ip_input.c revision 1.334
1 /* $NetBSD: ip_input.c,v 1.334 2016/07/06 05:27:52 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 1.334 2016/07/06 05:27:52 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_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; 451 int hlen = 0, len; 452 int downmatch; 453 int srcrt = 0; 454 ifnet_t *ifp; 455 struct psref psref; 456 457 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 458 "interrupt handler; synchronization assumptions violated"); 459 460 MCLAIM(m, &ip_rx_mowner); 461 KASSERT((m->m_flags & M_PKTHDR) != 0); 462 463 ifp = m_get_rcvif_psref(m, &psref); 464 if (__predict_false(ifp == NULL)) 465 goto out; 466 467 /* 468 * If no IP addresses have been set yet but the interfaces 469 * are receiving, can't do anything with incoming packets yet. 470 * Note: we pre-check without locks held. 471 */ 472 if (!TAILQ_FIRST(&in_ifaddrhead)) { 473 goto out; 474 } 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 ia = ip_match_our_address(ifp, ip, &downmatch); 670 if (ia != NULL) 671 goto ours; 672 673 if (ifp->if_flags & IFF_BROADCAST) { 674 ia = ip_match_our_address_broadcast(ifp, ip); 675 if (ia != NULL) 676 goto ours; 677 } 678 679 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 680 #ifdef MROUTING 681 extern struct socket *ip_mrouter; 682 683 if (ip_mrouter) { 684 /* 685 * If we are acting as a multicast router, all 686 * incoming multicast packets are passed to the 687 * kernel-level multicast forwarding function. 688 * The packet is returned (relatively) intact; if 689 * ip_mforward() returns a non-zero value, the packet 690 * must be discarded, else it may be accepted below. 691 * 692 * (The IP ident field is put in the same byte order 693 * as expected when ip_mforward() is called from 694 * ip_output().) 695 */ 696 SOFTNET_LOCK(); 697 if (ip_mforward(m, ifp) != 0) { 698 SOFTNET_UNLOCK(); 699 IP_STATINC(IP_STAT_CANTFORWARD); 700 goto out; 701 } 702 SOFTNET_UNLOCK(); 703 704 /* 705 * The process-level routing demon needs to receive 706 * all multicast IGMP packets, whether or not this 707 * host belongs to their destination groups. 708 */ 709 if (ip->ip_p == IPPROTO_IGMP) { 710 goto ours; 711 } 712 IP_STATINC(IP_STAT_CANTFORWARD); 713 } 714 #endif 715 /* 716 * See if we belong to the destination multicast group on the 717 * arrival interface. 718 */ 719 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 720 IP_STATINC(IP_STAT_CANTFORWARD); 721 goto out; 722 } 723 goto ours; 724 } 725 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 726 in_nullhost(ip->ip_dst)) 727 goto ours; 728 729 /* 730 * Not for us; forward if possible and desirable. 731 */ 732 if (ipforwarding == 0) { 733 m_put_rcvif_psref(ifp, &psref); 734 IP_STATINC(IP_STAT_CANTFORWARD); 735 m_freem(m); 736 } else { 737 /* 738 * If ip_dst matched any of my address on !IFF_UP interface, 739 * and there's no IFF_UP interface that matches ip_dst, 740 * send icmp unreach. Forwarding it will result in in-kernel 741 * forwarding loop till TTL goes to 0. 742 */ 743 if (downmatch) { 744 m_put_rcvif_psref(ifp, &psref); 745 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 746 IP_STATINC(IP_STAT_CANTFORWARD); 747 return; 748 } 749 #ifdef IPSEC 750 /* Perform IPsec, if any. */ 751 if (ipsec_used) { 752 SOFTNET_LOCK(); 753 if (ipsec4_input(m, IP_FORWARDING | 754 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 755 SOFTNET_UNLOCK(); 756 goto out; 757 } 758 SOFTNET_UNLOCK(); 759 } 760 #endif 761 ip_forward(m, srcrt, ifp); 762 m_put_rcvif_psref(ifp, &psref); 763 } 764 return; 765 766 ours: 767 m_put_rcvif_psref(ifp, &psref); 768 ifp = NULL; 769 770 /* 771 * If offset or IP_MF are set, must reassemble. 772 */ 773 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 774 /* 775 * Pass to IP reassembly mechanism. 776 */ 777 if (ip_reass_packet(&m, ip) != 0) { 778 /* Failed; invalid fragment(s) or packet. */ 779 goto out; 780 } 781 if (m == NULL) { 782 /* More fragments should come; silently return. */ 783 goto out; 784 } 785 /* 786 * Reassembly is done, we have the final packet. 787 * Updated cached data in local variable(s). 788 */ 789 ip = mtod(m, struct ip *); 790 hlen = ip->ip_hl << 2; 791 } 792 793 #ifdef IPSEC 794 /* 795 * Enforce IPsec policy checking if we are seeing last header. 796 * Note that we do not visit this with protocols with PCB layer 797 * code - like UDP/TCP/raw IP. 798 */ 799 if (ipsec_used && 800 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 801 SOFTNET_LOCK(); 802 if (ipsec4_input(m, 0) != 0) { 803 SOFTNET_UNLOCK(); 804 goto out; 805 } 806 SOFTNET_UNLOCK(); 807 } 808 #endif 809 810 /* 811 * Switch out to protocol's input routine. 812 */ 813 #if IFA_STATS 814 if (ia && ip) 815 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 816 #endif 817 IP_STATINC(IP_STAT_DELIVERED); 818 819 const int off = hlen, nh = ip->ip_p; 820 821 SOFTNET_LOCK(); 822 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 823 SOFTNET_UNLOCK(); 824 return; 825 826 out: 827 m_put_rcvif_psref(ifp, &psref); 828 if (m != NULL) 829 m_freem(m); 830 } 831 832 /* 833 * IP timer processing. 834 */ 835 void 836 ip_slowtimo(void) 837 { 838 839 mutex_enter(softnet_lock); 840 KERNEL_LOCK(1, NULL); 841 842 ip_reass_slowtimo(); 843 844 KERNEL_UNLOCK_ONE(NULL); 845 mutex_exit(softnet_lock); 846 } 847 848 /* 849 * IP drain processing. 850 */ 851 void 852 ip_drain(void) 853 { 854 855 KERNEL_LOCK(1, NULL); 856 ip_reass_drain(); 857 KERNEL_UNLOCK_ONE(NULL); 858 } 859 860 /* 861 * ip_dooptions: perform option processing on a datagram, possibly discarding 862 * it if bad options are encountered, or forwarding it if source-routed. 863 * 864 * => Returns true if packet has been forwarded/freed. 865 * => Returns false if the packet should be processed further. 866 */ 867 static bool 868 ip_dooptions(struct mbuf *m) 869 { 870 struct ip *ip = mtod(m, struct ip *); 871 u_char *cp, *cp0; 872 struct ip_timestamp *ipt; 873 struct in_ifaddr *ia; 874 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 875 struct in_addr dst; 876 n_time ntime; 877 878 dst = ip->ip_dst; 879 cp = (u_char *)(ip + 1); 880 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 881 for (; cnt > 0; cnt -= optlen, cp += optlen) { 882 opt = cp[IPOPT_OPTVAL]; 883 if (opt == IPOPT_EOL) 884 break; 885 if (opt == IPOPT_NOP) 886 optlen = 1; 887 else { 888 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 889 code = &cp[IPOPT_OLEN] - (u_char *)ip; 890 goto bad; 891 } 892 optlen = cp[IPOPT_OLEN]; 893 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 894 code = &cp[IPOPT_OLEN] - (u_char *)ip; 895 goto bad; 896 } 897 } 898 switch (opt) { 899 900 default: 901 break; 902 903 /* 904 * Source routing with record. 905 * Find interface with current destination address. 906 * If none on this machine then drop if strictly routed, 907 * or do nothing if loosely routed. 908 * Record interface address and bring up next address 909 * component. If strictly routed make sure next 910 * address is on directly accessible net. 911 */ 912 case IPOPT_LSRR: 913 case IPOPT_SSRR: 914 if (ip_allowsrcrt == 0) { 915 type = ICMP_UNREACH; 916 code = ICMP_UNREACH_NET_PROHIB; 917 goto bad; 918 } 919 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 920 code = &cp[IPOPT_OLEN] - (u_char *)ip; 921 goto bad; 922 } 923 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 924 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 925 goto bad; 926 } 927 ipaddr.sin_addr = ip->ip_dst; 928 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 929 if (ia == 0) { 930 if (opt == IPOPT_SSRR) { 931 type = ICMP_UNREACH; 932 code = ICMP_UNREACH_SRCFAIL; 933 goto bad; 934 } 935 /* 936 * Loose routing, and not at next destination 937 * yet; nothing to do except forward. 938 */ 939 break; 940 } 941 off--; /* 0 origin */ 942 if ((off + sizeof(struct in_addr)) > optlen) { 943 /* 944 * End of source route. Should be for us. 945 */ 946 save_rte(cp, ip->ip_src); 947 break; 948 } 949 /* 950 * locate outgoing interface 951 */ 952 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 953 sizeof(ipaddr.sin_addr)); 954 if (opt == IPOPT_SSRR) 955 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 956 else 957 ia = ip_rtaddr(ipaddr.sin_addr); 958 if (ia == 0) { 959 type = ICMP_UNREACH; 960 code = ICMP_UNREACH_SRCFAIL; 961 goto bad; 962 } 963 ip->ip_dst = ipaddr.sin_addr; 964 bcopy((void *)&ia->ia_addr.sin_addr, 965 (void *)(cp + off), sizeof(struct in_addr)); 966 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 967 /* 968 * Let ip_intr's mcast routing check handle mcast pkts 969 */ 970 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 971 break; 972 973 case IPOPT_RR: 974 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 975 code = &cp[IPOPT_OLEN] - (u_char *)ip; 976 goto bad; 977 } 978 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 979 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 980 goto bad; 981 } 982 /* 983 * If no space remains, ignore. 984 */ 985 off--; /* 0 origin */ 986 if ((off + sizeof(struct in_addr)) > optlen) 987 break; 988 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 989 sizeof(ipaddr.sin_addr)); 990 /* 991 * locate outgoing interface; if we're the destination, 992 * use the incoming interface (should be same). 993 */ 994 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 995 == NULL && 996 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 997 type = ICMP_UNREACH; 998 code = ICMP_UNREACH_HOST; 999 goto bad; 1000 } 1001 bcopy((void *)&ia->ia_addr.sin_addr, 1002 (void *)(cp + off), sizeof(struct in_addr)); 1003 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1004 break; 1005 1006 case IPOPT_TS: 1007 code = cp - (u_char *)ip; 1008 ipt = (struct ip_timestamp *)cp; 1009 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1010 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1011 goto bad; 1012 } 1013 if (ipt->ipt_ptr < 5) { 1014 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1015 goto bad; 1016 } 1017 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 1018 if (++ipt->ipt_oflw == 0) { 1019 code = (u_char *)&ipt->ipt_ptr - 1020 (u_char *)ip; 1021 goto bad; 1022 } 1023 break; 1024 } 1025 cp0 = (cp + ipt->ipt_ptr - 1); 1026 switch (ipt->ipt_flg) { 1027 1028 case IPOPT_TS_TSONLY: 1029 break; 1030 1031 case IPOPT_TS_TSANDADDR: { 1032 struct ifnet *rcvif; 1033 int s; 1034 1035 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1036 sizeof(struct in_addr) > ipt->ipt_len) { 1037 code = (u_char *)&ipt->ipt_ptr - 1038 (u_char *)ip; 1039 goto bad; 1040 } 1041 ipaddr.sin_addr = dst; 1042 rcvif = m_get_rcvif(m, &s); 1043 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1044 rcvif)); 1045 m_put_rcvif(rcvif, &s); 1046 if (ia == 0) 1047 continue; 1048 bcopy(&ia->ia_addr.sin_addr, 1049 cp0, sizeof(struct in_addr)); 1050 ipt->ipt_ptr += sizeof(struct in_addr); 1051 break; 1052 } 1053 1054 case IPOPT_TS_PRESPEC: 1055 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1056 sizeof(struct in_addr) > ipt->ipt_len) { 1057 code = (u_char *)&ipt->ipt_ptr - 1058 (u_char *)ip; 1059 goto bad; 1060 } 1061 memcpy(&ipaddr.sin_addr, cp0, 1062 sizeof(struct in_addr)); 1063 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1064 == NULL) 1065 continue; 1066 ipt->ipt_ptr += sizeof(struct in_addr); 1067 break; 1068 1069 default: 1070 /* XXX can't take &ipt->ipt_flg */ 1071 code = (u_char *)&ipt->ipt_ptr - 1072 (u_char *)ip + 1; 1073 goto bad; 1074 } 1075 ntime = iptime(); 1076 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1077 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1078 sizeof(n_time)); 1079 ipt->ipt_ptr += sizeof(n_time); 1080 } 1081 } 1082 if (forward) { 1083 struct ifnet *rcvif; 1084 struct psref psref; 1085 1086 if (ip_forwsrcrt == 0) { 1087 type = ICMP_UNREACH; 1088 code = ICMP_UNREACH_SRCFAIL; 1089 goto bad; 1090 } 1091 1092 rcvif = m_get_rcvif_psref(m, &psref); 1093 if (__predict_false(rcvif == NULL)) { 1094 type = ICMP_UNREACH; 1095 code = ICMP_UNREACH_HOST; 1096 goto bad; 1097 } 1098 ip_forward(m, 1, rcvif); 1099 m_put_rcvif_psref(rcvif, &psref); 1100 return true; 1101 } 1102 return false; 1103 bad: 1104 icmp_error(m, type, code, 0, 0); 1105 IP_STATINC(IP_STAT_BADOPTIONS); 1106 return true; 1107 } 1108 1109 /* 1110 * ip_rtaddr: given address of next destination (final or next hop), 1111 * return internet address info of interface to be used to get there. 1112 */ 1113 static struct in_ifaddr * 1114 ip_rtaddr(struct in_addr dst) 1115 { 1116 struct rtentry *rt; 1117 union { 1118 struct sockaddr dst; 1119 struct sockaddr_in dst4; 1120 } u; 1121 1122 sockaddr_in_init(&u.dst4, &dst, 0); 1123 1124 SOFTNET_LOCK(); 1125 rt = rtcache_lookup(&ipforward_rt, &u.dst); 1126 SOFTNET_UNLOCK(); 1127 if (rt == NULL) 1128 return NULL; 1129 1130 return ifatoia(rt->rt_ifa); 1131 } 1132 1133 /* 1134 * save_rte: save incoming source route for use in replies, to be picked 1135 * up later by ip_srcroute if the receiver is interested. 1136 */ 1137 static void 1138 save_rte(u_char *option, struct in_addr dst) 1139 { 1140 unsigned olen; 1141 1142 olen = option[IPOPT_OLEN]; 1143 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1144 return; 1145 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen); 1146 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1147 ip_srcrt.dst = dst; 1148 } 1149 1150 /* 1151 * Retrieve incoming source route for use in replies, 1152 * in the same form used by setsockopt. 1153 * The first hop is placed before the options, will be removed later. 1154 */ 1155 struct mbuf * 1156 ip_srcroute(void) 1157 { 1158 struct in_addr *p, *q; 1159 struct mbuf *m; 1160 1161 if (ip_nhops == 0) 1162 return NULL; 1163 m = m_get(M_DONTWAIT, MT_SOOPTS); 1164 if (m == 0) 1165 return NULL; 1166 1167 MCLAIM(m, &inetdomain.dom_mowner); 1168 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1169 1170 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1171 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1172 OPTSIZ; 1173 1174 /* 1175 * First save first hop for return route 1176 */ 1177 p = &ip_srcrt.route[ip_nhops - 1]; 1178 *(mtod(m, struct in_addr *)) = *p--; 1179 1180 /* 1181 * Copy option fields and padding (nop) to mbuf. 1182 */ 1183 ip_srcrt.nop = IPOPT_NOP; 1184 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1185 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1186 OPTSIZ); 1187 q = (struct in_addr *)(mtod(m, char *) + 1188 sizeof(struct in_addr) + OPTSIZ); 1189 #undef OPTSIZ 1190 /* 1191 * Record return path as an IP source route, 1192 * reversing the path (pointers are now aligned). 1193 */ 1194 while (p >= ip_srcrt.route) { 1195 *q++ = *p--; 1196 } 1197 /* 1198 * Last hop goes to final destination. 1199 */ 1200 *q = ip_srcrt.dst; 1201 return (m); 1202 } 1203 1204 const int inetctlerrmap[PRC_NCMDS] = { 1205 [PRC_MSGSIZE] = EMSGSIZE, 1206 [PRC_HOSTDEAD] = EHOSTDOWN, 1207 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1208 [PRC_UNREACH_NET] = EHOSTUNREACH, 1209 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1210 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1211 [PRC_UNREACH_PORT] = ECONNREFUSED, 1212 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1213 [PRC_PARAMPROB] = ENOPROTOOPT, 1214 }; 1215 1216 void 1217 ip_fasttimo(void) 1218 { 1219 if (ip_drainwanted) { 1220 ip_drain(); 1221 ip_drainwanted = 0; 1222 } 1223 } 1224 1225 void 1226 ip_drainstub(void) 1227 { 1228 ip_drainwanted = 1; 1229 } 1230 1231 /* 1232 * Forward a packet. If some error occurs return the sender 1233 * an icmp packet. Note we can't always generate a meaningful 1234 * icmp message because icmp doesn't have a large enough repertoire 1235 * of codes and types. 1236 * 1237 * If not forwarding, just drop the packet. This could be confusing 1238 * if ipforwarding was zero but some routing protocol was advancing 1239 * us as a gateway to somewhere. However, we must let the routing 1240 * protocol deal with that. 1241 * 1242 * The srcrt parameter indicates whether the packet is being forwarded 1243 * via a source route. 1244 */ 1245 static void 1246 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1247 { 1248 struct ip *ip = mtod(m, struct ip *); 1249 struct rtentry *rt; 1250 int error, type = 0, code = 0, destmtu = 0; 1251 struct mbuf *mcopy; 1252 n_long dest; 1253 union { 1254 struct sockaddr dst; 1255 struct sockaddr_in dst4; 1256 } u; 1257 uint64_t *ips; 1258 1259 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1260 "interrupt handler; synchronization assumptions violated"); 1261 1262 /* 1263 * We are now in the output path. 1264 */ 1265 MCLAIM(m, &ip_tx_mowner); 1266 1267 /* 1268 * Clear any in-bound checksum flags for this packet. 1269 */ 1270 m->m_pkthdr.csum_flags = 0; 1271 1272 dest = 0; 1273 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1274 IP_STATINC(IP_STAT_CANTFORWARD); 1275 m_freem(m); 1276 return; 1277 } 1278 1279 SOFTNET_LOCK(); 1280 1281 if (ip->ip_ttl <= IPTTLDEC) { 1282 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1283 SOFTNET_UNLOCK(); 1284 return; 1285 } 1286 1287 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1288 1289 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1290 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1291 SOFTNET_UNLOCK(); 1292 return; 1293 } 1294 1295 /* 1296 * Save at most 68 bytes of the packet in case 1297 * we need to generate an ICMP message to the src. 1298 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1299 */ 1300 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1301 if (mcopy) 1302 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1303 1304 ip->ip_ttl -= IPTTLDEC; 1305 1306 /* 1307 * If forwarding packet using same interface that it came in on, 1308 * perhaps should send a redirect to sender to shortcut a hop. 1309 * Only send redirect if source is sending directly to us, 1310 * and if packet was not source routed (or has any options). 1311 * Also, don't send redirect if forwarding using a default route 1312 * or a route modified by a redirect. 1313 */ 1314 if (rt->rt_ifp == rcvif && 1315 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1316 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1317 ipsendredirects && !srcrt) { 1318 if (rt->rt_ifa && 1319 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1320 ifatoia(rt->rt_ifa)->ia_subnet) { 1321 if (rt->rt_flags & RTF_GATEWAY) 1322 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1323 else 1324 dest = ip->ip_dst.s_addr; 1325 /* 1326 * Router requirements says to only send host 1327 * redirects. 1328 */ 1329 type = ICMP_REDIRECT; 1330 code = ICMP_REDIRECT_HOST; 1331 } 1332 } 1333 1334 error = ip_output(m, NULL, &ipforward_rt, 1335 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1336 NULL, NULL); 1337 1338 if (error) { 1339 IP_STATINC(IP_STAT_CANTFORWARD); 1340 goto error; 1341 } 1342 1343 ips = IP_STAT_GETREF(); 1344 ips[IP_STAT_FORWARD]++; 1345 1346 if (type) { 1347 ips[IP_STAT_REDIRECTSENT]++; 1348 IP_STAT_PUTREF(); 1349 goto redirect; 1350 } 1351 1352 IP_STAT_PUTREF(); 1353 if (mcopy) { 1354 #ifdef GATEWAY 1355 if (mcopy->m_flags & M_CANFASTFWD) 1356 ipflow_create(&ipforward_rt, mcopy); 1357 #endif 1358 m_freem(mcopy); 1359 } 1360 1361 SOFTNET_UNLOCK(); 1362 return; 1363 1364 redirect: 1365 error: 1366 if (mcopy == NULL) { 1367 SOFTNET_UNLOCK(); 1368 return; 1369 } 1370 1371 switch (error) { 1372 1373 case 0: /* forwarded, but need redirect */ 1374 /* type, code set above */ 1375 break; 1376 1377 case ENETUNREACH: /* shouldn't happen, checked above */ 1378 case EHOSTUNREACH: 1379 case ENETDOWN: 1380 case EHOSTDOWN: 1381 default: 1382 type = ICMP_UNREACH; 1383 code = ICMP_UNREACH_HOST; 1384 break; 1385 1386 case EMSGSIZE: 1387 type = ICMP_UNREACH; 1388 code = ICMP_UNREACH_NEEDFRAG; 1389 1390 if ((rt = rtcache_validate(&ipforward_rt)) != NULL) 1391 destmtu = rt->rt_ifp->if_mtu; 1392 #ifdef IPSEC 1393 if (ipsec_used) 1394 (void)ipsec4_forward(mcopy, &destmtu); 1395 #endif 1396 IP_STATINC(IP_STAT_CANTFRAG); 1397 break; 1398 1399 case ENOBUFS: 1400 /* 1401 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1402 * Requirements for IP Version 4 Routers. Source quench can 1403 * big problem under DoS attacks or if the underlying 1404 * interface is rate-limited. 1405 */ 1406 if (mcopy) 1407 m_freem(mcopy); 1408 SOFTNET_UNLOCK(); 1409 return; 1410 } 1411 icmp_error(mcopy, type, code, dest, destmtu); 1412 SOFTNET_UNLOCK(); 1413 } 1414 1415 void 1416 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1417 struct mbuf *m) 1418 { 1419 struct socket *so = inp->inp_socket; 1420 ifnet_t *ifp; 1421 int inpflags = inp->inp_flags; 1422 struct psref psref; 1423 1424 ifp = m_get_rcvif_psref(m, &psref); 1425 if (__predict_false(ifp == NULL)) 1426 return; /* XXX should report error? */ 1427 1428 if (so->so_options & SO_TIMESTAMP 1429 #ifdef SO_OTIMESTAMP 1430 || so->so_options & SO_OTIMESTAMP 1431 #endif 1432 ) { 1433 struct timeval tv; 1434 1435 microtime(&tv); 1436 #ifdef SO_OTIMESTAMP 1437 if (so->so_options & SO_OTIMESTAMP) { 1438 struct timeval50 tv50; 1439 timeval_to_timeval50(&tv, &tv50); 1440 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50), 1441 SCM_OTIMESTAMP, SOL_SOCKET); 1442 } else 1443 #endif 1444 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 1445 SCM_TIMESTAMP, SOL_SOCKET); 1446 if (*mp) 1447 mp = &(*mp)->m_next; 1448 } 1449 if (inpflags & INP_RECVDSTADDR) { 1450 *mp = sbcreatecontrol((void *) &ip->ip_dst, 1451 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1452 if (*mp) 1453 mp = &(*mp)->m_next; 1454 } 1455 if (inpflags & INP_RECVPKTINFO) { 1456 struct in_pktinfo ipi; 1457 ipi.ipi_addr = ip->ip_src; 1458 ipi.ipi_ifindex = ifp->if_index; 1459 *mp = sbcreatecontrol((void *) &ipi, 1460 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP); 1461 if (*mp) 1462 mp = &(*mp)->m_next; 1463 } 1464 if (inpflags & INP_PKTINFO) { 1465 struct in_pktinfo ipi; 1466 ipi.ipi_addr = ip->ip_dst; 1467 ipi.ipi_ifindex = ifp->if_index; 1468 *mp = sbcreatecontrol((void *) &ipi, 1469 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1470 if (*mp) 1471 mp = &(*mp)->m_next; 1472 } 1473 if (inpflags & INP_RECVIF) { 1474 struct sockaddr_dl sdl; 1475 1476 sockaddr_dl_init(&sdl, sizeof(sdl), ifp ? 1477 ifp->if_index : 0, 0, NULL, 0, NULL, 0); 1478 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1479 if (*mp) 1480 mp = &(*mp)->m_next; 1481 } 1482 if (inpflags & INP_RECVTTL) { 1483 *mp = sbcreatecontrol((void *) &ip->ip_ttl, 1484 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1485 if (*mp) 1486 mp = &(*mp)->m_next; 1487 } 1488 m_put_rcvif_psref(ifp, &psref); 1489 } 1490 1491 /* 1492 * sysctl helper routine for net.inet.ip.forwsrcrt. 1493 */ 1494 static int 1495 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1496 { 1497 int error, tmp; 1498 struct sysctlnode node; 1499 1500 node = *rnode; 1501 tmp = ip_forwsrcrt; 1502 node.sysctl_data = &tmp; 1503 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1504 if (error || newp == NULL) 1505 return (error); 1506 1507 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1508 0, NULL, NULL, NULL); 1509 if (error) 1510 return (error); 1511 1512 ip_forwsrcrt = tmp; 1513 1514 return (0); 1515 } 1516 1517 /* 1518 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1519 * range of the new value and tweaks timers if it changes. 1520 */ 1521 static int 1522 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1523 { 1524 int error, tmp; 1525 struct sysctlnode node; 1526 1527 node = *rnode; 1528 tmp = ip_mtudisc_timeout; 1529 node.sysctl_data = &tmp; 1530 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1531 if (error || newp == NULL) 1532 return (error); 1533 if (tmp < 0) 1534 return (EINVAL); 1535 1536 mutex_enter(softnet_lock); 1537 1538 ip_mtudisc_timeout = tmp; 1539 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1540 1541 mutex_exit(softnet_lock); 1542 1543 return (0); 1544 } 1545 1546 static int 1547 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1548 { 1549 1550 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1551 } 1552 1553 static void 1554 sysctl_net_inet_ip_setup(struct sysctllog **clog) 1555 { 1556 sysctl_createv(clog, 0, NULL, NULL, 1557 CTLFLAG_PERMANENT, 1558 CTLTYPE_NODE, "inet", 1559 SYSCTL_DESCR("PF_INET related settings"), 1560 NULL, 0, NULL, 0, 1561 CTL_NET, PF_INET, CTL_EOL); 1562 sysctl_createv(clog, 0, NULL, NULL, 1563 CTLFLAG_PERMANENT, 1564 CTLTYPE_NODE, "ip", 1565 SYSCTL_DESCR("IPv4 related settings"), 1566 NULL, 0, NULL, 0, 1567 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1568 1569 sysctl_createv(clog, 0, NULL, NULL, 1570 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1571 CTLTYPE_INT, "forwarding", 1572 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1573 NULL, 0, &ipforwarding, 0, 1574 CTL_NET, PF_INET, IPPROTO_IP, 1575 IPCTL_FORWARDING, CTL_EOL); 1576 sysctl_createv(clog, 0, NULL, NULL, 1577 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1578 CTLTYPE_INT, "redirect", 1579 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1580 NULL, 0, &ipsendredirects, 0, 1581 CTL_NET, PF_INET, IPPROTO_IP, 1582 IPCTL_SENDREDIRECTS, CTL_EOL); 1583 sysctl_createv(clog, 0, NULL, NULL, 1584 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1585 CTLTYPE_INT, "ttl", 1586 SYSCTL_DESCR("Default TTL for an INET datagram"), 1587 NULL, 0, &ip_defttl, 0, 1588 CTL_NET, PF_INET, IPPROTO_IP, 1589 IPCTL_DEFTTL, CTL_EOL); 1590 #ifdef IPCTL_DEFMTU 1591 sysctl_createv(clog, 0, NULL, NULL, 1592 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1593 CTLTYPE_INT, "mtu", 1594 SYSCTL_DESCR("Default MTA for an INET route"), 1595 NULL, 0, &ip_mtu, 0, 1596 CTL_NET, PF_INET, IPPROTO_IP, 1597 IPCTL_DEFMTU, CTL_EOL); 1598 #endif /* IPCTL_DEFMTU */ 1599 sysctl_createv(clog, 0, NULL, NULL, 1600 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1601 CTLTYPE_INT, "forwsrcrt", 1602 SYSCTL_DESCR("Enable forwarding of source-routed " 1603 "datagrams"), 1604 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1605 CTL_NET, PF_INET, IPPROTO_IP, 1606 IPCTL_FORWSRCRT, CTL_EOL); 1607 sysctl_createv(clog, 0, NULL, NULL, 1608 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1609 CTLTYPE_INT, "directed-broadcast", 1610 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1611 NULL, 0, &ip_directedbcast, 0, 1612 CTL_NET, PF_INET, IPPROTO_IP, 1613 IPCTL_DIRECTEDBCAST, CTL_EOL); 1614 sysctl_createv(clog, 0, NULL, NULL, 1615 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1616 CTLTYPE_INT, "allowsrcrt", 1617 SYSCTL_DESCR("Accept source-routed datagrams"), 1618 NULL, 0, &ip_allowsrcrt, 0, 1619 CTL_NET, PF_INET, IPPROTO_IP, 1620 IPCTL_ALLOWSRCRT, CTL_EOL); 1621 1622 sysctl_createv(clog, 0, NULL, NULL, 1623 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1624 CTLTYPE_INT, "mtudisc", 1625 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1626 NULL, 0, &ip_mtudisc, 0, 1627 CTL_NET, PF_INET, IPPROTO_IP, 1628 IPCTL_MTUDISC, CTL_EOL); 1629 sysctl_createv(clog, 0, NULL, NULL, 1630 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1631 CTLTYPE_INT, "anonportmin", 1632 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1633 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1634 CTL_NET, PF_INET, IPPROTO_IP, 1635 IPCTL_ANONPORTMIN, CTL_EOL); 1636 sysctl_createv(clog, 0, NULL, NULL, 1637 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1638 CTLTYPE_INT, "anonportmax", 1639 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1640 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1641 CTL_NET, PF_INET, IPPROTO_IP, 1642 IPCTL_ANONPORTMAX, CTL_EOL); 1643 sysctl_createv(clog, 0, NULL, NULL, 1644 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1645 CTLTYPE_INT, "mtudisctimeout", 1646 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1647 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1648 CTL_NET, PF_INET, IPPROTO_IP, 1649 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1650 #ifndef IPNOPRIVPORTS 1651 sysctl_createv(clog, 0, NULL, NULL, 1652 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1653 CTLTYPE_INT, "lowportmin", 1654 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1655 "to assign"), 1656 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1657 CTL_NET, PF_INET, IPPROTO_IP, 1658 IPCTL_LOWPORTMIN, CTL_EOL); 1659 sysctl_createv(clog, 0, NULL, NULL, 1660 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1661 CTLTYPE_INT, "lowportmax", 1662 SYSCTL_DESCR("Highest privileged ephemeral port number " 1663 "to assign"), 1664 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1665 CTL_NET, PF_INET, IPPROTO_IP, 1666 IPCTL_LOWPORTMAX, CTL_EOL); 1667 #endif /* IPNOPRIVPORTS */ 1668 #if NGRE > 0 1669 sysctl_createv(clog, 0, NULL, NULL, 1670 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1671 CTLTYPE_INT, "grettl", 1672 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1673 NULL, 0, &ip_gre_ttl, 0, 1674 CTL_NET, PF_INET, IPPROTO_IP, 1675 IPCTL_GRE_TTL, CTL_EOL); 1676 #endif /* NGRE */ 1677 sysctl_createv(clog, 0, NULL, NULL, 1678 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1679 CTLTYPE_INT, "checkinterface", 1680 SYSCTL_DESCR("Enable receive side of Strong ES model " 1681 "from RFC1122"), 1682 NULL, 0, &ip_checkinterface, 0, 1683 CTL_NET, PF_INET, IPPROTO_IP, 1684 IPCTL_CHECKINTERFACE, CTL_EOL); 1685 sysctl_createv(clog, 0, NULL, NULL, 1686 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1687 CTLTYPE_INT, "random_id", 1688 SYSCTL_DESCR("Assign random ip_id values"), 1689 NULL, 0, &ip_do_randomid, 0, 1690 CTL_NET, PF_INET, IPPROTO_IP, 1691 IPCTL_RANDOMID, CTL_EOL); 1692 sysctl_createv(clog, 0, NULL, NULL, 1693 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1694 CTLTYPE_INT, "do_loopback_cksum", 1695 SYSCTL_DESCR("Perform IP checksum on loopback"), 1696 NULL, 0, &ip_do_loopback_cksum, 0, 1697 CTL_NET, PF_INET, IPPROTO_IP, 1698 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1699 sysctl_createv(clog, 0, NULL, NULL, 1700 CTLFLAG_PERMANENT, 1701 CTLTYPE_STRUCT, "stats", 1702 SYSCTL_DESCR("IP statistics"), 1703 sysctl_net_inet_ip_stats, 0, NULL, 0, 1704 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1705 CTL_EOL); 1706 #if NARP 1707 sysctl_createv(clog, 0, NULL, NULL, 1708 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1709 CTLTYPE_INT, "dad_count", 1710 SYSCTL_DESCR("Number of Duplicate Address Detection " 1711 "probes to send"), 1712 NULL, 0, &ip_dad_count, 0, 1713 CTL_NET, PF_INET, IPPROTO_IP, 1714 IPCTL_DAD_COUNT, CTL_EOL); 1715 #endif 1716 1717 /* anonportalgo RFC6056 subtree */ 1718 const struct sysctlnode *portalgo_node; 1719 sysctl_createv(clog, 0, NULL, &portalgo_node, 1720 CTLFLAG_PERMANENT, 1721 CTLTYPE_NODE, "anonportalgo", 1722 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1723 NULL, 0, NULL, 0, 1724 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1725 sysctl_createv(clog, 0, &portalgo_node, NULL, 1726 CTLFLAG_PERMANENT, 1727 CTLTYPE_STRING, "available", 1728 SYSCTL_DESCR("available algorithms"), 1729 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1730 CTL_CREATE, CTL_EOL); 1731 sysctl_createv(clog, 0, &portalgo_node, NULL, 1732 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1733 CTLTYPE_STRING, "selected", 1734 SYSCTL_DESCR("selected algorithm"), 1735 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1736 CTL_CREATE, CTL_EOL); 1737 sysctl_createv(clog, 0, &portalgo_node, NULL, 1738 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1739 CTLTYPE_STRUCT, "reserve", 1740 SYSCTL_DESCR("bitmap of reserved ports"), 1741 sysctl_portalgo_reserve4, 0, NULL, 0, 1742 CTL_CREATE, CTL_EOL); 1743 } 1744 1745 void 1746 ip_statinc(u_int stat) 1747 { 1748 1749 KASSERT(stat < IP_NSTATS); 1750 IP_STATINC(stat); 1751 } 1752
Indexes created Thu Sep 25 02:09:54 GMT 2025