ip_input.c revision 1.370
1 /* $NetBSD: ip_input.c,v 1.370 2018/02/05 14:52:42 maxv Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 59 * POSSIBILITY OF SUCH DAMAGE. 60 */ 61 62 /* 63 * Copyright (c) 1982, 1986, 1988, 1993 64 * The Regents of the University of California. All rights reserved. 65 * 66 * Redistribution and use in source and binary forms, with or without 67 * modification, are permitted provided that the following conditions 68 * are met: 69 * 1. Redistributions of source code must retain the above copyright 70 * notice, this list of conditions and the following disclaimer. 71 * 2. Redistributions in binary form must reproduce the above copyright 72 * notice, this list of conditions and the following disclaimer in the 73 * documentation and/or other materials provided with the distribution. 74 * 3. Neither the name of the University nor the names of its contributors 75 * may be used to endorse or promote products derived from this software 76 * without specific prior written permission. 77 * 78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 88 * SUCH DAMAGE. 89 * 90 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 91 */ 92 93 #include <sys/cdefs.h> 94 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.370 2018/02/05 14:52:42 maxv Exp $"); 95 96 #ifdef _KERNEL_OPT 97 #include "opt_inet.h" 98 #include "opt_gateway.h" 99 #include "opt_ipsec.h" 100 #include "opt_mrouting.h" 101 #include "opt_mbuftrace.h" 102 #include "opt_inet_csum.h" 103 #include "opt_net_mpsafe.h" 104 #endif 105 106 #include "arp.h" 107 108 #include <sys/param.h> 109 #include <sys/systm.h> 110 #include <sys/cpu.h> 111 #include <sys/mbuf.h> 112 #include <sys/domain.h> 113 #include <sys/protosw.h> 114 #include <sys/socket.h> 115 #include <sys/socketvar.h> 116 #include <sys/errno.h> 117 #include <sys/time.h> 118 #include <sys/kernel.h> 119 #include <sys/pool.h> 120 #include <sys/sysctl.h> 121 #include <sys/kauth.h> 122 123 #include <net/if.h> 124 #include <net/if_dl.h> 125 #include <net/route.h> 126 #include <net/pktqueue.h> 127 #include <net/pfil.h> 128 129 #include <netinet/in.h> 130 #include <netinet/in_systm.h> 131 #include <netinet/ip.h> 132 #include <netinet/in_pcb.h> 133 #include <netinet/in_proto.h> 134 #include <netinet/in_var.h> 135 #include <netinet/ip_var.h> 136 #include <netinet/ip_private.h> 137 #include <netinet/ip_icmp.h> 138 /* just for gif_ttl */ 139 #include <netinet/in_gif.h> 140 #include "gif.h" 141 #include <net/if_gre.h> 142 #include "gre.h" 143 144 #ifdef MROUTING 145 #include <netinet/ip_mroute.h> 146 #endif 147 #include <netinet/portalgo.h> 148 149 #ifdef IPSEC 150 #include <netipsec/ipsec.h> 151 #endif 152 153 #ifndef IPFORWARDING 154 #ifdef GATEWAY 155 #define IPFORWARDING 1 /* forward IP packets not for us */ 156 #else 157 #define IPFORWARDING 0 /* don't forward IP packets not for us */ 158 #endif 159 #endif 160 161 #ifndef IPSENDREDIRECTS 162 #define IPSENDREDIRECTS 1 163 #endif 164 165 #ifndef IPMTUDISCTIMEOUT 166 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */ 167 #endif 168 169 #ifdef DIRECTED_BROADCAST 170 #define IPDIRECTEDBCAST 1 171 #else 172 #define IPDIRECTEDBCAST 0 173 #endif 174 175 int ipforwarding = IPFORWARDING; 176 int ipsendredirects = IPSENDREDIRECTS; 177 int ip_defttl = IPDEFTTL; 178 int ip_forwsrcrt = 0; 179 int ip_directedbcast = IPDIRECTEDBCAST; 180 int ip_allowsrcrt = 0; 181 int ip_mtudisc = 1; 182 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 183 int ip_do_randomid = 0; 184 185 /* 186 * XXX - Setting ip_checkinterface mostly implements the receive side of 187 * the Strong ES model described in RFC 1122, but since the routing table 188 * and transmit implementation do not implement the Strong ES model, 189 * setting this to 1 results in an odd hybrid. 190 * 191 * XXX - ip_checkinterface currently must be disabled if you use ipnat 192 * to translate the destination address to another local interface. 193 * 194 * XXX - ip_checkinterface must be disabled if you add IP aliases 195 * to the loopback interface instead of the interface where the 196 * packets for those addresses are received. 197 */ 198 static int ip_checkinterface __read_mostly = 0; 199 200 struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 201 202 pktqueue_t * ip_pktq __read_mostly; 203 pfil_head_t * inet_pfil_hook __read_mostly; 204 ipid_state_t * ip_ids __read_mostly; 205 percpu_t * ipstat_percpu __read_mostly; 206 207 static percpu_t *ipforward_rt_percpu __cacheline_aligned; 208 209 uint16_t ip_id; 210 211 #ifdef INET_CSUM_COUNTERS 212 #include <sys/device.h> 213 214 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 215 NULL, "inet", "hwcsum bad"); 216 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 217 NULL, "inet", "hwcsum ok"); 218 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, 219 NULL, "inet", "swcsum"); 220 221 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ 222 223 EVCNT_ATTACH_STATIC(ip_hwcsum_bad); 224 EVCNT_ATTACH_STATIC(ip_hwcsum_ok); 225 EVCNT_ATTACH_STATIC(ip_swcsum); 226 227 #else 228 229 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */ 230 231 #endif /* INET_CSUM_COUNTERS */ 232 233 /* 234 * Used to save the IP options in case a protocol wants to respond 235 * to an incoming packet over the same route if the packet got here 236 * using IP source routing. This allows connection establishment and 237 * maintenance when the remote end is on a network that is not known 238 * to us. 239 */ 240 struct ip_srcrt { 241 int isr_nhops; /* number of hops */ 242 struct in_addr isr_dst; /* final destination */ 243 char isr_nop; /* one NOP to align */ 244 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */ 245 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)]; 246 }; 247 248 static int ip_drainwanted; 249 250 static void save_rte(struct mbuf *, u_char *, struct in_addr); 251 252 #ifdef MBUFTRACE 253 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); 254 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); 255 #endif 256 257 static void ipintr(void *); 258 static void ip_input(struct mbuf *); 259 static void ip_forward(struct mbuf *, int, struct ifnet *); 260 static bool ip_dooptions(struct mbuf *); 261 static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *); 262 static void sysctl_net_inet_ip_setup(struct sysctllog **); 263 264 static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *, 265 int *); 266 static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *, 267 struct ip *); 268 269 #ifdef NET_MPSAFE 270 #define SOFTNET_LOCK() mutex_enter(softnet_lock) 271 #define SOFTNET_UNLOCK() mutex_exit(softnet_lock) 272 #else 273 #define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 274 #define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 275 #endif 276 277 /* 278 * IP initialization: fill in IP protocol switch table. 279 * All protocols not implemented in kernel go to raw IP protocol handler. 280 */ 281 void 282 ip_init(void) 283 { 284 const struct protosw *pr; 285 286 in_init(); 287 sysctl_net_inet_ip_setup(NULL); 288 289 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 290 KASSERT(pr != NULL); 291 292 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 293 KASSERT(ip_pktq != NULL); 294 295 for (u_int i = 0; i < IPPROTO_MAX; i++) { 296 ip_protox[i] = pr - inetsw; 297 } 298 for (pr = inetdomain.dom_protosw; 299 pr < inetdomain.dom_protoswNPROTOSW; pr++) 300 if (pr->pr_domain->dom_family == PF_INET && 301 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 302 ip_protox[pr->pr_protocol] = pr - inetsw; 303 304 ip_reass_init(); 305 306 ip_ids = ip_id_init(); 307 ip_id = time_uptime & 0xfffff; 308 309 #ifdef GATEWAY 310 ipflow_init(); 311 #endif 312 313 /* Register our Packet Filter hook. */ 314 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 315 KASSERT(inet_pfil_hook != NULL); 316 317 #ifdef MBUFTRACE 318 MOWNER_ATTACH(&ip_tx_mowner); 319 MOWNER_ATTACH(&ip_rx_mowner); 320 #endif /* MBUFTRACE */ 321 322 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 323 ipforward_rt_percpu = percpu_alloc(sizeof(struct route)); 324 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 325 } 326 327 static struct in_ifaddr * 328 ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch) 329 { 330 struct in_ifaddr *ia = NULL; 331 int checkif; 332 333 /* 334 * Enable a consistency check between the destination address 335 * and the arrival interface for a unicast packet (the RFC 1122 336 * strong ES model) if IP forwarding is disabled and the packet 337 * is not locally generated. 338 * 339 * XXX - Checking also should be disabled if the destination 340 * address is ipnat'ed to a different interface. 341 * 342 * XXX - Checking is incompatible with IP aliases added 343 * to the loopback interface instead of the interface where 344 * the packets are received. 345 * 346 * XXX - We need to add a per ifaddr flag for this so that 347 * we get finer grain control. 348 */ 349 checkif = ip_checkinterface && (ipforwarding == 0) && 350 (ifp->if_flags & IFF_LOOPBACK) == 0; 351 352 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) { 353 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 354 if (ia->ia4_flags & IN_IFF_NOTREADY) 355 continue; 356 if (checkif && ia->ia_ifp != ifp) 357 continue; 358 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) { 359 (*downmatch)++; 360 continue; 361 } 362 if (ia->ia4_flags & IN_IFF_DETACHED && 363 (ifp->if_flags & IFF_LOOPBACK) == 0) 364 continue; 365 break; 366 } 367 } 368 369 return ia; 370 } 371 372 static struct in_ifaddr * 373 ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip) 374 { 375 struct in_ifaddr *ia = NULL; 376 struct ifaddr *ifa; 377 378 IFADDR_READER_FOREACH(ifa, ifp) { 379 if (ifa->ifa_addr->sa_family != AF_INET) 380 continue; 381 ia = ifatoia(ifa); 382 if (ia->ia4_flags & IN_IFF_NOTREADY) 383 continue; 384 if (ia->ia4_flags & IN_IFF_DETACHED && 385 (ifp->if_flags & IFF_LOOPBACK) == 0) 386 continue; 387 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 388 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 389 /* 390 * Look for all-0's host part (old broadcast addr), 391 * either for subnet or net. 392 */ 393 ip->ip_dst.s_addr == ia->ia_subnet || 394 ip->ip_dst.s_addr == ia->ia_net) 395 goto matched; 396 /* 397 * An interface with IP address zero accepts 398 * all packets that arrive on that interface. 399 */ 400 if (in_nullhost(ia->ia_addr.sin_addr)) 401 goto matched; 402 } 403 ia = NULL; 404 405 matched: 406 return ia; 407 } 408 409 /* 410 * IP software interrupt routine. 411 */ 412 static void 413 ipintr(void *arg __unused) 414 { 415 struct mbuf *m; 416 417 KASSERT(cpu_softintr_p()); 418 419 SOFTNET_LOCK_UNLESS_NET_MPSAFE(); 420 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 421 ip_input(m); 422 } 423 SOFTNET_UNLOCK_UNLESS_NET_MPSAFE(); 424 } 425 426 /* 427 * IP input routine. Checksum and byte swap header. If fragmented 428 * try to reassemble. Process options. Pass to next level. 429 */ 430 static void 431 ip_input(struct mbuf *m) 432 { 433 struct ip *ip = NULL; 434 struct in_ifaddr *ia = NULL; 435 int hlen = 0, len; 436 int downmatch; 437 int srcrt = 0; 438 ifnet_t *ifp; 439 struct psref psref; 440 int s; 441 442 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 443 "interrupt handler; synchronization assumptions violated"); 444 445 MCLAIM(m, &ip_rx_mowner); 446 KASSERT((m->m_flags & M_PKTHDR) != 0); 447 448 ifp = m_get_rcvif_psref(m, &psref); 449 if (__predict_false(ifp == NULL)) 450 goto out; 451 452 /* 453 * If no IP addresses have been set yet but the interfaces 454 * are receiving, can't do anything with incoming packets yet. 455 * Note: we pre-check without locks held. 456 */ 457 if (IN_ADDRLIST_READER_EMPTY()) 458 goto out; 459 IP_STATINC(IP_STAT_TOTAL); 460 461 /* 462 * If the IP header is not aligned, slurp it up into a new 463 * mbuf with space for link headers, in the event we forward 464 * it. Otherwise, if it is aligned, make sure the entire 465 * base IP header is in the first mbuf of the chain. 466 */ 467 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 468 if ((m = m_copyup(m, sizeof(struct ip), 469 (max_linkhdr + 3) & ~3)) == NULL) { 470 /* XXXJRT new stat, please */ 471 IP_STATINC(IP_STAT_TOOSMALL); 472 goto out; 473 } 474 } else if (__predict_false(m->m_len < sizeof(struct ip))) { 475 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) { 476 IP_STATINC(IP_STAT_TOOSMALL); 477 goto out; 478 } 479 } 480 ip = mtod(m, struct ip *); 481 if (ip->ip_v != IPVERSION) { 482 IP_STATINC(IP_STAT_BADVERS); 483 goto out; 484 } 485 hlen = ip->ip_hl << 2; 486 if (hlen < sizeof(struct ip)) { /* minimum header length */ 487 IP_STATINC(IP_STAT_BADHLEN); 488 goto out; 489 } 490 if (hlen > m->m_len) { 491 if ((m = m_pullup(m, hlen)) == NULL) { 492 IP_STATINC(IP_STAT_BADHLEN); 493 goto out; 494 } 495 ip = mtod(m, struct ip *); 496 } 497 498 /* 499 * RFC1122: packets with a multicast source address are 500 * not allowed. 501 */ 502 if (IN_MULTICAST(ip->ip_src.s_addr)) { 503 IP_STATINC(IP_STAT_BADADDR); 504 goto out; 505 } 506 507 /* 127/8 must not appear on wire - RFC1122 */ 508 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 509 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 510 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 511 IP_STATINC(IP_STAT_BADADDR); 512 goto out; 513 } 514 } 515 516 switch (m->m_pkthdr.csum_flags & 517 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) { 518 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 519 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 520 IP_STATINC(IP_STAT_BADSUM); 521 goto out; 522 523 case M_CSUM_IPv4: 524 /* Checksum was okay. */ 525 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 526 break; 527 528 default: 529 /* 530 * Must compute it ourselves. Maybe skip checksum on 531 * loopback interfaces. 532 */ 533 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 534 ip_do_loopback_cksum)) { 535 INET_CSUM_COUNTER_INCR(&ip_swcsum); 536 if (in_cksum(m, hlen) != 0) { 537 IP_STATINC(IP_STAT_BADSUM); 538 goto out; 539 } 540 } 541 break; 542 } 543 544 /* Retrieve the packet length. */ 545 len = ntohs(ip->ip_len); 546 547 /* 548 * Check for additional length bogosity 549 */ 550 if (len < hlen) { 551 IP_STATINC(IP_STAT_BADLEN); 552 goto out; 553 } 554 555 /* 556 * Check that the amount of data in the buffers is at least as much 557 * as the IP header would have us expect. Trim mbufs if longer than 558 * we expect. Drop packet if shorter than we expect. 559 */ 560 if (m->m_pkthdr.len < len) { 561 IP_STATINC(IP_STAT_TOOSHORT); 562 goto out; 563 } 564 if (m->m_pkthdr.len > len) { 565 if (m->m_len == m->m_pkthdr.len) { 566 m->m_len = len; 567 m->m_pkthdr.len = len; 568 } else 569 m_adj(m, len - m->m_pkthdr.len); 570 } 571 572 /* 573 * Assume that we can create a fast-forward IP flow entry 574 * based on this packet. 575 */ 576 m->m_flags |= M_CANFASTFWD; 577 578 /* 579 * Run through list of hooks for input packets. If there are any 580 * filters which require that additional packets in the flow are 581 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 582 * Note that filters must _never_ set this flag, as another filter 583 * in the list may have previously cleared it. 584 */ 585 #if defined(IPSEC) 586 if (!ipsec_used || !ipsec_indone(m)) 587 #else 588 if (1) 589 #endif 590 { 591 struct in_addr odst = ip->ip_dst; 592 bool freed; 593 594 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 595 if (freed || m == NULL) { 596 m = NULL; 597 goto out; 598 } 599 ip = mtod(m, struct ip *); 600 hlen = ip->ip_hl << 2; 601 602 /* 603 * XXX The setting of "srcrt" here is to prevent ip_forward() 604 * from generating ICMP redirects for packets that have 605 * been redirected by a hook back out on to the same LAN that 606 * they came from and is not an indication that the packet 607 * is being inffluenced by source routing options. This 608 * allows things like 609 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 610 * where tlp0 is both on the 1.1.1.0/24 network and is the 611 * default route for hosts on 1.1.1.0/24. Of course this 612 * also requires a "map tlp0 ..." to complete the story. 613 * One might argue whether or not this kind of network config. 614 * should be supported in this manner... 615 */ 616 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 617 } 618 619 #ifdef ALTQ 620 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 621 if (altq_input) { 622 SOFTNET_LOCK(); 623 if ((*altq_input)(m, AF_INET) == 0) { 624 /* Packet dropped by traffic conditioner. */ 625 SOFTNET_UNLOCK(); 626 m = NULL; 627 goto out; 628 } 629 SOFTNET_UNLOCK(); 630 } 631 #endif 632 633 /* 634 * Process options and, if not destined for us, 635 * ship it on. ip_dooptions returns 1 when an 636 * error was detected (causing an icmp message 637 * to be sent and the original packet to be freed). 638 */ 639 if (hlen > sizeof(struct ip) && ip_dooptions(m)) { 640 m = NULL; 641 goto out; 642 } 643 644 /* 645 * Check our list of addresses, to see if the packet is for us. 646 * 647 * Traditional 4.4BSD did not consult IFF_UP at all. 648 * The behavior here is to treat addresses on !IFF_UP interface 649 * or IN_IFF_NOTREADY addresses as not mine. 650 */ 651 downmatch = 0; 652 s = pserialize_read_enter(); 653 ia = ip_match_our_address(ifp, ip, &downmatch); 654 if (ia != NULL) { 655 pserialize_read_exit(s); 656 goto ours; 657 } 658 659 if (ifp->if_flags & IFF_BROADCAST) { 660 ia = ip_match_our_address_broadcast(ifp, ip); 661 if (ia != NULL) { 662 pserialize_read_exit(s); 663 goto ours; 664 } 665 } 666 pserialize_read_exit(s); 667 668 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 669 #ifdef MROUTING 670 extern struct socket *ip_mrouter; 671 672 if (ip_mrouter) { 673 /* 674 * If we are acting as a multicast router, all 675 * incoming multicast packets are passed to the 676 * kernel-level multicast forwarding function. 677 * The packet is returned (relatively) intact; if 678 * ip_mforward() returns a non-zero value, the packet 679 * must be discarded, else it may be accepted below. 680 * 681 * (The IP ident field is put in the same byte order 682 * as expected when ip_mforward() is called from 683 * ip_output().) 684 */ 685 SOFTNET_LOCK(); 686 if (ip_mforward(m, ifp) != 0) { 687 SOFTNET_UNLOCK(); 688 IP_STATINC(IP_STAT_CANTFORWARD); 689 goto out; 690 } 691 SOFTNET_UNLOCK(); 692 693 /* 694 * The process-level routing demon needs to receive 695 * all multicast IGMP packets, whether or not this 696 * host belongs to their destination groups. 697 */ 698 if (ip->ip_p == IPPROTO_IGMP) { 699 goto ours; 700 } 701 IP_STATINC(IP_STAT_CANTFORWARD); 702 } 703 #endif 704 /* 705 * See if we belong to the destination multicast group on the 706 * arrival interface. 707 */ 708 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 709 IP_STATINC(IP_STAT_CANTFORWARD); 710 goto out; 711 } 712 goto ours; 713 } 714 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 715 in_nullhost(ip->ip_dst)) 716 goto ours; 717 718 /* 719 * Not for us; forward if possible and desirable. 720 */ 721 if (ipforwarding == 0) { 722 m_put_rcvif_psref(ifp, &psref); 723 IP_STATINC(IP_STAT_CANTFORWARD); 724 m_freem(m); 725 } else { 726 /* 727 * If ip_dst matched any of my address on !IFF_UP interface, 728 * and there's no IFF_UP interface that matches ip_dst, 729 * send icmp unreach. Forwarding it will result in in-kernel 730 * forwarding loop till TTL goes to 0. 731 */ 732 if (downmatch) { 733 m_put_rcvif_psref(ifp, &psref); 734 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 735 IP_STATINC(IP_STAT_CANTFORWARD); 736 return; 737 } 738 #ifdef IPSEC 739 /* Check the security policy (SP) for the packet */ 740 if (ipsec_used) { 741 if (ipsec4_input(m, IP_FORWARDING | 742 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 743 goto out; 744 } 745 } 746 #endif 747 ip_forward(m, srcrt, ifp); 748 m_put_rcvif_psref(ifp, &psref); 749 } 750 return; 751 752 ours: 753 m_put_rcvif_psref(ifp, &psref); 754 ifp = NULL; 755 756 /* 757 * If offset or IP_MF are set, must reassemble. 758 */ 759 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 760 /* 761 * Pass to IP reassembly mechanism. 762 */ 763 if (ip_reass_packet(&m, ip) != 0) { 764 /* Failed; invalid fragment(s) or packet. */ 765 goto out; 766 } 767 if (m == NULL) { 768 /* More fragments should come; silently return. */ 769 goto out; 770 } 771 /* 772 * Reassembly is done, we have the final packet. 773 * Update cached data in local variable(s). 774 */ 775 ip = mtod(m, struct ip *); 776 hlen = ip->ip_hl << 2; 777 } 778 779 #ifdef IPSEC 780 /* 781 * Enforce IPsec policy checking if we are seeing last header. 782 * Note that we do not visit this with protocols with PCB layer 783 * code - like UDP/TCP/raw IP. 784 */ 785 if (ipsec_used && 786 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 787 if (ipsec4_input(m, 0) != 0) { 788 goto out; 789 } 790 } 791 #endif 792 793 /* 794 * Switch out to protocol's input routine. 795 */ 796 #if IFA_STATS 797 if (ia && ip) { 798 struct in_ifaddr *_ia; 799 /* 800 * Keep a reference from ip_match_our_address with psref 801 * is expensive, so explore ia here again. 802 */ 803 s = pserialize_read_enter(); 804 _ia = in_get_ia(ip->ip_dst); 805 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 806 pserialize_read_exit(s); 807 } 808 #endif 809 IP_STATINC(IP_STAT_DELIVERED); 810 811 const int off = hlen, nh = ip->ip_p; 812 813 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 814 return; 815 816 out: 817 m_put_rcvif_psref(ifp, &psref); 818 if (m != NULL) 819 m_freem(m); 820 } 821 822 /* 823 * IP timer processing. 824 */ 825 void 826 ip_slowtimo(void) 827 { 828 829 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 830 831 ip_reass_slowtimo(); 832 833 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 834 } 835 836 /* 837 * IP drain processing. 838 */ 839 void 840 ip_drain(void) 841 { 842 843 KERNEL_LOCK(1, NULL); 844 ip_reass_drain(); 845 KERNEL_UNLOCK_ONE(NULL); 846 } 847 848 /* 849 * ip_dooptions: perform option processing on a datagram, possibly discarding 850 * it if bad options are encountered, or forwarding it if source-routed. 851 * 852 * => Returns true if packet has been forwarded/freed. 853 * => Returns false if the packet should be processed further. 854 */ 855 static bool 856 ip_dooptions(struct mbuf *m) 857 { 858 struct ip *ip = mtod(m, struct ip *); 859 u_char *cp, *cp0; 860 struct ip_timestamp *ipt; 861 struct in_ifaddr *ia; 862 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 863 int srr_present, rr_present, ts_present; 864 struct in_addr dst; 865 n_time ntime; 866 struct ifaddr *ifa = NULL; 867 int s; 868 869 srr_present = 0; 870 rr_present = 0; 871 ts_present = 0; 872 873 dst = ip->ip_dst; 874 cp = (u_char *)(ip + 1); 875 cnt = (ip->ip_hl << 2) - sizeof(struct ip); 876 for (; cnt > 0; cnt -= optlen, cp += optlen) { 877 opt = cp[IPOPT_OPTVAL]; 878 if (opt == IPOPT_EOL) 879 break; 880 if (opt == IPOPT_NOP) 881 optlen = 1; 882 else { 883 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 884 code = &cp[IPOPT_OLEN] - (u_char *)ip; 885 goto bad; 886 } 887 optlen = cp[IPOPT_OLEN]; 888 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 889 code = &cp[IPOPT_OLEN] - (u_char *)ip; 890 goto bad; 891 } 892 } 893 switch (opt) { 894 895 default: 896 break; 897 898 /* 899 * Source routing with record. 900 * Find interface with current destination address. 901 * If none on this machine then drop if strictly routed, 902 * or do nothing if loosely routed. 903 * Record interface address and bring up next address 904 * component. If strictly routed make sure next 905 * address is on directly accessible net. 906 */ 907 case IPOPT_LSRR: 908 case IPOPT_SSRR: { 909 struct psref psref; 910 struct sockaddr_in ipaddr = { 911 .sin_len = sizeof(ipaddr), 912 .sin_family = AF_INET, 913 }; 914 915 if (ip_allowsrcrt == 0) { 916 type = ICMP_UNREACH; 917 code = ICMP_UNREACH_NET_PROHIB; 918 goto bad; 919 } 920 if (srr_present++) { 921 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 922 goto bad; 923 } 924 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 925 code = &cp[IPOPT_OLEN] - (u_char *)ip; 926 goto bad; 927 } 928 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 929 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 930 goto bad; 931 } 932 ipaddr.sin_addr = ip->ip_dst; 933 934 s = pserialize_read_enter(); 935 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 936 if (ifa == NULL) { 937 pserialize_read_exit(s); 938 if (opt == IPOPT_SSRR) { 939 type = ICMP_UNREACH; 940 code = ICMP_UNREACH_SRCFAIL; 941 goto bad; 942 } 943 /* 944 * Loose routing, and not at next destination 945 * yet; nothing to do except forward. 946 */ 947 break; 948 } 949 pserialize_read_exit(s); 950 951 off--; /* 0 origin */ 952 if ((off + sizeof(struct in_addr)) > optlen) { 953 /* 954 * End of source route. Should be for us. 955 */ 956 save_rte(m, cp, ip->ip_src); 957 break; 958 } 959 /* 960 * locate outgoing interface 961 */ 962 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 963 sizeof(ipaddr.sin_addr)); 964 if (opt == IPOPT_SSRR) { 965 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr), 966 &psref); 967 if (ifa != NULL) 968 ia = ifatoia(ifa); 969 else 970 ia = NULL; 971 } else { 972 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 973 } 974 if (ia == NULL) { 975 type = ICMP_UNREACH; 976 code = ICMP_UNREACH_SRCFAIL; 977 goto bad; 978 } 979 ip->ip_dst = ipaddr.sin_addr; 980 bcopy((void *)&ia->ia_addr.sin_addr, 981 (void *)(cp + off), sizeof(struct in_addr)); 982 ia4_release(ia, &psref); 983 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 984 /* 985 * Let ip_intr's mcast routing check handle mcast pkts 986 */ 987 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 988 break; 989 } 990 991 case IPOPT_RR: { 992 struct psref psref; 993 struct sockaddr_in ipaddr = { 994 .sin_len = sizeof(ipaddr), 995 .sin_family = AF_INET, 996 }; 997 998 if (rr_present++) { 999 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1000 goto bad; 1001 } 1002 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1003 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1004 goto bad; 1005 } 1006 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1007 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1008 goto bad; 1009 } 1010 /* 1011 * If no space remains, ignore. 1012 */ 1013 off--; /* 0 origin */ 1014 if ((off + sizeof(struct in_addr)) > optlen) 1015 break; 1016 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 1017 sizeof(ipaddr.sin_addr)); 1018 /* 1019 * locate outgoing interface; if we're the destination, 1020 * use the incoming interface (should be same). 1021 */ 1022 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref); 1023 if (ifa == NULL) { 1024 ia = ip_rtaddr(ipaddr.sin_addr, &psref); 1025 if (ia == NULL) { 1026 type = ICMP_UNREACH; 1027 code = ICMP_UNREACH_HOST; 1028 goto bad; 1029 } 1030 } else { 1031 ia = ifatoia(ifa); 1032 } 1033 bcopy((void *)&ia->ia_addr.sin_addr, 1034 (void *)(cp + off), sizeof(struct in_addr)); 1035 ia4_release(ia, &psref); 1036 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1037 break; 1038 } 1039 1040 case IPOPT_TS: 1041 code = cp - (u_char *)ip; 1042 ipt = (struct ip_timestamp *)cp; 1043 if (ts_present++) { 1044 code = &cp[IPOPT_OPTVAL] - (u_char *)ip; 1045 goto bad; 1046 } 1047 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 1048 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 1049 goto bad; 1050 } 1051 if (ipt->ipt_ptr < 5) { 1052 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 1053 goto bad; 1054 } 1055 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) { 1056 if (++ipt->ipt_oflw == 0) { 1057 code = (u_char *)&ipt->ipt_ptr - 1058 (u_char *)ip; 1059 goto bad; 1060 } 1061 break; 1062 } 1063 cp0 = (cp + ipt->ipt_ptr - 1); 1064 switch (ipt->ipt_flg) { 1065 1066 case IPOPT_TS_TSONLY: 1067 break; 1068 1069 case IPOPT_TS_TSANDADDR: { 1070 struct ifnet *rcvif; 1071 int _s, _ss; 1072 struct sockaddr_in ipaddr = { 1073 .sin_len = sizeof(ipaddr), 1074 .sin_family = AF_INET, 1075 }; 1076 1077 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1078 sizeof(struct in_addr) > ipt->ipt_len) { 1079 code = (u_char *)&ipt->ipt_ptr - 1080 (u_char *)ip; 1081 goto bad; 1082 } 1083 ipaddr.sin_addr = dst; 1084 _ss = pserialize_read_enter(); 1085 rcvif = m_get_rcvif(m, &_s); 1086 if (__predict_true(rcvif != NULL)) { 1087 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), 1088 rcvif); 1089 } 1090 m_put_rcvif(rcvif, &_s); 1091 if (ifa == NULL) { 1092 pserialize_read_exit(_ss); 1093 break; 1094 } 1095 ia = ifatoia(ifa); 1096 bcopy(&ia->ia_addr.sin_addr, 1097 cp0, sizeof(struct in_addr)); 1098 pserialize_read_exit(_ss); 1099 ipt->ipt_ptr += sizeof(struct in_addr); 1100 break; 1101 } 1102 1103 case IPOPT_TS_PRESPEC: { 1104 struct sockaddr_in ipaddr = { 1105 .sin_len = sizeof(ipaddr), 1106 .sin_family = AF_INET, 1107 }; 1108 1109 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1110 sizeof(struct in_addr) > ipt->ipt_len) { 1111 code = (u_char *)&ipt->ipt_ptr - 1112 (u_char *)ip; 1113 goto bad; 1114 } 1115 memcpy(&ipaddr.sin_addr, cp0, 1116 sizeof(struct in_addr)); 1117 s = pserialize_read_enter(); 1118 ifa = ifa_ifwithaddr(sintosa(&ipaddr)); 1119 if (ifa == NULL) { 1120 pserialize_read_exit(s); 1121 continue; 1122 } 1123 pserialize_read_exit(s); 1124 ipt->ipt_ptr += sizeof(struct in_addr); 1125 break; 1126 } 1127 1128 default: 1129 /* XXX can't take &ipt->ipt_flg */ 1130 code = (u_char *)&ipt->ipt_ptr - 1131 (u_char *)ip + 1; 1132 goto bad; 1133 } 1134 ntime = iptime(); 1135 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1136 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1137 sizeof(n_time)); 1138 ipt->ipt_ptr += sizeof(n_time); 1139 } 1140 } 1141 if (forward) { 1142 struct ifnet *rcvif; 1143 struct psref _psref; 1144 1145 if (ip_forwsrcrt == 0) { 1146 type = ICMP_UNREACH; 1147 code = ICMP_UNREACH_SRCFAIL; 1148 goto bad; 1149 } 1150 1151 rcvif = m_get_rcvif_psref(m, &_psref); 1152 if (__predict_false(rcvif == NULL)) { 1153 type = ICMP_UNREACH; 1154 code = ICMP_UNREACH_HOST; 1155 goto bad; 1156 } 1157 ip_forward(m, 1, rcvif); 1158 m_put_rcvif_psref(rcvif, &_psref); 1159 return true; 1160 } 1161 return false; 1162 bad: 1163 icmp_error(m, type, code, 0, 0); 1164 IP_STATINC(IP_STAT_BADOPTIONS); 1165 return true; 1166 } 1167 1168 /* 1169 * ip_rtaddr: given address of next destination (final or next hop), 1170 * return internet address info of interface to be used to get there. 1171 */ 1172 static struct in_ifaddr * 1173 ip_rtaddr(struct in_addr dst, struct psref *psref) 1174 { 1175 struct rtentry *rt; 1176 union { 1177 struct sockaddr dst; 1178 struct sockaddr_in dst4; 1179 } u; 1180 struct route *ro; 1181 1182 sockaddr_in_init(&u.dst4, &dst, 0); 1183 1184 ro = percpu_getref(ipforward_rt_percpu); 1185 rt = rtcache_lookup(ro, &u.dst); 1186 if (rt == NULL) { 1187 percpu_putref(ipforward_rt_percpu); 1188 return NULL; 1189 } 1190 1191 ia4_acquire(ifatoia(rt->rt_ifa), psref); 1192 rtcache_unref(rt, ro); 1193 percpu_putref(ipforward_rt_percpu); 1194 1195 return ifatoia(rt->rt_ifa); 1196 } 1197 1198 /* 1199 * save_rte: save incoming source route for use in replies, to be picked 1200 * up later by ip_srcroute if the receiver is interested. 1201 */ 1202 static void 1203 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1204 { 1205 struct ip_srcrt *isr; 1206 struct m_tag *mtag; 1207 unsigned olen; 1208 1209 olen = option[IPOPT_OLEN]; 1210 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1211 return; 1212 1213 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1214 if (mtag == NULL) 1215 return; 1216 isr = (struct ip_srcrt *)(mtag + 1); 1217 1218 memcpy(isr->isr_hdr, option, olen); 1219 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1220 isr->isr_dst = dst; 1221 m_tag_prepend(m, mtag); 1222 } 1223 1224 /* 1225 * Retrieve incoming source route for use in replies, 1226 * in the same form used by setsockopt. 1227 * The first hop is placed before the options, will be removed later. 1228 */ 1229 struct mbuf * 1230 ip_srcroute(struct mbuf *m0) 1231 { 1232 struct in_addr *p, *q; 1233 struct mbuf *m; 1234 struct ip_srcrt *isr; 1235 struct m_tag *mtag; 1236 1237 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL); 1238 if (mtag == NULL) 1239 return NULL; 1240 isr = (struct ip_srcrt *)(mtag + 1); 1241 1242 if (isr->isr_nhops == 0) 1243 return NULL; 1244 1245 m = m_get(M_DONTWAIT, MT_SOOPTS); 1246 if (m == NULL) 1247 return NULL; 1248 1249 MCLAIM(m, &inetdomain.dom_mowner); 1250 #define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1251 1252 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1253 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1254 1255 /* 1256 * First save first hop for return route 1257 */ 1258 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1259 *(mtod(m, struct in_addr *)) = *p--; 1260 1261 /* 1262 * Copy option fields and padding (nop) to mbuf. 1263 */ 1264 isr->isr_nop = IPOPT_NOP; 1265 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1266 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop, 1267 OPTSIZ); 1268 q = (struct in_addr *)(mtod(m, char *) + 1269 sizeof(struct in_addr) + OPTSIZ); 1270 #undef OPTSIZ 1271 /* 1272 * Record return path as an IP source route, 1273 * reversing the path (pointers are now aligned). 1274 */ 1275 while (p >= isr->isr_routes) { 1276 *q++ = *p--; 1277 } 1278 /* 1279 * Last hop goes to final destination. 1280 */ 1281 *q = isr->isr_dst; 1282 m_tag_delete(m0, mtag); 1283 return m; 1284 } 1285 1286 const int inetctlerrmap[PRC_NCMDS] = { 1287 [PRC_MSGSIZE] = EMSGSIZE, 1288 [PRC_HOSTDEAD] = EHOSTDOWN, 1289 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1290 [PRC_UNREACH_NET] = EHOSTUNREACH, 1291 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1292 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1293 [PRC_UNREACH_PORT] = ECONNREFUSED, 1294 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1295 [PRC_PARAMPROB] = ENOPROTOOPT, 1296 }; 1297 1298 void 1299 ip_fasttimo(void) 1300 { 1301 if (ip_drainwanted) { 1302 ip_drain(); 1303 ip_drainwanted = 0; 1304 } 1305 } 1306 1307 void 1308 ip_drainstub(void) 1309 { 1310 ip_drainwanted = 1; 1311 } 1312 1313 /* 1314 * Forward a packet. If some error occurs return the sender 1315 * an icmp packet. Note we can't always generate a meaningful 1316 * icmp message because icmp doesn't have a large enough repertoire 1317 * of codes and types. 1318 * 1319 * If not forwarding, just drop the packet. This could be confusing 1320 * if ipforwarding was zero but some routing protocol was advancing 1321 * us as a gateway to somewhere. However, we must let the routing 1322 * protocol deal with that. 1323 * 1324 * The srcrt parameter indicates whether the packet is being forwarded 1325 * via a source route. 1326 */ 1327 static void 1328 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1329 { 1330 struct ip *ip = mtod(m, struct ip *); 1331 struct rtentry *rt; 1332 int error, type = 0, code = 0, destmtu = 0; 1333 struct mbuf *mcopy; 1334 n_long dest; 1335 union { 1336 struct sockaddr dst; 1337 struct sockaddr_in dst4; 1338 } u; 1339 uint64_t *ips; 1340 struct route *ro; 1341 1342 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1343 "interrupt handler; synchronization assumptions violated"); 1344 1345 /* 1346 * We are now in the output path. 1347 */ 1348 MCLAIM(m, &ip_tx_mowner); 1349 1350 /* 1351 * Clear any in-bound checksum flags for this packet. 1352 */ 1353 m->m_pkthdr.csum_flags = 0; 1354 1355 dest = 0; 1356 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1357 IP_STATINC(IP_STAT_CANTFORWARD); 1358 m_freem(m); 1359 return; 1360 } 1361 1362 if (ip->ip_ttl <= IPTTLDEC) { 1363 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1364 return; 1365 } 1366 1367 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1368 1369 ro = percpu_getref(ipforward_rt_percpu); 1370 rt = rtcache_lookup(ro, &u.dst); 1371 if (rt == NULL) { 1372 percpu_putref(ipforward_rt_percpu); 1373 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1374 return; 1375 } 1376 1377 /* 1378 * Save at most 68 bytes of the packet in case 1379 * we need to generate an ICMP message to the src. 1380 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1381 */ 1382 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1383 if (mcopy) 1384 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1385 1386 ip->ip_ttl -= IPTTLDEC; 1387 1388 /* 1389 * If forwarding packet using same interface that it came in on, 1390 * perhaps should send a redirect to sender to shortcut a hop. 1391 * Only send redirect if source is sending directly to us, 1392 * and if packet was not source routed (or has any options). 1393 * Also, don't send redirect if forwarding using a default route 1394 * or a route modified by a redirect. 1395 */ 1396 if (rt->rt_ifp == rcvif && 1397 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1398 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1399 ipsendredirects && !srcrt) { 1400 if (rt->rt_ifa && 1401 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1402 ifatoia(rt->rt_ifa)->ia_subnet) { 1403 if (rt->rt_flags & RTF_GATEWAY) 1404 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1405 else 1406 dest = ip->ip_dst.s_addr; 1407 /* 1408 * Router requirements says to only send host 1409 * redirects. 1410 */ 1411 type = ICMP_REDIRECT; 1412 code = ICMP_REDIRECT_HOST; 1413 } 1414 } 1415 rtcache_unref(rt, ro); 1416 1417 error = ip_output(m, NULL, ro, 1418 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1419 NULL, NULL); 1420 1421 if (error) { 1422 IP_STATINC(IP_STAT_CANTFORWARD); 1423 goto error; 1424 } 1425 1426 ips = IP_STAT_GETREF(); 1427 ips[IP_STAT_FORWARD]++; 1428 1429 if (type) { 1430 ips[IP_STAT_REDIRECTSENT]++; 1431 IP_STAT_PUTREF(); 1432 goto redirect; 1433 } 1434 1435 IP_STAT_PUTREF(); 1436 if (mcopy) { 1437 #ifdef GATEWAY 1438 if (mcopy->m_flags & M_CANFASTFWD) 1439 ipflow_create(ro, mcopy); 1440 #endif 1441 m_freem(mcopy); 1442 } 1443 1444 percpu_putref(ipforward_rt_percpu); 1445 return; 1446 1447 redirect: 1448 error: 1449 if (mcopy == NULL) { 1450 percpu_putref(ipforward_rt_percpu); 1451 return; 1452 } 1453 1454 switch (error) { 1455 1456 case 0: /* forwarded, but need redirect */ 1457 /* type, code set above */ 1458 break; 1459 1460 case ENETUNREACH: /* shouldn't happen, checked above */ 1461 case EHOSTUNREACH: 1462 case ENETDOWN: 1463 case EHOSTDOWN: 1464 default: 1465 type = ICMP_UNREACH; 1466 code = ICMP_UNREACH_HOST; 1467 break; 1468 1469 case EMSGSIZE: 1470 type = ICMP_UNREACH; 1471 code = ICMP_UNREACH_NEEDFRAG; 1472 1473 if ((rt = rtcache_validate(ro)) != NULL) { 1474 destmtu = rt->rt_ifp->if_mtu; 1475 rtcache_unref(rt, ro); 1476 } 1477 #ifdef IPSEC 1478 if (ipsec_used) 1479 (void)ipsec4_forward(mcopy, &destmtu); 1480 #endif 1481 IP_STATINC(IP_STAT_CANTFRAG); 1482 break; 1483 1484 case ENOBUFS: 1485 /* 1486 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1487 * Requirements for IP Version 4 Routers. Source quench can 1488 * big problem under DoS attacks or if the underlying 1489 * interface is rate-limited. 1490 */ 1491 if (mcopy) 1492 m_freem(mcopy); 1493 percpu_putref(ipforward_rt_percpu); 1494 return; 1495 } 1496 icmp_error(mcopy, type, code, dest, destmtu); 1497 percpu_putref(ipforward_rt_percpu); 1498 } 1499 1500 void 1501 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1502 struct mbuf *m) 1503 { 1504 struct socket *so = inp->inp_socket; 1505 int inpflags = inp->inp_flags; 1506 1507 if (SOOPT_TIMESTAMP(so->so_options)) 1508 mp = sbsavetimestamp(so->so_options, m, mp); 1509 1510 if (inpflags & INP_RECVDSTADDR) { 1511 *mp = sbcreatecontrol(&ip->ip_dst, 1512 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1513 if (*mp) 1514 mp = &(*mp)->m_next; 1515 } 1516 1517 if (inpflags & INP_RECVTTL) { 1518 *mp = sbcreatecontrol(&ip->ip_ttl, 1519 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1520 if (*mp) 1521 mp = &(*mp)->m_next; 1522 } 1523 1524 struct psref psref; 1525 ifnet_t *ifp = m_get_rcvif_psref(m, &psref); 1526 if (__predict_false(ifp == NULL)) { 1527 #ifdef DIAGNOSTIC 1528 printf("%s: missing receive interface\n", __func__); 1529 #endif 1530 return; /* XXX should report error? */ 1531 } 1532 1533 if (inpflags & INP_RECVPKTINFO) { 1534 struct in_pktinfo ipi; 1535 ipi.ipi_addr = ip->ip_dst; 1536 ipi.ipi_ifindex = ifp->if_index; 1537 *mp = sbcreatecontrol(&ipi, 1538 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1539 if (*mp) 1540 mp = &(*mp)->m_next; 1541 } 1542 if (inpflags & INP_RECVIF) { 1543 struct sockaddr_dl sdl; 1544 1545 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0, 1546 NULL, 0); 1547 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1548 if (*mp) 1549 mp = &(*mp)->m_next; 1550 } 1551 m_put_rcvif_psref(ifp, &psref); 1552 } 1553 1554 /* 1555 * sysctl helper routine for net.inet.ip.forwsrcrt. 1556 */ 1557 static int 1558 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1559 { 1560 int error, tmp; 1561 struct sysctlnode node; 1562 1563 node = *rnode; 1564 tmp = ip_forwsrcrt; 1565 node.sysctl_data = &tmp; 1566 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1567 if (error || newp == NULL) 1568 return (error); 1569 1570 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1571 0, NULL, NULL, NULL); 1572 if (error) 1573 return (error); 1574 1575 ip_forwsrcrt = tmp; 1576 1577 return (0); 1578 } 1579 1580 /* 1581 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1582 * range of the new value and tweaks timers if it changes. 1583 */ 1584 static int 1585 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1586 { 1587 int error, tmp; 1588 struct sysctlnode node; 1589 1590 icmp_mtudisc_lock(); 1591 1592 node = *rnode; 1593 tmp = ip_mtudisc_timeout; 1594 node.sysctl_data = &tmp; 1595 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1596 if (error || newp == NULL) 1597 goto out; 1598 if (tmp < 0) { 1599 error = EINVAL; 1600 goto out; 1601 } 1602 1603 ip_mtudisc_timeout = tmp; 1604 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1605 error = 0; 1606 out: 1607 icmp_mtudisc_unlock(); 1608 return error; 1609 } 1610 1611 static int 1612 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1613 { 1614 1615 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1616 } 1617 1618 static void 1619 sysctl_net_inet_ip_setup(struct sysctllog **clog) 1620 { 1621 sysctl_createv(clog, 0, NULL, NULL, 1622 CTLFLAG_PERMANENT, 1623 CTLTYPE_NODE, "inet", 1624 SYSCTL_DESCR("PF_INET related settings"), 1625 NULL, 0, NULL, 0, 1626 CTL_NET, PF_INET, CTL_EOL); 1627 sysctl_createv(clog, 0, NULL, NULL, 1628 CTLFLAG_PERMANENT, 1629 CTLTYPE_NODE, "ip", 1630 SYSCTL_DESCR("IPv4 related settings"), 1631 NULL, 0, NULL, 0, 1632 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1633 1634 sysctl_createv(clog, 0, NULL, NULL, 1635 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1636 CTLTYPE_INT, "forwarding", 1637 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1638 NULL, 0, &ipforwarding, 0, 1639 CTL_NET, PF_INET, IPPROTO_IP, 1640 IPCTL_FORWARDING, CTL_EOL); 1641 sysctl_createv(clog, 0, NULL, NULL, 1642 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1643 CTLTYPE_INT, "redirect", 1644 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1645 NULL, 0, &ipsendredirects, 0, 1646 CTL_NET, PF_INET, IPPROTO_IP, 1647 IPCTL_SENDREDIRECTS, CTL_EOL); 1648 sysctl_createv(clog, 0, NULL, NULL, 1649 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1650 CTLTYPE_INT, "ttl", 1651 SYSCTL_DESCR("Default TTL for an INET datagram"), 1652 NULL, 0, &ip_defttl, 0, 1653 CTL_NET, PF_INET, IPPROTO_IP, 1654 IPCTL_DEFTTL, CTL_EOL); 1655 #ifdef IPCTL_DEFMTU 1656 sysctl_createv(clog, 0, NULL, NULL, 1657 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1658 CTLTYPE_INT, "mtu", 1659 SYSCTL_DESCR("Default MTA for an INET route"), 1660 NULL, 0, &ip_mtu, 0, 1661 CTL_NET, PF_INET, IPPROTO_IP, 1662 IPCTL_DEFMTU, CTL_EOL); 1663 #endif /* IPCTL_DEFMTU */ 1664 sysctl_createv(clog, 0, NULL, NULL, 1665 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1666 CTLTYPE_INT, "forwsrcrt", 1667 SYSCTL_DESCR("Enable forwarding of source-routed " 1668 "datagrams"), 1669 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1670 CTL_NET, PF_INET, IPPROTO_IP, 1671 IPCTL_FORWSRCRT, CTL_EOL); 1672 sysctl_createv(clog, 0, NULL, NULL, 1673 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1674 CTLTYPE_INT, "directed-broadcast", 1675 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1676 NULL, 0, &ip_directedbcast, 0, 1677 CTL_NET, PF_INET, IPPROTO_IP, 1678 IPCTL_DIRECTEDBCAST, CTL_EOL); 1679 sysctl_createv(clog, 0, NULL, NULL, 1680 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1681 CTLTYPE_INT, "allowsrcrt", 1682 SYSCTL_DESCR("Accept source-routed datagrams"), 1683 NULL, 0, &ip_allowsrcrt, 0, 1684 CTL_NET, PF_INET, IPPROTO_IP, 1685 IPCTL_ALLOWSRCRT, CTL_EOL); 1686 1687 sysctl_createv(clog, 0, NULL, NULL, 1688 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1689 CTLTYPE_INT, "mtudisc", 1690 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1691 NULL, 0, &ip_mtudisc, 0, 1692 CTL_NET, PF_INET, IPPROTO_IP, 1693 IPCTL_MTUDISC, CTL_EOL); 1694 sysctl_createv(clog, 0, NULL, NULL, 1695 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1696 CTLTYPE_INT, "anonportmin", 1697 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1698 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1699 CTL_NET, PF_INET, IPPROTO_IP, 1700 IPCTL_ANONPORTMIN, CTL_EOL); 1701 sysctl_createv(clog, 0, NULL, NULL, 1702 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1703 CTLTYPE_INT, "anonportmax", 1704 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1705 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1706 CTL_NET, PF_INET, IPPROTO_IP, 1707 IPCTL_ANONPORTMAX, CTL_EOL); 1708 sysctl_createv(clog, 0, NULL, NULL, 1709 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1710 CTLTYPE_INT, "mtudisctimeout", 1711 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1712 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1713 CTL_NET, PF_INET, IPPROTO_IP, 1714 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1715 #ifndef IPNOPRIVPORTS 1716 sysctl_createv(clog, 0, NULL, NULL, 1717 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1718 CTLTYPE_INT, "lowportmin", 1719 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1720 "to assign"), 1721 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1722 CTL_NET, PF_INET, IPPROTO_IP, 1723 IPCTL_LOWPORTMIN, CTL_EOL); 1724 sysctl_createv(clog, 0, NULL, NULL, 1725 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1726 CTLTYPE_INT, "lowportmax", 1727 SYSCTL_DESCR("Highest privileged ephemeral port number " 1728 "to assign"), 1729 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1730 CTL_NET, PF_INET, IPPROTO_IP, 1731 IPCTL_LOWPORTMAX, CTL_EOL); 1732 #endif /* IPNOPRIVPORTS */ 1733 #if NGRE > 0 1734 sysctl_createv(clog, 0, NULL, NULL, 1735 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1736 CTLTYPE_INT, "grettl", 1737 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1738 NULL, 0, &ip_gre_ttl, 0, 1739 CTL_NET, PF_INET, IPPROTO_IP, 1740 IPCTL_GRE_TTL, CTL_EOL); 1741 #endif /* NGRE */ 1742 sysctl_createv(clog, 0, NULL, NULL, 1743 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1744 CTLTYPE_INT, "checkinterface", 1745 SYSCTL_DESCR("Enable receive side of Strong ES model " 1746 "from RFC1122"), 1747 NULL, 0, &ip_checkinterface, 0, 1748 CTL_NET, PF_INET, IPPROTO_IP, 1749 IPCTL_CHECKINTERFACE, CTL_EOL); 1750 sysctl_createv(clog, 0, NULL, NULL, 1751 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1752 CTLTYPE_INT, "random_id", 1753 SYSCTL_DESCR("Assign random ip_id values"), 1754 NULL, 0, &ip_do_randomid, 0, 1755 CTL_NET, PF_INET, IPPROTO_IP, 1756 IPCTL_RANDOMID, CTL_EOL); 1757 sysctl_createv(clog, 0, NULL, NULL, 1758 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1759 CTLTYPE_INT, "do_loopback_cksum", 1760 SYSCTL_DESCR("Perform IP checksum on loopback"), 1761 NULL, 0, &ip_do_loopback_cksum, 0, 1762 CTL_NET, PF_INET, IPPROTO_IP, 1763 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1764 sysctl_createv(clog, 0, NULL, NULL, 1765 CTLFLAG_PERMANENT, 1766 CTLTYPE_STRUCT, "stats", 1767 SYSCTL_DESCR("IP statistics"), 1768 sysctl_net_inet_ip_stats, 0, NULL, 0, 1769 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1770 CTL_EOL); 1771 #if NARP 1772 sysctl_createv(clog, 0, NULL, NULL, 1773 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1774 CTLTYPE_INT, "dad_count", 1775 SYSCTL_DESCR("Number of Duplicate Address Detection " 1776 "probes to send"), 1777 NULL, 0, &ip_dad_count, 0, 1778 CTL_NET, PF_INET, IPPROTO_IP, 1779 IPCTL_DAD_COUNT, CTL_EOL); 1780 #endif 1781 1782 /* anonportalgo RFC6056 subtree */ 1783 const struct sysctlnode *portalgo_node; 1784 sysctl_createv(clog, 0, NULL, &portalgo_node, 1785 CTLFLAG_PERMANENT, 1786 CTLTYPE_NODE, "anonportalgo", 1787 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1788 NULL, 0, NULL, 0, 1789 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1790 sysctl_createv(clog, 0, &portalgo_node, NULL, 1791 CTLFLAG_PERMANENT, 1792 CTLTYPE_STRING, "available", 1793 SYSCTL_DESCR("available algorithms"), 1794 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1795 CTL_CREATE, CTL_EOL); 1796 sysctl_createv(clog, 0, &portalgo_node, NULL, 1797 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1798 CTLTYPE_STRING, "selected", 1799 SYSCTL_DESCR("selected algorithm"), 1800 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1801 CTL_CREATE, CTL_EOL); 1802 sysctl_createv(clog, 0, &portalgo_node, NULL, 1803 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1804 CTLTYPE_STRUCT, "reserve", 1805 SYSCTL_DESCR("bitmap of reserved ports"), 1806 sysctl_portalgo_reserve4, 0, NULL, 0, 1807 CTL_CREATE, CTL_EOL); 1808 } 1809 1810 void 1811 ip_statinc(u_int stat) 1812 { 1813 1814 KASSERT(stat < IP_NSTATS); 1815 IP_STATINC(stat); 1816 } 1817
Indexes created Thu Sep 25 16:09:42 GMT 2025