1 /* $NetBSD: ip_input.c,v 1.385 2018/07/10 15:46:58 maxv Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 59 * POSSIBILITY OF SUCH DAMAGE. 60 */ 61 62 /* 63 * Copyright (c) 1982, 1986, 1988, 1993 64 * The Regents of the University of California. All rights reserved. 65 * 66 * Redistribution and use in source and binary forms, with or without 67 * modification, are permitted provided that the following conditions 68 * are met: 69 * 1. Redistributions of source code must retain the above copyright 70 * notice, this list of conditions and the following disclaimer. 71 * 2. Redistributions in binary form must reproduce the above copyright 72 * notice, this list of conditions and the following disclaimer in the 73 * documentation and/or other materials provided with the distribution. 74 * 3. Neither the name of the University nor the names of its contributors 75 * may be used to endorse or promote products derived from this software 76 * without specific prior written permission. 77 * 78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 88 * SUCH DAMAGE. 89 * 90 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 91 */ 92 93 #include <sys/cdefs.h> 94 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.385 2018/07/10 15:46:58 maxv Exp $"); 95 96 #ifdef _KERNEL_OPT 97 #include "opt_inet.h" 98 #include "opt_gateway.h" 99 #include "opt_ipsec.h" 100 #include "opt_mrouting.h" 101 #include "opt_mbuftrace.h" 102 #include "opt_inet_csum.h" 103 #include "opt_net_mpsafe.h" 104 #endif 105 106 #include "arp.h" 107 108 #include <sys/param.h> 109 #include <sys/systm.h> 110 #include <sys/cpu.h> 111 #include <sys/mbuf.h> 112 #include <sys/domain.h> 113 #include <sys/protosw.h> 114 #include <sys/socket.h> 115 #include <sys/socketvar.h> 116 #include <sys/errno.h> 117 #include <sys/time.h> 118 #include <sys/kernel.h> 119 #include <sys/pool.h> 120 #include <sys/sysctl.h> 121 #include <sys/kauth.h> 122 123 #include <net/if.h> 124 #include <net/if_dl.h> 125 #include <net/route.h> 126 #include <net/pktqueue.h> 127 #include <net/pfil.h> 128 129 #include <netinet/in.h> 130 #include <netinet/in_systm.h> 131 #include <netinet/ip.h> 132 #include <netinet/in_pcb.h> 133 #include <netinet/in_proto.h> 134 #include <netinet/in_var.h> 135 #include <netinet/ip_var.h> 136 #include <netinet/ip_private.h> 137 #include <netinet/ip_icmp.h> 138 /* just for gif_ttl */ 139 #include <netinet/in_gif.h> 140 #include "gif.h" 141 #include <net/if_gre.h> 142 #include "gre.h" 143 144 #ifdef MROUTING 145 #include <netinet/ip_mroute.h> 146 #endif 147 #include <netinet/portalgo.h> 148 149 #ifdef IPSEC 150 #include <netipsec/ipsec.h> 151 #endif 152 153 #ifndef IPFORWARDING 154 #ifdef GATEWAY 155 #define IPFORWARDING 1 /* forward IP packets not for us */ 156 #else 157 #define IPFORWARDING 0 /* don't forward IP packets not for us */ 158 #endif 159 #endif 160 161 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 162 163 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 = 0; 172 173 /* 174 * XXX - Setting ip_checkinterface mostly implements the receive side of 175 * the Strong ES model described in RFC 1122, but since the routing table 176 * and transmit implementation do not implement the Strong ES model, 177 * setting this to 1 results in an odd hybrid. 178 * 179 * XXX - ip_checkinterface currently must be disabled if you use ipnat 180 * to translate the destination address to another local interface. 181 * 182 * XXX - ip_checkinterface must be disabled if you add IP aliases 183 * to the loopback interface instead of the interface where the 184 * packets for those addresses are received. 185 */ 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 ipid_state_t * ip_ids __read_mostly; 193 percpu_t * ipstat_percpu __read_mostly; 194 195 static percpu_t *ipforward_rt_percpu __cacheline_aligned; 196 197 uint16_t ip_id; 198 199 #ifdef INET_CSUM_COUNTERS 200 #include <sys/device.h> 201 202 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 203 NULL, "inet", "hwcsum bad"); 204 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 205 NULL, "inet", "hwcsum ok"); 206 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 207 NULL, "inet", "swcsum"); 208 209 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 210 211 EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 212 EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 213 EVCNT_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 */ 228 struct 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 }; 235 236 static int ip_drainwanted; 237 238 static void save_rte(struct mbuf *, u_char *, struct in_addr); 239 240 #ifdef MBUFTRACE 241 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 242 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 243 #endif 244 245 static void ipintr(void *); 246 static void ip_input(struct mbuf *); 247 static void ip_forward(struct mbuf *, int, struct ifnet *); 248 static bool ip_dooptions(struct mbuf *); 249 static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *); 250 static void sysctl_net_inet_ip_setup(struct sysctllog **); 251 252 static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *, 253 int *); 254 static 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 */ 269 void 270 ip_init(void) 271 { 272 const struct protosw *pr; 273 274 in_init(); 275 sysctl_net_inet_ip_setup(NULL); 276 277 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 278 KASSERT(pr != NULL); 279 280 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 281 KASSERT(ip_pktq != NULL); 282 283 for (u_int i = 0; i < IPPROTO_MAX; i++) { 284 ip_protox[i] = pr - inetsw; 285 } 286 for (pr = inetdomain.dom_protosw; 287 pr < inetdomain.dom_protoswNPROTOSW; pr++) 288 if (pr->pr_domain->dom_family == PF_INET && 289 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 290 ip_protox[pr->pr_protocol] = pr - inetsw; 291 292 ip_reass_init(); 293 294 ip_ids = ip_id_init(); 295 ip_id = time_uptime & 0xfffff; 296 297 #ifdef GATEWAY 298 ipflow_init(); 299 #endif 300 301 /* Register our Packet Filter hook. */ 302 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 303 KASSERT(inet_pfil_hook != NULL); 304 305 #ifdef MBUFTRACE 306 MOWNER_ATTACH(&ip_tx_mowner); 307 MOWNER_ATTACH(&ip_rx_mowner); 308 #endif 309 310 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 311 ipforward_rt_percpu = percpu_alloc(sizeof(struct route)); 312 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 313 } 314 315 static struct in_ifaddr * 316 ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 317 { 318 struct in_ifaddr *ia = NULL; 319 int checkif; 320 321 /* 322 * Enable a consistency check between the destination address 323 * and the arrival interface for a unicast packet (the RFC 1122 324 * strong ES model) if IP forwarding is disabled and the packet 325 * is not locally generated. 326 * 327 * XXX - Checking also should be disabled if the destination 328 * address is ipnat'ed to a different interface. 329 * 330 * XXX - Checking is incompatible with IP aliases added 331 * to the loopback interface instead of the interface where 332 * the packets are received. 333 * 334 * XXX - We need to add a per ifaddr flag for this so that 335 * we get finer grain control. 336 */ 337 checkif = ip_checkinterface && (ipforwarding == 0) && 338 (ifp->if_flags & IFF_LOOPBACK) == 0; 339 340 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 341 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 342 if (ia->ia4_flags & IN_IFF_NOTREADY) 343 continue; 344 if (checkif && ia->ia_ifp != ifp) 345 continue; 346 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) { 347 (*downmatch)++; 348 continue; 349 } 350 if (ia->ia4_flags & IN_IFF_DETACHED && 351 (ifp->if_flags & IFF_LOOPBACK) == 0) 352 continue; 353 break; 354 } 355 } 356 357 return ia; 358 } 359 360 static struct in_ifaddr * 361 ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 362 { 363 struct in_ifaddr *ia = NULL; 364 struct ifaddr *ifa; 365 366 IFADDR_READER_FOREACH(ifa, ifp) { 367 if (ifa->ifa_addr->sa_family != AF_INET) 368 continue; 369 ia = ifatoia(ifa); 370 if (ia->ia4_flags & IN_IFF_NOTREADY) 371 continue; 372 if (ia->ia4_flags & IN_IFF_DETACHED && 373 (ifp->if_flags & IFF_LOOPBACK) == 0) 374 continue; 375 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 376 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 377 /* 378 * Look for all-0's host part (old broadcast addr), 379 * either for subnet or net. 380 */ 381 ip->ip_dst.s_addr == ia->ia_subnet || 382 ip->ip_dst.s_addr == ia->ia_net) 383 goto matched; 384 /* 385 * An interface with IP address zero accepts 386 * all packets that arrive on that interface. 387 */ 388 if (in_nullhost(ia->ia_addr.sin_addr)) 389 goto matched; 390 } 391 ia = NULL; 392 393 matched: 394 return ia; 395 } 396 397 /* 398 * IP software interrupt routine. 399 */ 400 static void 401 ipintr(void *arg __unused) 402 { 403 struct mbuf *m; 404 405 KASSERT(cpu_softintr_p()); 406 407 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 408 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 409 ip_input(m); 410 } 411 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 412 } 413 414 /* 415 * IP input routine. Checksum and byte swap header. If fragmented 416 * try to reassemble. Process options. Pass to next level. 417 */ 418 static void 419 ip_input(struct mbuf *m) 420 { 421 struct ip *ip = NULL; 422 struct in_ifaddr *ia = NULL; 423 int hlen = 0, len; 424 int downmatch; 425 int srcrt = 0; 426 ifnet_t *ifp; 427 struct psref psref; 428 int s; 429 430 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 431 "interrupt handler; synchronization assumptions violated"); 432 433 MCLAIM(m, &ip_rx_mowner); 434 KASSERT((m->m_flags & M_PKTHDR) != 0); 435 436 ifp = m_get_rcvif_psref(m, &psref); 437 if (__predict_false(ifp == NULL)) 438 goto out; 439 440 /* 441 * If no IP addresses have been set yet but the interfaces 442 * are receiving, can't do anything with incoming packets yet. 443 * Note: we pre-check without locks held. 444 */ 445 if (IN_ADDRLIST_READER_EMPTY()) 446 goto out; 447 IP_STATINC(IP_STAT_TOTAL); 448 449 /* 450 * If the IP header is not aligned, slurp it up into a new 451 * mbuf with space for link headers, in the event we forward 452 * it. Otherwise, if it is aligned, make sure the entire 453 * base IP header is in the first mbuf of the chain. 454 */ 455 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 456 if ((m = m_copyup(m, sizeof(struct ip), 457 (max_linkhdr + 3) & ~3)) == NULL) { 458 /* XXXJRT new stat, please */ 459 IP_STATINC(IP_STAT_TOOSMALL); 460 goto out; 461 } 462 } else if (__predict_false(m->m_len < sizeof(struct ip))) { 463 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) { 464 IP_STATINC(IP_STAT_TOOSMALL); 465 goto out; 466 } 467 } 468 ip = mtod(m, struct ip *); 469 if (ip->ip_v != IPVERSION) { 470 IP_STATINC(IP_STAT_BADVERS); 471 goto out; 472 } 473 hlen = ip->ip_hl << 2; 474 if (hlen < sizeof(struct ip)) { /* minimum header length */ 475 IP_STATINC(IP_STAT_BADHLEN); 476 goto out; 477 } 478 if (hlen > m->m_len) { 479 if ((m = m_pullup(m, hlen)) == NULL) { 480 IP_STATINC(IP_STAT_BADHLEN); 481 goto out; 482 } 483 ip = mtod(m, struct ip *); 484 } 485 486 /* 487 * RFC1122: packets with a multicast source address are 488 * not allowed. 489 */ 490 if (IN_MULTICAST(ip->ip_src.s_addr)) { 491 IP_STATINC(IP_STAT_BADADDR); 492 goto out; 493 } 494 495 /* 127/8 must not appear on wire - RFC1122 */ 496 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 497 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 498 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 499 IP_STATINC(IP_STAT_BADADDR); 500 goto out; 501 } 502 } 503 504 switch (m->m_pkthdr.csum_flags & 505 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) { 506 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 507 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 508 IP_STATINC(IP_STAT_BADSUM); 509 goto out; 510 511 case M_CSUM_IPv4: 512 /* Checksum was okay. */ 513 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 514 break; 515 516 default: 517 /* 518 * Must compute it ourselves. Maybe skip checksum on 519 * loopback interfaces. 520 */ 521 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 522 ip_do_loopback_cksum)) { 523 INET_CSUM_COUNTER_INCR(&ip_swcsum); 524 if (in_cksum(m, hlen) != 0) { 525 IP_STATINC(IP_STAT_BADSUM); 526 goto out; 527 } 528 } 529 break; 530 } 531 532 /* Retrieve the packet length. */ 533 len = ntohs(ip->ip_len); 534 535 /* 536 * Check for additional length bogosity 537 */ 538 if (len < hlen) { 539 IP_STATINC(IP_STAT_BADLEN); 540 goto out; 541 } 542 543 /* 544 * Check that the amount of data in the buffers is at least as much 545 * as the IP header would have us expect. Trim mbufs if longer than 546 * we expect. Drop packet if shorter than we expect. 547 */ 548 if (m->m_pkthdr.len < len) { 549 IP_STATINC(IP_STAT_TOOSHORT); 550 goto out; 551 } 552 if (m->m_pkthdr.len > len) { 553 if (m->m_len == m->m_pkthdr.len) { 554 m->m_len = len; 555 m->m_pkthdr.len = len; 556 } else 557 m_adj(m, len - m->m_pkthdr.len); 558 } 559 560 /* 561 * Assume that we can create a fast-forward IP flow entry 562 * based on this packet. 563 */ 564 m->m_flags |= M_CANFASTFWD; 565 566 /* 567 * Run through list of hooks for input packets. If there are any 568 * filters which require that additional packets in the flow are 569 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 570 * Note that filters must _never_ set this flag, as another filter 571 * in the list may have previously cleared it. 572 * 573 * Don't call hooks if the packet has already been processed by 574 * IPsec (encapsulated, tunnel mode). 575 */ 576 #if defined(IPSEC) 577 if (!ipsec_used || !ipsec_indone(m)) 578 #else 579 if (1) 580 #endif 581 { 582 struct in_addr odst = ip->ip_dst; 583 bool freed; 584 585 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 586 if (freed || m == NULL) { 587 m = NULL; 588 goto out; 589 } 590 KASSERT(m->m_len >= sizeof(struct ip)); 591 ip = mtod(m, struct ip *); 592 hlen = ip->ip_hl << 2; 593 KASSERT(m->m_len >= hlen); 594 595 /* 596 * XXX The setting of "srcrt" here is to prevent ip_forward() 597 * from generating ICMP redirects for packets that have 598 * been redirected by a hook back out on to the same LAN that 599 * they came from and is not an indication that the packet 600 * is being influenced by source routing options. This 601 * allows things like 602 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 603 * where tlp0 is both on the 1.1.1.0/24 network and is the 604 * default route for hosts on 1.1.1.0/24. Of course this 605 * also requires a "map tlp0 ..." to complete the story. 606 * One might argue whether or not this kind of network config. 607 * should be supported in this manner... 608 */ 609 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 610 } 611 612 #ifdef ALTQ 613 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 614 if (altq_input) { 615 SOFTNET_LOCK(); 616 if ((*altq_input)(m, AF_INET) == 0) { 617 /* Packet dropped by traffic conditioner. */ 618 SOFTNET_UNLOCK(); 619 m = NULL; 620 goto out; 621 } 622 SOFTNET_UNLOCK(); 623 } 624 #endif 625 626 /* 627 * Process options and, if not destined for us, 628 * ship it on. ip_dooptions returns 1 when an 629 * error was detected (causing an icmp message 630 * to be sent and the original packet to be freed). 631 */ 632 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 633 m = NULL; 634 goto out; 635 } 636 637 /* 638 * Check our list of addresses, to see if the packet is for us. 639 * 640 * Traditional 4.4BSD did not consult IFF_UP at all. 641 * The behavior here is to treat addresses on !IFF_UP interface 642 * or IN_IFF_NOTREADY addresses as not mine. 643 */ 644 downmatch = 0; 645 s = pserialize_read_enter(); 646 ia = ip_match_our_address(ifp, ip, &downmatch); 647 if (ia != NULL) { 648 pserialize_read_exit(s); 649 goto ours; 650 } 651 652 if (ifp->if_flags & IFF_BROADCAST) { 653 ia = ip_match_our_address_broadcast(ifp, ip); 654 if (ia != NULL) { 655 pserialize_read_exit(s); 656 goto ours; 657 } 658 } 659 pserialize_read_exit(s); 660 661 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 662 #ifdef MROUTING 663 extern struct socket *ip_mrouter; 664 665 if (ip_mrouter) { 666 /* 667 * If we are acting as a multicast router, all 668 * incoming multicast packets are passed to the 669 * kernel-level multicast forwarding function. 670 * The packet is returned (relatively) intact; if 671 * ip_mforward() returns a non-zero value, the packet 672 * must be discarded, else it may be accepted below. 673 * 674 * (The IP ident field is put in the same byte order 675 * as expected when ip_mforward() is called from 676 * ip_output().) 677 */ 678 SOFTNET_LOCK(); 679 if (ip_mforward(m, ifp) != 0) { 680 SOFTNET_UNLOCK(); 681 IP_STATINC(IP_STAT_CANTFORWARD); 682 goto out; 683 } 684 SOFTNET_UNLOCK(); 685 686 /* 687 * The process-level routing demon needs to receive 688 * all multicast IGMP packets, whether or not this 689 * host belongs to their destination groups. 690 */ 691 if (ip->ip_p == IPPROTO_IGMP) { 692 goto ours; 693 } 694 IP_STATINC(IP_STAT_CANTFORWARD); 695 } 696 #endif 697 /* 698 * See if we belong to the destination multicast group on the 699 * arrival interface. 700 */ 701 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 702 IP_STATINC(IP_STAT_CANTFORWARD); 703 goto out; 704 } 705 goto ours; 706 } 707 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 708 in_nullhost(ip->ip_dst)) 709 goto ours; 710 711 /* 712 * Not for us; forward if possible and desirable. 713 */ 714 if (ipforwarding == 0) { 715 m_put_rcvif_psref(ifp, &psref); 716 IP_STATINC(IP_STAT_CANTFORWARD); 717 m_freem(m); 718 } else { 719 /* 720 * If ip_dst matched any of my address on !IFF_UP interface, 721 * and there's no IFF_UP interface that matches ip_dst, 722 * send icmp unreach. Forwarding it will result in in-kernel 723 * forwarding loop till TTL goes to 0. 724 */ 725 if (downmatch) { 726 m_put_rcvif_psref(ifp, &psref); 727 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 728 IP_STATINC(IP_STAT_CANTFORWARD); 729 return; 730 } 731 #ifdef IPSEC 732 /* Check the security policy (SP) for the packet */ 733 if (ipsec_used) { 734 if (ipsec_ip_input(m, true) != 0) { 735 goto out; 736 } 737 } 738 #endif 739 ip_forward(m, srcrt, ifp); 740 m_put_rcvif_psref(ifp, &psref); 741 } 742 return; 743 744 ours: 745 m_put_rcvif_psref(ifp, &psref); 746 ifp = NULL; 747 748 /* 749 * If offset or IP_MF are set, must reassemble. 750 */ 751 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 752 /* 753 * Pass to IP reassembly mechanism. 754 */ 755 if (ip_reass_packet(&m) != 0) { 756 /* Failed; invalid fragment(s) or packet. */ 757 goto out; 758 } 759 if (m == NULL) { 760 /* More fragments should come; silently return. */ 761 goto out; 762 } 763 /* 764 * Reassembly is done, we have the final packet. 765 * Update cached data in local variable(s). 766 */ 767 ip = mtod(m, struct ip *); 768 hlen = ip->ip_hl << 2; 769 } 770 771 M_VERIFY_PACKET(m); 772 773 #ifdef IPSEC 774 /* 775 * Enforce IPsec policy checking if we are seeing last header. 776 * Note that we do not visit this with protocols with PCB layer 777 * code - like UDP/TCP/raw IP. 778 */ 779 if (ipsec_used && 780 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 781 if (ipsec_ip_input(m, false) != 0) { 782 goto out; 783 } 784 } 785 #endif 786 787 /* 788 * Switch out to protocol's input routine. 789 */ 790 #if IFA_STATS 791 if (ia) { 792 struct in_ifaddr *_ia; 793 /* 794 * Keep a reference from ip_match_our_address with psref 795 * is expensive, so explore ia here again. 796 */ 797 s = pserialize_read_enter(); 798 _ia = in_get_ia(ip->ip_dst); 799 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 800 pserialize_read_exit(s); 801 } 802 #endif 803 IP_STATINC(IP_STAT_DELIVERED); 804 805 const int off = hlen, nh = ip->ip_p; 806 807 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 808 return; 809 810 out: 811 m_put_rcvif_psref(ifp, &psref); 812 if (m != NULL) 813 m_freem(m); 814 } 815 816 /* 817 * IP timer processing. 818 */ 819 void 820 ip_slowtimo(void) 821 { 822 823 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 824 825 ip_reass_slowtimo(); 826 827 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 828 } 829 830 /* 831 * IP drain processing. 832 */ 833 void 834 ip_drain(void) 835 { 836 837 KERNEL_LOCK(1, NULL); 838 ip_reass_drain(); 839 KERNEL_UNLOCK_ONE(NULL); 840 } 841 842 /* 843 * ip_dooptions: perform option processing on a datagram, possibly discarding 844 * it if bad options are encountered, or forwarding it if source-routed. 845 * 846 * => Returns true if packet has been forwarded/freed. 847 * => Returns false if the packet should be processed further. 848 */ 849 static bool 850 ip_dooptions(struct mbuf *m) 851 { 852 struct ip *ip = mtod(m, struct ip *); 853 u_char *cp, *cp0; 854 struct ip_timestamp *ipt; 855 struct in_ifaddr *ia; 856 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 857 int srr_present, rr_present, ts_present; 858 struct in_addr dst; 859 n_time ntime; 860 struct ifaddr *ifa = NULL; 861 int s; 862 863 srr_present = 0; 864 rr_present = 0; 865 ts_present = 0; 866 867 dst = ip->ip_dst; 868 cp = (u_char *)(ip + 1); 869 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 870 for (; cnt > 0; cnt -= optlen, cp += optlen) { 871 opt = cp[IPOPT_OPTVAL]; 872 if (opt == IPOPT_EOL) 873 break; 874 if (opt == IPOPT_NOP) 875 optlen = 1; 876 else { 877 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 878 code = &cp[IPOPT_OLEN] - (u_char *)ip; 879 goto bad; 880 } 881 optlen = cp[IPOPT_OLEN]; 882 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 883 code = &cp[IPOPT_OLEN] - (u_char *)ip; 884 goto bad; 885 } 886 } 887 switch (opt) { 888 889 default: 890 break; 891 892 /* 893 * Source routing with record. 894 * Find interface with current destination address. 895 * If none on this machine then drop if strictly routed, 896 * or do nothing if loosely routed. 897 * Record interface address and bring up next address 898 * component. If strictly routed make sure next 899 * address is on directly accessible net. 900 */ 901 case IPOPT_LSRR: 902 case IPOPT_SSRR: { 903 struct psref psref; 904 struct sockaddr_in ipaddr = { 905 .sin_len = sizeof(ipaddr), 906 .sin_family = AF_INET, 907 }; 908 909 if (ip_allowsrcrt == 0) { 910 type = ICMP_UNREACH; 911 code = ICMP_UNREACH_NET_PROHIB; 912 goto bad; 913 } 914 if (srr_present++) { 915 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 916 goto bad; 917 } 918 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 919 code = &cp[IPOPT_OLEN] - (u_char *)ip; 920 goto bad; 921 } 922 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 923 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 924 goto bad; 925 } 926 ipaddr.sin_addr = ip->ip_dst; 927 928 s = pserialize_read_enter(); 929 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 930 if (ifa == NULL) { 931 pserialize_read_exit(s); 932 if (opt == IPOPT_SSRR) { 933 type = ICMP_UNREACH; 934 code = ICMP_UNREACH_SRCFAIL; 935 goto bad; 936 } 937 /* 938 * Loose routing, and not at next destination 939 * yet; nothing to do except forward. 940 */ 941 break; 942 } 943 pserialize_read_exit(s); 944 945 off--; /* 0 origin */ 946 if ((off + sizeof(struct in_addr)) > optlen) { 947 /* 948 * End of source route. Should be for us. 949 */ 950 save_rte(m, cp, ip->ip_src); 951 break; 952 } 953 /* 954 * locate outgoing interface 955 */ 956 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 957 sizeof(ipaddr.sin_addr)); 958 if (opt == IPOPT_SSRR) { 959 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 960 &psref); 961 if (ifa != NULL) 962 ia = ifatoia(ifa); 963 else 964 ia = NULL; 965 } else { 966 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 967 } 968 if (ia == NULL) { 969 type = ICMP_UNREACH; 970 code = ICMP_UNREACH_SRCFAIL; 971 goto bad; 972 } 973 ip->ip_dst = ipaddr.sin_addr; 974 memcpy(cp + off, &ia->ia_addr.sin_addr, 975 sizeof(struct in_addr)); 976 ia4_release(ia, &psref); 977 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 978 /* 979 * Let ip_intr's mcast routing check handle mcast pkts 980 */ 981 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 982 break; 983 } 984 985 case IPOPT_RR: { 986 struct psref psref; 987 struct sockaddr_in ipaddr = { 988 .sin_len = sizeof(ipaddr), 989 .sin_family = AF_INET, 990 }; 991 992 if (rr_present++) { 993 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 994 goto bad; 995 } 996 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 997 code = &cp[IPOPT_OLEN] - (u_char *)ip; 998 goto bad; 999 } 1000 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1001 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1002 goto bad; 1003 } 1004 /* 1005 * If no space remains, ignore. 1006 */ 1007 off--; /* 0 origin */ 1008 if ((off + sizeof(struct in_addr)) > optlen) 1009 break; 1010 memcpy((void *)&ipaddr.sin_addr, (void *)&ip->ip_dst, 1011 sizeof(ipaddr.sin_addr)); 1012 /* 1013 * locate outgoing interface; if we're the destination, 1014 * use the incoming interface (should be same). 1015 */ 1016 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1017 if (ifa == NULL) { 1018 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1019 if (ia == NULL) { 1020 type = ICMP_UNREACH; 1021 code = ICMP_UNREACH_HOST; 1022 goto bad; 1023 } 1024 } else { 1025 ia = ifatoia(ifa); 1026 } 1027 memcpy(cp + off, &ia->ia_addr.sin_addr, 1028 sizeof(struct in_addr)); 1029 ia4_release(ia, &psref); 1030 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1031 break; 1032 } 1033 1034 case IPOPT_TS: 1035 code = cp - (u_char *)ip; 1036 ipt = (struct ip_timestamp *)cp; 1037 if (ts_present++) { 1038 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1039 goto bad; 1040 } 1041 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1042 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1043 goto bad; 1044 } 1045 if (ipt->ipt_ptr < 5) { 1046 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1047 goto bad; 1048 } 1049 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1050 if (++ipt->ipt_oflw == 0) { 1051 code = (u_char *)&ipt->ipt_ptr - 1052 (u_char *)ip; 1053 goto bad; 1054 } 1055 break; 1056 } 1057 cp0 = (cp + ipt->ipt_ptr - 1); 1058 switch (ipt->ipt_flg) { 1059 1060 case IPOPT_TS_TSONLY: 1061 break; 1062 1063 case IPOPT_TS_TSANDADDR: { 1064 struct ifnet *rcvif; 1065 int _s, _ss; 1066 struct sockaddr_in ipaddr = { 1067 .sin_len = sizeof(ipaddr), 1068 .sin_family = AF_INET, 1069 }; 1070 1071 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1072 sizeof(struct in_addr) > ipt->ipt_len) { 1073 code = (u_char *)&ipt->ipt_ptr - 1074 (u_char *)ip; 1075 goto bad; 1076 } 1077 ipaddr.sin_addr = dst; 1078 _ss = pserialize_read_enter(); 1079 rcvif = m_get_rcvif(m, &_s); 1080 if (__predict_true(rcvif != NULL)) { 1081 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1082 rcvif); 1083 } 1084 m_put_rcvif(rcvif, &_s); 1085 if (ifa == NULL) { 1086 pserialize_read_exit(_ss); 1087 break; 1088 } 1089 ia = ifatoia(ifa); 1090 memcpy(cp0, &ia->ia_addr.sin_addr, 1091 sizeof(struct in_addr)); 1092 pserialize_read_exit(_ss); 1093 ipt->ipt_ptr += sizeof(struct in_addr); 1094 break; 1095 } 1096 1097 case IPOPT_TS_PRESPEC: { 1098 struct sockaddr_in ipaddr = { 1099 .sin_len = sizeof(ipaddr), 1100 .sin_family = AF_INET, 1101 }; 1102 1103 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1104 sizeof(struct in_addr) > ipt->ipt_len) { 1105 code = (u_char *)&ipt->ipt_ptr - 1106 (u_char *)ip; 1107 goto bad; 1108 } 1109 memcpy(&ipaddr.sin_addr, cp0, 1110 sizeof(struct in_addr)); 1111 s = pserialize_read_enter(); 1112 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1113 if (ifa == NULL) { 1114 pserialize_read_exit(s); 1115 continue; 1116 } 1117 pserialize_read_exit(s); 1118 ipt->ipt_ptr += sizeof(struct in_addr); 1119 break; 1120 } 1121 1122 default: 1123 /* XXX can't take &ipt->ipt_flg */ 1124 code = (u_char *)&ipt->ipt_ptr - 1125 (u_char *)ip + 1; 1126 goto bad; 1127 } 1128 ntime = iptime(); 1129 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1130 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1131 sizeof(n_time)); 1132 ipt->ipt_ptr += sizeof(n_time); 1133 } 1134 } 1135 if (forward) { 1136 struct ifnet *rcvif; 1137 struct psref _psref; 1138 1139 if (ip_forwsrcrt == 0) { 1140 type = ICMP_UNREACH; 1141 code = ICMP_UNREACH_SRCFAIL; 1142 goto bad; 1143 } 1144 1145 rcvif = m_get_rcvif_psref(m, &_psref); 1146 if (__predict_false(rcvif == NULL)) { 1147 type = ICMP_UNREACH; 1148 code = ICMP_UNREACH_HOST; 1149 goto bad; 1150 } 1151 ip_forward(m, 1, rcvif); 1152 m_put_rcvif_psref(rcvif, &_psref); 1153 return true; 1154 } 1155 return false; 1156 bad: 1157 icmp_error(m, type, code, 0, 0); 1158 IP_STATINC(IP_STAT_BADOPTIONS); 1159 return true; 1160 } 1161 1162 /* 1163 * ip_rtaddr: given address of next destination (final or next hop), 1164 * return internet address info of interface to be used to get there. 1165 */ 1166 static struct in_ifaddr * 1167 ip_rtaddr(struct in_addr dst, struct psref *psref) 1168 { 1169 struct rtentry *rt; 1170 union { 1171 struct sockaddr dst; 1172 struct sockaddr_in dst4; 1173 } u; 1174 struct route *ro; 1175 1176 sockaddr_in_init(&u.dst4, &dst, 0); 1177 1178 ro = percpu_getref(ipforward_rt_percpu); 1179 rt = rtcache_lookup(ro, &u.dst); 1180 if (rt == NULL) { 1181 percpu_putref(ipforward_rt_percpu); 1182 return NULL; 1183 } 1184 1185 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1186 rtcache_unref(rt, ro); 1187 percpu_putref(ipforward_rt_percpu); 1188 1189 return ifatoia(rt->rt_ifa); 1190 } 1191 1192 /* 1193 * save_rte: save incoming source route for use in replies, to be picked 1194 * up later by ip_srcroute if the receiver is interested. 1195 */ 1196 static void 1197 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1198 { 1199 struct ip_srcrt *isr; 1200 struct m_tag *mtag; 1201 unsigned olen; 1202 1203 olen = option[IPOPT_OLEN]; 1204 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1205 return; 1206 1207 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1208 if (mtag == NULL) 1209 return; 1210 isr = (struct ip_srcrt *)(mtag + 1); 1211 1212 memcpy(isr->isr_hdr, option, olen); 1213 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1214 isr->isr_dst = dst; 1215 m_tag_prepend(m, mtag); 1216 } 1217 1218 /* 1219 * Retrieve incoming source route for use in replies, 1220 * in the same form used by setsockopt. 1221 * The first hop is placed before the options, will be removed later. 1222 */ 1223 struct mbuf * 1224 ip_srcroute(struct mbuf *m0) 1225 { 1226 struct in_addr *p, *q; 1227 struct mbuf *m; 1228 struct ip_srcrt *isr; 1229 struct m_tag *mtag; 1230 1231 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL); 1232 if (mtag == NULL) 1233 return NULL; 1234 isr = (struct ip_srcrt *)(mtag + 1); 1235 1236 if (isr->isr_nhops == 0) 1237 return NULL; 1238 1239 m = m_get(M_DONTWAIT, MT_SOOPTS); 1240 if (m == NULL) 1241 return NULL; 1242 1243 MCLAIM(m, &inetdomain.dom_mowner); 1244 #define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1245 1246 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1247 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1248 1249 /* 1250 * First save first hop for return route 1251 */ 1252 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1253 *(mtod(m, struct in_addr *)) = *p--; 1254 1255 /* 1256 * Copy option fields and padding (nop) to mbuf. 1257 */ 1258 isr->isr_nop = IPOPT_NOP; 1259 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1260 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1261 OPTSIZ); 1262 q = (struct in_addr *)(mtod(m, char *) + 1263 sizeof(struct in_addr) + OPTSIZ); 1264 #undef OPTSIZ 1265 /* 1266 * Record return path as an IP source route, 1267 * reversing the path (pointers are now aligned). 1268 */ 1269 while (p >= isr->isr_routes) { 1270 *q++ = *p--; 1271 } 1272 /* 1273 * Last hop goes to final destination. 1274 */ 1275 *q = isr->isr_dst; 1276 m_tag_delete(m0, mtag); 1277 return m; 1278 } 1279 1280 const int inetctlerrmap[PRC_NCMDS] = { 1281 [PRC_MSGSIZE] = EMSGSIZE, 1282 [PRC_HOSTDEAD] = EHOSTDOWN, 1283 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1284 [PRC_UNREACH_NET] = EHOSTUNREACH, 1285 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1286 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1287 [PRC_UNREACH_PORT] = ECONNREFUSED, 1288 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1289 [PRC_PARAMPROB] = ENOPROTOOPT, 1290 }; 1291 1292 void 1293 ip_fasttimo(void) 1294 { 1295 if (ip_drainwanted) { 1296 ip_drain(); 1297 ip_drainwanted = 0; 1298 } 1299 } 1300 1301 void 1302 ip_drainstub(void) 1303 { 1304 ip_drainwanted = 1; 1305 } 1306 1307 /* 1308 * Forward a packet. If some error occurs return the sender 1309 * an icmp packet. Note we can't always generate a meaningful 1310 * icmp message because icmp doesn't have a large enough repertoire 1311 * of codes and types. 1312 * 1313 * If not forwarding, just drop the packet. This could be confusing 1314 * if ipforwarding was zero but some routing protocol was advancing 1315 * us as a gateway to somewhere. However, we must let the routing 1316 * protocol deal with that. 1317 * 1318 * The srcrt parameter indicates whether the packet is being forwarded 1319 * via a source route. 1320 */ 1321 static void 1322 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1323 { 1324 struct ip *ip = mtod(m, struct ip *); 1325 struct rtentry *rt; 1326 int error, type = 0, code = 0, destmtu = 0; 1327 struct mbuf *mcopy; 1328 n_long dest; 1329 union { 1330 struct sockaddr dst; 1331 struct sockaddr_in dst4; 1332 } u; 1333 uint64_t *ips; 1334 struct route *ro; 1335 1336 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1337 "interrupt handler; synchronization assumptions violated"); 1338 1339 /* 1340 * We are now in the output path. 1341 */ 1342 MCLAIM(m, &ip_tx_mowner); 1343 1344 /* 1345 * Clear any in-bound checksum flags for this packet. 1346 */ 1347 m->m_pkthdr.csum_flags = 0; 1348 1349 dest = 0; 1350 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1351 IP_STATINC(IP_STAT_CANTFORWARD); 1352 m_freem(m); 1353 return; 1354 } 1355 1356 if (ip->ip_ttl <= IPTTLDEC) { 1357 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1358 return; 1359 } 1360 1361 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1362 1363 ro = percpu_getref(ipforward_rt_percpu); 1364 rt = rtcache_lookup(ro, &u.dst); 1365 if (rt == NULL) { 1366 percpu_putref(ipforward_rt_percpu); 1367 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1368 return; 1369 } 1370 1371 /* 1372 * Save at most 68 bytes of the packet in case 1373 * we need to generate an ICMP message to the src. 1374 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1375 */ 1376 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1377 if (mcopy) 1378 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1379 1380 ip->ip_ttl -= IPTTLDEC; 1381 1382 /* 1383 * If forwarding packet using same interface that it came in on, 1384 * perhaps should send a redirect to sender to shortcut a hop. 1385 * Only send redirect if source is sending directly to us, 1386 * and if packet was not source routed (or has any options). 1387 * Also, don't send redirect if forwarding using a default route 1388 * or a route modified by a redirect. 1389 */ 1390 if (rt->rt_ifp == rcvif && 1391 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1392 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1393 ipsendredirects && !srcrt) { 1394 if (rt->rt_ifa && 1395 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1396 ifatoia(rt->rt_ifa)->ia_subnet) { 1397 if (rt->rt_flags & RTF_GATEWAY) 1398 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1399 else 1400 dest = ip->ip_dst.s_addr; 1401 /* 1402 * Router requirements says to only send host 1403 * redirects. 1404 */ 1405 type = ICMP_REDIRECT; 1406 code = ICMP_REDIRECT_HOST; 1407 } 1408 } 1409 rtcache_unref(rt, ro); 1410 1411 error = ip_output(m, NULL, ro, 1412 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1413 NULL, NULL); 1414 1415 if (error) { 1416 IP_STATINC(IP_STAT_CANTFORWARD); 1417 goto error; 1418 } 1419 1420 ips = IP_STAT_GETREF(); 1421 ips[IP_STAT_FORWARD]++; 1422 1423 if (type) { 1424 ips[IP_STAT_REDIRECTSENT]++; 1425 IP_STAT_PUTREF(); 1426 goto redirect; 1427 } 1428 1429 IP_STAT_PUTREF(); 1430 if (mcopy) { 1431 #ifdef GATEWAY 1432 if (mcopy->m_flags & M_CANFASTFWD) 1433 ipflow_create(ro, mcopy); 1434 #endif 1435 m_freem(mcopy); 1436 } 1437 1438 percpu_putref(ipforward_rt_percpu); 1439 return; 1440 1441 redirect: 1442 error: 1443 if (mcopy == NULL) { 1444 percpu_putref(ipforward_rt_percpu); 1445 return; 1446 } 1447 1448 switch (error) { 1449 1450 case 0: /* forwarded, but need redirect */ 1451 /* type, code set above */ 1452 break; 1453 1454 case ENETUNREACH: /* shouldn't happen, checked above */ 1455 case EHOSTUNREACH: 1456 case ENETDOWN: 1457 case EHOSTDOWN: 1458 default: 1459 type = ICMP_UNREACH; 1460 code = ICMP_UNREACH_HOST; 1461 break; 1462 1463 case EMSGSIZE: 1464 type = ICMP_UNREACH; 1465 code = ICMP_UNREACH_NEEDFRAG; 1466 1467 if ((rt = rtcache_validate(ro)) != NULL) { 1468 destmtu = rt->rt_ifp->if_mtu; 1469 rtcache_unref(rt, ro); 1470 } 1471 #ifdef IPSEC 1472 if (ipsec_used) 1473 ipsec_mtu(mcopy, &destmtu); 1474 #endif 1475 IP_STATINC(IP_STAT_CANTFRAG); 1476 break; 1477 1478 case ENOBUFS: 1479 /* 1480 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1481 * Requirements for IP Version 4 Routers. Source quench can 1482 * be a big problem under DoS attacks or if the underlying 1483 * interface is rate-limited. 1484 */ 1485 if (mcopy) 1486 m_freem(mcopy); 1487 percpu_putref(ipforward_rt_percpu); 1488 return; 1489 } 1490 icmp_error(mcopy, type, code, dest, destmtu); 1491 percpu_putref(ipforward_rt_percpu); 1492 } 1493 1494 void 1495 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1496 struct mbuf *m) 1497 { 1498 struct socket *so = inp->inp_socket; 1499 int inpflags = inp->inp_flags; 1500 1501 if (SOOPT_TIMESTAMP(so->so_options)) 1502 mp = sbsavetimestamp(so->so_options, mp); 1503 1504 if (inpflags & INP_RECVDSTADDR) { 1505 *mp = sbcreatecontrol(&ip->ip_dst, 1506 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1507 if (*mp) 1508 mp = &(*mp)->m_next; 1509 } 1510 1511 if (inpflags & INP_RECVTTL) { 1512 *mp = sbcreatecontrol(&ip->ip_ttl, 1513 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1514 if (*mp) 1515 mp = &(*mp)->m_next; 1516 } 1517 1518 struct psref psref; 1519 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1520 if (__predict_false(ifp == NULL)) { 1521 #ifdef DIAGNOSTIC 1522 printf("%s: missing receive interface\n", __func__); 1523 #endif 1524 return; /* XXX should report error? */ 1525 } 1526 1527 if (inpflags & INP_RECVPKTINFO) { 1528 struct in_pktinfo ipi; 1529 ipi.ipi_addr = ip->ip_dst; 1530 ipi.ipi_ifindex = ifp->if_index; 1531 *mp = sbcreatecontrol(&ipi, 1532 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1533 if (*mp) 1534 mp = &(*mp)->m_next; 1535 } 1536 if (inpflags & INP_RECVIF) { 1537 struct sockaddr_dl sdl; 1538 1539 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1540 NULL, 0); 1541 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1542 if (*mp) 1543 mp = &(*mp)->m_next; 1544 } 1545 m_put_rcvif_psref(ifp, &psref); 1546 } 1547 1548 /* 1549 * sysctl helper routine for net.inet.ip.forwsrcrt. 1550 */ 1551 static int 1552 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1553 { 1554 int error, tmp; 1555 struct sysctlnode node; 1556 1557 node = *rnode; 1558 tmp = ip_forwsrcrt; 1559 node.sysctl_data = &tmp; 1560 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1561 if (error || newp == NULL) 1562 return (error); 1563 1564 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1565 0, NULL, NULL, NULL); 1566 if (error) 1567 return (error); 1568 1569 ip_forwsrcrt = tmp; 1570 1571 return (0); 1572 } 1573 1574 /* 1575 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1576 * range of the new value and tweaks timers if it changes. 1577 */ 1578 static int 1579 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1580 { 1581 int error, tmp; 1582 struct sysctlnode node; 1583 1584 icmp_mtudisc_lock(); 1585 1586 node = *rnode; 1587 tmp = ip_mtudisc_timeout; 1588 node.sysctl_data = &tmp; 1589 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1590 if (error || newp == NULL) 1591 goto out; 1592 if (tmp < 0) { 1593 error = EINVAL; 1594 goto out; 1595 } 1596 1597 ip_mtudisc_timeout = tmp; 1598 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1599 error = 0; 1600 out: 1601 icmp_mtudisc_unlock(); 1602 return error; 1603 } 1604 1605 static int 1606 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1607 { 1608 1609 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1610 } 1611 1612 static void 1613 sysctl_net_inet_ip_setup(struct sysctllog **clog) 1614 { 1615 sysctl_createv(clog, 0, NULL, NULL, 1616 CTLFLAG_PERMANENT, 1617 CTLTYPE_NODE, "inet", 1618 SYSCTL_DESCR("PF_INET related settings"), 1619 NULL, 0, NULL, 0, 1620 CTL_NET, PF_INET, CTL_EOL); 1621 sysctl_createv(clog, 0, NULL, NULL, 1622 CTLFLAG_PERMANENT, 1623 CTLTYPE_NODE, "ip", 1624 SYSCTL_DESCR("IPv4 related settings"), 1625 NULL, 0, NULL, 0, 1626 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1627 1628 sysctl_createv(clog, 0, NULL, NULL, 1629 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1630 CTLTYPE_INT, "forwarding", 1631 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1632 NULL, 0, &ipforwarding, 0, 1633 CTL_NET, PF_INET, IPPROTO_IP, 1634 IPCTL_FORWARDING, CTL_EOL); 1635 sysctl_createv(clog, 0, NULL, NULL, 1636 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1637 CTLTYPE_INT, "redirect", 1638 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1639 NULL, 0, &ipsendredirects, 0, 1640 CTL_NET, PF_INET, IPPROTO_IP, 1641 IPCTL_SENDREDIRECTS, CTL_EOL); 1642 sysctl_createv(clog, 0, NULL, NULL, 1643 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1644 CTLTYPE_INT, "ttl", 1645 SYSCTL_DESCR("Default TTL for an INET datagram"), 1646 NULL, 0, &ip_defttl, 0, 1647 CTL_NET, PF_INET, IPPROTO_IP, 1648 IPCTL_DEFTTL, CTL_EOL); 1649 sysctl_createv(clog, 0, NULL, NULL, 1650 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1651 CTLTYPE_INT, "forwsrcrt", 1652 SYSCTL_DESCR("Enable forwarding of source-routed " 1653 "datagrams"), 1654 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1655 CTL_NET, PF_INET, IPPROTO_IP, 1656 IPCTL_FORWSRCRT, CTL_EOL); 1657 sysctl_createv(clog, 0, NULL, NULL, 1658 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1659 CTLTYPE_INT, "directed-broadcast", 1660 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1661 NULL, 0, &ip_directedbcast, 0, 1662 CTL_NET, PF_INET, IPPROTO_IP, 1663 IPCTL_DIRECTEDBCAST, CTL_EOL); 1664 sysctl_createv(clog, 0, NULL, NULL, 1665 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1666 CTLTYPE_INT, "allowsrcrt", 1667 SYSCTL_DESCR("Accept source-routed datagrams"), 1668 NULL, 0, &ip_allowsrcrt, 0, 1669 CTL_NET, PF_INET, IPPROTO_IP, 1670 IPCTL_ALLOWSRCRT, CTL_EOL); 1671 1672 sysctl_createv(clog, 0, NULL, NULL, 1673 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1674 CTLTYPE_INT, "mtudisc", 1675 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1676 NULL, 0, &ip_mtudisc, 0, 1677 CTL_NET, PF_INET, IPPROTO_IP, 1678 IPCTL_MTUDISC, CTL_EOL); 1679 sysctl_createv(clog, 0, NULL, NULL, 1680 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1681 CTLTYPE_INT, "anonportmin", 1682 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1683 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1684 CTL_NET, PF_INET, IPPROTO_IP, 1685 IPCTL_ANONPORTMIN, CTL_EOL); 1686 sysctl_createv(clog, 0, NULL, NULL, 1687 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1688 CTLTYPE_INT, "anonportmax", 1689 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1690 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1691 CTL_NET, PF_INET, IPPROTO_IP, 1692 IPCTL_ANONPORTMAX, CTL_EOL); 1693 sysctl_createv(clog, 0, NULL, NULL, 1694 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1695 CTLTYPE_INT, "mtudisctimeout", 1696 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1697 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1698 CTL_NET, PF_INET, IPPROTO_IP, 1699 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1700 #ifndef IPNOPRIVPORTS 1701 sysctl_createv(clog, 0, NULL, NULL, 1702 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1703 CTLTYPE_INT, "lowportmin", 1704 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1705 "to assign"), 1706 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1707 CTL_NET, PF_INET, IPPROTO_IP, 1708 IPCTL_LOWPORTMIN, CTL_EOL); 1709 sysctl_createv(clog, 0, NULL, NULL, 1710 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1711 CTLTYPE_INT, "lowportmax", 1712 SYSCTL_DESCR("Highest privileged ephemeral port number " 1713 "to assign"), 1714 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1715 CTL_NET, PF_INET, IPPROTO_IP, 1716 IPCTL_LOWPORTMAX, CTL_EOL); 1717 #endif /* IPNOPRIVPORTS */ 1718 #if NGRE > 0 1719 sysctl_createv(clog, 0, NULL, NULL, 1720 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1721 CTLTYPE_INT, "grettl", 1722 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1723 NULL, 0, &ip_gre_ttl, 0, 1724 CTL_NET, PF_INET, IPPROTO_IP, 1725 IPCTL_GRE_TTL, CTL_EOL); 1726 #endif /* NGRE */ 1727 sysctl_createv(clog, 0, NULL, NULL, 1728 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1729 CTLTYPE_INT, "checkinterface", 1730 SYSCTL_DESCR("Enable receive side of Strong ES model " 1731 "from RFC1122"), 1732 NULL, 0, &ip_checkinterface, 0, 1733 CTL_NET, PF_INET, IPPROTO_IP, 1734 IPCTL_CHECKINTERFACE, CTL_EOL); 1735 sysctl_createv(clog, 0, NULL, NULL, 1736 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1737 CTLTYPE_INT, "random_id", 1738 SYSCTL_DESCR("Assign random ip_id values"), 1739 NULL, 0, &ip_do_randomid, 0, 1740 CTL_NET, PF_INET, IPPROTO_IP, 1741 IPCTL_RANDOMID, CTL_EOL); 1742 sysctl_createv(clog, 0, NULL, NULL, 1743 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1744 CTLTYPE_INT, "do_loopback_cksum", 1745 SYSCTL_DESCR("Perform IP checksum on loopback"), 1746 NULL, 0, &ip_do_loopback_cksum, 0, 1747 CTL_NET, PF_INET, IPPROTO_IP, 1748 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1749 sysctl_createv(clog, 0, NULL, NULL, 1750 CTLFLAG_PERMANENT, 1751 CTLTYPE_STRUCT, "stats", 1752 SYSCTL_DESCR("IP statistics"), 1753 sysctl_net_inet_ip_stats, 0, NULL, 0, 1754 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1755 CTL_EOL); 1756 #if NARP 1757 sysctl_createv(clog, 0, NULL, NULL, 1758 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1759 CTLTYPE_INT, "dad_count", 1760 SYSCTL_DESCR("Number of Duplicate Address Detection " 1761 "probes to send"), 1762 NULL, 0, &ip_dad_count, 0, 1763 CTL_NET, PF_INET, IPPROTO_IP, 1764 IPCTL_DAD_COUNT, CTL_EOL); 1765 #endif 1766 1767 /* anonportalgo RFC6056 subtree */ 1768 const struct sysctlnode *portalgo_node; 1769 sysctl_createv(clog, 0, NULL, &portalgo_node, 1770 CTLFLAG_PERMANENT, 1771 CTLTYPE_NODE, "anonportalgo", 1772 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1773 NULL, 0, NULL, 0, 1774 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1775 sysctl_createv(clog, 0, &portalgo_node, NULL, 1776 CTLFLAG_PERMANENT, 1777 CTLTYPE_STRING, "available", 1778 SYSCTL_DESCR("available algorithms"), 1779 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1780 CTL_CREATE, CTL_EOL); 1781 sysctl_createv(clog, 0, &portalgo_node, NULL, 1782 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1783 CTLTYPE_STRING, "selected", 1784 SYSCTL_DESCR("selected algorithm"), 1785 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1786 CTL_CREATE, CTL_EOL); 1787 sysctl_createv(clog, 0, &portalgo_node, NULL, 1788 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1789 CTLTYPE_STRUCT, "reserve", 1790 SYSCTL_DESCR("bitmap of reserved ports"), 1791 sysctl_portalgo_reserve4, 0, NULL, 0, 1792 CTL_CREATE, CTL_EOL); 1793 } 1794 1795 void 1796 ip_statinc(u_int stat) 1797 { 1798 1799 KASSERT(stat < IP_NSTATS); 1800 IP_STATINC(stat); 1801 } 1802