ip_input.c revision 1.390
1/* $NetBSD: ip_input.c,v 1.390 2019/09/15 21:00:15 bouyer 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 1.390 2019/09/15 21:00:15 bouyer 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 IP_STATINC(IP_STAT_PFILDROP_IN); 582 goto out; 583 } 584 if (__predict_false(m->m_len < sizeof(struct ip))) { 585 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) { 586 IP_STATINC(IP_STAT_TOOSMALL); 587 goto out; 588 } 589 } 590 ip = mtod(m, struct ip *); 591 hlen = ip->ip_hl << 2; 592 if (hlen < sizeof(struct ip)) { /* minimum header length */ 593 IP_STATINC(IP_STAT_BADHLEN); 594 goto out; 595 } 596 if (hlen > m->m_len) { 597 if ((m = m_pullup(m, hlen)) == NULL) { 598 IP_STATINC(IP_STAT_BADHLEN); 599 goto out; 600 } 601 ip = mtod(m, struct ip *); 602 } 603 604 /* 605 * XXX The setting of "srcrt" here is to prevent ip_forward() 606 * from generating ICMP redirects for packets that have 607 * been redirected by a hook back out on to the same LAN that 608 * they came from and is not an indication that the packet 609 * is being influenced by source routing options. This 610 * allows things like 611 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 612 * where tlp0 is both on the 1.1.1.0/24 network and is the 613 * default route for hosts on 1.1.1.0/24. Of course this 614 * also requires a "map tlp0 ..." to complete the story. 615 * One might argue whether or not this kind of network config. 616 * should be supported in this manner... 617 */ 618 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 619 } 620 621#ifdef ALTQ 622 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 623 if (altq_input) { 624 SOFTNET_LOCK(); 625 if ((*altq_input)(m, AF_INET) == 0) { 626 /* Packet dropped by traffic conditioner. */ 627 SOFTNET_UNLOCK(); 628 m = NULL; 629 goto out; 630 } 631 SOFTNET_UNLOCK(); 632 } 633#endif 634 635 /* 636 * Process options and, if not destined for us, 637 * ship it on. ip_dooptions returns 1 when an 638 * error was detected (causing an icmp message 639 * to be sent and the original packet to be freed). 640 */ 641 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 642 m = NULL; 643 goto out; 644 } 645 646 /* 647 * Check our list of addresses, to see if the packet is for us. 648 * 649 * Traditional 4.4BSD did not consult IFF_UP at all. 650 * The behavior here is to treat addresses on !IFF_UP interface 651 * or IN_IFF_NOTREADY addresses as not mine. 652 */ 653 downmatch = 0; 654 s = pserialize_read_enter(); 655 ia = ip_match_our_address(ifp, ip, &downmatch); 656 if (ia != NULL) { 657 pserialize_read_exit(s); 658 goto ours; 659 } 660 661 if (ifp->if_flags & IFF_BROADCAST) { 662 ia = ip_match_our_address_broadcast(ifp, ip); 663 if (ia != NULL) { 664 pserialize_read_exit(s); 665 goto ours; 666 } 667 } 668 pserialize_read_exit(s); 669 670 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 671#ifdef MROUTING 672 extern struct socket *ip_mrouter; 673 674 if (ip_mrouter) { 675 /* 676 * If we are acting as a multicast router, all 677 * incoming multicast packets are passed to the 678 * kernel-level multicast forwarding function. 679 * The packet is returned (relatively) intact; if 680 * ip_mforward() returns a non-zero value, the packet 681 * must be discarded, else it may be accepted below. 682 * 683 * (The IP ident field is put in the same byte order 684 * as expected when ip_mforward() is called from 685 * ip_output().) 686 */ 687 SOFTNET_LOCK(); 688 if (ip_mforward(m, ifp) != 0) { 689 SOFTNET_UNLOCK(); 690 IP_STATINC(IP_STAT_CANTFORWARD); 691 goto out; 692 } 693 SOFTNET_UNLOCK(); 694 695 /* 696 * The process-level routing demon needs to receive 697 * all multicast IGMP packets, whether or not this 698 * host belongs to their destination groups. 699 */ 700 if (ip->ip_p == IPPROTO_IGMP) { 701 goto ours; 702 } 703 IP_STATINC(IP_STAT_CANTFORWARD); 704 } 705#endif 706 /* 707 * See if we belong to the destination multicast group on the 708 * arrival interface. 709 */ 710 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 711 IP_STATINC(IP_STAT_CANTFORWARD); 712 goto out; 713 } 714 goto ours; 715 } 716 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 717 in_nullhost(ip->ip_dst)) 718 goto ours; 719 720 /* 721 * Not for us; forward if possible and desirable. 722 */ 723 if (ipforwarding == 0) { 724 m_put_rcvif_psref(ifp, &psref); 725 IP_STATINC(IP_STAT_CANTFORWARD); 726 m_freem(m); 727 } else { 728 /* 729 * If ip_dst matched any of my address on !IFF_UP interface, 730 * and there's no IFF_UP interface that matches ip_dst, 731 * send icmp unreach. Forwarding it will result in in-kernel 732 * forwarding loop till TTL goes to 0. 733 */ 734 if (downmatch) { 735 m_put_rcvif_psref(ifp, &psref); 736 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 737 IP_STATINC(IP_STAT_CANTFORWARD); 738 return; 739 } 740#ifdef IPSEC 741 /* Check the security policy (SP) for the packet */ 742 if (ipsec_used) { 743 if (ipsec_ip_input(m, true) != 0) { 744 goto out; 745 } 746 } 747#endif 748 ip_forward(m, srcrt, ifp); 749 m_put_rcvif_psref(ifp, &psref); 750 } 751 return; 752 753ours: 754 m_put_rcvif_psref(ifp, &psref); 755 ifp = NULL; 756 757 /* 758 * If offset or IP_MF are set, must reassemble. 759 */ 760 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 761 /* 762 * Pass to IP reassembly mechanism. 763 */ 764 if (ip_reass_packet(&m) != 0) { 765 /* Failed; invalid fragment(s) or packet. */ 766 goto out; 767 } 768 if (m == NULL) { 769 /* More fragments should come; silently return. */ 770 goto out; 771 } 772 /* 773 * Reassembly is done, we have the final packet. 774 * Update cached data in local variable(s). 775 */ 776 ip = mtod(m, struct ip *); 777 hlen = ip->ip_hl << 2; 778 } 779 780 M_VERIFY_PACKET(m); 781 782#ifdef IPSEC 783 /* 784 * Enforce IPsec policy checking if we are seeing last header. 785 * Note that we do not visit this with protocols with PCB layer 786 * code - like UDP/TCP/raw IP. 787 */ 788 if (ipsec_used && 789 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 790 if (ipsec_ip_input(m, false) != 0) { 791 goto out; 792 } 793 } 794#endif 795 796 /* 797 * Switch out to protocol's input routine. 798 */ 799#if IFA_STATS 800 if (ia) { 801 struct in_ifaddr *_ia; 802 /* 803 * Keep a reference from ip_match_our_address with psref 804 * is expensive, so explore ia here again. 805 */ 806 s = pserialize_read_enter(); 807 _ia = in_get_ia(ip->ip_dst); 808 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 809 pserialize_read_exit(s); 810 } 811#endif 812 IP_STATINC(IP_STAT_DELIVERED); 813 814 const int off = hlen, nh = ip->ip_p; 815 816 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 817 return; 818 819out: 820 m_put_rcvif_psref(ifp, &psref); 821 if (m != NULL) 822 m_freem(m); 823} 824 825/* 826 * IP timer processing. 827 */ 828void 829ip_slowtimo(void) 830{ 831 832 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 833 834 ip_reass_slowtimo(); 835 836 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 837} 838 839/* 840 * IP drain processing. 841 */ 842void 843ip_drain(void) 844{ 845 846 KERNEL_LOCK(1, NULL); 847 ip_reass_drain(); 848 KERNEL_UNLOCK_ONE(NULL); 849} 850 851/* 852 * ip_dooptions: perform option processing on a datagram, possibly discarding 853 * it if bad options are encountered, or forwarding it if source-routed. 854 * 855 * => Returns true if packet has been forwarded/freed. 856 * => Returns false if the packet should be processed further. 857 */ 858static bool 859ip_dooptions(struct mbuf *m) 860{ 861 struct ip *ip = mtod(m, struct ip *); 862 u_char *cp, *cp0; 863 struct ip_timestamp *ipt; 864 struct in_ifaddr *ia; 865 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 866 int srr_present, rr_present, ts_present; 867 struct in_addr dst; 868 n_time ntime; 869 struct ifaddr *ifa = NULL; 870 int s; 871 872 srr_present = 0; 873 rr_present = 0; 874 ts_present = 0; 875 876 dst = ip->ip_dst; 877 cp = (u_char *)(ip + 1); 878 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 879 for (; cnt > 0; cnt -= optlen, cp += optlen) { 880 opt = cp[IPOPT_OPTVAL]; 881 if (opt == IPOPT_EOL) 882 break; 883 if (opt == IPOPT_NOP) 884 optlen = 1; 885 else { 886 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 887 code = &cp[IPOPT_OLEN] - (u_char *)ip; 888 goto bad; 889 } 890 optlen = cp[IPOPT_OLEN]; 891 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 892 code = &cp[IPOPT_OLEN] - (u_char *)ip; 893 goto bad; 894 } 895 } 896 switch (opt) { 897 898 default: 899 break; 900 901 /* 902 * Source routing with record. 903 * Find interface with current destination address. 904 * If none on this machine then drop if strictly routed, 905 * or do nothing if loosely routed. 906 * Record interface address and bring up next address 907 * component. If strictly routed make sure next 908 * address is on directly accessible net. 909 */ 910 case IPOPT_LSRR: 911 case IPOPT_SSRR: { 912 struct psref psref; 913 struct sockaddr_in ipaddr = { 914 .sin_len = sizeof(ipaddr), 915 .sin_family = AF_INET, 916 }; 917 918 if (ip_allowsrcrt == 0) { 919 type = ICMP_UNREACH; 920 code = ICMP_UNREACH_NET_PROHIB; 921 goto bad; 922 } 923 if (srr_present++) { 924 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 925 goto bad; 926 } 927 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 928 code = &cp[IPOPT_OLEN] - (u_char *)ip; 929 goto bad; 930 } 931 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 932 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 933 goto bad; 934 } 935 ipaddr.sin_addr = ip->ip_dst; 936 937 s = pserialize_read_enter(); 938 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 939 if (ifa == NULL) { 940 pserialize_read_exit(s); 941 if (opt == IPOPT_SSRR) { 942 type = ICMP_UNREACH; 943 code = ICMP_UNREACH_SRCFAIL; 944 goto bad; 945 } 946 /* 947 * Loose routing, and not at next destination 948 * yet; nothing to do except forward. 949 */ 950 break; 951 } 952 pserialize_read_exit(s); 953 954 off--; /* 0 origin */ 955 if ((off + sizeof(struct in_addr)) > optlen) { 956 /* 957 * End of source route. Should be for us. 958 */ 959 save_rte(m, cp, ip->ip_src); 960 break; 961 } 962 /* 963 * locate outgoing interface 964 */ 965 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 966 sizeof(ipaddr.sin_addr)); 967 if (opt == IPOPT_SSRR) { 968 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 969 &psref); 970 if (ifa != NULL) 971 ia = ifatoia(ifa); 972 else 973 ia = NULL; 974 } else { 975 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 976 } 977 if (ia == NULL) { 978 type = ICMP_UNREACH; 979 code = ICMP_UNREACH_SRCFAIL; 980 goto bad; 981 } 982 ip->ip_dst = ipaddr.sin_addr; 983 memcpy(cp + off, &ia->ia_addr.sin_addr, 984 sizeof(struct in_addr)); 985 ia4_release(ia, &psref); 986 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 987 /* 988 * Let ip_intr's mcast routing check handle mcast pkts 989 */ 990 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 991 break; 992 } 993 994 case IPOPT_RR: { 995 struct psref psref; 996 struct sockaddr_in ipaddr = { 997 .sin_len = sizeof(ipaddr), 998 .sin_family = AF_INET, 999 }; 1000 1001 if (rr_present++) { 1002 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1003 goto bad; 1004 } 1005 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1006 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1007 goto bad; 1008 } 1009 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1010 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1011 goto bad; 1012 } 1013 /* 1014 * If no space remains, ignore. 1015 */ 1016 off--; /* 0 origin */ 1017 if ((off + sizeof(struct in_addr)) > optlen) 1018 break; 1019 memcpy((void *)&ipaddr.sin_addr, (void *)&ip->ip_dst, 1020 sizeof(ipaddr.sin_addr)); 1021 /* 1022 * locate outgoing interface; if we're the destination, 1023 * use the incoming interface (should be same). 1024 */ 1025 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1026 if (ifa == NULL) { 1027 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1028 if (ia == NULL) { 1029 type = ICMP_UNREACH; 1030 code = ICMP_UNREACH_HOST; 1031 goto bad; 1032 } 1033 } else { 1034 ia = ifatoia(ifa); 1035 } 1036 memcpy(cp + off, &ia->ia_addr.sin_addr, 1037 sizeof(struct in_addr)); 1038 ia4_release(ia, &psref); 1039 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1040 break; 1041 } 1042 1043 case IPOPT_TS: 1044 code = cp - (u_char *)ip; 1045 ipt = (struct ip_timestamp *)cp; 1046 if (ts_present++) { 1047 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1048 goto bad; 1049 } 1050 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1051 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1052 goto bad; 1053 } 1054 if (ipt->ipt_ptr < 5) { 1055 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1056 goto bad; 1057 } 1058 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1059 if (++ipt->ipt_oflw == 0) { 1060 code = (u_char *)&ipt->ipt_ptr - 1061 (u_char *)ip; 1062 goto bad; 1063 } 1064 break; 1065 } 1066 cp0 = (cp + ipt->ipt_ptr - 1); 1067 switch (ipt->ipt_flg) { 1068 1069 case IPOPT_TS_TSONLY: 1070 break; 1071 1072 case IPOPT_TS_TSANDADDR: { 1073 struct ifnet *rcvif; 1074 int _s, _ss; 1075 struct sockaddr_in ipaddr = { 1076 .sin_len = sizeof(ipaddr), 1077 .sin_family = AF_INET, 1078 }; 1079 1080 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1081 sizeof(struct in_addr) > ipt->ipt_len) { 1082 code = (u_char *)&ipt->ipt_ptr - 1083 (u_char *)ip; 1084 goto bad; 1085 } 1086 ipaddr.sin_addr = dst; 1087 _ss = pserialize_read_enter(); 1088 rcvif = m_get_rcvif(m, &_s); 1089 if (__predict_true(rcvif != NULL)) { 1090 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1091 rcvif); 1092 } 1093 m_put_rcvif(rcvif, &_s); 1094 if (ifa == NULL) { 1095 pserialize_read_exit(_ss); 1096 break; 1097 } 1098 ia = ifatoia(ifa); 1099 memcpy(cp0, &ia->ia_addr.sin_addr, 1100 sizeof(struct in_addr)); 1101 pserialize_read_exit(_ss); 1102 ipt->ipt_ptr += sizeof(struct in_addr); 1103 break; 1104 } 1105 1106 case IPOPT_TS_PRESPEC: { 1107 struct sockaddr_in ipaddr = { 1108 .sin_len = sizeof(ipaddr), 1109 .sin_family = AF_INET, 1110 }; 1111 1112 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1113 sizeof(struct in_addr) > ipt->ipt_len) { 1114 code = (u_char *)&ipt->ipt_ptr - 1115 (u_char *)ip; 1116 goto bad; 1117 } 1118 memcpy(&ipaddr.sin_addr, cp0, 1119 sizeof(struct in_addr)); 1120 s = pserialize_read_enter(); 1121 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1122 if (ifa == NULL) { 1123 pserialize_read_exit(s); 1124 continue; 1125 } 1126 pserialize_read_exit(s); 1127 ipt->ipt_ptr += sizeof(struct in_addr); 1128 break; 1129 } 1130 1131 default: 1132 /* XXX can't take &ipt->ipt_flg */ 1133 code = (u_char *)&ipt->ipt_ptr - 1134 (u_char *)ip + 1; 1135 goto bad; 1136 } 1137 ntime = iptime(); 1138 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1139 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1140 sizeof(n_time)); 1141 ipt->ipt_ptr += sizeof(n_time); 1142 } 1143 } 1144 if (forward) { 1145 struct ifnet *rcvif; 1146 struct psref _psref; 1147 1148 if (ip_forwsrcrt == 0) { 1149 type = ICMP_UNREACH; 1150 code = ICMP_UNREACH_SRCFAIL; 1151 goto bad; 1152 } 1153 1154 rcvif = m_get_rcvif_psref(m, &_psref); 1155 if (__predict_false(rcvif == NULL)) { 1156 type = ICMP_UNREACH; 1157 code = ICMP_UNREACH_HOST; 1158 goto bad; 1159 } 1160 ip_forward(m, 1, rcvif); 1161 m_put_rcvif_psref(rcvif, &_psref); 1162 return true; 1163 } 1164 return false; 1165bad: 1166 icmp_error(m, type, code, 0, 0); 1167 IP_STATINC(IP_STAT_BADOPTIONS); 1168 return true; 1169} 1170 1171/* 1172 * ip_rtaddr: given address of next destination (final or next hop), 1173 * return internet address info of interface to be used to get there. 1174 */ 1175static struct in_ifaddr * 1176ip_rtaddr(struct in_addr dst, struct psref *psref) 1177{ 1178 struct rtentry *rt; 1179 union { 1180 struct sockaddr dst; 1181 struct sockaddr_in dst4; 1182 } u; 1183 struct route *ro; 1184 1185 sockaddr_in_init(&u.dst4, &dst, 0); 1186 1187 ro = percpu_getref(ipforward_rt_percpu); 1188 rt = rtcache_lookup(ro, &u.dst); 1189 if (rt == NULL) { 1190 percpu_putref(ipforward_rt_percpu); 1191 return NULL; 1192 } 1193 1194 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1195 rtcache_unref(rt, ro); 1196 percpu_putref(ipforward_rt_percpu); 1197 1198 return ifatoia(rt->rt_ifa); 1199} 1200 1201/* 1202 * save_rte: save incoming source route for use in replies, to be picked 1203 * up later by ip_srcroute if the receiver is interested. 1204 */ 1205static void 1206save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1207{ 1208 struct ip_srcrt *isr; 1209 struct m_tag *mtag; 1210 unsigned olen; 1211 1212 olen = option[IPOPT_OLEN]; 1213 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1214 return; 1215 1216 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1217 if (mtag == NULL) 1218 return; 1219 isr = (struct ip_srcrt *)(mtag + 1); 1220 1221 memcpy(isr->isr_hdr, option, olen); 1222 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1223 isr->isr_dst = dst; 1224 m_tag_prepend(m, mtag); 1225} 1226 1227/* 1228 * Retrieve incoming source route for use in replies, 1229 * in the same form used by setsockopt. 1230 * The first hop is placed before the options, will be removed later. 1231 */ 1232struct mbuf * 1233ip_srcroute(struct mbuf *m0) 1234{ 1235 struct in_addr *p, *q; 1236 struct mbuf *m; 1237 struct ip_srcrt *isr; 1238 struct m_tag *mtag; 1239 1240 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE); 1241 if (mtag == NULL) 1242 return NULL; 1243 isr = (struct ip_srcrt *)(mtag + 1); 1244 1245 if (isr->isr_nhops == 0) 1246 return NULL; 1247 1248 m = m_get(M_DONTWAIT, MT_SOOPTS); 1249 if (m == NULL) 1250 return NULL; 1251 1252 MCLAIM(m, &inetdomain.dom_mowner); 1253#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1254 1255 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1256 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1257 1258 /* 1259 * First save first hop for return route 1260 */ 1261 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1262 *(mtod(m, struct in_addr *)) = *p--; 1263 1264 /* 1265 * Copy option fields and padding (nop) to mbuf. 1266 */ 1267 isr->isr_nop = IPOPT_NOP; 1268 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1269 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1270 OPTSIZ); 1271 q = (struct in_addr *)(mtod(m, char *) + 1272 sizeof(struct in_addr) + OPTSIZ); 1273#undef OPTSIZ 1274 /* 1275 * Record return path as an IP source route, 1276 * reversing the path (pointers are now aligned). 1277 */ 1278 while (p >= isr->isr_routes) { 1279 *q++ = *p--; 1280 } 1281 /* 1282 * Last hop goes to final destination. 1283 */ 1284 *q = isr->isr_dst; 1285 m_tag_delete(m0, mtag); 1286 return m; 1287} 1288 1289const int inetctlerrmap[PRC_NCMDS] = { 1290 [PRC_MSGSIZE] = EMSGSIZE, 1291 [PRC_HOSTDEAD] = EHOSTDOWN, 1292 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1293 [PRC_UNREACH_NET] = EHOSTUNREACH, 1294 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1295 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1296 [PRC_UNREACH_PORT] = ECONNREFUSED, 1297 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1298 [PRC_PARAMPROB] = ENOPROTOOPT, 1299}; 1300 1301void 1302ip_fasttimo(void) 1303{ 1304 if (ip_drainwanted) { 1305 ip_drain(); 1306 ip_drainwanted = 0; 1307 } 1308} 1309 1310void 1311ip_drainstub(void) 1312{ 1313 ip_drainwanted = 1; 1314} 1315 1316/* 1317 * Forward a packet. If some error occurs return the sender 1318 * an icmp packet. Note we can't always generate a meaningful 1319 * icmp message because icmp doesn't have a large enough repertoire 1320 * of codes and types. 1321 * 1322 * If not forwarding, just drop the packet. This could be confusing 1323 * if ipforwarding was zero but some routing protocol was advancing 1324 * us as a gateway to somewhere. However, we must let the routing 1325 * protocol deal with that. 1326 * 1327 * The srcrt parameter indicates whether the packet is being forwarded 1328 * via a source route. 1329 */ 1330static void 1331ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1332{ 1333 struct ip *ip = mtod(m, struct ip *); 1334 struct rtentry *rt; 1335 int error, type = 0, code = 0, destmtu = 0; 1336 struct mbuf *mcopy; 1337 n_long dest; 1338 union { 1339 struct sockaddr dst; 1340 struct sockaddr_in dst4; 1341 } u; 1342 uint64_t *ips; 1343 struct route *ro; 1344 1345 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1346 "interrupt handler; synchronization assumptions violated"); 1347 1348 /* 1349 * We are now in the output path. 1350 */ 1351 MCLAIM(m, &ip_tx_mowner); 1352 1353 /* 1354 * Clear any in-bound checksum flags for this packet. 1355 */ 1356 m->m_pkthdr.csum_flags = 0; 1357 1358 dest = 0; 1359 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1360 IP_STATINC(IP_STAT_CANTFORWARD); 1361 m_freem(m); 1362 return; 1363 } 1364 1365 if (ip->ip_ttl <= IPTTLDEC) { 1366 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1367 return; 1368 } 1369 1370 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1371 1372 ro = percpu_getref(ipforward_rt_percpu); 1373 rt = rtcache_lookup(ro, &u.dst); 1374 if (rt == NULL) { 1375 percpu_putref(ipforward_rt_percpu); 1376 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1377 return; 1378 } 1379 1380 /* 1381 * Save at most 68 bytes of the packet in case 1382 * we need to generate an ICMP message to the src. 1383 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1384 */ 1385 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1386 if (mcopy) 1387 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1388 1389 ip->ip_ttl -= IPTTLDEC; 1390 1391 /* 1392 * If forwarding packet using same interface that it came in on, 1393 * perhaps should send a redirect to sender to shortcut a hop. 1394 * Only send redirect if source is sending directly to us, 1395 * and if packet was not source routed (or has any options). 1396 * Also, don't send redirect if forwarding using a default route 1397 * or a route modified by a redirect. 1398 */ 1399 if (rt->rt_ifp == rcvif && 1400 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1401 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1402 ipsendredirects && !srcrt) { 1403 if (rt->rt_ifa && 1404 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1405 ifatoia(rt->rt_ifa)->ia_subnet) { 1406 if (rt->rt_flags & RTF_GATEWAY) 1407 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1408 else 1409 dest = ip->ip_dst.s_addr; 1410 /* 1411 * Router requirements says to only send host 1412 * redirects. 1413 */ 1414 type = ICMP_REDIRECT; 1415 code = ICMP_REDIRECT_HOST; 1416 } 1417 } 1418 rtcache_unref(rt, ro); 1419 1420 error = ip_output(m, NULL, ro, 1421 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1422 NULL, NULL); 1423 1424 if (error) { 1425 IP_STATINC(IP_STAT_CANTFORWARD); 1426 goto error; 1427 } 1428 1429 ips = IP_STAT_GETREF(); 1430 ips[IP_STAT_FORWARD]++; 1431 1432 if (type) { 1433 ips[IP_STAT_REDIRECTSENT]++; 1434 IP_STAT_PUTREF(); 1435 goto redirect; 1436 } 1437 1438 IP_STAT_PUTREF(); 1439 if (mcopy) { 1440#ifdef GATEWAY 1441 if (mcopy->m_flags & M_CANFASTFWD) 1442 ipflow_create(ro, mcopy); 1443#endif 1444 m_freem(mcopy); 1445 } 1446 1447 percpu_putref(ipforward_rt_percpu); 1448 return; 1449 1450redirect: 1451error: 1452 if (mcopy == NULL) { 1453 percpu_putref(ipforward_rt_percpu); 1454 return; 1455 } 1456 1457 switch (error) { 1458 1459 case 0: /* forwarded, but need redirect */ 1460 /* type, code set above */ 1461 break; 1462 1463 case ENETUNREACH: /* shouldn't happen, checked above */ 1464 case EHOSTUNREACH: 1465 case ENETDOWN: 1466 case EHOSTDOWN: 1467 default: 1468 type = ICMP_UNREACH; 1469 code = ICMP_UNREACH_HOST; 1470 break; 1471 1472 case EMSGSIZE: 1473 type = ICMP_UNREACH; 1474 code = ICMP_UNREACH_NEEDFRAG; 1475 1476 if ((rt = rtcache_validate(ro)) != NULL) { 1477 destmtu = rt->rt_ifp->if_mtu; 1478 rtcache_unref(rt, ro); 1479 } 1480#ifdef IPSEC 1481 if (ipsec_used) 1482 ipsec_mtu(mcopy, &destmtu); 1483#endif 1484 IP_STATINC(IP_STAT_CANTFRAG); 1485 break; 1486 1487 case ENOBUFS: 1488 /* 1489 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1490 * Requirements for IP Version 4 Routers. Source quench can 1491 * be a big problem under DoS attacks or if the underlying 1492 * interface is rate-limited. 1493 */ 1494 if (mcopy) 1495 m_freem(mcopy); 1496 percpu_putref(ipforward_rt_percpu); 1497 return; 1498 } 1499 icmp_error(mcopy, type, code, dest, destmtu); 1500 percpu_putref(ipforward_rt_percpu); 1501} 1502 1503void 1504ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1505 struct mbuf *m) 1506{ 1507 struct socket *so = inp->inp_socket; 1508 int inpflags = inp->inp_flags; 1509 1510 if (SOOPT_TIMESTAMP(so->so_options)) 1511 mp = sbsavetimestamp(so->so_options, mp); 1512 1513 if (inpflags & INP_RECVDSTADDR) { 1514 *mp = sbcreatecontrol(&ip->ip_dst, 1515 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1516 if (*mp) 1517 mp = &(*mp)->m_next; 1518 } 1519 1520 if (inpflags & INP_RECVTTL) { 1521 *mp = sbcreatecontrol(&ip->ip_ttl, 1522 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1523 if (*mp) 1524 mp = &(*mp)->m_next; 1525 } 1526 1527 struct psref psref; 1528 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1529 if (__predict_false(ifp == NULL)) { 1530#ifdef DIAGNOSTIC 1531 printf("%s: missing receive interface\n", __func__); 1532#endif 1533 return; /* XXX should report error? */ 1534 } 1535 1536 if (inpflags & INP_RECVPKTINFO) { 1537 struct in_pktinfo ipi; 1538 ipi.ipi_addr = ip->ip_dst; 1539 ipi.ipi_ifindex = ifp->if_index; 1540 *mp = sbcreatecontrol(&ipi, 1541 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1542 if (*mp) 1543 mp = &(*mp)->m_next; 1544 } 1545 if (inpflags & INP_RECVIF) { 1546 struct sockaddr_dl sdl; 1547 1548 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1549 NULL, 0); 1550 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1551 if (*mp) 1552 mp = &(*mp)->m_next; 1553 } 1554 m_put_rcvif_psref(ifp, &psref); 1555} 1556 1557/* 1558 * sysctl helper routine for net.inet.ip.forwsrcrt. 1559 */ 1560static int 1561sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1562{ 1563 int error, tmp; 1564 struct sysctlnode node; 1565 1566 node = *rnode; 1567 tmp = ip_forwsrcrt; 1568 node.sysctl_data = &tmp; 1569 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1570 if (error || newp == NULL) 1571 return (error); 1572 1573 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1574 0, NULL, NULL, NULL); 1575 if (error) 1576 return (error); 1577 1578 ip_forwsrcrt = tmp; 1579 1580 return (0); 1581} 1582 1583/* 1584 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1585 * range of the new value and tweaks timers if it changes. 1586 */ 1587static int 1588sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1589{ 1590 int error, tmp; 1591 struct sysctlnode node; 1592 1593 icmp_mtudisc_lock(); 1594 1595 node = *rnode; 1596 tmp = ip_mtudisc_timeout; 1597 node.sysctl_data = &tmp; 1598 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1599 if (error || newp == NULL) 1600 goto out; 1601 if (tmp < 0) { 1602 error = EINVAL; 1603 goto out; 1604 } 1605 1606 ip_mtudisc_timeout = tmp; 1607 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1608 error = 0; 1609out: 1610 icmp_mtudisc_unlock(); 1611 return error; 1612} 1613 1614static int 1615sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1616{ 1617 1618 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1619} 1620 1621static void 1622sysctl_net_inet_ip_setup(struct sysctllog **clog) 1623{ 1624 sysctl_createv(clog, 0, NULL, NULL, 1625 CTLFLAG_PERMANENT, 1626 CTLTYPE_NODE, "inet", 1627 SYSCTL_DESCR("PF_INET related settings"), 1628 NULL, 0, NULL, 0, 1629 CTL_NET, PF_INET, CTL_EOL); 1630 sysctl_createv(clog, 0, NULL, NULL, 1631 CTLFLAG_PERMANENT, 1632 CTLTYPE_NODE, "ip", 1633 SYSCTL_DESCR("IPv4 related settings"), 1634 NULL, 0, NULL, 0, 1635 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1636 1637 sysctl_createv(clog, 0, NULL, NULL, 1638 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1639 CTLTYPE_INT, "forwarding", 1640 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1641 NULL, 0, &ipforwarding, 0, 1642 CTL_NET, PF_INET, IPPROTO_IP, 1643 IPCTL_FORWARDING, CTL_EOL); 1644 sysctl_createv(clog, 0, NULL, NULL, 1645 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1646 CTLTYPE_INT, "redirect", 1647 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1648 NULL, 0, &ipsendredirects, 0, 1649 CTL_NET, PF_INET, IPPROTO_IP, 1650 IPCTL_SENDREDIRECTS, CTL_EOL); 1651 sysctl_createv(clog, 0, NULL, NULL, 1652 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1653 CTLTYPE_INT, "ttl", 1654 SYSCTL_DESCR("Default TTL for an INET datagram"), 1655 NULL, 0, &ip_defttl, 0, 1656 CTL_NET, PF_INET, IPPROTO_IP, 1657 IPCTL_DEFTTL, CTL_EOL); 1658 sysctl_createv(clog, 0, NULL, NULL, 1659 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1660 CTLTYPE_INT, "forwsrcrt", 1661 SYSCTL_DESCR("Enable forwarding of source-routed " 1662 "datagrams"), 1663 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1664 CTL_NET, PF_INET, IPPROTO_IP, 1665 IPCTL_FORWSRCRT, CTL_EOL); 1666 sysctl_createv(clog, 0, NULL, NULL, 1667 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1668 CTLTYPE_INT, "directed-broadcast", 1669 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1670 NULL, 0, &ip_directedbcast, 0, 1671 CTL_NET, PF_INET, IPPROTO_IP, 1672 IPCTL_DIRECTEDBCAST, CTL_EOL); 1673 sysctl_createv(clog, 0, NULL, NULL, 1674 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1675 CTLTYPE_INT, "allowsrcrt", 1676 SYSCTL_DESCR("Accept source-routed datagrams"), 1677 NULL, 0, &ip_allowsrcrt, 0, 1678 CTL_NET, PF_INET, IPPROTO_IP, 1679 IPCTL_ALLOWSRCRT, CTL_EOL); 1680 1681 sysctl_createv(clog, 0, NULL, NULL, 1682 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1683 CTLTYPE_INT, "mtudisc", 1684 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1685 NULL, 0, &ip_mtudisc, 0, 1686 CTL_NET, PF_INET, IPPROTO_IP, 1687 IPCTL_MTUDISC, CTL_EOL); 1688 sysctl_createv(clog, 0, NULL, NULL, 1689 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1690 CTLTYPE_INT, "anonportmin", 1691 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1692 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1693 CTL_NET, PF_INET, IPPROTO_IP, 1694 IPCTL_ANONPORTMIN, CTL_EOL); 1695 sysctl_createv(clog, 0, NULL, NULL, 1696 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1697 CTLTYPE_INT, "anonportmax", 1698 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1699 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1700 CTL_NET, PF_INET, IPPROTO_IP, 1701 IPCTL_ANONPORTMAX, CTL_EOL); 1702 sysctl_createv(clog, 0, NULL, NULL, 1703 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1704 CTLTYPE_INT, "mtudisctimeout", 1705 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1706 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1707 CTL_NET, PF_INET, IPPROTO_IP, 1708 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1709#ifndef IPNOPRIVPORTS 1710 sysctl_createv(clog, 0, NULL, NULL, 1711 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1712 CTLTYPE_INT, "lowportmin", 1713 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1714 "to assign"), 1715 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1716 CTL_NET, PF_INET, IPPROTO_IP, 1717 IPCTL_LOWPORTMIN, CTL_EOL); 1718 sysctl_createv(clog, 0, NULL, NULL, 1719 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1720 CTLTYPE_INT, "lowportmax", 1721 SYSCTL_DESCR("Highest privileged ephemeral port number " 1722 "to assign"), 1723 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1724 CTL_NET, PF_INET, IPPROTO_IP, 1725 IPCTL_LOWPORTMAX, CTL_EOL); 1726#endif /* IPNOPRIVPORTS */ 1727#if NGRE > 0 1728 sysctl_createv(clog, 0, NULL, NULL, 1729 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1730 CTLTYPE_INT, "grettl", 1731 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1732 NULL, 0, &ip_gre_ttl, 0, 1733 CTL_NET, PF_INET, IPPROTO_IP, 1734 IPCTL_GRE_TTL, CTL_EOL); 1735#endif /* NGRE */ 1736 sysctl_createv(clog, 0, NULL, NULL, 1737 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1738 CTLTYPE_INT, "checkinterface", 1739 SYSCTL_DESCR("Enable receive side of Strong ES model " 1740 "from RFC1122"), 1741 NULL, 0, &ip_checkinterface, 0, 1742 CTL_NET, PF_INET, IPPROTO_IP, 1743 IPCTL_CHECKINTERFACE, CTL_EOL); 1744 sysctl_createv(clog, 0, NULL, NULL, 1745 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1746 CTLTYPE_INT, "random_id", 1747 SYSCTL_DESCR("Assign random ip_id values"), 1748 NULL, 0, &ip_do_randomid, 0, 1749 CTL_NET, PF_INET, IPPROTO_IP, 1750 IPCTL_RANDOMID, CTL_EOL); 1751 sysctl_createv(clog, 0, NULL, NULL, 1752 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1753 CTLTYPE_INT, "do_loopback_cksum", 1754 SYSCTL_DESCR("Perform IP checksum on loopback"), 1755 NULL, 0, &ip_do_loopback_cksum, 0, 1756 CTL_NET, PF_INET, IPPROTO_IP, 1757 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1758 sysctl_createv(clog, 0, NULL, NULL, 1759 CTLFLAG_PERMANENT, 1760 CTLTYPE_STRUCT, "stats", 1761 SYSCTL_DESCR("IP statistics"), 1762 sysctl_net_inet_ip_stats, 0, NULL, 0, 1763 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1764 CTL_EOL); 1765#if NARP 1766 sysctl_createv(clog, 0, NULL, NULL, 1767 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1768 CTLTYPE_INT, "dad_count", 1769 SYSCTL_DESCR("Number of Duplicate Address Detection " 1770 "probes to send"), 1771 NULL, 0, &ip_dad_count, 0, 1772 CTL_NET, PF_INET, IPPROTO_IP, 1773 IPCTL_DAD_COUNT, CTL_EOL); 1774#endif 1775 1776 /* anonportalgo RFC6056 subtree */ 1777 const struct sysctlnode *portalgo_node; 1778 sysctl_createv(clog, 0, NULL, &portalgo_node, 1779 CTLFLAG_PERMANENT, 1780 CTLTYPE_NODE, "anonportalgo", 1781 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1782 NULL, 0, NULL, 0, 1783 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1784 sysctl_createv(clog, 0, &portalgo_node, NULL, 1785 CTLFLAG_PERMANENT, 1786 CTLTYPE_STRING, "available", 1787 SYSCTL_DESCR("available algorithms"), 1788 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1789 CTL_CREATE, CTL_EOL); 1790 sysctl_createv(clog, 0, &portalgo_node, NULL, 1791 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1792 CTLTYPE_STRING, "selected", 1793 SYSCTL_DESCR("selected algorithm"), 1794 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1795 CTL_CREATE, CTL_EOL); 1796 sysctl_createv(clog, 0, &portalgo_node, NULL, 1797 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1798 CTLTYPE_STRUCT, "reserve", 1799 SYSCTL_DESCR("bitmap of reserved ports"), 1800 sysctl_portalgo_reserve4, 0, NULL, 0, 1801 CTL_CREATE, CTL_EOL); 1802} 1803 1804void 1805ip_statinc(u_int stat) 1806{ 1807 1808 KASSERT(stat < IP_NSTATS); 1809 IP_STATINC(stat); 1810} 1811