1/* 2 * Copyright (c) 2000-2012 Apple Inc. All rights reserved. 3 * 4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ 5 * 6 * This file contains Original Code and/or Modifications of Original Code 7 * as defined in and that are subject to the Apple Public Source License 8 * Version 2.0 (the 'License'). You may not use this file except in 9 * compliance with the License. The rights granted to you under the License 10 * may not be used to create, or enable the creation or redistribution of, 11 * unlawful or unlicensed copies of an Apple operating system, or to 12 * circumvent, violate, or enable the circumvention or violation of, any 13 * terms of an Apple operating system software license agreement. 14 * 15 * Please obtain a copy of the License at 16 * http://www.opensource.apple.com/apsl/ and read it before using this file. 17 * 18 * The Original Code and all software distributed under the License are 19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER 20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, 21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, 22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. 23 * Please see the License for the specific language governing rights and 24 * limitations under the License. 25 * 26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ 27 */ 28/* 29 * Copyright (c) 1982, 1986, 1988, 1993 30 * The Regents of the University of California. All rights reserved. 31 * 32 * Redistribution and use in source and binary forms, with or without 33 * modification, are permitted provided that the following conditions 34 * are met: 35 * 1. Redistributions of source code must retain the above copyright 36 * notice, this list of conditions and the following disclaimer. 37 * 2. Redistributions in binary form must reproduce the above copyright 38 * notice, this list of conditions and the following disclaimer in the 39 * documentation and/or other materials provided with the distribution. 40 * 3. All advertising materials mentioning features or use of this software 41 * must display the following acknowledgement: 42 * This product includes software developed by the University of 43 * California, Berkeley and its contributors. 44 * 4. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 61 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.25 2001/08/29 21:41:37 jesper Exp $ 62 */ 63/* 64 * NOTICE: This file was modified by SPARTA, Inc. in 2007 to introduce 65 * support for mandatory and extensible security protections. This notice 66 * is included in support of clause 2.2 (b) of the Apple Public License, 67 * Version 2.0. 68 */ 69 70#define _IP_VHL 71 72#include <sys/param.h> 73#include <sys/systm.h> 74#include <sys/mbuf.h> 75#include <sys/malloc.h> 76#include <sys/domain.h> 77#include <sys/protosw.h> 78#include <sys/socket.h> 79#include <sys/time.h> 80#include <sys/kernel.h> 81#include <sys/syslog.h> 82#include <sys/sysctl.h> 83#include <sys/mcache.h> 84#include <mach/mach_time.h> 85 86#include <machine/endian.h> 87 88#include <kern/queue.h> 89#include <kern/locks.h> 90 91#include <pexpert/pexpert.h> 92 93#include <net/if.h> 94#include <net/if_var.h> 95#include <net/if_dl.h> 96#include <net/route.h> 97#include <net/kpi_protocol.h> 98#include <net/ntstat.h> 99 100#include <netinet/in.h> 101#include <netinet/in_systm.h> 102#include <netinet/in_var.h> 103#include <netinet/in_arp.h> 104#include <netinet/ip.h> 105#include <netinet/in_pcb.h> 106#include <netinet/ip_var.h> 107#include <netinet/ip_icmp.h> 108#include <sys/socketvar.h> 109 110#include <netinet/ip_fw.h> 111#include <netinet/ip_divert.h> 112 113#include <netinet/kpi_ipfilter_var.h> 114 115/* needed for AUTOCONFIGURING: */ 116#include <netinet/udp.h> 117#include <netinet/udp_var.h> 118#include <netinet/bootp.h> 119#include <mach/sdt.h> 120 121#if CONFIG_MACF_NET 122#include <security/mac_framework.h> 123#endif 124 125#include <sys/kdebug.h> 126#include <libkern/OSAtomic.h> 127 128#define DBG_LAYER_BEG NETDBG_CODE(DBG_NETIP, 0) 129#define DBG_LAYER_END NETDBG_CODE(DBG_NETIP, 2) 130#define DBG_FNC_IP_INPUT NETDBG_CODE(DBG_NETIP, (2 << 8)) 131 132 133#if IPSEC 134#include <netinet6/ipsec.h> 135#include <netkey/key.h> 136#endif 137 138#if DUMMYNET 139#include <netinet/ip_dummynet.h> 140#endif 141 142#if PF 143#include <net/pfvar.h> 144#endif /* PF */ 145 146#include <netinet/lro_ext.h> 147 148#if IPSEC 149extern int ipsec_bypass; 150extern lck_mtx_t *sadb_mutex; 151 152lck_grp_t *sadb_stat_mutex_grp; 153lck_grp_attr_t *sadb_stat_mutex_grp_attr; 154lck_attr_t *sadb_stat_mutex_attr; 155decl_lck_mtx_data(, sadb_stat_mutex_data); 156lck_mtx_t *sadb_stat_mutex = &sadb_stat_mutex_data; 157 158#endif 159 160int rsvp_on = 0; 161static int ip_rsvp_on; 162struct socket *ip_rsvpd; 163 164static int sysctl_ipforwarding SYSCTL_HANDLER_ARGS; 165 166int ipforwarding = 0; 167SYSCTL_PROC(_net_inet_ip, IPCTL_FORWARDING, forwarding, 168 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &ipforwarding, 0, 169 sysctl_ipforwarding, "I", "Enable IP forwarding between interfaces"); 170 171static int ipsendredirects = 1; /* XXX */ 172SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW | CTLFLAG_LOCKED, 173 &ipsendredirects, 0, "Enable sending IP redirects"); 174 175int ip_defttl = IPDEFTTL; 176SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW | CTLFLAG_LOCKED, 177 &ip_defttl, 0, "Maximum TTL on IP packets"); 178 179static int ip_dosourceroute = 0; 180SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW | CTLFLAG_LOCKED, 181 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 182 183static int ip_acceptsourceroute = 0; 184SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 185 CTLFLAG_RW | CTLFLAG_LOCKED, &ip_acceptsourceroute, 0, 186 "Enable accepting source routed IP packets"); 187 188static int ip_keepfaith = 0; 189SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RD | CTLFLAG_LOCKED, 190 &ip_keepfaith, 0, ""); 191 192static int nipq = 0; /* total # of reass queues */ 193static int maxnipq; 194SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW | CTLFLAG_LOCKED, 195 &maxnipq, 0, 196 "Maximum number of IPv4 fragment reassembly queue entries"); 197 198static int maxfragsperpacket; 199SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW | CTLFLAG_LOCKED, 200 &maxfragsperpacket, 0, 201 "Maximum number of IPv4 fragments allowed per packet"); 202 203static int maxfrags; 204SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfrags, CTLFLAG_RW | CTLFLAG_LOCKED, 205 &maxfrags, 0, "Maximum number of IPv4 fragments allowed"); 206 207static int currentfrags = 0; 208 209int ip_doscopedroute = 1; 210SYSCTL_INT(_net_inet_ip, OID_AUTO, scopedroute, CTLFLAG_RD | CTLFLAG_LOCKED, 211 &ip_doscopedroute, 0, "Enable IPv4 scoped routing"); 212 213int ip_restrictrecvif = 1; 214SYSCTL_INT(_net_inet_ip, OID_AUTO, restrictrecvif, CTLFLAG_RW | CTLFLAG_LOCKED, 215 &ip_restrictrecvif, 0, "Enable inbound interface restrictions"); 216 217/* 218 * XXX - Setting ip_checkinterface mostly implements the receive side of 219 * the Strong ES model described in RFC 1122, but since the routing table 220 * and transmit implementation do not implement the Strong ES model, 221 * setting this to 1 results in an odd hybrid. 222 * 223 * XXX - ip_checkinterface currently must be disabled if you use ipnat 224 * to translate the destination address to another local interface. 225 * 226 * XXX - ip_checkinterface must be disabled if you add IP aliases 227 * to the loopback interface instead of the interface where the 228 * packets for those addresses are received. 229 */ 230static int ip_checkinterface = 0; 231SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW | CTLFLAG_LOCKED, 232 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 233 234#if DIAGNOSTIC 235static int ipprintfs = 0; 236#endif 237 238extern int in_proto_count; 239extern struct domain inetdomain; 240extern struct protosw inetsw[]; 241struct protosw *ip_protox[IPPROTO_MAX]; 242static int ipqmaxlen = IFQ_MAXLEN; 243 244static lck_grp_attr_t *in_ifaddr_rwlock_grp_attr; 245static lck_grp_t *in_ifaddr_rwlock_grp; 246static lck_attr_t *in_ifaddr_rwlock_attr; 247decl_lck_rw_data(, in_ifaddr_rwlock_data); 248lck_rw_t *in_ifaddr_rwlock = &in_ifaddr_rwlock_data; 249 250/* Protected by in_ifaddr_rwlock */ 251struct in_ifaddrhead in_ifaddrhead; /* first inet address */ 252struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */ 253 254#define INADDR_NHASH 61 255static u_int32_t inaddr_nhash; /* hash table size */ 256static u_int32_t inaddr_hashp; /* next largest prime */ 257 258struct ifqueue ipintrq; 259SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW | CTLFLAG_LOCKED, 260 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue"); 261SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD | CTLFLAG_LOCKED, 262 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue"); 263 264struct ipstat ipstat; 265SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD | CTLFLAG_LOCKED, 266 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)"); 267 268/* Packet reassembly stuff */ 269#define IPREASS_NHASH_LOG2 6 270#define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 271#define IPREASS_HMASK (IPREASS_NHASH - 1) 272#define IPREASS_HASH(x,y) \ 273 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 274 275static struct ipq ipq[IPREASS_NHASH]; 276static TAILQ_HEAD(ipq_list, ipq) ipq_list = 277 TAILQ_HEAD_INITIALIZER(ipq_list); 278const int ipintrq_present = 1; 279lck_mtx_t *ip_mutex; 280lck_attr_t *ip_mutex_attr; 281lck_grp_t *ip_mutex_grp; 282lck_grp_attr_t *ip_mutex_grp_attr; 283lck_mtx_t *inet_domain_mutex; 284 285#if IPCTL_DEFMTU 286SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW | CTLFLAG_LOCKED, 287 &ip_mtu, 0, "Default MTU"); 288#endif 289 290#if IPSTEALTH 291static int ipstealth = 0; 292SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW | CTLFLAG_LOCKED, 293 &ipstealth, 0, ""); 294#endif 295 296 297/* Firewall hooks */ 298#if IPFIREWALL 299ip_fw_chk_t *ip_fw_chk_ptr; 300int fw_enable = 1; 301int fw_bypass = 1; 302int fw_one_pass = 0; 303#endif /* IPFIREWALL */ 304 305#if DUMMYNET 306ip_dn_io_t *ip_dn_io_ptr; 307#endif 308 309SYSCTL_NODE(_net_inet_ip, OID_AUTO, linklocal, CTLFLAG_RW | CTLFLAG_LOCKED, 0, "link local"); 310 311struct ip_linklocal_stat ip_linklocal_stat; 312SYSCTL_STRUCT(_net_inet_ip_linklocal, OID_AUTO, stat, CTLFLAG_RD | CTLFLAG_LOCKED, 313 &ip_linklocal_stat, ip_linklocal_stat, 314 "Number of link local packets with TTL less than 255"); 315 316SYSCTL_NODE(_net_inet_ip_linklocal, OID_AUTO, in, CTLFLAG_RW | CTLFLAG_LOCKED, 0, "link local input"); 317 318int ip_linklocal_in_allowbadttl = 1; 319SYSCTL_INT(_net_inet_ip_linklocal_in, OID_AUTO, allowbadttl, CTLFLAG_RW | CTLFLAG_LOCKED, 320 &ip_linklocal_in_allowbadttl, 0, 321 "Allow incoming link local packets with TTL less than 255"); 322 323 324/* 325 * We need to save the IP options in case a protocol wants to respond 326 * to an incoming packet over the same route if the packet got here 327 * using IP source routing. This allows connection establishment and 328 * maintenance when the remote end is on a network that is not known 329 * to us. 330 */ 331static int ip_nhops = 0; 332static struct ip_srcrt { 333 struct in_addr dst; /* final destination */ 334 char nop; /* one NOP to align */ 335 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 336 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 337} ip_srcrt; 338 339static void in_ifaddrhashtbl_init(void); 340static void save_rte(u_char *, struct in_addr); 341static int ip_dooptions(struct mbuf *, int, struct sockaddr_in *); 342static void ip_forward(struct mbuf *, int, struct sockaddr_in *); 343static void ip_freef(struct ipq *); 344#if IPDIVERT 345#ifdef IPDIVERT_44 346static struct mbuf *ip_reass(struct mbuf *, 347 struct ipq *, struct ipq *, u_int32_t *, u_int16_t *); 348#else 349static struct mbuf *ip_reass(struct mbuf *, 350 struct ipq *, struct ipq *, u_int16_t *, u_int16_t *); 351#endif 352#else 353static struct mbuf *ip_reass(struct mbuf *, struct ipq *, struct ipq *); 354#endif 355static void ip_fwd_route_copyout(struct ifnet *, struct route *); 356static void ip_fwd_route_copyin(struct ifnet *, struct route *); 357void ipintr(void); 358void in_dinit(void); 359static inline u_short ip_cksum(struct mbuf *, int); 360 361#if RANDOM_IP_ID 362extern u_short ip_id; 363 364int ip_use_randomid = 1; 365SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW | CTLFLAG_LOCKED, 366 &ip_use_randomid, 0, "Randomize IP packets IDs"); 367#endif 368 369/* 370 * On platforms which require strict alignment (currently for anything but 371 * i386 or x86_64), check if the IP header pointer is 32-bit aligned; if not, 372 * copy the contents of the mbuf chain into a new chain, and free the original 373 * one. Create some head room in the first mbuf of the new chain, in case 374 * it's needed later on. 375 */ 376#if defined(__i386__) || defined(__x86_64__) 377#define IP_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do { } while (0) 378#else /* !__i386__ && !__x86_64__ */ 379#define IP_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do { \ 380 if (!IP_HDR_ALIGNED_P(mtod(_m, caddr_t))) { \ 381 struct mbuf *_n; \ 382 struct ifnet *__ifp = (_ifp); \ 383 atomic_add_64(&(__ifp)->if_alignerrs, 1); \ 384 if (((_m)->m_flags & M_PKTHDR) && \ 385 (_m)->m_pkthdr.header != NULL) \ 386 (_m)->m_pkthdr.header = NULL; \ 387 _n = m_defrag_offset(_m, max_linkhdr, M_NOWAIT); \ 388 if (_n == NULL) { \ 389 atomic_add_32(&ipstat.ips_toosmall, 1); \ 390 m_freem(_m); \ 391 (_m) = NULL; \ 392 _action \ 393 } else { \ 394 VERIFY(_n != (_m)); \ 395 (_m) = _n; \ 396 } \ 397 } \ 398} while (0) 399#endif /* !__i386__ && !__x86_64__ */ 400 401/* 402 * IP initialization: fill in IP protocol switch table. 403 * All protocols not implemented in kernel go to raw IP protocol handler. 404 */ 405void 406ip_init(void) 407{ 408 struct protosw *pr; 409 int i; 410 static int ip_initialized = 0; 411 412 if (!ip_initialized) 413 { 414 PE_parse_boot_argn("net.inet.ip.scopedroute", 415 &ip_doscopedroute, sizeof (ip_doscopedroute)); 416 417 in_ifaddr_init(); 418 419 in_ifaddr_rwlock_grp_attr = lck_grp_attr_alloc_init(); 420 in_ifaddr_rwlock_grp = lck_grp_alloc_init("in_ifaddr_rwlock", 421 in_ifaddr_rwlock_grp_attr); 422 in_ifaddr_rwlock_attr = lck_attr_alloc_init(); 423 lck_rw_init(in_ifaddr_rwlock, in_ifaddr_rwlock_grp, 424 in_ifaddr_rwlock_attr); 425 426 TAILQ_INIT(&in_ifaddrhead); 427 in_ifaddrhashtbl_init(); 428 429 ip_moptions_init(); 430 431 pr = pffindproto_locked(PF_INET, IPPROTO_RAW, SOCK_RAW); 432 if (pr == 0) 433 panic("ip_init"); 434 for (i = 0; i < IPPROTO_MAX; i++) 435 ip_protox[i] = pr; 436 for (pr = inetdomain.dom_protosw; pr; pr = pr->pr_next) { 437 if (pr->pr_domain == NULL) 438 continue; /* If uninitialized, skip */ 439 if (pr->pr_domain->dom_family == PF_INET && 440 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) 441 ip_protox[pr->pr_protocol] = pr; 442 } 443 for (i = 0; i < IPREASS_NHASH; i++) 444 ipq[i].next = ipq[i].prev = &ipq[i]; 445 446 maxnipq = nmbclusters / 32; 447 maxfrags = maxnipq * 2; 448 maxfragsperpacket = 128; /* enough for 64k in 512 byte fragments */ 449 450#if RANDOM_IP_ID 451 { 452 struct timeval timenow; 453 getmicrotime(&timenow); 454 ip_id = timenow.tv_sec & 0xffff; 455 } 456#endif 457 ipintrq.ifq_maxlen = ipqmaxlen; 458 459 ipf_init(); 460 461 ip_mutex_grp_attr = lck_grp_attr_alloc_init(); 462 463 ip_mutex_grp = lck_grp_alloc_init("ip", ip_mutex_grp_attr); 464 465 ip_mutex_attr = lck_attr_alloc_init(); 466 467 if ((ip_mutex = lck_mtx_alloc_init(ip_mutex_grp, ip_mutex_attr)) == NULL) { 468 printf("ip_init: can't alloc ip_mutex\n"); 469 return; 470 } 471 472#if IPSEC 473 474 sadb_stat_mutex_grp_attr = lck_grp_attr_alloc_init(); 475 sadb_stat_mutex_grp = lck_grp_alloc_init("sadb_stat", sadb_stat_mutex_grp_attr); 476 sadb_stat_mutex_attr = lck_attr_alloc_init(); 477 478 lck_mtx_init(sadb_stat_mutex, sadb_stat_mutex_grp, sadb_stat_mutex_attr); 479 480#endif 481 arp_init(); 482 483 ip_initialized = 1; 484 } 485} 486 487/* 488 * Initialize IPv4 source address hash table. 489 */ 490static void 491in_ifaddrhashtbl_init(void) 492{ 493 int i, k, p; 494 495 if (in_ifaddrhashtbl != NULL) 496 return; 497 498 PE_parse_boot_argn("inaddr_nhash", &inaddr_nhash, sizeof (inaddr_nhash)); 499 if (inaddr_nhash == 0) 500 inaddr_nhash = INADDR_NHASH; 501 502 MALLOC(in_ifaddrhashtbl, struct in_ifaddrhashhead *, 503 inaddr_nhash * sizeof (*in_ifaddrhashtbl), 504 M_IFADDR, M_WAITOK | M_ZERO); 505 if (in_ifaddrhashtbl == NULL) 506 panic("in_ifaddrhashtbl_init allocation failed"); 507 508 /* 509 * Generate the next largest prime greater than inaddr_nhash. 510 */ 511 k = (inaddr_nhash % 2 == 0) ? inaddr_nhash + 1 : inaddr_nhash + 2; 512 for (;;) { 513 p = 1; 514 for (i = 3; i * i <= k; i += 2) { 515 if (k % i == 0) 516 p = 0; 517 } 518 if (p == 1) 519 break; 520 k += 2; 521 } 522 inaddr_hashp = k; 523} 524 525u_int32_t 526inaddr_hashval(u_int32_t key) 527{ 528 /* 529 * The hash index is the computed prime times the key modulo 530 * the hash size, as documented in "Introduction to Algorithms" 531 * (Cormen, Leiserson, Rivest). 532 */ 533 if (inaddr_nhash > 1) 534 return ((key * inaddr_hashp) % inaddr_nhash); 535 else 536 return (0); 537} 538 539static void 540ip_proto_input( 541 protocol_family_t __unused protocol, 542 mbuf_t packet_list) 543{ 544 mbuf_t packet; 545 int how_many = 0 ; 546 547 /* ip_input should handle a list of packets but does not yet */ 548 549 for (packet = packet_list; packet; packet = packet_list) { 550 how_many++; 551 packet_list = mbuf_nextpkt(packet); 552 mbuf_setnextpkt(packet, NULL); 553 ip_input(packet); 554 } 555} 556 557/* Initialize the PF_INET domain, and add in the pre-defined protos */ 558void 559in_dinit(void) 560{ 561 int i; 562 struct protosw *pr; 563 struct domain *dp; 564 static int inetdomain_initted = 0; 565 566 if (!inetdomain_initted) 567 { 568 dp = &inetdomain; 569 dp->dom_flags = DOM_REENTRANT; 570 571 for (i=0, pr = &inetsw[0]; i<in_proto_count; i++, pr++) 572 net_add_proto(pr, dp); 573 inet_domain_mutex = dp->dom_mtx; 574 inetdomain_initted = 1; 575 576 domain_proto_mtx_unlock(TRUE); 577 proto_register_input(PF_INET, ip_proto_input, NULL, 1); 578 domain_proto_mtx_lock(); 579 } 580} 581 582void 583ip_proto_dispatch_in_wrapper(struct mbuf *m, int hlen, u_int8_t proto) 584{ 585 ip_proto_dispatch_in(m, hlen, proto, 0); 586} 587 588__private_extern__ void 589ip_proto_dispatch_in(struct mbuf *m, int hlen, u_int8_t proto, 590 ipfilter_t inject_ipfref) 591{ 592 struct ipfilter *filter; 593 int seen = (inject_ipfref == 0); 594 int changed_header = 0; 595 struct ip *ip; 596 void (*pr_input)(struct mbuf *, int len); 597 598 if (!TAILQ_EMPTY(&ipv4_filters)) { 599 ipf_ref(); 600 TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) { 601 if (seen == 0) { 602 if ((struct ipfilter *)inject_ipfref == filter) 603 seen = 1; 604 } else if (filter->ipf_filter.ipf_input) { 605 errno_t result; 606 607 if (changed_header == 0) { 608 /* 609 * Perform IP header alignment fixup, 610 * if needed, before passing packet 611 * into filter(s). 612 */ 613 IP_HDR_ALIGNMENT_FIXUP(m, 614 m->m_pkthdr.rcvif, 615 ipf_unref(); return;); 616 617 changed_header = 1; 618 ip = mtod(m, struct ip *); 619 ip->ip_len = htons(ip->ip_len + hlen); 620 ip->ip_off = htons(ip->ip_off); 621 ip->ip_sum = 0; 622 ip->ip_sum = in_cksum(m, hlen); 623 } 624 result = filter->ipf_filter.ipf_input( 625 filter->ipf_filter.cookie, (mbuf_t*)&m, 626 hlen, proto); 627 if (result == EJUSTRETURN) { 628 ipf_unref(); 629 return; 630 } 631 if (result != 0) { 632 ipf_unref(); 633 m_freem(m); 634 return; 635 } 636 } 637 } 638 ipf_unref(); 639 } 640 641 /* Perform IP header alignment fixup (post-filters), if needed */ 642 IP_HDR_ALIGNMENT_FIXUP(m, m->m_pkthdr.rcvif, return;); 643 644 /* 645 * If there isn't a specific lock for the protocol 646 * we're about to call, use the generic lock for AF_INET. 647 * otherwise let the protocol deal with its own locking 648 */ 649 ip = mtod(m, struct ip *); 650 651 if (changed_header) { 652 ip->ip_len = ntohs(ip->ip_len) - hlen; 653 ip->ip_off = ntohs(ip->ip_off); 654 } 655 656 if ((pr_input = ip_protox[ip->ip_p]->pr_input) == NULL) { 657 m_freem(m); 658 } else if (!(ip_protox[ip->ip_p]->pr_flags & PR_PROTOLOCK)) { 659 lck_mtx_lock(inet_domain_mutex); 660 pr_input(m, hlen); 661 lck_mtx_unlock(inet_domain_mutex); 662 } else { 663 pr_input(m, hlen); 664 } 665} 666 667/* 668 * Ip input routine. Checksum and byte swap header. If fragmented 669 * try to reassemble. Process options. Pass to next level. 670 */ 671void 672ip_input(struct mbuf *m) 673{ 674 struct ip *ip; 675 struct ipq *fp; 676 struct in_ifaddr *ia = NULL; 677 unsigned int hlen, checkif; 678 u_short sum = 0; 679 struct in_addr pkt_dst; 680#if IPFIREWALL 681 int i; 682 u_int32_t div_info = 0; /* packet divert/tee info */ 683#endif 684#if IPFIREWALL || DUMMYNET 685 struct ip_fw_args args; 686 struct m_tag *tag; 687#endif 688 ipfilter_t inject_filter_ref = 0; 689 690 /* Check if the mbuf is still valid after interface filter processing */ 691 MBUF_INPUT_CHECK(m, m->m_pkthdr.rcvif); 692 693 /* Perform IP header alignment fixup, if needed */ 694 IP_HDR_ALIGNMENT_FIXUP(m, m->m_pkthdr.rcvif, goto bad;); 695 696#if IPFIREWALL || DUMMYNET 697 bzero(&args, sizeof(struct ip_fw_args)); 698 699 /* 700 * Don't bother searching for tag(s) if there's none. 701 */ 702 if (SLIST_EMPTY(&m->m_pkthdr.tags)) 703 goto ipfw_tags_done; 704 705 /* Grab info from mtags prepended to the chain */ 706#if DUMMYNET 707 if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, 708 KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) { 709 struct dn_pkt_tag *dn_tag; 710 711 dn_tag = (struct dn_pkt_tag *)(tag+1); 712 args.fwa_ipfw_rule = dn_tag->dn_ipfw_rule; 713 args.fwa_pf_rule = dn_tag->dn_pf_rule; 714 715 m_tag_delete(m, tag); 716 } 717#endif /* DUMMYNET */ 718 719#if IPDIVERT 720 if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, 721 KERNEL_TAG_TYPE_DIVERT, NULL)) != NULL) { 722 struct divert_tag *div_tag; 723 724 div_tag = (struct divert_tag *)(tag+1); 725 args.fwa_divert_rule = div_tag->cookie; 726 727 m_tag_delete(m, tag); 728 } 729#endif 730 731 if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, 732 KERNEL_TAG_TYPE_IPFORWARD, NULL)) != NULL) { 733 struct ip_fwd_tag *ipfwd_tag; 734 735 ipfwd_tag = (struct ip_fwd_tag *)(tag+1); 736 args.fwa_next_hop = ipfwd_tag->next_hop; 737 738 m_tag_delete(m, tag); 739 } 740 741#if DIAGNOSTIC 742 if (m == NULL || (m->m_flags & M_PKTHDR) == 0) 743 panic("ip_input no HDR"); 744#endif 745 746#if DUMMYNET 747 if (args.fwa_ipfw_rule || args.fwa_pf_rule) { /* dummynet already filtered us */ 748 ip = mtod(m, struct ip *); 749 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 750 inject_filter_ref = ipf_get_inject_filter(m); 751#if IPFIREWALL 752 if (args.fwa_ipfw_rule) 753 goto iphack; 754#endif /* IPFIREWALL */ 755 if (args.fwa_pf_rule) 756 goto check_with_pf; 757 } 758#endif /* DUMMYNET */ 759ipfw_tags_done: 760#endif /* IPFIREWALL || DUMMYNET*/ 761 762 /* 763 * No need to process packet twice if we've already seen it. 764 */ 765 if (!SLIST_EMPTY(&m->m_pkthdr.tags)) 766 inject_filter_ref = ipf_get_inject_filter(m); 767 if (inject_filter_ref != 0) { 768 ip = mtod(m, struct ip *); 769 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 770 771 DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL, 772 struct ip *, ip, struct ifnet *, m->m_pkthdr.rcvif, 773 struct ip *, ip, struct ip6_hdr *, NULL); 774 775 ip->ip_len = ntohs(ip->ip_len) - hlen; 776 ip->ip_off = ntohs(ip->ip_off); 777 ip_proto_dispatch_in(m, hlen, ip->ip_p, inject_filter_ref); 778 return; 779 } 780 781 OSAddAtomic(1, &ipstat.ips_total); 782 if (m->m_pkthdr.len < sizeof(struct ip)) 783 goto tooshort; 784 785 if (m->m_len < sizeof (struct ip) && 786 (m = m_pullup(m, sizeof (struct ip))) == 0) { 787 OSAddAtomic(1, &ipstat.ips_toosmall); 788 return; 789 } 790 ip = mtod(m, struct ip *); 791 792 KERNEL_DEBUG(DBG_LAYER_BEG, ip->ip_dst.s_addr, 793 ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len); 794 795 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 796 OSAddAtomic(1, &ipstat.ips_badvers); 797 goto bad; 798 } 799 800 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 801 if (hlen < sizeof(struct ip)) { /* minimum header length */ 802 OSAddAtomic(1, &ipstat.ips_badhlen); 803 goto bad; 804 } 805 if (hlen > m->m_len) { 806 if ((m = m_pullup(m, hlen)) == 0) { 807 OSAddAtomic(1, &ipstat.ips_badhlen); 808 return; 809 } 810 ip = mtod(m, struct ip *); 811 } 812 813 /* 127/8 must not appear on wire - RFC1122 */ 814 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 815 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 816 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) { 817 OSAddAtomic(1, &ipstat.ips_badaddr); 818 goto bad; 819 } 820 } 821 822 /* IPv4 Link-Local Addresses as defined in <draft-ietf-zeroconf-ipv4-linklocal-05.txt> */ 823 if ((IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) || 824 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr)))) { 825 ip_linklocal_stat.iplls_in_total++; 826 if (ip->ip_ttl != MAXTTL) { 827 OSAddAtomic(1, &ip_linklocal_stat.iplls_in_badttl); 828 /* Silently drop link local traffic with bad TTL */ 829 if (!ip_linklocal_in_allowbadttl) 830 goto bad; 831 } 832 } 833 834 sum = ip_cksum(m, hlen); 835 if (sum) { 836 goto bad; 837 } 838 839 DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL, 840 struct ip *, ip, struct ifnet *, m->m_pkthdr.rcvif, 841 struct ip *, ip, struct ip6_hdr *, NULL); 842 843 /* 844 * Naively assume we can attribute inbound data to the route we would 845 * use to send to this destination. Asymetric routing breaks this 846 * assumption, but it still allows us to account for traffic from 847 * a remote node in the routing table. 848 * this has a very significant performance impact so we bypass 849 * if nstat_collect is disabled. We may also bypass if the 850 * protocol is tcp in the future because tcp will have a route that 851 * we can use to attribute the data to. That does mean we would not 852 * account for forwarded tcp traffic. 853 */ 854 if (nstat_collect) { 855 struct rtentry *rt = 856 ifnet_cached_rtlookup_inet(m->m_pkthdr.rcvif, ip->ip_src); 857 if (rt != NULL) { 858 nstat_route_rx(rt, 1, m->m_pkthdr.len, 0); 859 rtfree(rt); 860 } 861 } 862 863 /* 864 * Convert fields to host representation. 865 */ 866#if BYTE_ORDER != BIG_ENDIAN 867 NTOHS(ip->ip_len); 868#endif 869 870 if (ip->ip_len < hlen) { 871 OSAddAtomic(1, &ipstat.ips_badlen); 872 goto bad; 873 } 874 875#if BYTE_ORDER != BIG_ENDIAN 876 NTOHS(ip->ip_off); 877#endif 878 /* 879 * Check that the amount of data in the buffers 880 * is as at least much as the IP header would have us expect. 881 * Trim mbufs if longer than we expect. 882 * Drop packet if shorter than we expect. 883 */ 884 if (m->m_pkthdr.len < ip->ip_len) { 885tooshort: 886 OSAddAtomic(1, &ipstat.ips_tooshort); 887 goto bad; 888 } 889 if (m->m_pkthdr.len > ip->ip_len) { 890 /* Invalidate hwcksuming */ 891 m->m_pkthdr.csum_flags = 0; 892 m->m_pkthdr.csum_data = 0; 893 894 if (m->m_len == m->m_pkthdr.len) { 895 m->m_len = ip->ip_len; 896 m->m_pkthdr.len = ip->ip_len; 897 } else 898 m_adj(m, ip->ip_len - m->m_pkthdr.len); 899 } 900 901 902#if DUMMYNET 903check_with_pf: 904#endif 905#if PF 906 /* Invoke inbound packet filter */ 907 if (PF_IS_ENABLED) { 908 int error; 909#if DUMMYNET 910 error = pf_af_hook(m->m_pkthdr.rcvif, NULL, &m, AF_INET, TRUE, &args); 911#else 912 error = pf_af_hook(m->m_pkthdr.rcvif, NULL, &m, AF_INET, TRUE, NULL); 913#endif /* DUMMYNET */ 914 if (error != 0 || m == NULL) { 915 if (m != NULL) { 916 panic("%s: unexpected packet %p\n", __func__, m); 917 /* NOTREACHED */ 918 } 919 /* Already freed by callee */ 920 return; 921 } 922 ip = mtod(m, struct ip *); 923 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 924 } 925#endif /* PF */ 926 927#if IPSEC 928 if (ipsec_bypass == 0 && ipsec_gethist(m, NULL)) 929 goto pass; 930#endif 931 932#if IPFIREWALL 933#if DUMMYNET 934iphack: 935#endif /* DUMMYNET */ 936 /* 937 * Check if we want to allow this packet to be processed. 938 * Consider it to be bad if not. 939 */ 940 if (fw_enable && IPFW_LOADED) { 941#if IPFIREWALL_FORWARD 942 /* 943 * If we've been forwarded from the output side, then 944 * skip the firewall a second time 945 */ 946 if (args.fwa_next_hop) 947 goto ours; 948#endif /* IPFIREWALL_FORWARD */ 949 950 args.fwa_m = m; 951 952 i = ip_fw_chk_ptr(&args); 953 m = args.fwa_m; 954 955 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */ 956 if (m) 957 m_freem(m); 958 return; 959 } 960 ip = mtod(m, struct ip *); /* just in case m changed */ 961 962 if (i == 0 && args.fwa_next_hop == NULL) { /* common case */ 963 goto pass; 964 } 965#if DUMMYNET 966 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) { 967 /* Send packet to the appropriate pipe */ 968 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args, DN_CLIENT_IPFW); 969 return; 970 } 971#endif /* DUMMYNET */ 972#if IPDIVERT 973 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) { 974 /* Divert or tee packet */ 975 div_info = i; 976 goto ours; 977 } 978#endif 979#if IPFIREWALL_FORWARD 980 if (i == 0 && args.fwa_next_hop != NULL) { 981 goto pass; 982 } 983#endif 984 /* 985 * if we get here, the packet must be dropped 986 */ 987 m_freem(m); 988 return; 989 } 990#endif /* IPFIREWALL */ 991pass: 992 993 /* 994 * Process options and, if not destined for us, 995 * ship it on. ip_dooptions returns 1 when an 996 * error was detected (causing an icmp message 997 * to be sent and the original packet to be freed). 998 */ 999 ip_nhops = 0; /* for source routed packets */ 1000#if IPFIREWALL 1001 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.fwa_next_hop)) { 1002#else 1003 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, NULL)) { 1004#endif 1005 return; 1006 } 1007 1008 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 1009 * matter if it is destined to another node, or whether it is 1010 * a multicast one, RSVP wants it! and prevents it from being forwarded 1011 * anywhere else. Also checks if the rsvp daemon is running before 1012 * grabbing the packet. 1013 */ 1014 if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 1015 goto ours; 1016 1017 /* 1018 * Check our list of addresses, to see if the packet is for us. 1019 * If we don't have any addresses, assume any unicast packet 1020 * we receive might be for us (and let the upper layers deal 1021 * with it). 1022 */ 1023 if (TAILQ_EMPTY(&in_ifaddrhead) && 1024 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 1025 goto ours; 1026 1027 /* 1028 * Cache the destination address of the packet; this may be 1029 * changed by use of 'ipfw fwd'. 1030 */ 1031#if IPFIREWALL 1032 pkt_dst = args.fwa_next_hop == NULL ? 1033 ip->ip_dst : args.fwa_next_hop->sin_addr; 1034#else 1035 pkt_dst = ip->ip_dst; 1036#endif 1037 1038 /* 1039 * Enable a consistency check between the destination address 1040 * and the arrival interface for a unicast packet (the RFC 1122 1041 * strong ES model) if IP forwarding is disabled and the packet 1042 * is not locally generated and the packet is not subject to 1043 * 'ipfw fwd'. 1044 * 1045 * XXX - Checking also should be disabled if the destination 1046 * address is ipnat'ed to a different interface. 1047 * 1048 * XXX - Checking is incompatible with IP aliases added 1049 * to the loopback interface instead of the interface where 1050 * the packets are received. 1051 */ 1052 checkif = ip_checkinterface && (ipforwarding == 0) && 1053 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) 1054#if IPFIREWALL 1055 && (args.fwa_next_hop == NULL); 1056#else 1057 ; 1058#endif 1059 1060 /* 1061 * Check for exact addresses in the hash bucket. 1062 */ 1063 lck_rw_lock_shared(in_ifaddr_rwlock); 1064 TAILQ_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 1065 /* 1066 * If the address matches, verify that the packet 1067 * arrived via the correct interface if checking is 1068 * enabled. 1069 */ 1070 IFA_LOCK_SPIN(&ia->ia_ifa); 1071 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 1072 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) { 1073 IFA_UNLOCK(&ia->ia_ifa); 1074 lck_rw_done(in_ifaddr_rwlock); 1075 goto ours; 1076 } 1077 IFA_UNLOCK(&ia->ia_ifa); 1078 } 1079 lck_rw_done(in_ifaddr_rwlock); 1080 1081 /* 1082 * Check for broadcast addresses. 1083 * 1084 * Only accept broadcast packets that arrive via the matching 1085 * interface. Reception of forwarded directed broadcasts would be 1086 * handled via ip_forward() and ether_frameout() with the loopback 1087 * into the stack for SIMPLEX interfaces handled by ether_frameout(). 1088 */ 1089 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 1090 struct ifaddr *ifa; 1091 struct ifnet *ifp = m->m_pkthdr.rcvif; 1092 ifnet_lock_shared(ifp); 1093 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1094 IFA_LOCK_SPIN(ifa); 1095 if (ifa->ifa_addr->sa_family != AF_INET) { 1096 IFA_UNLOCK(ifa); 1097 continue; 1098 } 1099 ia = ifatoia(ifa); 1100 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 1101 pkt_dst.s_addr || ia->ia_netbroadcast.s_addr == 1102 pkt_dst.s_addr) { 1103 IFA_UNLOCK(ifa); 1104 ifnet_lock_done(ifp); 1105 goto ours; 1106 } 1107 IFA_UNLOCK(ifa); 1108 } 1109 ifnet_lock_done(ifp); 1110 } 1111 1112 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 1113 struct in_multi *inm; 1114 struct ifnet *ifp = m->m_pkthdr.rcvif; 1115#if MROUTING 1116 if (ip_mrouter) { 1117 /* 1118 * If we are acting as a multicast router, all 1119 * incoming multicast packets are passed to the 1120 * kernel-level multicast forwarding function. 1121 * The packet is returned (relatively) intact; if 1122 * ip_mforward() returns a non-zero value, the packet 1123 * must be discarded, else it may be accepted below. 1124 */ 1125 lck_mtx_lock(ip_mutex); 1126 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 1127 OSAddAtomic(1, &ipstat.ips_cantforward); 1128 m_freem(m); 1129 lck_mtx_unlock(ip_mutex); 1130 return; 1131 } 1132 1133 /* 1134 * The process-level routing daemon needs to receive 1135 * all multicast IGMP packets, whether or not this 1136 * host belongs to their destination groups. 1137 */ 1138 if (ip->ip_p == IPPROTO_IGMP) 1139 goto ours; 1140 OSAddAtomic(1, &ipstat.ips_forward); 1141 } 1142#endif /* MROUTING */ 1143 /* 1144 * See if we belong to the destination multicast group on the 1145 * arrival interface. 1146 */ 1147 in_multihead_lock_shared(); 1148 IN_LOOKUP_MULTI(&ip->ip_dst, ifp, inm); 1149 in_multihead_lock_done(); 1150 if (inm == NULL) { 1151 OSAddAtomic(1, &ipstat.ips_notmember); 1152 m_freem(m); 1153 return; 1154 } 1155 INM_REMREF(inm); 1156 goto ours; 1157 } 1158 if (ip->ip_dst.s_addr == (u_int32_t)INADDR_BROADCAST) 1159 goto ours; 1160 if (ip->ip_dst.s_addr == INADDR_ANY) 1161 goto ours; 1162 1163 /* Allow DHCP/BootP responses through */ 1164 if (m->m_pkthdr.rcvif != NULL 1165 && (m->m_pkthdr.rcvif->if_eflags & IFEF_AUTOCONFIGURING) 1166 && hlen == sizeof(struct ip) 1167 && ip->ip_p == IPPROTO_UDP) { 1168 struct udpiphdr *ui; 1169 if (m->m_len < sizeof(struct udpiphdr) 1170 && (m = m_pullup(m, sizeof(struct udpiphdr))) == 0) { 1171 OSAddAtomic(1, &udpstat.udps_hdrops); 1172 return; 1173 } 1174 ui = mtod(m, struct udpiphdr *); 1175 if (ntohs(ui->ui_dport) == IPPORT_BOOTPC) { 1176 goto ours; 1177 } 1178 ip = mtod(m, struct ip *); /* in case it changed */ 1179 } 1180 1181 /* 1182 * Not for us; forward if possible and desirable. 1183 */ 1184 if (ipforwarding == 0) { 1185 OSAddAtomic(1, &ipstat.ips_cantforward); 1186 m_freem(m); 1187 } else { 1188#if IPFIREWALL 1189 ip_forward(m, 0, args.fwa_next_hop); 1190#else 1191 ip_forward(m, 0, NULL); 1192#endif 1193 } 1194 return; 1195 1196ours: 1197 /* 1198 * If offset or IP_MF are set, must reassemble. 1199 * Otherwise, nothing need be done. 1200 * (We could look in the reassembly queue to see 1201 * if the packet was previously fragmented, 1202 * but it's not worth the time; just let them time out.) 1203 */ 1204 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) { 1205 1206 /* If maxnipq is 0, never accept fragments. */ 1207 if (maxnipq == 0) { 1208 1209 OSAddAtomic(1, &ipstat.ips_fragments); 1210 OSAddAtomic(1, &ipstat.ips_fragdropped); 1211 goto bad; 1212 } 1213 1214 /* 1215 * If we will exceed the number of fragments in queues, timeout the 1216 * oldest fragemented packet to make space. 1217 */ 1218 lck_mtx_lock(ip_mutex); 1219 if (currentfrags >= maxfrags) { 1220 fp = TAILQ_LAST(&ipq_list, ipq_list); 1221 OSAddAtomic(fp->ipq_nfrags, &ipstat.ips_fragtimeout); 1222 1223 if (ip->ip_id == fp->ipq_id && 1224 ip->ip_src.s_addr == fp->ipq_src.s_addr && 1225 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 1226 ip->ip_p == fp->ipq_p) { 1227 /* 1228 * If we match the fragment queue we were going to 1229 * discard, drop this packet too. 1230 */ 1231 OSAddAtomic(1, &ipstat.ips_fragdropped); 1232 ip_freef(fp); 1233 lck_mtx_unlock(ip_mutex); 1234 goto bad; 1235 } 1236 1237 ip_freef(fp); 1238 } 1239 1240 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 1241 /* 1242 * Look for queue of fragments 1243 * of this datagram. 1244 */ 1245 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) 1246 if (ip->ip_id == fp->ipq_id && 1247 ip->ip_src.s_addr == fp->ipq_src.s_addr && 1248 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 1249#if CONFIG_MACF_NET 1250 mac_ipq_label_compare(m, fp) && 1251#endif 1252 ip->ip_p == fp->ipq_p) 1253 goto found; 1254 1255 /* 1256 * Enforce upper bound on number of fragmented packets 1257 * for which we attempt reassembly; 1258 * If maxnipq is -1, accept all fragments without limitation. 1259 */ 1260 if ((nipq > maxnipq) && (maxnipq > 0)) { 1261 /* 1262 * drop the oldest fragment before proceeding further 1263 */ 1264 fp = TAILQ_LAST(&ipq_list, ipq_list); 1265 OSAddAtomic(fp->ipq_nfrags, &ipstat.ips_fragtimeout); 1266 ip_freef(fp); 1267 } 1268 1269 fp = NULL; 1270 1271found: 1272 /* 1273 * Adjust ip_len to not reflect header, 1274 * convert offset of this to bytes. 1275 */ 1276 ip->ip_len -= hlen; 1277 if (ip->ip_off & IP_MF) { 1278 /* 1279 * Make sure that fragments have a data length 1280 * that's a non-zero multiple of 8 bytes. 1281 */ 1282 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 1283 OSAddAtomic(1, &ipstat.ips_toosmall); 1284 lck_mtx_unlock(ip_mutex); 1285 goto bad; 1286 } 1287 m->m_flags |= M_FRAG; 1288 } else { 1289 /* Clear the flag in case packet comes from loopback */ 1290 m->m_flags &= ~M_FRAG; 1291 } 1292 ip->ip_off <<= 3; 1293 1294 /* 1295 * Attempt reassembly; if it succeeds, proceed. 1296 * ip_reass() will return a different mbuf, and update 1297 * the divert info in div_info and args.fwa_divert_rule. 1298 */ 1299 OSAddAtomic(1, &ipstat.ips_fragments); 1300 m->m_pkthdr.header = ip; 1301#if IPDIVERT 1302 m = ip_reass(m, fp, &ipq[sum], 1303 (u_int16_t *)&div_info, &args.fwa_divert_rule); 1304#else 1305 m = ip_reass(m, fp, &ipq[sum]); 1306#endif 1307 if (m == 0) { 1308 lck_mtx_unlock(ip_mutex); 1309 return; 1310 } 1311 OSAddAtomic(1, &ipstat.ips_reassembled); 1312 ip = mtod(m, struct ip *); 1313 /* Get the header length of the reassembled packet */ 1314 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 1315 1316#if IPDIVERT 1317 /* Restore original checksum before diverting packet */ 1318 if (div_info != 0) { 1319 ip->ip_len += hlen; 1320 1321#if BYTE_ORDER != BIG_ENDIAN 1322 HTONS(ip->ip_len); 1323 HTONS(ip->ip_off); 1324#endif 1325 1326 ip->ip_sum = 0; 1327 ip->ip_sum = in_cksum(m, hlen); 1328 1329#if BYTE_ORDER != BIG_ENDIAN 1330 NTOHS(ip->ip_off); 1331 NTOHS(ip->ip_len); 1332#endif 1333 1334 ip->ip_len -= hlen; 1335 } 1336#endif 1337 lck_mtx_unlock(ip_mutex); 1338 } else 1339 ip->ip_len -= hlen; 1340 1341#if IPDIVERT 1342 /* 1343 * Divert or tee packet to the divert protocol if required. 1344 * 1345 * If div_info is zero then cookie should be too, so we shouldn't 1346 * need to clear them here. Assume divert_packet() does so also. 1347 */ 1348 if (div_info != 0) { 1349 struct mbuf *clone = NULL; 1350 1351 /* Clone packet if we're doing a 'tee' */ 1352 if ((div_info & IP_FW_PORT_TEE_FLAG) != 0) 1353 clone = m_dup(m, M_DONTWAIT); 1354 1355 /* Restore packet header fields to original values */ 1356 ip->ip_len += hlen; 1357 1358#if BYTE_ORDER != BIG_ENDIAN 1359 HTONS(ip->ip_len); 1360 HTONS(ip->ip_off); 1361#endif 1362 /* Deliver packet to divert input routine */ 1363 OSAddAtomic(1, &ipstat.ips_delivered); 1364 divert_packet(m, 1, div_info & 0xffff, args.fwa_divert_rule); 1365 1366 /* If 'tee', continue with original packet */ 1367 if (clone == NULL) { 1368 return; 1369 } 1370 m = clone; 1371 ip = mtod(m, struct ip *); 1372 } 1373#endif 1374 1375#if IPSEC 1376 /* 1377 * enforce IPsec policy checking if we are seeing last header. 1378 * note that we do not visit this with protocols with pcb layer 1379 * code - like udp/tcp/raw ip. 1380 */ 1381 if (ipsec_bypass == 0 && (ip_protox[ip->ip_p]->pr_flags & PR_LASTHDR) != 0) { 1382 if (ipsec4_in_reject(m, NULL)) { 1383 IPSEC_STAT_INCREMENT(ipsecstat.in_polvio); 1384 goto bad; 1385 } 1386 } 1387#endif 1388 1389 /* 1390 * Switch out to protocol's input routine. 1391 */ 1392 OSAddAtomic(1, &ipstat.ips_delivered); 1393 { 1394#if IPFIREWALL 1395 if (args.fwa_next_hop && ip->ip_p == IPPROTO_TCP) { 1396 /* TCP needs IPFORWARD info if available */ 1397 struct m_tag *fwd_tag; 1398 struct ip_fwd_tag *ipfwd_tag; 1399 1400 fwd_tag = m_tag_create(KERNEL_MODULE_TAG_ID, 1401 KERNEL_TAG_TYPE_IPFORWARD, sizeof (*ipfwd_tag), 1402 M_NOWAIT, m); 1403 if (fwd_tag == NULL) { 1404 goto bad; 1405 } 1406 1407 ipfwd_tag = (struct ip_fwd_tag *)(fwd_tag+1); 1408 ipfwd_tag->next_hop = args.fwa_next_hop; 1409 1410 m_tag_prepend(m, fwd_tag); 1411 1412 KERNEL_DEBUG(DBG_LAYER_END, ip->ip_dst.s_addr, 1413 ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len); 1414 1415 if (sw_lro) { 1416 m = tcp_lro(m, hlen); 1417 if (m == NULL) 1418 return; 1419 } 1420 /* TCP deals with its own locking */ 1421 ip_proto_dispatch_in(m, hlen, ip->ip_p, 0); 1422 } else { 1423 KERNEL_DEBUG(DBG_LAYER_END, ip->ip_dst.s_addr, 1424 ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len); 1425 1426 ip_proto_dispatch_in(m, hlen, ip->ip_p, 0); 1427 } 1428#else 1429 if ((sw_lro) && (ip->ip_p == IPPROTO_TCP)) { 1430 m = tcp_lro(m, hlen); 1431 if (m == NULL) 1432 return; 1433 } 1434 ip_proto_dispatch_in(m, hlen, ip->ip_p, 0); 1435#endif 1436 1437 return; 1438 } 1439bad: 1440 KERNEL_DEBUG(DBG_LAYER_END, 0,0,0,0,0); 1441 m_freem(m); 1442} 1443 1444/* 1445 * Take incoming datagram fragment and try to reassemble it into 1446 * whole datagram. If a chain for reassembly of this datagram already 1447 * exists, then it is given as fp; otherwise have to make a chain. 1448 * 1449 * When IPDIVERT enabled, keep additional state with each packet that 1450 * tells us if we need to divert or tee the packet we're building. 1451 */ 1452 1453static struct mbuf * 1454#if IPDIVERT 1455ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where, 1456#ifdef IPDIVERT_44 1457 u_int32_t *divinfo, 1458#else /* IPDIVERT_44 */ 1459 u_int16_t *divinfo, 1460#endif /* IPDIVERT_44 */ 1461 u_int16_t *divcookie) 1462#else /* IPDIVERT */ 1463ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where) 1464#endif /* IPDIVERT */ 1465{ 1466 struct ip *ip = mtod(m, struct ip *); 1467 struct mbuf *p = 0, *q, *nq; 1468 struct mbuf *t; 1469 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 1470 int i, next; 1471 u_int8_t ecn, ecn0; 1472 1473 lck_mtx_assert(ip_mutex, LCK_MTX_ASSERT_OWNED); 1474 /* 1475 * Presence of header sizes in mbufs 1476 * would confuse code below. 1477 */ 1478 m->m_data += hlen; 1479 m->m_len -= hlen; 1480 1481 if (m->m_pkthdr.csum_flags & CSUM_TCP_SUM16) 1482 m->m_pkthdr.csum_flags = 0; 1483 /* 1484 * If first fragment to arrive, create a reassembly queue. 1485 */ 1486 if (fp == 0) { 1487 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL) 1488 goto dropfrag; 1489 fp = mtod(t, struct ipq *); 1490#if CONFIG_MACF_NET 1491 if (mac_ipq_label_init(fp, M_NOWAIT) != 0) { 1492 m_free(t); 1493 fp = NULL; 1494 goto dropfrag; 1495 } 1496 mac_ipq_label_associate(m, fp); 1497#endif 1498 insque((void*)fp, (void*)where); 1499 nipq++; 1500 fp->ipq_nfrags = 1; 1501 fp->ipq_ttl = IPFRAGTTL; 1502 fp->ipq_p = ip->ip_p; 1503 fp->ipq_id = ip->ip_id; 1504 fp->ipq_src = ip->ip_src; 1505 fp->ipq_dst = ip->ip_dst; 1506 fp->ipq_frags = m; 1507 m->m_nextpkt = NULL; 1508#if IPDIVERT 1509#ifdef IPDIVERT_44 1510 fp->ipq_div_info = 0; 1511#else 1512 fp->ipq_divert = 0; 1513#endif 1514 fp->ipq_div_cookie = 0; 1515#endif 1516 TAILQ_INSERT_HEAD(&ipq_list, fp, ipq_list); 1517 goto inserted; 1518 } else { 1519 fp->ipq_nfrags++; 1520#if CONFIG_MACF_NET 1521 mac_ipq_label_update(m, fp); 1522#endif 1523 } 1524 1525#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1526 1527 /* 1528 * Handle ECN by comparing this segment with the first one; 1529 * if CE is set, do not lose CE. 1530 * drop if CE and not-ECT are mixed for the same packet. 1531 */ 1532 ecn = ip->ip_tos & IPTOS_ECN_MASK; 1533 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK; 1534 if (ecn == IPTOS_ECN_CE) { 1535 if (ecn0 == IPTOS_ECN_NOTECT) 1536 goto dropfrag; 1537 if (ecn0 != IPTOS_ECN_CE) 1538 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE; 1539 } 1540 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 1541 goto dropfrag; 1542 1543 /* 1544 * Find a segment which begins after this one does. 1545 */ 1546 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) 1547 if (GETIP(q)->ip_off > ip->ip_off) 1548 break; 1549 1550 /* 1551 * If there is a preceding segment, it may provide some of 1552 * our data already. If so, drop the data from the incoming 1553 * segment. If it provides all of our data, drop us, otherwise 1554 * stick new segment in the proper place. 1555 * 1556 * If some of the data is dropped from the the preceding 1557 * segment, then it's checksum is invalidated. 1558 */ 1559 if (p) { 1560 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1561 if (i > 0) { 1562 if (i >= ip->ip_len) 1563 goto dropfrag; 1564 m_adj(m, i); 1565 m->m_pkthdr.csum_flags = 0; 1566 ip->ip_off += i; 1567 ip->ip_len -= i; 1568 } 1569 m->m_nextpkt = p->m_nextpkt; 1570 p->m_nextpkt = m; 1571 } else { 1572 m->m_nextpkt = fp->ipq_frags; 1573 fp->ipq_frags = m; 1574 } 1575 1576 /* 1577 * While we overlap succeeding segments trim them or, 1578 * if they are completely covered, dequeue them. 1579 */ 1580 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1581 q = nq) { 1582 i = (ip->ip_off + ip->ip_len) - 1583 GETIP(q)->ip_off; 1584 if (i < GETIP(q)->ip_len) { 1585 GETIP(q)->ip_len -= i; 1586 GETIP(q)->ip_off += i; 1587 m_adj(q, i); 1588 q->m_pkthdr.csum_flags = 0; 1589 break; 1590 } 1591 nq = q->m_nextpkt; 1592 m->m_nextpkt = nq; 1593 OSAddAtomic(1, &ipstat.ips_fragdropped); 1594 fp->ipq_nfrags--; 1595 m_freem(q); 1596 } 1597 1598inserted: 1599 currentfrags++; 1600 1601#if IPDIVERT 1602 /* 1603 * Transfer firewall instructions to the fragment structure. 1604 * Only trust info in the fragment at offset 0. 1605 */ 1606 if (ip->ip_off == 0) { 1607#ifdef IPDIVERT_44 1608 fp->ipq_div_info = *divinfo; 1609#else 1610 fp->ipq_divert = *divinfo; 1611#endif 1612 fp->ipq_div_cookie = *divcookie; 1613 } 1614 *divinfo = 0; 1615 *divcookie = 0; 1616#endif 1617 1618 /* 1619 * Check for complete reassembly and perform frag per packet 1620 * limiting. 1621 * 1622 * Frag limiting is performed here so that the nth frag has 1623 * a chance to complete the packet before we drop the packet. 1624 * As a result, n+1 frags are actually allowed per packet, but 1625 * only n will ever be stored. (n = maxfragsperpacket.) 1626 * 1627 */ 1628 next = 0; 1629 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1630 if (GETIP(q)->ip_off != next) { 1631 if (fp->ipq_nfrags > maxfragsperpacket) { 1632 OSAddAtomic(fp->ipq_nfrags, &ipstat.ips_fragdropped); 1633 ip_freef(fp); 1634 } 1635 return (0); 1636 } 1637 next += GETIP(q)->ip_len; 1638 } 1639 /* Make sure the last packet didn't have the IP_MF flag */ 1640 if (p->m_flags & M_FRAG) { 1641 if (fp->ipq_nfrags > maxfragsperpacket) { 1642 OSAddAtomic(fp->ipq_nfrags, &ipstat.ips_fragdropped); 1643 ip_freef(fp); 1644 } 1645 return (0); 1646 } 1647 1648 /* 1649 * Reassembly is complete. Make sure the packet is a sane size. 1650 */ 1651 q = fp->ipq_frags; 1652 ip = GETIP(q); 1653 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1654 OSAddAtomic(1, &ipstat.ips_toolong); 1655 OSAddAtomic(fp->ipq_nfrags, &ipstat.ips_fragdropped); 1656 ip_freef(fp); 1657 return (0); 1658 } 1659 1660 /* 1661 * Concatenate fragments. 1662 */ 1663 m = q; 1664 t = m->m_next; 1665 m->m_next = 0; 1666 m_cat(m, t); 1667 nq = q->m_nextpkt; 1668 q->m_nextpkt = 0; 1669 for (q = nq; q != NULL; q = nq) { 1670 nq = q->m_nextpkt; 1671 q->m_nextpkt = NULL; 1672 if (q->m_pkthdr.csum_flags & CSUM_TCP_SUM16) 1673 m->m_pkthdr.csum_flags = 0; 1674 else { 1675 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1676 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1677 } 1678 m_cat(m, q); 1679 } 1680 1681#if IPDIVERT 1682 /* 1683 * Extract firewall instructions from the fragment structure. 1684 */ 1685#ifdef IPDIVERT_44 1686 *divinfo = fp->ipq_div_info; 1687#else 1688 *divinfo = fp->ipq_divert; 1689#endif 1690 *divcookie = fp->ipq_div_cookie; 1691#endif 1692 1693#if CONFIG_MACF_NET 1694 mac_mbuf_label_associate_ipq(fp, m); 1695 mac_ipq_label_destroy(fp); 1696#endif 1697 /* 1698 * Create header for new ip packet by 1699 * modifying header of first packet; 1700 * dequeue and discard fragment reassembly header. 1701 * Make header visible. 1702 */ 1703 ip->ip_len = next; 1704 ip->ip_src = fp->ipq_src; 1705 ip->ip_dst = fp->ipq_dst; 1706 remque((void*)fp); 1707 TAILQ_REMOVE(&ipq_list, fp, ipq_list); 1708 currentfrags -= fp->ipq_nfrags; 1709 nipq--; 1710 (void) m_free(dtom(fp)); 1711 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1712 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1713 /* some debugging cruft by sklower, below, will go away soon */ 1714 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1715 int plen = 0; 1716 for (t = m; t; t = t->m_next) 1717 plen += t->m_len; 1718 m->m_pkthdr.len = plen; 1719 } 1720 return (m); 1721 1722dropfrag: 1723#if IPDIVERT 1724 *divinfo = 0; 1725 *divcookie = 0; 1726#endif 1727 OSAddAtomic(1, &ipstat.ips_fragdropped); 1728 if (fp != 0) 1729 fp->ipq_nfrags--; 1730 m_freem(m); 1731 return (0); 1732 1733#undef GETIP 1734} 1735 1736/* 1737 * Free a fragment reassembly header and all 1738 * associated datagrams. 1739 */ 1740static void 1741ip_freef(struct ipq *fp) 1742{ 1743 lck_mtx_assert(ip_mutex, LCK_MTX_ASSERT_OWNED); 1744 currentfrags -= fp->ipq_nfrags; 1745 m_freem_list(fp->ipq_frags); 1746 remque((void*)fp); 1747 TAILQ_REMOVE(&ipq_list, fp, ipq_list); 1748 (void) m_free(dtom(fp)); 1749 nipq--; 1750} 1751 1752/* 1753 * IP timer processing; 1754 * if a timer expires on a reassembly 1755 * queue, discard it. 1756 */ 1757void 1758ip_slowtimo(void) 1759{ 1760 struct ipq *fp; 1761 int i; 1762 lck_mtx_lock(ip_mutex); 1763 for (i = 0; i < IPREASS_NHASH; i++) { 1764 fp = ipq[i].next; 1765 if (fp == 0) 1766 continue; 1767 while (fp != &ipq[i]) { 1768 --fp->ipq_ttl; 1769 fp = fp->next; 1770 if (fp->prev->ipq_ttl == 0) { 1771 OSAddAtomic(fp->ipq_nfrags, &ipstat.ips_fragtimeout); 1772 ip_freef(fp->prev); 1773 } 1774 } 1775 } 1776 /* 1777 * If we are over the maximum number of fragments 1778 * (due to the limit being lowered), drain off 1779 * enough to get down to the new limit. 1780 */ 1781 if (maxnipq >= 0 && nipq > maxnipq) { 1782 for (i = 0; i < IPREASS_NHASH; i++) { 1783 while (nipq > maxnipq && 1784 (ipq[i].next != &ipq[i])) { 1785 OSAddAtomic(ipq[i].next->ipq_nfrags, &ipstat.ips_fragdropped); 1786 ip_freef(ipq[i].next); 1787 } 1788 } 1789 } 1790 lck_mtx_unlock(ip_mutex); 1791} 1792 1793/* 1794 * Drain off all datagram fragments. 1795 */ 1796void 1797ip_drain(void) 1798{ 1799 int i; 1800 1801 lck_mtx_lock(ip_mutex); 1802 for (i = 0; i < IPREASS_NHASH; i++) { 1803 while (ipq[i].next != &ipq[i]) { 1804 OSAddAtomic(ipq[i].next->ipq_nfrags, &ipstat.ips_fragdropped); 1805 ip_freef(ipq[i].next); 1806 } 1807 } 1808 lck_mtx_unlock(ip_mutex); 1809 in_rtqdrain(); 1810} 1811 1812/* 1813 * Do option processing on a datagram, 1814 * possibly discarding it if bad options are encountered, 1815 * or forwarding it if source-routed. 1816 * The pass argument is used when operating in the IPSTEALTH 1817 * mode to tell what options to process: 1818 * [LS]SRR (pass 0) or the others (pass 1). 1819 * The reason for as many as two passes is that when doing IPSTEALTH, 1820 * non-routing options should be processed only if the packet is for us. 1821 * Returns 1 if packet has been forwarded/freed, 1822 * 0 if the packet should be processed further. 1823 */ 1824static int 1825ip_dooptions(struct mbuf *m, __unused int pass, struct sockaddr_in *next_hop) 1826{ 1827 struct ip *ip = mtod(m, struct ip *); 1828 u_char *cp; 1829 struct ip_timestamp *ipt; 1830 struct in_ifaddr *ia; 1831 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1832 struct in_addr *sin, dst; 1833 n_time ntime; 1834 struct sockaddr_in ipaddr = { 1835 sizeof (ipaddr), AF_INET , 0 , { 0 }, { 0, } }; 1836 1837 /* Expect 32-bit aligned data pointer on strict-align platforms */ 1838 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m); 1839 1840 dst = ip->ip_dst; 1841 cp = (u_char *)(ip + 1); 1842 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1843 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1844 opt = cp[IPOPT_OPTVAL]; 1845 if (opt == IPOPT_EOL) 1846 break; 1847 if (opt == IPOPT_NOP) 1848 optlen = 1; 1849 else { 1850 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1851 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1852 goto bad; 1853 } 1854 optlen = cp[IPOPT_OLEN]; 1855 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1856 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1857 goto bad; 1858 } 1859 } 1860 switch (opt) { 1861 1862 default: 1863 break; 1864 1865 /* 1866 * Source routing with record. 1867 * Find interface with current destination address. 1868 * If none on this machine then drop if strictly routed, 1869 * or do nothing if loosely routed. 1870 * Record interface address and bring up next address 1871 * component. If strictly routed make sure next 1872 * address is on directly accessible net. 1873 */ 1874 case IPOPT_LSRR: 1875 case IPOPT_SSRR: 1876 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1877 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1878 goto bad; 1879 } 1880 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1881 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1882 goto bad; 1883 } 1884 ipaddr.sin_addr = ip->ip_dst; 1885 ia = (struct in_ifaddr *) 1886 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1887 if (ia == 0) { 1888 if (opt == IPOPT_SSRR) { 1889 type = ICMP_UNREACH; 1890 code = ICMP_UNREACH_SRCFAIL; 1891 goto bad; 1892 } 1893 if (!ip_dosourceroute) 1894 goto nosourcerouting; 1895 /* 1896 * Loose routing, and not at next destination 1897 * yet; nothing to do except forward. 1898 */ 1899 break; 1900 } 1901 else { 1902 IFA_REMREF(&ia->ia_ifa); 1903 ia = NULL; 1904 } 1905 off--; /* 0 origin */ 1906 if (off > optlen - (int)sizeof(struct in_addr)) { 1907 /* 1908 * End of source route. Should be for us. 1909 */ 1910 if (!ip_acceptsourceroute) 1911 goto nosourcerouting; 1912 save_rte(cp, ip->ip_src); 1913 break; 1914 } 1915 1916 if (!ip_dosourceroute) { 1917 if (ipforwarding) { 1918 char buf[MAX_IPv4_STR_LEN]; 1919 char buf2[MAX_IPv4_STR_LEN]; 1920 /* 1921 * Acting as a router, so generate ICMP 1922 */ 1923nosourcerouting: 1924 log(LOG_WARNING, 1925 "attempted source route from %s to %s\n", 1926 inet_ntop(AF_INET, &ip->ip_src, buf, sizeof(buf)), 1927 inet_ntop(AF_INET, &ip->ip_dst, buf2, sizeof(buf2))); 1928 type = ICMP_UNREACH; 1929 code = ICMP_UNREACH_SRCFAIL; 1930 goto bad; 1931 } else { 1932 /* 1933 * Not acting as a router, so silently drop. 1934 */ 1935 OSAddAtomic(1, &ipstat.ips_cantforward); 1936 m_freem(m); 1937 return (1); 1938 } 1939 } 1940 1941 /* 1942 * locate outgoing interface 1943 */ 1944 (void)memcpy(&ipaddr.sin_addr, cp + off, 1945 sizeof(ipaddr.sin_addr)); 1946 1947 if (opt == IPOPT_SSRR) { 1948#define INA struct in_ifaddr * 1949 if ((ia = (INA)ifa_ifwithdstaddr( 1950 (struct sockaddr *)&ipaddr)) == 0) { 1951 ia = (INA)ifa_ifwithnet( 1952 (struct sockaddr *)&ipaddr); 1953 } 1954 } else { 1955 ia = ip_rtaddr(ipaddr.sin_addr); 1956 } 1957 if (ia == 0) { 1958 type = ICMP_UNREACH; 1959 code = ICMP_UNREACH_SRCFAIL; 1960 goto bad; 1961 } 1962 ip->ip_dst = ipaddr.sin_addr; 1963 IFA_LOCK(&ia->ia_ifa); 1964 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 1965 sizeof(struct in_addr)); 1966 IFA_UNLOCK(&ia->ia_ifa); 1967 IFA_REMREF(&ia->ia_ifa); 1968 ia = NULL; 1969 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1970 /* 1971 * Let ip_intr's mcast routing check handle mcast pkts 1972 */ 1973 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1974 break; 1975 1976 case IPOPT_RR: 1977 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1978 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1979 goto bad; 1980 } 1981 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1982 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1983 goto bad; 1984 } 1985 /* 1986 * If no space remains, ignore. 1987 */ 1988 off--; /* 0 origin */ 1989 if (off > optlen - (int)sizeof(struct in_addr)) 1990 break; 1991 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1992 sizeof(ipaddr.sin_addr)); 1993 /* 1994 * locate outgoing interface; if we're the destination, 1995 * use the incoming interface (should be same). 1996 */ 1997 if ((ia = (INA)ifa_ifwithaddr((struct sockaddr *) 1998 &ipaddr)) == 0) { 1999 if ((ia = ip_rtaddr(ipaddr.sin_addr)) == 0) { 2000 type = ICMP_UNREACH; 2001 code = ICMP_UNREACH_HOST; 2002 goto bad; 2003 } 2004 } 2005 IFA_LOCK(&ia->ia_ifa); 2006 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr), 2007 sizeof(struct in_addr)); 2008 IFA_UNLOCK(&ia->ia_ifa); 2009 IFA_REMREF(&ia->ia_ifa); 2010 ia = NULL; 2011 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 2012 break; 2013 2014 case IPOPT_TS: 2015 code = cp - (u_char *)ip; 2016 ipt = (struct ip_timestamp *)(void *)cp; 2017 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) { 2018 code = (u_char *)&ipt->ipt_len - (u_char *)ip; 2019 goto bad; 2020 } 2021 if (ipt->ipt_ptr < 5) { 2022 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip; 2023 goto bad; 2024 } 2025 if (ipt->ipt_ptr > 2026 ipt->ipt_len - (int)sizeof(int32_t)) { 2027 if (++ipt->ipt_oflw == 0) { 2028 code = (u_char *)&ipt->ipt_ptr - 2029 (u_char *)ip; 2030 goto bad; 2031 } 2032 break; 2033 } 2034 sin = (struct in_addr *)(void *)(cp + ipt->ipt_ptr - 1); 2035 switch (ipt->ipt_flg) { 2036 2037 case IPOPT_TS_TSONLY: 2038 break; 2039 2040 case IPOPT_TS_TSANDADDR: 2041 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 2042 sizeof(struct in_addr) > ipt->ipt_len) { 2043 code = (u_char *)&ipt->ipt_ptr - 2044 (u_char *)ip; 2045 goto bad; 2046 } 2047 ipaddr.sin_addr = dst; 2048 ia = (INA)ifaof_ifpforaddr((struct sockaddr *) 2049 &ipaddr, m->m_pkthdr.rcvif); 2050 if (ia == 0) 2051 continue; 2052 IFA_LOCK(&ia->ia_ifa); 2053 (void)memcpy(sin, &IA_SIN(ia)->sin_addr, 2054 sizeof(struct in_addr)); 2055 IFA_UNLOCK(&ia->ia_ifa); 2056 ipt->ipt_ptr += sizeof(struct in_addr); 2057 IFA_REMREF(&ia->ia_ifa); 2058 ia = NULL; 2059 break; 2060 2061 case IPOPT_TS_PRESPEC: 2062 if (ipt->ipt_ptr - 1 + sizeof(n_time) + 2063 sizeof(struct in_addr) > ipt->ipt_len) { 2064 code = (u_char *)&ipt->ipt_ptr - 2065 (u_char *)ip; 2066 goto bad; 2067 } 2068 (void)memcpy(&ipaddr.sin_addr, sin, 2069 sizeof(struct in_addr)); 2070 if ((ia = (struct in_ifaddr*)ifa_ifwithaddr( 2071 (struct sockaddr *)&ipaddr)) == 0) 2072 continue; 2073 IFA_REMREF(&ia->ia_ifa); 2074 ia = NULL; 2075 ipt->ipt_ptr += sizeof(struct in_addr); 2076 break; 2077 2078 default: 2079 /* XXX can't take &ipt->ipt_flg */ 2080 code = (u_char *)&ipt->ipt_ptr - 2081 (u_char *)ip + 1; 2082 goto bad; 2083 } 2084 ntime = iptime(); 2085 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime, 2086 sizeof(n_time)); 2087 ipt->ipt_ptr += sizeof(n_time); 2088 } 2089 } 2090 if (forward && ipforwarding) { 2091 ip_forward(m, 1, next_hop); 2092 return (1); 2093 } 2094 return (0); 2095bad: 2096 ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2; /* XXX icmp_error adds in hdr length */ 2097 icmp_error(m, type, code, 0, 0); 2098 OSAddAtomic(1, &ipstat.ips_badoptions); 2099 return (1); 2100} 2101 2102/* 2103 * Given address of next destination (final or next hop), 2104 * return internet address info of interface to be used to get there. 2105 */ 2106struct in_ifaddr * 2107ip_rtaddr(struct in_addr dst) 2108{ 2109 struct sockaddr_in *sin; 2110 struct ifaddr *rt_ifa; 2111 struct route ro; 2112 2113 bzero(&ro, sizeof (ro)); 2114 sin = (struct sockaddr_in *)(void *)&ro.ro_dst; 2115 sin->sin_family = AF_INET; 2116 sin->sin_len = sizeof (*sin); 2117 sin->sin_addr = dst; 2118 2119 rtalloc_ign(&ro, RTF_PRCLONING); 2120 if (ro.ro_rt == NULL) 2121 return (NULL); 2122 2123 RT_LOCK(ro.ro_rt); 2124 if ((rt_ifa = ro.ro_rt->rt_ifa) != NULL) 2125 IFA_ADDREF(rt_ifa); 2126 RT_UNLOCK(ro.ro_rt); 2127 rtfree(ro.ro_rt); 2128 2129 return ((struct in_ifaddr *)rt_ifa); 2130} 2131 2132/* 2133 * Save incoming source route for use in replies, 2134 * to be picked up later by ip_srcroute if the receiver is interested. 2135 */ 2136void 2137save_rte(u_char *option, struct in_addr dst) 2138{ 2139 unsigned olen; 2140 2141 olen = option[IPOPT_OLEN]; 2142#if DIAGNOSTIC 2143 if (ipprintfs) 2144 printf("save_rte: olen %d\n", olen); 2145#endif 2146 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) 2147 return; 2148 bcopy(option, ip_srcrt.srcopt, olen); 2149 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 2150 ip_srcrt.dst = dst; 2151} 2152 2153/* 2154 * Retrieve incoming source route for use in replies, 2155 * in the same form used by setsockopt. 2156 * The first hop is placed before the options, will be removed later. 2157 */ 2158struct mbuf * 2159ip_srcroute(void) 2160{ 2161 struct in_addr *p, *q; 2162 struct mbuf *m; 2163 2164 if (ip_nhops == 0) 2165 return ((struct mbuf *)0); 2166 m = m_get(M_DONTWAIT, MT_HEADER); 2167 if (m == 0) 2168 return ((struct mbuf *)0); 2169 2170#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) 2171 2172 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 2173 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) + 2174 OPTSIZ; 2175#if DIAGNOSTIC 2176 if (ipprintfs) 2177 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len); 2178#endif 2179 2180 /* 2181 * First save first hop for return route 2182 */ 2183 p = &ip_srcrt.route[ip_nhops - 1]; 2184 *(mtod(m, struct in_addr *)) = *p--; 2185#if DIAGNOSTIC 2186 if (ipprintfs) 2187 printf(" hops %lx", (u_int32_t)ntohl(mtod(m, struct in_addr *)->s_addr)); 2188#endif 2189 2190 /* 2191 * Copy option fields and padding (nop) to mbuf. 2192 */ 2193 ip_srcrt.nop = IPOPT_NOP; 2194 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 2195 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), 2196 &ip_srcrt.nop, OPTSIZ); 2197 q = (struct in_addr *)(void *)(mtod(m, caddr_t) + 2198 sizeof(struct in_addr) + OPTSIZ); 2199#undef OPTSIZ 2200 /* 2201 * Record return path as an IP source route, 2202 * reversing the path (pointers are now aligned). 2203 */ 2204 while (p >= ip_srcrt.route) { 2205#if DIAGNOSTIC 2206 if (ipprintfs) 2207 printf(" %lx", (u_int32_t)ntohl(q->s_addr)); 2208#endif 2209 *q++ = *p--; 2210 } 2211 /* 2212 * Last hop goes to final destination. 2213 */ 2214 *q = ip_srcrt.dst; 2215#if DIAGNOSTIC 2216 if (ipprintfs) 2217 printf(" %lx\n", (u_int32_t)ntohl(q->s_addr)); 2218#endif 2219 return (m); 2220} 2221 2222/* 2223 * Strip out IP options, at higher 2224 * level protocol in the kernel. 2225 * Second argument is buffer to which options 2226 * will be moved, and return value is their length. 2227 * XXX should be deleted; last arg currently ignored. 2228 */ 2229void 2230ip_stripoptions(struct mbuf *m, __unused struct mbuf *mopt) 2231{ 2232 int i; 2233 struct ip *ip = mtod(m, struct ip *); 2234 caddr_t opts; 2235 int olen; 2236 2237 /* Expect 32-bit aligned data pointer on strict-align platforms */ 2238 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m); 2239 2240 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 2241 opts = (caddr_t)(ip + 1); 2242 i = m->m_len - (sizeof (struct ip) + olen); 2243 bcopy(opts + olen, opts, (unsigned)i); 2244 m->m_len -= olen; 2245 if (m->m_flags & M_PKTHDR) 2246 m->m_pkthdr.len -= olen; 2247 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 2248} 2249 2250u_char inetctlerrmap[PRC_NCMDS] = { 2251 0, 0, 0, 0, 2252 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 2253 ENETUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 2254 EMSGSIZE, EHOSTUNREACH, 0, 0, 2255 0, 0, 0, 0, 2256 ENOPROTOOPT, ECONNREFUSED 2257}; 2258 2259static int 2260sysctl_ipforwarding SYSCTL_HANDLER_ARGS 2261{ 2262#pragma unused(arg1, arg2) 2263 int i, was_ipforwarding = ipforwarding; 2264 2265 i = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req); 2266 if (i != 0 || req->newptr == USER_ADDR_NULL) 2267 return (i); 2268 2269 if (was_ipforwarding && !ipforwarding) { 2270 /* clean up IPv4 forwarding cached routes */ 2271 ifnet_head_lock_shared(); 2272 for (i = 0; i <= if_index; i++) { 2273 struct ifnet *ifp = ifindex2ifnet[i]; 2274 if (ifp != NULL) { 2275 lck_mtx_lock(&ifp->if_cached_route_lock); 2276 if (ifp->if_fwd_route.ro_rt != NULL) 2277 rtfree(ifp->if_fwd_route.ro_rt); 2278 bzero(&ifp->if_fwd_route, 2279 sizeof (ifp->if_fwd_route)); 2280 lck_mtx_unlock(&ifp->if_cached_route_lock); 2281 } 2282 } 2283 ifnet_head_done(); 2284 } 2285 2286 return (0); 2287} 2288 2289/* 2290 * Similar to inp_route_{copyout,copyin} routines except that these copy 2291 * out the cached IPv4 forwarding route from struct ifnet instead of the 2292 * inpcb. See comments for those routines for explanations. 2293 */ 2294static void 2295ip_fwd_route_copyout(struct ifnet *ifp, struct route *dst) 2296{ 2297 struct route *src = &ifp->if_fwd_route; 2298 2299 lck_mtx_lock_spin(&ifp->if_cached_route_lock); 2300 lck_mtx_convert_spin(&ifp->if_cached_route_lock); 2301 2302 /* Minor sanity check */ 2303 if (src->ro_rt != NULL && rt_key(src->ro_rt)->sa_family != AF_INET) 2304 panic("%s: wrong or corrupted route: %p", __func__, src); 2305 2306 route_copyout(dst, src, sizeof(*dst)); 2307 2308 lck_mtx_unlock(&ifp->if_cached_route_lock); 2309} 2310 2311static void 2312ip_fwd_route_copyin(struct ifnet *ifp, struct route *src) 2313{ 2314 struct route *dst = &ifp->if_fwd_route; 2315 2316 lck_mtx_lock_spin(&ifp->if_cached_route_lock); 2317 lck_mtx_convert_spin(&ifp->if_cached_route_lock); 2318 2319 /* Minor sanity check */ 2320 if (src->ro_rt != NULL && rt_key(src->ro_rt)->sa_family != AF_INET) 2321 panic("%s: wrong or corrupted route: %p", __func__, src); 2322 2323 if (ifp->if_fwd_cacheok) 2324 route_copyin(src, dst, sizeof(*src)); 2325 2326 lck_mtx_unlock(&ifp->if_cached_route_lock); 2327} 2328 2329/* 2330 * Forward a packet. If some error occurs return the sender 2331 * an icmp packet. Note we can't always generate a meaningful 2332 * icmp message because icmp doesn't have a large enough repertoire 2333 * of codes and types. 2334 * 2335 * If not forwarding, just drop the packet. This could be confusing 2336 * if ipforwarding was zero but some routing protocol was advancing 2337 * us as a gateway to somewhere. However, we must let the routing 2338 * protocol deal with that. 2339 * 2340 * The srcrt parameter indicates whether the packet is being forwarded 2341 * via a source route. 2342 */ 2343static void 2344ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop) 2345{ 2346#if !IPFIREWALL 2347#pragma unused(next_hop) 2348#endif 2349 struct ip *ip = mtod(m, struct ip *); 2350 struct sockaddr_in *sin; 2351 struct rtentry *rt; 2352 struct route fwd_rt; 2353 int error, type = 0, code = 0; 2354 struct mbuf *mcopy; 2355 n_long dest; 2356 struct in_addr pkt_dst; 2357 u_int32_t nextmtu = 0; 2358 struct ip_out_args ipoa = { IFSCOPE_NONE, { 0 }, 0 }; 2359 struct ifnet *ifp = m->m_pkthdr.rcvif; 2360#if PF 2361 struct pf_mtag *pf_mtag; 2362#endif /* PF */ 2363 2364 dest = 0; 2365#if IPFIREWALL 2366 /* 2367 * Cache the destination address of the packet; this may be 2368 * changed by use of 'ipfw fwd'. 2369 */ 2370 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 2371#else 2372 pkt_dst = ip->ip_dst; 2373#endif 2374 2375#if DIAGNOSTIC 2376 if (ipprintfs) 2377 printf("forward: src %lx dst %lx ttl %x\n", 2378 (u_int32_t)ip->ip_src.s_addr, (u_int32_t)pkt_dst.s_addr, 2379 ip->ip_ttl); 2380#endif 2381 2382 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) { 2383 OSAddAtomic(1, &ipstat.ips_cantforward); 2384 m_freem(m); 2385 return; 2386 } 2387#if IPSTEALTH 2388 if (!ipstealth) { 2389#endif 2390 if (ip->ip_ttl <= IPTTLDEC) { 2391 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 2392 dest, 0); 2393 return; 2394 } 2395#if IPSTEALTH 2396 } 2397#endif 2398 2399#if PF 2400 pf_mtag = pf_find_mtag(m); 2401 if (pf_mtag != NULL && pf_mtag->pftag_rtableid != IFSCOPE_NONE) { 2402 ipoa.ipoa_boundif = pf_mtag->pftag_rtableid; 2403 ipoa.ipoa_flags |= IPOAF_BOUND_IF; 2404 } 2405#endif /* PF */ 2406 2407 ip_fwd_route_copyout(ifp, &fwd_rt); 2408 2409 sin = (struct sockaddr_in *)(void *)&fwd_rt.ro_dst; 2410 if (fwd_rt.ro_rt == NULL || 2411 fwd_rt.ro_rt->generation_id != route_generation || 2412 pkt_dst.s_addr != sin->sin_addr.s_addr) { 2413 if (fwd_rt.ro_rt != NULL) { 2414 rtfree(fwd_rt.ro_rt); 2415 fwd_rt.ro_rt = NULL; 2416 } 2417 sin->sin_family = AF_INET; 2418 sin->sin_len = sizeof (*sin); 2419 sin->sin_addr = pkt_dst; 2420 2421 rtalloc_scoped_ign(&fwd_rt, RTF_PRCLONING, ipoa.ipoa_boundif); 2422 if (fwd_rt.ro_rt == NULL) { 2423 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 2424 goto done; 2425 } 2426 } 2427 rt = fwd_rt.ro_rt; 2428 2429 /* 2430 * Save the IP header and at most 8 bytes of the payload, 2431 * in case we need to generate an ICMP message to the src. 2432 * 2433 * We don't use m_copy() because it might return a reference 2434 * to a shared cluster. Both this function and ip_output() 2435 * assume exclusive access to the IP header in `m', so any 2436 * data in a cluster may change before we reach icmp_error(). 2437 */ 2438 MGET(mcopy, M_DONTWAIT, m->m_type); 2439 if (mcopy != NULL) { 2440 M_COPY_PKTHDR(mcopy, m); 2441 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 2442 (int)ip->ip_len); 2443 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 2444 } 2445 2446#if IPSTEALTH 2447 if (!ipstealth) { 2448#endif 2449 ip->ip_ttl -= IPTTLDEC; 2450#if IPSTEALTH 2451 } 2452#endif 2453 2454 /* 2455 * If forwarding packet using same interface that it came in on, 2456 * perhaps should send a redirect to sender to shortcut a hop. 2457 * Only send redirect if source is sending directly to us, 2458 * and if packet was not source routed (or has any options). 2459 * Also, don't send redirect if forwarding using a default route 2460 * or a route modified by a redirect. 2461 */ 2462 RT_LOCK_SPIN(rt); 2463 if (rt->rt_ifp == m->m_pkthdr.rcvif && 2464 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 2465 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 2466 ipsendredirects && !srcrt && rt->rt_ifa != NULL) { 2467 struct in_ifaddr *ia = (struct in_ifaddr *)rt->rt_ifa; 2468 u_int32_t src = ntohl(ip->ip_src.s_addr); 2469 2470 /* Become a regular mutex */ 2471 RT_CONVERT_LOCK(rt); 2472 IFA_LOCK_SPIN(&ia->ia_ifa); 2473 if ((src & ia->ia_subnetmask) == ia->ia_subnet) { 2474 if (rt->rt_flags & RTF_GATEWAY) 2475 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 2476 else 2477 dest = pkt_dst.s_addr; 2478 /* Router requirements says to only send host redirects */ 2479 type = ICMP_REDIRECT; 2480 code = ICMP_REDIRECT_HOST; 2481#if DIAGNOSTIC 2482 if (ipprintfs) 2483 printf("redirect (%d) to %lx\n", code, (u_int32_t)dest); 2484#endif 2485 } 2486 IFA_UNLOCK(&ia->ia_ifa); 2487 } 2488 RT_UNLOCK(rt); 2489 2490#if IPFIREWALL 2491 if (next_hop) { 2492 /* Pass IPFORWARD info if available */ 2493 struct m_tag *tag; 2494 struct ip_fwd_tag *ipfwd_tag; 2495 2496 tag = m_tag_create(KERNEL_MODULE_TAG_ID, 2497 KERNEL_TAG_TYPE_IPFORWARD, 2498 sizeof (*ipfwd_tag), M_NOWAIT, m); 2499 if (tag == NULL) { 2500 error = ENOBUFS; 2501 m_freem(m); 2502 goto done; 2503 } 2504 2505 ipfwd_tag = (struct ip_fwd_tag *)(tag+1); 2506 ipfwd_tag->next_hop = next_hop; 2507 2508 m_tag_prepend(m, tag); 2509 } 2510#endif 2511 error = ip_output_list(m, 0, NULL, &fwd_rt, 2512 IP_FORWARDING | IP_OUTARGS, 0, &ipoa); 2513 2514 /* Refresh rt since the route could have changed while in IP */ 2515 rt = fwd_rt.ro_rt; 2516 2517 if (error) { 2518 OSAddAtomic(1, &ipstat.ips_cantforward); 2519 } else { 2520 OSAddAtomic(1, &ipstat.ips_forward); 2521 if (type) 2522 OSAddAtomic(1, &ipstat.ips_redirectsent); 2523 else { 2524 if (mcopy) { 2525 /* 2526 * If we didn't have to go thru ipflow and 2527 * the packet was successfully consumed by 2528 * ip_output, the mcopy is rather a waste; 2529 * this could be further optimized. 2530 */ 2531 m_freem(mcopy); 2532 } 2533 goto done; 2534 } 2535 } 2536 if (mcopy == NULL) 2537 goto done; 2538 2539 switch (error) { 2540 2541 case 0: /* forwarded, but need redirect */ 2542 /* type, code set above */ 2543 break; 2544 2545 case ENETUNREACH: /* shouldn't happen, checked above */ 2546 case EHOSTUNREACH: 2547 case ENETDOWN: 2548 case EHOSTDOWN: 2549 default: 2550 type = ICMP_UNREACH; 2551 code = ICMP_UNREACH_HOST; 2552 break; 2553 2554 case EMSGSIZE: 2555 type = ICMP_UNREACH; 2556 code = ICMP_UNREACH_NEEDFRAG; 2557#ifndef IPSEC 2558 if (rt != NULL) { 2559 RT_LOCK_SPIN(rt); 2560 if (rt->rt_ifp != NULL) 2561 nextmtu = rt->rt_ifp->if_mtu; 2562 RT_UNLOCK(rt); 2563 } 2564#else 2565 /* 2566 * If the packet is routed over IPsec tunnel, tell the 2567 * originator the tunnel MTU. 2568 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2569 * XXX quickhack!!! 2570 */ 2571 if (rt != NULL) { 2572 struct secpolicy *sp = NULL; 2573 int ipsecerror; 2574 int ipsechdr; 2575 struct route *ro; 2576 2577 RT_LOCK_SPIN(rt); 2578 if (rt->rt_ifp != NULL) 2579 nextmtu = rt->rt_ifp->if_mtu; 2580 RT_UNLOCK(rt); 2581 2582 if (ipsec_bypass) { 2583 OSAddAtomic(1, &ipstat.ips_cantfrag); 2584 break; 2585 } 2586 sp = ipsec4_getpolicybyaddr(mcopy, 2587 IPSEC_DIR_OUTBOUND, 2588 IP_FORWARDING, 2589 &ipsecerror); 2590 2591 if (sp != NULL) { 2592 /* count IPsec header size */ 2593 ipsechdr = ipsec_hdrsiz(sp); 2594 2595 /* 2596 * find the correct route for outer IPv4 2597 * header, compute tunnel MTU. 2598 */ 2599 nextmtu = 0; 2600 2601 if (sp->req != NULL) { 2602 if (sp->req->saidx.mode == IPSEC_MODE_TUNNEL) { 2603 struct secasindex saidx; 2604 struct ip *ipm; 2605 struct secasvar *sav; 2606 2607 ipm = mtod(mcopy, struct ip *); 2608 bcopy(&sp->req->saidx, &saidx, sizeof(saidx)); 2609 saidx.mode = sp->req->saidx.mode; 2610 saidx.reqid = sp->req->saidx.reqid; 2611 sin = (struct sockaddr_in *)&saidx.src; 2612 if (sin->sin_len == 0) { 2613 sin->sin_len = sizeof(*sin); 2614 sin->sin_family = AF_INET; 2615 sin->sin_port = IPSEC_PORT_ANY; 2616 bcopy(&ipm->ip_src, &sin->sin_addr, 2617 sizeof(sin->sin_addr)); 2618 } 2619 sin = (struct sockaddr_in *)&saidx.dst; 2620 if (sin->sin_len == 0) { 2621 sin->sin_len = sizeof(*sin); 2622 sin->sin_family = AF_INET; 2623 sin->sin_port = IPSEC_PORT_ANY; 2624 bcopy(&ipm->ip_dst, &sin->sin_addr, 2625 sizeof(sin->sin_addr)); 2626 } 2627 sav = key_allocsa_policy(&saidx); 2628 if (sav != NULL) { 2629 lck_mtx_lock(sadb_mutex); 2630 if (sav->sah != NULL) { 2631 ro = &sav->sah->sa_route; 2632 if (ro->ro_rt != NULL) { 2633 RT_LOCK(ro->ro_rt); 2634 if (ro->ro_rt->rt_ifp != NULL) { 2635 nextmtu = ro->ro_rt->rt_ifp->if_mtu; 2636 nextmtu -= ipsechdr; 2637 } 2638 RT_UNLOCK(ro->ro_rt); 2639 } 2640 } 2641 key_freesav(sav, KEY_SADB_LOCKED); 2642 lck_mtx_unlock(sadb_mutex); 2643 } 2644 } 2645 } 2646 key_freesp(sp, KEY_SADB_UNLOCKED); 2647 } 2648 } 2649#endif /*IPSEC*/ 2650 OSAddAtomic(1, &ipstat.ips_cantfrag); 2651 break; 2652 2653 case ENOBUFS: 2654 type = ICMP_SOURCEQUENCH; 2655 code = 0; 2656 break; 2657 2658 case EACCES: /* ipfw denied packet */ 2659 m_freem(mcopy); 2660 goto done; 2661 } 2662 2663 icmp_error(mcopy, type, code, dest, nextmtu); 2664done: 2665 ip_fwd_route_copyin(ifp, &fwd_rt); 2666} 2667 2668int 2669ip_savecontrol( 2670 struct inpcb *inp, 2671 struct mbuf **mp, 2672 struct ip *ip, 2673 struct mbuf *m) 2674{ 2675 *mp = NULL; 2676 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2677 struct timeval tv; 2678 2679 microtime(&tv); 2680 mp = sbcreatecontrol_mbuf((caddr_t) &tv, sizeof(tv), 2681 SCM_TIMESTAMP, SOL_SOCKET, mp); 2682 if (*mp == NULL) { 2683 goto no_mbufs; 2684 } 2685 } 2686 if ((inp->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) { 2687 uint64_t time; 2688 2689 time = mach_absolute_time(); 2690 mp = sbcreatecontrol_mbuf((caddr_t) &time, sizeof(time), 2691 SCM_TIMESTAMP_MONOTONIC, SOL_SOCKET, mp); 2692 2693 if (*mp == NULL) { 2694 goto no_mbufs; 2695 } 2696 } 2697 if (inp->inp_flags & INP_RECVDSTADDR) { 2698 mp = sbcreatecontrol_mbuf((caddr_t) &ip->ip_dst, 2699 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP, mp); 2700 if (*mp == NULL) { 2701 goto no_mbufs; 2702 } 2703 } 2704#ifdef notyet 2705 /* XXX 2706 * Moving these out of udp_input() made them even more broken 2707 * than they already were. 2708 */ 2709 /* options were tossed already */ 2710 if (inp->inp_flags & INP_RECVOPTS) { 2711 mp = sbcreatecontrol_mbuf((caddr_t) opts_deleted_above, 2712 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, mp); 2713 if (*mp == NULL) { 2714 goto no_mbufs; 2715 } 2716 } 2717 /* ip_srcroute doesn't do what we want here, need to fix */ 2718 if (inp->inp_flags & INP_RECVRETOPTS) { 2719 mp = sbcreatecontrol_mbuf((caddr_t) ip_srcroute(), 2720 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP, mp); 2721 if (*mp == NULL) { 2722 goto no_mbufs; 2723 } 2724 } 2725#endif 2726 if (inp->inp_flags & INP_RECVIF) { 2727 struct ifnet *ifp; 2728 struct sdlbuf { 2729 struct sockaddr_dl sdl; 2730 u_char pad[32]; 2731 } sdlbuf; 2732 struct sockaddr_dl *sdp; 2733 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2734 2735 ifnet_head_lock_shared(); 2736 if ((ifp = m->m_pkthdr.rcvif) != NULL && 2737 ifp->if_index && (ifp->if_index <= if_index)) { 2738 struct ifaddr *ifa = ifnet_addrs[ifp->if_index - 1]; 2739 2740 if (!ifa || !ifa->ifa_addr) 2741 goto makedummy; 2742 2743 IFA_LOCK_SPIN(ifa); 2744 sdp = (struct sockaddr_dl *)(void *)ifa->ifa_addr; 2745 /* 2746 * Change our mind and don't try copy. 2747 */ 2748 if ((sdp->sdl_family != AF_LINK) || 2749 (sdp->sdl_len > sizeof(sdlbuf))) { 2750 IFA_UNLOCK(ifa); 2751 goto makedummy; 2752 } 2753 bcopy(sdp, sdl2, sdp->sdl_len); 2754 IFA_UNLOCK(ifa); 2755 } else { 2756makedummy: 2757 sdl2->sdl_len 2758 = offsetof(struct sockaddr_dl, sdl_data[0]); 2759 sdl2->sdl_family = AF_LINK; 2760 sdl2->sdl_index = 0; 2761 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2762 } 2763 ifnet_head_done(); 2764 mp = sbcreatecontrol_mbuf((caddr_t) sdl2, sdl2->sdl_len, 2765 IP_RECVIF, IPPROTO_IP, mp); 2766 if (*mp == NULL) { 2767 goto no_mbufs; 2768 } 2769 } 2770 if (inp->inp_flags & INP_RECVTTL) { 2771 mp = sbcreatecontrol_mbuf((caddr_t)&ip->ip_ttl, sizeof(ip->ip_ttl), 2772 IP_RECVTTL, IPPROTO_IP, mp); 2773 if (*mp == NULL) { 2774 goto no_mbufs; 2775 } 2776 } 2777 if ((inp->inp_socket->so_flags & SOF_RECV_TRAFFIC_CLASS) != 0) { 2778 int tc = m_get_traffic_class(m); 2779 2780 mp = sbcreatecontrol_mbuf((caddr_t) &tc, sizeof(tc), 2781 SO_TRAFFIC_CLASS, SOL_SOCKET, mp); 2782 if (*mp == NULL) { 2783 goto no_mbufs; 2784 } 2785 } 2786 if (inp->inp_flags & INP_PKTINFO) { 2787 struct in_pktinfo pi; 2788 2789 bzero(&pi, sizeof(struct in_pktinfo)); 2790 bcopy(&ip->ip_dst, &pi.ipi_addr, sizeof(struct in_addr)); 2791 pi.ipi_ifindex = (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 2792 2793 mp = sbcreatecontrol_mbuf((caddr_t)&pi, sizeof(struct in_pktinfo), 2794 IP_RECVPKTINFO, IPPROTO_IP, mp); 2795 if (*mp == NULL) { 2796 goto no_mbufs; 2797 } 2798 } 2799 return 0; 2800 2801no_mbufs: 2802 ipstat.ips_pktdropcntrl++; 2803 return ENOBUFS; 2804} 2805 2806int 2807ip_rsvp_init(struct socket *so) 2808{ 2809 if (so->so_type != SOCK_RAW || 2810 so->so_proto->pr_protocol != IPPROTO_RSVP) 2811 return EOPNOTSUPP; 2812 2813 if (ip_rsvpd != NULL) 2814 return EADDRINUSE; 2815 2816 ip_rsvpd = so; 2817 /* 2818 * This may seem silly, but we need to be sure we don't over-increment 2819 * the RSVP counter, in case something slips up. 2820 */ 2821 if (!ip_rsvp_on) { 2822 ip_rsvp_on = 1; 2823 rsvp_on++; 2824 } 2825 2826 return 0; 2827} 2828 2829int 2830ip_rsvp_done(void) 2831{ 2832 ip_rsvpd = NULL; 2833 /* 2834 * This may seem silly, but we need to be sure we don't over-decrement 2835 * the RSVP counter, in case something slips up. 2836 */ 2837 if (ip_rsvp_on) { 2838 ip_rsvp_on = 0; 2839 rsvp_on--; 2840 } 2841 return 0; 2842} 2843 2844static inline u_short 2845ip_cksum(struct mbuf *m, int hlen) 2846{ 2847 2848 u_short sum; 2849 struct ip *ip; 2850 2851 ip = mtod(m, struct ip *); 2852 2853 if ((IF_HWASSIST_CSUM_FLAGS(m->m_pkthdr.rcvif->if_hwassist) == 0) 2854 || (apple_hwcksum_rx == 0) || 2855 ((m->m_pkthdr.csum_flags & CSUM_TCP_SUM16) && ip->ip_p != IPPROTO_TCP)) { 2856 m->m_pkthdr.csum_flags = 0; /* invalidate HW generated checksum flags */ 2857 2858 } 2859 2860 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 2861 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 2862 } else if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) || 2863 apple_hwcksum_tx == 0) { 2864 /* 2865 * Either this is not loopback packet coming from an interface 2866 * that does not support checksum offloading, or it is loopback 2867 * packet that has undergone software checksumming at the send 2868 * side because apple_hwcksum_tx was set to 0. In this case, 2869 * calculate the checksum in software to validate the packet. 2870 */ 2871 sum = in_cksum(m, hlen); 2872 } else { 2873 /* 2874 * This is a loopback packet without any valid checksum since 2875 * the send side has bypassed it (apple_hwcksum_tx set to 1). 2876 * We get here because apple_hwcksum_rx was set to 0, and so 2877 * we pretend that all is well. 2878 */ 2879 sum = 0; 2880 m->m_pkthdr.csum_flags |= CSUM_DATA_VALID | CSUM_PSEUDO_HDR | 2881 CSUM_IP_CHECKED | CSUM_IP_VALID; 2882 m->m_pkthdr.csum_data = 0xffff; 2883 } 2884 2885 if (sum) { 2886 OSAddAtomic(1, &ipstat.ips_badsum); 2887 } 2888 2889 return sum; 2890} 2891