ip_fw2.c revision 136714
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 136714 2004-10-19 21:14:57Z andre $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdn.h" 38#include "opt_inet.h" 39#include "opt_ipsec.h" 40#ifndef INET 41#error IPFIREWALL requires INET. 42#endif /* INET */ 43#endif 44 45#define IPFW2 1 46#if IPFW2 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/malloc.h> 50#include <sys/mbuf.h> 51#include <sys/kernel.h> 52#include <sys/jail.h> 53#include <sys/module.h> 54#include <sys/proc.h> 55#include <sys/socket.h> 56#include <sys/socketvar.h> 57#include <sys/sysctl.h> 58#include <sys/syslog.h> 59#include <sys/ucred.h> 60#include <net/if.h> 61#include <net/radix.h> 62#include <net/route.h> 63#include <netinet/in.h> 64#include <netinet/in_systm.h> 65#include <netinet/in_var.h> 66#include <netinet/in_pcb.h> 67#include <netinet/ip.h> 68#include <netinet/ip_var.h> 69#include <netinet/ip_icmp.h> 70#include <netinet/ip_fw.h> 71#include <netinet/ip_divert.h> 72#include <netinet/ip_dummynet.h> 73#include <netinet/tcp.h> 74#include <netinet/tcp_timer.h> 75#include <netinet/tcp_var.h> 76#include <netinet/tcpip.h> 77#include <netinet/udp.h> 78#include <netinet/udp_var.h> 79#include <altq/if_altq.h> 80 81#ifdef IPSEC 82#include <netinet6/ipsec.h> 83#endif 84 85#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 86 87#include <machine/in_cksum.h> /* XXX for in_cksum */ 88 89/* 90 * set_disable contains one bit per set value (0..31). 91 * If the bit is set, all rules with the corresponding set 92 * are disabled. Set RESVD_SET(31) is reserved for the default rule 93 * and rules that are not deleted by the flush command, 94 * and CANNOT be disabled. 95 * Rules in set RESVD_SET can only be deleted explicitly. 96 */ 97static u_int32_t set_disable; 98 99static int fw_verbose; 100static int verbose_limit; 101 102static struct callout ipfw_timeout; 103#define IPFW_DEFAULT_RULE 65535 104 105/* 106 * Data structure to cache our ucred related 107 * information. This structure only gets used if 108 * the user specified UID/GID based constraints in 109 * a firewall rule. 110 */ 111struct ip_fw_ugid { 112 gid_t fw_groups[NGROUPS]; 113 int fw_ngroups; 114 uid_t fw_uid; 115 int fw_prid; 116}; 117 118struct ip_fw_chain { 119 struct ip_fw *rules; /* list of rules */ 120 struct ip_fw *reap; /* list of rules to reap */ 121 struct mtx mtx; /* lock guarding rule list */ 122}; 123#define IPFW_LOCK_INIT(_chain) \ 124 mtx_init(&(_chain)->mtx, "IPFW static rules", NULL, \ 125 MTX_DEF | MTX_RECURSE) 126#define IPFW_LOCK_DESTROY(_chain) mtx_destroy(&(_chain)->mtx) 127#define IPFW_LOCK(_chain) mtx_lock(&(_chain)->mtx) 128#define IPFW_UNLOCK(_chain) mtx_unlock(&(_chain)->mtx) 129#define IPFW_LOCK_ASSERT(_chain) do { \ 130 mtx_assert(&(_chain)->mtx, MA_OWNED); \ 131 NET_ASSERT_GIANT(); \ 132} while (0) 133 134/* 135 * list of rules for layer 3 136 */ 137static struct ip_fw_chain layer3_chain; 138 139MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 140MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 141 142struct table_entry { 143 struct radix_node rn[2]; 144 struct sockaddr_in addr, mask; 145 u_int32_t value; 146}; 147 148#define IPFW_TABLES_MAX 128 149static struct { 150 struct radix_node_head *rnh; 151 int modified; 152} ipfw_tables[IPFW_TABLES_MAX]; 153 154static int fw_debug = 1; 155static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 156 157#ifdef SYSCTL_NODE 158SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 159SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 160 CTLFLAG_RW | CTLFLAG_SECURE3, 161 &fw_enable, 0, "Enable ipfw"); 162SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 163 &autoinc_step, 0, "Rule number autincrement step"); 164SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 165 CTLFLAG_RW | CTLFLAG_SECURE3, 166 &fw_one_pass, 0, 167 "Only do a single pass through ipfw when using dummynet(4)"); 168SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 169 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 170SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 171 CTLFLAG_RW | CTLFLAG_SECURE3, 172 &fw_verbose, 0, "Log matches to ipfw rules"); 173SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 174 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 175 176/* 177 * Description of dynamic rules. 178 * 179 * Dynamic rules are stored in lists accessed through a hash table 180 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 181 * be modified through the sysctl variable dyn_buckets which is 182 * updated when the table becomes empty. 183 * 184 * XXX currently there is only one list, ipfw_dyn. 185 * 186 * When a packet is received, its address fields are first masked 187 * with the mask defined for the rule, then hashed, then matched 188 * against the entries in the corresponding list. 189 * Dynamic rules can be used for different purposes: 190 * + stateful rules; 191 * + enforcing limits on the number of sessions; 192 * + in-kernel NAT (not implemented yet) 193 * 194 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 195 * measured in seconds and depending on the flags. 196 * 197 * The total number of dynamic rules is stored in dyn_count. 198 * The max number of dynamic rules is dyn_max. When we reach 199 * the maximum number of rules we do not create anymore. This is 200 * done to avoid consuming too much memory, but also too much 201 * time when searching on each packet (ideally, we should try instead 202 * to put a limit on the length of the list on each bucket...). 203 * 204 * Each dynamic rule holds a pointer to the parent ipfw rule so 205 * we know what action to perform. Dynamic rules are removed when 206 * the parent rule is deleted. XXX we should make them survive. 207 * 208 * There are some limitations with dynamic rules -- we do not 209 * obey the 'randomized match', and we do not do multiple 210 * passes through the firewall. XXX check the latter!!! 211 */ 212static ipfw_dyn_rule **ipfw_dyn_v = NULL; 213static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 214static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 215 216static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 217#define IPFW_DYN_LOCK_INIT() \ 218 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 219#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 220#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 221#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 222#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 223 224/* 225 * Timeouts for various events in handing dynamic rules. 226 */ 227static u_int32_t dyn_ack_lifetime = 300; 228static u_int32_t dyn_syn_lifetime = 20; 229static u_int32_t dyn_fin_lifetime = 1; 230static u_int32_t dyn_rst_lifetime = 1; 231static u_int32_t dyn_udp_lifetime = 10; 232static u_int32_t dyn_short_lifetime = 5; 233 234/* 235 * Keepalives are sent if dyn_keepalive is set. They are sent every 236 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 237 * seconds of lifetime of a rule. 238 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 239 * than dyn_keepalive_period. 240 */ 241 242static u_int32_t dyn_keepalive_interval = 20; 243static u_int32_t dyn_keepalive_period = 5; 244static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 245 246static u_int32_t static_count; /* # of static rules */ 247static u_int32_t static_len; /* size in bytes of static rules */ 248static u_int32_t dyn_count; /* # of dynamic rules */ 249static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 250 251SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 252 &dyn_buckets, 0, "Number of dyn. buckets"); 253SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 254 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 255SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 256 &dyn_count, 0, "Number of dyn. rules"); 257SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 258 &dyn_max, 0, "Max number of dyn. rules"); 259SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 260 &static_count, 0, "Number of static rules"); 261SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 262 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 263SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 264 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 265SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 266 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 267SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 268 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 269SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 270 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 271SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 272 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 273SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 274 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 275 276#endif /* SYSCTL_NODE */ 277 278 279/* 280 * This macro maps an ip pointer into a layer3 header pointer of type T 281 */ 282#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 283 284static __inline int 285icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 286{ 287 int type = L3HDR(struct icmp,ip)->icmp_type; 288 289 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 290} 291 292#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 293 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 294 295static int 296is_icmp_query(struct ip *ip) 297{ 298 int type = L3HDR(struct icmp, ip)->icmp_type; 299 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 300} 301#undef TT 302 303/* 304 * The following checks use two arrays of 8 or 16 bits to store the 305 * bits that we want set or clear, respectively. They are in the 306 * low and high half of cmd->arg1 or cmd->d[0]. 307 * 308 * We scan options and store the bits we find set. We succeed if 309 * 310 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 311 * 312 * The code is sometimes optimized not to store additional variables. 313 */ 314 315static int 316flags_match(ipfw_insn *cmd, u_int8_t bits) 317{ 318 u_char want_clear; 319 bits = ~bits; 320 321 if ( ((cmd->arg1 & 0xff) & bits) != 0) 322 return 0; /* some bits we want set were clear */ 323 want_clear = (cmd->arg1 >> 8) & 0xff; 324 if ( (want_clear & bits) != want_clear) 325 return 0; /* some bits we want clear were set */ 326 return 1; 327} 328 329static int 330ipopts_match(struct ip *ip, ipfw_insn *cmd) 331{ 332 int optlen, bits = 0; 333 u_char *cp = (u_char *)(ip + 1); 334 int x = (ip->ip_hl << 2) - sizeof (struct ip); 335 336 for (; x > 0; x -= optlen, cp += optlen) { 337 int opt = cp[IPOPT_OPTVAL]; 338 339 if (opt == IPOPT_EOL) 340 break; 341 if (opt == IPOPT_NOP) 342 optlen = 1; 343 else { 344 optlen = cp[IPOPT_OLEN]; 345 if (optlen <= 0 || optlen > x) 346 return 0; /* invalid or truncated */ 347 } 348 switch (opt) { 349 350 default: 351 break; 352 353 case IPOPT_LSRR: 354 bits |= IP_FW_IPOPT_LSRR; 355 break; 356 357 case IPOPT_SSRR: 358 bits |= IP_FW_IPOPT_SSRR; 359 break; 360 361 case IPOPT_RR: 362 bits |= IP_FW_IPOPT_RR; 363 break; 364 365 case IPOPT_TS: 366 bits |= IP_FW_IPOPT_TS; 367 break; 368 } 369 } 370 return (flags_match(cmd, bits)); 371} 372 373static int 374tcpopts_match(struct ip *ip, ipfw_insn *cmd) 375{ 376 int optlen, bits = 0; 377 struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 378 u_char *cp = (u_char *)(tcp + 1); 379 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 380 381 for (; x > 0; x -= optlen, cp += optlen) { 382 int opt = cp[0]; 383 if (opt == TCPOPT_EOL) 384 break; 385 if (opt == TCPOPT_NOP) 386 optlen = 1; 387 else { 388 optlen = cp[1]; 389 if (optlen <= 0) 390 break; 391 } 392 393 switch (opt) { 394 395 default: 396 break; 397 398 case TCPOPT_MAXSEG: 399 bits |= IP_FW_TCPOPT_MSS; 400 break; 401 402 case TCPOPT_WINDOW: 403 bits |= IP_FW_TCPOPT_WINDOW; 404 break; 405 406 case TCPOPT_SACK_PERMITTED: 407 case TCPOPT_SACK: 408 bits |= IP_FW_TCPOPT_SACK; 409 break; 410 411 case TCPOPT_TIMESTAMP: 412 bits |= IP_FW_TCPOPT_TS; 413 break; 414 415 case TCPOPT_CC: 416 case TCPOPT_CCNEW: 417 case TCPOPT_CCECHO: 418 bits |= IP_FW_TCPOPT_CC; 419 break; 420 } 421 } 422 return (flags_match(cmd, bits)); 423} 424 425static int 426iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 427{ 428 if (ifp == NULL) /* no iface with this packet, match fails */ 429 return 0; 430 /* Check by name or by IP address */ 431 if (cmd->name[0] != '\0') { /* match by name */ 432 /* Check name */ 433 if (cmd->p.glob) { 434 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 435 return(1); 436 } else { 437 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 438 return(1); 439 } 440 } else { 441 struct ifaddr *ia; 442 443 /* XXX lock? */ 444 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 445 if (ia->ifa_addr == NULL) 446 continue; 447 if (ia->ifa_addr->sa_family != AF_INET) 448 continue; 449 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 450 (ia->ifa_addr))->sin_addr.s_addr) 451 return(1); /* match */ 452 } 453 } 454 return(0); /* no match, fail ... */ 455} 456 457/* 458 * The verify_path function checks if a route to the src exists and 459 * if it is reachable via ifp (when provided). 460 * 461 * The 'verrevpath' option checks that the interface that an IP packet 462 * arrives on is the same interface that traffic destined for the 463 * packet's source address would be routed out of. The 'versrcreach' 464 * option just checks that the source address is reachable via any route 465 * (except default) in the routing table. These two are a measure to block 466 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 467 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 468 * is purposely reminiscent of the Cisco IOS command, 469 * 470 * ip verify unicast reverse-path 471 * ip verify unicast source reachable-via any 472 * 473 * which implements the same functionality. But note that syntax is 474 * misleading. The check may be performed on all IP packets whether unicast, 475 * multicast, or broadcast. 476 */ 477static int 478verify_path(struct in_addr src, struct ifnet *ifp) 479{ 480 struct route ro; 481 struct sockaddr_in *dst; 482 483 bzero(&ro, sizeof(ro)); 484 485 dst = (struct sockaddr_in *)&(ro.ro_dst); 486 dst->sin_family = AF_INET; 487 dst->sin_len = sizeof(*dst); 488 dst->sin_addr = src; 489 rtalloc_ign(&ro, RTF_CLONING); 490 491 if (ro.ro_rt == NULL) 492 return 0; 493 494 /* if ifp is provided, check for equality with rtentry */ 495 if (ifp != NULL && ro.ro_rt->rt_ifp != ifp) { 496 RTFREE(ro.ro_rt); 497 return 0; 498 } 499 500 /* if no ifp provided, check if rtentry is not default route */ 501 if (ifp == NULL && 502 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 503 RTFREE(ro.ro_rt); 504 return 0; 505 } 506 507 /* or if this is a blackhole/reject route */ 508 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 509 RTFREE(ro.ro_rt); 510 return 0; 511 } 512 513 /* found valid route */ 514 RTFREE(ro.ro_rt); 515 return 1; 516} 517 518 519static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 520 521#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 522#define SNP(buf) buf, sizeof(buf) 523 524/* 525 * We enter here when we have a rule with O_LOG. 526 * XXX this function alone takes about 2Kbytes of code! 527 */ 528static void 529ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 530 struct mbuf *m, struct ifnet *oif) 531{ 532 char *action; 533 int limit_reached = 0; 534 char action2[40], proto[48], fragment[28]; 535 536 fragment[0] = '\0'; 537 proto[0] = '\0'; 538 539 if (f == NULL) { /* bogus pkt */ 540 if (verbose_limit != 0 && norule_counter >= verbose_limit) 541 return; 542 norule_counter++; 543 if (norule_counter == verbose_limit) 544 limit_reached = verbose_limit; 545 action = "Refuse"; 546 } else { /* O_LOG is the first action, find the real one */ 547 ipfw_insn *cmd = ACTION_PTR(f); 548 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 549 550 if (l->max_log != 0 && l->log_left == 0) 551 return; 552 l->log_left--; 553 if (l->log_left == 0) 554 limit_reached = l->max_log; 555 cmd += F_LEN(cmd); /* point to first action */ 556 if (cmd->opcode == O_ALTQ) { 557 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 558 559 snprintf(SNPARGS(action2, 0), "Altq %d", 560 altq->qid); 561 cmd += F_LEN(cmd); 562 } 563 if (cmd->opcode == O_PROB) 564 cmd += F_LEN(cmd); 565 566 action = action2; 567 switch (cmd->opcode) { 568 case O_DENY: 569 action = "Deny"; 570 break; 571 572 case O_REJECT: 573 if (cmd->arg1==ICMP_REJECT_RST) 574 action = "Reset"; 575 else if (cmd->arg1==ICMP_UNREACH_HOST) 576 action = "Reject"; 577 else 578 snprintf(SNPARGS(action2, 0), "Unreach %d", 579 cmd->arg1); 580 break; 581 582 case O_ACCEPT: 583 action = "Accept"; 584 break; 585 case O_COUNT: 586 action = "Count"; 587 break; 588 case O_DIVERT: 589 snprintf(SNPARGS(action2, 0), "Divert %d", 590 cmd->arg1); 591 break; 592 case O_TEE: 593 snprintf(SNPARGS(action2, 0), "Tee %d", 594 cmd->arg1); 595 break; 596 case O_SKIPTO: 597 snprintf(SNPARGS(action2, 0), "SkipTo %d", 598 cmd->arg1); 599 break; 600 case O_PIPE: 601 snprintf(SNPARGS(action2, 0), "Pipe %d", 602 cmd->arg1); 603 break; 604 case O_QUEUE: 605 snprintf(SNPARGS(action2, 0), "Queue %d", 606 cmd->arg1); 607 break; 608 case O_FORWARD_IP: { 609 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 610 int len; 611 612 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 613 inet_ntoa(sa->sa.sin_addr)); 614 if (sa->sa.sin_port) 615 snprintf(SNPARGS(action2, len), ":%d", 616 sa->sa.sin_port); 617 } 618 break; 619 default: 620 action = "UNKNOWN"; 621 break; 622 } 623 } 624 625 if (hlen == 0) { /* non-ip */ 626 snprintf(SNPARGS(proto, 0), "MAC"); 627 } else { 628 struct ip *ip = mtod(m, struct ip *); 629 /* these three are all aliases to the same thing */ 630 struct icmp *const icmp = L3HDR(struct icmp, ip); 631 struct tcphdr *const tcp = (struct tcphdr *)icmp; 632 struct udphdr *const udp = (struct udphdr *)icmp; 633 634 int ip_off, offset, ip_len; 635 636 int len; 637 638 if (eh != NULL) { /* layer 2 packets are as on the wire */ 639 ip_off = ntohs(ip->ip_off); 640 ip_len = ntohs(ip->ip_len); 641 } else { 642 ip_off = ip->ip_off; 643 ip_len = ip->ip_len; 644 } 645 offset = ip_off & IP_OFFMASK; 646 switch (ip->ip_p) { 647 case IPPROTO_TCP: 648 len = snprintf(SNPARGS(proto, 0), "TCP %s", 649 inet_ntoa(ip->ip_src)); 650 if (offset == 0) 651 snprintf(SNPARGS(proto, len), ":%d %s:%d", 652 ntohs(tcp->th_sport), 653 inet_ntoa(ip->ip_dst), 654 ntohs(tcp->th_dport)); 655 else 656 snprintf(SNPARGS(proto, len), " %s", 657 inet_ntoa(ip->ip_dst)); 658 break; 659 660 case IPPROTO_UDP: 661 len = snprintf(SNPARGS(proto, 0), "UDP %s", 662 inet_ntoa(ip->ip_src)); 663 if (offset == 0) 664 snprintf(SNPARGS(proto, len), ":%d %s:%d", 665 ntohs(udp->uh_sport), 666 inet_ntoa(ip->ip_dst), 667 ntohs(udp->uh_dport)); 668 else 669 snprintf(SNPARGS(proto, len), " %s", 670 inet_ntoa(ip->ip_dst)); 671 break; 672 673 case IPPROTO_ICMP: 674 if (offset == 0) 675 len = snprintf(SNPARGS(proto, 0), 676 "ICMP:%u.%u ", 677 icmp->icmp_type, icmp->icmp_code); 678 else 679 len = snprintf(SNPARGS(proto, 0), "ICMP "); 680 len += snprintf(SNPARGS(proto, len), "%s", 681 inet_ntoa(ip->ip_src)); 682 snprintf(SNPARGS(proto, len), " %s", 683 inet_ntoa(ip->ip_dst)); 684 break; 685 686 default: 687 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 688 inet_ntoa(ip->ip_src)); 689 snprintf(SNPARGS(proto, len), " %s", 690 inet_ntoa(ip->ip_dst)); 691 break; 692 } 693 694 if (ip_off & (IP_MF | IP_OFFMASK)) 695 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 696 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 697 offset << 3, 698 (ip_off & IP_MF) ? "+" : ""); 699 } 700 if (oif || m->m_pkthdr.rcvif) 701 log(LOG_SECURITY | LOG_INFO, 702 "ipfw: %d %s %s %s via %s%s\n", 703 f ? f->rulenum : -1, 704 action, proto, oif ? "out" : "in", 705 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 706 fragment); 707 else 708 log(LOG_SECURITY | LOG_INFO, 709 "ipfw: %d %s %s [no if info]%s\n", 710 f ? f->rulenum : -1, 711 action, proto, fragment); 712 if (limit_reached) 713 log(LOG_SECURITY | LOG_NOTICE, 714 "ipfw: limit %d reached on entry %d\n", 715 limit_reached, f ? f->rulenum : -1); 716} 717 718/* 719 * IMPORTANT: the hash function for dynamic rules must be commutative 720 * in source and destination (ip,port), because rules are bidirectional 721 * and we want to find both in the same bucket. 722 */ 723static __inline int 724hash_packet(struct ipfw_flow_id *id) 725{ 726 u_int32_t i; 727 728 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 729 i &= (curr_dyn_buckets - 1); 730 return i; 731} 732 733/** 734 * unlink a dynamic rule from a chain. prev is a pointer to 735 * the previous one, q is a pointer to the rule to delete, 736 * head is a pointer to the head of the queue. 737 * Modifies q and potentially also head. 738 */ 739#define UNLINK_DYN_RULE(prev, head, q) { \ 740 ipfw_dyn_rule *old_q = q; \ 741 \ 742 /* remove a refcount to the parent */ \ 743 if (q->dyn_type == O_LIMIT) \ 744 q->parent->count--; \ 745 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 746 (q->id.src_ip), (q->id.src_port), \ 747 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 748 if (prev != NULL) \ 749 prev->next = q = q->next; \ 750 else \ 751 head = q = q->next; \ 752 dyn_count--; \ 753 free(old_q, M_IPFW); } 754 755#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 756 757/** 758 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 759 * 760 * If keep_me == NULL, rules are deleted even if not expired, 761 * otherwise only expired rules are removed. 762 * 763 * The value of the second parameter is also used to point to identify 764 * a rule we absolutely do not want to remove (e.g. because we are 765 * holding a reference to it -- this is the case with O_LIMIT_PARENT 766 * rules). The pointer is only used for comparison, so any non-null 767 * value will do. 768 */ 769static void 770remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 771{ 772 static u_int32_t last_remove = 0; 773 774#define FORCE (keep_me == NULL) 775 776 ipfw_dyn_rule *prev, *q; 777 int i, pass = 0, max_pass = 0; 778 779 IPFW_DYN_LOCK_ASSERT(); 780 781 if (ipfw_dyn_v == NULL || dyn_count == 0) 782 return; 783 /* do not expire more than once per second, it is useless */ 784 if (!FORCE && last_remove == time_second) 785 return; 786 last_remove = time_second; 787 788 /* 789 * because O_LIMIT refer to parent rules, during the first pass only 790 * remove child and mark any pending LIMIT_PARENT, and remove 791 * them in a second pass. 792 */ 793next_pass: 794 for (i = 0 ; i < curr_dyn_buckets ; i++) { 795 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 796 /* 797 * Logic can become complex here, so we split tests. 798 */ 799 if (q == keep_me) 800 goto next; 801 if (rule != NULL && rule != q->rule) 802 goto next; /* not the one we are looking for */ 803 if (q->dyn_type == O_LIMIT_PARENT) { 804 /* 805 * handle parent in the second pass, 806 * record we need one. 807 */ 808 max_pass = 1; 809 if (pass == 0) 810 goto next; 811 if (FORCE && q->count != 0 ) { 812 /* XXX should not happen! */ 813 printf("ipfw: OUCH! cannot remove rule," 814 " count %d\n", q->count); 815 } 816 } else { 817 if (!FORCE && 818 !TIME_LEQ( q->expire, time_second )) 819 goto next; 820 } 821 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 822 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 823 continue; 824 } 825next: 826 prev=q; 827 q=q->next; 828 } 829 } 830 if (pass++ < max_pass) 831 goto next_pass; 832} 833 834 835/** 836 * lookup a dynamic rule. 837 */ 838static ipfw_dyn_rule * 839lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 840 struct tcphdr *tcp) 841{ 842 /* 843 * stateful ipfw extensions. 844 * Lookup into dynamic session queue 845 */ 846#define MATCH_REVERSE 0 847#define MATCH_FORWARD 1 848#define MATCH_NONE 2 849#define MATCH_UNKNOWN 3 850 int i, dir = MATCH_NONE; 851 ipfw_dyn_rule *prev, *q=NULL; 852 853 IPFW_DYN_LOCK_ASSERT(); 854 855 if (ipfw_dyn_v == NULL) 856 goto done; /* not found */ 857 i = hash_packet( pkt ); 858 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 859 if (q->dyn_type == O_LIMIT_PARENT && q->count) 860 goto next; 861 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 862 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 863 continue; 864 } 865 if (pkt->proto == q->id.proto && 866 q->dyn_type != O_LIMIT_PARENT) { 867 if (pkt->src_ip == q->id.src_ip && 868 pkt->dst_ip == q->id.dst_ip && 869 pkt->src_port == q->id.src_port && 870 pkt->dst_port == q->id.dst_port ) { 871 dir = MATCH_FORWARD; 872 break; 873 } 874 if (pkt->src_ip == q->id.dst_ip && 875 pkt->dst_ip == q->id.src_ip && 876 pkt->src_port == q->id.dst_port && 877 pkt->dst_port == q->id.src_port ) { 878 dir = MATCH_REVERSE; 879 break; 880 } 881 } 882next: 883 prev = q; 884 q = q->next; 885 } 886 if (q == NULL) 887 goto done; /* q = NULL, not found */ 888 889 if ( prev != NULL) { /* found and not in front */ 890 prev->next = q->next; 891 q->next = ipfw_dyn_v[i]; 892 ipfw_dyn_v[i] = q; 893 } 894 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 895 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 896 897#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 898#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 899 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 900 switch (q->state) { 901 case TH_SYN: /* opening */ 902 q->expire = time_second + dyn_syn_lifetime; 903 break; 904 905 case BOTH_SYN: /* move to established */ 906 case BOTH_SYN | TH_FIN : /* one side tries to close */ 907 case BOTH_SYN | (TH_FIN << 8) : 908 if (tcp) { 909#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 910 u_int32_t ack = ntohl(tcp->th_ack); 911 if (dir == MATCH_FORWARD) { 912 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 913 q->ack_fwd = ack; 914 else { /* ignore out-of-sequence */ 915 break; 916 } 917 } else { 918 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 919 q->ack_rev = ack; 920 else { /* ignore out-of-sequence */ 921 break; 922 } 923 } 924 } 925 q->expire = time_second + dyn_ack_lifetime; 926 break; 927 928 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 929 if (dyn_fin_lifetime >= dyn_keepalive_period) 930 dyn_fin_lifetime = dyn_keepalive_period - 1; 931 q->expire = time_second + dyn_fin_lifetime; 932 break; 933 934 default: 935#if 0 936 /* 937 * reset or some invalid combination, but can also 938 * occur if we use keep-state the wrong way. 939 */ 940 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 941 printf("invalid state: 0x%x\n", q->state); 942#endif 943 if (dyn_rst_lifetime >= dyn_keepalive_period) 944 dyn_rst_lifetime = dyn_keepalive_period - 1; 945 q->expire = time_second + dyn_rst_lifetime; 946 break; 947 } 948 } else if (pkt->proto == IPPROTO_UDP) { 949 q->expire = time_second + dyn_udp_lifetime; 950 } else { 951 /* other protocols */ 952 q->expire = time_second + dyn_short_lifetime; 953 } 954done: 955 if (match_direction) 956 *match_direction = dir; 957 return q; 958} 959 960static ipfw_dyn_rule * 961lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 962 struct tcphdr *tcp) 963{ 964 ipfw_dyn_rule *q; 965 966 IPFW_DYN_LOCK(); 967 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 968 if (q == NULL) 969 IPFW_DYN_UNLOCK(); 970 /* NB: return table locked when q is not NULL */ 971 return q; 972} 973 974static void 975realloc_dynamic_table(void) 976{ 977 IPFW_DYN_LOCK_ASSERT(); 978 979 /* 980 * Try reallocation, make sure we have a power of 2 and do 981 * not allow more than 64k entries. In case of overflow, 982 * default to 1024. 983 */ 984 985 if (dyn_buckets > 65536) 986 dyn_buckets = 1024; 987 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 988 dyn_buckets = curr_dyn_buckets; /* reset */ 989 return; 990 } 991 curr_dyn_buckets = dyn_buckets; 992 if (ipfw_dyn_v != NULL) 993 free(ipfw_dyn_v, M_IPFW); 994 for (;;) { 995 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 996 M_IPFW, M_NOWAIT | M_ZERO); 997 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 998 break; 999 curr_dyn_buckets /= 2; 1000 } 1001} 1002 1003/** 1004 * Install state of type 'type' for a dynamic session. 1005 * The hash table contains two type of rules: 1006 * - regular rules (O_KEEP_STATE) 1007 * - rules for sessions with limited number of sess per user 1008 * (O_LIMIT). When they are created, the parent is 1009 * increased by 1, and decreased on delete. In this case, 1010 * the third parameter is the parent rule and not the chain. 1011 * - "parent" rules for the above (O_LIMIT_PARENT). 1012 */ 1013static ipfw_dyn_rule * 1014add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1015{ 1016 ipfw_dyn_rule *r; 1017 int i; 1018 1019 IPFW_DYN_LOCK_ASSERT(); 1020 1021 if (ipfw_dyn_v == NULL || 1022 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1023 realloc_dynamic_table(); 1024 if (ipfw_dyn_v == NULL) 1025 return NULL; /* failed ! */ 1026 } 1027 i = hash_packet(id); 1028 1029 r = malloc(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO); 1030 if (r == NULL) { 1031 printf ("ipfw: sorry cannot allocate state\n"); 1032 return NULL; 1033 } 1034 1035 /* increase refcount on parent, and set pointer */ 1036 if (dyn_type == O_LIMIT) { 1037 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1038 if ( parent->dyn_type != O_LIMIT_PARENT) 1039 panic("invalid parent"); 1040 parent->count++; 1041 r->parent = parent; 1042 rule = parent->rule; 1043 } 1044 1045 r->id = *id; 1046 r->expire = time_second + dyn_syn_lifetime; 1047 r->rule = rule; 1048 r->dyn_type = dyn_type; 1049 r->pcnt = r->bcnt = 0; 1050 r->count = 0; 1051 1052 r->bucket = i; 1053 r->next = ipfw_dyn_v[i]; 1054 ipfw_dyn_v[i] = r; 1055 dyn_count++; 1056 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1057 dyn_type, 1058 (r->id.src_ip), (r->id.src_port), 1059 (r->id.dst_ip), (r->id.dst_port), 1060 dyn_count ); ) 1061 return r; 1062} 1063 1064/** 1065 * lookup dynamic parent rule using pkt and rule as search keys. 1066 * If the lookup fails, then install one. 1067 */ 1068static ipfw_dyn_rule * 1069lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1070{ 1071 ipfw_dyn_rule *q; 1072 int i; 1073 1074 IPFW_DYN_LOCK_ASSERT(); 1075 1076 if (ipfw_dyn_v) { 1077 i = hash_packet( pkt ); 1078 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1079 if (q->dyn_type == O_LIMIT_PARENT && 1080 rule== q->rule && 1081 pkt->proto == q->id.proto && 1082 pkt->src_ip == q->id.src_ip && 1083 pkt->dst_ip == q->id.dst_ip && 1084 pkt->src_port == q->id.src_port && 1085 pkt->dst_port == q->id.dst_port) { 1086 q->expire = time_second + dyn_short_lifetime; 1087 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1088 return q; 1089 } 1090 } 1091 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1092} 1093 1094/** 1095 * Install dynamic state for rule type cmd->o.opcode 1096 * 1097 * Returns 1 (failure) if state is not installed because of errors or because 1098 * session limitations are enforced. 1099 */ 1100static int 1101install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1102 struct ip_fw_args *args) 1103{ 1104 static int last_log; 1105 1106 ipfw_dyn_rule *q; 1107 1108 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 1109 cmd->o.opcode, 1110 (args->f_id.src_ip), (args->f_id.src_port), 1111 (args->f_id.dst_ip), (args->f_id.dst_port) );) 1112 1113 IPFW_DYN_LOCK(); 1114 1115 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1116 1117 if (q != NULL) { /* should never occur */ 1118 if (last_log != time_second) { 1119 last_log = time_second; 1120 printf("ipfw: install_state: entry already present, done\n"); 1121 } 1122 IPFW_DYN_UNLOCK(); 1123 return 0; 1124 } 1125 1126 if (dyn_count >= dyn_max) 1127 /* 1128 * Run out of slots, try to remove any expired rule. 1129 */ 1130 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1131 1132 if (dyn_count >= dyn_max) { 1133 if (last_log != time_second) { 1134 last_log = time_second; 1135 printf("ipfw: install_state: Too many dynamic rules\n"); 1136 } 1137 IPFW_DYN_UNLOCK(); 1138 return 1; /* cannot install, notify caller */ 1139 } 1140 1141 switch (cmd->o.opcode) { 1142 case O_KEEP_STATE: /* bidir rule */ 1143 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1144 break; 1145 1146 case O_LIMIT: /* limit number of sessions */ 1147 { 1148 u_int16_t limit_mask = cmd->limit_mask; 1149 struct ipfw_flow_id id; 1150 ipfw_dyn_rule *parent; 1151 1152 DEB(printf("ipfw: installing dyn-limit rule %d\n", 1153 cmd->conn_limit);) 1154 1155 id.dst_ip = id.src_ip = 0; 1156 id.dst_port = id.src_port = 0; 1157 id.proto = args->f_id.proto; 1158 1159 if (limit_mask & DYN_SRC_ADDR) 1160 id.src_ip = args->f_id.src_ip; 1161 if (limit_mask & DYN_DST_ADDR) 1162 id.dst_ip = args->f_id.dst_ip; 1163 if (limit_mask & DYN_SRC_PORT) 1164 id.src_port = args->f_id.src_port; 1165 if (limit_mask & DYN_DST_PORT) 1166 id.dst_port = args->f_id.dst_port; 1167 parent = lookup_dyn_parent(&id, rule); 1168 if (parent == NULL) { 1169 printf("ipfw: add parent failed\n"); 1170 return 1; 1171 } 1172 if (parent->count >= cmd->conn_limit) { 1173 /* 1174 * See if we can remove some expired rule. 1175 */ 1176 remove_dyn_rule(rule, parent); 1177 if (parent->count >= cmd->conn_limit) { 1178 if (fw_verbose && last_log != time_second) { 1179 last_log = time_second; 1180 log(LOG_SECURITY | LOG_DEBUG, 1181 "drop session, too many entries\n"); 1182 } 1183 IPFW_DYN_UNLOCK(); 1184 return 1; 1185 } 1186 } 1187 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1188 } 1189 break; 1190 default: 1191 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 1192 IPFW_DYN_UNLOCK(); 1193 return 1; 1194 } 1195 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1196 IPFW_DYN_UNLOCK(); 1197 return 0; 1198} 1199 1200/* 1201 * Transmit a TCP packet, containing either a RST or a keepalive. 1202 * When flags & TH_RST, we are sending a RST packet, because of a 1203 * "reset" action matched the packet. 1204 * Otherwise we are sending a keepalive, and flags & TH_ 1205 */ 1206static void 1207send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1208{ 1209 struct mbuf *m; 1210 struct ip *ip; 1211 struct tcphdr *tcp; 1212 1213 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1214 if (m == 0) 1215 return; 1216 m->m_pkthdr.rcvif = (struct ifnet *)0; 1217 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1218 m->m_data += max_linkhdr; 1219 1220 ip = mtod(m, struct ip *); 1221 bzero(ip, m->m_len); 1222 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1223 ip->ip_p = IPPROTO_TCP; 1224 tcp->th_off = 5; 1225 /* 1226 * Assume we are sending a RST (or a keepalive in the reverse 1227 * direction), swap src and destination addresses and ports. 1228 */ 1229 ip->ip_src.s_addr = htonl(id->dst_ip); 1230 ip->ip_dst.s_addr = htonl(id->src_ip); 1231 tcp->th_sport = htons(id->dst_port); 1232 tcp->th_dport = htons(id->src_port); 1233 if (flags & TH_RST) { /* we are sending a RST */ 1234 if (flags & TH_ACK) { 1235 tcp->th_seq = htonl(ack); 1236 tcp->th_ack = htonl(0); 1237 tcp->th_flags = TH_RST; 1238 } else { 1239 if (flags & TH_SYN) 1240 seq++; 1241 tcp->th_seq = htonl(0); 1242 tcp->th_ack = htonl(seq); 1243 tcp->th_flags = TH_RST | TH_ACK; 1244 } 1245 } else { 1246 /* 1247 * We are sending a keepalive. flags & TH_SYN determines 1248 * the direction, forward if set, reverse if clear. 1249 * NOTE: seq and ack are always assumed to be correct 1250 * as set by the caller. This may be confusing... 1251 */ 1252 if (flags & TH_SYN) { 1253 /* 1254 * we have to rewrite the correct addresses! 1255 */ 1256 ip->ip_dst.s_addr = htonl(id->dst_ip); 1257 ip->ip_src.s_addr = htonl(id->src_ip); 1258 tcp->th_dport = htons(id->dst_port); 1259 tcp->th_sport = htons(id->src_port); 1260 } 1261 tcp->th_seq = htonl(seq); 1262 tcp->th_ack = htonl(ack); 1263 tcp->th_flags = TH_ACK; 1264 } 1265 /* 1266 * set ip_len to the payload size so we can compute 1267 * the tcp checksum on the pseudoheader 1268 * XXX check this, could save a couple of words ? 1269 */ 1270 ip->ip_len = htons(sizeof(struct tcphdr)); 1271 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1272 /* 1273 * now fill fields left out earlier 1274 */ 1275 ip->ip_ttl = ip_defttl; 1276 ip->ip_len = m->m_pkthdr.len; 1277 m->m_flags |= M_SKIP_FIREWALL; 1278 ip_output(m, NULL, NULL, 0, NULL, NULL); 1279} 1280 1281/* 1282 * sends a reject message, consuming the mbuf passed as an argument. 1283 */ 1284static void 1285send_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 1286{ 1287 1288 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1289 /* We need the IP header in host order for icmp_error(). */ 1290 if (args->eh != NULL) { 1291 struct ip *ip = mtod(args->m, struct ip *); 1292 ip->ip_len = ntohs(ip->ip_len); 1293 ip->ip_off = ntohs(ip->ip_off); 1294 } 1295 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1296 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1297 struct tcphdr *const tcp = 1298 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1299 if ( (tcp->th_flags & TH_RST) == 0) 1300 send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1301 ntohl(tcp->th_ack), 1302 tcp->th_flags | TH_RST); 1303 m_freem(args->m); 1304 } else 1305 m_freem(args->m); 1306 args->m = NULL; 1307} 1308 1309/** 1310 * 1311 * Given an ip_fw *, lookup_next_rule will return a pointer 1312 * to the next rule, which can be either the jump 1313 * target (for skipto instructions) or the next one in the list (in 1314 * all other cases including a missing jump target). 1315 * The result is also written in the "next_rule" field of the rule. 1316 * Backward jumps are not allowed, so start looking from the next 1317 * rule... 1318 * 1319 * This never returns NULL -- in case we do not have an exact match, 1320 * the next rule is returned. When the ruleset is changed, 1321 * pointers are flushed so we are always correct. 1322 */ 1323 1324static struct ip_fw * 1325lookup_next_rule(struct ip_fw *me) 1326{ 1327 struct ip_fw *rule = NULL; 1328 ipfw_insn *cmd; 1329 1330 /* look for action, in case it is a skipto */ 1331 cmd = ACTION_PTR(me); 1332 if (cmd->opcode == O_LOG) 1333 cmd += F_LEN(cmd); 1334 if (cmd->opcode == O_ALTQ) 1335 cmd += F_LEN(cmd); 1336 if ( cmd->opcode == O_SKIPTO ) 1337 for (rule = me->next; rule ; rule = rule->next) 1338 if (rule->rulenum >= cmd->arg1) 1339 break; 1340 if (rule == NULL) /* failure or not a skipto */ 1341 rule = me->next; 1342 me->next_rule = rule; 1343 return rule; 1344} 1345 1346static void 1347init_tables(void) 1348{ 1349 int i; 1350 1351 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1352 rn_inithead((void **)&ipfw_tables[i].rnh, 32); 1353 ipfw_tables[i].modified = 1; 1354 } 1355} 1356 1357static int 1358add_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen, u_int32_t value) 1359{ 1360 struct radix_node_head *rnh; 1361 struct table_entry *ent; 1362 1363 if (tbl >= IPFW_TABLES_MAX) 1364 return (EINVAL); 1365 rnh = ipfw_tables[tbl].rnh; 1366 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1367 if (ent == NULL) 1368 return (ENOMEM); 1369 ent->value = value; 1370 ent->addr.sin_len = ent->mask.sin_len = 8; 1371 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1372 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1373 RADIX_NODE_HEAD_LOCK(rnh); 1374 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1375 NULL) { 1376 RADIX_NODE_HEAD_UNLOCK(rnh); 1377 free(ent, M_IPFW_TBL); 1378 return (EEXIST); 1379 } 1380 ipfw_tables[tbl].modified = 1; 1381 RADIX_NODE_HEAD_UNLOCK(rnh); 1382 return (0); 1383} 1384 1385static int 1386del_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen) 1387{ 1388 struct radix_node_head *rnh; 1389 struct table_entry *ent; 1390 struct sockaddr_in sa, mask; 1391 1392 if (tbl >= IPFW_TABLES_MAX) 1393 return (EINVAL); 1394 rnh = ipfw_tables[tbl].rnh; 1395 sa.sin_len = mask.sin_len = 8; 1396 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1397 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1398 RADIX_NODE_HEAD_LOCK(rnh); 1399 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1400 if (ent == NULL) { 1401 RADIX_NODE_HEAD_UNLOCK(rnh); 1402 return (ESRCH); 1403 } 1404 ipfw_tables[tbl].modified = 1; 1405 RADIX_NODE_HEAD_UNLOCK(rnh); 1406 free(ent, M_IPFW_TBL); 1407 return (0); 1408} 1409 1410static int 1411flush_table_entry(struct radix_node *rn, void *arg) 1412{ 1413 struct radix_node_head * const rnh = arg; 1414 struct table_entry *ent; 1415 1416 ent = (struct table_entry *) 1417 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1418 if (ent != NULL) 1419 free(ent, M_IPFW_TBL); 1420 return (0); 1421} 1422 1423static int 1424flush_table(u_int16_t tbl) 1425{ 1426 struct radix_node_head *rnh; 1427 1428 if (tbl >= IPFW_TABLES_MAX) 1429 return (EINVAL); 1430 rnh = ipfw_tables[tbl].rnh; 1431 RADIX_NODE_HEAD_LOCK(rnh); 1432 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1433 ipfw_tables[tbl].modified = 1; 1434 RADIX_NODE_HEAD_UNLOCK(rnh); 1435 return (0); 1436} 1437 1438static void 1439flush_tables(void) 1440{ 1441 u_int16_t tbl; 1442 1443 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1444 flush_table(tbl); 1445} 1446 1447static int 1448lookup_table(u_int16_t tbl, in_addr_t addr, u_int32_t *val) 1449{ 1450 struct radix_node_head *rnh; 1451 struct table_entry *ent; 1452 struct sockaddr_in sa; 1453 static in_addr_t last_addr; 1454 static int last_tbl; 1455 static int last_match; 1456 static u_int32_t last_value; 1457 1458 if (tbl >= IPFW_TABLES_MAX) 1459 return (0); 1460 if (tbl == last_tbl && addr == last_addr && 1461 !ipfw_tables[tbl].modified) { 1462 if (last_match) 1463 *val = last_value; 1464 return (last_match); 1465 } 1466 rnh = ipfw_tables[tbl].rnh; 1467 sa.sin_len = 8; 1468 sa.sin_addr.s_addr = addr; 1469 RADIX_NODE_HEAD_LOCK(rnh); 1470 ipfw_tables[tbl].modified = 0; 1471 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1472 RADIX_NODE_HEAD_UNLOCK(rnh); 1473 last_addr = addr; 1474 last_tbl = tbl; 1475 if (ent != NULL) { 1476 last_value = *val = ent->value; 1477 last_match = 1; 1478 return (1); 1479 } 1480 last_match = 0; 1481 return (0); 1482} 1483 1484static int 1485count_table_entry(struct radix_node *rn, void *arg) 1486{ 1487 u_int32_t * const cnt = arg; 1488 1489 (*cnt)++; 1490 return (0); 1491} 1492 1493static int 1494count_table(u_int32_t tbl, u_int32_t *cnt) 1495{ 1496 struct radix_node_head *rnh; 1497 1498 if (tbl >= IPFW_TABLES_MAX) 1499 return (EINVAL); 1500 rnh = ipfw_tables[tbl].rnh; 1501 *cnt = 0; 1502 RADIX_NODE_HEAD_LOCK(rnh); 1503 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1504 RADIX_NODE_HEAD_UNLOCK(rnh); 1505 return (0); 1506} 1507 1508static int 1509dump_table_entry(struct radix_node *rn, void *arg) 1510{ 1511 struct table_entry * const n = (struct table_entry *)rn; 1512 ipfw_table * const tbl = arg; 1513 ipfw_table_entry *ent; 1514 1515 if (tbl->cnt == tbl->size) 1516 return (1); 1517 ent = &tbl->ent[tbl->cnt]; 1518 ent->tbl = tbl->tbl; 1519 if (in_nullhost(n->mask.sin_addr)) 1520 ent->masklen = 0; 1521 else 1522 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1523 ent->addr = n->addr.sin_addr.s_addr; 1524 ent->value = n->value; 1525 tbl->cnt++; 1526 return (0); 1527} 1528 1529static int 1530dump_table(ipfw_table *tbl) 1531{ 1532 struct radix_node_head *rnh; 1533 1534 if (tbl->tbl >= IPFW_TABLES_MAX) 1535 return (EINVAL); 1536 rnh = ipfw_tables[tbl->tbl].rnh; 1537 tbl->cnt = 0; 1538 RADIX_NODE_HEAD_LOCK(rnh); 1539 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1540 RADIX_NODE_HEAD_UNLOCK(rnh); 1541 return (0); 1542} 1543 1544static void 1545fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1546{ 1547 struct ucred *cr; 1548 1549 if (inp->inp_socket != NULL) { 1550 cr = inp->inp_socket->so_cred; 1551 ugp->fw_prid = jailed(cr) ? 1552 cr->cr_prison->pr_id : -1; 1553 ugp->fw_uid = cr->cr_uid; 1554 ugp->fw_ngroups = cr->cr_ngroups; 1555 bcopy(cr->cr_groups, ugp->fw_groups, 1556 sizeof(ugp->fw_groups)); 1557 } 1558} 1559 1560static int 1561check_uidgid(ipfw_insn_u32 *insn, 1562 int proto, struct ifnet *oif, 1563 struct in_addr dst_ip, u_int16_t dst_port, 1564 struct in_addr src_ip, u_int16_t src_port, 1565 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1566{ 1567 struct inpcbinfo *pi; 1568 int wildcard; 1569 struct inpcb *pcb; 1570 int match; 1571 gid_t *gp; 1572 1573 /* 1574 * Check to see if the UDP or TCP stack supplied us with 1575 * the PCB. If so, rather then holding a lock and looking 1576 * up the PCB, we can use the one that was supplied. 1577 */ 1578 if (inp && *lookup == 0) { 1579 INP_LOCK_ASSERT(inp); 1580 if (inp->inp_socket != NULL) { 1581 fill_ugid_cache(inp, ugp); 1582 *lookup = 1; 1583 } 1584 } 1585 /* 1586 * If we have already been here and the packet has no 1587 * PCB entry associated with it, then we can safely 1588 * assume that this is a no match. 1589 */ 1590 if (*lookup == -1) 1591 return (0); 1592 if (proto == IPPROTO_TCP) { 1593 wildcard = 0; 1594 pi = &tcbinfo; 1595 } else if (proto == IPPROTO_UDP) { 1596 wildcard = 1; 1597 pi = &udbinfo; 1598 } else 1599 return 0; 1600 match = 0; 1601 if (*lookup == 0) { 1602 INP_INFO_RLOCK(pi); 1603 pcb = (oif) ? 1604 in_pcblookup_hash(pi, 1605 dst_ip, htons(dst_port), 1606 src_ip, htons(src_port), 1607 wildcard, oif) : 1608 in_pcblookup_hash(pi, 1609 src_ip, htons(src_port), 1610 dst_ip, htons(dst_port), 1611 wildcard, NULL); 1612 if (pcb != NULL) { 1613 INP_LOCK(pcb); 1614 if (pcb->inp_socket != NULL) { 1615 fill_ugid_cache(pcb, ugp); 1616 *lookup = 1; 1617 } 1618 INP_UNLOCK(pcb); 1619 } 1620 INP_INFO_RUNLOCK(pi); 1621 if (*lookup == 0) { 1622 /* 1623 * If the lookup did not yield any results, there 1624 * is no sense in coming back and trying again. So 1625 * we can set lookup to -1 and ensure that we wont 1626 * bother the pcb system again. 1627 */ 1628 *lookup = -1; 1629 return (0); 1630 } 1631 } 1632 if (insn->o.opcode == O_UID) 1633 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1634 else if (insn->o.opcode == O_GID) { 1635 for (gp = ugp->fw_groups; 1636 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1637 if (*gp == (gid_t)insn->d[0]) { 1638 match = 1; 1639 break; 1640 } 1641 } else if (insn->o.opcode == O_JAIL) 1642 match = (ugp->fw_prid == (int)insn->d[0]); 1643 return match; 1644} 1645 1646/* 1647 * The main check routine for the firewall. 1648 * 1649 * All arguments are in args so we can modify them and return them 1650 * back to the caller. 1651 * 1652 * Parameters: 1653 * 1654 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1655 * Starts with the IP header. 1656 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1657 * args->oif Outgoing interface, or NULL if packet is incoming. 1658 * The incoming interface is in the mbuf. (in) 1659 * args->divert_rule (in/out) 1660 * Skip up to the first rule past this rule number; 1661 * upon return, non-zero port number for divert or tee. 1662 * 1663 * args->rule Pointer to the last matching rule (in/out) 1664 * args->next_hop Socket we are forwarding to (out). 1665 * args->f_id Addresses grabbed from the packet (out) 1666 * 1667 * Return value: 1668 * 1669 * IP_FW_PORT_DENY_FLAG the packet must be dropped. 1670 * 0 The packet is to be accepted and routed normally OR 1671 * the packet was denied/rejected and has been dropped; 1672 * in the latter case, *m is equal to NULL upon return. 1673 * port Divert the packet to port, with these caveats: 1674 * 1675 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead 1676 * of diverting it (ie, 'ipfw tee'). 1677 * 1678 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower 1679 * 16 bits as a dummynet pipe number instead of diverting 1680 */ 1681 1682int 1683ipfw_chk(struct ip_fw_args *args) 1684{ 1685 /* 1686 * Local variables hold state during the processing of a packet. 1687 * 1688 * IMPORTANT NOTE: to speed up the processing of rules, there 1689 * are some assumption on the values of the variables, which 1690 * are documented here. Should you change them, please check 1691 * the implementation of the various instructions to make sure 1692 * that they still work. 1693 * 1694 * args->eh The MAC header. It is non-null for a layer2 1695 * packet, it is NULL for a layer-3 packet. 1696 * 1697 * m | args->m Pointer to the mbuf, as received from the caller. 1698 * It may change if ipfw_chk() does an m_pullup, or if it 1699 * consumes the packet because it calls send_reject(). 1700 * XXX This has to change, so that ipfw_chk() never modifies 1701 * or consumes the buffer. 1702 * ip is simply an alias of the value of m, and it is kept 1703 * in sync with it (the packet is supposed to start with 1704 * the ip header). 1705 */ 1706 struct mbuf *m = args->m; 1707 struct ip *ip = mtod(m, struct ip *); 1708 1709 /* 1710 * For rules which contain uid/gid or jail constraints, cache 1711 * a copy of the users credentials after the pcb lookup has been 1712 * executed. This will speed up the processing of rules with 1713 * these types of constraints, as well as decrease contention 1714 * on pcb related locks. 1715 */ 1716 struct ip_fw_ugid fw_ugid_cache; 1717 int ugid_lookup = 0; 1718 1719 /* 1720 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 1721 * associated with a packet input on a divert socket. This 1722 * will allow to distinguish traffic and its direction when 1723 * it originates from a divert socket. 1724 */ 1725 u_int divinput_flags = 0; 1726 1727 /* 1728 * oif | args->oif If NULL, ipfw_chk has been called on the 1729 * inbound path (ether_input, bdg_forward, ip_input). 1730 * If non-NULL, ipfw_chk has been called on the outbound path 1731 * (ether_output, ip_output). 1732 */ 1733 struct ifnet *oif = args->oif; 1734 1735 struct ip_fw *f = NULL; /* matching rule */ 1736 int retval = 0; 1737 1738 /* 1739 * hlen The length of the IPv4 header. 1740 * hlen >0 means we have an IPv4 packet. 1741 */ 1742 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1743 1744 /* 1745 * offset The offset of a fragment. offset != 0 means that 1746 * we have a fragment at this offset of an IPv4 packet. 1747 * offset == 0 means that (if this is an IPv4 packet) 1748 * this is the first or only fragment. 1749 */ 1750 u_short offset = 0; 1751 1752 /* 1753 * Local copies of addresses. They are only valid if we have 1754 * an IP packet. 1755 * 1756 * proto The protocol. Set to 0 for non-ip packets, 1757 * or to the protocol read from the packet otherwise. 1758 * proto != 0 means that we have an IPv4 packet. 1759 * 1760 * src_port, dst_port port numbers, in HOST format. Only 1761 * valid for TCP and UDP packets. 1762 * 1763 * src_ip, dst_ip ip addresses, in NETWORK format. 1764 * Only valid for IPv4 packets. 1765 */ 1766 u_int8_t proto; 1767 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1768 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1769 u_int16_t ip_len=0; 1770 int pktlen; 1771 int dyn_dir = MATCH_UNKNOWN; 1772 ipfw_dyn_rule *q = NULL; 1773 struct ip_fw_chain *chain = &layer3_chain; 1774 struct m_tag *mtag; 1775 1776 if (m->m_flags & M_SKIP_FIREWALL) 1777 return 0; /* accept */ 1778 /* 1779 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1780 * MATCH_NONE when checked and not matched (q = NULL), 1781 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1782 */ 1783 1784 pktlen = m->m_pkthdr.len; 1785 if (args->eh == NULL || /* layer 3 packet */ 1786 ( m->m_pkthdr.len >= sizeof(struct ip) && 1787 ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 1788 hlen = ip->ip_hl << 2; 1789 1790 /* 1791 * Collect parameters into local variables for faster matching. 1792 */ 1793 if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 1794 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1795 goto after_ip_checks; 1796 } 1797 1798 proto = args->f_id.proto = ip->ip_p; 1799 src_ip = ip->ip_src; 1800 dst_ip = ip->ip_dst; 1801 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1802 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1803 ip_len = ntohs(ip->ip_len); 1804 } else { 1805 offset = ip->ip_off & IP_OFFMASK; 1806 ip_len = ip->ip_len; 1807 } 1808 pktlen = ip_len < pktlen ? ip_len : pktlen; 1809 1810#define PULLUP_TO(len) \ 1811 do { \ 1812 if ((m)->m_len < (len)) { \ 1813 args->m = m = m_pullup(m, (len)); \ 1814 if (m == 0) \ 1815 goto pullup_failed; \ 1816 ip = mtod(m, struct ip *); \ 1817 } \ 1818 } while (0) 1819 1820 if (offset == 0) { 1821 switch (proto) { 1822 case IPPROTO_TCP: 1823 { 1824 struct tcphdr *tcp; 1825 1826 PULLUP_TO(hlen + sizeof(struct tcphdr)); 1827 tcp = L3HDR(struct tcphdr, ip); 1828 dst_port = tcp->th_dport; 1829 src_port = tcp->th_sport; 1830 args->f_id.flags = tcp->th_flags; 1831 } 1832 break; 1833 1834 case IPPROTO_UDP: 1835 { 1836 struct udphdr *udp; 1837 1838 PULLUP_TO(hlen + sizeof(struct udphdr)); 1839 udp = L3HDR(struct udphdr, ip); 1840 dst_port = udp->uh_dport; 1841 src_port = udp->uh_sport; 1842 } 1843 break; 1844 1845 case IPPROTO_ICMP: 1846 PULLUP_TO(hlen + 4); /* type, code and checksum. */ 1847 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 1848 break; 1849 1850 default: 1851 break; 1852 } 1853#undef PULLUP_TO 1854 } 1855 1856 args->f_id.src_ip = ntohl(src_ip.s_addr); 1857 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1858 args->f_id.src_port = src_port = ntohs(src_port); 1859 args->f_id.dst_port = dst_port = ntohs(dst_port); 1860 1861after_ip_checks: 1862 IPFW_LOCK(chain); /* XXX expensive? can we run lock free? */ 1863 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 1864 if (args->rule) { 1865 /* 1866 * Packet has already been tagged. Look for the next rule 1867 * to restart processing. 1868 * 1869 * If fw_one_pass != 0 then just accept it. 1870 * XXX should not happen here, but optimized out in 1871 * the caller. 1872 */ 1873 if (fw_one_pass) { 1874 IPFW_UNLOCK(chain); /* XXX optimize */ 1875 return 0; 1876 } 1877 1878 f = args->rule->next_rule; 1879 if (f == NULL) 1880 f = lookup_next_rule(args->rule); 1881 } else { 1882 /* 1883 * Find the starting rule. It can be either the first 1884 * one, or the one after divert_rule if asked so. 1885 */ 1886 int skipto = mtag ? divert_cookie(mtag) : 0; 1887 1888 f = chain->rules; 1889 if (args->eh == NULL && skipto != 0) { 1890 if (skipto >= IPFW_DEFAULT_RULE) { 1891 IPFW_UNLOCK(chain); 1892 return(IP_FW_PORT_DENY_FLAG); /* invalid */ 1893 } 1894 while (f && f->rulenum <= skipto) 1895 f = f->next; 1896 if (f == NULL) { /* drop packet */ 1897 IPFW_UNLOCK(chain); 1898 return(IP_FW_PORT_DENY_FLAG); 1899 } 1900 } 1901 } 1902 /* reset divert rule to avoid confusion later */ 1903 if (mtag) { 1904 divinput_flags = divert_info(mtag) & 1905 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 1906 m_tag_delete(m, mtag); 1907 } 1908 1909 /* 1910 * Now scan the rules, and parse microinstructions for each rule. 1911 */ 1912 for (; f; f = f->next) { 1913 int l, cmdlen; 1914 ipfw_insn *cmd; 1915 int skip_or; /* skip rest of OR block */ 1916 1917again: 1918 if (set_disable & (1 << f->set) ) 1919 continue; 1920 1921 skip_or = 0; 1922 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1923 l -= cmdlen, cmd += cmdlen) { 1924 int match; 1925 1926 /* 1927 * check_body is a jump target used when we find a 1928 * CHECK_STATE, and need to jump to the body of 1929 * the target rule. 1930 */ 1931 1932check_body: 1933 cmdlen = F_LEN(cmd); 1934 /* 1935 * An OR block (insn_1 || .. || insn_n) has the 1936 * F_OR bit set in all but the last instruction. 1937 * The first match will set "skip_or", and cause 1938 * the following instructions to be skipped until 1939 * past the one with the F_OR bit clear. 1940 */ 1941 if (skip_or) { /* skip this instruction */ 1942 if ((cmd->len & F_OR) == 0) 1943 skip_or = 0; /* next one is good */ 1944 continue; 1945 } 1946 match = 0; /* set to 1 if we succeed */ 1947 1948 switch (cmd->opcode) { 1949 /* 1950 * The first set of opcodes compares the packet's 1951 * fields with some pattern, setting 'match' if a 1952 * match is found. At the end of the loop there is 1953 * logic to deal with F_NOT and F_OR flags associated 1954 * with the opcode. 1955 */ 1956 case O_NOP: 1957 match = 1; 1958 break; 1959 1960 case O_FORWARD_MAC: 1961 printf("ipfw: opcode %d unimplemented\n", 1962 cmd->opcode); 1963 break; 1964 1965 case O_GID: 1966 case O_UID: 1967 case O_JAIL: 1968 /* 1969 * We only check offset == 0 && proto != 0, 1970 * as this ensures that we have an IPv4 1971 * packet with the ports info. 1972 */ 1973 if (offset!=0) 1974 break; 1975 if (proto == IPPROTO_TCP || 1976 proto == IPPROTO_UDP) 1977 match = check_uidgid( 1978 (ipfw_insn_u32 *)cmd, 1979 proto, oif, 1980 dst_ip, dst_port, 1981 src_ip, src_port, &fw_ugid_cache, 1982 &ugid_lookup, args->inp); 1983 break; 1984 1985 case O_RECV: 1986 match = iface_match(m->m_pkthdr.rcvif, 1987 (ipfw_insn_if *)cmd); 1988 break; 1989 1990 case O_XMIT: 1991 match = iface_match(oif, (ipfw_insn_if *)cmd); 1992 break; 1993 1994 case O_VIA: 1995 match = iface_match(oif ? oif : 1996 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1997 break; 1998 1999 case O_MACADDR2: 2000 if (args->eh != NULL) { /* have MAC header */ 2001 u_int32_t *want = (u_int32_t *) 2002 ((ipfw_insn_mac *)cmd)->addr; 2003 u_int32_t *mask = (u_int32_t *) 2004 ((ipfw_insn_mac *)cmd)->mask; 2005 u_int32_t *hdr = (u_int32_t *)args->eh; 2006 2007 match = 2008 ( want[0] == (hdr[0] & mask[0]) && 2009 want[1] == (hdr[1] & mask[1]) && 2010 want[2] == (hdr[2] & mask[2]) ); 2011 } 2012 break; 2013 2014 case O_MAC_TYPE: 2015 if (args->eh != NULL) { 2016 u_int16_t t = 2017 ntohs(args->eh->ether_type); 2018 u_int16_t *p = 2019 ((ipfw_insn_u16 *)cmd)->ports; 2020 int i; 2021 2022 for (i = cmdlen - 1; !match && i>0; 2023 i--, p += 2) 2024 match = (t>=p[0] && t<=p[1]); 2025 } 2026 break; 2027 2028 case O_FRAG: 2029 match = (hlen > 0 && offset != 0); 2030 break; 2031 2032 case O_IN: /* "out" is "not in" */ 2033 match = (oif == NULL); 2034 break; 2035 2036 case O_LAYER2: 2037 match = (args->eh != NULL); 2038 break; 2039 2040 case O_DIVERTED: 2041 match = (cmd->arg1 & 1 && divinput_flags & 2042 IP_FW_DIVERT_LOOPBACK_FLAG) || 2043 (cmd->arg1 & 2 && divinput_flags & 2044 IP_FW_DIVERT_OUTPUT_FLAG); 2045 break; 2046 2047 case O_PROTO: 2048 /* 2049 * We do not allow an arg of 0 so the 2050 * check of "proto" only suffices. 2051 */ 2052 match = (proto == cmd->arg1); 2053 break; 2054 2055 case O_IP_SRC: 2056 match = (hlen > 0 && 2057 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2058 src_ip.s_addr); 2059 break; 2060 2061 case O_IP_SRC_LOOKUP: 2062 case O_IP_DST_LOOKUP: 2063 if (hlen > 0) { 2064 uint32_t a = 2065 (cmd->opcode == O_IP_DST_LOOKUP) ? 2066 dst_ip.s_addr : src_ip.s_addr; 2067 uint32_t v; 2068 2069 match = lookup_table(cmd->arg1, a, &v); 2070 if (!match) 2071 break; 2072 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2073 match = 2074 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2075 } 2076 break; 2077 2078 case O_IP_SRC_MASK: 2079 case O_IP_DST_MASK: 2080 if (hlen > 0) { 2081 uint32_t a = 2082 (cmd->opcode == O_IP_DST_MASK) ? 2083 dst_ip.s_addr : src_ip.s_addr; 2084 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2085 int i = cmdlen-1; 2086 2087 for (; !match && i>0; i-= 2, p+= 2) 2088 match = (p[0] == (a & p[1])); 2089 } 2090 break; 2091 2092 case O_IP_SRC_ME: 2093 if (hlen > 0) { 2094 struct ifnet *tif; 2095 2096 INADDR_TO_IFP(src_ip, tif); 2097 match = (tif != NULL); 2098 } 2099 break; 2100 2101 case O_IP_DST_SET: 2102 case O_IP_SRC_SET: 2103 if (hlen > 0) { 2104 u_int32_t *d = (u_int32_t *)(cmd+1); 2105 u_int32_t addr = 2106 cmd->opcode == O_IP_DST_SET ? 2107 args->f_id.dst_ip : 2108 args->f_id.src_ip; 2109 2110 if (addr < d[0]) 2111 break; 2112 addr -= d[0]; /* subtract base */ 2113 match = (addr < cmd->arg1) && 2114 ( d[ 1 + (addr>>5)] & 2115 (1<<(addr & 0x1f)) ); 2116 } 2117 break; 2118 2119 case O_IP_DST: 2120 match = (hlen > 0 && 2121 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2122 dst_ip.s_addr); 2123 break; 2124 2125 case O_IP_DST_ME: 2126 if (hlen > 0) { 2127 struct ifnet *tif; 2128 2129 INADDR_TO_IFP(dst_ip, tif); 2130 match = (tif != NULL); 2131 } 2132 break; 2133 2134 case O_IP_SRCPORT: 2135 case O_IP_DSTPORT: 2136 /* 2137 * offset == 0 && proto != 0 is enough 2138 * to guarantee that we have an IPv4 2139 * packet with port info. 2140 */ 2141 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2142 && offset == 0) { 2143 u_int16_t x = 2144 (cmd->opcode == O_IP_SRCPORT) ? 2145 src_port : dst_port ; 2146 u_int16_t *p = 2147 ((ipfw_insn_u16 *)cmd)->ports; 2148 int i; 2149 2150 for (i = cmdlen - 1; !match && i>0; 2151 i--, p += 2) 2152 match = (x>=p[0] && x<=p[1]); 2153 } 2154 break; 2155 2156 case O_ICMPTYPE: 2157 match = (offset == 0 && proto==IPPROTO_ICMP && 2158 icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 2159 break; 2160 2161 case O_IPOPT: 2162 match = (hlen > 0 && ipopts_match(ip, cmd) ); 2163 break; 2164 2165 case O_IPVER: 2166 match = (hlen > 0 && cmd->arg1 == ip->ip_v); 2167 break; 2168 2169 case O_IPID: 2170 case O_IPLEN: 2171 case O_IPTTL: 2172 if (hlen > 0) { /* only for IP packets */ 2173 uint16_t x; 2174 uint16_t *p; 2175 int i; 2176 2177 if (cmd->opcode == O_IPLEN) 2178 x = ip_len; 2179 else if (cmd->opcode == O_IPTTL) 2180 x = ip->ip_ttl; 2181 else /* must be IPID */ 2182 x = ntohs(ip->ip_id); 2183 if (cmdlen == 1) { 2184 match = (cmd->arg1 == x); 2185 break; 2186 } 2187 /* otherwise we have ranges */ 2188 p = ((ipfw_insn_u16 *)cmd)->ports; 2189 i = cmdlen - 1; 2190 for (; !match && i>0; i--, p += 2) 2191 match = (x >= p[0] && x <= p[1]); 2192 } 2193 break; 2194 2195 case O_IPPRECEDENCE: 2196 match = (hlen > 0 && 2197 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2198 break; 2199 2200 case O_IPTOS: 2201 match = (hlen > 0 && 2202 flags_match(cmd, ip->ip_tos)); 2203 break; 2204 2205 case O_TCPDATALEN: 2206 if (proto == IPPROTO_TCP && offset == 0) { 2207 struct tcphdr *tcp; 2208 uint16_t x; 2209 uint16_t *p; 2210 int i; 2211 2212 tcp = L3HDR(struct tcphdr,ip); 2213 x = ip_len - 2214 ((ip->ip_hl + tcp->th_off) << 2); 2215 if (cmdlen == 1) { 2216 match = (cmd->arg1 == x); 2217 break; 2218 } 2219 /* otherwise we have ranges */ 2220 p = ((ipfw_insn_u16 *)cmd)->ports; 2221 i = cmdlen - 1; 2222 for (; !match && i>0; i--, p += 2) 2223 match = (x >= p[0] && x <= p[1]); 2224 } 2225 break; 2226 2227 case O_TCPFLAGS: 2228 match = (proto == IPPROTO_TCP && offset == 0 && 2229 flags_match(cmd, 2230 L3HDR(struct tcphdr,ip)->th_flags)); 2231 break; 2232 2233 case O_TCPOPTS: 2234 match = (proto == IPPROTO_TCP && offset == 0 && 2235 tcpopts_match(ip, cmd)); 2236 break; 2237 2238 case O_TCPSEQ: 2239 match = (proto == IPPROTO_TCP && offset == 0 && 2240 ((ipfw_insn_u32 *)cmd)->d[0] == 2241 L3HDR(struct tcphdr,ip)->th_seq); 2242 break; 2243 2244 case O_TCPACK: 2245 match = (proto == IPPROTO_TCP && offset == 0 && 2246 ((ipfw_insn_u32 *)cmd)->d[0] == 2247 L3HDR(struct tcphdr,ip)->th_ack); 2248 break; 2249 2250 case O_TCPWIN: 2251 match = (proto == IPPROTO_TCP && offset == 0 && 2252 cmd->arg1 == 2253 L3HDR(struct tcphdr,ip)->th_win); 2254 break; 2255 2256 case O_ESTAB: 2257 /* reject packets which have SYN only */ 2258 /* XXX should i also check for TH_ACK ? */ 2259 match = (proto == IPPROTO_TCP && offset == 0 && 2260 (L3HDR(struct tcphdr,ip)->th_flags & 2261 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2262 break; 2263 2264 case O_ALTQ: { 2265 struct altq_tag *at; 2266 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2267 2268 match = 1; 2269 mtag = m_tag_get(PACKET_TAG_PF_QID, 2270 sizeof(struct altq_tag), 2271 M_NOWAIT); 2272 if (mtag == NULL) { 2273 /* 2274 * Let the packet fall back to the 2275 * default ALTQ. 2276 */ 2277 break; 2278 } 2279 at = (struct altq_tag *)(mtag+1); 2280 at->qid = altq->qid; 2281 if (hlen != 0) 2282 at->af = AF_INET; 2283 else 2284 at->af = AF_LINK; 2285 at->hdr = ip; 2286 m_tag_prepend(m, mtag); 2287 break; 2288 } 2289 2290 case O_LOG: 2291 if (fw_verbose) 2292 ipfw_log(f, hlen, args->eh, m, oif); 2293 match = 1; 2294 break; 2295 2296 case O_PROB: 2297 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2298 break; 2299 2300 case O_VERREVPATH: 2301 /* Outgoing packets automatically pass/match */ 2302 match = (hlen > 0 && ((oif != NULL) || 2303 (m->m_pkthdr.rcvif == NULL) || 2304 verify_path(src_ip, m->m_pkthdr.rcvif))); 2305 break; 2306 2307 case O_VERSRCREACH: 2308 /* Outgoing packets automatically pass/match */ 2309 match = (hlen > 0 && ((oif != NULL) || 2310 verify_path(src_ip, NULL))); 2311 break; 2312 2313 case O_ANTISPOOF: 2314 /* Outgoing packets automatically pass/match */ 2315 if (oif == NULL && hlen > 0 && 2316 in_localaddr(src_ip)) 2317 match = verify_path(src_ip, 2318 m->m_pkthdr.rcvif); 2319 else 2320 match = 1; 2321 break; 2322 2323 case O_IPSEC: 2324#ifdef FAST_IPSEC 2325 match = (m_tag_find(m, 2326 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2327#endif 2328#ifdef IPSEC 2329 match = (ipsec_getnhist(m) != 0); 2330#endif 2331 /* otherwise no match */ 2332 break; 2333 2334 /* 2335 * The second set of opcodes represents 'actions', 2336 * i.e. the terminal part of a rule once the packet 2337 * matches all previous patterns. 2338 * Typically there is only one action for each rule, 2339 * and the opcode is stored at the end of the rule 2340 * (but there are exceptions -- see below). 2341 * 2342 * In general, here we set retval and terminate the 2343 * outer loop (would be a 'break 3' in some language, 2344 * but we need to do a 'goto done'). 2345 * 2346 * Exceptions: 2347 * O_COUNT and O_SKIPTO actions: 2348 * instead of terminating, we jump to the next rule 2349 * ('goto next_rule', equivalent to a 'break 2'), 2350 * or to the SKIPTO target ('goto again' after 2351 * having set f, cmd and l), respectively. 2352 * 2353 * O_LOG and O_ALTQ action parameters: 2354 * perform some action and set match = 1; 2355 * 2356 * O_LIMIT and O_KEEP_STATE: these opcodes are 2357 * not real 'actions', and are stored right 2358 * before the 'action' part of the rule. 2359 * These opcodes try to install an entry in the 2360 * state tables; if successful, we continue with 2361 * the next opcode (match=1; break;), otherwise 2362 * the packet * must be dropped 2363 * ('goto done' after setting retval); 2364 * 2365 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2366 * cause a lookup of the state table, and a jump 2367 * to the 'action' part of the parent rule 2368 * ('goto check_body') if an entry is found, or 2369 * (CHECK_STATE only) a jump to the next rule if 2370 * the entry is not found ('goto next_rule'). 2371 * The result of the lookup is cached to make 2372 * further instances of these opcodes are 2373 * effectively NOPs. 2374 */ 2375 case O_LIMIT: 2376 case O_KEEP_STATE: 2377 if (install_state(f, 2378 (ipfw_insn_limit *)cmd, args)) { 2379 retval = IP_FW_PORT_DENY_FLAG; 2380 goto done; /* error/limit violation */ 2381 } 2382 match = 1; 2383 break; 2384 2385 case O_PROBE_STATE: 2386 case O_CHECK_STATE: 2387 /* 2388 * dynamic rules are checked at the first 2389 * keep-state or check-state occurrence, 2390 * with the result being stored in dyn_dir. 2391 * The compiler introduces a PROBE_STATE 2392 * instruction for us when we have a 2393 * KEEP_STATE (because PROBE_STATE needs 2394 * to be run first). 2395 */ 2396 if (dyn_dir == MATCH_UNKNOWN && 2397 (q = lookup_dyn_rule(&args->f_id, 2398 &dyn_dir, proto == IPPROTO_TCP ? 2399 L3HDR(struct tcphdr, ip) : NULL)) 2400 != NULL) { 2401 /* 2402 * Found dynamic entry, update stats 2403 * and jump to the 'action' part of 2404 * the parent rule. 2405 */ 2406 q->pcnt++; 2407 q->bcnt += pktlen; 2408 f = q->rule; 2409 cmd = ACTION_PTR(f); 2410 l = f->cmd_len - f->act_ofs; 2411 IPFW_DYN_UNLOCK(); 2412 goto check_body; 2413 } 2414 /* 2415 * Dynamic entry not found. If CHECK_STATE, 2416 * skip to next rule, if PROBE_STATE just 2417 * ignore and continue with next opcode. 2418 */ 2419 if (cmd->opcode == O_CHECK_STATE) 2420 goto next_rule; 2421 match = 1; 2422 break; 2423 2424 case O_ACCEPT: 2425 retval = 0; /* accept */ 2426 goto done; 2427 2428 case O_PIPE: 2429 case O_QUEUE: 2430 args->rule = f; /* report matching rule */ 2431 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG; 2432 goto done; 2433 2434 case O_DIVERT: 2435 case O_TEE: { 2436 struct divert_tag *dt; 2437 2438 if (args->eh) /* not on layer 2 */ 2439 break; 2440 mtag = m_tag_get(PACKET_TAG_DIVERT, 2441 sizeof(struct divert_tag), 2442 M_NOWAIT); 2443 if (mtag == NULL) { 2444 /* XXX statistic */ 2445 /* drop packet */ 2446 IPFW_UNLOCK(chain); 2447 return IP_FW_PORT_DENY_FLAG; 2448 } 2449 dt = (struct divert_tag *)(mtag+1); 2450 dt->cookie = f->rulenum; 2451 dt->info = (cmd->opcode == O_DIVERT) ? 2452 cmd->arg1 : 2453 cmd->arg1 | IP_FW_PORT_TEE_FLAG; 2454 m_tag_prepend(m, mtag); 2455 retval = dt->info; 2456 goto done; 2457 } 2458 2459 case O_COUNT: 2460 case O_SKIPTO: 2461 f->pcnt++; /* update stats */ 2462 f->bcnt += pktlen; 2463 f->timestamp = time_second; 2464 if (cmd->opcode == O_COUNT) 2465 goto next_rule; 2466 /* handle skipto */ 2467 if (f->next_rule == NULL) 2468 lookup_next_rule(f); 2469 f = f->next_rule; 2470 goto again; 2471 2472 case O_REJECT: 2473 /* 2474 * Drop the packet and send a reject notice 2475 * if the packet is not ICMP (or is an ICMP 2476 * query), and it is not multicast/broadcast. 2477 */ 2478 if (hlen > 0 && 2479 (proto != IPPROTO_ICMP || 2480 is_icmp_query(ip)) && 2481 !(m->m_flags & (M_BCAST|M_MCAST)) && 2482 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2483 send_reject(args, cmd->arg1, 2484 offset,ip_len); 2485 m = args->m; 2486 } 2487 /* FALLTHROUGH */ 2488 case O_DENY: 2489 retval = IP_FW_PORT_DENY_FLAG; 2490 goto done; 2491 2492 case O_FORWARD_IP: 2493 if (args->eh) /* not valid on layer2 pkts */ 2494 break; 2495 if (!q || dyn_dir == MATCH_FORWARD) 2496 args->next_hop = 2497 &((ipfw_insn_sa *)cmd)->sa; 2498 retval = 0; 2499 goto done; 2500 2501 default: 2502 panic("-- unknown opcode %d\n", cmd->opcode); 2503 } /* end of switch() on opcodes */ 2504 2505 if (cmd->len & F_NOT) 2506 match = !match; 2507 2508 if (match) { 2509 if (cmd->len & F_OR) 2510 skip_or = 1; 2511 } else { 2512 if (!(cmd->len & F_OR)) /* not an OR block, */ 2513 break; /* try next rule */ 2514 } 2515 2516 } /* end of inner for, scan opcodes */ 2517 2518next_rule:; /* try next rule */ 2519 2520 } /* end of outer for, scan rules */ 2521 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2522 IPFW_UNLOCK(chain); 2523 return(IP_FW_PORT_DENY_FLAG); 2524 2525done: 2526 /* Update statistics */ 2527 f->pcnt++; 2528 f->bcnt += pktlen; 2529 f->timestamp = time_second; 2530 IPFW_UNLOCK(chain); 2531 return retval; 2532 2533pullup_failed: 2534 if (fw_verbose) 2535 printf("ipfw: pullup failed\n"); 2536 return(IP_FW_PORT_DENY_FLAG); 2537} 2538 2539/* 2540 * When a rule is added/deleted, clear the next_rule pointers in all rules. 2541 * These will be reconstructed on the fly as packets are matched. 2542 */ 2543static void 2544flush_rule_ptrs(struct ip_fw_chain *chain) 2545{ 2546 struct ip_fw *rule; 2547 2548 IPFW_LOCK_ASSERT(chain); 2549 2550 for (rule = chain->rules; rule; rule = rule->next) 2551 rule->next_rule = NULL; 2552} 2553 2554/* 2555 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 2556 * pipe/queue, or to all of them (match == NULL). 2557 */ 2558void 2559flush_pipe_ptrs(struct dn_flow_set *match) 2560{ 2561 struct ip_fw *rule; 2562 2563 IPFW_LOCK(&layer3_chain); 2564 for (rule = layer3_chain.rules; rule; rule = rule->next) { 2565 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 2566 2567 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 2568 continue; 2569 /* 2570 * XXX Use bcmp/bzero to handle pipe_ptr to overcome 2571 * possible alignment problems on 64-bit architectures. 2572 * This code is seldom used so we do not worry too 2573 * much about efficiency. 2574 */ 2575 if (match == NULL || 2576 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) ) 2577 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr)); 2578 } 2579 IPFW_UNLOCK(&layer3_chain); 2580} 2581 2582/* 2583 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 2584 * possibly create a rule number and add the rule to the list. 2585 * Update the rule_number in the input struct so the caller knows it as well. 2586 */ 2587static int 2588add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 2589{ 2590 struct ip_fw *rule, *f, *prev; 2591 int l = RULESIZE(input_rule); 2592 2593 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 2594 return (EINVAL); 2595 2596 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 2597 if (rule == NULL) 2598 return (ENOSPC); 2599 2600 bcopy(input_rule, rule, l); 2601 2602 rule->next = NULL; 2603 rule->next_rule = NULL; 2604 2605 rule->pcnt = 0; 2606 rule->bcnt = 0; 2607 rule->timestamp = 0; 2608 2609 IPFW_LOCK(chain); 2610 2611 if (chain->rules == NULL) { /* default rule */ 2612 chain->rules = rule; 2613 goto done; 2614 } 2615 2616 /* 2617 * If rulenum is 0, find highest numbered rule before the 2618 * default rule, and add autoinc_step 2619 */ 2620 if (autoinc_step < 1) 2621 autoinc_step = 1; 2622 else if (autoinc_step > 1000) 2623 autoinc_step = 1000; 2624 if (rule->rulenum == 0) { 2625 /* 2626 * locate the highest numbered rule before default 2627 */ 2628 for (f = chain->rules; f; f = f->next) { 2629 if (f->rulenum == IPFW_DEFAULT_RULE) 2630 break; 2631 rule->rulenum = f->rulenum; 2632 } 2633 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2634 rule->rulenum += autoinc_step; 2635 input_rule->rulenum = rule->rulenum; 2636 } 2637 2638 /* 2639 * Now insert the new rule in the right place in the sorted list. 2640 */ 2641 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 2642 if (f->rulenum > rule->rulenum) { /* found the location */ 2643 if (prev) { 2644 rule->next = f; 2645 prev->next = rule; 2646 } else { /* head insert */ 2647 rule->next = chain->rules; 2648 chain->rules = rule; 2649 } 2650 break; 2651 } 2652 } 2653 flush_rule_ptrs(chain); 2654done: 2655 static_count++; 2656 static_len += l; 2657 IPFW_UNLOCK(chain); 2658 DEB(printf("ipfw: installed rule %d, static count now %d\n", 2659 rule->rulenum, static_count);) 2660 return (0); 2661} 2662 2663/** 2664 * Remove a static rule (including derived * dynamic rules) 2665 * and place it on the ``reap list'' for later reclamation. 2666 * The caller is in charge of clearing rule pointers to avoid 2667 * dangling pointers. 2668 * @return a pointer to the next entry. 2669 * Arguments are not checked, so they better be correct. 2670 */ 2671static struct ip_fw * 2672remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 2673{ 2674 struct ip_fw *n; 2675 int l = RULESIZE(rule); 2676 2677 IPFW_LOCK_ASSERT(chain); 2678 2679 n = rule->next; 2680 IPFW_DYN_LOCK(); 2681 remove_dyn_rule(rule, NULL /* force removal */); 2682 IPFW_DYN_UNLOCK(); 2683 if (prev == NULL) 2684 chain->rules = n; 2685 else 2686 prev->next = n; 2687 static_count--; 2688 static_len -= l; 2689 2690 rule->next = chain->reap; 2691 chain->reap = rule; 2692 2693 return n; 2694} 2695 2696/** 2697 * Reclaim storage associated with a list of rules. This is 2698 * typically the list created using remove_rule. 2699 */ 2700static void 2701reap_rules(struct ip_fw *head) 2702{ 2703 struct ip_fw *rule; 2704 2705 while ((rule = head) != NULL) { 2706 head = head->next; 2707 if (DUMMYNET_LOADED) 2708 ip_dn_ruledel_ptr(rule); 2709 free(rule, M_IPFW); 2710 } 2711} 2712 2713/* 2714 * Remove all rules from a chain (except rules in set RESVD_SET 2715 * unless kill_default = 1). The caller is responsible for 2716 * reclaiming storage for the rules left in chain->reap. 2717 */ 2718static void 2719free_chain(struct ip_fw_chain *chain, int kill_default) 2720{ 2721 struct ip_fw *prev, *rule; 2722 2723 IPFW_LOCK_ASSERT(chain); 2724 2725 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 2726 for (prev = NULL, rule = chain->rules; rule ; ) 2727 if (kill_default || rule->set != RESVD_SET) 2728 rule = remove_rule(chain, rule, prev); 2729 else { 2730 prev = rule; 2731 rule = rule->next; 2732 } 2733} 2734 2735/** 2736 * Remove all rules with given number, and also do set manipulation. 2737 * Assumes chain != NULL && *chain != NULL. 2738 * 2739 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2740 * the next 8 bits are the new set, the top 8 bits are the command: 2741 * 2742 * 0 delete rules with given number 2743 * 1 delete rules with given set number 2744 * 2 move rules with given number to new set 2745 * 3 move rules with given set number to new set 2746 * 4 swap sets with given numbers 2747 */ 2748static int 2749del_entry(struct ip_fw_chain *chain, u_int32_t arg) 2750{ 2751 struct ip_fw *prev = NULL, *rule; 2752 u_int16_t rulenum; /* rule or old_set */ 2753 u_int8_t cmd, new_set; 2754 2755 rulenum = arg & 0xffff; 2756 cmd = (arg >> 24) & 0xff; 2757 new_set = (arg >> 16) & 0xff; 2758 2759 if (cmd > 4) 2760 return EINVAL; 2761 if (new_set > RESVD_SET) 2762 return EINVAL; 2763 if (cmd == 0 || cmd == 2) { 2764 if (rulenum >= IPFW_DEFAULT_RULE) 2765 return EINVAL; 2766 } else { 2767 if (rulenum > RESVD_SET) /* old_set */ 2768 return EINVAL; 2769 } 2770 2771 IPFW_LOCK(chain); 2772 rule = chain->rules; 2773 chain->reap = NULL; 2774 switch (cmd) { 2775 case 0: /* delete rules with given number */ 2776 /* 2777 * locate first rule to delete 2778 */ 2779 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 2780 ; 2781 if (rule->rulenum != rulenum) { 2782 IPFW_UNLOCK(chain); 2783 return EINVAL; 2784 } 2785 2786 /* 2787 * flush pointers outside the loop, then delete all matching 2788 * rules. prev remains the same throughout the cycle. 2789 */ 2790 flush_rule_ptrs(chain); 2791 while (rule->rulenum == rulenum) 2792 rule = remove_rule(chain, rule, prev); 2793 break; 2794 2795 case 1: /* delete all rules with given set number */ 2796 flush_rule_ptrs(chain); 2797 rule = chain->rules; 2798 while (rule->rulenum < IPFW_DEFAULT_RULE) 2799 if (rule->set == rulenum) 2800 rule = remove_rule(chain, rule, prev); 2801 else { 2802 prev = rule; 2803 rule = rule->next; 2804 } 2805 break; 2806 2807 case 2: /* move rules with given number to new set */ 2808 rule = chain->rules; 2809 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2810 if (rule->rulenum == rulenum) 2811 rule->set = new_set; 2812 break; 2813 2814 case 3: /* move rules with given set number to new set */ 2815 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2816 if (rule->set == rulenum) 2817 rule->set = new_set; 2818 break; 2819 2820 case 4: /* swap two sets */ 2821 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2822 if (rule->set == rulenum) 2823 rule->set = new_set; 2824 else if (rule->set == new_set) 2825 rule->set = rulenum; 2826 break; 2827 } 2828 /* 2829 * Look for rules to reclaim. We grab the list before 2830 * releasing the lock then reclaim them w/o the lock to 2831 * avoid a LOR with dummynet. 2832 */ 2833 rule = chain->reap; 2834 chain->reap = NULL; 2835 IPFW_UNLOCK(chain); 2836 if (rule) 2837 reap_rules(rule); 2838 return 0; 2839} 2840 2841/* 2842 * Clear counters for a specific rule. 2843 * The enclosing "table" is assumed locked. 2844 */ 2845static void 2846clear_counters(struct ip_fw *rule, int log_only) 2847{ 2848 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 2849 2850 if (log_only == 0) { 2851 rule->bcnt = rule->pcnt = 0; 2852 rule->timestamp = 0; 2853 } 2854 if (l->o.opcode == O_LOG) 2855 l->log_left = l->max_log; 2856} 2857 2858/** 2859 * Reset some or all counters on firewall rules. 2860 * @arg frwl is null to clear all entries, or contains a specific 2861 * rule number. 2862 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 2863 */ 2864static int 2865zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 2866{ 2867 struct ip_fw *rule; 2868 char *msg; 2869 2870 IPFW_LOCK(chain); 2871 if (rulenum == 0) { 2872 norule_counter = 0; 2873 for (rule = chain->rules; rule; rule = rule->next) 2874 clear_counters(rule, log_only); 2875 msg = log_only ? "ipfw: All logging counts reset.\n" : 2876 "ipfw: Accounting cleared.\n"; 2877 } else { 2878 int cleared = 0; 2879 /* 2880 * We can have multiple rules with the same number, so we 2881 * need to clear them all. 2882 */ 2883 for (rule = chain->rules; rule; rule = rule->next) 2884 if (rule->rulenum == rulenum) { 2885 while (rule && rule->rulenum == rulenum) { 2886 clear_counters(rule, log_only); 2887 rule = rule->next; 2888 } 2889 cleared = 1; 2890 break; 2891 } 2892 if (!cleared) { /* we did not find any matching rules */ 2893 IPFW_UNLOCK(chain); 2894 return (EINVAL); 2895 } 2896 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 2897 "ipfw: Entry %d cleared.\n"; 2898 } 2899 IPFW_UNLOCK(chain); 2900 2901 if (fw_verbose) 2902 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 2903 return (0); 2904} 2905 2906/* 2907 * Check validity of the structure before insert. 2908 * Fortunately rules are simple, so this mostly need to check rule sizes. 2909 */ 2910static int 2911check_ipfw_struct(struct ip_fw *rule, int size) 2912{ 2913 int l, cmdlen = 0; 2914 int have_action=0; 2915 ipfw_insn *cmd; 2916 2917 if (size < sizeof(*rule)) { 2918 printf("ipfw: rule too short\n"); 2919 return (EINVAL); 2920 } 2921 /* first, check for valid size */ 2922 l = RULESIZE(rule); 2923 if (l != size) { 2924 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 2925 return (EINVAL); 2926 } 2927 if (rule->act_ofs >= rule->cmd_len) { 2928 printf("ipfw: bogus action offset (%u > %u)\n", 2929 rule->act_ofs, rule->cmd_len - 1); 2930 return (EINVAL); 2931 } 2932 /* 2933 * Now go for the individual checks. Very simple ones, basically only 2934 * instruction sizes. 2935 */ 2936 for (l = rule->cmd_len, cmd = rule->cmd ; 2937 l > 0 ; l -= cmdlen, cmd += cmdlen) { 2938 cmdlen = F_LEN(cmd); 2939 if (cmdlen > l) { 2940 printf("ipfw: opcode %d size truncated\n", 2941 cmd->opcode); 2942 return EINVAL; 2943 } 2944 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 2945 switch (cmd->opcode) { 2946 case O_PROBE_STATE: 2947 case O_KEEP_STATE: 2948 case O_PROTO: 2949 case O_IP_SRC_ME: 2950 case O_IP_DST_ME: 2951 case O_LAYER2: 2952 case O_IN: 2953 case O_FRAG: 2954 case O_DIVERTED: 2955 case O_IPOPT: 2956 case O_IPTOS: 2957 case O_IPPRECEDENCE: 2958 case O_IPVER: 2959 case O_TCPWIN: 2960 case O_TCPFLAGS: 2961 case O_TCPOPTS: 2962 case O_ESTAB: 2963 case O_VERREVPATH: 2964 case O_VERSRCREACH: 2965 case O_ANTISPOOF: 2966 case O_IPSEC: 2967 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2968 goto bad_size; 2969 break; 2970 2971 case O_UID: 2972 case O_GID: 2973 case O_JAIL: 2974 case O_IP_SRC: 2975 case O_IP_DST: 2976 case O_TCPSEQ: 2977 case O_TCPACK: 2978 case O_PROB: 2979 case O_ICMPTYPE: 2980 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2981 goto bad_size; 2982 break; 2983 2984 case O_LIMIT: 2985 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 2986 goto bad_size; 2987 break; 2988 2989 case O_LOG: 2990 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 2991 goto bad_size; 2992 2993 ((ipfw_insn_log *)cmd)->log_left = 2994 ((ipfw_insn_log *)cmd)->max_log; 2995 2996 break; 2997 2998 case O_IP_SRC_MASK: 2999 case O_IP_DST_MASK: 3000 /* only odd command lengths */ 3001 if ( !(cmdlen & 1) || cmdlen > 31) 3002 goto bad_size; 3003 break; 3004 3005 case O_IP_SRC_SET: 3006 case O_IP_DST_SET: 3007 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3008 printf("ipfw: invalid set size %d\n", 3009 cmd->arg1); 3010 return EINVAL; 3011 } 3012 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3013 (cmd->arg1+31)/32 ) 3014 goto bad_size; 3015 break; 3016 3017 case O_IP_SRC_LOOKUP: 3018 case O_IP_DST_LOOKUP: 3019 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3020 printf("ipfw: invalid table number %d\n", 3021 cmd->arg1); 3022 return (EINVAL); 3023 } 3024 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3025 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3026 goto bad_size; 3027 break; 3028 3029 case O_MACADDR2: 3030 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3031 goto bad_size; 3032 break; 3033 3034 case O_NOP: 3035 case O_IPID: 3036 case O_IPTTL: 3037 case O_IPLEN: 3038 case O_TCPDATALEN: 3039 if (cmdlen < 1 || cmdlen > 31) 3040 goto bad_size; 3041 break; 3042 3043 case O_MAC_TYPE: 3044 case O_IP_SRCPORT: 3045 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3046 if (cmdlen < 2 || cmdlen > 31) 3047 goto bad_size; 3048 break; 3049 3050 case O_RECV: 3051 case O_XMIT: 3052 case O_VIA: 3053 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3054 goto bad_size; 3055 break; 3056 3057 case O_ALTQ: 3058 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3059 goto bad_size; 3060 break; 3061 3062 case O_PIPE: 3063 case O_QUEUE: 3064 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 3065 goto bad_size; 3066 goto check_action; 3067 3068 case O_FORWARD_IP: 3069#ifdef IPFIREWALL_FORWARD 3070 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3071 goto bad_size; 3072 goto check_action; 3073#else 3074 return EINVAL; 3075#endif 3076 3077 case O_DIVERT: 3078 case O_TEE: 3079 if (ip_divert_ptr == NULL) 3080 return EINVAL; 3081 case O_FORWARD_MAC: /* XXX not implemented yet */ 3082 case O_CHECK_STATE: 3083 case O_COUNT: 3084 case O_ACCEPT: 3085 case O_DENY: 3086 case O_REJECT: 3087 case O_SKIPTO: 3088 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3089 goto bad_size; 3090check_action: 3091 if (have_action) { 3092 printf("ipfw: opcode %d, multiple actions" 3093 " not allowed\n", 3094 cmd->opcode); 3095 return EINVAL; 3096 } 3097 have_action = 1; 3098 if (l != cmdlen) { 3099 printf("ipfw: opcode %d, action must be" 3100 " last opcode\n", 3101 cmd->opcode); 3102 return EINVAL; 3103 } 3104 break; 3105 default: 3106 printf("ipfw: opcode %d, unknown opcode\n", 3107 cmd->opcode); 3108 return EINVAL; 3109 } 3110 } 3111 if (have_action == 0) { 3112 printf("ipfw: missing action\n"); 3113 return EINVAL; 3114 } 3115 return 0; 3116 3117bad_size: 3118 printf("ipfw: opcode %d size %d wrong\n", 3119 cmd->opcode, cmdlen); 3120 return EINVAL; 3121} 3122 3123/* 3124 * Copy the static and dynamic rules to the supplied buffer 3125 * and return the amount of space actually used. 3126 */ 3127static size_t 3128ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3129{ 3130 char *bp = buf; 3131 char *ep = bp + space; 3132 struct ip_fw *rule; 3133 int i; 3134 3135 /* XXX this can take a long time and locking will block packet flow */ 3136 IPFW_LOCK(chain); 3137 for (rule = chain->rules; rule ; rule = rule->next) { 3138 /* 3139 * Verify the entry fits in the buffer in case the 3140 * rules changed between calculating buffer space and 3141 * now. This would be better done using a generation 3142 * number but should suffice for now. 3143 */ 3144 i = RULESIZE(rule); 3145 if (bp + i <= ep) { 3146 bcopy(rule, bp, i); 3147 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3148 sizeof(set_disable)); 3149 bp += i; 3150 } 3151 } 3152 IPFW_UNLOCK(chain); 3153 if (ipfw_dyn_v) { 3154 ipfw_dyn_rule *p, *last = NULL; 3155 3156 IPFW_DYN_LOCK(); 3157 for (i = 0 ; i < curr_dyn_buckets; i++) 3158 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3159 if (bp + sizeof *p <= ep) { 3160 ipfw_dyn_rule *dst = 3161 (ipfw_dyn_rule *)bp; 3162 bcopy(p, dst, sizeof *p); 3163 bcopy(&(p->rule->rulenum), &(dst->rule), 3164 sizeof(p->rule->rulenum)); 3165 /* 3166 * store a non-null value in "next". 3167 * The userland code will interpret a 3168 * NULL here as a marker 3169 * for the last dynamic rule. 3170 */ 3171 bcopy(&dst, &dst->next, sizeof(dst)); 3172 last = dst; 3173 dst->expire = 3174 TIME_LEQ(dst->expire, time_second) ? 3175 0 : dst->expire - time_second ; 3176 bp += sizeof(ipfw_dyn_rule); 3177 } 3178 } 3179 IPFW_DYN_UNLOCK(); 3180 if (last != NULL) /* mark last dynamic rule */ 3181 bzero(&last->next, sizeof(last)); 3182 } 3183 return (bp - (char *)buf); 3184} 3185 3186 3187/** 3188 * {set|get}sockopt parser. 3189 */ 3190static int 3191ipfw_ctl(struct sockopt *sopt) 3192{ 3193#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3194 int error, rule_num; 3195 size_t size; 3196 struct ip_fw *buf, *rule; 3197 u_int32_t rulenum[2]; 3198 3199 error = suser(sopt->sopt_td); 3200 if (error) 3201 return (error); 3202 3203 /* 3204 * Disallow modifications in really-really secure mode, but still allow 3205 * the logging counters to be reset. 3206 */ 3207 if (sopt->sopt_name == IP_FW_ADD || 3208 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3209#if __FreeBSD_version >= 500034 3210 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3211 if (error) 3212 return (error); 3213#else /* FreeBSD 4.x */ 3214 if (securelevel >= 3) 3215 return (EPERM); 3216#endif 3217 } 3218 3219 error = 0; 3220 3221 switch (sopt->sopt_name) { 3222 case IP_FW_GET: 3223 /* 3224 * pass up a copy of the current rules. Static rules 3225 * come first (the last of which has number IPFW_DEFAULT_RULE), 3226 * followed by a possibly empty list of dynamic rule. 3227 * The last dynamic rule has NULL in the "next" field. 3228 * 3229 * Note that the calculated size is used to bound the 3230 * amount of data returned to the user. The rule set may 3231 * change between calculating the size and returning the 3232 * data in which case we'll just return what fits. 3233 */ 3234 size = static_len; /* size of static rules */ 3235 if (ipfw_dyn_v) /* add size of dyn.rules */ 3236 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3237 3238 /* 3239 * XXX todo: if the user passes a short length just to know 3240 * how much room is needed, do not bother filling up the 3241 * buffer, just jump to the sooptcopyout. 3242 */ 3243 buf = malloc(size, M_TEMP, M_WAITOK); 3244 error = sooptcopyout(sopt, buf, 3245 ipfw_getrules(&layer3_chain, buf, size)); 3246 free(buf, M_TEMP); 3247 break; 3248 3249 case IP_FW_FLUSH: 3250 /* 3251 * Normally we cannot release the lock on each iteration. 3252 * We could do it here only because we start from the head all 3253 * the times so there is no risk of missing some entries. 3254 * On the other hand, the risk is that we end up with 3255 * a very inconsistent ruleset, so better keep the lock 3256 * around the whole cycle. 3257 * 3258 * XXX this code can be improved by resetting the head of 3259 * the list to point to the default rule, and then freeing 3260 * the old list without the need for a lock. 3261 */ 3262 3263 IPFW_LOCK(&layer3_chain); 3264 layer3_chain.reap = NULL; 3265 free_chain(&layer3_chain, 0 /* keep default rule */); 3266 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3267 IPFW_UNLOCK(&layer3_chain); 3268 if (layer3_chain.reap != NULL) 3269 reap_rules(rule); 3270 break; 3271 3272 case IP_FW_ADD: 3273 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3274 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3275 sizeof(struct ip_fw) ); 3276 if (error == 0) 3277 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3278 if (error == 0) { 3279 error = add_rule(&layer3_chain, rule); 3280 size = RULESIZE(rule); 3281 if (!error && sopt->sopt_dir == SOPT_GET) 3282 error = sooptcopyout(sopt, rule, size); 3283 } 3284 free(rule, M_TEMP); 3285 break; 3286 3287 case IP_FW_DEL: 3288 /* 3289 * IP_FW_DEL is used for deleting single rules or sets, 3290 * and (ab)used to atomically manipulate sets. Argument size 3291 * is used to distinguish between the two: 3292 * sizeof(u_int32_t) 3293 * delete single rule or set of rules, 3294 * or reassign rules (or sets) to a different set. 3295 * 2*sizeof(u_int32_t) 3296 * atomic disable/enable sets. 3297 * first u_int32_t contains sets to be disabled, 3298 * second u_int32_t contains sets to be enabled. 3299 */ 3300 error = sooptcopyin(sopt, rulenum, 3301 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3302 if (error) 3303 break; 3304 size = sopt->sopt_valsize; 3305 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3306 error = del_entry(&layer3_chain, rulenum[0]); 3307 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3308 set_disable = 3309 (set_disable | rulenum[0]) & ~rulenum[1] & 3310 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3311 else 3312 error = EINVAL; 3313 break; 3314 3315 case IP_FW_ZERO: 3316 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3317 rule_num = 0; 3318 if (sopt->sopt_val != 0) { 3319 error = sooptcopyin(sopt, &rule_num, 3320 sizeof(int), sizeof(int)); 3321 if (error) 3322 break; 3323 } 3324 error = zero_entry(&layer3_chain, rule_num, 3325 sopt->sopt_name == IP_FW_RESETLOG); 3326 break; 3327 3328 case IP_FW_TABLE_ADD: 3329 { 3330 ipfw_table_entry ent; 3331 3332 error = sooptcopyin(sopt, &ent, 3333 sizeof(ent), sizeof(ent)); 3334 if (error) 3335 break; 3336 error = add_table_entry(ent.tbl, ent.addr, 3337 ent.masklen, ent.value); 3338 } 3339 break; 3340 3341 case IP_FW_TABLE_DEL: 3342 { 3343 ipfw_table_entry ent; 3344 3345 error = sooptcopyin(sopt, &ent, 3346 sizeof(ent), sizeof(ent)); 3347 if (error) 3348 break; 3349 error = del_table_entry(ent.tbl, ent.addr, ent.masklen); 3350 } 3351 break; 3352 3353 case IP_FW_TABLE_FLUSH: 3354 { 3355 u_int16_t tbl; 3356 3357 error = sooptcopyin(sopt, &tbl, 3358 sizeof(tbl), sizeof(tbl)); 3359 if (error) 3360 break; 3361 error = flush_table(tbl); 3362 } 3363 break; 3364 3365 case IP_FW_TABLE_GETSIZE: 3366 { 3367 u_int32_t tbl, cnt; 3368 3369 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 3370 sizeof(tbl)))) 3371 break; 3372 if ((error = count_table(tbl, &cnt))) 3373 break; 3374 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 3375 } 3376 break; 3377 3378 case IP_FW_TABLE_LIST: 3379 { 3380 ipfw_table *tbl; 3381 3382 if (sopt->sopt_valsize < sizeof(*tbl)) { 3383 error = EINVAL; 3384 break; 3385 } 3386 size = sopt->sopt_valsize; 3387 tbl = malloc(size, M_TEMP, M_WAITOK); 3388 if (tbl == NULL) { 3389 error = ENOMEM; 3390 break; 3391 } 3392 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 3393 if (error) { 3394 free(tbl, M_TEMP); 3395 break; 3396 } 3397 tbl->size = (size - sizeof(*tbl)) / 3398 sizeof(ipfw_table_entry); 3399 error = dump_table(tbl); 3400 if (error) { 3401 free(tbl, M_TEMP); 3402 break; 3403 } 3404 error = sooptcopyout(sopt, tbl, size); 3405 free(tbl, M_TEMP); 3406 } 3407 break; 3408 3409 default: 3410 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 3411 error = EINVAL; 3412 } 3413 3414 return (error); 3415#undef RULE_MAXSIZE 3416} 3417 3418/** 3419 * dummynet needs a reference to the default rule, because rules can be 3420 * deleted while packets hold a reference to them. When this happens, 3421 * dummynet changes the reference to the default rule (it could well be a 3422 * NULL pointer, but this way we do not need to check for the special 3423 * case, plus here he have info on the default behaviour). 3424 */ 3425struct ip_fw *ip_fw_default_rule; 3426 3427/* 3428 * This procedure is only used to handle keepalives. It is invoked 3429 * every dyn_keepalive_period 3430 */ 3431static void 3432ipfw_tick(void * __unused unused) 3433{ 3434 int i; 3435 ipfw_dyn_rule *q; 3436 3437 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 3438 goto done; 3439 3440 IPFW_DYN_LOCK(); 3441 for (i = 0 ; i < curr_dyn_buckets ; i++) { 3442 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 3443 if (q->dyn_type == O_LIMIT_PARENT) 3444 continue; 3445 if (q->id.proto != IPPROTO_TCP) 3446 continue; 3447 if ( (q->state & BOTH_SYN) != BOTH_SYN) 3448 continue; 3449 if (TIME_LEQ( time_second+dyn_keepalive_interval, 3450 q->expire)) 3451 continue; /* too early */ 3452 if (TIME_LEQ(q->expire, time_second)) 3453 continue; /* too late, rule expired */ 3454 3455 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 3456 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 3457 } 3458 } 3459 IPFW_DYN_UNLOCK(); 3460done: 3461 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 3462} 3463 3464int 3465ipfw_init(void) 3466{ 3467 struct ip_fw default_rule; 3468 int error; 3469 3470 layer3_chain.rules = NULL; 3471 IPFW_LOCK_INIT(&layer3_chain); 3472 IPFW_DYN_LOCK_INIT(); 3473 callout_init(&ipfw_timeout, debug_mpsafenet ? CALLOUT_MPSAFE : 0); 3474 3475 bzero(&default_rule, sizeof default_rule); 3476 3477 default_rule.act_ofs = 0; 3478 default_rule.rulenum = IPFW_DEFAULT_RULE; 3479 default_rule.cmd_len = 1; 3480 default_rule.set = RESVD_SET; 3481 3482 default_rule.cmd[0].len = 1; 3483 default_rule.cmd[0].opcode = 3484#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 3485 1 ? O_ACCEPT : 3486#endif 3487 O_DENY; 3488 3489 error = add_rule(&layer3_chain, &default_rule); 3490 if (error != 0) { 3491 printf("ipfw2: error %u initializing default rule " 3492 "(support disabled)\n", error); 3493 IPFW_DYN_LOCK_DESTROY(); 3494 IPFW_LOCK_DESTROY(&layer3_chain); 3495 return (error); 3496 } 3497 3498 ip_fw_default_rule = layer3_chain.rules; 3499 printf("ipfw2 initialized, divert %s, " 3500 "rule-based forwarding " 3501#ifdef IPFIREWALL_FORWARD 3502 "enabled, " 3503#else 3504 "disabled, " 3505#endif 3506 "default to %s, logging ", 3507#ifdef IPDIVERT 3508 "enabled", 3509#else 3510 "disabled", 3511#endif 3512 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 3513 3514#ifdef IPFIREWALL_VERBOSE 3515 fw_verbose = 1; 3516#endif 3517#ifdef IPFIREWALL_VERBOSE_LIMIT 3518 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3519#endif 3520 if (fw_verbose == 0) 3521 printf("disabled\n"); 3522 else if (verbose_limit == 0) 3523 printf("unlimited\n"); 3524 else 3525 printf("limited to %d packets/entry by default\n", 3526 verbose_limit); 3527 3528 init_tables(); 3529 ip_fw_ctl_ptr = ipfw_ctl; 3530 ip_fw_chk_ptr = ipfw_chk; 3531 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 3532 3533 return (0); 3534} 3535 3536void 3537ipfw_destroy(void) 3538{ 3539 struct ip_fw *reap; 3540 3541 ip_fw_chk_ptr = NULL; 3542 ip_fw_ctl_ptr = NULL; 3543 callout_drain(&ipfw_timeout); 3544 IPFW_LOCK(&layer3_chain); 3545 layer3_chain.reap = NULL; 3546 free_chain(&layer3_chain, 1 /* kill default rule */); 3547 reap = layer3_chain.reap, layer3_chain.reap = NULL; 3548 IPFW_UNLOCK(&layer3_chain); 3549 if (reap != NULL) 3550 reap_rules(reap); 3551 flush_tables(); 3552 IPFW_DYN_LOCK_DESTROY(); 3553 IPFW_LOCK_DESTROY(&layer3_chain); 3554 printf("IP firewall unloaded\n"); 3555} 3556 3557#endif /* IPFW2 */ 3558