exit(_ss); 1086 ipt->ipt_ptr += sizeof(struct in_addr); 1087 break; 1088 } 1089 1090 case IPOPT_TS_PRESPEC: { 1091 struct sockaddr_in ipaddr = { 1092 .sin_len = sizeof(ipaddr), 1093 .sin_family = AF_INET, 1094 }; 1095 1096 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1097 sizeof(struct in_addr) > ipt->ipt_len) { 1098 code = (u_char *)&ipt->ipt_ptr - 1099 (u_char *)ip; 1100 goto bad; 1101 } 1102 memcpy(&ipaddr.sin_addr, cp0, 1103 sizeof(struct in_addr)); 1104 s = pserialize_read_enter(); 1105 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1106 if (ifa == NULL) { 1107 pserialize_read_exit(s); 1108 continue; 1109 } 1110 pserialize_read_exit(s); 1111 ipt->ipt_ptr += sizeof(struct in_addr); 1112 break; 1113 } 1114 1115 default: 1116 /* XXX can't take &ipt->ipt_flg */ 1117 code = (u_char *)&ipt->ipt_ptr - 1118 (u_char *)ip + 1; 1119 goto bad; 1120 } 1121 ntime = iptime(); 1122 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1123 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1124 sizeof(n_time)); 1125 ipt->ipt_ptr += sizeof(n_time); 1126 } 1127 } 1128 if (forward) { 1129 struct ifnet *rcvif; 1130 struct psref _psref; 1131 1132 if (ip_forwsrcrt == 0) { 1133 type = ICMP_UNREACH; 1134 code = ICMP_UNREACH_SRCFAIL; 1135 goto bad; 1136 } 1137 1138 rcvif = m_get_rcvif_psref(m, &_psref); 1139 if (__predict_false(rcvif == NULL)) { 1140 type = ICMP_UNREACH; 1141 code = ICMP_UNREACH_HOST; 1142 goto bad; 1143 } 1144 ip_forward(m, 1, rcvif); 1145 m_put_rcvif_psref(rcvif, &_psref); 1146 return true; 1147 } 1148 return false; 1149bad: 1150 icmp_error(m, type, code, 0, 0); 1151 IP_STATINC(IP_STAT_BADOPTIONS); 1152 return true; 1153} 1154 1155/* 1156 * ip_rtaddr: given address of next destination (final or next hop), 1157 * return internet address info of interface to be used to get there. 1158 */ 1159static struct in_ifaddr * 1160ip_rtaddr(struct in_addr dst, struct psref *psref) 1161{ 1162 struct rtentry *rt; 1163 union { 1164 struct sockaddr dst; 1165 struct sockaddr_in dst4; 1166 } u; 1167 struct route *ro; 1168 1169 sockaddr_in_init(&u.dst4, &dst, 0); 1170 1171 ro = percpu_getref(ipforward_rt_percpu); 1172 rt = rtcache_lookup(ro, &u.dst); 1173 if (rt == NULL) { 1174 percpu_putref(ipforward_rt_percpu); 1175 return NULL; 1176 } 1177 1178 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1179 rtcache_unref(rt, ro); 1180 percpu_putref(ipforward_rt_percpu); 1181 1182 return ifatoia(rt->rt_ifa); 1183} 1184 1185/* 1186 * save_rte: save incoming source route for use in replies, to be picked 1187 * up later by ip_srcroute if the receiver is interested. 1188 */ 1189static void 1190save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1191{ 1192 struct ip_srcrt *isr; 1193 struct m_tag *mtag; 1194 unsigned olen; 1195 1196 olen = option[IPOPT_OLEN]; 1197 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1198 return; 1199 1200 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1201 if (mtag == NULL) 1202 return; 1203 isr = (struct ip_srcrt *)(mtag + 1); 1204 1205 memcpy(isr->isr_hdr, option, olen); 1206 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1207 isr->isr_dst = dst; 1208 m_tag_prepend(m, mtag); 1209} 1210 1211/* 1212 * Retrieve incoming source route for use in replies, 1213 * in the same form used by setsockopt. 1214 * The first hop is placed before the options, will be removed later. 1215 */ 1216struct mbuf * 1217ip_srcroute(struct mbuf *m0) 1218{ 1219 struct in_addr *p, *q; 1220 struct mbuf *m; 1221 struct ip_srcrt *isr; 1222 struct m_tag *mtag; 1223 1224 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE); 1225 if (mtag == NULL) 1226 return NULL; 1227 isr = (struct ip_srcrt *)(mtag + 1); 1228 1229 if (isr->isr_nhops == 0) 1230 return NULL; 1231 1232 m = m_get(M_DONTWAIT, MT_SOOPTS); 1233 if (m == NULL) 1234 return NULL; 1235 1236 MCLAIM(m, &inetdomain.dom_mowner); 1237#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1238 1239 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1240 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1241 1242 /* 1243 * First save first hop for return route 1244 */ 1245 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1246 *(mtod(m, struct in_addr *)) = *p--; 1247 1248 /* 1249 * Copy option fields and padding (nop) to mbuf. 1250 */ 1251 isr->isr_nop = IPOPT_NOP; 1252 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1253 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1254 OPTSIZ); 1255 q = (struct in_addr *)(mtod(m, char *) + 1256 sizeof(struct in_addr) + OPTSIZ); 1257#undef OPTSIZ 1258 /* 1259 * Record return path as an IP source route, 1260 * reversing the path (pointers are now aligned). 1261 */ 1262 while (p >= isr->isr_routes) { 1263 *q++ = *p--; 1264 } 1265 /* 1266 * Last hop goes to final destination. 1267 */ 1268 *q = isr->isr_dst; 1269 m_tag_delete(m0, mtag); 1270 return m; 1271} 1272 1273const int inetctlerrmap[PRC_NCMDS] = { 1274 [PRC_MSGSIZE] = EMSGSIZE, 1275 [PRC_HOSTDEAD] = EHOSTDOWN, 1276 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1277 [PRC_UNREACH_NET] = EHOSTUNREACH, 1278 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1279 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1280 [PRC_UNREACH_PORT] = ECONNREFUSED, 1281 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1282 [PRC_PARAMPROB] = ENOPROTOOPT, 1283}; 1284 1285void 1286ip_fasttimo(void) 1287{ 1288 if (ip_drainwanted) { 1289 ip_drain(); 1290 ip_drainwanted = 0; 1291 } 1292} 1293 1294void 1295ip_drainstub(void) 1296{ 1297 ip_drainwanted = 1; 1298} 1299 1300/* 1301 * Forward a packet. If some error occurs return the sender 1302 * an icmp packet. Note we can't always generate a meaningful 1303 * icmp message because icmp doesn't have a large enough repertoire 1304 * of codes and types. 1305 * 1306 * If not forwarding, just drop the packet. This could be confusing 1307 * if ipforwarding was zero but some routing protocol was advancing 1308 * us as a gateway to somewhere. However, we must let the routing 1309 * protocol deal with that. 1310 * 1311 * The srcrt parameter indicates whether the packet is being forwarded 1312 * via a source route. 1313 */ 1314static void 1315ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1316{ 1317 struct ip *ip = mtod(m, struct ip *); 1318 struct rtentry *rt; 1319 int error, type = 0, code = 0, destmtu = 0; 1320 struct mbuf *mcopy; 1321 n_long dest; 1322 union { 1323 struct sockaddr dst; 1324 struct sockaddr_in dst4; 1325 } u; 1326 uint64_t *ips; 1327 struct route *ro; 1328 1329 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1330 "interrupt handler; synchronization assumptions violated"); 1331 1332 /* 1333 * We are now in the output path. 1334 */ 1335 MCLAIM(m, &ip_tx_mowner); 1336 1337 /* 1338 * Clear any in-bound checksum flags for this packet. 1339 */ 1340 m->m_pkthdr.csum_flags = 0; 1341 1342 dest = 0; 1343 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1344 IP_STATINC(IP_STAT_CANTFORWARD); 1345 m_freem(m); 1346 return; 1347 } 1348 1349 if (ip->ip_ttl <= IPTTLDEC) { 1350 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1351 return; 1352 } 1353 1354 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1355 1356 ro = percpu_getref(ipforward_rt_percpu); 1357 rt = rtcache_lookup(ro, &u.dst); 1358 if (rt == NULL) { 1359 percpu_putref(ipforward_rt_percpu); 1360 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1361 return; 1362 } 1363 1364 /* 1365 * Save at most 68 bytes of the packet in case 1366 * we need to generate an ICMP message to the src. 1367 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1368 */ 1369 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1370 if (mcopy) 1371 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1372 1373 ip->ip_ttl -= IPTTLDEC; 1374 1375 /* 1376 * If forwarding packet using same interface that it came in on, 1377 * perhaps should send a redirect to sender to shortcut a hop. 1378 * Only send redirect if source is sending directly to us, 1379 * and if packet was not source routed (or has any options). 1380 * Also, don't send redirect if forwarding using a default route 1381 * or a route modified by a redirect. 1382 */ 1383 if (rt->rt_ifp == rcvif && 1384 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1385 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1386 ipsendredirects && !srcrt) { 1387 if (rt->rt_ifa && 1388 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1389 ifatoia(rt->rt_ifa)->ia_subnet) { 1390 if (rt->rt_flags & RTF_GATEWAY) 1391 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1392 else 1393 dest = ip->ip_dst.s_addr; 1394 /* 1395 * Router requirements says to only send host 1396 * redirects. 1397 */ 1398 type = ICMP_REDIRECT; 1399 code = ICMP_REDIRECT_HOST; 1400 } 1401 } 1402 rtcache_unref(rt, ro); 1403 1404 error = ip_output(m, NULL, ro, 1405 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1406 NULL, NULL); 1407 1408 if (error) { 1409 IP_STATINC(IP_STAT_CANTFORWARD); 1410 goto error; 1411 } 1412 1413 ips = IP_STAT_GETREF(); 1414 ips[IP_STAT_FORWARD]++; 1415 1416 if (type) { 1417 ips[IP_STAT_REDIRECTSENT]++; 1418 IP_STAT_PUTREF(); 1419 goto redirect; 1420 } 1421 1422 IP_STAT_PUTREF(); 1423 if (mcopy) { 1424#ifdef GATEWAY 1425 if (mcopy->m_flags & M_CANFASTFWD) 1426 ipflow_create(ro, mcopy); 1427#endif 1428 m_freem(mcopy); 1429 } 1430 1431 percpu_putref(ipforward_rt_percpu); 1432 return; 1433 1434redirect: 1435error: 1436 if (mcopy == NULL) { 1437 percpu_putref(ipforward_rt_percpu); 1438 return; 1439 } 1440 1441 switch (error) { 1442 1443 case 0: /* forwarded, but need redirect */ 1444 /* type, code set above */ 1445 break; 1446 1447 case ENETUNREACH: /* shouldn't happen, checked above */ 1448 case EHOSTUNREACH: 1449 case ENETDOWN: 1450 case EHOSTDOWN: 1451 default: 1452 type = ICMP_UNREACH; 1453 code = ICMP_UNREACH_HOST; 1454 break; 1455 1456 case EMSGSIZE: 1457 type = ICMP_UNREACH; 1458 code = ICMP_UNREACH_NEEDFRAG; 1459 1460 if ((rt = rtcache_validate(ro)) != NULL) { 1461 destmtu = rt->rt_ifp->if_mtu; 1462 rtcache_unref(rt, ro); 1463 } 1464#ifdef IPSEC 1465 if (ipsec_used) 1466 ipsec_mtu(mcopy, &destmtu); 1467#endif 1468 IP_STATINC(IP_STAT_CANTFRAG); 1469 break; 1470 1471 case ENOBUFS: 1472 /* 1473 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1474 * Requirements for IP Version 4 Routers. Source quench can 1475 * be a big problem under DoS attacks or if the underlying 1476 * interface is rate-limited. 1477 */ 1478 if (mcopy) 1479 m_freem(mcopy); 1480 percpu_putref(ipforward_rt_percpu); 1481 return; 1482 } 1483 icmp_error(mcopy, type, code, dest, destmtu); 1484 percpu_putref(ipforward_rt_percpu); 1485} 1486 1487void 1488ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1489 struct mbuf *m) 1490{ 1491 struct socket *so = inp->inp_socket; 1492 int inpflags = inp->inp_flags; 1493 1494 if (SOOPT_TIMESTAMP(so->so_options)) 1495 mp = sbsavetimestamp(so->so_options, mp); 1496 1497 if (inpflags & INP_RECVDSTADDR) { 1498 *mp = sbcreatecontrol(&ip->ip_dst, 1499 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1500 if (*mp) 1501 mp = &(*mp)->m_next; 1502 } 1503 1504 if (inpflags & INP_RECVTTL) { 1505 *mp = sbcreatecontrol(&ip->ip_ttl, 1506 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1507 if (*mp) 1508 mp = &(*mp)->m_next; 1509 } 1510 1511 struct psref psref; 1512 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1513 if (__predict_false(ifp == NULL)) { 1514#ifdef DIAGNOSTIC 1515 printf("%s: missing receive interface\n", __func__); 1516#endif 1517 return; /* XXX should report error? */ 1518 } 1519 1520 if (inpflags & INP_RECVPKTINFO) { 1521 struct in_pktinfo ipi; 1522 ipi.ipi_addr = ip->ip_dst; 1523 ipi.ipi_ifindex = ifp->if_index; 1524 *mp = sbcreatecontrol(&ipi, 1525 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1526 if (*mp) 1527 mp = &(*mp)->m_next; 1528 } 1529 if (inpflags & INP_RECVIF) { 1530 struct sockaddr_dl sdl; 1531 1532 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1533 NULL, 0); 1534 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1535 if (*mp) 1536 mp = &(*mp)->m_next; 1537 } 1538 m_put_rcvif_psref(ifp, &psref); 1539} 1540 1541/* 1542 * sysctl helper routine for net.inet.ip.forwsrcrt. 1543 */ 1544static int 1545sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1546{ 1547 int error, tmp; 1548 struct sysctlnode node; 1549 1550 node = *rnode; 1551 tmp = ip_forwsrcrt; 1552 node.sysctl_data = &tmp; 1553 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1554 if (error || newp == NULL) 1555 return (error); 1556 1557 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1558 0, NULL, NULL, NULL); 1559 if (error) 1560 return (error); 1561 1562 ip_forwsrcrt = tmp; 1563 1564 return (0); 1565} 1566 1567/* 1568 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1569 * range of the new value and tweaks timers if it changes. 1570 */ 1571static int 1572sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1573{ 1574 int error, tmp; 1575 struct sysctlnode node; 1576 1577 icmp_mtudisc_lock(); 1578 1579 node = *rnode; 1580 tmp = ip_mtudisc_timeout; 1581 node.sysctl_data = &tmp; 1582 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1583 if (error || newp == NULL) 1584 goto out; 1585 if (tmp < 0) { 1586 error = EINVAL; 1587 goto out; 1588 } 1589 1590 ip_mtudisc_timeout = tmp; 1591 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1592 error = 0; 1593out: 1594 icmp_mtudisc_unlock(); 1595 return error; 1596} 1597 1598static int 1599sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1600{ 1601 1602 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1603} 1604 1605static void 1606sysctl_net_inet_ip_setup(struct sysctllog **clog) 1607{ 1608 sysctl_createv(clog, 0, NULL, NULL, 1609 CTLFLAG_PERMANENT, 1610 CTLTYPE_NODE, "inet", 1611 SYSCTL_DESCR("PF_INET related settings"), 1612 NULL, 0, NULL, 0, 1613 CTL_NET, PF_INET, CTL_EOL); 1614 sysctl_createv(clog, 0, NULL, NULL, 1615 CTLFLAG_PERMANENT, 1616 CTLTYPE_NODE, "ip", 1617 SYSCTL_DESCR("IPv4 related settings"), 1618 NULL, 0, NULL, 0, 1619 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1620 1621 sysctl_createv(clog, 0, NULL, NULL, 1622 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1623 CTLTYPE_INT, "forwarding", 1624 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1625 NULL, 0, &ipforwarding, 0, 1626 CTL_NET, PF_INET, IPPROTO_IP, 1627 IPCTL_FORWARDING, CTL_EOL); 1628 sysctl_createv(clog, 0, NULL, NULL, 1629 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1630 CTLTYPE_INT, "redirect", 1631 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1632 NULL, 0, &ipsendredirects, 0, 1633 CTL_NET, PF_INET, IPPROTO_IP, 1634 IPCTL_SENDREDIRECTS, CTL_EOL); 1635 sysctl_createv(clog, 0, NULL, NULL, 1636 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1637 CTLTYPE_INT, "ttl", 1638 SYSCTL_DESCR("Default TTL for an INET datagram"), 1639 NULL, 0, &ip_defttl, 0, 1640 CTL_NET, PF_INET, IPPROTO_IP, 1641 IPCTL_DEFTTL, CTL_EOL); 1642 sysctl_createv(clog, 0, NULL, NULL, 1643 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1644 CTLTYPE_INT, "forwsrcrt", 1645 SYSCTL_DESCR("Enable forwarding of source-routed " 1646 "datagrams"), 1647 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1648 CTL_NET, PF_INET, IPPROTO_IP, 1649 IPCTL_FORWSRCRT, CTL_EOL); 1650 sysctl_createv(clog, 0, NULL, NULL, 1651 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1652 CTLTYPE_INT, "directed-broadcast", 1653 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1654 NULL, 0, &ip_directedbcast, 0, 1655 CTL_NET, PF_INET, IPPROTO_IP, 1656 IPCTL_DIRECTEDBCAST, CTL_EOL); 1657 sysctl_createv(clog, 0, NULL, NULL, 1658 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1659 CTLTYPE_INT, "allowsrcrt", 1660 SYSCTL_DESCR("Accept source-routed datagrams"), 1661 NULL, 0, &ip_allowsrcrt, 0, 1662 CTL_NET, PF_INET, IPPROTO_IP, 1663 IPCTL_ALLOWSRCRT, CTL_EOL); 1664 1665 sysctl_createv(clog, 0, NULL, NULL, 1666 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1667 CTLTYPE_INT, "mtudisc", 1668 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1669 NULL, 0, &ip_mtudisc, 0, 1670 CTL_NET, PF_INET, IPPROTO_IP, 1671 IPCTL_MTUDISC, CTL_EOL); 1672 sysctl_createv(clog, 0, NULL, NULL, 1673 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1674 CTLTYPE_INT, "anonportmin", 1675 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1676 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1677 CTL_NET, PF_INET, IPPROTO_IP, 1678 IPCTL_ANONPORTMIN, CTL_EOL); 1679 sysctl_createv(clog, 0, NULL, NULL, 1680 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1681 CTLTYPE_INT, "anonportmax", 1682 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1683 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1684 CTL_NET, PF_INET, IPPROTO_IP, 1685 IPCTL_ANONPORTMAX, CTL_EOL); 1686 sysctl_createv(clog, 0, NULL, NULL, 1687 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1688 CTLTYPE_INT, "mtudisctimeout", 1689 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1690 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1691 CTL_NET, PF_INET, IPPROTO_IP, 1692 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1693#ifndef IPNOPRIVPORTS 1694 sysctl_createv(clog, 0, NULL, NULL, 1695 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1696 CTLTYPE_INT, "lowportmin", 1697 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1698 "to assign"), 1699 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1700 CTL_NET, PF_INET, IPPROTO_IP, 1701 IPCTL_LOWPORTMIN, CTL_EOL); 1702 sysctl_createv(clog, 0, NULL, NULL, 1703 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1704 CTLTYPE_INT, "lowportmax", 1705 SYSCTL_DESCR("Highest privileged ephemeral port number " 1706 "to assign"), 1707 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1708 CTL_NET, PF_INET, IPPROTO_IP, 1709 IPCTL_LOWPORTMAX, CTL_EOL); 1710#endif /* IPNOPRIVPORTS */ 1711#if NGRE > 0 1712 sysctl_createv(clog, 0, NULL, NULL, 1713 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1714 CTLTYPE_INT, "grettl", 1715 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1716 NULL, 0, &ip_gre_ttl, 0, 1717 CTL_NET, PF_INET, IPPROTO_IP, 1718 IPCTL_GRE_TTL, CTL_EOL); 1719#endif /* NGRE */ 1720 sysctl_createv(clog, 0, NULL, NULL, 1721 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1722 CTLTYPE_INT, "checkinterface", 1723 SYSCTL_DESCR("Enable receive side of Strong ES model " 1724 "from RFC1122"), 1725 NULL, 0, &ip_checkinterface, 0, 1726 CTL_NET, PF_INET, IPPROTO_IP, 1727 IPCTL_CHECKINTERFACE, CTL_EOL); 1728 sysctl_createv(clog, 0, NULL, NULL, 1729 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1730 CTLTYPE_INT, "random_id", 1731 SYSCTL_DESCR("Assign random ip_id values"), 1732 NULL, 0, &ip_do_randomid, 0, 1733 CTL_NET, PF_INET, IPPROTO_IP, 1734 IPCTL_RANDOMID, CTL_EOL); 1735 sysctl_createv(clog, 0, NULL, NULL, 1736 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1737 CTLTYPE_INT, "do_loopback_cksum", 1738 SYSCTL_DESCR("Perform IP checksum on loopback"), 1739 NULL, 0, &ip_do_loopback_cksum, 0, 1740 CTL_NET, PF_INET, IPPROTO_IP, 1741 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1742 sysctl_createv(clog, 0, NULL, NULL, 1743 CTLFLAG_PERMANENT, 1744 CTLTYPE_STRUCT, "stats", 1745 SYSCTL_DESCR("IP statistics"), 1746 sysctl_net_inet_ip_stats, 0, NULL, 0, 1747 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1748 CTL_EOL); 1749#if NARP 1750 sysctl_createv(clog, 0, NULL, NULL, 1751 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1752 CTLTYPE_INT, "dad_count", 1753 SYSCTL_DESCR("Number of Duplicate Address Detection " 1754 "probes to send"), 1755 NULL, 0, &ip_dad_count, 0, 1756 CTL_NET, PF_INET, IPPROTO_IP, 1757 IPCTL_DAD_COUNT, CTL_EOL); 1758#endif 1759 1760 /* anonportalgo RFC6056 subtree */ 1761 const struct sysctlnode *portalgo_node; 1762 sysctl_createv(clog, 0, NULL, &portalgo_node, 1763 CTLFLAG_PERMANENT, 1764 CTLTYPE_NODE, "anonportalgo", 1765 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1766 NULL, 0, NULL, 0, 1767 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1768 sysctl_createv(clog, 0, &portalgo_node, NULL, 1769 CTLFLAG_PERMANENT, 1770 CTLTYPE_STRING, "available", 1771 SYSCTL_DESCR("available algorithms"), 1772 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1773 CTL_CREATE, CTL_EOL); 1774 sysctl_createv(clog, 0, &portalgo_node, NULL, 1775 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1776 CTLTYPE_STRING, "selected", 1777 SYSCTL_DESCR("selected algorithm"), 1778 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1779 CTL_CREATE, CTL_EOL); 1780 sysctl_createv(clog, 0, &portalgo_node, NULL, 1781 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1782 CTLTYPE_STRUCT, "reserve", 1783 SYSCTL_DESCR("bitmap of reserved ports"), 1784 sysctl_portalgo_reserve4, 0, NULL, 0, 1785 CTL_CREATE, CTL_EOL); 1786} 1787 1788void 1789ip_statinc(u_int stat) 1790{ 1791 1792 KASSERT(stat < IP_NSTATS); 1793 IP_STATINC(stat); 1794} 1795