1 /* $NetBSD: ip_input.c,v 1.404 2024/07/05 04:31:54 rin 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 1.404 2024/07/05 04:31:54 rin 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 */ 162 163 int ipforwarding = IPFORWARDING; 164 int ipsendredirects = 1; 165 int ip_defttl = IPDEFTTL; 166 int ip_forwsrcrt = 0; 167 int ip_directedbcast = 0; 168 int ip_allowsrcrt = 0; 169 int ip_mtudisc = 1; 170 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 171 int ip_do_randomid = 1; 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 */ 186 static int ip_checkinterface __read_mostly = 0; 187 188 struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 189 190 pktqueue_t * ip_pktq __read_mostly; 191 pfil_head_t * inet_pfil_hook __read_mostly; 192 percpu_t * ipstat_percpu __read_mostly; 193 194 static percpu_t *ipforward_rt_percpu __cacheline_aligned; 195 196 uint16_t ip_id; 197 198 #ifdef INET_CSUM_COUNTERS 199 #include <sys/device.h> 200 201 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 202 NULL, "inet", "hwcsum bad"); 203 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 204 NULL, "inet", "hwcsum ok"); 205 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 206 NULL, "inet", "swcsum"); 207 208 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 209 210 EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 211 EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 212 EVCNT_ATTACH_STATIC(ip_swcsum); 213 214 #else 215 216 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 217 218 #endif /* INET_CSUM_COUNTERS */ 219 220 /* 221 * Used to save the IP options in case a protocol wants to respond 222 * to an incoming packet over the same route if the packet got here 223 * using IP source routing. This allows connection establishment and 224 * maintenance when the remote end is on a network that is not known 225 * to us. 226 */ 227 struct ip_srcrt { 228 int isr_nhops; /* number of hops */ 229 struct in_addr isr_dst; /* final destination */ 230 char isr_nop; /* one NOP to align */ 231 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */ 232 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)]; 233 }; 234 235 static int ip_drainwanted; 236 237 static void save_rte(struct mbuf *, u_char *, struct in_addr); 238 239 #ifdef MBUFTRACE 240 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 241 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 242 #endif 243 244 static void ipintr(void *); 245 static void ip_input(struct mbuf *, struct ifnet *); 246 static void ip_forward(struct mbuf *, int, struct ifnet *); 247 static bool ip_dooptions(struct mbuf *); 248 static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *); 249 static void sysctl_net_inet_ip_setup(struct sysctllog **); 250 251 static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *, 252 int *); 253 static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *, 254 struct ip *); 255 256 #ifdef NET_MPSAFE 257 #define SOFTNET_LOCK() mutex_enter(softnet_lock) 258 #define SOFTNET_UNLOCK() mutex_exit(softnet_lock) 259 #else 260 #define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 261 #define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 262 #endif 263 264 /* 265 * IP initialization: fill in IP protocol switch table. 266 * All protocols not implemented in kernel go to raw IP protocol handler. 267 */ 268 void 269 ip_init(void) 270 { 271 const struct protosw *pr; 272 273 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 274 KASSERT(ip_pktq != NULL); 275 276 in_init(); 277 sysctl_net_inet_ip_setup(NULL); 278 279 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 280 KASSERT(pr != NULL); 281 282 for (u_int i = 0; i < IPPROTO_MAX; i++) { 283 ip_protox[i] = pr - inetsw; 284 } 285 for (pr = inetdomain.dom_protosw; 286 pr < inetdomain.dom_protoswNPROTOSW; pr++) 287 if (pr->pr_domain->dom_family == PF_INET && 288 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 289 ip_protox[pr->pr_protocol] = pr - inetsw; 290 291 ip_reass_init(); 292 293 ip_id = time_uptime & 0xfffff; 294 295 #ifdef GATEWAY 296 ipflow_init(); 297 #endif 298 299 /* Register our Packet Filter hook. */ 300 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 301 KASSERT(inet_pfil_hook != NULL); 302 303 #ifdef MBUFTRACE 304 MOWNER_ATTACH(&ip_tx_mowner); 305 MOWNER_ATTACH(&ip_rx_mowner); 306 #endif 307 308 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 309 ipforward_rt_percpu = rtcache_percpu_alloc(); 310 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 311 } 312 313 static struct in_ifaddr * 314 ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 315 { 316 struct in_ifaddr *ia = NULL; 317 int checkif; 318 319 /* 320 * Enable a consistency check between the destination address 321 * and the arrival interface for a unicast packet (the RFC 1122 322 * strong ES model) if IP forwarding is disabled and the packet 323 * is not locally generated. 324 * 325 * XXX - We need to add a per ifaddr flag for this so that 326 * we get finer grain control. 327 */ 328 checkif = ip_checkinterface && (ipforwarding == 0) && 329 (ifp->if_flags & IFF_LOOPBACK) == 0; 330 331 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 332 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 333 if (ia->ia4_flags & IN_IFF_NOTREADY) 334 continue; 335 if (checkif && ia->ia_ifp != ifp) 336 continue; 337 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) { 338 (*downmatch)++; 339 continue; 340 } 341 if (ia->ia4_flags & IN_IFF_DETACHED && 342 (ifp->if_flags & IFF_LOOPBACK) == 0) 343 continue; 344 break; 345 } 346 } 347 348 return ia; 349 } 350 351 static struct in_ifaddr * 352 ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 353 { 354 struct in_ifaddr *ia = NULL; 355 struct ifaddr *ifa; 356 357 IFADDR_READER_FOREACH(ifa, ifp) { 358 if (ifa->ifa_addr->sa_family != AF_INET) 359 continue; 360 ia = ifatoia(ifa); 361 if (ia->ia4_flags & IN_IFF_NOTREADY) 362 continue; 363 if (ia->ia4_flags & IN_IFF_DETACHED && 364 (ifp->if_flags & IFF_LOOPBACK) == 0) 365 continue; 366 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 367 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 368 /* 369 * Look for all-0's host part (old broadcast addr), 370 * either for subnet or net. 371 */ 372 ip->ip_dst.s_addr == ia->ia_subnet || 373 ip->ip_dst.s_addr == ia->ia_net) 374 goto matched; 375 /* 376 * An interface with IP address zero accepts 377 * all packets that arrive on that interface. 378 */ 379 if (in_nullhost(ia->ia_addr.sin_addr)) 380 goto matched; 381 } 382 ia = NULL; 383 384 matched: 385 return ia; 386 } 387 388 /* 389 * IP software interrupt routine. 390 */ 391 static void 392 ipintr(void *arg __unused) 393 { 394 struct mbuf *m; 395 396 KASSERT(cpu_softintr_p()); 397 398 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 399 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 400 struct ifnet *ifp; 401 struct psref psref; 402 403 ifp = m_get_rcvif_psref(m, &psref); 404 if (__predict_false(ifp == NULL)) { 405 IP_STATINC(IP_STAT_IFDROP); 406 m_freem(m); 407 continue; 408 } 409 410 ip_input(m, ifp); 411 412 m_put_rcvif_psref(ifp, &psref); 413 } 414 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 415 } 416 417 /* 418 * IP input routine. Checksum and byte swap header. If fragmented 419 * try to reassemble. Process options. Pass to next level. 420 */ 421 static void 422 ip_input(struct mbuf *m, struct ifnet *ifp) 423 { 424 struct ip *ip = NULL; 425 struct in_ifaddr *ia = NULL; 426 int hlen = 0, len; 427 int downmatch; 428 int srcrt = 0; 429 int s; 430 431 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 432 "interrupt handler; synchronization assumptions violated"); 433 434 MCLAIM(m, &ip_rx_mowner); 435 KASSERT((m->m_flags & M_PKTHDR) != 0); 436 437 /* 438 * If no IP addresses have been set yet but the interfaces 439 * are receiving, can't do anything with incoming packets yet. 440 * Note: we pre-check without locks held. 441 */ 442 if (IN_ADDRLIST_READER_EMPTY()) { 443 IP_STATINC(IP_STAT_IFDROP); 444 goto out; 445 } 446 447 IP_STATINC(IP_STAT_TOTAL); 448 449 /* 450 * If the IP header is not aligned, slurp it up into a new 451 * mbuf with space for link headers, in the event we forward 452 * it. Otherwise, if it is aligned, make sure the entire 453 * base IP header is in the first mbuf of the chain. 454 */ 455 if (M_GET_ALIGNED_HDR(&m, struct ip, true) != 0) { 456 /* XXXJRT new stat, please */ 457 IP_STATINC(IP_STAT_TOOSMALL); 458 goto out; 459 } 460 ip = mtod(m, struct ip *); 461 if (ip->ip_v != IPVERSION) { 462 IP_STATINC(IP_STAT_BADVERS); 463 goto out; 464 } 465 hlen = ip->ip_hl << 2; 466 if (hlen < sizeof(struct ip)) { /* minimum header length */ 467 IP_STATINC(IP_STAT_BADHLEN); 468 goto out; 469 } 470 if (hlen > m->m_len) { 471 if ((m = m_pullup(m, hlen)) == NULL) { 472 IP_STATINC(IP_STAT_BADHLEN); 473 goto out; 474 } 475 ip = mtod(m, struct ip *); 476 } 477 478 /* 479 * RFC1122: packets with a multicast source address are 480 * not allowed. 481 */ 482 if (IN_MULTICAST(ip->ip_src.s_addr)) { 483 IP_STATINC(IP_STAT_BADADDR); 484 goto out; 485 } 486 487 /* 127/8 must not appear on wire - RFC1122 */ 488 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 489 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 490 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 491 IP_STATINC(IP_STAT_BADADDR); 492 goto out; 493 } 494 } 495 496 switch (m->m_pkthdr.csum_flags & 497 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) { 498 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 499 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 500 IP_STATINC(IP_STAT_BADSUM); 501 goto out; 502 503 case M_CSUM_IPv4: 504 /* Checksum was okay. */ 505 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 506 break; 507 508 default: 509 /* 510 * Must compute it ourselves. Maybe skip checksum on 511 * loopback interfaces. 512 */ 513 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 514 ip_do_loopback_cksum)) { 515 INET_CSUM_COUNTER_INCR(&ip_swcsum); 516 if (in_cksum(m, hlen) != 0) { 517 IP_STATINC(IP_STAT_BADSUM); 518 goto out; 519 } 520 } 521 break; 522 } 523 524 /* Retrieve the packet length. */ 525 len = ntohs(ip->ip_len); 526 527 /* 528 * Check for additional length bogosity 529 */ 530 if (len < hlen) { 531 IP_STATINC(IP_STAT_BADLEN); 532 goto out; 533 } 534 535 /* 536 * Check that the amount of data in the buffers is at least as much 537 * as the IP header would have us expect. Trim mbufs if longer than 538 * we expect. Drop packet if shorter than we expect. 539 */ 540 if (m->m_pkthdr.len < len) { 541 IP_STATINC(IP_STAT_TOOSHORT); 542 goto out; 543 } 544 if (m->m_pkthdr.len > len) { 545 if (m->m_len == m->m_pkthdr.len) { 546 m->m_len = len; 547 m->m_pkthdr.len = len; 548 } else 549 m_adj(m, len - m->m_pkthdr.len); 550 } 551 552 /* 553 * Assume that we can create a fast-forward IP flow entry 554 * based on this packet. 555 */ 556 m->m_flags |= M_CANFASTFWD; 557 558 /* 559 * Run through list of hooks for input packets. If there are any 560 * filters which require that additional packets in the flow are 561 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 562 * Note that filters must _never_ set this flag, as another filter 563 * in the list may have previously cleared it. 564 * 565 * Don't call hooks if the packet has already been processed by 566 * IPsec (encapsulated, tunnel mode). 567 */ 568 #if defined(IPSEC) 569 if (!ipsec_used || !ipsec_skip_pfil(m)) 570 #else 571 if (1) 572 #endif 573 { 574 struct in_addr odst = ip->ip_dst; 575 bool freed; 576 577 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 578 if (freed || m == NULL) { 579 m = NULL; 580 IP_STATINC(IP_STAT_PFILDROP_IN); 581 goto out; 582 } 583 if (__predict_false(m->m_len < sizeof(struct ip))) { 584 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) { 585 IP_STATINC(IP_STAT_TOOSMALL); 586 goto out; 587 } 588 } 589 ip = mtod(m, struct ip *); 590 hlen = ip->ip_hl << 2; 591 if (hlen < sizeof(struct ip)) { /* minimum header length */ 592 IP_STATINC(IP_STAT_BADHLEN); 593 goto out; 594 } 595 if (hlen > m->m_len) { 596 if ((m = m_pullup(m, hlen)) == NULL) { 597 IP_STATINC(IP_STAT_BADHLEN); 598 goto out; 599 } 600 ip = mtod(m, struct ip *); 601 } 602 603 /* 604 * XXX The setting of "srcrt" here is to prevent ip_forward() 605 * from generating ICMP redirects for packets that have 606 * been redirected by a hook back out on to the same LAN that 607 * they came from and is not an indication that the packet 608 * is being influenced by source routing options. This 609 * allows things like 610 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 611 * where tlp0 is both on the 1.1.1.0/24 network and is the 612 * default route for hosts on 1.1.1.0/24. Of course this 613 * also requires a "map tlp0 ..." to complete the story. 614 * One might argue whether or not this kind of network config. 615 * should be supported in this manner... 616 */ 617 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 618 } 619 620 #ifdef ALTQ 621 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 622 if (altq_input) { 623 SOFTNET_LOCK(); 624 if ((*altq_input)(m, AF_INET) == 0) { 625 /* Packet dropped by traffic conditioner. */ 626 SOFTNET_UNLOCK(); 627 m = NULL; 628 goto out; 629 } 630 SOFTNET_UNLOCK(); 631 } 632 #endif 633 634 /* 635 * Process options and, if not destined for us, 636 * ship it on. ip_dooptions returns 1 when an 637 * error was detected (causing an icmp message 638 * to be sent and the original packet to be freed). 639 */ 640 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 641 m = NULL; 642 goto out; 643 } 644 645 /* 646 * Check our list of addresses, to see if the packet is for us. 647 * 648 * Traditional 4.4BSD did not consult IFF_UP at all. 649 * The behavior here is to treat addresses on !IFF_UP interface 650 * or IN_IFF_NOTREADY addresses as not mine. 651 */ 652 downmatch = 0; 653 s = pserialize_read_enter(); 654 ia = ip_match_our_address(ifp, ip, &downmatch); 655 if (ia != NULL) { 656 pserialize_read_exit(s); 657 goto ours; 658 } 659 660 if (ifp->if_flags & IFF_BROADCAST) { 661 ia = ip_match_our_address_broadcast(ifp, ip); 662 if (ia != NULL) { 663 pserialize_read_exit(s); 664 goto ours; 665 } 666 } 667 pserialize_read_exit(s); 668 669 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 670 #ifdef MROUTING 671 extern struct socket *ip_mrouter; 672 673 if (ip_mrouter) { 674 /* 675 * If we are acting as a multicast router, all 676 * incoming multicast packets are passed to the 677 * kernel-level multicast forwarding function. 678 * The packet is returned (relatively) intact; if 679 * ip_mforward() returns a non-zero value, the packet 680 * must be discarded, else it may be accepted below. 681 * 682 * (The IP ident field is put in the same byte order 683 * as expected when ip_mforward() is called from 684 * ip_output().) 685 */ 686 SOFTNET_LOCK(); 687 if (ip_mforward(m, ifp) != 0) { 688 SOFTNET_UNLOCK(); 689 IP_STATINC(IP_STAT_CANTFORWARD); 690 goto out; 691 } 692 SOFTNET_UNLOCK(); 693 694 /* 695 * The process-level routing demon needs to receive 696 * all multicast IGMP packets, whether or not this 697 * host belongs to their destination groups. 698 */ 699 if (ip->ip_p == IPPROTO_IGMP) { 700 goto ours; 701 } 702 IP_STATINC(IP_STAT_CANTFORWARD); 703 } 704 #endif 705 /* 706 * See if we belong to the destination multicast group on the 707 * arrival interface. 708 */ 709 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 710 IP_STATINC(IP_STAT_CANTFORWARD); 711 goto out; 712 } 713 goto ours; 714 } 715 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 716 in_nullhost(ip->ip_dst)) 717 goto ours; 718 719 /* 720 * Not for us; forward if possible and desirable. 721 */ 722 if (ipforwarding == 0) { 723 IP_STATINC(IP_STAT_CANTFORWARD); 724 m_freem(m); 725 } else { 726 /* 727 * If ip_dst matched any of my address on !IFF_UP interface, 728 * and there's no IFF_UP interface that matches ip_dst, 729 * send icmp unreach. Forwarding it will result in in-kernel 730 * forwarding loop till TTL goes to 0. 731 */ 732 if (downmatch) { 733 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 734 IP_STATINC(IP_STAT_CANTFORWARD); 735 return; 736 } 737 #ifdef IPSEC 738 /* Check the security policy (SP) for the packet */ 739 if (ipsec_used) { 740 if (ipsec_ip_input_checkpolicy(m, true) != 0) { 741 IP_STATINC(IP_STAT_IPSECDROP_IN); 742 goto out; 743 } 744 } 745 #endif 746 ip_forward(m, srcrt, ifp); 747 } 748 return; 749 750 ours: 751 /* 752 * If offset or IP_MF are set, must reassemble. 753 */ 754 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 755 /* 756 * Pass to IP reassembly mechanism. 757 */ 758 if (ip_reass_packet(&m) != 0) { 759 /* Failed; invalid fragment(s) or packet. */ 760 goto out; 761 } 762 if (m == NULL) { 763 /* More fragments should come; silently return. */ 764 goto out; 765 } 766 /* 767 * Reassembly is done, we have the final packet. 768 * Update cached data in local variable(s). 769 */ 770 ip = mtod(m, struct ip *); 771 hlen = ip->ip_hl << 2; 772 } 773 774 M_VERIFY_PACKET(m); 775 776 #ifdef IPSEC 777 /* 778 * Enforce IPsec policy checking if we are seeing last header. 779 * Note that we do not visit this with protocols with PCB layer 780 * code - like UDP/TCP/raw IP. 781 */ 782 if (ipsec_used && 783 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 784 if (ipsec_ip_input_checkpolicy(m, false) != 0) { 785 IP_STATINC(IP_STAT_IPSECDROP_IN); 786 goto out; 787 } 788 } 789 #endif 790 791 /* 792 * Switch out to protocol's input routine. 793 */ 794 #if IFA_STATS 795 if (ia) { 796 struct in_ifaddr *_ia; 797 /* 798 * Keep a reference from ip_match_our_address with psref 799 * is expensive, so explore ia here again. 800 */ 801 s = pserialize_read_enter(); 802 _ia = in_get_ia(ip->ip_dst); 803 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 804 pserialize_read_exit(s); 805 } 806 #endif 807 IP_STATINC(IP_STAT_DELIVERED); 808 809 const int off = hlen, nh = ip->ip_p; 810 811 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 812 return; 813 814 out: 815 m_freem(m); 816 } 817 818 /* 819 * IP timer processing. 820 */ 821 void 822 ip_slowtimo(void) 823 { 824 825 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 826 827 ip_reass_slowtimo(); 828 829 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 830 } 831 832 /* 833 * IP drain processing. 834 */ 835 void 836 ip_drain(void) 837 { 838 839 KERNEL_LOCK(1, NULL); 840 ip_reass_drain(); 841 KERNEL_UNLOCK_ONE(NULL); 842 } 843 844 /* 845 * ip_dooptions: perform option processing on a datagram, possibly discarding 846 * it if bad options are encountered, or forwarding it if source-routed. 847 * 848 * => Returns true if packet has been forwarded/freed. 849 * => Returns false if the packet should be processed further. 850 */ 851 static bool 852 ip_dooptions(struct mbuf *m) 853 { 854 struct ip *ip = mtod(m, struct ip *); 855 u_char *cp, *cp0; 856 struct ip_timestamp *ipt; 857 struct in_ifaddr *ia; 858 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 859 int srr_present, rr_present, ts_present; 860 struct in_addr dst; 861 n_time ntime; 862 struct ifaddr *ifa = NULL; 863 int s; 864 865 srr_present = 0; 866 rr_present = 0; 867 ts_present = 0; 868 869 dst = ip->ip_dst; 870 cp = (u_char *)(ip + 1); 871 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 872 for (; cnt > 0; cnt -= optlen, cp += optlen) { 873 opt = cp[IPOPT_OPTVAL]; 874 if (opt == IPOPT_EOL) 875 break; 876 if (opt == IPOPT_NOP) 877 optlen = 1; 878 else { 879 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 880 code = &cp[IPOPT_OLEN] - (u_char *)ip; 881 goto bad; 882 } 883 optlen = cp[IPOPT_OLEN]; 884 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 885 code = &cp[IPOPT_OLEN] - (u_char *)ip; 886 goto bad; 887 } 888 } 889 switch (opt) { 890 891 default: 892 break; 893 894 /* 895 * Source routing with record. 896 * Find interface with current destination address. 897 * If none on this machine then drop if strictly routed, 898 * or do nothing if loosely routed. 899 * Record interface address and bring up next address 900 * component. If strictly routed make sure next 901 * address is on directly accessible net. 902 */ 903 case IPOPT_LSRR: 904 case IPOPT_SSRR: { 905 struct psref psref; 906 struct sockaddr_in ipaddr = { 907 .sin_len = sizeof(ipaddr), 908 .sin_family = AF_INET, 909 }; 910 911 if (ip_allowsrcrt == 0) { 912 type = ICMP_UNREACH; 913 code = ICMP_UNREACH_NET_PROHIB; 914 goto bad; 915 } 916 if (srr_present++) { 917 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 918 goto bad; 919 } 920 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 921 code = &cp[IPOPT_OLEN] - (u_char *)ip; 922 goto bad; 923 } 924 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 925 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 926 goto bad; 927 } 928 ipaddr.sin_addr = ip->ip_dst; 929 930 s = pserialize_read_enter(); 931 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 932 if (ifa == NULL) { 933 pserialize_read_exit(s); 934 if (opt == IPOPT_SSRR) { 935 type = ICMP_UNREACH; 936 code = ICMP_UNREACH_SRCFAIL; 937 goto bad; 938 } 939 /* 940 * Loose routing, and not at next destination 941 * yet; nothing to do except forward. 942 */ 943 break; 944 } 945 pserialize_read_exit(s); 946 947 off--; /* 0 origin */ 948 if ((off + sizeof(struct in_addr)) > optlen) { 949 /* 950 * End of source route. Should be for us. 951 */ 952 save_rte(m, cp, ip->ip_src); 953 break; 954 } 955 /* 956 * locate outgoing interface 957 */ 958 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 959 sizeof(ipaddr.sin_addr)); 960 if (opt == IPOPT_SSRR) { 961 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 962 &psref); 963 if (ifa != NULL) 964 ia = ifatoia(ifa); 965 else 966 ia = NULL; 967 } else { 968 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 969 } 970 if (ia == NULL) { 971 type = ICMP_UNREACH; 972 code = ICMP_UNREACH_SRCFAIL; 973 goto bad; 974 } 975 ip->ip_dst = ipaddr.sin_addr; 976 memcpy(cp + off, &ia->ia_addr.sin_addr, 977 sizeof(struct in_addr)); 978 ia4_release(ia, &psref); 979 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 980 /* 981 * Let ip_intr's mcast routing check handle mcast pkts 982 */ 983 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 984 break; 985 } 986 987 case IPOPT_RR: { 988 struct psref psref; 989 struct sockaddr_in ipaddr = { 990 .sin_len = sizeof(ipaddr), 991 .sin_family = AF_INET, 992 }; 993 994 if (rr_present++) { 995 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 996 goto bad; 997 } 998 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 999 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1000 goto bad; 1001 } 1002 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1003 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1004 goto bad; 1005 } 1006 /* 1007 * If no space remains, ignore. 1008 */ 1009 off--; /* 0 origin */ 1010 if ((off + sizeof(struct in_addr)) > optlen) 1011 break; 1012 memcpy((void *)&ipaddr.sin_addr, (void *)&ip->ip_dst, 1013 sizeof(ipaddr.sin_addr)); 1014 /* 1015 * locate outgoing interface; if we're the destination, 1016 * use the incoming interface (should be same). 1017 */ 1018 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1019 if (ifa == NULL) { 1020 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1021 if (ia == NULL) { 1022 type = ICMP_UNREACH; 1023 code = ICMP_UNREACH_HOST; 1024 goto bad; 1025 } 1026 } else { 1027 ia = ifatoia(ifa); 1028 } 1029 memcpy(cp + off, &ia->ia_addr.sin_addr, 1030 sizeof(struct in_addr)); 1031 ia4_release(ia, &psref); 1032 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1033 break; 1034 } 1035 1036 case IPOPT_TS: 1037 code = cp - (u_char *)ip; 1038 ipt = (struct ip_timestamp *)cp; 1039 if (ts_present++) { 1040 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1041 goto bad; 1042 } 1043 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1044 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1045 goto bad; 1046 } 1047 if (ipt->ipt_ptr < 5) { 1048 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1049 goto bad; 1050 } 1051 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1052 if (++ipt->ipt_oflw == 0) { 1053 code = (u_char *)&ipt->ipt_ptr - 1054 (u_char *)ip; 1055 goto bad; 1056 } 1057 break; 1058 } 1059 cp0 = (cp + ipt->ipt_ptr - 1); 1060 switch (ipt->ipt_flg) { 1061 1062 case IPOPT_TS_TSONLY: 1063 break; 1064 1065 case IPOPT_TS_TSANDADDR: { 1066 struct ifnet *rcvif; 1067 int _s, _ss; 1068 struct sockaddr_in ipaddr = { 1069 .sin_len = sizeof(ipaddr), 1070 .sin_family = AF_INET, 1071 }; 1072 1073 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1074 sizeof(struct in_addr) > ipt->ipt_len) { 1075 code = (u_char *)&ipt->ipt_ptr - 1076 (u_char *)ip; 1077 goto bad; 1078 } 1079 ipaddr.sin_addr = dst; 1080 _ss = pserialize_read_enter(); 1081 rcvif = m_get_rcvif(m, &_s); 1082 if (__predict_true(rcvif != NULL)) { 1083 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1084 rcvif); 1085 } 1086 m_put_rcvif(rcvif, &_s); 1087 if (ifa == NULL) { 1088 pserialize_read_exit(_ss); 1089 break; 1090 } 1091 ia = ifatoia(ifa); 1092 memcpy(cp0, &ia->ia_addr.sin_addr, 1093 sizeof(struct in_addr)); 1094 pserialize_read_exit(_ss); 1095 ipt->ipt_ptr += sizeof(struct in_addr); 1096 break; 1097 } 1098 1099 case IPOPT_TS_PRESPEC: { 1100 struct sockaddr_in ipaddr = { 1101 .sin_len = sizeof(ipaddr), 1102 .sin_family = AF_INET, 1103 }; 1104 1105 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1106 sizeof(struct in_addr) > ipt->ipt_len) { 1107 code = (u_char *)&ipt->ipt_ptr - 1108 (u_char *)ip; 1109 goto bad; 1110 } 1111 memcpy(&ipaddr.sin_addr, cp0, 1112 sizeof(struct in_addr)); 1113 s = pserialize_read_enter(); 1114 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1115 if (ifa == NULL) { 1116 pserialize_read_exit(s); 1117 continue; 1118 } 1119 pserialize_read_exit(s); 1120 ipt->ipt_ptr += sizeof(struct in_addr); 1121 break; 1122 } 1123 1124 default: 1125 /* XXX can't take &ipt->ipt_flg */ 1126 code = (u_char *)&ipt->ipt_ptr - 1127 (u_char *)ip + 1; 1128 goto bad; 1129 } 1130 ntime = iptime(); 1131 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1132 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1133 sizeof(n_time)); 1134 ipt->ipt_ptr += sizeof(n_time); 1135 } 1136 } 1137 if (forward) { 1138 struct ifnet *rcvif; 1139 struct psref _psref; 1140 1141 if (ip_forwsrcrt == 0) { 1142 type = ICMP_UNREACH; 1143 code = ICMP_UNREACH_SRCFAIL; 1144 goto bad; 1145 } 1146 1147 rcvif = m_get_rcvif_psref(m, &_psref); 1148 if (__predict_false(rcvif == NULL)) { 1149 type = ICMP_UNREACH; 1150 code = ICMP_UNREACH_HOST; 1151 goto bad; 1152 } 1153 ip_forward(m, 1, rcvif); 1154 m_put_rcvif_psref(rcvif, &_psref); 1155 return true; 1156 } 1157 return false; 1158 bad: 1159 icmp_error(m, type, code, 0, 0); 1160 IP_STATINC(IP_STAT_BADOPTIONS); 1161 return true; 1162 } 1163 1164 /* 1165 * ip_rtaddr: given address of next destination (final or next hop), 1166 * return internet address info of interface to be used to get there. 1167 */ 1168 static struct in_ifaddr * 1169 ip_rtaddr(struct in_addr dst, struct psref *psref) 1170 { 1171 struct rtentry *rt; 1172 union { 1173 struct sockaddr dst; 1174 struct sockaddr_in dst4; 1175 } u; 1176 struct route *ro; 1177 1178 sockaddr_in_init(&u.dst4, &dst, 0); 1179 1180 ro = rtcache_percpu_getref(ipforward_rt_percpu); 1181 rt = rtcache_lookup(ro, &u.dst); 1182 if (rt == NULL) { 1183 rtcache_percpu_putref(ipforward_rt_percpu); 1184 return NULL; 1185 } 1186 1187 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1188 rtcache_unref(rt, ro); 1189 rtcache_percpu_putref(ipforward_rt_percpu); 1190 1191 return ifatoia(rt->rt_ifa); 1192 } 1193 1194 /* 1195 * save_rte: save incoming source route for use in replies, to be picked 1196 * up later by ip_srcroute if the receiver is interested. 1197 */ 1198 static void 1199 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1200 { 1201 struct ip_srcrt *isr; 1202 struct m_tag *mtag; 1203 unsigned olen; 1204 1205 olen = option[IPOPT_OLEN]; 1206 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1207 return; 1208 1209 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1210 if (mtag == NULL) 1211 return; 1212 isr = (struct ip_srcrt *)(mtag + 1); 1213 1214 memcpy(isr->isr_hdr, option, olen); 1215 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1216 isr->isr_dst = dst; 1217 m_tag_prepend(m, mtag); 1218 } 1219 1220 /* 1221 * Retrieve incoming source route for use in replies, 1222 * in the same form used by setsockopt. 1223 * The first hop is placed before the options, will be removed later. 1224 */ 1225 struct mbuf * 1226 ip_srcroute(struct mbuf *m0) 1227 { 1228 struct in_addr *p, *q; 1229 struct mbuf *m; 1230 struct ip_srcrt *isr; 1231 struct m_tag *mtag; 1232 1233 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE); 1234 if (mtag == NULL) 1235 return NULL; 1236 isr = (struct ip_srcrt *)(mtag + 1); 1237 1238 if (isr->isr_nhops == 0) 1239 return NULL; 1240 1241 m = m_get(M_DONTWAIT, MT_SOOPTS); 1242 if (m == NULL) 1243 return NULL; 1244 1245 MCLAIM(m, &inetdomain.dom_mowner); 1246 #define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1247 1248 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1249 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1250 1251 /* 1252 * First save first hop for return route 1253 */ 1254 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1255 *(mtod(m, struct in_addr *)) = *p--; 1256 1257 /* 1258 * Copy option fields and padding (nop) to mbuf. 1259 */ 1260 isr->isr_nop = IPOPT_NOP; 1261 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1262 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1263 OPTSIZ); 1264 q = (struct in_addr *)(mtod(m, char *) + 1265 sizeof(struct in_addr) + OPTSIZ); 1266 #undef OPTSIZ 1267 /* 1268 * Record return path as an IP source route, 1269 * reversing the path (pointers are now aligned). 1270 */ 1271 while (p >= isr->isr_routes) { 1272 *q++ = *p--; 1273 } 1274 /* 1275 * Last hop goes to final destination. 1276 */ 1277 *q = isr->isr_dst; 1278 m_tag_delete(m0, mtag); 1279 return m; 1280 } 1281 1282 const int inetctlerrmap[PRC_NCMDS] = { 1283 [PRC_MSGSIZE] = EMSGSIZE, 1284 [PRC_HOSTDEAD] = EHOSTDOWN, 1285 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1286 [PRC_UNREACH_NET] = EHOSTUNREACH, 1287 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1288 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1289 [PRC_UNREACH_PORT] = ECONNREFUSED, 1290 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1291 [PRC_PARAMPROB] = ENOPROTOOPT, 1292 }; 1293 1294 void 1295 ip_fasttimo(void) 1296 { 1297 if (ip_drainwanted) { 1298 ip_drain(); 1299 ip_drainwanted = 0; 1300 } 1301 } 1302 1303 void 1304 ip_drainstub(void) 1305 { 1306 ip_drainwanted = 1; 1307 } 1308 1309 /* 1310 * Forward a packet. If some error occurs return the sender 1311 * an icmp packet. Note we can't always generate a meaningful 1312 * icmp message because icmp doesn't have a large enough repertoire 1313 * of codes and types. 1314 * 1315 * If not forwarding, just drop the packet. This could be confusing 1316 * if ipforwarding was zero but some routing protocol was advancing 1317 * us as a gateway to somewhere. However, we must let the routing 1318 * protocol deal with that. 1319 * 1320 * The srcrt parameter indicates whether the packet is being forwarded 1321 * via a source route. 1322 */ 1323 static void 1324 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1325 { 1326 struct ip *ip = mtod(m, struct ip *); 1327 struct rtentry *rt; 1328 int error, type = 0, code = 0, destmtu = 0; 1329 struct mbuf *mcopy; 1330 n_long dest; 1331 union { 1332 struct sockaddr dst; 1333 struct sockaddr_in dst4; 1334 } u; 1335 net_stat_ref_t ips; 1336 struct route *ro; 1337 1338 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1339 "interrupt handler; synchronization assumptions violated"); 1340 1341 /* 1342 * We are now in the output path. 1343 */ 1344 MCLAIM(m, &ip_tx_mowner); 1345 1346 /* 1347 * Clear any in-bound checksum flags for this packet. 1348 */ 1349 m->m_pkthdr.csum_flags = 0; 1350 1351 dest = 0; 1352 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1353 IP_STATINC(IP_STAT_CANTFORWARD); 1354 m_freem(m); 1355 return; 1356 } 1357 1358 if (ip->ip_ttl <= IPTTLDEC) { 1359 IP_STATINC(IP_STAT_TIMXCEED); 1360 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1361 return; 1362 } 1363 1364 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1365 1366 ro = rtcache_percpu_getref(ipforward_rt_percpu); 1367 rt = rtcache_lookup(ro, &u.dst); 1368 if (rt == NULL) { 1369 rtcache_percpu_putref(ipforward_rt_percpu); 1370 IP_STATINC(IP_STAT_NOROUTE); 1371 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1372 return; 1373 } 1374 1375 /* 1376 * Save at most 68 bytes of the packet in case 1377 * we need to generate an ICMP message to the src. 1378 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1379 */ 1380 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1381 if (mcopy) 1382 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1383 1384 ip->ip_ttl -= IPTTLDEC; 1385 1386 /* 1387 * If forwarding packet using same interface that it came in on, 1388 * perhaps should send a redirect to sender to shortcut a hop. 1389 * Only send redirect if source is sending directly to us, 1390 * and if packet was not source routed (or has any options). 1391 * Also, don't send redirect if forwarding using a default route 1392 * or a route modified by a redirect. 1393 */ 1394 if (rt->rt_ifp == rcvif && 1395 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1396 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1397 ipsendredirects && !srcrt) { 1398 if ((ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1399 ifatoia(rt->rt_ifa)->ia_subnet) { 1400 if (rt->rt_flags & RTF_GATEWAY) 1401 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1402 else 1403 dest = ip->ip_dst.s_addr; 1404 /* 1405 * Router requirements says to only send host 1406 * redirects. 1407 */ 1408 type = ICMP_REDIRECT; 1409 code = ICMP_REDIRECT_HOST; 1410 } 1411 } 1412 rtcache_unref(rt, ro); 1413 1414 error = ip_output(m, NULL, ro, 1415 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1416 NULL, NULL); 1417 1418 if (error) { 1419 IP_STATINC(IP_STAT_CANTFORWARD); 1420 goto error; 1421 } 1422 1423 ips = IP_STAT_GETREF(); 1424 _NET_STATINC_REF(ips, IP_STAT_FORWARD); 1425 1426 if (type) { 1427 _NET_STATINC_REF(ips, IP_STAT_REDIRECTSENT); 1428 IP_STAT_PUTREF(); 1429 goto redirect; 1430 } 1431 1432 IP_STAT_PUTREF(); 1433 if (mcopy) { 1434 #ifdef GATEWAY 1435 if (mcopy->m_flags & M_CANFASTFWD) 1436 ipflow_create(ro, mcopy); 1437 #endif 1438 m_freem(mcopy); 1439 } 1440 1441 rtcache_percpu_putref(ipforward_rt_percpu); 1442 return; 1443 1444 redirect: 1445 error: 1446 if (mcopy == NULL) { 1447 rtcache_percpu_putref(ipforward_rt_percpu); 1448 return; 1449 } 1450 1451 switch (error) { 1452 1453 case 0: /* forwarded, but need redirect */ 1454 /* type, code set above */ 1455 break; 1456 1457 case ENETUNREACH: /* shouldn't happen, checked above */ 1458 case EHOSTUNREACH: 1459 case ENETDOWN: 1460 case EHOSTDOWN: 1461 default: 1462 type = ICMP_UNREACH; 1463 code = ICMP_UNREACH_HOST; 1464 break; 1465 1466 case EMSGSIZE: 1467 type = ICMP_UNREACH; 1468 code = ICMP_UNREACH_NEEDFRAG; 1469 1470 if ((rt = rtcache_validate(ro)) != NULL) { 1471 destmtu = rt->rt_ifp->if_mtu; 1472 rtcache_unref(rt, ro); 1473 } 1474 #ifdef IPSEC 1475 if (ipsec_used) 1476 ipsec_mtu(mcopy, &destmtu); 1477 #endif 1478 IP_STATINC(IP_STAT_CANTFRAG); 1479 break; 1480 1481 case ENOBUFS: 1482 /* 1483 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1484 * Requirements for IP Version 4 Routers. Source quench can 1485 * be a big problem under DoS attacks or if the underlying 1486 * interface is rate-limited. 1487 */ 1488 m_freem(mcopy); 1489 rtcache_percpu_putref(ipforward_rt_percpu); 1490 return; 1491 } 1492 icmp_error(mcopy, type, code, dest, destmtu); 1493 rtcache_percpu_putref(ipforward_rt_percpu); 1494 } 1495 1496 void 1497 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1498 struct mbuf *m) 1499 { 1500 struct socket *so = inp->inp_socket; 1501 int inpflags = inp->inp_flags; 1502 1503 if (SOOPT_TIMESTAMP(so->so_options)) 1504 mp = sbsavetimestamp(so->so_options, mp); 1505 1506 if (inpflags & INP_RECVDSTADDR) { 1507 *mp = sbcreatecontrol(&ip->ip_dst, 1508 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1509 if (*mp) 1510 mp = &(*mp)->m_next; 1511 } 1512 1513 if (inpflags & INP_RECVTTL) { 1514 *mp = sbcreatecontrol(&ip->ip_ttl, 1515 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1516 if (*mp) 1517 mp = &(*mp)->m_next; 1518 } 1519 1520 struct psref psref; 1521 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1522 if (__predict_false(ifp == NULL)) { 1523 #ifdef DIAGNOSTIC 1524 printf("%s: missing receive interface\n", __func__); 1525 #endif 1526 return; /* XXX should report error? */ 1527 } 1528 1529 if (inpflags & INP_RECVPKTINFO) { 1530 struct in_pktinfo ipi; 1531 ipi.ipi_addr = ip->ip_dst; 1532 ipi.ipi_ifindex = ifp->if_index; 1533 *mp = sbcreatecontrol(&ipi, 1534 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1535 if (*mp) 1536 mp = &(*mp)->m_next; 1537 } 1538 if (inpflags & INP_RECVIF) { 1539 struct sockaddr_dl sdl; 1540 1541 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1542 NULL, 0); 1543 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1544 if (*mp) 1545 mp = &(*mp)->m_next; 1546 } 1547 m_put_rcvif_psref(ifp, &psref); 1548 } 1549 1550 /* 1551 * sysctl helper routine for net.inet.ip.forwsrcrt. 1552 */ 1553 static int 1554 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1555 { 1556 int error, tmp; 1557 struct sysctlnode node; 1558 1559 node = *rnode; 1560 tmp = ip_forwsrcrt; 1561 node.sysctl_data = &tmp; 1562 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1563 if (error || newp == NULL) 1564 return (error); 1565 1566 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1567 0, NULL, NULL, NULL); 1568 if (error) 1569 return (error); 1570 1571 ip_forwsrcrt = tmp; 1572 1573 return (0); 1574 } 1575 1576 /* 1577 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1578 * range of the new value and tweaks timers if it changes. 1579 */ 1580 static int 1581 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1582 { 1583 int error, tmp; 1584 struct sysctlnode node; 1585 1586 icmp_mtudisc_lock(); 1587 1588 node = *rnode; 1589 tmp = ip_mtudisc_timeout; 1590 node.sysctl_data = &tmp; 1591 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1592 if (error || newp == NULL) 1593 goto out; 1594 if (tmp < 0) { 1595 error = EINVAL; 1596 goto out; 1597 } 1598 1599 ip_mtudisc_timeout = tmp; 1600 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1601 error = 0; 1602 out: 1603 icmp_mtudisc_unlock(); 1604 return error; 1605 } 1606 1607 static int 1608 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1609 { 1610 1611 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1612 } 1613 1614 static void 1615 sysctl_net_inet_ip_setup(struct sysctllog **clog) 1616 { 1617 const struct sysctlnode *ip_node; 1618 1619 sysctl_createv(clog, 0, NULL, NULL, 1620 CTLFLAG_PERMANENT, 1621 CTLTYPE_NODE, "inet", 1622 SYSCTL_DESCR("PF_INET related settings"), 1623 NULL, 0, NULL, 0, 1624 CTL_NET, PF_INET, CTL_EOL); 1625 sysctl_createv(clog, 0, NULL, &ip_node, 1626 CTLFLAG_PERMANENT, 1627 CTLTYPE_NODE, "ip", 1628 SYSCTL_DESCR("IPv4 related settings"), 1629 NULL, 0, NULL, 0, 1630 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1631 1632 sysctl_createv(clog, 0, NULL, NULL, 1633 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1634 CTLTYPE_INT, "forwarding", 1635 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1636 NULL, 0, &ipforwarding, 0, 1637 CTL_NET, PF_INET, IPPROTO_IP, 1638 IPCTL_FORWARDING, CTL_EOL); 1639 sysctl_createv(clog, 0, NULL, NULL, 1640 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1641 CTLTYPE_INT, "redirect", 1642 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1643 NULL, 0, &ipsendredirects, 0, 1644 CTL_NET, PF_INET, IPPROTO_IP, 1645 IPCTL_SENDREDIRECTS, CTL_EOL); 1646 sysctl_createv(clog, 0, NULL, NULL, 1647 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1648 CTLTYPE_INT, "ttl", 1649 SYSCTL_DESCR("Default TTL for an INET datagram"), 1650 NULL, 0, &ip_defttl, 0, 1651 CTL_NET, PF_INET, IPPROTO_IP, 1652 IPCTL_DEFTTL, CTL_EOL); 1653 sysctl_createv(clog, 0, NULL, NULL, 1654 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1655 CTLTYPE_INT, "forwsrcrt", 1656 SYSCTL_DESCR("Enable forwarding of source-routed " 1657 "datagrams"), 1658 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1659 CTL_NET, PF_INET, IPPROTO_IP, 1660 IPCTL_FORWSRCRT, CTL_EOL); 1661 sysctl_createv(clog, 0, NULL, NULL, 1662 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1663 CTLTYPE_INT, "directed-broadcast", 1664 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1665 NULL, 0, &ip_directedbcast, 0, 1666 CTL_NET, PF_INET, IPPROTO_IP, 1667 IPCTL_DIRECTEDBCAST, CTL_EOL); 1668 sysctl_createv(clog, 0, NULL, NULL, 1669 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1670 CTLTYPE_INT, "allowsrcrt", 1671 SYSCTL_DESCR("Accept source-routed datagrams"), 1672 NULL, 0, &ip_allowsrcrt, 0, 1673 CTL_NET, PF_INET, IPPROTO_IP, 1674 IPCTL_ALLOWSRCRT, CTL_EOL); 1675 1676 sysctl_createv(clog, 0, NULL, NULL, 1677 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1678 CTLTYPE_INT, "mtudisc", 1679 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1680 NULL, 0, &ip_mtudisc, 0, 1681 CTL_NET, PF_INET, IPPROTO_IP, 1682 IPCTL_MTUDISC, CTL_EOL); 1683 sysctl_createv(clog, 0, NULL, NULL, 1684 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1685 CTLTYPE_INT, "anonportmin", 1686 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1687 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1688 CTL_NET, PF_INET, IPPROTO_IP, 1689 IPCTL_ANONPORTMIN, CTL_EOL); 1690 sysctl_createv(clog, 0, NULL, NULL, 1691 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1692 CTLTYPE_INT, "anonportmax", 1693 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1694 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1695 CTL_NET, PF_INET, IPPROTO_IP, 1696 IPCTL_ANONPORTMAX, CTL_EOL); 1697 sysctl_createv(clog, 0, NULL, NULL, 1698 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1699 CTLTYPE_INT, "mtudisctimeout", 1700 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1701 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1702 CTL_NET, PF_INET, IPPROTO_IP, 1703 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1704 #ifndef IPNOPRIVPORTS 1705 sysctl_createv(clog, 0, NULL, NULL, 1706 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1707 CTLTYPE_INT, "lowportmin", 1708 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1709 "to assign"), 1710 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1711 CTL_NET, PF_INET, IPPROTO_IP, 1712 IPCTL_LOWPORTMIN, CTL_EOL); 1713 sysctl_createv(clog, 0, NULL, NULL, 1714 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1715 CTLTYPE_INT, "lowportmax", 1716 SYSCTL_DESCR("Highest privileged ephemeral port number " 1717 "to assign"), 1718 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1719 CTL_NET, PF_INET, IPPROTO_IP, 1720 IPCTL_LOWPORTMAX, CTL_EOL); 1721 #endif /* IPNOPRIVPORTS */ 1722 #if NGRE > 0 1723 sysctl_createv(clog, 0, NULL, NULL, 1724 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1725 CTLTYPE_INT, "grettl", 1726 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1727 NULL, 0, &ip_gre_ttl, 0, 1728 CTL_NET, PF_INET, IPPROTO_IP, 1729 IPCTL_GRE_TTL, CTL_EOL); 1730 #endif /* NGRE */ 1731 sysctl_createv(clog, 0, NULL, NULL, 1732 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1733 CTLTYPE_INT, "checkinterface", 1734 SYSCTL_DESCR("Enable receive side of Strong ES model " 1735 "from RFC1122"), 1736 NULL, 0, &ip_checkinterface, 0, 1737 CTL_NET, PF_INET, IPPROTO_IP, 1738 IPCTL_CHECKINTERFACE, CTL_EOL); 1739 1740 pktq_sysctl_setup(ip_pktq, clog, ip_node, IPCTL_IFQ); 1741 1742 sysctl_createv(clog, 0, NULL, NULL, 1743 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1744 CTLTYPE_INT, "random_id", 1745 SYSCTL_DESCR("Assign random ip_id values"), 1746 NULL, 0, &ip_do_randomid, 0, 1747 CTL_NET, PF_INET, IPPROTO_IP, 1748 IPCTL_RANDOMID, CTL_EOL); 1749 sysctl_createv(clog, 0, NULL, NULL, 1750 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1751 CTLTYPE_INT, "do_loopback_cksum", 1752 SYSCTL_DESCR("Perform IP checksum on loopback"), 1753 NULL, 0, &ip_do_loopback_cksum, 0, 1754 CTL_NET, PF_INET, IPPROTO_IP, 1755 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1756 sysctl_createv(clog, 0, NULL, NULL, 1757 CTLFLAG_PERMANENT, 1758 CTLTYPE_STRUCT, "stats", 1759 SYSCTL_DESCR("IP statistics"), 1760 sysctl_net_inet_ip_stats, 0, NULL, 0, 1761 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1762 CTL_EOL); 1763 #if NARP 1764 sysctl_createv(clog, 0, NULL, NULL, 1765 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1766 CTLTYPE_INT, "dad_count", 1767 SYSCTL_DESCR("Number of Duplicate Address Detection " 1768 "probes to send"), 1769 NULL, 0, &ip_dad_count, 0, 1770 CTL_NET, PF_INET, IPPROTO_IP, 1771 IPCTL_DAD_COUNT, CTL_EOL); 1772 #endif 1773 1774 /* anonportalgo RFC6056 subtree */ 1775 const struct sysctlnode *portalgo_node; 1776 sysctl_createv(clog, 0, NULL, &portalgo_node, 1777 CTLFLAG_PERMANENT, 1778 CTLTYPE_NODE, "anonportalgo", 1779 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1780 NULL, 0, NULL, 0, 1781 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1782 sysctl_createv(clog, 0, &portalgo_node, NULL, 1783 CTLFLAG_PERMANENT, 1784 CTLTYPE_STRING, "available", 1785 SYSCTL_DESCR("available algorithms"), 1786 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1787 CTL_CREATE, CTL_EOL); 1788 sysctl_createv(clog, 0, &portalgo_node, NULL, 1789 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1790 CTLTYPE_STRING, "selected", 1791 SYSCTL_DESCR("selected algorithm"), 1792 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1793 CTL_CREATE, CTL_EOL); 1794 sysctl_createv(clog, 0, &portalgo_node, NULL, 1795 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1796 CTLTYPE_STRUCT, "reserve", 1797 SYSCTL_DESCR("bitmap of reserved ports"), 1798 sysctl_portalgo_reserve4, 0, NULL, 0, 1799 CTL_CREATE, CTL_EOL); 1800 } 1801 1802 void 1803 ip_statinc(u_int stat) 1804 { 1805 1806 KASSERT(stat < IP_NSTATS); 1807 IP_STATINC(stat); 1808 } 1809