1 /* $NetBSD: ip_input.c,v 1.339 2016/08/01 03:15:30 ozaki-r Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 59 * POSSIBILITY OF SUCH DAMAGE. 60 */ 61 62 /* 63 * Copyright (c) 1982, 1986, 1988, 1993 64 * The Regents of the University of California. All rights reserved. 65 * 66 * Redistribution and use in source and binary forms, with or without 67 * modification, are permitted provided that the following conditions 68 * are met: 69 * 1. Redistributions of source code must retain the above copyright 70 * notice, this list of conditions and the following disclaimer. 71 * 2. Redistributions in binary form must reproduce the above copyright 72 * notice, this list of conditions and the following disclaimer in the 73 * documentation and/or other materials provided with the distribution. 74 * 3. Neither the name of the University nor the names of its contributors 75 * may be used to endorse or promote products derived from this software 76 * without specific prior written permission. 77 * 78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 88 * SUCH DAMAGE. 89 * 90 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 91 */ 92 93 #include <sys/cdefs.h> 94 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.339 2016/08/01 03:15:30 ozaki-r Exp $"); 95 96 #ifdef _KERNEL_OPT 97 #include "opt_inet.h" 98 #include "opt_compat_netbsd.h" 99 #include "opt_gateway.h" 100 #include "opt_ipsec.h" 101 #include "opt_mrouting.h" 102 #include "opt_mbuftrace.h" 103 #include "opt_inet_csum.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 /* GATEWAY */ 157 #define IPFORWARDING 0 /* don't forward IP packets not for us */ 158 #endif /* GATEWAY */ 159 #endif /* IPFORWARDING */ 160 #ifndef IPSENDREDIRECTS 161 #define IPSENDREDIRECTS 1 162 #endif 163 #ifndef IPFORWSRCRT 164 #define IPFORWSRCRT 1 /* forward source-routed packets */ 165 #endif 166 #ifndef IPALLOWSRCRT 167 #define IPALLOWSRCRT 1 /* allow source-routed packets */ 168 #endif 169 #ifndef IPMTUDISC 170 #define IPMTUDISC 1 171 #endif 172 #ifndef IPMTUDISCTIMEOUT 173 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 174 #endif 175 176 #ifdef COMPAT_50 177 #include <compat/sys/time.h> 178 #include <compat/sys/socket.h> 179 #endif 180 181 /* 182 * Note: DIRECTED_BROADCAST is handled this way so that previous 183 * configuration using this option will Just Work. 184 */ 185 #ifndef IPDIRECTEDBCAST 186 #ifdef DIRECTED_BROADCAST 187 #define IPDIRECTEDBCAST 1 188 #else 189 #define IPDIRECTEDBCAST 0 190 #endif /* DIRECTED_BROADCAST */ 191 #endif /* IPDIRECTEDBCAST */ 192 int ipforwarding = IPFORWARDING; 193 int ipsendredirects = IPSENDREDIRECTS; 194 int ip_defttl = IPDEFTTL; 195 int ip_forwsrcrt = IPFORWSRCRT; 196 int ip_directedbcast = IPDIRECTEDBCAST; 197 int ip_allowsrcrt = IPALLOWSRCRT; 198 int ip_mtudisc = IPMTUDISC; 199 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 200 #ifdef DIAGNOSTIC 201 int ipprintfs = 0; 202 #endif 203 204 int ip_do_randomid = 0; 205 206 /* 207 * XXX - Setting ip_checkinterface mostly implements the receive side of 208 * the Strong ES model described in RFC 1122, but since the routing table 209 * and transmit implementation do not implement the Strong ES model, 210 * setting this to 1 results in an odd hybrid. 211 * 212 * XXX - ip_checkinterface currently must be disabled if you use ipnat 213 * to translate the destination address to another local interface. 214 * 215 * XXX - ip_checkinterface must be disabled if you add IP aliases 216 * to the loopback interface instead of the interface where the 217 * packets for those addresses are received. 218 */ 219 static int ip_checkinterface __read_mostly = 0; 220 221 struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 222 223 pktqueue_t * ip_pktq __read_mostly; 224 pfil_head_t * inet_pfil_hook __read_mostly; 225 ipid_state_t * ip_ids __read_mostly; 226 percpu_t * ipstat_percpu __read_mostly; 227 228 static struct route ipforward_rt __cacheline_aligned; 229 230 uint16_t ip_id; 231 232 #ifdef INET_CSUM_COUNTERS 233 #include <sys/device.h> 234 235 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 236 NULL, "inet", "hwcsum bad"); 237 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 238 NULL, "inet", "hwcsum ok"); 239 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 240 NULL, "inet", "swcsum"); 241 242 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 243 244 EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 245 EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 246 EVCNT_ATTACH_STATIC(ip_swcsum); 247 248 #else 249 250 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 251 252 #endif /* INET_CSUM_COUNTERS */ 253 254 /* 255 * We need to save the IP options in case a protocol wants to respond 256 * to an incoming packet over the same route if the packet got here 257 * using IP source routing. This allows connection establishment and 258 * maintenance when the remote end is on a network that is not known 259 * to us. 260 */ 261 262 static int ip_nhops = 0; 263 264 static struct ip_srcrt { 265 struct in_addr dst; /* final destination */ 266 char nop; /* one NOP to align */ 267 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 268 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 269 } ip_srcrt; 270 271 static int ip_drainwanted; 272 273 struct sockaddr_in ipaddr = { 274 .sin_len = sizeof(ipaddr), 275 .sin_family = AF_INET, 276 }; 277 278 static void save_rte(u_char *, struct in_addr); 279 280 #ifdef MBUFTRACE 281 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 282 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 283 #endif 284 285 static void ipintr(void *); 286 static void ip_input(struct mbuf *); 287 static void ip_forward(struct mbuf *, int, struct ifnet *); 288 static bool ip_dooptions(struct mbuf *); 289 static struct in_ifaddr *ip_rtaddr(struct in_addr); 290 static void sysctl_net_inet_ip_setup(struct sysctllog **); 291 292 static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *, 293 int *); 294 static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *, 295 struct ip *); 296 297 /* XXX: Not yet enabled. */ 298 #define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 299 #define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 300 301 /* 302 * IP initialization: fill in IP protocol switch table. 303 * All protocols not implemented in kernel go to raw IP protocol handler. 304 */ 305 void 306 ip_init(void) 307 { 308 const struct protosw *pr; 309 310 in_init(); 311 sysctl_net_inet_ip_setup(NULL); 312 313 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 314 KASSERT(pr != NULL); 315 316 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 317 KASSERT(ip_pktq != NULL); 318 319 for (u_int i = 0; i < IPPROTO_MAX; i++) { 320 ip_protox[i] = pr - inetsw; 321 } 322 for (pr = inetdomain.dom_protosw; 323 pr < inetdomain.dom_protoswNPROTOSW; pr++) 324 if (pr->pr_domain->dom_family == PF_INET && 325 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 326 ip_protox[pr->pr_protocol] = pr - inetsw; 327 328 ip_reass_init(); 329 330 ip_ids = ip_id_init(); 331 ip_id = time_uptime & 0xfffff; 332 333 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 334 #ifdef GATEWAY 335 ipflow_init(); 336 #endif 337 338 /* Register our Packet Filter hook. */ 339 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 340 KASSERT(inet_pfil_hook != NULL); 341 342 #ifdef MBUFTRACE 343 MOWNER_ATTACH(&ip_tx_mowner); 344 MOWNER_ATTACH(&ip_rx_mowner); 345 #endif /* MBUFTRACE */ 346 347 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 348 } 349 350 static struct in_ifaddr * 351 ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 352 { 353 struct in_ifaddr *ia = NULL; 354 int checkif; 355 356 /* 357 * Enable a consistency check between the destination address 358 * and the arrival interface for a unicast packet (the RFC 1122 359 * strong ES model) if IP forwarding is disabled and the packet 360 * is not locally generated. 361 * 362 * XXX - Checking also should be disabled if the destination 363 * address is ipnat'ed to a different interface. 364 * 365 * XXX - Checking is incompatible with IP aliases added 366 * to the loopback interface instead of the interface where 367 * the packets are received. 368 * 369 * XXX - We need to add a per ifaddr flag for this so that 370 * we get finer grain control. 371 */ 372 checkif = ip_checkinterface && (ipforwarding == 0) && 373 (ifp->if_flags & IFF_LOOPBACK) == 0; 374 375 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 376 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 377 if (ia->ia4_flags & IN_IFF_NOTREADY) 378 continue; 379 if (checkif && ia->ia_ifp != ifp) 380 continue; 381 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 382 break; 383 else 384 (*downmatch)++; 385 } 386 } 387 388 return ia; 389 } 390 391 static struct in_ifaddr * 392 ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 393 { 394 struct in_ifaddr *ia = NULL; 395 struct ifaddr *ifa; 396 397 IFADDR_READER_FOREACH(ifa, ifp) { 398 if (ifa->ifa_addr->sa_family != AF_INET) 399 continue; 400 ia = ifatoia(ifa); 401 if (ia->ia4_flags & IN_IFF_NOTREADY) 402 continue; 403 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 404 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 405 /* 406 * Look for all-0's host part (old broadcast addr), 407 * either for subnet or net. 408 */ 409 ip->ip_dst.s_addr == ia->ia_subnet || 410 ip->ip_dst.s_addr == ia->ia_net) 411 goto matched; 412 /* 413 * An interface with IP address zero accepts 414 * all packets that arrive on that interface. 415 */ 416 if (in_nullhost(ia->ia_addr.sin_addr)) 417 goto matched; 418 } 419 ia = NULL; 420 421 matched: 422 return ia; 423 } 424 425 /* 426 * IP software interrupt routine. 427 */ 428 static void 429 ipintr(void *arg __unused) 430 { 431 struct mbuf *m; 432 433 KASSERT(cpu_softintr_p()); 434 435 mutex_enter(softnet_lock); 436 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 437 ip_input(m); 438 } 439 mutex_exit(softnet_lock); 440 } 441 442 /* 443 * IP input routine. Checksum and byte swap header. If fragmented 444 * try to reassemble. Process options. Pass to next level. 445 */ 446 static void 447 ip_input(struct mbuf *m) 448 { 449 struct ip *ip = NULL; 450 struct in_ifaddr *ia = NULL; 451 int hlen = 0, len; 452 int downmatch; 453 int srcrt = 0; 454 ifnet_t *ifp; 455 struct psref psref; 456 int s; 457 458 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 459 "interrupt handler; synchronization assumptions violated"); 460 461 MCLAIM(m, &ip_rx_mowner); 462 KASSERT((m->m_flags & M_PKTHDR) != 0); 463 464 ifp = m_get_rcvif_psref(m, &psref); 465 if (__predict_false(ifp == NULL)) 466 goto out; 467 468 /* 469 * If no IP addresses have been set yet but the interfaces 470 * are receiving, can't do anything with incoming packets yet. 471 * Note: we pre-check without locks held. 472 */ 473 if (IN_ADDRLIST_READER_EMPTY()) 474 goto out; 475 IP_STATINC(IP_STAT_TOTAL); 476 477 /* 478 * If the IP header is not aligned, slurp it up into a new 479 * mbuf with space for link headers, in the event we forward 480 * it. Otherwise, if it is aligned, make sure the entire 481 * base IP header is in the first mbuf of the chain. 482 */ 483 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 484 if ((m = m_copyup(m, sizeof(struct ip), 485 (max_linkhdr + 3) & ~3)) == NULL) { 486 /* XXXJRT new stat, please */ 487 IP_STATINC(IP_STAT_TOOSMALL); 488 goto out; 489 } 490 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 491 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 492 IP_STATINC(IP_STAT_TOOSMALL); 493 goto out; 494 } 495 } 496 ip = mtod(m, struct ip *); 497 if (ip->ip_v != IPVERSION) { 498 IP_STATINC(IP_STAT_BADVERS); 499 goto out; 500 } 501 hlen = ip->ip_hl << 2; 502 if (hlen < sizeof(struct ip)) { /* minimum header length */ 503 IP_STATINC(IP_STAT_BADHLEN); 504 goto out; 505 } 506 if (hlen > m->m_len) { 507 if ((m = m_pullup(m, hlen)) == NULL) { 508 IP_STATINC(IP_STAT_BADHLEN); 509 goto out; 510 } 511 ip = mtod(m, struct ip *); 512 } 513 514 /* 515 * RFC1122: packets with a multicast source address are 516 * not allowed. 517 */ 518 if (IN_MULTICAST(ip->ip_src.s_addr)) { 519 IP_STATINC(IP_STAT_BADADDR); 520 goto out; 521 } 522 523 /* 127/8 must not appear on wire - RFC1122 */ 524 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 525 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 526 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 527 IP_STATINC(IP_STAT_BADADDR); 528 goto out; 529 } 530 } 531 532 switch (m->m_pkthdr.csum_flags & 533 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | 534 M_CSUM_IPv4_BAD)) { 535 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 536 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 537 IP_STATINC(IP_STAT_BADSUM); 538 goto out; 539 540 case M_CSUM_IPv4: 541 /* Checksum was okay. */ 542 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 543 break; 544 545 default: 546 /* 547 * Must compute it ourselves. Maybe skip checksum on 548 * loopback interfaces. 549 */ 550 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 551 ip_do_loopback_cksum)) { 552 INET_CSUM_COUNTER_INCR(&ip_swcsum); 553 if (in_cksum(m, hlen) != 0) { 554 IP_STATINC(IP_STAT_BADSUM); 555 goto out; 556 } 557 } 558 break; 559 } 560 561 /* Retrieve the packet length. */ 562 len = ntohs(ip->ip_len); 563 564 /* 565 * Check for additional length bogosity 566 */ 567 if (len < hlen) { 568 IP_STATINC(IP_STAT_BADLEN); 569 goto out; 570 } 571 572 /* 573 * Check that the amount of data in the buffers 574 * is as at least much as the IP header would have us expect. 575 * Trim mbufs if longer than we expect. 576 * Drop packet if shorter than we expect. 577 */ 578 if (m->m_pkthdr.len < len) { 579 IP_STATINC(IP_STAT_TOOSHORT); 580 goto out; 581 } 582 if (m->m_pkthdr.len > len) { 583 if (m->m_len == m->m_pkthdr.len) { 584 m->m_len = len; 585 m->m_pkthdr.len = len; 586 } else 587 m_adj(m, len - m->m_pkthdr.len); 588 } 589 590 /* 591 * Assume that we can create a fast-forward IP flow entry 592 * based on this packet. 593 */ 594 m->m_flags |= M_CANFASTFWD; 595 596 /* 597 * Run through list of hooks for input packets. If there are any 598 * filters which require that additional packets in the flow are 599 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 600 * Note that filters must _never_ set this flag, as another filter 601 * in the list may have previously cleared it. 602 */ 603 #if defined(IPSEC) 604 if (!ipsec_used || !ipsec_indone(m)) 605 #else 606 if (1) 607 #endif 608 { 609 struct in_addr odst = ip->ip_dst; 610 bool freed; 611 612 SOFTNET_LOCK(); 613 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 614 SOFTNET_UNLOCK(); 615 if (freed || m == NULL) { 616 goto out; 617 } 618 ip = mtod(m, struct ip *); 619 hlen = ip->ip_hl << 2; 620 621 /* 622 * XXX The setting of "srcrt" here is to prevent ip_forward() 623 * from generating ICMP redirects for packets that have 624 * been redirected by a hook back out on to the same LAN that 625 * they came from and is not an indication that the packet 626 * is being inffluenced by source routing options. This 627 * allows things like 628 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 629 * where tlp0 is both on the 1.1.1.0/24 network and is the 630 * default route for hosts on 1.1.1.0/24. Of course this 631 * also requires a "map tlp0 ..." to complete the story. 632 * One might argue whether or not this kind of network config. 633 * should be supported in this manner... 634 */ 635 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 636 } 637 638 #ifdef ALTQ 639 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 640 if (altq_input) { 641 SOFTNET_LOCK(); 642 if ((*altq_input)(m, AF_INET) == 0) { 643 /* Packet dropped by traffic conditioner. */ 644 SOFTNET_UNLOCK(); 645 goto out; 646 } 647 SOFTNET_UNLOCK(); 648 } 649 #endif 650 651 /* 652 * Process options and, if not destined for us, 653 * ship it on. ip_dooptions returns 1 when an 654 * error was detected (causing an icmp message 655 * to be sent and the original packet to be freed). 656 */ 657 ip_nhops = 0; /* for source routed packets */ 658 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 659 goto out; 660 661 /* 662 * Check our list of addresses, to see if the packet is for us. 663 * 664 * Traditional 4.4BSD did not consult IFF_UP at all. 665 * The behavior here is to treat addresses on !IFF_UP interface 666 * or IN_IFF_NOTREADY addresses as not mine. 667 */ 668 downmatch = 0; 669 s = pserialize_read_enter(); 670 ia = ip_match_our_address(ifp, ip, &downmatch); 671 if (ia != NULL) { 672 pserialize_read_exit(s); 673 goto ours; 674 } 675 676 if (ifp->if_flags & IFF_BROADCAST) { 677 ia = ip_match_our_address_broadcast(ifp, ip); 678 if (ia != NULL) { 679 pserialize_read_exit(s); 680 goto ours; 681 } 682 } 683 pserialize_read_exit(s); 684 685 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 686 #ifdef MROUTING 687 extern struct socket *ip_mrouter; 688 689 if (ip_mrouter) { 690 /* 691 * If we are acting as a multicast router, all 692 * incoming multicast packets are passed to the 693 * kernel-level multicast forwarding function. 694 * The packet is returned (relatively) intact; if 695 * ip_mforward() returns a non-zero value, the packet 696 * must be discarded, else it may be accepted below. 697 * 698 * (The IP ident field is put in the same byte order 699 * as expected when ip_mforward() is called from 700 * ip_output().) 701 */ 702 SOFTNET_LOCK(); 703 if (ip_mforward(m, ifp) != 0) { 704 SOFTNET_UNLOCK(); 705 IP_STATINC(IP_STAT_CANTFORWARD); 706 goto out; 707 } 708 SOFTNET_UNLOCK(); 709 710 /* 711 * The process-level routing demon needs to receive 712 * all multicast IGMP packets, whether or not this 713 * host belongs to their destination groups. 714 */ 715 if (ip->ip_p == IPPROTO_IGMP) { 716 goto ours; 717 } 718 IP_STATINC(IP_STAT_CANTFORWARD); 719 } 720 #endif 721 /* 722 * See if we belong to the destination multicast group on the 723 * arrival interface. 724 */ 725 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 726 IP_STATINC(IP_STAT_CANTFORWARD); 727 goto out; 728 } 729 goto ours; 730 } 731 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 732 in_nullhost(ip->ip_dst)) 733 goto ours; 734 735 /* 736 * Not for us; forward if possible and desirable. 737 */ 738 if (ipforwarding == 0) { 739 m_put_rcvif_psref(ifp, &psref); 740 IP_STATINC(IP_STAT_CANTFORWARD); 741 m_freem(m); 742 } else { 743 /* 744 * If ip_dst matched any of my address on !IFF_UP interface, 745 * and there's no IFF_UP interface that matches ip_dst, 746 * send icmp unreach. Forwarding it will result in in-kernel 747 * forwarding loop till TTL goes to 0. 748 */ 749 if (downmatch) { 750 m_put_rcvif_psref(ifp, &psref); 751 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 752 IP_STATINC(IP_STAT_CANTFORWARD); 753 return; 754 } 755 #ifdef IPSEC 756 /* Perform IPsec, if any. */ 757 if (ipsec_used) { 758 SOFTNET_LOCK(); 759 if (ipsec4_input(m, IP_FORWARDING | 760 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 761 SOFTNET_UNLOCK(); 762 goto out; 763 } 764 SOFTNET_UNLOCK(); 765 } 766 #endif 767 ip_forward(m, srcrt, ifp); 768 m_put_rcvif_psref(ifp, &psref); 769 } 770 return; 771 772 ours: 773 m_put_rcvif_psref(ifp, &psref); 774 ifp = NULL; 775 776 /* 777 * If offset or IP_MF are set, must reassemble. 778 */ 779 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 780 /* 781 * Pass to IP reassembly mechanism. 782 */ 783 if (ip_reass_packet(&m, ip) != 0) { 784 /* Failed; invalid fragment(s) or packet. */ 785 goto out; 786 } 787 if (m == NULL) { 788 /* More fragments should come; silently return. */ 789 goto out; 790 } 791 /* 792 * Reassembly is done, we have the final packet. 793 * Updated cached data in local variable(s). 794 */ 795 ip = mtod(m, struct ip *); 796 hlen = ip->ip_hl << 2; 797 } 798 799 #ifdef IPSEC 800 /* 801 * Enforce IPsec policy checking if we are seeing last header. 802 * Note that we do not visit this with protocols with PCB layer 803 * code - like UDP/TCP/raw IP. 804 */ 805 if (ipsec_used && 806 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 807 SOFTNET_LOCK(); 808 if (ipsec4_input(m, 0) != 0) { 809 SOFTNET_UNLOCK(); 810 goto out; 811 } 812 SOFTNET_UNLOCK(); 813 } 814 #endif 815 816 /* 817 * Switch out to protocol's input routine. 818 */ 819 #if IFA_STATS 820 if (ia && ip) { 821 struct in_ifaddr *_ia; 822 /* 823 * Keep a reference from ip_match_our_address with psref 824 * is expensive, so explore ia here again. 825 */ 826 s = pserialize_read_enter(); 827 _ia = in_get_ia(ip->ip_dst.s_addr); 828 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 829 pserialize_read_exit(s); 830 } 831 #endif 832 IP_STATINC(IP_STAT_DELIVERED); 833 834 const int off = hlen, nh = ip->ip_p; 835 836 SOFTNET_LOCK(); 837 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 838 SOFTNET_UNLOCK(); 839 return; 840 841 out: 842 m_put_rcvif_psref(ifp, &psref); 843 if (m != NULL) 844 m_freem(m); 845 } 846 847 /* 848 * IP timer processing. 849 */ 850 void 851 ip_slowtimo(void) 852 { 853 854 mutex_enter(softnet_lock); 855 KERNEL_LOCK(1, NULL); 856 857 ip_reass_slowtimo(); 858 859 KERNEL_UNLOCK_ONE(NULL); 860 mutex_exit(softnet_lock); 861 } 862 863 /* 864 * IP drain processing. 865 */ 866 void 867 ip_drain(void) 868 { 869 870 KERNEL_LOCK(1, NULL); 871 ip_reass_drain(); 872 KERNEL_UNLOCK_ONE(NULL); 873 } 874 875 /* 876 * ip_dooptions: perform option processing on a datagram, possibly discarding 877 * it if bad options are encountered, or forwarding it if source-routed. 878 * 879 * => Returns true if packet has been forwarded/freed. 880 * => Returns false if the packet should be processed further. 881 */ 882 static bool 883 ip_dooptions(struct mbuf *m) 884 { 885 struct ip *ip = mtod(m, struct ip *); 886 u_char *cp, *cp0; 887 struct ip_timestamp *ipt; 888 struct in_ifaddr *ia; 889 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 890 struct in_addr dst; 891 n_time ntime; 892 struct ifaddr *ifa; 893 int s; 894 895 dst = ip->ip_dst; 896 cp = (u_char *)(ip + 1); 897 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 898 for (; cnt > 0; cnt -= optlen, cp += optlen) { 899 opt = cp[IPOPT_OPTVAL]; 900 if (opt == IPOPT_EOL) 901 break; 902 if (opt == IPOPT_NOP) 903 optlen = 1; 904 else { 905 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 906 code = &cp[IPOPT_OLEN] - (u_char *)ip; 907 goto bad; 908 } 909 optlen = cp[IPOPT_OLEN]; 910 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 911 code = &cp[IPOPT_OLEN] - (u_char *)ip; 912 goto bad; 913 } 914 } 915 switch (opt) { 916 917 default: 918 break; 919 920 /* 921 * Source routing with record. 922 * Find interface with current destination address. 923 * If none on this machine then drop if strictly routed, 924 * or do nothing if loosely routed. 925 * Record interface address and bring up next address 926 * component. If strictly routed make sure next 927 * address is on directly accessible net. 928 */ 929 case IPOPT_LSRR: 930 case IPOPT_SSRR: 931 if (ip_allowsrcrt == 0) { 932 type = ICMP_UNREACH; 933 code = ICMP_UNREACH_NET_PROHIB; 934 goto bad; 935 } 936 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 937 code = &cp[IPOPT_OLEN] - (u_char *)ip; 938 goto bad; 939 } 940 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 941 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 942 goto bad; 943 } 944 ipaddr.sin_addr = ip->ip_dst; 945 946 s = pserialize_read_enter(); 947 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 948 if (ifa == NULL) { 949 pserialize_read_exit(s); 950 if (opt == IPOPT_SSRR) { 951 type = ICMP_UNREACH; 952 code = ICMP_UNREACH_SRCFAIL; 953 goto bad; 954 } 955 /* 956 * Loose routing, and not at next destination 957 * yet; nothing to do except forward. 958 */ 959 break; 960 } 961 pserialize_read_exit(s); 962 963 off--; /* 0 origin */ 964 if ((off + sizeof(struct in_addr)) > optlen) { 965 /* 966 * End of source route. Should be for us. 967 */ 968 save_rte(cp, ip->ip_src); 969 break; 970 } 971 /* 972 * locate outgoing interface 973 */ 974 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 975 sizeof(ipaddr.sin_addr)); 976 s = pserialize_read_enter(); 977 if (opt == IPOPT_SSRR) { 978 ifa = ifa_ifwithladdr(sintosa(&ipaddr)); 979 if (ifa != NULL) 980 ia = ifatoia(ifa); 981 else 982 ia = NULL; 983 } else { 984 ia = ip_rtaddr(ipaddr.sin_addr); 985 } 986 if (ia == NULL) { 987 type = ICMP_UNREACH; 988 code = ICMP_UNREACH_SRCFAIL; 989 pserialize_read_exit(s); 990 goto bad; 991 } 992 ip->ip_dst = ipaddr.sin_addr; 993 bcopy((void *)&ia->ia_addr.sin_addr, 994 (void *)(cp + off), sizeof(struct in_addr)); 995 pserialize_read_exit(s); 996 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 997 /* 998 * Let ip_intr's mcast routing check handle mcast pkts 999 */ 1000 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1001 break; 1002 1003 case IPOPT_RR: 1004 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1005 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1006 goto bad; 1007 } 1008 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1009 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1010 goto bad; 1011 } 1012 /* 1013 * If no space remains, ignore. 1014 */ 1015 off--; /* 0 origin */ 1016 if ((off + sizeof(struct in_addr)) > optlen) 1017 break; 1018 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1019 sizeof(ipaddr.sin_addr)); 1020 /* 1021 * locate outgoing interface; if we're the destination, 1022 * use the incoming interface (should be same). 1023 */ 1024 s = pserialize_read_enter(); 1025 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1026 if (ifa == NULL) { 1027 ia = ip_rtaddr(ipaddr.sin_addr); 1028 if (ia == NULL) { 1029 pserialize_read_exit(s); 1030 type = ICMP_UNREACH; 1031 code = ICMP_UNREACH_HOST; 1032 goto bad; 1033 } 1034 } else { 1035 ia = ifatoia(ifa); 1036 } 1037 bcopy((void *)&ia->ia_addr.sin_addr, 1038 (void *)(cp + off), sizeof(struct in_addr)); 1039 pserialize_read_exit(s); 1040 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1041 break; 1042 1043 case IPOPT_TS: 1044 code = cp - (u_char *)ip; 1045 ipt = (struct ip_timestamp *)cp; 1046 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1047 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1048 goto bad; 1049 } 1050 if (ipt->ipt_ptr < 5) { 1051 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1052 goto bad; 1053 } 1054 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1055 if (++ipt->ipt_oflw == 0) { 1056 code = (u_char *)&ipt->ipt_ptr - 1057 (u_char *)ip; 1058 goto bad; 1059 } 1060 break; 1061 } 1062 cp0 = (cp + ipt->ipt_ptr - 1); 1063 switch (ipt->ipt_flg) { 1064 1065 case IPOPT_TS_TSONLY: 1066 break; 1067 1068 case IPOPT_TS_TSANDADDR: { 1069 struct ifnet *rcvif; 1070 int _s, _ss; 1071 1072 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1073 sizeof(struct in_addr) > ipt->ipt_len) { 1074 code = (u_char *)&ipt->ipt_ptr - 1075 (u_char *)ip; 1076 goto bad; 1077 } 1078 ipaddr.sin_addr = dst; 1079 _ss = pserialize_read_enter(); 1080 rcvif = m_get_rcvif(m, &_s); 1081 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), rcvif); 1082 m_put_rcvif(rcvif, &_s); 1083 if (ifa == NULL) { 1084 pserialize_read_exit(_ss); 1085 break; 1086 } 1087 ia = ifatoia(ifa); 1088 bcopy(&ia->ia_addr.sin_addr, 1089 cp0, sizeof(struct in_addr)); 1090 pserialize_read_exit(_ss); 1091 ipt->ipt_ptr += sizeof(struct in_addr); 1092 break; 1093 } 1094 1095 case IPOPT_TS_PRESPEC: 1096 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1097 sizeof(struct in_addr) > ipt->ipt_len) { 1098 code = (u_char *)&ipt->ipt_ptr - 1099 (u_char *)ip; 1100 goto bad; 1101 } 1102 memcpy(&ipaddr.sin_addr, cp0, 1103 sizeof(struct in_addr)); 1104 s = pserialize_read_enter(); 1105 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1106 if (ifa == NULL) { 1107 pserialize_read_exit(s); 1108 continue; 1109 } 1110 pserialize_read_exit(s); 1111 ipt->ipt_ptr += sizeof(struct in_addr); 1112 break; 1113 1114 default: 1115 /* XXX can't take &ipt->ipt_flg */ 1116 code = (u_char *)&ipt->ipt_ptr - 1117 (u_char *)ip + 1; 1118 goto bad; 1119 } 1120 ntime = iptime(); 1121 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1122 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1123 sizeof(n_time)); 1124 ipt->ipt_ptr += sizeof(n_time); 1125 } 1126 } 1127 if (forward) { 1128 struct ifnet *rcvif; 1129 struct psref _psref; 1130 1131 if (ip_forwsrcrt == 0) { 1132 type = ICMP_UNREACH; 1133 code = ICMP_UNREACH_SRCFAIL; 1134 goto bad; 1135 } 1136 1137 rcvif = m_get_rcvif_psref(m, &_psref); 1138 if (__predict_false(rcvif == NULL)) { 1139 type = ICMP_UNREACH; 1140 code = ICMP_UNREACH_HOST; 1141 goto bad; 1142 } 1143 ip_forward(m, 1, rcvif); 1144 m_put_rcvif_psref(rcvif, &_psref); 1145 return true; 1146 } 1147 return false; 1148 bad: 1149 icmp_error(m, type, code, 0, 0); 1150 IP_STATINC(IP_STAT_BADOPTIONS); 1151 return true; 1152 } 1153 1154 /* 1155 * ip_rtaddr: given address of next destination (final or next hop), 1156 * return internet address info of interface to be used to get there. 1157 */ 1158 static struct in_ifaddr * 1159 ip_rtaddr(struct in_addr dst) 1160 { 1161 struct rtentry *rt; 1162 union { 1163 struct sockaddr dst; 1164 struct sockaddr_in dst4; 1165 } u; 1166 1167 sockaddr_in_init(&u.dst4, &dst, 0); 1168 1169 SOFTNET_LOCK(); 1170 rt = rtcache_lookup(&ipforward_rt, &u.dst); 1171 SOFTNET_UNLOCK(); 1172 if (rt == NULL) 1173 return NULL; 1174 1175 return ifatoia(rt->rt_ifa); 1176 } 1177 1178 /* 1179 * save_rte: save incoming source route for use in replies, to be picked 1180 * up later by ip_srcroute if the receiver is interested. 1181 */ 1182 static void 1183 save_rte(u_char *option, struct in_addr dst) 1184 { 1185 unsigned olen; 1186 1187 olen = option[IPOPT_OLEN]; 1188 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1189 return; 1190 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen); 1191 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1192 ip_srcrt.dst = dst; 1193 } 1194 1195 /* 1196 * Retrieve incoming source route for use in replies, 1197 * in the same form used by setsockopt. 1198 * The first hop is placed before the options, will be removed later. 1199 */ 1200 struct mbuf * 1201 ip_srcroute(void) 1202 { 1203 struct in_addr *p, *q; 1204 struct mbuf *m; 1205 1206 if (ip_nhops == 0) 1207 return NULL; 1208 m = m_get(M_DONTWAIT, MT_SOOPTS); 1209 if (m == 0) 1210 return NULL; 1211 1212 MCLAIM(m, &inetdomain.dom_mowner); 1213 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1214 1215 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1216 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1217 OPTSIZ; 1218 1219 /* 1220 * First save first hop for return route 1221 */ 1222 p = &ip_srcrt.route[ip_nhops - 1]; 1223 *(mtod(m, struct in_addr *)) = *p--; 1224 1225 /* 1226 * Copy option fields and padding (nop) to mbuf. 1227 */ 1228 ip_srcrt.nop = IPOPT_NOP; 1229 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1230 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1231 OPTSIZ); 1232 q = (struct in_addr *)(mtod(m, char *) + 1233 sizeof(struct in_addr) + OPTSIZ); 1234 #undef OPTSIZ 1235 /* 1236 * Record return path as an IP source route, 1237 * reversing the path (pointers are now aligned). 1238 */ 1239 while (p >= ip_srcrt.route) { 1240 *q++ = *p--; 1241 } 1242 /* 1243 * Last hop goes to final destination. 1244 */ 1245 *q = ip_srcrt.dst; 1246 return (m); 1247 } 1248 1249 const int inetctlerrmap[PRC_NCMDS] = { 1250 [PRC_MSGSIZE] = EMSGSIZE, 1251 [PRC_HOSTDEAD] = EHOSTDOWN, 1252 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1253 [PRC_UNREACH_NET] = EHOSTUNREACH, 1254 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1255 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1256 [PRC_UNREACH_PORT] = ECONNREFUSED, 1257 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1258 [PRC_PARAMPROB] = ENOPROTOOPT, 1259 }; 1260 1261 void 1262 ip_fasttimo(void) 1263 { 1264 if (ip_drainwanted) { 1265 ip_drain(); 1266 ip_drainwanted = 0; 1267 } 1268 } 1269 1270 void 1271 ip_drainstub(void) 1272 { 1273 ip_drainwanted = 1; 1274 } 1275 1276 /* 1277 * Forward a packet. If some error occurs return the sender 1278 * an icmp packet. Note we can't always generate a meaningful 1279 * icmp message because icmp doesn't have a large enough repertoire 1280 * of codes and types. 1281 * 1282 * If not forwarding, just drop the packet. This could be confusing 1283 * if ipforwarding was zero but some routing protocol was advancing 1284 * us as a gateway to somewhere. However, we must let the routing 1285 * protocol deal with that. 1286 * 1287 * The srcrt parameter indicates whether the packet is being forwarded 1288 * via a source route. 1289 */ 1290 static void 1291 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1292 { 1293 struct ip *ip = mtod(m, struct ip *); 1294 struct rtentry *rt; 1295 int error, type = 0, code = 0, destmtu = 0; 1296 struct mbuf *mcopy; 1297 n_long dest; 1298 union { 1299 struct sockaddr dst; 1300 struct sockaddr_in dst4; 1301 } u; 1302 uint64_t *ips; 1303 1304 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1305 "interrupt handler; synchronization assumptions violated"); 1306 1307 /* 1308 * We are now in the output path. 1309 */ 1310 MCLAIM(m, &ip_tx_mowner); 1311 1312 /* 1313 * Clear any in-bound checksum flags for this packet. 1314 */ 1315 m->m_pkthdr.csum_flags = 0; 1316 1317 dest = 0; 1318 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1319 IP_STATINC(IP_STAT_CANTFORWARD); 1320 m_freem(m); 1321 return; 1322 } 1323 1324 SOFTNET_LOCK(); 1325 1326 if (ip->ip_ttl <= IPTTLDEC) { 1327 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1328 SOFTNET_UNLOCK(); 1329 return; 1330 } 1331 1332 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1333 1334 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1335 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1336 SOFTNET_UNLOCK(); 1337 return; 1338 } 1339 1340 /* 1341 * Save at most 68 bytes of the packet in case 1342 * we need to generate an ICMP message to the src. 1343 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1344 */ 1345 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1346 if (mcopy) 1347 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1348 1349 ip->ip_ttl -= IPTTLDEC; 1350 1351 /* 1352 * If forwarding packet using same interface that it came in on, 1353 * perhaps should send a redirect to sender to shortcut a hop. 1354 * Only send redirect if source is sending directly to us, 1355 * and if packet was not source routed (or has any options). 1356 * Also, don't send redirect if forwarding using a default route 1357 * or a route modified by a redirect. 1358 */ 1359 if (rt->rt_ifp == rcvif && 1360 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1361 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1362 ipsendredirects && !srcrt) { 1363 if (rt->rt_ifa && 1364 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1365 ifatoia(rt->rt_ifa)->ia_subnet) { 1366 if (rt->rt_flags & RTF_GATEWAY) 1367 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1368 else 1369 dest = ip->ip_dst.s_addr; 1370 /* 1371 * Router requirements says to only send host 1372 * redirects. 1373 */ 1374 type = ICMP_REDIRECT; 1375 code = ICMP_REDIRECT_HOST; 1376 } 1377 } 1378 1379 error = ip_output(m, NULL, &ipforward_rt, 1380 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1381 NULL, NULL); 1382 1383 if (error) { 1384 IP_STATINC(IP_STAT_CANTFORWARD); 1385 goto error; 1386 } 1387 1388 ips = IP_STAT_GETREF(); 1389 ips[IP_STAT_FORWARD]++; 1390 1391 if (type) { 1392 ips[IP_STAT_REDIRECTSENT]++; 1393 IP_STAT_PUTREF(); 1394 goto redirect; 1395 } 1396 1397 IP_STAT_PUTREF(); 1398 if (mcopy) { 1399 #ifdef GATEWAY 1400 if (mcopy->m_flags & M_CANFASTFWD) 1401 ipflow_create(&ipforward_rt, mcopy); 1402 #endif 1403 m_freem(mcopy); 1404 } 1405 1406 SOFTNET_UNLOCK(); 1407 return; 1408 1409 redirect: 1410 error: 1411 if (mcopy == NULL) { 1412 SOFTNET_UNLOCK(); 1413 return; 1414 } 1415 1416 switch (error) { 1417 1418 case 0: /* forwarded, but need redirect */ 1419 /* type, code set above */ 1420 break; 1421 1422 case ENETUNREACH: /* shouldn't happen, checked above */ 1423 case EHOSTUNREACH: 1424 case ENETDOWN: 1425 case EHOSTDOWN: 1426 default: 1427 type = ICMP_UNREACH; 1428 code = ICMP_UNREACH_HOST; 1429 break; 1430 1431 case EMSGSIZE: 1432 type = ICMP_UNREACH; 1433 code = ICMP_UNREACH_NEEDFRAG; 1434 1435 if ((rt = rtcache_validate(&ipforward_rt)) != NULL) 1436 destmtu = rt->rt_ifp->if_mtu; 1437 #ifdef IPSEC 1438 if (ipsec_used) 1439 (void)ipsec4_forward(mcopy, &destmtu); 1440 #endif 1441 IP_STATINC(IP_STAT_CANTFRAG); 1442 break; 1443 1444 case ENOBUFS: 1445 /* 1446 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1447 * Requirements for IP Version 4 Routers. Source quench can 1448 * big problem under DoS attacks or if the underlying 1449 * interface is rate-limited. 1450 */ 1451 if (mcopy) 1452 m_freem(mcopy); 1453 SOFTNET_UNLOCK(); 1454 return; 1455 } 1456 icmp_error(mcopy, type, code, dest, destmtu); 1457 SOFTNET_UNLOCK(); 1458 } 1459 1460 void 1461 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1462 struct mbuf *m) 1463 { 1464 struct socket *so = inp->inp_socket; 1465 ifnet_t *ifp; 1466 int inpflags = inp->inp_flags; 1467 struct psref psref; 1468 1469 ifp = m_get_rcvif_psref(m, &psref); 1470 if (__predict_false(ifp == NULL)) 1471 return; /* XXX should report error? */ 1472 1473 if (so->so_options & SO_TIMESTAMP 1474 #ifdef SO_OTIMESTAMP 1475 || so->so_options & SO_OTIMESTAMP 1476 #endif 1477 ) { 1478 struct timeval tv; 1479 1480 microtime(&tv); 1481 #ifdef SO_OTIMESTAMP 1482 if (so->so_options & SO_OTIMESTAMP) { 1483 struct timeval50 tv50; 1484 timeval_to_timeval50(&tv, &tv50); 1485 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50), 1486 SCM_OTIMESTAMP, SOL_SOCKET); 1487 } else 1488 #endif 1489 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 1490 SCM_TIMESTAMP, SOL_SOCKET); 1491 if (*mp) 1492 mp = &(*mp)->m_next; 1493 } 1494 if (inpflags & INP_RECVDSTADDR) { 1495 *mp = sbcreatecontrol((void *) &ip->ip_dst, 1496 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1497 if (*mp) 1498 mp = &(*mp)->m_next; 1499 } 1500 if (inpflags & INP_RECVPKTINFO) { 1501 struct in_pktinfo ipi; 1502 ipi.ipi_addr = ip->ip_src; 1503 ipi.ipi_ifindex = ifp->if_index; 1504 *mp = sbcreatecontrol((void *) &ipi, 1505 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP); 1506 if (*mp) 1507 mp = &(*mp)->m_next; 1508 } 1509 if (inpflags & INP_PKTINFO) { 1510 struct in_pktinfo ipi; 1511 ipi.ipi_addr = ip->ip_dst; 1512 ipi.ipi_ifindex = ifp->if_index; 1513 *mp = sbcreatecontrol((void *) &ipi, 1514 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1515 if (*mp) 1516 mp = &(*mp)->m_next; 1517 } 1518 if (inpflags & INP_RECVIF) { 1519 struct sockaddr_dl sdl; 1520 1521 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1522 NULL, 0); 1523 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1524 if (*mp) 1525 mp = &(*mp)->m_next; 1526 } 1527 if (inpflags & INP_RECVTTL) { 1528 *mp = sbcreatecontrol((void *) &ip->ip_ttl, 1529 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1530 if (*mp) 1531 mp = &(*mp)->m_next; 1532 } 1533 m_put_rcvif_psref(ifp, &psref); 1534 } 1535 1536 /* 1537 * sysctl helper routine for net.inet.ip.forwsrcrt. 1538 */ 1539 static int 1540 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1541 { 1542 int error, tmp; 1543 struct sysctlnode node; 1544 1545 node = *rnode; 1546 tmp = ip_forwsrcrt; 1547 node.sysctl_data = &tmp; 1548 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1549 if (error || newp == NULL) 1550 return (error); 1551 1552 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1553 0, NULL, NULL, NULL); 1554 if (error) 1555 return (error); 1556 1557 ip_forwsrcrt = tmp; 1558 1559 return (0); 1560 } 1561 1562 /* 1563 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1564 * range of the new value and tweaks timers if it changes. 1565 */ 1566 static int 1567 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1568 { 1569 int error, tmp; 1570 struct sysctlnode node; 1571 1572 node = *rnode; 1573 tmp = ip_mtudisc_timeout; 1574 node.sysctl_data = &tmp; 1575 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1576 if (error || newp == NULL) 1577 return (error); 1578 if (tmp < 0) 1579 return (EINVAL); 1580 1581 mutex_enter(softnet_lock); 1582 1583 ip_mtudisc_timeout = tmp; 1584 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1585 1586 mutex_exit(softnet_lock); 1587 1588 return (0); 1589 } 1590 1591 static int 1592 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1593 { 1594 1595 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1596 } 1597 1598 static void 1599 sysctl_net_inet_ip_setup(struct sysctllog **clog) 1600 { 1601 sysctl_createv(clog, 0, NULL, NULL, 1602 CTLFLAG_PERMANENT, 1603 CTLTYPE_NODE, "inet", 1604 SYSCTL_DESCR("PF_INET related settings"), 1605 NULL, 0, NULL, 0, 1606 CTL_NET, PF_INET, CTL_EOL); 1607 sysctl_createv(clog, 0, NULL, NULL, 1608 CTLFLAG_PERMANENT, 1609 CTLTYPE_NODE, "ip", 1610 SYSCTL_DESCR("IPv4 related settings"), 1611 NULL, 0, NULL, 0, 1612 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1613 1614 sysctl_createv(clog, 0, NULL, NULL, 1615 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1616 CTLTYPE_INT, "forwarding", 1617 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1618 NULL, 0, &ipforwarding, 0, 1619 CTL_NET, PF_INET, IPPROTO_IP, 1620 IPCTL_FORWARDING, CTL_EOL); 1621 sysctl_createv(clog, 0, NULL, NULL, 1622 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1623 CTLTYPE_INT, "redirect", 1624 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1625 NULL, 0, &ipsendredirects, 0, 1626 CTL_NET, PF_INET, IPPROTO_IP, 1627 IPCTL_SENDREDIRECTS, CTL_EOL); 1628 sysctl_createv(clog, 0, NULL, NULL, 1629 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1630 CTLTYPE_INT, "ttl", 1631 SYSCTL_DESCR("Default TTL for an INET datagram"), 1632 NULL, 0, &ip_defttl, 0, 1633 CTL_NET, PF_INET, IPPROTO_IP, 1634 IPCTL_DEFTTL, CTL_EOL); 1635 #ifdef IPCTL_DEFMTU 1636 sysctl_createv(clog, 0, NULL, NULL, 1637 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1638 CTLTYPE_INT, "mtu", 1639 SYSCTL_DESCR("Default MTA for an INET route"), 1640 NULL, 0, &ip_mtu, 0, 1641 CTL_NET, PF_INET, IPPROTO_IP, 1642 IPCTL_DEFMTU, CTL_EOL); 1643 #endif /* IPCTL_DEFMTU */ 1644 sysctl_createv(clog, 0, NULL, NULL, 1645 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1646 CTLTYPE_INT, "forwsrcrt", 1647 SYSCTL_DESCR("Enable forwarding of source-routed " 1648 "datagrams"), 1649 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1650 CTL_NET, PF_INET, IPPROTO_IP, 1651 IPCTL_FORWSRCRT, CTL_EOL); 1652 sysctl_createv(clog, 0, NULL, NULL, 1653 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1654 CTLTYPE_INT, "directed-broadcast", 1655 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1656 NULL, 0, &ip_directedbcast, 0, 1657 CTL_NET, PF_INET, IPPROTO_IP, 1658 IPCTL_DIRECTEDBCAST, CTL_EOL); 1659 sysctl_createv(clog, 0, NULL, NULL, 1660 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1661 CTLTYPE_INT, "allowsrcrt", 1662 SYSCTL_DESCR("Accept source-routed datagrams"), 1663 NULL, 0, &ip_allowsrcrt, 0, 1664 CTL_NET, PF_INET, IPPROTO_IP, 1665 IPCTL_ALLOWSRCRT, CTL_EOL); 1666 1667 sysctl_createv(clog, 0, NULL, NULL, 1668 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1669 CTLTYPE_INT, "mtudisc", 1670 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1671 NULL, 0, &ip_mtudisc, 0, 1672 CTL_NET, PF_INET, IPPROTO_IP, 1673 IPCTL_MTUDISC, CTL_EOL); 1674 sysctl_createv(clog, 0, NULL, NULL, 1675 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1676 CTLTYPE_INT, "anonportmin", 1677 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1678 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1679 CTL_NET, PF_INET, IPPROTO_IP, 1680 IPCTL_ANONPORTMIN, CTL_EOL); 1681 sysctl_createv(clog, 0, NULL, NULL, 1682 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1683 CTLTYPE_INT, "anonportmax", 1684 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1685 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1686 CTL_NET, PF_INET, IPPROTO_IP, 1687 IPCTL_ANONPORTMAX, CTL_EOL); 1688 sysctl_createv(clog, 0, NULL, NULL, 1689 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1690 CTLTYPE_INT, "mtudisctimeout", 1691 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1692 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1693 CTL_NET, PF_INET, IPPROTO_IP, 1694 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1695 #ifndef IPNOPRIVPORTS 1696 sysctl_createv(clog, 0, NULL, NULL, 1697 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1698 CTLTYPE_INT, "lowportmin", 1699 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1700 "to assign"), 1701 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1702 CTL_NET, PF_INET, IPPROTO_IP, 1703 IPCTL_LOWPORTMIN, CTL_EOL); 1704 sysctl_createv(clog, 0, NULL, NULL, 1705 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1706 CTLTYPE_INT, "lowportmax", 1707 SYSCTL_DESCR("Highest privileged ephemeral port number " 1708 "to assign"), 1709 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1710 CTL_NET, PF_INET, IPPROTO_IP, 1711 IPCTL_LOWPORTMAX, CTL_EOL); 1712 #endif /* IPNOPRIVPORTS */ 1713 #if NGRE > 0 1714 sysctl_createv(clog, 0, NULL, NULL, 1715 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1716 CTLTYPE_INT, "grettl", 1717 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1718 NULL, 0, &ip_gre_ttl, 0, 1719 CTL_NET, PF_INET, IPPROTO_IP, 1720 IPCTL_GRE_TTL, CTL_EOL); 1721 #endif /* NGRE */ 1722 sysctl_createv(clog, 0, NULL, NULL, 1723 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1724 CTLTYPE_INT, "checkinterface", 1725 SYSCTL_DESCR("Enable receive side of Strong ES model " 1726 "from RFC1122"), 1727 NULL, 0, &ip_checkinterface, 0, 1728 CTL_NET, PF_INET, IPPROTO_IP, 1729 IPCTL_CHECKINTERFACE, CTL_EOL); 1730 sysctl_createv(clog, 0, NULL, NULL, 1731 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1732 CTLTYPE_INT, "random_id", 1733 SYSCTL_DESCR("Assign random ip_id values"), 1734 NULL, 0, &ip_do_randomid, 0, 1735 CTL_NET, PF_INET, IPPROTO_IP, 1736 IPCTL_RANDOMID, CTL_EOL); 1737 sysctl_createv(clog, 0, NULL, NULL, 1738 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1739 CTLTYPE_INT, "do_loopback_cksum", 1740 SYSCTL_DESCR("Perform IP checksum on loopback"), 1741 NULL, 0, &ip_do_loopback_cksum, 0, 1742 CTL_NET, PF_INET, IPPROTO_IP, 1743 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1744 sysctl_createv(clog, 0, NULL, NULL, 1745 CTLFLAG_PERMANENT, 1746 CTLTYPE_STRUCT, "stats", 1747 SYSCTL_DESCR("IP statistics"), 1748 sysctl_net_inet_ip_stats, 0, NULL, 0, 1749 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1750 CTL_EOL); 1751 #if NARP 1752 sysctl_createv(clog, 0, NULL, NULL, 1753 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1754 CTLTYPE_INT, "dad_count", 1755 SYSCTL_DESCR("Number of Duplicate Address Detection " 1756 "probes to send"), 1757 NULL, 0, &ip_dad_count, 0, 1758 CTL_NET, PF_INET, IPPROTO_IP, 1759 IPCTL_DAD_COUNT, CTL_EOL); 1760 #endif 1761 1762 /* anonportalgo RFC6056 subtree */ 1763 const struct sysctlnode *portalgo_node; 1764 sysctl_createv(clog, 0, NULL, &portalgo_node, 1765 CTLFLAG_PERMANENT, 1766 CTLTYPE_NODE, "anonportalgo", 1767 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1768 NULL, 0, NULL, 0, 1769 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1770 sysctl_createv(clog, 0, &portalgo_node, NULL, 1771 CTLFLAG_PERMANENT, 1772 CTLTYPE_STRING, "available", 1773 SYSCTL_DESCR("available algorithms"), 1774 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1775 CTL_CREATE, CTL_EOL); 1776 sysctl_createv(clog, 0, &portalgo_node, NULL, 1777 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1778 CTLTYPE_STRING, "selected", 1779 SYSCTL_DESCR("selected algorithm"), 1780 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1781 CTL_CREATE, CTL_EOL); 1782 sysctl_createv(clog, 0, &portalgo_node, NULL, 1783 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1784 CTLTYPE_STRUCT, "reserve", 1785 SYSCTL_DESCR("bitmap of reserved ports"), 1786 sysctl_portalgo_reserve4, 0, NULL, 0, 1787 CTL_CREATE, CTL_EOL); 1788 } 1789 1790 void 1791 ip_statinc(u_int stat) 1792 { 1793 1794 KASSERT(stat < IP_NSTATS); 1795 IP_STATINC(stat); 1796 } 1797