ip_input.c revision 1.397
1/* $NetBSD: ip_input.c,v 1.397 2020/08/28 06:31:42 ozaki-r 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 1.397 2020/08/28 06:31:42 ozaki-r 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 *, struct ifnet *); 247static void ip_forward(struct mbuf *, int, struct ifnet *); 248static bool ip_dooptions(struct mbuf *); 249static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *); 250static void sysctl_net_inet_ip_setup(struct sysctllog **); 251 252static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *, 253 int *); 254static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *, 255 struct ip *); 256 257#ifdef NET_MPSAFE 258#define SOFTNET_LOCK() mutex_enter(softnet_lock) 259#define SOFTNET_UNLOCK() mutex_exit(softnet_lock) 260#else 261#define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 262#define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 263#endif 264 265/* 266 * IP initialization: fill in IP protocol switch table. 267 * All protocols not implemented in kernel go to raw IP protocol handler. 268 */ 269void 270ip_init(void) 271{ 272 const struct protosw *pr; 273 274 in_init(); 275 sysctl_net_inet_ip_setup(NULL); 276 277 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 278 KASSERT(pr != NULL); 279 280 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 281 KASSERT(ip_pktq != NULL); 282 283 for (u_int i = 0; i < IPPROTO_MAX; i++) { 284 ip_protox[i] = pr - inetsw; 285 } 286 for (pr = inetdomain.dom_protosw; 287 pr < inetdomain.dom_protoswNPROTOSW; pr++) 288 if (pr->pr_domain->dom_family == PF_INET && 289 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 290 ip_protox[pr->pr_protocol] = pr - inetsw; 291 292 ip_reass_init(); 293 294 ip_ids = ip_id_init(); 295 ip_id = time_uptime & 0xfffff; 296 297#ifdef GATEWAY 298 ipflow_init(); 299#endif 300 301 /* Register our Packet Filter hook. */ 302 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 303 KASSERT(inet_pfil_hook != NULL); 304 305#ifdef MBUFTRACE 306 MOWNER_ATTACH(&ip_tx_mowner); 307 MOWNER_ATTACH(&ip_rx_mowner); 308#endif 309 310 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 311 ipforward_rt_percpu = rtcache_percpu_alloc(); 312 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 313} 314 315static struct in_ifaddr * 316ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 317{ 318 struct in_ifaddr *ia = NULL; 319 int checkif; 320 321 /* 322 * Enable a consistency check between the destination address 323 * and the arrival interface for a unicast packet (the RFC 1122 324 * strong ES model) if IP forwarding is disabled and the packet 325 * is not locally generated. 326 * 327 * XXX - We need to add a per ifaddr flag for this so that 328 * we get finer grain control. 329 */ 330 checkif = ip_checkinterface && (ipforwarding == 0) && 331 (ifp->if_flags & IFF_LOOPBACK) == 0; 332 333 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 334 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 335 if (ia->ia4_flags & IN_IFF_NOTREADY) 336 continue; 337 if (checkif && ia->ia_ifp != ifp) 338 continue; 339 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) { 340 (*downmatch)++; 341 continue; 342 } 343 if (ia->ia4_flags & IN_IFF_DETACHED && 344 (ifp->if_flags & IFF_LOOPBACK) == 0) 345 continue; 346 break; 347 } 348 } 349 350 return ia; 351} 352 353static struct in_ifaddr * 354ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 355{ 356 struct in_ifaddr *ia = NULL; 357 struct ifaddr *ifa; 358 359 IFADDR_READER_FOREACH(ifa, ifp) { 360 if (ifa->ifa_addr->sa_family != AF_INET) 361 continue; 362 ia = ifatoia(ifa); 363 if (ia->ia4_flags & IN_IFF_NOTREADY) 364 continue; 365 if (ia->ia4_flags & IN_IFF_DETACHED && 366 (ifp->if_flags & IFF_LOOPBACK) == 0) 367 continue; 368 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 369 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 370 /* 371 * Look for all-0's host part (old broadcast addr), 372 * either for subnet or net. 373 */ 374 ip->ip_dst.s_addr == ia->ia_subnet || 375 ip->ip_dst.s_addr == ia->ia_net) 376 goto matched; 377 /* 378 * An interface with IP address zero accepts 379 * all packets that arrive on that interface. 380 */ 381 if (in_nullhost(ia->ia_addr.sin_addr)) 382 goto matched; 383 } 384 ia = NULL; 385 386matched: 387 return ia; 388} 389 390/* 391 * IP software interrupt routine. 392 */ 393static void 394ipintr(void *arg __unused) 395{ 396 struct mbuf *m; 397 398 KASSERT(cpu_softintr_p()); 399 400 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 401 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 402 struct ifnet *ifp; 403 struct psref psref; 404 405 ifp = m_get_rcvif_psref(m, &psref); 406 if (__predict_false(ifp == NULL)) { 407 IP_STATINC(IP_STAT_IFDROP); 408 m_freem(m); 409 continue; 410 } 411 412 ip_input(m, ifp); 413 414 m_put_rcvif_psref(ifp, &psref); 415 } 416 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 417} 418 419/* 420 * IP input routine. Checksum and byte swap header. If fragmented 421 * try to reassemble. Process options. Pass to next level. 422 */ 423static void 424ip_input(struct mbuf *m, struct ifnet *ifp) 425{ 426 struct ip *ip = NULL; 427 struct in_ifaddr *ia = NULL; 428 int hlen = 0, len; 429 int downmatch; 430 int srcrt = 0; 431 int s; 432 433 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 434 "interrupt handler; synchronization assumptions violated"); 435 436 MCLAIM(m, &ip_rx_mowner); 437 KASSERT((m->m_flags & M_PKTHDR) != 0); 438 439 /* 440 * If no IP addresses have been set yet but the interfaces 441 * are receiving, can't do anything with incoming packets yet. 442 * Note: we pre-check without locks held. 443 */ 444 if (IN_ADDRLIST_READER_EMPTY()) { 445 IP_STATINC(IP_STAT_IFDROP); 446 goto out; 447 } 448 449 IP_STATINC(IP_STAT_TOTAL); 450 451 /* 452 * If the IP header is not aligned, slurp it up into a new 453 * mbuf with space for link headers, in the event we forward 454 * it. Otherwise, if it is aligned, make sure the entire 455 * base IP header is in the first mbuf of the chain. 456 */ 457 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 458 if ((m = m_copyup(m, sizeof(struct ip), 459 (max_linkhdr + 3) & ~3)) == NULL) { 460 /* XXXJRT new stat, please */ 461 IP_STATINC(IP_STAT_TOOSMALL); 462 goto out; 463 } 464 } else if (__predict_false(m->m_len < sizeof(struct ip))) { 465 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) { 466 IP_STATINC(IP_STAT_TOOSMALL); 467 goto out; 468 } 469 } 470 ip = mtod(m, struct ip *); 471 if (ip->ip_v != IPVERSION) { 472 IP_STATINC(IP_STAT_BADVERS); 473 goto out; 474 } 475 hlen = ip->ip_hl << 2; 476 if (hlen < sizeof(struct ip)) { /* minimum header length */ 477 IP_STATINC(IP_STAT_BADHLEN); 478 goto out; 479 } 480 if (hlen > m->m_len) { 481 if ((m = m_pullup(m, hlen)) == NULL) { 482 IP_STATINC(IP_STAT_BADHLEN); 483 goto out; 484 } 485 ip = mtod(m, struct ip *); 486 } 487 488 /* 489 * RFC1122: packets with a multicast source address are 490 * not allowed. 491 */ 492 if (IN_MULTICAST(ip->ip_src.s_addr)) { 493 IP_STATINC(IP_STAT_BADADDR); 494 goto out; 495 } 496 497 /* 127/8 must not appear on wire - RFC1122 */ 498 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 499 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 500 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 501 IP_STATINC(IP_STAT_BADADDR); 502 goto out; 503 } 504 } 505 506 switch (m->m_pkthdr.csum_flags & 507 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) { 508 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 509 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 510 IP_STATINC(IP_STAT_BADSUM); 511 goto out; 512 513 case M_CSUM_IPv4: 514 /* Checksum was okay. */ 515 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 516 break; 517 518 default: 519 /* 520 * Must compute it ourselves. Maybe skip checksum on 521 * loopback interfaces. 522 */ 523 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 524 ip_do_loopback_cksum)) { 525 INET_CSUM_COUNTER_INCR(&ip_swcsum); 526 if (in_cksum(m, hlen) != 0) { 527 IP_STATINC(IP_STAT_BADSUM); 528 goto out; 529 } 530 } 531 break; 532 } 533 534 /* Retrieve the packet length. */ 535 len = ntohs(ip->ip_len); 536 537 /* 538 * Check for additional length bogosity 539 */ 540 if (len < hlen) { 541 IP_STATINC(IP_STAT_BADLEN); 542 goto out; 543 } 544 545 /* 546 * Check that the amount of data in the buffers is at least as much 547 * as the IP header would have us expect. Trim mbufs if longer than 548 * we expect. Drop packet if shorter than we expect. 549 */ 550 if (m->m_pkthdr.len < len) { 551 IP_STATINC(IP_STAT_TOOSHORT); 552 goto out; 553 } 554 if (m->m_pkthdr.len > len) { 555 if (m->m_len == m->m_pkthdr.len) { 556 m->m_len = len; 557 m->m_pkthdr.len = len; 558 } else 559 m_adj(m, len - m->m_pkthdr.len); 560 } 561 562 /* 563 * Assume that we can create a fast-forward IP flow entry 564 * based on this packet. 565 */ 566 m->m_flags |= M_CANFASTFWD; 567 568 /* 569 * Run through list of hooks for input packets. If there are any 570 * filters which require that additional packets in the flow are 571 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 572 * Note that filters must _never_ set this flag, as another filter 573 * in the list may have previously cleared it. 574 * 575 * Don't call hooks if the packet has already been processed by 576 * IPsec (encapsulated, tunnel mode). 577 */ 578#if defined(IPSEC) 579 if (!ipsec_used || !ipsec_skip_pfil(m)) 580#else 581 if (1) 582#endif 583 { 584 struct in_addr odst = ip->ip_dst; 585 bool freed; 586 587 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 588 if (freed || m == NULL) { 589 m = NULL; 590 IP_STATINC(IP_STAT_PFILDROP_IN); 591 goto out; 592 } 593 if (__predict_false(m->m_len < sizeof(struct ip))) { 594 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) { 595 IP_STATINC(IP_STAT_TOOSMALL); 596 goto out; 597 } 598 } 599 ip = mtod(m, struct ip *); 600 hlen = ip->ip_hl << 2; 601 if (hlen < sizeof(struct ip)) { /* minimum header length */ 602 IP_STATINC(IP_STAT_BADHLEN); 603 goto out; 604 } 605 if (hlen > m->m_len) { 606 if ((m = m_pullup(m, hlen)) == NULL) { 607 IP_STATINC(IP_STAT_BADHLEN); 608 goto out; 609 } 610 ip = mtod(m, struct ip *); 611 } 612 613 /* 614 * XXX The setting of "srcrt" here is to prevent ip_forward() 615 * from generating ICMP redirects for packets that have 616 * been redirected by a hook back out on to the same LAN that 617 * they came from and is not an indication that the packet 618 * is being influenced by source routing options. This 619 * allows things like 620 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 621 * where tlp0 is both on the 1.1.1.0/24 network and is the 622 * default route for hosts on 1.1.1.0/24. Of course this 623 * also requires a "map tlp0 ..." to complete the story. 624 * One might argue whether or not this kind of network config. 625 * should be supported in this manner... 626 */ 627 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 628 } 629 630#ifdef ALTQ 631 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 632 if (altq_input) { 633 SOFTNET_LOCK(); 634 if ((*altq_input)(m, AF_INET) == 0) { 635 /* Packet dropped by traffic conditioner. */ 636 SOFTNET_UNLOCK(); 637 m = NULL; 638 goto out; 639 } 640 SOFTNET_UNLOCK(); 641 } 642#endif 643 644 /* 645 * Process options and, if not destined for us, 646 * ship it on. ip_dooptions returns 1 when an 647 * error was detected (causing an icmp message 648 * to be sent and the original packet to be freed). 649 */ 650 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 651 m = NULL; 652 goto out; 653 } 654 655 /* 656 * Check our list of addresses, to see if the packet is for us. 657 * 658 * Traditional 4.4BSD did not consult IFF_UP at all. 659 * The behavior here is to treat addresses on !IFF_UP interface 660 * or IN_IFF_NOTREADY addresses as not mine. 661 */ 662 downmatch = 0; 663 s = pserialize_read_enter(); 664 ia = ip_match_our_address(ifp, ip, &downmatch); 665 if (ia != NULL) { 666 pserialize_read_exit(s); 667 goto ours; 668 } 669 670 if (ifp->if_flags & IFF_BROADCAST) { 671 ia = ip_match_our_address_broadcast(ifp, ip); 672 if (ia != NULL) { 673 pserialize_read_exit(s); 674 goto ours; 675 } 676 } 677 pserialize_read_exit(s); 678 679 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 680#ifdef MROUTING 681 extern struct socket *ip_mrouter; 682 683 if (ip_mrouter) { 684 /* 685 * If we are acting as a multicast router, all 686 * incoming multicast packets are passed to the 687 * kernel-level multicast forwarding function. 688 * The packet is returned (relatively) intact; if 689 * ip_mforward() returns a non-zero value, the packet 690 * must be discarded, else it may be accepted below. 691 * 692 * (The IP ident field is put in the same byte order 693 * as expected when ip_mforward() is called from 694 * ip_output().) 695 */ 696 SOFTNET_LOCK(); 697 if (ip_mforward(m, ifp) != 0) { 698 SOFTNET_UNLOCK(); 699 IP_STATINC(IP_STAT_CANTFORWARD); 700 goto out; 701 } 702 SOFTNET_UNLOCK(); 703 704 /* 705 * The process-level routing demon needs to receive 706 * all multicast IGMP packets, whether or not this 707 * host belongs to their destination groups. 708 */ 709 if (ip->ip_p == IPPROTO_IGMP) { 710 goto ours; 711 } 712 IP_STATINC(IP_STAT_CANTFORWARD); 713 } 714#endif 715 /* 716 * See if we belong to the destination multicast group on the 717 * arrival interface. 718 */ 719 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 720 IP_STATINC(IP_STAT_CANTFORWARD); 721 goto out; 722 } 723 goto ours; 724 } 725 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 726 in_nullhost(ip->ip_dst)) 727 goto ours; 728 729 /* 730 * Not for us; forward if possible and desirable. 731 */ 732 if (ipforwarding == 0) { 733 IP_STATINC(IP_STAT_CANTFORWARD); 734 m_freem(m); 735 } else { 736 /* 737 * If ip_dst matched any of my address on !IFF_UP interface, 738 * and there's no IFF_UP interface that matches ip_dst, 739 * send icmp unreach. Forwarding it will result in in-kernel 740 * forwarding loop till TTL goes to 0. 741 */ 742 if (downmatch) { 743 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 744 IP_STATINC(IP_STAT_CANTFORWARD); 745 return; 746 } 747#ifdef IPSEC 748 /* Check the security policy (SP) for the packet */ 749 if (ipsec_used) { 750 if (ipsec_ip_input_checkpolicy(m, true) != 0) { 751 IP_STATINC(IP_STAT_IPSECDROP_IN); 752 goto out; 753 } 754 } 755#endif 756 ip_forward(m, srcrt, ifp); 757 } 758 return; 759 760ours: 761 /* 762 * If offset or IP_MF are set, must reassemble. 763 */ 764 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 765 /* 766 * Pass to IP reassembly mechanism. 767 */ 768 if (ip_reass_packet(&m) != 0) { 769 /* Failed; invalid fragment(s) or packet. */ 770 goto out; 771 } 772 if (m == NULL) { 773 /* More fragments should come; silently return. */ 774 goto out; 775 } 776 /* 777 * Reassembly is done, we have the final packet. 778 * Update cached data in local variable(s). 779 */ 780 ip = mtod(m, struct ip *); 781 hlen = ip->ip_hl << 2; 782 } 783 784 M_VERIFY_PACKET(m); 785 786#ifdef IPSEC 787 /* 788 * Enforce IPsec policy checking if we are seeing last header. 789 * Note that we do not visit this with protocols with PCB layer 790 * code - like UDP/TCP/raw IP. 791 */ 792 if (ipsec_used && 793 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 794 if (ipsec_ip_input_checkpolicy(m, false) != 0) { 795 IP_STATINC(IP_STAT_IPSECDROP_IN); 796 goto out; 797 } 798 } 799#endif 800 801 /* 802 * Switch out to protocol's input routine. 803 */ 804#if IFA_STATS 805 if (ia) { 806 struct in_ifaddr *_ia; 807 /* 808 * Keep a reference from ip_match_our_address with psref 809 * is expensive, so explore ia here again. 810 */ 811 s = pserialize_read_enter(); 812 _ia = in_get_ia(ip->ip_dst); 813 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 814 pserialize_read_exit(s); 815 } 816#endif 817 IP_STATINC(IP_STAT_DELIVERED); 818 819 const int off = hlen, nh = ip->ip_p; 820 821 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 822 return; 823 824out: 825 if (m != NULL) 826 m_freem(m); 827} 828 829/* 830 * IP timer processing. 831 */ 832void 833ip_slowtimo(void) 834{ 835 836 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 837 838 ip_reass_slowtimo(); 839 840 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 841} 842 843/* 844 * IP drain processing. 845 */ 846void 847ip_drain(void) 848{ 849 850 KERNEL_LOCK(1, NULL); 851 ip_reass_drain(); 852 KERNEL_UNLOCK_ONE(NULL); 853} 854 855/* 856 * ip_dooptions: perform option processing on a datagram, possibly discarding 857 * it if bad options are encountered, or forwarding it if source-routed. 858 * 859 * => Returns true if packet has been forwarded/freed. 860 * => Returns false if the packet should be processed further. 861 */ 862static bool 863ip_dooptions(struct mbuf *m) 864{ 865 struct ip *ip = mtod(m, struct ip *); 866 u_char *cp, *cp0; 867 struct ip_timestamp *ipt; 868 struct in_ifaddr *ia; 869 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 870 int srr_present, rr_present, ts_present; 871 struct in_addr dst; 872 n_time ntime; 873 struct ifaddr *ifa = NULL; 874 int s; 875 876 srr_present = 0; 877 rr_present = 0; 878 ts_present = 0; 879 880 dst = ip->ip_dst; 881 cp = (u_char *)(ip + 1); 882 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 883 for (; cnt > 0; cnt -= optlen, cp += optlen) { 884 opt = cp[IPOPT_OPTVAL]; 885 if (opt == IPOPT_EOL) 886 break; 887 if (opt == IPOPT_NOP) 888 optlen = 1; 889 else { 890 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 891 code = &cp[IPOPT_OLEN] - (u_char *)ip; 892 goto bad; 893 } 894 optlen = cp[IPOPT_OLEN]; 895 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 896 code = &cp[IPOPT_OLEN] - (u_char *)ip; 897 goto bad; 898 } 899 } 900 switch (opt) { 901 902 default: 903 break; 904 905 /* 906 * Source routing with record. 907 * Find interface with current destination address. 908 * If none on this machine then drop if strictly routed, 909 * or do nothing if loosely routed. 910 * Record interface address and bring up next address 911 * component. If strictly routed make sure next 912 * address is on directly accessible net. 913 */ 914 case IPOPT_LSRR: 915 case IPOPT_SSRR: { 916 struct psref psref; 917 struct sockaddr_in ipaddr = { 918 .sin_len = sizeof(ipaddr), 919 .sin_family = AF_INET, 920 }; 921 922 if (ip_allowsrcrt == 0) { 923 type = ICMP_UNREACH; 924 code = ICMP_UNREACH_NET_PROHIB; 925 goto bad; 926 } 927 if (srr_present++) { 928 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 929 goto bad; 930 } 931 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 932 code = &cp[IPOPT_OLEN] - (u_char *)ip; 933 goto bad; 934 } 935 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 936 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 937 goto bad; 938 } 939 ipaddr.sin_addr = ip->ip_dst; 940 941 s = pserialize_read_enter(); 942 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 943 if (ifa == NULL) { 944 pserialize_read_exit(s); 945 if (opt == IPOPT_SSRR) { 946 type = ICMP_UNREACH; 947 code = ICMP_UNREACH_SRCFAIL; 948 goto bad; 949 } 950 /* 951 * Loose routing, and not at next destination 952 * yet; nothing to do except forward. 953 */ 954 break; 955 } 956 pserialize_read_exit(s); 957 958 off--; /* 0 origin */ 959 if ((off + sizeof(struct in_addr)) > optlen) { 960 /* 961 * End of source route. Should be for us. 962 */ 963 save_rte(m, cp, ip->ip_src); 964 break; 965 } 966 /* 967 * locate outgoing interface 968 */ 969 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 970 sizeof(ipaddr.sin_addr)); 971 if (opt == IPOPT_SSRR) { 972 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 973 &psref); 974 if (ifa != NULL) 975 ia = ifatoia(ifa); 976 else 977 ia = NULL; 978 } else { 979 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 980 } 981 if (ia == NULL) { 982 type = ICMP_UNREACH; 983 code = ICMP_UNREACH_SRCFAIL; 984 goto bad; 985 } 986 ip->ip_dst = ipaddr.sin_addr; 987 memcpy(cp + off, &ia->ia_addr.sin_addr, 988 sizeof(struct in_addr)); 989 ia4_release(ia, &psref); 990 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 991 /* 992 * Let ip_intr's mcast routing check handle mcast pkts 993 */ 994 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 995 break; 996 } 997 998 case IPOPT_RR: { 999 struct psref psref; 1000 struct sockaddr_in ipaddr = { 1001 .sin_len = sizeof(ipaddr), 1002 .sin_family = AF_INET, 1003 }; 1004 1005 if (rr_present++) { 1006 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1007 goto bad; 1008 } 1009 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1010 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1011 goto bad; 1012 } 1013 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1014 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1015 goto bad; 1016 } 1017 /* 1018 * If no space remains, ignore. 1019 */ 1020 off--; /* 0 origin */ 1021 if ((off + sizeof(struct in_addr)) > optlen) 1022 break; 1023 memcpy((void *)&ipaddr.sin_addr, (void *)&ip->ip_dst, 1024 sizeof(ipaddr.sin_addr)); 1025 /* 1026 * locate outgoing interface; if we're the destination, 1027 * use the incoming interface (should be same). 1028 */ 1029 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1030 if (ifa == NULL) { 1031 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1032 if (ia == NULL) { 1033 type = ICMP_UNREACH; 1034 code = ICMP_UNREACH_HOST; 1035 goto bad; 1036 } 1037 } else { 1038 ia = ifatoia(ifa); 1039 } 1040 memcpy(cp + off, &ia->ia_addr.sin_addr, 1041 sizeof(struct in_addr)); 1042 ia4_release(ia, &psref); 1043 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1044 break; 1045 } 1046 1047 case IPOPT_TS: 1048 code = cp - (u_char *)ip; 1049 ipt = (struct ip_timestamp *)cp; 1050 if (ts_present++) { 1051 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1052 goto bad; 1053 } 1054 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1055 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1056 goto bad; 1057 } 1058 if (ipt->ipt_ptr < 5) { 1059 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1060 goto bad; 1061 } 1062 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1063 if (++ipt->ipt_oflw == 0) { 1064 code = (u_char *)&ipt->ipt_ptr - 1065 (u_char *)ip; 1066 goto bad; 1067 } 1068 break; 1069 } 1070 cp0 = (cp + ipt->ipt_ptr - 1); 1071 switch (ipt->ipt_flg) { 1072 1073 case IPOPT_TS_TSONLY: 1074 break; 1075 1076 case IPOPT_TS_TSANDADDR: { 1077 struct ifnet *rcvif; 1078 int _s, _ss; 1079 struct sockaddr_in ipaddr = { 1080 .sin_len = sizeof(ipaddr), 1081 .sin_family = AF_INET, 1082 }; 1083 1084 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1085 sizeof(struct in_addr) > ipt->ipt_len) { 1086 code = (u_char *)&ipt->ipt_ptr - 1087 (u_char *)ip; 1088 goto bad; 1089 } 1090 ipaddr.sin_addr = dst; 1091 _ss = pserialize_read_enter(); 1092 rcvif = m_get_rcvif(m, &_s); 1093 if (__predict_true(rcvif != NULL)) { 1094 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1095 rcvif); 1096 } 1097 m_put_rcvif(rcvif, &_s); 1098 if (ifa == NULL) { 1099 pserialize_read_exit(_ss); 1100 break; 1101 } 1102 ia = ifatoia(ifa); 1103 memcpy(cp0, &ia->ia_addr.sin_addr, 1104 sizeof(struct in_addr)); 1105 pserialize_read_exit(_ss); 1106 ipt->ipt_ptr += sizeof(struct in_addr); 1107 break; 1108 } 1109 1110 case IPOPT_TS_PRESPEC: { 1111 struct sockaddr_in ipaddr = { 1112 .sin_len = sizeof(ipaddr), 1113 .sin_family = AF_INET, 1114 }; 1115 1116 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1117 sizeof(struct in_addr) > ipt->ipt_len) { 1118 code = (u_char *)&ipt->ipt_ptr - 1119 (u_char *)ip; 1120 goto bad; 1121 } 1122 memcpy(&ipaddr.sin_addr, cp0, 1123 sizeof(struct in_addr)); 1124 s = pserialize_read_enter(); 1125 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1126 if (ifa == NULL) { 1127 pserialize_read_exit(s); 1128 continue; 1129 } 1130 pserialize_read_exit(s); 1131 ipt->ipt_ptr += sizeof(struct in_addr); 1132 break; 1133 } 1134 1135 default: 1136 /* XXX can't take &ipt->ipt_flg */ 1137 code = (u_char *)&ipt->ipt_ptr - 1138 (u_char *)ip + 1; 1139 goto bad; 1140 } 1141 ntime = iptime(); 1142 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1143 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1144 sizeof(n_time)); 1145 ipt->ipt_ptr += sizeof(n_time); 1146 } 1147 } 1148 if (forward) { 1149 struct ifnet *rcvif; 1150 struct psref _psref; 1151 1152 if (ip_forwsrcrt == 0) { 1153 type = ICMP_UNREACH; 1154 code = ICMP_UNREACH_SRCFAIL; 1155 goto bad; 1156 } 1157 1158 rcvif = m_get_rcvif_psref(m, &_psref); 1159 if (__predict_false(rcvif == NULL)) { 1160 type = ICMP_UNREACH; 1161 code = ICMP_UNREACH_HOST; 1162 goto bad; 1163 } 1164 ip_forward(m, 1, rcvif); 1165 m_put_rcvif_psref(rcvif, &_psref); 1166 return true; 1167 } 1168 return false; 1169bad: 1170 icmp_error(m, type, code, 0, 0); 1171 IP_STATINC(IP_STAT_BADOPTIONS); 1172 return true; 1173} 1174 1175/* 1176 * ip_rtaddr: given address of next destination (final or next hop), 1177 * return internet address info of interface to be used to get there. 1178 */ 1179static struct in_ifaddr * 1180ip_rtaddr(struct in_addr dst, struct psref *psref) 1181{ 1182 struct rtentry *rt; 1183 union { 1184 struct sockaddr dst; 1185 struct sockaddr_in dst4; 1186 } u; 1187 struct route *ro; 1188 1189 sockaddr_in_init(&u.dst4, &dst, 0); 1190 1191 ro = rtcache_percpu_getref(ipforward_rt_percpu); 1192 rt = rtcache_lookup(ro, &u.dst); 1193 if (rt == NULL) { 1194 rtcache_percpu_putref(ipforward_rt_percpu); 1195 return NULL; 1196 } 1197 1198 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1199 rtcache_unref(rt, ro); 1200 rtcache_percpu_putref(ipforward_rt_percpu); 1201 1202 return ifatoia(rt->rt_ifa); 1203} 1204 1205/* 1206 * save_rte: save incoming source route for use in replies, to be picked 1207 * up later by ip_srcroute if the receiver is interested. 1208 */ 1209static void 1210save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1211{ 1212 struct ip_srcrt *isr; 1213 struct m_tag *mtag; 1214 unsigned olen; 1215 1216 olen = option[IPOPT_OLEN]; 1217 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1218 return; 1219 1220 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1221 if (mtag == NULL) 1222 return; 1223 isr = (struct ip_srcrt *)(mtag + 1); 1224 1225 memcpy(isr->isr_hdr, option, olen); 1226 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1227 isr->isr_dst = dst; 1228 m_tag_prepend(m, mtag); 1229} 1230 1231/* 1232 * Retrieve incoming source route for use in replies, 1233 * in the same form used by setsockopt. 1234 * The first hop is placed before the options, will be removed later. 1235 */ 1236struct mbuf * 1237ip_srcroute(struct mbuf *m0) 1238{ 1239 struct in_addr *p, *q; 1240 struct mbuf *m; 1241 struct ip_srcrt *isr; 1242 struct m_tag *mtag; 1243 1244 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE); 1245 if (mtag == NULL) 1246 return NULL; 1247 isr = (struct ip_srcrt *)(mtag + 1); 1248 1249 if (isr->isr_nhops == 0) 1250 return NULL; 1251 1252 m = m_get(M_DONTWAIT, MT_SOOPTS); 1253 if (m == NULL) 1254 return NULL; 1255 1256 MCLAIM(m, &inetdomain.dom_mowner); 1257#define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1258 1259 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1260 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1261 1262 /* 1263 * First save first hop for return route 1264 */ 1265 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1266 *(mtod(m, struct in_addr *)) = *p--; 1267 1268 /* 1269 * Copy option fields and padding (nop) to mbuf. 1270 */ 1271 isr->isr_nop = IPOPT_NOP; 1272 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1273 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1274 OPTSIZ); 1275 q = (struct in_addr *)(mtod(m, char *) + 1276 sizeof(struct in_addr) + OPTSIZ); 1277#undef OPTSIZ 1278 /* 1279 * Record return path as an IP source route, 1280 * reversing the path (pointers are now aligned). 1281 */ 1282 while (p >= isr->isr_routes) { 1283 *q++ = *p--; 1284 } 1285 /* 1286 * Last hop goes to final destination. 1287 */ 1288 *q = isr->isr_dst; 1289 m_tag_delete(m0, mtag); 1290 return m; 1291} 1292 1293const int inetctlerrmap[PRC_NCMDS] = { 1294 [PRC_MSGSIZE] = EMSGSIZE, 1295 [PRC_HOSTDEAD] = EHOSTDOWN, 1296 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1297 [PRC_UNREACH_NET] = EHOSTUNREACH, 1298 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1299 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1300 [PRC_UNREACH_PORT] = ECONNREFUSED, 1301 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1302 [PRC_PARAMPROB] = ENOPROTOOPT, 1303}; 1304 1305void 1306ip_fasttimo(void) 1307{ 1308 if (ip_drainwanted) { 1309 ip_drain(); 1310 ip_drainwanted = 0; 1311 } 1312} 1313 1314void 1315ip_drainstub(void) 1316{ 1317 ip_drainwanted = 1; 1318} 1319 1320/* 1321 * Forward a packet. If some error occurs return the sender 1322 * an icmp packet. Note we can't always generate a meaningful 1323 * icmp message because icmp doesn't have a large enough repertoire 1324 * of codes and types. 1325 * 1326 * If not forwarding, just drop the packet. This could be confusing 1327 * if ipforwarding was zero but some routing protocol was advancing 1328 * us as a gateway to somewhere. However, we must let the routing 1329 * protocol deal with that. 1330 * 1331 * The srcrt parameter indicates whether the packet is being forwarded 1332 * via a source route. 1333 */ 1334static void 1335ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1336{ 1337 struct ip *ip = mtod(m, struct ip *); 1338 struct rtentry *rt; 1339 int error, type = 0, code = 0, destmtu = 0; 1340 struct mbuf *mcopy; 1341 n_long dest; 1342 union { 1343 struct sockaddr dst; 1344 struct sockaddr_in dst4; 1345 } u; 1346 uint64_t *ips; 1347 struct route *ro; 1348 1349 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1350 "interrupt handler; synchronization assumptions violated"); 1351 1352 /* 1353 * We are now in the output path. 1354 */ 1355 MCLAIM(m, &ip_tx_mowner); 1356 1357 /* 1358 * Clear any in-bound checksum flags for this packet. 1359 */ 1360 m->m_pkthdr.csum_flags = 0; 1361 1362 dest = 0; 1363 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1364 IP_STATINC(IP_STAT_CANTFORWARD); 1365 m_freem(m); 1366 return; 1367 } 1368 1369 if (ip->ip_ttl <= IPTTLDEC) { 1370 IP_STATINC(IP_STAT_TIMXCEED); 1371 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1372 return; 1373 } 1374 1375 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1376 1377 ro = rtcache_percpu_getref(ipforward_rt_percpu); 1378 rt = rtcache_lookup(ro, &u.dst); 1379 if (rt == NULL) { 1380 rtcache_percpu_putref(ipforward_rt_percpu); 1381 IP_STATINC(IP_STAT_NOROUTE); 1382 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1383 return; 1384 } 1385 1386 /* 1387 * Save at most 68 bytes of the packet in case 1388 * we need to generate an ICMP message to the src. 1389 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1390 */ 1391 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1392 if (mcopy) 1393 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1394 1395 ip->ip_ttl -= IPTTLDEC; 1396 1397 /* 1398 * If forwarding packet using same interface that it came in on, 1399 * perhaps should send a redirect to sender to shortcut a hop. 1400 * Only send redirect if source is sending directly to us, 1401 * and if packet was not source routed (or has any options). 1402 * Also, don't send redirect if forwarding using a default route 1403 * or a route modified by a redirect. 1404 */ 1405 if (rt->rt_ifp == rcvif && 1406 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1407 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1408 ipsendredirects && !srcrt) { 1409 if ((ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1410 ifatoia(rt->rt_ifa)->ia_subnet) { 1411 if (rt->rt_flags & RTF_GATEWAY) 1412 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1413 else 1414 dest = ip->ip_dst.s_addr; 1415 /* 1416 * Router requirements says to only send host 1417 * redirects. 1418 */ 1419 type = ICMP_REDIRECT; 1420 code = ICMP_REDIRECT_HOST; 1421 } 1422 } 1423 rtcache_unref(rt, ro); 1424 1425 error = ip_output(m, NULL, ro, 1426 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1427 NULL, NULL); 1428 1429 if (error) { 1430 IP_STATINC(IP_STAT_CANTFORWARD); 1431 goto error; 1432 } 1433 1434 ips = IP_STAT_GETREF(); 1435 ips[IP_STAT_FORWARD]++; 1436 1437 if (type) { 1438 ips[IP_STAT_REDIRECTSENT]++; 1439 IP_STAT_PUTREF(); 1440 goto redirect; 1441 } 1442 1443 IP_STAT_PUTREF(); 1444 if (mcopy) { 1445#ifdef GATEWAY 1446 if (mcopy->m_flags & M_CANFASTFWD) 1447 ipflow_create(ro, mcopy); 1448#endif 1449 m_freem(mcopy); 1450 } 1451 1452 rtcache_percpu_putref(ipforward_rt_percpu); 1453 return; 1454 1455redirect: 1456error: 1457 if (mcopy == NULL) { 1458 rtcache_percpu_putref(ipforward_rt_percpu); 1459 return; 1460 } 1461 1462 switch (error) { 1463 1464 case 0: /* forwarded, but need redirect */ 1465 /* type, code set above */ 1466 break; 1467 1468 case ENETUNREACH: /* shouldn't happen, checked above */ 1469 case EHOSTUNREACH: 1470 case ENETDOWN: 1471 case EHOSTDOWN: 1472 default: 1473 type = ICMP_UNREACH; 1474 code = ICMP_UNREACH_HOST; 1475 break; 1476 1477 case EMSGSIZE: 1478 type = ICMP_UNREACH; 1479 code = ICMP_UNREACH_NEEDFRAG; 1480 1481 if ((rt = rtcache_validate(ro)) != NULL) { 1482 destmtu = rt->rt_ifp->if_mtu; 1483 rtcache_unref(rt, ro); 1484 } 1485#ifdef IPSEC 1486 if (ipsec_used) 1487 ipsec_mtu(mcopy, &destmtu); 1488#endif 1489 IP_STATINC(IP_STAT_CANTFRAG); 1490 break; 1491 1492 case ENOBUFS: 1493 /* 1494 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1495 * Requirements for IP Version 4 Routers. Source quench can 1496 * be a big problem under DoS attacks or if the underlying 1497 * interface is rate-limited. 1498 */ 1499 if (mcopy) 1500 m_freem(mcopy); 1501 rtcache_percpu_putref(ipforward_rt_percpu); 1502 return; 1503 } 1504 icmp_error(mcopy, type, code, dest, destmtu); 1505 rtcache_percpu_putref(ipforward_rt_percpu); 1506} 1507 1508void 1509ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1510 struct mbuf *m) 1511{ 1512 struct socket *so = inp->inp_socket; 1513 int inpflags = inp->inp_flags; 1514 1515 if (SOOPT_TIMESTAMP(so->so_options)) 1516 mp = sbsavetimestamp(so->so_options, mp); 1517 1518 if (inpflags & INP_RECVDSTADDR) { 1519 *mp = sbcreatecontrol(&ip->ip_dst, 1520 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1521 if (*mp) 1522 mp = &(*mp)->m_next; 1523 } 1524 1525 if (inpflags & INP_RECVTTL) { 1526 *mp = sbcreatecontrol(&ip->ip_ttl, 1527 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1528 if (*mp) 1529 mp = &(*mp)->m_next; 1530 } 1531 1532 struct psref psref; 1533 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1534 if (__predict_false(ifp == NULL)) { 1535#ifdef DIAGNOSTIC 1536 printf("%s: missing receive interface\n", __func__); 1537#endif 1538 return; /* XXX should report error? */ 1539 } 1540 1541 if (inpflags & INP_RECVPKTINFO) { 1542 struct in_pktinfo ipi; 1543 ipi.ipi_addr = ip->ip_dst; 1544 ipi.ipi_ifindex = ifp->if_index; 1545 *mp = sbcreatecontrol(&ipi, 1546 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1547 if (*mp) 1548 mp = &(*mp)->m_next; 1549 } 1550 if (inpflags & INP_RECVIF) { 1551 struct sockaddr_dl sdl; 1552 1553 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1554 NULL, 0); 1555 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1556 if (*mp) 1557 mp = &(*mp)->m_next; 1558 } 1559 m_put_rcvif_psref(ifp, &psref); 1560} 1561 1562/* 1563 * sysctl helper routine for net.inet.ip.forwsrcrt. 1564 */ 1565static int 1566sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1567{ 1568 int error, tmp; 1569 struct sysctlnode node; 1570 1571 node = *rnode; 1572 tmp = ip_forwsrcrt; 1573 node.sysctl_data = &tmp; 1574 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1575 if (error || newp == NULL) 1576 return (error); 1577 1578 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1579 0, NULL, NULL, NULL); 1580 if (error) 1581 return (error); 1582 1583 ip_forwsrcrt = tmp; 1584 1585 return (0); 1586} 1587 1588/* 1589 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1590 * range of the new value and tweaks timers if it changes. 1591 */ 1592static int 1593sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1594{ 1595 int error, tmp; 1596 struct sysctlnode node; 1597 1598 icmp_mtudisc_lock(); 1599 1600 node = *rnode; 1601 tmp = ip_mtudisc_timeout; 1602 node.sysctl_data = &tmp; 1603 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1604 if (error || newp == NULL) 1605 goto out; 1606 if (tmp < 0) { 1607 error = EINVAL; 1608 goto out; 1609 } 1610 1611 ip_mtudisc_timeout = tmp; 1612 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1613 error = 0; 1614out: 1615 icmp_mtudisc_unlock(); 1616 return error; 1617} 1618 1619static int 1620sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1621{ 1622 1623 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1624} 1625 1626static void 1627sysctl_net_inet_ip_setup(struct sysctllog **clog) 1628{ 1629 sysctl_createv(clog, 0, NULL, NULL, 1630 CTLFLAG_PERMANENT, 1631 CTLTYPE_NODE, "inet", 1632 SYSCTL_DESCR("PF_INET related settings"), 1633 NULL, 0, NULL, 0, 1634 CTL_NET, PF_INET, CTL_EOL); 1635 sysctl_createv(clog, 0, NULL, NULL, 1636 CTLFLAG_PERMANENT, 1637 CTLTYPE_NODE, "ip", 1638 SYSCTL_DESCR("IPv4 related settings"), 1639 NULL, 0, NULL, 0, 1640 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1641 1642 sysctl_createv(clog, 0, NULL, NULL, 1643 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1644 CTLTYPE_INT, "forwarding", 1645 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1646 NULL, 0, &ipforwarding, 0, 1647 CTL_NET, PF_INET, IPPROTO_IP, 1648 IPCTL_FORWARDING, CTL_EOL); 1649 sysctl_createv(clog, 0, NULL, NULL, 1650 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1651 CTLTYPE_INT, "redirect", 1652 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1653 NULL, 0, &ipsendredirects, 0, 1654 CTL_NET, PF_INET, IPPROTO_IP, 1655 IPCTL_SENDREDIRECTS, CTL_EOL); 1656 sysctl_createv(clog, 0, NULL, NULL, 1657 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1658 CTLTYPE_INT, "ttl", 1659 SYSCTL_DESCR("Default TTL for an INET datagram"), 1660 NULL, 0, &ip_defttl, 0, 1661 CTL_NET, PF_INET, IPPROTO_IP, 1662 IPCTL_DEFTTL, CTL_EOL); 1663 sysctl_createv(clog, 0, NULL, NULL, 1664 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1665 CTLTYPE_INT, "forwsrcrt", 1666 SYSCTL_DESCR("Enable forwarding of source-routed " 1667 "datagrams"), 1668 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1669 CTL_NET, PF_INET, IPPROTO_IP, 1670 IPCTL_FORWSRCRT, CTL_EOL); 1671 sysctl_createv(clog, 0, NULL, NULL, 1672 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1673 CTLTYPE_INT, "directed-broadcast", 1674 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1675 NULL, 0, &ip_directedbcast, 0, 1676 CTL_NET, PF_INET, IPPROTO_IP, 1677 IPCTL_DIRECTEDBCAST, CTL_EOL); 1678 sysctl_createv(clog, 0, NULL, NULL, 1679 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1680 CTLTYPE_INT, "allowsrcrt", 1681 SYSCTL_DESCR("Accept source-routed datagrams"), 1682 NULL, 0, &ip_allowsrcrt, 0, 1683 CTL_NET, PF_INET, IPPROTO_IP, 1684 IPCTL_ALLOWSRCRT, CTL_EOL); 1685 1686 sysctl_createv(clog, 0, NULL, NULL, 1687 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1688 CTLTYPE_INT, "mtudisc", 1689 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1690 NULL, 0, &ip_mtudisc, 0, 1691 CTL_NET, PF_INET, IPPROTO_IP, 1692 IPCTL_MTUDISC, CTL_EOL); 1693 sysctl_createv(clog, 0, NULL, NULL, 1694 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1695 CTLTYPE_INT, "anonportmin", 1696 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1697 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1698 CTL_NET, PF_INET, IPPROTO_IP, 1699 IPCTL_ANONPORTMIN, CTL_EOL); 1700 sysctl_createv(clog, 0, NULL, NULL, 1701 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1702 CTLTYPE_INT, "anonportmax", 1703 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1704 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1705 CTL_NET, PF_INET, IPPROTO_IP, 1706 IPCTL_ANONPORTMAX, CTL_EOL); 1707 sysctl_createv(clog, 0, NULL, NULL, 1708 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1709 CTLTYPE_INT, "mtudisctimeout", 1710 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1711 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1712 CTL_NET, PF_INET, IPPROTO_IP, 1713 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1714#ifndef IPNOPRIVPORTS 1715 sysctl_createv(clog, 0, NULL, NULL, 1716 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1717 CTLTYPE_INT, "lowportmin", 1718 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1719 "to assign"), 1720 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1721 CTL_NET, PF_INET, IPPROTO_IP, 1722 IPCTL_LOWPORTMIN, CTL_EOL); 1723 sysctl_createv(clog, 0, NULL, NULL, 1724 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1725 CTLTYPE_INT, "lowportmax", 1726 SYSCTL_DESCR("Highest privileged ephemeral port number " 1727 "to assign"), 1728 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1729 CTL_NET, PF_INET, IPPROTO_IP, 1730 IPCTL_LOWPORTMAX, CTL_EOL); 1731#endif /* IPNOPRIVPORTS */ 1732#if NGRE > 0 1733 sysctl_createv(clog, 0, NULL, NULL, 1734 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1735 CTLTYPE_INT, "grettl", 1736 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1737 NULL, 0, &ip_gre_ttl, 0, 1738 CTL_NET, PF_INET, IPPROTO_IP, 1739 IPCTL_GRE_TTL, CTL_EOL); 1740#endif /* NGRE */ 1741 sysctl_createv(clog, 0, NULL, NULL, 1742 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1743 CTLTYPE_INT, "checkinterface", 1744 SYSCTL_DESCR("Enable receive side of Strong ES model " 1745 "from RFC1122"), 1746 NULL, 0, &ip_checkinterface, 0, 1747 CTL_NET, PF_INET, IPPROTO_IP, 1748 IPCTL_CHECKINTERFACE, CTL_EOL); 1749 sysctl_createv(clog, 0, NULL, NULL, 1750 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1751 CTLTYPE_INT, "random_id", 1752 SYSCTL_DESCR("Assign random ip_id values"), 1753 NULL, 0, &ip_do_randomid, 0, 1754 CTL_NET, PF_INET, IPPROTO_IP, 1755 IPCTL_RANDOMID, CTL_EOL); 1756 sysctl_createv(clog, 0, NULL, NULL, 1757 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1758 CTLTYPE_INT, "do_loopback_cksum", 1759 SYSCTL_DESCR("Perform IP checksum on loopback"), 1760 NULL, 0, &ip_do_loopback_cksum, 0, 1761 CTL_NET, PF_INET, IPPROTO_IP, 1762 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1763 sysctl_createv(clog, 0, NULL, NULL, 1764 CTLFLAG_PERMANENT, 1765 CTLTYPE_STRUCT, "stats", 1766 SYSCTL_DESCR("IP statistics"), 1767 sysctl_net_inet_ip_stats, 0, NULL, 0, 1768 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1769 CTL_EOL); 1770#if NARP 1771 sysctl_createv(clog, 0, NULL, NULL, 1772 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1773 CTLTYPE_INT, "dad_count", 1774 SYSCTL_DESCR("Number of Duplicate Address Detection " 1775 "probes to send"), 1776 NULL, 0, &ip_dad_count, 0, 1777 CTL_NET, PF_INET, IPPROTO_IP, 1778 IPCTL_DAD_COUNT, CTL_EOL); 1779#endif 1780 1781 /* anonportalgo RFC6056 subtree */ 1782 const struct sysctlnode *portalgo_node; 1783 sysctl_createv(clog, 0, NULL, &portalgo_node, 1784 CTLFLAG_PERMANENT, 1785 CTLTYPE_NODE, "anonportalgo", 1786 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1787 NULL, 0, NULL, 0, 1788 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1789 sysctl_createv(clog, 0, &portalgo_node, NULL, 1790 CTLFLAG_PERMANENT, 1791 CTLTYPE_STRING, "available", 1792 SYSCTL_DESCR("available algorithms"), 1793 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1794 CTL_CREATE, CTL_EOL); 1795 sysctl_createv(clog, 0, &portalgo_node, NULL, 1796 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1797 CTLTYPE_STRING, "selected", 1798 SYSCTL_DESCR("selected algorithm"), 1799 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1800 CTL_CREATE, CTL_EOL); 1801 sysctl_createv(clog, 0, &portalgo_node, NULL, 1802 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1803 CTLTYPE_STRUCT, "reserve", 1804 SYSCTL_DESCR("bitmap of reserved ports"), 1805 sysctl_portalgo_reserve4, 0, NULL, 0, 1806 CTL_CREATE, CTL_EOL); 1807} 1808 1809void 1810ip_statinc(u_int stat) 1811{ 1812 1813 KASSERT(stat < IP_NSTATS); 1814 IP_STATINC(stat); 1815} 1816