ip_fw2.c revision 105886
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 105886 2002-10-24 18:01:53Z 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; 110int fw_one_pass = 1; 111static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 112 113#ifdef SYSCTL_NODE 114SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 115SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 116 CTLFLAG_RW | CTLFLAG_SECURE, 117 &fw_enable, 0, "Enable ipfw"); 118SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 119 &autoinc_step, 0, "Rule number autincrement step"); 120SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 121 CTLFLAG_RW | CTLFLAG_SECURE, 122 &fw_one_pass, 0, 123 "Only do a single pass through ipfw when using dummynet(4)"); 124SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 125 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 126SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 127 CTLFLAG_RW | CTLFLAG_SECURE, 128 &fw_verbose, 0, "Log matches to ipfw rules"); 129SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 130 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 131 132/* 133 * Description of dynamic rules. 134 * 135 * Dynamic rules are stored in lists accessed through a hash table 136 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 137 * be modified through the sysctl variable dyn_buckets which is 138 * updated when the table becomes empty. 139 * 140 * XXX currently there is only one list, ipfw_dyn. 141 * 142 * When a packet is received, its address fields are first masked 143 * with the mask defined for the rule, then hashed, then matched 144 * against the entries in the corresponding list. 145 * Dynamic rules can be used for different purposes: 146 * + stateful rules; 147 * + enforcing limits on the number of sessions; 148 * + in-kernel NAT (not implemented yet) 149 * 150 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 151 * measured in seconds and depending on the flags. 152 * 153 * The total number of dynamic rules is stored in dyn_count. 154 * The max number of dynamic rules is dyn_max. When we reach 155 * the maximum number of rules we do not create anymore. This is 156 * done to avoid consuming too much memory, but also too much 157 * time when searching on each packet (ideally, we should try instead 158 * to put a limit on the length of the list on each bucket...). 159 * 160 * Each dynamic rule holds a pointer to the parent ipfw rule so 161 * we know what action to perform. Dynamic rules are removed when 162 * the parent rule is deleted. XXX we should make them survive. 163 * 164 * There are some limitations with dynamic rules -- we do not 165 * obey the 'randomized match', and we do not do multiple 166 * passes through the firewall. XXX check the latter!!! 167 */ 168static ipfw_dyn_rule **ipfw_dyn_v = NULL; 169static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 170static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 171 172/* 173 * Timeouts for various events in handing dynamic rules. 174 */ 175static u_int32_t dyn_ack_lifetime = 300; 176static u_int32_t dyn_syn_lifetime = 20; 177static u_int32_t dyn_fin_lifetime = 1; 178static u_int32_t dyn_rst_lifetime = 1; 179static u_int32_t dyn_udp_lifetime = 10; 180static u_int32_t dyn_short_lifetime = 5; 181 182/* 183 * Keepalives are sent if dyn_keepalive is set. They are sent every 184 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 185 * seconds of lifetime of a rule. 186 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 187 * than dyn_keepalive_period. 188 */ 189 190static u_int32_t dyn_keepalive_interval = 20; 191static u_int32_t dyn_keepalive_period = 5; 192static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 193 194static u_int32_t static_count; /* # of static rules */ 195static u_int32_t static_len; /* size in bytes of static rules */ 196static u_int32_t dyn_count; /* # of dynamic rules */ 197static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 198 199SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 200 &dyn_buckets, 0, "Number of dyn. buckets"); 201SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 202 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 203SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 204 &dyn_count, 0, "Number of dyn. rules"); 205SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 206 &dyn_max, 0, "Max number of dyn. rules"); 207SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 208 &static_count, 0, "Number of static rules"); 209SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 210 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 211SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 212 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 213SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 214 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 215SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 216 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 217SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 218 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 219SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 220 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 221SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 222 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 223 224#endif /* SYSCTL_NODE */ 225 226 227static ip_fw_chk_t ipfw_chk; 228 229ip_dn_ruledel_t *ip_dn_ruledel_ptr = NULL; /* hook into dummynet */ 230 231/* 232 * This macro maps an ip pointer into a layer3 header pointer of type T 233 */ 234#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 235 236static __inline int 237icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 238{ 239 int type = L3HDR(struct icmp,ip)->icmp_type; 240 241 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 242} 243 244#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 245 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 246 247static int 248is_icmp_query(struct ip *ip) 249{ 250 int type = L3HDR(struct icmp, ip)->icmp_type; 251 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 252} 253#undef TT 254 255/* 256 * The following checks use two arrays of 8 or 16 bits to store the 257 * bits that we want set or clear, respectively. They are in the 258 * low and high half of cmd->arg1 or cmd->d[0]. 259 * 260 * We scan options and store the bits we find set. We succeed if 261 * 262 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 263 * 264 * The code is sometimes optimized not to store additional variables. 265 */ 266 267static int 268flags_match(ipfw_insn *cmd, u_int8_t bits) 269{ 270 u_char want_clear; 271 bits = ~bits; 272 273 if ( ((cmd->arg1 & 0xff) & bits) != 0) 274 return 0; /* some bits we want set were clear */ 275 want_clear = (cmd->arg1 >> 8) & 0xff; 276 if ( (want_clear & bits) != want_clear) 277 return 0; /* some bits we want clear were set */ 278 return 1; 279} 280 281static int 282ipopts_match(struct ip *ip, ipfw_insn *cmd) 283{ 284 int optlen, bits = 0; 285 u_char *cp = (u_char *)(ip + 1); 286 int x = (ip->ip_hl << 2) - sizeof (struct ip); 287 288 for (; x > 0; x -= optlen, cp += optlen) { 289 int opt = cp[IPOPT_OPTVAL]; 290 291 if (opt == IPOPT_EOL) 292 break; 293 if (opt == IPOPT_NOP) 294 optlen = 1; 295 else { 296 optlen = cp[IPOPT_OLEN]; 297 if (optlen <= 0 || optlen > x) 298 return 0; /* invalid or truncated */ 299 } 300 switch (opt) { 301 302 default: 303 break; 304 305 case IPOPT_LSRR: 306 bits |= IP_FW_IPOPT_LSRR; 307 break; 308 309 case IPOPT_SSRR: 310 bits |= IP_FW_IPOPT_SSRR; 311 break; 312 313 case IPOPT_RR: 314 bits |= IP_FW_IPOPT_RR; 315 break; 316 317 case IPOPT_TS: 318 bits |= IP_FW_IPOPT_TS; 319 break; 320 } 321 } 322 return (flags_match(cmd, bits)); 323} 324 325static int 326tcpopts_match(struct ip *ip, ipfw_insn *cmd) 327{ 328 int optlen, bits = 0; 329 struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 330 u_char *cp = (u_char *)(tcp + 1); 331 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 332 333 for (; x > 0; x -= optlen, cp += optlen) { 334 int opt = cp[0]; 335 if (opt == TCPOPT_EOL) 336 break; 337 if (opt == TCPOPT_NOP) 338 optlen = 1; 339 else { 340 optlen = cp[1]; 341 if (optlen <= 0) 342 break; 343 } 344 345 switch (opt) { 346 347 default: 348 break; 349 350 case TCPOPT_MAXSEG: 351 bits |= IP_FW_TCPOPT_MSS; 352 break; 353 354 case TCPOPT_WINDOW: 355 bits |= IP_FW_TCPOPT_WINDOW; 356 break; 357 358 case TCPOPT_SACK_PERMITTED: 359 case TCPOPT_SACK: 360 bits |= IP_FW_TCPOPT_SACK; 361 break; 362 363 case TCPOPT_TIMESTAMP: 364 bits |= IP_FW_TCPOPT_TS; 365 break; 366 367 case TCPOPT_CC: 368 case TCPOPT_CCNEW: 369 case TCPOPT_CCECHO: 370 bits |= IP_FW_TCPOPT_CC; 371 break; 372 } 373 } 374 return (flags_match(cmd, bits)); 375} 376 377static int 378iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 379{ 380 if (ifp == NULL) /* no iface with this packet, match fails */ 381 return 0; 382 /* Check by name or by IP address */ 383 if (cmd->name[0] != '\0') { /* match by name */ 384 /* Check unit number (-1 is wildcard) */ 385 if (cmd->p.unit != -1 && cmd->p.unit != ifp->if_unit) 386 return(0); 387 /* Check name */ 388 if (!strncmp(ifp->if_name, cmd->name, IFNAMSIZ)) 389 return(1); 390 } else { 391 struct ifaddr *ia; 392 393 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 394 if (ia->ifa_addr == NULL) 395 continue; 396 if (ia->ifa_addr->sa_family != AF_INET) 397 continue; 398 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 399 (ia->ifa_addr))->sin_addr.s_addr) 400 return(1); /* match */ 401 } 402 } 403 return(0); /* no match, fail ... */ 404} 405 406static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 407 408#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 409#define SNP(buf) buf, sizeof(buf) 410 411/* 412 * We enter here when we have a rule with O_LOG. 413 * XXX this function alone takes about 2Kbytes of code! 414 */ 415static void 416ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 417 struct mbuf *m, struct ifnet *oif) 418{ 419 char *action; 420 int limit_reached = 0; 421 char action2[40], proto[48], fragment[28]; 422 423 fragment[0] = '\0'; 424 proto[0] = '\0'; 425 426 if (f == NULL) { /* bogus pkt */ 427 if (verbose_limit != 0 && norule_counter >= verbose_limit) 428 return; 429 norule_counter++; 430 if (norule_counter == verbose_limit) 431 limit_reached = verbose_limit; 432 action = "Refuse"; 433 } else { /* O_LOG is the first action, find the real one */ 434 ipfw_insn *cmd = ACTION_PTR(f); 435 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 436 437 if (l->max_log != 0 && l->log_left == 0) 438 return; 439 l->log_left--; 440 if (l->log_left == 0) 441 limit_reached = l->max_log; 442 cmd += F_LEN(cmd); /* point to first action */ 443 if (cmd->opcode == O_PROB) 444 cmd += F_LEN(cmd); 445 446 action = action2; 447 switch (cmd->opcode) { 448 case O_DENY: 449 action = "Deny"; 450 break; 451 452 case O_REJECT: 453 if (cmd->arg1==ICMP_REJECT_RST) 454 action = "Reset"; 455 else if (cmd->arg1==ICMP_UNREACH_HOST) 456 action = "Reject"; 457 else 458 snprintf(SNPARGS(action2, 0), "Unreach %d", 459 cmd->arg1); 460 break; 461 462 case O_ACCEPT: 463 action = "Accept"; 464 break; 465 case O_COUNT: 466 action = "Count"; 467 break; 468 case O_DIVERT: 469 snprintf(SNPARGS(action2, 0), "Divert %d", 470 cmd->arg1); 471 break; 472 case O_TEE: 473 snprintf(SNPARGS(action2, 0), "Tee %d", 474 cmd->arg1); 475 break; 476 case O_SKIPTO: 477 snprintf(SNPARGS(action2, 0), "SkipTo %d", 478 cmd->arg1); 479 break; 480 case O_PIPE: 481 snprintf(SNPARGS(action2, 0), "Pipe %d", 482 cmd->arg1); 483 break; 484 case O_QUEUE: 485 snprintf(SNPARGS(action2, 0), "Queue %d", 486 cmd->arg1); 487 break; 488 case O_FORWARD_IP: { 489 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 490 int len; 491 492 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 493 inet_ntoa(sa->sa.sin_addr)); 494 if (sa->sa.sin_port) 495 snprintf(SNPARGS(action2, len), ":%d", 496 ntohs(sa->sa.sin_port)); 497 } 498 break; 499 default: 500 action = "UNKNOWN"; 501 break; 502 } 503 } 504 505 if (hlen == 0) { /* non-ip */ 506 snprintf(SNPARGS(proto, 0), "MAC"); 507 } else { 508 struct ip *ip = mtod(m, struct ip *); 509 /* these three are all aliases to the same thing */ 510 struct icmp *const icmp = L3HDR(struct icmp, ip); 511 struct tcphdr *const tcp = (struct tcphdr *)icmp; 512 struct udphdr *const udp = (struct udphdr *)icmp; 513 514 int ip_off, offset, ip_len; 515 516 int len; 517 518 if (eh != NULL) { /* layer 2 packets are as on the wire */ 519 ip_off = ntohs(ip->ip_off); 520 ip_len = ntohs(ip->ip_len); 521 } else { 522 ip_off = ip->ip_off; 523 ip_len = ip->ip_len; 524 } 525 offset = ip_off & IP_OFFMASK; 526 switch (ip->ip_p) { 527 case IPPROTO_TCP: 528 len = snprintf(SNPARGS(proto, 0), "TCP %s", 529 inet_ntoa(ip->ip_src)); 530 if (offset == 0) 531 snprintf(SNPARGS(proto, len), ":%d %s:%d", 532 ntohs(tcp->th_sport), 533 inet_ntoa(ip->ip_dst), 534 ntohs(tcp->th_dport)); 535 else 536 snprintf(SNPARGS(proto, len), " %s", 537 inet_ntoa(ip->ip_dst)); 538 break; 539 540 case IPPROTO_UDP: 541 len = snprintf(SNPARGS(proto, 0), "UDP %s", 542 inet_ntoa(ip->ip_src)); 543 if (offset == 0) 544 snprintf(SNPARGS(proto, len), ":%d %s:%d", 545 ntohs(udp->uh_sport), 546 inet_ntoa(ip->ip_dst), 547 ntohs(udp->uh_dport)); 548 else 549 snprintf(SNPARGS(proto, len), " %s", 550 inet_ntoa(ip->ip_dst)); 551 break; 552 553 case IPPROTO_ICMP: 554 if (offset == 0) 555 len = snprintf(SNPARGS(proto, 0), 556 "ICMP:%u.%u ", 557 icmp->icmp_type, icmp->icmp_code); 558 else 559 len = snprintf(SNPARGS(proto, 0), "ICMP "); 560 len += snprintf(SNPARGS(proto, len), "%s", 561 inet_ntoa(ip->ip_src)); 562 snprintf(SNPARGS(proto, len), " %s", 563 inet_ntoa(ip->ip_dst)); 564 break; 565 566 default: 567 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 568 inet_ntoa(ip->ip_src)); 569 snprintf(SNPARGS(proto, len), " %s", 570 inet_ntoa(ip->ip_dst)); 571 break; 572 } 573 574 if (ip_off & (IP_MF | IP_OFFMASK)) 575 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 576 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 577 offset << 3, 578 (ip_off & IP_MF) ? "+" : ""); 579 } 580 if (oif || m->m_pkthdr.rcvif) 581 log(LOG_SECURITY | LOG_INFO, 582 "ipfw: %d %s %s %s via %s%d%s\n", 583 f ? f->rulenum : -1, 584 action, proto, oif ? "out" : "in", 585 oif ? oif->if_name : m->m_pkthdr.rcvif->if_name, 586 oif ? oif->if_unit : m->m_pkthdr.rcvif->if_unit, 587 fragment); 588 else 589 log(LOG_SECURITY | LOG_INFO, 590 "ipfw: %d %s %s [no if info]%s\n", 591 f ? f->rulenum : -1, 592 action, proto, fragment); 593 if (limit_reached) 594 log(LOG_SECURITY | LOG_NOTICE, 595 "ipfw: limit %d reached on entry %d\n", 596 limit_reached, f ? f->rulenum : -1); 597} 598 599/* 600 * IMPORTANT: the hash function for dynamic rules must be commutative 601 * in source and destination (ip,port), because rules are bidirectional 602 * and we want to find both in the same bucket. 603 */ 604static __inline int 605hash_packet(struct ipfw_flow_id *id) 606{ 607 u_int32_t i; 608 609 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 610 i &= (curr_dyn_buckets - 1); 611 return i; 612} 613 614/** 615 * unlink a dynamic rule from a chain. prev is a pointer to 616 * the previous one, q is a pointer to the rule to delete, 617 * head is a pointer to the head of the queue. 618 * Modifies q and potentially also head. 619 */ 620#define UNLINK_DYN_RULE(prev, head, q) { \ 621 ipfw_dyn_rule *old_q = q; \ 622 \ 623 /* remove a refcount to the parent */ \ 624 if (q->dyn_type == O_LIMIT) \ 625 q->parent->count--; \ 626 DEB(printf("-- unlink entry 0x%08x %d -> 0x%08x %d, %d left\n", \ 627 (q->id.src_ip), (q->id.src_port), \ 628 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 629 if (prev != NULL) \ 630 prev->next = q = q->next; \ 631 else \ 632 head = q = q->next; \ 633 dyn_count--; \ 634 free(old_q, M_IPFW); } 635 636#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 637 638/** 639 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 640 * 641 * If keep_me == NULL, rules are deleted even if not expired, 642 * otherwise only expired rules are removed. 643 * 644 * The value of the second parameter is also used to point to identify 645 * a rule we absolutely do not want to remove (e.g. because we are 646 * holding a reference to it -- this is the case with O_LIMIT_PARENT 647 * rules). The pointer is only used for comparison, so any non-null 648 * value will do. 649 */ 650static void 651remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 652{ 653 static u_int32_t last_remove = 0; 654 655#define FORCE (keep_me == NULL) 656 657 ipfw_dyn_rule *prev, *q; 658 int i, pass = 0, max_pass = 0; 659 660 if (ipfw_dyn_v == NULL || dyn_count == 0) 661 return; 662 /* do not expire more than once per second, it is useless */ 663 if (!FORCE && last_remove == time_second) 664 return; 665 last_remove = time_second; 666 667 /* 668 * because O_LIMIT refer to parent rules, during the first pass only 669 * remove child and mark any pending LIMIT_PARENT, and remove 670 * them in a second pass. 671 */ 672next_pass: 673 for (i = 0 ; i < curr_dyn_buckets ; i++) { 674 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 675 /* 676 * Logic can become complex here, so we split tests. 677 */ 678 if (q == keep_me) 679 goto next; 680 if (rule != NULL && rule != q->rule) 681 goto next; /* not the one we are looking for */ 682 if (q->dyn_type == O_LIMIT_PARENT) { 683 /* 684 * handle parent in the second pass, 685 * record we need one. 686 */ 687 max_pass = 1; 688 if (pass == 0) 689 goto next; 690 if (FORCE && q->count != 0 ) { 691 /* XXX should not happen! */ 692 printf( "OUCH! cannot remove rule," 693 " count %d\n", q->count); 694 } 695 } else { 696 if (!FORCE && 697 !TIME_LEQ( q->expire, time_second )) 698 goto next; 699 } 700 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 701 continue; 702next: 703 prev=q; 704 q=q->next; 705 } 706 } 707 if (pass++ < max_pass) 708 goto next_pass; 709} 710 711 712/** 713 * lookup a dynamic rule. 714 */ 715static ipfw_dyn_rule * 716lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 717 struct tcphdr *tcp) 718{ 719 /* 720 * stateful ipfw extensions. 721 * Lookup into dynamic session queue 722 */ 723#define MATCH_REVERSE 0 724#define MATCH_FORWARD 1 725#define MATCH_NONE 2 726#define MATCH_UNKNOWN 3 727 int i, dir = MATCH_NONE; 728 ipfw_dyn_rule *prev, *q=NULL; 729 730 if (ipfw_dyn_v == NULL) 731 goto done; /* not found */ 732 i = hash_packet( pkt ); 733 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 734 if (q->dyn_type == O_LIMIT_PARENT) 735 goto next; 736 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 737 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 738 continue; 739 } 740 if ( pkt->proto == q->id.proto) { 741 if (pkt->src_ip == q->id.src_ip && 742 pkt->dst_ip == q->id.dst_ip && 743 pkt->src_port == q->id.src_port && 744 pkt->dst_port == q->id.dst_port ) { 745 dir = MATCH_FORWARD; 746 break; 747 } 748 if (pkt->src_ip == q->id.dst_ip && 749 pkt->dst_ip == q->id.src_ip && 750 pkt->src_port == q->id.dst_port && 751 pkt->dst_port == q->id.src_port ) { 752 dir = MATCH_REVERSE; 753 break; 754 } 755 } 756next: 757 prev = q; 758 q = q->next; 759 } 760 if (q == NULL) 761 goto done; /* q = NULL, not found */ 762 763 if ( prev != NULL) { /* found and not in front */ 764 prev->next = q->next; 765 q->next = ipfw_dyn_v[i]; 766 ipfw_dyn_v[i] = q; 767 } 768 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 769 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 770 771#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 772#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 773 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 774 switch (q->state) { 775 case TH_SYN: /* opening */ 776 q->expire = time_second + dyn_syn_lifetime; 777 break; 778 779 case BOTH_SYN: /* move to established */ 780 case BOTH_SYN | TH_FIN : /* one side tries to close */ 781 case BOTH_SYN | (TH_FIN << 8) : 782 if (tcp) { 783#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 784 u_int32_t ack = ntohl(tcp->th_ack); 785 if (dir == MATCH_FORWARD) { 786 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 787 q->ack_fwd = ack; 788 else { /* ignore out-of-sequence */ 789 break; 790 } 791 } else { 792 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 793 q->ack_rev = ack; 794 else { /* ignore out-of-sequence */ 795 break; 796 } 797 } 798 } 799 q->expire = time_second + dyn_ack_lifetime; 800 break; 801 802 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 803 if (dyn_fin_lifetime >= dyn_keepalive_period) 804 dyn_fin_lifetime = dyn_keepalive_period - 1; 805 q->expire = time_second + dyn_fin_lifetime; 806 break; 807 808 default: 809#if 0 810 /* 811 * reset or some invalid combination, but can also 812 * occur if we use keep-state the wrong way. 813 */ 814 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 815 printf("invalid state: 0x%x\n", q->state); 816#endif 817 if (dyn_rst_lifetime >= dyn_keepalive_period) 818 dyn_rst_lifetime = dyn_keepalive_period - 1; 819 q->expire = time_second + dyn_rst_lifetime; 820 break; 821 } 822 } else if (pkt->proto == IPPROTO_UDP) { 823 q->expire = time_second + dyn_udp_lifetime; 824 } else { 825 /* other protocols */ 826 q->expire = time_second + dyn_short_lifetime; 827 } 828done: 829 if (match_direction) 830 *match_direction = dir; 831 return q; 832} 833 834static void 835realloc_dynamic_table(void) 836{ 837 /* 838 * Try reallocation, make sure we have a power of 2 and do 839 * not allow more than 64k entries. In case of overflow, 840 * default to 1024. 841 */ 842 843 if (dyn_buckets > 65536) 844 dyn_buckets = 1024; 845 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 846 dyn_buckets = curr_dyn_buckets; /* reset */ 847 return; 848 } 849 curr_dyn_buckets = dyn_buckets; 850 if (ipfw_dyn_v != NULL) 851 free(ipfw_dyn_v, M_IPFW); 852 for (;;) { 853 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 854 M_IPFW, M_NOWAIT | M_ZERO); 855 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 856 break; 857 curr_dyn_buckets /= 2; 858 } 859} 860 861/** 862 * Install state of type 'type' for a dynamic session. 863 * The hash table contains two type of rules: 864 * - regular rules (O_KEEP_STATE) 865 * - rules for sessions with limited number of sess per user 866 * (O_LIMIT). When they are created, the parent is 867 * increased by 1, and decreased on delete. In this case, 868 * the third parameter is the parent rule and not the chain. 869 * - "parent" rules for the above (O_LIMIT_PARENT). 870 */ 871static ipfw_dyn_rule * 872add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 873{ 874 ipfw_dyn_rule *r; 875 int i; 876 877 if (ipfw_dyn_v == NULL || 878 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 879 realloc_dynamic_table(); 880 if (ipfw_dyn_v == NULL) 881 return NULL; /* failed ! */ 882 } 883 i = hash_packet(id); 884 885 r = malloc(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO); 886 if (r == NULL) { 887 printf ("sorry cannot allocate state\n"); 888 return NULL; 889 } 890 891 /* increase refcount on parent, and set pointer */ 892 if (dyn_type == O_LIMIT) { 893 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 894 if ( parent->dyn_type != O_LIMIT_PARENT) 895 panic("invalid parent"); 896 parent->count++; 897 r->parent = parent; 898 rule = parent->rule; 899 } 900 901 r->id = *id; 902 r->expire = time_second + dyn_syn_lifetime; 903 r->rule = rule; 904 r->dyn_type = dyn_type; 905 r->pcnt = r->bcnt = 0; 906 r->count = 0; 907 908 r->bucket = i; 909 r->next = ipfw_dyn_v[i]; 910 ipfw_dyn_v[i] = r; 911 dyn_count++; 912 DEB(printf("-- add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 913 dyn_type, 914 (r->id.src_ip), (r->id.src_port), 915 (r->id.dst_ip), (r->id.dst_port), 916 dyn_count ); ) 917 return r; 918} 919 920/** 921 * lookup dynamic parent rule using pkt and rule as search keys. 922 * If the lookup fails, then install one. 923 */ 924static ipfw_dyn_rule * 925lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 926{ 927 ipfw_dyn_rule *q; 928 int i; 929 930 if (ipfw_dyn_v) { 931 i = hash_packet( pkt ); 932 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 933 if (q->dyn_type == O_LIMIT_PARENT && 934 rule== q->rule && 935 pkt->proto == q->id.proto && 936 pkt->src_ip == q->id.src_ip && 937 pkt->dst_ip == q->id.dst_ip && 938 pkt->src_port == q->id.src_port && 939 pkt->dst_port == q->id.dst_port) { 940 q->expire = time_second + dyn_short_lifetime; 941 DEB(printf("lookup_dyn_parent found 0x%p\n",q);) 942 return q; 943 } 944 } 945 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 946} 947 948/** 949 * Install dynamic state for rule type cmd->o.opcode 950 * 951 * Returns 1 (failure) if state is not installed because of errors or because 952 * session limitations are enforced. 953 */ 954static int 955install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 956 struct ip_fw_args *args) 957{ 958 static int last_log; 959 960 ipfw_dyn_rule *q; 961 962 DEB(printf("-- install state type %d 0x%08x %u -> 0x%08x %u\n", 963 cmd->o.opcode, 964 (args->f_id.src_ip), (args->f_id.src_port), 965 (args->f_id.dst_ip), (args->f_id.dst_port) );) 966 967 q = lookup_dyn_rule(&args->f_id, NULL, NULL); 968 969 if (q != NULL) { /* should never occur */ 970 if (last_log != time_second) { 971 last_log = time_second; 972 printf(" install_state: entry already present, done\n"); 973 } 974 return 0; 975 } 976 977 if (dyn_count >= dyn_max) 978 /* 979 * Run out of slots, try to remove any expired rule. 980 */ 981 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 982 983 if (dyn_count >= dyn_max) { 984 if (last_log != time_second) { 985 last_log = time_second; 986 printf("install_state: Too many dynamic rules\n"); 987 } 988 return 1; /* cannot install, notify caller */ 989 } 990 991 switch (cmd->o.opcode) { 992 case O_KEEP_STATE: /* bidir rule */ 993 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 994 break; 995 996 case O_LIMIT: /* limit number of sessions */ 997 { 998 u_int16_t limit_mask = cmd->limit_mask; 999 struct ipfw_flow_id id; 1000 ipfw_dyn_rule *parent; 1001 1002 DEB(printf("installing dyn-limit rule %d\n", cmd->conn_limit);) 1003 1004 id.dst_ip = id.src_ip = 0; 1005 id.dst_port = id.src_port = 0; 1006 id.proto = args->f_id.proto; 1007 1008 if (limit_mask & DYN_SRC_ADDR) 1009 id.src_ip = args->f_id.src_ip; 1010 if (limit_mask & DYN_DST_ADDR) 1011 id.dst_ip = args->f_id.dst_ip; 1012 if (limit_mask & DYN_SRC_PORT) 1013 id.src_port = args->f_id.src_port; 1014 if (limit_mask & DYN_DST_PORT) 1015 id.dst_port = args->f_id.dst_port; 1016 parent = lookup_dyn_parent(&id, rule); 1017 if (parent == NULL) { 1018 printf("add parent failed\n"); 1019 return 1; 1020 } 1021 if (parent->count >= cmd->conn_limit) { 1022 /* 1023 * See if we can remove some expired rule. 1024 */ 1025 remove_dyn_rule(rule, parent); 1026 if (parent->count >= cmd->conn_limit) { 1027 if (fw_verbose && last_log != time_second) { 1028 last_log = time_second; 1029 log(LOG_SECURITY | LOG_INFO, 1030 "drop session, too many entries\n"); 1031 } 1032 return 1; 1033 } 1034 } 1035 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1036 } 1037 break; 1038 default: 1039 printf("unknown dynamic rule type %u\n", cmd->o.opcode); 1040 return 1; 1041 } 1042 lookup_dyn_rule(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1043 return 0; 1044} 1045 1046/* 1047 * Transmit a TCP packet, containing either a RST or a keepalive. 1048 * When flags & TH_RST, we are sending a RST packet, because of a 1049 * "reset" action matched the packet. 1050 * Otherwise we are sending a keepalive, and flags & TH_ 1051 */ 1052static void 1053send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1054{ 1055 struct mbuf *m; 1056 struct ip *ip; 1057 struct tcphdr *tcp; 1058 struct route sro; /* fake route */ 1059 1060 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1061 if (m == 0) 1062 return; 1063 m->m_pkthdr.rcvif = (struct ifnet *)0; 1064 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1065 m->m_data += max_linkhdr; 1066 1067 ip = mtod(m, struct ip *); 1068 bzero(ip, m->m_len); 1069 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1070 ip->ip_p = IPPROTO_TCP; 1071 tcp->th_off = 5; 1072 /* 1073 * Assume we are sending a RST (or a keepalive in the reverse 1074 * direction), swap src and destination addresses and ports. 1075 */ 1076 ip->ip_src.s_addr = htonl(id->dst_ip); 1077 ip->ip_dst.s_addr = htonl(id->src_ip); 1078 tcp->th_sport = htons(id->dst_port); 1079 tcp->th_dport = htons(id->src_port); 1080 if (flags & TH_RST) { /* we are sending a RST */ 1081 if (flags & TH_ACK) { 1082 tcp->th_seq = htonl(ack); 1083 tcp->th_ack = htonl(0); 1084 tcp->th_flags = TH_RST; 1085 } else { 1086 if (flags & TH_SYN) 1087 seq++; 1088 tcp->th_seq = htonl(0); 1089 tcp->th_ack = htonl(seq); 1090 tcp->th_flags = TH_RST | TH_ACK; 1091 } 1092 } else { 1093 /* 1094 * We are sending a keepalive. flags & TH_SYN determines 1095 * the direction, forward if set, reverse if clear. 1096 * NOTE: seq and ack are always assumed to be correct 1097 * as set by the caller. This may be confusing... 1098 */ 1099 if (flags & TH_SYN) { 1100 /* 1101 * we have to rewrite the correct addresses! 1102 */ 1103 ip->ip_dst.s_addr = htonl(id->dst_ip); 1104 ip->ip_src.s_addr = htonl(id->src_ip); 1105 tcp->th_dport = htons(id->dst_port); 1106 tcp->th_sport = htons(id->src_port); 1107 } 1108 tcp->th_seq = htonl(seq); 1109 tcp->th_ack = htonl(ack); 1110 tcp->th_flags = TH_ACK; 1111 } 1112 /* 1113 * set ip_len to the payload size so we can compute 1114 * the tcp checksum on the pseudoheader 1115 * XXX check this, could save a couple of words ? 1116 */ 1117 ip->ip_len = htons(sizeof(struct tcphdr)); 1118 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1119 /* 1120 * now fill fields left out earlier 1121 */ 1122 ip->ip_ttl = ip_defttl; 1123 ip->ip_len = m->m_pkthdr.len; 1124 bzero (&sro, sizeof (sro)); 1125 ip_rtaddr(ip->ip_dst, &sro); 1126 m->m_flags |= M_SKIP_FIREWALL; 1127 ip_output(m, NULL, &sro, 0, NULL, NULL); 1128 if (sro.ro_rt) 1129 RTFREE(sro.ro_rt); 1130} 1131 1132/* 1133 * sends a reject message, consuming the mbuf passed as an argument. 1134 */ 1135static void 1136send_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 1137{ 1138 1139 if (code != ICMP_REJECT_RST) /* Send an ICMP unreach */ 1140 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1141 else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1142 struct tcphdr *const tcp = 1143 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1144 if ( (tcp->th_flags & TH_RST) == 0) 1145 send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1146 ntohl(tcp->th_ack), 1147 tcp->th_flags | TH_RST); 1148 m_freem(args->m); 1149 } else 1150 m_freem(args->m); 1151 args->m = NULL; 1152} 1153 1154/** 1155 * 1156 * Given an ip_fw *, lookup_next_rule will return a pointer 1157 * to the next rule, which can be either the jump 1158 * target (for skipto instructions) or the next one in the list (in 1159 * all other cases including a missing jump target). 1160 * The result is also written in the "next_rule" field of the rule. 1161 * Backward jumps are not allowed, so start looking from the next 1162 * rule... 1163 * 1164 * This never returns NULL -- in case we do not have an exact match, 1165 * the next rule is returned. When the ruleset is changed, 1166 * pointers are flushed so we are always correct. 1167 */ 1168 1169static struct ip_fw * 1170lookup_next_rule(struct ip_fw *me) 1171{ 1172 struct ip_fw *rule = NULL; 1173 ipfw_insn *cmd; 1174 1175 /* look for action, in case it is a skipto */ 1176 cmd = ACTION_PTR(me); 1177 if ( cmd->opcode == O_SKIPTO ) 1178 for (rule = me->next; rule ; rule = rule->next) 1179 if (rule->rulenum >= cmd->arg1) 1180 break; 1181 if (rule == NULL) /* failure or not a skipto */ 1182 rule = me->next; 1183 me->next_rule = rule; 1184 return rule; 1185} 1186 1187/* 1188 * The main check routine for the firewall. 1189 * 1190 * All arguments are in args so we can modify them and return them 1191 * back to the caller. 1192 * 1193 * Parameters: 1194 * 1195 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1196 * Starts with the IP header. 1197 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1198 * args->oif Outgoing interface, or NULL if packet is incoming. 1199 * The incoming interface is in the mbuf. (in) 1200 * args->divert_rule (in/out) 1201 * Skip up to the first rule past this rule number; 1202 * upon return, non-zero port number for divert or tee. 1203 * 1204 * args->rule Pointer to the last matching rule (in/out) 1205 * args->next_hop Socket we are forwarding to (out). 1206 * args->f_id Addresses grabbed from the packet (out) 1207 * 1208 * Return value: 1209 * 1210 * IP_FW_PORT_DENY_FLAG the packet must be dropped. 1211 * 0 The packet is to be accepted and routed normally OR 1212 * the packet was denied/rejected and has been dropped; 1213 * in the latter case, *m is equal to NULL upon return. 1214 * port Divert the packet to port, with these caveats: 1215 * 1216 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead 1217 * of diverting it (ie, 'ipfw tee'). 1218 * 1219 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower 1220 * 16 bits as a dummynet pipe number instead of diverting 1221 */ 1222 1223static int 1224ipfw_chk(struct ip_fw_args *args) 1225{ 1226 /* 1227 * Local variables hold state during the processing of a packet. 1228 * 1229 * IMPORTANT NOTE: to speed up the processing of rules, there 1230 * are some assumption on the values of the variables, which 1231 * are documented here. Should you change them, please check 1232 * the implementation of the various instructions to make sure 1233 * that they still work. 1234 * 1235 * args->eh The MAC header. It is non-null for a layer2 1236 * packet, it is NULL for a layer-3 packet. 1237 * 1238 * m | args->m Pointer to the mbuf, as received from the caller. 1239 * It may change if ipfw_chk() does an m_pullup, or if it 1240 * consumes the packet because it calls send_reject(). 1241 * XXX This has to change, so that ipfw_chk() never modifies 1242 * or consumes the buffer. 1243 * ip is simply an alias of the value of m, and it is kept 1244 * in sync with it (the packet is supposed to start with 1245 * the ip header). 1246 */ 1247 struct mbuf *m = args->m; 1248 struct ip *ip = mtod(m, struct ip *); 1249 1250 /* 1251 * oif | args->oif If NULL, ipfw_chk has been called on the 1252 * inbound path (ether_input, bdg_forward, ip_input). 1253 * If non-NULL, ipfw_chk has been called on the outbound path 1254 * (ether_output, ip_output). 1255 */ 1256 struct ifnet *oif = args->oif; 1257 1258 struct ip_fw *f = NULL; /* matching rule */ 1259 int retval = 0; 1260 1261 /* 1262 * hlen The length of the IPv4 header. 1263 * hlen >0 means we have an IPv4 packet. 1264 */ 1265 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1266 1267 /* 1268 * offset The offset of a fragment. offset != 0 means that 1269 * we have a fragment at this offset of an IPv4 packet. 1270 * offset == 0 means that (if this is an IPv4 packet) 1271 * this is the first or only fragment. 1272 */ 1273 u_short offset = 0; 1274 1275 /* 1276 * Local copies of addresses. They are only valid if we have 1277 * an IP packet. 1278 * 1279 * proto The protocol. Set to 0 for non-ip packets, 1280 * or to the protocol read from the packet otherwise. 1281 * proto != 0 means that we have an IPv4 packet. 1282 * 1283 * src_port, dst_port port numbers, in HOST format. Only 1284 * valid for TCP and UDP packets. 1285 * 1286 * src_ip, dst_ip ip addresses, in NETWORK format. 1287 * Only valid for IPv4 packets. 1288 */ 1289 u_int8_t proto; 1290 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1291 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1292 u_int16_t ip_len=0; 1293 int dyn_dir = MATCH_UNKNOWN; 1294 ipfw_dyn_rule *q = NULL; 1295 1296 if (m->m_flags & M_SKIP_FIREWALL) 1297 return 0; /* accept */ 1298 /* 1299 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1300 * MATCH_NONE when checked and not matched (q = NULL), 1301 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1302 */ 1303 1304 if (args->eh == NULL || /* layer 3 packet */ 1305 ( m->m_pkthdr.len >= sizeof(struct ip) && 1306 ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 1307 hlen = ip->ip_hl << 2; 1308 1309 /* 1310 * Collect parameters into local variables for faster matching. 1311 */ 1312 if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 1313 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1314 goto after_ip_checks; 1315 } 1316 1317 proto = args->f_id.proto = ip->ip_p; 1318 src_ip = ip->ip_src; 1319 dst_ip = ip->ip_dst; 1320 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1321 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1322 ip_len = ntohs(ip->ip_len); 1323 } else { 1324 offset = ip->ip_off & IP_OFFMASK; 1325 ip_len = ip->ip_len; 1326 } 1327 1328#define PULLUP_TO(len) \ 1329 do { \ 1330 if ((m)->m_len < (len)) { \ 1331 args->m = m = m_pullup(m, (len)); \ 1332 if (m == 0) \ 1333 goto pullup_failed; \ 1334 ip = mtod(m, struct ip *); \ 1335 } \ 1336 } while (0) 1337 1338 if (offset == 0) { 1339 switch (proto) { 1340 case IPPROTO_TCP: 1341 { 1342 struct tcphdr *tcp; 1343 1344 PULLUP_TO(hlen + sizeof(struct tcphdr)); 1345 tcp = L3HDR(struct tcphdr, ip); 1346 dst_port = tcp->th_dport; 1347 src_port = tcp->th_sport; 1348 args->f_id.flags = tcp->th_flags; 1349 } 1350 break; 1351 1352 case IPPROTO_UDP: 1353 { 1354 struct udphdr *udp; 1355 1356 PULLUP_TO(hlen + sizeof(struct udphdr)); 1357 udp = L3HDR(struct udphdr, ip); 1358 dst_port = udp->uh_dport; 1359 src_port = udp->uh_sport; 1360 } 1361 break; 1362 1363 case IPPROTO_ICMP: 1364 PULLUP_TO(hlen + 4); /* type, code and checksum. */ 1365 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 1366 break; 1367 1368 default: 1369 break; 1370 } 1371#undef PULLUP_TO 1372 } 1373 1374 args->f_id.src_ip = ntohl(src_ip.s_addr); 1375 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1376 args->f_id.src_port = src_port = ntohs(src_port); 1377 args->f_id.dst_port = dst_port = ntohs(dst_port); 1378 1379after_ip_checks: 1380 if (args->rule) { 1381 /* 1382 * Packet has already been tagged. Look for the next rule 1383 * to restart processing. 1384 * 1385 * If fw_one_pass != 0 then just accept it. 1386 * XXX should not happen here, but optimized out in 1387 * the caller. 1388 */ 1389 if (fw_one_pass) 1390 return 0; 1391 1392 f = args->rule->next_rule; 1393 if (f == NULL) 1394 f = lookup_next_rule(args->rule); 1395 } else { 1396 /* 1397 * Find the starting rule. It can be either the first 1398 * one, or the one after divert_rule if asked so. 1399 */ 1400 int skipto = args->divert_rule; 1401 1402 f = layer3_chain; 1403 if (args->eh == NULL && skipto != 0) { 1404 if (skipto >= IPFW_DEFAULT_RULE) 1405 return(IP_FW_PORT_DENY_FLAG); /* invalid */ 1406 while (f && f->rulenum <= skipto) 1407 f = f->next; 1408 if (f == NULL) /* drop packet */ 1409 return(IP_FW_PORT_DENY_FLAG); 1410 } 1411 } 1412 args->divert_rule = 0; /* reset to avoid confusion later */ 1413 1414 /* 1415 * Now scan the rules, and parse microinstructions for each rule. 1416 */ 1417 for (; f; f = f->next) { 1418 int l, cmdlen; 1419 ipfw_insn *cmd; 1420 int skip_or; /* skip rest of OR block */ 1421 1422again: 1423 if (set_disable & (1 << f->set) ) 1424 continue; 1425 1426 skip_or = 0; 1427 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1428 l -= cmdlen, cmd += cmdlen) { 1429 int match; 1430 1431 /* 1432 * check_body is a jump target used when we find a 1433 * CHECK_STATE, and need to jump to the body of 1434 * the target rule. 1435 */ 1436 1437check_body: 1438 cmdlen = F_LEN(cmd); 1439 /* 1440 * An OR block (insn_1 || .. || insn_n) has the 1441 * F_OR bit set in all but the last instruction. 1442 * The first match will set "skip_or", and cause 1443 * the following instructions to be skipped until 1444 * past the one with the F_OR bit clear. 1445 */ 1446 if (skip_or) { /* skip this instruction */ 1447 if ((cmd->len & F_OR) == 0) 1448 skip_or = 0; /* next one is good */ 1449 continue; 1450 } 1451 match = 0; /* set to 1 if we succeed */ 1452 1453 switch (cmd->opcode) { 1454 /* 1455 * The first set of opcodes compares the packet's 1456 * fields with some pattern, setting 'match' if a 1457 * match is found. At the end of the loop there is 1458 * logic to deal with F_NOT and F_OR flags associated 1459 * with the opcode. 1460 */ 1461 case O_NOP: 1462 match = 1; 1463 break; 1464 1465 case O_FORWARD_MAC: 1466 printf("ipfw: opcode %d unimplemented\n", 1467 cmd->opcode); 1468 break; 1469 1470 case O_GID: 1471 case O_UID: 1472 /* 1473 * We only check offset == 0 && proto != 0, 1474 * as this ensures that we have an IPv4 1475 * packet with the ports info. 1476 */ 1477 if (offset!=0) 1478 break; 1479 { 1480 struct inpcbinfo *pi; 1481 int wildcard; 1482 struct inpcb *pcb; 1483 1484 if (proto == IPPROTO_TCP) { 1485 wildcard = 0; 1486 pi = &tcbinfo; 1487 } else if (proto == IPPROTO_UDP) { 1488 wildcard = 1; 1489 pi = &udbinfo; 1490 } else 1491 break; 1492 1493 pcb = (oif) ? 1494 in_pcblookup_hash(pi, 1495 dst_ip, htons(dst_port), 1496 src_ip, htons(src_port), 1497 wildcard, oif) : 1498 in_pcblookup_hash(pi, 1499 src_ip, htons(src_port), 1500 dst_ip, htons(dst_port), 1501 wildcard, NULL); 1502 1503 if (pcb == NULL || pcb->inp_socket == NULL) 1504 break; 1505#if __FreeBSD_version < 500034 1506#define socheckuid(a,b) ((a)->so_cred->cr_uid == (b)) 1507#endif 1508 if (cmd->opcode == O_UID) { 1509 match = 1510 socheckuid(pcb->inp_socket, 1511 (uid_t)((ipfw_insn_u32 *)cmd)->d[0]); 1512 } else { 1513 match = groupmember( 1514 (uid_t)((ipfw_insn_u32 *)cmd)->d[0], 1515 pcb->inp_socket->so_cred); 1516 } 1517 } 1518 break; 1519 1520 case O_RECV: 1521 match = iface_match(m->m_pkthdr.rcvif, 1522 (ipfw_insn_if *)cmd); 1523 break; 1524 1525 case O_XMIT: 1526 match = iface_match(oif, (ipfw_insn_if *)cmd); 1527 break; 1528 1529 case O_VIA: 1530 match = iface_match(oif ? oif : 1531 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1532 break; 1533 1534 case O_MACADDR2: 1535 if (args->eh != NULL) { /* have MAC header */ 1536 u_int32_t *want = (u_int32_t *) 1537 ((ipfw_insn_mac *)cmd)->addr; 1538 u_int32_t *mask = (u_int32_t *) 1539 ((ipfw_insn_mac *)cmd)->mask; 1540 u_int32_t *hdr = (u_int32_t *)args->eh; 1541 1542 match = 1543 ( want[0] == (hdr[0] & mask[0]) && 1544 want[1] == (hdr[1] & mask[1]) && 1545 want[2] == (hdr[2] & mask[2]) ); 1546 } 1547 break; 1548 1549 case O_MAC_TYPE: 1550 if (args->eh != NULL) { 1551 u_int16_t t = 1552 ntohs(args->eh->ether_type); 1553 u_int16_t *p = 1554 ((ipfw_insn_u16 *)cmd)->ports; 1555 int i; 1556 1557 for (i = cmdlen - 1; !match && i>0; 1558 i--, p += 2) 1559 match = (t>=p[0] && t<=p[1]); 1560 } 1561 break; 1562 1563 case O_FRAG: 1564 match = (hlen > 0 && offset != 0); 1565 break; 1566 1567 case O_IN: /* "out" is "not in" */ 1568 match = (oif == NULL); 1569 break; 1570 1571 case O_LAYER2: 1572 match = (args->eh != NULL); 1573 break; 1574 1575 case O_PROTO: 1576 /* 1577 * We do not allow an arg of 0 so the 1578 * check of "proto" only suffices. 1579 */ 1580 match = (proto == cmd->arg1); 1581 break; 1582 1583 case O_IP_SRC: 1584 match = (hlen > 0 && 1585 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1586 src_ip.s_addr); 1587 break; 1588 1589 case O_IP_SRC_MASK: 1590 match = (hlen > 0 && 1591 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1592 (src_ip.s_addr & 1593 ((ipfw_insn_ip *)cmd)->mask.s_addr)); 1594 break; 1595 1596 case O_IP_SRC_ME: 1597 if (hlen > 0) { 1598 struct ifnet *tif; 1599 1600 INADDR_TO_IFP(src_ip, tif); 1601 match = (tif != NULL); 1602 } 1603 break; 1604 1605 case O_IP_DST_SET: 1606 case O_IP_SRC_SET: 1607 if (hlen > 0) { 1608 u_int32_t *d = (u_int32_t *)(cmd+1); 1609 u_int32_t addr = 1610 cmd->opcode == O_IP_DST_SET ? 1611 args->f_id.dst_ip : 1612 args->f_id.src_ip; 1613 1614 if (addr < d[0]) 1615 break; 1616 addr -= d[0]; /* subtract base */ 1617 match = (addr < cmd->arg1) && 1618 ( d[ 1 + (addr>>5)] & 1619 (1<<(addr & 0x1f)) ); 1620 } 1621 break; 1622 1623 case O_IP_DST: 1624 match = (hlen > 0 && 1625 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1626 dst_ip.s_addr); 1627 break; 1628 1629 case O_IP_DST_MASK: 1630 match = (hlen > 0) && 1631 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1632 (dst_ip.s_addr & 1633 ((ipfw_insn_ip *)cmd)->mask.s_addr)); 1634 break; 1635 1636 case O_IP_DST_ME: 1637 if (hlen > 0) { 1638 struct ifnet *tif; 1639 1640 INADDR_TO_IFP(dst_ip, tif); 1641 match = (tif != NULL); 1642 } 1643 break; 1644 1645 case O_IP_SRCPORT: 1646 case O_IP_DSTPORT: 1647 /* 1648 * offset == 0 && proto != 0 is enough 1649 * to guarantee that we have an IPv4 1650 * packet with port info. 1651 */ 1652 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 1653 && offset == 0) { 1654 u_int16_t x = 1655 (cmd->opcode == O_IP_SRCPORT) ? 1656 src_port : dst_port ; 1657 u_int16_t *p = 1658 ((ipfw_insn_u16 *)cmd)->ports; 1659 int i; 1660 1661 for (i = cmdlen - 1; !match && i>0; 1662 i--, p += 2) 1663 match = (x>=p[0] && x<=p[1]); 1664 } 1665 break; 1666 1667 case O_ICMPTYPE: 1668 match = (offset == 0 && proto==IPPROTO_ICMP && 1669 icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 1670 break; 1671 1672 case O_IPOPT: 1673 match = (hlen > 0 && ipopts_match(ip, cmd) ); 1674 break; 1675 1676 case O_IPVER: 1677 match = (hlen > 0 && cmd->arg1 == ip->ip_v); 1678 break; 1679 1680 case O_IPTTL: 1681 match = (hlen > 0 && cmd->arg1 == ip->ip_ttl); 1682 break; 1683 1684 case O_IPID: 1685 match = (hlen > 0 && 1686 cmd->arg1 == ntohs(ip->ip_id)); 1687 break; 1688 1689 case O_IPLEN: 1690 match = (hlen > 0 && cmd->arg1 == ip_len); 1691 break; 1692 1693 case O_IPPRECEDENCE: 1694 match = (hlen > 0 && 1695 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 1696 break; 1697 1698 case O_IPTOS: 1699 match = (hlen > 0 && 1700 flags_match(cmd, ip->ip_tos)); 1701 break; 1702 1703 case O_TCPFLAGS: 1704 match = (proto == IPPROTO_TCP && offset == 0 && 1705 flags_match(cmd, 1706 L3HDR(struct tcphdr,ip)->th_flags)); 1707 break; 1708 1709 case O_TCPOPTS: 1710 match = (proto == IPPROTO_TCP && offset == 0 && 1711 tcpopts_match(ip, cmd)); 1712 break; 1713 1714 case O_TCPSEQ: 1715 match = (proto == IPPROTO_TCP && offset == 0 && 1716 ((ipfw_insn_u32 *)cmd)->d[0] == 1717 L3HDR(struct tcphdr,ip)->th_seq); 1718 break; 1719 1720 case O_TCPACK: 1721 match = (proto == IPPROTO_TCP && offset == 0 && 1722 ((ipfw_insn_u32 *)cmd)->d[0] == 1723 L3HDR(struct tcphdr,ip)->th_ack); 1724 break; 1725 1726 case O_TCPWIN: 1727 match = (proto == IPPROTO_TCP && offset == 0 && 1728 cmd->arg1 == 1729 L3HDR(struct tcphdr,ip)->th_win); 1730 break; 1731 1732 case O_ESTAB: 1733 /* reject packets which have SYN only */ 1734 /* XXX should i also check for TH_ACK ? */ 1735 match = (proto == IPPROTO_TCP && offset == 0 && 1736 (L3HDR(struct tcphdr,ip)->th_flags & 1737 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 1738 break; 1739 1740 case O_LOG: 1741 if (fw_verbose) 1742 ipfw_log(f, hlen, args->eh, m, oif); 1743 match = 1; 1744 break; 1745 1746 case O_PROB: 1747 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 1748 break; 1749 1750 /* 1751 * The second set of opcodes represents 'actions', 1752 * i.e. the terminal part of a rule once the packet 1753 * matches all previous patterns. 1754 * Typically there is only one action for each rule, 1755 * and the opcode is stored at the end of the rule 1756 * (but there are exceptions -- see below). 1757 * 1758 * In general, here we set retval and terminate the 1759 * outer loop (would be a 'break 3' in some language, 1760 * but we need to do a 'goto done'). 1761 * 1762 * Exceptions: 1763 * O_COUNT and O_SKIPTO actions: 1764 * instead of terminating, we jump to the next rule 1765 * ('goto next_rule', equivalent to a 'break 2'), 1766 * or to the SKIPTO target ('goto again' after 1767 * having set f, cmd and l), respectively. 1768 * 1769 * O_LIMIT and O_KEEP_STATE: these opcodes are 1770 * not real 'actions', and are stored right 1771 * before the 'action' part of the rule. 1772 * These opcodes try to install an entry in the 1773 * state tables; if successful, we continue with 1774 * the next opcode (match=1; break;), otherwise 1775 * the packet * must be dropped 1776 * ('goto done' after setting retval); 1777 * 1778 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 1779 * cause a lookup of the state table, and a jump 1780 * to the 'action' part of the parent rule 1781 * ('goto check_body') if an entry is found, or 1782 * (CHECK_STATE only) a jump to the next rule if 1783 * the entry is not found ('goto next_rule'). 1784 * The result of the lookup is cached to make 1785 * further instances of these opcodes are 1786 * effectively NOPs. 1787 */ 1788 case O_LIMIT: 1789 case O_KEEP_STATE: 1790 if (install_state(f, 1791 (ipfw_insn_limit *)cmd, args)) { 1792 retval = IP_FW_PORT_DENY_FLAG; 1793 goto done; /* error/limit violation */ 1794 } 1795 match = 1; 1796 break; 1797 1798 case O_PROBE_STATE: 1799 case O_CHECK_STATE: 1800 /* 1801 * dynamic rules are checked at the first 1802 * keep-state or check-state occurrence, 1803 * with the result being stored in dyn_dir. 1804 * The compiler introduces a PROBE_STATE 1805 * instruction for us when we have a 1806 * KEEP_STATE (because PROBE_STATE needs 1807 * to be run first). 1808 */ 1809 if (dyn_dir == MATCH_UNKNOWN && 1810 (q = lookup_dyn_rule(&args->f_id, 1811 &dyn_dir, proto == IPPROTO_TCP ? 1812 L3HDR(struct tcphdr, ip) : NULL)) 1813 != NULL) { 1814 /* 1815 * Found dynamic entry, update stats 1816 * and jump to the 'action' part of 1817 * the parent rule. 1818 */ 1819 q->pcnt++; 1820 q->bcnt += ip_len; 1821 f = q->rule; 1822 cmd = ACTION_PTR(f); 1823 l = f->cmd_len - f->act_ofs; 1824 goto check_body; 1825 } 1826 /* 1827 * Dynamic entry not found. If CHECK_STATE, 1828 * skip to next rule, if PROBE_STATE just 1829 * ignore and continue with next opcode. 1830 */ 1831 if (cmd->opcode == O_CHECK_STATE) 1832 goto next_rule; 1833 match = 1; 1834 break; 1835 1836 case O_ACCEPT: 1837 retval = 0; /* accept */ 1838 goto done; 1839 1840 case O_PIPE: 1841 case O_QUEUE: 1842 args->rule = f; /* report matching rule */ 1843 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG; 1844 goto done; 1845 1846 case O_DIVERT: 1847 case O_TEE: 1848 if (args->eh) /* not on layer 2 */ 1849 break; 1850 args->divert_rule = f->rulenum; 1851 retval = (cmd->opcode == O_DIVERT) ? 1852 cmd->arg1 : 1853 cmd->arg1 | IP_FW_PORT_TEE_FLAG; 1854 goto done; 1855 1856 case O_COUNT: 1857 case O_SKIPTO: 1858 f->pcnt++; /* update stats */ 1859 f->bcnt += ip_len; 1860 f->timestamp = time_second; 1861 if (cmd->opcode == O_COUNT) 1862 goto next_rule; 1863 /* handle skipto */ 1864 if (f->next_rule == NULL) 1865 lookup_next_rule(f); 1866 f = f->next_rule; 1867 goto again; 1868 1869 case O_REJECT: 1870 /* 1871 * Drop the packet and send a reject notice 1872 * if the packet is not ICMP (or is an ICMP 1873 * query), and it is not multicast/broadcast. 1874 */ 1875 if (hlen > 0 && 1876 (proto != IPPROTO_ICMP || 1877 is_icmp_query(ip)) && 1878 !(m->m_flags & (M_BCAST|M_MCAST)) && 1879 !IN_MULTICAST(dst_ip.s_addr)) { 1880 send_reject(args, cmd->arg1, 1881 offset,ip_len); 1882 m = args->m; 1883 } 1884 /* FALLTHROUGH */ 1885 case O_DENY: 1886 retval = IP_FW_PORT_DENY_FLAG; 1887 goto done; 1888 1889 case O_FORWARD_IP: 1890 if (args->eh) /* not valid on layer2 pkts */ 1891 break; 1892 if (!q || dyn_dir == MATCH_FORWARD) 1893 args->next_hop = 1894 &((ipfw_insn_sa *)cmd)->sa; 1895 retval = 0; 1896 goto done; 1897 1898 default: 1899 panic("-- unknown opcode %d\n", cmd->opcode); 1900 } /* end of switch() on opcodes */ 1901 1902 if (cmd->len & F_NOT) 1903 match = !match; 1904 1905 if (match) { 1906 if (cmd->len & F_OR) 1907 skip_or = 1; 1908 } else { 1909 if (!(cmd->len & F_OR)) /* not an OR block, */ 1910 break; /* try next rule */ 1911 } 1912 1913 } /* end of inner for, scan opcodes */ 1914 1915next_rule:; /* try next rule */ 1916 1917 } /* end of outer for, scan rules */ 1918 printf("+++ ipfw: ouch!, skip past end of rules, denying packet\n"); 1919 return(IP_FW_PORT_DENY_FLAG); 1920 1921done: 1922 /* Update statistics */ 1923 f->pcnt++; 1924 f->bcnt += ip_len; 1925 f->timestamp = time_second; 1926 return retval; 1927 1928pullup_failed: 1929 if (fw_verbose) 1930 printf("pullup failed\n"); 1931 return(IP_FW_PORT_DENY_FLAG); 1932} 1933 1934/* 1935 * When a rule is added/deleted, clear the next_rule pointers in all rules. 1936 * These will be reconstructed on the fly as packets are matched. 1937 * Must be called at splimp(). 1938 */ 1939static void 1940flush_rule_ptrs(void) 1941{ 1942 struct ip_fw *rule; 1943 1944 for (rule = layer3_chain; rule; rule = rule->next) 1945 rule->next_rule = NULL; 1946} 1947 1948/* 1949 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 1950 * pipe/queue, or to all of them (match == NULL). 1951 * Must be called at splimp(). 1952 */ 1953void 1954flush_pipe_ptrs(struct dn_flow_set *match) 1955{ 1956 struct ip_fw *rule; 1957 1958 for (rule = layer3_chain; rule; rule = rule->next) { 1959 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 1960 1961 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 1962 continue; 1963 if (match == NULL || cmd->pipe_ptr == match) 1964 cmd->pipe_ptr = NULL; 1965 } 1966} 1967 1968/* 1969 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 1970 * possibly create a rule number and add the rule to the list. 1971 * Update the rule_number in the input struct so the caller knows it as well. 1972 */ 1973static int 1974add_rule(struct ip_fw **head, struct ip_fw *input_rule) 1975{ 1976 struct ip_fw *rule, *f, *prev; 1977 int s; 1978 int l = RULESIZE(input_rule); 1979 1980 if (*head == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 1981 return (EINVAL); 1982 1983 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 1984 if (rule == NULL) 1985 return (ENOSPC); 1986 1987 bcopy(input_rule, rule, l); 1988 1989 rule->next = NULL; 1990 rule->next_rule = NULL; 1991 1992 rule->pcnt = 0; 1993 rule->bcnt = 0; 1994 rule->timestamp = 0; 1995 1996 s = splimp(); 1997 1998 if (*head == NULL) { /* default rule */ 1999 *head = rule; 2000 goto done; 2001 } 2002 2003 /* 2004 * If rulenum is 0, find highest numbered rule before the 2005 * default rule, and add autoinc_step 2006 */ 2007 if (autoinc_step < 1) 2008 autoinc_step = 1; 2009 else if (autoinc_step > 1000) 2010 autoinc_step = 1000; 2011 if (rule->rulenum == 0) { 2012 /* 2013 * locate the highest numbered rule before default 2014 */ 2015 for (f = *head; f; f = f->next) { 2016 if (f->rulenum == IPFW_DEFAULT_RULE) 2017 break; 2018 rule->rulenum = f->rulenum; 2019 } 2020 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2021 rule->rulenum += autoinc_step; 2022 input_rule->rulenum = rule->rulenum; 2023 } 2024 2025 /* 2026 * Now insert the new rule in the right place in the sorted list. 2027 */ 2028 for (prev = NULL, f = *head; f; prev = f, f = f->next) { 2029 if (f->rulenum > rule->rulenum) { /* found the location */ 2030 if (prev) { 2031 rule->next = f; 2032 prev->next = rule; 2033 } else { /* head insert */ 2034 rule->next = *head; 2035 *head = rule; 2036 } 2037 break; 2038 } 2039 } 2040 flush_rule_ptrs(); 2041done: 2042 static_count++; 2043 static_len += l; 2044 splx(s); 2045 DEB(printf("++ installed rule %d, static count now %d\n", 2046 rule->rulenum, static_count);) 2047 return (0); 2048} 2049 2050/** 2051 * Free storage associated with a static rule (including derived 2052 * dynamic rules). 2053 * The caller is in charge of clearing rule pointers to avoid 2054 * dangling pointers. 2055 * @return a pointer to the next entry. 2056 * Arguments are not checked, so they better be correct. 2057 * Must be called at splimp(). 2058 */ 2059static struct ip_fw * 2060delete_rule(struct ip_fw **head, struct ip_fw *prev, struct ip_fw *rule) 2061{ 2062 struct ip_fw *n; 2063 int l = RULESIZE(rule); 2064 2065 n = rule->next; 2066 remove_dyn_rule(rule, NULL /* force removal */); 2067 if (prev == NULL) 2068 *head = n; 2069 else 2070 prev->next = n; 2071 static_count--; 2072 static_len -= l; 2073 2074 if (DUMMYNET_LOADED) 2075 ip_dn_ruledel_ptr(rule); 2076 free(rule, M_IPFW); 2077 return n; 2078} 2079 2080/* 2081 * Deletes all rules from a chain (including the default rule 2082 * if the second argument is set). 2083 * Must be called at splimp(). 2084 */ 2085static void 2086free_chain(struct ip_fw **chain, int kill_default) 2087{ 2088 struct ip_fw *rule; 2089 2090 flush_rule_ptrs(); /* more efficient to do outside the loop */ 2091 2092 while ( (rule = *chain) != NULL && 2093 (kill_default || rule->rulenum != IPFW_DEFAULT_RULE) ) 2094 delete_rule(chain, NULL, rule); 2095} 2096 2097/** 2098 * Remove all rules with given number, and also do set manipulation. 2099 * 2100 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2101 * the next 8 bits are the new set, the top 8 bits are the command: 2102 * 2103 * 0 delete rules with given number 2104 * 1 delete rules with given set number 2105 * 2 move rules with given number to new set 2106 * 3 move rules with given set number to new set 2107 * 4 swap sets with given numbers 2108 */ 2109static int 2110del_entry(struct ip_fw **chain, u_int32_t arg) 2111{ 2112 struct ip_fw *prev, *rule; 2113 int s; 2114 u_int16_t rulenum; 2115 u_int8_t cmd, new_set; 2116 2117 rulenum = arg & 0xffff; 2118 cmd = (arg >> 24) & 0xff; 2119 new_set = (arg >> 16) & 0xff; 2120 2121 if (cmd > 4) 2122 return EINVAL; 2123 if (new_set > 30) 2124 return EINVAL; 2125 if (cmd == 0 || cmd == 2) { 2126 if (rulenum == IPFW_DEFAULT_RULE) 2127 return EINVAL; 2128 } else { 2129 if (rulenum > 30) 2130 return EINVAL; 2131 } 2132 2133 switch (cmd) { 2134 case 0: /* delete rules with given number */ 2135 /* 2136 * locate first rule to delete 2137 */ 2138 for (prev = NULL, rule = *chain; 2139 rule && rule->rulenum < rulenum; 2140 prev = rule, rule = rule->next) 2141 ; 2142 if (rule->rulenum != rulenum) 2143 return EINVAL; 2144 2145 s = splimp(); /* no access to rules while removing */ 2146 /* 2147 * flush pointers outside the loop, then delete all matching 2148 * rules. prev remains the same throughout the cycle. 2149 */ 2150 flush_rule_ptrs(); 2151 while (rule && rule->rulenum == rulenum) 2152 rule = delete_rule(chain, prev, rule); 2153 splx(s); 2154 break; 2155 2156 case 1: /* delete all rules with given set number */ 2157 s = splimp(); 2158 flush_rule_ptrs(); 2159 for (prev = NULL, rule = *chain; rule ; ) 2160 if (rule->set == rulenum) 2161 rule = delete_rule(chain, prev, rule); 2162 else { 2163 prev = rule; 2164 rule = rule->next; 2165 } 2166 splx(s); 2167 break; 2168 2169 case 2: /* move rules with given number to new set */ 2170 s = splimp(); 2171 for (rule = *chain; rule ; rule = rule->next) 2172 if (rule->rulenum == rulenum) 2173 rule->set = new_set; 2174 splx(s); 2175 break; 2176 2177 case 3: /* move rules with given set number to new set */ 2178 s = splimp(); 2179 for (rule = *chain; rule ; rule = rule->next) 2180 if (rule->set == rulenum) 2181 rule->set = new_set; 2182 splx(s); 2183 break; 2184 2185 case 4: /* swap two sets */ 2186 s = splimp(); 2187 for (rule = *chain; rule ; rule = rule->next) 2188 if (rule->set == rulenum) 2189 rule->set = new_set; 2190 else if (rule->set == new_set) 2191 rule->set = rulenum; 2192 splx(s); 2193 break; 2194 } 2195 return 0; 2196} 2197 2198/* 2199 * Clear counters for a specific rule. 2200 */ 2201static void 2202clear_counters(struct ip_fw *rule, int log_only) 2203{ 2204 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 2205 2206 if (log_only == 0) { 2207 rule->bcnt = rule->pcnt = 0; 2208 rule->timestamp = 0; 2209 } 2210 if (l->o.opcode == O_LOG) 2211 l->log_left = l->max_log; 2212} 2213 2214/** 2215 * Reset some or all counters on firewall rules. 2216 * @arg frwl is null to clear all entries, or contains a specific 2217 * rule number. 2218 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 2219 */ 2220static int 2221zero_entry(int rulenum, int log_only) 2222{ 2223 struct ip_fw *rule; 2224 int s; 2225 char *msg; 2226 2227 if (rulenum == 0) { 2228 s = splimp(); 2229 norule_counter = 0; 2230 for (rule = layer3_chain; rule; rule = rule->next) 2231 clear_counters(rule, log_only); 2232 splx(s); 2233 msg = log_only ? "ipfw: All logging counts reset.\n" : 2234 "ipfw: Accounting cleared.\n"; 2235 } else { 2236 int cleared = 0; 2237 /* 2238 * We can have multiple rules with the same number, so we 2239 * need to clear them all. 2240 */ 2241 for (rule = layer3_chain; rule; rule = rule->next) 2242 if (rule->rulenum == rulenum) { 2243 s = splimp(); 2244 while (rule && rule->rulenum == rulenum) { 2245 clear_counters(rule, log_only); 2246 rule = rule->next; 2247 } 2248 splx(s); 2249 cleared = 1; 2250 break; 2251 } 2252 if (!cleared) /* we did not find any matching rules */ 2253 return (EINVAL); 2254 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 2255 "ipfw: Entry %d cleared.\n"; 2256 } 2257 if (fw_verbose) 2258 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 2259 return (0); 2260} 2261 2262/* 2263 * Check validity of the structure before insert. 2264 * Fortunately rules are simple, so this mostly need to check rule sizes. 2265 */ 2266static int 2267check_ipfw_struct(struct ip_fw *rule, int size) 2268{ 2269 int l, cmdlen = 0; 2270 int have_action=0; 2271 ipfw_insn *cmd; 2272 2273 if (size < sizeof(*rule)) { 2274 printf("ipfw: rule too short\n"); 2275 return (EINVAL); 2276 } 2277 /* first, check for valid size */ 2278 l = RULESIZE(rule); 2279 if (l != size) { 2280 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 2281 return (EINVAL); 2282 } 2283 /* 2284 * Now go for the individual checks. Very simple ones, basically only 2285 * instruction sizes. 2286 */ 2287 for (l = rule->cmd_len, cmd = rule->cmd ; 2288 l > 0 ; l -= cmdlen, cmd += cmdlen) { 2289 cmdlen = F_LEN(cmd); 2290 if (cmdlen > l) { 2291 printf("ipfw: opcode %d size truncated\n", 2292 cmd->opcode); 2293 return EINVAL; 2294 } 2295 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 2296 switch (cmd->opcode) { 2297 case O_NOP: 2298 case O_PROBE_STATE: 2299 case O_KEEP_STATE: 2300 case O_PROTO: 2301 case O_IP_SRC_ME: 2302 case O_IP_DST_ME: 2303 case O_LAYER2: 2304 case O_IN: 2305 case O_FRAG: 2306 case O_IPOPT: 2307 case O_IPLEN: 2308 case O_IPID: 2309 case O_IPTOS: 2310 case O_IPPRECEDENCE: 2311 case O_IPTTL: 2312 case O_IPVER: 2313 case O_TCPWIN: 2314 case O_TCPFLAGS: 2315 case O_TCPOPTS: 2316 case O_ESTAB: 2317 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2318 goto bad_size; 2319 break; 2320 2321 case O_UID: 2322 case O_GID: 2323 case O_IP_SRC: 2324 case O_IP_DST: 2325 case O_TCPSEQ: 2326 case O_TCPACK: 2327 case O_PROB: 2328 case O_ICMPTYPE: 2329 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2330 goto bad_size; 2331 break; 2332 2333 case O_LIMIT: 2334 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 2335 goto bad_size; 2336 break; 2337 2338 case O_LOG: 2339 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 2340 goto bad_size; 2341 2342 ((ipfw_insn_log *)cmd)->log_left = 2343 ((ipfw_insn_log *)cmd)->max_log; 2344 2345 break; 2346 2347 case O_IP_SRC_MASK: 2348 case O_IP_DST_MASK: 2349 if (cmdlen != F_INSN_SIZE(ipfw_insn_ip)) 2350 goto bad_size; 2351 if (((ipfw_insn_ip *)cmd)->mask.s_addr == 0) { 2352 printf("ipfw: opcode %d, useless rule\n", 2353 cmd->opcode); 2354 return EINVAL; 2355 } 2356 break; 2357 2358 case O_IP_SRC_SET: 2359 case O_IP_DST_SET: 2360 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 2361 printf("ipfw: invalid set size %d\n", 2362 cmd->arg1); 2363 return EINVAL; 2364 } 2365 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 2366 (cmd->arg1+31)/32 ) 2367 goto bad_size; 2368 break; 2369 2370 case O_MACADDR2: 2371 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 2372 goto bad_size; 2373 break; 2374 2375 case O_MAC_TYPE: 2376 case O_IP_SRCPORT: 2377 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 2378 if (cmdlen < 2 || cmdlen > 31) 2379 goto bad_size; 2380 break; 2381 2382 case O_RECV: 2383 case O_XMIT: 2384 case O_VIA: 2385 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 2386 goto bad_size; 2387 break; 2388 2389 case O_PIPE: 2390 case O_QUEUE: 2391 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 2392 goto bad_size; 2393 goto check_action; 2394 2395 case O_FORWARD_IP: 2396 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 2397 goto bad_size; 2398 goto check_action; 2399 2400 case O_FORWARD_MAC: /* XXX not implemented yet */ 2401 case O_CHECK_STATE: 2402 case O_COUNT: 2403 case O_ACCEPT: 2404 case O_DENY: 2405 case O_REJECT: 2406 case O_SKIPTO: 2407 case O_DIVERT: 2408 case O_TEE: 2409 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2410 goto bad_size; 2411check_action: 2412 if (have_action) { 2413 printf("ipfw: opcode %d, multiple actions" 2414 " not allowed\n", 2415 cmd->opcode); 2416 return EINVAL; 2417 } 2418 have_action = 1; 2419 if (l != cmdlen) { 2420 printf("ipfw: opcode %d, action must be" 2421 " last opcode\n", 2422 cmd->opcode); 2423 return EINVAL; 2424 } 2425 break; 2426 default: 2427 printf("ipfw: opcode %d, unknown opcode\n", 2428 cmd->opcode); 2429 return EINVAL; 2430 } 2431 } 2432 if (have_action == 0) { 2433 printf("ipfw: missing action\n"); 2434 return EINVAL; 2435 } 2436 return 0; 2437 2438bad_size: 2439 printf("ipfw: opcode %d size %d wrong\n", 2440 cmd->opcode, cmdlen); 2441 return EINVAL; 2442} 2443 2444 2445/** 2446 * {set|get}sockopt parser. 2447 */ 2448static int 2449ipfw_ctl(struct sockopt *sopt) 2450{ 2451 int error, s, rulenum; 2452 size_t size; 2453 struct ip_fw *bp , *buf, *rule; 2454 2455 static u_int32_t rule_buf[255]; /* we copy the data here */ 2456 2457 /* 2458 * Disallow modifications in really-really secure mode, but still allow 2459 * the logging counters to be reset. 2460 */ 2461 if (sopt->sopt_name == IP_FW_ADD || 2462 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 2463#if __FreeBSD_version >= 500034 2464 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 2465 if (error) 2466 return (error); 2467#else /* FreeBSD 4.x */ 2468 if (securelevel >= 3) 2469 return (EPERM); 2470#endif 2471 } 2472 2473 error = 0; 2474 2475 switch (sopt->sopt_name) { 2476 case IP_FW_GET: 2477 /* 2478 * pass up a copy of the current rules. Static rules 2479 * come first (the last of which has number IPFW_DEFAULT_RULE), 2480 * followed by a possibly empty list of dynamic rule. 2481 * The last dynamic rule has NULL in the "next" field. 2482 */ 2483 s = splimp(); 2484 size = static_len; /* size of static rules */ 2485 if (ipfw_dyn_v) /* add size of dyn.rules */ 2486 size += (dyn_count * sizeof(ipfw_dyn_rule)); 2487 2488 /* 2489 * XXX todo: if the user passes a short length just to know 2490 * how much room is needed, do not bother filling up the 2491 * buffer, just jump to the sooptcopyout. 2492 */ 2493 buf = malloc(size, M_TEMP, M_WAITOK); 2494 if (buf == 0) { 2495 splx(s); 2496 error = ENOBUFS; 2497 break; 2498 } 2499 2500 bp = buf; 2501 for (rule = layer3_chain; rule ; rule = rule->next) { 2502 int i = RULESIZE(rule); 2503 bcopy(rule, bp, i); 2504 /* 2505 * abuse 'next_rule' to store the set_disable word 2506 */ 2507 (u_int32_t)(((struct ip_fw *)bp)->next_rule) = 2508 set_disable; 2509 bp = (struct ip_fw *)((char *)bp + i); 2510 } 2511 if (ipfw_dyn_v) { 2512 int i; 2513 ipfw_dyn_rule *p, *dst, *last = NULL; 2514 2515 dst = (ipfw_dyn_rule *)bp; 2516 for (i = 0 ; i < curr_dyn_buckets ; i++ ) 2517 for ( p = ipfw_dyn_v[i] ; p != NULL ; 2518 p = p->next, dst++ ) { 2519 bcopy(p, dst, sizeof *p); 2520 (int)dst->rule = p->rule->rulenum ; 2521 /* 2522 * store a non-null value in "next". 2523 * The userland code will interpret a 2524 * NULL here as a marker 2525 * for the last dynamic rule. 2526 */ 2527 dst->next = dst ; 2528 last = dst ; 2529 dst->expire = 2530 TIME_LEQ(dst->expire, time_second) ? 2531 0 : dst->expire - time_second ; 2532 } 2533 if (last != NULL) /* mark last dynamic rule */ 2534 last->next = NULL; 2535 } 2536 splx(s); 2537 2538 error = sooptcopyout(sopt, buf, size); 2539 free(buf, M_TEMP); 2540 break; 2541 2542 case IP_FW_FLUSH: 2543 /* 2544 * Normally we cannot release the lock on each iteration. 2545 * We could do it here only because we start from the head all 2546 * the times so there is no risk of missing some entries. 2547 * On the other hand, the risk is that we end up with 2548 * a very inconsistent ruleset, so better keep the lock 2549 * around the whole cycle. 2550 * 2551 * XXX this code can be improved by resetting the head of 2552 * the list to point to the default rule, and then freeing 2553 * the old list without the need for a lock. 2554 */ 2555 2556 s = splimp(); 2557 free_chain(&layer3_chain, 0 /* keep default rule */); 2558 splx(s); 2559 break; 2560 2561 case IP_FW_ADD: 2562 rule = (struct ip_fw *)rule_buf; /* XXX do a malloc */ 2563 error = sooptcopyin(sopt, rule, sizeof(rule_buf), 2564 sizeof(struct ip_fw) ); 2565 size = sopt->sopt_valsize; 2566 if (error || (error = check_ipfw_struct(rule, size))) 2567 break; 2568 2569 error = add_rule(&layer3_chain, rule); 2570 size = RULESIZE(rule); 2571 if (!error && sopt->sopt_dir == SOPT_GET) 2572 error = sooptcopyout(sopt, rule, size); 2573 break; 2574 2575 case IP_FW_DEL: 2576 /* 2577 * IP_FW_DEL is used for deleting single rules or sets, 2578 * and (ab)used to atomically manipulate sets. Argument size 2579 * is used to distinguish between the two: 2580 * sizeof(u_int32_t) 2581 * delete single rule or set of rules, 2582 * or reassign rules (or sets) to a different set. 2583 * 2*sizeof(u_int32_t) 2584 * atomic disable/enable sets. 2585 * first u_int32_t contains sets to be disabled, 2586 * second u_int32_t contains sets to be enabled. 2587 */ 2588 error = sooptcopyin(sopt, rule_buf, 2589 2*sizeof(u_int32_t), sizeof(u_int32_t)); 2590 if (error) 2591 break; 2592 size = sopt->sopt_valsize; 2593 if (size == sizeof(u_int32_t)) /* delete or reassign */ 2594 error = del_entry(&layer3_chain, rule_buf[0]); 2595 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 2596 set_disable = 2597 (set_disable | rule_buf[0]) & ~rule_buf[1] & 2598 ~(1<<31); /* set 31 always enabled */ 2599 else 2600 error = EINVAL; 2601 break; 2602 2603 case IP_FW_ZERO: 2604 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 2605 rulenum=0; 2606 2607 if (sopt->sopt_val != 0) { 2608 error = sooptcopyin(sopt, &rulenum, 2609 sizeof(int), sizeof(int)); 2610 if (error) 2611 break; 2612 } 2613 error = zero_entry(rulenum, sopt->sopt_name == IP_FW_RESETLOG); 2614 break; 2615 2616 default: 2617 printf("ipfw_ctl invalid option %d\n", sopt->sopt_name); 2618 error = EINVAL; 2619 } 2620 2621 return (error); 2622} 2623 2624/** 2625 * dummynet needs a reference to the default rule, because rules can be 2626 * deleted while packets hold a reference to them. When this happens, 2627 * dummynet changes the reference to the default rule (it could well be a 2628 * NULL pointer, but this way we do not need to check for the special 2629 * case, plus here he have info on the default behaviour). 2630 */ 2631struct ip_fw *ip_fw_default_rule; 2632 2633/* 2634 * This procedure is only used to handle keepalives. It is invoked 2635 * every dyn_keepalive_period 2636 */ 2637static void 2638ipfw_tick(void * __unused unused) 2639{ 2640 int i; 2641 int s; 2642 ipfw_dyn_rule *q; 2643 2644 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 2645 goto done; 2646 2647 s = splimp(); 2648 for (i = 0 ; i < curr_dyn_buckets ; i++) { 2649 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 2650 if (q->dyn_type == O_LIMIT_PARENT) 2651 continue; 2652 if (q->id.proto != IPPROTO_TCP) 2653 continue; 2654 if ( (q->state & BOTH_SYN) != BOTH_SYN) 2655 continue; 2656 if (TIME_LEQ( time_second+dyn_keepalive_interval, 2657 q->expire)) 2658 continue; /* too early */ 2659 if (TIME_LEQ(q->expire, time_second)) 2660 continue; /* too late, rule expired */ 2661 2662 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 2663 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 2664 } 2665 } 2666 splx(s); 2667done: 2668 ipfw_timeout_h = timeout(ipfw_tick, NULL, dyn_keepalive_period*hz); 2669} 2670 2671static void 2672ipfw_init(void) 2673{ 2674 struct ip_fw default_rule; 2675 2676 ip_fw_chk_ptr = ipfw_chk; 2677 ip_fw_ctl_ptr = ipfw_ctl; 2678 layer3_chain = NULL; 2679 2680 bzero(&default_rule, sizeof default_rule); 2681 2682 default_rule.act_ofs = 0; 2683 default_rule.rulenum = IPFW_DEFAULT_RULE; 2684 default_rule.cmd_len = 1; 2685 default_rule.set = 31; 2686 2687 default_rule.cmd[0].len = 1; 2688 default_rule.cmd[0].opcode = 2689#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 2690 1 ? O_ACCEPT : 2691#endif 2692 O_DENY; 2693 2694 add_rule(&layer3_chain, &default_rule); 2695 2696 ip_fw_default_rule = layer3_chain; 2697 printf("ipfw2 initialized, divert %s, " 2698 "rule-based forwarding enabled, default to %s, logging ", 2699#ifdef IPDIVERT 2700 "enabled", 2701#else 2702 "disabled", 2703#endif 2704 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 2705 2706#ifdef IPFIREWALL_VERBOSE 2707 fw_verbose = 1; 2708#endif 2709#ifdef IPFIREWALL_VERBOSE_LIMIT 2710 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 2711#endif 2712 if (fw_verbose == 0) 2713 printf("disabled\n"); 2714 else if (verbose_limit == 0) 2715 printf("unlimited\n"); 2716 else 2717 printf("limited to %d packets/entry by default\n", 2718 verbose_limit); 2719 bzero(&ipfw_timeout_h, sizeof(struct callout_handle)); 2720 ipfw_timeout_h = timeout(ipfw_tick, NULL, hz); 2721} 2722 2723static int 2724ipfw_modevent(module_t mod, int type, void *unused) 2725{ 2726 int s; 2727 int err = 0; 2728 2729 switch (type) { 2730 case MOD_LOAD: 2731 s = splimp(); 2732 if (IPFW_LOADED) { 2733 splx(s); 2734 printf("IP firewall already loaded\n"); 2735 err = EEXIST; 2736 } else { 2737 ipfw_init(); 2738 splx(s); 2739 } 2740 break; 2741 2742 case MOD_UNLOAD: 2743#if !defined(KLD_MODULE) 2744 printf("ipfw statically compiled, cannot unload\n"); 2745 err = EBUSY; 2746#else 2747 s = splimp(); 2748 untimeout(ipfw_tick, NULL, ipfw_timeout_h); 2749 ip_fw_chk_ptr = NULL; 2750 ip_fw_ctl_ptr = NULL; 2751 free_chain(&layer3_chain, 1 /* kill default rule */); 2752 splx(s); 2753 printf("IP firewall unloaded\n"); 2754#endif 2755 break; 2756 default: 2757 break; 2758 } 2759 return err; 2760} 2761 2762static moduledata_t ipfwmod = { 2763 "ipfw", 2764 ipfw_modevent, 2765 0 2766}; 2767DECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PSEUDO, SI_ORDER_ANY); 2768MODULE_VERSION(ipfw, 1); 2769#endif /* IPFW2 */ 2770