ip_fw2.c revision 136073
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 136073 2004-10-03 00:26:35Z green $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdn.h" 38#include "opt_ipdivert.h" 39#include "opt_inet.h" 40#include "opt_ipsec.h" 41#ifndef INET 42#error IPFIREWALL requires INET. 43#endif /* INET */ 44#endif 45 46#define IPFW2 1 47#if IPFW2 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/malloc.h> 51#include <sys/mbuf.h> 52#include <sys/kernel.h> 53#include <sys/jail.h> 54#include <sys/module.h> 55#include <sys/proc.h> 56#include <sys/socket.h> 57#include <sys/socketvar.h> 58#include <sys/sysctl.h> 59#include <sys/syslog.h> 60#include <sys/ucred.h> 61#include <net/if.h> 62#include <net/radix.h> 63#include <net/route.h> 64#include <netinet/in.h> 65#include <netinet/in_systm.h> 66#include <netinet/in_var.h> 67#include <netinet/in_pcb.h> 68#include <netinet/ip.h> 69#include <netinet/ip_var.h> 70#include <netinet/ip_icmp.h> 71#include <netinet/ip_fw.h> 72#include <netinet/ip_divert.h> 73#include <netinet/ip_dummynet.h> 74#include <netinet/tcp.h> 75#include <netinet/tcp_timer.h> 76#include <netinet/tcp_var.h> 77#include <netinet/tcpip.h> 78#include <netinet/udp.h> 79#include <netinet/udp_var.h> 80#include <altq/if_altq.h> 81 82#ifdef IPSEC 83#include <netinet6/ipsec.h> 84#endif 85 86#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 87 88#include <machine/in_cksum.h> /* XXX for in_cksum */ 89 90/* 91 * set_disable contains one bit per set value (0..31). 92 * If the bit is set, all rules with the corresponding set 93 * are disabled. Set RESVD_SET(31) is reserved for the default rule 94 * and rules that are not deleted by the flush command, 95 * and CANNOT be disabled. 96 * Rules in set RESVD_SET can only be deleted explicitly. 97 */ 98static u_int32_t set_disable; 99 100static int fw_verbose; 101static int verbose_limit; 102 103static struct callout ipfw_timeout; 104#define IPFW_DEFAULT_RULE 65535 105 106/* 107 * Data structure to cache our ucred related 108 * information. This structure only gets used if 109 * the user specified UID/GID based constraints in 110 * a firewall rule. 111 */ 112struct ip_fw_ugid { 113 gid_t fw_groups[NGROUPS]; 114 int fw_ngroups; 115 uid_t fw_uid; 116 int fw_prid; 117}; 118 119struct ip_fw_chain { 120 struct ip_fw *rules; /* list of rules */ 121 struct ip_fw *reap; /* list of rules to reap */ 122 struct mtx mtx; /* lock guarding rule list */ 123}; 124#define IPFW_LOCK_INIT(_chain) \ 125 mtx_init(&(_chain)->mtx, "IPFW static rules", NULL, \ 126 MTX_DEF | MTX_RECURSE) 127#define IPFW_LOCK_DESTROY(_chain) mtx_destroy(&(_chain)->mtx) 128#define IPFW_LOCK(_chain) mtx_lock(&(_chain)->mtx) 129#define IPFW_UNLOCK(_chain) mtx_unlock(&(_chain)->mtx) 130#define IPFW_LOCK_ASSERT(_chain) do { \ 131 mtx_assert(&(_chain)->mtx, MA_OWNED); \ 132 NET_ASSERT_GIANT(); \ 133} while (0) 134 135/* 136 * list of rules for layer 3 137 */ 138static struct ip_fw_chain layer3_chain; 139 140MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 141MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 142 143struct table_entry { 144 struct radix_node rn[2]; 145 struct sockaddr_in addr, mask; 146 u_int32_t value; 147}; 148 149#define IPFW_TABLES_MAX 128 150static struct { 151 struct radix_node_head *rnh; 152 int modified; 153} ipfw_tables[IPFW_TABLES_MAX]; 154 155static int fw_debug = 1; 156static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 157 158#ifdef SYSCTL_NODE 159SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 160SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 161 CTLFLAG_RW | CTLFLAG_SECURE3, 162 &fw_enable, 0, "Enable ipfw"); 163SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 164 &autoinc_step, 0, "Rule number autincrement step"); 165SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 166 CTLFLAG_RW | CTLFLAG_SECURE3, 167 &fw_one_pass, 0, 168 "Only do a single pass through ipfw when using dummynet(4)"); 169SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 170 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 171SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 172 CTLFLAG_RW | CTLFLAG_SECURE3, 173 &fw_verbose, 0, "Log matches to ipfw rules"); 174SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 175 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 176 177/* 178 * Description of dynamic rules. 179 * 180 * Dynamic rules are stored in lists accessed through a hash table 181 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 182 * be modified through the sysctl variable dyn_buckets which is 183 * updated when the table becomes empty. 184 * 185 * XXX currently there is only one list, ipfw_dyn. 186 * 187 * When a packet is received, its address fields are first masked 188 * with the mask defined for the rule, then hashed, then matched 189 * against the entries in the corresponding list. 190 * Dynamic rules can be used for different purposes: 191 * + stateful rules; 192 * + enforcing limits on the number of sessions; 193 * + in-kernel NAT (not implemented yet) 194 * 195 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 196 * measured in seconds and depending on the flags. 197 * 198 * The total number of dynamic rules is stored in dyn_count. 199 * The max number of dynamic rules is dyn_max. When we reach 200 * the maximum number of rules we do not create anymore. This is 201 * done to avoid consuming too much memory, but also too much 202 * time when searching on each packet (ideally, we should try instead 203 * to put a limit on the length of the list on each bucket...). 204 * 205 * Each dynamic rule holds a pointer to the parent ipfw rule so 206 * we know what action to perform. Dynamic rules are removed when 207 * the parent rule is deleted. XXX we should make them survive. 208 * 209 * There are some limitations with dynamic rules -- we do not 210 * obey the 'randomized match', and we do not do multiple 211 * passes through the firewall. XXX check the latter!!! 212 */ 213static ipfw_dyn_rule **ipfw_dyn_v = NULL; 214static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 215static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 216 217static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 218#define IPFW_DYN_LOCK_INIT() \ 219 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 220#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 221#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 222#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 223#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 224 225/* 226 * Timeouts for various events in handing dynamic rules. 227 */ 228static u_int32_t dyn_ack_lifetime = 300; 229static u_int32_t dyn_syn_lifetime = 20; 230static u_int32_t dyn_fin_lifetime = 1; 231static u_int32_t dyn_rst_lifetime = 1; 232static u_int32_t dyn_udp_lifetime = 10; 233static u_int32_t dyn_short_lifetime = 5; 234 235/* 236 * Keepalives are sent if dyn_keepalive is set. They are sent every 237 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 238 * seconds of lifetime of a rule. 239 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 240 * than dyn_keepalive_period. 241 */ 242 243static u_int32_t dyn_keepalive_interval = 20; 244static u_int32_t dyn_keepalive_period = 5; 245static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 246 247static u_int32_t static_count; /* # of static rules */ 248static u_int32_t static_len; /* size in bytes of static rules */ 249static u_int32_t dyn_count; /* # of dynamic rules */ 250static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 251 252SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 253 &dyn_buckets, 0, "Number of dyn. buckets"); 254SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 255 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 256SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 257 &dyn_count, 0, "Number of dyn. rules"); 258SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 259 &dyn_max, 0, "Max number of dyn. rules"); 260SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 261 &static_count, 0, "Number of static rules"); 262SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 263 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 264SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 265 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 266SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 267 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 268SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 269 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 270SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 271 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 272SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 273 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 274SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 275 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 276 277#endif /* SYSCTL_NODE */ 278 279 280/* 281 * This macro maps an ip pointer into a layer3 header pointer of type T 282 */ 283#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 284 285static __inline int 286icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 287{ 288 int type = L3HDR(struct icmp,ip)->icmp_type; 289 290 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 291} 292 293#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 294 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 295 296static int 297is_icmp_query(struct ip *ip) 298{ 299 int type = L3HDR(struct icmp, ip)->icmp_type; 300 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 301} 302#undef TT 303 304/* 305 * The following checks use two arrays of 8 or 16 bits to store the 306 * bits that we want set or clear, respectively. They are in the 307 * low and high half of cmd->arg1 or cmd->d[0]. 308 * 309 * We scan options and store the bits we find set. We succeed if 310 * 311 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 312 * 313 * The code is sometimes optimized not to store additional variables. 314 */ 315 316static int 317flags_match(ipfw_insn *cmd, u_int8_t bits) 318{ 319 u_char want_clear; 320 bits = ~bits; 321 322 if ( ((cmd->arg1 & 0xff) & bits) != 0) 323 return 0; /* some bits we want set were clear */ 324 want_clear = (cmd->arg1 >> 8) & 0xff; 325 if ( (want_clear & bits) != want_clear) 326 return 0; /* some bits we want clear were set */ 327 return 1; 328} 329 330static int 331ipopts_match(struct ip *ip, ipfw_insn *cmd) 332{ 333 int optlen, bits = 0; 334 u_char *cp = (u_char *)(ip + 1); 335 int x = (ip->ip_hl << 2) - sizeof (struct ip); 336 337 for (; x > 0; x -= optlen, cp += optlen) { 338 int opt = cp[IPOPT_OPTVAL]; 339 340 if (opt == IPOPT_EOL) 341 break; 342 if (opt == IPOPT_NOP) 343 optlen = 1; 344 else { 345 optlen = cp[IPOPT_OLEN]; 346 if (optlen <= 0 || optlen > x) 347 return 0; /* invalid or truncated */ 348 } 349 switch (opt) { 350 351 default: 352 break; 353 354 case IPOPT_LSRR: 355 bits |= IP_FW_IPOPT_LSRR; 356 break; 357 358 case IPOPT_SSRR: 359 bits |= IP_FW_IPOPT_SSRR; 360 break; 361 362 case IPOPT_RR: 363 bits |= IP_FW_IPOPT_RR; 364 break; 365 366 case IPOPT_TS: 367 bits |= IP_FW_IPOPT_TS; 368 break; 369 } 370 } 371 return (flags_match(cmd, bits)); 372} 373 374static int 375tcpopts_match(struct ip *ip, ipfw_insn *cmd) 376{ 377 int optlen, bits = 0; 378 struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 379 u_char *cp = (u_char *)(tcp + 1); 380 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 381 382 for (; x > 0; x -= optlen, cp += optlen) { 383 int opt = cp[0]; 384 if (opt == TCPOPT_EOL) 385 break; 386 if (opt == TCPOPT_NOP) 387 optlen = 1; 388 else { 389 optlen = cp[1]; 390 if (optlen <= 0) 391 break; 392 } 393 394 switch (opt) { 395 396 default: 397 break; 398 399 case TCPOPT_MAXSEG: 400 bits |= IP_FW_TCPOPT_MSS; 401 break; 402 403 case TCPOPT_WINDOW: 404 bits |= IP_FW_TCPOPT_WINDOW; 405 break; 406 407 case TCPOPT_SACK_PERMITTED: 408 case TCPOPT_SACK: 409 bits |= IP_FW_TCPOPT_SACK; 410 break; 411 412 case TCPOPT_TIMESTAMP: 413 bits |= IP_FW_TCPOPT_TS; 414 break; 415 416 case TCPOPT_CC: 417 case TCPOPT_CCNEW: 418 case TCPOPT_CCECHO: 419 bits |= IP_FW_TCPOPT_CC; 420 break; 421 } 422 } 423 return (flags_match(cmd, bits)); 424} 425 426static int 427iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 428{ 429 if (ifp == NULL) /* no iface with this packet, match fails */ 430 return 0; 431 /* Check by name or by IP address */ 432 if (cmd->name[0] != '\0') { /* match by name */ 433 /* Check name */ 434 if (cmd->p.glob) { 435 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 436 return(1); 437 } else { 438 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 439 return(1); 440 } 441 } else { 442 struct ifaddr *ia; 443 444 /* XXX lock? */ 445 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 446 if (ia->ifa_addr == NULL) 447 continue; 448 if (ia->ifa_addr->sa_family != AF_INET) 449 continue; 450 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 451 (ia->ifa_addr))->sin_addr.s_addr) 452 return(1); /* match */ 453 } 454 } 455 return(0); /* no match, fail ... */ 456} 457 458/* 459 * The verify_path function checks if a route to the src exists and 460 * if it is reachable via ifp (when provided). 461 * 462 * The 'verrevpath' option checks that the interface that an IP packet 463 * arrives on is the same interface that traffic destined for the 464 * packet's source address would be routed out of. The 'versrcreach' 465 * option just checks that the source address is reachable via any route 466 * (except default) in the routing table. These two are a measure to block 467 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 468 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 469 * is purposely reminiscent of the Cisco IOS command, 470 * 471 * ip verify unicast reverse-path 472 * ip verify unicast source reachable-via any 473 * 474 * which implements the same functionality. But note that syntax is 475 * misleading. The check may be performed on all IP packets whether unicast, 476 * multicast, or broadcast. 477 */ 478static int 479verify_path(struct in_addr src, struct ifnet *ifp) 480{ 481 struct route ro; 482 struct sockaddr_in *dst; 483 484 bzero(&ro, sizeof(ro)); 485 486 dst = (struct sockaddr_in *)&(ro.ro_dst); 487 dst->sin_family = AF_INET; 488 dst->sin_len = sizeof(*dst); 489 dst->sin_addr = src; 490 rtalloc_ign(&ro, RTF_CLONING); 491 492 if (ro.ro_rt == NULL) 493 return 0; 494 495 /* if ifp is provided, check for equality with rtentry */ 496 if (ifp != NULL && ro.ro_rt->rt_ifp != ifp) { 497 RTFREE(ro.ro_rt); 498 return 0; 499 } 500 501 /* if no ifp provided, check if rtentry is not default route */ 502 if (ifp == NULL && 503 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 504 RTFREE(ro.ro_rt); 505 return 0; 506 } 507 508 /* or if this is a blackhole/reject route */ 509 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 510 RTFREE(ro.ro_rt); 511 return 0; 512 } 513 514 /* found valid route */ 515 RTFREE(ro.ro_rt); 516 return 1; 517} 518 519 520static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 521 522#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 523#define SNP(buf) buf, sizeof(buf) 524 525/* 526 * We enter here when we have a rule with O_LOG. 527 * XXX this function alone takes about 2Kbytes of code! 528 */ 529static void 530ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 531 struct mbuf *m, struct ifnet *oif) 532{ 533 char *action; 534 int limit_reached = 0; 535 char action2[40], proto[48], fragment[28]; 536 537 fragment[0] = '\0'; 538 proto[0] = '\0'; 539 540 if (f == NULL) { /* bogus pkt */ 541 if (verbose_limit != 0 && norule_counter >= verbose_limit) 542 return; 543 norule_counter++; 544 if (norule_counter == verbose_limit) 545 limit_reached = verbose_limit; 546 action = "Refuse"; 547 } else { /* O_LOG is the first action, find the real one */ 548 ipfw_insn *cmd = ACTION_PTR(f); 549 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 550 551 if (l->max_log != 0 && l->log_left == 0) 552 return; 553 l->log_left--; 554 if (l->log_left == 0) 555 limit_reached = l->max_log; 556 cmd += F_LEN(cmd); /* point to first action */ 557 if (cmd->opcode == O_ALTQ) { 558 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 559 560 snprintf(SNPARGS(action2, 0), "Altq %d", 561 altq->qid); 562 cmd += F_LEN(cmd); 563 } 564 if (cmd->opcode == O_PROB) 565 cmd += F_LEN(cmd); 566 567 action = action2; 568 switch (cmd->opcode) { 569 case O_DENY: 570 action = "Deny"; 571 break; 572 573 case O_REJECT: 574 if (cmd->arg1==ICMP_REJECT_RST) 575 action = "Reset"; 576 else if (cmd->arg1==ICMP_UNREACH_HOST) 577 action = "Reject"; 578 else 579 snprintf(SNPARGS(action2, 0), "Unreach %d", 580 cmd->arg1); 581 break; 582 583 case O_ACCEPT: 584 action = "Accept"; 585 break; 586 case O_COUNT: 587 action = "Count"; 588 break; 589 case O_DIVERT: 590 snprintf(SNPARGS(action2, 0), "Divert %d", 591 cmd->arg1); 592 break; 593 case O_TEE: 594 snprintf(SNPARGS(action2, 0), "Tee %d", 595 cmd->arg1); 596 break; 597 case O_SKIPTO: 598 snprintf(SNPARGS(action2, 0), "SkipTo %d", 599 cmd->arg1); 600 break; 601 case O_PIPE: 602 snprintf(SNPARGS(action2, 0), "Pipe %d", 603 cmd->arg1); 604 break; 605 case O_QUEUE: 606 snprintf(SNPARGS(action2, 0), "Queue %d", 607 cmd->arg1); 608 break; 609 case O_FORWARD_IP: { 610 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 611 int len; 612 613 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 614 inet_ntoa(sa->sa.sin_addr)); 615 if (sa->sa.sin_port) 616 snprintf(SNPARGS(action2, len), ":%d", 617 sa->sa.sin_port); 618 } 619 break; 620 default: 621 action = "UNKNOWN"; 622 break; 623 } 624 } 625 626 if (hlen == 0) { /* non-ip */ 627 snprintf(SNPARGS(proto, 0), "MAC"); 628 } else { 629 struct ip *ip = mtod(m, struct ip *); 630 /* these three are all aliases to the same thing */ 631 struct icmp *const icmp = L3HDR(struct icmp, ip); 632 struct tcphdr *const tcp = (struct tcphdr *)icmp; 633 struct udphdr *const udp = (struct udphdr *)icmp; 634 635 int ip_off, offset, ip_len; 636 637 int len; 638 639 if (eh != NULL) { /* layer 2 packets are as on the wire */ 640 ip_off = ntohs(ip->ip_off); 641 ip_len = ntohs(ip->ip_len); 642 } else { 643 ip_off = ip->ip_off; 644 ip_len = ip->ip_len; 645 } 646 offset = ip_off & IP_OFFMASK; 647 switch (ip->ip_p) { 648 case IPPROTO_TCP: 649 len = snprintf(SNPARGS(proto, 0), "TCP %s", 650 inet_ntoa(ip->ip_src)); 651 if (offset == 0) 652 snprintf(SNPARGS(proto, len), ":%d %s:%d", 653 ntohs(tcp->th_sport), 654 inet_ntoa(ip->ip_dst), 655 ntohs(tcp->th_dport)); 656 else 657 snprintf(SNPARGS(proto, len), " %s", 658 inet_ntoa(ip->ip_dst)); 659 break; 660 661 case IPPROTO_UDP: 662 len = snprintf(SNPARGS(proto, 0), "UDP %s", 663 inet_ntoa(ip->ip_src)); 664 if (offset == 0) 665 snprintf(SNPARGS(proto, len), ":%d %s:%d", 666 ntohs(udp->uh_sport), 667 inet_ntoa(ip->ip_dst), 668 ntohs(udp->uh_dport)); 669 else 670 snprintf(SNPARGS(proto, len), " %s", 671 inet_ntoa(ip->ip_dst)); 672 break; 673 674 case IPPROTO_ICMP: 675 if (offset == 0) 676 len = snprintf(SNPARGS(proto, 0), 677 "ICMP:%u.%u ", 678 icmp->icmp_type, icmp->icmp_code); 679 else 680 len = snprintf(SNPARGS(proto, 0), "ICMP "); 681 len += snprintf(SNPARGS(proto, len), "%s", 682 inet_ntoa(ip->ip_src)); 683 snprintf(SNPARGS(proto, len), " %s", 684 inet_ntoa(ip->ip_dst)); 685 break; 686 687 default: 688 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 689 inet_ntoa(ip->ip_src)); 690 snprintf(SNPARGS(proto, len), " %s", 691 inet_ntoa(ip->ip_dst)); 692 break; 693 } 694 695 if (ip_off & (IP_MF | IP_OFFMASK)) 696 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 697 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 698 offset << 3, 699 (ip_off & IP_MF) ? "+" : ""); 700 } 701 if (oif || m->m_pkthdr.rcvif) 702 log(LOG_SECURITY | LOG_INFO, 703 "ipfw: %d %s %s %s via %s%s\n", 704 f ? f->rulenum : -1, 705 action, proto, oif ? "out" : "in", 706 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 707 fragment); 708 else 709 log(LOG_SECURITY | LOG_INFO, 710 "ipfw: %d %s %s [no if info]%s\n", 711 f ? f->rulenum : -1, 712 action, proto, fragment); 713 if (limit_reached) 714 log(LOG_SECURITY | LOG_NOTICE, 715 "ipfw: limit %d reached on entry %d\n", 716 limit_reached, f ? f->rulenum : -1); 717} 718 719/* 720 * IMPORTANT: the hash function for dynamic rules must be commutative 721 * in source and destination (ip,port), because rules are bidirectional 722 * and we want to find both in the same bucket. 723 */ 724static __inline int 725hash_packet(struct ipfw_flow_id *id) 726{ 727 u_int32_t i; 728 729 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 730 i &= (curr_dyn_buckets - 1); 731 return i; 732} 733 734/** 735 * unlink a dynamic rule from a chain. prev is a pointer to 736 * the previous one, q is a pointer to the rule to delete, 737 * head is a pointer to the head of the queue. 738 * Modifies q and potentially also head. 739 */ 740#define UNLINK_DYN_RULE(prev, head, q) { \ 741 ipfw_dyn_rule *old_q = q; \ 742 \ 743 /* remove a refcount to the parent */ \ 744 if (q->dyn_type == O_LIMIT) \ 745 q->parent->count--; \ 746 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 747 (q->id.src_ip), (q->id.src_port), \ 748 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 749 if (prev != NULL) \ 750 prev->next = q = q->next; \ 751 else \ 752 head = q = q->next; \ 753 dyn_count--; \ 754 free(old_q, M_IPFW); } 755 756#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 757 758/** 759 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 760 * 761 * If keep_me == NULL, rules are deleted even if not expired, 762 * otherwise only expired rules are removed. 763 * 764 * The value of the second parameter is also used to point to identify 765 * a rule we absolutely do not want to remove (e.g. because we are 766 * holding a reference to it -- this is the case with O_LIMIT_PARENT 767 * rules). The pointer is only used for comparison, so any non-null 768 * value will do. 769 */ 770static void 771remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 772{ 773 static u_int32_t last_remove = 0; 774 775#define FORCE (keep_me == NULL) 776 777 ipfw_dyn_rule *prev, *q; 778 int i, pass = 0, max_pass = 0; 779 780 IPFW_DYN_LOCK_ASSERT(); 781 782 if (ipfw_dyn_v == NULL || dyn_count == 0) 783 return; 784 /* do not expire more than once per second, it is useless */ 785 if (!FORCE && last_remove == time_second) 786 return; 787 last_remove = time_second; 788 789 /* 790 * because O_LIMIT refer to parent rules, during the first pass only 791 * remove child and mark any pending LIMIT_PARENT, and remove 792 * them in a second pass. 793 */ 794next_pass: 795 for (i = 0 ; i < curr_dyn_buckets ; i++) { 796 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 797 /* 798 * Logic can become complex here, so we split tests. 799 */ 800 if (q == keep_me) 801 goto next; 802 if (rule != NULL && rule != q->rule) 803 goto next; /* not the one we are looking for */ 804 if (q->dyn_type == O_LIMIT_PARENT) { 805 /* 806 * handle parent in the second pass, 807 * record we need one. 808 */ 809 max_pass = 1; 810 if (pass == 0) 811 goto next; 812 if (FORCE && q->count != 0 ) { 813 /* XXX should not happen! */ 814 printf("ipfw: OUCH! cannot remove rule," 815 " count %d\n", q->count); 816 } 817 } else { 818 if (!FORCE && 819 !TIME_LEQ( q->expire, time_second )) 820 goto next; 821 } 822 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 823 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 824 continue; 825 } 826next: 827 prev=q; 828 q=q->next; 829 } 830 } 831 if (pass++ < max_pass) 832 goto next_pass; 833} 834 835 836/** 837 * lookup a dynamic rule. 838 */ 839static ipfw_dyn_rule * 840lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 841 struct tcphdr *tcp) 842{ 843 /* 844 * stateful ipfw extensions. 845 * Lookup into dynamic session queue 846 */ 847#define MATCH_REVERSE 0 848#define MATCH_FORWARD 1 849#define MATCH_NONE 2 850#define MATCH_UNKNOWN 3 851 int i, dir = MATCH_NONE; 852 ipfw_dyn_rule *prev, *q=NULL; 853 854 IPFW_DYN_LOCK_ASSERT(); 855 856 if (ipfw_dyn_v == NULL) 857 goto done; /* not found */ 858 i = hash_packet( pkt ); 859 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 860 if (q->dyn_type == O_LIMIT_PARENT && q->count) 861 goto next; 862 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 863 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 864 continue; 865 } 866 if (pkt->proto == q->id.proto && 867 q->dyn_type != O_LIMIT_PARENT) { 868 if (pkt->src_ip == q->id.src_ip && 869 pkt->dst_ip == q->id.dst_ip && 870 pkt->src_port == q->id.src_port && 871 pkt->dst_port == q->id.dst_port ) { 872 dir = MATCH_FORWARD; 873 break; 874 } 875 if (pkt->src_ip == q->id.dst_ip && 876 pkt->dst_ip == q->id.src_ip && 877 pkt->src_port == q->id.dst_port && 878 pkt->dst_port == q->id.src_port ) { 879 dir = MATCH_REVERSE; 880 break; 881 } 882 } 883next: 884 prev = q; 885 q = q->next; 886 } 887 if (q == NULL) 888 goto done; /* q = NULL, not found */ 889 890 if ( prev != NULL) { /* found and not in front */ 891 prev->next = q->next; 892 q->next = ipfw_dyn_v[i]; 893 ipfw_dyn_v[i] = q; 894 } 895 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 896 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 897 898#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 899#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 900 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 901 switch (q->state) { 902 case TH_SYN: /* opening */ 903 q->expire = time_second + dyn_syn_lifetime; 904 break; 905 906 case BOTH_SYN: /* move to established */ 907 case BOTH_SYN | TH_FIN : /* one side tries to close */ 908 case BOTH_SYN | (TH_FIN << 8) : 909 if (tcp) { 910#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 911 u_int32_t ack = ntohl(tcp->th_ack); 912 if (dir == MATCH_FORWARD) { 913 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 914 q->ack_fwd = ack; 915 else { /* ignore out-of-sequence */ 916 break; 917 } 918 } else { 919 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 920 q->ack_rev = ack; 921 else { /* ignore out-of-sequence */ 922 break; 923 } 924 } 925 } 926 q->expire = time_second + dyn_ack_lifetime; 927 break; 928 929 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 930 if (dyn_fin_lifetime >= dyn_keepalive_period) 931 dyn_fin_lifetime = dyn_keepalive_period - 1; 932 q->expire = time_second + dyn_fin_lifetime; 933 break; 934 935 default: 936#if 0 937 /* 938 * reset or some invalid combination, but can also 939 * occur if we use keep-state the wrong way. 940 */ 941 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 942 printf("invalid state: 0x%x\n", q->state); 943#endif 944 if (dyn_rst_lifetime >= dyn_keepalive_period) 945 dyn_rst_lifetime = dyn_keepalive_period - 1; 946 q->expire = time_second + dyn_rst_lifetime; 947 break; 948 } 949 } else if (pkt->proto == IPPROTO_UDP) { 950 q->expire = time_second + dyn_udp_lifetime; 951 } else { 952 /* other protocols */ 953 q->expire = time_second + dyn_short_lifetime; 954 } 955done: 956 if (match_direction) 957 *match_direction = dir; 958 return q; 959} 960 961static ipfw_dyn_rule * 962lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 963 struct tcphdr *tcp) 964{ 965 ipfw_dyn_rule *q; 966 967 IPFW_DYN_LOCK(); 968 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 969 if (q == NULL) 970 IPFW_DYN_UNLOCK(); 971 /* NB: return table locked when q is not NULL */ 972 return q; 973} 974 975static void 976realloc_dynamic_table(void) 977{ 978 IPFW_DYN_LOCK_ASSERT(); 979 980 /* 981 * Try reallocation, make sure we have a power of 2 and do 982 * not allow more than 64k entries. In case of overflow, 983 * default to 1024. 984 */ 985 986 if (dyn_buckets > 65536) 987 dyn_buckets = 1024; 988 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 989 dyn_buckets = curr_dyn_buckets; /* reset */ 990 return; 991 } 992 curr_dyn_buckets = dyn_buckets; 993 if (ipfw_dyn_v != NULL) 994 free(ipfw_dyn_v, M_IPFW); 995 for (;;) { 996 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 997 M_IPFW, M_NOWAIT | M_ZERO); 998 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 999 break; 1000 curr_dyn_buckets /= 2; 1001 } 1002} 1003 1004/** 1005 * Install state of type 'type' for a dynamic session. 1006 * The hash table contains two type of rules: 1007 * - regular rules (O_KEEP_STATE) 1008 * - rules for sessions with limited number of sess per user 1009 * (O_LIMIT). When they are created, the parent is 1010 * increased by 1, and decreased on delete. In this case, 1011 * the third parameter is the parent rule and not the chain. 1012 * - "parent" rules for the above (O_LIMIT_PARENT). 1013 */ 1014static ipfw_dyn_rule * 1015add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1016{ 1017 ipfw_dyn_rule *r; 1018 int i; 1019 1020 IPFW_DYN_LOCK_ASSERT(); 1021 1022 if (ipfw_dyn_v == NULL || 1023 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1024 realloc_dynamic_table(); 1025 if (ipfw_dyn_v == NULL) 1026 return NULL; /* failed ! */ 1027 } 1028 i = hash_packet(id); 1029 1030 r = malloc(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO); 1031 if (r == NULL) { 1032 printf ("ipfw: sorry cannot allocate state\n"); 1033 return NULL; 1034 } 1035 1036 /* increase refcount on parent, and set pointer */ 1037 if (dyn_type == O_LIMIT) { 1038 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1039 if ( parent->dyn_type != O_LIMIT_PARENT) 1040 panic("invalid parent"); 1041 parent->count++; 1042 r->parent = parent; 1043 rule = parent->rule; 1044 } 1045 1046 r->id = *id; 1047 r->expire = time_second + dyn_syn_lifetime; 1048 r->rule = rule; 1049 r->dyn_type = dyn_type; 1050 r->pcnt = r->bcnt = 0; 1051 r->count = 0; 1052 1053 r->bucket = i; 1054 r->next = ipfw_dyn_v[i]; 1055 ipfw_dyn_v[i] = r; 1056 dyn_count++; 1057 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1058 dyn_type, 1059 (r->id.src_ip), (r->id.src_port), 1060 (r->id.dst_ip), (r->id.dst_port), 1061 dyn_count ); ) 1062 return r; 1063} 1064 1065/** 1066 * lookup dynamic parent rule using pkt and rule as search keys. 1067 * If the lookup fails, then install one. 1068 */ 1069static ipfw_dyn_rule * 1070lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1071{ 1072 ipfw_dyn_rule *q; 1073 int i; 1074 1075 IPFW_DYN_LOCK_ASSERT(); 1076 1077 if (ipfw_dyn_v) { 1078 i = hash_packet( pkt ); 1079 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1080 if (q->dyn_type == O_LIMIT_PARENT && 1081 rule== q->rule && 1082 pkt->proto == q->id.proto && 1083 pkt->src_ip == q->id.src_ip && 1084 pkt->dst_ip == q->id.dst_ip && 1085 pkt->src_port == q->id.src_port && 1086 pkt->dst_port == q->id.dst_port) { 1087 q->expire = time_second + dyn_short_lifetime; 1088 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1089 return q; 1090 } 1091 } 1092 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1093} 1094 1095/** 1096 * Install dynamic state for rule type cmd->o.opcode 1097 * 1098 * Returns 1 (failure) if state is not installed because of errors or because 1099 * session limitations are enforced. 1100 */ 1101static int 1102install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1103 struct ip_fw_args *args) 1104{ 1105 static int last_log; 1106 1107 ipfw_dyn_rule *q; 1108 1109 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 1110 cmd->o.opcode, 1111 (args->f_id.src_ip), (args->f_id.src_port), 1112 (args->f_id.dst_ip), (args->f_id.dst_port) );) 1113 1114 IPFW_DYN_LOCK(); 1115 1116 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1117 1118 if (q != NULL) { /* should never occur */ 1119 if (last_log != time_second) { 1120 last_log = time_second; 1121 printf("ipfw: install_state: entry already present, done\n"); 1122 } 1123 IPFW_DYN_UNLOCK(); 1124 return 0; 1125 } 1126 1127 if (dyn_count >= dyn_max) 1128 /* 1129 * Run out of slots, try to remove any expired rule. 1130 */ 1131 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1132 1133 if (dyn_count >= dyn_max) { 1134 if (last_log != time_second) { 1135 last_log = time_second; 1136 printf("ipfw: install_state: Too many dynamic rules\n"); 1137 } 1138 IPFW_DYN_UNLOCK(); 1139 return 1; /* cannot install, notify caller */ 1140 } 1141 1142 switch (cmd->o.opcode) { 1143 case O_KEEP_STATE: /* bidir rule */ 1144 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1145 break; 1146 1147 case O_LIMIT: /* limit number of sessions */ 1148 { 1149 u_int16_t limit_mask = cmd->limit_mask; 1150 struct ipfw_flow_id id; 1151 ipfw_dyn_rule *parent; 1152 1153 DEB(printf("ipfw: installing dyn-limit rule %d\n", 1154 cmd->conn_limit);) 1155 1156 id.dst_ip = id.src_ip = 0; 1157 id.dst_port = id.src_port = 0; 1158 id.proto = args->f_id.proto; 1159 1160 if (limit_mask & DYN_SRC_ADDR) 1161 id.src_ip = args->f_id.src_ip; 1162 if (limit_mask & DYN_DST_ADDR) 1163 id.dst_ip = args->f_id.dst_ip; 1164 if (limit_mask & DYN_SRC_PORT) 1165 id.src_port = args->f_id.src_port; 1166 if (limit_mask & DYN_DST_PORT) 1167 id.dst_port = args->f_id.dst_port; 1168 parent = lookup_dyn_parent(&id, rule); 1169 if (parent == NULL) { 1170 printf("ipfw: add parent failed\n"); 1171 return 1; 1172 } 1173 if (parent->count >= cmd->conn_limit) { 1174 /* 1175 * See if we can remove some expired rule. 1176 */ 1177 remove_dyn_rule(rule, parent); 1178 if (parent->count >= cmd->conn_limit) { 1179 if (fw_verbose && last_log != time_second) { 1180 last_log = time_second; 1181 log(LOG_SECURITY | LOG_DEBUG, 1182 "drop session, too many entries\n"); 1183 } 1184 IPFW_DYN_UNLOCK(); 1185 return 1; 1186 } 1187 } 1188 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1189 } 1190 break; 1191 default: 1192 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 1193 IPFW_DYN_UNLOCK(); 1194 return 1; 1195 } 1196 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1197 IPFW_DYN_UNLOCK(); 1198 return 0; 1199} 1200 1201/* 1202 * Transmit a TCP packet, containing either a RST or a keepalive. 1203 * When flags & TH_RST, we are sending a RST packet, because of a 1204 * "reset" action matched the packet. 1205 * Otherwise we are sending a keepalive, and flags & TH_ 1206 */ 1207static void 1208send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1209{ 1210 struct mbuf *m; 1211 struct ip *ip; 1212 struct tcphdr *tcp; 1213 1214 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1215 if (m == 0) 1216 return; 1217 m->m_pkthdr.rcvif = (struct ifnet *)0; 1218 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1219 m->m_data += max_linkhdr; 1220 1221 ip = mtod(m, struct ip *); 1222 bzero(ip, m->m_len); 1223 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1224 ip->ip_p = IPPROTO_TCP; 1225 tcp->th_off = 5; 1226 /* 1227 * Assume we are sending a RST (or a keepalive in the reverse 1228 * direction), swap src and destination addresses and ports. 1229 */ 1230 ip->ip_src.s_addr = htonl(id->dst_ip); 1231 ip->ip_dst.s_addr = htonl(id->src_ip); 1232 tcp->th_sport = htons(id->dst_port); 1233 tcp->th_dport = htons(id->src_port); 1234 if (flags & TH_RST) { /* we are sending a RST */ 1235 if (flags & TH_ACK) { 1236 tcp->th_seq = htonl(ack); 1237 tcp->th_ack = htonl(0); 1238 tcp->th_flags = TH_RST; 1239 } else { 1240 if (flags & TH_SYN) 1241 seq++; 1242 tcp->th_seq = htonl(0); 1243 tcp->th_ack = htonl(seq); 1244 tcp->th_flags = TH_RST | TH_ACK; 1245 } 1246 } else { 1247 /* 1248 * We are sending a keepalive. flags & TH_SYN determines 1249 * the direction, forward if set, reverse if clear. 1250 * NOTE: seq and ack are always assumed to be correct 1251 * as set by the caller. This may be confusing... 1252 */ 1253 if (flags & TH_SYN) { 1254 /* 1255 * we have to rewrite the correct addresses! 1256 */ 1257 ip->ip_dst.s_addr = htonl(id->dst_ip); 1258 ip->ip_src.s_addr = htonl(id->src_ip); 1259 tcp->th_dport = htons(id->dst_port); 1260 tcp->th_sport = htons(id->src_port); 1261 } 1262 tcp->th_seq = htonl(seq); 1263 tcp->th_ack = htonl(ack); 1264 tcp->th_flags = TH_ACK; 1265 } 1266 /* 1267 * set ip_len to the payload size so we can compute 1268 * the tcp checksum on the pseudoheader 1269 * XXX check this, could save a couple of words ? 1270 */ 1271 ip->ip_len = htons(sizeof(struct tcphdr)); 1272 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1273 /* 1274 * now fill fields left out earlier 1275 */ 1276 ip->ip_ttl = ip_defttl; 1277 ip->ip_len = m->m_pkthdr.len; 1278 m->m_flags |= M_SKIP_FIREWALL; 1279 ip_output(m, NULL, NULL, 0, NULL, NULL); 1280} 1281 1282/* 1283 * sends a reject message, consuming the mbuf passed as an argument. 1284 */ 1285static void 1286send_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 1287{ 1288 1289 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1290 /* We need the IP header in host order for icmp_error(). */ 1291 if (args->eh != NULL) { 1292 struct ip *ip = mtod(args->m, struct ip *); 1293 ip->ip_len = ntohs(ip->ip_len); 1294 ip->ip_off = ntohs(ip->ip_off); 1295 } 1296 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1297 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1298 struct tcphdr *const tcp = 1299 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1300 if ( (tcp->th_flags & TH_RST) == 0) 1301 send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1302 ntohl(tcp->th_ack), 1303 tcp->th_flags | TH_RST); 1304 m_freem(args->m); 1305 } else 1306 m_freem(args->m); 1307 args->m = NULL; 1308} 1309 1310/** 1311 * 1312 * Given an ip_fw *, lookup_next_rule will return a pointer 1313 * to the next rule, which can be either the jump 1314 * target (for skipto instructions) or the next one in the list (in 1315 * all other cases including a missing jump target). 1316 * The result is also written in the "next_rule" field of the rule. 1317 * Backward jumps are not allowed, so start looking from the next 1318 * rule... 1319 * 1320 * This never returns NULL -- in case we do not have an exact match, 1321 * the next rule is returned. When the ruleset is changed, 1322 * pointers are flushed so we are always correct. 1323 */ 1324 1325static struct ip_fw * 1326lookup_next_rule(struct ip_fw *me) 1327{ 1328 struct ip_fw *rule = NULL; 1329 ipfw_insn *cmd; 1330 1331 /* look for action, in case it is a skipto */ 1332 cmd = ACTION_PTR(me); 1333 if (cmd->opcode == O_LOG) 1334 cmd += F_LEN(cmd); 1335 if (cmd->opcode == O_ALTQ) 1336 cmd += F_LEN(cmd); 1337 if ( cmd->opcode == O_SKIPTO ) 1338 for (rule = me->next; rule ; rule = rule->next) 1339 if (rule->rulenum >= cmd->arg1) 1340 break; 1341 if (rule == NULL) /* failure or not a skipto */ 1342 rule = me->next; 1343 me->next_rule = rule; 1344 return rule; 1345} 1346 1347static void 1348init_tables(void) 1349{ 1350 int i; 1351 1352 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1353 rn_inithead((void **)&ipfw_tables[i].rnh, 32); 1354 ipfw_tables[i].modified = 1; 1355 } 1356} 1357 1358static int 1359add_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen, u_int32_t value) 1360{ 1361 struct radix_node_head *rnh; 1362 struct table_entry *ent; 1363 1364 if (tbl >= IPFW_TABLES_MAX) 1365 return (EINVAL); 1366 rnh = ipfw_tables[tbl].rnh; 1367 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1368 if (ent == NULL) 1369 return (ENOMEM); 1370 ent->value = value; 1371 ent->addr.sin_len = ent->mask.sin_len = 8; 1372 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1373 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1374 RADIX_NODE_HEAD_LOCK(rnh); 1375 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1376 NULL) { 1377 RADIX_NODE_HEAD_UNLOCK(rnh); 1378 free(ent, M_IPFW_TBL); 1379 return (EEXIST); 1380 } 1381 ipfw_tables[tbl].modified = 1; 1382 RADIX_NODE_HEAD_UNLOCK(rnh); 1383 return (0); 1384} 1385 1386static int 1387del_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen) 1388{ 1389 struct radix_node_head *rnh; 1390 struct table_entry *ent; 1391 struct sockaddr_in sa, mask; 1392 1393 if (tbl >= IPFW_TABLES_MAX) 1394 return (EINVAL); 1395 rnh = ipfw_tables[tbl].rnh; 1396 sa.sin_len = mask.sin_len = 8; 1397 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1398 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1399 RADIX_NODE_HEAD_LOCK(rnh); 1400 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1401 if (ent == NULL) { 1402 RADIX_NODE_HEAD_UNLOCK(rnh); 1403 return (ESRCH); 1404 } 1405 ipfw_tables[tbl].modified = 1; 1406 RADIX_NODE_HEAD_UNLOCK(rnh); 1407 free(ent, M_IPFW_TBL); 1408 return (0); 1409} 1410 1411static int 1412flush_table_entry(struct radix_node *rn, void *arg) 1413{ 1414 struct radix_node_head * const rnh = arg; 1415 struct table_entry *ent; 1416 1417 ent = (struct table_entry *) 1418 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1419 if (ent != NULL) 1420 free(ent, M_IPFW_TBL); 1421 return (0); 1422} 1423 1424static int 1425flush_table(u_int16_t tbl) 1426{ 1427 struct radix_node_head *rnh; 1428 1429 if (tbl >= IPFW_TABLES_MAX) 1430 return (EINVAL); 1431 rnh = ipfw_tables[tbl].rnh; 1432 RADIX_NODE_HEAD_LOCK(rnh); 1433 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1434 ipfw_tables[tbl].modified = 1; 1435 RADIX_NODE_HEAD_UNLOCK(rnh); 1436 return (0); 1437} 1438 1439static void 1440flush_tables(void) 1441{ 1442 u_int16_t tbl; 1443 1444 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1445 flush_table(tbl); 1446} 1447 1448static int 1449lookup_table(u_int16_t tbl, in_addr_t addr, u_int32_t *val) 1450{ 1451 struct radix_node_head *rnh; 1452 struct table_entry *ent; 1453 struct sockaddr_in sa; 1454 static in_addr_t last_addr; 1455 static int last_tbl; 1456 static int last_match; 1457 static u_int32_t last_value; 1458 1459 if (tbl >= IPFW_TABLES_MAX) 1460 return (0); 1461 if (tbl == last_tbl && addr == last_addr && 1462 !ipfw_tables[tbl].modified) { 1463 if (last_match) 1464 *val = last_value; 1465 return (last_match); 1466 } 1467 rnh = ipfw_tables[tbl].rnh; 1468 sa.sin_len = 8; 1469 sa.sin_addr.s_addr = addr; 1470 RADIX_NODE_HEAD_LOCK(rnh); 1471 ipfw_tables[tbl].modified = 0; 1472 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1473 RADIX_NODE_HEAD_UNLOCK(rnh); 1474 last_addr = addr; 1475 last_tbl = tbl; 1476 if (ent != NULL) { 1477 last_value = *val = ent->value; 1478 last_match = 1; 1479 return (1); 1480 } 1481 last_match = 0; 1482 return (0); 1483} 1484 1485static int 1486count_table_entry(struct radix_node *rn, void *arg) 1487{ 1488 u_int32_t * const cnt = arg; 1489 1490 (*cnt)++; 1491 return (0); 1492} 1493 1494static int 1495count_table(u_int32_t tbl, u_int32_t *cnt) 1496{ 1497 struct radix_node_head *rnh; 1498 1499 if (tbl >= IPFW_TABLES_MAX) 1500 return (EINVAL); 1501 rnh = ipfw_tables[tbl].rnh; 1502 *cnt = 0; 1503 RADIX_NODE_HEAD_LOCK(rnh); 1504 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1505 RADIX_NODE_HEAD_UNLOCK(rnh); 1506 return (0); 1507} 1508 1509static int 1510dump_table_entry(struct radix_node *rn, void *arg) 1511{ 1512 struct table_entry * const n = (struct table_entry *)rn; 1513 ipfw_table * const tbl = arg; 1514 ipfw_table_entry *ent; 1515 1516 if (tbl->cnt == tbl->size) 1517 return (1); 1518 ent = &tbl->ent[tbl->cnt]; 1519 ent->tbl = tbl->tbl; 1520 if (in_nullhost(n->mask.sin_addr)) 1521 ent->masklen = 0; 1522 else 1523 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1524 ent->addr = n->addr.sin_addr.s_addr; 1525 ent->value = n->value; 1526 tbl->cnt++; 1527 return (0); 1528} 1529 1530static int 1531dump_table(ipfw_table *tbl) 1532{ 1533 struct radix_node_head *rnh; 1534 1535 if (tbl->tbl >= IPFW_TABLES_MAX) 1536 return (EINVAL); 1537 rnh = ipfw_tables[tbl->tbl].rnh; 1538 tbl->cnt = 0; 1539 RADIX_NODE_HEAD_LOCK(rnh); 1540 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1541 RADIX_NODE_HEAD_UNLOCK(rnh); 1542 return (0); 1543} 1544 1545static void 1546fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1547{ 1548 struct ucred *cr; 1549 1550 if (inp->inp_socket != NULL) { 1551 cr = inp->inp_socket->so_cred; 1552 ugp->fw_prid = jailed(cr) ? 1553 cr->cr_prison->pr_id : -1; 1554 ugp->fw_uid = cr->cr_uid; 1555 ugp->fw_ngroups = cr->cr_ngroups; 1556 bcopy(cr->cr_groups, ugp->fw_groups, 1557 sizeof(ugp->fw_groups)); 1558 } 1559} 1560 1561static int 1562check_uidgid(ipfw_insn_u32 *insn, 1563 int proto, struct ifnet *oif, 1564 struct in_addr dst_ip, u_int16_t dst_port, 1565 struct in_addr src_ip, u_int16_t src_port, 1566 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1567{ 1568 struct inpcbinfo *pi; 1569 int wildcard; 1570 struct inpcb *pcb; 1571 int match; 1572 gid_t *gp; 1573 1574 /* 1575 * Check to see if the UDP or TCP stack supplied us with 1576 * the PCB. If so, rather then holding a lock and looking 1577 * up the PCB, we can use the one that was supplied. 1578 */ 1579 if (inp && *lookup == 0) { 1580 INP_LOCK_ASSERT(inp); 1581 if (inp->inp_socket != NULL) { 1582 fill_ugid_cache(inp, ugp); 1583 *lookup = 1; 1584 } 1585 } 1586 /* 1587 * If we have already been here and the packet has no 1588 * PCB entry associated with it, then we can safely 1589 * assume that this is a no match. 1590 */ 1591 if (*lookup == -1) 1592 return (0); 1593 if (proto == IPPROTO_TCP) { 1594 wildcard = 0; 1595 pi = &tcbinfo; 1596 } else if (proto == IPPROTO_UDP) { 1597 wildcard = 1; 1598 pi = &udbinfo; 1599 } else 1600 return 0; 1601 match = 0; 1602 if (*lookup == 0) { 1603 INP_INFO_RLOCK(pi); 1604 pcb = (oif) ? 1605 in_pcblookup_hash(pi, 1606 dst_ip, htons(dst_port), 1607 src_ip, htons(src_port), 1608 wildcard, oif) : 1609 in_pcblookup_hash(pi, 1610 src_ip, htons(src_port), 1611 dst_ip, htons(dst_port), 1612 wildcard, NULL); 1613 if (pcb != NULL) { 1614 INP_LOCK(pcb); 1615 if (pcb->inp_socket != NULL) { 1616 fill_ugid_cache(pcb, ugp); 1617 *lookup = 1; 1618 } 1619 INP_UNLOCK(pcb); 1620 } 1621 INP_INFO_RUNLOCK(pi); 1622 if (*lookup == 0) { 1623 /* 1624 * If the lookup did not yield any results, there 1625 * is no sense in coming back and trying again. So 1626 * we can set lookup to -1 and ensure that we wont 1627 * bother the pcb system again. 1628 */ 1629 *lookup = -1; 1630 return (0); 1631 } 1632 } 1633 if (insn->o.opcode == O_UID) 1634 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1635 else if (insn->o.opcode == O_GID) { 1636 for (gp = ugp->fw_groups; 1637 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1638 if (*gp == (gid_t)insn->d[0]) { 1639 match = 1; 1640 break; 1641 } 1642 } else if (insn->o.opcode == O_JAIL) 1643 match = (ugp->fw_prid == (int)insn->d[0]); 1644 return match; 1645} 1646 1647/* 1648 * The main check routine for the firewall. 1649 * 1650 * All arguments are in args so we can modify them and return them 1651 * back to the caller. 1652 * 1653 * Parameters: 1654 * 1655 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1656 * Starts with the IP header. 1657 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1658 * args->oif Outgoing interface, or NULL if packet is incoming. 1659 * The incoming interface is in the mbuf. (in) 1660 * args->divert_rule (in/out) 1661 * Skip up to the first rule past this rule number; 1662 * upon return, non-zero port number for divert or tee. 1663 * 1664 * args->rule Pointer to the last matching rule (in/out) 1665 * args->next_hop Socket we are forwarding to (out). 1666 * args->f_id Addresses grabbed from the packet (out) 1667 * 1668 * Return value: 1669 * 1670 * IP_FW_PORT_DENY_FLAG the packet must be dropped. 1671 * 0 The packet is to be accepted and routed normally OR 1672 * the packet was denied/rejected and has been dropped; 1673 * in the latter case, *m is equal to NULL upon return. 1674 * port Divert the packet to port, with these caveats: 1675 * 1676 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead 1677 * of diverting it (ie, 'ipfw tee'). 1678 * 1679 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower 1680 * 16 bits as a dummynet pipe number instead of diverting 1681 */ 1682 1683int 1684ipfw_chk(struct ip_fw_args *args) 1685{ 1686 /* 1687 * Local variables hold state during the processing of a packet. 1688 * 1689 * IMPORTANT NOTE: to speed up the processing of rules, there 1690 * are some assumption on the values of the variables, which 1691 * are documented here. Should you change them, please check 1692 * the implementation of the various instructions to make sure 1693 * that they still work. 1694 * 1695 * args->eh The MAC header. It is non-null for a layer2 1696 * packet, it is NULL for a layer-3 packet. 1697 * 1698 * m | args->m Pointer to the mbuf, as received from the caller. 1699 * It may change if ipfw_chk() does an m_pullup, or if it 1700 * consumes the packet because it calls send_reject(). 1701 * XXX This has to change, so that ipfw_chk() never modifies 1702 * or consumes the buffer. 1703 * ip is simply an alias of the value of m, and it is kept 1704 * in sync with it (the packet is supposed to start with 1705 * the ip header). 1706 */ 1707 struct mbuf *m = args->m; 1708 struct ip *ip = mtod(m, struct ip *); 1709 1710 /* 1711 * For rules which contain uid/gid or jail constraints, cache 1712 * a copy of the users credentials after the pcb lookup has been 1713 * executed. This will speed up the processing of rules with 1714 * these types of constraints, as well as decrease contention 1715 * on pcb related locks. 1716 */ 1717 struct ip_fw_ugid fw_ugid_cache; 1718 int ugid_lookup = 0; 1719 1720 /* 1721 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 1722 * associated with a packet input on a divert socket. This 1723 * will allow to distinguish traffic and its direction when 1724 * it originates from a divert socket. 1725 */ 1726 u_int divinput_flags = 0; 1727 1728 /* 1729 * oif | args->oif If NULL, ipfw_chk has been called on the 1730 * inbound path (ether_input, bdg_forward, ip_input). 1731 * If non-NULL, ipfw_chk has been called on the outbound path 1732 * (ether_output, ip_output). 1733 */ 1734 struct ifnet *oif = args->oif; 1735 1736 struct ip_fw *f = NULL; /* matching rule */ 1737 int retval = 0; 1738 1739 /* 1740 * hlen The length of the IPv4 header. 1741 * hlen >0 means we have an IPv4 packet. 1742 */ 1743 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1744 1745 /* 1746 * offset The offset of a fragment. offset != 0 means that 1747 * we have a fragment at this offset of an IPv4 packet. 1748 * offset == 0 means that (if this is an IPv4 packet) 1749 * this is the first or only fragment. 1750 */ 1751 u_short offset = 0; 1752 1753 /* 1754 * Local copies of addresses. They are only valid if we have 1755 * an IP packet. 1756 * 1757 * proto The protocol. Set to 0 for non-ip packets, 1758 * or to the protocol read from the packet otherwise. 1759 * proto != 0 means that we have an IPv4 packet. 1760 * 1761 * src_port, dst_port port numbers, in HOST format. Only 1762 * valid for TCP and UDP packets. 1763 * 1764 * src_ip, dst_ip ip addresses, in NETWORK format. 1765 * Only valid for IPv4 packets. 1766 */ 1767 u_int8_t proto; 1768 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1769 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1770 u_int16_t ip_len=0; 1771 int pktlen; 1772 int dyn_dir = MATCH_UNKNOWN; 1773 ipfw_dyn_rule *q = NULL; 1774 struct ip_fw_chain *chain = &layer3_chain; 1775 struct m_tag *mtag; 1776 1777 if (m->m_flags & M_SKIP_FIREWALL) 1778 return 0; /* accept */ 1779 /* 1780 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1781 * MATCH_NONE when checked and not matched (q = NULL), 1782 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1783 */ 1784 1785 pktlen = m->m_pkthdr.len; 1786 if (args->eh == NULL || /* layer 3 packet */ 1787 ( m->m_pkthdr.len >= sizeof(struct ip) && 1788 ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 1789 hlen = ip->ip_hl << 2; 1790 1791 /* 1792 * Collect parameters into local variables for faster matching. 1793 */ 1794 if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 1795 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1796 goto after_ip_checks; 1797 } 1798 1799 proto = args->f_id.proto = ip->ip_p; 1800 src_ip = ip->ip_src; 1801 dst_ip = ip->ip_dst; 1802 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1803 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1804 ip_len = ntohs(ip->ip_len); 1805 } else { 1806 offset = ip->ip_off & IP_OFFMASK; 1807 ip_len = ip->ip_len; 1808 } 1809 pktlen = ip_len < pktlen ? ip_len : pktlen; 1810 1811#define PULLUP_TO(len) \ 1812 do { \ 1813 if ((m)->m_len < (len)) { \ 1814 args->m = m = m_pullup(m, (len)); \ 1815 if (m == 0) \ 1816 goto pullup_failed; \ 1817 ip = mtod(m, struct ip *); \ 1818 } \ 1819 } while (0) 1820 1821 if (offset == 0) { 1822 switch (proto) { 1823 case IPPROTO_TCP: 1824 { 1825 struct tcphdr *tcp; 1826 1827 PULLUP_TO(hlen + sizeof(struct tcphdr)); 1828 tcp = L3HDR(struct tcphdr, ip); 1829 dst_port = tcp->th_dport; 1830 src_port = tcp->th_sport; 1831 args->f_id.flags = tcp->th_flags; 1832 } 1833 break; 1834 1835 case IPPROTO_UDP: 1836 { 1837 struct udphdr *udp; 1838 1839 PULLUP_TO(hlen + sizeof(struct udphdr)); 1840 udp = L3HDR(struct udphdr, ip); 1841 dst_port = udp->uh_dport; 1842 src_port = udp->uh_sport; 1843 } 1844 break; 1845 1846 case IPPROTO_ICMP: 1847 PULLUP_TO(hlen + 4); /* type, code and checksum. */ 1848 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 1849 break; 1850 1851 default: 1852 break; 1853 } 1854#undef PULLUP_TO 1855 } 1856 1857 args->f_id.src_ip = ntohl(src_ip.s_addr); 1858 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1859 args->f_id.src_port = src_port = ntohs(src_port); 1860 args->f_id.dst_port = dst_port = ntohs(dst_port); 1861 1862after_ip_checks: 1863 IPFW_LOCK(chain); /* XXX expensive? can we run lock free? */ 1864 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 1865 if (args->rule) { 1866 /* 1867 * Packet has already been tagged. Look for the next rule 1868 * to restart processing. 1869 * 1870 * If fw_one_pass != 0 then just accept it. 1871 * XXX should not happen here, but optimized out in 1872 * the caller. 1873 */ 1874 if (fw_one_pass) { 1875 IPFW_UNLOCK(chain); /* XXX optimize */ 1876 return 0; 1877 } 1878 1879 f = args->rule->next_rule; 1880 if (f == NULL) 1881 f = lookup_next_rule(args->rule); 1882 } else { 1883 /* 1884 * Find the starting rule. It can be either the first 1885 * one, or the one after divert_rule if asked so. 1886 */ 1887 int skipto = mtag ? divert_cookie(mtag) : 0; 1888 1889 f = chain->rules; 1890 if (args->eh == NULL && skipto != 0) { 1891 if (skipto >= IPFW_DEFAULT_RULE) { 1892 IPFW_UNLOCK(chain); 1893 return(IP_FW_PORT_DENY_FLAG); /* invalid */ 1894 } 1895 while (f && f->rulenum <= skipto) 1896 f = f->next; 1897 if (f == NULL) { /* drop packet */ 1898 IPFW_UNLOCK(chain); 1899 return(IP_FW_PORT_DENY_FLAG); 1900 } 1901 } 1902 } 1903 /* reset divert rule to avoid confusion later */ 1904 if (mtag) { 1905 divinput_flags = divert_info(mtag) & 1906 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 1907 m_tag_delete(m, mtag); 1908 } 1909 1910 /* 1911 * Now scan the rules, and parse microinstructions for each rule. 1912 */ 1913 for (; f; f = f->next) { 1914 int l, cmdlen; 1915 ipfw_insn *cmd; 1916 int skip_or; /* skip rest of OR block */ 1917 1918again: 1919 if (set_disable & (1 << f->set) ) 1920 continue; 1921 1922 skip_or = 0; 1923 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1924 l -= cmdlen, cmd += cmdlen) { 1925 int match; 1926 1927 /* 1928 * check_body is a jump target used when we find a 1929 * CHECK_STATE, and need to jump to the body of 1930 * the target rule. 1931 */ 1932 1933check_body: 1934 cmdlen = F_LEN(cmd); 1935 /* 1936 * An OR block (insn_1 || .. || insn_n) has the 1937 * F_OR bit set in all but the last instruction. 1938 * The first match will set "skip_or", and cause 1939 * the following instructions to be skipped until 1940 * past the one with the F_OR bit clear. 1941 */ 1942 if (skip_or) { /* skip this instruction */ 1943 if ((cmd->len & F_OR) == 0) 1944 skip_or = 0; /* next one is good */ 1945 continue; 1946 } 1947 match = 0; /* set to 1 if we succeed */ 1948 1949 switch (cmd->opcode) { 1950 /* 1951 * The first set of opcodes compares the packet's 1952 * fields with some pattern, setting 'match' if a 1953 * match is found. At the end of the loop there is 1954 * logic to deal with F_NOT and F_OR flags associated 1955 * with the opcode. 1956 */ 1957 case O_NOP: 1958 match = 1; 1959 break; 1960 1961 case O_FORWARD_MAC: 1962 printf("ipfw: opcode %d unimplemented\n", 1963 cmd->opcode); 1964 break; 1965 1966 case O_GID: 1967 case O_UID: 1968 case O_JAIL: 1969 /* 1970 * We only check offset == 0 && proto != 0, 1971 * as this ensures that we have an IPv4 1972 * packet with the ports info. 1973 */ 1974 if (offset!=0) 1975 break; 1976 if (proto == IPPROTO_TCP || 1977 proto == IPPROTO_UDP) 1978 match = check_uidgid( 1979 (ipfw_insn_u32 *)cmd, 1980 proto, oif, 1981 dst_ip, dst_port, 1982 src_ip, src_port, &fw_ugid_cache, 1983 &ugid_lookup, args->inp); 1984 break; 1985 1986 case O_RECV: 1987 match = iface_match(m->m_pkthdr.rcvif, 1988 (ipfw_insn_if *)cmd); 1989 break; 1990 1991 case O_XMIT: 1992 match = iface_match(oif, (ipfw_insn_if *)cmd); 1993 break; 1994 1995 case O_VIA: 1996 match = iface_match(oif ? oif : 1997 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1998 break; 1999 2000 case O_MACADDR2: 2001 if (args->eh != NULL) { /* have MAC header */ 2002 u_int32_t *want = (u_int32_t *) 2003 ((ipfw_insn_mac *)cmd)->addr; 2004 u_int32_t *mask = (u_int32_t *) 2005 ((ipfw_insn_mac *)cmd)->mask; 2006 u_int32_t *hdr = (u_int32_t *)args->eh; 2007 2008 match = 2009 ( want[0] == (hdr[0] & mask[0]) && 2010 want[1] == (hdr[1] & mask[1]) && 2011 want[2] == (hdr[2] & mask[2]) ); 2012 } 2013 break; 2014 2015 case O_MAC_TYPE: 2016 if (args->eh != NULL) { 2017 u_int16_t t = 2018 ntohs(args->eh->ether_type); 2019 u_int16_t *p = 2020 ((ipfw_insn_u16 *)cmd)->ports; 2021 int i; 2022 2023 for (i = cmdlen - 1; !match && i>0; 2024 i--, p += 2) 2025 match = (t>=p[0] && t<=p[1]); 2026 } 2027 break; 2028 2029 case O_FRAG: 2030 match = (hlen > 0 && offset != 0); 2031 break; 2032 2033 case O_IN: /* "out" is "not in" */ 2034 match = (oif == NULL); 2035 break; 2036 2037 case O_LAYER2: 2038 match = (args->eh != NULL); 2039 break; 2040 2041 case O_DIVERTED: 2042 match = (cmd->arg1 & 1 && divinput_flags & 2043 IP_FW_DIVERT_LOOPBACK_FLAG) || 2044 (cmd->arg1 & 2 && divinput_flags & 2045 IP_FW_DIVERT_OUTPUT_FLAG); 2046 break; 2047 2048 case O_PROTO: 2049 /* 2050 * We do not allow an arg of 0 so the 2051 * check of "proto" only suffices. 2052 */ 2053 match = (proto == cmd->arg1); 2054 break; 2055 2056 case O_IP_SRC: 2057 match = (hlen > 0 && 2058 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2059 src_ip.s_addr); 2060 break; 2061 2062 case O_IP_SRC_LOOKUP: 2063 case O_IP_DST_LOOKUP: 2064 if (hlen > 0) { 2065 uint32_t a = 2066 (cmd->opcode == O_IP_DST_LOOKUP) ? 2067 dst_ip.s_addr : src_ip.s_addr; 2068 uint32_t v; 2069 2070 match = lookup_table(cmd->arg1, a, &v); 2071 if (!match) 2072 break; 2073 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2074 match = 2075 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2076 } 2077 break; 2078 2079 case O_IP_SRC_MASK: 2080 case O_IP_DST_MASK: 2081 if (hlen > 0) { 2082 uint32_t a = 2083 (cmd->opcode == O_IP_DST_MASK) ? 2084 dst_ip.s_addr : src_ip.s_addr; 2085 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2086 int i = cmdlen-1; 2087 2088 for (; !match && i>0; i-= 2, p+= 2) 2089 match = (p[0] == (a & p[1])); 2090 } 2091 break; 2092 2093 case O_IP_SRC_ME: 2094 if (hlen > 0) { 2095 struct ifnet *tif; 2096 2097 INADDR_TO_IFP(src_ip, tif); 2098 match = (tif != NULL); 2099 } 2100 break; 2101 2102 case O_IP_DST_SET: 2103 case O_IP_SRC_SET: 2104 if (hlen > 0) { 2105 u_int32_t *d = (u_int32_t *)(cmd+1); 2106 u_int32_t addr = 2107 cmd->opcode == O_IP_DST_SET ? 2108 args->f_id.dst_ip : 2109 args->f_id.src_ip; 2110 2111 if (addr < d[0]) 2112 break; 2113 addr -= d[0]; /* subtract base */ 2114 match = (addr < cmd->arg1) && 2115 ( d[ 1 + (addr>>5)] & 2116 (1<<(addr & 0x1f)) ); 2117 } 2118 break; 2119 2120 case O_IP_DST: 2121 match = (hlen > 0 && 2122 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2123 dst_ip.s_addr); 2124 break; 2125 2126 case O_IP_DST_ME: 2127 if (hlen > 0) { 2128 struct ifnet *tif; 2129 2130 INADDR_TO_IFP(dst_ip, tif); 2131 match = (tif != NULL); 2132 } 2133 break; 2134 2135 case O_IP_SRCPORT: 2136 case O_IP_DSTPORT: 2137 /* 2138 * offset == 0 && proto != 0 is enough 2139 * to guarantee that we have an IPv4 2140 * packet with port info. 2141 */ 2142 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2143 && offset == 0) { 2144 u_int16_t x = 2145 (cmd->opcode == O_IP_SRCPORT) ? 2146 src_port : dst_port ; 2147 u_int16_t *p = 2148 ((ipfw_insn_u16 *)cmd)->ports; 2149 int i; 2150 2151 for (i = cmdlen - 1; !match && i>0; 2152 i--, p += 2) 2153 match = (x>=p[0] && x<=p[1]); 2154 } 2155 break; 2156 2157 case O_ICMPTYPE: 2158 match = (offset == 0 && proto==IPPROTO_ICMP && 2159 icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 2160 break; 2161 2162 case O_IPOPT: 2163 match = (hlen > 0 && ipopts_match(ip, cmd) ); 2164 break; 2165 2166 case O_IPVER: 2167 match = (hlen > 0 && cmd->arg1 == ip->ip_v); 2168 break; 2169 2170 case O_IPID: 2171 case O_IPLEN: 2172 case O_IPTTL: 2173 if (hlen > 0) { /* only for IP packets */ 2174 uint16_t x; 2175 uint16_t *p; 2176 int i; 2177 2178 if (cmd->opcode == O_IPLEN) 2179 x = ip_len; 2180 else if (cmd->opcode == O_IPTTL) 2181 x = ip->ip_ttl; 2182 else /* must be IPID */ 2183 x = ntohs(ip->ip_id); 2184 if (cmdlen == 1) { 2185 match = (cmd->arg1 == x); 2186 break; 2187 } 2188 /* otherwise we have ranges */ 2189 p = ((ipfw_insn_u16 *)cmd)->ports; 2190 i = cmdlen - 1; 2191 for (; !match && i>0; i--, p += 2) 2192 match = (x >= p[0] && x <= p[1]); 2193 } 2194 break; 2195 2196 case O_IPPRECEDENCE: 2197 match = (hlen > 0 && 2198 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2199 break; 2200 2201 case O_IPTOS: 2202 match = (hlen > 0 && 2203 flags_match(cmd, ip->ip_tos)); 2204 break; 2205 2206 case O_TCPFLAGS: 2207 match = (proto == IPPROTO_TCP && offset == 0 && 2208 flags_match(cmd, 2209 L3HDR(struct tcphdr,ip)->th_flags)); 2210 break; 2211 2212 case O_TCPOPTS: 2213 match = (proto == IPPROTO_TCP && offset == 0 && 2214 tcpopts_match(ip, cmd)); 2215 break; 2216 2217 case O_TCPSEQ: 2218 match = (proto == IPPROTO_TCP && offset == 0 && 2219 ((ipfw_insn_u32 *)cmd)->d[0] == 2220 L3HDR(struct tcphdr,ip)->th_seq); 2221 break; 2222 2223 case O_TCPACK: 2224 match = (proto == IPPROTO_TCP && offset == 0 && 2225 ((ipfw_insn_u32 *)cmd)->d[0] == 2226 L3HDR(struct tcphdr,ip)->th_ack); 2227 break; 2228 2229 case O_TCPWIN: 2230 match = (proto == IPPROTO_TCP && offset == 0 && 2231 cmd->arg1 == 2232 L3HDR(struct tcphdr,ip)->th_win); 2233 break; 2234 2235 case O_ESTAB: 2236 /* reject packets which have SYN only */ 2237 /* XXX should i also check for TH_ACK ? */ 2238 match = (proto == IPPROTO_TCP && offset == 0 && 2239 (L3HDR(struct tcphdr,ip)->th_flags & 2240 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2241 break; 2242 2243 case O_ALTQ: { 2244 struct altq_tag *at; 2245 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2246 2247 match = 1; 2248 mtag = m_tag_get(PACKET_TAG_PF_QID, 2249 sizeof(struct altq_tag), 2250 M_NOWAIT); 2251 if (mtag == NULL) { 2252 /* 2253 * Let the packet fall back to the 2254 * default ALTQ. 2255 */ 2256 break; 2257 } 2258 at = (struct altq_tag *)(mtag+1); 2259 at->qid = altq->qid; 2260 if (hlen != 0) 2261 at->af = AF_INET; 2262 else 2263 at->af = AF_LINK; 2264 at->hdr = ip; 2265 m_tag_prepend(m, mtag); 2266 break; 2267 } 2268 2269 case O_LOG: 2270 if (fw_verbose) 2271 ipfw_log(f, hlen, args->eh, m, oif); 2272 match = 1; 2273 break; 2274 2275 case O_PROB: 2276 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2277 break; 2278 2279 case O_VERREVPATH: 2280 /* Outgoing packets automatically pass/match */ 2281 match = (hlen > 0 && ((oif != NULL) || 2282 (m->m_pkthdr.rcvif == NULL) || 2283 verify_path(src_ip, m->m_pkthdr.rcvif))); 2284 break; 2285 2286 case O_VERSRCREACH: 2287 /* Outgoing packets automatically pass/match */ 2288 match = (hlen > 0 && ((oif != NULL) || 2289 verify_path(src_ip, NULL))); 2290 break; 2291 2292 case O_ANTISPOOF: 2293 /* Outgoing packets automatically pass/match */ 2294 if (oif == NULL && hlen > 0 && 2295 in_localaddr(src_ip)) 2296 match = verify_path(src_ip, 2297 m->m_pkthdr.rcvif); 2298 else 2299 match = 1; 2300 break; 2301 2302 case O_IPSEC: 2303#ifdef FAST_IPSEC 2304 match = (m_tag_find(m, 2305 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2306#endif 2307#ifdef IPSEC 2308 match = (ipsec_getnhist(m) != 0); 2309#endif 2310 /* otherwise no match */ 2311 break; 2312 2313 /* 2314 * The second set of opcodes represents 'actions', 2315 * i.e. the terminal part of a rule once the packet 2316 * matches all previous patterns. 2317 * Typically there is only one action for each rule, 2318 * and the opcode is stored at the end of the rule 2319 * (but there are exceptions -- see below). 2320 * 2321 * In general, here we set retval and terminate the 2322 * outer loop (would be a 'break 3' in some language, 2323 * but we need to do a 'goto done'). 2324 * 2325 * Exceptions: 2326 * O_COUNT and O_SKIPTO actions: 2327 * instead of terminating, we jump to the next rule 2328 * ('goto next_rule', equivalent to a 'break 2'), 2329 * or to the SKIPTO target ('goto again' after 2330 * having set f, cmd and l), respectively. 2331 * 2332 * O_LOG and O_ALTQ action parameters: 2333 * perform some action and set match = 1; 2334 * 2335 * O_LIMIT and O_KEEP_STATE: these opcodes are 2336 * not real 'actions', and are stored right 2337 * before the 'action' part of the rule. 2338 * These opcodes try to install an entry in the 2339 * state tables; if successful, we continue with 2340 * the next opcode (match=1; break;), otherwise 2341 * the packet * must be dropped 2342 * ('goto done' after setting retval); 2343 * 2344 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2345 * cause a lookup of the state table, and a jump 2346 * to the 'action' part of the parent rule 2347 * ('goto check_body') if an entry is found, or 2348 * (CHECK_STATE only) a jump to the next rule if 2349 * the entry is not found ('goto next_rule'). 2350 * The result of the lookup is cached to make 2351 * further instances of these opcodes are 2352 * effectively NOPs. 2353 */ 2354 case O_LIMIT: 2355 case O_KEEP_STATE: 2356 if (install_state(f, 2357 (ipfw_insn_limit *)cmd, args)) { 2358 retval = IP_FW_PORT_DENY_FLAG; 2359 goto done; /* error/limit violation */ 2360 } 2361 match = 1; 2362 break; 2363 2364 case O_PROBE_STATE: 2365 case O_CHECK_STATE: 2366 /* 2367 * dynamic rules are checked at the first 2368 * keep-state or check-state occurrence, 2369 * with the result being stored in dyn_dir. 2370 * The compiler introduces a PROBE_STATE 2371 * instruction for us when we have a 2372 * KEEP_STATE (because PROBE_STATE needs 2373 * to be run first). 2374 */ 2375 if (dyn_dir == MATCH_UNKNOWN && 2376 (q = lookup_dyn_rule(&args->f_id, 2377 &dyn_dir, proto == IPPROTO_TCP ? 2378 L3HDR(struct tcphdr, ip) : NULL)) 2379 != NULL) { 2380 /* 2381 * Found dynamic entry, update stats 2382 * and jump to the 'action' part of 2383 * the parent rule. 2384 */ 2385 q->pcnt++; 2386 q->bcnt += pktlen; 2387 f = q->rule; 2388 cmd = ACTION_PTR(f); 2389 l = f->cmd_len - f->act_ofs; 2390 IPFW_DYN_UNLOCK(); 2391 goto check_body; 2392 } 2393 /* 2394 * Dynamic entry not found. If CHECK_STATE, 2395 * skip to next rule, if PROBE_STATE just 2396 * ignore and continue with next opcode. 2397 */ 2398 if (cmd->opcode == O_CHECK_STATE) 2399 goto next_rule; 2400 match = 1; 2401 break; 2402 2403 case O_ACCEPT: 2404 retval = 0; /* accept */ 2405 goto done; 2406 2407 case O_PIPE: 2408 case O_QUEUE: 2409 args->rule = f; /* report matching rule */ 2410 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG; 2411 goto done; 2412 2413 case O_DIVERT: 2414 case O_TEE: { 2415 struct divert_tag *dt; 2416 2417 if (args->eh) /* not on layer 2 */ 2418 break; 2419 mtag = m_tag_get(PACKET_TAG_DIVERT, 2420 sizeof(struct divert_tag), 2421 M_NOWAIT); 2422 if (mtag == NULL) { 2423 /* XXX statistic */ 2424 /* drop packet */ 2425 IPFW_UNLOCK(chain); 2426 return IP_FW_PORT_DENY_FLAG; 2427 } 2428 dt = (struct divert_tag *)(mtag+1); 2429 dt->cookie = f->rulenum; 2430 dt->info = (cmd->opcode == O_DIVERT) ? 2431 cmd->arg1 : 2432 cmd->arg1 | IP_FW_PORT_TEE_FLAG; 2433 m_tag_prepend(m, mtag); 2434 retval = dt->info; 2435 goto done; 2436 } 2437 2438 case O_COUNT: 2439 case O_SKIPTO: 2440 f->pcnt++; /* update stats */ 2441 f->bcnt += pktlen; 2442 f->timestamp = time_second; 2443 if (cmd->opcode == O_COUNT) 2444 goto next_rule; 2445 /* handle skipto */ 2446 if (f->next_rule == NULL) 2447 lookup_next_rule(f); 2448 f = f->next_rule; 2449 goto again; 2450 2451 case O_REJECT: 2452 /* 2453 * Drop the packet and send a reject notice 2454 * if the packet is not ICMP (or is an ICMP 2455 * query), and it is not multicast/broadcast. 2456 */ 2457 if (hlen > 0 && 2458 (proto != IPPROTO_ICMP || 2459 is_icmp_query(ip)) && 2460 !(m->m_flags & (M_BCAST|M_MCAST)) && 2461 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2462 send_reject(args, cmd->arg1, 2463 offset,ip_len); 2464 m = args->m; 2465 } 2466 /* FALLTHROUGH */ 2467 case O_DENY: 2468 retval = IP_FW_PORT_DENY_FLAG; 2469 goto done; 2470 2471 case O_FORWARD_IP: 2472 if (args->eh) /* not valid on layer2 pkts */ 2473 break; 2474 if (!q || dyn_dir == MATCH_FORWARD) 2475 args->next_hop = 2476 &((ipfw_insn_sa *)cmd)->sa; 2477 retval = 0; 2478 goto done; 2479 2480 default: 2481 panic("-- unknown opcode %d\n", cmd->opcode); 2482 } /* end of switch() on opcodes */ 2483 2484 if (cmd->len & F_NOT) 2485 match = !match; 2486 2487 if (match) { 2488 if (cmd->len & F_OR) 2489 skip_or = 1; 2490 } else { 2491 if (!(cmd->len & F_OR)) /* not an OR block, */ 2492 break; /* try next rule */ 2493 } 2494 2495 } /* end of inner for, scan opcodes */ 2496 2497next_rule:; /* try next rule */ 2498 2499 } /* end of outer for, scan rules */ 2500 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2501 IPFW_UNLOCK(chain); 2502 return(IP_FW_PORT_DENY_FLAG); 2503 2504done: 2505 /* Update statistics */ 2506 f->pcnt++; 2507 f->bcnt += pktlen; 2508 f->timestamp = time_second; 2509 IPFW_UNLOCK(chain); 2510 return retval; 2511 2512pullup_failed: 2513 if (fw_verbose) 2514 printf("ipfw: pullup failed\n"); 2515 return(IP_FW_PORT_DENY_FLAG); 2516} 2517 2518/* 2519 * When a rule is added/deleted, clear the next_rule pointers in all rules. 2520 * These will be reconstructed on the fly as packets are matched. 2521 */ 2522static void 2523flush_rule_ptrs(struct ip_fw_chain *chain) 2524{ 2525 struct ip_fw *rule; 2526 2527 IPFW_LOCK_ASSERT(chain); 2528 2529 for (rule = chain->rules; rule; rule = rule->next) 2530 rule->next_rule = NULL; 2531} 2532 2533/* 2534 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 2535 * pipe/queue, or to all of them (match == NULL). 2536 */ 2537void 2538flush_pipe_ptrs(struct dn_flow_set *match) 2539{ 2540 struct ip_fw *rule; 2541 2542 IPFW_LOCK(&layer3_chain); 2543 for (rule = layer3_chain.rules; rule; rule = rule->next) { 2544 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 2545 2546 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 2547 continue; 2548 /* 2549 * XXX Use bcmp/bzero to handle pipe_ptr to overcome 2550 * possible alignment problems on 64-bit architectures. 2551 * This code is seldom used so we do not worry too 2552 * much about efficiency. 2553 */ 2554 if (match == NULL || 2555 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) ) 2556 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr)); 2557 } 2558 IPFW_UNLOCK(&layer3_chain); 2559} 2560 2561/* 2562 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 2563 * possibly create a rule number and add the rule to the list. 2564 * Update the rule_number in the input struct so the caller knows it as well. 2565 */ 2566static int 2567add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 2568{ 2569 struct ip_fw *rule, *f, *prev; 2570 int l = RULESIZE(input_rule); 2571 2572 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 2573 return (EINVAL); 2574 2575 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 2576 if (rule == NULL) 2577 return (ENOSPC); 2578 2579 bcopy(input_rule, rule, l); 2580 2581 rule->next = NULL; 2582 rule->next_rule = NULL; 2583 2584 rule->pcnt = 0; 2585 rule->bcnt = 0; 2586 rule->timestamp = 0; 2587 2588 IPFW_LOCK(chain); 2589 2590 if (chain->rules == NULL) { /* default rule */ 2591 chain->rules = rule; 2592 goto done; 2593 } 2594 2595 /* 2596 * If rulenum is 0, find highest numbered rule before the 2597 * default rule, and add autoinc_step 2598 */ 2599 if (autoinc_step < 1) 2600 autoinc_step = 1; 2601 else if (autoinc_step > 1000) 2602 autoinc_step = 1000; 2603 if (rule->rulenum == 0) { 2604 /* 2605 * locate the highest numbered rule before default 2606 */ 2607 for (f = chain->rules; f; f = f->next) { 2608 if (f->rulenum == IPFW_DEFAULT_RULE) 2609 break; 2610 rule->rulenum = f->rulenum; 2611 } 2612 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2613 rule->rulenum += autoinc_step; 2614 input_rule->rulenum = rule->rulenum; 2615 } 2616 2617 /* 2618 * Now insert the new rule in the right place in the sorted list. 2619 */ 2620 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 2621 if (f->rulenum > rule->rulenum) { /* found the location */ 2622 if (prev) { 2623 rule->next = f; 2624 prev->next = rule; 2625 } else { /* head insert */ 2626 rule->next = chain->rules; 2627 chain->rules = rule; 2628 } 2629 break; 2630 } 2631 } 2632 flush_rule_ptrs(chain); 2633done: 2634 static_count++; 2635 static_len += l; 2636 IPFW_UNLOCK(chain); 2637 DEB(printf("ipfw: installed rule %d, static count now %d\n", 2638 rule->rulenum, static_count);) 2639 return (0); 2640} 2641 2642/** 2643 * Remove a static rule (including derived * dynamic rules) 2644 * and place it on the ``reap list'' for later reclamation. 2645 * The caller is in charge of clearing rule pointers to avoid 2646 * dangling pointers. 2647 * @return a pointer to the next entry. 2648 * Arguments are not checked, so they better be correct. 2649 */ 2650static struct ip_fw * 2651remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 2652{ 2653 struct ip_fw *n; 2654 int l = RULESIZE(rule); 2655 2656 IPFW_LOCK_ASSERT(chain); 2657 2658 n = rule->next; 2659 IPFW_DYN_LOCK(); 2660 remove_dyn_rule(rule, NULL /* force removal */); 2661 IPFW_DYN_UNLOCK(); 2662 if (prev == NULL) 2663 chain->rules = n; 2664 else 2665 prev->next = n; 2666 static_count--; 2667 static_len -= l; 2668 2669 rule->next = chain->reap; 2670 chain->reap = rule; 2671 2672 return n; 2673} 2674 2675/** 2676 * Reclaim storage associated with a list of rules. This is 2677 * typically the list created using remove_rule. 2678 */ 2679static void 2680reap_rules(struct ip_fw *head) 2681{ 2682 struct ip_fw *rule; 2683 2684 while ((rule = head) != NULL) { 2685 head = head->next; 2686 if (DUMMYNET_LOADED) 2687 ip_dn_ruledel_ptr(rule); 2688 free(rule, M_IPFW); 2689 } 2690} 2691 2692/* 2693 * Remove all rules from a chain (except rules in set RESVD_SET 2694 * unless kill_default = 1). The caller is responsible for 2695 * reclaiming storage for the rules left in chain->reap. 2696 */ 2697static void 2698free_chain(struct ip_fw_chain *chain, int kill_default) 2699{ 2700 struct ip_fw *prev, *rule; 2701 2702 IPFW_LOCK_ASSERT(chain); 2703 2704 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 2705 for (prev = NULL, rule = chain->rules; rule ; ) 2706 if (kill_default || rule->set != RESVD_SET) 2707 rule = remove_rule(chain, rule, prev); 2708 else { 2709 prev = rule; 2710 rule = rule->next; 2711 } 2712} 2713 2714/** 2715 * Remove all rules with given number, and also do set manipulation. 2716 * Assumes chain != NULL && *chain != NULL. 2717 * 2718 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2719 * the next 8 bits are the new set, the top 8 bits are the command: 2720 * 2721 * 0 delete rules with given number 2722 * 1 delete rules with given set number 2723 * 2 move rules with given number to new set 2724 * 3 move rules with given set number to new set 2725 * 4 swap sets with given numbers 2726 */ 2727static int 2728del_entry(struct ip_fw_chain *chain, u_int32_t arg) 2729{ 2730 struct ip_fw *prev = NULL, *rule; 2731 u_int16_t rulenum; /* rule or old_set */ 2732 u_int8_t cmd, new_set; 2733 2734 rulenum = arg & 0xffff; 2735 cmd = (arg >> 24) & 0xff; 2736 new_set = (arg >> 16) & 0xff; 2737 2738 if (cmd > 4) 2739 return EINVAL; 2740 if (new_set > RESVD_SET) 2741 return EINVAL; 2742 if (cmd == 0 || cmd == 2) { 2743 if (rulenum >= IPFW_DEFAULT_RULE) 2744 return EINVAL; 2745 } else { 2746 if (rulenum > RESVD_SET) /* old_set */ 2747 return EINVAL; 2748 } 2749 2750 IPFW_LOCK(chain); 2751 rule = chain->rules; 2752 chain->reap = NULL; 2753 switch (cmd) { 2754 case 0: /* delete rules with given number */ 2755 /* 2756 * locate first rule to delete 2757 */ 2758 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 2759 ; 2760 if (rule->rulenum != rulenum) { 2761 IPFW_UNLOCK(chain); 2762 return EINVAL; 2763 } 2764 2765 /* 2766 * flush pointers outside the loop, then delete all matching 2767 * rules. prev remains the same throughout the cycle. 2768 */ 2769 flush_rule_ptrs(chain); 2770 while (rule->rulenum == rulenum) 2771 rule = remove_rule(chain, rule, prev); 2772 break; 2773 2774 case 1: /* delete all rules with given set number */ 2775 flush_rule_ptrs(chain); 2776 rule = chain->rules; 2777 while (rule->rulenum < IPFW_DEFAULT_RULE) 2778 if (rule->set == rulenum) 2779 rule = remove_rule(chain, rule, prev); 2780 else { 2781 prev = rule; 2782 rule = rule->next; 2783 } 2784 break; 2785 2786 case 2: /* move rules with given number to new set */ 2787 rule = chain->rules; 2788 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2789 if (rule->rulenum == rulenum) 2790 rule->set = new_set; 2791 break; 2792 2793 case 3: /* move rules with given set number to new set */ 2794 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2795 if (rule->set == rulenum) 2796 rule->set = new_set; 2797 break; 2798 2799 case 4: /* swap two sets */ 2800 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2801 if (rule->set == rulenum) 2802 rule->set = new_set; 2803 else if (rule->set == new_set) 2804 rule->set = rulenum; 2805 break; 2806 } 2807 /* 2808 * Look for rules to reclaim. We grab the list before 2809 * releasing the lock then reclaim them w/o the lock to 2810 * avoid a LOR with dummynet. 2811 */ 2812 rule = chain->reap; 2813 chain->reap = NULL; 2814 IPFW_UNLOCK(chain); 2815 if (rule) 2816 reap_rules(rule); 2817 return 0; 2818} 2819 2820/* 2821 * Clear counters for a specific rule. 2822 * The enclosing "table" is assumed locked. 2823 */ 2824static void 2825clear_counters(struct ip_fw *rule, int log_only) 2826{ 2827 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 2828 2829 if (log_only == 0) { 2830 rule->bcnt = rule->pcnt = 0; 2831 rule->timestamp = 0; 2832 } 2833 if (l->o.opcode == O_LOG) 2834 l->log_left = l->max_log; 2835} 2836 2837/** 2838 * Reset some or all counters on firewall rules. 2839 * @arg frwl is null to clear all entries, or contains a specific 2840 * rule number. 2841 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 2842 */ 2843static int 2844zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 2845{ 2846 struct ip_fw *rule; 2847 char *msg; 2848 2849 IPFW_LOCK(chain); 2850 if (rulenum == 0) { 2851 norule_counter = 0; 2852 for (rule = chain->rules; rule; rule = rule->next) 2853 clear_counters(rule, log_only); 2854 msg = log_only ? "ipfw: All logging counts reset.\n" : 2855 "ipfw: Accounting cleared.\n"; 2856 } else { 2857 int cleared = 0; 2858 /* 2859 * We can have multiple rules with the same number, so we 2860 * need to clear them all. 2861 */ 2862 for (rule = chain->rules; rule; rule = rule->next) 2863 if (rule->rulenum == rulenum) { 2864 while (rule && rule->rulenum == rulenum) { 2865 clear_counters(rule, log_only); 2866 rule = rule->next; 2867 } 2868 cleared = 1; 2869 break; 2870 } 2871 if (!cleared) { /* we did not find any matching rules */ 2872 IPFW_UNLOCK(chain); 2873 return (EINVAL); 2874 } 2875 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 2876 "ipfw: Entry %d cleared.\n"; 2877 } 2878 IPFW_UNLOCK(chain); 2879 2880 if (fw_verbose) 2881 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 2882 return (0); 2883} 2884 2885/* 2886 * Check validity of the structure before insert. 2887 * Fortunately rules are simple, so this mostly need to check rule sizes. 2888 */ 2889static int 2890check_ipfw_struct(struct ip_fw *rule, int size) 2891{ 2892 int l, cmdlen = 0; 2893 int have_action=0; 2894 ipfw_insn *cmd; 2895 2896 if (size < sizeof(*rule)) { 2897 printf("ipfw: rule too short\n"); 2898 return (EINVAL); 2899 } 2900 /* first, check for valid size */ 2901 l = RULESIZE(rule); 2902 if (l != size) { 2903 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 2904 return (EINVAL); 2905 } 2906 if (rule->act_ofs >= rule->cmd_len) { 2907 printf("ipfw: bogus action offset (%u > %u)\n", 2908 rule->act_ofs, rule->cmd_len - 1); 2909 return (EINVAL); 2910 } 2911 /* 2912 * Now go for the individual checks. Very simple ones, basically only 2913 * instruction sizes. 2914 */ 2915 for (l = rule->cmd_len, cmd = rule->cmd ; 2916 l > 0 ; l -= cmdlen, cmd += cmdlen) { 2917 cmdlen = F_LEN(cmd); 2918 if (cmdlen > l) { 2919 printf("ipfw: opcode %d size truncated\n", 2920 cmd->opcode); 2921 return EINVAL; 2922 } 2923 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 2924 switch (cmd->opcode) { 2925 case O_PROBE_STATE: 2926 case O_KEEP_STATE: 2927 case O_PROTO: 2928 case O_IP_SRC_ME: 2929 case O_IP_DST_ME: 2930 case O_LAYER2: 2931 case O_IN: 2932 case O_FRAG: 2933 case O_DIVERTED: 2934 case O_IPOPT: 2935 case O_IPTOS: 2936 case O_IPPRECEDENCE: 2937 case O_IPVER: 2938 case O_TCPWIN: 2939 case O_TCPFLAGS: 2940 case O_TCPOPTS: 2941 case O_ESTAB: 2942 case O_VERREVPATH: 2943 case O_VERSRCREACH: 2944 case O_ANTISPOOF: 2945 case O_IPSEC: 2946 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2947 goto bad_size; 2948 break; 2949 2950 case O_UID: 2951 case O_GID: 2952 case O_JAIL: 2953 case O_IP_SRC: 2954 case O_IP_DST: 2955 case O_TCPSEQ: 2956 case O_TCPACK: 2957 case O_PROB: 2958 case O_ICMPTYPE: 2959 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2960 goto bad_size; 2961 break; 2962 2963 case O_LIMIT: 2964 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 2965 goto bad_size; 2966 break; 2967 2968 case O_LOG: 2969 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 2970 goto bad_size; 2971 2972 ((ipfw_insn_log *)cmd)->log_left = 2973 ((ipfw_insn_log *)cmd)->max_log; 2974 2975 break; 2976 2977 case O_IP_SRC_MASK: 2978 case O_IP_DST_MASK: 2979 /* only odd command lengths */ 2980 if ( !(cmdlen & 1) || cmdlen > 31) 2981 goto bad_size; 2982 break; 2983 2984 case O_IP_SRC_SET: 2985 case O_IP_DST_SET: 2986 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 2987 printf("ipfw: invalid set size %d\n", 2988 cmd->arg1); 2989 return EINVAL; 2990 } 2991 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 2992 (cmd->arg1+31)/32 ) 2993 goto bad_size; 2994 break; 2995 2996 case O_IP_SRC_LOOKUP: 2997 case O_IP_DST_LOOKUP: 2998 if (cmd->arg1 >= IPFW_TABLES_MAX) { 2999 printf("ipfw: invalid table number %d\n", 3000 cmd->arg1); 3001 return (EINVAL); 3002 } 3003 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3004 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3005 goto bad_size; 3006 break; 3007 3008 case O_MACADDR2: 3009 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3010 goto bad_size; 3011 break; 3012 3013 case O_NOP: 3014 case O_IPID: 3015 case O_IPTTL: 3016 case O_IPLEN: 3017 if (cmdlen < 1 || cmdlen > 31) 3018 goto bad_size; 3019 break; 3020 3021 case O_MAC_TYPE: 3022 case O_IP_SRCPORT: 3023 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3024 if (cmdlen < 2 || cmdlen > 31) 3025 goto bad_size; 3026 break; 3027 3028 case O_RECV: 3029 case O_XMIT: 3030 case O_VIA: 3031 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3032 goto bad_size; 3033 break; 3034 3035 case O_ALTQ: 3036 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3037 goto bad_size; 3038 break; 3039 3040 case O_PIPE: 3041 case O_QUEUE: 3042 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 3043 goto bad_size; 3044 goto check_action; 3045 3046 case O_FORWARD_IP: 3047#ifdef IPFIREWALL_FORWARD 3048 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3049 goto bad_size; 3050 goto check_action; 3051#else 3052 return EINVAL; 3053#endif 3054 3055 case O_DIVERT: 3056 case O_TEE: 3057#ifndef IPDIVERT 3058 return EINVAL; 3059#endif 3060 case O_FORWARD_MAC: /* XXX not implemented yet */ 3061 case O_CHECK_STATE: 3062 case O_COUNT: 3063 case O_ACCEPT: 3064 case O_DENY: 3065 case O_REJECT: 3066 case O_SKIPTO: 3067 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3068 goto bad_size; 3069check_action: 3070 if (have_action) { 3071 printf("ipfw: opcode %d, multiple actions" 3072 " not allowed\n", 3073 cmd->opcode); 3074 return EINVAL; 3075 } 3076 have_action = 1; 3077 if (l != cmdlen) { 3078 printf("ipfw: opcode %d, action must be" 3079 " last opcode\n", 3080 cmd->opcode); 3081 return EINVAL; 3082 } 3083 break; 3084 default: 3085 printf("ipfw: opcode %d, unknown opcode\n", 3086 cmd->opcode); 3087 return EINVAL; 3088 } 3089 } 3090 if (have_action == 0) { 3091 printf("ipfw: missing action\n"); 3092 return EINVAL; 3093 } 3094 return 0; 3095 3096bad_size: 3097 printf("ipfw: opcode %d size %d wrong\n", 3098 cmd->opcode, cmdlen); 3099 return EINVAL; 3100} 3101 3102/* 3103 * Copy the static and dynamic rules to the supplied buffer 3104 * and return the amount of space actually used. 3105 */ 3106static size_t 3107ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3108{ 3109 char *bp = buf; 3110 char *ep = bp + space; 3111 struct ip_fw *rule; 3112 int i; 3113 3114 /* XXX this can take a long time and locking will block packet flow */ 3115 IPFW_LOCK(chain); 3116 for (rule = chain->rules; rule ; rule = rule->next) { 3117 /* 3118 * Verify the entry fits in the buffer in case the 3119 * rules changed between calculating buffer space and 3120 * now. This would be better done using a generation 3121 * number but should suffice for now. 3122 */ 3123 i = RULESIZE(rule); 3124 if (bp + i <= ep) { 3125 bcopy(rule, bp, i); 3126 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3127 sizeof(set_disable)); 3128 bp += i; 3129 } 3130 } 3131 IPFW_UNLOCK(chain); 3132 if (ipfw_dyn_v) { 3133 ipfw_dyn_rule *p, *last = NULL; 3134 3135 IPFW_DYN_LOCK(); 3136 for (i = 0 ; i < curr_dyn_buckets; i++) 3137 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3138 if (bp + sizeof *p <= ep) { 3139 ipfw_dyn_rule *dst = 3140 (ipfw_dyn_rule *)bp; 3141 bcopy(p, dst, sizeof *p); 3142 bcopy(&(p->rule->rulenum), &(dst->rule), 3143 sizeof(p->rule->rulenum)); 3144 /* 3145 * store a non-null value in "next". 3146 * The userland code will interpret a 3147 * NULL here as a marker 3148 * for the last dynamic rule. 3149 */ 3150 bcopy(&dst, &dst->next, sizeof(dst)); 3151 last = dst; 3152 dst->expire = 3153 TIME_LEQ(dst->expire, time_second) ? 3154 0 : dst->expire - time_second ; 3155 bp += sizeof(ipfw_dyn_rule); 3156 } 3157 } 3158 IPFW_DYN_UNLOCK(); 3159 if (last != NULL) /* mark last dynamic rule */ 3160 bzero(&last->next, sizeof(last)); 3161 } 3162 return (bp - (char *)buf); 3163} 3164 3165 3166/** 3167 * {set|get}sockopt parser. 3168 */ 3169static int 3170ipfw_ctl(struct sockopt *sopt) 3171{ 3172#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3173 int error, rule_num; 3174 size_t size; 3175 struct ip_fw *buf, *rule; 3176 u_int32_t rulenum[2]; 3177 3178 error = suser(sopt->sopt_td); 3179 if (error) 3180 return (error); 3181 3182 /* 3183 * Disallow modifications in really-really secure mode, but still allow 3184 * the logging counters to be reset. 3185 */ 3186 if (sopt->sopt_name == IP_FW_ADD || 3187 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3188#if __FreeBSD_version >= 500034 3189 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3190 if (error) 3191 return (error); 3192#else /* FreeBSD 4.x */ 3193 if (securelevel >= 3) 3194 return (EPERM); 3195#endif 3196 } 3197 3198 error = 0; 3199 3200 switch (sopt->sopt_name) { 3201 case IP_FW_GET: 3202 /* 3203 * pass up a copy of the current rules. Static rules 3204 * come first (the last of which has number IPFW_DEFAULT_RULE), 3205 * followed by a possibly empty list of dynamic rule. 3206 * The last dynamic rule has NULL in the "next" field. 3207 * 3208 * Note that the calculated size is used to bound the 3209 * amount of data returned to the user. The rule set may 3210 * change between calculating the size and returning the 3211 * data in which case we'll just return what fits. 3212 */ 3213 size = static_len; /* size of static rules */ 3214 if (ipfw_dyn_v) /* add size of dyn.rules */ 3215 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3216 3217 /* 3218 * XXX todo: if the user passes a short length just to know 3219 * how much room is needed, do not bother filling up the 3220 * buffer, just jump to the sooptcopyout. 3221 */ 3222 buf = malloc(size, M_TEMP, M_WAITOK); 3223 error = sooptcopyout(sopt, buf, 3224 ipfw_getrules(&layer3_chain, buf, size)); 3225 free(buf, M_TEMP); 3226 break; 3227 3228 case IP_FW_FLUSH: 3229 /* 3230 * Normally we cannot release the lock on each iteration. 3231 * We could do it here only because we start from the head all 3232 * the times so there is no risk of missing some entries. 3233 * On the other hand, the risk is that we end up with 3234 * a very inconsistent ruleset, so better keep the lock 3235 * around the whole cycle. 3236 * 3237 * XXX this code can be improved by resetting the head of 3238 * the list to point to the default rule, and then freeing 3239 * the old list without the need for a lock. 3240 */ 3241 3242 IPFW_LOCK(&layer3_chain); 3243 layer3_chain.reap = NULL; 3244 free_chain(&layer3_chain, 0 /* keep default rule */); 3245 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3246 IPFW_UNLOCK(&layer3_chain); 3247 if (layer3_chain.reap != NULL) 3248 reap_rules(rule); 3249 break; 3250 3251 case IP_FW_ADD: 3252 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3253 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3254 sizeof(struct ip_fw) ); 3255 if (error == 0) 3256 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3257 if (error == 0) { 3258 error = add_rule(&layer3_chain, rule); 3259 size = RULESIZE(rule); 3260 if (!error && sopt->sopt_dir == SOPT_GET) 3261 error = sooptcopyout(sopt, rule, size); 3262 } 3263 free(rule, M_TEMP); 3264 break; 3265 3266 case IP_FW_DEL: 3267 /* 3268 * IP_FW_DEL is used for deleting single rules or sets, 3269 * and (ab)used to atomically manipulate sets. Argument size 3270 * is used to distinguish between the two: 3271 * sizeof(u_int32_t) 3272 * delete single rule or set of rules, 3273 * or reassign rules (or sets) to a different set. 3274 * 2*sizeof(u_int32_t) 3275 * atomic disable/enable sets. 3276 * first u_int32_t contains sets to be disabled, 3277 * second u_int32_t contains sets to be enabled. 3278 */ 3279 error = sooptcopyin(sopt, rulenum, 3280 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3281 if (error) 3282 break; 3283 size = sopt->sopt_valsize; 3284 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3285 error = del_entry(&layer3_chain, rulenum[0]); 3286 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3287 set_disable = 3288 (set_disable | rulenum[0]) & ~rulenum[1] & 3289 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3290 else 3291 error = EINVAL; 3292 break; 3293 3294 case IP_FW_ZERO: 3295 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3296 rule_num = 0; 3297 if (sopt->sopt_val != 0) { 3298 error = sooptcopyin(sopt, &rule_num, 3299 sizeof(int), sizeof(int)); 3300 if (error) 3301 break; 3302 } 3303 error = zero_entry(&layer3_chain, rule_num, 3304 sopt->sopt_name == IP_FW_RESETLOG); 3305 break; 3306 3307 case IP_FW_TABLE_ADD: 3308 { 3309 ipfw_table_entry ent; 3310 3311 error = sooptcopyin(sopt, &ent, 3312 sizeof(ent), sizeof(ent)); 3313 if (error) 3314 break; 3315 error = add_table_entry(ent.tbl, ent.addr, 3316 ent.masklen, ent.value); 3317 } 3318 break; 3319 3320 case IP_FW_TABLE_DEL: 3321 { 3322 ipfw_table_entry ent; 3323 3324 error = sooptcopyin(sopt, &ent, 3325 sizeof(ent), sizeof(ent)); 3326 if (error) 3327 break; 3328 error = del_table_entry(ent.tbl, ent.addr, ent.masklen); 3329 } 3330 break; 3331 3332 case IP_FW_TABLE_FLUSH: 3333 { 3334 u_int16_t tbl; 3335 3336 error = sooptcopyin(sopt, &tbl, 3337 sizeof(tbl), sizeof(tbl)); 3338 if (error) 3339 break; 3340 error = flush_table(tbl); 3341 } 3342 break; 3343 3344 case IP_FW_TABLE_GETSIZE: 3345 { 3346 u_int32_t tbl, cnt; 3347 3348 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 3349 sizeof(tbl)))) 3350 break; 3351 if ((error = count_table(tbl, &cnt))) 3352 break; 3353 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 3354 } 3355 break; 3356 3357 case IP_FW_TABLE_LIST: 3358 { 3359 ipfw_table *tbl; 3360 3361 if (sopt->sopt_valsize < sizeof(*tbl)) { 3362 error = EINVAL; 3363 break; 3364 } 3365 size = sopt->sopt_valsize; 3366 tbl = malloc(size, M_TEMP, M_WAITOK); 3367 if (tbl == NULL) { 3368 error = ENOMEM; 3369 break; 3370 } 3371 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 3372 if (error) { 3373 free(tbl, M_TEMP); 3374 break; 3375 } 3376 tbl->size = (size - sizeof(*tbl)) / 3377 sizeof(ipfw_table_entry); 3378 error = dump_table(tbl); 3379 if (error) { 3380 free(tbl, M_TEMP); 3381 break; 3382 } 3383 error = sooptcopyout(sopt, tbl, size); 3384 free(tbl, M_TEMP); 3385 } 3386 break; 3387 3388 default: 3389 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 3390 error = EINVAL; 3391 } 3392 3393 return (error); 3394#undef RULE_MAXSIZE 3395} 3396 3397/** 3398 * dummynet needs a reference to the default rule, because rules can be 3399 * deleted while packets hold a reference to them. When this happens, 3400 * dummynet changes the reference to the default rule (it could well be a 3401 * NULL pointer, but this way we do not need to check for the special 3402 * case, plus here he have info on the default behaviour). 3403 */ 3404struct ip_fw *ip_fw_default_rule; 3405 3406/* 3407 * This procedure is only used to handle keepalives. It is invoked 3408 * every dyn_keepalive_period 3409 */ 3410static void 3411ipfw_tick(void * __unused unused) 3412{ 3413 int i; 3414 ipfw_dyn_rule *q; 3415 3416 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 3417 goto done; 3418 3419 IPFW_DYN_LOCK(); 3420 for (i = 0 ; i < curr_dyn_buckets ; i++) { 3421 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 3422 if (q->dyn_type == O_LIMIT_PARENT) 3423 continue; 3424 if (q->id.proto != IPPROTO_TCP) 3425 continue; 3426 if ( (q->state & BOTH_SYN) != BOTH_SYN) 3427 continue; 3428 if (TIME_LEQ( time_second+dyn_keepalive_interval, 3429 q->expire)) 3430 continue; /* too early */ 3431 if (TIME_LEQ(q->expire, time_second)) 3432 continue; /* too late, rule expired */ 3433 3434 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 3435 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 3436 } 3437 } 3438 IPFW_DYN_UNLOCK(); 3439done: 3440 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 3441} 3442 3443int 3444ipfw_init(void) 3445{ 3446 struct ip_fw default_rule; 3447 int error; 3448 3449 layer3_chain.rules = NULL; 3450 IPFW_LOCK_INIT(&layer3_chain); 3451 IPFW_DYN_LOCK_INIT(); 3452 callout_init(&ipfw_timeout, debug_mpsafenet ? CALLOUT_MPSAFE : 0); 3453 3454 bzero(&default_rule, sizeof default_rule); 3455 3456 default_rule.act_ofs = 0; 3457 default_rule.rulenum = IPFW_DEFAULT_RULE; 3458 default_rule.cmd_len = 1; 3459 default_rule.set = RESVD_SET; 3460 3461 default_rule.cmd[0].len = 1; 3462 default_rule.cmd[0].opcode = 3463#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 3464 1 ? O_ACCEPT : 3465#endif 3466 O_DENY; 3467 3468 error = add_rule(&layer3_chain, &default_rule); 3469 if (error != 0) { 3470 printf("ipfw2: error %u initializing default rule " 3471 "(support disabled)\n", error); 3472 IPFW_DYN_LOCK_DESTROY(); 3473 IPFW_LOCK_DESTROY(&layer3_chain); 3474 return (error); 3475 } 3476 3477 ip_fw_default_rule = layer3_chain.rules; 3478 printf("ipfw2 initialized, divert %s, " 3479 "rule-based forwarding " 3480#ifdef IPFIREWALL_FORWARD 3481 "enabled, " 3482#else 3483 "disabled, " 3484#endif 3485 "default to %s, logging ", 3486#ifdef IPDIVERT 3487 "enabled", 3488#else 3489 "disabled", 3490#endif 3491 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 3492 3493#ifdef IPFIREWALL_VERBOSE 3494 fw_verbose = 1; 3495#endif 3496#ifdef IPFIREWALL_VERBOSE_LIMIT 3497 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3498#endif 3499 if (fw_verbose == 0) 3500 printf("disabled\n"); 3501 else if (verbose_limit == 0) 3502 printf("unlimited\n"); 3503 else 3504 printf("limited to %d packets/entry by default\n", 3505 verbose_limit); 3506 3507 init_tables(); 3508 ip_fw_ctl_ptr = ipfw_ctl; 3509 ip_fw_chk_ptr = ipfw_chk; 3510 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 3511 3512 return (0); 3513} 3514 3515void 3516ipfw_destroy(void) 3517{ 3518 struct ip_fw *reap; 3519 3520 ip_fw_chk_ptr = NULL; 3521 ip_fw_ctl_ptr = NULL; 3522 callout_drain(&ipfw_timeout); 3523 IPFW_LOCK(&layer3_chain); 3524 layer3_chain.reap = NULL; 3525 free_chain(&layer3_chain, 1 /* kill default rule */); 3526 reap = layer3_chain.reap, layer3_chain.reap = NULL; 3527 IPFW_UNLOCK(&layer3_chain); 3528 if (reap != NULL) 3529 reap_rules(reap); 3530 flush_tables(); 3531 IPFW_DYN_LOCK_DESTROY(); 3532 IPFW_LOCK_DESTROY(&layer3_chain); 3533 printf("IP firewall unloaded\n"); 3534} 3535 3536#endif /* IPFW2 */ 3537