ip_fw2.c revision 102086
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 102086 2002-08-19 04:45:01Z 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, CTLFLAG_RW, 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,CTLFLAG_RW, 120 &fw_one_pass, 0, 121 "Only do a single pass through ipfw when using dummynet(4)"); 122SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 123 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 124SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, CTLFLAG_RW, 125 &fw_verbose, 0, "Log matches to ipfw rules"); 126SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 127 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 128 129/* 130 * Description of dynamic rules. 131 * 132 * Dynamic rules are stored in lists accessed through a hash table 133 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 134 * be modified through the sysctl variable dyn_buckets which is 135 * updated when the table becomes empty. 136 * 137 * XXX currently there is only one list, ipfw_dyn. 138 * 139 * When a packet is received, its address fields are first masked 140 * with the mask defined for the rule, then hashed, then matched 141 * against the entries in the corresponding list. 142 * Dynamic rules can be used for different purposes: 143 * + stateful rules; 144 * + enforcing limits on the number of sessions; 145 * + in-kernel NAT (not implemented yet) 146 * 147 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 148 * measured in seconds and depending on the flags. 149 * 150 * The total number of dynamic rules is stored in dyn_count. 151 * The max number of dynamic rules is dyn_max. When we reach 152 * the maximum number of rules we do not create anymore. This is 153 * done to avoid consuming too much memory, but also too much 154 * time when searching on each packet (ideally, we should try instead 155 * to put a limit on the length of the list on each bucket...). 156 * 157 * Each dynamic rule holds a pointer to the parent ipfw rule so 158 * we know what action to perform. Dynamic rules are removed when 159 * the parent rule is deleted. XXX we should make them survive. 160 * 161 * There are some limitations with dynamic rules -- we do not 162 * obey the 'randomized match', and we do not do multiple 163 * passes through the firewall. XXX check the latter!!! 164 */ 165static ipfw_dyn_rule **ipfw_dyn_v = NULL; 166static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 167static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 168 169/* 170 * Timeouts for various events in handing dynamic rules. 171 */ 172static u_int32_t dyn_ack_lifetime = 300; 173static u_int32_t dyn_syn_lifetime = 20; 174static u_int32_t dyn_fin_lifetime = 1; 175static u_int32_t dyn_rst_lifetime = 1; 176static u_int32_t dyn_udp_lifetime = 10; 177static u_int32_t dyn_short_lifetime = 5; 178 179/* 180 * Keepalives are sent if dyn_keepalive is set. They are sent every 181 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 182 * seconds of lifetime of a rule. 183 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 184 * than dyn_keepalive_period. 185 */ 186 187static u_int32_t dyn_keepalive_interval = 20; 188static u_int32_t dyn_keepalive_period = 5; 189static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 190 191static u_int32_t static_count; /* # of static rules */ 192static u_int32_t static_len; /* size in bytes of static rules */ 193static u_int32_t dyn_count; /* # of dynamic rules */ 194static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 195 196SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 197 &dyn_buckets, 0, "Number of dyn. buckets"); 198SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 199 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 200SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 201 &dyn_count, 0, "Number of dyn. rules"); 202SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 203 &dyn_max, 0, "Max number of dyn. rules"); 204SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 205 &static_count, 0, "Number of static rules"); 206SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 207 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 208SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 209 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 210SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 211 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 212SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 213 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 214SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 215 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 216SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 217 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 218SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 219 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 220 221#endif /* SYSCTL_NODE */ 222 223 224static ip_fw_chk_t ipfw_chk; 225 226ip_dn_ruledel_t *ip_dn_ruledel_ptr = NULL; /* hook into dummynet */ 227 228/* 229 * This macro maps an ip pointer into a layer3 header pointer of type T 230 */ 231#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 232 233static __inline int 234icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 235{ 236 int type = L3HDR(struct icmp,ip)->icmp_type; 237 238 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 239} 240 241#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 242 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 243 244static int 245is_icmp_query(struct ip *ip) 246{ 247 int type = L3HDR(struct icmp, ip)->icmp_type; 248 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 249} 250#undef TT 251 252/* 253 * The following checks use two arrays of 8 or 16 bits to store the 254 * bits that we want set or clear, respectively. They are in the 255 * low and high half of cmd->arg1 or cmd->d[0]. 256 * 257 * We scan options and store the bits we find set. We succeed if 258 * 259 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 260 * 261 * The code is sometimes optimized not to store additional variables. 262 */ 263 264static int 265flags_match(ipfw_insn *cmd, u_int8_t bits) 266{ 267 u_char want_clear; 268 bits = ~bits; 269 270 if ( ((cmd->arg1 & 0xff) & bits) != 0) 271 return 0; /* some bits we want set were clear */ 272 want_clear = (cmd->arg1 >> 8) & 0xff; 273 if ( (want_clear & bits) != want_clear) 274 return 0; /* some bits we want clear were set */ 275 return 1; 276} 277 278static int 279ipopts_match(struct ip *ip, ipfw_insn *cmd) 280{ 281 int optlen, bits = 0; 282 u_char *cp = (u_char *)(ip + 1); 283 int x = (ip->ip_hl << 2) - sizeof (struct ip); 284 285 for (; x > 0; x -= optlen, cp += optlen) { 286 int opt = cp[IPOPT_OPTVAL]; 287 288 if (opt == IPOPT_EOL) 289 break; 290 if (opt == IPOPT_NOP) 291 optlen = 1; 292 else { 293 optlen = cp[IPOPT_OLEN]; 294 if (optlen <= 0 || optlen > x) 295 return 0; /* invalid or truncated */ 296 } 297 switch (opt) { 298 299 default: 300 break; 301 302 case IPOPT_LSRR: 303 bits |= IP_FW_IPOPT_LSRR; 304 break; 305 306 case IPOPT_SSRR: 307 bits |= IP_FW_IPOPT_SSRR; 308 break; 309 310 case IPOPT_RR: 311 bits |= IP_FW_IPOPT_RR; 312 break; 313 314 case IPOPT_TS: 315 bits |= IP_FW_IPOPT_TS; 316 break; 317 } 318 } 319 return (flags_match(cmd, bits)); 320} 321 322static int 323tcpopts_match(struct ip *ip, ipfw_insn *cmd) 324{ 325 int optlen, bits = 0; 326 struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 327 u_char *cp = (u_char *)(tcp + 1); 328 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 329 330 for (; x > 0; x -= optlen, cp += optlen) { 331 int opt = cp[0]; 332 if (opt == TCPOPT_EOL) 333 break; 334 if (opt == TCPOPT_NOP) 335 optlen = 1; 336 else { 337 optlen = cp[1]; 338 if (optlen <= 0) 339 break; 340 } 341 342 switch (opt) { 343 344 default: 345 break; 346 347 case TCPOPT_MAXSEG: 348 bits |= IP_FW_TCPOPT_MSS; 349 break; 350 351 case TCPOPT_WINDOW: 352 bits |= IP_FW_TCPOPT_WINDOW; 353 break; 354 355 case TCPOPT_SACK_PERMITTED: 356 case TCPOPT_SACK: 357 bits |= IP_FW_TCPOPT_SACK; 358 break; 359 360 case TCPOPT_TIMESTAMP: 361 bits |= IP_FW_TCPOPT_TS; 362 break; 363 364 case TCPOPT_CC: 365 case TCPOPT_CCNEW: 366 case TCPOPT_CCECHO: 367 bits |= IP_FW_TCPOPT_CC; 368 break; 369 } 370 } 371 return (flags_match(cmd, bits)); 372} 373 374static int 375iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 376{ 377 if (ifp == NULL) /* no iface with this packet, match fails */ 378 return 0; 379 /* Check by name or by IP address */ 380 if (cmd->name[0] != '\0') { /* match by name */ 381 /* Check unit number (-1 is wildcard) */ 382 if (cmd->p.unit != -1 && cmd->p.unit != ifp->if_unit) 383 return(0); 384 /* Check name */ 385 if (!strncmp(ifp->if_name, cmd->name, IFNAMSIZ)) 386 return(1); 387 } else { 388 struct ifaddr *ia; 389 390 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 391 if (ia->ifa_addr == NULL) 392 continue; 393 if (ia->ifa_addr->sa_family != AF_INET) 394 continue; 395 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 396 (ia->ifa_addr))->sin_addr.s_addr) 397 return(1); /* match */ 398 } 399 } 400 return(0); /* no match, fail ... */ 401} 402 403static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 404 405#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 406#define SNP(buf) buf, sizeof(buf) 407 408/* 409 * We enter here when we have a rule with O_LOG. 410 * XXX this function alone takes about 2Kbytes of code! 411 */ 412static void 413ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 414 struct mbuf *m, struct ifnet *oif) 415{ 416 char *action; 417 int limit_reached = 0; 418 char action2[40], proto[48], fragment[28]; 419 420 fragment[0] = '\0'; 421 proto[0] = '\0'; 422 423 if (f == NULL) { /* bogus pkt */ 424 if (verbose_limit != 0 && norule_counter >= verbose_limit) 425 return; 426 norule_counter++; 427 if (norule_counter == verbose_limit) 428 limit_reached = verbose_limit; 429 action = "Refuse"; 430 } else { /* O_LOG is the first action, find the real one */ 431 ipfw_insn *cmd = ACTION_PTR(f); 432 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 433 434 if (l->max_log != 0 && l->log_left == 0) 435 return; 436 l->log_left--; 437 if (l->log_left == 0) 438 limit_reached = l->max_log; 439 cmd += F_LEN(cmd); /* point to first action */ 440 if (cmd->opcode == O_PROB) 441 cmd += F_LEN(cmd); 442 443 action = action2; 444 switch (cmd->opcode) { 445 case O_DENY: 446 action = "Deny"; 447 break; 448 449 case O_REJECT: 450 if (cmd->arg1==ICMP_REJECT_RST) 451 action = "Reset"; 452 else if (cmd->arg1==ICMP_UNREACH_HOST) 453 action = "Reject"; 454 else 455 snprintf(SNPARGS(action2, 0), "Unreach %d", 456 cmd->arg1); 457 break; 458 459 case O_ACCEPT: 460 action = "Accept"; 461 break; 462 case O_COUNT: 463 action = "Count"; 464 break; 465 case O_DIVERT: 466 snprintf(SNPARGS(action2, 0), "Divert %d", 467 cmd->arg1); 468 break; 469 case O_TEE: 470 snprintf(SNPARGS(action2, 0), "Tee %d", 471 cmd->arg1); 472 break; 473 case O_SKIPTO: 474 snprintf(SNPARGS(action2, 0), "SkipTo %d", 475 cmd->arg1); 476 break; 477 case O_PIPE: 478 snprintf(SNPARGS(action2, 0), "Pipe %d", 479 cmd->arg1); 480 break; 481 case O_QUEUE: 482 snprintf(SNPARGS(action2, 0), "Queue %d", 483 cmd->arg1); 484 break; 485 case O_FORWARD_IP: { 486 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 487 int len; 488 489 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 490 inet_ntoa(sa->sa.sin_addr)); 491 if (sa->sa.sin_port) 492 snprintf(SNPARGS(action2, len), ":%d", 493 ntohs(sa->sa.sin_port)); 494 } 495 break; 496 default: 497 action = "UNKNOWN"; 498 break; 499 } 500 } 501 502 if (hlen == 0) { /* non-ip */ 503 snprintf(SNPARGS(proto, 0), "MAC"); 504 } else { 505 struct ip *ip = mtod(m, struct ip *); 506 /* these three are all aliases to the same thing */ 507 struct icmp *const icmp = L3HDR(struct icmp, ip); 508 struct tcphdr *const tcp = (struct tcphdr *)icmp; 509 struct udphdr *const udp = (struct udphdr *)icmp; 510 511 int ip_off, offset, ip_len; 512 513 int len; 514 515 if (eh != NULL) { /* layer 2 packets are as on the wire */ 516 ip_off = ntohs(ip->ip_off); 517 ip_len = ntohs(ip->ip_len); 518 } else { 519 ip_off = ip->ip_off; 520 ip_len = ip->ip_len; 521 } 522 offset = ip_off & IP_OFFMASK; 523 switch (ip->ip_p) { 524 case IPPROTO_TCP: 525 len = snprintf(SNPARGS(proto, 0), "TCP %s", 526 inet_ntoa(ip->ip_src)); 527 if (offset == 0) 528 snprintf(SNPARGS(proto, len), ":%d %s:%d", 529 ntohs(tcp->th_sport), 530 inet_ntoa(ip->ip_dst), 531 ntohs(tcp->th_dport)); 532 else 533 snprintf(SNPARGS(proto, len), " %s", 534 inet_ntoa(ip->ip_dst)); 535 break; 536 537 case IPPROTO_UDP: 538 len = snprintf(SNPARGS(proto, 0), "UDP %s", 539 inet_ntoa(ip->ip_src)); 540 if (offset == 0) 541 snprintf(SNPARGS(proto, len), ":%d %s:%d", 542 ntohs(udp->uh_sport), 543 inet_ntoa(ip->ip_dst), 544 ntohs(udp->uh_dport)); 545 else 546 snprintf(SNPARGS(proto, len), " %s", 547 inet_ntoa(ip->ip_dst)); 548 break; 549 550 case IPPROTO_ICMP: 551 if (offset == 0) 552 len = snprintf(SNPARGS(proto, 0), 553 "ICMP:%u.%u ", 554 icmp->icmp_type, icmp->icmp_code); 555 else 556 len = snprintf(SNPARGS(proto, 0), "ICMP "); 557 len += snprintf(SNPARGS(proto, len), "%s", 558 inet_ntoa(ip->ip_src)); 559 snprintf(SNPARGS(proto, len), " %s", 560 inet_ntoa(ip->ip_dst)); 561 break; 562 563 default: 564 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 565 inet_ntoa(ip->ip_src)); 566 snprintf(SNPARGS(proto, len), " %s", 567 inet_ntoa(ip->ip_dst)); 568 break; 569 } 570 571 if (ip_off & (IP_MF | IP_OFFMASK)) 572 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 573 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 574 offset << 3, 575 (ip_off & IP_MF) ? "+" : ""); 576 } 577 if (oif || m->m_pkthdr.rcvif) 578 log(LOG_SECURITY | LOG_INFO, 579 "ipfw: %d %s %s %s via %s%d%s\n", 580 f ? f->rulenum : -1, 581 action, proto, oif ? "out" : "in", 582 oif ? oif->if_name : m->m_pkthdr.rcvif->if_name, 583 oif ? oif->if_unit : m->m_pkthdr.rcvif->if_unit, 584 fragment); 585 else 586 log(LOG_SECURITY | LOG_INFO, 587 "ipfw: %d %s %s [no if info]%s\n", 588 f ? f->rulenum : -1, 589 action, proto, fragment); 590 if (limit_reached) 591 log(LOG_SECURITY | LOG_NOTICE, 592 "ipfw: limit %d reached on entry %d\n", 593 limit_reached, f ? f->rulenum : -1); 594} 595 596/* 597 * IMPORTANT: the hash function for dynamic rules must be commutative 598 * in source and destination (ip,port), because rules are bidirectional 599 * and we want to find both in the same bucket. 600 */ 601static __inline int 602hash_packet(struct ipfw_flow_id *id) 603{ 604 u_int32_t i; 605 606 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 607 i &= (curr_dyn_buckets - 1); 608 return i; 609} 610 611/** 612 * unlink a dynamic rule from a chain. prev is a pointer to 613 * the previous one, q is a pointer to the rule to delete, 614 * head is a pointer to the head of the queue. 615 * Modifies q and potentially also head. 616 */ 617#define UNLINK_DYN_RULE(prev, head, q) { \ 618 ipfw_dyn_rule *old_q = q; \ 619 \ 620 /* remove a refcount to the parent */ \ 621 if (q->dyn_type == O_LIMIT) \ 622 q->parent->count--; \ 623 DEB(printf("-- unlink entry 0x%08x %d -> 0x%08x %d, %d left\n", \ 624 (q->id.src_ip), (q->id.src_port), \ 625 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 626 if (prev != NULL) \ 627 prev->next = q = q->next; \ 628 else \ 629 head = q = q->next; \ 630 dyn_count--; \ 631 free(old_q, M_IPFW); } 632 633#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 634 635/** 636 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 637 * 638 * If keep_me == NULL, rules are deleted even if not expired, 639 * otherwise only expired rules are removed. 640 * 641 * The value of the second parameter is also used to point to identify 642 * a rule we absolutely do not want to remove (e.g. because we are 643 * holding a reference to it -- this is the case with O_LIMIT_PARENT 644 * rules). The pointer is only used for comparison, so any non-null 645 * value will do. 646 */ 647static void 648remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 649{ 650 static u_int32_t last_remove = 0; 651 652#define FORCE (keep_me == NULL) 653 654 ipfw_dyn_rule *prev, *q; 655 int i, pass = 0, max_pass = 0; 656 657 if (ipfw_dyn_v == NULL || dyn_count == 0) 658 return; 659 /* do not expire more than once per second, it is useless */ 660 if (!FORCE && last_remove == time_second) 661 return; 662 last_remove = time_second; 663 664 /* 665 * because O_LIMIT refer to parent rules, during the first pass only 666 * remove child and mark any pending LIMIT_PARENT, and remove 667 * them in a second pass. 668 */ 669next_pass: 670 for (i = 0 ; i < curr_dyn_buckets ; i++) { 671 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 672 /* 673 * Logic can become complex here, so we split tests. 674 */ 675 if (q == keep_me) 676 goto next; 677 if (rule != NULL && rule != q->rule) 678 goto next; /* not the one we are looking for */ 679 if (q->dyn_type == O_LIMIT_PARENT) { 680 /* 681 * handle parent in the second pass, 682 * record we need one. 683 */ 684 max_pass = 1; 685 if (pass == 0) 686 goto next; 687 if (FORCE && q->count != 0 ) { 688 /* XXX should not happen! */ 689 printf( "OUCH! cannot remove rule," 690 " count %d\n", q->count); 691 } 692 } else { 693 if (!FORCE && 694 !TIME_LEQ( q->expire, time_second )) 695 goto next; 696 } 697 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 698 continue; 699next: 700 prev=q; 701 q=q->next; 702 } 703 } 704 if (pass++ < max_pass) 705 goto next_pass; 706} 707 708 709/** 710 * lookup a dynamic rule. 711 */ 712static ipfw_dyn_rule * 713lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 714 struct tcphdr *tcp) 715{ 716 /* 717 * stateful ipfw extensions. 718 * Lookup into dynamic session queue 719 */ 720#define MATCH_REVERSE 0 721#define MATCH_FORWARD 1 722#define MATCH_NONE 2 723#define MATCH_UNKNOWN 3 724 int i, dir = MATCH_NONE; 725 ipfw_dyn_rule *prev, *q=NULL; 726 727 if (ipfw_dyn_v == NULL) 728 goto done; /* not found */ 729 i = hash_packet( pkt ); 730 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 731 if (q->dyn_type == O_LIMIT_PARENT) 732 goto next; 733 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 734 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 735 continue; 736 } 737 if ( pkt->proto == q->id.proto) { 738 if (pkt->src_ip == q->id.src_ip && 739 pkt->dst_ip == q->id.dst_ip && 740 pkt->src_port == q->id.src_port && 741 pkt->dst_port == q->id.dst_port ) { 742 dir = MATCH_FORWARD; 743 break; 744 } 745 if (pkt->src_ip == q->id.dst_ip && 746 pkt->dst_ip == q->id.src_ip && 747 pkt->src_port == q->id.dst_port && 748 pkt->dst_port == q->id.src_port ) { 749 dir = MATCH_REVERSE; 750 break; 751 } 752 } 753next: 754 prev = q; 755 q = q->next; 756 } 757 if (q == NULL) 758 goto done; /* q = NULL, not found */ 759 760 if ( prev != NULL) { /* found and not in front */ 761 prev->next = q->next; 762 q->next = ipfw_dyn_v[i]; 763 ipfw_dyn_v[i] = q; 764 } 765 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 766 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 767 768#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 769#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 770 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 771 switch (q->state) { 772 case TH_SYN: /* opening */ 773 q->expire = time_second + dyn_syn_lifetime; 774 break; 775 776 case BOTH_SYN: /* move to established */ 777 case BOTH_SYN | TH_FIN : /* one side tries to close */ 778 case BOTH_SYN | (TH_FIN << 8) : 779 if (tcp) { 780#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 781 u_int32_t ack = ntohl(tcp->th_ack); 782 if (dir == MATCH_FORWARD) { 783 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 784 q->ack_fwd = ack; 785 else { /* ignore out-of-sequence */ 786 break; 787 } 788 } else { 789 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 790 q->ack_rev = ack; 791 else { /* ignore out-of-sequence */ 792 break; 793 } 794 } 795 } 796 q->expire = time_second + dyn_ack_lifetime; 797 break; 798 799 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 800 if (dyn_fin_lifetime >= dyn_keepalive_period) 801 dyn_fin_lifetime = dyn_keepalive_period - 1; 802 q->expire = time_second + dyn_fin_lifetime; 803 break; 804 805 default: 806#if 0 807 /* 808 * reset or some invalid combination, but can also 809 * occur if we use keep-state the wrong way. 810 */ 811 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 812 printf("invalid state: 0x%x\n", q->state); 813#endif 814 if (dyn_rst_lifetime >= dyn_keepalive_period) 815 dyn_rst_lifetime = dyn_keepalive_period - 1; 816 q->expire = time_second + dyn_rst_lifetime; 817 break; 818 } 819 } else if (pkt->proto == IPPROTO_UDP) { 820 q->expire = time_second + dyn_udp_lifetime; 821 } else { 822 /* other protocols */ 823 q->expire = time_second + dyn_short_lifetime; 824 } 825done: 826 if (match_direction) 827 *match_direction = dir; 828 return q; 829} 830 831static void 832realloc_dynamic_table(void) 833{ 834 /* 835 * Try reallocation, make sure we have a power of 2 and do 836 * not allow more than 64k entries. In case of overflow, 837 * default to 1024. 838 */ 839 840 if (dyn_buckets > 65536) 841 dyn_buckets = 1024; 842 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 843 dyn_buckets = curr_dyn_buckets; /* reset */ 844 return; 845 } 846 curr_dyn_buckets = dyn_buckets; 847 if (ipfw_dyn_v != NULL) 848 free(ipfw_dyn_v, M_IPFW); 849 for (;;) { 850 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 851 M_IPFW, M_DONTWAIT | M_ZERO); 852 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 853 break; 854 curr_dyn_buckets /= 2; 855 } 856} 857 858/** 859 * Install state of type 'type' for a dynamic session. 860 * The hash table contains two type of rules: 861 * - regular rules (O_KEEP_STATE) 862 * - rules for sessions with limited number of sess per user 863 * (O_LIMIT). When they are created, the parent is 864 * increased by 1, and decreased on delete. In this case, 865 * the third parameter is the parent rule and not the chain. 866 * - "parent" rules for the above (O_LIMIT_PARENT). 867 */ 868static ipfw_dyn_rule * 869add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 870{ 871 ipfw_dyn_rule *r; 872 int i; 873 874 if (ipfw_dyn_v == NULL || 875 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 876 realloc_dynamic_table(); 877 if (ipfw_dyn_v == NULL) 878 return NULL; /* failed ! */ 879 } 880 i = hash_packet(id); 881 882 r = malloc(sizeof *r, M_IPFW, M_DONTWAIT | M_ZERO); 883 if (r == NULL) { 884 printf ("sorry cannot allocate state\n"); 885 return NULL; 886 } 887 888 /* increase refcount on parent, and set pointer */ 889 if (dyn_type == O_LIMIT) { 890 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 891 if ( parent->dyn_type != O_LIMIT_PARENT) 892 panic("invalid parent"); 893 parent->count++; 894 r->parent = parent; 895 rule = parent->rule; 896 } 897 898 r->id = *id; 899 r->expire = time_second + dyn_syn_lifetime; 900 r->rule = rule; 901 r->dyn_type = dyn_type; 902 r->pcnt = r->bcnt = 0; 903 r->count = 0; 904 905 r->bucket = i; 906 r->next = ipfw_dyn_v[i]; 907 ipfw_dyn_v[i] = r; 908 dyn_count++; 909 DEB(printf("-- add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 910 dyn_type, 911 (r->id.src_ip), (r->id.src_port), 912 (r->id.dst_ip), (r->id.dst_port), 913 dyn_count ); ) 914 return r; 915} 916 917/** 918 * lookup dynamic parent rule using pkt and rule as search keys. 919 * If the lookup fails, then install one. 920 */ 921static ipfw_dyn_rule * 922lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 923{ 924 ipfw_dyn_rule *q; 925 int i; 926 927 if (ipfw_dyn_v) { 928 i = hash_packet( pkt ); 929 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 930 if (q->dyn_type == O_LIMIT_PARENT && 931 rule== q->rule && 932 pkt->proto == q->id.proto && 933 pkt->src_ip == q->id.src_ip && 934 pkt->dst_ip == q->id.dst_ip && 935 pkt->src_port == q->id.src_port && 936 pkt->dst_port == q->id.dst_port) { 937 q->expire = time_second + dyn_short_lifetime; 938 DEB(printf("lookup_dyn_parent found 0x%p\n",q);) 939 return q; 940 } 941 } 942 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 943} 944 945/** 946 * Install dynamic state for rule type cmd->o.opcode 947 * 948 * Returns 1 (failure) if state is not installed because of errors or because 949 * session limitations are enforced. 950 */ 951static int 952install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 953 struct ip_fw_args *args) 954{ 955 static int last_log; 956 957 ipfw_dyn_rule *q; 958 959 DEB(printf("-- install state type %d 0x%08x %u -> 0x%08x %u\n", 960 cmd->o.opcode, 961 (args->f_id.src_ip), (args->f_id.src_port), 962 (args->f_id.dst_ip), (args->f_id.dst_port) );) 963 964 q = lookup_dyn_rule(&args->f_id, NULL, NULL); 965 966 if (q != NULL) { /* should never occur */ 967 if (last_log != time_second) { 968 last_log = time_second; 969 printf(" install_state: entry already present, done\n"); 970 } 971 return 0; 972 } 973 974 if (dyn_count >= dyn_max) 975 /* 976 * Run out of slots, try to remove any expired rule. 977 */ 978 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 979 980 if (dyn_count >= dyn_max) { 981 if (last_log != time_second) { 982 last_log = time_second; 983 printf("install_state: Too many dynamic rules\n"); 984 } 985 return 1; /* cannot install, notify caller */ 986 } 987 988 switch (cmd->o.opcode) { 989 case O_KEEP_STATE: /* bidir rule */ 990 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 991 break; 992 993 case O_LIMIT: /* limit number of sessions */ 994 { 995 u_int16_t limit_mask = cmd->limit_mask; 996 struct ipfw_flow_id id; 997 ipfw_dyn_rule *parent; 998 999 DEB(printf("installing dyn-limit rule %d\n", cmd->conn_limit);) 1000 1001 id.dst_ip = id.src_ip = 0; 1002 id.dst_port = id.src_port = 0; 1003 id.proto = args->f_id.proto; 1004 1005 if (limit_mask & DYN_SRC_ADDR) 1006 id.src_ip = args->f_id.src_ip; 1007 if (limit_mask & DYN_DST_ADDR) 1008 id.dst_ip = args->f_id.dst_ip; 1009 if (limit_mask & DYN_SRC_PORT) 1010 id.src_port = args->f_id.src_port; 1011 if (limit_mask & DYN_DST_PORT) 1012 id.dst_port = args->f_id.dst_port; 1013 parent = lookup_dyn_parent(&id, rule); 1014 if (parent == NULL) { 1015 printf("add parent failed\n"); 1016 return 1; 1017 } 1018 if (parent->count >= cmd->conn_limit) { 1019 /* 1020 * See if we can remove some expired rule. 1021 */ 1022 remove_dyn_rule(rule, parent); 1023 if (parent->count >= cmd->conn_limit) { 1024 if (fw_verbose && last_log != time_second) { 1025 last_log = time_second; 1026 printf( 1027 "drop session, too many entries\n"); 1028 } 1029 return 1; 1030 } 1031 } 1032 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1033 } 1034 break; 1035 default: 1036 printf("unknown dynamic rule type %u\n", cmd->o.opcode); 1037 return 1; 1038 } 1039 lookup_dyn_rule(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1040 return 0; 1041} 1042 1043/* 1044 * Transmit a TCP packet, containing either a RST or a keepalive. 1045 * When flags & TH_RST, we are sending a RST packet, because of a 1046 * "reset" action matched the packet. 1047 * Otherwise we are sending a keepalive, and flags & TH_ 1048 */ 1049static void 1050send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1051{ 1052 struct mbuf *m; 1053 struct ip *ip; 1054 struct tcphdr *tcp; 1055 struct route sro; /* fake route */ 1056 1057 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1058 if (m == 0) 1059 return; 1060 m->m_pkthdr.rcvif = (struct ifnet *)0; 1061 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1062 m->m_data += max_linkhdr; 1063 1064 ip = mtod(m, struct ip *); 1065 bzero(ip, m->m_len); 1066 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1067 ip->ip_p = IPPROTO_TCP; 1068 tcp->th_off = 5; 1069 /* 1070 * Assume we are sending a RST (or a keepalive in the reverse 1071 * direction), swap src and destination addresses and ports. 1072 */ 1073 ip->ip_src.s_addr = htonl(id->dst_ip); 1074 ip->ip_dst.s_addr = htonl(id->src_ip); 1075 tcp->th_sport = htons(id->dst_port); 1076 tcp->th_dport = htons(id->src_port); 1077 if (flags & TH_RST) { /* we are sending a RST */ 1078 if (flags & TH_ACK) { 1079 tcp->th_seq = htonl(ack); 1080 tcp->th_ack = htonl(0); 1081 tcp->th_flags = TH_RST; 1082 } else { 1083 if (flags & TH_SYN) 1084 seq++; 1085 tcp->th_seq = htonl(0); 1086 tcp->th_ack = htonl(seq); 1087 tcp->th_flags = TH_RST | TH_ACK; 1088 } 1089 } else { 1090 /* 1091 * We are sending a keepalive. flags & TH_SYN determines 1092 * the direction, forward if set, reverse if clear. 1093 * NOTE: seq and ack are always assumed to be correct 1094 * as set by the caller. This may be confusing... 1095 */ 1096 if (flags & TH_SYN) { 1097 /* 1098 * we have to rewrite the correct addresses! 1099 */ 1100 ip->ip_dst.s_addr = htonl(id->dst_ip); 1101 ip->ip_src.s_addr = htonl(id->src_ip); 1102 tcp->th_dport = htons(id->dst_port); 1103 tcp->th_sport = htons(id->src_port); 1104 } 1105 tcp->th_seq = htonl(seq); 1106 tcp->th_ack = htonl(ack); 1107 tcp->th_flags = TH_ACK; 1108 } 1109 /* 1110 * set ip_len to the payload size so we can compute 1111 * the tcp checksum on the pseudoheader 1112 * XXX check this, could save a couple of words ? 1113 */ 1114 ip->ip_len = htons(sizeof(struct tcphdr)); 1115 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1116 /* 1117 * now fill fields left out earlier 1118 */ 1119 ip->ip_ttl = ip_defttl; 1120 ip->ip_len = m->m_pkthdr.len; 1121 bzero (&sro, sizeof (sro)); 1122 ip_rtaddr(ip->ip_dst, &sro); 1123 m->m_flags |= M_SKIP_FIREWALL; 1124 ip_output(m, NULL, &sro, 0, NULL); 1125 if (sro.ro_rt) 1126 RTFREE(sro.ro_rt); 1127} 1128 1129/* 1130 * sends a reject message, consuming the mbuf passed as an argument. 1131 */ 1132static void 1133send_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 1134{ 1135 1136 if (code != ICMP_REJECT_RST) /* Send an ICMP unreach */ 1137 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1138 else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1139 struct tcphdr *const tcp = 1140 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1141 if ( (tcp->th_flags & TH_RST) == 0) 1142 send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1143 ntohl(tcp->th_ack), 1144 tcp->th_flags | TH_RST); 1145 m_freem(args->m); 1146 } else 1147 m_freem(args->m); 1148 args->m = NULL; 1149} 1150 1151/** 1152 * 1153 * Given an ip_fw *, lookup_next_rule will return a pointer 1154 * to the next rule, which can be either the jump 1155 * target (for skipto instructions) or the next one in the list (in 1156 * all other cases including a missing jump target). 1157 * The result is also written in the "next_rule" field of the rule. 1158 * Backward jumps are not allowed, so start looking from the next 1159 * rule... 1160 * 1161 * This never returns NULL -- in case we do not have an exact match, 1162 * the next rule is returned. When the ruleset is changed, 1163 * pointers are flushed so we are always correct. 1164 */ 1165 1166static struct ip_fw * 1167lookup_next_rule(struct ip_fw *me) 1168{ 1169 struct ip_fw *rule = NULL; 1170 ipfw_insn *cmd; 1171 1172 /* look for action, in case it is a skipto */ 1173 cmd = ACTION_PTR(me); 1174 if ( cmd->opcode == O_SKIPTO ) 1175 for (rule = me->next; rule ; rule = rule->next) 1176 if (rule->rulenum >= cmd->arg1) 1177 break; 1178 if (rule == NULL) /* failure or not a skipto */ 1179 rule = me->next; 1180 me->next_rule = rule; 1181 return rule; 1182} 1183 1184/* 1185 * The main check routine for the firewall. 1186 * 1187 * All arguments are in args so we can modify them and return them 1188 * back to the caller. 1189 * 1190 * Parameters: 1191 * 1192 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1193 * Starts with the IP header. 1194 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1195 * args->oif Outgoing interface, or NULL if packet is incoming. 1196 * The incoming interface is in the mbuf. (in) 1197 * args->divert_rule (in/out) 1198 * Skip up to the first rule past this rule number; 1199 * upon return, non-zero port number for divert or tee. 1200 * 1201 * args->rule Pointer to the last matching rule (in/out) 1202 * args->next_hop Socket we are forwarding to (out). 1203 * args->f_id Addresses grabbed from the packet (out) 1204 * 1205 * Return value: 1206 * 1207 * IP_FW_PORT_DENY_FLAG the packet must be dropped. 1208 * 0 The packet is to be accepted and routed normally OR 1209 * the packet was denied/rejected and has been dropped; 1210 * in the latter case, *m is equal to NULL upon return. 1211 * port Divert the packet to port, with these caveats: 1212 * 1213 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead 1214 * of diverting it (ie, 'ipfw tee'). 1215 * 1216 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower 1217 * 16 bits as a dummynet pipe number instead of diverting 1218 */ 1219 1220static int 1221ipfw_chk(struct ip_fw_args *args) 1222{ 1223 /* 1224 * Local variables hold state during the processing of a packet. 1225 * 1226 * IMPORTANT NOTE: to speed up the processing of rules, there 1227 * are some assumption on the values of the variables, which 1228 * are documented here. Should you change them, please check 1229 * the implementation of the various instructions to make sure 1230 * that they still work. 1231 * 1232 * args->eh The MAC header. It is non-null for a layer2 1233 * packet, it is NULL for a layer-3 packet. 1234 * 1235 * m | args->m Pointer to the mbuf, as received from the caller. 1236 * It may change if ipfw_chk() does an m_pullup, or if it 1237 * consumes the packet because it calls send_reject(). 1238 * XXX This has to change, so that ipfw_chk() never modifies 1239 * or consumes the buffer. 1240 * ip is simply an alias of the value of m, and it is kept 1241 * in sync with it (the packet is supposed to start with 1242 * the ip header). 1243 */ 1244 struct mbuf *m = args->m; 1245 struct ip *ip = mtod(m, struct ip *); 1246 1247 /* 1248 * oif | args->oif If NULL, ipfw_chk has been called on the 1249 * inbound path (ether_input, bdg_forward, ip_input). 1250 * If non-NULL, ipfw_chk has been called on the outbound path 1251 * (ether_output, ip_output). 1252 */ 1253 struct ifnet *oif = args->oif; 1254 1255 struct ip_fw *f = NULL; /* matching rule */ 1256 int retval = 0; 1257 1258 /* 1259 * hlen The length of the IPv4 header. 1260 * hlen >0 means we have an IPv4 packet. 1261 */ 1262 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1263 1264 /* 1265 * offset The offset of a fragment. offset != 0 means that 1266 * we have a fragment at this offset of an IPv4 packet. 1267 * offset == 0 means that (if this is an IPv4 packet) 1268 * this is the first or only fragment. 1269 */ 1270 u_short offset = 0; 1271 1272 /* 1273 * Local copies of addresses. They are only valid if we have 1274 * an IP packet. 1275 * 1276 * proto The protocol. Set to 0 for non-ip packets, 1277 * or to the protocol read from the packet otherwise. 1278 * proto != 0 means that we have an IPv4 packet. 1279 * 1280 * src_port, dst_port port numbers, in HOST format. Only 1281 * valid for TCP and UDP packets. 1282 * 1283 * src_ip, dst_ip ip addresses, in NETWORK format. 1284 * Only valid for IPv4 packets. 1285 */ 1286 u_int8_t proto; 1287 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1288 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1289 u_int16_t ip_len=0; 1290 int dyn_dir = MATCH_UNKNOWN; 1291 ipfw_dyn_rule *q = NULL; 1292 1293 if (m->m_flags & M_SKIP_FIREWALL) 1294 return 0; /* accept */ 1295 /* 1296 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1297 * MATCH_NONE when checked and not matched (q = NULL), 1298 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1299 */ 1300 1301 if (args->eh == NULL || /* layer 3 packet */ 1302 ( m->m_pkthdr.len >= sizeof(struct ip) && 1303 ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 1304 hlen = ip->ip_hl << 2; 1305 1306 /* 1307 * Collect parameters into local variables for faster matching. 1308 */ 1309 if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 1310 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1311 goto after_ip_checks; 1312 } 1313 1314 proto = args->f_id.proto = ip->ip_p; 1315 src_ip = ip->ip_src; 1316 dst_ip = ip->ip_dst; 1317 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1318 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1319 ip_len = ntohs(ip->ip_len); 1320 } else { 1321 offset = ip->ip_off & IP_OFFMASK; 1322 ip_len = ip->ip_len; 1323 } 1324 1325#define PULLUP_TO(len) \ 1326 do { \ 1327 if ((m)->m_len < (len)) { \ 1328 args->m = m = m_pullup(m, (len)); \ 1329 if (m == 0) \ 1330 goto pullup_failed; \ 1331 ip = mtod(m, struct ip *); \ 1332 } \ 1333 } while (0) 1334 1335 if (offset == 0) { 1336 switch (proto) { 1337 case IPPROTO_TCP: 1338 { 1339 struct tcphdr *tcp; 1340 1341 PULLUP_TO(hlen + sizeof(struct tcphdr)); 1342 tcp = L3HDR(struct tcphdr, ip); 1343 dst_port = tcp->th_dport; 1344 src_port = tcp->th_sport; 1345 args->f_id.flags = tcp->th_flags; 1346 } 1347 break; 1348 1349 case IPPROTO_UDP: 1350 { 1351 struct udphdr *udp; 1352 1353 PULLUP_TO(hlen + sizeof(struct udphdr)); 1354 udp = L3HDR(struct udphdr, ip); 1355 dst_port = udp->uh_dport; 1356 src_port = udp->uh_sport; 1357 } 1358 break; 1359 1360 case IPPROTO_ICMP: 1361 PULLUP_TO(hlen + 4); /* type, code and checksum. */ 1362 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 1363 break; 1364 1365 default: 1366 break; 1367 } 1368#undef PULLUP_TO 1369 } 1370 1371 args->f_id.src_ip = ntohl(src_ip.s_addr); 1372 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1373 args->f_id.src_port = src_port = ntohs(src_port); 1374 args->f_id.dst_port = dst_port = ntohs(dst_port); 1375 1376after_ip_checks: 1377 if (args->rule) { 1378 /* 1379 * Packet has already been tagged. Look for the next rule 1380 * to restart processing. 1381 * 1382 * If fw_one_pass != 0 then just accept it. 1383 * XXX should not happen here, but optimized out in 1384 * the caller. 1385 */ 1386 if (fw_one_pass) 1387 return 0; 1388 1389 f = args->rule->next_rule; 1390 if (f == NULL) 1391 f = lookup_next_rule(args->rule); 1392 } else { 1393 /* 1394 * Find the starting rule. It can be either the first 1395 * one, or the one after divert_rule if asked so. 1396 */ 1397 int skipto = args->divert_rule; 1398 1399 f = layer3_chain; 1400 if (args->eh == NULL && skipto != 0) { 1401 if (skipto >= IPFW_DEFAULT_RULE) 1402 return(IP_FW_PORT_DENY_FLAG); /* invalid */ 1403 while (f && f->rulenum <= skipto) 1404 f = f->next; 1405 if (f == NULL) /* drop packet */ 1406 return(IP_FW_PORT_DENY_FLAG); 1407 } 1408 } 1409 args->divert_rule = 0; /* reset to avoid confusion later */ 1410 1411 /* 1412 * Now scan the rules, and parse microinstructions for each rule. 1413 */ 1414 for (; f; f = f->next) { 1415 int l, cmdlen; 1416 ipfw_insn *cmd; 1417 int skip_or; /* skip rest of OR block */ 1418 1419again: 1420 if (set_disable & (1 << f->set) ) 1421 continue; 1422 1423 skip_or = 0; 1424 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1425 l -= cmdlen, cmd += cmdlen) { 1426 int match; 1427 1428 /* 1429 * check_body is a jump target used when we find a 1430 * CHECK_STATE, and need to jump to the body of 1431 * the target rule. 1432 */ 1433 1434check_body: 1435 cmdlen = F_LEN(cmd); 1436 /* 1437 * An OR block (insn_1 || .. || insn_n) has the 1438 * F_OR bit set in all but the last instruction. 1439 * The first match will set "skip_or", and cause 1440 * the following instructions to be skipped until 1441 * past the one with the F_OR bit clear. 1442 */ 1443 if (skip_or) { /* skip this instruction */ 1444 if ((cmd->len & F_OR) == 0) 1445 skip_or = 0; /* next one is good */ 1446 continue; 1447 } 1448 match = 0; /* set to 1 if we succeed */ 1449 1450 switch (cmd->opcode) { 1451 /* 1452 * The first set of opcodes compares the packet's 1453 * fields with some pattern, setting 'match' if a 1454 * match is found. At the end of the loop there is 1455 * logic to deal with F_NOT and F_OR flags associated 1456 * with the opcode. 1457 */ 1458 case O_NOP: 1459 match = 1; 1460 break; 1461 1462 case O_FORWARD_MAC: 1463 printf("ipfw: opcode %d unimplemented\n", 1464 cmd->opcode); 1465 break; 1466 1467 case O_GID: 1468 case O_UID: 1469 /* 1470 * We only check offset == 0 && proto != 0, 1471 * as this ensures that we have an IPv4 1472 * packet with the ports info. 1473 */ 1474 if (offset!=0) 1475 break; 1476 { 1477 struct inpcbinfo *pi; 1478 int wildcard; 1479 struct inpcb *pcb; 1480 1481 if (proto == IPPROTO_TCP) { 1482 wildcard = 0; 1483 pi = &tcbinfo; 1484 } else if (proto == IPPROTO_UDP) { 1485 wildcard = 1; 1486 pi = &udbinfo; 1487 } else 1488 break; 1489 1490 pcb = (oif) ? 1491 in_pcblookup_hash(pi, 1492 dst_ip, htons(dst_port), 1493 src_ip, htons(src_port), 1494 wildcard, oif) : 1495 in_pcblookup_hash(pi, 1496 src_ip, htons(src_port), 1497 dst_ip, htons(dst_port), 1498 wildcard, NULL); 1499 1500 if (pcb == NULL || pcb->inp_socket == NULL) 1501 break; 1502#if __FreeBSD_version < 500034 1503#define socheckuid(a,b) ((a)->so_cred->cr_uid == (b)) 1504#endif 1505 if (cmd->opcode == O_UID) { 1506 match = 1507 socheckuid(pcb->inp_socket, 1508 (uid_t)((ipfw_insn_u32 *)cmd)->d[0]); 1509 } else { 1510 match = groupmember( 1511 (uid_t)((ipfw_insn_u32 *)cmd)->d[0], 1512 pcb->inp_socket->so_cred); 1513 } 1514 } 1515 break; 1516 1517 case O_RECV: 1518 match = iface_match(m->m_pkthdr.rcvif, 1519 (ipfw_insn_if *)cmd); 1520 break; 1521 1522 case O_XMIT: 1523 match = iface_match(oif, (ipfw_insn_if *)cmd); 1524 break; 1525 1526 case O_VIA: 1527 match = iface_match(oif ? oif : 1528 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1529 break; 1530 1531 case O_MACADDR2: 1532 if (args->eh != NULL) { /* have MAC header */ 1533 u_int32_t *want = (u_int32_t *) 1534 ((ipfw_insn_mac *)cmd)->addr; 1535 u_int32_t *mask = (u_int32_t *) 1536 ((ipfw_insn_mac *)cmd)->mask; 1537 u_int32_t *hdr = (u_int32_t *)args->eh; 1538 1539 match = 1540 ( want[0] == (hdr[0] & mask[0]) && 1541 want[1] == (hdr[1] & mask[1]) && 1542 want[2] == (hdr[2] & mask[2]) ); 1543 } 1544 break; 1545 1546 case O_MAC_TYPE: 1547 if (args->eh != NULL) { 1548 u_int16_t t = 1549 ntohs(args->eh->ether_type); 1550 u_int16_t *p = 1551 ((ipfw_insn_u16 *)cmd)->ports; 1552 int i; 1553 1554 for (i = cmdlen - 1; !match && i>0; 1555 i--, p += 2) 1556 match = (t>=p[0] && t<=p[1]); 1557 } 1558 break; 1559 1560 case O_FRAG: 1561 match = (hlen > 0 && offset != 0); 1562 break; 1563 1564 case O_IN: /* "out" is "not in" */ 1565 match = (oif == NULL); 1566 break; 1567 1568 case O_LAYER2: 1569 match = (args->eh != NULL); 1570 break; 1571 1572 case O_PROTO: 1573 /* 1574 * We do not allow an arg of 0 so the 1575 * check of "proto" only suffices. 1576 */ 1577 match = (proto == cmd->arg1); 1578 break; 1579 1580 case O_IP_SRC: 1581 match = (hlen > 0 && 1582 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1583 src_ip.s_addr); 1584 break; 1585 1586 case O_IP_SRC_MASK: 1587 match = (hlen > 0 && 1588 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1589 (src_ip.s_addr & 1590 ((ipfw_insn_ip *)cmd)->mask.s_addr)); 1591 break; 1592 1593 case O_IP_SRC_ME: 1594 if (hlen > 0) { 1595 struct ifnet *tif; 1596 1597 INADDR_TO_IFP(src_ip, tif); 1598 match = (tif != NULL); 1599 } 1600 break; 1601 1602 case O_IP_DST_SET: 1603 case O_IP_SRC_SET: 1604 if (hlen > 0) { 1605 u_int32_t *d = (u_int32_t *)(cmd+1); 1606 u_int32_t addr = 1607 cmd->opcode == O_IP_DST_SET ? 1608 args->f_id.src_ip : 1609 args->f_id.dst_ip; 1610 1611 if (addr < d[0]) 1612 break; 1613 addr -= d[0]; /* subtract base */ 1614 match = (addr < cmd->arg1) && 1615 ( d[ 1 + (addr>>5)] & 1616 (1<<(addr & 0x1f)) ); 1617 } 1618 break; 1619 1620 case O_IP_DST: 1621 match = (hlen > 0 && 1622 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1623 dst_ip.s_addr); 1624 break; 1625 1626 case O_IP_DST_MASK: 1627 match = (hlen > 0) && 1628 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1629 (dst_ip.s_addr & 1630 ((ipfw_insn_ip *)cmd)->mask.s_addr)); 1631 break; 1632 1633 case O_IP_DST_ME: 1634 if (hlen > 0) { 1635 struct ifnet *tif; 1636 1637 INADDR_TO_IFP(dst_ip, tif); 1638 match = (tif != NULL); 1639 } 1640 break; 1641 1642 case O_IP_SRCPORT: 1643 case O_IP_DSTPORT: 1644 /* 1645 * offset == 0 && proto != 0 is enough 1646 * to guarantee that we have an IPv4 1647 * packet with port info. 1648 */ 1649 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 1650 && offset == 0) { 1651 u_int16_t x = 1652 (cmd->opcode == O_IP_SRCPORT) ? 1653 src_port : dst_port ; 1654 u_int16_t *p = 1655 ((ipfw_insn_u16 *)cmd)->ports; 1656 int i; 1657 1658 for (i = cmdlen - 1; !match && i>0; 1659 i--, p += 2) 1660 match = (x>=p[0] && x<=p[1]); 1661 } 1662 break; 1663 1664 case O_ICMPTYPE: 1665 match = (offset == 0 && proto==IPPROTO_ICMP && 1666 icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 1667 break; 1668 1669 case O_IPOPT: 1670 match = (hlen > 0 && ipopts_match(ip, cmd) ); 1671 break; 1672 1673 case O_IPVER: 1674 match = (hlen > 0 && cmd->arg1 == ip->ip_v); 1675 break; 1676 1677 case O_IPTTL: 1678 match = (hlen > 0 && cmd->arg1 == ip->ip_ttl); 1679 break; 1680 1681 case O_IPID: 1682 match = (hlen > 0 && 1683 cmd->arg1 == ntohs(ip->ip_id)); 1684 break; 1685 1686 case O_IPLEN: 1687 match = (hlen > 0 && cmd->arg1 == ip_len); 1688 break; 1689 1690 case O_IPPRECEDENCE: 1691 match = (hlen > 0 && 1692 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 1693 break; 1694 1695 case O_IPTOS: 1696 match = (hlen > 0 && 1697 flags_match(cmd, ip->ip_tos)); 1698 break; 1699 1700 case O_TCPFLAGS: 1701 match = (proto == IPPROTO_TCP && offset == 0 && 1702 flags_match(cmd, 1703 L3HDR(struct tcphdr,ip)->th_flags)); 1704 break; 1705 1706 case O_TCPOPTS: 1707 match = (proto == IPPROTO_TCP && offset == 0 && 1708 tcpopts_match(ip, cmd)); 1709 break; 1710 1711 case O_TCPSEQ: 1712 match = (proto == IPPROTO_TCP && offset == 0 && 1713 ((ipfw_insn_u32 *)cmd)->d[0] == 1714 L3HDR(struct tcphdr,ip)->th_seq); 1715 break; 1716 1717 case O_TCPACK: 1718 match = (proto == IPPROTO_TCP && offset == 0 && 1719 ((ipfw_insn_u32 *)cmd)->d[0] == 1720 L3HDR(struct tcphdr,ip)->th_ack); 1721 break; 1722 1723 case O_TCPWIN: 1724 match = (proto == IPPROTO_TCP && offset == 0 && 1725 cmd->arg1 == 1726 L3HDR(struct tcphdr,ip)->th_win); 1727 break; 1728 1729 case O_ESTAB: 1730 /* reject packets which have SYN only */ 1731 /* XXX should i also check for TH_ACK ? */ 1732 match = (proto == IPPROTO_TCP && offset == 0 && 1733 (L3HDR(struct tcphdr,ip)->th_flags & 1734 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 1735 break; 1736 1737 case O_LOG: 1738 if (fw_verbose) 1739 ipfw_log(f, hlen, args->eh, m, oif); 1740 match = 1; 1741 break; 1742 1743 case O_PROB: 1744 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 1745 break; 1746 1747 /* 1748 * The second set of opcodes represents 'actions', 1749 * i.e. the terminal part of a rule once the packet 1750 * matches all previous patterns. 1751 * Typically there is only one action for each rule, 1752 * and the opcode is stored at the end of the rule 1753 * (but there are exceptions -- see below). 1754 * 1755 * In general, here we set retval and terminate the 1756 * outer loop (would be a 'break 3' in some language, 1757 * but we need to do a 'goto done'). 1758 * 1759 * Exceptions: 1760 * O_COUNT and O_SKIPTO actions: 1761 * instead of terminating, we jump to the next rule 1762 * ('goto next_rule', equivalent to a 'break 2'), 1763 * or to the SKIPTO target ('goto again' after 1764 * having set f, cmd and l), respectively. 1765 * 1766 * O_LIMIT and O_KEEP_STATE: these opcodes are 1767 * not real 'actions', and are stored right 1768 * before the 'action' part of the rule. 1769 * These opcodes try to install an entry in the 1770 * state tables; if successful, we continue with 1771 * the next opcode (match=1; break;), otherwise 1772 * the packet * must be dropped 1773 * ('goto done' after setting retval); 1774 * 1775 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 1776 * cause a lookup of the state table, and a jump 1777 * to the 'action' part of the parent rule 1778 * ('goto check_body') if an entry is found, or 1779 * (CHECK_STATE only) a jump to the next rule if 1780 * the entry is not found ('goto next_rule'). 1781 * The result of the lookup is cached to make 1782 * further instances of these opcodes are 1783 * effectively NOPs. 1784 */ 1785 case O_LIMIT: 1786 case O_KEEP_STATE: 1787 if (install_state(f, 1788 (ipfw_insn_limit *)cmd, args)) { 1789 retval = IP_FW_PORT_DENY_FLAG; 1790 goto done; /* error/limit violation */ 1791 } 1792 match = 1; 1793 break; 1794 1795 case O_PROBE_STATE: 1796 case O_CHECK_STATE: 1797 /* 1798 * dynamic rules are checked at the first 1799 * keep-state or check-state occurrence, 1800 * with the result being stored in dyn_dir. 1801 * The compiler introduces a PROBE_STATE 1802 * instruction for us when we have a 1803 * KEEP_STATE (because PROBE_STATE needs 1804 * to be run first). 1805 */ 1806 if (dyn_dir == MATCH_UNKNOWN && 1807 (q = lookup_dyn_rule(&args->f_id, 1808 &dyn_dir, proto == IPPROTO_TCP ? 1809 L3HDR(struct tcphdr, ip) : NULL)) 1810 != NULL) { 1811 /* 1812 * Found dynamic entry, update stats 1813 * and jump to the 'action' part of 1814 * the parent rule. 1815 */ 1816 q->pcnt++; 1817 q->bcnt += ip_len; 1818 f = q->rule; 1819 cmd = ACTION_PTR(f); 1820 l = f->cmd_len - f->act_ofs; 1821 goto check_body; 1822 } 1823 /* 1824 * Dynamic entry not found. If CHECK_STATE, 1825 * skip to next rule, if PROBE_STATE just 1826 * ignore and continue with next opcode. 1827 */ 1828 if (cmd->opcode == O_CHECK_STATE) 1829 goto next_rule; 1830 match = 1; 1831 break; 1832 1833 case O_ACCEPT: 1834 retval = 0; /* accept */ 1835 goto done; 1836 1837 case O_PIPE: 1838 case O_QUEUE: 1839 args->rule = f; /* report matching rule */ 1840 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG; 1841 goto done; 1842 1843 case O_DIVERT: 1844 case O_TEE: 1845 if (args->eh) /* not on layer 2 */ 1846 break; 1847 args->divert_rule = f->rulenum; 1848 retval = (cmd->opcode == O_DIVERT) ? 1849 cmd->arg1 : 1850 cmd->arg1 | IP_FW_PORT_TEE_FLAG; 1851 goto done; 1852 1853 case O_COUNT: 1854 case O_SKIPTO: 1855 f->pcnt++; /* update stats */ 1856 f->bcnt += ip_len; 1857 f->timestamp = time_second; 1858 if (cmd->opcode == O_COUNT) 1859 goto next_rule; 1860 /* handle skipto */ 1861 if (f->next_rule == NULL) 1862 lookup_next_rule(f); 1863 f = f->next_rule; 1864 goto again; 1865 1866 case O_REJECT: 1867 /* 1868 * Drop the packet and send a reject notice 1869 * if the packet is not ICMP (or is an ICMP 1870 * query), and it is not multicast/broadcast. 1871 */ 1872 if (hlen > 0 && 1873 (proto != IPPROTO_ICMP || 1874 is_icmp_query(ip)) && 1875 !(m->m_flags & (M_BCAST|M_MCAST)) && 1876 !IN_MULTICAST(dst_ip.s_addr)) { 1877 send_reject(args, cmd->arg1, 1878 offset,ip_len); 1879 m = args->m; 1880 } 1881 /* FALLTHROUGH */ 1882 case O_DENY: 1883 retval = IP_FW_PORT_DENY_FLAG; 1884 goto done; 1885 1886 case O_FORWARD_IP: 1887 if (args->eh) /* not valid on layer2 pkts */ 1888 break; 1889 if (!q || dyn_dir == MATCH_FORWARD) 1890 args->next_hop = 1891 &((ipfw_insn_sa *)cmd)->sa; 1892 retval = 0; 1893 goto done; 1894 1895 default: 1896 panic("-- unknown opcode %d\n", cmd->opcode); 1897 } /* end of switch() on opcodes */ 1898 1899 if (cmd->len & F_NOT) 1900 match = !match; 1901 1902 if (match) { 1903 if (cmd->len & F_OR) 1904 skip_or = 1; 1905 } else { 1906 if (!(cmd->len & F_OR)) /* not an OR block, */ 1907 break; /* try next rule */ 1908 } 1909 1910 } /* end of inner for, scan opcodes */ 1911 1912next_rule:; /* try next rule */ 1913 1914 } /* end of outer for, scan rules */ 1915 printf("+++ ipfw: ouch!, skip past end of rules, denying packet\n"); 1916 return(IP_FW_PORT_DENY_FLAG); 1917 1918done: 1919 /* Update statistics */ 1920 f->pcnt++; 1921 f->bcnt += ip_len; 1922 f->timestamp = time_second; 1923 return retval; 1924 1925pullup_failed: 1926 if (fw_verbose) 1927 printf("pullup failed\n"); 1928 return(IP_FW_PORT_DENY_FLAG); 1929} 1930 1931/* 1932 * When a rule is added/deleted, clear the next_rule pointers in all rules. 1933 * These will be reconstructed on the fly as packets are matched. 1934 * Must be called at splimp(). 1935 */ 1936static void 1937flush_rule_ptrs(void) 1938{ 1939 struct ip_fw *rule; 1940 1941 for (rule = layer3_chain; rule; rule = rule->next) 1942 rule->next_rule = NULL; 1943} 1944 1945/* 1946 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 1947 * pipe/queue, or to all of them (match == NULL). 1948 * Must be called at splimp(). 1949 */ 1950void 1951flush_pipe_ptrs(struct dn_flow_set *match) 1952{ 1953 struct ip_fw *rule; 1954 1955 for (rule = layer3_chain; rule; rule = rule->next) { 1956 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 1957 1958 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 1959 continue; 1960 1961 if (match == NULL || cmd->pipe_ptr == match) 1962 cmd->pipe_ptr = NULL; 1963 } 1964} 1965 1966/* 1967 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 1968 * possibly create a rule number and add the rule to the list. 1969 * Update the rule_number in the input struct so the caller knows it as well. 1970 */ 1971static int 1972add_rule(struct ip_fw **head, struct ip_fw *input_rule) 1973{ 1974 struct ip_fw *rule, *f, *prev; 1975 int s; 1976 int l = RULESIZE(input_rule); 1977 1978 if (*head == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 1979 return (EINVAL); 1980 1981 rule = malloc(l, M_IPFW, M_DONTWAIT | M_ZERO); 1982 if (rule == NULL) 1983 return (ENOSPC); 1984 1985 bcopy(input_rule, rule, l); 1986 1987 rule->next = NULL; 1988 rule->next_rule = NULL; 1989 1990 rule->pcnt = 0; 1991 rule->bcnt = 0; 1992 rule->timestamp = 0; 1993 1994 s = splimp(); 1995 1996 if (*head == NULL) { /* default rule */ 1997 *head = rule; 1998 goto done; 1999 } 2000 2001 /* 2002 * If rulenum is 0, find highest numbered rule before the 2003 * default rule, and add autoinc_step 2004 */ 2005 if (autoinc_step < 1) 2006 autoinc_step = 1; 2007 else if (autoinc_step > 1000) 2008 autoinc_step = 1000; 2009 if (rule->rulenum == 0) { 2010 /* 2011 * locate the highest numbered rule before default 2012 */ 2013 for (f = *head; f; f = f->next) { 2014 if (f->rulenum == IPFW_DEFAULT_RULE) 2015 break; 2016 rule->rulenum = f->rulenum; 2017 } 2018 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2019 rule->rulenum += autoinc_step; 2020 input_rule->rulenum = rule->rulenum; 2021 } 2022 2023 /* 2024 * Now insert the new rule in the right place in the sorted list. 2025 */ 2026 for (prev = NULL, f = *head; f; prev = f, f = f->next) { 2027 if (f->rulenum > rule->rulenum) { /* found the location */ 2028 if (prev) { 2029 rule->next = f; 2030 prev->next = rule; 2031 } else { /* head insert */ 2032 rule->next = *head; 2033 *head = rule; 2034 } 2035 break; 2036 } 2037 } 2038 flush_rule_ptrs(); 2039done: 2040 static_count++; 2041 static_len += l; 2042 splx(s); 2043 DEB(printf("++ installed rule %d, static count now %d\n", 2044 rule->rulenum, static_count);) 2045 return (0); 2046} 2047 2048/** 2049 * Free storage associated with a static rule (including derived 2050 * dynamic rules). 2051 * The caller is in charge of clearing rule pointers to avoid 2052 * dangling pointers. 2053 * @return a pointer to the next entry. 2054 * Arguments are not checked, so they better be correct. 2055 * Must be called at splimp(). 2056 */ 2057static struct ip_fw * 2058delete_rule(struct ip_fw **head, struct ip_fw *prev, struct ip_fw *rule) 2059{ 2060 struct ip_fw *n; 2061 int l = RULESIZE(rule); 2062 2063 n = rule->next; 2064 remove_dyn_rule(rule, NULL /* force removal */); 2065 if (prev == NULL) 2066 *head = n; 2067 else 2068 prev->next = n; 2069 static_count--; 2070 static_len -= l; 2071 2072 if (DUMMYNET_LOADED) 2073 ip_dn_ruledel_ptr(rule); 2074 free(rule, M_IPFW); 2075 return n; 2076} 2077 2078/* 2079 * Deletes all rules from a chain (including the default rule 2080 * if the second argument is set). 2081 * Must be called at splimp(). 2082 */ 2083static void 2084free_chain(struct ip_fw **chain, int kill_default) 2085{ 2086 struct ip_fw *rule; 2087 2088 flush_rule_ptrs(); /* more efficient to do outside the loop */ 2089 2090 while ( (rule = *chain) != NULL && 2091 (kill_default || rule->rulenum != IPFW_DEFAULT_RULE) ) 2092 delete_rule(chain, NULL, rule); 2093} 2094 2095/** 2096 * Remove all rules with given number, and also do set manipulation. 2097 * 2098 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2099 * the next 8 bits are the new set, the top 8 bits are the command: 2100 * 2101 * 0 delete rules with given number 2102 * 1 delete rules with given set number 2103 * 2 move rules with given number to new set 2104 * 3 move rules with given set number to new set 2105 * 4 swap sets with given numbers 2106 */ 2107static int 2108del_entry(struct ip_fw **chain, u_int32_t arg) 2109{ 2110 struct ip_fw *prev, *rule; 2111 int s; 2112 u_int16_t rulenum; 2113 u_int8_t cmd, new_set; 2114 2115 rulenum = arg & 0xffff; 2116 cmd = (arg >> 24) & 0xff; 2117 new_set = (arg >> 16) & 0xff; 2118 2119 if (cmd > 4) 2120 return EINVAL; 2121 if (new_set > 30) 2122 return EINVAL; 2123 if (cmd == 0 || cmd == 2) { 2124 if (rulenum == IPFW_DEFAULT_RULE) 2125 return EINVAL; 2126 } else { 2127 if (rulenum > 30) 2128 return EINVAL; 2129 } 2130 2131 switch (cmd) { 2132 case 0: /* delete rules with given number */ 2133 /* 2134 * locate first rule to delete 2135 */ 2136 for (prev = NULL, rule = *chain; 2137 rule && rule->rulenum < rulenum; 2138 prev = rule, rule = rule->next) 2139 ; 2140 if (rule->rulenum != rulenum) 2141 return EINVAL; 2142 2143 s = splimp(); /* no access to rules while removing */ 2144 /* 2145 * flush pointers outside the loop, then delete all matching 2146 * rules. prev remains the same throughout the cycle. 2147 */ 2148 flush_rule_ptrs(); 2149 while (rule && rule->rulenum == rulenum) 2150 rule = delete_rule(chain, prev, rule); 2151 splx(s); 2152 break; 2153 2154 case 1: /* delete all rules with given set number */ 2155 s = splimp(); 2156 flush_rule_ptrs(); 2157 for (prev = NULL, rule = *chain; rule ; ) 2158 if (rule->set == rulenum) 2159 rule = delete_rule(chain, prev, rule); 2160 else { 2161 prev = rule; 2162 rule = rule->next; 2163 } 2164 splx(s); 2165 break; 2166 2167 case 2: /* move rules with given number to new set */ 2168 s = splimp(); 2169 for (rule = *chain; rule ; rule = rule->next) 2170 if (rule->rulenum == rulenum) 2171 rule->set = new_set; 2172 splx(s); 2173 break; 2174 2175 case 3: /* move rules with given set number to new set */ 2176 s = splimp(); 2177 for (rule = *chain; rule ; rule = rule->next) 2178 if (rule->set == rulenum) 2179 rule->set = new_set; 2180 splx(s); 2181 break; 2182 2183 case 4: /* swap two sets */ 2184 s = splimp(); 2185 for (rule = *chain; rule ; rule = rule->next) 2186 if (rule->set == rulenum) 2187 rule->set = new_set; 2188 else if (rule->set == new_set) 2189 rule->set = rulenum; 2190 splx(s); 2191 break; 2192 } 2193 return 0; 2194} 2195 2196/* 2197 * Clear counters for a specific rule. 2198 */ 2199static void 2200clear_counters(struct ip_fw *rule, int log_only) 2201{ 2202 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 2203 2204 if (log_only == 0) { 2205 rule->bcnt = rule->pcnt = 0; 2206 rule->timestamp = 0; 2207 } 2208 if (l->o.opcode == O_LOG) 2209 l->log_left = l->max_log; 2210} 2211 2212/** 2213 * Reset some or all counters on firewall rules. 2214 * @arg frwl is null to clear all entries, or contains a specific 2215 * rule number. 2216 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 2217 */ 2218static int 2219zero_entry(int rulenum, int log_only) 2220{ 2221 struct ip_fw *rule; 2222 int s; 2223 char *msg; 2224 2225 if (rulenum == 0) { 2226 s = splimp(); 2227 norule_counter = 0; 2228 for (rule = layer3_chain; rule; rule = rule->next) 2229 clear_counters(rule, log_only); 2230 splx(s); 2231 msg = log_only ? "ipfw: All logging counts reset.\n" : 2232 "ipfw: Accounting cleared.\n"; 2233 } else { 2234 int cleared = 0; 2235 /* 2236 * We can have multiple rules with the same number, so we 2237 * need to clear them all. 2238 */ 2239 for (rule = layer3_chain; rule; rule = rule->next) 2240 if (rule->rulenum == rulenum) { 2241 s = splimp(); 2242 while (rule && rule->rulenum == rulenum) { 2243 clear_counters(rule, log_only); 2244 rule = rule->next; 2245 } 2246 splx(s); 2247 cleared = 1; 2248 break; 2249 } 2250 if (!cleared) /* we did not find any matching rules */ 2251 return (EINVAL); 2252 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 2253 "ipfw: Entry %d cleared.\n"; 2254 } 2255 if (fw_verbose) 2256 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 2257 return (0); 2258} 2259 2260/* 2261 * Check validity of the structure before insert. 2262 * Fortunately rules are simple, so this mostly need to check rule sizes. 2263 */ 2264static int 2265check_ipfw_struct(struct ip_fw *rule, int size) 2266{ 2267 int l, cmdlen = 0; 2268 int have_action=0; 2269 ipfw_insn *cmd; 2270 2271 if (size < sizeof(*rule)) { 2272 printf("ipfw: rule too short\n"); 2273 return (EINVAL); 2274 } 2275 /* first, check for valid size */ 2276 l = RULESIZE(rule); 2277 if (l != size) { 2278 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 2279 return (EINVAL); 2280 } 2281 /* 2282 * Now go for the individual checks. Very simple ones, basically only 2283 * instruction sizes. 2284 */ 2285 for (l = rule->cmd_len, cmd = rule->cmd ; 2286 l > 0 ; l -= cmdlen, cmd += cmdlen) { 2287 cmdlen = F_LEN(cmd); 2288 if (cmdlen > l) { 2289 printf("ipfw: opcode %d size truncated\n", 2290 cmd->opcode); 2291 return EINVAL; 2292 } 2293 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 2294 switch (cmd->opcode) { 2295 case O_NOP: 2296 case O_PROBE_STATE: 2297 case O_KEEP_STATE: 2298 case O_PROTO: 2299 case O_IP_SRC_ME: 2300 case O_IP_DST_ME: 2301 case O_LAYER2: 2302 case O_IN: 2303 case O_FRAG: 2304 case O_IPOPT: 2305 case O_IPLEN: 2306 case O_IPID: 2307 case O_IPTOS: 2308 case O_IPPRECEDENCE: 2309 case O_IPTTL: 2310 case O_IPVER: 2311 case O_TCPWIN: 2312 case O_TCPFLAGS: 2313 case O_TCPOPTS: 2314 case O_ESTAB: 2315 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2316 goto bad_size; 2317 break; 2318 2319 case O_UID: 2320 case O_GID: 2321 case O_IP_SRC: 2322 case O_IP_DST: 2323 case O_TCPSEQ: 2324 case O_TCPACK: 2325 case O_PROB: 2326 case O_ICMPTYPE: 2327 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2328 goto bad_size; 2329 break; 2330 2331 case O_LIMIT: 2332 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 2333 goto bad_size; 2334 break; 2335 2336 case O_LOG: 2337 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 2338 goto bad_size; 2339 2340 ((ipfw_insn_log *)cmd)->log_left = 2341 ((ipfw_insn_log *)cmd)->max_log; 2342 2343 break; 2344 2345 case O_IP_SRC_MASK: 2346 case O_IP_DST_MASK: 2347 if (cmdlen != F_INSN_SIZE(ipfw_insn_ip)) 2348 goto bad_size; 2349 if (((ipfw_insn_ip *)cmd)->mask.s_addr == 0) { 2350 printf("ipfw: opcode %d, useless rule\n", 2351 cmd->opcode); 2352 return EINVAL; 2353 } 2354 break; 2355 2356 case O_IP_SRC_SET: 2357 case O_IP_DST_SET: 2358 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 2359 printf("ipfw: invalid set size %d\n", 2360 cmd->arg1); 2361 return EINVAL; 2362 } 2363 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 2364 (cmd->arg1+31)/32 ) 2365 goto bad_size; 2366 break; 2367 2368 case O_MACADDR2: 2369 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 2370 goto bad_size; 2371 break; 2372 2373 case O_MAC_TYPE: 2374 case O_IP_SRCPORT: 2375 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 2376 if (cmdlen < 2 || cmdlen > 31) 2377 goto bad_size; 2378 break; 2379 2380 case O_RECV: 2381 case O_XMIT: 2382 case O_VIA: 2383 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 2384 goto bad_size; 2385 break; 2386 2387 case O_PIPE: 2388 case O_QUEUE: 2389 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 2390 goto bad_size; 2391 goto check_action; 2392 2393 case O_FORWARD_IP: 2394 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 2395 goto bad_size; 2396 goto check_action; 2397 2398 case O_FORWARD_MAC: /* XXX not implemented yet */ 2399 case O_CHECK_STATE: 2400 case O_COUNT: 2401 case O_ACCEPT: 2402 case O_DENY: 2403 case O_REJECT: 2404 case O_SKIPTO: 2405 case O_DIVERT: 2406 case O_TEE: 2407 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2408 goto bad_size; 2409check_action: 2410 if (have_action) { 2411 printf("ipfw: opcode %d, multiple actions" 2412 " not allowed\n", 2413 cmd->opcode); 2414 return EINVAL; 2415 } 2416 have_action = 1; 2417 if (l != cmdlen) { 2418 printf("ipfw: opcode %d, action must be" 2419 " last opcode\n", 2420 cmd->opcode); 2421 return EINVAL; 2422 } 2423 break; 2424 default: 2425 printf("ipfw: opcode %d, unknown opcode\n", 2426 cmd->opcode); 2427 return EINVAL; 2428 } 2429 } 2430 if (have_action == 0) { 2431 printf("ipfw: missing action\n"); 2432 return EINVAL; 2433 } 2434 return 0; 2435 2436bad_size: 2437 printf("ipfw: opcode %d size %d wrong\n", 2438 cmd->opcode, cmdlen); 2439 return EINVAL; 2440} 2441 2442 2443/** 2444 * {set|get}sockopt parser. 2445 */ 2446static int 2447ipfw_ctl(struct sockopt *sopt) 2448{ 2449 int error, s, rulenum; 2450 size_t size; 2451 struct ip_fw *bp , *buf, *rule; 2452 2453 static u_int32_t rule_buf[255]; /* we copy the data here */ 2454 2455 /* 2456 * Disallow modifications in really-really secure mode, but still allow 2457 * the logging counters to be reset. 2458 */ 2459 if (sopt->sopt_name == IP_FW_ADD || 2460 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 2461#if __FreeBSD_version >= 500034 2462 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 2463 if (error) 2464 return (error); 2465#else /* FreeBSD 4.x */ 2466 if (securelevel >= 3) 2467 return (EPERM); 2468#endif 2469 } 2470 2471 error = 0; 2472 2473 switch (sopt->sopt_name) { 2474 case IP_FW_GET: 2475 /* 2476 * pass up a copy of the current rules. Static rules 2477 * come first (the last of which has number IPFW_DEFAULT_RULE), 2478 * followed by a possibly empty list of dynamic rule. 2479 * The last dynamic rule has NULL in the "next" field. 2480 */ 2481 s = splimp(); 2482 size = static_len; /* size of static rules */ 2483 if (ipfw_dyn_v) /* add size of dyn.rules */ 2484 size += (dyn_count * sizeof(ipfw_dyn_rule)); 2485 2486 /* 2487 * XXX todo: if the user passes a short length just to know 2488 * how much room is needed, do not bother filling up the 2489 * buffer, just jump to the sooptcopyout. 2490 */ 2491 buf = malloc(size, M_TEMP, M_WAITOK); 2492 if (buf == 0) { 2493 splx(s); 2494 error = ENOBUFS; 2495 break; 2496 } 2497 2498 bp = buf; 2499 for (rule = layer3_chain; rule ; rule = rule->next) { 2500 int i = RULESIZE(rule); 2501 bcopy(rule, bp, i); 2502 /* 2503 * abuse 'next_rule' to store the set_disable word 2504 */ 2505 (u_int32_t)(((struct ip_fw *)bp)->next_rule) = 2506 set_disable; 2507 bp = (struct ip_fw *)((char *)bp + i); 2508 } 2509 if (ipfw_dyn_v) { 2510 int i; 2511 ipfw_dyn_rule *p, *dst, *last = NULL; 2512 2513 dst = (ipfw_dyn_rule *)bp; 2514 for (i = 0 ; i < curr_dyn_buckets ; i++ ) 2515 for ( p = ipfw_dyn_v[i] ; p != NULL ; 2516 p = p->next, dst++ ) { 2517 bcopy(p, dst, sizeof *p); 2518 (int)dst->rule = p->rule->rulenum ; 2519 /* 2520 * store a non-null value in "next". 2521 * The userland code will interpret a 2522 * NULL here as a marker 2523 * for the last dynamic rule. 2524 */ 2525 dst->next = dst ; 2526 last = dst ; 2527 dst->expire = 2528 TIME_LEQ(dst->expire, time_second) ? 2529 0 : dst->expire - time_second ; 2530 } 2531 if (last != NULL) /* mark last dynamic rule */ 2532 last->next = NULL; 2533 } 2534 splx(s); 2535 2536 error = sooptcopyout(sopt, buf, size); 2537 free(buf, M_TEMP); 2538 break; 2539 2540 case IP_FW_FLUSH: 2541 /* 2542 * Normally we cannot release the lock on each iteration. 2543 * We could do it here only because we start from the head all 2544 * the times so there is no risk of missing some entries. 2545 * On the other hand, the risk is that we end up with 2546 * a very inconsistent ruleset, so better keep the lock 2547 * around the whole cycle. 2548 * 2549 * XXX this code can be improved by resetting the head of 2550 * the list to point to the default rule, and then freeing 2551 * the old list without the need for a lock. 2552 */ 2553 2554 s = splimp(); 2555 free_chain(&layer3_chain, 0 /* keep default rule */); 2556 splx(s); 2557 break; 2558 2559 case IP_FW_ADD: 2560 rule = (struct ip_fw *)rule_buf; /* XXX do a malloc */ 2561 error = sooptcopyin(sopt, rule, sizeof(rule_buf), 2562 sizeof(struct ip_fw) ); 2563 size = sopt->sopt_valsize; 2564 if (error || (error = check_ipfw_struct(rule, size))) 2565 break; 2566 2567 error = add_rule(&layer3_chain, rule); 2568 size = RULESIZE(rule); 2569 if (!error && sopt->sopt_dir == SOPT_GET) 2570 error = sooptcopyout(sopt, rule, size); 2571 break; 2572 2573 case IP_FW_DEL: 2574 /* 2575 * IP_FW_DEL is used for deleting single rules or sets, 2576 * and (ab)used to atomically manipulate sets. Argument size 2577 * is used to distinguish between the two: 2578 * sizeof(u_int32_t) 2579 * delete single rule or set of rules, 2580 * or reassign rules (or sets) to a different set. 2581 * 2*sizeof(u_int32_t) 2582 * atomic disable/enable sets. 2583 * first u_int32_t contains sets to be disabled, 2584 * second u_int32_t contains sets to be enabled. 2585 */ 2586 error = sooptcopyin(sopt, rule_buf, 2587 2*sizeof(u_int32_t), sizeof(u_int32_t)); 2588 if (error) 2589 break; 2590 size = sopt->sopt_valsize; 2591 if (size == sizeof(u_int32_t)) /* delete or reassign */ 2592 error = del_entry(&layer3_chain, rule_buf[0]); 2593 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 2594 set_disable = 2595 (set_disable | rule_buf[0]) & ~rule_buf[1] & 2596 ~(1<<31); /* set 31 always enabled */ 2597 else 2598 error = EINVAL; 2599 break; 2600 2601 case IP_FW_ZERO: 2602 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 2603 rulenum=0; 2604 2605 if (sopt->sopt_val != 0) { 2606 error = sooptcopyin(sopt, &rulenum, 2607 sizeof(int), sizeof(int)); 2608 if (error) 2609 break; 2610 } 2611 error = zero_entry(rulenum, sopt->sopt_name == IP_FW_RESETLOG); 2612 break; 2613 2614 default: 2615 printf("ipfw_ctl invalid option %d\n", sopt->sopt_name); 2616 error = EINVAL; 2617 } 2618 2619 return (error); 2620} 2621 2622/** 2623 * dummynet needs a reference to the default rule, because rules can be 2624 * deleted while packets hold a reference to them. When this happens, 2625 * dummynet changes the reference to the default rule (it could well be a 2626 * NULL pointer, but this way we do not need to check for the special 2627 * case, plus here he have info on the default behaviour). 2628 */ 2629struct ip_fw *ip_fw_default_rule; 2630 2631/* 2632 * This procedure is only used to handle keepalives. It is invoked 2633 * every dyn_keepalive_period 2634 */ 2635static void 2636ipfw_tick(void * __unused unused) 2637{ 2638 int i; 2639 int s; 2640 ipfw_dyn_rule *q; 2641 2642 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 2643 goto done; 2644 2645 s = splimp(); 2646 for (i = 0 ; i < curr_dyn_buckets ; i++) { 2647 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 2648 if (q->dyn_type == O_LIMIT_PARENT) 2649 continue; 2650 if (q->id.proto != IPPROTO_TCP) 2651 continue; 2652 if ( (q->state & BOTH_SYN) != BOTH_SYN) 2653 continue; 2654 if (TIME_LEQ( time_second+dyn_keepalive_interval, 2655 q->expire)) 2656 continue; /* too early */ 2657 if (TIME_LEQ(q->expire, time_second)) 2658 continue; /* too late, rule expired */ 2659 2660 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 2661 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 2662 } 2663 } 2664 splx(s); 2665done: 2666 ipfw_timeout_h = timeout(ipfw_tick, NULL, dyn_keepalive_period*hz); 2667} 2668 2669static void 2670ipfw_init(void) 2671{ 2672 struct ip_fw default_rule; 2673 2674 ip_fw_chk_ptr = ipfw_chk; 2675 ip_fw_ctl_ptr = ipfw_ctl; 2676 layer3_chain = NULL; 2677 2678 bzero(&default_rule, sizeof default_rule); 2679 2680 default_rule.act_ofs = 0; 2681 default_rule.rulenum = IPFW_DEFAULT_RULE; 2682 default_rule.cmd_len = 1; 2683 default_rule.set = 31; 2684 2685 default_rule.cmd[0].len = 1; 2686 default_rule.cmd[0].opcode = 2687#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 2688 1 ? O_ACCEPT : 2689#endif 2690 O_DENY; 2691 2692 add_rule(&layer3_chain, &default_rule); 2693 2694 ip_fw_default_rule = layer3_chain; 2695 printf("ipfw2 initialized, divert %s, " 2696 "rule-based forwarding enabled, default to %s, logging ", 2697#ifdef IPDIVERT 2698 "enabled", 2699#else 2700 "disabled", 2701#endif 2702 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 2703 2704#ifdef IPFIREWALL_VERBOSE 2705 fw_verbose = 1; 2706#endif 2707#ifdef IPFIREWALL_VERBOSE_LIMIT 2708 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 2709#endif 2710 if (fw_verbose == 0) 2711 printf("disabled\n"); 2712 else if (verbose_limit == 0) 2713 printf("unlimited\n"); 2714 else 2715 printf("limited to %d packets/entry by default\n", 2716 verbose_limit); 2717 bzero(&ipfw_timeout_h, sizeof(struct callout_handle)); 2718 ipfw_timeout_h = timeout(ipfw_tick, NULL, hz); 2719} 2720 2721static int 2722ipfw_modevent(module_t mod, int type, void *unused) 2723{ 2724 int s; 2725 int err = 0; 2726 2727 switch (type) { 2728 case MOD_LOAD: 2729 s = splimp(); 2730 if (IPFW_LOADED) { 2731 splx(s); 2732 printf("IP firewall already loaded\n"); 2733 err = EEXIST; 2734 } else { 2735 ipfw_init(); 2736 splx(s); 2737 } 2738 break; 2739 2740 case MOD_UNLOAD: 2741#if !defined(KLD_MODULE) 2742 printf("ipfw statically compiled, cannot unload\n"); 2743 err = EBUSY; 2744#else 2745 s = splimp(); 2746 untimeout(ipfw_tick, NULL, ipfw_timeout_h); 2747 ip_fw_chk_ptr = NULL; 2748 ip_fw_ctl_ptr = NULL; 2749 free_chain(&layer3_chain, 1 /* kill default rule */); 2750 splx(s); 2751 printf("IP firewall unloaded\n"); 2752#endif 2753 break; 2754 default: 2755 break; 2756 } 2757 return err; 2758} 2759 2760static moduledata_t ipfwmod = { 2761 "ipfw", 2762 ipfw_modevent, 2763 0 2764}; 2765DECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PSEUDO, SI_ORDER_ANY); 2766MODULE_VERSION(ipfw, 1); 2767#endif /* IPFW2 */ 2768