ip_fw2.c revision 137139
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 137139 2004-11-02 22:22:22Z 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_inet.h" 39#include "opt_ipsec.h" 40#ifndef INET 41#error IPFIREWALL requires INET. 42#endif /* INET */ 43#endif 44 45#define IPFW2 1 46#if IPFW2 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/malloc.h> 50#include <sys/mbuf.h> 51#include <sys/kernel.h> 52#include <sys/jail.h> 53#include <sys/module.h> 54#include <sys/proc.h> 55#include <sys/socket.h> 56#include <sys/socketvar.h> 57#include <sys/sysctl.h> 58#include <sys/syslog.h> 59#include <sys/ucred.h> 60#include <net/if.h> 61#include <net/radix.h> 62#include <net/route.h> 63#include <netinet/in.h> 64#include <netinet/in_systm.h> 65#include <netinet/in_var.h> 66#include <netinet/in_pcb.h> 67#include <netinet/ip.h> 68#include <netinet/ip_var.h> 69#include <netinet/ip_icmp.h> 70#include <netinet/ip_fw.h> 71#include <netinet/ip_divert.h> 72#include <netinet/ip_dummynet.h> 73#include <netinet/tcp.h> 74#include <netinet/tcp_timer.h> 75#include <netinet/tcp_var.h> 76#include <netinet/tcpip.h> 77#include <netinet/udp.h> 78#include <netinet/udp_var.h> 79#include <altq/if_altq.h> 80 81#ifdef IPSEC 82#include <netinet6/ipsec.h> 83#endif 84 85#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 86 87#include <machine/in_cksum.h> /* XXX for in_cksum */ 88 89/* 90 * set_disable contains one bit per set value (0..31). 91 * If the bit is set, all rules with the corresponding set 92 * are disabled. Set RESVD_SET(31) is reserved for the default rule 93 * and rules that are not deleted by the flush command, 94 * and CANNOT be disabled. 95 * Rules in set RESVD_SET can only be deleted explicitly. 96 */ 97static u_int32_t set_disable; 98 99static int fw_verbose; 100static int verbose_limit; 101 102static struct callout ipfw_timeout; 103#define IPFW_DEFAULT_RULE 65535 104 105/* 106 * Data structure to cache our ucred related 107 * information. This structure only gets used if 108 * the user specified UID/GID based constraints in 109 * a firewall rule. 110 */ 111struct ip_fw_ugid { 112 gid_t fw_groups[NGROUPS]; 113 int fw_ngroups; 114 uid_t fw_uid; 115 int fw_prid; 116}; 117 118struct ip_fw_chain { 119 struct ip_fw *rules; /* list of rules */ 120 struct ip_fw *reap; /* list of rules to reap */ 121 struct mtx mtx; /* lock guarding rule list */ 122}; 123#define IPFW_LOCK_INIT(_chain) \ 124 mtx_init(&(_chain)->mtx, "IPFW static rules", NULL, \ 125 MTX_DEF | MTX_RECURSE) 126#define IPFW_LOCK_DESTROY(_chain) mtx_destroy(&(_chain)->mtx) 127#define IPFW_LOCK(_chain) mtx_lock(&(_chain)->mtx) 128#define IPFW_UNLOCK(_chain) mtx_unlock(&(_chain)->mtx) 129#define IPFW_LOCK_ASSERT(_chain) do { \ 130 mtx_assert(&(_chain)->mtx, MA_OWNED); \ 131 NET_ASSERT_GIANT(); \ 132} while (0) 133 134/* 135 * list of rules for layer 3 136 */ 137static struct ip_fw_chain layer3_chain; 138 139MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 140MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 141 142struct table_entry { 143 struct radix_node rn[2]; 144 struct sockaddr_in addr, mask; 145 u_int32_t value; 146}; 147 148#define IPFW_TABLES_MAX 128 149static struct { 150 struct radix_node_head *rnh; 151 int modified; 152} ipfw_tables[IPFW_TABLES_MAX]; 153 154static int fw_debug = 1; 155static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 156 157#ifdef SYSCTL_NODE 158SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 159SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 160 CTLFLAG_RW | CTLFLAG_SECURE3, 161 &fw_enable, 0, "Enable ipfw"); 162SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 163 &autoinc_step, 0, "Rule number autincrement step"); 164SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 165 CTLFLAG_RW | CTLFLAG_SECURE3, 166 &fw_one_pass, 0, 167 "Only do a single pass through ipfw when using dummynet(4)"); 168SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 169 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 170SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 171 CTLFLAG_RW | CTLFLAG_SECURE3, 172 &fw_verbose, 0, "Log matches to ipfw rules"); 173SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 174 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 175 176/* 177 * Description of dynamic rules. 178 * 179 * Dynamic rules are stored in lists accessed through a hash table 180 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 181 * be modified through the sysctl variable dyn_buckets which is 182 * updated when the table becomes empty. 183 * 184 * XXX currently there is only one list, ipfw_dyn. 185 * 186 * When a packet is received, its address fields are first masked 187 * with the mask defined for the rule, then hashed, then matched 188 * against the entries in the corresponding list. 189 * Dynamic rules can be used for different purposes: 190 * + stateful rules; 191 * + enforcing limits on the number of sessions; 192 * + in-kernel NAT (not implemented yet) 193 * 194 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 195 * measured in seconds and depending on the flags. 196 * 197 * The total number of dynamic rules is stored in dyn_count. 198 * The max number of dynamic rules is dyn_max. When we reach 199 * the maximum number of rules we do not create anymore. This is 200 * done to avoid consuming too much memory, but also too much 201 * time when searching on each packet (ideally, we should try instead 202 * to put a limit on the length of the list on each bucket...). 203 * 204 * Each dynamic rule holds a pointer to the parent ipfw rule so 205 * we know what action to perform. Dynamic rules are removed when 206 * the parent rule is deleted. XXX we should make them survive. 207 * 208 * There are some limitations with dynamic rules -- we do not 209 * obey the 'randomized match', and we do not do multiple 210 * passes through the firewall. XXX check the latter!!! 211 */ 212static ipfw_dyn_rule **ipfw_dyn_v = NULL; 213static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 214static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 215 216static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 217#define IPFW_DYN_LOCK_INIT() \ 218 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 219#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 220#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 221#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 222#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 223 224/* 225 * Timeouts for various events in handing dynamic rules. 226 */ 227static u_int32_t dyn_ack_lifetime = 300; 228static u_int32_t dyn_syn_lifetime = 20; 229static u_int32_t dyn_fin_lifetime = 1; 230static u_int32_t dyn_rst_lifetime = 1; 231static u_int32_t dyn_udp_lifetime = 10; 232static u_int32_t dyn_short_lifetime = 5; 233 234/* 235 * Keepalives are sent if dyn_keepalive is set. They are sent every 236 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 237 * seconds of lifetime of a rule. 238 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 239 * than dyn_keepalive_period. 240 */ 241 242static u_int32_t dyn_keepalive_interval = 20; 243static u_int32_t dyn_keepalive_period = 5; 244static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 245 246static u_int32_t static_count; /* # of static rules */ 247static u_int32_t static_len; /* size in bytes of static rules */ 248static u_int32_t dyn_count; /* # of dynamic rules */ 249static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 250 251SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 252 &dyn_buckets, 0, "Number of dyn. buckets"); 253SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 254 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 255SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 256 &dyn_count, 0, "Number of dyn. rules"); 257SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 258 &dyn_max, 0, "Max number of dyn. rules"); 259SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 260 &static_count, 0, "Number of static rules"); 261SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 262 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 263SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 264 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 265SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 266 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 267SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 268 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 269SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 270 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 271SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 272 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 273SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 274 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 275 276#endif /* SYSCTL_NODE */ 277 278 279/* 280 * This macro maps an ip pointer into a layer3 header pointer of type T 281 */ 282#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 283 284static __inline int 285icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 286{ 287 int type = L3HDR(struct icmp,ip)->icmp_type; 288 289 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 290} 291 292#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 293 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 294 295static int 296is_icmp_query(struct ip *ip) 297{ 298 int type = L3HDR(struct icmp, ip)->icmp_type; 299 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 300} 301#undef TT 302 303/* 304 * The following checks use two arrays of 8 or 16 bits to store the 305 * bits that we want set or clear, respectively. They are in the 306 * low and high half of cmd->arg1 or cmd->d[0]. 307 * 308 * We scan options and store the bits we find set. We succeed if 309 * 310 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 311 * 312 * The code is sometimes optimized not to store additional variables. 313 */ 314 315static int 316flags_match(ipfw_insn *cmd, u_int8_t bits) 317{ 318 u_char want_clear; 319 bits = ~bits; 320 321 if ( ((cmd->arg1 & 0xff) & bits) != 0) 322 return 0; /* some bits we want set were clear */ 323 want_clear = (cmd->arg1 >> 8) & 0xff; 324 if ( (want_clear & bits) != want_clear) 325 return 0; /* some bits we want clear were set */ 326 return 1; 327} 328 329static int 330ipopts_match(struct ip *ip, ipfw_insn *cmd) 331{ 332 int optlen, bits = 0; 333 u_char *cp = (u_char *)(ip + 1); 334 int x = (ip->ip_hl << 2) - sizeof (struct ip); 335 336 for (; x > 0; x -= optlen, cp += optlen) { 337 int opt = cp[IPOPT_OPTVAL]; 338 339 if (opt == IPOPT_EOL) 340 break; 341 if (opt == IPOPT_NOP) 342 optlen = 1; 343 else { 344 optlen = cp[IPOPT_OLEN]; 345 if (optlen <= 0 || optlen > x) 346 return 0; /* invalid or truncated */ 347 } 348 switch (opt) { 349 350 default: 351 break; 352 353 case IPOPT_LSRR: 354 bits |= IP_FW_IPOPT_LSRR; 355 break; 356 357 case IPOPT_SSRR: 358 bits |= IP_FW_IPOPT_SSRR; 359 break; 360 361 case IPOPT_RR: 362 bits |= IP_FW_IPOPT_RR; 363 break; 364 365 case IPOPT_TS: 366 bits |= IP_FW_IPOPT_TS; 367 break; 368 } 369 } 370 return (flags_match(cmd, bits)); 371} 372 373static int 374tcpopts_match(struct ip *ip, ipfw_insn *cmd) 375{ 376 int optlen, bits = 0; 377 struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 378 u_char *cp = (u_char *)(tcp + 1); 379 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 380 381 for (; x > 0; x -= optlen, cp += optlen) { 382 int opt = cp[0]; 383 if (opt == TCPOPT_EOL) 384 break; 385 if (opt == TCPOPT_NOP) 386 optlen = 1; 387 else { 388 optlen = cp[1]; 389 if (optlen <= 0) 390 break; 391 } 392 393 switch (opt) { 394 395 default: 396 break; 397 398 case TCPOPT_MAXSEG: 399 bits |= IP_FW_TCPOPT_MSS; 400 break; 401 402 case TCPOPT_WINDOW: 403 bits |= IP_FW_TCPOPT_WINDOW; 404 break; 405 406 case TCPOPT_SACK_PERMITTED: 407 case TCPOPT_SACK: 408 bits |= IP_FW_TCPOPT_SACK; 409 break; 410 411 case TCPOPT_TIMESTAMP: 412 bits |= IP_FW_TCPOPT_TS; 413 break; 414 415 } 416 } 417 return (flags_match(cmd, bits)); 418} 419 420static int 421iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 422{ 423 if (ifp == NULL) /* no iface with this packet, match fails */ 424 return 0; 425 /* Check by name or by IP address */ 426 if (cmd->name[0] != '\0') { /* match by name */ 427 /* Check name */ 428 if (cmd->p.glob) { 429 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 430 return(1); 431 } else { 432 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 433 return(1); 434 } 435 } else { 436 struct ifaddr *ia; 437 438 /* XXX lock? */ 439 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 440 if (ia->ifa_addr == NULL) 441 continue; 442 if (ia->ifa_addr->sa_family != AF_INET) 443 continue; 444 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 445 (ia->ifa_addr))->sin_addr.s_addr) 446 return(1); /* match */ 447 } 448 } 449 return(0); /* no match, fail ... */ 450} 451 452/* 453 * The verify_path function checks if a route to the src exists and 454 * if it is reachable via ifp (when provided). 455 * 456 * The 'verrevpath' option checks that the interface that an IP packet 457 * arrives on is the same interface that traffic destined for the 458 * packet's source address would be routed out of. The 'versrcreach' 459 * option just checks that the source address is reachable via any route 460 * (except default) in the routing table. These two are a measure to block 461 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 462 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 463 * is purposely reminiscent of the Cisco IOS command, 464 * 465 * ip verify unicast reverse-path 466 * ip verify unicast source reachable-via any 467 * 468 * which implements the same functionality. But note that syntax is 469 * misleading. The check may be performed on all IP packets whether unicast, 470 * multicast, or broadcast. 471 */ 472static int 473verify_path(struct in_addr src, struct ifnet *ifp) 474{ 475 struct route ro; 476 struct sockaddr_in *dst; 477 478 bzero(&ro, sizeof(ro)); 479 480 dst = (struct sockaddr_in *)&(ro.ro_dst); 481 dst->sin_family = AF_INET; 482 dst->sin_len = sizeof(*dst); 483 dst->sin_addr = src; 484 rtalloc_ign(&ro, RTF_CLONING); 485 486 if (ro.ro_rt == NULL) 487 return 0; 488 489 /* if ifp is provided, check for equality with rtentry */ 490 if (ifp != NULL && ro.ro_rt->rt_ifp != ifp) { 491 RTFREE(ro.ro_rt); 492 return 0; 493 } 494 495 /* if no ifp provided, check if rtentry is not default route */ 496 if (ifp == NULL && 497 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 498 RTFREE(ro.ro_rt); 499 return 0; 500 } 501 502 /* or if this is a blackhole/reject route */ 503 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 504 RTFREE(ro.ro_rt); 505 return 0; 506 } 507 508 /* found valid route */ 509 RTFREE(ro.ro_rt); 510 return 1; 511} 512 513 514static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 515 516#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 517#define SNP(buf) buf, sizeof(buf) 518 519/* 520 * We enter here when we have a rule with O_LOG. 521 * XXX this function alone takes about 2Kbytes of code! 522 */ 523static void 524ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 525 struct mbuf *m, struct ifnet *oif) 526{ 527 char *action; 528 int limit_reached = 0; 529 char action2[40], proto[48], fragment[28]; 530 531 fragment[0] = '\0'; 532 proto[0] = '\0'; 533 534 if (f == NULL) { /* bogus pkt */ 535 if (verbose_limit != 0 && norule_counter >= verbose_limit) 536 return; 537 norule_counter++; 538 if (norule_counter == verbose_limit) 539 limit_reached = verbose_limit; 540 action = "Refuse"; 541 } else { /* O_LOG is the first action, find the real one */ 542 ipfw_insn *cmd = ACTION_PTR(f); 543 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 544 545 if (l->max_log != 0 && l->log_left == 0) 546 return; 547 l->log_left--; 548 if (l->log_left == 0) 549 limit_reached = l->max_log; 550 cmd += F_LEN(cmd); /* point to first action */ 551 if (cmd->opcode == O_ALTQ) { 552 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 553 554 snprintf(SNPARGS(action2, 0), "Altq %d", 555 altq->qid); 556 cmd += F_LEN(cmd); 557 } 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_ALTQ) 1330 cmd += F_LEN(cmd); 1331 if ( cmd->opcode == O_SKIPTO ) 1332 for (rule = me->next; rule ; rule = rule->next) 1333 if (rule->rulenum >= cmd->arg1) 1334 break; 1335 if (rule == NULL) /* failure or not a skipto */ 1336 rule = me->next; 1337 me->next_rule = rule; 1338 return rule; 1339} 1340 1341static void 1342init_tables(void) 1343{ 1344 int i; 1345 1346 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1347 rn_inithead((void **)&ipfw_tables[i].rnh, 32); 1348 ipfw_tables[i].modified = 1; 1349 } 1350} 1351 1352static int 1353add_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen, u_int32_t value) 1354{ 1355 struct radix_node_head *rnh; 1356 struct table_entry *ent; 1357 1358 if (tbl >= IPFW_TABLES_MAX) 1359 return (EINVAL); 1360 rnh = ipfw_tables[tbl].rnh; 1361 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1362 if (ent == NULL) 1363 return (ENOMEM); 1364 ent->value = value; 1365 ent->addr.sin_len = ent->mask.sin_len = 8; 1366 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1367 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1368 RADIX_NODE_HEAD_LOCK(rnh); 1369 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1370 NULL) { 1371 RADIX_NODE_HEAD_UNLOCK(rnh); 1372 free(ent, M_IPFW_TBL); 1373 return (EEXIST); 1374 } 1375 ipfw_tables[tbl].modified = 1; 1376 RADIX_NODE_HEAD_UNLOCK(rnh); 1377 return (0); 1378} 1379 1380static int 1381del_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen) 1382{ 1383 struct radix_node_head *rnh; 1384 struct table_entry *ent; 1385 struct sockaddr_in sa, mask; 1386 1387 if (tbl >= IPFW_TABLES_MAX) 1388 return (EINVAL); 1389 rnh = ipfw_tables[tbl].rnh; 1390 sa.sin_len = mask.sin_len = 8; 1391 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1392 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1393 RADIX_NODE_HEAD_LOCK(rnh); 1394 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1395 if (ent == NULL) { 1396 RADIX_NODE_HEAD_UNLOCK(rnh); 1397 return (ESRCH); 1398 } 1399 ipfw_tables[tbl].modified = 1; 1400 RADIX_NODE_HEAD_UNLOCK(rnh); 1401 free(ent, M_IPFW_TBL); 1402 return (0); 1403} 1404 1405static int 1406flush_table_entry(struct radix_node *rn, void *arg) 1407{ 1408 struct radix_node_head * const rnh = arg; 1409 struct table_entry *ent; 1410 1411 ent = (struct table_entry *) 1412 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1413 if (ent != NULL) 1414 free(ent, M_IPFW_TBL); 1415 return (0); 1416} 1417 1418static int 1419flush_table(u_int16_t tbl) 1420{ 1421 struct radix_node_head *rnh; 1422 1423 if (tbl >= IPFW_TABLES_MAX) 1424 return (EINVAL); 1425 rnh = ipfw_tables[tbl].rnh; 1426 RADIX_NODE_HEAD_LOCK(rnh); 1427 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1428 ipfw_tables[tbl].modified = 1; 1429 RADIX_NODE_HEAD_UNLOCK(rnh); 1430 return (0); 1431} 1432 1433static void 1434flush_tables(void) 1435{ 1436 u_int16_t tbl; 1437 1438 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1439 flush_table(tbl); 1440} 1441 1442static int 1443lookup_table(u_int16_t tbl, in_addr_t addr, u_int32_t *val) 1444{ 1445 struct radix_node_head *rnh; 1446 struct table_entry *ent; 1447 struct sockaddr_in sa; 1448 static in_addr_t last_addr; 1449 static int last_tbl; 1450 static int last_match; 1451 static u_int32_t last_value; 1452 1453 if (tbl >= IPFW_TABLES_MAX) 1454 return (0); 1455 if (tbl == last_tbl && addr == last_addr && 1456 !ipfw_tables[tbl].modified) { 1457 if (last_match) 1458 *val = last_value; 1459 return (last_match); 1460 } 1461 rnh = ipfw_tables[tbl].rnh; 1462 sa.sin_len = 8; 1463 sa.sin_addr.s_addr = addr; 1464 RADIX_NODE_HEAD_LOCK(rnh); 1465 ipfw_tables[tbl].modified = 0; 1466 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1467 RADIX_NODE_HEAD_UNLOCK(rnh); 1468 last_addr = addr; 1469 last_tbl = tbl; 1470 if (ent != NULL) { 1471 last_value = *val = ent->value; 1472 last_match = 1; 1473 return (1); 1474 } 1475 last_match = 0; 1476 return (0); 1477} 1478 1479static int 1480count_table_entry(struct radix_node *rn, void *arg) 1481{ 1482 u_int32_t * const cnt = arg; 1483 1484 (*cnt)++; 1485 return (0); 1486} 1487 1488static int 1489count_table(u_int32_t tbl, u_int32_t *cnt) 1490{ 1491 struct radix_node_head *rnh; 1492 1493 if (tbl >= IPFW_TABLES_MAX) 1494 return (EINVAL); 1495 rnh = ipfw_tables[tbl].rnh; 1496 *cnt = 0; 1497 RADIX_NODE_HEAD_LOCK(rnh); 1498 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1499 RADIX_NODE_HEAD_UNLOCK(rnh); 1500 return (0); 1501} 1502 1503static int 1504dump_table_entry(struct radix_node *rn, void *arg) 1505{ 1506 struct table_entry * const n = (struct table_entry *)rn; 1507 ipfw_table * const tbl = arg; 1508 ipfw_table_entry *ent; 1509 1510 if (tbl->cnt == tbl->size) 1511 return (1); 1512 ent = &tbl->ent[tbl->cnt]; 1513 ent->tbl = tbl->tbl; 1514 if (in_nullhost(n->mask.sin_addr)) 1515 ent->masklen = 0; 1516 else 1517 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1518 ent->addr = n->addr.sin_addr.s_addr; 1519 ent->value = n->value; 1520 tbl->cnt++; 1521 return (0); 1522} 1523 1524static int 1525dump_table(ipfw_table *tbl) 1526{ 1527 struct radix_node_head *rnh; 1528 1529 if (tbl->tbl >= IPFW_TABLES_MAX) 1530 return (EINVAL); 1531 rnh = ipfw_tables[tbl->tbl].rnh; 1532 tbl->cnt = 0; 1533 RADIX_NODE_HEAD_LOCK(rnh); 1534 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1535 RADIX_NODE_HEAD_UNLOCK(rnh); 1536 return (0); 1537} 1538 1539static void 1540fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1541{ 1542 struct ucred *cr; 1543 1544 if (inp->inp_socket != NULL) { 1545 cr = inp->inp_socket->so_cred; 1546 ugp->fw_prid = jailed(cr) ? 1547 cr->cr_prison->pr_id : -1; 1548 ugp->fw_uid = cr->cr_uid; 1549 ugp->fw_ngroups = cr->cr_ngroups; 1550 bcopy(cr->cr_groups, ugp->fw_groups, 1551 sizeof(ugp->fw_groups)); 1552 } 1553} 1554 1555static int 1556check_uidgid(ipfw_insn_u32 *insn, 1557 int proto, struct ifnet *oif, 1558 struct in_addr dst_ip, u_int16_t dst_port, 1559 struct in_addr src_ip, u_int16_t src_port, 1560 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1561{ 1562 struct inpcbinfo *pi; 1563 int wildcard; 1564 struct inpcb *pcb; 1565 int match; 1566 gid_t *gp; 1567 1568 /* 1569 * Check to see if the UDP or TCP stack supplied us with 1570 * the PCB. If so, rather then holding a lock and looking 1571 * up the PCB, we can use the one that was supplied. 1572 */ 1573 if (inp && *lookup == 0) { 1574 INP_LOCK_ASSERT(inp); 1575 if (inp->inp_socket != NULL) { 1576 fill_ugid_cache(inp, ugp); 1577 *lookup = 1; 1578 } 1579 } 1580 /* 1581 * If we have already been here and the packet has no 1582 * PCB entry associated with it, then we can safely 1583 * assume that this is a no match. 1584 */ 1585 if (*lookup == -1) 1586 return (0); 1587 if (proto == IPPROTO_TCP) { 1588 wildcard = 0; 1589 pi = &tcbinfo; 1590 } else if (proto == IPPROTO_UDP) { 1591 wildcard = 1; 1592 pi = &udbinfo; 1593 } else 1594 return 0; 1595 match = 0; 1596 if (*lookup == 0) { 1597 INP_INFO_RLOCK(pi); 1598 pcb = (oif) ? 1599 in_pcblookup_hash(pi, 1600 dst_ip, htons(dst_port), 1601 src_ip, htons(src_port), 1602 wildcard, oif) : 1603 in_pcblookup_hash(pi, 1604 src_ip, htons(src_port), 1605 dst_ip, htons(dst_port), 1606 wildcard, NULL); 1607 if (pcb != NULL) { 1608 INP_LOCK(pcb); 1609 if (pcb->inp_socket != NULL) { 1610 fill_ugid_cache(pcb, ugp); 1611 *lookup = 1; 1612 } 1613 INP_UNLOCK(pcb); 1614 } 1615 INP_INFO_RUNLOCK(pi); 1616 if (*lookup == 0) { 1617 /* 1618 * If the lookup did not yield any results, there 1619 * is no sense in coming back and trying again. So 1620 * we can set lookup to -1 and ensure that we wont 1621 * bother the pcb system again. 1622 */ 1623 *lookup = -1; 1624 return (0); 1625 } 1626 } 1627 if (insn->o.opcode == O_UID) 1628 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1629 else if (insn->o.opcode == O_GID) { 1630 for (gp = ugp->fw_groups; 1631 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1632 if (*gp == (gid_t)insn->d[0]) { 1633 match = 1; 1634 break; 1635 } 1636 } else if (insn->o.opcode == O_JAIL) 1637 match = (ugp->fw_prid == (int)insn->d[0]); 1638 return match; 1639} 1640 1641/* 1642 * The main check routine for the firewall. 1643 * 1644 * All arguments are in args so we can modify them and return them 1645 * back to the caller. 1646 * 1647 * Parameters: 1648 * 1649 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1650 * Starts with the IP header. 1651 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1652 * args->oif Outgoing interface, or NULL if packet is incoming. 1653 * The incoming interface is in the mbuf. (in) 1654 * args->divert_rule (in/out) 1655 * Skip up to the first rule past this rule number; 1656 * upon return, non-zero port number for divert or tee. 1657 * 1658 * args->rule Pointer to the last matching rule (in/out) 1659 * args->next_hop Socket we are forwarding to (out). 1660 * args->f_id Addresses grabbed from the packet (out) 1661 * 1662 * Return value: 1663 * 1664 * IP_FW_PORT_DENY_FLAG the packet must be dropped. 1665 * 0 The packet is to be accepted and routed normally OR 1666 * the packet was denied/rejected and has been dropped; 1667 * in the latter case, *m is equal to NULL upon return. 1668 * port Divert the packet to port, with these caveats: 1669 * 1670 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead 1671 * of diverting it (ie, 'ipfw tee'). 1672 * 1673 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower 1674 * 16 bits as a dummynet pipe number instead of diverting 1675 */ 1676 1677int 1678ipfw_chk(struct ip_fw_args *args) 1679{ 1680 /* 1681 * Local variables hold state during the processing of a packet. 1682 * 1683 * IMPORTANT NOTE: to speed up the processing of rules, there 1684 * are some assumption on the values of the variables, which 1685 * are documented here. Should you change them, please check 1686 * the implementation of the various instructions to make sure 1687 * that they still work. 1688 * 1689 * args->eh The MAC header. It is non-null for a layer2 1690 * packet, it is NULL for a layer-3 packet. 1691 * 1692 * m | args->m Pointer to the mbuf, as received from the caller. 1693 * It may change if ipfw_chk() does an m_pullup, or if it 1694 * consumes the packet because it calls send_reject(). 1695 * XXX This has to change, so that ipfw_chk() never modifies 1696 * or consumes the buffer. 1697 * ip is simply an alias of the value of m, and it is kept 1698 * in sync with it (the packet is supposed to start with 1699 * the ip header). 1700 */ 1701 struct mbuf *m = args->m; 1702 struct ip *ip = mtod(m, struct ip *); 1703 1704 /* 1705 * For rules which contain uid/gid or jail constraints, cache 1706 * a copy of the users credentials after the pcb lookup has been 1707 * executed. This will speed up the processing of rules with 1708 * these types of constraints, as well as decrease contention 1709 * on pcb related locks. 1710 */ 1711 struct ip_fw_ugid fw_ugid_cache; 1712 int ugid_lookup = 0; 1713 1714 /* 1715 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 1716 * associated with a packet input on a divert socket. This 1717 * will allow to distinguish traffic and its direction when 1718 * it originates from a divert socket. 1719 */ 1720 u_int divinput_flags = 0; 1721 1722 /* 1723 * oif | args->oif If NULL, ipfw_chk has been called on the 1724 * inbound path (ether_input, bdg_forward, ip_input). 1725 * If non-NULL, ipfw_chk has been called on the outbound path 1726 * (ether_output, ip_output). 1727 */ 1728 struct ifnet *oif = args->oif; 1729 1730 struct ip_fw *f = NULL; /* matching rule */ 1731 int retval = 0; 1732 1733 /* 1734 * hlen The length of the IPv4 header. 1735 * hlen >0 means we have an IPv4 packet. 1736 */ 1737 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1738 1739 /* 1740 * offset The offset of a fragment. offset != 0 means that 1741 * we have a fragment at this offset of an IPv4 packet. 1742 * offset == 0 means that (if this is an IPv4 packet) 1743 * this is the first or only fragment. 1744 */ 1745 u_short offset = 0; 1746 1747 /* 1748 * Local copies of addresses. They are only valid if we have 1749 * an IP packet. 1750 * 1751 * proto The protocol. Set to 0 for non-ip packets, 1752 * or to the protocol read from the packet otherwise. 1753 * proto != 0 means that we have an IPv4 packet. 1754 * 1755 * src_port, dst_port port numbers, in HOST format. Only 1756 * valid for TCP and UDP packets. 1757 * 1758 * src_ip, dst_ip ip addresses, in NETWORK format. 1759 * Only valid for IPv4 packets. 1760 */ 1761 u_int8_t proto; 1762 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1763 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1764 u_int16_t ip_len=0; 1765 int pktlen; 1766 int dyn_dir = MATCH_UNKNOWN; 1767 ipfw_dyn_rule *q = NULL; 1768 struct ip_fw_chain *chain = &layer3_chain; 1769 struct m_tag *mtag; 1770 1771 if (m->m_flags & M_SKIP_FIREWALL) 1772 return 0; /* accept */ 1773 /* 1774 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1775 * MATCH_NONE when checked and not matched (q = NULL), 1776 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1777 */ 1778 1779 pktlen = m->m_pkthdr.len; 1780 if (args->eh == NULL || /* layer 3 packet */ 1781 ( m->m_pkthdr.len >= sizeof(struct ip) && 1782 ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 1783 hlen = ip->ip_hl << 2; 1784 1785 /* 1786 * Collect parameters into local variables for faster matching. 1787 */ 1788 if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 1789 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1790 goto after_ip_checks; 1791 } 1792 1793 proto = args->f_id.proto = ip->ip_p; 1794 src_ip = ip->ip_src; 1795 dst_ip = ip->ip_dst; 1796 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1797 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1798 ip_len = ntohs(ip->ip_len); 1799 } else { 1800 offset = ip->ip_off & IP_OFFMASK; 1801 ip_len = ip->ip_len; 1802 } 1803 pktlen = ip_len < pktlen ? ip_len : pktlen; 1804 1805#define PULLUP_TO(len) \ 1806 do { \ 1807 if ((m)->m_len < (len)) { \ 1808 args->m = m = m_pullup(m, (len)); \ 1809 if (m == 0) \ 1810 goto pullup_failed; \ 1811 ip = mtod(m, struct ip *); \ 1812 } \ 1813 } while (0) 1814 1815 if (offset == 0) { 1816 switch (proto) { 1817 case IPPROTO_TCP: 1818 { 1819 struct tcphdr *tcp; 1820 1821 PULLUP_TO(hlen + sizeof(struct tcphdr)); 1822 tcp = L3HDR(struct tcphdr, ip); 1823 dst_port = tcp->th_dport; 1824 src_port = tcp->th_sport; 1825 args->f_id.flags = tcp->th_flags; 1826 } 1827 break; 1828 1829 case IPPROTO_UDP: 1830 { 1831 struct udphdr *udp; 1832 1833 PULLUP_TO(hlen + sizeof(struct udphdr)); 1834 udp = L3HDR(struct udphdr, ip); 1835 dst_port = udp->uh_dport; 1836 src_port = udp->uh_sport; 1837 } 1838 break; 1839 1840 case IPPROTO_ICMP: 1841 PULLUP_TO(hlen + 4); /* type, code and checksum. */ 1842 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 1843 break; 1844 1845 default: 1846 break; 1847 } 1848#undef PULLUP_TO 1849 } 1850 1851 args->f_id.src_ip = ntohl(src_ip.s_addr); 1852 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1853 args->f_id.src_port = src_port = ntohs(src_port); 1854 args->f_id.dst_port = dst_port = ntohs(dst_port); 1855 1856after_ip_checks: 1857 IPFW_LOCK(chain); /* XXX expensive? can we run lock free? */ 1858 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 1859 if (args->rule) { 1860 /* 1861 * Packet has already been tagged. Look for the next rule 1862 * to restart processing. 1863 * 1864 * If fw_one_pass != 0 then just accept it. 1865 * XXX should not happen here, but optimized out in 1866 * the caller. 1867 */ 1868 if (fw_one_pass) { 1869 IPFW_UNLOCK(chain); /* XXX optimize */ 1870 return 0; 1871 } 1872 1873 f = args->rule->next_rule; 1874 if (f == NULL) 1875 f = lookup_next_rule(args->rule); 1876 } else { 1877 /* 1878 * Find the starting rule. It can be either the first 1879 * one, or the one after divert_rule if asked so. 1880 */ 1881 int skipto = mtag ? divert_cookie(mtag) : 0; 1882 1883 f = chain->rules; 1884 if (args->eh == NULL && skipto != 0) { 1885 if (skipto >= IPFW_DEFAULT_RULE) { 1886 IPFW_UNLOCK(chain); 1887 return(IP_FW_PORT_DENY_FLAG); /* invalid */ 1888 } 1889 while (f && f->rulenum <= skipto) 1890 f = f->next; 1891 if (f == NULL) { /* drop packet */ 1892 IPFW_UNLOCK(chain); 1893 return(IP_FW_PORT_DENY_FLAG); 1894 } 1895 } 1896 } 1897 /* reset divert rule to avoid confusion later */ 1898 if (mtag) { 1899 divinput_flags = divert_info(mtag) & 1900 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 1901 m_tag_delete(m, mtag); 1902 } 1903 1904 /* 1905 * Now scan the rules, and parse microinstructions for each rule. 1906 */ 1907 for (; f; f = f->next) { 1908 int l, cmdlen; 1909 ipfw_insn *cmd; 1910 int skip_or; /* skip rest of OR block */ 1911 1912again: 1913 if (set_disable & (1 << f->set) ) 1914 continue; 1915 1916 skip_or = 0; 1917 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1918 l -= cmdlen, cmd += cmdlen) { 1919 int match; 1920 1921 /* 1922 * check_body is a jump target used when we find a 1923 * CHECK_STATE, and need to jump to the body of 1924 * the target rule. 1925 */ 1926 1927check_body: 1928 cmdlen = F_LEN(cmd); 1929 /* 1930 * An OR block (insn_1 || .. || insn_n) has the 1931 * F_OR bit set in all but the last instruction. 1932 * The first match will set "skip_or", and cause 1933 * the following instructions to be skipped until 1934 * past the one with the F_OR bit clear. 1935 */ 1936 if (skip_or) { /* skip this instruction */ 1937 if ((cmd->len & F_OR) == 0) 1938 skip_or = 0; /* next one is good */ 1939 continue; 1940 } 1941 match = 0; /* set to 1 if we succeed */ 1942 1943 switch (cmd->opcode) { 1944 /* 1945 * The first set of opcodes compares the packet's 1946 * fields with some pattern, setting 'match' if a 1947 * match is found. At the end of the loop there is 1948 * logic to deal with F_NOT and F_OR flags associated 1949 * with the opcode. 1950 */ 1951 case O_NOP: 1952 match = 1; 1953 break; 1954 1955 case O_FORWARD_MAC: 1956 printf("ipfw: opcode %d unimplemented\n", 1957 cmd->opcode); 1958 break; 1959 1960 case O_GID: 1961 case O_UID: 1962 case O_JAIL: 1963 /* 1964 * We only check offset == 0 && proto != 0, 1965 * as this ensures that we have an IPv4 1966 * packet with the ports info. 1967 */ 1968 if (offset!=0) 1969 break; 1970 if (proto == IPPROTO_TCP || 1971 proto == IPPROTO_UDP) 1972 match = check_uidgid( 1973 (ipfw_insn_u32 *)cmd, 1974 proto, oif, 1975 dst_ip, dst_port, 1976 src_ip, src_port, &fw_ugid_cache, 1977 &ugid_lookup, args->inp); 1978 break; 1979 1980 case O_RECV: 1981 match = iface_match(m->m_pkthdr.rcvif, 1982 (ipfw_insn_if *)cmd); 1983 break; 1984 1985 case O_XMIT: 1986 match = iface_match(oif, (ipfw_insn_if *)cmd); 1987 break; 1988 1989 case O_VIA: 1990 match = iface_match(oif ? oif : 1991 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1992 break; 1993 1994 case O_MACADDR2: 1995 if (args->eh != NULL) { /* have MAC header */ 1996 u_int32_t *want = (u_int32_t *) 1997 ((ipfw_insn_mac *)cmd)->addr; 1998 u_int32_t *mask = (u_int32_t *) 1999 ((ipfw_insn_mac *)cmd)->mask; 2000 u_int32_t *hdr = (u_int32_t *)args->eh; 2001 2002 match = 2003 ( want[0] == (hdr[0] & mask[0]) && 2004 want[1] == (hdr[1] & mask[1]) && 2005 want[2] == (hdr[2] & mask[2]) ); 2006 } 2007 break; 2008 2009 case O_MAC_TYPE: 2010 if (args->eh != NULL) { 2011 u_int16_t t = 2012 ntohs(args->eh->ether_type); 2013 u_int16_t *p = 2014 ((ipfw_insn_u16 *)cmd)->ports; 2015 int i; 2016 2017 for (i = cmdlen - 1; !match && i>0; 2018 i--, p += 2) 2019 match = (t>=p[0] && t<=p[1]); 2020 } 2021 break; 2022 2023 case O_FRAG: 2024 match = (hlen > 0 && offset != 0); 2025 break; 2026 2027 case O_IN: /* "out" is "not in" */ 2028 match = (oif == NULL); 2029 break; 2030 2031 case O_LAYER2: 2032 match = (args->eh != NULL); 2033 break; 2034 2035 case O_DIVERTED: 2036 match = (cmd->arg1 & 1 && divinput_flags & 2037 IP_FW_DIVERT_LOOPBACK_FLAG) || 2038 (cmd->arg1 & 2 && divinput_flags & 2039 IP_FW_DIVERT_OUTPUT_FLAG); 2040 break; 2041 2042 case O_PROTO: 2043 /* 2044 * We do not allow an arg of 0 so the 2045 * check of "proto" only suffices. 2046 */ 2047 match = (proto == cmd->arg1); 2048 break; 2049 2050 case O_IP_SRC: 2051 match = (hlen > 0 && 2052 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2053 src_ip.s_addr); 2054 break; 2055 2056 case O_IP_SRC_LOOKUP: 2057 case O_IP_DST_LOOKUP: 2058 if (hlen > 0) { 2059 uint32_t a = 2060 (cmd->opcode == O_IP_DST_LOOKUP) ? 2061 dst_ip.s_addr : src_ip.s_addr; 2062 uint32_t v; 2063 2064 match = lookup_table(cmd->arg1, a, &v); 2065 if (!match) 2066 break; 2067 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2068 match = 2069 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2070 } 2071 break; 2072 2073 case O_IP_SRC_MASK: 2074 case O_IP_DST_MASK: 2075 if (hlen > 0) { 2076 uint32_t a = 2077 (cmd->opcode == O_IP_DST_MASK) ? 2078 dst_ip.s_addr : src_ip.s_addr; 2079 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2080 int i = cmdlen-1; 2081 2082 for (; !match && i>0; i-= 2, p+= 2) 2083 match = (p[0] == (a & p[1])); 2084 } 2085 break; 2086 2087 case O_IP_SRC_ME: 2088 if (hlen > 0) { 2089 struct ifnet *tif; 2090 2091 INADDR_TO_IFP(src_ip, tif); 2092 match = (tif != NULL); 2093 } 2094 break; 2095 2096 case O_IP_DST_SET: 2097 case O_IP_SRC_SET: 2098 if (hlen > 0) { 2099 u_int32_t *d = (u_int32_t *)(cmd+1); 2100 u_int32_t addr = 2101 cmd->opcode == O_IP_DST_SET ? 2102 args->f_id.dst_ip : 2103 args->f_id.src_ip; 2104 2105 if (addr < d[0]) 2106 break; 2107 addr -= d[0]; /* subtract base */ 2108 match = (addr < cmd->arg1) && 2109 ( d[ 1 + (addr>>5)] & 2110 (1<<(addr & 0x1f)) ); 2111 } 2112 break; 2113 2114 case O_IP_DST: 2115 match = (hlen > 0 && 2116 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2117 dst_ip.s_addr); 2118 break; 2119 2120 case O_IP_DST_ME: 2121 if (hlen > 0) { 2122 struct ifnet *tif; 2123 2124 INADDR_TO_IFP(dst_ip, tif); 2125 match = (tif != NULL); 2126 } 2127 break; 2128 2129 case O_IP_SRCPORT: 2130 case O_IP_DSTPORT: 2131 /* 2132 * offset == 0 && proto != 0 is enough 2133 * to guarantee that we have an IPv4 2134 * packet with port info. 2135 */ 2136 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2137 && offset == 0) { 2138 u_int16_t x = 2139 (cmd->opcode == O_IP_SRCPORT) ? 2140 src_port : dst_port ; 2141 u_int16_t *p = 2142 ((ipfw_insn_u16 *)cmd)->ports; 2143 int i; 2144 2145 for (i = cmdlen - 1; !match && i>0; 2146 i--, p += 2) 2147 match = (x>=p[0] && x<=p[1]); 2148 } 2149 break; 2150 2151 case O_ICMPTYPE: 2152 match = (offset == 0 && proto==IPPROTO_ICMP && 2153 icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 2154 break; 2155 2156 case O_IPOPT: 2157 match = (hlen > 0 && ipopts_match(ip, cmd) ); 2158 break; 2159 2160 case O_IPVER: 2161 match = (hlen > 0 && cmd->arg1 == ip->ip_v); 2162 break; 2163 2164 case O_IPID: 2165 case O_IPLEN: 2166 case O_IPTTL: 2167 if (hlen > 0) { /* only for IP packets */ 2168 uint16_t x; 2169 uint16_t *p; 2170 int i; 2171 2172 if (cmd->opcode == O_IPLEN) 2173 x = ip_len; 2174 else if (cmd->opcode == O_IPTTL) 2175 x = ip->ip_ttl; 2176 else /* must be IPID */ 2177 x = ntohs(ip->ip_id); 2178 if (cmdlen == 1) { 2179 match = (cmd->arg1 == x); 2180 break; 2181 } 2182 /* otherwise we have ranges */ 2183 p = ((ipfw_insn_u16 *)cmd)->ports; 2184 i = cmdlen - 1; 2185 for (; !match && i>0; i--, p += 2) 2186 match = (x >= p[0] && x <= p[1]); 2187 } 2188 break; 2189 2190 case O_IPPRECEDENCE: 2191 match = (hlen > 0 && 2192 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2193 break; 2194 2195 case O_IPTOS: 2196 match = (hlen > 0 && 2197 flags_match(cmd, ip->ip_tos)); 2198 break; 2199 2200 case O_TCPDATALEN: 2201 if (proto == IPPROTO_TCP && offset == 0) { 2202 struct tcphdr *tcp; 2203 uint16_t x; 2204 uint16_t *p; 2205 int i; 2206 2207 tcp = L3HDR(struct tcphdr,ip); 2208 x = ip_len - 2209 ((ip->ip_hl + tcp->th_off) << 2); 2210 if (cmdlen == 1) { 2211 match = (cmd->arg1 == x); 2212 break; 2213 } 2214 /* otherwise we have ranges */ 2215 p = ((ipfw_insn_u16 *)cmd)->ports; 2216 i = cmdlen - 1; 2217 for (; !match && i>0; i--, p += 2) 2218 match = (x >= p[0] && x <= p[1]); 2219 } 2220 break; 2221 2222 case O_TCPFLAGS: 2223 match = (proto == IPPROTO_TCP && offset == 0 && 2224 flags_match(cmd, 2225 L3HDR(struct tcphdr,ip)->th_flags)); 2226 break; 2227 2228 case O_TCPOPTS: 2229 match = (proto == IPPROTO_TCP && offset == 0 && 2230 tcpopts_match(ip, cmd)); 2231 break; 2232 2233 case O_TCPSEQ: 2234 match = (proto == IPPROTO_TCP && offset == 0 && 2235 ((ipfw_insn_u32 *)cmd)->d[0] == 2236 L3HDR(struct tcphdr,ip)->th_seq); 2237 break; 2238 2239 case O_TCPACK: 2240 match = (proto == IPPROTO_TCP && offset == 0 && 2241 ((ipfw_insn_u32 *)cmd)->d[0] == 2242 L3HDR(struct tcphdr,ip)->th_ack); 2243 break; 2244 2245 case O_TCPWIN: 2246 match = (proto == IPPROTO_TCP && offset == 0 && 2247 cmd->arg1 == 2248 L3HDR(struct tcphdr,ip)->th_win); 2249 break; 2250 2251 case O_ESTAB: 2252 /* reject packets which have SYN only */ 2253 /* XXX should i also check for TH_ACK ? */ 2254 match = (proto == IPPROTO_TCP && offset == 0 && 2255 (L3HDR(struct tcphdr,ip)->th_flags & 2256 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2257 break; 2258 2259 case O_ALTQ: { 2260 struct altq_tag *at; 2261 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2262 2263 match = 1; 2264 mtag = m_tag_get(PACKET_TAG_PF_QID, 2265 sizeof(struct altq_tag), 2266 M_NOWAIT); 2267 if (mtag == NULL) { 2268 /* 2269 * Let the packet fall back to the 2270 * default ALTQ. 2271 */ 2272 break; 2273 } 2274 at = (struct altq_tag *)(mtag+1); 2275 at->qid = altq->qid; 2276 if (hlen != 0) 2277 at->af = AF_INET; 2278 else 2279 at->af = AF_LINK; 2280 at->hdr = ip; 2281 m_tag_prepend(m, mtag); 2282 break; 2283 } 2284 2285 case O_LOG: 2286 if (fw_verbose) 2287 ipfw_log(f, hlen, args->eh, m, oif); 2288 match = 1; 2289 break; 2290 2291 case O_PROB: 2292 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2293 break; 2294 2295 case O_VERREVPATH: 2296 /* Outgoing packets automatically pass/match */ 2297 match = (hlen > 0 && ((oif != NULL) || 2298 (m->m_pkthdr.rcvif == NULL) || 2299 verify_path(src_ip, m->m_pkthdr.rcvif))); 2300 break; 2301 2302 case O_VERSRCREACH: 2303 /* Outgoing packets automatically pass/match */ 2304 match = (hlen > 0 && ((oif != NULL) || 2305 verify_path(src_ip, NULL))); 2306 break; 2307 2308 case O_ANTISPOOF: 2309 /* Outgoing packets automatically pass/match */ 2310 if (oif == NULL && hlen > 0 && 2311 in_localaddr(src_ip)) 2312 match = verify_path(src_ip, 2313 m->m_pkthdr.rcvif); 2314 else 2315 match = 1; 2316 break; 2317 2318 case O_IPSEC: 2319#ifdef FAST_IPSEC 2320 match = (m_tag_find(m, 2321 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2322#endif 2323#ifdef IPSEC 2324 match = (ipsec_getnhist(m) != 0); 2325#endif 2326 /* otherwise no match */ 2327 break; 2328 2329 /* 2330 * The second set of opcodes represents 'actions', 2331 * i.e. the terminal part of a rule once the packet 2332 * matches all previous patterns. 2333 * Typically there is only one action for each rule, 2334 * and the opcode is stored at the end of the rule 2335 * (but there are exceptions -- see below). 2336 * 2337 * In general, here we set retval and terminate the 2338 * outer loop (would be a 'break 3' in some language, 2339 * but we need to do a 'goto done'). 2340 * 2341 * Exceptions: 2342 * O_COUNT and O_SKIPTO actions: 2343 * instead of terminating, we jump to the next rule 2344 * ('goto next_rule', equivalent to a 'break 2'), 2345 * or to the SKIPTO target ('goto again' after 2346 * having set f, cmd and l), respectively. 2347 * 2348 * O_LOG and O_ALTQ action parameters: 2349 * perform some action and set match = 1; 2350 * 2351 * O_LIMIT and O_KEEP_STATE: these opcodes are 2352 * not real 'actions', and are stored right 2353 * before the 'action' part of the rule. 2354 * These opcodes try to install an entry in the 2355 * state tables; if successful, we continue with 2356 * the next opcode (match=1; break;), otherwise 2357 * the packet * must be dropped 2358 * ('goto done' after setting retval); 2359 * 2360 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2361 * cause a lookup of the state table, and a jump 2362 * to the 'action' part of the parent rule 2363 * ('goto check_body') if an entry is found, or 2364 * (CHECK_STATE only) a jump to the next rule if 2365 * the entry is not found ('goto next_rule'). 2366 * The result of the lookup is cached to make 2367 * further instances of these opcodes are 2368 * effectively NOPs. 2369 */ 2370 case O_LIMIT: 2371 case O_KEEP_STATE: 2372 if (install_state(f, 2373 (ipfw_insn_limit *)cmd, args)) { 2374 retval = IP_FW_PORT_DENY_FLAG; 2375 goto done; /* error/limit violation */ 2376 } 2377 match = 1; 2378 break; 2379 2380 case O_PROBE_STATE: 2381 case O_CHECK_STATE: 2382 /* 2383 * dynamic rules are checked at the first 2384 * keep-state or check-state occurrence, 2385 * with the result being stored in dyn_dir. 2386 * The compiler introduces a PROBE_STATE 2387 * instruction for us when we have a 2388 * KEEP_STATE (because PROBE_STATE needs 2389 * to be run first). 2390 */ 2391 if (dyn_dir == MATCH_UNKNOWN && 2392 (q = lookup_dyn_rule(&args->f_id, 2393 &dyn_dir, proto == IPPROTO_TCP ? 2394 L3HDR(struct tcphdr, ip) : NULL)) 2395 != NULL) { 2396 /* 2397 * Found dynamic entry, update stats 2398 * and jump to the 'action' part of 2399 * the parent rule. 2400 */ 2401 q->pcnt++; 2402 q->bcnt += pktlen; 2403 f = q->rule; 2404 cmd = ACTION_PTR(f); 2405 l = f->cmd_len - f->act_ofs; 2406 IPFW_DYN_UNLOCK(); 2407 goto check_body; 2408 } 2409 /* 2410 * Dynamic entry not found. If CHECK_STATE, 2411 * skip to next rule, if PROBE_STATE just 2412 * ignore and continue with next opcode. 2413 */ 2414 if (cmd->opcode == O_CHECK_STATE) 2415 goto next_rule; 2416 match = 1; 2417 break; 2418 2419 case O_ACCEPT: 2420 retval = 0; /* accept */ 2421 goto done; 2422 2423 case O_PIPE: 2424 case O_QUEUE: 2425 args->rule = f; /* report matching rule */ 2426 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG; 2427 goto done; 2428 2429 case O_DIVERT: 2430 case O_TEE: { 2431 struct divert_tag *dt; 2432 2433 if (args->eh) /* not on layer 2 */ 2434 break; 2435 mtag = m_tag_get(PACKET_TAG_DIVERT, 2436 sizeof(struct divert_tag), 2437 M_NOWAIT); 2438 if (mtag == NULL) { 2439 /* XXX statistic */ 2440 /* drop packet */ 2441 IPFW_UNLOCK(chain); 2442 return IP_FW_PORT_DENY_FLAG; 2443 } 2444 dt = (struct divert_tag *)(mtag+1); 2445 dt->cookie = f->rulenum; 2446 dt->info = (cmd->opcode == O_DIVERT) ? 2447 cmd->arg1 : 2448 cmd->arg1 | IP_FW_PORT_TEE_FLAG; 2449 m_tag_prepend(m, mtag); 2450 retval = dt->info; 2451 goto done; 2452 } 2453 2454 case O_COUNT: 2455 case O_SKIPTO: 2456 f->pcnt++; /* update stats */ 2457 f->bcnt += pktlen; 2458 f->timestamp = time_second; 2459 if (cmd->opcode == O_COUNT) 2460 goto next_rule; 2461 /* handle skipto */ 2462 if (f->next_rule == NULL) 2463 lookup_next_rule(f); 2464 f = f->next_rule; 2465 goto again; 2466 2467 case O_REJECT: 2468 /* 2469 * Drop the packet and send a reject notice 2470 * if the packet is not ICMP (or is an ICMP 2471 * query), and it is not multicast/broadcast. 2472 */ 2473 if (hlen > 0 && 2474 (proto != IPPROTO_ICMP || 2475 is_icmp_query(ip)) && 2476 !(m->m_flags & (M_BCAST|M_MCAST)) && 2477 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2478 send_reject(args, cmd->arg1, 2479 offset,ip_len); 2480 m = args->m; 2481 } 2482 /* FALLTHROUGH */ 2483 case O_DENY: 2484 retval = IP_FW_PORT_DENY_FLAG; 2485 goto done; 2486 2487 case O_FORWARD_IP: 2488 if (args->eh) /* not valid on layer2 pkts */ 2489 break; 2490 if (!q || dyn_dir == MATCH_FORWARD) 2491 args->next_hop = 2492 &((ipfw_insn_sa *)cmd)->sa; 2493 retval = 0; 2494 goto done; 2495 2496 default: 2497 panic("-- unknown opcode %d\n", cmd->opcode); 2498 } /* end of switch() on opcodes */ 2499 2500 if (cmd->len & F_NOT) 2501 match = !match; 2502 2503 if (match) { 2504 if (cmd->len & F_OR) 2505 skip_or = 1; 2506 } else { 2507 if (!(cmd->len & F_OR)) /* not an OR block, */ 2508 break; /* try next rule */ 2509 } 2510 2511 } /* end of inner for, scan opcodes */ 2512 2513next_rule:; /* try next rule */ 2514 2515 } /* end of outer for, scan rules */ 2516 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2517 IPFW_UNLOCK(chain); 2518 return(IP_FW_PORT_DENY_FLAG); 2519 2520done: 2521 /* Update statistics */ 2522 f->pcnt++; 2523 f->bcnt += pktlen; 2524 f->timestamp = time_second; 2525 IPFW_UNLOCK(chain); 2526 return retval; 2527 2528pullup_failed: 2529 if (fw_verbose) 2530 printf("ipfw: pullup failed\n"); 2531 return(IP_FW_PORT_DENY_FLAG); 2532} 2533 2534/* 2535 * When a rule is added/deleted, clear the next_rule pointers in all rules. 2536 * These will be reconstructed on the fly as packets are matched. 2537 */ 2538static void 2539flush_rule_ptrs(struct ip_fw_chain *chain) 2540{ 2541 struct ip_fw *rule; 2542 2543 IPFW_LOCK_ASSERT(chain); 2544 2545 for (rule = chain->rules; rule; rule = rule->next) 2546 rule->next_rule = NULL; 2547} 2548 2549/* 2550 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 2551 * pipe/queue, or to all of them (match == NULL). 2552 */ 2553void 2554flush_pipe_ptrs(struct dn_flow_set *match) 2555{ 2556 struct ip_fw *rule; 2557 2558 IPFW_LOCK(&layer3_chain); 2559 for (rule = layer3_chain.rules; rule; rule = rule->next) { 2560 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 2561 2562 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 2563 continue; 2564 /* 2565 * XXX Use bcmp/bzero to handle pipe_ptr to overcome 2566 * possible alignment problems on 64-bit architectures. 2567 * This code is seldom used so we do not worry too 2568 * much about efficiency. 2569 */ 2570 if (match == NULL || 2571 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) ) 2572 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr)); 2573 } 2574 IPFW_UNLOCK(&layer3_chain); 2575} 2576 2577/* 2578 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 2579 * possibly create a rule number and add the rule to the list. 2580 * Update the rule_number in the input struct so the caller knows it as well. 2581 */ 2582static int 2583add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 2584{ 2585 struct ip_fw *rule, *f, *prev; 2586 int l = RULESIZE(input_rule); 2587 2588 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 2589 return (EINVAL); 2590 2591 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 2592 if (rule == NULL) 2593 return (ENOSPC); 2594 2595 bcopy(input_rule, rule, l); 2596 2597 rule->next = NULL; 2598 rule->next_rule = NULL; 2599 2600 rule->pcnt = 0; 2601 rule->bcnt = 0; 2602 rule->timestamp = 0; 2603 2604 IPFW_LOCK(chain); 2605 2606 if (chain->rules == NULL) { /* default rule */ 2607 chain->rules = rule; 2608 goto done; 2609 } 2610 2611 /* 2612 * If rulenum is 0, find highest numbered rule before the 2613 * default rule, and add autoinc_step 2614 */ 2615 if (autoinc_step < 1) 2616 autoinc_step = 1; 2617 else if (autoinc_step > 1000) 2618 autoinc_step = 1000; 2619 if (rule->rulenum == 0) { 2620 /* 2621 * locate the highest numbered rule before default 2622 */ 2623 for (f = chain->rules; f; f = f->next) { 2624 if (f->rulenum == IPFW_DEFAULT_RULE) 2625 break; 2626 rule->rulenum = f->rulenum; 2627 } 2628 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2629 rule->rulenum += autoinc_step; 2630 input_rule->rulenum = rule->rulenum; 2631 } 2632 2633 /* 2634 * Now insert the new rule in the right place in the sorted list. 2635 */ 2636 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 2637 if (f->rulenum > rule->rulenum) { /* found the location */ 2638 if (prev) { 2639 rule->next = f; 2640 prev->next = rule; 2641 } else { /* head insert */ 2642 rule->next = chain->rules; 2643 chain->rules = rule; 2644 } 2645 break; 2646 } 2647 } 2648 flush_rule_ptrs(chain); 2649done: 2650 static_count++; 2651 static_len += l; 2652 IPFW_UNLOCK(chain); 2653 DEB(printf("ipfw: installed rule %d, static count now %d\n", 2654 rule->rulenum, static_count);) 2655 return (0); 2656} 2657 2658/** 2659 * Remove a static rule (including derived * dynamic rules) 2660 * and place it on the ``reap list'' for later reclamation. 2661 * The caller is in charge of clearing rule pointers to avoid 2662 * dangling pointers. 2663 * @return a pointer to the next entry. 2664 * Arguments are not checked, so they better be correct. 2665 */ 2666static struct ip_fw * 2667remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 2668{ 2669 struct ip_fw *n; 2670 int l = RULESIZE(rule); 2671 2672 IPFW_LOCK_ASSERT(chain); 2673 2674 n = rule->next; 2675 IPFW_DYN_LOCK(); 2676 remove_dyn_rule(rule, NULL /* force removal */); 2677 IPFW_DYN_UNLOCK(); 2678 if (prev == NULL) 2679 chain->rules = n; 2680 else 2681 prev->next = n; 2682 static_count--; 2683 static_len -= l; 2684 2685 rule->next = chain->reap; 2686 chain->reap = rule; 2687 2688 return n; 2689} 2690 2691/** 2692 * Reclaim storage associated with a list of rules. This is 2693 * typically the list created using remove_rule. 2694 */ 2695static void 2696reap_rules(struct ip_fw *head) 2697{ 2698 struct ip_fw *rule; 2699 2700 while ((rule = head) != NULL) { 2701 head = head->next; 2702 if (DUMMYNET_LOADED) 2703 ip_dn_ruledel_ptr(rule); 2704 free(rule, M_IPFW); 2705 } 2706} 2707 2708/* 2709 * Remove all rules from a chain (except rules in set RESVD_SET 2710 * unless kill_default = 1). The caller is responsible for 2711 * reclaiming storage for the rules left in chain->reap. 2712 */ 2713static void 2714free_chain(struct ip_fw_chain *chain, int kill_default) 2715{ 2716 struct ip_fw *prev, *rule; 2717 2718 IPFW_LOCK_ASSERT(chain); 2719 2720 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 2721 for (prev = NULL, rule = chain->rules; rule ; ) 2722 if (kill_default || rule->set != RESVD_SET) 2723 rule = remove_rule(chain, rule, prev); 2724 else { 2725 prev = rule; 2726 rule = rule->next; 2727 } 2728} 2729 2730/** 2731 * Remove all rules with given number, and also do set manipulation. 2732 * Assumes chain != NULL && *chain != NULL. 2733 * 2734 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2735 * the next 8 bits are the new set, the top 8 bits are the command: 2736 * 2737 * 0 delete rules with given number 2738 * 1 delete rules with given set number 2739 * 2 move rules with given number to new set 2740 * 3 move rules with given set number to new set 2741 * 4 swap sets with given numbers 2742 */ 2743static int 2744del_entry(struct ip_fw_chain *chain, u_int32_t arg) 2745{ 2746 struct ip_fw *prev = NULL, *rule; 2747 u_int16_t rulenum; /* rule or old_set */ 2748 u_int8_t cmd, new_set; 2749 2750 rulenum = arg & 0xffff; 2751 cmd = (arg >> 24) & 0xff; 2752 new_set = (arg >> 16) & 0xff; 2753 2754 if (cmd > 4) 2755 return EINVAL; 2756 if (new_set > RESVD_SET) 2757 return EINVAL; 2758 if (cmd == 0 || cmd == 2) { 2759 if (rulenum >= IPFW_DEFAULT_RULE) 2760 return EINVAL; 2761 } else { 2762 if (rulenum > RESVD_SET) /* old_set */ 2763 return EINVAL; 2764 } 2765 2766 IPFW_LOCK(chain); 2767 rule = chain->rules; 2768 chain->reap = NULL; 2769 switch (cmd) { 2770 case 0: /* delete rules with given number */ 2771 /* 2772 * locate first rule to delete 2773 */ 2774 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 2775 ; 2776 if (rule->rulenum != rulenum) { 2777 IPFW_UNLOCK(chain); 2778 return EINVAL; 2779 } 2780 2781 /* 2782 * flush pointers outside the loop, then delete all matching 2783 * rules. prev remains the same throughout the cycle. 2784 */ 2785 flush_rule_ptrs(chain); 2786 while (rule->rulenum == rulenum) 2787 rule = remove_rule(chain, rule, prev); 2788 break; 2789 2790 case 1: /* delete all rules with given set number */ 2791 flush_rule_ptrs(chain); 2792 rule = chain->rules; 2793 while (rule->rulenum < IPFW_DEFAULT_RULE) 2794 if (rule->set == rulenum) 2795 rule = remove_rule(chain, rule, prev); 2796 else { 2797 prev = rule; 2798 rule = rule->next; 2799 } 2800 break; 2801 2802 case 2: /* move rules with given number to new set */ 2803 rule = chain->rules; 2804 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2805 if (rule->rulenum == rulenum) 2806 rule->set = new_set; 2807 break; 2808 2809 case 3: /* move rules with given set number to new set */ 2810 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2811 if (rule->set == rulenum) 2812 rule->set = new_set; 2813 break; 2814 2815 case 4: /* swap two sets */ 2816 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2817 if (rule->set == rulenum) 2818 rule->set = new_set; 2819 else if (rule->set == new_set) 2820 rule->set = rulenum; 2821 break; 2822 } 2823 /* 2824 * Look for rules to reclaim. We grab the list before 2825 * releasing the lock then reclaim them w/o the lock to 2826 * avoid a LOR with dummynet. 2827 */ 2828 rule = chain->reap; 2829 chain->reap = NULL; 2830 IPFW_UNLOCK(chain); 2831 if (rule) 2832 reap_rules(rule); 2833 return 0; 2834} 2835 2836/* 2837 * Clear counters for a specific rule. 2838 * The enclosing "table" is assumed locked. 2839 */ 2840static void 2841clear_counters(struct ip_fw *rule, int log_only) 2842{ 2843 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 2844 2845 if (log_only == 0) { 2846 rule->bcnt = rule->pcnt = 0; 2847 rule->timestamp = 0; 2848 } 2849 if (l->o.opcode == O_LOG) 2850 l->log_left = l->max_log; 2851} 2852 2853/** 2854 * Reset some or all counters on firewall rules. 2855 * @arg frwl is null to clear all entries, or contains a specific 2856 * rule number. 2857 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 2858 */ 2859static int 2860zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 2861{ 2862 struct ip_fw *rule; 2863 char *msg; 2864 2865 IPFW_LOCK(chain); 2866 if (rulenum == 0) { 2867 norule_counter = 0; 2868 for (rule = chain->rules; rule; rule = rule->next) 2869 clear_counters(rule, log_only); 2870 msg = log_only ? "ipfw: All logging counts reset.\n" : 2871 "ipfw: Accounting cleared.\n"; 2872 } else { 2873 int cleared = 0; 2874 /* 2875 * We can have multiple rules with the same number, so we 2876 * need to clear them all. 2877 */ 2878 for (rule = chain->rules; rule; rule = rule->next) 2879 if (rule->rulenum == rulenum) { 2880 while (rule && rule->rulenum == rulenum) { 2881 clear_counters(rule, log_only); 2882 rule = rule->next; 2883 } 2884 cleared = 1; 2885 break; 2886 } 2887 if (!cleared) { /* we did not find any matching rules */ 2888 IPFW_UNLOCK(chain); 2889 return (EINVAL); 2890 } 2891 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 2892 "ipfw: Entry %d cleared.\n"; 2893 } 2894 IPFW_UNLOCK(chain); 2895 2896 if (fw_verbose) 2897 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 2898 return (0); 2899} 2900 2901/* 2902 * Check validity of the structure before insert. 2903 * Fortunately rules are simple, so this mostly need to check rule sizes. 2904 */ 2905static int 2906check_ipfw_struct(struct ip_fw *rule, int size) 2907{ 2908 int l, cmdlen = 0; 2909 int have_action=0; 2910 ipfw_insn *cmd; 2911 2912 if (size < sizeof(*rule)) { 2913 printf("ipfw: rule too short\n"); 2914 return (EINVAL); 2915 } 2916 /* first, check for valid size */ 2917 l = RULESIZE(rule); 2918 if (l != size) { 2919 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 2920 return (EINVAL); 2921 } 2922 if (rule->act_ofs >= rule->cmd_len) { 2923 printf("ipfw: bogus action offset (%u > %u)\n", 2924 rule->act_ofs, rule->cmd_len - 1); 2925 return (EINVAL); 2926 } 2927 /* 2928 * Now go for the individual checks. Very simple ones, basically only 2929 * instruction sizes. 2930 */ 2931 for (l = rule->cmd_len, cmd = rule->cmd ; 2932 l > 0 ; l -= cmdlen, cmd += cmdlen) { 2933 cmdlen = F_LEN(cmd); 2934 if (cmdlen > l) { 2935 printf("ipfw: opcode %d size truncated\n", 2936 cmd->opcode); 2937 return EINVAL; 2938 } 2939 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 2940 switch (cmd->opcode) { 2941 case O_PROBE_STATE: 2942 case O_KEEP_STATE: 2943 case O_PROTO: 2944 case O_IP_SRC_ME: 2945 case O_IP_DST_ME: 2946 case O_LAYER2: 2947 case O_IN: 2948 case O_FRAG: 2949 case O_DIVERTED: 2950 case O_IPOPT: 2951 case O_IPTOS: 2952 case O_IPPRECEDENCE: 2953 case O_IPVER: 2954 case O_TCPWIN: 2955 case O_TCPFLAGS: 2956 case O_TCPOPTS: 2957 case O_ESTAB: 2958 case O_VERREVPATH: 2959 case O_VERSRCREACH: 2960 case O_ANTISPOOF: 2961 case O_IPSEC: 2962 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2963 goto bad_size; 2964 break; 2965 2966 case O_UID: 2967 case O_GID: 2968 case O_JAIL: 2969 case O_IP_SRC: 2970 case O_IP_DST: 2971 case O_TCPSEQ: 2972 case O_TCPACK: 2973 case O_PROB: 2974 case O_ICMPTYPE: 2975 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2976 goto bad_size; 2977 break; 2978 2979 case O_LIMIT: 2980 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 2981 goto bad_size; 2982 break; 2983 2984 case O_LOG: 2985 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 2986 goto bad_size; 2987 2988 ((ipfw_insn_log *)cmd)->log_left = 2989 ((ipfw_insn_log *)cmd)->max_log; 2990 2991 break; 2992 2993 case O_IP_SRC_MASK: 2994 case O_IP_DST_MASK: 2995 /* only odd command lengths */ 2996 if ( !(cmdlen & 1) || cmdlen > 31) 2997 goto bad_size; 2998 break; 2999 3000 case O_IP_SRC_SET: 3001 case O_IP_DST_SET: 3002 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3003 printf("ipfw: invalid set size %d\n", 3004 cmd->arg1); 3005 return EINVAL; 3006 } 3007 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3008 (cmd->arg1+31)/32 ) 3009 goto bad_size; 3010 break; 3011 3012 case O_IP_SRC_LOOKUP: 3013 case O_IP_DST_LOOKUP: 3014 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3015 printf("ipfw: invalid table number %d\n", 3016 cmd->arg1); 3017 return (EINVAL); 3018 } 3019 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3020 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3021 goto bad_size; 3022 break; 3023 3024 case O_MACADDR2: 3025 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3026 goto bad_size; 3027 break; 3028 3029 case O_NOP: 3030 case O_IPID: 3031 case O_IPTTL: 3032 case O_IPLEN: 3033 case O_TCPDATALEN: 3034 if (cmdlen < 1 || cmdlen > 31) 3035 goto bad_size; 3036 break; 3037 3038 case O_MAC_TYPE: 3039 case O_IP_SRCPORT: 3040 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3041 if (cmdlen < 2 || cmdlen > 31) 3042 goto bad_size; 3043 break; 3044 3045 case O_RECV: 3046 case O_XMIT: 3047 case O_VIA: 3048 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3049 goto bad_size; 3050 break; 3051 3052 case O_ALTQ: 3053 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3054 goto bad_size; 3055 break; 3056 3057 case O_PIPE: 3058 case O_QUEUE: 3059 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 3060 goto bad_size; 3061 goto check_action; 3062 3063 case O_FORWARD_IP: 3064#ifdef IPFIREWALL_FORWARD 3065 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3066 goto bad_size; 3067 goto check_action; 3068#else 3069 return EINVAL; 3070#endif 3071 3072 case O_DIVERT: 3073 case O_TEE: 3074 if (ip_divert_ptr == NULL) 3075 return EINVAL; 3076 case O_FORWARD_MAC: /* XXX not implemented yet */ 3077 case O_CHECK_STATE: 3078 case O_COUNT: 3079 case O_ACCEPT: 3080 case O_DENY: 3081 case O_REJECT: 3082 case O_SKIPTO: 3083 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3084 goto bad_size; 3085check_action: 3086 if (have_action) { 3087 printf("ipfw: opcode %d, multiple actions" 3088 " not allowed\n", 3089 cmd->opcode); 3090 return EINVAL; 3091 } 3092 have_action = 1; 3093 if (l != cmdlen) { 3094 printf("ipfw: opcode %d, action must be" 3095 " last opcode\n", 3096 cmd->opcode); 3097 return EINVAL; 3098 } 3099 break; 3100 default: 3101 printf("ipfw: opcode %d, unknown opcode\n", 3102 cmd->opcode); 3103 return EINVAL; 3104 } 3105 } 3106 if (have_action == 0) { 3107 printf("ipfw: missing action\n"); 3108 return EINVAL; 3109 } 3110 return 0; 3111 3112bad_size: 3113 printf("ipfw: opcode %d size %d wrong\n", 3114 cmd->opcode, cmdlen); 3115 return EINVAL; 3116} 3117 3118/* 3119 * Copy the static and dynamic rules to the supplied buffer 3120 * and return the amount of space actually used. 3121 */ 3122static size_t 3123ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3124{ 3125 char *bp = buf; 3126 char *ep = bp + space; 3127 struct ip_fw *rule; 3128 int i; 3129 3130 /* XXX this can take a long time and locking will block packet flow */ 3131 IPFW_LOCK(chain); 3132 for (rule = chain->rules; rule ; rule = rule->next) { 3133 /* 3134 * Verify the entry fits in the buffer in case the 3135 * rules changed between calculating buffer space and 3136 * now. This would be better done using a generation 3137 * number but should suffice for now. 3138 */ 3139 i = RULESIZE(rule); 3140 if (bp + i <= ep) { 3141 bcopy(rule, bp, i); 3142 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3143 sizeof(set_disable)); 3144 bp += i; 3145 } 3146 } 3147 IPFW_UNLOCK(chain); 3148 if (ipfw_dyn_v) { 3149 ipfw_dyn_rule *p, *last = NULL; 3150 3151 IPFW_DYN_LOCK(); 3152 for (i = 0 ; i < curr_dyn_buckets; i++) 3153 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3154 if (bp + sizeof *p <= ep) { 3155 ipfw_dyn_rule *dst = 3156 (ipfw_dyn_rule *)bp; 3157 bcopy(p, dst, sizeof *p); 3158 bcopy(&(p->rule->rulenum), &(dst->rule), 3159 sizeof(p->rule->rulenum)); 3160 /* 3161 * store a non-null value in "next". 3162 * The userland code will interpret a 3163 * NULL here as a marker 3164 * for the last dynamic rule. 3165 */ 3166 bcopy(&dst, &dst->next, sizeof(dst)); 3167 last = dst; 3168 dst->expire = 3169 TIME_LEQ(dst->expire, time_second) ? 3170 0 : dst->expire - time_second ; 3171 bp += sizeof(ipfw_dyn_rule); 3172 } 3173 } 3174 IPFW_DYN_UNLOCK(); 3175 if (last != NULL) /* mark last dynamic rule */ 3176 bzero(&last->next, sizeof(last)); 3177 } 3178 return (bp - (char *)buf); 3179} 3180 3181 3182/** 3183 * {set|get}sockopt parser. 3184 */ 3185static int 3186ipfw_ctl(struct sockopt *sopt) 3187{ 3188#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3189 int error, rule_num; 3190 size_t size; 3191 struct ip_fw *buf, *rule; 3192 u_int32_t rulenum[2]; 3193 3194 error = suser(sopt->sopt_td); 3195 if (error) 3196 return (error); 3197 3198 /* 3199 * Disallow modifications in really-really secure mode, but still allow 3200 * the logging counters to be reset. 3201 */ 3202 if (sopt->sopt_name == IP_FW_ADD || 3203 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3204#if __FreeBSD_version >= 500034 3205 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3206 if (error) 3207 return (error); 3208#else /* FreeBSD 4.x */ 3209 if (securelevel >= 3) 3210 return (EPERM); 3211#endif 3212 } 3213 3214 error = 0; 3215 3216 switch (sopt->sopt_name) { 3217 case IP_FW_GET: 3218 /* 3219 * pass up a copy of the current rules. Static rules 3220 * come first (the last of which has number IPFW_DEFAULT_RULE), 3221 * followed by a possibly empty list of dynamic rule. 3222 * The last dynamic rule has NULL in the "next" field. 3223 * 3224 * Note that the calculated size is used to bound the 3225 * amount of data returned to the user. The rule set may 3226 * change between calculating the size and returning the 3227 * data in which case we'll just return what fits. 3228 */ 3229 size = static_len; /* size of static rules */ 3230 if (ipfw_dyn_v) /* add size of dyn.rules */ 3231 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3232 3233 /* 3234 * XXX todo: if the user passes a short length just to know 3235 * how much room is needed, do not bother filling up the 3236 * buffer, just jump to the sooptcopyout. 3237 */ 3238 buf = malloc(size, M_TEMP, M_WAITOK); 3239 error = sooptcopyout(sopt, buf, 3240 ipfw_getrules(&layer3_chain, buf, size)); 3241 free(buf, M_TEMP); 3242 break; 3243 3244 case IP_FW_FLUSH: 3245 /* 3246 * Normally we cannot release the lock on each iteration. 3247 * We could do it here only because we start from the head all 3248 * the times so there is no risk of missing some entries. 3249 * On the other hand, the risk is that we end up with 3250 * a very inconsistent ruleset, so better keep the lock 3251 * around the whole cycle. 3252 * 3253 * XXX this code can be improved by resetting the head of 3254 * the list to point to the default rule, and then freeing 3255 * the old list without the need for a lock. 3256 */ 3257 3258 IPFW_LOCK(&layer3_chain); 3259 layer3_chain.reap = NULL; 3260 free_chain(&layer3_chain, 0 /* keep default rule */); 3261 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3262 IPFW_UNLOCK(&layer3_chain); 3263 if (layer3_chain.reap != NULL) 3264 reap_rules(rule); 3265 break; 3266 3267 case IP_FW_ADD: 3268 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3269 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3270 sizeof(struct ip_fw) ); 3271 if (error == 0) 3272 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3273 if (error == 0) { 3274 error = add_rule(&layer3_chain, rule); 3275 size = RULESIZE(rule); 3276 if (!error && sopt->sopt_dir == SOPT_GET) 3277 error = sooptcopyout(sopt, rule, size); 3278 } 3279 free(rule, M_TEMP); 3280 break; 3281 3282 case IP_FW_DEL: 3283 /* 3284 * IP_FW_DEL is used for deleting single rules or sets, 3285 * and (ab)used to atomically manipulate sets. Argument size 3286 * is used to distinguish between the two: 3287 * sizeof(u_int32_t) 3288 * delete single rule or set of rules, 3289 * or reassign rules (or sets) to a different set. 3290 * 2*sizeof(u_int32_t) 3291 * atomic disable/enable sets. 3292 * first u_int32_t contains sets to be disabled, 3293 * second u_int32_t contains sets to be enabled. 3294 */ 3295 error = sooptcopyin(sopt, rulenum, 3296 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3297 if (error) 3298 break; 3299 size = sopt->sopt_valsize; 3300 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3301 error = del_entry(&layer3_chain, rulenum[0]); 3302 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3303 set_disable = 3304 (set_disable | rulenum[0]) & ~rulenum[1] & 3305 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3306 else 3307 error = EINVAL; 3308 break; 3309 3310 case IP_FW_ZERO: 3311 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3312 rule_num = 0; 3313 if (sopt->sopt_val != 0) { 3314 error = sooptcopyin(sopt, &rule_num, 3315 sizeof(int), sizeof(int)); 3316 if (error) 3317 break; 3318 } 3319 error = zero_entry(&layer3_chain, rule_num, 3320 sopt->sopt_name == IP_FW_RESETLOG); 3321 break; 3322 3323 case IP_FW_TABLE_ADD: 3324 { 3325 ipfw_table_entry ent; 3326 3327 error = sooptcopyin(sopt, &ent, 3328 sizeof(ent), sizeof(ent)); 3329 if (error) 3330 break; 3331 error = add_table_entry(ent.tbl, ent.addr, 3332 ent.masklen, ent.value); 3333 } 3334 break; 3335 3336 case IP_FW_TABLE_DEL: 3337 { 3338 ipfw_table_entry ent; 3339 3340 error = sooptcopyin(sopt, &ent, 3341 sizeof(ent), sizeof(ent)); 3342 if (error) 3343 break; 3344 error = del_table_entry(ent.tbl, ent.addr, ent.masklen); 3345 } 3346 break; 3347 3348 case IP_FW_TABLE_FLUSH: 3349 { 3350 u_int16_t tbl; 3351 3352 error = sooptcopyin(sopt, &tbl, 3353 sizeof(tbl), sizeof(tbl)); 3354 if (error) 3355 break; 3356 error = flush_table(tbl); 3357 } 3358 break; 3359 3360 case IP_FW_TABLE_GETSIZE: 3361 { 3362 u_int32_t tbl, cnt; 3363 3364 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 3365 sizeof(tbl)))) 3366 break; 3367 if ((error = count_table(tbl, &cnt))) 3368 break; 3369 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 3370 } 3371 break; 3372 3373 case IP_FW_TABLE_LIST: 3374 { 3375 ipfw_table *tbl; 3376 3377 if (sopt->sopt_valsize < sizeof(*tbl)) { 3378 error = EINVAL; 3379 break; 3380 } 3381 size = sopt->sopt_valsize; 3382 tbl = malloc(size, M_TEMP, M_WAITOK); 3383 if (tbl == NULL) { 3384 error = ENOMEM; 3385 break; 3386 } 3387 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 3388 if (error) { 3389 free(tbl, M_TEMP); 3390 break; 3391 } 3392 tbl->size = (size - sizeof(*tbl)) / 3393 sizeof(ipfw_table_entry); 3394 error = dump_table(tbl); 3395 if (error) { 3396 free(tbl, M_TEMP); 3397 break; 3398 } 3399 error = sooptcopyout(sopt, tbl, size); 3400 free(tbl, M_TEMP); 3401 } 3402 break; 3403 3404 default: 3405 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 3406 error = EINVAL; 3407 } 3408 3409 return (error); 3410#undef RULE_MAXSIZE 3411} 3412 3413/** 3414 * dummynet needs a reference to the default rule, because rules can be 3415 * deleted while packets hold a reference to them. When this happens, 3416 * dummynet changes the reference to the default rule (it could well be a 3417 * NULL pointer, but this way we do not need to check for the special 3418 * case, plus here he have info on the default behaviour). 3419 */ 3420struct ip_fw *ip_fw_default_rule; 3421 3422/* 3423 * This procedure is only used to handle keepalives. It is invoked 3424 * every dyn_keepalive_period 3425 */ 3426static void 3427ipfw_tick(void * __unused unused) 3428{ 3429 int i; 3430 ipfw_dyn_rule *q; 3431 3432 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 3433 goto done; 3434 3435 IPFW_DYN_LOCK(); 3436 for (i = 0 ; i < curr_dyn_buckets ; i++) { 3437 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 3438 if (q->dyn_type == O_LIMIT_PARENT) 3439 continue; 3440 if (q->id.proto != IPPROTO_TCP) 3441 continue; 3442 if ( (q->state & BOTH_SYN) != BOTH_SYN) 3443 continue; 3444 if (TIME_LEQ( time_second+dyn_keepalive_interval, 3445 q->expire)) 3446 continue; /* too early */ 3447 if (TIME_LEQ(q->expire, time_second)) 3448 continue; /* too late, rule expired */ 3449 3450 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 3451 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 3452 } 3453 } 3454 IPFW_DYN_UNLOCK(); 3455done: 3456 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 3457} 3458 3459int 3460ipfw_init(void) 3461{ 3462 struct ip_fw default_rule; 3463 int error; 3464 3465 layer3_chain.rules = NULL; 3466 IPFW_LOCK_INIT(&layer3_chain); 3467 IPFW_DYN_LOCK_INIT(); 3468 callout_init(&ipfw_timeout, debug_mpsafenet ? CALLOUT_MPSAFE : 0); 3469 3470 bzero(&default_rule, sizeof default_rule); 3471 3472 default_rule.act_ofs = 0; 3473 default_rule.rulenum = IPFW_DEFAULT_RULE; 3474 default_rule.cmd_len = 1; 3475 default_rule.set = RESVD_SET; 3476 3477 default_rule.cmd[0].len = 1; 3478 default_rule.cmd[0].opcode = 3479#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 3480 1 ? O_ACCEPT : 3481#endif 3482 O_DENY; 3483 3484 error = add_rule(&layer3_chain, &default_rule); 3485 if (error != 0) { 3486 printf("ipfw2: error %u initializing default rule " 3487 "(support disabled)\n", error); 3488 IPFW_DYN_LOCK_DESTROY(); 3489 IPFW_LOCK_DESTROY(&layer3_chain); 3490 return (error); 3491 } 3492 3493 ip_fw_default_rule = layer3_chain.rules; 3494 printf("ipfw2 initialized, divert %s, " 3495 "rule-based forwarding " 3496#ifdef IPFIREWALL_FORWARD 3497 "enabled, " 3498#else 3499 "disabled, " 3500#endif 3501 "default to %s, logging ", 3502#ifdef IPDIVERT 3503 "enabled", 3504#else 3505 "loadable", 3506#endif 3507 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 3508 3509#ifdef IPFIREWALL_VERBOSE 3510 fw_verbose = 1; 3511#endif 3512#ifdef IPFIREWALL_VERBOSE_LIMIT 3513 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3514#endif 3515 if (fw_verbose == 0) 3516 printf("disabled\n"); 3517 else if (verbose_limit == 0) 3518 printf("unlimited\n"); 3519 else 3520 printf("limited to %d packets/entry by default\n", 3521 verbose_limit); 3522 3523 init_tables(); 3524 ip_fw_ctl_ptr = ipfw_ctl; 3525 ip_fw_chk_ptr = ipfw_chk; 3526 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 3527 3528 return (0); 3529} 3530 3531void 3532ipfw_destroy(void) 3533{ 3534 struct ip_fw *reap; 3535 3536 ip_fw_chk_ptr = NULL; 3537 ip_fw_ctl_ptr = NULL; 3538 callout_drain(&ipfw_timeout); 3539 IPFW_LOCK(&layer3_chain); 3540 layer3_chain.reap = NULL; 3541 free_chain(&layer3_chain, 1 /* kill default rule */); 3542 reap = layer3_chain.reap, layer3_chain.reap = NULL; 3543 IPFW_UNLOCK(&layer3_chain); 3544 if (reap != NULL) 3545 reap_rules(reap); 3546 flush_tables(); 3547 IPFW_DYN_LOCK_DESTROY(); 3548 IPFW_LOCK_DESTROY(&layer3_chain); 3549 printf("IP firewall unloaded\n"); 3550} 3551 3552#endif /* IPFW2 */ 3553