ip_fw2.c revision 132510
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 132510 2004-07-21 19:55:14Z andre $ 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_ipdivert.h" 39#include "opt_inet.h" 40#include "opt_ipsec.h" 41#ifndef INET 42#error IPFIREWALL requires INET. 43#endif /* INET */ 44#endif 45 46#define IPFW2 1 47#if IPFW2 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/malloc.h> 51#include <sys/mbuf.h> 52#include <sys/kernel.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#ifdef IPSEC 81#include <netinet6/ipsec.h> 82#endif 83 84#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 85 86#include <machine/in_cksum.h> /* XXX for in_cksum */ 87 88/* 89 * set_disable contains one bit per set value (0..31). 90 * If the bit is set, all rules with the corresponding set 91 * are disabled. Set RESVD_SET(31) is reserved for the default rule 92 * and rules that are not deleted by the flush command, 93 * and CANNOT be disabled. 94 * Rules in set RESVD_SET can only be deleted explicitly. 95 */ 96static u_int32_t set_disable; 97 98static int fw_verbose; 99static int verbose_limit; 100 101static struct callout ipfw_timeout; 102#define IPFW_DEFAULT_RULE 65535 103 104/* 105 * Data structure to cache our ucred related 106 * information. This structure only gets used if 107 * the user specified UID/GID based constraints in 108 * a firewall rule. 109 */ 110struct ip_fw_ugid { 111 gid_t fw_groups[NGROUPS]; 112 int fw_ngroups; 113 uid_t fw_uid; 114}; 115 116struct ip_fw_chain { 117 struct ip_fw *rules; /* list of rules */ 118 struct ip_fw *reap; /* list of rules to reap */ 119 struct mtx mtx; /* lock guarding rule list */ 120}; 121#define IPFW_LOCK_INIT(_chain) \ 122 mtx_init(&(_chain)->mtx, "IPFW static rules", NULL, \ 123 MTX_DEF | MTX_RECURSE) 124#define IPFW_LOCK_DESTROY(_chain) mtx_destroy(&(_chain)->mtx) 125#define IPFW_LOCK(_chain) mtx_lock(&(_chain)->mtx) 126#define IPFW_UNLOCK(_chain) mtx_unlock(&(_chain)->mtx) 127#define IPFW_LOCK_ASSERT(_chain) do { \ 128 mtx_assert(&(_chain)->mtx, MA_OWNED); \ 129 NET_ASSERT_GIANT(); \ 130} while (0) 131 132/* 133 * list of rules for layer 3 134 */ 135static struct ip_fw_chain layer3_chain; 136 137MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 138MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 139 140struct table_entry { 141 struct radix_node rn[2]; 142 struct sockaddr_in addr, mask; 143 u_int32_t value; 144}; 145 146#define IPFW_TABLES_MAX 128 147static struct { 148 struct radix_node_head *rnh; 149 int modified; 150} ipfw_tables[IPFW_TABLES_MAX]; 151 152static int fw_debug = 1; 153static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 154 155#ifdef SYSCTL_NODE 156SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 157SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 158 CTLFLAG_RW | CTLFLAG_SECURE3, 159 &fw_enable, 0, "Enable ipfw"); 160SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 161 &autoinc_step, 0, "Rule number autincrement step"); 162SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 163 CTLFLAG_RW | CTLFLAG_SECURE3, 164 &fw_one_pass, 0, 165 "Only do a single pass through ipfw when using dummynet(4)"); 166SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 167 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 168SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 169 CTLFLAG_RW | CTLFLAG_SECURE3, 170 &fw_verbose, 0, "Log matches to ipfw rules"); 171SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 172 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 173 174/* 175 * Description of dynamic rules. 176 * 177 * Dynamic rules are stored in lists accessed through a hash table 178 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 179 * be modified through the sysctl variable dyn_buckets which is 180 * updated when the table becomes empty. 181 * 182 * XXX currently there is only one list, ipfw_dyn. 183 * 184 * When a packet is received, its address fields are first masked 185 * with the mask defined for the rule, then hashed, then matched 186 * against the entries in the corresponding list. 187 * Dynamic rules can be used for different purposes: 188 * + stateful rules; 189 * + enforcing limits on the number of sessions; 190 * + in-kernel NAT (not implemented yet) 191 * 192 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 193 * measured in seconds and depending on the flags. 194 * 195 * The total number of dynamic rules is stored in dyn_count. 196 * The max number of dynamic rules is dyn_max. When we reach 197 * the maximum number of rules we do not create anymore. This is 198 * done to avoid consuming too much memory, but also too much 199 * time when searching on each packet (ideally, we should try instead 200 * to put a limit on the length of the list on each bucket...). 201 * 202 * Each dynamic rule holds a pointer to the parent ipfw rule so 203 * we know what action to perform. Dynamic rules are removed when 204 * the parent rule is deleted. XXX we should make them survive. 205 * 206 * There are some limitations with dynamic rules -- we do not 207 * obey the 'randomized match', and we do not do multiple 208 * passes through the firewall. XXX check the latter!!! 209 */ 210static ipfw_dyn_rule **ipfw_dyn_v = NULL; 211static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 212static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 213 214static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 215#define IPFW_DYN_LOCK_INIT() \ 216 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 217#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 218#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 219#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 220#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 221 222/* 223 * Timeouts for various events in handing dynamic rules. 224 */ 225static u_int32_t dyn_ack_lifetime = 300; 226static u_int32_t dyn_syn_lifetime = 20; 227static u_int32_t dyn_fin_lifetime = 1; 228static u_int32_t dyn_rst_lifetime = 1; 229static u_int32_t dyn_udp_lifetime = 10; 230static u_int32_t dyn_short_lifetime = 5; 231 232/* 233 * Keepalives are sent if dyn_keepalive is set. They are sent every 234 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 235 * seconds of lifetime of a rule. 236 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 237 * than dyn_keepalive_period. 238 */ 239 240static u_int32_t dyn_keepalive_interval = 20; 241static u_int32_t dyn_keepalive_period = 5; 242static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 243 244static u_int32_t static_count; /* # of static rules */ 245static u_int32_t static_len; /* size in bytes of static rules */ 246static u_int32_t dyn_count; /* # of dynamic rules */ 247static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 248 249SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 250 &dyn_buckets, 0, "Number of dyn. buckets"); 251SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 252 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 253SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 254 &dyn_count, 0, "Number of dyn. rules"); 255SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 256 &dyn_max, 0, "Max number of dyn. rules"); 257SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 258 &static_count, 0, "Number of static rules"); 259SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 260 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 261SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 262 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 263SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 264 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 265SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 266 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 267SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 268 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 269SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 270 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 271SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 272 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 273 274#endif /* SYSCTL_NODE */ 275 276 277static ip_fw_chk_t ipfw_chk; 278 279ip_dn_ruledel_t *ip_dn_ruledel_ptr = NULL; /* hook into dummynet */ 280 281/* 282 * This macro maps an ip pointer into a layer3 header pointer of type T 283 */ 284#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 285 286static __inline int 287icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 288{ 289 int type = L3HDR(struct icmp,ip)->icmp_type; 290 291 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 292} 293 294#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 295 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 296 297static int 298is_icmp_query(struct ip *ip) 299{ 300 int type = L3HDR(struct icmp, ip)->icmp_type; 301 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 302} 303#undef TT 304 305/* 306 * The following checks use two arrays of 8 or 16 bits to store the 307 * bits that we want set or clear, respectively. They are in the 308 * low and high half of cmd->arg1 or cmd->d[0]. 309 * 310 * We scan options and store the bits we find set. We succeed if 311 * 312 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 313 * 314 * The code is sometimes optimized not to store additional variables. 315 */ 316 317static int 318flags_match(ipfw_insn *cmd, u_int8_t bits) 319{ 320 u_char want_clear; 321 bits = ~bits; 322 323 if ( ((cmd->arg1 & 0xff) & bits) != 0) 324 return 0; /* some bits we want set were clear */ 325 want_clear = (cmd->arg1 >> 8) & 0xff; 326 if ( (want_clear & bits) != want_clear) 327 return 0; /* some bits we want clear were set */ 328 return 1; 329} 330 331static int 332ipopts_match(struct ip *ip, ipfw_insn *cmd) 333{ 334 int optlen, bits = 0; 335 u_char *cp = (u_char *)(ip + 1); 336 int x = (ip->ip_hl << 2) - sizeof (struct ip); 337 338 for (; x > 0; x -= optlen, cp += optlen) { 339 int opt = cp[IPOPT_OPTVAL]; 340 341 if (opt == IPOPT_EOL) 342 break; 343 if (opt == IPOPT_NOP) 344 optlen = 1; 345 else { 346 optlen = cp[IPOPT_OLEN]; 347 if (optlen <= 0 || optlen > x) 348 return 0; /* invalid or truncated */ 349 } 350 switch (opt) { 351 352 default: 353 break; 354 355 case IPOPT_LSRR: 356 bits |= IP_FW_IPOPT_LSRR; 357 break; 358 359 case IPOPT_SSRR: 360 bits |= IP_FW_IPOPT_SSRR; 361 break; 362 363 case IPOPT_RR: 364 bits |= IP_FW_IPOPT_RR; 365 break; 366 367 case IPOPT_TS: 368 bits |= IP_FW_IPOPT_TS; 369 break; 370 } 371 } 372 return (flags_match(cmd, bits)); 373} 374 375static int 376tcpopts_match(struct ip *ip, ipfw_insn *cmd) 377{ 378 int optlen, bits = 0; 379 struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 380 u_char *cp = (u_char *)(tcp + 1); 381 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 382 383 for (; x > 0; x -= optlen, cp += optlen) { 384 int opt = cp[0]; 385 if (opt == TCPOPT_EOL) 386 break; 387 if (opt == TCPOPT_NOP) 388 optlen = 1; 389 else { 390 optlen = cp[1]; 391 if (optlen <= 0) 392 break; 393 } 394 395 switch (opt) { 396 397 default: 398 break; 399 400 case TCPOPT_MAXSEG: 401 bits |= IP_FW_TCPOPT_MSS; 402 break; 403 404 case TCPOPT_WINDOW: 405 bits |= IP_FW_TCPOPT_WINDOW; 406 break; 407 408 case TCPOPT_SACK_PERMITTED: 409 case TCPOPT_SACK: 410 bits |= IP_FW_TCPOPT_SACK; 411 break; 412 413 case TCPOPT_TIMESTAMP: 414 bits |= IP_FW_TCPOPT_TS; 415 break; 416 417 case TCPOPT_CC: 418 case TCPOPT_CCNEW: 419 case TCPOPT_CCECHO: 420 bits |= IP_FW_TCPOPT_CC; 421 break; 422 } 423 } 424 return (flags_match(cmd, bits)); 425} 426 427static int 428iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 429{ 430 if (ifp == NULL) /* no iface with this packet, match fails */ 431 return 0; 432 /* Check by name or by IP address */ 433 if (cmd->name[0] != '\0') { /* match by name */ 434 /* Check name */ 435 if (cmd->p.glob) { 436 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 437 return(1); 438 } else { 439 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 440 return(1); 441 } 442 } else { 443 struct ifaddr *ia; 444 445 /* XXX lock? */ 446 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 447 if (ia->ifa_addr == NULL) 448 continue; 449 if (ia->ifa_addr->sa_family != AF_INET) 450 continue; 451 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 452 (ia->ifa_addr))->sin_addr.s_addr) 453 return(1); /* match */ 454 } 455 } 456 return(0); /* no match, fail ... */ 457} 458 459/* 460 * The verify_path function checks if a route to the src exists and 461 * if it is reachable via ifp (when provided). 462 * 463 * The 'verrevpath' option checks that the interface that an IP packet 464 * arrives on is the same interface that traffic destined for the 465 * packet's source address would be routed out of. The 'versrcreach' 466 * option just checks that the source address is reachable via any route 467 * (except default) in the routing table. These two are a measure to block 468 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 469 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 470 * is purposely reminiscent of the Cisco IOS command, 471 * 472 * ip verify unicast reverse-path 473 * ip verify unicast source reachable-via any 474 * 475 * which implements the same functionality. But note that syntax is 476 * misleading. The check may be performed on all IP packets whether unicast, 477 * multicast, or broadcast. 478 */ 479static int 480verify_path(struct in_addr src, struct ifnet *ifp) 481{ 482 struct route ro; 483 struct sockaddr_in *dst; 484 485 bzero(&ro, sizeof(ro)); 486 487 dst = (struct sockaddr_in *)&(ro.ro_dst); 488 dst->sin_family = AF_INET; 489 dst->sin_len = sizeof(*dst); 490 dst->sin_addr = src; 491 rtalloc_ign(&ro, RTF_CLONING); 492 493 if (ro.ro_rt == NULL) 494 return 0; 495 496 /* if ifp is provided, check for equality with rtentry */ 497 if (ifp != NULL && ro.ro_rt->rt_ifp != ifp) { 498 RTFREE(ro.ro_rt); 499 return 0; 500 } 501 502 /* if no ifp provided, check if rtentry is not default route */ 503 if (ifp == NULL && 504 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 505 RTFREE(ro.ro_rt); 506 return 0; 507 } 508 509 /* or if this is a blackhole/reject route */ 510 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 511 RTFREE(ro.ro_rt); 512 return 0; 513 } 514 515 /* found valid route */ 516 RTFREE(ro.ro_rt); 517 return 1; 518} 519 520 521static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 522 523#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 524#define SNP(buf) buf, sizeof(buf) 525 526/* 527 * We enter here when we have a rule with O_LOG. 528 * XXX this function alone takes about 2Kbytes of code! 529 */ 530static void 531ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 532 struct mbuf *m, struct ifnet *oif) 533{ 534 char *action; 535 int limit_reached = 0; 536 char action2[40], proto[48], fragment[28]; 537 538 fragment[0] = '\0'; 539 proto[0] = '\0'; 540 541 if (f == NULL) { /* bogus pkt */ 542 if (verbose_limit != 0 && norule_counter >= verbose_limit) 543 return; 544 norule_counter++; 545 if (norule_counter == verbose_limit) 546 limit_reached = verbose_limit; 547 action = "Refuse"; 548 } else { /* O_LOG is the first action, find the real one */ 549 ipfw_insn *cmd = ACTION_PTR(f); 550 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 551 552 if (l->max_log != 0 && l->log_left == 0) 553 return; 554 l->log_left--; 555 if (l->log_left == 0) 556 limit_reached = l->max_log; 557 cmd += F_LEN(cmd); /* point to first action */ 558 if (cmd->opcode == O_PROB) 559 cmd += F_LEN(cmd); 560 561 action = action2; 562 switch (cmd->opcode) { 563 case O_DENY: 564 action = "Deny"; 565 break; 566 567 case O_REJECT: 568 if (cmd->arg1==ICMP_REJECT_RST) 569 action = "Reset"; 570 else if (cmd->arg1==ICMP_UNREACH_HOST) 571 action = "Reject"; 572 else 573 snprintf(SNPARGS(action2, 0), "Unreach %d", 574 cmd->arg1); 575 break; 576 577 case O_ACCEPT: 578 action = "Accept"; 579 break; 580 case O_COUNT: 581 action = "Count"; 582 break; 583 case O_DIVERT: 584 snprintf(SNPARGS(action2, 0), "Divert %d", 585 cmd->arg1); 586 break; 587 case O_TEE: 588 snprintf(SNPARGS(action2, 0), "Tee %d", 589 cmd->arg1); 590 break; 591 case O_SKIPTO: 592 snprintf(SNPARGS(action2, 0), "SkipTo %d", 593 cmd->arg1); 594 break; 595 case O_PIPE: 596 snprintf(SNPARGS(action2, 0), "Pipe %d", 597 cmd->arg1); 598 break; 599 case O_QUEUE: 600 snprintf(SNPARGS(action2, 0), "Queue %d", 601 cmd->arg1); 602 break; 603 case O_FORWARD_IP: { 604 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 605 int len; 606 607 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 608 inet_ntoa(sa->sa.sin_addr)); 609 if (sa->sa.sin_port) 610 snprintf(SNPARGS(action2, len), ":%d", 611 sa->sa.sin_port); 612 } 613 break; 614 default: 615 action = "UNKNOWN"; 616 break; 617 } 618 } 619 620 if (hlen == 0) { /* non-ip */ 621 snprintf(SNPARGS(proto, 0), "MAC"); 622 } else { 623 struct ip *ip = mtod(m, struct ip *); 624 /* these three are all aliases to the same thing */ 625 struct icmp *const icmp = L3HDR(struct icmp, ip); 626 struct tcphdr *const tcp = (struct tcphdr *)icmp; 627 struct udphdr *const udp = (struct udphdr *)icmp; 628 629 int ip_off, offset, ip_len; 630 631 int len; 632 633 if (eh != NULL) { /* layer 2 packets are as on the wire */ 634 ip_off = ntohs(ip->ip_off); 635 ip_len = ntohs(ip->ip_len); 636 } else { 637 ip_off = ip->ip_off; 638 ip_len = ip->ip_len; 639 } 640 offset = ip_off & IP_OFFMASK; 641 switch (ip->ip_p) { 642 case IPPROTO_TCP: 643 len = snprintf(SNPARGS(proto, 0), "TCP %s", 644 inet_ntoa(ip->ip_src)); 645 if (offset == 0) 646 snprintf(SNPARGS(proto, len), ":%d %s:%d", 647 ntohs(tcp->th_sport), 648 inet_ntoa(ip->ip_dst), 649 ntohs(tcp->th_dport)); 650 else 651 snprintf(SNPARGS(proto, len), " %s", 652 inet_ntoa(ip->ip_dst)); 653 break; 654 655 case IPPROTO_UDP: 656 len = snprintf(SNPARGS(proto, 0), "UDP %s", 657 inet_ntoa(ip->ip_src)); 658 if (offset == 0) 659 snprintf(SNPARGS(proto, len), ":%d %s:%d", 660 ntohs(udp->uh_sport), 661 inet_ntoa(ip->ip_dst), 662 ntohs(udp->uh_dport)); 663 else 664 snprintf(SNPARGS(proto, len), " %s", 665 inet_ntoa(ip->ip_dst)); 666 break; 667 668 case IPPROTO_ICMP: 669 if (offset == 0) 670 len = snprintf(SNPARGS(proto, 0), 671 "ICMP:%u.%u ", 672 icmp->icmp_type, icmp->icmp_code); 673 else 674 len = snprintf(SNPARGS(proto, 0), "ICMP "); 675 len += snprintf(SNPARGS(proto, len), "%s", 676 inet_ntoa(ip->ip_src)); 677 snprintf(SNPARGS(proto, len), " %s", 678 inet_ntoa(ip->ip_dst)); 679 break; 680 681 default: 682 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 683 inet_ntoa(ip->ip_src)); 684 snprintf(SNPARGS(proto, len), " %s", 685 inet_ntoa(ip->ip_dst)); 686 break; 687 } 688 689 if (ip_off & (IP_MF | IP_OFFMASK)) 690 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 691 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 692 offset << 3, 693 (ip_off & IP_MF) ? "+" : ""); 694 } 695 if (oif || m->m_pkthdr.rcvif) 696 log(LOG_SECURITY | LOG_INFO, 697 "ipfw: %d %s %s %s via %s%s\n", 698 f ? f->rulenum : -1, 699 action, proto, oif ? "out" : "in", 700 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 701 fragment); 702 else 703 log(LOG_SECURITY | LOG_INFO, 704 "ipfw: %d %s %s [no if info]%s\n", 705 f ? f->rulenum : -1, 706 action, proto, fragment); 707 if (limit_reached) 708 log(LOG_SECURITY | LOG_NOTICE, 709 "ipfw: limit %d reached on entry %d\n", 710 limit_reached, f ? f->rulenum : -1); 711} 712 713/* 714 * IMPORTANT: the hash function for dynamic rules must be commutative 715 * in source and destination (ip,port), because rules are bidirectional 716 * and we want to find both in the same bucket. 717 */ 718static __inline int 719hash_packet(struct ipfw_flow_id *id) 720{ 721 u_int32_t i; 722 723 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 724 i &= (curr_dyn_buckets - 1); 725 return i; 726} 727 728/** 729 * unlink a dynamic rule from a chain. prev is a pointer to 730 * the previous one, q is a pointer to the rule to delete, 731 * head is a pointer to the head of the queue. 732 * Modifies q and potentially also head. 733 */ 734#define UNLINK_DYN_RULE(prev, head, q) { \ 735 ipfw_dyn_rule *old_q = q; \ 736 \ 737 /* remove a refcount to the parent */ \ 738 if (q->dyn_type == O_LIMIT) \ 739 q->parent->count--; \ 740 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 741 (q->id.src_ip), (q->id.src_port), \ 742 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 743 if (prev != NULL) \ 744 prev->next = q = q->next; \ 745 else \ 746 head = q = q->next; \ 747 dyn_count--; \ 748 free(old_q, M_IPFW); } 749 750#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 751 752/** 753 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 754 * 755 * If keep_me == NULL, rules are deleted even if not expired, 756 * otherwise only expired rules are removed. 757 * 758 * The value of the second parameter is also used to point to identify 759 * a rule we absolutely do not want to remove (e.g. because we are 760 * holding a reference to it -- this is the case with O_LIMIT_PARENT 761 * rules). The pointer is only used for comparison, so any non-null 762 * value will do. 763 */ 764static void 765remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 766{ 767 static u_int32_t last_remove = 0; 768 769#define FORCE (keep_me == NULL) 770 771 ipfw_dyn_rule *prev, *q; 772 int i, pass = 0, max_pass = 0; 773 774 IPFW_DYN_LOCK_ASSERT(); 775 776 if (ipfw_dyn_v == NULL || dyn_count == 0) 777 return; 778 /* do not expire more than once per second, it is useless */ 779 if (!FORCE && last_remove == time_second) 780 return; 781 last_remove = time_second; 782 783 /* 784 * because O_LIMIT refer to parent rules, during the first pass only 785 * remove child and mark any pending LIMIT_PARENT, and remove 786 * them in a second pass. 787 */ 788next_pass: 789 for (i = 0 ; i < curr_dyn_buckets ; i++) { 790 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 791 /* 792 * Logic can become complex here, so we split tests. 793 */ 794 if (q == keep_me) 795 goto next; 796 if (rule != NULL && rule != q->rule) 797 goto next; /* not the one we are looking for */ 798 if (q->dyn_type == O_LIMIT_PARENT) { 799 /* 800 * handle parent in the second pass, 801 * record we need one. 802 */ 803 max_pass = 1; 804 if (pass == 0) 805 goto next; 806 if (FORCE && q->count != 0 ) { 807 /* XXX should not happen! */ 808 printf("ipfw: OUCH! cannot remove rule," 809 " count %d\n", q->count); 810 } 811 } else { 812 if (!FORCE && 813 !TIME_LEQ( q->expire, time_second )) 814 goto next; 815 } 816 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 817 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 818 continue; 819 } 820next: 821 prev=q; 822 q=q->next; 823 } 824 } 825 if (pass++ < max_pass) 826 goto next_pass; 827} 828 829 830/** 831 * lookup a dynamic rule. 832 */ 833static ipfw_dyn_rule * 834lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 835 struct tcphdr *tcp) 836{ 837 /* 838 * stateful ipfw extensions. 839 * Lookup into dynamic session queue 840 */ 841#define MATCH_REVERSE 0 842#define MATCH_FORWARD 1 843#define MATCH_NONE 2 844#define MATCH_UNKNOWN 3 845 int i, dir = MATCH_NONE; 846 ipfw_dyn_rule *prev, *q=NULL; 847 848 IPFW_DYN_LOCK_ASSERT(); 849 850 if (ipfw_dyn_v == NULL) 851 goto done; /* not found */ 852 i = hash_packet( pkt ); 853 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 854 if (q->dyn_type == O_LIMIT_PARENT && q->count) 855 goto next; 856 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 857 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 858 continue; 859 } 860 if (pkt->proto == q->id.proto && 861 q->dyn_type != O_LIMIT_PARENT) { 862 if (pkt->src_ip == q->id.src_ip && 863 pkt->dst_ip == q->id.dst_ip && 864 pkt->src_port == q->id.src_port && 865 pkt->dst_port == q->id.dst_port ) { 866 dir = MATCH_FORWARD; 867 break; 868 } 869 if (pkt->src_ip == q->id.dst_ip && 870 pkt->dst_ip == q->id.src_ip && 871 pkt->src_port == q->id.dst_port && 872 pkt->dst_port == q->id.src_port ) { 873 dir = MATCH_REVERSE; 874 break; 875 } 876 } 877next: 878 prev = q; 879 q = q->next; 880 } 881 if (q == NULL) 882 goto done; /* q = NULL, not found */ 883 884 if ( prev != NULL) { /* found and not in front */ 885 prev->next = q->next; 886 q->next = ipfw_dyn_v[i]; 887 ipfw_dyn_v[i] = q; 888 } 889 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 890 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 891 892#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 893#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 894 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 895 switch (q->state) { 896 case TH_SYN: /* opening */ 897 q->expire = time_second + dyn_syn_lifetime; 898 break; 899 900 case BOTH_SYN: /* move to established */ 901 case BOTH_SYN | TH_FIN : /* one side tries to close */ 902 case BOTH_SYN | (TH_FIN << 8) : 903 if (tcp) { 904#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 905 u_int32_t ack = ntohl(tcp->th_ack); 906 if (dir == MATCH_FORWARD) { 907 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 908 q->ack_fwd = ack; 909 else { /* ignore out-of-sequence */ 910 break; 911 } 912 } else { 913 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 914 q->ack_rev = ack; 915 else { /* ignore out-of-sequence */ 916 break; 917 } 918 } 919 } 920 q->expire = time_second + dyn_ack_lifetime; 921 break; 922 923 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 924 if (dyn_fin_lifetime >= dyn_keepalive_period) 925 dyn_fin_lifetime = dyn_keepalive_period - 1; 926 q->expire = time_second + dyn_fin_lifetime; 927 break; 928 929 default: 930#if 0 931 /* 932 * reset or some invalid combination, but can also 933 * occur if we use keep-state the wrong way. 934 */ 935 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 936 printf("invalid state: 0x%x\n", q->state); 937#endif 938 if (dyn_rst_lifetime >= dyn_keepalive_period) 939 dyn_rst_lifetime = dyn_keepalive_period - 1; 940 q->expire = time_second + dyn_rst_lifetime; 941 break; 942 } 943 } else if (pkt->proto == IPPROTO_UDP) { 944 q->expire = time_second + dyn_udp_lifetime; 945 } else { 946 /* other protocols */ 947 q->expire = time_second + dyn_short_lifetime; 948 } 949done: 950 if (match_direction) 951 *match_direction = dir; 952 return q; 953} 954 955static ipfw_dyn_rule * 956lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 957 struct tcphdr *tcp) 958{ 959 ipfw_dyn_rule *q; 960 961 IPFW_DYN_LOCK(); 962 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 963 if (q == NULL) 964 IPFW_DYN_UNLOCK(); 965 /* NB: return table locked when q is not NULL */ 966 return q; 967} 968 969static void 970realloc_dynamic_table(void) 971{ 972 IPFW_DYN_LOCK_ASSERT(); 973 974 /* 975 * Try reallocation, make sure we have a power of 2 and do 976 * not allow more than 64k entries. In case of overflow, 977 * default to 1024. 978 */ 979 980 if (dyn_buckets > 65536) 981 dyn_buckets = 1024; 982 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 983 dyn_buckets = curr_dyn_buckets; /* reset */ 984 return; 985 } 986 curr_dyn_buckets = dyn_buckets; 987 if (ipfw_dyn_v != NULL) 988 free(ipfw_dyn_v, M_IPFW); 989 for (;;) { 990 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 991 M_IPFW, M_NOWAIT | M_ZERO); 992 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 993 break; 994 curr_dyn_buckets /= 2; 995 } 996} 997 998/** 999 * Install state of type 'type' for a dynamic session. 1000 * The hash table contains two type of rules: 1001 * - regular rules (O_KEEP_STATE) 1002 * - rules for sessions with limited number of sess per user 1003 * (O_LIMIT). When they are created, the parent is 1004 * increased by 1, and decreased on delete. In this case, 1005 * the third parameter is the parent rule and not the chain. 1006 * - "parent" rules for the above (O_LIMIT_PARENT). 1007 */ 1008static ipfw_dyn_rule * 1009add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1010{ 1011 ipfw_dyn_rule *r; 1012 int i; 1013 1014 IPFW_DYN_LOCK_ASSERT(); 1015 1016 if (ipfw_dyn_v == NULL || 1017 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1018 realloc_dynamic_table(); 1019 if (ipfw_dyn_v == NULL) 1020 return NULL; /* failed ! */ 1021 } 1022 i = hash_packet(id); 1023 1024 r = malloc(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO); 1025 if (r == NULL) { 1026 printf ("ipfw: sorry cannot allocate state\n"); 1027 return NULL; 1028 } 1029 1030 /* increase refcount on parent, and set pointer */ 1031 if (dyn_type == O_LIMIT) { 1032 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1033 if ( parent->dyn_type != O_LIMIT_PARENT) 1034 panic("invalid parent"); 1035 parent->count++; 1036 r->parent = parent; 1037 rule = parent->rule; 1038 } 1039 1040 r->id = *id; 1041 r->expire = time_second + dyn_syn_lifetime; 1042 r->rule = rule; 1043 r->dyn_type = dyn_type; 1044 r->pcnt = r->bcnt = 0; 1045 r->count = 0; 1046 1047 r->bucket = i; 1048 r->next = ipfw_dyn_v[i]; 1049 ipfw_dyn_v[i] = r; 1050 dyn_count++; 1051 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1052 dyn_type, 1053 (r->id.src_ip), (r->id.src_port), 1054 (r->id.dst_ip), (r->id.dst_port), 1055 dyn_count ); ) 1056 return r; 1057} 1058 1059/** 1060 * lookup dynamic parent rule using pkt and rule as search keys. 1061 * If the lookup fails, then install one. 1062 */ 1063static ipfw_dyn_rule * 1064lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1065{ 1066 ipfw_dyn_rule *q; 1067 int i; 1068 1069 IPFW_DYN_LOCK_ASSERT(); 1070 1071 if (ipfw_dyn_v) { 1072 i = hash_packet( pkt ); 1073 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1074 if (q->dyn_type == O_LIMIT_PARENT && 1075 rule== q->rule && 1076 pkt->proto == q->id.proto && 1077 pkt->src_ip == q->id.src_ip && 1078 pkt->dst_ip == q->id.dst_ip && 1079 pkt->src_port == q->id.src_port && 1080 pkt->dst_port == q->id.dst_port) { 1081 q->expire = time_second + dyn_short_lifetime; 1082 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1083 return q; 1084 } 1085 } 1086 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1087} 1088 1089/** 1090 * Install dynamic state for rule type cmd->o.opcode 1091 * 1092 * Returns 1 (failure) if state is not installed because of errors or because 1093 * session limitations are enforced. 1094 */ 1095static int 1096install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1097 struct ip_fw_args *args) 1098{ 1099 static int last_log; 1100 1101 ipfw_dyn_rule *q; 1102 1103 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 1104 cmd->o.opcode, 1105 (args->f_id.src_ip), (args->f_id.src_port), 1106 (args->f_id.dst_ip), (args->f_id.dst_port) );) 1107 1108 IPFW_DYN_LOCK(); 1109 1110 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1111 1112 if (q != NULL) { /* should never occur */ 1113 if (last_log != time_second) { 1114 last_log = time_second; 1115 printf("ipfw: install_state: entry already present, done\n"); 1116 } 1117 IPFW_DYN_UNLOCK(); 1118 return 0; 1119 } 1120 1121 if (dyn_count >= dyn_max) 1122 /* 1123 * Run out of slots, try to remove any expired rule. 1124 */ 1125 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1126 1127 if (dyn_count >= dyn_max) { 1128 if (last_log != time_second) { 1129 last_log = time_second; 1130 printf("ipfw: install_state: Too many dynamic rules\n"); 1131 } 1132 IPFW_DYN_UNLOCK(); 1133 return 1; /* cannot install, notify caller */ 1134 } 1135 1136 switch (cmd->o.opcode) { 1137 case O_KEEP_STATE: /* bidir rule */ 1138 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1139 break; 1140 1141 case O_LIMIT: /* limit number of sessions */ 1142 { 1143 u_int16_t limit_mask = cmd->limit_mask; 1144 struct ipfw_flow_id id; 1145 ipfw_dyn_rule *parent; 1146 1147 DEB(printf("ipfw: installing dyn-limit rule %d\n", 1148 cmd->conn_limit);) 1149 1150 id.dst_ip = id.src_ip = 0; 1151 id.dst_port = id.src_port = 0; 1152 id.proto = args->f_id.proto; 1153 1154 if (limit_mask & DYN_SRC_ADDR) 1155 id.src_ip = args->f_id.src_ip; 1156 if (limit_mask & DYN_DST_ADDR) 1157 id.dst_ip = args->f_id.dst_ip; 1158 if (limit_mask & DYN_SRC_PORT) 1159 id.src_port = args->f_id.src_port; 1160 if (limit_mask & DYN_DST_PORT) 1161 id.dst_port = args->f_id.dst_port; 1162 parent = lookup_dyn_parent(&id, rule); 1163 if (parent == NULL) { 1164 printf("ipfw: add parent failed\n"); 1165 return 1; 1166 } 1167 if (parent->count >= cmd->conn_limit) { 1168 /* 1169 * See if we can remove some expired rule. 1170 */ 1171 remove_dyn_rule(rule, parent); 1172 if (parent->count >= cmd->conn_limit) { 1173 if (fw_verbose && last_log != time_second) { 1174 last_log = time_second; 1175 log(LOG_SECURITY | LOG_DEBUG, 1176 "drop session, too many entries\n"); 1177 } 1178 IPFW_DYN_UNLOCK(); 1179 return 1; 1180 } 1181 } 1182 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1183 } 1184 break; 1185 default: 1186 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 1187 IPFW_DYN_UNLOCK(); 1188 return 1; 1189 } 1190 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1191 IPFW_DYN_UNLOCK(); 1192 return 0; 1193} 1194 1195/* 1196 * Transmit a TCP packet, containing either a RST or a keepalive. 1197 * When flags & TH_RST, we are sending a RST packet, because of a 1198 * "reset" action matched the packet. 1199 * Otherwise we are sending a keepalive, and flags & TH_ 1200 */ 1201static void 1202send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1203{ 1204 struct mbuf *m; 1205 struct ip *ip; 1206 struct tcphdr *tcp; 1207 1208 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1209 if (m == 0) 1210 return; 1211 m->m_pkthdr.rcvif = (struct ifnet *)0; 1212 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1213 m->m_data += max_linkhdr; 1214 1215 ip = mtod(m, struct ip *); 1216 bzero(ip, m->m_len); 1217 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1218 ip->ip_p = IPPROTO_TCP; 1219 tcp->th_off = 5; 1220 /* 1221 * Assume we are sending a RST (or a keepalive in the reverse 1222 * direction), swap src and destination addresses and ports. 1223 */ 1224 ip->ip_src.s_addr = htonl(id->dst_ip); 1225 ip->ip_dst.s_addr = htonl(id->src_ip); 1226 tcp->th_sport = htons(id->dst_port); 1227 tcp->th_dport = htons(id->src_port); 1228 if (flags & TH_RST) { /* we are sending a RST */ 1229 if (flags & TH_ACK) { 1230 tcp->th_seq = htonl(ack); 1231 tcp->th_ack = htonl(0); 1232 tcp->th_flags = TH_RST; 1233 } else { 1234 if (flags & TH_SYN) 1235 seq++; 1236 tcp->th_seq = htonl(0); 1237 tcp->th_ack = htonl(seq); 1238 tcp->th_flags = TH_RST | TH_ACK; 1239 } 1240 } else { 1241 /* 1242 * We are sending a keepalive. flags & TH_SYN determines 1243 * the direction, forward if set, reverse if clear. 1244 * NOTE: seq and ack are always assumed to be correct 1245 * as set by the caller. This may be confusing... 1246 */ 1247 if (flags & TH_SYN) { 1248 /* 1249 * we have to rewrite the correct addresses! 1250 */ 1251 ip->ip_dst.s_addr = htonl(id->dst_ip); 1252 ip->ip_src.s_addr = htonl(id->src_ip); 1253 tcp->th_dport = htons(id->dst_port); 1254 tcp->th_sport = htons(id->src_port); 1255 } 1256 tcp->th_seq = htonl(seq); 1257 tcp->th_ack = htonl(ack); 1258 tcp->th_flags = TH_ACK; 1259 } 1260 /* 1261 * set ip_len to the payload size so we can compute 1262 * the tcp checksum on the pseudoheader 1263 * XXX check this, could save a couple of words ? 1264 */ 1265 ip->ip_len = htons(sizeof(struct tcphdr)); 1266 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1267 /* 1268 * now fill fields left out earlier 1269 */ 1270 ip->ip_ttl = ip_defttl; 1271 ip->ip_len = m->m_pkthdr.len; 1272 m->m_flags |= M_SKIP_FIREWALL; 1273 ip_output(m, NULL, NULL, 0, NULL, NULL); 1274} 1275 1276/* 1277 * sends a reject message, consuming the mbuf passed as an argument. 1278 */ 1279static void 1280send_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 1281{ 1282 1283 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1284 /* We need the IP header in host order for icmp_error(). */ 1285 if (args->eh != NULL) { 1286 struct ip *ip = mtod(args->m, struct ip *); 1287 ip->ip_len = ntohs(ip->ip_len); 1288 ip->ip_off = ntohs(ip->ip_off); 1289 } 1290 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1291 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1292 struct tcphdr *const tcp = 1293 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1294 if ( (tcp->th_flags & TH_RST) == 0) 1295 send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1296 ntohl(tcp->th_ack), 1297 tcp->th_flags | TH_RST); 1298 m_freem(args->m); 1299 } else 1300 m_freem(args->m); 1301 args->m = NULL; 1302} 1303 1304/** 1305 * 1306 * Given an ip_fw *, lookup_next_rule will return a pointer 1307 * to the next rule, which can be either the jump 1308 * target (for skipto instructions) or the next one in the list (in 1309 * all other cases including a missing jump target). 1310 * The result is also written in the "next_rule" field of the rule. 1311 * Backward jumps are not allowed, so start looking from the next 1312 * rule... 1313 * 1314 * This never returns NULL -- in case we do not have an exact match, 1315 * the next rule is returned. When the ruleset is changed, 1316 * pointers are flushed so we are always correct. 1317 */ 1318 1319static struct ip_fw * 1320lookup_next_rule(struct ip_fw *me) 1321{ 1322 struct ip_fw *rule = NULL; 1323 ipfw_insn *cmd; 1324 1325 /* look for action, in case it is a skipto */ 1326 cmd = ACTION_PTR(me); 1327 if (cmd->opcode == O_LOG) 1328 cmd += F_LEN(cmd); 1329 if ( cmd->opcode == O_SKIPTO ) 1330 for (rule = me->next; rule ; rule = rule->next) 1331 if (rule->rulenum >= cmd->arg1) 1332 break; 1333 if (rule == NULL) /* failure or not a skipto */ 1334 rule = me->next; 1335 me->next_rule = rule; 1336 return rule; 1337} 1338 1339static void 1340init_tables(void) 1341{ 1342 int i; 1343 1344 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1345 rn_inithead((void **)&ipfw_tables[i].rnh, 32); 1346 ipfw_tables[i].modified = 1; 1347 } 1348} 1349 1350static int 1351add_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen, u_int32_t value) 1352{ 1353 struct radix_node_head *rnh; 1354 struct table_entry *ent; 1355 1356 if (tbl >= IPFW_TABLES_MAX) 1357 return (EINVAL); 1358 rnh = ipfw_tables[tbl].rnh; 1359 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1360 if (ent == NULL) 1361 return (ENOMEM); 1362 ent->value = value; 1363 ent->addr.sin_len = ent->mask.sin_len = 8; 1364 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1365 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1366 RADIX_NODE_HEAD_LOCK(rnh); 1367 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1368 NULL) { 1369 RADIX_NODE_HEAD_UNLOCK(rnh); 1370 free(ent, M_IPFW_TBL); 1371 return (EEXIST); 1372 } 1373 ipfw_tables[tbl].modified = 1; 1374 RADIX_NODE_HEAD_UNLOCK(rnh); 1375 return (0); 1376} 1377 1378static int 1379del_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen) 1380{ 1381 struct radix_node_head *rnh; 1382 struct table_entry *ent; 1383 struct sockaddr_in sa, mask; 1384 1385 if (tbl >= IPFW_TABLES_MAX) 1386 return (EINVAL); 1387 rnh = ipfw_tables[tbl].rnh; 1388 sa.sin_len = mask.sin_len = 8; 1389 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1390 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1391 RADIX_NODE_HEAD_LOCK(rnh); 1392 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1393 if (ent == NULL) { 1394 RADIX_NODE_HEAD_UNLOCK(rnh); 1395 return (ESRCH); 1396 } 1397 ipfw_tables[tbl].modified = 1; 1398 RADIX_NODE_HEAD_UNLOCK(rnh); 1399 free(ent, M_IPFW_TBL); 1400 return (0); 1401} 1402 1403static int 1404flush_table_entry(struct radix_node *rn, void *arg) 1405{ 1406 struct radix_node_head * const rnh = arg; 1407 struct table_entry *ent; 1408 1409 ent = (struct table_entry *) 1410 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1411 if (ent != NULL) 1412 free(ent, M_IPFW_TBL); 1413 return (0); 1414} 1415 1416static int 1417flush_table(u_int16_t tbl) 1418{ 1419 struct radix_node_head *rnh; 1420 1421 if (tbl >= IPFW_TABLES_MAX) 1422 return (EINVAL); 1423 rnh = ipfw_tables[tbl].rnh; 1424 RADIX_NODE_HEAD_LOCK(rnh); 1425 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1426 ipfw_tables[tbl].modified = 1; 1427 RADIX_NODE_HEAD_UNLOCK(rnh); 1428 return (0); 1429} 1430 1431static void 1432flush_tables(void) 1433{ 1434 u_int16_t tbl; 1435 1436 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1437 flush_table(tbl); 1438} 1439 1440static int 1441lookup_table(u_int16_t tbl, in_addr_t addr, u_int32_t *val) 1442{ 1443 struct radix_node_head *rnh; 1444 struct table_entry *ent; 1445 struct sockaddr_in sa; 1446 static in_addr_t last_addr; 1447 static int last_tbl; 1448 static int last_match; 1449 static u_int32_t last_value; 1450 1451 if (tbl >= IPFW_TABLES_MAX) 1452 return (0); 1453 if (tbl == last_tbl && addr == last_addr && 1454 !ipfw_tables[tbl].modified) { 1455 if (last_match) 1456 *val = last_value; 1457 return (last_match); 1458 } 1459 rnh = ipfw_tables[tbl].rnh; 1460 sa.sin_len = 8; 1461 sa.sin_addr.s_addr = addr; 1462 RADIX_NODE_HEAD_LOCK(rnh); 1463 ipfw_tables[tbl].modified = 0; 1464 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1465 RADIX_NODE_HEAD_UNLOCK(rnh); 1466 last_addr = addr; 1467 last_tbl = tbl; 1468 if (ent != NULL) { 1469 last_value = *val = ent->value; 1470 last_match = 1; 1471 return (1); 1472 } 1473 last_match = 0; 1474 return (0); 1475} 1476 1477static int 1478count_table_entry(struct radix_node *rn, void *arg) 1479{ 1480 u_int32_t * const cnt = arg; 1481 1482 (*cnt)++; 1483 return (0); 1484} 1485 1486static int 1487count_table(u_int32_t tbl, u_int32_t *cnt) 1488{ 1489 struct radix_node_head *rnh; 1490 1491 if (tbl >= IPFW_TABLES_MAX) 1492 return (EINVAL); 1493 rnh = ipfw_tables[tbl].rnh; 1494 *cnt = 0; 1495 RADIX_NODE_HEAD_LOCK(rnh); 1496 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1497 RADIX_NODE_HEAD_UNLOCK(rnh); 1498 return (0); 1499} 1500 1501static int 1502dump_table_entry(struct radix_node *rn, void *arg) 1503{ 1504 struct table_entry * const n = (struct table_entry *)rn; 1505 ipfw_table * const tbl = arg; 1506 ipfw_table_entry *ent; 1507 1508 if (tbl->cnt == tbl->size) 1509 return (1); 1510 ent = &tbl->ent[tbl->cnt]; 1511 ent->tbl = tbl->tbl; 1512 if (in_nullhost(n->mask.sin_addr)) 1513 ent->masklen = 0; 1514 else 1515 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1516 ent->addr = n->addr.sin_addr.s_addr; 1517 ent->value = n->value; 1518 tbl->cnt++; 1519 return (0); 1520} 1521 1522static int 1523dump_table(ipfw_table *tbl) 1524{ 1525 struct radix_node_head *rnh; 1526 1527 if (tbl->tbl >= IPFW_TABLES_MAX) 1528 return (EINVAL); 1529 rnh = ipfw_tables[tbl->tbl].rnh; 1530 tbl->cnt = 0; 1531 RADIX_NODE_HEAD_LOCK(rnh); 1532 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1533 RADIX_NODE_HEAD_UNLOCK(rnh); 1534 return (0); 1535} 1536 1537static int 1538check_uidgid(ipfw_insn_u32 *insn, 1539 int proto, struct ifnet *oif, 1540 struct in_addr dst_ip, u_int16_t dst_port, 1541 struct in_addr src_ip, u_int16_t src_port, 1542 struct ip_fw_ugid *ugp, int *lookup) 1543{ 1544 struct inpcbinfo *pi; 1545 int wildcard; 1546 struct inpcb *pcb; 1547 int match; 1548 struct ucred *cr; 1549 gid_t *gp; 1550 1551 /* 1552 * If we have already been here and the packet has no 1553 * PCB entry associated with it, then we can safely 1554 * assume that this is a no match. 1555 */ 1556 if (*lookup == -1) 1557 return (0); 1558 if (proto == IPPROTO_TCP) { 1559 wildcard = 0; 1560 pi = &tcbinfo; 1561 } else if (proto == IPPROTO_UDP) { 1562 wildcard = 1; 1563 pi = &udbinfo; 1564 } else 1565 return 0; 1566 match = 0; 1567 if (*lookup == 0) { 1568 INP_INFO_RLOCK(pi); /* XXX LOR with IPFW */ 1569 pcb = (oif) ? 1570 in_pcblookup_hash(pi, 1571 dst_ip, htons(dst_port), 1572 src_ip, htons(src_port), 1573 wildcard, oif) : 1574 in_pcblookup_hash(pi, 1575 src_ip, htons(src_port), 1576 dst_ip, htons(dst_port), 1577 wildcard, NULL); 1578 if (pcb != NULL) { 1579 INP_LOCK(pcb); 1580 if (pcb->inp_socket != NULL) { 1581 cr = pcb->inp_socket->so_cred; 1582 ugp->fw_uid = cr->cr_uid; 1583 ugp->fw_ngroups = cr->cr_ngroups; 1584 bcopy(cr->cr_groups, ugp->fw_groups, 1585 sizeof(ugp->fw_groups)); 1586 *lookup = 1; 1587 } 1588 INP_UNLOCK(pcb); 1589 } 1590 INP_INFO_RUNLOCK(pi); 1591 if (*lookup == 0) { 1592 /* 1593 * If the lookup did not yield any results, there 1594 * is no sense in coming back and trying again. So 1595 * we can set lookup to -1 and ensure that we wont 1596 * bother the pcb system again. 1597 */ 1598 *lookup = -1; 1599 return (0); 1600 } 1601 } 1602 if (insn->o.opcode == O_UID) 1603 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1604 else if (insn->o.opcode == O_GID) 1605 for (gp = ugp->fw_groups; 1606 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1607 if (*gp == (gid_t)insn->d[0]) { 1608 match = 1; 1609 break; 1610 } 1611 return match; 1612} 1613 1614/* 1615 * The main check routine for the firewall. 1616 * 1617 * All arguments are in args so we can modify them and return them 1618 * back to the caller. 1619 * 1620 * Parameters: 1621 * 1622 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1623 * Starts with the IP header. 1624 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1625 * args->oif Outgoing interface, or NULL if packet is incoming. 1626 * The incoming interface is in the mbuf. (in) 1627 * args->divert_rule (in/out) 1628 * Skip up to the first rule past this rule number; 1629 * upon return, non-zero port number for divert or tee. 1630 * 1631 * args->rule Pointer to the last matching rule (in/out) 1632 * args->next_hop Socket we are forwarding to (out). 1633 * args->f_id Addresses grabbed from the packet (out) 1634 * 1635 * Return value: 1636 * 1637 * IP_FW_PORT_DENY_FLAG the packet must be dropped. 1638 * 0 The packet is to be accepted and routed normally OR 1639 * the packet was denied/rejected and has been dropped; 1640 * in the latter case, *m is equal to NULL upon return. 1641 * port Divert the packet to port, with these caveats: 1642 * 1643 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead 1644 * of diverting it (ie, 'ipfw tee'). 1645 * 1646 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower 1647 * 16 bits as a dummynet pipe number instead of diverting 1648 */ 1649 1650static int 1651ipfw_chk(struct ip_fw_args *args) 1652{ 1653 /* 1654 * Local variables hold state during the processing of a packet. 1655 * 1656 * IMPORTANT NOTE: to speed up the processing of rules, there 1657 * are some assumption on the values of the variables, which 1658 * are documented here. Should you change them, please check 1659 * the implementation of the various instructions to make sure 1660 * that they still work. 1661 * 1662 * args->eh The MAC header. It is non-null for a layer2 1663 * packet, it is NULL for a layer-3 packet. 1664 * 1665 * m | args->m Pointer to the mbuf, as received from the caller. 1666 * It may change if ipfw_chk() does an m_pullup, or if it 1667 * consumes the packet because it calls send_reject(). 1668 * XXX This has to change, so that ipfw_chk() never modifies 1669 * or consumes the buffer. 1670 * ip is simply an alias of the value of m, and it is kept 1671 * in sync with it (the packet is supposed to start with 1672 * the ip header). 1673 */ 1674 struct mbuf *m = args->m; 1675 struct ip *ip = mtod(m, struct ip *); 1676 1677 /* 1678 * For rules which contain uid/gid or jail constraints, cache 1679 * a copy of the users credentials after the pcb lookup has been 1680 * executed. This will speed up the processing of rules with 1681 * these types of constraints, as well as decrease contention 1682 * on pcb related locks. 1683 */ 1684 struct ip_fw_ugid fw_ugid_cache; 1685 int ugid_lookup = 0; 1686 1687 /* 1688 * oif | args->oif If NULL, ipfw_chk has been called on the 1689 * inbound path (ether_input, bdg_forward, ip_input). 1690 * If non-NULL, ipfw_chk has been called on the outbound path 1691 * (ether_output, ip_output). 1692 */ 1693 struct ifnet *oif = args->oif; 1694 1695 struct ip_fw *f = NULL; /* matching rule */ 1696 int retval = 0; 1697 1698 /* 1699 * hlen The length of the IPv4 header. 1700 * hlen >0 means we have an IPv4 packet. 1701 */ 1702 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1703 1704 /* 1705 * offset The offset of a fragment. offset != 0 means that 1706 * we have a fragment at this offset of an IPv4 packet. 1707 * offset == 0 means that (if this is an IPv4 packet) 1708 * this is the first or only fragment. 1709 */ 1710 u_short offset = 0; 1711 1712 /* 1713 * Local copies of addresses. They are only valid if we have 1714 * an IP packet. 1715 * 1716 * proto The protocol. Set to 0 for non-ip packets, 1717 * or to the protocol read from the packet otherwise. 1718 * proto != 0 means that we have an IPv4 packet. 1719 * 1720 * src_port, dst_port port numbers, in HOST format. Only 1721 * valid for TCP and UDP packets. 1722 * 1723 * src_ip, dst_ip ip addresses, in NETWORK format. 1724 * Only valid for IPv4 packets. 1725 */ 1726 u_int8_t proto; 1727 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1728 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1729 u_int16_t ip_len=0; 1730 int pktlen; 1731 int dyn_dir = MATCH_UNKNOWN; 1732 ipfw_dyn_rule *q = NULL; 1733 struct ip_fw_chain *chain = &layer3_chain; 1734 struct m_tag *mtag; 1735 1736 if (m->m_flags & M_SKIP_FIREWALL) 1737 return 0; /* accept */ 1738 /* 1739 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1740 * MATCH_NONE when checked and not matched (q = NULL), 1741 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1742 */ 1743 1744 pktlen = m->m_pkthdr.len; 1745 if (args->eh == NULL || /* layer 3 packet */ 1746 ( m->m_pkthdr.len >= sizeof(struct ip) && 1747 ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 1748 hlen = ip->ip_hl << 2; 1749 1750 /* 1751 * Collect parameters into local variables for faster matching. 1752 */ 1753 if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 1754 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1755 goto after_ip_checks; 1756 } 1757 1758 proto = args->f_id.proto = ip->ip_p; 1759 src_ip = ip->ip_src; 1760 dst_ip = ip->ip_dst; 1761 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1762 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1763 ip_len = ntohs(ip->ip_len); 1764 } else { 1765 offset = ip->ip_off & IP_OFFMASK; 1766 ip_len = ip->ip_len; 1767 } 1768 pktlen = ip_len < pktlen ? ip_len : pktlen; 1769 1770#define PULLUP_TO(len) \ 1771 do { \ 1772 if ((m)->m_len < (len)) { \ 1773 args->m = m = m_pullup(m, (len)); \ 1774 if (m == 0) \ 1775 goto pullup_failed; \ 1776 ip = mtod(m, struct ip *); \ 1777 } \ 1778 } while (0) 1779 1780 if (offset == 0) { 1781 switch (proto) { 1782 case IPPROTO_TCP: 1783 { 1784 struct tcphdr *tcp; 1785 1786 PULLUP_TO(hlen + sizeof(struct tcphdr)); 1787 tcp = L3HDR(struct tcphdr, ip); 1788 dst_port = tcp->th_dport; 1789 src_port = tcp->th_sport; 1790 args->f_id.flags = tcp->th_flags; 1791 } 1792 break; 1793 1794 case IPPROTO_UDP: 1795 { 1796 struct udphdr *udp; 1797 1798 PULLUP_TO(hlen + sizeof(struct udphdr)); 1799 udp = L3HDR(struct udphdr, ip); 1800 dst_port = udp->uh_dport; 1801 src_port = udp->uh_sport; 1802 } 1803 break; 1804 1805 case IPPROTO_ICMP: 1806 PULLUP_TO(hlen + 4); /* type, code and checksum. */ 1807 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 1808 break; 1809 1810 default: 1811 break; 1812 } 1813#undef PULLUP_TO 1814 } 1815 1816 args->f_id.src_ip = ntohl(src_ip.s_addr); 1817 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1818 args->f_id.src_port = src_port = ntohs(src_port); 1819 args->f_id.dst_port = dst_port = ntohs(dst_port); 1820 1821after_ip_checks: 1822 IPFW_LOCK(chain); /* XXX expensive? can we run lock free? */ 1823 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 1824 if (args->rule) { 1825 /* 1826 * Packet has already been tagged. Look for the next rule 1827 * to restart processing. 1828 * 1829 * If fw_one_pass != 0 then just accept it. 1830 * XXX should not happen here, but optimized out in 1831 * the caller. 1832 */ 1833 if (fw_one_pass) { 1834 IPFW_UNLOCK(chain); /* XXX optimize */ 1835 return 0; 1836 } 1837 1838 f = args->rule->next_rule; 1839 if (f == NULL) 1840 f = lookup_next_rule(args->rule); 1841 } else { 1842 /* 1843 * Find the starting rule. It can be either the first 1844 * one, or the one after divert_rule if asked so. 1845 */ 1846 int skipto = mtag ? divert_cookie(mtag) : 0; 1847 1848 f = chain->rules; 1849 if (args->eh == NULL && skipto != 0) { 1850 if (skipto >= IPFW_DEFAULT_RULE) { 1851 IPFW_UNLOCK(chain); 1852 return(IP_FW_PORT_DENY_FLAG); /* invalid */ 1853 } 1854 while (f && f->rulenum <= skipto) 1855 f = f->next; 1856 if (f == NULL) { /* drop packet */ 1857 IPFW_UNLOCK(chain); 1858 return(IP_FW_PORT_DENY_FLAG); 1859 } 1860 } 1861 } 1862 /* reset divert rule to avoid confusion later */ 1863 if (mtag) 1864 m_tag_delete(m, mtag); 1865 1866 /* 1867 * Now scan the rules, and parse microinstructions for each rule. 1868 */ 1869 for (; f; f = f->next) { 1870 int l, cmdlen; 1871 ipfw_insn *cmd; 1872 int skip_or; /* skip rest of OR block */ 1873 1874again: 1875 if (set_disable & (1 << f->set) ) 1876 continue; 1877 1878 skip_or = 0; 1879 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1880 l -= cmdlen, cmd += cmdlen) { 1881 int match; 1882 1883 /* 1884 * check_body is a jump target used when we find a 1885 * CHECK_STATE, and need to jump to the body of 1886 * the target rule. 1887 */ 1888 1889check_body: 1890 cmdlen = F_LEN(cmd); 1891 /* 1892 * An OR block (insn_1 || .. || insn_n) has the 1893 * F_OR bit set in all but the last instruction. 1894 * The first match will set "skip_or", and cause 1895 * the following instructions to be skipped until 1896 * past the one with the F_OR bit clear. 1897 */ 1898 if (skip_or) { /* skip this instruction */ 1899 if ((cmd->len & F_OR) == 0) 1900 skip_or = 0; /* next one is good */ 1901 continue; 1902 } 1903 match = 0; /* set to 1 if we succeed */ 1904 1905 switch (cmd->opcode) { 1906 /* 1907 * The first set of opcodes compares the packet's 1908 * fields with some pattern, setting 'match' if a 1909 * match is found. At the end of the loop there is 1910 * logic to deal with F_NOT and F_OR flags associated 1911 * with the opcode. 1912 */ 1913 case O_NOP: 1914 match = 1; 1915 break; 1916 1917 case O_FORWARD_MAC: 1918 printf("ipfw: opcode %d unimplemented\n", 1919 cmd->opcode); 1920 break; 1921 1922 case O_GID: 1923 case O_UID: 1924 /* 1925 * We only check offset == 0 && proto != 0, 1926 * as this ensures that we have an IPv4 1927 * packet with the ports info. 1928 */ 1929 if (offset!=0) 1930 break; 1931 if (proto == IPPROTO_TCP || 1932 proto == IPPROTO_UDP) 1933 match = check_uidgid( 1934 (ipfw_insn_u32 *)cmd, 1935 proto, oif, 1936 dst_ip, dst_port, 1937 src_ip, src_port, &fw_ugid_cache, 1938 &ugid_lookup); 1939 break; 1940 1941 case O_RECV: 1942 match = iface_match(m->m_pkthdr.rcvif, 1943 (ipfw_insn_if *)cmd); 1944 break; 1945 1946 case O_XMIT: 1947 match = iface_match(oif, (ipfw_insn_if *)cmd); 1948 break; 1949 1950 case O_VIA: 1951 match = iface_match(oif ? oif : 1952 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1953 break; 1954 1955 case O_MACADDR2: 1956 if (args->eh != NULL) { /* have MAC header */ 1957 u_int32_t *want = (u_int32_t *) 1958 ((ipfw_insn_mac *)cmd)->addr; 1959 u_int32_t *mask = (u_int32_t *) 1960 ((ipfw_insn_mac *)cmd)->mask; 1961 u_int32_t *hdr = (u_int32_t *)args->eh; 1962 1963 match = 1964 ( want[0] == (hdr[0] & mask[0]) && 1965 want[1] == (hdr[1] & mask[1]) && 1966 want[2] == (hdr[2] & mask[2]) ); 1967 } 1968 break; 1969 1970 case O_MAC_TYPE: 1971 if (args->eh != NULL) { 1972 u_int16_t t = 1973 ntohs(args->eh->ether_type); 1974 u_int16_t *p = 1975 ((ipfw_insn_u16 *)cmd)->ports; 1976 int i; 1977 1978 for (i = cmdlen - 1; !match && i>0; 1979 i--, p += 2) 1980 match = (t>=p[0] && t<=p[1]); 1981 } 1982 break; 1983 1984 case O_FRAG: 1985 match = (hlen > 0 && offset != 0); 1986 break; 1987 1988 case O_IN: /* "out" is "not in" */ 1989 match = (oif == NULL); 1990 break; 1991 1992 case O_LAYER2: 1993 match = (args->eh != NULL); 1994 break; 1995 1996 case O_PROTO: 1997 /* 1998 * We do not allow an arg of 0 so the 1999 * check of "proto" only suffices. 2000 */ 2001 match = (proto == cmd->arg1); 2002 break; 2003 2004 case O_IP_SRC: 2005 match = (hlen > 0 && 2006 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2007 src_ip.s_addr); 2008 break; 2009 2010 case O_IP_SRC_LOOKUP: 2011 case O_IP_DST_LOOKUP: 2012 if (hlen > 0) { 2013 uint32_t a = 2014 (cmd->opcode == O_IP_DST_LOOKUP) ? 2015 dst_ip.s_addr : src_ip.s_addr; 2016 uint32_t v; 2017 2018 match = lookup_table(cmd->arg1, a, &v); 2019 if (!match) 2020 break; 2021 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2022 match = 2023 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2024 } 2025 break; 2026 2027 case O_IP_SRC_MASK: 2028 case O_IP_DST_MASK: 2029 if (hlen > 0) { 2030 uint32_t a = 2031 (cmd->opcode == O_IP_DST_MASK) ? 2032 dst_ip.s_addr : src_ip.s_addr; 2033 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2034 int i = cmdlen-1; 2035 2036 for (; !match && i>0; i-= 2, p+= 2) 2037 match = (p[0] == (a & p[1])); 2038 } 2039 break; 2040 2041 case O_IP_SRC_ME: 2042 if (hlen > 0) { 2043 struct ifnet *tif; 2044 2045 INADDR_TO_IFP(src_ip, tif); 2046 match = (tif != NULL); 2047 } 2048 break; 2049 2050 case O_IP_DST_SET: 2051 case O_IP_SRC_SET: 2052 if (hlen > 0) { 2053 u_int32_t *d = (u_int32_t *)(cmd+1); 2054 u_int32_t addr = 2055 cmd->opcode == O_IP_DST_SET ? 2056 args->f_id.dst_ip : 2057 args->f_id.src_ip; 2058 2059 if (addr < d[0]) 2060 break; 2061 addr -= d[0]; /* subtract base */ 2062 match = (addr < cmd->arg1) && 2063 ( d[ 1 + (addr>>5)] & 2064 (1<<(addr & 0x1f)) ); 2065 } 2066 break; 2067 2068 case O_IP_DST: 2069 match = (hlen > 0 && 2070 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2071 dst_ip.s_addr); 2072 break; 2073 2074 case O_IP_DST_ME: 2075 if (hlen > 0) { 2076 struct ifnet *tif; 2077 2078 INADDR_TO_IFP(dst_ip, tif); 2079 match = (tif != NULL); 2080 } 2081 break; 2082 2083 case O_IP_SRCPORT: 2084 case O_IP_DSTPORT: 2085 /* 2086 * offset == 0 && proto != 0 is enough 2087 * to guarantee that we have an IPv4 2088 * packet with port info. 2089 */ 2090 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2091 && offset == 0) { 2092 u_int16_t x = 2093 (cmd->opcode == O_IP_SRCPORT) ? 2094 src_port : dst_port ; 2095 u_int16_t *p = 2096 ((ipfw_insn_u16 *)cmd)->ports; 2097 int i; 2098 2099 for (i = cmdlen - 1; !match && i>0; 2100 i--, p += 2) 2101 match = (x>=p[0] && x<=p[1]); 2102 } 2103 break; 2104 2105 case O_ICMPTYPE: 2106 match = (offset == 0 && proto==IPPROTO_ICMP && 2107 icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 2108 break; 2109 2110 case O_IPOPT: 2111 match = (hlen > 0 && ipopts_match(ip, cmd) ); 2112 break; 2113 2114 case O_IPVER: 2115 match = (hlen > 0 && cmd->arg1 == ip->ip_v); 2116 break; 2117 2118 case O_IPID: 2119 case O_IPLEN: 2120 case O_IPTTL: 2121 if (hlen > 0) { /* only for IP packets */ 2122 uint16_t x; 2123 uint16_t *p; 2124 int i; 2125 2126 if (cmd->opcode == O_IPLEN) 2127 x = ip_len; 2128 else if (cmd->opcode == O_IPTTL) 2129 x = ip->ip_ttl; 2130 else /* must be IPID */ 2131 x = ntohs(ip->ip_id); 2132 if (cmdlen == 1) { 2133 match = (cmd->arg1 == x); 2134 break; 2135 } 2136 /* otherwise we have ranges */ 2137 p = ((ipfw_insn_u16 *)cmd)->ports; 2138 i = cmdlen - 1; 2139 for (; !match && i>0; i--, p += 2) 2140 match = (x >= p[0] && x <= p[1]); 2141 } 2142 break; 2143 2144 case O_IPPRECEDENCE: 2145 match = (hlen > 0 && 2146 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2147 break; 2148 2149 case O_IPTOS: 2150 match = (hlen > 0 && 2151 flags_match(cmd, ip->ip_tos)); 2152 break; 2153 2154 case O_TCPFLAGS: 2155 match = (proto == IPPROTO_TCP && offset == 0 && 2156 flags_match(cmd, 2157 L3HDR(struct tcphdr,ip)->th_flags)); 2158 break; 2159 2160 case O_TCPOPTS: 2161 match = (proto == IPPROTO_TCP && offset == 0 && 2162 tcpopts_match(ip, cmd)); 2163 break; 2164 2165 case O_TCPSEQ: 2166 match = (proto == IPPROTO_TCP && offset == 0 && 2167 ((ipfw_insn_u32 *)cmd)->d[0] == 2168 L3HDR(struct tcphdr,ip)->th_seq); 2169 break; 2170 2171 case O_TCPACK: 2172 match = (proto == IPPROTO_TCP && offset == 0 && 2173 ((ipfw_insn_u32 *)cmd)->d[0] == 2174 L3HDR(struct tcphdr,ip)->th_ack); 2175 break; 2176 2177 case O_TCPWIN: 2178 match = (proto == IPPROTO_TCP && offset == 0 && 2179 cmd->arg1 == 2180 L3HDR(struct tcphdr,ip)->th_win); 2181 break; 2182 2183 case O_ESTAB: 2184 /* reject packets which have SYN only */ 2185 /* XXX should i also check for TH_ACK ? */ 2186 match = (proto == IPPROTO_TCP && offset == 0 && 2187 (L3HDR(struct tcphdr,ip)->th_flags & 2188 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2189 break; 2190 2191 case O_LOG: 2192 if (fw_verbose) 2193 ipfw_log(f, hlen, args->eh, m, oif); 2194 match = 1; 2195 break; 2196 2197 case O_PROB: 2198 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2199 break; 2200 2201 case O_VERREVPATH: 2202 /* Outgoing packets automatically pass/match */ 2203 match = ((oif != NULL) || 2204 (m->m_pkthdr.rcvif == NULL) || 2205 verify_path(src_ip, m->m_pkthdr.rcvif)); 2206 break; 2207 2208 case O_VERSRCREACH: 2209 /* Outgoing packets automatically pass/match */ 2210 match = ((oif != NULL) || 2211 verify_path(src_ip, NULL)); 2212 break; 2213 2214 case O_IPSEC: 2215#ifdef FAST_IPSEC 2216 match = (m_tag_find(m, 2217 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2218#endif 2219#ifdef IPSEC 2220 match = (ipsec_getnhist(m) != 0); 2221#endif 2222 /* otherwise no match */ 2223 break; 2224 2225 /* 2226 * The second set of opcodes represents 'actions', 2227 * i.e. the terminal part of a rule once the packet 2228 * matches all previous patterns. 2229 * Typically there is only one action for each rule, 2230 * and the opcode is stored at the end of the rule 2231 * (but there are exceptions -- see below). 2232 * 2233 * In general, here we set retval and terminate the 2234 * outer loop (would be a 'break 3' in some language, 2235 * but we need to do a 'goto done'). 2236 * 2237 * Exceptions: 2238 * O_COUNT and O_SKIPTO actions: 2239 * instead of terminating, we jump to the next rule 2240 * ('goto next_rule', equivalent to a 'break 2'), 2241 * or to the SKIPTO target ('goto again' after 2242 * having set f, cmd and l), respectively. 2243 * 2244 * O_LIMIT and O_KEEP_STATE: these opcodes are 2245 * not real 'actions', and are stored right 2246 * before the 'action' part of the rule. 2247 * These opcodes try to install an entry in the 2248 * state tables; if successful, we continue with 2249 * the next opcode (match=1; break;), otherwise 2250 * the packet * must be dropped 2251 * ('goto done' after setting retval); 2252 * 2253 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2254 * cause a lookup of the state table, and a jump 2255 * to the 'action' part of the parent rule 2256 * ('goto check_body') if an entry is found, or 2257 * (CHECK_STATE only) a jump to the next rule if 2258 * the entry is not found ('goto next_rule'). 2259 * The result of the lookup is cached to make 2260 * further instances of these opcodes are 2261 * effectively NOPs. 2262 */ 2263 case O_LIMIT: 2264 case O_KEEP_STATE: 2265 if (install_state(f, 2266 (ipfw_insn_limit *)cmd, args)) { 2267 retval = IP_FW_PORT_DENY_FLAG; 2268 goto done; /* error/limit violation */ 2269 } 2270 match = 1; 2271 break; 2272 2273 case O_PROBE_STATE: 2274 case O_CHECK_STATE: 2275 /* 2276 * dynamic rules are checked at the first 2277 * keep-state or check-state occurrence, 2278 * with the result being stored in dyn_dir. 2279 * The compiler introduces a PROBE_STATE 2280 * instruction for us when we have a 2281 * KEEP_STATE (because PROBE_STATE needs 2282 * to be run first). 2283 */ 2284 if (dyn_dir == MATCH_UNKNOWN && 2285 (q = lookup_dyn_rule(&args->f_id, 2286 &dyn_dir, proto == IPPROTO_TCP ? 2287 L3HDR(struct tcphdr, ip) : NULL)) 2288 != NULL) { 2289 /* 2290 * Found dynamic entry, update stats 2291 * and jump to the 'action' part of 2292 * the parent rule. 2293 */ 2294 q->pcnt++; 2295 q->bcnt += pktlen; 2296 f = q->rule; 2297 cmd = ACTION_PTR(f); 2298 l = f->cmd_len - f->act_ofs; 2299 IPFW_DYN_UNLOCK(); 2300 goto check_body; 2301 } 2302 /* 2303 * Dynamic entry not found. If CHECK_STATE, 2304 * skip to next rule, if PROBE_STATE just 2305 * ignore and continue with next opcode. 2306 */ 2307 if (cmd->opcode == O_CHECK_STATE) 2308 goto next_rule; 2309 match = 1; 2310 break; 2311 2312 case O_ACCEPT: 2313 retval = 0; /* accept */ 2314 goto done; 2315 2316 case O_PIPE: 2317 case O_QUEUE: 2318 args->rule = f; /* report matching rule */ 2319 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG; 2320 goto done; 2321 2322 case O_DIVERT: 2323 case O_TEE: { 2324 struct divert_tag *dt; 2325 2326 if (args->eh) /* not on layer 2 */ 2327 break; 2328 mtag = m_tag_get(PACKET_TAG_DIVERT, 2329 sizeof(struct divert_tag), 2330 M_NOWAIT); 2331 if (mtag == NULL) { 2332 /* XXX statistic */ 2333 /* drop packet */ 2334 IPFW_UNLOCK(chain); 2335 return IP_FW_PORT_DENY_FLAG; 2336 } 2337 dt = (struct divert_tag *)(mtag+1); 2338 dt->cookie = f->rulenum; 2339 dt->info = (cmd->opcode == O_DIVERT) ? 2340 cmd->arg1 : 2341 cmd->arg1 | IP_FW_PORT_TEE_FLAG; 2342 m_tag_prepend(m, mtag); 2343 retval = dt->info; 2344 goto done; 2345 } 2346 2347 case O_COUNT: 2348 case O_SKIPTO: 2349 f->pcnt++; /* update stats */ 2350 f->bcnt += pktlen; 2351 f->timestamp = time_second; 2352 if (cmd->opcode == O_COUNT) 2353 goto next_rule; 2354 /* handle skipto */ 2355 if (f->next_rule == NULL) 2356 lookup_next_rule(f); 2357 f = f->next_rule; 2358 goto again; 2359 2360 case O_REJECT: 2361 /* 2362 * Drop the packet and send a reject notice 2363 * if the packet is not ICMP (or is an ICMP 2364 * query), and it is not multicast/broadcast. 2365 */ 2366 if (hlen > 0 && 2367 (proto != IPPROTO_ICMP || 2368 is_icmp_query(ip)) && 2369 !(m->m_flags & (M_BCAST|M_MCAST)) && 2370 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2371 send_reject(args, cmd->arg1, 2372 offset,ip_len); 2373 m = args->m; 2374 } 2375 /* FALLTHROUGH */ 2376 case O_DENY: 2377 retval = IP_FW_PORT_DENY_FLAG; 2378 goto done; 2379 2380 case O_FORWARD_IP: 2381 if (args->eh) /* not valid on layer2 pkts */ 2382 break; 2383 if (!q || dyn_dir == MATCH_FORWARD) 2384 args->next_hop = 2385 &((ipfw_insn_sa *)cmd)->sa; 2386 retval = 0; 2387 goto done; 2388 2389 default: 2390 panic("-- unknown opcode %d\n", cmd->opcode); 2391 } /* end of switch() on opcodes */ 2392 2393 if (cmd->len & F_NOT) 2394 match = !match; 2395 2396 if (match) { 2397 if (cmd->len & F_OR) 2398 skip_or = 1; 2399 } else { 2400 if (!(cmd->len & F_OR)) /* not an OR block, */ 2401 break; /* try next rule */ 2402 } 2403 2404 } /* end of inner for, scan opcodes */ 2405 2406next_rule:; /* try next rule */ 2407 2408 } /* end of outer for, scan rules */ 2409 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2410 IPFW_UNLOCK(chain); 2411 return(IP_FW_PORT_DENY_FLAG); 2412 2413done: 2414 /* Update statistics */ 2415 f->pcnt++; 2416 f->bcnt += pktlen; 2417 f->timestamp = time_second; 2418 IPFW_UNLOCK(chain); 2419 return retval; 2420 2421pullup_failed: 2422 if (fw_verbose) 2423 printf("ipfw: pullup failed\n"); 2424 return(IP_FW_PORT_DENY_FLAG); 2425} 2426 2427/* 2428 * When a rule is added/deleted, clear the next_rule pointers in all rules. 2429 * These will be reconstructed on the fly as packets are matched. 2430 */ 2431static void 2432flush_rule_ptrs(struct ip_fw_chain *chain) 2433{ 2434 struct ip_fw *rule; 2435 2436 IPFW_LOCK_ASSERT(chain); 2437 2438 for (rule = chain->rules; rule; rule = rule->next) 2439 rule->next_rule = NULL; 2440} 2441 2442/* 2443 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 2444 * pipe/queue, or to all of them (match == NULL). 2445 */ 2446void 2447flush_pipe_ptrs(struct dn_flow_set *match) 2448{ 2449 struct ip_fw *rule; 2450 2451 IPFW_LOCK(&layer3_chain); 2452 for (rule = layer3_chain.rules; rule; rule = rule->next) { 2453 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 2454 2455 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 2456 continue; 2457 /* 2458 * XXX Use bcmp/bzero to handle pipe_ptr to overcome 2459 * possible alignment problems on 64-bit architectures. 2460 * This code is seldom used so we do not worry too 2461 * much about efficiency. 2462 */ 2463 if (match == NULL || 2464 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) ) 2465 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr)); 2466 } 2467 IPFW_UNLOCK(&layer3_chain); 2468} 2469 2470/* 2471 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 2472 * possibly create a rule number and add the rule to the list. 2473 * Update the rule_number in the input struct so the caller knows it as well. 2474 */ 2475static int 2476add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 2477{ 2478 struct ip_fw *rule, *f, *prev; 2479 int l = RULESIZE(input_rule); 2480 2481 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 2482 return (EINVAL); 2483 2484 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 2485 if (rule == NULL) 2486 return (ENOSPC); 2487 2488 bcopy(input_rule, rule, l); 2489 2490 rule->next = NULL; 2491 rule->next_rule = NULL; 2492 2493 rule->pcnt = 0; 2494 rule->bcnt = 0; 2495 rule->timestamp = 0; 2496 2497 IPFW_LOCK(chain); 2498 2499 if (chain->rules == NULL) { /* default rule */ 2500 chain->rules = rule; 2501 goto done; 2502 } 2503 2504 /* 2505 * If rulenum is 0, find highest numbered rule before the 2506 * default rule, and add autoinc_step 2507 */ 2508 if (autoinc_step < 1) 2509 autoinc_step = 1; 2510 else if (autoinc_step > 1000) 2511 autoinc_step = 1000; 2512 if (rule->rulenum == 0) { 2513 /* 2514 * locate the highest numbered rule before default 2515 */ 2516 for (f = chain->rules; f; f = f->next) { 2517 if (f->rulenum == IPFW_DEFAULT_RULE) 2518 break; 2519 rule->rulenum = f->rulenum; 2520 } 2521 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2522 rule->rulenum += autoinc_step; 2523 input_rule->rulenum = rule->rulenum; 2524 } 2525 2526 /* 2527 * Now insert the new rule in the right place in the sorted list. 2528 */ 2529 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 2530 if (f->rulenum > rule->rulenum) { /* found the location */ 2531 if (prev) { 2532 rule->next = f; 2533 prev->next = rule; 2534 } else { /* head insert */ 2535 rule->next = chain->rules; 2536 chain->rules = rule; 2537 } 2538 break; 2539 } 2540 } 2541 flush_rule_ptrs(chain); 2542done: 2543 static_count++; 2544 static_len += l; 2545 IPFW_UNLOCK(chain); 2546 DEB(printf("ipfw: installed rule %d, static count now %d\n", 2547 rule->rulenum, static_count);) 2548 return (0); 2549} 2550 2551/** 2552 * Remove a static rule (including derived * dynamic rules) 2553 * and place it on the ``reap list'' for later reclamation. 2554 * The caller is in charge of clearing rule pointers to avoid 2555 * dangling pointers. 2556 * @return a pointer to the next entry. 2557 * Arguments are not checked, so they better be correct. 2558 */ 2559static struct ip_fw * 2560remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 2561{ 2562 struct ip_fw *n; 2563 int l = RULESIZE(rule); 2564 2565 IPFW_LOCK_ASSERT(chain); 2566 2567 n = rule->next; 2568 IPFW_DYN_LOCK(); 2569 remove_dyn_rule(rule, NULL /* force removal */); 2570 IPFW_DYN_UNLOCK(); 2571 if (prev == NULL) 2572 chain->rules = n; 2573 else 2574 prev->next = n; 2575 static_count--; 2576 static_len -= l; 2577 2578 rule->next = chain->reap; 2579 chain->reap = rule; 2580 2581 return n; 2582} 2583 2584/** 2585 * Reclaim storage associated with a list of rules. This is 2586 * typically the list created using remove_rule. 2587 */ 2588static void 2589reap_rules(struct ip_fw *head) 2590{ 2591 struct ip_fw *rule; 2592 2593 while ((rule = head) != NULL) { 2594 head = head->next; 2595 if (DUMMYNET_LOADED) 2596 ip_dn_ruledel_ptr(rule); 2597 free(rule, M_IPFW); 2598 } 2599} 2600 2601/* 2602 * Remove all rules from a chain (except rules in set RESVD_SET 2603 * unless kill_default = 1). The caller is responsible for 2604 * reclaiming storage for the rules left in chain->reap. 2605 */ 2606static void 2607free_chain(struct ip_fw_chain *chain, int kill_default) 2608{ 2609 struct ip_fw *prev, *rule; 2610 2611 IPFW_LOCK_ASSERT(chain); 2612 2613 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 2614 for (prev = NULL, rule = chain->rules; rule ; ) 2615 if (kill_default || rule->set != RESVD_SET) 2616 rule = remove_rule(chain, rule, prev); 2617 else { 2618 prev = rule; 2619 rule = rule->next; 2620 } 2621} 2622 2623/** 2624 * Remove all rules with given number, and also do set manipulation. 2625 * Assumes chain != NULL && *chain != NULL. 2626 * 2627 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2628 * the next 8 bits are the new set, the top 8 bits are the command: 2629 * 2630 * 0 delete rules with given number 2631 * 1 delete rules with given set number 2632 * 2 move rules with given number to new set 2633 * 3 move rules with given set number to new set 2634 * 4 swap sets with given numbers 2635 */ 2636static int 2637del_entry(struct ip_fw_chain *chain, u_int32_t arg) 2638{ 2639 struct ip_fw *prev = NULL, *rule; 2640 u_int16_t rulenum; /* rule or old_set */ 2641 u_int8_t cmd, new_set; 2642 2643 rulenum = arg & 0xffff; 2644 cmd = (arg >> 24) & 0xff; 2645 new_set = (arg >> 16) & 0xff; 2646 2647 if (cmd > 4) 2648 return EINVAL; 2649 if (new_set > RESVD_SET) 2650 return EINVAL; 2651 if (cmd == 0 || cmd == 2) { 2652 if (rulenum >= IPFW_DEFAULT_RULE) 2653 return EINVAL; 2654 } else { 2655 if (rulenum > RESVD_SET) /* old_set */ 2656 return EINVAL; 2657 } 2658 2659 IPFW_LOCK(chain); 2660 rule = chain->rules; 2661 chain->reap = NULL; 2662 switch (cmd) { 2663 case 0: /* delete rules with given number */ 2664 /* 2665 * locate first rule to delete 2666 */ 2667 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 2668 ; 2669 if (rule->rulenum != rulenum) { 2670 IPFW_UNLOCK(chain); 2671 return EINVAL; 2672 } 2673 2674 /* 2675 * flush pointers outside the loop, then delete all matching 2676 * rules. prev remains the same throughout the cycle. 2677 */ 2678 flush_rule_ptrs(chain); 2679 while (rule->rulenum == rulenum) 2680 rule = remove_rule(chain, rule, prev); 2681 break; 2682 2683 case 1: /* delete all rules with given set number */ 2684 flush_rule_ptrs(chain); 2685 rule = chain->rules; 2686 while (rule->rulenum < IPFW_DEFAULT_RULE) 2687 if (rule->set == rulenum) 2688 rule = remove_rule(chain, rule, prev); 2689 else { 2690 prev = rule; 2691 rule = rule->next; 2692 } 2693 break; 2694 2695 case 2: /* move rules with given number to new set */ 2696 rule = chain->rules; 2697 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2698 if (rule->rulenum == rulenum) 2699 rule->set = new_set; 2700 break; 2701 2702 case 3: /* move rules with given set number to new set */ 2703 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2704 if (rule->set == rulenum) 2705 rule->set = new_set; 2706 break; 2707 2708 case 4: /* swap two sets */ 2709 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2710 if (rule->set == rulenum) 2711 rule->set = new_set; 2712 else if (rule->set == new_set) 2713 rule->set = rulenum; 2714 break; 2715 } 2716 /* 2717 * Look for rules to reclaim. We grab the list before 2718 * releasing the lock then reclaim them w/o the lock to 2719 * avoid a LOR with dummynet. 2720 */ 2721 rule = chain->reap; 2722 chain->reap = NULL; 2723 IPFW_UNLOCK(chain); 2724 if (rule) 2725 reap_rules(rule); 2726 return 0; 2727} 2728 2729/* 2730 * Clear counters for a specific rule. 2731 * The enclosing "table" is assumed locked. 2732 */ 2733static void 2734clear_counters(struct ip_fw *rule, int log_only) 2735{ 2736 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 2737 2738 if (log_only == 0) { 2739 rule->bcnt = rule->pcnt = 0; 2740 rule->timestamp = 0; 2741 } 2742 if (l->o.opcode == O_LOG) 2743 l->log_left = l->max_log; 2744} 2745 2746/** 2747 * Reset some or all counters on firewall rules. 2748 * @arg frwl is null to clear all entries, or contains a specific 2749 * rule number. 2750 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 2751 */ 2752static int 2753zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 2754{ 2755 struct ip_fw *rule; 2756 char *msg; 2757 2758 IPFW_LOCK(chain); 2759 if (rulenum == 0) { 2760 norule_counter = 0; 2761 for (rule = chain->rules; rule; rule = rule->next) 2762 clear_counters(rule, log_only); 2763 msg = log_only ? "ipfw: All logging counts reset.\n" : 2764 "ipfw: Accounting cleared.\n"; 2765 } else { 2766 int cleared = 0; 2767 /* 2768 * We can have multiple rules with the same number, so we 2769 * need to clear them all. 2770 */ 2771 for (rule = chain->rules; rule; rule = rule->next) 2772 if (rule->rulenum == rulenum) { 2773 while (rule && rule->rulenum == rulenum) { 2774 clear_counters(rule, log_only); 2775 rule = rule->next; 2776 } 2777 cleared = 1; 2778 break; 2779 } 2780 if (!cleared) { /* we did not find any matching rules */ 2781 IPFW_UNLOCK(chain); 2782 return (EINVAL); 2783 } 2784 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 2785 "ipfw: Entry %d cleared.\n"; 2786 } 2787 IPFW_UNLOCK(chain); 2788 2789 if (fw_verbose) 2790 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 2791 return (0); 2792} 2793 2794/* 2795 * Check validity of the structure before insert. 2796 * Fortunately rules are simple, so this mostly need to check rule sizes. 2797 */ 2798static int 2799check_ipfw_struct(struct ip_fw *rule, int size) 2800{ 2801 int l, cmdlen = 0; 2802 int have_action=0; 2803 ipfw_insn *cmd; 2804 2805 if (size < sizeof(*rule)) { 2806 printf("ipfw: rule too short\n"); 2807 return (EINVAL); 2808 } 2809 /* first, check for valid size */ 2810 l = RULESIZE(rule); 2811 if (l != size) { 2812 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 2813 return (EINVAL); 2814 } 2815 /* 2816 * Now go for the individual checks. Very simple ones, basically only 2817 * instruction sizes. 2818 */ 2819 for (l = rule->cmd_len, cmd = rule->cmd ; 2820 l > 0 ; l -= cmdlen, cmd += cmdlen) { 2821 cmdlen = F_LEN(cmd); 2822 if (cmdlen > l) { 2823 printf("ipfw: opcode %d size truncated\n", 2824 cmd->opcode); 2825 return EINVAL; 2826 } 2827 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 2828 switch (cmd->opcode) { 2829 case O_PROBE_STATE: 2830 case O_KEEP_STATE: 2831 case O_PROTO: 2832 case O_IP_SRC_ME: 2833 case O_IP_DST_ME: 2834 case O_LAYER2: 2835 case O_IN: 2836 case O_FRAG: 2837 case O_IPOPT: 2838 case O_IPTOS: 2839 case O_IPPRECEDENCE: 2840 case O_IPVER: 2841 case O_TCPWIN: 2842 case O_TCPFLAGS: 2843 case O_TCPOPTS: 2844 case O_ESTAB: 2845 case O_VERREVPATH: 2846 case O_VERSRCREACH: 2847 case O_IPSEC: 2848 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2849 goto bad_size; 2850 break; 2851 2852 case O_UID: 2853 case O_GID: 2854 case O_IP_SRC: 2855 case O_IP_DST: 2856 case O_TCPSEQ: 2857 case O_TCPACK: 2858 case O_PROB: 2859 case O_ICMPTYPE: 2860 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2861 goto bad_size; 2862 break; 2863 2864 case O_LIMIT: 2865 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 2866 goto bad_size; 2867 break; 2868 2869 case O_LOG: 2870 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 2871 goto bad_size; 2872 2873 ((ipfw_insn_log *)cmd)->log_left = 2874 ((ipfw_insn_log *)cmd)->max_log; 2875 2876 break; 2877 2878 case O_IP_SRC_MASK: 2879 case O_IP_DST_MASK: 2880 /* only odd command lengths */ 2881 if ( !(cmdlen & 1) || cmdlen > 31) 2882 goto bad_size; 2883 break; 2884 2885 case O_IP_SRC_SET: 2886 case O_IP_DST_SET: 2887 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 2888 printf("ipfw: invalid set size %d\n", 2889 cmd->arg1); 2890 return EINVAL; 2891 } 2892 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 2893 (cmd->arg1+31)/32 ) 2894 goto bad_size; 2895 break; 2896 2897 case O_IP_SRC_LOOKUP: 2898 case O_IP_DST_LOOKUP: 2899 if (cmd->arg1 >= IPFW_TABLES_MAX) { 2900 printf("ipfw: invalid table number %d\n", 2901 cmd->arg1); 2902 return (EINVAL); 2903 } 2904 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 2905 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2906 goto bad_size; 2907 break; 2908 2909 case O_MACADDR2: 2910 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 2911 goto bad_size; 2912 break; 2913 2914 case O_NOP: 2915 case O_IPID: 2916 case O_IPTTL: 2917 case O_IPLEN: 2918 if (cmdlen < 1 || cmdlen > 31) 2919 goto bad_size; 2920 break; 2921 2922 case O_MAC_TYPE: 2923 case O_IP_SRCPORT: 2924 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 2925 if (cmdlen < 2 || cmdlen > 31) 2926 goto bad_size; 2927 break; 2928 2929 case O_RECV: 2930 case O_XMIT: 2931 case O_VIA: 2932 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 2933 goto bad_size; 2934 break; 2935 2936 case O_PIPE: 2937 case O_QUEUE: 2938 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 2939 goto bad_size; 2940 goto check_action; 2941 2942 case O_FORWARD_IP: 2943 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 2944 goto bad_size; 2945 goto check_action; 2946 2947 case O_FORWARD_MAC: /* XXX not implemented yet */ 2948 case O_CHECK_STATE: 2949 case O_COUNT: 2950 case O_ACCEPT: 2951 case O_DENY: 2952 case O_REJECT: 2953 case O_SKIPTO: 2954 case O_DIVERT: 2955 case O_TEE: 2956 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2957 goto bad_size; 2958check_action: 2959 if (have_action) { 2960 printf("ipfw: opcode %d, multiple actions" 2961 " not allowed\n", 2962 cmd->opcode); 2963 return EINVAL; 2964 } 2965 have_action = 1; 2966 if (l != cmdlen) { 2967 printf("ipfw: opcode %d, action must be" 2968 " last opcode\n", 2969 cmd->opcode); 2970 return EINVAL; 2971 } 2972 break; 2973 default: 2974 printf("ipfw: opcode %d, unknown opcode\n", 2975 cmd->opcode); 2976 return EINVAL; 2977 } 2978 } 2979 if (have_action == 0) { 2980 printf("ipfw: missing action\n"); 2981 return EINVAL; 2982 } 2983 return 0; 2984 2985bad_size: 2986 printf("ipfw: opcode %d size %d wrong\n", 2987 cmd->opcode, cmdlen); 2988 return EINVAL; 2989} 2990 2991/* 2992 * Copy the static and dynamic rules to the supplied buffer 2993 * and return the amount of space actually used. 2994 */ 2995static size_t 2996ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 2997{ 2998 char *bp = buf; 2999 char *ep = bp + space; 3000 struct ip_fw *rule; 3001 int i; 3002 3003 /* XXX this can take a long time and locking will block packet flow */ 3004 IPFW_LOCK(chain); 3005 for (rule = chain->rules; rule ; rule = rule->next) { 3006 /* 3007 * Verify the entry fits in the buffer in case the 3008 * rules changed between calculating buffer space and 3009 * now. This would be better done using a generation 3010 * number but should suffice for now. 3011 */ 3012 i = RULESIZE(rule); 3013 if (bp + i <= ep) { 3014 bcopy(rule, bp, i); 3015 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3016 sizeof(set_disable)); 3017 bp += i; 3018 } 3019 } 3020 IPFW_UNLOCK(chain); 3021 if (ipfw_dyn_v) { 3022 ipfw_dyn_rule *p, *last = NULL; 3023 3024 IPFW_DYN_LOCK(); 3025 for (i = 0 ; i < curr_dyn_buckets; i++) 3026 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3027 if (bp + sizeof *p <= ep) { 3028 ipfw_dyn_rule *dst = 3029 (ipfw_dyn_rule *)bp; 3030 bcopy(p, dst, sizeof *p); 3031 bcopy(&(p->rule->rulenum), &(dst->rule), 3032 sizeof(p->rule->rulenum)); 3033 /* 3034 * store a non-null value in "next". 3035 * The userland code will interpret a 3036 * NULL here as a marker 3037 * for the last dynamic rule. 3038 */ 3039 bcopy(&dst, &dst->next, sizeof(dst)); 3040 last = dst; 3041 dst->expire = 3042 TIME_LEQ(dst->expire, time_second) ? 3043 0 : dst->expire - time_second ; 3044 bp += sizeof(ipfw_dyn_rule); 3045 } 3046 } 3047 IPFW_DYN_UNLOCK(); 3048 if (last != NULL) /* mark last dynamic rule */ 3049 bzero(&last->next, sizeof(last)); 3050 } 3051 return (bp - (char *)buf); 3052} 3053 3054 3055/** 3056 * {set|get}sockopt parser. 3057 */ 3058static int 3059ipfw_ctl(struct sockopt *sopt) 3060{ 3061#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3062 int error, rule_num; 3063 size_t size; 3064 struct ip_fw *buf, *rule; 3065 u_int32_t rulenum[2]; 3066 3067 error = suser(sopt->sopt_td); 3068 if (error) 3069 return (error); 3070 3071 /* 3072 * Disallow modifications in really-really secure mode, but still allow 3073 * the logging counters to be reset. 3074 */ 3075 if (sopt->sopt_name == IP_FW_ADD || 3076 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3077#if __FreeBSD_version >= 500034 3078 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3079 if (error) 3080 return (error); 3081#else /* FreeBSD 4.x */ 3082 if (securelevel >= 3) 3083 return (EPERM); 3084#endif 3085 } 3086 3087 error = 0; 3088 3089 switch (sopt->sopt_name) { 3090 case IP_FW_GET: 3091 /* 3092 * pass up a copy of the current rules. Static rules 3093 * come first (the last of which has number IPFW_DEFAULT_RULE), 3094 * followed by a possibly empty list of dynamic rule. 3095 * The last dynamic rule has NULL in the "next" field. 3096 * 3097 * Note that the calculated size is used to bound the 3098 * amount of data returned to the user. The rule set may 3099 * change between calculating the size and returning the 3100 * data in which case we'll just return what fits. 3101 */ 3102 size = static_len; /* size of static rules */ 3103 if (ipfw_dyn_v) /* add size of dyn.rules */ 3104 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3105 3106 /* 3107 * XXX todo: if the user passes a short length just to know 3108 * how much room is needed, do not bother filling up the 3109 * buffer, just jump to the sooptcopyout. 3110 */ 3111 buf = malloc(size, M_TEMP, M_WAITOK); 3112 error = sooptcopyout(sopt, buf, 3113 ipfw_getrules(&layer3_chain, buf, size)); 3114 free(buf, M_TEMP); 3115 break; 3116 3117 case IP_FW_FLUSH: 3118 /* 3119 * Normally we cannot release the lock on each iteration. 3120 * We could do it here only because we start from the head all 3121 * the times so there is no risk of missing some entries. 3122 * On the other hand, the risk is that we end up with 3123 * a very inconsistent ruleset, so better keep the lock 3124 * around the whole cycle. 3125 * 3126 * XXX this code can be improved by resetting the head of 3127 * the list to point to the default rule, and then freeing 3128 * the old list without the need for a lock. 3129 */ 3130 3131 IPFW_LOCK(&layer3_chain); 3132 layer3_chain.reap = NULL; 3133 free_chain(&layer3_chain, 0 /* keep default rule */); 3134 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3135 IPFW_UNLOCK(&layer3_chain); 3136 if (layer3_chain.reap != NULL) 3137 reap_rules(rule); 3138 break; 3139 3140 case IP_FW_ADD: 3141 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3142 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3143 sizeof(struct ip_fw) ); 3144 if (error == 0) 3145 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3146 if (error == 0) { 3147 error = add_rule(&layer3_chain, rule); 3148 size = RULESIZE(rule); 3149 if (!error && sopt->sopt_dir == SOPT_GET) 3150 error = sooptcopyout(sopt, rule, size); 3151 } 3152 free(rule, M_TEMP); 3153 break; 3154 3155 case IP_FW_DEL: 3156 /* 3157 * IP_FW_DEL is used for deleting single rules or sets, 3158 * and (ab)used to atomically manipulate sets. Argument size 3159 * is used to distinguish between the two: 3160 * sizeof(u_int32_t) 3161 * delete single rule or set of rules, 3162 * or reassign rules (or sets) to a different set. 3163 * 2*sizeof(u_int32_t) 3164 * atomic disable/enable sets. 3165 * first u_int32_t contains sets to be disabled, 3166 * second u_int32_t contains sets to be enabled. 3167 */ 3168 error = sooptcopyin(sopt, rulenum, 3169 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3170 if (error) 3171 break; 3172 size = sopt->sopt_valsize; 3173 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3174 error = del_entry(&layer3_chain, rulenum[0]); 3175 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3176 set_disable = 3177 (set_disable | rulenum[0]) & ~rulenum[1] & 3178 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3179 else 3180 error = EINVAL; 3181 break; 3182 3183 case IP_FW_ZERO: 3184 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3185 rule_num = 0; 3186 if (sopt->sopt_val != 0) { 3187 error = sooptcopyin(sopt, &rule_num, 3188 sizeof(int), sizeof(int)); 3189 if (error) 3190 break; 3191 } 3192 error = zero_entry(&layer3_chain, rule_num, 3193 sopt->sopt_name == IP_FW_RESETLOG); 3194 break; 3195 3196 case IP_FW_TABLE_ADD: 3197 { 3198 ipfw_table_entry ent; 3199 3200 error = sooptcopyin(sopt, &ent, 3201 sizeof(ent), sizeof(ent)); 3202 if (error) 3203 break; 3204 error = add_table_entry(ent.tbl, ent.addr, 3205 ent.masklen, ent.value); 3206 } 3207 break; 3208 3209 case IP_FW_TABLE_DEL: 3210 { 3211 ipfw_table_entry ent; 3212 3213 error = sooptcopyin(sopt, &ent, 3214 sizeof(ent), sizeof(ent)); 3215 if (error) 3216 break; 3217 error = del_table_entry(ent.tbl, ent.addr, ent.masklen); 3218 } 3219 break; 3220 3221 case IP_FW_TABLE_FLUSH: 3222 { 3223 u_int16_t tbl; 3224 3225 error = sooptcopyin(sopt, &tbl, 3226 sizeof(tbl), sizeof(tbl)); 3227 if (error) 3228 break; 3229 error = flush_table(tbl); 3230 } 3231 break; 3232 3233 case IP_FW_TABLE_GETSIZE: 3234 { 3235 u_int32_t tbl, cnt; 3236 3237 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 3238 sizeof(tbl)))) 3239 break; 3240 if ((error = count_table(tbl, &cnt))) 3241 break; 3242 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 3243 } 3244 break; 3245 3246 case IP_FW_TABLE_LIST: 3247 { 3248 ipfw_table *tbl; 3249 3250 if (sopt->sopt_valsize < sizeof(*tbl)) { 3251 error = EINVAL; 3252 break; 3253 } 3254 size = sopt->sopt_valsize; 3255 tbl = malloc(size, M_TEMP, M_WAITOK); 3256 if (tbl == NULL) { 3257 error = ENOMEM; 3258 break; 3259 } 3260 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 3261 if (error) { 3262 free(tbl, M_TEMP); 3263 break; 3264 } 3265 tbl->size = (size - sizeof(*tbl)) / 3266 sizeof(ipfw_table_entry); 3267 error = dump_table(tbl); 3268 if (error) { 3269 free(tbl, M_TEMP); 3270 break; 3271 } 3272 error = sooptcopyout(sopt, tbl, size); 3273 free(tbl, M_TEMP); 3274 } 3275 break; 3276 3277 default: 3278 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 3279 error = EINVAL; 3280 } 3281 3282 return (error); 3283#undef RULE_MAXSIZE 3284} 3285 3286/** 3287 * dummynet needs a reference to the default rule, because rules can be 3288 * deleted while packets hold a reference to them. When this happens, 3289 * dummynet changes the reference to the default rule (it could well be a 3290 * NULL pointer, but this way we do not need to check for the special 3291 * case, plus here he have info on the default behaviour). 3292 */ 3293struct ip_fw *ip_fw_default_rule; 3294 3295/* 3296 * This procedure is only used to handle keepalives. It is invoked 3297 * every dyn_keepalive_period 3298 */ 3299static void 3300ipfw_tick(void * __unused unused) 3301{ 3302 int i; 3303 ipfw_dyn_rule *q; 3304 3305 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 3306 goto done; 3307 3308 IPFW_DYN_LOCK(); 3309 for (i = 0 ; i < curr_dyn_buckets ; i++) { 3310 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 3311 if (q->dyn_type == O_LIMIT_PARENT) 3312 continue; 3313 if (q->id.proto != IPPROTO_TCP) 3314 continue; 3315 if ( (q->state & BOTH_SYN) != BOTH_SYN) 3316 continue; 3317 if (TIME_LEQ( time_second+dyn_keepalive_interval, 3318 q->expire)) 3319 continue; /* too early */ 3320 if (TIME_LEQ(q->expire, time_second)) 3321 continue; /* too late, rule expired */ 3322 3323 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 3324 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 3325 } 3326 } 3327 IPFW_DYN_UNLOCK(); 3328done: 3329 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 3330} 3331 3332static int 3333ipfw_init(void) 3334{ 3335 struct ip_fw default_rule; 3336 int error; 3337 3338 layer3_chain.rules = NULL; 3339 IPFW_LOCK_INIT(&layer3_chain); 3340 IPFW_DYN_LOCK_INIT(); 3341 callout_init(&ipfw_timeout, debug_mpsafenet ? CALLOUT_MPSAFE : 0); 3342 3343 bzero(&default_rule, sizeof default_rule); 3344 3345 default_rule.act_ofs = 0; 3346 default_rule.rulenum = IPFW_DEFAULT_RULE; 3347 default_rule.cmd_len = 1; 3348 default_rule.set = RESVD_SET; 3349 3350 default_rule.cmd[0].len = 1; 3351 default_rule.cmd[0].opcode = 3352#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 3353 1 ? O_ACCEPT : 3354#endif 3355 O_DENY; 3356 3357 error = add_rule(&layer3_chain, &default_rule); 3358 if (error != 0) { 3359 printf("ipfw2: error %u initializing default rule " 3360 "(support disabled)\n", error); 3361 IPFW_DYN_LOCK_DESTROY(); 3362 IPFW_LOCK_DESTROY(&layer3_chain); 3363 return (error); 3364 } 3365 3366 ip_fw_default_rule = layer3_chain.rules; 3367 printf("ipfw2 initialized, divert %s, " 3368 "rule-based forwarding enabled, default to %s, logging ", 3369#ifdef IPDIVERT 3370 "enabled", 3371#else 3372 "disabled", 3373#endif 3374 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 3375 3376#ifdef IPFIREWALL_VERBOSE 3377 fw_verbose = 1; 3378#endif 3379#ifdef IPFIREWALL_VERBOSE_LIMIT 3380 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3381#endif 3382 if (fw_verbose == 0) 3383 printf("disabled\n"); 3384 else if (verbose_limit == 0) 3385 printf("unlimited\n"); 3386 else 3387 printf("limited to %d packets/entry by default\n", 3388 verbose_limit); 3389 3390 ip_fw_chk_ptr = ipfw_chk; 3391 ip_fw_ctl_ptr = ipfw_ctl; 3392 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 3393 3394 return (0); 3395} 3396 3397static void 3398ipfw_destroy(void) 3399{ 3400 struct ip_fw *reap; 3401 3402 IPFW_LOCK(&layer3_chain); 3403 callout_stop(&ipfw_timeout); 3404 ip_fw_chk_ptr = NULL; 3405 ip_fw_ctl_ptr = NULL; 3406 layer3_chain.reap = NULL; 3407 free_chain(&layer3_chain, 1 /* kill default rule */); 3408 reap = layer3_chain.reap, layer3_chain.reap = NULL; 3409 IPFW_UNLOCK(&layer3_chain); 3410 if (reap != NULL) 3411 reap_rules(reap); 3412 flush_tables(); 3413 IPFW_DYN_LOCK_DESTROY(); 3414 IPFW_LOCK_DESTROY(&layer3_chain); 3415 printf("IP firewall unloaded\n"); 3416} 3417 3418static int 3419ipfw_modevent(module_t mod, int type, void *unused) 3420{ 3421 int err = 0; 3422 3423 switch (type) { 3424 case MOD_LOAD: 3425 if (IPFW_LOADED) { 3426 printf("IP firewall already loaded\n"); 3427 err = EEXIST; 3428 } else { 3429 err = ipfw_init(); 3430 } 3431 break; 3432 3433 case MOD_UNLOAD: 3434 ipfw_destroy(); 3435 err = 0; 3436 break; 3437 default: 3438 return EOPNOTSUPP; 3439 break; 3440 } 3441 return err; 3442} 3443 3444static moduledata_t ipfwmod = { 3445 "ipfw", 3446 ipfw_modevent, 3447 0 3448}; 3449DECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3450MODULE_VERSION(ipfw, 1); 3451 3452/* Must be run after route_init(). */ 3453SYSINIT(ipfw, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, init_tables, 0) 3454 3455#endif /* IPFW2 */ 3456