ip_fw2.c revision 107114
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 107114 2002-11-20 19:07:27Z luigi $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdn.h" 38#include "opt_ipdivert.h" 39#include "opt_inet.h" 40#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/proc.h> 53#include <sys/socket.h> 54#include <sys/socketvar.h> 55#include <sys/sysctl.h> 56#include <sys/syslog.h> 57#include <sys/ucred.h> 58#include <net/if.h> 59#include <net/route.h> 60#include <netinet/in.h> 61#include <netinet/in_systm.h> 62#include <netinet/in_var.h> 63#include <netinet/in_pcb.h> 64#include <netinet/ip.h> 65#include <netinet/ip_var.h> 66#include <netinet/ip_icmp.h> 67#include <netinet/ip_fw.h> 68#include <netinet/ip_dummynet.h> 69#include <netinet/tcp.h> 70#include <netinet/tcp_timer.h> 71#include <netinet/tcp_var.h> 72#include <netinet/tcpip.h> 73#include <netinet/udp.h> 74#include <netinet/udp_var.h> 75 76#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 77 78#include <machine/in_cksum.h> /* XXX for in_cksum */ 79 80/* 81 * XXX This one should go in sys/mbuf.h. It is used to avoid that 82 * a firewall-generated packet loops forever through the firewall. 83 */ 84#ifndef M_SKIP_FIREWALL 85#define M_SKIP_FIREWALL 0x4000 86#endif 87 88/* 89 * set_disable contains one bit per set value (0..31). 90 * If the bit is set, all rules with the corresponding set 91 * are disabled. Set 31 is reserved for the default rule 92 * and CANNOT be disabled. 93 */ 94static u_int32_t set_disable; 95 96static int fw_verbose; 97static int verbose_limit; 98 99static struct callout_handle ipfw_timeout_h; 100#define IPFW_DEFAULT_RULE 65535 101 102/* 103 * list of rules for layer 3 104 */ 105static struct ip_fw *layer3_chain; 106 107MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 108 109static int fw_debug = 1; 110static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 111 112#ifdef SYSCTL_NODE 113SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 114SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 115 CTLFLAG_RW | CTLFLAG_SECURE, 116 &fw_enable, 0, "Enable ipfw"); 117SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 118 &autoinc_step, 0, "Rule number autincrement step"); 119SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 120 CTLFLAG_RW | CTLFLAG_SECURE, 121 &fw_one_pass, 0, 122 "Only do a single pass through ipfw when using dummynet(4)"); 123SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 124 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 125SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 126 CTLFLAG_RW | CTLFLAG_SECURE, 127 &fw_verbose, 0, "Log matches to ipfw rules"); 128SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 129 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 130 131/* 132 * Description of dynamic rules. 133 * 134 * Dynamic rules are stored in lists accessed through a hash table 135 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 136 * be modified through the sysctl variable dyn_buckets which is 137 * updated when the table becomes empty. 138 * 139 * XXX currently there is only one list, ipfw_dyn. 140 * 141 * When a packet is received, its address fields are first masked 142 * with the mask defined for the rule, then hashed, then matched 143 * against the entries in the corresponding list. 144 * Dynamic rules can be used for different purposes: 145 * + stateful rules; 146 * + enforcing limits on the number of sessions; 147 * + in-kernel NAT (not implemented yet) 148 * 149 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 150 * measured in seconds and depending on the flags. 151 * 152 * The total number of dynamic rules is stored in dyn_count. 153 * The max number of dynamic rules is dyn_max. When we reach 154 * the maximum number of rules we do not create anymore. This is 155 * done to avoid consuming too much memory, but also too much 156 * time when searching on each packet (ideally, we should try instead 157 * to put a limit on the length of the list on each bucket...). 158 * 159 * Each dynamic rule holds a pointer to the parent ipfw rule so 160 * we know what action to perform. Dynamic rules are removed when 161 * the parent rule is deleted. XXX we should make them survive. 162 * 163 * There are some limitations with dynamic rules -- we do not 164 * obey the 'randomized match', and we do not do multiple 165 * passes through the firewall. XXX check the latter!!! 166 */ 167static ipfw_dyn_rule **ipfw_dyn_v = NULL; 168static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 169static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 170 171/* 172 * Timeouts for various events in handing dynamic rules. 173 */ 174static u_int32_t dyn_ack_lifetime = 300; 175static u_int32_t dyn_syn_lifetime = 20; 176static u_int32_t dyn_fin_lifetime = 1; 177static u_int32_t dyn_rst_lifetime = 1; 178static u_int32_t dyn_udp_lifetime = 10; 179static u_int32_t dyn_short_lifetime = 5; 180 181/* 182 * Keepalives are sent if dyn_keepalive is set. They are sent every 183 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 184 * seconds of lifetime of a rule. 185 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 186 * than dyn_keepalive_period. 187 */ 188 189static u_int32_t dyn_keepalive_interval = 20; 190static u_int32_t dyn_keepalive_period = 5; 191static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 192 193static u_int32_t static_count; /* # of static rules */ 194static u_int32_t static_len; /* size in bytes of static rules */ 195static u_int32_t dyn_count; /* # of dynamic rules */ 196static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 197 198SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 199 &dyn_buckets, 0, "Number of dyn. buckets"); 200SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 201 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 202SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 203 &dyn_count, 0, "Number of dyn. rules"); 204SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 205 &dyn_max, 0, "Max number of dyn. rules"); 206SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 207 &static_count, 0, "Number of static rules"); 208SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 209 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 210SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 211 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 212SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 213 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 214SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 215 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 216SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 217 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 218SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 219 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 220SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 221 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 222 223#endif /* SYSCTL_NODE */ 224 225 226static ip_fw_chk_t ipfw_chk; 227 228ip_dn_ruledel_t *ip_dn_ruledel_ptr = NULL; /* hook into dummynet */ 229 230/* 231 * This macro maps an ip pointer into a layer3 header pointer of type T 232 */ 233#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 234 235static __inline int 236icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 237{ 238 int type = L3HDR(struct icmp,ip)->icmp_type; 239 240 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 241} 242 243#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 244 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 245 246static int 247is_icmp_query(struct ip *ip) 248{ 249 int type = L3HDR(struct icmp, ip)->icmp_type; 250 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 251} 252#undef TT 253 254/* 255 * The following checks use two arrays of 8 or 16 bits to store the 256 * bits that we want set or clear, respectively. They are in the 257 * low and high half of cmd->arg1 or cmd->d[0]. 258 * 259 * We scan options and store the bits we find set. We succeed if 260 * 261 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 262 * 263 * The code is sometimes optimized not to store additional variables. 264 */ 265 266static int 267flags_match(ipfw_insn *cmd, u_int8_t bits) 268{ 269 u_char want_clear; 270 bits = ~bits; 271 272 if ( ((cmd->arg1 & 0xff) & bits) != 0) 273 return 0; /* some bits we want set were clear */ 274 want_clear = (cmd->arg1 >> 8) & 0xff; 275 if ( (want_clear & bits) != want_clear) 276 return 0; /* some bits we want clear were set */ 277 return 1; 278} 279 280static int 281ipopts_match(struct ip *ip, ipfw_insn *cmd) 282{ 283 int optlen, bits = 0; 284 u_char *cp = (u_char *)(ip + 1); 285 int x = (ip->ip_hl << 2) - sizeof (struct ip); 286 287 for (; x > 0; x -= optlen, cp += optlen) { 288 int opt = cp[IPOPT_OPTVAL]; 289 290 if (opt == IPOPT_EOL) 291 break; 292 if (opt == IPOPT_NOP) 293 optlen = 1; 294 else { 295 optlen = cp[IPOPT_OLEN]; 296 if (optlen <= 0 || optlen > x) 297 return 0; /* invalid or truncated */ 298 } 299 switch (opt) { 300 301 default: 302 break; 303 304 case IPOPT_LSRR: 305 bits |= IP_FW_IPOPT_LSRR; 306 break; 307 308 case IPOPT_SSRR: 309 bits |= IP_FW_IPOPT_SSRR; 310 break; 311 312 case IPOPT_RR: 313 bits |= IP_FW_IPOPT_RR; 314 break; 315 316 case IPOPT_TS: 317 bits |= IP_FW_IPOPT_TS; 318 break; 319 } 320 } 321 return (flags_match(cmd, bits)); 322} 323 324static int 325tcpopts_match(struct ip *ip, ipfw_insn *cmd) 326{ 327 int optlen, bits = 0; 328 struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 329 u_char *cp = (u_char *)(tcp + 1); 330 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 331 332 for (; x > 0; x -= optlen, cp += optlen) { 333 int opt = cp[0]; 334 if (opt == TCPOPT_EOL) 335 break; 336 if (opt == TCPOPT_NOP) 337 optlen = 1; 338 else { 339 optlen = cp[1]; 340 if (optlen <= 0) 341 break; 342 } 343 344 switch (opt) { 345 346 default: 347 break; 348 349 case TCPOPT_MAXSEG: 350 bits |= IP_FW_TCPOPT_MSS; 351 break; 352 353 case TCPOPT_WINDOW: 354 bits |= IP_FW_TCPOPT_WINDOW; 355 break; 356 357 case TCPOPT_SACK_PERMITTED: 358 case TCPOPT_SACK: 359 bits |= IP_FW_TCPOPT_SACK; 360 break; 361 362 case TCPOPT_TIMESTAMP: 363 bits |= IP_FW_TCPOPT_TS; 364 break; 365 366 case TCPOPT_CC: 367 case TCPOPT_CCNEW: 368 case TCPOPT_CCECHO: 369 bits |= IP_FW_TCPOPT_CC; 370 break; 371 } 372 } 373 return (flags_match(cmd, bits)); 374} 375 376static int 377iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 378{ 379 if (ifp == NULL) /* no iface with this packet, match fails */ 380 return 0; 381 /* Check by name or by IP address */ 382 if (cmd->name[0] != '\0') { /* match by name */ 383 /* Check unit number (-1 is wildcard) */ 384 if (cmd->p.unit != -1 && cmd->p.unit != ifp->if_unit) 385 return(0); 386 /* Check name */ 387 if (!strncmp(ifp->if_name, cmd->name, IFNAMSIZ)) 388 return(1); 389 } else { 390 struct ifaddr *ia; 391 392 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 393 if (ia->ifa_addr == NULL) 394 continue; 395 if (ia->ifa_addr->sa_family != AF_INET) 396 continue; 397 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 398 (ia->ifa_addr))->sin_addr.s_addr) 399 return(1); /* match */ 400 } 401 } 402 return(0); /* no match, fail ... */ 403} 404 405static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 406 407#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 408#define SNP(buf) buf, sizeof(buf) 409 410/* 411 * We enter here when we have a rule with O_LOG. 412 * XXX this function alone takes about 2Kbytes of code! 413 */ 414static void 415ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 416 struct mbuf *m, struct ifnet *oif) 417{ 418 char *action; 419 int limit_reached = 0; 420 char action2[40], proto[48], fragment[28]; 421 422 fragment[0] = '\0'; 423 proto[0] = '\0'; 424 425 if (f == NULL) { /* bogus pkt */ 426 if (verbose_limit != 0 && norule_counter >= verbose_limit) 427 return; 428 norule_counter++; 429 if (norule_counter == verbose_limit) 430 limit_reached = verbose_limit; 431 action = "Refuse"; 432 } else { /* O_LOG is the first action, find the real one */ 433 ipfw_insn *cmd = ACTION_PTR(f); 434 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 435 436 if (l->max_log != 0 && l->log_left == 0) 437 return; 438 l->log_left--; 439 if (l->log_left == 0) 440 limit_reached = l->max_log; 441 cmd += F_LEN(cmd); /* point to first action */ 442 if (cmd->opcode == O_PROB) 443 cmd += F_LEN(cmd); 444 445 action = action2; 446 switch (cmd->opcode) { 447 case O_DENY: 448 action = "Deny"; 449 break; 450 451 case O_REJECT: 452 if (cmd->arg1==ICMP_REJECT_RST) 453 action = "Reset"; 454 else if (cmd->arg1==ICMP_UNREACH_HOST) 455 action = "Reject"; 456 else 457 snprintf(SNPARGS(action2, 0), "Unreach %d", 458 cmd->arg1); 459 break; 460 461 case O_ACCEPT: 462 action = "Accept"; 463 break; 464 case O_COUNT: 465 action = "Count"; 466 break; 467 case O_DIVERT: 468 snprintf(SNPARGS(action2, 0), "Divert %d", 469 cmd->arg1); 470 break; 471 case O_TEE: 472 snprintf(SNPARGS(action2, 0), "Tee %d", 473 cmd->arg1); 474 break; 475 case O_SKIPTO: 476 snprintf(SNPARGS(action2, 0), "SkipTo %d", 477 cmd->arg1); 478 break; 479 case O_PIPE: 480 snprintf(SNPARGS(action2, 0), "Pipe %d", 481 cmd->arg1); 482 break; 483 case O_QUEUE: 484 snprintf(SNPARGS(action2, 0), "Queue %d", 485 cmd->arg1); 486 break; 487 case O_FORWARD_IP: { 488 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 489 int len; 490 491 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 492 inet_ntoa(sa->sa.sin_addr)); 493 if (sa->sa.sin_port) 494 snprintf(SNPARGS(action2, len), ":%d", 495 ntohs(sa->sa.sin_port)); 496 } 497 break; 498 default: 499 action = "UNKNOWN"; 500 break; 501 } 502 } 503 504 if (hlen == 0) { /* non-ip */ 505 snprintf(SNPARGS(proto, 0), "MAC"); 506 } else { 507 struct ip *ip = mtod(m, struct ip *); 508 /* these three are all aliases to the same thing */ 509 struct icmp *const icmp = L3HDR(struct icmp, ip); 510 struct tcphdr *const tcp = (struct tcphdr *)icmp; 511 struct udphdr *const udp = (struct udphdr *)icmp; 512 513 int ip_off, offset, ip_len; 514 515 int len; 516 517 if (eh != NULL) { /* layer 2 packets are as on the wire */ 518 ip_off = ntohs(ip->ip_off); 519 ip_len = ntohs(ip->ip_len); 520 } else { 521 ip_off = ip->ip_off; 522 ip_len = ip->ip_len; 523 } 524 offset = ip_off & IP_OFFMASK; 525 switch (ip->ip_p) { 526 case IPPROTO_TCP: 527 len = snprintf(SNPARGS(proto, 0), "TCP %s", 528 inet_ntoa(ip->ip_src)); 529 if (offset == 0) 530 snprintf(SNPARGS(proto, len), ":%d %s:%d", 531 ntohs(tcp->th_sport), 532 inet_ntoa(ip->ip_dst), 533 ntohs(tcp->th_dport)); 534 else 535 snprintf(SNPARGS(proto, len), " %s", 536 inet_ntoa(ip->ip_dst)); 537 break; 538 539 case IPPROTO_UDP: 540 len = snprintf(SNPARGS(proto, 0), "UDP %s", 541 inet_ntoa(ip->ip_src)); 542 if (offset == 0) 543 snprintf(SNPARGS(proto, len), ":%d %s:%d", 544 ntohs(udp->uh_sport), 545 inet_ntoa(ip->ip_dst), 546 ntohs(udp->uh_dport)); 547 else 548 snprintf(SNPARGS(proto, len), " %s", 549 inet_ntoa(ip->ip_dst)); 550 break; 551 552 case IPPROTO_ICMP: 553 if (offset == 0) 554 len = snprintf(SNPARGS(proto, 0), 555 "ICMP:%u.%u ", 556 icmp->icmp_type, icmp->icmp_code); 557 else 558 len = snprintf(SNPARGS(proto, 0), "ICMP "); 559 len += snprintf(SNPARGS(proto, len), "%s", 560 inet_ntoa(ip->ip_src)); 561 snprintf(SNPARGS(proto, len), " %s", 562 inet_ntoa(ip->ip_dst)); 563 break; 564 565 default: 566 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 567 inet_ntoa(ip->ip_src)); 568 snprintf(SNPARGS(proto, len), " %s", 569 inet_ntoa(ip->ip_dst)); 570 break; 571 } 572 573 if (ip_off & (IP_MF | IP_OFFMASK)) 574 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 575 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 576 offset << 3, 577 (ip_off & IP_MF) ? "+" : ""); 578 } 579 if (oif || m->m_pkthdr.rcvif) 580 log(LOG_SECURITY | LOG_INFO, 581 "ipfw: %d %s %s %s via %s%d%s\n", 582 f ? f->rulenum : -1, 583 action, proto, oif ? "out" : "in", 584 oif ? oif->if_name : m->m_pkthdr.rcvif->if_name, 585 oif ? oif->if_unit : m->m_pkthdr.rcvif->if_unit, 586 fragment); 587 else 588 log(LOG_SECURITY | LOG_INFO, 589 "ipfw: %d %s %s [no if info]%s\n", 590 f ? f->rulenum : -1, 591 action, proto, fragment); 592 if (limit_reached) 593 log(LOG_SECURITY | LOG_NOTICE, 594 "ipfw: limit %d reached on entry %d\n", 595 limit_reached, f ? f->rulenum : -1); 596} 597 598/* 599 * IMPORTANT: the hash function for dynamic rules must be commutative 600 * in source and destination (ip,port), because rules are bidirectional 601 * and we want to find both in the same bucket. 602 */ 603static __inline int 604hash_packet(struct ipfw_flow_id *id) 605{ 606 u_int32_t i; 607 608 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 609 i &= (curr_dyn_buckets - 1); 610 return i; 611} 612 613/** 614 * unlink a dynamic rule from a chain. prev is a pointer to 615 * the previous one, q is a pointer to the rule to delete, 616 * head is a pointer to the head of the queue. 617 * Modifies q and potentially also head. 618 */ 619#define UNLINK_DYN_RULE(prev, head, q) { \ 620 ipfw_dyn_rule *old_q = q; \ 621 \ 622 /* remove a refcount to the parent */ \ 623 if (q->dyn_type == O_LIMIT) \ 624 q->parent->count--; \ 625 DEB(printf("-- unlink entry 0x%08x %d -> 0x%08x %d, %d left\n", \ 626 (q->id.src_ip), (q->id.src_port), \ 627 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 628 if (prev != NULL) \ 629 prev->next = q = q->next; \ 630 else \ 631 head = q = q->next; \ 632 dyn_count--; \ 633 free(old_q, M_IPFW); } 634 635#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 636 637/** 638 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 639 * 640 * If keep_me == NULL, rules are deleted even if not expired, 641 * otherwise only expired rules are removed. 642 * 643 * The value of the second parameter is also used to point to identify 644 * a rule we absolutely do not want to remove (e.g. because we are 645 * holding a reference to it -- this is the case with O_LIMIT_PARENT 646 * rules). The pointer is only used for comparison, so any non-null 647 * value will do. 648 */ 649static void 650remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 651{ 652 static u_int32_t last_remove = 0; 653 654#define FORCE (keep_me == NULL) 655 656 ipfw_dyn_rule *prev, *q; 657 int i, pass = 0, max_pass = 0; 658 659 if (ipfw_dyn_v == NULL || dyn_count == 0) 660 return; 661 /* do not expire more than once per second, it is useless */ 662 if (!FORCE && last_remove == time_second) 663 return; 664 last_remove = time_second; 665 666 /* 667 * because O_LIMIT refer to parent rules, during the first pass only 668 * remove child and mark any pending LIMIT_PARENT, and remove 669 * them in a second pass. 670 */ 671next_pass: 672 for (i = 0 ; i < curr_dyn_buckets ; i++) { 673 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 674 /* 675 * Logic can become complex here, so we split tests. 676 */ 677 if (q == keep_me) 678 goto next; 679 if (rule != NULL && rule != q->rule) 680 goto next; /* not the one we are looking for */ 681 if (q->dyn_type == O_LIMIT_PARENT) { 682 /* 683 * handle parent in the second pass, 684 * record we need one. 685 */ 686 max_pass = 1; 687 if (pass == 0) 688 goto next; 689 if (FORCE && q->count != 0 ) { 690 /* XXX should not happen! */ 691 printf( "OUCH! cannot remove rule," 692 " count %d\n", q->count); 693 } 694 } else { 695 if (!FORCE && 696 !TIME_LEQ( q->expire, time_second )) 697 goto next; 698 } 699 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 700 continue; 701next: 702 prev=q; 703 q=q->next; 704 } 705 } 706 if (pass++ < max_pass) 707 goto next_pass; 708} 709 710 711/** 712 * lookup a dynamic rule. 713 */ 714static ipfw_dyn_rule * 715lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 716 struct tcphdr *tcp) 717{ 718 /* 719 * stateful ipfw extensions. 720 * Lookup into dynamic session queue 721 */ 722#define MATCH_REVERSE 0 723#define MATCH_FORWARD 1 724#define MATCH_NONE 2 725#define MATCH_UNKNOWN 3 726 int i, dir = MATCH_NONE; 727 ipfw_dyn_rule *prev, *q=NULL; 728 729 if (ipfw_dyn_v == NULL) 730 goto done; /* not found */ 731 i = hash_packet( pkt ); 732 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 733 if (q->dyn_type == O_LIMIT_PARENT) 734 goto next; 735 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 736 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 737 continue; 738 } 739 if ( pkt->proto == q->id.proto) { 740 if (pkt->src_ip == q->id.src_ip && 741 pkt->dst_ip == q->id.dst_ip && 742 pkt->src_port == q->id.src_port && 743 pkt->dst_port == q->id.dst_port ) { 744 dir = MATCH_FORWARD; 745 break; 746 } 747 if (pkt->src_ip == q->id.dst_ip && 748 pkt->dst_ip == q->id.src_ip && 749 pkt->src_port == q->id.dst_port && 750 pkt->dst_port == q->id.src_port ) { 751 dir = MATCH_REVERSE; 752 break; 753 } 754 } 755next: 756 prev = q; 757 q = q->next; 758 } 759 if (q == NULL) 760 goto done; /* q = NULL, not found */ 761 762 if ( prev != NULL) { /* found and not in front */ 763 prev->next = q->next; 764 q->next = ipfw_dyn_v[i]; 765 ipfw_dyn_v[i] = q; 766 } 767 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 768 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 769 770#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 771#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 772 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 773 switch (q->state) { 774 case TH_SYN: /* opening */ 775 q->expire = time_second + dyn_syn_lifetime; 776 break; 777 778 case BOTH_SYN: /* move to established */ 779 case BOTH_SYN | TH_FIN : /* one side tries to close */ 780 case BOTH_SYN | (TH_FIN << 8) : 781 if (tcp) { 782#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 783 u_int32_t ack = ntohl(tcp->th_ack); 784 if (dir == MATCH_FORWARD) { 785 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 786 q->ack_fwd = ack; 787 else { /* ignore out-of-sequence */ 788 break; 789 } 790 } else { 791 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 792 q->ack_rev = ack; 793 else { /* ignore out-of-sequence */ 794 break; 795 } 796 } 797 } 798 q->expire = time_second + dyn_ack_lifetime; 799 break; 800 801 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 802 if (dyn_fin_lifetime >= dyn_keepalive_period) 803 dyn_fin_lifetime = dyn_keepalive_period - 1; 804 q->expire = time_second + dyn_fin_lifetime; 805 break; 806 807 default: 808#if 0 809 /* 810 * reset or some invalid combination, but can also 811 * occur if we use keep-state the wrong way. 812 */ 813 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 814 printf("invalid state: 0x%x\n", q->state); 815#endif 816 if (dyn_rst_lifetime >= dyn_keepalive_period) 817 dyn_rst_lifetime = dyn_keepalive_period - 1; 818 q->expire = time_second + dyn_rst_lifetime; 819 break; 820 } 821 } else if (pkt->proto == IPPROTO_UDP) { 822 q->expire = time_second + dyn_udp_lifetime; 823 } else { 824 /* other protocols */ 825 q->expire = time_second + dyn_short_lifetime; 826 } 827done: 828 if (match_direction) 829 *match_direction = dir; 830 return q; 831} 832 833static void 834realloc_dynamic_table(void) 835{ 836 /* 837 * Try reallocation, make sure we have a power of 2 and do 838 * not allow more than 64k entries. In case of overflow, 839 * default to 1024. 840 */ 841 842 if (dyn_buckets > 65536) 843 dyn_buckets = 1024; 844 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 845 dyn_buckets = curr_dyn_buckets; /* reset */ 846 return; 847 } 848 curr_dyn_buckets = dyn_buckets; 849 if (ipfw_dyn_v != NULL) 850 free(ipfw_dyn_v, M_IPFW); 851 for (;;) { 852 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 853 M_IPFW, M_NOWAIT | M_ZERO); 854 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 855 break; 856 curr_dyn_buckets /= 2; 857 } 858} 859 860/** 861 * Install state of type 'type' for a dynamic session. 862 * The hash table contains two type of rules: 863 * - regular rules (O_KEEP_STATE) 864 * - rules for sessions with limited number of sess per user 865 * (O_LIMIT). When they are created, the parent is 866 * increased by 1, and decreased on delete. In this case, 867 * the third parameter is the parent rule and not the chain. 868 * - "parent" rules for the above (O_LIMIT_PARENT). 869 */ 870static ipfw_dyn_rule * 871add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 872{ 873 ipfw_dyn_rule *r; 874 int i; 875 876 if (ipfw_dyn_v == NULL || 877 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 878 realloc_dynamic_table(); 879 if (ipfw_dyn_v == NULL) 880 return NULL; /* failed ! */ 881 } 882 i = hash_packet(id); 883 884 r = malloc(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO); 885 if (r == NULL) { 886 printf ("sorry cannot allocate state\n"); 887 return NULL; 888 } 889 890 /* increase refcount on parent, and set pointer */ 891 if (dyn_type == O_LIMIT) { 892 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 893 if ( parent->dyn_type != O_LIMIT_PARENT) 894 panic("invalid parent"); 895 parent->count++; 896 r->parent = parent; 897 rule = parent->rule; 898 } 899 900 r->id = *id; 901 r->expire = time_second + dyn_syn_lifetime; 902 r->rule = rule; 903 r->dyn_type = dyn_type; 904 r->pcnt = r->bcnt = 0; 905 r->count = 0; 906 907 r->bucket = i; 908 r->next = ipfw_dyn_v[i]; 909 ipfw_dyn_v[i] = r; 910 dyn_count++; 911 DEB(printf("-- add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 912 dyn_type, 913 (r->id.src_ip), (r->id.src_port), 914 (r->id.dst_ip), (r->id.dst_port), 915 dyn_count ); ) 916 return r; 917} 918 919/** 920 * lookup dynamic parent rule using pkt and rule as search keys. 921 * If the lookup fails, then install one. 922 */ 923static ipfw_dyn_rule * 924lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 925{ 926 ipfw_dyn_rule *q; 927 int i; 928 929 if (ipfw_dyn_v) { 930 i = hash_packet( pkt ); 931 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 932 if (q->dyn_type == O_LIMIT_PARENT && 933 rule== q->rule && 934 pkt->proto == q->id.proto && 935 pkt->src_ip == q->id.src_ip && 936 pkt->dst_ip == q->id.dst_ip && 937 pkt->src_port == q->id.src_port && 938 pkt->dst_port == q->id.dst_port) { 939 q->expire = time_second + dyn_short_lifetime; 940 DEB(printf("lookup_dyn_parent found 0x%p\n",q);) 941 return q; 942 } 943 } 944 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 945} 946 947/** 948 * Install dynamic state for rule type cmd->o.opcode 949 * 950 * Returns 1 (failure) if state is not installed because of errors or because 951 * session limitations are enforced. 952 */ 953static int 954install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 955 struct ip_fw_args *args) 956{ 957 static int last_log; 958 959 ipfw_dyn_rule *q; 960 961 DEB(printf("-- install state type %d 0x%08x %u -> 0x%08x %u\n", 962 cmd->o.opcode, 963 (args->f_id.src_ip), (args->f_id.src_port), 964 (args->f_id.dst_ip), (args->f_id.dst_port) );) 965 966 q = lookup_dyn_rule(&args->f_id, NULL, NULL); 967 968 if (q != NULL) { /* should never occur */ 969 if (last_log != time_second) { 970 last_log = time_second; 971 printf(" install_state: entry already present, done\n"); 972 } 973 return 0; 974 } 975 976 if (dyn_count >= dyn_max) 977 /* 978 * Run out of slots, try to remove any expired rule. 979 */ 980 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 981 982 if (dyn_count >= dyn_max) { 983 if (last_log != time_second) { 984 last_log = time_second; 985 printf("install_state: Too many dynamic rules\n"); 986 } 987 return 1; /* cannot install, notify caller */ 988 } 989 990 switch (cmd->o.opcode) { 991 case O_KEEP_STATE: /* bidir rule */ 992 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 993 break; 994 995 case O_LIMIT: /* limit number of sessions */ 996 { 997 u_int16_t limit_mask = cmd->limit_mask; 998 struct ipfw_flow_id id; 999 ipfw_dyn_rule *parent; 1000 1001 DEB(printf("installing dyn-limit rule %d\n", cmd->conn_limit);) 1002 1003 id.dst_ip = id.src_ip = 0; 1004 id.dst_port = id.src_port = 0; 1005 id.proto = args->f_id.proto; 1006 1007 if (limit_mask & DYN_SRC_ADDR) 1008 id.src_ip = args->f_id.src_ip; 1009 if (limit_mask & DYN_DST_ADDR) 1010 id.dst_ip = args->f_id.dst_ip; 1011 if (limit_mask & DYN_SRC_PORT) 1012 id.src_port = args->f_id.src_port; 1013 if (limit_mask & DYN_DST_PORT) 1014 id.dst_port = args->f_id.dst_port; 1015 parent = lookup_dyn_parent(&id, rule); 1016 if (parent == NULL) { 1017 printf("add parent failed\n"); 1018 return 1; 1019 } 1020 if (parent->count >= cmd->conn_limit) { 1021 /* 1022 * See if we can remove some expired rule. 1023 */ 1024 remove_dyn_rule(rule, parent); 1025 if (parent->count >= cmd->conn_limit) { 1026 if (fw_verbose && last_log != time_second) { 1027 last_log = time_second; 1028 log(LOG_SECURITY | LOG_DEBUG, 1029 "drop session, too many entries\n"); 1030 } 1031 return 1; 1032 } 1033 } 1034 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1035 } 1036 break; 1037 default: 1038 printf("unknown dynamic rule type %u\n", cmd->o.opcode); 1039 return 1; 1040 } 1041 lookup_dyn_rule(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1042 return 0; 1043} 1044 1045/* 1046 * Transmit a TCP packet, containing either a RST or a keepalive. 1047 * When flags & TH_RST, we are sending a RST packet, because of a 1048 * "reset" action matched the packet. 1049 * Otherwise we are sending a keepalive, and flags & TH_ 1050 */ 1051static void 1052send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1053{ 1054 struct mbuf *m; 1055 struct ip *ip; 1056 struct tcphdr *tcp; 1057 struct route sro; /* fake route */ 1058 1059 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1060 if (m == 0) 1061 return; 1062 m->m_pkthdr.rcvif = (struct ifnet *)0; 1063 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1064 m->m_data += max_linkhdr; 1065 1066 ip = mtod(m, struct ip *); 1067 bzero(ip, m->m_len); 1068 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1069 ip->ip_p = IPPROTO_TCP; 1070 tcp->th_off = 5; 1071 /* 1072 * Assume we are sending a RST (or a keepalive in the reverse 1073 * direction), swap src and destination addresses and ports. 1074 */ 1075 ip->ip_src.s_addr = htonl(id->dst_ip); 1076 ip->ip_dst.s_addr = htonl(id->src_ip); 1077 tcp->th_sport = htons(id->dst_port); 1078 tcp->th_dport = htons(id->src_port); 1079 if (flags & TH_RST) { /* we are sending a RST */ 1080 if (flags & TH_ACK) { 1081 tcp->th_seq = htonl(ack); 1082 tcp->th_ack = htonl(0); 1083 tcp->th_flags = TH_RST; 1084 } else { 1085 if (flags & TH_SYN) 1086 seq++; 1087 tcp->th_seq = htonl(0); 1088 tcp->th_ack = htonl(seq); 1089 tcp->th_flags = TH_RST | TH_ACK; 1090 } 1091 } else { 1092 /* 1093 * We are sending a keepalive. flags & TH_SYN determines 1094 * the direction, forward if set, reverse if clear. 1095 * NOTE: seq and ack are always assumed to be correct 1096 * as set by the caller. This may be confusing... 1097 */ 1098 if (flags & TH_SYN) { 1099 /* 1100 * we have to rewrite the correct addresses! 1101 */ 1102 ip->ip_dst.s_addr = htonl(id->dst_ip); 1103 ip->ip_src.s_addr = htonl(id->src_ip); 1104 tcp->th_dport = htons(id->dst_port); 1105 tcp->th_sport = htons(id->src_port); 1106 } 1107 tcp->th_seq = htonl(seq); 1108 tcp->th_ack = htonl(ack); 1109 tcp->th_flags = TH_ACK; 1110 } 1111 /* 1112 * set ip_len to the payload size so we can compute 1113 * the tcp checksum on the pseudoheader 1114 * XXX check this, could save a couple of words ? 1115 */ 1116 ip->ip_len = htons(sizeof(struct tcphdr)); 1117 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1118 /* 1119 * now fill fields left out earlier 1120 */ 1121 ip->ip_ttl = ip_defttl; 1122 ip->ip_len = m->m_pkthdr.len; 1123 bzero (&sro, sizeof (sro)); 1124 ip_rtaddr(ip->ip_dst, &sro); 1125 m->m_flags |= M_SKIP_FIREWALL; 1126 ip_output(m, NULL, &sro, 0, NULL, NULL); 1127 if (sro.ro_rt) 1128 RTFREE(sro.ro_rt); 1129} 1130 1131/* 1132 * sends a reject message, consuming the mbuf passed as an argument. 1133 */ 1134static void 1135send_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 1136{ 1137 1138 if (code != ICMP_REJECT_RST) /* Send an ICMP unreach */ 1139 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1140 else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1141 struct tcphdr *const tcp = 1142 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1143 if ( (tcp->th_flags & TH_RST) == 0) 1144 send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1145 ntohl(tcp->th_ack), 1146 tcp->th_flags | TH_RST); 1147 m_freem(args->m); 1148 } else 1149 m_freem(args->m); 1150 args->m = NULL; 1151} 1152 1153/** 1154 * 1155 * Given an ip_fw *, lookup_next_rule will return a pointer 1156 * to the next rule, which can be either the jump 1157 * target (for skipto instructions) or the next one in the list (in 1158 * all other cases including a missing jump target). 1159 * The result is also written in the "next_rule" field of the rule. 1160 * Backward jumps are not allowed, so start looking from the next 1161 * rule... 1162 * 1163 * This never returns NULL -- in case we do not have an exact match, 1164 * the next rule is returned. When the ruleset is changed, 1165 * pointers are flushed so we are always correct. 1166 */ 1167 1168static struct ip_fw * 1169lookup_next_rule(struct ip_fw *me) 1170{ 1171 struct ip_fw *rule = NULL; 1172 ipfw_insn *cmd; 1173 1174 /* look for action, in case it is a skipto */ 1175 cmd = ACTION_PTR(me); 1176 if ( cmd->opcode == O_SKIPTO ) 1177 for (rule = me->next; rule ; rule = rule->next) 1178 if (rule->rulenum >= cmd->arg1) 1179 break; 1180 if (rule == NULL) /* failure or not a skipto */ 1181 rule = me->next; 1182 me->next_rule = rule; 1183 return rule; 1184} 1185 1186/* 1187 * The main check routine for the firewall. 1188 * 1189 * All arguments are in args so we can modify them and return them 1190 * back to the caller. 1191 * 1192 * Parameters: 1193 * 1194 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1195 * Starts with the IP header. 1196 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1197 * args->oif Outgoing interface, or NULL if packet is incoming. 1198 * The incoming interface is in the mbuf. (in) 1199 * args->divert_rule (in/out) 1200 * Skip up to the first rule past this rule number; 1201 * upon return, non-zero port number for divert or tee. 1202 * 1203 * args->rule Pointer to the last matching rule (in/out) 1204 * args->next_hop Socket we are forwarding to (out). 1205 * args->f_id Addresses grabbed from the packet (out) 1206 * 1207 * Return value: 1208 * 1209 * IP_FW_PORT_DENY_FLAG the packet must be dropped. 1210 * 0 The packet is to be accepted and routed normally OR 1211 * the packet was denied/rejected and has been dropped; 1212 * in the latter case, *m is equal to NULL upon return. 1213 * port Divert the packet to port, with these caveats: 1214 * 1215 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead 1216 * of diverting it (ie, 'ipfw tee'). 1217 * 1218 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower 1219 * 16 bits as a dummynet pipe number instead of diverting 1220 */ 1221 1222static int 1223ipfw_chk(struct ip_fw_args *args) 1224{ 1225 /* 1226 * Local variables hold state during the processing of a packet. 1227 * 1228 * IMPORTANT NOTE: to speed up the processing of rules, there 1229 * are some assumption on the values of the variables, which 1230 * are documented here. Should you change them, please check 1231 * the implementation of the various instructions to make sure 1232 * that they still work. 1233 * 1234 * args->eh The MAC header. It is non-null for a layer2 1235 * packet, it is NULL for a layer-3 packet. 1236 * 1237 * m | args->m Pointer to the mbuf, as received from the caller. 1238 * It may change if ipfw_chk() does an m_pullup, or if it 1239 * consumes the packet because it calls send_reject(). 1240 * XXX This has to change, so that ipfw_chk() never modifies 1241 * or consumes the buffer. 1242 * ip is simply an alias of the value of m, and it is kept 1243 * in sync with it (the packet is supposed to start with 1244 * the ip header). 1245 */ 1246 struct mbuf *m = args->m; 1247 struct ip *ip = mtod(m, struct ip *); 1248 1249 /* 1250 * oif | args->oif If NULL, ipfw_chk has been called on the 1251 * inbound path (ether_input, bdg_forward, ip_input). 1252 * If non-NULL, ipfw_chk has been called on the outbound path 1253 * (ether_output, ip_output). 1254 */ 1255 struct ifnet *oif = args->oif; 1256 1257 struct ip_fw *f = NULL; /* matching rule */ 1258 int retval = 0; 1259 1260 /* 1261 * hlen The length of the IPv4 header. 1262 * hlen >0 means we have an IPv4 packet. 1263 */ 1264 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1265 1266 /* 1267 * offset The offset of a fragment. offset != 0 means that 1268 * we have a fragment at this offset of an IPv4 packet. 1269 * offset == 0 means that (if this is an IPv4 packet) 1270 * this is the first or only fragment. 1271 */ 1272 u_short offset = 0; 1273 1274 /* 1275 * Local copies of addresses. They are only valid if we have 1276 * an IP packet. 1277 * 1278 * proto The protocol. Set to 0 for non-ip packets, 1279 * or to the protocol read from the packet otherwise. 1280 * proto != 0 means that we have an IPv4 packet. 1281 * 1282 * src_port, dst_port port numbers, in HOST format. Only 1283 * valid for TCP and UDP packets. 1284 * 1285 * src_ip, dst_ip ip addresses, in NETWORK format. 1286 * Only valid for IPv4 packets. 1287 */ 1288 u_int8_t proto; 1289 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1290 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1291 u_int16_t ip_len=0; 1292 int dyn_dir = MATCH_UNKNOWN; 1293 ipfw_dyn_rule *q = NULL; 1294 1295 if (m->m_flags & M_SKIP_FIREWALL) 1296 return 0; /* accept */ 1297 /* 1298 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1299 * MATCH_NONE when checked and not matched (q = NULL), 1300 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1301 */ 1302 1303 if (args->eh == NULL || /* layer 3 packet */ 1304 ( m->m_pkthdr.len >= sizeof(struct ip) && 1305 ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 1306 hlen = ip->ip_hl << 2; 1307 1308 /* 1309 * Collect parameters into local variables for faster matching. 1310 */ 1311 if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 1312 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1313 goto after_ip_checks; 1314 } 1315 1316 proto = args->f_id.proto = ip->ip_p; 1317 src_ip = ip->ip_src; 1318 dst_ip = ip->ip_dst; 1319 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1320 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1321 ip_len = ntohs(ip->ip_len); 1322 } else { 1323 offset = ip->ip_off & IP_OFFMASK; 1324 ip_len = ip->ip_len; 1325 } 1326 1327#define PULLUP_TO(len) \ 1328 do { \ 1329 if ((m)->m_len < (len)) { \ 1330 args->m = m = m_pullup(m, (len)); \ 1331 if (m == 0) \ 1332 goto pullup_failed; \ 1333 ip = mtod(m, struct ip *); \ 1334 } \ 1335 } while (0) 1336 1337 if (offset == 0) { 1338 switch (proto) { 1339 case IPPROTO_TCP: 1340 { 1341 struct tcphdr *tcp; 1342 1343 PULLUP_TO(hlen + sizeof(struct tcphdr)); 1344 tcp = L3HDR(struct tcphdr, ip); 1345 dst_port = tcp->th_dport; 1346 src_port = tcp->th_sport; 1347 args->f_id.flags = tcp->th_flags; 1348 } 1349 break; 1350 1351 case IPPROTO_UDP: 1352 { 1353 struct udphdr *udp; 1354 1355 PULLUP_TO(hlen + sizeof(struct udphdr)); 1356 udp = L3HDR(struct udphdr, ip); 1357 dst_port = udp->uh_dport; 1358 src_port = udp->uh_sport; 1359 } 1360 break; 1361 1362 case IPPROTO_ICMP: 1363 PULLUP_TO(hlen + 4); /* type, code and checksum. */ 1364 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 1365 break; 1366 1367 default: 1368 break; 1369 } 1370#undef PULLUP_TO 1371 } 1372 1373 args->f_id.src_ip = ntohl(src_ip.s_addr); 1374 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1375 args->f_id.src_port = src_port = ntohs(src_port); 1376 args->f_id.dst_port = dst_port = ntohs(dst_port); 1377 1378after_ip_checks: 1379 if (args->rule) { 1380 /* 1381 * Packet has already been tagged. Look for the next rule 1382 * to restart processing. 1383 * 1384 * If fw_one_pass != 0 then just accept it. 1385 * XXX should not happen here, but optimized out in 1386 * the caller. 1387 */ 1388 if (fw_one_pass) 1389 return 0; 1390 1391 f = args->rule->next_rule; 1392 if (f == NULL) 1393 f = lookup_next_rule(args->rule); 1394 } else { 1395 /* 1396 * Find the starting rule. It can be either the first 1397 * one, or the one after divert_rule if asked so. 1398 */ 1399 int skipto = args->divert_rule; 1400 1401 f = layer3_chain; 1402 if (args->eh == NULL && skipto != 0) { 1403 if (skipto >= IPFW_DEFAULT_RULE) 1404 return(IP_FW_PORT_DENY_FLAG); /* invalid */ 1405 while (f && f->rulenum <= skipto) 1406 f = f->next; 1407 if (f == NULL) /* drop packet */ 1408 return(IP_FW_PORT_DENY_FLAG); 1409 } 1410 } 1411 args->divert_rule = 0; /* reset to avoid confusion later */ 1412 1413 /* 1414 * Now scan the rules, and parse microinstructions for each rule. 1415 */ 1416 for (; f; f = f->next) { 1417 int l, cmdlen; 1418 ipfw_insn *cmd; 1419 int skip_or; /* skip rest of OR block */ 1420 1421again: 1422 if (set_disable & (1 << f->set) ) 1423 continue; 1424 1425 skip_or = 0; 1426 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1427 l -= cmdlen, cmd += cmdlen) { 1428 int match; 1429 1430 /* 1431 * check_body is a jump target used when we find a 1432 * CHECK_STATE, and need to jump to the body of 1433 * the target rule. 1434 */ 1435 1436check_body: 1437 cmdlen = F_LEN(cmd); 1438 /* 1439 * An OR block (insn_1 || .. || insn_n) has the 1440 * F_OR bit set in all but the last instruction. 1441 * The first match will set "skip_or", and cause 1442 * the following instructions to be skipped until 1443 * past the one with the F_OR bit clear. 1444 */ 1445 if (skip_or) { /* skip this instruction */ 1446 if ((cmd->len & F_OR) == 0) 1447 skip_or = 0; /* next one is good */ 1448 continue; 1449 } 1450 match = 0; /* set to 1 if we succeed */ 1451 1452 switch (cmd->opcode) { 1453 /* 1454 * The first set of opcodes compares the packet's 1455 * fields with some pattern, setting 'match' if a 1456 * match is found. At the end of the loop there is 1457 * logic to deal with F_NOT and F_OR flags associated 1458 * with the opcode. 1459 */ 1460 case O_NOP: 1461 match = 1; 1462 break; 1463 1464 case O_FORWARD_MAC: 1465 printf("ipfw: opcode %d unimplemented\n", 1466 cmd->opcode); 1467 break; 1468 1469 case O_GID: 1470 case O_UID: 1471 /* 1472 * We only check offset == 0 && proto != 0, 1473 * as this ensures that we have an IPv4 1474 * packet with the ports info. 1475 */ 1476 if (offset!=0) 1477 break; 1478 { 1479 struct inpcbinfo *pi; 1480 int wildcard; 1481 struct inpcb *pcb; 1482 1483 if (proto == IPPROTO_TCP) { 1484 wildcard = 0; 1485 pi = &tcbinfo; 1486 } else if (proto == IPPROTO_UDP) { 1487 wildcard = 1; 1488 pi = &udbinfo; 1489 } else 1490 break; 1491 1492 pcb = (oif) ? 1493 in_pcblookup_hash(pi, 1494 dst_ip, htons(dst_port), 1495 src_ip, htons(src_port), 1496 wildcard, oif) : 1497 in_pcblookup_hash(pi, 1498 src_ip, htons(src_port), 1499 dst_ip, htons(dst_port), 1500 wildcard, NULL); 1501 1502 if (pcb == NULL || pcb->inp_socket == NULL) 1503 break; 1504#if __FreeBSD_version < 500034 1505#define socheckuid(a,b) ((a)->so_cred->cr_uid == (b)) 1506#endif 1507 if (cmd->opcode == O_UID) { 1508 match = 1509 socheckuid(pcb->inp_socket, 1510 (uid_t)((ipfw_insn_u32 *)cmd)->d[0]); 1511 } else { 1512 match = groupmember( 1513 (uid_t)((ipfw_insn_u32 *)cmd)->d[0], 1514 pcb->inp_socket->so_cred); 1515 } 1516 } 1517 break; 1518 1519 case O_RECV: 1520 match = iface_match(m->m_pkthdr.rcvif, 1521 (ipfw_insn_if *)cmd); 1522 break; 1523 1524 case O_XMIT: 1525 match = iface_match(oif, (ipfw_insn_if *)cmd); 1526 break; 1527 1528 case O_VIA: 1529 match = iface_match(oif ? oif : 1530 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1531 break; 1532 1533 case O_MACADDR2: 1534 if (args->eh != NULL) { /* have MAC header */ 1535 u_int32_t *want = (u_int32_t *) 1536 ((ipfw_insn_mac *)cmd)->addr; 1537 u_int32_t *mask = (u_int32_t *) 1538 ((ipfw_insn_mac *)cmd)->mask; 1539 u_int32_t *hdr = (u_int32_t *)args->eh; 1540 1541 match = 1542 ( want[0] == (hdr[0] & mask[0]) && 1543 want[1] == (hdr[1] & mask[1]) && 1544 want[2] == (hdr[2] & mask[2]) ); 1545 } 1546 break; 1547 1548 case O_MAC_TYPE: 1549 if (args->eh != NULL) { 1550 u_int16_t t = 1551 ntohs(args->eh->ether_type); 1552 u_int16_t *p = 1553 ((ipfw_insn_u16 *)cmd)->ports; 1554 int i; 1555 1556 for (i = cmdlen - 1; !match && i>0; 1557 i--, p += 2) 1558 match = (t>=p[0] && t<=p[1]); 1559 } 1560 break; 1561 1562 case O_FRAG: 1563 match = (hlen > 0 && offset != 0); 1564 break; 1565 1566 case O_IN: /* "out" is "not in" */ 1567 match = (oif == NULL); 1568 break; 1569 1570 case O_LAYER2: 1571 match = (args->eh != NULL); 1572 break; 1573 1574 case O_PROTO: 1575 /* 1576 * We do not allow an arg of 0 so the 1577 * check of "proto" only suffices. 1578 */ 1579 match = (proto == cmd->arg1); 1580 break; 1581 1582 case O_IP_SRC: 1583 match = (hlen > 0 && 1584 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1585 src_ip.s_addr); 1586 break; 1587 1588 case O_IP_SRC_MASK: 1589 match = (hlen > 0 && 1590 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1591 (src_ip.s_addr & 1592 ((ipfw_insn_ip *)cmd)->mask.s_addr)); 1593 break; 1594 1595 case O_IP_SRC_ME: 1596 if (hlen > 0) { 1597 struct ifnet *tif; 1598 1599 INADDR_TO_IFP(src_ip, tif); 1600 match = (tif != NULL); 1601 } 1602 break; 1603 1604 case O_IP_DST_SET: 1605 case O_IP_SRC_SET: 1606 if (hlen > 0) { 1607 u_int32_t *d = (u_int32_t *)(cmd+1); 1608 u_int32_t addr = 1609 cmd->opcode == O_IP_DST_SET ? 1610 args->f_id.dst_ip : 1611 args->f_id.src_ip; 1612 1613 if (addr < d[0]) 1614 break; 1615 addr -= d[0]; /* subtract base */ 1616 match = (addr < cmd->arg1) && 1617 ( d[ 1 + (addr>>5)] & 1618 (1<<(addr & 0x1f)) ); 1619 } 1620 break; 1621 1622 case O_IP_DST: 1623 match = (hlen > 0 && 1624 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1625 dst_ip.s_addr); 1626 break; 1627 1628 case O_IP_DST_MASK: 1629 match = (hlen > 0) && 1630 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1631 (dst_ip.s_addr & 1632 ((ipfw_insn_ip *)cmd)->mask.s_addr)); 1633 break; 1634 1635 case O_IP_DST_ME: 1636 if (hlen > 0) { 1637 struct ifnet *tif; 1638 1639 INADDR_TO_IFP(dst_ip, tif); 1640 match = (tif != NULL); 1641 } 1642 break; 1643 1644 case O_IP_SRCPORT: 1645 case O_IP_DSTPORT: 1646 /* 1647 * offset == 0 && proto != 0 is enough 1648 * to guarantee that we have an IPv4 1649 * packet with port info. 1650 */ 1651 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 1652 && offset == 0) { 1653 u_int16_t x = 1654 (cmd->opcode == O_IP_SRCPORT) ? 1655 src_port : dst_port ; 1656 u_int16_t *p = 1657 ((ipfw_insn_u16 *)cmd)->ports; 1658 int i; 1659 1660 for (i = cmdlen - 1; !match && i>0; 1661 i--, p += 2) 1662 match = (x>=p[0] && x<=p[1]); 1663 } 1664 break; 1665 1666 case O_ICMPTYPE: 1667 match = (offset == 0 && proto==IPPROTO_ICMP && 1668 icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 1669 break; 1670 1671 case O_IPOPT: 1672 match = (hlen > 0 && ipopts_match(ip, cmd) ); 1673 break; 1674 1675 case O_IPVER: 1676 match = (hlen > 0 && cmd->arg1 == ip->ip_v); 1677 break; 1678 1679 case O_IPTTL: 1680 match = (hlen > 0 && cmd->arg1 == ip->ip_ttl); 1681 break; 1682 1683 case O_IPID: 1684 match = (hlen > 0 && 1685 cmd->arg1 == ntohs(ip->ip_id)); 1686 break; 1687 1688 case O_IPLEN: 1689 match = (hlen > 0 && cmd->arg1 == ip_len); 1690 break; 1691 1692 case O_IPPRECEDENCE: 1693 match = (hlen > 0 && 1694 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 1695 break; 1696 1697 case O_IPTOS: 1698 match = (hlen > 0 && 1699 flags_match(cmd, ip->ip_tos)); 1700 break; 1701 1702 case O_TCPFLAGS: 1703 match = (proto == IPPROTO_TCP && offset == 0 && 1704 flags_match(cmd, 1705 L3HDR(struct tcphdr,ip)->th_flags)); 1706 break; 1707 1708 case O_TCPOPTS: 1709 match = (proto == IPPROTO_TCP && offset == 0 && 1710 tcpopts_match(ip, cmd)); 1711 break; 1712 1713 case O_TCPSEQ: 1714 match = (proto == IPPROTO_TCP && offset == 0 && 1715 ((ipfw_insn_u32 *)cmd)->d[0] == 1716 L3HDR(struct tcphdr,ip)->th_seq); 1717 break; 1718 1719 case O_TCPACK: 1720 match = (proto == IPPROTO_TCP && offset == 0 && 1721 ((ipfw_insn_u32 *)cmd)->d[0] == 1722 L3HDR(struct tcphdr,ip)->th_ack); 1723 break; 1724 1725 case O_TCPWIN: 1726 match = (proto == IPPROTO_TCP && offset == 0 && 1727 cmd->arg1 == 1728 L3HDR(struct tcphdr,ip)->th_win); 1729 break; 1730 1731 case O_ESTAB: 1732 /* reject packets which have SYN only */ 1733 /* XXX should i also check for TH_ACK ? */ 1734 match = (proto == IPPROTO_TCP && offset == 0 && 1735 (L3HDR(struct tcphdr,ip)->th_flags & 1736 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 1737 break; 1738 1739 case O_LOG: 1740 if (fw_verbose) 1741 ipfw_log(f, hlen, args->eh, m, oif); 1742 match = 1; 1743 break; 1744 1745 case O_PROB: 1746 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 1747 break; 1748 1749 /* 1750 * The second set of opcodes represents 'actions', 1751 * i.e. the terminal part of a rule once the packet 1752 * matches all previous patterns. 1753 * Typically there is only one action for each rule, 1754 * and the opcode is stored at the end of the rule 1755 * (but there are exceptions -- see below). 1756 * 1757 * In general, here we set retval and terminate the 1758 * outer loop (would be a 'break 3' in some language, 1759 * but we need to do a 'goto done'). 1760 * 1761 * Exceptions: 1762 * O_COUNT and O_SKIPTO actions: 1763 * instead of terminating, we jump to the next rule 1764 * ('goto next_rule', equivalent to a 'break 2'), 1765 * or to the SKIPTO target ('goto again' after 1766 * having set f, cmd and l), respectively. 1767 * 1768 * O_LIMIT and O_KEEP_STATE: these opcodes are 1769 * not real 'actions', and are stored right 1770 * before the 'action' part of the rule. 1771 * These opcodes try to install an entry in the 1772 * state tables; if successful, we continue with 1773 * the next opcode (match=1; break;), otherwise 1774 * the packet * must be dropped 1775 * ('goto done' after setting retval); 1776 * 1777 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 1778 * cause a lookup of the state table, and a jump 1779 * to the 'action' part of the parent rule 1780 * ('goto check_body') if an entry is found, or 1781 * (CHECK_STATE only) a jump to the next rule if 1782 * the entry is not found ('goto next_rule'). 1783 * The result of the lookup is cached to make 1784 * further instances of these opcodes are 1785 * effectively NOPs. 1786 */ 1787 case O_LIMIT: 1788 case O_KEEP_STATE: 1789 if (install_state(f, 1790 (ipfw_insn_limit *)cmd, args)) { 1791 retval = IP_FW_PORT_DENY_FLAG; 1792 goto done; /* error/limit violation */ 1793 } 1794 match = 1; 1795 break; 1796 1797 case O_PROBE_STATE: 1798 case O_CHECK_STATE: 1799 /* 1800 * dynamic rules are checked at the first 1801 * keep-state or check-state occurrence, 1802 * with the result being stored in dyn_dir. 1803 * The compiler introduces a PROBE_STATE 1804 * instruction for us when we have a 1805 * KEEP_STATE (because PROBE_STATE needs 1806 * to be run first). 1807 */ 1808 if (dyn_dir == MATCH_UNKNOWN && 1809 (q = lookup_dyn_rule(&args->f_id, 1810 &dyn_dir, proto == IPPROTO_TCP ? 1811 L3HDR(struct tcphdr, ip) : NULL)) 1812 != NULL) { 1813 /* 1814 * Found dynamic entry, update stats 1815 * and jump to the 'action' part of 1816 * the parent rule. 1817 */ 1818 q->pcnt++; 1819 q->bcnt += ip_len; 1820 f = q->rule; 1821 cmd = ACTION_PTR(f); 1822 l = f->cmd_len - f->act_ofs; 1823 goto check_body; 1824 } 1825 /* 1826 * Dynamic entry not found. If CHECK_STATE, 1827 * skip to next rule, if PROBE_STATE just 1828 * ignore and continue with next opcode. 1829 */ 1830 if (cmd->opcode == O_CHECK_STATE) 1831 goto next_rule; 1832 match = 1; 1833 break; 1834 1835 case O_ACCEPT: 1836 retval = 0; /* accept */ 1837 goto done; 1838 1839 case O_PIPE: 1840 case O_QUEUE: 1841 args->rule = f; /* report matching rule */ 1842 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG; 1843 goto done; 1844 1845 case O_DIVERT: 1846 case O_TEE: 1847 if (args->eh) /* not on layer 2 */ 1848 break; 1849 args->divert_rule = f->rulenum; 1850 retval = (cmd->opcode == O_DIVERT) ? 1851 cmd->arg1 : 1852 cmd->arg1 | IP_FW_PORT_TEE_FLAG; 1853 goto done; 1854 1855 case O_COUNT: 1856 case O_SKIPTO: 1857 f->pcnt++; /* update stats */ 1858 f->bcnt += ip_len; 1859 f->timestamp = time_second; 1860 if (cmd->opcode == O_COUNT) 1861 goto next_rule; 1862 /* handle skipto */ 1863 if (f->next_rule == NULL) 1864 lookup_next_rule(f); 1865 f = f->next_rule; 1866 goto again; 1867 1868 case O_REJECT: 1869 /* 1870 * Drop the packet and send a reject notice 1871 * if the packet is not ICMP (or is an ICMP 1872 * query), and it is not multicast/broadcast. 1873 */ 1874 if (hlen > 0 && 1875 (proto != IPPROTO_ICMP || 1876 is_icmp_query(ip)) && 1877 !(m->m_flags & (M_BCAST|M_MCAST)) && 1878 !IN_MULTICAST(dst_ip.s_addr)) { 1879 send_reject(args, cmd->arg1, 1880 offset,ip_len); 1881 m = args->m; 1882 } 1883 /* FALLTHROUGH */ 1884 case O_DENY: 1885 retval = IP_FW_PORT_DENY_FLAG; 1886 goto done; 1887 1888 case O_FORWARD_IP: 1889 if (args->eh) /* not valid on layer2 pkts */ 1890 break; 1891 if (!q || dyn_dir == MATCH_FORWARD) 1892 args->next_hop = 1893 &((ipfw_insn_sa *)cmd)->sa; 1894 retval = 0; 1895 goto done; 1896 1897 default: 1898 panic("-- unknown opcode %d\n", cmd->opcode); 1899 } /* end of switch() on opcodes */ 1900 1901 if (cmd->len & F_NOT) 1902 match = !match; 1903 1904 if (match) { 1905 if (cmd->len & F_OR) 1906 skip_or = 1; 1907 } else { 1908 if (!(cmd->len & F_OR)) /* not an OR block, */ 1909 break; /* try next rule */ 1910 } 1911 1912 } /* end of inner for, scan opcodes */ 1913 1914next_rule:; /* try next rule */ 1915 1916 } /* end of outer for, scan rules */ 1917 printf("+++ ipfw: ouch!, skip past end of rules, denying packet\n"); 1918 return(IP_FW_PORT_DENY_FLAG); 1919 1920done: 1921 /* Update statistics */ 1922 f->pcnt++; 1923 f->bcnt += ip_len; 1924 f->timestamp = time_second; 1925 return retval; 1926 1927pullup_failed: 1928 if (fw_verbose) 1929 printf("pullup failed\n"); 1930 return(IP_FW_PORT_DENY_FLAG); 1931} 1932 1933/* 1934 * When a rule is added/deleted, clear the next_rule pointers in all rules. 1935 * These will be reconstructed on the fly as packets are matched. 1936 * Must be called at splimp(). 1937 */ 1938static void 1939flush_rule_ptrs(void) 1940{ 1941 struct ip_fw *rule; 1942 1943 for (rule = layer3_chain; rule; rule = rule->next) 1944 rule->next_rule = NULL; 1945} 1946 1947/* 1948 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 1949 * pipe/queue, or to all of them (match == NULL). 1950 * Must be called at splimp(). 1951 */ 1952void 1953flush_pipe_ptrs(struct dn_flow_set *match) 1954{ 1955 struct ip_fw *rule; 1956 1957 for (rule = layer3_chain; rule; rule = rule->next) { 1958 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 1959 1960 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 1961 continue; 1962 if (match == NULL || cmd->pipe_ptr == match) 1963 cmd->pipe_ptr = NULL; 1964 } 1965} 1966 1967/* 1968 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 1969 * possibly create a rule number and add the rule to the list. 1970 * Update the rule_number in the input struct so the caller knows it as well. 1971 */ 1972static int 1973add_rule(struct ip_fw **head, struct ip_fw *input_rule) 1974{ 1975 struct ip_fw *rule, *f, *prev; 1976 int s; 1977 int l = RULESIZE(input_rule); 1978 1979 if (*head == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 1980 return (EINVAL); 1981 1982 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 1983 if (rule == NULL) 1984 return (ENOSPC); 1985 1986 bcopy(input_rule, rule, l); 1987 1988 rule->next = NULL; 1989 rule->next_rule = NULL; 1990 1991 rule->pcnt = 0; 1992 rule->bcnt = 0; 1993 rule->timestamp = 0; 1994 1995 s = splimp(); 1996 1997 if (*head == NULL) { /* default rule */ 1998 *head = rule; 1999 goto done; 2000 } 2001 2002 /* 2003 * If rulenum is 0, find highest numbered rule before the 2004 * default rule, and add autoinc_step 2005 */ 2006 if (autoinc_step < 1) 2007 autoinc_step = 1; 2008 else if (autoinc_step > 1000) 2009 autoinc_step = 1000; 2010 if (rule->rulenum == 0) { 2011 /* 2012 * locate the highest numbered rule before default 2013 */ 2014 for (f = *head; f; f = f->next) { 2015 if (f->rulenum == IPFW_DEFAULT_RULE) 2016 break; 2017 rule->rulenum = f->rulenum; 2018 } 2019 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2020 rule->rulenum += autoinc_step; 2021 input_rule->rulenum = rule->rulenum; 2022 } 2023 2024 /* 2025 * Now insert the new rule in the right place in the sorted list. 2026 */ 2027 for (prev = NULL, f = *head; f; prev = f, f = f->next) { 2028 if (f->rulenum > rule->rulenum) { /* found the location */ 2029 if (prev) { 2030 rule->next = f; 2031 prev->next = rule; 2032 } else { /* head insert */ 2033 rule->next = *head; 2034 *head = rule; 2035 } 2036 break; 2037 } 2038 } 2039 flush_rule_ptrs(); 2040done: 2041 static_count++; 2042 static_len += l; 2043 splx(s); 2044 DEB(printf("++ installed rule %d, static count now %d\n", 2045 rule->rulenum, static_count);) 2046 return (0); 2047} 2048 2049/** 2050 * Free storage associated with a static rule (including derived 2051 * dynamic rules). 2052 * The caller is in charge of clearing rule pointers to avoid 2053 * dangling pointers. 2054 * @return a pointer to the next entry. 2055 * Arguments are not checked, so they better be correct. 2056 * Must be called at splimp(). 2057 */ 2058static struct ip_fw * 2059delete_rule(struct ip_fw **head, struct ip_fw *prev, struct ip_fw *rule) 2060{ 2061 struct ip_fw *n; 2062 int l = RULESIZE(rule); 2063 2064 n = rule->next; 2065 remove_dyn_rule(rule, NULL /* force removal */); 2066 if (prev == NULL) 2067 *head = n; 2068 else 2069 prev->next = n; 2070 static_count--; 2071 static_len -= l; 2072 2073 if (DUMMYNET_LOADED) 2074 ip_dn_ruledel_ptr(rule); 2075 free(rule, M_IPFW); 2076 return n; 2077} 2078 2079/* 2080 * Deletes all rules from a chain (including the default rule 2081 * if the second argument is set). 2082 * Must be called at splimp(). 2083 */ 2084static void 2085free_chain(struct ip_fw **chain, int kill_default) 2086{ 2087 struct ip_fw *rule; 2088 2089 flush_rule_ptrs(); /* more efficient to do outside the loop */ 2090 2091 while ( (rule = *chain) != NULL && 2092 (kill_default || rule->rulenum != IPFW_DEFAULT_RULE) ) 2093 delete_rule(chain, NULL, rule); 2094} 2095 2096/** 2097 * Remove all rules with given number, and also do set manipulation. 2098 * 2099 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2100 * the next 8 bits are the new set, the top 8 bits are the command: 2101 * 2102 * 0 delete rules with given number 2103 * 1 delete rules with given set number 2104 * 2 move rules with given number to new set 2105 * 3 move rules with given set number to new set 2106 * 4 swap sets with given numbers 2107 */ 2108static int 2109del_entry(struct ip_fw **chain, u_int32_t arg) 2110{ 2111 struct ip_fw *prev, *rule; 2112 int s; 2113 u_int16_t rulenum; 2114 u_int8_t cmd, new_set; 2115 2116 rulenum = arg & 0xffff; 2117 cmd = (arg >> 24) & 0xff; 2118 new_set = (arg >> 16) & 0xff; 2119 2120 if (cmd > 4) 2121 return EINVAL; 2122 if (new_set > 30) 2123 return EINVAL; 2124 if (cmd == 0 || cmd == 2) { 2125 if (rulenum == IPFW_DEFAULT_RULE) 2126 return EINVAL; 2127 } else { 2128 if (rulenum > 30) 2129 return EINVAL; 2130 } 2131 2132 switch (cmd) { 2133 case 0: /* delete rules with given number */ 2134 /* 2135 * locate first rule to delete 2136 */ 2137 for (prev = NULL, rule = *chain; 2138 rule && rule->rulenum < rulenum; 2139 prev = rule, rule = rule->next) 2140 ; 2141 if (rule->rulenum != rulenum) 2142 return EINVAL; 2143 2144 s = splimp(); /* no access to rules while removing */ 2145 /* 2146 * flush pointers outside the loop, then delete all matching 2147 * rules. prev remains the same throughout the cycle. 2148 */ 2149 flush_rule_ptrs(); 2150 while (rule && rule->rulenum == rulenum) 2151 rule = delete_rule(chain, prev, rule); 2152 splx(s); 2153 break; 2154 2155 case 1: /* delete all rules with given set number */ 2156 s = splimp(); 2157 flush_rule_ptrs(); 2158 for (prev = NULL, rule = *chain; rule ; ) 2159 if (rule->set == rulenum) 2160 rule = delete_rule(chain, prev, rule); 2161 else { 2162 prev = rule; 2163 rule = rule->next; 2164 } 2165 splx(s); 2166 break; 2167 2168 case 2: /* move rules with given number to new set */ 2169 s = splimp(); 2170 for (rule = *chain; rule ; rule = rule->next) 2171 if (rule->rulenum == rulenum) 2172 rule->set = new_set; 2173 splx(s); 2174 break; 2175 2176 case 3: /* move rules with given set number to new set */ 2177 s = splimp(); 2178 for (rule = *chain; rule ; rule = rule->next) 2179 if (rule->set == rulenum) 2180 rule->set = new_set; 2181 splx(s); 2182 break; 2183 2184 case 4: /* swap two sets */ 2185 s = splimp(); 2186 for (rule = *chain; rule ; rule = rule->next) 2187 if (rule->set == rulenum) 2188 rule->set = new_set; 2189 else if (rule->set == new_set) 2190 rule->set = rulenum; 2191 splx(s); 2192 break; 2193 } 2194 return 0; 2195} 2196 2197/* 2198 * Clear counters for a specific rule. 2199 */ 2200static void 2201clear_counters(struct ip_fw *rule, int log_only) 2202{ 2203 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 2204 2205 if (log_only == 0) { 2206 rule->bcnt = rule->pcnt = 0; 2207 rule->timestamp = 0; 2208 } 2209 if (l->o.opcode == O_LOG) 2210 l->log_left = l->max_log; 2211} 2212 2213/** 2214 * Reset some or all counters on firewall rules. 2215 * @arg frwl is null to clear all entries, or contains a specific 2216 * rule number. 2217 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 2218 */ 2219static int 2220zero_entry(int rulenum, int log_only) 2221{ 2222 struct ip_fw *rule; 2223 int s; 2224 char *msg; 2225 2226 if (rulenum == 0) { 2227 s = splimp(); 2228 norule_counter = 0; 2229 for (rule = layer3_chain; rule; rule = rule->next) 2230 clear_counters(rule, log_only); 2231 splx(s); 2232 msg = log_only ? "ipfw: All logging counts reset.\n" : 2233 "ipfw: Accounting cleared.\n"; 2234 } else { 2235 int cleared = 0; 2236 /* 2237 * We can have multiple rules with the same number, so we 2238 * need to clear them all. 2239 */ 2240 for (rule = layer3_chain; rule; rule = rule->next) 2241 if (rule->rulenum == rulenum) { 2242 s = splimp(); 2243 while (rule && rule->rulenum == rulenum) { 2244 clear_counters(rule, log_only); 2245 rule = rule->next; 2246 } 2247 splx(s); 2248 cleared = 1; 2249 break; 2250 } 2251 if (!cleared) /* we did not find any matching rules */ 2252 return (EINVAL); 2253 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 2254 "ipfw: Entry %d cleared.\n"; 2255 } 2256 if (fw_verbose) 2257 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 2258 return (0); 2259} 2260 2261/* 2262 * Check validity of the structure before insert. 2263 * Fortunately rules are simple, so this mostly need to check rule sizes. 2264 */ 2265static int 2266check_ipfw_struct(struct ip_fw *rule, int size) 2267{ 2268 int l, cmdlen = 0; 2269 int have_action=0; 2270 ipfw_insn *cmd; 2271 2272 if (size < sizeof(*rule)) { 2273 printf("ipfw: rule too short\n"); 2274 return (EINVAL); 2275 } 2276 /* first, check for valid size */ 2277 l = RULESIZE(rule); 2278 if (l != size) { 2279 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 2280 return (EINVAL); 2281 } 2282 /* 2283 * Now go for the individual checks. Very simple ones, basically only 2284 * instruction sizes. 2285 */ 2286 for (l = rule->cmd_len, cmd = rule->cmd ; 2287 l > 0 ; l -= cmdlen, cmd += cmdlen) { 2288 cmdlen = F_LEN(cmd); 2289 if (cmdlen > l) { 2290 printf("ipfw: opcode %d size truncated\n", 2291 cmd->opcode); 2292 return EINVAL; 2293 } 2294 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 2295 switch (cmd->opcode) { 2296 case O_NOP: 2297 case O_PROBE_STATE: 2298 case O_KEEP_STATE: 2299 case O_PROTO: 2300 case O_IP_SRC_ME: 2301 case O_IP_DST_ME: 2302 case O_LAYER2: 2303 case O_IN: 2304 case O_FRAG: 2305 case O_IPOPT: 2306 case O_IPLEN: 2307 case O_IPID: 2308 case O_IPTOS: 2309 case O_IPPRECEDENCE: 2310 case O_IPTTL: 2311 case O_IPVER: 2312 case O_TCPWIN: 2313 case O_TCPFLAGS: 2314 case O_TCPOPTS: 2315 case O_ESTAB: 2316 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2317 goto bad_size; 2318 break; 2319 2320 case O_UID: 2321 case O_GID: 2322 case O_IP_SRC: 2323 case O_IP_DST: 2324 case O_TCPSEQ: 2325 case O_TCPACK: 2326 case O_PROB: 2327 case O_ICMPTYPE: 2328 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2329 goto bad_size; 2330 break; 2331 2332 case O_LIMIT: 2333 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 2334 goto bad_size; 2335 break; 2336 2337 case O_LOG: 2338 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 2339 goto bad_size; 2340 2341 ((ipfw_insn_log *)cmd)->log_left = 2342 ((ipfw_insn_log *)cmd)->max_log; 2343 2344 break; 2345 2346 case O_IP_SRC_MASK: 2347 case O_IP_DST_MASK: 2348 if (cmdlen != F_INSN_SIZE(ipfw_insn_ip)) 2349 goto bad_size; 2350 if (((ipfw_insn_ip *)cmd)->mask.s_addr == 0) { 2351 printf("ipfw: opcode %d, useless rule\n", 2352 cmd->opcode); 2353 return EINVAL; 2354 } 2355 break; 2356 2357 case O_IP_SRC_SET: 2358 case O_IP_DST_SET: 2359 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 2360 printf("ipfw: invalid set size %d\n", 2361 cmd->arg1); 2362 return EINVAL; 2363 } 2364 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 2365 (cmd->arg1+31)/32 ) 2366 goto bad_size; 2367 break; 2368 2369 case O_MACADDR2: 2370 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 2371 goto bad_size; 2372 break; 2373 2374 case O_MAC_TYPE: 2375 case O_IP_SRCPORT: 2376 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 2377 if (cmdlen < 2 || cmdlen > 31) 2378 goto bad_size; 2379 break; 2380 2381 case O_RECV: 2382 case O_XMIT: 2383 case O_VIA: 2384 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 2385 goto bad_size; 2386 break; 2387 2388 case O_PIPE: 2389 case O_QUEUE: 2390 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 2391 goto bad_size; 2392 goto check_action; 2393 2394 case O_FORWARD_IP: 2395 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 2396 goto bad_size; 2397 goto check_action; 2398 2399 case O_FORWARD_MAC: /* XXX not implemented yet */ 2400 case O_CHECK_STATE: 2401 case O_COUNT: 2402 case O_ACCEPT: 2403 case O_DENY: 2404 case O_REJECT: 2405 case O_SKIPTO: 2406 case O_DIVERT: 2407 case O_TEE: 2408 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2409 goto bad_size; 2410check_action: 2411 if (have_action) { 2412 printf("ipfw: opcode %d, multiple actions" 2413 " not allowed\n", 2414 cmd->opcode); 2415 return EINVAL; 2416 } 2417 have_action = 1; 2418 if (l != cmdlen) { 2419 printf("ipfw: opcode %d, action must be" 2420 " last opcode\n", 2421 cmd->opcode); 2422 return EINVAL; 2423 } 2424 break; 2425 default: 2426 printf("ipfw: opcode %d, unknown opcode\n", 2427 cmd->opcode); 2428 return EINVAL; 2429 } 2430 } 2431 if (have_action == 0) { 2432 printf("ipfw: missing action\n"); 2433 return EINVAL; 2434 } 2435 return 0; 2436 2437bad_size: 2438 printf("ipfw: opcode %d size %d wrong\n", 2439 cmd->opcode, cmdlen); 2440 return EINVAL; 2441} 2442 2443 2444/** 2445 * {set|get}sockopt parser. 2446 */ 2447static int 2448ipfw_ctl(struct sockopt *sopt) 2449{ 2450 int error, s, rulenum; 2451 size_t size; 2452 struct ip_fw *bp , *buf, *rule; 2453 2454 static u_int32_t rule_buf[255]; /* we copy the data here */ 2455 2456 /* 2457 * Disallow modifications in really-really secure mode, but still allow 2458 * the logging counters to be reset. 2459 */ 2460 if (sopt->sopt_name == IP_FW_ADD || 2461 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 2462#if __FreeBSD_version >= 500034 2463 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 2464 if (error) 2465 return (error); 2466#else /* FreeBSD 4.x */ 2467 if (securelevel >= 3) 2468 return (EPERM); 2469#endif 2470 } 2471 2472 error = 0; 2473 2474 switch (sopt->sopt_name) { 2475 case IP_FW_GET: 2476 /* 2477 * pass up a copy of the current rules. Static rules 2478 * come first (the last of which has number IPFW_DEFAULT_RULE), 2479 * followed by a possibly empty list of dynamic rule. 2480 * The last dynamic rule has NULL in the "next" field. 2481 */ 2482 s = splimp(); 2483 size = static_len; /* size of static rules */ 2484 if (ipfw_dyn_v) /* add size of dyn.rules */ 2485 size += (dyn_count * sizeof(ipfw_dyn_rule)); 2486 2487 /* 2488 * XXX todo: if the user passes a short length just to know 2489 * how much room is needed, do not bother filling up the 2490 * buffer, just jump to the sooptcopyout. 2491 */ 2492 buf = malloc(size, M_TEMP, M_WAITOK); 2493 if (buf == 0) { 2494 splx(s); 2495 error = ENOBUFS; 2496 break; 2497 } 2498 2499 bp = buf; 2500 for (rule = layer3_chain; rule ; rule = rule->next) { 2501 int i = RULESIZE(rule); 2502 bcopy(rule, bp, i); 2503 ((struct ip_fw *)bp)->set_disable = set_disable; 2504 bp = (struct ip_fw *)((char *)bp + i); 2505 } 2506 if (ipfw_dyn_v) { 2507 int i; 2508 ipfw_dyn_rule *p, *dst, *last = NULL; 2509 2510 dst = (ipfw_dyn_rule *)bp; 2511 for (i = 0 ; i < curr_dyn_buckets ; i++ ) 2512 for ( p = ipfw_dyn_v[i] ; p != NULL ; 2513 p = p->next, dst++ ) { 2514 bcopy(p, dst, sizeof *p); 2515 dst->rulenum = p->rule->rulenum; 2516 /* 2517 * store a non-null value in "next". 2518 * The userland code will interpret a 2519 * NULL here as a marker 2520 * for the last dynamic rule. 2521 */ 2522 dst->next = dst ; 2523 last = dst ; 2524 dst->expire = 2525 TIME_LEQ(dst->expire, time_second) ? 2526 0 : dst->expire - time_second ; 2527 } 2528 if (last != NULL) /* mark last dynamic rule */ 2529 last->next = NULL; 2530 } 2531 splx(s); 2532 2533 error = sooptcopyout(sopt, buf, size); 2534 free(buf, M_TEMP); 2535 break; 2536 2537 case IP_FW_FLUSH: 2538 /* 2539 * Normally we cannot release the lock on each iteration. 2540 * We could do it here only because we start from the head all 2541 * the times so there is no risk of missing some entries. 2542 * On the other hand, the risk is that we end up with 2543 * a very inconsistent ruleset, so better keep the lock 2544 * around the whole cycle. 2545 * 2546 * XXX this code can be improved by resetting the head of 2547 * the list to point to the default rule, and then freeing 2548 * the old list without the need for a lock. 2549 */ 2550 2551 s = splimp(); 2552 free_chain(&layer3_chain, 0 /* keep default rule */); 2553 splx(s); 2554 break; 2555 2556 case IP_FW_ADD: 2557 rule = (struct ip_fw *)rule_buf; /* XXX do a malloc */ 2558 error = sooptcopyin(sopt, rule, sizeof(rule_buf), 2559 sizeof(struct ip_fw) ); 2560 size = sopt->sopt_valsize; 2561 if (error || (error = check_ipfw_struct(rule, size))) 2562 break; 2563 2564 error = add_rule(&layer3_chain, rule); 2565 size = RULESIZE(rule); 2566 if (!error && sopt->sopt_dir == SOPT_GET) 2567 error = sooptcopyout(sopt, rule, size); 2568 break; 2569 2570 case IP_FW_DEL: 2571 /* 2572 * IP_FW_DEL is used for deleting single rules or sets, 2573 * and (ab)used to atomically manipulate sets. Argument size 2574 * is used to distinguish between the two: 2575 * sizeof(u_int32_t) 2576 * delete single rule or set of rules, 2577 * or reassign rules (or sets) to a different set. 2578 * 2*sizeof(u_int32_t) 2579 * atomic disable/enable sets. 2580 * first u_int32_t contains sets to be disabled, 2581 * second u_int32_t contains sets to be enabled. 2582 */ 2583 error = sooptcopyin(sopt, rule_buf, 2584 2*sizeof(u_int32_t), sizeof(u_int32_t)); 2585 if (error) 2586 break; 2587 size = sopt->sopt_valsize; 2588 if (size == sizeof(u_int32_t)) /* delete or reassign */ 2589 error = del_entry(&layer3_chain, rule_buf[0]); 2590 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 2591 set_disable = 2592 (set_disable | rule_buf[0]) & ~rule_buf[1] & 2593 ~(1<<31); /* set 31 always enabled */ 2594 else 2595 error = EINVAL; 2596 break; 2597 2598 case IP_FW_ZERO: 2599 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 2600 rulenum=0; 2601 2602 if (sopt->sopt_val != 0) { 2603 error = sooptcopyin(sopt, &rulenum, 2604 sizeof(int), sizeof(int)); 2605 if (error) 2606 break; 2607 } 2608 error = zero_entry(rulenum, sopt->sopt_name == IP_FW_RESETLOG); 2609 break; 2610 2611 default: 2612 printf("ipfw_ctl invalid option %d\n", sopt->sopt_name); 2613 error = EINVAL; 2614 } 2615 2616 return (error); 2617} 2618 2619/** 2620 * dummynet needs a reference to the default rule, because rules can be 2621 * deleted while packets hold a reference to them. When this happens, 2622 * dummynet changes the reference to the default rule (it could well be a 2623 * NULL pointer, but this way we do not need to check for the special 2624 * case, plus here he have info on the default behaviour). 2625 */ 2626struct ip_fw *ip_fw_default_rule; 2627 2628/* 2629 * This procedure is only used to handle keepalives. It is invoked 2630 * every dyn_keepalive_period 2631 */ 2632static void 2633ipfw_tick(void * __unused unused) 2634{ 2635 int i; 2636 int s; 2637 ipfw_dyn_rule *q; 2638 2639 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 2640 goto done; 2641 2642 s = splimp(); 2643 for (i = 0 ; i < curr_dyn_buckets ; i++) { 2644 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 2645 if (q->dyn_type == O_LIMIT_PARENT) 2646 continue; 2647 if (q->id.proto != IPPROTO_TCP) 2648 continue; 2649 if ( (q->state & BOTH_SYN) != BOTH_SYN) 2650 continue; 2651 if (TIME_LEQ( time_second+dyn_keepalive_interval, 2652 q->expire)) 2653 continue; /* too early */ 2654 if (TIME_LEQ(q->expire, time_second)) 2655 continue; /* too late, rule expired */ 2656 2657 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 2658 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 2659 } 2660 } 2661 splx(s); 2662done: 2663 ipfw_timeout_h = timeout(ipfw_tick, NULL, dyn_keepalive_period*hz); 2664} 2665 2666static void 2667ipfw_init(void) 2668{ 2669 struct ip_fw default_rule; 2670 2671 ip_fw_chk_ptr = ipfw_chk; 2672 ip_fw_ctl_ptr = ipfw_ctl; 2673 layer3_chain = NULL; 2674 2675 bzero(&default_rule, sizeof default_rule); 2676 2677 default_rule.act_ofs = 0; 2678 default_rule.rulenum = IPFW_DEFAULT_RULE; 2679 default_rule.cmd_len = 1; 2680 default_rule.set = 31; 2681 2682 default_rule.cmd[0].len = 1; 2683 default_rule.cmd[0].opcode = 2684#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 2685 1 ? O_ACCEPT : 2686#endif 2687 O_DENY; 2688 2689 add_rule(&layer3_chain, &default_rule); 2690 2691 ip_fw_default_rule = layer3_chain; 2692 printf("ipfw2 initialized, divert %s, " 2693 "rule-based forwarding enabled, default to %s, logging ", 2694#ifdef IPDIVERT 2695 "enabled", 2696#else 2697 "disabled", 2698#endif 2699 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 2700 2701#ifdef IPFIREWALL_VERBOSE 2702 fw_verbose = 1; 2703#endif 2704#ifdef IPFIREWALL_VERBOSE_LIMIT 2705 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 2706#endif 2707 if (fw_verbose == 0) 2708 printf("disabled\n"); 2709 else if (verbose_limit == 0) 2710 printf("unlimited\n"); 2711 else 2712 printf("limited to %d packets/entry by default\n", 2713 verbose_limit); 2714 bzero(&ipfw_timeout_h, sizeof(struct callout_handle)); 2715 ipfw_timeout_h = timeout(ipfw_tick, NULL, hz); 2716} 2717 2718static int 2719ipfw_modevent(module_t mod, int type, void *unused) 2720{ 2721 int s; 2722 int err = 0; 2723 2724 switch (type) { 2725 case MOD_LOAD: 2726 s = splimp(); 2727 if (IPFW_LOADED) { 2728 splx(s); 2729 printf("IP firewall already loaded\n"); 2730 err = EEXIST; 2731 } else { 2732 ipfw_init(); 2733 splx(s); 2734 } 2735 break; 2736 2737 case MOD_UNLOAD: 2738#if !defined(KLD_MODULE) 2739 printf("ipfw statically compiled, cannot unload\n"); 2740 err = EBUSY; 2741#else 2742 s = splimp(); 2743 untimeout(ipfw_tick, NULL, ipfw_timeout_h); 2744 ip_fw_chk_ptr = NULL; 2745 ip_fw_ctl_ptr = NULL; 2746 free_chain(&layer3_chain, 1 /* kill default rule */); 2747 splx(s); 2748 printf("IP firewall unloaded\n"); 2749#endif 2750 break; 2751 default: 2752 break; 2753 } 2754 return err; 2755} 2756 2757static moduledata_t ipfwmod = { 2758 "ipfw", 2759 ipfw_modevent, 2760 0 2761}; 2762DECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PSEUDO, SI_ORDER_ANY); 2763MODULE_VERSION(ipfw, 1); 2764#endif /* IPFW2 */ 2765