ip_fw2.c revision 146704
1/*- 2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions 6 * are met: 7 * 1. Redistributions of source code must retain the above copyright 8 * notice, this list of conditions and the following disclaimer. 9 * 2. Redistributions in binary form must reproduce the above copyright 10 * notice, this list of conditions and the following disclaimer in the 11 * documentation and/or other materials provided with the distribution. 12 * 13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23 * SUCH DAMAGE. 24 * 25 * $FreeBSD: head/sys/netinet/ip_fw2.c 146704 2005-05-28 07:46:44Z tanimura $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdn.h" 38#include "opt_inet.h" 39#include "opt_inet6.h" 40#include "opt_ipsec.h" 41#ifndef INET 42#error IPFIREWALL requires INET. 43#endif /* INET */ 44#endif 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/condvar.h> 49#include <sys/malloc.h> 50#include <sys/mbuf.h> 51#include <sys/kernel.h> 52#include <sys/jail.h> 53#include <sys/module.h> 54#include <sys/proc.h> 55#include <sys/socket.h> 56#include <sys/socketvar.h> 57#include <sys/sysctl.h> 58#include <sys/syslog.h> 59#include <sys/ucred.h> 60#include <net/if.h> 61#include <net/radix.h> 62#include <net/route.h> 63#include <netinet/in.h> 64#include <netinet/in_systm.h> 65#include <netinet/in_var.h> 66#include <netinet/in_pcb.h> 67#include <netinet/ip.h> 68#include <netinet/ip_var.h> 69#include <netinet/ip_icmp.h> 70#include <netinet/ip_fw.h> 71#include <netinet/ip_divert.h> 72#include <netinet/ip_dummynet.h> 73#include <netinet/tcp.h> 74#include <netinet/tcp_timer.h> 75#include <netinet/tcp_var.h> 76#include <netinet/tcpip.h> 77#include <netinet/udp.h> 78#include <netinet/udp_var.h> 79 80#include <netgraph/ng_ipfw.h> 81 82#include <altq/if_altq.h> 83 84#ifdef IPSEC 85#include <netinet6/ipsec.h> 86#endif 87 88#include <netinet/ip6.h> 89#include <netinet/icmp6.h> 90 91#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 92 93#include <machine/in_cksum.h> /* XXX for in_cksum */ 94 95/* 96 * set_disable contains one bit per set value (0..31). 97 * If the bit is set, all rules with the corresponding set 98 * are disabled. Set RESVD_SET(31) is reserved for the default rule 99 * and rules that are not deleted by the flush command, 100 * and CANNOT be disabled. 101 * Rules in set RESVD_SET can only be deleted explicitly. 102 */ 103static u_int32_t set_disable; 104 105static int fw_verbose; 106static int verbose_limit; 107 108static struct callout ipfw_timeout; 109static uma_zone_t ipfw_dyn_rule_zone; 110#define IPFW_DEFAULT_RULE 65535 111 112/* 113 * Data structure to cache our ucred related 114 * information. This structure only gets used if 115 * the user specified UID/GID based constraints in 116 * a firewall rule. 117 */ 118struct ip_fw_ugid { 119 gid_t fw_groups[NGROUPS]; 120 int fw_ngroups; 121 uid_t fw_uid; 122 int fw_prid; 123}; 124 125struct ip_fw_chain { 126 struct ip_fw *rules; /* list of rules */ 127 struct ip_fw *reap; /* list of rules to reap */ 128 struct mtx mtx; /* lock guarding rule list */ 129 int busy_count; /* busy count for rw locks */ 130 int want_write; 131 struct cv cv; 132}; 133#define IPFW_LOCK_INIT(_chain) \ 134 mtx_init(&(_chain)->mtx, "IPFW static rules", NULL, \ 135 MTX_DEF | MTX_RECURSE) 136#define IPFW_LOCK_DESTROY(_chain) mtx_destroy(&(_chain)->mtx) 137#define IPFW_WLOCK_ASSERT(_chain) do { \ 138 mtx_assert(&(_chain)->mtx, MA_OWNED); \ 139 NET_ASSERT_GIANT(); \ 140} while (0) 141 142static __inline void 143IPFW_RLOCK(struct ip_fw_chain *chain) 144{ 145 mtx_lock(&chain->mtx); 146 chain->busy_count++; 147 mtx_unlock(&chain->mtx); 148} 149 150static __inline void 151IPFW_RUNLOCK(struct ip_fw_chain *chain) 152{ 153 mtx_lock(&chain->mtx); 154 chain->busy_count--; 155 if (chain->busy_count == 0 && chain->want_write) 156 cv_signal(&chain->cv); 157 mtx_unlock(&chain->mtx); 158} 159 160static __inline void 161IPFW_WLOCK(struct ip_fw_chain *chain) 162{ 163 mtx_lock(&chain->mtx); 164 chain->want_write++; 165 while (chain->busy_count > 0) 166 cv_wait(&chain->cv, &chain->mtx); 167} 168 169static __inline void 170IPFW_WUNLOCK(struct ip_fw_chain *chain) 171{ 172 chain->want_write--; 173 cv_signal(&chain->cv); 174 mtx_unlock(&chain->mtx); 175} 176 177/* 178 * list of rules for layer 3 179 */ 180static struct ip_fw_chain layer3_chain; 181 182MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 183MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 184 185struct table_entry { 186 struct radix_node rn[2]; 187 struct sockaddr_in addr, mask; 188 u_int32_t value; 189}; 190 191#define IPFW_TABLES_MAX 128 192static struct { 193 struct radix_node_head *rnh; 194 int modified; 195} ipfw_tables[IPFW_TABLES_MAX]; 196 197static int fw_debug = 1; 198static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 199 200#ifdef SYSCTL_NODE 201SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 202SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 203 CTLFLAG_RW | CTLFLAG_SECURE3, 204 &fw_enable, 0, "Enable ipfw"); 205SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 206 &autoinc_step, 0, "Rule number autincrement step"); 207SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 208 CTLFLAG_RW | CTLFLAG_SECURE3, 209 &fw_one_pass, 0, 210 "Only do a single pass through ipfw when using dummynet(4)"); 211SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 212 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 213SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 214 CTLFLAG_RW | CTLFLAG_SECURE3, 215 &fw_verbose, 0, "Log matches to ipfw rules"); 216SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 217 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 218 219/* 220 * Description of dynamic rules. 221 * 222 * Dynamic rules are stored in lists accessed through a hash table 223 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 224 * be modified through the sysctl variable dyn_buckets which is 225 * updated when the table becomes empty. 226 * 227 * XXX currently there is only one list, ipfw_dyn. 228 * 229 * When a packet is received, its address fields are first masked 230 * with the mask defined for the rule, then hashed, then matched 231 * against the entries in the corresponding list. 232 * Dynamic rules can be used for different purposes: 233 * + stateful rules; 234 * + enforcing limits on the number of sessions; 235 * + in-kernel NAT (not implemented yet) 236 * 237 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 238 * measured in seconds and depending on the flags. 239 * 240 * The total number of dynamic rules is stored in dyn_count. 241 * The max number of dynamic rules is dyn_max. When we reach 242 * the maximum number of rules we do not create anymore. This is 243 * done to avoid consuming too much memory, but also too much 244 * time when searching on each packet (ideally, we should try instead 245 * to put a limit on the length of the list on each bucket...). 246 * 247 * Each dynamic rule holds a pointer to the parent ipfw rule so 248 * we know what action to perform. Dynamic rules are removed when 249 * the parent rule is deleted. XXX we should make them survive. 250 * 251 * There are some limitations with dynamic rules -- we do not 252 * obey the 'randomized match', and we do not do multiple 253 * passes through the firewall. XXX check the latter!!! 254 */ 255static ipfw_dyn_rule **ipfw_dyn_v = NULL; 256static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 257static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 258 259static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 260#define IPFW_DYN_LOCK_INIT() \ 261 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 262#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 263#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 264#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 265#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 266 267/* 268 * Timeouts for various events in handing dynamic rules. 269 */ 270static u_int32_t dyn_ack_lifetime = 300; 271static u_int32_t dyn_syn_lifetime = 20; 272static u_int32_t dyn_fin_lifetime = 1; 273static u_int32_t dyn_rst_lifetime = 1; 274static u_int32_t dyn_udp_lifetime = 10; 275static u_int32_t dyn_short_lifetime = 5; 276 277/* 278 * Keepalives are sent if dyn_keepalive is set. They are sent every 279 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 280 * seconds of lifetime of a rule. 281 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 282 * than dyn_keepalive_period. 283 */ 284 285static u_int32_t dyn_keepalive_interval = 20; 286static u_int32_t dyn_keepalive_period = 5; 287static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 288 289static u_int32_t static_count; /* # of static rules */ 290static u_int32_t static_len; /* size in bytes of static rules */ 291static u_int32_t dyn_count; /* # of dynamic rules */ 292static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 293 294SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 295 &dyn_buckets, 0, "Number of dyn. buckets"); 296SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 297 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 298SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 299 &dyn_count, 0, "Number of dyn. rules"); 300SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 301 &dyn_max, 0, "Max number of dyn. rules"); 302SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 303 &static_count, 0, "Number of static rules"); 304SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 305 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 306SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 307 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 308SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 309 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 310SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 311 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 312SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 313 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 314SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 315 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 316SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 317 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 318 319#endif /* SYSCTL_NODE */ 320 321 322/* 323 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 324 * Other macros just cast void * into the appropriate type 325 */ 326#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 327#define TCP(p) ((struct tcphdr *)(p)) 328#define UDP(p) ((struct udphdr *)(p)) 329#define ICMP(p) ((struct icmphdr *)(p)) 330#define ICMP6(p) ((struct icmp6_hdr *)(p)) 331 332static __inline int 333icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 334{ 335 int type = icmp->icmp_type; 336 337 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 338} 339 340#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 341 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 342 343static int 344is_icmp_query(struct icmphdr *icmp) 345{ 346 int type = icmp->icmp_type; 347 348 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 349} 350#undef TT 351 352/* 353 * The following checks use two arrays of 8 or 16 bits to store the 354 * bits that we want set or clear, respectively. They are in the 355 * low and high half of cmd->arg1 or cmd->d[0]. 356 * 357 * We scan options and store the bits we find set. We succeed if 358 * 359 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 360 * 361 * The code is sometimes optimized not to store additional variables. 362 */ 363 364static int 365flags_match(ipfw_insn *cmd, u_int8_t bits) 366{ 367 u_char want_clear; 368 bits = ~bits; 369 370 if ( ((cmd->arg1 & 0xff) & bits) != 0) 371 return 0; /* some bits we want set were clear */ 372 want_clear = (cmd->arg1 >> 8) & 0xff; 373 if ( (want_clear & bits) != want_clear) 374 return 0; /* some bits we want clear were set */ 375 return 1; 376} 377 378static int 379ipopts_match(struct ip *ip, ipfw_insn *cmd) 380{ 381 int optlen, bits = 0; 382 u_char *cp = (u_char *)(ip + 1); 383 int x = (ip->ip_hl << 2) - sizeof (struct ip); 384 385 for (; x > 0; x -= optlen, cp += optlen) { 386 int opt = cp[IPOPT_OPTVAL]; 387 388 if (opt == IPOPT_EOL) 389 break; 390 if (opt == IPOPT_NOP) 391 optlen = 1; 392 else { 393 optlen = cp[IPOPT_OLEN]; 394 if (optlen <= 0 || optlen > x) 395 return 0; /* invalid or truncated */ 396 } 397 switch (opt) { 398 399 default: 400 break; 401 402 case IPOPT_LSRR: 403 bits |= IP_FW_IPOPT_LSRR; 404 break; 405 406 case IPOPT_SSRR: 407 bits |= IP_FW_IPOPT_SSRR; 408 break; 409 410 case IPOPT_RR: 411 bits |= IP_FW_IPOPT_RR; 412 break; 413 414 case IPOPT_TS: 415 bits |= IP_FW_IPOPT_TS; 416 break; 417 } 418 } 419 return (flags_match(cmd, bits)); 420} 421 422static int 423tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 424{ 425 int optlen, bits = 0; 426 u_char *cp = (u_char *)(tcp + 1); 427 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 428 429 for (; x > 0; x -= optlen, cp += optlen) { 430 int opt = cp[0]; 431 if (opt == TCPOPT_EOL) 432 break; 433 if (opt == TCPOPT_NOP) 434 optlen = 1; 435 else { 436 optlen = cp[1]; 437 if (optlen <= 0) 438 break; 439 } 440 441 switch (opt) { 442 443 default: 444 break; 445 446 case TCPOPT_MAXSEG: 447 bits |= IP_FW_TCPOPT_MSS; 448 break; 449 450 case TCPOPT_WINDOW: 451 bits |= IP_FW_TCPOPT_WINDOW; 452 break; 453 454 case TCPOPT_SACK_PERMITTED: 455 case TCPOPT_SACK: 456 bits |= IP_FW_TCPOPT_SACK; 457 break; 458 459 case TCPOPT_TIMESTAMP: 460 bits |= IP_FW_TCPOPT_TS; 461 break; 462 463 } 464 } 465 return (flags_match(cmd, bits)); 466} 467 468static int 469iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 470{ 471 if (ifp == NULL) /* no iface with this packet, match fails */ 472 return 0; 473 /* Check by name or by IP address */ 474 if (cmd->name[0] != '\0') { /* match by name */ 475 /* Check name */ 476 if (cmd->p.glob) { 477 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 478 return(1); 479 } else { 480 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 481 return(1); 482 } 483 } else { 484 struct ifaddr *ia; 485 486 /* XXX lock? */ 487 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 488 if (ia->ifa_addr == NULL) 489 continue; 490 if (ia->ifa_addr->sa_family != AF_INET) 491 continue; 492 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 493 (ia->ifa_addr))->sin_addr.s_addr) 494 return(1); /* match */ 495 } 496 } 497 return(0); /* no match, fail ... */ 498} 499 500/* 501 * The verify_path function checks if a route to the src exists and 502 * if it is reachable via ifp (when provided). 503 * 504 * The 'verrevpath' option checks that the interface that an IP packet 505 * arrives on is the same interface that traffic destined for the 506 * packet's source address would be routed out of. The 'versrcreach' 507 * option just checks that the source address is reachable via any route 508 * (except default) in the routing table. These two are a measure to block 509 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 510 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 511 * is purposely reminiscent of the Cisco IOS command, 512 * 513 * ip verify unicast reverse-path 514 * ip verify unicast source reachable-via any 515 * 516 * which implements the same functionality. But note that syntax is 517 * misleading. The check may be performed on all IP packets whether unicast, 518 * multicast, or broadcast. 519 */ 520static int 521verify_path(struct in_addr src, struct ifnet *ifp) 522{ 523 struct route ro; 524 struct sockaddr_in *dst; 525 526 bzero(&ro, sizeof(ro)); 527 528 dst = (struct sockaddr_in *)&(ro.ro_dst); 529 dst->sin_family = AF_INET; 530 dst->sin_len = sizeof(*dst); 531 dst->sin_addr = src; 532 rtalloc_ign(&ro, RTF_CLONING); 533 534 if (ro.ro_rt == NULL) 535 return 0; 536 537 /* if ifp is provided, check for equality with rtentry */ 538 if (ifp != NULL && ro.ro_rt->rt_ifp != ifp) { 539 RTFREE(ro.ro_rt); 540 return 0; 541 } 542 543 /* if no ifp provided, check if rtentry is not default route */ 544 if (ifp == NULL && 545 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 546 RTFREE(ro.ro_rt); 547 return 0; 548 } 549 550 /* or if this is a blackhole/reject route */ 551 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 552 RTFREE(ro.ro_rt); 553 return 0; 554 } 555 556 /* found valid route */ 557 RTFREE(ro.ro_rt); 558 return 1; 559} 560 561#ifdef INET6 562/* 563 * ipv6 specific rules here... 564 */ 565static __inline int 566icmp6type_match (int type, ipfw_insn_u32 *cmd) 567{ 568 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 569} 570 571static int 572flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 573{ 574 int i; 575 for (i=0; i <= cmd->o.arg1; ++i ) 576 if (curr_flow == cmd->d[i] ) 577 return 1; 578 return 0; 579} 580 581/* support for IP6_*_ME opcodes */ 582static int 583search_ip6_addr_net (struct in6_addr * ip6_addr) 584{ 585 struct ifnet *mdc; 586 struct ifaddr *mdc2; 587 struct in6_ifaddr *fdm; 588 struct in6_addr copia; 589 590 TAILQ_FOREACH(mdc, &ifnet, if_link) 591 for (mdc2 = mdc->if_addrlist.tqh_first; mdc2; 592 mdc2 = mdc2->ifa_list.tqe_next) { 593 if (!mdc2->ifa_addr) 594 continue; 595 if (mdc2->ifa_addr->sa_family == AF_INET6) { 596 fdm = (struct in6_ifaddr *)mdc2; 597 copia = fdm->ia_addr.sin6_addr; 598 /* need for leaving scope_id in the sock_addr */ 599 in6_clearscope(&copia); 600 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 601 return 1; 602 } 603 } 604 return 0; 605} 606 607static int 608verify_rev_path6(struct in6_addr *src, struct ifnet *ifp) 609{ 610 static struct route_in6 ro; 611 struct sockaddr_in6 *dst; 612 613 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 614 615 if ( !(IN6_ARE_ADDR_EQUAL (src, &dst->sin6_addr) )) { 616 bzero(dst, sizeof(*dst)); 617 dst->sin6_family = AF_INET6; 618 dst->sin6_len = sizeof(*dst); 619 dst->sin6_addr = *src; 620 rtalloc_ign((struct route *)&ro, RTF_CLONING); 621 } 622 if ((ro.ro_rt == NULL) || (ifp == NULL) || 623 (ro.ro_rt->rt_ifp->if_index != ifp->if_index)) 624 return 0; 625 return 1; 626} 627static __inline int 628hash_packet6(struct ipfw_flow_id *id) 629{ 630 u_int32_t i; 631 i= (id->dst_ip6.__u6_addr.__u6_addr32[0]) ^ 632 (id->dst_ip6.__u6_addr.__u6_addr32[1]) ^ 633 (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 634 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 635 (id->dst_port) ^ (id->src_port) ^ (id->flow_id6); 636 return i; 637} 638/* end of ipv6 opcodes */ 639 640#endif /* INET6 */ 641 642static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 643 644#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 645#define SNP(buf) buf, sizeof(buf) 646 647/* 648 * We enter here when we have a rule with O_LOG. 649 * XXX this function alone takes about 2Kbytes of code! 650 */ 651static void 652ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 653 struct mbuf *m, struct ifnet *oif) 654{ 655 char *action; 656 int limit_reached = 0; 657 char action2[40], proto[48], fragment[28]; 658 659 fragment[0] = '\0'; 660 proto[0] = '\0'; 661 662 if (f == NULL) { /* bogus pkt */ 663 if (verbose_limit != 0 && norule_counter >= verbose_limit) 664 return; 665 norule_counter++; 666 if (norule_counter == verbose_limit) 667 limit_reached = verbose_limit; 668 action = "Refuse"; 669 } else { /* O_LOG is the first action, find the real one */ 670 ipfw_insn *cmd = ACTION_PTR(f); 671 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 672 673 if (l->max_log != 0 && l->log_left == 0) 674 return; 675 l->log_left--; 676 if (l->log_left == 0) 677 limit_reached = l->max_log; 678 cmd += F_LEN(cmd); /* point to first action */ 679 if (cmd->opcode == O_ALTQ) { 680 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 681 682 snprintf(SNPARGS(action2, 0), "Altq %d", 683 altq->qid); 684 cmd += F_LEN(cmd); 685 } 686 if (cmd->opcode == O_PROB) 687 cmd += F_LEN(cmd); 688 689 action = action2; 690 switch (cmd->opcode) { 691 case O_DENY: 692 action = "Deny"; 693 break; 694 695 case O_REJECT: 696 if (cmd->arg1==ICMP_REJECT_RST) 697 action = "Reset"; 698 else if (cmd->arg1==ICMP_UNREACH_HOST) 699 action = "Reject"; 700 else 701 snprintf(SNPARGS(action2, 0), "Unreach %d", 702 cmd->arg1); 703 break; 704 705 case O_ACCEPT: 706 action = "Accept"; 707 break; 708 case O_COUNT: 709 action = "Count"; 710 break; 711 case O_DIVERT: 712 snprintf(SNPARGS(action2, 0), "Divert %d", 713 cmd->arg1); 714 break; 715 case O_TEE: 716 snprintf(SNPARGS(action2, 0), "Tee %d", 717 cmd->arg1); 718 break; 719 case O_SKIPTO: 720 snprintf(SNPARGS(action2, 0), "SkipTo %d", 721 cmd->arg1); 722 break; 723 case O_PIPE: 724 snprintf(SNPARGS(action2, 0), "Pipe %d", 725 cmd->arg1); 726 break; 727 case O_QUEUE: 728 snprintf(SNPARGS(action2, 0), "Queue %d", 729 cmd->arg1); 730 break; 731 case O_FORWARD_IP: { 732 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 733 int len; 734 735 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 736 inet_ntoa(sa->sa.sin_addr)); 737 if (sa->sa.sin_port) 738 snprintf(SNPARGS(action2, len), ":%d", 739 sa->sa.sin_port); 740 } 741 break; 742 case O_NETGRAPH: 743 snprintf(SNPARGS(action2, 0), "Netgraph %d", 744 cmd->arg1); 745 break; 746 case O_NGTEE: 747 snprintf(SNPARGS(action2, 0), "Ngtee %d", 748 cmd->arg1); 749 break; 750 default: 751 action = "UNKNOWN"; 752 break; 753 } 754 } 755 756 if (hlen == 0) { /* non-ip */ 757 snprintf(SNPARGS(proto, 0), "MAC"); 758 } else { 759 struct ip *ip = mtod(m, struct ip *); 760 /* these three are all aliases to the same thing */ 761 struct icmphdr *const icmp = L3HDR(struct icmphdr, ip); 762 struct tcphdr *const tcp = (struct tcphdr *)icmp; 763 struct udphdr *const udp = (struct udphdr *)icmp; 764 765 int ip_off, offset, ip_len; 766 767 int len; 768 769 if (eh != NULL) { /* layer 2 packets are as on the wire */ 770 ip_off = ntohs(ip->ip_off); 771 ip_len = ntohs(ip->ip_len); 772 } else { 773 ip_off = ip->ip_off; 774 ip_len = ip->ip_len; 775 } 776 offset = ip_off & IP_OFFMASK; 777 switch (ip->ip_p) { 778 case IPPROTO_TCP: 779 len = snprintf(SNPARGS(proto, 0), "TCP %s", 780 inet_ntoa(ip->ip_src)); 781 if (offset == 0) 782 snprintf(SNPARGS(proto, len), ":%d %s:%d", 783 ntohs(tcp->th_sport), 784 inet_ntoa(ip->ip_dst), 785 ntohs(tcp->th_dport)); 786 else 787 snprintf(SNPARGS(proto, len), " %s", 788 inet_ntoa(ip->ip_dst)); 789 break; 790 791 case IPPROTO_UDP: 792 len = snprintf(SNPARGS(proto, 0), "UDP %s", 793 inet_ntoa(ip->ip_src)); 794 if (offset == 0) 795 snprintf(SNPARGS(proto, len), ":%d %s:%d", 796 ntohs(udp->uh_sport), 797 inet_ntoa(ip->ip_dst), 798 ntohs(udp->uh_dport)); 799 else 800 snprintf(SNPARGS(proto, len), " %s", 801 inet_ntoa(ip->ip_dst)); 802 break; 803 804 case IPPROTO_ICMP: 805 if (offset == 0) 806 len = snprintf(SNPARGS(proto, 0), 807 "ICMP:%u.%u ", 808 icmp->icmp_type, icmp->icmp_code); 809 else 810 len = snprintf(SNPARGS(proto, 0), "ICMP "); 811 len += snprintf(SNPARGS(proto, len), "%s", 812 inet_ntoa(ip->ip_src)); 813 snprintf(SNPARGS(proto, len), " %s", 814 inet_ntoa(ip->ip_dst)); 815 break; 816 817 default: 818 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 819 inet_ntoa(ip->ip_src)); 820 snprintf(SNPARGS(proto, len), " %s", 821 inet_ntoa(ip->ip_dst)); 822 break; 823 } 824 825 if (ip_off & (IP_MF | IP_OFFMASK)) 826 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 827 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 828 offset << 3, 829 (ip_off & IP_MF) ? "+" : ""); 830 } 831 if (oif || m->m_pkthdr.rcvif) 832 log(LOG_SECURITY | LOG_INFO, 833 "ipfw: %d %s %s %s via %s%s\n", 834 f ? f->rulenum : -1, 835 action, proto, oif ? "out" : "in", 836 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 837 fragment); 838 else 839 log(LOG_SECURITY | LOG_INFO, 840 "ipfw: %d %s %s [no if info]%s\n", 841 f ? f->rulenum : -1, 842 action, proto, fragment); 843 if (limit_reached) 844 log(LOG_SECURITY | LOG_NOTICE, 845 "ipfw: limit %d reached on entry %d\n", 846 limit_reached, f ? f->rulenum : -1); 847} 848 849/* 850 * IMPORTANT: the hash function for dynamic rules must be commutative 851 * in source and destination (ip,port), because rules are bidirectional 852 * and we want to find both in the same bucket. 853 */ 854static __inline int 855hash_packet(struct ipfw_flow_id *id) 856{ 857 u_int32_t i; 858 859#ifdef INET6 860 if (IS_IP6_FLOW_ID(id)) 861 i = hash_packet6(id); 862 else 863#endif /* INET6 */ 864 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 865 i &= (curr_dyn_buckets - 1); 866 return i; 867} 868 869/** 870 * unlink a dynamic rule from a chain. prev is a pointer to 871 * the previous one, q is a pointer to the rule to delete, 872 * head is a pointer to the head of the queue. 873 * Modifies q and potentially also head. 874 */ 875#define UNLINK_DYN_RULE(prev, head, q) { \ 876 ipfw_dyn_rule *old_q = q; \ 877 \ 878 /* remove a refcount to the parent */ \ 879 if (q->dyn_type == O_LIMIT) \ 880 q->parent->count--; \ 881 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 882 (q->id.src_ip), (q->id.src_port), \ 883 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 884 if (prev != NULL) \ 885 prev->next = q = q->next; \ 886 else \ 887 head = q = q->next; \ 888 dyn_count--; \ 889 uma_zfree(ipfw_dyn_rule_zone, old_q); } 890 891#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 892 893/** 894 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 895 * 896 * If keep_me == NULL, rules are deleted even if not expired, 897 * otherwise only expired rules are removed. 898 * 899 * The value of the second parameter is also used to point to identify 900 * a rule we absolutely do not want to remove (e.g. because we are 901 * holding a reference to it -- this is the case with O_LIMIT_PARENT 902 * rules). The pointer is only used for comparison, so any non-null 903 * value will do. 904 */ 905static void 906remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 907{ 908 static u_int32_t last_remove = 0; 909 910#define FORCE (keep_me == NULL) 911 912 ipfw_dyn_rule *prev, *q; 913 int i, pass = 0, max_pass = 0; 914 915 IPFW_DYN_LOCK_ASSERT(); 916 917 if (ipfw_dyn_v == NULL || dyn_count == 0) 918 return; 919 /* do not expire more than once per second, it is useless */ 920 if (!FORCE && last_remove == time_second) 921 return; 922 last_remove = time_second; 923 924 /* 925 * because O_LIMIT refer to parent rules, during the first pass only 926 * remove child and mark any pending LIMIT_PARENT, and remove 927 * them in a second pass. 928 */ 929next_pass: 930 for (i = 0 ; i < curr_dyn_buckets ; i++) { 931 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 932 /* 933 * Logic can become complex here, so we split tests. 934 */ 935 if (q == keep_me) 936 goto next; 937 if (rule != NULL && rule != q->rule) 938 goto next; /* not the one we are looking for */ 939 if (q->dyn_type == O_LIMIT_PARENT) { 940 /* 941 * handle parent in the second pass, 942 * record we need one. 943 */ 944 max_pass = 1; 945 if (pass == 0) 946 goto next; 947 if (FORCE && q->count != 0 ) { 948 /* XXX should not happen! */ 949 printf("ipfw: OUCH! cannot remove rule," 950 " count %d\n", q->count); 951 } 952 } else { 953 if (!FORCE && 954 !TIME_LEQ( q->expire, time_second )) 955 goto next; 956 } 957 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 958 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 959 continue; 960 } 961next: 962 prev=q; 963 q=q->next; 964 } 965 } 966 if (pass++ < max_pass) 967 goto next_pass; 968} 969 970 971/** 972 * lookup a dynamic rule. 973 */ 974static ipfw_dyn_rule * 975lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 976 struct tcphdr *tcp) 977{ 978 /* 979 * stateful ipfw extensions. 980 * Lookup into dynamic session queue 981 */ 982#define MATCH_REVERSE 0 983#define MATCH_FORWARD 1 984#define MATCH_NONE 2 985#define MATCH_UNKNOWN 3 986 int i, dir = MATCH_NONE; 987 ipfw_dyn_rule *prev, *q=NULL; 988 989 IPFW_DYN_LOCK_ASSERT(); 990 991 if (ipfw_dyn_v == NULL) 992 goto done; /* not found */ 993 i = hash_packet( pkt ); 994 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 995 if (q->dyn_type == O_LIMIT_PARENT && q->count) 996 goto next; 997 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 998 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 999 continue; 1000 } 1001 if (pkt->proto == q->id.proto && 1002 q->dyn_type != O_LIMIT_PARENT) { 1003 if (IS_IP6_FLOW_ID(pkt)) { 1004 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1005 &(q->id.src_ip6)) && 1006 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1007 &(q->id.dst_ip6)) && 1008 pkt->src_port == q->id.src_port && 1009 pkt->dst_port == q->id.dst_port ) { 1010 dir = MATCH_FORWARD; 1011 break; 1012 } 1013 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1014 &(q->id.dst_ip6)) && 1015 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1016 &(q->id.src_ip6)) && 1017 pkt->src_port == q->id.dst_port && 1018 pkt->dst_port == q->id.src_port ) { 1019 dir = MATCH_REVERSE; 1020 break; 1021 } 1022 } else { 1023 if (pkt->src_ip == q->id.src_ip && 1024 pkt->dst_ip == q->id.dst_ip && 1025 pkt->src_port == q->id.src_port && 1026 pkt->dst_port == q->id.dst_port ) { 1027 dir = MATCH_FORWARD; 1028 break; 1029 } 1030 if (pkt->src_ip == q->id.dst_ip && 1031 pkt->dst_ip == q->id.src_ip && 1032 pkt->src_port == q->id.dst_port && 1033 pkt->dst_port == q->id.src_port ) { 1034 dir = MATCH_REVERSE; 1035 break; 1036 } 1037 } 1038 } 1039next: 1040 prev = q; 1041 q = q->next; 1042 } 1043 if (q == NULL) 1044 goto done; /* q = NULL, not found */ 1045 1046 if ( prev != NULL) { /* found and not in front */ 1047 prev->next = q->next; 1048 q->next = ipfw_dyn_v[i]; 1049 ipfw_dyn_v[i] = q; 1050 } 1051 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1052 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1053 1054#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1055#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1056 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1057 switch (q->state) { 1058 case TH_SYN: /* opening */ 1059 q->expire = time_second + dyn_syn_lifetime; 1060 break; 1061 1062 case BOTH_SYN: /* move to established */ 1063 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1064 case BOTH_SYN | (TH_FIN << 8) : 1065 if (tcp) { 1066#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1067 u_int32_t ack = ntohl(tcp->th_ack); 1068 if (dir == MATCH_FORWARD) { 1069 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1070 q->ack_fwd = ack; 1071 else { /* ignore out-of-sequence */ 1072 break; 1073 } 1074 } else { 1075 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1076 q->ack_rev = ack; 1077 else { /* ignore out-of-sequence */ 1078 break; 1079 } 1080 } 1081 } 1082 q->expire = time_second + dyn_ack_lifetime; 1083 break; 1084 1085 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1086 if (dyn_fin_lifetime >= dyn_keepalive_period) 1087 dyn_fin_lifetime = dyn_keepalive_period - 1; 1088 q->expire = time_second + dyn_fin_lifetime; 1089 break; 1090 1091 default: 1092#if 0 1093 /* 1094 * reset or some invalid combination, but can also 1095 * occur if we use keep-state the wrong way. 1096 */ 1097 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1098 printf("invalid state: 0x%x\n", q->state); 1099#endif 1100 if (dyn_rst_lifetime >= dyn_keepalive_period) 1101 dyn_rst_lifetime = dyn_keepalive_period - 1; 1102 q->expire = time_second + dyn_rst_lifetime; 1103 break; 1104 } 1105 } else if (pkt->proto == IPPROTO_UDP) { 1106 q->expire = time_second + dyn_udp_lifetime; 1107 } else { 1108 /* other protocols */ 1109 q->expire = time_second + dyn_short_lifetime; 1110 } 1111done: 1112 if (match_direction) 1113 *match_direction = dir; 1114 return q; 1115} 1116 1117static ipfw_dyn_rule * 1118lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1119 struct tcphdr *tcp) 1120{ 1121 ipfw_dyn_rule *q; 1122 1123 IPFW_DYN_LOCK(); 1124 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1125 if (q == NULL) 1126 IPFW_DYN_UNLOCK(); 1127 /* NB: return table locked when q is not NULL */ 1128 return q; 1129} 1130 1131static void 1132realloc_dynamic_table(void) 1133{ 1134 IPFW_DYN_LOCK_ASSERT(); 1135 1136 /* 1137 * Try reallocation, make sure we have a power of 2 and do 1138 * not allow more than 64k entries. In case of overflow, 1139 * default to 1024. 1140 */ 1141 1142 if (dyn_buckets > 65536) 1143 dyn_buckets = 1024; 1144 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1145 dyn_buckets = curr_dyn_buckets; /* reset */ 1146 return; 1147 } 1148 curr_dyn_buckets = dyn_buckets; 1149 if (ipfw_dyn_v != NULL) 1150 free(ipfw_dyn_v, M_IPFW); 1151 for (;;) { 1152 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1153 M_IPFW, M_NOWAIT | M_ZERO); 1154 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1155 break; 1156 curr_dyn_buckets /= 2; 1157 } 1158} 1159 1160/** 1161 * Install state of type 'type' for a dynamic session. 1162 * The hash table contains two type of rules: 1163 * - regular rules (O_KEEP_STATE) 1164 * - rules for sessions with limited number of sess per user 1165 * (O_LIMIT). When they are created, the parent is 1166 * increased by 1, and decreased on delete. In this case, 1167 * the third parameter is the parent rule and not the chain. 1168 * - "parent" rules for the above (O_LIMIT_PARENT). 1169 */ 1170static ipfw_dyn_rule * 1171add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1172{ 1173 ipfw_dyn_rule *r; 1174 int i; 1175 1176 IPFW_DYN_LOCK_ASSERT(); 1177 1178 if (ipfw_dyn_v == NULL || 1179 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1180 realloc_dynamic_table(); 1181 if (ipfw_dyn_v == NULL) 1182 return NULL; /* failed ! */ 1183 } 1184 i = hash_packet(id); 1185 1186 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1187 if (r == NULL) { 1188 printf ("ipfw: sorry cannot allocate state\n"); 1189 return NULL; 1190 } 1191 1192 /* increase refcount on parent, and set pointer */ 1193 if (dyn_type == O_LIMIT) { 1194 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1195 if ( parent->dyn_type != O_LIMIT_PARENT) 1196 panic("invalid parent"); 1197 parent->count++; 1198 r->parent = parent; 1199 rule = parent->rule; 1200 } 1201 1202 r->id = *id; 1203 r->expire = time_second + dyn_syn_lifetime; 1204 r->rule = rule; 1205 r->dyn_type = dyn_type; 1206 r->pcnt = r->bcnt = 0; 1207 r->count = 0; 1208 1209 r->bucket = i; 1210 r->next = ipfw_dyn_v[i]; 1211 ipfw_dyn_v[i] = r; 1212 dyn_count++; 1213 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1214 dyn_type, 1215 (r->id.src_ip), (r->id.src_port), 1216 (r->id.dst_ip), (r->id.dst_port), 1217 dyn_count ); ) 1218 return r; 1219} 1220 1221/** 1222 * lookup dynamic parent rule using pkt and rule as search keys. 1223 * If the lookup fails, then install one. 1224 */ 1225static ipfw_dyn_rule * 1226lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1227{ 1228 ipfw_dyn_rule *q; 1229 int i; 1230 1231 IPFW_DYN_LOCK_ASSERT(); 1232 1233 if (ipfw_dyn_v) { 1234 int is_v6 = IS_IP6_FLOW_ID(pkt); 1235 i = hash_packet( pkt ); 1236 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1237 if (q->dyn_type == O_LIMIT_PARENT && 1238 rule== q->rule && 1239 pkt->proto == q->id.proto && 1240 pkt->src_port == q->id.src_port && 1241 pkt->dst_port == q->id.dst_port && 1242 ( 1243 (is_v6 && 1244 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1245 &(q->id.src_ip6)) && 1246 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1247 &(q->id.dst_ip6))) || 1248 (!is_v6 && 1249 pkt->src_ip == q->id.src_ip && 1250 pkt->dst_ip == q->id.dst_ip) 1251 ) 1252 ) { 1253 q->expire = time_second + dyn_short_lifetime; 1254 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1255 return q; 1256 } 1257 } 1258 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1259} 1260 1261/** 1262 * Install dynamic state for rule type cmd->o.opcode 1263 * 1264 * Returns 1 (failure) if state is not installed because of errors or because 1265 * session limitations are enforced. 1266 */ 1267static int 1268install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1269 struct ip_fw_args *args) 1270{ 1271 static int last_log; 1272 1273 ipfw_dyn_rule *q; 1274 1275 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 1276 cmd->o.opcode, 1277 (args->f_id.src_ip), (args->f_id.src_port), 1278 (args->f_id.dst_ip), (args->f_id.dst_port) );) 1279 1280 IPFW_DYN_LOCK(); 1281 1282 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1283 1284 if (q != NULL) { /* should never occur */ 1285 if (last_log != time_second) { 1286 last_log = time_second; 1287 printf("ipfw: install_state: entry already present, done\n"); 1288 } 1289 IPFW_DYN_UNLOCK(); 1290 return 0; 1291 } 1292 1293 if (dyn_count >= dyn_max) 1294 /* 1295 * Run out of slots, try to remove any expired rule. 1296 */ 1297 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1298 1299 if (dyn_count >= dyn_max) { 1300 if (last_log != time_second) { 1301 last_log = time_second; 1302 printf("ipfw: install_state: Too many dynamic rules\n"); 1303 } 1304 IPFW_DYN_UNLOCK(); 1305 return 1; /* cannot install, notify caller */ 1306 } 1307 1308 switch (cmd->o.opcode) { 1309 case O_KEEP_STATE: /* bidir rule */ 1310 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1311 break; 1312 1313 case O_LIMIT: /* limit number of sessions */ 1314 { 1315 u_int16_t limit_mask = cmd->limit_mask; 1316 struct ipfw_flow_id id; 1317 ipfw_dyn_rule *parent; 1318 1319 DEB(printf("ipfw: installing dyn-limit rule %d\n", 1320 cmd->conn_limit);) 1321 1322 id.dst_ip = id.src_ip = 0; 1323 id.dst_port = id.src_port = 0; 1324 id.proto = args->f_id.proto; 1325 1326 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1327 if (limit_mask & DYN_SRC_ADDR) 1328 id.src_ip6 = args->f_id.src_ip6; 1329 if (limit_mask & DYN_DST_ADDR) 1330 id.dst_ip6 = args->f_id.dst_ip6; 1331 } else { 1332 if (limit_mask & DYN_SRC_ADDR) 1333 id.src_ip = args->f_id.src_ip; 1334 if (limit_mask & DYN_DST_ADDR) 1335 id.dst_ip = args->f_id.dst_ip; 1336 } 1337 if (limit_mask & DYN_SRC_PORT) 1338 id.src_port = args->f_id.src_port; 1339 if (limit_mask & DYN_DST_PORT) 1340 id.dst_port = args->f_id.dst_port; 1341 parent = lookup_dyn_parent(&id, rule); 1342 if (parent == NULL) { 1343 printf("ipfw: add parent failed\n"); 1344 return 1; 1345 } 1346 if (parent->count >= cmd->conn_limit) { 1347 /* 1348 * See if we can remove some expired rule. 1349 */ 1350 remove_dyn_rule(rule, parent); 1351 if (parent->count >= cmd->conn_limit) { 1352 if (fw_verbose && last_log != time_second) { 1353 last_log = time_second; 1354 log(LOG_SECURITY | LOG_DEBUG, 1355 "drop session, too many entries\n"); 1356 } 1357 IPFW_DYN_UNLOCK(); 1358 return 1; 1359 } 1360 } 1361 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1362 } 1363 break; 1364 default: 1365 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 1366 IPFW_DYN_UNLOCK(); 1367 return 1; 1368 } 1369 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1370 IPFW_DYN_UNLOCK(); 1371 return 0; 1372} 1373 1374/* 1375 * Transmit a TCP packet, containing either a RST or a keepalive. 1376 * When flags & TH_RST, we are sending a RST packet, because of a 1377 * "reset" action matched the packet. 1378 * Otherwise we are sending a keepalive, and flags & TH_ 1379 */ 1380static void 1381send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1382{ 1383 struct mbuf *m; 1384 struct ip *ip; 1385 struct tcphdr *tcp; 1386 1387 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1388 if (m == 0) 1389 return; 1390 m->m_pkthdr.rcvif = (struct ifnet *)0; 1391 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1392 m->m_data += max_linkhdr; 1393 1394 ip = mtod(m, struct ip *); 1395 bzero(ip, m->m_len); 1396 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1397 ip->ip_p = IPPROTO_TCP; 1398 tcp->th_off = 5; 1399 /* 1400 * Assume we are sending a RST (or a keepalive in the reverse 1401 * direction), swap src and destination addresses and ports. 1402 */ 1403 ip->ip_src.s_addr = htonl(id->dst_ip); 1404 ip->ip_dst.s_addr = htonl(id->src_ip); 1405 tcp->th_sport = htons(id->dst_port); 1406 tcp->th_dport = htons(id->src_port); 1407 if (flags & TH_RST) { /* we are sending a RST */ 1408 if (flags & TH_ACK) { 1409 tcp->th_seq = htonl(ack); 1410 tcp->th_ack = htonl(0); 1411 tcp->th_flags = TH_RST; 1412 } else { 1413 if (flags & TH_SYN) 1414 seq++; 1415 tcp->th_seq = htonl(0); 1416 tcp->th_ack = htonl(seq); 1417 tcp->th_flags = TH_RST | TH_ACK; 1418 } 1419 } else { 1420 /* 1421 * We are sending a keepalive. flags & TH_SYN determines 1422 * the direction, forward if set, reverse if clear. 1423 * NOTE: seq and ack are always assumed to be correct 1424 * as set by the caller. This may be confusing... 1425 */ 1426 if (flags & TH_SYN) { 1427 /* 1428 * we have to rewrite the correct addresses! 1429 */ 1430 ip->ip_dst.s_addr = htonl(id->dst_ip); 1431 ip->ip_src.s_addr = htonl(id->src_ip); 1432 tcp->th_dport = htons(id->dst_port); 1433 tcp->th_sport = htons(id->src_port); 1434 } 1435 tcp->th_seq = htonl(seq); 1436 tcp->th_ack = htonl(ack); 1437 tcp->th_flags = TH_ACK; 1438 } 1439 /* 1440 * set ip_len to the payload size so we can compute 1441 * the tcp checksum on the pseudoheader 1442 * XXX check this, could save a couple of words ? 1443 */ 1444 ip->ip_len = htons(sizeof(struct tcphdr)); 1445 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1446 /* 1447 * now fill fields left out earlier 1448 */ 1449 ip->ip_ttl = ip_defttl; 1450 ip->ip_len = m->m_pkthdr.len; 1451 m->m_flags |= M_SKIP_FIREWALL; 1452 ip_output(m, NULL, NULL, 0, NULL, NULL); 1453} 1454 1455/* 1456 * sends a reject message, consuming the mbuf passed as an argument. 1457 */ 1458static void 1459send_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 1460{ 1461 1462 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1463 /* We need the IP header in host order for icmp_error(). */ 1464 if (args->eh != NULL) { 1465 struct ip *ip = mtod(args->m, struct ip *); 1466 ip->ip_len = ntohs(ip->ip_len); 1467 ip->ip_off = ntohs(ip->ip_off); 1468 } 1469 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1470 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1471 struct tcphdr *const tcp = 1472 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1473 if ( (tcp->th_flags & TH_RST) == 0) 1474 send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1475 ntohl(tcp->th_ack), 1476 tcp->th_flags | TH_RST); 1477 m_freem(args->m); 1478 } else 1479 m_freem(args->m); 1480 args->m = NULL; 1481} 1482 1483/** 1484 * 1485 * Given an ip_fw *, lookup_next_rule will return a pointer 1486 * to the next rule, which can be either the jump 1487 * target (for skipto instructions) or the next one in the list (in 1488 * all other cases including a missing jump target). 1489 * The result is also written in the "next_rule" field of the rule. 1490 * Backward jumps are not allowed, so start looking from the next 1491 * rule... 1492 * 1493 * This never returns NULL -- in case we do not have an exact match, 1494 * the next rule is returned. When the ruleset is changed, 1495 * pointers are flushed so we are always correct. 1496 */ 1497 1498static struct ip_fw * 1499lookup_next_rule(struct ip_fw *me) 1500{ 1501 struct ip_fw *rule = NULL; 1502 ipfw_insn *cmd; 1503 1504 /* look for action, in case it is a skipto */ 1505 cmd = ACTION_PTR(me); 1506 if (cmd->opcode == O_LOG) 1507 cmd += F_LEN(cmd); 1508 if (cmd->opcode == O_ALTQ) 1509 cmd += F_LEN(cmd); 1510 if ( cmd->opcode == O_SKIPTO ) 1511 for (rule = me->next; rule ; rule = rule->next) 1512 if (rule->rulenum >= cmd->arg1) 1513 break; 1514 if (rule == NULL) /* failure or not a skipto */ 1515 rule = me->next; 1516 me->next_rule = rule; 1517 return rule; 1518} 1519 1520static void 1521init_tables(void) 1522{ 1523 int i; 1524 1525 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1526 rn_inithead((void **)&ipfw_tables[i].rnh, 32); 1527 ipfw_tables[i].modified = 1; 1528 } 1529} 1530 1531static int 1532add_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen, u_int32_t value) 1533{ 1534 struct radix_node_head *rnh; 1535 struct table_entry *ent; 1536 1537 if (tbl >= IPFW_TABLES_MAX) 1538 return (EINVAL); 1539 rnh = ipfw_tables[tbl].rnh; 1540 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1541 if (ent == NULL) 1542 return (ENOMEM); 1543 ent->value = value; 1544 ent->addr.sin_len = ent->mask.sin_len = 8; 1545 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1546 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1547 RADIX_NODE_HEAD_LOCK(rnh); 1548 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1549 NULL) { 1550 RADIX_NODE_HEAD_UNLOCK(rnh); 1551 free(ent, M_IPFW_TBL); 1552 return (EEXIST); 1553 } 1554 ipfw_tables[tbl].modified = 1; 1555 RADIX_NODE_HEAD_UNLOCK(rnh); 1556 return (0); 1557} 1558 1559static int 1560del_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen) 1561{ 1562 struct radix_node_head *rnh; 1563 struct table_entry *ent; 1564 struct sockaddr_in sa, mask; 1565 1566 if (tbl >= IPFW_TABLES_MAX) 1567 return (EINVAL); 1568 rnh = ipfw_tables[tbl].rnh; 1569 sa.sin_len = mask.sin_len = 8; 1570 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1571 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1572 RADIX_NODE_HEAD_LOCK(rnh); 1573 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1574 if (ent == NULL) { 1575 RADIX_NODE_HEAD_UNLOCK(rnh); 1576 return (ESRCH); 1577 } 1578 ipfw_tables[tbl].modified = 1; 1579 RADIX_NODE_HEAD_UNLOCK(rnh); 1580 free(ent, M_IPFW_TBL); 1581 return (0); 1582} 1583 1584static int 1585flush_table_entry(struct radix_node *rn, void *arg) 1586{ 1587 struct radix_node_head * const rnh = arg; 1588 struct table_entry *ent; 1589 1590 ent = (struct table_entry *) 1591 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1592 if (ent != NULL) 1593 free(ent, M_IPFW_TBL); 1594 return (0); 1595} 1596 1597static int 1598flush_table(u_int16_t tbl) 1599{ 1600 struct radix_node_head *rnh; 1601 1602 if (tbl >= IPFW_TABLES_MAX) 1603 return (EINVAL); 1604 rnh = ipfw_tables[tbl].rnh; 1605 RADIX_NODE_HEAD_LOCK(rnh); 1606 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1607 ipfw_tables[tbl].modified = 1; 1608 RADIX_NODE_HEAD_UNLOCK(rnh); 1609 return (0); 1610} 1611 1612static void 1613flush_tables(void) 1614{ 1615 u_int16_t tbl; 1616 1617 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1618 flush_table(tbl); 1619} 1620 1621static int 1622lookup_table(u_int16_t tbl, in_addr_t addr, u_int32_t *val) 1623{ 1624 struct radix_node_head *rnh; 1625 struct table_entry *ent; 1626 struct sockaddr_in sa; 1627 static in_addr_t last_addr; 1628 static int last_tbl; 1629 static int last_match; 1630 static u_int32_t last_value; 1631 1632 if (tbl >= IPFW_TABLES_MAX) 1633 return (0); 1634 if (tbl == last_tbl && addr == last_addr && 1635 !ipfw_tables[tbl].modified) { 1636 if (last_match) 1637 *val = last_value; 1638 return (last_match); 1639 } 1640 rnh = ipfw_tables[tbl].rnh; 1641 sa.sin_len = 8; 1642 sa.sin_addr.s_addr = addr; 1643 RADIX_NODE_HEAD_LOCK(rnh); 1644 ipfw_tables[tbl].modified = 0; 1645 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1646 RADIX_NODE_HEAD_UNLOCK(rnh); 1647 last_addr = addr; 1648 last_tbl = tbl; 1649 if (ent != NULL) { 1650 last_value = *val = ent->value; 1651 last_match = 1; 1652 return (1); 1653 } 1654 last_match = 0; 1655 return (0); 1656} 1657 1658static int 1659count_table_entry(struct radix_node *rn, void *arg) 1660{ 1661 u_int32_t * const cnt = arg; 1662 1663 (*cnt)++; 1664 return (0); 1665} 1666 1667static int 1668count_table(u_int32_t tbl, u_int32_t *cnt) 1669{ 1670 struct radix_node_head *rnh; 1671 1672 if (tbl >= IPFW_TABLES_MAX) 1673 return (EINVAL); 1674 rnh = ipfw_tables[tbl].rnh; 1675 *cnt = 0; 1676 RADIX_NODE_HEAD_LOCK(rnh); 1677 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1678 RADIX_NODE_HEAD_UNLOCK(rnh); 1679 return (0); 1680} 1681 1682static int 1683dump_table_entry(struct radix_node *rn, void *arg) 1684{ 1685 struct table_entry * const n = (struct table_entry *)rn; 1686 ipfw_table * const tbl = arg; 1687 ipfw_table_entry *ent; 1688 1689 if (tbl->cnt == tbl->size) 1690 return (1); 1691 ent = &tbl->ent[tbl->cnt]; 1692 ent->tbl = tbl->tbl; 1693 if (in_nullhost(n->mask.sin_addr)) 1694 ent->masklen = 0; 1695 else 1696 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1697 ent->addr = n->addr.sin_addr.s_addr; 1698 ent->value = n->value; 1699 tbl->cnt++; 1700 return (0); 1701} 1702 1703static int 1704dump_table(ipfw_table *tbl) 1705{ 1706 struct radix_node_head *rnh; 1707 1708 if (tbl->tbl >= IPFW_TABLES_MAX) 1709 return (EINVAL); 1710 rnh = ipfw_tables[tbl->tbl].rnh; 1711 tbl->cnt = 0; 1712 RADIX_NODE_HEAD_LOCK(rnh); 1713 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1714 RADIX_NODE_HEAD_UNLOCK(rnh); 1715 return (0); 1716} 1717 1718static void 1719fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1720{ 1721 struct ucred *cr; 1722 1723 if (inp->inp_socket != NULL) { 1724 cr = inp->inp_socket->so_cred; 1725 ugp->fw_prid = jailed(cr) ? 1726 cr->cr_prison->pr_id : -1; 1727 ugp->fw_uid = cr->cr_uid; 1728 ugp->fw_ngroups = cr->cr_ngroups; 1729 bcopy(cr->cr_groups, ugp->fw_groups, 1730 sizeof(ugp->fw_groups)); 1731 } 1732} 1733 1734static int 1735check_uidgid(ipfw_insn_u32 *insn, 1736 int proto, struct ifnet *oif, 1737 struct in_addr dst_ip, u_int16_t dst_port, 1738 struct in_addr src_ip, u_int16_t src_port, 1739 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1740{ 1741 struct inpcbinfo *pi; 1742 int wildcard; 1743 struct inpcb *pcb; 1744 int match; 1745 gid_t *gp; 1746 1747 /* 1748 * Check to see if the UDP or TCP stack supplied us with 1749 * the PCB. If so, rather then holding a lock and looking 1750 * up the PCB, we can use the one that was supplied. 1751 */ 1752 if (inp && *lookup == 0) { 1753 INP_LOCK_ASSERT(inp); 1754 if (inp->inp_socket != NULL) { 1755 fill_ugid_cache(inp, ugp); 1756 *lookup = 1; 1757 } 1758 } 1759 /* 1760 * If we have already been here and the packet has no 1761 * PCB entry associated with it, then we can safely 1762 * assume that this is a no match. 1763 */ 1764 if (*lookup == -1) 1765 return (0); 1766 if (proto == IPPROTO_TCP) { 1767 wildcard = 0; 1768 pi = &tcbinfo; 1769 } else if (proto == IPPROTO_UDP) { 1770 wildcard = 1; 1771 pi = &udbinfo; 1772 } else 1773 return 0; 1774 match = 0; 1775 if (*lookup == 0) { 1776 INP_INFO_RLOCK(pi); 1777 pcb = (oif) ? 1778 in_pcblookup_hash(pi, 1779 dst_ip, htons(dst_port), 1780 src_ip, htons(src_port), 1781 wildcard, oif) : 1782 in_pcblookup_hash(pi, 1783 src_ip, htons(src_port), 1784 dst_ip, htons(dst_port), 1785 wildcard, NULL); 1786 if (pcb != NULL) { 1787 INP_LOCK(pcb); 1788 if (pcb->inp_socket != NULL) { 1789 fill_ugid_cache(pcb, ugp); 1790 *lookup = 1; 1791 } 1792 INP_UNLOCK(pcb); 1793 } 1794 INP_INFO_RUNLOCK(pi); 1795 if (*lookup == 0) { 1796 /* 1797 * If the lookup did not yield any results, there 1798 * is no sense in coming back and trying again. So 1799 * we can set lookup to -1 and ensure that we wont 1800 * bother the pcb system again. 1801 */ 1802 *lookup = -1; 1803 return (0); 1804 } 1805 } 1806 if (insn->o.opcode == O_UID) 1807 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1808 else if (insn->o.opcode == O_GID) { 1809 for (gp = ugp->fw_groups; 1810 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1811 if (*gp == (gid_t)insn->d[0]) { 1812 match = 1; 1813 break; 1814 } 1815 } else if (insn->o.opcode == O_JAIL) 1816 match = (ugp->fw_prid == (int)insn->d[0]); 1817 return match; 1818} 1819 1820/* 1821 * The main check routine for the firewall. 1822 * 1823 * All arguments are in args so we can modify them and return them 1824 * back to the caller. 1825 * 1826 * Parameters: 1827 * 1828 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1829 * Starts with the IP header. 1830 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1831 * args->oif Outgoing interface, or NULL if packet is incoming. 1832 * The incoming interface is in the mbuf. (in) 1833 * args->divert_rule (in/out) 1834 * Skip up to the first rule past this rule number; 1835 * upon return, non-zero port number for divert or tee. 1836 * 1837 * args->rule Pointer to the last matching rule (in/out) 1838 * args->next_hop Socket we are forwarding to (out). 1839 * args->f_id Addresses grabbed from the packet (out) 1840 * args->cookie a cookie depending on rule action 1841 * 1842 * Return value: 1843 * 1844 * IP_FW_PASS the packet must be accepted 1845 * IP_FW_DENY the packet must be dropped 1846 * IP_FW_DIVERT divert packet, port in m_tag 1847 * IP_FW_TEE tee packet, port in m_tag 1848 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 1849 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 1850 * 1851 */ 1852 1853int 1854ipfw_chk(struct ip_fw_args *args) 1855{ 1856 /* 1857 * Local variables hold state during the processing of a packet. 1858 * 1859 * IMPORTANT NOTE: to speed up the processing of rules, there 1860 * are some assumption on the values of the variables, which 1861 * are documented here. Should you change them, please check 1862 * the implementation of the various instructions to make sure 1863 * that they still work. 1864 * 1865 * args->eh The MAC header. It is non-null for a layer2 1866 * packet, it is NULL for a layer-3 packet. 1867 * 1868 * m | args->m Pointer to the mbuf, as received from the caller. 1869 * It may change if ipfw_chk() does an m_pullup, or if it 1870 * consumes the packet because it calls send_reject(). 1871 * XXX This has to change, so that ipfw_chk() never modifies 1872 * or consumes the buffer. 1873 * ip is simply an alias of the value of m, and it is kept 1874 * in sync with it (the packet is supposed to start with 1875 * the ip header). 1876 */ 1877 struct mbuf *m = args->m; 1878 struct ip *ip = mtod(m, struct ip *); 1879 1880 /* 1881 * For rules which contain uid/gid or jail constraints, cache 1882 * a copy of the users credentials after the pcb lookup has been 1883 * executed. This will speed up the processing of rules with 1884 * these types of constraints, as well as decrease contention 1885 * on pcb related locks. 1886 */ 1887 struct ip_fw_ugid fw_ugid_cache; 1888 int ugid_lookup = 0; 1889 1890 /* 1891 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 1892 * associated with a packet input on a divert socket. This 1893 * will allow to distinguish traffic and its direction when 1894 * it originates from a divert socket. 1895 */ 1896 u_int divinput_flags = 0; 1897 1898 /* 1899 * oif | args->oif If NULL, ipfw_chk has been called on the 1900 * inbound path (ether_input, bdg_forward, ip_input). 1901 * If non-NULL, ipfw_chk has been called on the outbound path 1902 * (ether_output, ip_output). 1903 */ 1904 struct ifnet *oif = args->oif; 1905 1906 struct ip_fw *f = NULL; /* matching rule */ 1907 int retval = 0; 1908 1909 /* 1910 * hlen The length of the IPv4 header. 1911 * hlen >0 means we have an IPv4 packet. 1912 */ 1913 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1914 1915 /* 1916 * offset The offset of a fragment. offset != 0 means that 1917 * we have a fragment at this offset of an IPv4 packet. 1918 * offset == 0 means that (if this is an IPv4 packet) 1919 * this is the first or only fragment. 1920 */ 1921 u_short offset = 0; 1922 1923 /* 1924 * Local copies of addresses. They are only valid if we have 1925 * an IP packet. 1926 * 1927 * proto The protocol. Set to 0 for non-ip packets, 1928 * or to the protocol read from the packet otherwise. 1929 * proto != 0 means that we have an IPv4 packet. 1930 * 1931 * src_port, dst_port port numbers, in HOST format. Only 1932 * valid for TCP and UDP packets. 1933 * 1934 * src_ip, dst_ip ip addresses, in NETWORK format. 1935 * Only valid for IPv4 packets. 1936 */ 1937 u_int8_t proto; 1938 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1939 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1940 u_int16_t ip_len=0; 1941 int pktlen; 1942 1943 /* 1944 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1945 * MATCH_NONE when checked and not matched (q = NULL), 1946 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1947 */ 1948 int dyn_dir = MATCH_UNKNOWN; 1949 ipfw_dyn_rule *q = NULL; 1950 struct ip_fw_chain *chain = &layer3_chain; 1951 struct m_tag *mtag; 1952 1953 /* 1954 * We store in ulp a pointer to the upper layer protocol header. 1955 * In the ipv4 case this is easy to determine from the header, 1956 * but for ipv6 we might have some additional headers in the middle. 1957 * ulp is NULL if not found. 1958 */ 1959 void *ulp = NULL; /* upper layer protocol pointer. */ 1960 /* XXX ipv6 variables */ 1961 int is_ipv6 = 0; 1962 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 1963 /* end of ipv6 variables */ 1964 1965 if (m->m_flags & M_SKIP_FIREWALL) 1966 return (IP_FW_PASS); /* accept */ 1967 1968 pktlen = m->m_pkthdr.len; 1969 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1970 1971/* 1972 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 1973 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 1974 * pointer might become stale after other pullups (but we never use it 1975 * this way). 1976 */ 1977#define PULLUP_TO(len, p, T) \ 1978do { \ 1979 int x = (len) + sizeof(T); \ 1980 if ((m)->m_len < x) { \ 1981 args->m = m = m_pullup(m, x); \ 1982 if (m == NULL) \ 1983 goto pullup_failed; \ 1984 } \ 1985 p = (mtod(m, char *) + (len)); \ 1986} while (0) 1987 1988 /* Identify IP packets and fill up variables. */ 1989 if (pktlen >= sizeof(struct ip6_hdr) && 1990 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 1991 mtod(m, struct ip *)->ip_v == 6) { 1992 is_ipv6 = 1; 1993 args->f_id.addr_type = 6; 1994 hlen = sizeof(struct ip6_hdr); 1995 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 1996 1997 /* Search extension headers to find upper layer protocols */ 1998 while (ulp == NULL) { 1999 switch (proto) { 2000 case IPPROTO_ICMPV6: 2001 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2002 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2003 break; 2004 2005 case IPPROTO_TCP: 2006 PULLUP_TO(hlen, ulp, struct tcphdr); 2007 dst_port = TCP(ulp)->th_dport; 2008 src_port = TCP(ulp)->th_sport; 2009 args->f_id.flags = TCP(ulp)->th_flags; 2010 break; 2011 2012 case IPPROTO_UDP: 2013 PULLUP_TO(hlen, ulp, struct udphdr); 2014 dst_port = UDP(ulp)->uh_dport; 2015 src_port = UDP(ulp)->uh_sport; 2016 break; 2017 2018 case IPPROTO_HOPOPTS: 2019 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2020 ext_hd |= EXT_HOPOPTS; 2021 hlen += sizeof(struct ip6_hbh); 2022 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2023 ulp = NULL; 2024 break; 2025 2026 case IPPROTO_ROUTING: 2027 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2028 ext_hd |= EXT_ROUTING; 2029 hlen += sizeof(struct ip6_rthdr); 2030 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2031 ulp = NULL; 2032 break; 2033 2034 case IPPROTO_FRAGMENT: 2035 PULLUP_TO(hlen, ulp, struct ip6_frag); 2036 ext_hd |= EXT_FRAGMENT; 2037 hlen += sizeof (struct ip6_frag); 2038 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2039 offset = 1; 2040 ulp = NULL; /* XXX is it correct ? */ 2041 break; 2042 2043 case IPPROTO_AH: 2044 case IPPROTO_NONE: 2045 case IPPROTO_ESP: 2046 PULLUP_TO(hlen, ulp, struct ip6_ext); 2047 if (proto == IPPROTO_AH) 2048 ext_hd |= EXT_AH; 2049 else if (proto == IPPROTO_ESP) 2050 ext_hd |= EXT_ESP; 2051 hlen += ((struct ip6_ext *)ulp)->ip6e_len + 2052 sizeof (struct ip6_ext); 2053 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2054 ulp = NULL; 2055 break; 2056 2057 case IPPROTO_OSPFIGP: 2058 /* XXX OSPF header check? */ 2059 PULLUP_TO(hlen, ulp, struct ip6_ext); 2060 break; 2061 2062 default: 2063 printf( "IPFW2: IPV6 - Unknown Extension Header (%d)\n", 2064 proto); 2065 return 0; /* deny */ 2066 break; 2067 } /*switch */ 2068 } 2069 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2070 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2071 args->f_id.src_ip = 0; 2072 args->f_id.dst_ip = 0; 2073 args->f_id.flow_id6 = ntohs(mtod(m, struct ip6_hdr *)->ip6_flow); 2074 /* hlen != 0 is used to detect ipv4 packets, so clear it now */ 2075 hlen = 0; 2076 } else if (pktlen >= sizeof(struct ip) && 2077 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2078 mtod(m, struct ip *)->ip_v == 4) { 2079 ip = mtod(m, struct ip *); 2080 hlen = ip->ip_hl << 2; 2081 args->f_id.addr_type = 4; 2082 2083 /* 2084 * Collect parameters into local variables for faster matching. 2085 */ 2086 proto = ip->ip_p; 2087 src_ip = ip->ip_src; 2088 dst_ip = ip->ip_dst; 2089 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2090 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2091 ip_len = ntohs(ip->ip_len); 2092 } else { 2093 offset = ip->ip_off & IP_OFFMASK; 2094 ip_len = ip->ip_len; 2095 } 2096 pktlen = ip_len < pktlen ? ip_len : pktlen; 2097 2098 if (offset == 0) { 2099 switch (proto) { 2100 case IPPROTO_TCP: 2101 PULLUP_TO(hlen, ulp, struct tcphdr); 2102 dst_port = TCP(ulp)->th_dport; 2103 src_port = TCP(ulp)->th_sport; 2104 args->f_id.flags = TCP(ulp)->th_flags; 2105 break; 2106 2107 case IPPROTO_UDP: 2108 PULLUP_TO(hlen, ulp, struct udphdr); 2109 dst_port = UDP(ulp)->uh_dport; 2110 src_port = UDP(ulp)->uh_sport; 2111 break; 2112 2113 case IPPROTO_ICMP: 2114 PULLUP_TO(hlen, ulp, struct icmphdr); 2115 args->f_id.flags = ICMP(ulp)->icmp_type; 2116 break; 2117 2118 default: 2119 break; 2120 } 2121 } 2122 2123 args->f_id.src_ip = ntohl(src_ip.s_addr); 2124 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2125 } 2126#undef PULLUP_TO 2127 if (proto) { /* we may have port numbers, store them */ 2128 args->f_id.proto = proto; 2129 args->f_id.src_port = src_port = ntohs(src_port); 2130 args->f_id.dst_port = dst_port = ntohs(dst_port); 2131 } 2132 2133 IPFW_RLOCK(chain); 2134 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2135 if (args->rule) { 2136 /* 2137 * Packet has already been tagged. Look for the next rule 2138 * to restart processing. 2139 * 2140 * If fw_one_pass != 0 then just accept it. 2141 * XXX should not happen here, but optimized out in 2142 * the caller. 2143 */ 2144 if (fw_one_pass) { 2145 IPFW_RUNLOCK(chain); 2146 return (IP_FW_PASS); 2147 } 2148 2149 f = args->rule->next_rule; 2150 if (f == NULL) 2151 f = lookup_next_rule(args->rule); 2152 } else { 2153 /* 2154 * Find the starting rule. It can be either the first 2155 * one, or the one after divert_rule if asked so. 2156 */ 2157 int skipto = mtag ? divert_cookie(mtag) : 0; 2158 2159 f = chain->rules; 2160 if (args->eh == NULL && skipto != 0) { 2161 if (skipto >= IPFW_DEFAULT_RULE) { 2162 IPFW_RUNLOCK(chain); 2163 return (IP_FW_DENY); /* invalid */ 2164 } 2165 while (f && f->rulenum <= skipto) 2166 f = f->next; 2167 if (f == NULL) { /* drop packet */ 2168 IPFW_RUNLOCK(chain); 2169 return (IP_FW_DENY); 2170 } 2171 } 2172 } 2173 /* reset divert rule to avoid confusion later */ 2174 if (mtag) { 2175 divinput_flags = divert_info(mtag) & 2176 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2177 m_tag_delete(m, mtag); 2178 } 2179 2180 /* 2181 * Now scan the rules, and parse microinstructions for each rule. 2182 */ 2183 for (; f; f = f->next) { 2184 int l, cmdlen; 2185 ipfw_insn *cmd; 2186 int skip_or; /* skip rest of OR block */ 2187 2188again: 2189 if (set_disable & (1 << f->set) ) 2190 continue; 2191 2192 skip_or = 0; 2193 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2194 l -= cmdlen, cmd += cmdlen) { 2195 int match; 2196 2197 /* 2198 * check_body is a jump target used when we find a 2199 * CHECK_STATE, and need to jump to the body of 2200 * the target rule. 2201 */ 2202 2203check_body: 2204 cmdlen = F_LEN(cmd); 2205 /* 2206 * An OR block (insn_1 || .. || insn_n) has the 2207 * F_OR bit set in all but the last instruction. 2208 * The first match will set "skip_or", and cause 2209 * the following instructions to be skipped until 2210 * past the one with the F_OR bit clear. 2211 */ 2212 if (skip_or) { /* skip this instruction */ 2213 if ((cmd->len & F_OR) == 0) 2214 skip_or = 0; /* next one is good */ 2215 continue; 2216 } 2217 match = 0; /* set to 1 if we succeed */ 2218 2219 switch (cmd->opcode) { 2220 /* 2221 * The first set of opcodes compares the packet's 2222 * fields with some pattern, setting 'match' if a 2223 * match is found. At the end of the loop there is 2224 * logic to deal with F_NOT and F_OR flags associated 2225 * with the opcode. 2226 */ 2227 case O_NOP: 2228 match = 1; 2229 break; 2230 2231 case O_FORWARD_MAC: 2232 printf("ipfw: opcode %d unimplemented\n", 2233 cmd->opcode); 2234 break; 2235 2236 case O_GID: 2237 case O_UID: 2238 case O_JAIL: 2239 /* 2240 * We only check offset == 0 && proto != 0, 2241 * as this ensures that we have a 2242 * packet with the ports info. 2243 */ 2244 if (offset!=0) 2245 break; 2246 if (is_ipv6) /* XXX to be fixed later */ 2247 break; 2248 if (proto == IPPROTO_TCP || 2249 proto == IPPROTO_UDP) 2250 match = check_uidgid( 2251 (ipfw_insn_u32 *)cmd, 2252 proto, oif, 2253 dst_ip, dst_port, 2254 src_ip, src_port, &fw_ugid_cache, 2255 &ugid_lookup, args->inp); 2256 break; 2257 2258 case O_RECV: 2259 match = iface_match(m->m_pkthdr.rcvif, 2260 (ipfw_insn_if *)cmd); 2261 break; 2262 2263 case O_XMIT: 2264 match = iface_match(oif, (ipfw_insn_if *)cmd); 2265 break; 2266 2267 case O_VIA: 2268 match = iface_match(oif ? oif : 2269 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2270 break; 2271 2272 case O_MACADDR2: 2273 if (args->eh != NULL) { /* have MAC header */ 2274 u_int32_t *want = (u_int32_t *) 2275 ((ipfw_insn_mac *)cmd)->addr; 2276 u_int32_t *mask = (u_int32_t *) 2277 ((ipfw_insn_mac *)cmd)->mask; 2278 u_int32_t *hdr = (u_int32_t *)args->eh; 2279 2280 match = 2281 ( want[0] == (hdr[0] & mask[0]) && 2282 want[1] == (hdr[1] & mask[1]) && 2283 want[2] == (hdr[2] & mask[2]) ); 2284 } 2285 break; 2286 2287 case O_MAC_TYPE: 2288 if (args->eh != NULL) { 2289 u_int16_t t = 2290 ntohs(args->eh->ether_type); 2291 u_int16_t *p = 2292 ((ipfw_insn_u16 *)cmd)->ports; 2293 int i; 2294 2295 for (i = cmdlen - 1; !match && i>0; 2296 i--, p += 2) 2297 match = (t>=p[0] && t<=p[1]); 2298 } 2299 break; 2300 2301 case O_FRAG: 2302 match = (offset != 0); 2303 break; 2304 2305 case O_IN: /* "out" is "not in" */ 2306 match = (oif == NULL); 2307 break; 2308 2309 case O_LAYER2: 2310 match = (args->eh != NULL); 2311 break; 2312 2313 case O_DIVERTED: 2314 match = (cmd->arg1 & 1 && divinput_flags & 2315 IP_FW_DIVERT_LOOPBACK_FLAG) || 2316 (cmd->arg1 & 2 && divinput_flags & 2317 IP_FW_DIVERT_OUTPUT_FLAG); 2318 break; 2319 2320 case O_PROTO: 2321 /* 2322 * We do not allow an arg of 0 so the 2323 * check of "proto" only suffices. 2324 */ 2325 match = (proto == cmd->arg1); 2326 break; 2327 2328 case O_IP_SRC: 2329 match = (hlen > 0 && 2330 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2331 src_ip.s_addr); 2332 break; 2333 2334 case O_IP_SRC_LOOKUP: 2335 case O_IP_DST_LOOKUP: 2336 if (hlen > 0) { 2337 uint32_t a = 2338 (cmd->opcode == O_IP_DST_LOOKUP) ? 2339 dst_ip.s_addr : src_ip.s_addr; 2340 uint32_t v; 2341 2342 match = lookup_table(cmd->arg1, a, &v); 2343 if (!match) 2344 break; 2345 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2346 match = 2347 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2348 } 2349 break; 2350 2351 case O_IP_SRC_MASK: 2352 case O_IP_DST_MASK: 2353 if (hlen > 0) { 2354 uint32_t a = 2355 (cmd->opcode == O_IP_DST_MASK) ? 2356 dst_ip.s_addr : src_ip.s_addr; 2357 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2358 int i = cmdlen-1; 2359 2360 for (; !match && i>0; i-= 2, p+= 2) 2361 match = (p[0] == (a & p[1])); 2362 } 2363 break; 2364 2365 case O_IP_SRC_ME: 2366 if (hlen > 0) { 2367 struct ifnet *tif; 2368 2369 INADDR_TO_IFP(src_ip, tif); 2370 match = (tif != NULL); 2371 } 2372 break; 2373 2374 case O_IP_DST_SET: 2375 case O_IP_SRC_SET: 2376 if (hlen > 0) { 2377 u_int32_t *d = (u_int32_t *)(cmd+1); 2378 u_int32_t addr = 2379 cmd->opcode == O_IP_DST_SET ? 2380 args->f_id.dst_ip : 2381 args->f_id.src_ip; 2382 2383 if (addr < d[0]) 2384 break; 2385 addr -= d[0]; /* subtract base */ 2386 match = (addr < cmd->arg1) && 2387 ( d[ 1 + (addr>>5)] & 2388 (1<<(addr & 0x1f)) ); 2389 } 2390 break; 2391 2392 case O_IP_DST: 2393 match = (hlen > 0 && 2394 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2395 dst_ip.s_addr); 2396 break; 2397 2398 case O_IP_DST_ME: 2399 if (hlen > 0) { 2400 struct ifnet *tif; 2401 2402 INADDR_TO_IFP(dst_ip, tif); 2403 match = (tif != NULL); 2404 } 2405 break; 2406 2407 case O_IP_SRCPORT: 2408 case O_IP_DSTPORT: 2409 /* 2410 * offset == 0 && proto != 0 is enough 2411 * to guarantee that we have a 2412 * packet with port info. 2413 */ 2414 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2415 && offset == 0) { 2416 u_int16_t x = 2417 (cmd->opcode == O_IP_SRCPORT) ? 2418 src_port : dst_port ; 2419 u_int16_t *p = 2420 ((ipfw_insn_u16 *)cmd)->ports; 2421 int i; 2422 2423 for (i = cmdlen - 1; !match && i>0; 2424 i--, p += 2) 2425 match = (x>=p[0] && x<=p[1]); 2426 } 2427 break; 2428 2429 case O_ICMPTYPE: 2430 match = (offset == 0 && proto==IPPROTO_ICMP && 2431 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2432 break; 2433 2434#ifdef INET6 2435 case O_ICMP6TYPE: 2436 match = is_ipv6 && offset == 0 && 2437 proto==IPPROTO_ICMPV6 && 2438 icmp6type_match( 2439 ICMP6(ulp)->icmp6_type, 2440 (ipfw_insn_u32 *)cmd); 2441 break; 2442#endif /* INET6 */ 2443 2444 case O_IPOPT: 2445 match = (hlen > 0 && 2446 ipopts_match(mtod(m, struct ip *), cmd) ); 2447 break; 2448 2449 case O_IPVER: 2450 match = (hlen > 0 && 2451 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2452 break; 2453 2454 case O_IPID: 2455 case O_IPLEN: 2456 case O_IPTTL: 2457 if (hlen > 0) { /* only for IP packets */ 2458 uint16_t x; 2459 uint16_t *p; 2460 int i; 2461 2462 if (cmd->opcode == O_IPLEN) 2463 x = ip_len; 2464 else if (cmd->opcode == O_IPTTL) 2465 x = mtod(m, struct ip *)->ip_ttl; 2466 else /* must be IPID */ 2467 x = ntohs(mtod(m, struct ip *)->ip_id); 2468 if (cmdlen == 1) { 2469 match = (cmd->arg1 == x); 2470 break; 2471 } 2472 /* otherwise we have ranges */ 2473 p = ((ipfw_insn_u16 *)cmd)->ports; 2474 i = cmdlen - 1; 2475 for (; !match && i>0; i--, p += 2) 2476 match = (x >= p[0] && x <= p[1]); 2477 } 2478 break; 2479 2480 case O_IPPRECEDENCE: 2481 match = (hlen > 0 && 2482 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2483 break; 2484 2485 case O_IPTOS: 2486 match = (hlen > 0 && 2487 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2488 break; 2489 2490 case O_TCPDATALEN: 2491 if (proto == IPPROTO_TCP && offset == 0) { 2492 struct tcphdr *tcp; 2493 uint16_t x; 2494 uint16_t *p; 2495 int i; 2496 2497 tcp = TCP(ulp); 2498 x = ip_len - 2499 ((ip->ip_hl + tcp->th_off) << 2); 2500 if (cmdlen == 1) { 2501 match = (cmd->arg1 == x); 2502 break; 2503 } 2504 /* otherwise we have ranges */ 2505 p = ((ipfw_insn_u16 *)cmd)->ports; 2506 i = cmdlen - 1; 2507 for (; !match && i>0; i--, p += 2) 2508 match = (x >= p[0] && x <= p[1]); 2509 } 2510 break; 2511 2512 case O_TCPFLAGS: 2513 match = (proto == IPPROTO_TCP && offset == 0 && 2514 flags_match(cmd, TCP(ulp)->th_flags)); 2515 break; 2516 2517 case O_TCPOPTS: 2518 match = (proto == IPPROTO_TCP && offset == 0 && 2519 tcpopts_match(TCP(ulp), cmd)); 2520 break; 2521 2522 case O_TCPSEQ: 2523 match = (proto == IPPROTO_TCP && offset == 0 && 2524 ((ipfw_insn_u32 *)cmd)->d[0] == 2525 TCP(ulp)->th_seq); 2526 break; 2527 2528 case O_TCPACK: 2529 match = (proto == IPPROTO_TCP && offset == 0 && 2530 ((ipfw_insn_u32 *)cmd)->d[0] == 2531 TCP(ulp)->th_ack); 2532 break; 2533 2534 case O_TCPWIN: 2535 match = (proto == IPPROTO_TCP && offset == 0 && 2536 cmd->arg1 == TCP(ulp)->th_win); 2537 break; 2538 2539 case O_ESTAB: 2540 /* reject packets which have SYN only */ 2541 /* XXX should i also check for TH_ACK ? */ 2542 match = (proto == IPPROTO_TCP && offset == 0 && 2543 (TCP(ulp)->th_flags & 2544 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2545 break; 2546 2547 case O_ALTQ: { 2548 struct altq_tag *at; 2549 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2550 2551 match = 1; 2552 mtag = m_tag_get(PACKET_TAG_PF_QID, 2553 sizeof(struct altq_tag), 2554 M_NOWAIT); 2555 if (mtag == NULL) { 2556 /* 2557 * Let the packet fall back to the 2558 * default ALTQ. 2559 */ 2560 break; 2561 } 2562 at = (struct altq_tag *)(mtag+1); 2563 at->qid = altq->qid; 2564 if (hlen != 0) 2565 at->af = AF_INET; 2566 else 2567 at->af = AF_LINK; 2568 at->hdr = ip; 2569 m_tag_prepend(m, mtag); 2570 break; 2571 } 2572 2573 case O_LOG: 2574 if (fw_verbose) 2575 ipfw_log(f, hlen, args->eh, m, oif); 2576 match = 1; 2577 break; 2578 2579 case O_PROB: 2580 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2581 break; 2582 2583 case O_VERREVPATH: 2584 /* Outgoing packets automatically pass/match */ 2585 /* XXX BED: verify_path was verify_rev_path in the diff... */ 2586 match = ((oif != NULL) || 2587 (m->m_pkthdr.rcvif == NULL) || 2588 ( 2589#ifdef INET6 2590 is_ipv6 ? 2591 verify_rev_path6(&(args->f_id.src_ip6), 2592 m->m_pkthdr.rcvif) : 2593#endif 2594 verify_path(src_ip, m->m_pkthdr.rcvif))); 2595 break; 2596 2597 case O_VERSRCREACH: 2598 /* Outgoing packets automatically pass/match */ 2599 match = (hlen > 0 && ((oif != NULL) || 2600 verify_path(src_ip, NULL))); 2601 break; 2602 2603 case O_ANTISPOOF: 2604 /* Outgoing packets automatically pass/match */ 2605 if (oif == NULL && hlen > 0 && 2606 in_localaddr(src_ip)) 2607 match = verify_path(src_ip, 2608 m->m_pkthdr.rcvif); 2609 else 2610 match = 1; 2611 break; 2612 2613 case O_IPSEC: 2614#ifdef FAST_IPSEC 2615 match = (m_tag_find(m, 2616 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2617#endif 2618#ifdef IPSEC 2619 match = (ipsec_getnhist(m) != 0); 2620#endif 2621 /* otherwise no match */ 2622 break; 2623 2624 case O_IP6_SRC: 2625 match = is_ipv6 && 2626 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2627 &((ipfw_insn_ip6 *)cmd)->addr6); 2628 break; 2629 2630 case O_IP6_DST: 2631 match = is_ipv6 && 2632 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2633 &((ipfw_insn_ip6 *)cmd)->addr6); 2634 break; 2635 case O_IP6_SRC_MASK: 2636 if (is_ipv6) { 2637 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2638 struct in6_addr p = args->f_id.src_ip6; 2639 2640 APPLY_MASK(&p, &te->mask6); 2641 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2642 } 2643 break; 2644 2645 case O_IP6_DST_MASK: 2646 if (is_ipv6) { 2647 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2648 struct in6_addr p = args->f_id.dst_ip6; 2649 2650 APPLY_MASK(&p, &te->mask6); 2651 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2652 } 2653 break; 2654 2655#ifdef INET6 2656 case O_IP6_SRC_ME: 2657 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2658 break; 2659 2660 case O_IP6_DST_ME: 2661 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2662 break; 2663 2664 case O_FLOW6ID: 2665 match = is_ipv6 && 2666 flow6id_match(args->f_id.flow_id6, 2667 (ipfw_insn_u32 *) cmd); 2668 break; 2669 2670 case O_EXT_HDR: 2671 match = is_ipv6 && 2672 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2673 break; 2674 2675 case O_IP6: 2676 match = is_ipv6; 2677 break; 2678#endif 2679 2680 /* 2681 * The second set of opcodes represents 'actions', 2682 * i.e. the terminal part of a rule once the packet 2683 * matches all previous patterns. 2684 * Typically there is only one action for each rule, 2685 * and the opcode is stored at the end of the rule 2686 * (but there are exceptions -- see below). 2687 * 2688 * In general, here we set retval and terminate the 2689 * outer loop (would be a 'break 3' in some language, 2690 * but we need to do a 'goto done'). 2691 * 2692 * Exceptions: 2693 * O_COUNT and O_SKIPTO actions: 2694 * instead of terminating, we jump to the next rule 2695 * ('goto next_rule', equivalent to a 'break 2'), 2696 * or to the SKIPTO target ('goto again' after 2697 * having set f, cmd and l), respectively. 2698 * 2699 * O_LOG and O_ALTQ action parameters: 2700 * perform some action and set match = 1; 2701 * 2702 * O_LIMIT and O_KEEP_STATE: these opcodes are 2703 * not real 'actions', and are stored right 2704 * before the 'action' part of the rule. 2705 * These opcodes try to install an entry in the 2706 * state tables; if successful, we continue with 2707 * the next opcode (match=1; break;), otherwise 2708 * the packet * must be dropped 2709 * ('goto done' after setting retval); 2710 * 2711 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2712 * cause a lookup of the state table, and a jump 2713 * to the 'action' part of the parent rule 2714 * ('goto check_body') if an entry is found, or 2715 * (CHECK_STATE only) a jump to the next rule if 2716 * the entry is not found ('goto next_rule'). 2717 * The result of the lookup is cached to make 2718 * further instances of these opcodes are 2719 * effectively NOPs. 2720 */ 2721 case O_LIMIT: 2722 case O_KEEP_STATE: 2723 if (install_state(f, 2724 (ipfw_insn_limit *)cmd, args)) { 2725 retval = IP_FW_DENY; 2726 goto done; /* error/limit violation */ 2727 } 2728 match = 1; 2729 break; 2730 2731 case O_PROBE_STATE: 2732 case O_CHECK_STATE: 2733 /* 2734 * dynamic rules are checked at the first 2735 * keep-state or check-state occurrence, 2736 * with the result being stored in dyn_dir. 2737 * The compiler introduces a PROBE_STATE 2738 * instruction for us when we have a 2739 * KEEP_STATE (because PROBE_STATE needs 2740 * to be run first). 2741 */ 2742 if (dyn_dir == MATCH_UNKNOWN && 2743 (q = lookup_dyn_rule(&args->f_id, 2744 &dyn_dir, proto == IPPROTO_TCP ? 2745 TCP(ulp) : NULL)) 2746 != NULL) { 2747 /* 2748 * Found dynamic entry, update stats 2749 * and jump to the 'action' part of 2750 * the parent rule. 2751 */ 2752 q->pcnt++; 2753 q->bcnt += pktlen; 2754 f = q->rule; 2755 cmd = ACTION_PTR(f); 2756 l = f->cmd_len - f->act_ofs; 2757 IPFW_DYN_UNLOCK(); 2758 goto check_body; 2759 } 2760 /* 2761 * Dynamic entry not found. If CHECK_STATE, 2762 * skip to next rule, if PROBE_STATE just 2763 * ignore and continue with next opcode. 2764 */ 2765 if (cmd->opcode == O_CHECK_STATE) 2766 goto next_rule; 2767 match = 1; 2768 break; 2769 2770 case O_ACCEPT: 2771 retval = 0; /* accept */ 2772 goto done; 2773 2774 case O_PIPE: 2775 case O_QUEUE: 2776 args->rule = f; /* report matching rule */ 2777 args->cookie = cmd->arg1; 2778 retval = IP_FW_DUMMYNET; 2779 goto done; 2780 2781 case O_DIVERT: 2782 case O_TEE: { 2783 struct divert_tag *dt; 2784 2785 if (args->eh) /* not on layer 2 */ 2786 break; 2787 mtag = m_tag_get(PACKET_TAG_DIVERT, 2788 sizeof(struct divert_tag), 2789 M_NOWAIT); 2790 if (mtag == NULL) { 2791 /* XXX statistic */ 2792 /* drop packet */ 2793 IPFW_RUNLOCK(chain); 2794 return (IP_FW_DENY); 2795 } 2796 dt = (struct divert_tag *)(mtag+1); 2797 dt->cookie = f->rulenum; 2798 dt->info = cmd->arg1; 2799 m_tag_prepend(m, mtag); 2800 retval = (cmd->opcode == O_DIVERT) ? 2801 IP_FW_DIVERT : IP_FW_TEE; 2802 goto done; 2803 } 2804 2805 case O_COUNT: 2806 case O_SKIPTO: 2807 f->pcnt++; /* update stats */ 2808 f->bcnt += pktlen; 2809 f->timestamp = time_second; 2810 if (cmd->opcode == O_COUNT) 2811 goto next_rule; 2812 /* handle skipto */ 2813 if (f->next_rule == NULL) 2814 lookup_next_rule(f); 2815 f = f->next_rule; 2816 goto again; 2817 2818 case O_REJECT: 2819 /* 2820 * Drop the packet and send a reject notice 2821 * if the packet is not ICMP (or is an ICMP 2822 * query), and it is not multicast/broadcast. 2823 */ 2824 if (hlen > 0 && 2825 (proto != IPPROTO_ICMP || 2826 is_icmp_query(ICMP(ulp))) && 2827 !(m->m_flags & (M_BCAST|M_MCAST)) && 2828 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2829 send_reject(args, cmd->arg1, 2830 offset,ip_len); 2831 m = args->m; 2832 } 2833 /* FALLTHROUGH */ 2834 case O_DENY: 2835 retval = IP_FW_DENY; 2836 goto done; 2837 2838 case O_FORWARD_IP: 2839 if (args->eh) /* not valid on layer2 pkts */ 2840 break; 2841 if (!q || dyn_dir == MATCH_FORWARD) 2842 args->next_hop = 2843 &((ipfw_insn_sa *)cmd)->sa; 2844 retval = IP_FW_PASS; 2845 goto done; 2846 2847 case O_NETGRAPH: 2848 case O_NGTEE: 2849 args->rule = f; /* report matching rule */ 2850 args->cookie = cmd->arg1; 2851 retval = (cmd->opcode == O_NETGRAPH) ? 2852 IP_FW_NETGRAPH : IP_FW_NGTEE; 2853 goto done; 2854 2855 default: 2856 panic("-- unknown opcode %d\n", cmd->opcode); 2857 } /* end of switch() on opcodes */ 2858 2859 if (cmd->len & F_NOT) 2860 match = !match; 2861 2862 if (match) { 2863 if (cmd->len & F_OR) 2864 skip_or = 1; 2865 } else { 2866 if (!(cmd->len & F_OR)) /* not an OR block, */ 2867 break; /* try next rule */ 2868 } 2869 2870 } /* end of inner for, scan opcodes */ 2871 2872next_rule:; /* try next rule */ 2873 2874 } /* end of outer for, scan rules */ 2875 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2876 IPFW_RUNLOCK(chain); 2877 return (IP_FW_DENY); 2878 2879done: 2880 /* Update statistics */ 2881 f->pcnt++; 2882 f->bcnt += pktlen; 2883 f->timestamp = time_second; 2884 IPFW_RUNLOCK(chain); 2885 return (retval); 2886 2887pullup_failed: 2888 if (fw_verbose) 2889 printf("ipfw: pullup failed\n"); 2890 return (IP_FW_DENY); 2891} 2892 2893/* 2894 * When a rule is added/deleted, clear the next_rule pointers in all rules. 2895 * These will be reconstructed on the fly as packets are matched. 2896 */ 2897static void 2898flush_rule_ptrs(struct ip_fw_chain *chain) 2899{ 2900 struct ip_fw *rule; 2901 2902 IPFW_WLOCK_ASSERT(chain); 2903 2904 for (rule = chain->rules; rule; rule = rule->next) 2905 rule->next_rule = NULL; 2906} 2907 2908/* 2909 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 2910 * pipe/queue, or to all of them (match == NULL). 2911 */ 2912void 2913flush_pipe_ptrs(struct dn_flow_set *match) 2914{ 2915 struct ip_fw *rule; 2916 2917 IPFW_WLOCK(&layer3_chain); 2918 for (rule = layer3_chain.rules; rule; rule = rule->next) { 2919 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 2920 2921 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 2922 continue; 2923 /* 2924 * XXX Use bcmp/bzero to handle pipe_ptr to overcome 2925 * possible alignment problems on 64-bit architectures. 2926 * This code is seldom used so we do not worry too 2927 * much about efficiency. 2928 */ 2929 if (match == NULL || 2930 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) ) 2931 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr)); 2932 } 2933 IPFW_WUNLOCK(&layer3_chain); 2934} 2935 2936/* 2937 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 2938 * possibly create a rule number and add the rule to the list. 2939 * Update the rule_number in the input struct so the caller knows it as well. 2940 */ 2941static int 2942add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 2943{ 2944 struct ip_fw *rule, *f, *prev; 2945 int l = RULESIZE(input_rule); 2946 2947 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 2948 return (EINVAL); 2949 2950 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 2951 if (rule == NULL) 2952 return (ENOSPC); 2953 2954 bcopy(input_rule, rule, l); 2955 2956 rule->next = NULL; 2957 rule->next_rule = NULL; 2958 2959 rule->pcnt = 0; 2960 rule->bcnt = 0; 2961 rule->timestamp = 0; 2962 2963 IPFW_WLOCK(chain); 2964 2965 if (chain->rules == NULL) { /* default rule */ 2966 chain->rules = rule; 2967 goto done; 2968 } 2969 2970 /* 2971 * If rulenum is 0, find highest numbered rule before the 2972 * default rule, and add autoinc_step 2973 */ 2974 if (autoinc_step < 1) 2975 autoinc_step = 1; 2976 else if (autoinc_step > 1000) 2977 autoinc_step = 1000; 2978 if (rule->rulenum == 0) { 2979 /* 2980 * locate the highest numbered rule before default 2981 */ 2982 for (f = chain->rules; f; f = f->next) { 2983 if (f->rulenum == IPFW_DEFAULT_RULE) 2984 break; 2985 rule->rulenum = f->rulenum; 2986 } 2987 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2988 rule->rulenum += autoinc_step; 2989 input_rule->rulenum = rule->rulenum; 2990 } 2991 2992 /* 2993 * Now insert the new rule in the right place in the sorted list. 2994 */ 2995 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 2996 if (f->rulenum > rule->rulenum) { /* found the location */ 2997 if (prev) { 2998 rule->next = f; 2999 prev->next = rule; 3000 } else { /* head insert */ 3001 rule->next = chain->rules; 3002 chain->rules = rule; 3003 } 3004 break; 3005 } 3006 } 3007 flush_rule_ptrs(chain); 3008done: 3009 static_count++; 3010 static_len += l; 3011 IPFW_WUNLOCK(chain); 3012 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3013 rule->rulenum, static_count);) 3014 return (0); 3015} 3016 3017/** 3018 * Remove a static rule (including derived * dynamic rules) 3019 * and place it on the ``reap list'' for later reclamation. 3020 * The caller is in charge of clearing rule pointers to avoid 3021 * dangling pointers. 3022 * @return a pointer to the next entry. 3023 * Arguments are not checked, so they better be correct. 3024 */ 3025static struct ip_fw * 3026remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3027{ 3028 struct ip_fw *n; 3029 int l = RULESIZE(rule); 3030 3031 IPFW_WLOCK_ASSERT(chain); 3032 3033 n = rule->next; 3034 IPFW_DYN_LOCK(); 3035 remove_dyn_rule(rule, NULL /* force removal */); 3036 IPFW_DYN_UNLOCK(); 3037 if (prev == NULL) 3038 chain->rules = n; 3039 else 3040 prev->next = n; 3041 static_count--; 3042 static_len -= l; 3043 3044 rule->next = chain->reap; 3045 chain->reap = rule; 3046 3047 return n; 3048} 3049 3050/** 3051 * Reclaim storage associated with a list of rules. This is 3052 * typically the list created using remove_rule. 3053 */ 3054static void 3055reap_rules(struct ip_fw *head) 3056{ 3057 struct ip_fw *rule; 3058 3059 while ((rule = head) != NULL) { 3060 head = head->next; 3061 if (DUMMYNET_LOADED) 3062 ip_dn_ruledel_ptr(rule); 3063 free(rule, M_IPFW); 3064 } 3065} 3066 3067/* 3068 * Remove all rules from a chain (except rules in set RESVD_SET 3069 * unless kill_default = 1). The caller is responsible for 3070 * reclaiming storage for the rules left in chain->reap. 3071 */ 3072static void 3073free_chain(struct ip_fw_chain *chain, int kill_default) 3074{ 3075 struct ip_fw *prev, *rule; 3076 3077 IPFW_WLOCK_ASSERT(chain); 3078 3079 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3080 for (prev = NULL, rule = chain->rules; rule ; ) 3081 if (kill_default || rule->set != RESVD_SET) 3082 rule = remove_rule(chain, rule, prev); 3083 else { 3084 prev = rule; 3085 rule = rule->next; 3086 } 3087} 3088 3089/** 3090 * Remove all rules with given number, and also do set manipulation. 3091 * Assumes chain != NULL && *chain != NULL. 3092 * 3093 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3094 * the next 8 bits are the new set, the top 8 bits are the command: 3095 * 3096 * 0 delete rules with given number 3097 * 1 delete rules with given set number 3098 * 2 move rules with given number to new set 3099 * 3 move rules with given set number to new set 3100 * 4 swap sets with given numbers 3101 */ 3102static int 3103del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3104{ 3105 struct ip_fw *prev = NULL, *rule; 3106 u_int16_t rulenum; /* rule or old_set */ 3107 u_int8_t cmd, new_set; 3108 3109 rulenum = arg & 0xffff; 3110 cmd = (arg >> 24) & 0xff; 3111 new_set = (arg >> 16) & 0xff; 3112 3113 if (cmd > 4) 3114 return EINVAL; 3115 if (new_set > RESVD_SET) 3116 return EINVAL; 3117 if (cmd == 0 || cmd == 2) { 3118 if (rulenum >= IPFW_DEFAULT_RULE) 3119 return EINVAL; 3120 } else { 3121 if (rulenum > RESVD_SET) /* old_set */ 3122 return EINVAL; 3123 } 3124 3125 IPFW_WLOCK(chain); 3126 rule = chain->rules; 3127 chain->reap = NULL; 3128 switch (cmd) { 3129 case 0: /* delete rules with given number */ 3130 /* 3131 * locate first rule to delete 3132 */ 3133 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3134 ; 3135 if (rule->rulenum != rulenum) { 3136 IPFW_WUNLOCK(chain); 3137 return EINVAL; 3138 } 3139 3140 /* 3141 * flush pointers outside the loop, then delete all matching 3142 * rules. prev remains the same throughout the cycle. 3143 */ 3144 flush_rule_ptrs(chain); 3145 while (rule->rulenum == rulenum) 3146 rule = remove_rule(chain, rule, prev); 3147 break; 3148 3149 case 1: /* delete all rules with given set number */ 3150 flush_rule_ptrs(chain); 3151 rule = chain->rules; 3152 while (rule->rulenum < IPFW_DEFAULT_RULE) 3153 if (rule->set == rulenum) 3154 rule = remove_rule(chain, rule, prev); 3155 else { 3156 prev = rule; 3157 rule = rule->next; 3158 } 3159 break; 3160 3161 case 2: /* move rules with given number to new set */ 3162 rule = chain->rules; 3163 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3164 if (rule->rulenum == rulenum) 3165 rule->set = new_set; 3166 break; 3167 3168 case 3: /* move rules with given set number to new set */ 3169 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3170 if (rule->set == rulenum) 3171 rule->set = new_set; 3172 break; 3173 3174 case 4: /* swap two sets */ 3175 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3176 if (rule->set == rulenum) 3177 rule->set = new_set; 3178 else if (rule->set == new_set) 3179 rule->set = rulenum; 3180 break; 3181 } 3182 /* 3183 * Look for rules to reclaim. We grab the list before 3184 * releasing the lock then reclaim them w/o the lock to 3185 * avoid a LOR with dummynet. 3186 */ 3187 rule = chain->reap; 3188 chain->reap = NULL; 3189 IPFW_WUNLOCK(chain); 3190 if (rule) 3191 reap_rules(rule); 3192 return 0; 3193} 3194 3195/* 3196 * Clear counters for a specific rule. 3197 * The enclosing "table" is assumed locked. 3198 */ 3199static void 3200clear_counters(struct ip_fw *rule, int log_only) 3201{ 3202 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3203 3204 if (log_only == 0) { 3205 rule->bcnt = rule->pcnt = 0; 3206 rule->timestamp = 0; 3207 } 3208 if (l->o.opcode == O_LOG) 3209 l->log_left = l->max_log; 3210} 3211 3212/** 3213 * Reset some or all counters on firewall rules. 3214 * @arg frwl is null to clear all entries, or contains a specific 3215 * rule number. 3216 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3217 */ 3218static int 3219zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3220{ 3221 struct ip_fw *rule; 3222 char *msg; 3223 3224 IPFW_WLOCK(chain); 3225 if (rulenum == 0) { 3226 norule_counter = 0; 3227 for (rule = chain->rules; rule; rule = rule->next) 3228 clear_counters(rule, log_only); 3229 msg = log_only ? "ipfw: All logging counts reset.\n" : 3230 "ipfw: Accounting cleared.\n"; 3231 } else { 3232 int cleared = 0; 3233 /* 3234 * We can have multiple rules with the same number, so we 3235 * need to clear them all. 3236 */ 3237 for (rule = chain->rules; rule; rule = rule->next) 3238 if (rule->rulenum == rulenum) { 3239 while (rule && rule->rulenum == rulenum) { 3240 clear_counters(rule, log_only); 3241 rule = rule->next; 3242 } 3243 cleared = 1; 3244 break; 3245 } 3246 if (!cleared) { /* we did not find any matching rules */ 3247 IPFW_WUNLOCK(chain); 3248 return (EINVAL); 3249 } 3250 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3251 "ipfw: Entry %d cleared.\n"; 3252 } 3253 IPFW_WUNLOCK(chain); 3254 3255 if (fw_verbose) 3256 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3257 return (0); 3258} 3259 3260/* 3261 * Check validity of the structure before insert. 3262 * Fortunately rules are simple, so this mostly need to check rule sizes. 3263 */ 3264static int 3265check_ipfw_struct(struct ip_fw *rule, int size) 3266{ 3267 int l, cmdlen = 0; 3268 int have_action=0; 3269 ipfw_insn *cmd; 3270 3271 if (size < sizeof(*rule)) { 3272 printf("ipfw: rule too short\n"); 3273 return (EINVAL); 3274 } 3275 /* first, check for valid size */ 3276 l = RULESIZE(rule); 3277 if (l != size) { 3278 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3279 return (EINVAL); 3280 } 3281 if (rule->act_ofs >= rule->cmd_len) { 3282 printf("ipfw: bogus action offset (%u > %u)\n", 3283 rule->act_ofs, rule->cmd_len - 1); 3284 return (EINVAL); 3285 } 3286 /* 3287 * Now go for the individual checks. Very simple ones, basically only 3288 * instruction sizes. 3289 */ 3290 for (l = rule->cmd_len, cmd = rule->cmd ; 3291 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3292 cmdlen = F_LEN(cmd); 3293 if (cmdlen > l) { 3294 printf("ipfw: opcode %d size truncated\n", 3295 cmd->opcode); 3296 return EINVAL; 3297 } 3298 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3299 switch (cmd->opcode) { 3300 case O_PROBE_STATE: 3301 case O_KEEP_STATE: 3302 case O_PROTO: 3303 case O_IP_SRC_ME: 3304 case O_IP_DST_ME: 3305 case O_LAYER2: 3306 case O_IN: 3307 case O_FRAG: 3308 case O_DIVERTED: 3309 case O_IPOPT: 3310 case O_IPTOS: 3311 case O_IPPRECEDENCE: 3312 case O_IPVER: 3313 case O_TCPWIN: 3314 case O_TCPFLAGS: 3315 case O_TCPOPTS: 3316 case O_ESTAB: 3317 case O_VERREVPATH: 3318 case O_VERSRCREACH: 3319 case O_ANTISPOOF: 3320 case O_IPSEC: 3321 case O_IP6_SRC_ME: 3322 case O_IP6_DST_ME: 3323 case O_EXT_HDR: 3324 case O_IP6: 3325 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3326 goto bad_size; 3327 break; 3328 3329 case O_UID: 3330 case O_GID: 3331 case O_JAIL: 3332 case O_IP_SRC: 3333 case O_IP_DST: 3334 case O_TCPSEQ: 3335 case O_TCPACK: 3336 case O_PROB: 3337 case O_ICMPTYPE: 3338 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3339 goto bad_size; 3340 break; 3341 3342 case O_LIMIT: 3343 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3344 goto bad_size; 3345 break; 3346 3347 case O_LOG: 3348 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3349 goto bad_size; 3350 3351 ((ipfw_insn_log *)cmd)->log_left = 3352 ((ipfw_insn_log *)cmd)->max_log; 3353 3354 break; 3355 3356 case O_IP_SRC_MASK: 3357 case O_IP_DST_MASK: 3358 /* only odd command lengths */ 3359 if ( !(cmdlen & 1) || cmdlen > 31) 3360 goto bad_size; 3361 break; 3362 3363 case O_IP_SRC_SET: 3364 case O_IP_DST_SET: 3365 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3366 printf("ipfw: invalid set size %d\n", 3367 cmd->arg1); 3368 return EINVAL; 3369 } 3370 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3371 (cmd->arg1+31)/32 ) 3372 goto bad_size; 3373 break; 3374 3375 case O_IP_SRC_LOOKUP: 3376 case O_IP_DST_LOOKUP: 3377 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3378 printf("ipfw: invalid table number %d\n", 3379 cmd->arg1); 3380 return (EINVAL); 3381 } 3382 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3383 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3384 goto bad_size; 3385 break; 3386 3387 case O_MACADDR2: 3388 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3389 goto bad_size; 3390 break; 3391 3392 case O_NOP: 3393 case O_IPID: 3394 case O_IPTTL: 3395 case O_IPLEN: 3396 case O_TCPDATALEN: 3397 if (cmdlen < 1 || cmdlen > 31) 3398 goto bad_size; 3399 break; 3400 3401 case O_MAC_TYPE: 3402 case O_IP_SRCPORT: 3403 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3404 if (cmdlen < 2 || cmdlen > 31) 3405 goto bad_size; 3406 break; 3407 3408 case O_RECV: 3409 case O_XMIT: 3410 case O_VIA: 3411 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3412 goto bad_size; 3413 break; 3414 3415 case O_ALTQ: 3416 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3417 goto bad_size; 3418 break; 3419 3420 case O_PIPE: 3421 case O_QUEUE: 3422 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 3423 goto bad_size; 3424 goto check_action; 3425 3426 case O_FORWARD_IP: 3427#ifdef IPFIREWALL_FORWARD 3428 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3429 goto bad_size; 3430 goto check_action; 3431#else 3432 return EINVAL; 3433#endif 3434 3435 case O_DIVERT: 3436 case O_TEE: 3437 if (ip_divert_ptr == NULL) 3438 return EINVAL; 3439 else 3440 goto check_size; 3441 case O_NETGRAPH: 3442 case O_NGTEE: 3443 if (!NG_IPFW_LOADED) 3444 return EINVAL; 3445 else 3446 goto check_size; 3447 case O_FORWARD_MAC: /* XXX not implemented yet */ 3448 case O_CHECK_STATE: 3449 case O_COUNT: 3450 case O_ACCEPT: 3451 case O_DENY: 3452 case O_REJECT: 3453 case O_SKIPTO: 3454check_size: 3455 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3456 goto bad_size; 3457check_action: 3458 if (have_action) { 3459 printf("ipfw: opcode %d, multiple actions" 3460 " not allowed\n", 3461 cmd->opcode); 3462 return EINVAL; 3463 } 3464 have_action = 1; 3465 if (l != cmdlen) { 3466 printf("ipfw: opcode %d, action must be" 3467 " last opcode\n", 3468 cmd->opcode); 3469 return EINVAL; 3470 } 3471 break; 3472 case O_IP6_SRC: 3473 case O_IP6_DST: 3474 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3475 F_INSN_SIZE(ipfw_insn)) 3476 goto bad_size; 3477 break; 3478 3479 case O_FLOW6ID: 3480 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3481 ((ipfw_insn_u32 *)cmd)->o.arg1) 3482 goto bad_size; 3483 break; 3484 3485 case O_IP6_SRC_MASK: 3486 case O_IP6_DST_MASK: 3487 if ( !(cmdlen & 1) || cmdlen > 127) 3488 goto bad_size; 3489 break; 3490 case O_ICMP6TYPE: 3491 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3492 goto bad_size; 3493 break; 3494 3495 default: 3496 printf("ipfw: opcode %d, unknown opcode\n", 3497 cmd->opcode); 3498 return EINVAL; 3499 } 3500 } 3501 if (have_action == 0) { 3502 printf("ipfw: missing action\n"); 3503 return EINVAL; 3504 } 3505 return 0; 3506 3507bad_size: 3508 printf("ipfw: opcode %d size %d wrong\n", 3509 cmd->opcode, cmdlen); 3510 return EINVAL; 3511} 3512 3513/* 3514 * Copy the static and dynamic rules to the supplied buffer 3515 * and return the amount of space actually used. 3516 */ 3517static size_t 3518ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3519{ 3520 char *bp = buf; 3521 char *ep = bp + space; 3522 struct ip_fw *rule; 3523 int i; 3524 3525 /* XXX this can take a long time and locking will block packet flow */ 3526 IPFW_RLOCK(chain); 3527 for (rule = chain->rules; rule ; rule = rule->next) { 3528 /* 3529 * Verify the entry fits in the buffer in case the 3530 * rules changed between calculating buffer space and 3531 * now. This would be better done using a generation 3532 * number but should suffice for now. 3533 */ 3534 i = RULESIZE(rule); 3535 if (bp + i <= ep) { 3536 bcopy(rule, bp, i); 3537 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3538 sizeof(set_disable)); 3539 bp += i; 3540 } 3541 } 3542 IPFW_RUNLOCK(chain); 3543 if (ipfw_dyn_v) { 3544 ipfw_dyn_rule *p, *last = NULL; 3545 3546 IPFW_DYN_LOCK(); 3547 for (i = 0 ; i < curr_dyn_buckets; i++) 3548 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3549 if (bp + sizeof *p <= ep) { 3550 ipfw_dyn_rule *dst = 3551 (ipfw_dyn_rule *)bp; 3552 bcopy(p, dst, sizeof *p); 3553 bcopy(&(p->rule->rulenum), &(dst->rule), 3554 sizeof(p->rule->rulenum)); 3555 /* 3556 * store a non-null value in "next". 3557 * The userland code will interpret a 3558 * NULL here as a marker 3559 * for the last dynamic rule. 3560 */ 3561 bcopy(&dst, &dst->next, sizeof(dst)); 3562 last = dst; 3563 dst->expire = 3564 TIME_LEQ(dst->expire, time_second) ? 3565 0 : dst->expire - time_second ; 3566 bp += sizeof(ipfw_dyn_rule); 3567 } 3568 } 3569 IPFW_DYN_UNLOCK(); 3570 if (last != NULL) /* mark last dynamic rule */ 3571 bzero(&last->next, sizeof(last)); 3572 } 3573 return (bp - (char *)buf); 3574} 3575 3576 3577/** 3578 * {set|get}sockopt parser. 3579 */ 3580static int 3581ipfw_ctl(struct sockopt *sopt) 3582{ 3583#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3584 int error, rule_num; 3585 size_t size; 3586 struct ip_fw *buf, *rule; 3587 u_int32_t rulenum[2]; 3588 3589 error = suser(sopt->sopt_td); 3590 if (error) 3591 return (error); 3592 3593 /* 3594 * Disallow modifications in really-really secure mode, but still allow 3595 * the logging counters to be reset. 3596 */ 3597 if (sopt->sopt_name == IP_FW_ADD || 3598 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3599#if __FreeBSD_version >= 500034 3600 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3601 if (error) 3602 return (error); 3603#else /* FreeBSD 4.x */ 3604 if (securelevel >= 3) 3605 return (EPERM); 3606#endif 3607 } 3608 3609 error = 0; 3610 3611 switch (sopt->sopt_name) { 3612 case IP_FW_GET: 3613 /* 3614 * pass up a copy of the current rules. Static rules 3615 * come first (the last of which has number IPFW_DEFAULT_RULE), 3616 * followed by a possibly empty list of dynamic rule. 3617 * The last dynamic rule has NULL in the "next" field. 3618 * 3619 * Note that the calculated size is used to bound the 3620 * amount of data returned to the user. The rule set may 3621 * change between calculating the size and returning the 3622 * data in which case we'll just return what fits. 3623 */ 3624 size = static_len; /* size of static rules */ 3625 if (ipfw_dyn_v) /* add size of dyn.rules */ 3626 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3627 3628 /* 3629 * XXX todo: if the user passes a short length just to know 3630 * how much room is needed, do not bother filling up the 3631 * buffer, just jump to the sooptcopyout. 3632 */ 3633 buf = malloc(size, M_TEMP, M_WAITOK); 3634 error = sooptcopyout(sopt, buf, 3635 ipfw_getrules(&layer3_chain, buf, size)); 3636 free(buf, M_TEMP); 3637 break; 3638 3639 case IP_FW_FLUSH: 3640 /* 3641 * Normally we cannot release the lock on each iteration. 3642 * We could do it here only because we start from the head all 3643 * the times so there is no risk of missing some entries. 3644 * On the other hand, the risk is that we end up with 3645 * a very inconsistent ruleset, so better keep the lock 3646 * around the whole cycle. 3647 * 3648 * XXX this code can be improved by resetting the head of 3649 * the list to point to the default rule, and then freeing 3650 * the old list without the need for a lock. 3651 */ 3652 3653 IPFW_WLOCK(&layer3_chain); 3654 layer3_chain.reap = NULL; 3655 free_chain(&layer3_chain, 0 /* keep default rule */); 3656 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3657 IPFW_WUNLOCK(&layer3_chain); 3658 if (layer3_chain.reap != NULL) 3659 reap_rules(rule); 3660 break; 3661 3662 case IP_FW_ADD: 3663 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3664 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3665 sizeof(struct ip_fw) ); 3666 if (error == 0) 3667 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3668 if (error == 0) { 3669 error = add_rule(&layer3_chain, rule); 3670 size = RULESIZE(rule); 3671 if (!error && sopt->sopt_dir == SOPT_GET) 3672 error = sooptcopyout(sopt, rule, size); 3673 } 3674 free(rule, M_TEMP); 3675 break; 3676 3677 case IP_FW_DEL: 3678 /* 3679 * IP_FW_DEL is used for deleting single rules or sets, 3680 * and (ab)used to atomically manipulate sets. Argument size 3681 * is used to distinguish between the two: 3682 * sizeof(u_int32_t) 3683 * delete single rule or set of rules, 3684 * or reassign rules (or sets) to a different set. 3685 * 2*sizeof(u_int32_t) 3686 * atomic disable/enable sets. 3687 * first u_int32_t contains sets to be disabled, 3688 * second u_int32_t contains sets to be enabled. 3689 */ 3690 error = sooptcopyin(sopt, rulenum, 3691 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3692 if (error) 3693 break; 3694 size = sopt->sopt_valsize; 3695 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3696 error = del_entry(&layer3_chain, rulenum[0]); 3697 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3698 set_disable = 3699 (set_disable | rulenum[0]) & ~rulenum[1] & 3700 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3701 else 3702 error = EINVAL; 3703 break; 3704 3705 case IP_FW_ZERO: 3706 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3707 rule_num = 0; 3708 if (sopt->sopt_val != 0) { 3709 error = sooptcopyin(sopt, &rule_num, 3710 sizeof(int), sizeof(int)); 3711 if (error) 3712 break; 3713 } 3714 error = zero_entry(&layer3_chain, rule_num, 3715 sopt->sopt_name == IP_FW_RESETLOG); 3716 break; 3717 3718 case IP_FW_TABLE_ADD: 3719 { 3720 ipfw_table_entry ent; 3721 3722 error = sooptcopyin(sopt, &ent, 3723 sizeof(ent), sizeof(ent)); 3724 if (error) 3725 break; 3726 error = add_table_entry(ent.tbl, ent.addr, 3727 ent.masklen, ent.value); 3728 } 3729 break; 3730 3731 case IP_FW_TABLE_DEL: 3732 { 3733 ipfw_table_entry ent; 3734 3735 error = sooptcopyin(sopt, &ent, 3736 sizeof(ent), sizeof(ent)); 3737 if (error) 3738 break; 3739 error = del_table_entry(ent.tbl, ent.addr, ent.masklen); 3740 } 3741 break; 3742 3743 case IP_FW_TABLE_FLUSH: 3744 { 3745 u_int16_t tbl; 3746 3747 error = sooptcopyin(sopt, &tbl, 3748 sizeof(tbl), sizeof(tbl)); 3749 if (error) 3750 break; 3751 error = flush_table(tbl); 3752 } 3753 break; 3754 3755 case IP_FW_TABLE_GETSIZE: 3756 { 3757 u_int32_t tbl, cnt; 3758 3759 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 3760 sizeof(tbl)))) 3761 break; 3762 if ((error = count_table(tbl, &cnt))) 3763 break; 3764 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 3765 } 3766 break; 3767 3768 case IP_FW_TABLE_LIST: 3769 { 3770 ipfw_table *tbl; 3771 3772 if (sopt->sopt_valsize < sizeof(*tbl)) { 3773 error = EINVAL; 3774 break; 3775 } 3776 size = sopt->sopt_valsize; 3777 tbl = malloc(size, M_TEMP, M_WAITOK); 3778 if (tbl == NULL) { 3779 error = ENOMEM; 3780 break; 3781 } 3782 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 3783 if (error) { 3784 free(tbl, M_TEMP); 3785 break; 3786 } 3787 tbl->size = (size - sizeof(*tbl)) / 3788 sizeof(ipfw_table_entry); 3789 error = dump_table(tbl); 3790 if (error) { 3791 free(tbl, M_TEMP); 3792 break; 3793 } 3794 error = sooptcopyout(sopt, tbl, size); 3795 free(tbl, M_TEMP); 3796 } 3797 break; 3798 3799 default: 3800 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 3801 error = EINVAL; 3802 } 3803 3804 return (error); 3805#undef RULE_MAXSIZE 3806} 3807 3808/** 3809 * dummynet needs a reference to the default rule, because rules can be 3810 * deleted while packets hold a reference to them. When this happens, 3811 * dummynet changes the reference to the default rule (it could well be a 3812 * NULL pointer, but this way we do not need to check for the special 3813 * case, plus here he have info on the default behaviour). 3814 */ 3815struct ip_fw *ip_fw_default_rule; 3816 3817/* 3818 * This procedure is only used to handle keepalives. It is invoked 3819 * every dyn_keepalive_period 3820 */ 3821static void 3822ipfw_tick(void * __unused unused) 3823{ 3824 int i; 3825 ipfw_dyn_rule *q; 3826 3827 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 3828 goto done; 3829 3830 IPFW_DYN_LOCK(); 3831 for (i = 0 ; i < curr_dyn_buckets ; i++) { 3832 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 3833 if (q->dyn_type == O_LIMIT_PARENT) 3834 continue; 3835 if (q->id.proto != IPPROTO_TCP) 3836 continue; 3837 if ( (q->state & BOTH_SYN) != BOTH_SYN) 3838 continue; 3839 if (TIME_LEQ( time_second+dyn_keepalive_interval, 3840 q->expire)) 3841 continue; /* too early */ 3842 if (TIME_LEQ(q->expire, time_second)) 3843 continue; /* too late, rule expired */ 3844 3845 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 3846 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 3847 } 3848 } 3849 IPFW_DYN_UNLOCK(); 3850done: 3851 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 3852} 3853 3854int 3855ipfw_init(void) 3856{ 3857 struct ip_fw default_rule; 3858 int error; 3859 3860 layer3_chain.rules = NULL; 3861 layer3_chain.want_write = 0; 3862 layer3_chain.busy_count = 0; 3863 cv_init(&layer3_chain.cv, "Condition variable for IPFW rw locks"); 3864 IPFW_LOCK_INIT(&layer3_chain); 3865 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 3866 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 3867 UMA_ALIGN_PTR, 0); 3868 IPFW_DYN_LOCK_INIT(); 3869 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 3870 3871 bzero(&default_rule, sizeof default_rule); 3872 3873 default_rule.act_ofs = 0; 3874 default_rule.rulenum = IPFW_DEFAULT_RULE; 3875 default_rule.cmd_len = 1; 3876 default_rule.set = RESVD_SET; 3877 3878 default_rule.cmd[0].len = 1; 3879 default_rule.cmd[0].opcode = 3880#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 3881 1 ? O_ACCEPT : 3882#endif 3883 O_DENY; 3884 3885 error = add_rule(&layer3_chain, &default_rule); 3886 if (error != 0) { 3887 printf("ipfw2: error %u initializing default rule " 3888 "(support disabled)\n", error); 3889 IPFW_DYN_LOCK_DESTROY(); 3890 IPFW_LOCK_DESTROY(&layer3_chain); 3891 return (error); 3892 } 3893 3894 ip_fw_default_rule = layer3_chain.rules; 3895 printf("ipfw2 (+ipv6) initialized, divert %s, " 3896 "rule-based forwarding " 3897#ifdef IPFIREWALL_FORWARD 3898 "enabled, " 3899#else 3900 "disabled, " 3901#endif 3902 "default to %s, logging ", 3903#ifdef IPDIVERT 3904 "enabled", 3905#else 3906 "loadable", 3907#endif 3908 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 3909 3910#ifdef IPFIREWALL_VERBOSE 3911 fw_verbose = 1; 3912#endif 3913#ifdef IPFIREWALL_VERBOSE_LIMIT 3914 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3915#endif 3916 if (fw_verbose == 0) 3917 printf("disabled\n"); 3918 else if (verbose_limit == 0) 3919 printf("unlimited\n"); 3920 else 3921 printf("limited to %d packets/entry by default\n", 3922 verbose_limit); 3923 3924 init_tables(); 3925 ip_fw_ctl_ptr = ipfw_ctl; 3926 ip_fw_chk_ptr = ipfw_chk; 3927 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 3928 3929 return (0); 3930} 3931 3932void 3933ipfw_destroy(void) 3934{ 3935 struct ip_fw *reap; 3936 3937 ip_fw_chk_ptr = NULL; 3938 ip_fw_ctl_ptr = NULL; 3939 callout_drain(&ipfw_timeout); 3940 IPFW_WLOCK(&layer3_chain); 3941 layer3_chain.reap = NULL; 3942 free_chain(&layer3_chain, 1 /* kill default rule */); 3943 reap = layer3_chain.reap, layer3_chain.reap = NULL; 3944 IPFW_WUNLOCK(&layer3_chain); 3945 if (reap != NULL) 3946 reap_rules(reap); 3947 flush_tables(); 3948 IPFW_DYN_LOCK_DESTROY(); 3949 uma_zdestroy(ipfw_dyn_rule_zone); 3950 IPFW_LOCK_DESTROY(&layer3_chain); 3951 printf("IP firewall unloaded\n"); 3952} 3953