ip_input.c revision 1.330
1 /* $NetBSD: ip_input.c,v 1.330 2016/06/28 02:02:56 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 1.330 2016/06/28 02:02:56 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 /* XXX: Not yet enabled. */ 293 #define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock)) 294 #define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock)) 295 296 /* 297 * IP initialization: fill in IP protocol switch table. 298 * All protocols not implemented in kernel go to raw IP protocol handler. 299 */ 300 void 301 ip_init(void) 302 { 303 const struct protosw *pr; 304 305 in_init(); 306 sysctl_net_inet_ip_setup(NULL); 307 308 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 309 KASSERT(pr != NULL); 310 311 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL); 312 KASSERT(ip_pktq != NULL); 313 314 for (u_int i = 0; i < IPPROTO_MAX; i++) { 315 ip_protox[i] = pr - inetsw; 316 } 317 for (pr = inetdomain.dom_protosw; 318 pr < inetdomain.dom_protoswNPROTOSW; pr++) 319 if (pr->pr_domain->dom_family == PF_INET && 320 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 321 ip_protox[pr->pr_protocol] = pr - inetsw; 322 323 ip_reass_init(); 324 325 ip_ids = ip_id_init(); 326 ip_id = time_uptime & 0xfffff; 327 328 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); 329 #ifdef GATEWAY 330 ipflow_init(); 331 #endif 332 333 /* Register our Packet Filter hook. */ 334 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET); 335 KASSERT(inet_pfil_hook != NULL); 336 337 #ifdef MBUFTRACE 338 MOWNER_ATTACH(&ip_tx_mowner); 339 MOWNER_ATTACH(&ip_rx_mowner); 340 #endif /* MBUFTRACE */ 341 342 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS); 343 } 344 345 /* 346 * IP software interrupt routine. 347 */ 348 static void 349 ipintr(void *arg __unused) 350 { 351 struct mbuf *m; 352 353 KASSERT(cpu_softintr_p()); 354 355 mutex_enter(softnet_lock); 356 while ((m = pktq_dequeue(ip_pktq)) != NULL) { 357 ip_input(m); 358 } 359 mutex_exit(softnet_lock); 360 } 361 362 /* 363 * IP input routine. Checksum and byte swap header. If fragmented 364 * try to reassemble. Process options. Pass to next level. 365 */ 366 static void 367 ip_input(struct mbuf *m) 368 { 369 struct ip *ip = NULL; 370 struct in_ifaddr *ia; 371 struct ifaddr *ifa; 372 int hlen = 0, len; 373 int downmatch; 374 int checkif; 375 int srcrt = 0; 376 ifnet_t *ifp; 377 struct psref psref; 378 379 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software " 380 "interrupt handler; synchronization assumptions violated"); 381 382 MCLAIM(m, &ip_rx_mowner); 383 KASSERT((m->m_flags & M_PKTHDR) != 0); 384 385 ifp = m_get_rcvif_psref(m, &psref); 386 if (__predict_false(ifp == NULL)) 387 goto bad; 388 389 /* 390 * If no IP addresses have been set yet but the interfaces 391 * are receiving, can't do anything with incoming packets yet. 392 * Note: we pre-check without locks held. 393 */ 394 if (!TAILQ_FIRST(&in_ifaddrhead)) { 395 goto bad; 396 } 397 IP_STATINC(IP_STAT_TOTAL); 398 399 /* 400 * If the IP header is not aligned, slurp it up into a new 401 * mbuf with space for link headers, in the event we forward 402 * it. Otherwise, if it is aligned, make sure the entire 403 * base IP header is in the first mbuf of the chain. 404 */ 405 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 406 if ((m = m_copyup(m, sizeof(struct ip), 407 (max_linkhdr + 3) & ~3)) == NULL) { 408 /* XXXJRT new stat, please */ 409 IP_STATINC(IP_STAT_TOOSMALL); 410 goto out; 411 } 412 } else if (__predict_false(m->m_len < sizeof (struct ip))) { 413 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) { 414 IP_STATINC(IP_STAT_TOOSMALL); 415 goto out; 416 } 417 } 418 ip = mtod(m, struct ip *); 419 if (ip->ip_v != IPVERSION) { 420 IP_STATINC(IP_STAT_BADVERS); 421 goto bad; 422 } 423 hlen = ip->ip_hl << 2; 424 if (hlen < sizeof(struct ip)) { /* minimum header length */ 425 IP_STATINC(IP_STAT_BADHLEN); 426 goto bad; 427 } 428 if (hlen > m->m_len) { 429 if ((m = m_pullup(m, hlen)) == NULL) { 430 IP_STATINC(IP_STAT_BADHLEN); 431 goto out; 432 } 433 ip = mtod(m, struct ip *); 434 } 435 436 /* 437 * RFC1122: packets with a multicast source address are 438 * not allowed. 439 */ 440 if (IN_MULTICAST(ip->ip_src.s_addr)) { 441 IP_STATINC(IP_STAT_BADADDR); 442 goto bad; 443 } 444 445 /* 127/8 must not appear on wire - RFC1122 */ 446 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 447 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 448 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 449 IP_STATINC(IP_STAT_BADADDR); 450 goto bad; 451 } 452 } 453 454 switch (m->m_pkthdr.csum_flags & 455 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | 456 M_CSUM_IPv4_BAD)) { 457 case M_CSUM_IPv4|M_CSUM_IPv4_BAD: 458 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); 459 goto badcsum; 460 461 case M_CSUM_IPv4: 462 /* Checksum was okay. */ 463 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); 464 break; 465 466 default: 467 /* 468 * Must compute it ourselves. Maybe skip checksum on 469 * loopback interfaces. 470 */ 471 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) || 472 ip_do_loopback_cksum)) { 473 INET_CSUM_COUNTER_INCR(&ip_swcsum); 474 if (in_cksum(m, hlen) != 0) 475 goto badcsum; 476 } 477 break; 478 } 479 480 /* Retrieve the packet length. */ 481 len = ntohs(ip->ip_len); 482 483 /* 484 * Check for additional length bogosity 485 */ 486 if (len < hlen) { 487 IP_STATINC(IP_STAT_BADLEN); 488 goto bad; 489 } 490 491 /* 492 * Check that the amount of data in the buffers 493 * is as at least much as the IP header would have us expect. 494 * Trim mbufs if longer than we expect. 495 * Drop packet if shorter than we expect. 496 */ 497 if (m->m_pkthdr.len < len) { 498 IP_STATINC(IP_STAT_TOOSHORT); 499 goto bad; 500 } 501 if (m->m_pkthdr.len > len) { 502 if (m->m_len == m->m_pkthdr.len) { 503 m->m_len = len; 504 m->m_pkthdr.len = len; 505 } else 506 m_adj(m, len - m->m_pkthdr.len); 507 } 508 509 /* 510 * Assume that we can create a fast-forward IP flow entry 511 * based on this packet. 512 */ 513 m->m_flags |= M_CANFASTFWD; 514 515 /* 516 * Run through list of hooks for input packets. If there are any 517 * filters which require that additional packets in the flow are 518 * not fast-forwarded, they must clear the M_CANFASTFWD flag. 519 * Note that filters must _never_ set this flag, as another filter 520 * in the list may have previously cleared it. 521 */ 522 #if defined(IPSEC) 523 if (!ipsec_used || !ipsec_indone(m)) 524 #else 525 if (1) 526 #endif 527 { 528 struct in_addr odst = ip->ip_dst; 529 bool freed; 530 531 SOFTNET_LOCK(); 532 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0; 533 SOFTNET_UNLOCK(); 534 if (freed || m == NULL) { 535 goto out; 536 } 537 ip = mtod(m, struct ip *); 538 hlen = ip->ip_hl << 2; 539 540 /* 541 * XXX The setting of "srcrt" here is to prevent ip_forward() 542 * from generating ICMP redirects for packets that have 543 * been redirected by a hook back out on to the same LAN that 544 * they came from and is not an indication that the packet 545 * is being inffluenced by source routing options. This 546 * allows things like 547 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" 548 * where tlp0 is both on the 1.1.1.0/24 network and is the 549 * default route for hosts on 1.1.1.0/24. Of course this 550 * also requires a "map tlp0 ..." to complete the story. 551 * One might argue whether or not this kind of network config. 552 * should be supported in this manner... 553 */ 554 srcrt = (odst.s_addr != ip->ip_dst.s_addr); 555 } 556 557 #ifdef ALTQ 558 /* XXX Temporary until ALTQ is changed to use a pfil hook */ 559 if (altq_input) { 560 SOFTNET_LOCK(); 561 if ((*altq_input)(m, AF_INET) == 0) { 562 /* Packet dropped by traffic conditioner. */ 563 SOFTNET_UNLOCK(); 564 goto out; 565 } 566 SOFTNET_UNLOCK(); 567 } 568 #endif 569 570 /* 571 * Process options and, if not destined for us, 572 * ship it on. ip_dooptions returns 1 when an 573 * error was detected (causing an icmp message 574 * to be sent and the original packet to be freed). 575 */ 576 ip_nhops = 0; /* for source routed packets */ 577 if (hlen > sizeof (struct ip) && ip_dooptions(m)) 578 goto out; 579 580 /* 581 * Enable a consistency check between the destination address 582 * and the arrival interface for a unicast packet (the RFC 1122 583 * strong ES model) if IP forwarding is disabled and the packet 584 * is not locally generated. 585 * 586 * XXX - Checking also should be disabled if the destination 587 * address is ipnat'ed to a different interface. 588 * 589 * XXX - Checking is incompatible with IP aliases added 590 * to the loopback interface instead of the interface where 591 * the packets are received. 592 * 593 * XXX - We need to add a per ifaddr flag for this so that 594 * we get finer grain control. 595 */ 596 checkif = ip_checkinterface && (ipforwarding == 0) && 597 (ifp->if_flags & IFF_LOOPBACK) == 0; 598 599 /* 600 * Check our list of addresses, to see if the packet is for us. 601 * 602 * Traditional 4.4BSD did not consult IFF_UP at all. 603 * The behavior here is to treat addresses on !IFF_UP interface 604 * or IN_IFF_NOTREADY addresses as not mine. 605 */ 606 downmatch = 0; 607 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 608 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { 609 if (ia->ia4_flags & IN_IFF_NOTREADY) 610 continue; 611 if (checkif && ia->ia_ifp != ifp) 612 continue; 613 if ((ia->ia_ifp->if_flags & IFF_UP) != 0) 614 break; 615 else 616 downmatch++; 617 } 618 } 619 if (ia != NULL) 620 goto ours; 621 if (ifp->if_flags & IFF_BROADCAST) { 622 IFADDR_FOREACH(ifa, ifp) { 623 if (ifa->ifa_addr->sa_family != AF_INET) 624 continue; 625 ia = ifatoia(ifa); 626 if (ia->ia4_flags & IN_IFF_NOTREADY) 627 continue; 628 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || 629 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || 630 /* 631 * Look for all-0's host part (old broadcast addr), 632 * either for subnet or net. 633 */ 634 ip->ip_dst.s_addr == ia->ia_subnet || 635 ip->ip_dst.s_addr == ia->ia_net) 636 goto ours; 637 /* 638 * An interface with IP address zero accepts 639 * all packets that arrive on that interface. 640 */ 641 if (in_nullhost(ia->ia_addr.sin_addr)) 642 goto ours; 643 } 644 } 645 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 646 #ifdef MROUTING 647 extern struct socket *ip_mrouter; 648 649 if (ip_mrouter) { 650 /* 651 * If we are acting as a multicast router, all 652 * incoming multicast packets are passed to the 653 * kernel-level multicast forwarding function. 654 * The packet is returned (relatively) intact; if 655 * ip_mforward() returns a non-zero value, the packet 656 * must be discarded, else it may be accepted below. 657 * 658 * (The IP ident field is put in the same byte order 659 * as expected when ip_mforward() is called from 660 * ip_output().) 661 */ 662 SOFTNET_LOCK(); 663 if (ip_mforward(m, ifp) != 0) { 664 SOFTNET_UNLOCK(); 665 IP_STATINC(IP_STAT_CANTFORWARD); 666 goto bad; 667 } 668 SOFTNET_UNLOCK(); 669 670 /* 671 * The process-level routing demon needs to receive 672 * all multicast IGMP packets, whether or not this 673 * host belongs to their destination groups. 674 */ 675 if (ip->ip_p == IPPROTO_IGMP) { 676 goto ours; 677 } 678 IP_STATINC(IP_STAT_CANTFORWARD); 679 } 680 #endif 681 /* 682 * See if we belong to the destination multicast group on the 683 * arrival interface. 684 */ 685 if (!in_multi_group(ip->ip_dst, ifp, 0)) { 686 IP_STATINC(IP_STAT_CANTFORWARD); 687 goto bad; 688 } 689 goto ours; 690 } 691 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 692 in_nullhost(ip->ip_dst)) 693 goto ours; 694 695 /* 696 * Not for us; forward if possible and desirable. 697 */ 698 if (ipforwarding == 0) { 699 m_put_rcvif_psref(ifp, &psref); 700 IP_STATINC(IP_STAT_CANTFORWARD); 701 m_freem(m); 702 } else { 703 /* 704 * If ip_dst matched any of my address on !IFF_UP interface, 705 * and there's no IFF_UP interface that matches ip_dst, 706 * send icmp unreach. Forwarding it will result in in-kernel 707 * forwarding loop till TTL goes to 0. 708 */ 709 if (downmatch) { 710 m_put_rcvif_psref(ifp, &psref); 711 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 712 IP_STATINC(IP_STAT_CANTFORWARD); 713 return; 714 } 715 #ifdef IPSEC 716 /* Perform IPsec, if any. */ 717 if (ipsec_used) { 718 SOFTNET_LOCK(); 719 if (ipsec4_input(m, IP_FORWARDING | 720 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) { 721 SOFTNET_UNLOCK(); 722 goto bad; 723 } 724 SOFTNET_UNLOCK(); 725 } 726 #endif 727 ip_forward(m, srcrt, ifp); 728 m_put_rcvif_psref(ifp, &psref); 729 } 730 return; 731 732 ours: 733 m_put_rcvif_psref(ifp, &psref); 734 735 /* 736 * If offset or IP_MF are set, must reassemble. 737 */ 738 if (ip->ip_off & ~htons(IP_DF|IP_RF)) { 739 /* 740 * Pass to IP reassembly mechanism. 741 */ 742 if (ip_reass_packet(&m, ip) != 0) { 743 /* Failed; invalid fragment(s) or packet. */ 744 goto bad; 745 } 746 if (m == NULL) { 747 /* More fragments should come; silently return. */ 748 return; 749 } 750 /* 751 * Reassembly is done, we have the final packet. 752 * Updated cached data in local variable(s). 753 */ 754 ip = mtod(m, struct ip *); 755 hlen = ip->ip_hl << 2; 756 } 757 758 #ifdef IPSEC 759 /* 760 * Enforce IPsec policy checking if we are seeing last header. 761 * Note that we do not visit this with protocols with PCB layer 762 * code - like UDP/TCP/raw IP. 763 */ 764 if (ipsec_used && 765 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) { 766 SOFTNET_LOCK(); 767 if (ipsec4_input(m, 0) != 0) { 768 SOFTNET_UNLOCK(); 769 goto bad; 770 } 771 SOFTNET_UNLOCK(); 772 } 773 #endif 774 775 /* 776 * Switch out to protocol's input routine. 777 */ 778 #if IFA_STATS 779 if (ia && ip) 780 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len); 781 #endif 782 IP_STATINC(IP_STAT_DELIVERED); 783 784 const int off = hlen, nh = ip->ip_p; 785 786 SOFTNET_LOCK(); 787 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh); 788 SOFTNET_UNLOCK(); 789 return; 790 bad: 791 m_put_rcvif_psref(ifp, &psref); 792 m_freem(m); 793 return; 794 795 badcsum: 796 m_put_rcvif_psref(ifp, &psref); 797 IP_STATINC(IP_STAT_BADSUM); 798 m_freem(m); 799 return; 800 out: 801 m_put_rcvif_psref(ifp, &psref); 802 } 803 804 /* 805 * IP timer processing. 806 */ 807 void 808 ip_slowtimo(void) 809 { 810 811 mutex_enter(softnet_lock); 812 KERNEL_LOCK(1, NULL); 813 814 ip_reass_slowtimo(); 815 816 KERNEL_UNLOCK_ONE(NULL); 817 mutex_exit(softnet_lock); 818 } 819 820 /* 821 * IP drain processing. 822 */ 823 void 824 ip_drain(void) 825 { 826 827 KERNEL_LOCK(1, NULL); 828 ip_reass_drain(); 829 KERNEL_UNLOCK_ONE(NULL); 830 } 831 832 /* 833 * ip_dooptions: perform option processing on a datagram, possibly discarding 834 * it if bad options are encountered, or forwarding it if source-routed. 835 * 836 * => Returns true if packet has been forwarded/freed. 837 * => Returns false if the packet should be processed further. 838 */ 839 static bool 840 ip_dooptions(struct mbuf *m) 841 { 842 struct ip *ip = mtod(m, struct ip *); 843 u_char *cp, *cp0; 844 struct ip_timestamp *ipt; 845 struct in_ifaddr *ia; 846 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 847 struct in_addr dst; 848 n_time ntime; 849 850 dst = ip->ip_dst; 851 cp = (u_char *)(ip + 1); 852 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 853 for (; cnt > 0; cnt -= optlen, cp += optlen) { 854 opt = cp[IPOPT_OPTVAL]; 855 if (opt == IPOPT_EOL) 856 break; 857 if (opt == IPOPT_NOP) 858 optlen = 1; 859 else { 860 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 861 code = &cp[IPOPT_OLEN] - (u_char *)ip; 862 goto bad; 863 } 864 optlen = cp[IPOPT_OLEN]; 865 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 866 code = &cp[IPOPT_OLEN] - (u_char *)ip; 867 goto bad; 868 } 869 } 870 switch (opt) { 871 872 default: 873 break; 874 875 /* 876 * Source routing with record. 877 * Find interface with current destination address. 878 * If none on this machine then drop if strictly routed, 879 * or do nothing if loosely routed. 880 * Record interface address and bring up next address 881 * component. If strictly routed make sure next 882 * address is on directly accessible net. 883 */ 884 case IPOPT_LSRR: 885 case IPOPT_SSRR: 886 if (ip_allowsrcrt == 0) { 887 type = ICMP_UNREACH; 888 code = ICMP_UNREACH_NET_PROHIB; 889 goto bad; 890 } 891 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 892 code = &cp[IPOPT_OLEN] - (u_char *)ip; 893 goto bad; 894 } 895 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 896 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 897 goto bad; 898 } 899 ipaddr.sin_addr = ip->ip_dst; 900 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); 901 if (ia == 0) { 902 if (opt == IPOPT_SSRR) { 903 type = ICMP_UNREACH; 904 code = ICMP_UNREACH_SRCFAIL; 905 goto bad; 906 } 907 /* 908 * Loose routing, and not at next destination 909 * yet; nothing to do except forward. 910 */ 911 break; 912 } 913 off--; /* 0 origin */ 914 if ((off + sizeof(struct in_addr)) > optlen) { 915 /* 916 * End of source route. Should be for us. 917 */ 918 save_rte(cp, ip->ip_src); 919 break; 920 } 921 /* 922 * locate outgoing interface 923 */ 924 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off), 925 sizeof(ipaddr.sin_addr)); 926 if (opt == IPOPT_SSRR) 927 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); 928 else 929 ia = ip_rtaddr(ipaddr.sin_addr); 930 if (ia == 0) { 931 type = ICMP_UNREACH; 932 code = ICMP_UNREACH_SRCFAIL; 933 goto bad; 934 } 935 ip->ip_dst = ipaddr.sin_addr; 936 bcopy((void *)&ia->ia_addr.sin_addr, 937 (void *)(cp + off), sizeof(struct in_addr)); 938 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 939 /* 940 * Let ip_intr's mcast routing check handle mcast pkts 941 */ 942 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 943 break; 944 945 case IPOPT_RR: 946 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 947 code = &cp[IPOPT_OLEN] - (u_char *)ip; 948 goto bad; 949 } 950 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 951 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 952 goto bad; 953 } 954 /* 955 * If no space remains, ignore. 956 */ 957 off--; /* 0 origin */ 958 if ((off + sizeof(struct in_addr)) > optlen) 959 break; 960 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst), 961 sizeof(ipaddr.sin_addr)); 962 /* 963 * locate outgoing interface; if we're the destination, 964 * use the incoming interface (should be same). 965 */ 966 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))) 967 == NULL && 968 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) { 969 type = ICMP_UNREACH; 970 code = ICMP_UNREACH_HOST; 971 goto bad; 972 } 973 bcopy((void *)&ia->ia_addr.sin_addr, 974 (void *)(cp + off), sizeof(struct in_addr)); 975 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 976 break; 977 978 case IPOPT_TS: 979 code = cp - (u_char *)ip; 980 ipt = (struct ip_timestamp *)cp; 981 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 982 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 983 goto bad; 984 } 985 if (ipt->ipt_ptr < 5) { 986 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 987 goto bad; 988 } 989 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) { 990 if (++ipt->ipt_oflw == 0) { 991 code = (u_char *)&ipt->ipt_ptr - 992 (u_char *)ip; 993 goto bad; 994 } 995 break; 996 } 997 cp0 = (cp + ipt->ipt_ptr - 1); 998 switch (ipt->ipt_flg) { 999 1000 case IPOPT_TS_TSONLY: 1001 break; 1002 1003 case IPOPT_TS_TSANDADDR: { 1004 struct ifnet *rcvif; 1005 int s; 1006 1007 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1008 sizeof(struct in_addr) > ipt->ipt_len) { 1009 code = (u_char *)&ipt->ipt_ptr - 1010 (u_char *)ip; 1011 goto bad; 1012 } 1013 ipaddr.sin_addr = dst; 1014 rcvif = m_get_rcvif(m, &s); 1015 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1016 rcvif)); 1017 m_put_rcvif(rcvif, &s); 1018 if (ia == 0) 1019 continue; 1020 bcopy(&ia->ia_addr.sin_addr, 1021 cp0, sizeof(struct in_addr)); 1022 ipt->ipt_ptr += sizeof(struct in_addr); 1023 break; 1024 } 1025 1026 case IPOPT_TS_PRESPEC: 1027 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 1028 sizeof(struct in_addr) > ipt->ipt_len) { 1029 code = (u_char *)&ipt->ipt_ptr - 1030 (u_char *)ip; 1031 goto bad; 1032 } 1033 memcpy(&ipaddr.sin_addr, cp0, 1034 sizeof(struct in_addr)); 1035 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))) 1036 == NULL) 1037 continue; 1038 ipt->ipt_ptr += sizeof(struct in_addr); 1039 break; 1040 1041 default: 1042 /* XXX can't take &ipt->ipt_flg */ 1043 code = (u_char *)&ipt->ipt_ptr - 1044 (u_char *)ip + 1; 1045 goto bad; 1046 } 1047 ntime = iptime(); 1048 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ 1049 memmove((char *)cp + ipt->ipt_ptr - 1, cp0, 1050 sizeof(n_time)); 1051 ipt->ipt_ptr += sizeof(n_time); 1052 } 1053 } 1054 if (forward) { 1055 struct ifnet *rcvif; 1056 struct psref psref; 1057 1058 if (ip_forwsrcrt == 0) { 1059 type = ICMP_UNREACH; 1060 code = ICMP_UNREACH_SRCFAIL; 1061 goto bad; 1062 } 1063 1064 rcvif = m_get_rcvif_psref(m, &psref); 1065 if (__predict_false(rcvif == NULL)) { 1066 type = ICMP_UNREACH; 1067 code = ICMP_UNREACH_HOST; 1068 goto bad; 1069 } 1070 ip_forward(m, 1, rcvif); 1071 m_put_rcvif_psref(rcvif, &psref); 1072 return true; 1073 } 1074 return false; 1075 bad: 1076 icmp_error(m, type, code, 0, 0); 1077 IP_STATINC(IP_STAT_BADOPTIONS); 1078 return true; 1079 } 1080 1081 /* 1082 * ip_rtaddr: given address of next destination (final or next hop), 1083 * return internet address info of interface to be used to get there. 1084 */ 1085 static struct in_ifaddr * 1086 ip_rtaddr(struct in_addr dst) 1087 { 1088 struct rtentry *rt; 1089 union { 1090 struct sockaddr dst; 1091 struct sockaddr_in dst4; 1092 } u; 1093 1094 sockaddr_in_init(&u.dst4, &dst, 0); 1095 1096 SOFTNET_LOCK(); 1097 rt = rtcache_lookup(&ipforward_rt, &u.dst); 1098 SOFTNET_UNLOCK(); 1099 if (rt == NULL) 1100 return NULL; 1101 1102 return ifatoia(rt->rt_ifa); 1103 } 1104 1105 /* 1106 * save_rte: save incoming source route for use in replies, to be picked 1107 * up later by ip_srcroute if the receiver is interested. 1108 */ 1109 static void 1110 save_rte(u_char *option, struct in_addr dst) 1111 { 1112 unsigned olen; 1113 1114 olen = option[IPOPT_OLEN]; 1115 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 1116 return; 1117 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen); 1118 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1119 ip_srcrt.dst = dst; 1120 } 1121 1122 /* 1123 * Retrieve incoming source route for use in replies, 1124 * in the same form used by setsockopt. 1125 * The first hop is placed before the options, will be removed later. 1126 */ 1127 struct mbuf * 1128 ip_srcroute(void) 1129 { 1130 struct in_addr *p, *q; 1131 struct mbuf *m; 1132 1133 if (ip_nhops == 0) 1134 return NULL; 1135 m = m_get(M_DONTWAIT, MT_SOOPTS); 1136 if (m == 0) 1137 return NULL; 1138 1139 MCLAIM(m, &inetdomain.dom_mowner); 1140 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 1141 1142 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1143 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 1144 OPTSIZ; 1145 1146 /* 1147 * First save first hop for return route 1148 */ 1149 p = &ip_srcrt.route[ip_nhops - 1]; 1150 *(mtod(m, struct in_addr *)) = *p--; 1151 1152 /* 1153 * Copy option fields and padding (nop) to mbuf. 1154 */ 1155 ip_srcrt.nop = IPOPT_NOP; 1156 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1157 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, 1158 OPTSIZ); 1159 q = (struct in_addr *)(mtod(m, char *) + 1160 sizeof(struct in_addr) + OPTSIZ); 1161 #undef OPTSIZ 1162 /* 1163 * Record return path as an IP source route, 1164 * reversing the path (pointers are now aligned). 1165 */ 1166 while (p >= ip_srcrt.route) { 1167 *q++ = *p--; 1168 } 1169 /* 1170 * Last hop goes to final destination. 1171 */ 1172 *q = ip_srcrt.dst; 1173 return (m); 1174 } 1175 1176 const int inetctlerrmap[PRC_NCMDS] = { 1177 [PRC_MSGSIZE] = EMSGSIZE, 1178 [PRC_HOSTDEAD] = EHOSTDOWN, 1179 [PRC_HOSTUNREACH] = EHOSTUNREACH, 1180 [PRC_UNREACH_NET] = EHOSTUNREACH, 1181 [PRC_UNREACH_HOST] = EHOSTUNREACH, 1182 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED, 1183 [PRC_UNREACH_PORT] = ECONNREFUSED, 1184 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH, 1185 [PRC_PARAMPROB] = ENOPROTOOPT, 1186 }; 1187 1188 void 1189 ip_fasttimo(void) 1190 { 1191 if (ip_drainwanted) { 1192 ip_drain(); 1193 ip_drainwanted = 0; 1194 } 1195 } 1196 1197 void 1198 ip_drainstub(void) 1199 { 1200 ip_drainwanted = 1; 1201 } 1202 1203 /* 1204 * Forward a packet. If some error occurs return the sender 1205 * an icmp packet. Note we can't always generate a meaningful 1206 * icmp message because icmp doesn't have a large enough repertoire 1207 * of codes and types. 1208 * 1209 * If not forwarding, just drop the packet. This could be confusing 1210 * if ipforwarding was zero but some routing protocol was advancing 1211 * us as a gateway to somewhere. However, we must let the routing 1212 * protocol deal with that. 1213 * 1214 * The srcrt parameter indicates whether the packet is being forwarded 1215 * via a source route. 1216 */ 1217 static void 1218 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif) 1219 { 1220 struct ip *ip = mtod(m, struct ip *); 1221 struct rtentry *rt; 1222 int error, type = 0, code = 0, destmtu = 0; 1223 struct mbuf *mcopy; 1224 n_long dest; 1225 union { 1226 struct sockaddr dst; 1227 struct sockaddr_in dst4; 1228 } u; 1229 uint64_t *ips; 1230 1231 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software " 1232 "interrupt handler; synchronization assumptions violated"); 1233 1234 /* 1235 * We are now in the output path. 1236 */ 1237 MCLAIM(m, &ip_tx_mowner); 1238 1239 /* 1240 * Clear any in-bound checksum flags for this packet. 1241 */ 1242 m->m_pkthdr.csum_flags = 0; 1243 1244 dest = 0; 1245 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1246 IP_STATINC(IP_STAT_CANTFORWARD); 1247 m_freem(m); 1248 return; 1249 } 1250 1251 SOFTNET_LOCK(); 1252 1253 if (ip->ip_ttl <= IPTTLDEC) { 1254 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1255 SOFTNET_UNLOCK(); 1256 return; 1257 } 1258 1259 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 1260 1261 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { 1262 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); 1263 SOFTNET_UNLOCK(); 1264 return; 1265 } 1266 1267 /* 1268 * Save at most 68 bytes of the packet in case 1269 * we need to generate an ICMP message to the src. 1270 * Pullup to avoid sharing mbuf cluster between m and mcopy. 1271 */ 1272 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT); 1273 if (mcopy) 1274 mcopy = m_pullup(mcopy, ip->ip_hl << 2); 1275 1276 ip->ip_ttl -= IPTTLDEC; 1277 1278 /* 1279 * If forwarding packet using same interface that it came in on, 1280 * perhaps should send a redirect to sender to shortcut a hop. 1281 * Only send redirect if source is sending directly to us, 1282 * and if packet was not source routed (or has any options). 1283 * Also, don't send redirect if forwarding using a default route 1284 * or a route modified by a redirect. 1285 */ 1286 if (rt->rt_ifp == rcvif && 1287 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1288 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && 1289 ipsendredirects && !srcrt) { 1290 if (rt->rt_ifa && 1291 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == 1292 ifatoia(rt->rt_ifa)->ia_subnet) { 1293 if (rt->rt_flags & RTF_GATEWAY) 1294 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1295 else 1296 dest = ip->ip_dst.s_addr; 1297 /* 1298 * Router requirements says to only send host 1299 * redirects. 1300 */ 1301 type = ICMP_REDIRECT; 1302 code = ICMP_REDIRECT_HOST; 1303 } 1304 } 1305 1306 error = ip_output(m, NULL, &ipforward_rt, 1307 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1308 NULL, NULL); 1309 1310 if (error) { 1311 IP_STATINC(IP_STAT_CANTFORWARD); 1312 goto error; 1313 } 1314 1315 ips = IP_STAT_GETREF(); 1316 ips[IP_STAT_FORWARD]++; 1317 1318 if (type) { 1319 ips[IP_STAT_REDIRECTSENT]++; 1320 IP_STAT_PUTREF(); 1321 goto redirect; 1322 } 1323 1324 IP_STAT_PUTREF(); 1325 if (mcopy) { 1326 #ifdef GATEWAY 1327 if (mcopy->m_flags & M_CANFASTFWD) 1328 ipflow_create(&ipforward_rt, mcopy); 1329 #endif 1330 m_freem(mcopy); 1331 } 1332 1333 SOFTNET_UNLOCK(); 1334 return; 1335 1336 redirect: 1337 error: 1338 if (mcopy == NULL) { 1339 SOFTNET_UNLOCK(); 1340 return; 1341 } 1342 1343 switch (error) { 1344 1345 case 0: /* forwarded, but need redirect */ 1346 /* type, code set above */ 1347 break; 1348 1349 case ENETUNREACH: /* shouldn't happen, checked above */ 1350 case EHOSTUNREACH: 1351 case ENETDOWN: 1352 case EHOSTDOWN: 1353 default: 1354 type = ICMP_UNREACH; 1355 code = ICMP_UNREACH_HOST; 1356 break; 1357 1358 case EMSGSIZE: 1359 type = ICMP_UNREACH; 1360 code = ICMP_UNREACH_NEEDFRAG; 1361 1362 if ((rt = rtcache_validate(&ipforward_rt)) != NULL) 1363 destmtu = rt->rt_ifp->if_mtu; 1364 #ifdef IPSEC 1365 if (ipsec_used) 1366 (void)ipsec4_forward(mcopy, &destmtu); 1367 #endif 1368 IP_STATINC(IP_STAT_CANTFRAG); 1369 break; 1370 1371 case ENOBUFS: 1372 /* 1373 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812, 1374 * Requirements for IP Version 4 Routers. Source quench can 1375 * big problem under DoS attacks or if the underlying 1376 * interface is rate-limited. 1377 */ 1378 if (mcopy) 1379 m_freem(mcopy); 1380 SOFTNET_UNLOCK(); 1381 return; 1382 } 1383 icmp_error(mcopy, type, code, dest, destmtu); 1384 SOFTNET_UNLOCK(); 1385 } 1386 1387 void 1388 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1389 struct mbuf *m) 1390 { 1391 struct socket *so = inp->inp_socket; 1392 ifnet_t *ifp; 1393 int inpflags = inp->inp_flags; 1394 struct psref psref; 1395 1396 ifp = m_get_rcvif_psref(m, &psref); 1397 if (__predict_false(ifp == NULL)) 1398 return; /* XXX should report error? */ 1399 1400 if (so->so_options & SO_TIMESTAMP 1401 #ifdef SO_OTIMESTAMP 1402 || so->so_options & SO_OTIMESTAMP 1403 #endif 1404 ) { 1405 struct timeval tv; 1406 1407 microtime(&tv); 1408 #ifdef SO_OTIMESTAMP 1409 if (so->so_options & SO_OTIMESTAMP) { 1410 struct timeval50 tv50; 1411 timeval_to_timeval50(&tv, &tv50); 1412 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50), 1413 SCM_OTIMESTAMP, SOL_SOCKET); 1414 } else 1415 #endif 1416 *mp = sbcreatecontrol((void *) &tv, sizeof(tv), 1417 SCM_TIMESTAMP, SOL_SOCKET); 1418 if (*mp) 1419 mp = &(*mp)->m_next; 1420 } 1421 if (inpflags & INP_RECVDSTADDR) { 1422 *mp = sbcreatecontrol((void *) &ip->ip_dst, 1423 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1424 if (*mp) 1425 mp = &(*mp)->m_next; 1426 } 1427 if (inpflags & INP_RECVPKTINFO) { 1428 struct in_pktinfo ipi; 1429 ipi.ipi_addr = ip->ip_src; 1430 ipi.ipi_ifindex = ifp->if_index; 1431 *mp = sbcreatecontrol((void *) &ipi, 1432 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP); 1433 if (*mp) 1434 mp = &(*mp)->m_next; 1435 } 1436 if (inpflags & INP_PKTINFO) { 1437 struct in_pktinfo ipi; 1438 ipi.ipi_addr = ip->ip_dst; 1439 ipi.ipi_ifindex = ifp->if_index; 1440 *mp = sbcreatecontrol((void *) &ipi, 1441 sizeof(ipi), IP_PKTINFO, IPPROTO_IP); 1442 if (*mp) 1443 mp = &(*mp)->m_next; 1444 } 1445 if (inpflags & INP_RECVIF) { 1446 struct sockaddr_dl sdl; 1447 1448 sockaddr_dl_init(&sdl, sizeof(sdl), ifp ? 1449 ifp->if_index : 0, 0, NULL, 0, NULL, 0); 1450 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); 1451 if (*mp) 1452 mp = &(*mp)->m_next; 1453 } 1454 if (inpflags & INP_RECVTTL) { 1455 *mp = sbcreatecontrol((void *) &ip->ip_ttl, 1456 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP); 1457 if (*mp) 1458 mp = &(*mp)->m_next; 1459 } 1460 m_put_rcvif_psref(ifp, &psref); 1461 } 1462 1463 /* 1464 * sysctl helper routine for net.inet.ip.forwsrcrt. 1465 */ 1466 static int 1467 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) 1468 { 1469 int error, tmp; 1470 struct sysctlnode node; 1471 1472 node = *rnode; 1473 tmp = ip_forwsrcrt; 1474 node.sysctl_data = &tmp; 1475 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1476 if (error || newp == NULL) 1477 return (error); 1478 1479 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, 1480 0, NULL, NULL, NULL); 1481 if (error) 1482 return (error); 1483 1484 ip_forwsrcrt = tmp; 1485 1486 return (0); 1487 } 1488 1489 /* 1490 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the 1491 * range of the new value and tweaks timers if it changes. 1492 */ 1493 static int 1494 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) 1495 { 1496 int error, tmp; 1497 struct sysctlnode node; 1498 1499 node = *rnode; 1500 tmp = ip_mtudisc_timeout; 1501 node.sysctl_data = &tmp; 1502 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1503 if (error || newp == NULL) 1504 return (error); 1505 if (tmp < 0) 1506 return (EINVAL); 1507 1508 mutex_enter(softnet_lock); 1509 1510 ip_mtudisc_timeout = tmp; 1511 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); 1512 1513 mutex_exit(softnet_lock); 1514 1515 return (0); 1516 } 1517 1518 static int 1519 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS) 1520 { 1521 1522 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS)); 1523 } 1524 1525 static void 1526 sysctl_net_inet_ip_setup(struct sysctllog **clog) 1527 { 1528 sysctl_createv(clog, 0, NULL, NULL, 1529 CTLFLAG_PERMANENT, 1530 CTLTYPE_NODE, "inet", 1531 SYSCTL_DESCR("PF_INET related settings"), 1532 NULL, 0, NULL, 0, 1533 CTL_NET, PF_INET, CTL_EOL); 1534 sysctl_createv(clog, 0, NULL, NULL, 1535 CTLFLAG_PERMANENT, 1536 CTLTYPE_NODE, "ip", 1537 SYSCTL_DESCR("IPv4 related settings"), 1538 NULL, 0, NULL, 0, 1539 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 1540 1541 sysctl_createv(clog, 0, NULL, NULL, 1542 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1543 CTLTYPE_INT, "forwarding", 1544 SYSCTL_DESCR("Enable forwarding of INET datagrams"), 1545 NULL, 0, &ipforwarding, 0, 1546 CTL_NET, PF_INET, IPPROTO_IP, 1547 IPCTL_FORWARDING, CTL_EOL); 1548 sysctl_createv(clog, 0, NULL, NULL, 1549 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1550 CTLTYPE_INT, "redirect", 1551 SYSCTL_DESCR("Enable sending of ICMP redirect messages"), 1552 NULL, 0, &ipsendredirects, 0, 1553 CTL_NET, PF_INET, IPPROTO_IP, 1554 IPCTL_SENDREDIRECTS, CTL_EOL); 1555 sysctl_createv(clog, 0, NULL, NULL, 1556 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1557 CTLTYPE_INT, "ttl", 1558 SYSCTL_DESCR("Default TTL for an INET datagram"), 1559 NULL, 0, &ip_defttl, 0, 1560 CTL_NET, PF_INET, IPPROTO_IP, 1561 IPCTL_DEFTTL, CTL_EOL); 1562 #ifdef IPCTL_DEFMTU 1563 sysctl_createv(clog, 0, NULL, NULL, 1564 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, 1565 CTLTYPE_INT, "mtu", 1566 SYSCTL_DESCR("Default MTA for an INET route"), 1567 NULL, 0, &ip_mtu, 0, 1568 CTL_NET, PF_INET, IPPROTO_IP, 1569 IPCTL_DEFMTU, CTL_EOL); 1570 #endif /* IPCTL_DEFMTU */ 1571 sysctl_createv(clog, 0, NULL, NULL, 1572 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1573 CTLTYPE_INT, "forwsrcrt", 1574 SYSCTL_DESCR("Enable forwarding of source-routed " 1575 "datagrams"), 1576 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, 1577 CTL_NET, PF_INET, IPPROTO_IP, 1578 IPCTL_FORWSRCRT, CTL_EOL); 1579 sysctl_createv(clog, 0, NULL, NULL, 1580 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1581 CTLTYPE_INT, "directed-broadcast", 1582 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), 1583 NULL, 0, &ip_directedbcast, 0, 1584 CTL_NET, PF_INET, IPPROTO_IP, 1585 IPCTL_DIRECTEDBCAST, CTL_EOL); 1586 sysctl_createv(clog, 0, NULL, NULL, 1587 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1588 CTLTYPE_INT, "allowsrcrt", 1589 SYSCTL_DESCR("Accept source-routed datagrams"), 1590 NULL, 0, &ip_allowsrcrt, 0, 1591 CTL_NET, PF_INET, IPPROTO_IP, 1592 IPCTL_ALLOWSRCRT, CTL_EOL); 1593 1594 sysctl_createv(clog, 0, NULL, NULL, 1595 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1596 CTLTYPE_INT, "mtudisc", 1597 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), 1598 NULL, 0, &ip_mtudisc, 0, 1599 CTL_NET, PF_INET, IPPROTO_IP, 1600 IPCTL_MTUDISC, CTL_EOL); 1601 sysctl_createv(clog, 0, NULL, NULL, 1602 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1603 CTLTYPE_INT, "anonportmin", 1604 SYSCTL_DESCR("Lowest ephemeral port number to assign"), 1605 sysctl_net_inet_ip_ports, 0, &anonportmin, 0, 1606 CTL_NET, PF_INET, IPPROTO_IP, 1607 IPCTL_ANONPORTMIN, CTL_EOL); 1608 sysctl_createv(clog, 0, NULL, NULL, 1609 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1610 CTLTYPE_INT, "anonportmax", 1611 SYSCTL_DESCR("Highest ephemeral port number to assign"), 1612 sysctl_net_inet_ip_ports, 0, &anonportmax, 0, 1613 CTL_NET, PF_INET, IPPROTO_IP, 1614 IPCTL_ANONPORTMAX, CTL_EOL); 1615 sysctl_createv(clog, 0, NULL, NULL, 1616 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1617 CTLTYPE_INT, "mtudisctimeout", 1618 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), 1619 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0, 1620 CTL_NET, PF_INET, IPPROTO_IP, 1621 IPCTL_MTUDISCTIMEOUT, CTL_EOL); 1622 #ifndef IPNOPRIVPORTS 1623 sysctl_createv(clog, 0, NULL, NULL, 1624 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1625 CTLTYPE_INT, "lowportmin", 1626 SYSCTL_DESCR("Lowest privileged ephemeral port number " 1627 "to assign"), 1628 sysctl_net_inet_ip_ports, 0, &lowportmin, 0, 1629 CTL_NET, PF_INET, IPPROTO_IP, 1630 IPCTL_LOWPORTMIN, CTL_EOL); 1631 sysctl_createv(clog, 0, NULL, NULL, 1632 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1633 CTLTYPE_INT, "lowportmax", 1634 SYSCTL_DESCR("Highest privileged ephemeral port number " 1635 "to assign"), 1636 sysctl_net_inet_ip_ports, 0, &lowportmax, 0, 1637 CTL_NET, PF_INET, IPPROTO_IP, 1638 IPCTL_LOWPORTMAX, CTL_EOL); 1639 #endif /* IPNOPRIVPORTS */ 1640 #if NGRE > 0 1641 sysctl_createv(clog, 0, NULL, NULL, 1642 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1643 CTLTYPE_INT, "grettl", 1644 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), 1645 NULL, 0, &ip_gre_ttl, 0, 1646 CTL_NET, PF_INET, IPPROTO_IP, 1647 IPCTL_GRE_TTL, CTL_EOL); 1648 #endif /* NGRE */ 1649 sysctl_createv(clog, 0, NULL, NULL, 1650 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1651 CTLTYPE_INT, "checkinterface", 1652 SYSCTL_DESCR("Enable receive side of Strong ES model " 1653 "from RFC1122"), 1654 NULL, 0, &ip_checkinterface, 0, 1655 CTL_NET, PF_INET, IPPROTO_IP, 1656 IPCTL_CHECKINTERFACE, CTL_EOL); 1657 sysctl_createv(clog, 0, NULL, NULL, 1658 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1659 CTLTYPE_INT, "random_id", 1660 SYSCTL_DESCR("Assign random ip_id values"), 1661 NULL, 0, &ip_do_randomid, 0, 1662 CTL_NET, PF_INET, IPPROTO_IP, 1663 IPCTL_RANDOMID, CTL_EOL); 1664 sysctl_createv(clog, 0, NULL, NULL, 1665 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1666 CTLTYPE_INT, "do_loopback_cksum", 1667 SYSCTL_DESCR("Perform IP checksum on loopback"), 1668 NULL, 0, &ip_do_loopback_cksum, 0, 1669 CTL_NET, PF_INET, IPPROTO_IP, 1670 IPCTL_LOOPBACKCKSUM, CTL_EOL); 1671 sysctl_createv(clog, 0, NULL, NULL, 1672 CTLFLAG_PERMANENT, 1673 CTLTYPE_STRUCT, "stats", 1674 SYSCTL_DESCR("IP statistics"), 1675 sysctl_net_inet_ip_stats, 0, NULL, 0, 1676 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, 1677 CTL_EOL); 1678 #if NARP 1679 sysctl_createv(clog, 0, NULL, NULL, 1680 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1681 CTLTYPE_INT, "dad_count", 1682 SYSCTL_DESCR("Number of Duplicate Address Detection " 1683 "probes to send"), 1684 NULL, 0, &ip_dad_count, 0, 1685 CTL_NET, PF_INET, IPPROTO_IP, 1686 IPCTL_DAD_COUNT, CTL_EOL); 1687 #endif 1688 1689 /* anonportalgo RFC6056 subtree */ 1690 const struct sysctlnode *portalgo_node; 1691 sysctl_createv(clog, 0, NULL, &portalgo_node, 1692 CTLFLAG_PERMANENT, 1693 CTLTYPE_NODE, "anonportalgo", 1694 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"), 1695 NULL, 0, NULL, 0, 1696 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL); 1697 sysctl_createv(clog, 0, &portalgo_node, NULL, 1698 CTLFLAG_PERMANENT, 1699 CTLTYPE_STRING, "available", 1700 SYSCTL_DESCR("available algorithms"), 1701 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN, 1702 CTL_CREATE, CTL_EOL); 1703 sysctl_createv(clog, 0, &portalgo_node, NULL, 1704 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1705 CTLTYPE_STRING, "selected", 1706 SYSCTL_DESCR("selected algorithm"), 1707 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN, 1708 CTL_CREATE, CTL_EOL); 1709 sysctl_createv(clog, 0, &portalgo_node, NULL, 1710 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1711 CTLTYPE_STRUCT, "reserve", 1712 SYSCTL_DESCR("bitmap of reserved ports"), 1713 sysctl_portalgo_reserve4, 0, NULL, 0, 1714 CTL_CREATE, CTL_EOL); 1715 } 1716 1717 void 1718 ip_statinc(u_int stat) 1719 { 1720 1721 KASSERT(stat < IP_NSTATS); 1722 IP_STATINC(stat); 1723 } 1724
Indexes created Wed Sep 24 05:09:52 GMT 2025