ip_fw2.c revision 178888
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 26#include <sys/cdefs.h> 27__FBSDID("$FreeBSD: head/sys/netinet/ip_fw2.c 178888 2008-05-09 23:03:00Z julian $"); 28 29#define DEB(x) 30#define DDB(x) x 31 32/* 33 * Implement IP packet firewall (new version) 34 */ 35 36#if !defined(KLD_MODULE) 37#include "opt_ipfw.h" 38#include "opt_ipdivert.h" 39#include "opt_ipdn.h" 40#include "opt_inet.h" 41#ifndef INET 42#error IPFIREWALL requires INET. 43#endif /* INET */ 44#endif 45#include "opt_inet6.h" 46#include "opt_ipsec.h" 47#include "opt_mac.h" 48 49#include <sys/param.h> 50#include <sys/systm.h> 51#include <sys/condvar.h> 52#include <sys/eventhandler.h> 53#include <sys/malloc.h> 54#include <sys/mbuf.h> 55#include <sys/kernel.h> 56#include <sys/lock.h> 57#include <sys/jail.h> 58#include <sys/module.h> 59#include <sys/priv.h> 60#include <sys/proc.h> 61#include <sys/rwlock.h> 62#include <sys/socket.h> 63#include <sys/socketvar.h> 64#include <sys/sysctl.h> 65#include <sys/syslog.h> 66#include <sys/ucred.h> 67#include <net/if.h> 68#include <net/radix.h> 69#include <net/route.h> 70#include <net/pf_mtag.h> 71 72#define IPFW_INTERNAL /* Access to protected data structures in ip_fw.h. */ 73 74#include <netinet/in.h> 75#include <netinet/in_systm.h> 76#include <netinet/in_var.h> 77#include <netinet/in_pcb.h> 78#include <netinet/ip.h> 79#include <netinet/ip_var.h> 80#include <netinet/ip_icmp.h> 81#include <netinet/ip_fw.h> 82#include <netinet/ip_divert.h> 83#include <netinet/ip_dummynet.h> 84#include <netinet/ip_carp.h> 85#include <netinet/pim.h> 86#include <netinet/tcp.h> 87#include <netinet/tcp_timer.h> 88#include <netinet/tcp_var.h> 89#include <netinet/tcpip.h> 90#include <netinet/udp.h> 91#include <netinet/udp_var.h> 92#include <netinet/sctp.h> 93#include <netgraph/ng_ipfw.h> 94 95#include <altq/if_altq.h> 96 97#include <netinet/ip6.h> 98#include <netinet/icmp6.h> 99#ifdef INET6 100#include <netinet6/scope6_var.h> 101#endif 102 103#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 104 105#include <machine/in_cksum.h> /* XXX for in_cksum */ 106 107#include <security/mac/mac_framework.h> 108 109/* 110 * set_disable contains one bit per set value (0..31). 111 * If the bit is set, all rules with the corresponding set 112 * are disabled. Set RESVD_SET(31) is reserved for the default rule 113 * and rules that are not deleted by the flush command, 114 * and CANNOT be disabled. 115 * Rules in set RESVD_SET can only be deleted explicitly. 116 */ 117static u_int32_t set_disable; 118 119static int fw_verbose; 120static int verbose_limit; 121 122static struct callout ipfw_timeout; 123static uma_zone_t ipfw_dyn_rule_zone; 124#define IPFW_DEFAULT_RULE 65535 125 126/* 127 * Data structure to cache our ucred related 128 * information. This structure only gets used if 129 * the user specified UID/GID based constraints in 130 * a firewall rule. 131 */ 132struct ip_fw_ugid { 133 gid_t fw_groups[NGROUPS]; 134 int fw_ngroups; 135 uid_t fw_uid; 136 int fw_prid; 137}; 138 139/* 140 * list of rules for layer 3 141 */ 142struct ip_fw_chain layer3_chain; 143 144MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 145MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 146#define IPFW_NAT_LOADED (ipfw_nat_ptr != NULL) 147ipfw_nat_t *ipfw_nat_ptr = NULL; 148ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 149ipfw_nat_cfg_t *ipfw_nat_del_ptr; 150ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 151ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 152 153struct table_entry { 154 struct radix_node rn[2]; 155 struct sockaddr_in addr, mask; 156 u_int32_t value; 157}; 158 159static int fw_debug = 1; 160static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 161 162extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 163 164#ifdef SYSCTL_NODE 165SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 166SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 167 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 168 ipfw_chg_hook, "I", "Enable ipfw"); 169SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 170 &autoinc_step, 0, "Rule number autincrement step"); 171SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 172 CTLFLAG_RW | CTLFLAG_SECURE3, 173 &fw_one_pass, 0, 174 "Only do a single pass through ipfw when using dummynet(4)"); 175SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 176 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 177SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 178 CTLFLAG_RW | CTLFLAG_SECURE3, 179 &fw_verbose, 0, "Log matches to ipfw rules"); 180SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 181 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 182 183/* 184 * Description of dynamic rules. 185 * 186 * Dynamic rules are stored in lists accessed through a hash table 187 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 188 * be modified through the sysctl variable dyn_buckets which is 189 * updated when the table becomes empty. 190 * 191 * XXX currently there is only one list, ipfw_dyn. 192 * 193 * When a packet is received, its address fields are first masked 194 * with the mask defined for the rule, then hashed, then matched 195 * against the entries in the corresponding list. 196 * Dynamic rules can be used for different purposes: 197 * + stateful rules; 198 * + enforcing limits on the number of sessions; 199 * + in-kernel NAT (not implemented yet) 200 * 201 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 202 * measured in seconds and depending on the flags. 203 * 204 * The total number of dynamic rules is stored in dyn_count. 205 * The max number of dynamic rules is dyn_max. When we reach 206 * the maximum number of rules we do not create anymore. This is 207 * done to avoid consuming too much memory, but also too much 208 * time when searching on each packet (ideally, we should try instead 209 * to put a limit on the length of the list on each bucket...). 210 * 211 * Each dynamic rule holds a pointer to the parent ipfw rule so 212 * we know what action to perform. Dynamic rules are removed when 213 * the parent rule is deleted. XXX we should make them survive. 214 * 215 * There are some limitations with dynamic rules -- we do not 216 * obey the 'randomized match', and we do not do multiple 217 * passes through the firewall. XXX check the latter!!! 218 */ 219static ipfw_dyn_rule **ipfw_dyn_v = NULL; 220static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 221static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 222 223static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 224#define IPFW_DYN_LOCK_INIT() \ 225 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 226#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 227#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 228#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 229#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 230 231/* 232 * Timeouts for various events in handing dynamic rules. 233 */ 234static u_int32_t dyn_ack_lifetime = 300; 235static u_int32_t dyn_syn_lifetime = 20; 236static u_int32_t dyn_fin_lifetime = 1; 237static u_int32_t dyn_rst_lifetime = 1; 238static u_int32_t dyn_udp_lifetime = 10; 239static u_int32_t dyn_short_lifetime = 5; 240 241/* 242 * Keepalives are sent if dyn_keepalive is set. They are sent every 243 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 244 * seconds of lifetime of a rule. 245 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 246 * than dyn_keepalive_period. 247 */ 248 249static u_int32_t dyn_keepalive_interval = 20; 250static u_int32_t dyn_keepalive_period = 5; 251static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 252 253static u_int32_t static_count; /* # of static rules */ 254static u_int32_t static_len; /* size in bytes of static rules */ 255static u_int32_t dyn_count; /* # of dynamic rules */ 256static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 257 258SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 259 &dyn_buckets, 0, "Number of dyn. buckets"); 260SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 261 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 262SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 263 &dyn_count, 0, "Number of dyn. rules"); 264SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 265 &dyn_max, 0, "Max number of dyn. rules"); 266SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 267 &static_count, 0, "Number of static rules"); 268SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 269 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 270SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 271 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 272SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 273 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 274SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 275 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 276SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 277 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 278SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 279 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 280SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 281 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 282 283#ifdef INET6 284/* 285 * IPv6 specific variables 286 */ 287SYSCTL_DECL(_net_inet6_ip6); 288 289static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 290static struct sysctl_oid *ip6_fw_sysctl_tree; 291#endif /* INET6 */ 292#endif /* SYSCTL_NODE */ 293 294static int fw_deny_unknown_exthdrs = 1; 295 296 297/* 298 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 299 * Other macros just cast void * into the appropriate type 300 */ 301#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 302#define TCP(p) ((struct tcphdr *)(p)) 303#define SCTP(p) ((struct sctphdr *)(p)) 304#define UDP(p) ((struct udphdr *)(p)) 305#define ICMP(p) ((struct icmphdr *)(p)) 306#define ICMP6(p) ((struct icmp6_hdr *)(p)) 307 308static __inline int 309icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 310{ 311 int type = icmp->icmp_type; 312 313 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 314} 315 316#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 317 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 318 319static int 320is_icmp_query(struct icmphdr *icmp) 321{ 322 int type = icmp->icmp_type; 323 324 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 325} 326#undef TT 327 328/* 329 * The following checks use two arrays of 8 or 16 bits to store the 330 * bits that we want set or clear, respectively. They are in the 331 * low and high half of cmd->arg1 or cmd->d[0]. 332 * 333 * We scan options and store the bits we find set. We succeed if 334 * 335 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 336 * 337 * The code is sometimes optimized not to store additional variables. 338 */ 339 340static int 341flags_match(ipfw_insn *cmd, u_int8_t bits) 342{ 343 u_char want_clear; 344 bits = ~bits; 345 346 if ( ((cmd->arg1 & 0xff) & bits) != 0) 347 return 0; /* some bits we want set were clear */ 348 want_clear = (cmd->arg1 >> 8) & 0xff; 349 if ( (want_clear & bits) != want_clear) 350 return 0; /* some bits we want clear were set */ 351 return 1; 352} 353 354static int 355ipopts_match(struct ip *ip, ipfw_insn *cmd) 356{ 357 int optlen, bits = 0; 358 u_char *cp = (u_char *)(ip + 1); 359 int x = (ip->ip_hl << 2) - sizeof (struct ip); 360 361 for (; x > 0; x -= optlen, cp += optlen) { 362 int opt = cp[IPOPT_OPTVAL]; 363 364 if (opt == IPOPT_EOL) 365 break; 366 if (opt == IPOPT_NOP) 367 optlen = 1; 368 else { 369 optlen = cp[IPOPT_OLEN]; 370 if (optlen <= 0 || optlen > x) 371 return 0; /* invalid or truncated */ 372 } 373 switch (opt) { 374 375 default: 376 break; 377 378 case IPOPT_LSRR: 379 bits |= IP_FW_IPOPT_LSRR; 380 break; 381 382 case IPOPT_SSRR: 383 bits |= IP_FW_IPOPT_SSRR; 384 break; 385 386 case IPOPT_RR: 387 bits |= IP_FW_IPOPT_RR; 388 break; 389 390 case IPOPT_TS: 391 bits |= IP_FW_IPOPT_TS; 392 break; 393 } 394 } 395 return (flags_match(cmd, bits)); 396} 397 398static int 399tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 400{ 401 int optlen, bits = 0; 402 u_char *cp = (u_char *)(tcp + 1); 403 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 404 405 for (; x > 0; x -= optlen, cp += optlen) { 406 int opt = cp[0]; 407 if (opt == TCPOPT_EOL) 408 break; 409 if (opt == TCPOPT_NOP) 410 optlen = 1; 411 else { 412 optlen = cp[1]; 413 if (optlen <= 0) 414 break; 415 } 416 417 switch (opt) { 418 419 default: 420 break; 421 422 case TCPOPT_MAXSEG: 423 bits |= IP_FW_TCPOPT_MSS; 424 break; 425 426 case TCPOPT_WINDOW: 427 bits |= IP_FW_TCPOPT_WINDOW; 428 break; 429 430 case TCPOPT_SACK_PERMITTED: 431 case TCPOPT_SACK: 432 bits |= IP_FW_TCPOPT_SACK; 433 break; 434 435 case TCPOPT_TIMESTAMP: 436 bits |= IP_FW_TCPOPT_TS; 437 break; 438 439 } 440 } 441 return (flags_match(cmd, bits)); 442} 443 444static int 445iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 446{ 447 if (ifp == NULL) /* no iface with this packet, match fails */ 448 return 0; 449 /* Check by name or by IP address */ 450 if (cmd->name[0] != '\0') { /* match by name */ 451 /* Check name */ 452 if (cmd->p.glob) { 453 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 454 return(1); 455 } else { 456 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 457 return(1); 458 } 459 } else { 460 struct ifaddr *ia; 461 462 /* XXX lock? */ 463 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 464 if (ia->ifa_addr->sa_family != AF_INET) 465 continue; 466 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 467 (ia->ifa_addr))->sin_addr.s_addr) 468 return(1); /* match */ 469 } 470 } 471 return(0); /* no match, fail ... */ 472} 473 474/* 475 * The verify_path function checks if a route to the src exists and 476 * if it is reachable via ifp (when provided). 477 * 478 * The 'verrevpath' option checks that the interface that an IP packet 479 * arrives on is the same interface that traffic destined for the 480 * packet's source address would be routed out of. The 'versrcreach' 481 * option just checks that the source address is reachable via any route 482 * (except default) in the routing table. These two are a measure to block 483 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 484 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 485 * is purposely reminiscent of the Cisco IOS command, 486 * 487 * ip verify unicast reverse-path 488 * ip verify unicast source reachable-via any 489 * 490 * which implements the same functionality. But note that syntax is 491 * misleading. The check may be performed on all IP packets whether unicast, 492 * multicast, or broadcast. 493 */ 494static int 495verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 496{ 497 struct route ro; 498 struct sockaddr_in *dst; 499 500 bzero(&ro, sizeof(ro)); 501 502 dst = (struct sockaddr_in *)&(ro.ro_dst); 503 dst->sin_family = AF_INET; 504 dst->sin_len = sizeof(*dst); 505 dst->sin_addr = src; 506 in_rtalloc_ign(&ro, RTF_CLONING, fib); 507 508 if (ro.ro_rt == NULL) 509 return 0; 510 511 /* 512 * If ifp is provided, check for equality with rtentry. 513 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 514 * in order to pass packets injected back by if_simloop(): 515 * if useloopback == 1 routing entry (via lo0) for our own address 516 * may exist, so we need to handle routing assymetry. 517 */ 518 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 519 RTFREE(ro.ro_rt); 520 return 0; 521 } 522 523 /* if no ifp provided, check if rtentry is not default route */ 524 if (ifp == NULL && 525 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 526 RTFREE(ro.ro_rt); 527 return 0; 528 } 529 530 /* or if this is a blackhole/reject route */ 531 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 532 RTFREE(ro.ro_rt); 533 return 0; 534 } 535 536 /* found valid route */ 537 RTFREE(ro.ro_rt); 538 return 1; 539} 540 541#ifdef INET6 542/* 543 * ipv6 specific rules here... 544 */ 545static __inline int 546icmp6type_match (int type, ipfw_insn_u32 *cmd) 547{ 548 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 549} 550 551static int 552flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 553{ 554 int i; 555 for (i=0; i <= cmd->o.arg1; ++i ) 556 if (curr_flow == cmd->d[i] ) 557 return 1; 558 return 0; 559} 560 561/* support for IP6_*_ME opcodes */ 562static int 563search_ip6_addr_net (struct in6_addr * ip6_addr) 564{ 565 struct ifnet *mdc; 566 struct ifaddr *mdc2; 567 struct in6_ifaddr *fdm; 568 struct in6_addr copia; 569 570 TAILQ_FOREACH(mdc, &ifnet, if_link) 571 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 572 if (mdc2->ifa_addr->sa_family == AF_INET6) { 573 fdm = (struct in6_ifaddr *)mdc2; 574 copia = fdm->ia_addr.sin6_addr; 575 /* need for leaving scope_id in the sock_addr */ 576 in6_clearscope(&copia); 577 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 578 return 1; 579 } 580 } 581 return 0; 582} 583 584static int 585verify_path6(struct in6_addr *src, struct ifnet *ifp) 586{ 587 struct route_in6 ro; 588 struct sockaddr_in6 *dst; 589 590 bzero(&ro, sizeof(ro)); 591 592 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 593 dst->sin6_family = AF_INET6; 594 dst->sin6_len = sizeof(*dst); 595 dst->sin6_addr = *src; 596 /* XXX MRT 0 for ipv6 at this time */ 597 rtalloc_ign((struct route *)&ro, RTF_CLONING); 598 599 if (ro.ro_rt == NULL) 600 return 0; 601 602 /* 603 * if ifp is provided, check for equality with rtentry 604 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 605 * to support the case of sending packets to an address of our own. 606 * (where the former interface is the first argument of if_simloop() 607 * (=ifp), the latter is lo0) 608 */ 609 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 610 RTFREE(ro.ro_rt); 611 return 0; 612 } 613 614 /* if no ifp provided, check if rtentry is not default route */ 615 if (ifp == NULL && 616 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 617 RTFREE(ro.ro_rt); 618 return 0; 619 } 620 621 /* or if this is a blackhole/reject route */ 622 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 623 RTFREE(ro.ro_rt); 624 return 0; 625 } 626 627 /* found valid route */ 628 RTFREE(ro.ro_rt); 629 return 1; 630 631} 632static __inline int 633hash_packet6(struct ipfw_flow_id *id) 634{ 635 u_int32_t i; 636 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 637 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 638 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 639 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 640 (id->dst_port) ^ (id->src_port); 641 return i; 642} 643 644static int 645is_icmp6_query(int icmp6_type) 646{ 647 if ((icmp6_type <= ICMP6_MAXTYPE) && 648 (icmp6_type == ICMP6_ECHO_REQUEST || 649 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 650 icmp6_type == ICMP6_WRUREQUEST || 651 icmp6_type == ICMP6_FQDN_QUERY || 652 icmp6_type == ICMP6_NI_QUERY)) 653 return (1); 654 655 return (0); 656} 657 658static void 659send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 660{ 661 struct mbuf *m; 662 663 m = args->m; 664 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 665 struct tcphdr *tcp; 666 tcp_seq ack, seq; 667 int flags; 668 struct { 669 struct ip6_hdr ip6; 670 struct tcphdr th; 671 } ti; 672 tcp = (struct tcphdr *)((char *)ip6 + hlen); 673 674 if ((tcp->th_flags & TH_RST) != 0) { 675 m_freem(m); 676 args->m = NULL; 677 return; 678 } 679 680 ti.ip6 = *ip6; 681 ti.th = *tcp; 682 ti.th.th_seq = ntohl(ti.th.th_seq); 683 ti.th.th_ack = ntohl(ti.th.th_ack); 684 ti.ip6.ip6_nxt = IPPROTO_TCP; 685 686 if (ti.th.th_flags & TH_ACK) { 687 ack = 0; 688 seq = ti.th.th_ack; 689 flags = TH_RST; 690 } else { 691 ack = ti.th.th_seq; 692 if ((m->m_flags & M_PKTHDR) != 0) { 693 /* 694 * total new data to ACK is: 695 * total packet length, 696 * minus the header length, 697 * minus the tcp header length. 698 */ 699 ack += m->m_pkthdr.len - hlen 700 - (ti.th.th_off << 2); 701 } else if (ip6->ip6_plen) { 702 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 703 hlen - (ti.th.th_off << 2); 704 } else { 705 m_freem(m); 706 return; 707 } 708 if (tcp->th_flags & TH_SYN) 709 ack++; 710 seq = 0; 711 flags = TH_RST|TH_ACK; 712 } 713 bcopy(&ti, ip6, sizeof(ti)); 714 /* 715 * m is only used to recycle the mbuf 716 * The data in it is never read so we don't need 717 * to correct the offsets or anything 718 */ 719 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 720 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 721#if 0 722 /* 723 * Unlike above, the mbufs need to line up with the ip6 hdr, 724 * as the contents are read. We need to m_adj() the 725 * needed amount. 726 * The mbuf will however be thrown away so we can adjust it. 727 * Remember we did an m_pullup on it already so we 728 * can make some assumptions about contiguousness. 729 */ 730 if (args->L3offset) 731 m_adj(m, args->L3offset); 732#endif 733 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 734 } else 735 m_freem(m); 736 737 args->m = NULL; 738} 739 740#endif /* INET6 */ 741 742static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 743 744#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 745#define SNP(buf) buf, sizeof(buf) 746 747/* 748 * We enter here when we have a rule with O_LOG. 749 * XXX this function alone takes about 2Kbytes of code! 750 */ 751static void 752ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 753 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 754 struct ip *ip) 755{ 756 struct ether_header *eh = args->eh; 757 char *action; 758 int limit_reached = 0; 759 char action2[40], proto[128], fragment[32]; 760 761 fragment[0] = '\0'; 762 proto[0] = '\0'; 763 764 if (f == NULL) { /* bogus pkt */ 765 if (verbose_limit != 0 && norule_counter >= verbose_limit) 766 return; 767 norule_counter++; 768 if (norule_counter == verbose_limit) 769 limit_reached = verbose_limit; 770 action = "Refuse"; 771 } else { /* O_LOG is the first action, find the real one */ 772 ipfw_insn *cmd = ACTION_PTR(f); 773 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 774 775 if (l->max_log != 0 && l->log_left == 0) 776 return; 777 l->log_left--; 778 if (l->log_left == 0) 779 limit_reached = l->max_log; 780 cmd += F_LEN(cmd); /* point to first action */ 781 if (cmd->opcode == O_ALTQ) { 782 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 783 784 snprintf(SNPARGS(action2, 0), "Altq %d", 785 altq->qid); 786 cmd += F_LEN(cmd); 787 } 788 if (cmd->opcode == O_PROB) 789 cmd += F_LEN(cmd); 790 791 if (cmd->opcode == O_TAG) 792 cmd += F_LEN(cmd); 793 794 action = action2; 795 switch (cmd->opcode) { 796 case O_DENY: 797 action = "Deny"; 798 break; 799 800 case O_REJECT: 801 if (cmd->arg1==ICMP_REJECT_RST) 802 action = "Reset"; 803 else if (cmd->arg1==ICMP_UNREACH_HOST) 804 action = "Reject"; 805 else 806 snprintf(SNPARGS(action2, 0), "Unreach %d", 807 cmd->arg1); 808 break; 809 810 case O_UNREACH6: 811 if (cmd->arg1==ICMP6_UNREACH_RST) 812 action = "Reset"; 813 else 814 snprintf(SNPARGS(action2, 0), "Unreach %d", 815 cmd->arg1); 816 break; 817 818 case O_ACCEPT: 819 action = "Accept"; 820 break; 821 case O_COUNT: 822 action = "Count"; 823 break; 824 case O_DIVERT: 825 snprintf(SNPARGS(action2, 0), "Divert %d", 826 cmd->arg1); 827 break; 828 case O_TEE: 829 snprintf(SNPARGS(action2, 0), "Tee %d", 830 cmd->arg1); 831 break; 832 case O_SETFIB: 833 snprintf(SNPARGS(action2, 0), "SetFib %d", 834 cmd->arg1); 835 break; 836 case O_SKIPTO: 837 snprintf(SNPARGS(action2, 0), "SkipTo %d", 838 cmd->arg1); 839 break; 840 case O_PIPE: 841 snprintf(SNPARGS(action2, 0), "Pipe %d", 842 cmd->arg1); 843 break; 844 case O_QUEUE: 845 snprintf(SNPARGS(action2, 0), "Queue %d", 846 cmd->arg1); 847 break; 848 case O_FORWARD_IP: { 849 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 850 int len; 851 struct in_addr dummyaddr; 852 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 853 dummyaddr.s_addr = htonl(tablearg); 854 else 855 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 856 857 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 858 inet_ntoa(dummyaddr)); 859 860 if (sa->sa.sin_port) 861 snprintf(SNPARGS(action2, len), ":%d", 862 sa->sa.sin_port); 863 } 864 break; 865 case O_NETGRAPH: 866 snprintf(SNPARGS(action2, 0), "Netgraph %d", 867 cmd->arg1); 868 break; 869 case O_NGTEE: 870 snprintf(SNPARGS(action2, 0), "Ngtee %d", 871 cmd->arg1); 872 break; 873 case O_NAT: 874 action = "Nat"; 875 break; 876 default: 877 action = "UNKNOWN"; 878 break; 879 } 880 } 881 882 if (hlen == 0) { /* non-ip */ 883 snprintf(SNPARGS(proto, 0), "MAC"); 884 885 } else { 886 int len; 887 char src[48], dst[48]; 888 struct icmphdr *icmp; 889 struct tcphdr *tcp; 890 struct udphdr *udp; 891#ifdef INET6 892 struct ip6_hdr *ip6 = NULL; 893 struct icmp6_hdr *icmp6; 894#endif 895 src[0] = '\0'; 896 dst[0] = '\0'; 897#ifdef INET6 898 if (IS_IP6_FLOW_ID(&(args->f_id))) { 899 char ip6buf[INET6_ADDRSTRLEN]; 900 snprintf(src, sizeof(src), "[%s]", 901 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 902 snprintf(dst, sizeof(dst), "[%s]", 903 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 904 905 ip6 = (struct ip6_hdr *)ip; 906 tcp = (struct tcphdr *)(((char *)ip) + hlen); 907 udp = (struct udphdr *)(((char *)ip) + hlen); 908 } else 909#endif 910 { 911 tcp = L3HDR(struct tcphdr, ip); 912 udp = L3HDR(struct udphdr, ip); 913 914 inet_ntoa_r(ip->ip_src, src); 915 inet_ntoa_r(ip->ip_dst, dst); 916 } 917 918 switch (args->f_id.proto) { 919 case IPPROTO_TCP: 920 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 921 if (offset == 0) 922 snprintf(SNPARGS(proto, len), ":%d %s:%d", 923 ntohs(tcp->th_sport), 924 dst, 925 ntohs(tcp->th_dport)); 926 else 927 snprintf(SNPARGS(proto, len), " %s", dst); 928 break; 929 930 case IPPROTO_UDP: 931 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 932 if (offset == 0) 933 snprintf(SNPARGS(proto, len), ":%d %s:%d", 934 ntohs(udp->uh_sport), 935 dst, 936 ntohs(udp->uh_dport)); 937 else 938 snprintf(SNPARGS(proto, len), " %s", dst); 939 break; 940 941 case IPPROTO_ICMP: 942 icmp = L3HDR(struct icmphdr, ip); 943 if (offset == 0) 944 len = snprintf(SNPARGS(proto, 0), 945 "ICMP:%u.%u ", 946 icmp->icmp_type, icmp->icmp_code); 947 else 948 len = snprintf(SNPARGS(proto, 0), "ICMP "); 949 len += snprintf(SNPARGS(proto, len), "%s", src); 950 snprintf(SNPARGS(proto, len), " %s", dst); 951 break; 952#ifdef INET6 953 case IPPROTO_ICMPV6: 954 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 955 if (offset == 0) 956 len = snprintf(SNPARGS(proto, 0), 957 "ICMPv6:%u.%u ", 958 icmp6->icmp6_type, icmp6->icmp6_code); 959 else 960 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 961 len += snprintf(SNPARGS(proto, len), "%s", src); 962 snprintf(SNPARGS(proto, len), " %s", dst); 963 break; 964#endif 965 default: 966 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 967 args->f_id.proto, src); 968 snprintf(SNPARGS(proto, len), " %s", dst); 969 break; 970 } 971 972#ifdef INET6 973 if (IS_IP6_FLOW_ID(&(args->f_id))) { 974 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 975 snprintf(SNPARGS(fragment, 0), 976 " (frag %08x:%d@%d%s)", 977 args->f_id.frag_id6, 978 ntohs(ip6->ip6_plen) - hlen, 979 ntohs(offset & IP6F_OFF_MASK) << 3, 980 (offset & IP6F_MORE_FRAG) ? "+" : ""); 981 } else 982#endif 983 { 984 int ip_off, ip_len; 985 if (eh != NULL) { /* layer 2 packets are as on the wire */ 986 ip_off = ntohs(ip->ip_off); 987 ip_len = ntohs(ip->ip_len); 988 } else { 989 ip_off = ip->ip_off; 990 ip_len = ip->ip_len; 991 } 992 if (ip_off & (IP_MF | IP_OFFMASK)) 993 snprintf(SNPARGS(fragment, 0), 994 " (frag %d:%d@%d%s)", 995 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 996 offset << 3, 997 (ip_off & IP_MF) ? "+" : ""); 998 } 999 } 1000 if (oif || m->m_pkthdr.rcvif) 1001 log(LOG_SECURITY | LOG_INFO, 1002 "ipfw: %d %s %s %s via %s%s\n", 1003 f ? f->rulenum : -1, 1004 action, proto, oif ? "out" : "in", 1005 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1006 fragment); 1007 else 1008 log(LOG_SECURITY | LOG_INFO, 1009 "ipfw: %d %s %s [no if info]%s\n", 1010 f ? f->rulenum : -1, 1011 action, proto, fragment); 1012 if (limit_reached) 1013 log(LOG_SECURITY | LOG_NOTICE, 1014 "ipfw: limit %d reached on entry %d\n", 1015 limit_reached, f ? f->rulenum : -1); 1016} 1017 1018/* 1019 * IMPORTANT: the hash function for dynamic rules must be commutative 1020 * in source and destination (ip,port), because rules are bidirectional 1021 * and we want to find both in the same bucket. 1022 */ 1023static __inline int 1024hash_packet(struct ipfw_flow_id *id) 1025{ 1026 u_int32_t i; 1027 1028#ifdef INET6 1029 if (IS_IP6_FLOW_ID(id)) 1030 i = hash_packet6(id); 1031 else 1032#endif /* INET6 */ 1033 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1034 i &= (curr_dyn_buckets - 1); 1035 return i; 1036} 1037 1038/** 1039 * unlink a dynamic rule from a chain. prev is a pointer to 1040 * the previous one, q is a pointer to the rule to delete, 1041 * head is a pointer to the head of the queue. 1042 * Modifies q and potentially also head. 1043 */ 1044#define UNLINK_DYN_RULE(prev, head, q) { \ 1045 ipfw_dyn_rule *old_q = q; \ 1046 \ 1047 /* remove a refcount to the parent */ \ 1048 if (q->dyn_type == O_LIMIT) \ 1049 q->parent->count--; \ 1050 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1051 (q->id.src_ip), (q->id.src_port), \ 1052 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1053 if (prev != NULL) \ 1054 prev->next = q = q->next; \ 1055 else \ 1056 head = q = q->next; \ 1057 dyn_count--; \ 1058 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1059 1060#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1061 1062/** 1063 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1064 * 1065 * If keep_me == NULL, rules are deleted even if not expired, 1066 * otherwise only expired rules are removed. 1067 * 1068 * The value of the second parameter is also used to point to identify 1069 * a rule we absolutely do not want to remove (e.g. because we are 1070 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1071 * rules). The pointer is only used for comparison, so any non-null 1072 * value will do. 1073 */ 1074static void 1075remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1076{ 1077 static u_int32_t last_remove = 0; 1078 1079#define FORCE (keep_me == NULL) 1080 1081 ipfw_dyn_rule *prev, *q; 1082 int i, pass = 0, max_pass = 0; 1083 1084 IPFW_DYN_LOCK_ASSERT(); 1085 1086 if (ipfw_dyn_v == NULL || dyn_count == 0) 1087 return; 1088 /* do not expire more than once per second, it is useless */ 1089 if (!FORCE && last_remove == time_uptime) 1090 return; 1091 last_remove = time_uptime; 1092 1093 /* 1094 * because O_LIMIT refer to parent rules, during the first pass only 1095 * remove child and mark any pending LIMIT_PARENT, and remove 1096 * them in a second pass. 1097 */ 1098next_pass: 1099 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1100 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1101 /* 1102 * Logic can become complex here, so we split tests. 1103 */ 1104 if (q == keep_me) 1105 goto next; 1106 if (rule != NULL && rule != q->rule) 1107 goto next; /* not the one we are looking for */ 1108 if (q->dyn_type == O_LIMIT_PARENT) { 1109 /* 1110 * handle parent in the second pass, 1111 * record we need one. 1112 */ 1113 max_pass = 1; 1114 if (pass == 0) 1115 goto next; 1116 if (FORCE && q->count != 0 ) { 1117 /* XXX should not happen! */ 1118 printf("ipfw: OUCH! cannot remove rule," 1119 " count %d\n", q->count); 1120 } 1121 } else { 1122 if (!FORCE && 1123 !TIME_LEQ( q->expire, time_uptime )) 1124 goto next; 1125 } 1126 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1127 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1128 continue; 1129 } 1130next: 1131 prev=q; 1132 q=q->next; 1133 } 1134 } 1135 if (pass++ < max_pass) 1136 goto next_pass; 1137} 1138 1139 1140/** 1141 * lookup a dynamic rule. 1142 */ 1143static ipfw_dyn_rule * 1144lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1145 struct tcphdr *tcp) 1146{ 1147 /* 1148 * stateful ipfw extensions. 1149 * Lookup into dynamic session queue 1150 */ 1151#define MATCH_REVERSE 0 1152#define MATCH_FORWARD 1 1153#define MATCH_NONE 2 1154#define MATCH_UNKNOWN 3 1155 int i, dir = MATCH_NONE; 1156 ipfw_dyn_rule *prev, *q=NULL; 1157 1158 IPFW_DYN_LOCK_ASSERT(); 1159 1160 if (ipfw_dyn_v == NULL) 1161 goto done; /* not found */ 1162 i = hash_packet( pkt ); 1163 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1164 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1165 goto next; 1166 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1167 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1168 continue; 1169 } 1170 if (pkt->proto == q->id.proto && 1171 q->dyn_type != O_LIMIT_PARENT) { 1172 if (IS_IP6_FLOW_ID(pkt)) { 1173 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1174 &(q->id.src_ip6)) && 1175 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1176 &(q->id.dst_ip6)) && 1177 pkt->src_port == q->id.src_port && 1178 pkt->dst_port == q->id.dst_port ) { 1179 dir = MATCH_FORWARD; 1180 break; 1181 } 1182 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1183 &(q->id.dst_ip6)) && 1184 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1185 &(q->id.src_ip6)) && 1186 pkt->src_port == q->id.dst_port && 1187 pkt->dst_port == q->id.src_port ) { 1188 dir = MATCH_REVERSE; 1189 break; 1190 } 1191 } else { 1192 if (pkt->src_ip == q->id.src_ip && 1193 pkt->dst_ip == q->id.dst_ip && 1194 pkt->src_port == q->id.src_port && 1195 pkt->dst_port == q->id.dst_port ) { 1196 dir = MATCH_FORWARD; 1197 break; 1198 } 1199 if (pkt->src_ip == q->id.dst_ip && 1200 pkt->dst_ip == q->id.src_ip && 1201 pkt->src_port == q->id.dst_port && 1202 pkt->dst_port == q->id.src_port ) { 1203 dir = MATCH_REVERSE; 1204 break; 1205 } 1206 } 1207 } 1208next: 1209 prev = q; 1210 q = q->next; 1211 } 1212 if (q == NULL) 1213 goto done; /* q = NULL, not found */ 1214 1215 if ( prev != NULL) { /* found and not in front */ 1216 prev->next = q->next; 1217 q->next = ipfw_dyn_v[i]; 1218 ipfw_dyn_v[i] = q; 1219 } 1220 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1221 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1222 1223#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1224#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1225 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1226 switch (q->state) { 1227 case TH_SYN: /* opening */ 1228 q->expire = time_uptime + dyn_syn_lifetime; 1229 break; 1230 1231 case BOTH_SYN: /* move to established */ 1232 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1233 case BOTH_SYN | (TH_FIN << 8) : 1234 if (tcp) { 1235#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1236 u_int32_t ack = ntohl(tcp->th_ack); 1237 if (dir == MATCH_FORWARD) { 1238 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1239 q->ack_fwd = ack; 1240 else { /* ignore out-of-sequence */ 1241 break; 1242 } 1243 } else { 1244 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1245 q->ack_rev = ack; 1246 else { /* ignore out-of-sequence */ 1247 break; 1248 } 1249 } 1250 } 1251 q->expire = time_uptime + dyn_ack_lifetime; 1252 break; 1253 1254 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1255 if (dyn_fin_lifetime >= dyn_keepalive_period) 1256 dyn_fin_lifetime = dyn_keepalive_period - 1; 1257 q->expire = time_uptime + dyn_fin_lifetime; 1258 break; 1259 1260 default: 1261#if 0 1262 /* 1263 * reset or some invalid combination, but can also 1264 * occur if we use keep-state the wrong way. 1265 */ 1266 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1267 printf("invalid state: 0x%x\n", q->state); 1268#endif 1269 if (dyn_rst_lifetime >= dyn_keepalive_period) 1270 dyn_rst_lifetime = dyn_keepalive_period - 1; 1271 q->expire = time_uptime + dyn_rst_lifetime; 1272 break; 1273 } 1274 } else if (pkt->proto == IPPROTO_UDP) { 1275 q->expire = time_uptime + dyn_udp_lifetime; 1276 } else { 1277 /* other protocols */ 1278 q->expire = time_uptime + dyn_short_lifetime; 1279 } 1280done: 1281 if (match_direction) 1282 *match_direction = dir; 1283 return q; 1284} 1285 1286static ipfw_dyn_rule * 1287lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1288 struct tcphdr *tcp) 1289{ 1290 ipfw_dyn_rule *q; 1291 1292 IPFW_DYN_LOCK(); 1293 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1294 if (q == NULL) 1295 IPFW_DYN_UNLOCK(); 1296 /* NB: return table locked when q is not NULL */ 1297 return q; 1298} 1299 1300static void 1301realloc_dynamic_table(void) 1302{ 1303 IPFW_DYN_LOCK_ASSERT(); 1304 1305 /* 1306 * Try reallocation, make sure we have a power of 2 and do 1307 * not allow more than 64k entries. In case of overflow, 1308 * default to 1024. 1309 */ 1310 1311 if (dyn_buckets > 65536) 1312 dyn_buckets = 1024; 1313 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1314 dyn_buckets = curr_dyn_buckets; /* reset */ 1315 return; 1316 } 1317 curr_dyn_buckets = dyn_buckets; 1318 if (ipfw_dyn_v != NULL) 1319 free(ipfw_dyn_v, M_IPFW); 1320 for (;;) { 1321 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1322 M_IPFW, M_NOWAIT | M_ZERO); 1323 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1324 break; 1325 curr_dyn_buckets /= 2; 1326 } 1327} 1328 1329/** 1330 * Install state of type 'type' for a dynamic session. 1331 * The hash table contains two type of rules: 1332 * - regular rules (O_KEEP_STATE) 1333 * - rules for sessions with limited number of sess per user 1334 * (O_LIMIT). When they are created, the parent is 1335 * increased by 1, and decreased on delete. In this case, 1336 * the third parameter is the parent rule and not the chain. 1337 * - "parent" rules for the above (O_LIMIT_PARENT). 1338 */ 1339static ipfw_dyn_rule * 1340add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1341{ 1342 ipfw_dyn_rule *r; 1343 int i; 1344 1345 IPFW_DYN_LOCK_ASSERT(); 1346 1347 if (ipfw_dyn_v == NULL || 1348 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1349 realloc_dynamic_table(); 1350 if (ipfw_dyn_v == NULL) 1351 return NULL; /* failed ! */ 1352 } 1353 i = hash_packet(id); 1354 1355 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1356 if (r == NULL) { 1357 printf ("ipfw: sorry cannot allocate state\n"); 1358 return NULL; 1359 } 1360 1361 /* increase refcount on parent, and set pointer */ 1362 if (dyn_type == O_LIMIT) { 1363 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1364 if ( parent->dyn_type != O_LIMIT_PARENT) 1365 panic("invalid parent"); 1366 parent->count++; 1367 r->parent = parent; 1368 rule = parent->rule; 1369 } 1370 1371 r->id = *id; 1372 r->expire = time_uptime + dyn_syn_lifetime; 1373 r->rule = rule; 1374 r->dyn_type = dyn_type; 1375 r->pcnt = r->bcnt = 0; 1376 r->count = 0; 1377 1378 r->bucket = i; 1379 r->next = ipfw_dyn_v[i]; 1380 ipfw_dyn_v[i] = r; 1381 dyn_count++; 1382 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1383 dyn_type, 1384 (r->id.src_ip), (r->id.src_port), 1385 (r->id.dst_ip), (r->id.dst_port), 1386 dyn_count ); ) 1387 return r; 1388} 1389 1390/** 1391 * lookup dynamic parent rule using pkt and rule as search keys. 1392 * If the lookup fails, then install one. 1393 */ 1394static ipfw_dyn_rule * 1395lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1396{ 1397 ipfw_dyn_rule *q; 1398 int i; 1399 1400 IPFW_DYN_LOCK_ASSERT(); 1401 1402 if (ipfw_dyn_v) { 1403 int is_v6 = IS_IP6_FLOW_ID(pkt); 1404 i = hash_packet( pkt ); 1405 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1406 if (q->dyn_type == O_LIMIT_PARENT && 1407 rule== q->rule && 1408 pkt->proto == q->id.proto && 1409 pkt->src_port == q->id.src_port && 1410 pkt->dst_port == q->id.dst_port && 1411 ( 1412 (is_v6 && 1413 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1414 &(q->id.src_ip6)) && 1415 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1416 &(q->id.dst_ip6))) || 1417 (!is_v6 && 1418 pkt->src_ip == q->id.src_ip && 1419 pkt->dst_ip == q->id.dst_ip) 1420 ) 1421 ) { 1422 q->expire = time_uptime + dyn_short_lifetime; 1423 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1424 return q; 1425 } 1426 } 1427 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1428} 1429 1430/** 1431 * Install dynamic state for rule type cmd->o.opcode 1432 * 1433 * Returns 1 (failure) if state is not installed because of errors or because 1434 * session limitations are enforced. 1435 */ 1436static int 1437install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1438 struct ip_fw_args *args, uint32_t tablearg) 1439{ 1440 static int last_log; 1441 ipfw_dyn_rule *q; 1442 struct in_addr da; 1443 char src[48], dst[48]; 1444 1445 src[0] = '\0'; 1446 dst[0] = '\0'; 1447 1448 DEB( 1449 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1450 __func__, cmd->o.opcode, 1451 (args->f_id.src_ip), (args->f_id.src_port), 1452 (args->f_id.dst_ip), (args->f_id.dst_port)); 1453 ) 1454 1455 IPFW_DYN_LOCK(); 1456 1457 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1458 1459 if (q != NULL) { /* should never occur */ 1460 if (last_log != time_uptime) { 1461 last_log = time_uptime; 1462 printf("ipfw: %s: entry already present, done\n", 1463 __func__); 1464 } 1465 IPFW_DYN_UNLOCK(); 1466 return (0); 1467 } 1468 1469 if (dyn_count >= dyn_max) 1470 /* Run out of slots, try to remove any expired rule. */ 1471 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1472 1473 if (dyn_count >= dyn_max) { 1474 if (last_log != time_uptime) { 1475 last_log = time_uptime; 1476 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1477 } 1478 IPFW_DYN_UNLOCK(); 1479 return (1); /* cannot install, notify caller */ 1480 } 1481 1482 switch (cmd->o.opcode) { 1483 case O_KEEP_STATE: /* bidir rule */ 1484 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1485 break; 1486 1487 case O_LIMIT: { /* limit number of sessions */ 1488 struct ipfw_flow_id id; 1489 ipfw_dyn_rule *parent; 1490 uint32_t conn_limit; 1491 uint16_t limit_mask = cmd->limit_mask; 1492 1493 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1494 tablearg : cmd->conn_limit; 1495 1496 DEB( 1497 if (cmd->conn_limit == IP_FW_TABLEARG) 1498 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1499 "(tablearg)\n", __func__, conn_limit); 1500 else 1501 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1502 __func__, conn_limit); 1503 ) 1504 1505 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1506 id.proto = args->f_id.proto; 1507 id.addr_type = args->f_id.addr_type; 1508 id.fib = M_GETFIB(args->m); 1509 1510 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1511 if (limit_mask & DYN_SRC_ADDR) 1512 id.src_ip6 = args->f_id.src_ip6; 1513 if (limit_mask & DYN_DST_ADDR) 1514 id.dst_ip6 = args->f_id.dst_ip6; 1515 } else { 1516 if (limit_mask & DYN_SRC_ADDR) 1517 id.src_ip = args->f_id.src_ip; 1518 if (limit_mask & DYN_DST_ADDR) 1519 id.dst_ip = args->f_id.dst_ip; 1520 } 1521 if (limit_mask & DYN_SRC_PORT) 1522 id.src_port = args->f_id.src_port; 1523 if (limit_mask & DYN_DST_PORT) 1524 id.dst_port = args->f_id.dst_port; 1525 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1526 printf("ipfw: %s: add parent failed\n", __func__); 1527 IPFW_DYN_UNLOCK(); 1528 return (1); 1529 } 1530 1531 if (parent->count >= conn_limit) { 1532 /* See if we can remove some expired rule. */ 1533 remove_dyn_rule(rule, parent); 1534 if (parent->count >= conn_limit) { 1535 if (fw_verbose && last_log != time_uptime) { 1536 last_log = time_uptime; 1537#ifdef INET6 1538 /* 1539 * XXX IPv6 flows are not 1540 * supported yet. 1541 */ 1542 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1543 char ip6buf[INET6_ADDRSTRLEN]; 1544 snprintf(src, sizeof(src), 1545 "[%s]", ip6_sprintf(ip6buf, 1546 &args->f_id.src_ip6)); 1547 snprintf(dst, sizeof(dst), 1548 "[%s]", ip6_sprintf(ip6buf, 1549 &args->f_id.dst_ip6)); 1550 } else 1551#endif 1552 { 1553 da.s_addr = 1554 htonl(args->f_id.src_ip); 1555 inet_ntoa_r(da, src); 1556 da.s_addr = 1557 htonl(args->f_id.dst_ip); 1558 inet_ntoa_r(da, dst); 1559 } 1560 log(LOG_SECURITY | LOG_DEBUG, 1561 "ipfw: %d %s %s:%u -> %s:%u, %s\n", 1562 parent->rule->rulenum, 1563 "drop session", 1564 src, (args->f_id.src_port), 1565 dst, (args->f_id.dst_port), 1566 "too many entries"); 1567 } 1568 IPFW_DYN_UNLOCK(); 1569 return (1); 1570 } 1571 } 1572 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1573 break; 1574 } 1575 default: 1576 printf("ipfw: %s: unknown dynamic rule type %u\n", 1577 __func__, cmd->o.opcode); 1578 IPFW_DYN_UNLOCK(); 1579 return (1); 1580 } 1581 1582 /* XXX just set lifetime */ 1583 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1584 1585 IPFW_DYN_UNLOCK(); 1586 return (0); 1587} 1588 1589/* 1590 * Generate a TCP packet, containing either a RST or a keepalive. 1591 * When flags & TH_RST, we are sending a RST packet, because of a 1592 * "reset" action matched the packet. 1593 * Otherwise we are sending a keepalive, and flags & TH_ 1594 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1595 * so that MAC can label the reply appropriately. 1596 */ 1597static struct mbuf * 1598send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1599 u_int32_t ack, int flags) 1600{ 1601 struct mbuf *m; 1602 struct ip *ip; 1603 struct tcphdr *tcp; 1604 1605 MGETHDR(m, M_DONTWAIT, MT_DATA); 1606 if (m == 0) 1607 return (NULL); 1608 m->m_pkthdr.rcvif = (struct ifnet *)0; 1609 1610 M_SETFIB(m, id->fib); 1611#ifdef MAC 1612 if (replyto != NULL) 1613 mac_netinet_firewall_reply(replyto, m); 1614 else 1615 mac_netinet_firewall_send(m); 1616#else 1617 (void)replyto; /* don't warn about unused arg */ 1618#endif 1619 1620 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1621 m->m_data += max_linkhdr; 1622 1623 ip = mtod(m, struct ip *); 1624 bzero(ip, m->m_len); 1625 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1626 ip->ip_p = IPPROTO_TCP; 1627 tcp->th_off = 5; 1628 /* 1629 * Assume we are sending a RST (or a keepalive in the reverse 1630 * direction), swap src and destination addresses and ports. 1631 */ 1632 ip->ip_src.s_addr = htonl(id->dst_ip); 1633 ip->ip_dst.s_addr = htonl(id->src_ip); 1634 tcp->th_sport = htons(id->dst_port); 1635 tcp->th_dport = htons(id->src_port); 1636 if (flags & TH_RST) { /* we are sending a RST */ 1637 if (flags & TH_ACK) { 1638 tcp->th_seq = htonl(ack); 1639 tcp->th_ack = htonl(0); 1640 tcp->th_flags = TH_RST; 1641 } else { 1642 if (flags & TH_SYN) 1643 seq++; 1644 tcp->th_seq = htonl(0); 1645 tcp->th_ack = htonl(seq); 1646 tcp->th_flags = TH_RST | TH_ACK; 1647 } 1648 } else { 1649 /* 1650 * We are sending a keepalive. flags & TH_SYN determines 1651 * the direction, forward if set, reverse if clear. 1652 * NOTE: seq and ack are always assumed to be correct 1653 * as set by the caller. This may be confusing... 1654 */ 1655 if (flags & TH_SYN) { 1656 /* 1657 * we have to rewrite the correct addresses! 1658 */ 1659 ip->ip_dst.s_addr = htonl(id->dst_ip); 1660 ip->ip_src.s_addr = htonl(id->src_ip); 1661 tcp->th_dport = htons(id->dst_port); 1662 tcp->th_sport = htons(id->src_port); 1663 } 1664 tcp->th_seq = htonl(seq); 1665 tcp->th_ack = htonl(ack); 1666 tcp->th_flags = TH_ACK; 1667 } 1668 /* 1669 * set ip_len to the payload size so we can compute 1670 * the tcp checksum on the pseudoheader 1671 * XXX check this, could save a couple of words ? 1672 */ 1673 ip->ip_len = htons(sizeof(struct tcphdr)); 1674 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1675 /* 1676 * now fill fields left out earlier 1677 */ 1678 ip->ip_ttl = ip_defttl; 1679 ip->ip_len = m->m_pkthdr.len; 1680 m->m_flags |= M_SKIP_FIREWALL; 1681 return (m); 1682} 1683 1684/* 1685 * sends a reject message, consuming the mbuf passed as an argument. 1686 */ 1687static void 1688send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1689{ 1690 1691#if 0 1692 /* XXX When ip is not guaranteed to be at mtod() we will 1693 * need to account for this */ 1694 * The mbuf will however be thrown away so we can adjust it. 1695 * Remember we did an m_pullup on it already so we 1696 * can make some assumptions about contiguousness. 1697 */ 1698 if (args->L3offset) 1699 m_adj(m, args->L3offset); 1700#endif 1701 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1702 /* We need the IP header in host order for icmp_error(). */ 1703 if (args->eh != NULL) { 1704 ip->ip_len = ntohs(ip->ip_len); 1705 ip->ip_off = ntohs(ip->ip_off); 1706 } 1707 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1708 } else if (args->f_id.proto == IPPROTO_TCP) { 1709 struct tcphdr *const tcp = 1710 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1711 if ( (tcp->th_flags & TH_RST) == 0) { 1712 struct mbuf *m; 1713 m = send_pkt(args->m, &(args->f_id), 1714 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1715 tcp->th_flags | TH_RST); 1716 if (m != NULL) 1717 ip_output(m, NULL, NULL, 0, NULL, NULL); 1718 } 1719 m_freem(args->m); 1720 } else 1721 m_freem(args->m); 1722 args->m = NULL; 1723} 1724 1725/** 1726 * 1727 * Given an ip_fw *, lookup_next_rule will return a pointer 1728 * to the next rule, which can be either the jump 1729 * target (for skipto instructions) or the next one in the list (in 1730 * all other cases including a missing jump target). 1731 * The result is also written in the "next_rule" field of the rule. 1732 * Backward jumps are not allowed, so start looking from the next 1733 * rule... 1734 * 1735 * This never returns NULL -- in case we do not have an exact match, 1736 * the next rule is returned. When the ruleset is changed, 1737 * pointers are flushed so we are always correct. 1738 */ 1739 1740static struct ip_fw * 1741lookup_next_rule(struct ip_fw *me) 1742{ 1743 struct ip_fw *rule = NULL; 1744 ipfw_insn *cmd; 1745 1746 /* look for action, in case it is a skipto */ 1747 cmd = ACTION_PTR(me); 1748 if (cmd->opcode == O_LOG) 1749 cmd += F_LEN(cmd); 1750 if (cmd->opcode == O_ALTQ) 1751 cmd += F_LEN(cmd); 1752 if (cmd->opcode == O_TAG) 1753 cmd += F_LEN(cmd); 1754 if ( cmd->opcode == O_SKIPTO ) 1755 for (rule = me->next; rule ; rule = rule->next) 1756 if (rule->rulenum >= cmd->arg1) 1757 break; 1758 if (rule == NULL) /* failure or not a skipto */ 1759 rule = me->next; 1760 me->next_rule = rule; 1761 return rule; 1762} 1763 1764static int 1765add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1766 uint8_t mlen, uint32_t value) 1767{ 1768 struct radix_node_head *rnh; 1769 struct table_entry *ent; 1770 1771 if (tbl >= IPFW_TABLES_MAX) 1772 return (EINVAL); 1773 rnh = ch->tables[tbl]; 1774 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1775 if (ent == NULL) 1776 return (ENOMEM); 1777 ent->value = value; 1778 ent->addr.sin_len = ent->mask.sin_len = 8; 1779 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1780 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1781 IPFW_WLOCK(&layer3_chain); 1782 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1783 NULL) { 1784 IPFW_WUNLOCK(&layer3_chain); 1785 free(ent, M_IPFW_TBL); 1786 return (EEXIST); 1787 } 1788 IPFW_WUNLOCK(&layer3_chain); 1789 return (0); 1790} 1791 1792static int 1793del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1794 uint8_t mlen) 1795{ 1796 struct radix_node_head *rnh; 1797 struct table_entry *ent; 1798 struct sockaddr_in sa, mask; 1799 1800 if (tbl >= IPFW_TABLES_MAX) 1801 return (EINVAL); 1802 rnh = ch->tables[tbl]; 1803 sa.sin_len = mask.sin_len = 8; 1804 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1805 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1806 IPFW_WLOCK(ch); 1807 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1808 if (ent == NULL) { 1809 IPFW_WUNLOCK(ch); 1810 return (ESRCH); 1811 } 1812 IPFW_WUNLOCK(ch); 1813 free(ent, M_IPFW_TBL); 1814 return (0); 1815} 1816 1817static int 1818flush_table_entry(struct radix_node *rn, void *arg) 1819{ 1820 struct radix_node_head * const rnh = arg; 1821 struct table_entry *ent; 1822 1823 ent = (struct table_entry *) 1824 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1825 if (ent != NULL) 1826 free(ent, M_IPFW_TBL); 1827 return (0); 1828} 1829 1830static int 1831flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1832{ 1833 struct radix_node_head *rnh; 1834 1835 IPFW_WLOCK_ASSERT(ch); 1836 1837 if (tbl >= IPFW_TABLES_MAX) 1838 return (EINVAL); 1839 rnh = ch->tables[tbl]; 1840 KASSERT(rnh != NULL, ("NULL IPFW table")); 1841 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1842 return (0); 1843} 1844 1845static void 1846flush_tables(struct ip_fw_chain *ch) 1847{ 1848 uint16_t tbl; 1849 1850 IPFW_WLOCK_ASSERT(ch); 1851 1852 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1853 flush_table(ch, tbl); 1854} 1855 1856static int 1857init_tables(struct ip_fw_chain *ch) 1858{ 1859 int i; 1860 uint16_t j; 1861 1862 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1863 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1864 for (j = 0; j < i; j++) { 1865 (void) flush_table(ch, j); 1866 } 1867 return (ENOMEM); 1868 } 1869 } 1870 return (0); 1871} 1872 1873static int 1874lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1875 uint32_t *val) 1876{ 1877 struct radix_node_head *rnh; 1878 struct table_entry *ent; 1879 struct sockaddr_in sa; 1880 1881 if (tbl >= IPFW_TABLES_MAX) 1882 return (0); 1883 rnh = ch->tables[tbl]; 1884 sa.sin_len = 8; 1885 sa.sin_addr.s_addr = addr; 1886 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1887 if (ent != NULL) { 1888 *val = ent->value; 1889 return (1); 1890 } 1891 return (0); 1892} 1893 1894static int 1895count_table_entry(struct radix_node *rn, void *arg) 1896{ 1897 u_int32_t * const cnt = arg; 1898 1899 (*cnt)++; 1900 return (0); 1901} 1902 1903static int 1904count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1905{ 1906 struct radix_node_head *rnh; 1907 1908 if (tbl >= IPFW_TABLES_MAX) 1909 return (EINVAL); 1910 rnh = ch->tables[tbl]; 1911 *cnt = 0; 1912 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1913 return (0); 1914} 1915 1916static int 1917dump_table_entry(struct radix_node *rn, void *arg) 1918{ 1919 struct table_entry * const n = (struct table_entry *)rn; 1920 ipfw_table * const tbl = arg; 1921 ipfw_table_entry *ent; 1922 1923 if (tbl->cnt == tbl->size) 1924 return (1); 1925 ent = &tbl->ent[tbl->cnt]; 1926 ent->tbl = tbl->tbl; 1927 if (in_nullhost(n->mask.sin_addr)) 1928 ent->masklen = 0; 1929 else 1930 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1931 ent->addr = n->addr.sin_addr.s_addr; 1932 ent->value = n->value; 1933 tbl->cnt++; 1934 return (0); 1935} 1936 1937static int 1938dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1939{ 1940 struct radix_node_head *rnh; 1941 1942 if (tbl->tbl >= IPFW_TABLES_MAX) 1943 return (EINVAL); 1944 rnh = ch->tables[tbl->tbl]; 1945 tbl->cnt = 0; 1946 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1947 return (0); 1948} 1949 1950static void 1951fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1952{ 1953 struct ucred *cr; 1954 1955 if (inp->inp_socket != NULL) { 1956 cr = inp->inp_socket->so_cred; 1957 ugp->fw_prid = jailed(cr) ? 1958 cr->cr_prison->pr_id : -1; 1959 ugp->fw_uid = cr->cr_uid; 1960 ugp->fw_ngroups = cr->cr_ngroups; 1961 bcopy(cr->cr_groups, ugp->fw_groups, 1962 sizeof(ugp->fw_groups)); 1963 } 1964} 1965 1966static int 1967check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 1968 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 1969 u_int16_t src_port, struct ip_fw_ugid *ugp, int *lookup, 1970 struct inpcb *inp) 1971{ 1972 struct inpcbinfo *pi; 1973 int wildcard; 1974 struct inpcb *pcb; 1975 int match; 1976 gid_t *gp; 1977 1978 /* 1979 * Check to see if the UDP or TCP stack supplied us with 1980 * the PCB. If so, rather then holding a lock and looking 1981 * up the PCB, we can use the one that was supplied. 1982 */ 1983 if (inp && *lookup == 0) { 1984 INP_LOCK_ASSERT(inp); 1985 if (inp->inp_socket != NULL) { 1986 fill_ugid_cache(inp, ugp); 1987 *lookup = 1; 1988 } 1989 } 1990 /* 1991 * If we have already been here and the packet has no 1992 * PCB entry associated with it, then we can safely 1993 * assume that this is a no match. 1994 */ 1995 if (*lookup == -1) 1996 return (0); 1997 if (proto == IPPROTO_TCP) { 1998 wildcard = 0; 1999 pi = &tcbinfo; 2000 } else if (proto == IPPROTO_UDP) { 2001 wildcard = INPLOOKUP_WILDCARD; 2002 pi = &udbinfo; 2003 } else 2004 return 0; 2005 match = 0; 2006 if (*lookup == 0) { 2007 INP_INFO_RLOCK(pi); 2008 pcb = (oif) ? 2009 in_pcblookup_hash(pi, 2010 dst_ip, htons(dst_port), 2011 src_ip, htons(src_port), 2012 wildcard, oif) : 2013 in_pcblookup_hash(pi, 2014 src_ip, htons(src_port), 2015 dst_ip, htons(dst_port), 2016 wildcard, NULL); 2017 if (pcb != NULL) { 2018 INP_RLOCK(pcb); 2019 if (pcb->inp_socket != NULL) { 2020 fill_ugid_cache(pcb, ugp); 2021 *lookup = 1; 2022 } 2023 INP_RUNLOCK(pcb); 2024 } 2025 INP_INFO_RUNLOCK(pi); 2026 if (*lookup == 0) { 2027 /* 2028 * If the lookup did not yield any results, there 2029 * is no sense in coming back and trying again. So 2030 * we can set lookup to -1 and ensure that we wont 2031 * bother the pcb system again. 2032 */ 2033 *lookup = -1; 2034 return (0); 2035 } 2036 } 2037 if (insn->o.opcode == O_UID) 2038 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2039 else if (insn->o.opcode == O_GID) { 2040 for (gp = ugp->fw_groups; 2041 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2042 if (*gp == (gid_t)insn->d[0]) { 2043 match = 1; 2044 break; 2045 } 2046 } else if (insn->o.opcode == O_JAIL) 2047 match = (ugp->fw_prid == (int)insn->d[0]); 2048 return match; 2049} 2050 2051/* 2052 * The main check routine for the firewall. 2053 * 2054 * All arguments are in args so we can modify them and return them 2055 * back to the caller. 2056 * 2057 * Parameters: 2058 * 2059 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2060 * Starts with the IP header. 2061 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2062 * args->L3offset Number of bytes bypassed if we came from L2. 2063 * e.g. often sizeof(eh) ** NOTYET ** 2064 * args->oif Outgoing interface, or NULL if packet is incoming. 2065 * The incoming interface is in the mbuf. (in) 2066 * args->divert_rule (in/out) 2067 * Skip up to the first rule past this rule number; 2068 * upon return, non-zero port number for divert or tee. 2069 * 2070 * args->rule Pointer to the last matching rule (in/out) 2071 * args->next_hop Socket we are forwarding to (out). 2072 * args->f_id Addresses grabbed from the packet (out) 2073 * args->cookie a cookie depending on rule action 2074 * 2075 * Return value: 2076 * 2077 * IP_FW_PASS the packet must be accepted 2078 * IP_FW_DENY the packet must be dropped 2079 * IP_FW_DIVERT divert packet, port in m_tag 2080 * IP_FW_TEE tee packet, port in m_tag 2081 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2082 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2083 * 2084 */ 2085int 2086ipfw_chk(struct ip_fw_args *args) 2087{ 2088 /* 2089 * Local variables holding state during the processing of a packet: 2090 * 2091 * IMPORTANT NOTE: to speed up the processing of rules, there 2092 * are some assumption on the values of the variables, which 2093 * are documented here. Should you change them, please check 2094 * the implementation of the various instructions to make sure 2095 * that they still work. 2096 * 2097 * args->eh The MAC header. It is non-null for a layer2 2098 * packet, it is NULL for a layer-3 packet. 2099 * **notyet** 2100 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2101 * 2102 * m | args->m Pointer to the mbuf, as received from the caller. 2103 * It may change if ipfw_chk() does an m_pullup, or if it 2104 * consumes the packet because it calls send_reject(). 2105 * XXX This has to change, so that ipfw_chk() never modifies 2106 * or consumes the buffer. 2107 * ip is the beginning of the ip(4 or 6) header. 2108 * Calculated by adding the L3offset to the start of data. 2109 * (Until we start using L3offset, the packet is 2110 * supposed to start with the ip header). 2111 */ 2112 struct mbuf *m = args->m; 2113 struct ip *ip = mtod(m, struct ip *); 2114 2115 /* 2116 * For rules which contain uid/gid or jail constraints, cache 2117 * a copy of the users credentials after the pcb lookup has been 2118 * executed. This will speed up the processing of rules with 2119 * these types of constraints, as well as decrease contention 2120 * on pcb related locks. 2121 */ 2122 struct ip_fw_ugid fw_ugid_cache; 2123 int ugid_lookup = 0; 2124 2125 /* 2126 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2127 * associated with a packet input on a divert socket. This 2128 * will allow to distinguish traffic and its direction when 2129 * it originates from a divert socket. 2130 */ 2131 u_int divinput_flags = 0; 2132 2133 /* 2134 * oif | args->oif If NULL, ipfw_chk has been called on the 2135 * inbound path (ether_input, ip_input). 2136 * If non-NULL, ipfw_chk has been called on the outbound path 2137 * (ether_output, ip_output). 2138 */ 2139 struct ifnet *oif = args->oif; 2140 2141 struct ip_fw *f = NULL; /* matching rule */ 2142 int retval = 0; 2143 2144 /* 2145 * hlen The length of the IP header. 2146 */ 2147 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2148 2149 /* 2150 * offset The offset of a fragment. offset != 0 means that 2151 * we have a fragment at this offset of an IPv4 packet. 2152 * offset == 0 means that (if this is an IPv4 packet) 2153 * this is the first or only fragment. 2154 * For IPv6 offset == 0 means there is no Fragment Header. 2155 * If offset != 0 for IPv6 always use correct mask to 2156 * get the correct offset because we add IP6F_MORE_FRAG 2157 * to be able to dectect the first fragment which would 2158 * otherwise have offset = 0. 2159 */ 2160 u_short offset = 0; 2161 2162 /* 2163 * Local copies of addresses. They are only valid if we have 2164 * an IP packet. 2165 * 2166 * proto The protocol. Set to 0 for non-ip packets, 2167 * or to the protocol read from the packet otherwise. 2168 * proto != 0 means that we have an IPv4 packet. 2169 * 2170 * src_port, dst_port port numbers, in HOST format. Only 2171 * valid for TCP and UDP packets. 2172 * 2173 * src_ip, dst_ip ip addresses, in NETWORK format. 2174 * Only valid for IPv4 packets. 2175 */ 2176 u_int8_t proto; 2177 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2178 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2179 u_int16_t ip_len=0; 2180 int pktlen; 2181 u_int16_t etype = 0; /* Host order stored ether type */ 2182 2183 /* 2184 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2185 * MATCH_NONE when checked and not matched (q = NULL), 2186 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2187 */ 2188 int dyn_dir = MATCH_UNKNOWN; 2189 ipfw_dyn_rule *q = NULL; 2190 struct ip_fw_chain *chain = &layer3_chain; 2191 struct m_tag *mtag; 2192 2193 /* 2194 * We store in ulp a pointer to the upper layer protocol header. 2195 * In the ipv4 case this is easy to determine from the header, 2196 * but for ipv6 we might have some additional headers in the middle. 2197 * ulp is NULL if not found. 2198 */ 2199 void *ulp = NULL; /* upper layer protocol pointer. */ 2200 /* XXX ipv6 variables */ 2201 int is_ipv6 = 0; 2202 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2203 /* end of ipv6 variables */ 2204 int is_ipv4 = 0; 2205 2206 if (m->m_flags & M_SKIP_FIREWALL) 2207 return (IP_FW_PASS); /* accept */ 2208 2209 pktlen = m->m_pkthdr.len; 2210 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 2211 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2212 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2213 2214/* 2215 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2216 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2217 * pointer might become stale after other pullups (but we never use it 2218 * this way). 2219 */ 2220#define PULLUP_TO(len, p, T) \ 2221do { \ 2222 int x = (len) + sizeof(T); \ 2223 if ((m)->m_len < x) { \ 2224 args->m = m = m_pullup(m, x); \ 2225 if (m == NULL) \ 2226 goto pullup_failed; \ 2227 } \ 2228 p = (mtod(m, char *) + (len)); \ 2229} while (0) 2230 2231 /* 2232 * if we have an ether header, 2233 */ 2234 if (args->eh) 2235 etype = ntohs(args->eh->ether_type); 2236 2237 /* Identify IP packets and fill up variables. */ 2238 if (pktlen >= sizeof(struct ip6_hdr) && 2239 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2240 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2241 is_ipv6 = 1; 2242 args->f_id.addr_type = 6; 2243 hlen = sizeof(struct ip6_hdr); 2244 proto = ip6->ip6_nxt; 2245 2246 /* Search extension headers to find upper layer protocols */ 2247 while (ulp == NULL) { 2248 switch (proto) { 2249 case IPPROTO_ICMPV6: 2250 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2251 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2252 break; 2253 2254 case IPPROTO_TCP: 2255 PULLUP_TO(hlen, ulp, struct tcphdr); 2256 dst_port = TCP(ulp)->th_dport; 2257 src_port = TCP(ulp)->th_sport; 2258 args->f_id.flags = TCP(ulp)->th_flags; 2259 break; 2260 2261 case IPPROTO_SCTP: 2262 PULLUP_TO(hlen, ulp, struct sctphdr); 2263 src_port = SCTP(ulp)->src_port; 2264 dst_port = SCTP(ulp)->dest_port; 2265 break; 2266 2267 case IPPROTO_UDP: 2268 PULLUP_TO(hlen, ulp, struct udphdr); 2269 dst_port = UDP(ulp)->uh_dport; 2270 src_port = UDP(ulp)->uh_sport; 2271 break; 2272 2273 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2274 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2275 ext_hd |= EXT_HOPOPTS; 2276 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2277 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2278 ulp = NULL; 2279 break; 2280 2281 case IPPROTO_ROUTING: /* RFC 2460 */ 2282 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2283 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2284 case 0: 2285 ext_hd |= EXT_RTHDR0; 2286 break; 2287 case 2: 2288 ext_hd |= EXT_RTHDR2; 2289 break; 2290 default: 2291 printf("IPFW2: IPV6 - Unknown Routing " 2292 "Header type(%d)\n", 2293 ((struct ip6_rthdr *)ulp)->ip6r_type); 2294 if (fw_deny_unknown_exthdrs) 2295 return (IP_FW_DENY); 2296 break; 2297 } 2298 ext_hd |= EXT_ROUTING; 2299 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2300 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2301 ulp = NULL; 2302 break; 2303 2304 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2305 PULLUP_TO(hlen, ulp, struct ip6_frag); 2306 ext_hd |= EXT_FRAGMENT; 2307 hlen += sizeof (struct ip6_frag); 2308 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2309 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2310 IP6F_OFF_MASK; 2311 /* Add IP6F_MORE_FRAG for offset of first 2312 * fragment to be != 0. */ 2313 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2314 IP6F_MORE_FRAG; 2315 if (offset == 0) { 2316 printf("IPFW2: IPV6 - Invalid Fragment " 2317 "Header\n"); 2318 if (fw_deny_unknown_exthdrs) 2319 return (IP_FW_DENY); 2320 break; 2321 } 2322 args->f_id.frag_id6 = 2323 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2324 ulp = NULL; 2325 break; 2326 2327 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2328 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2329 ext_hd |= EXT_DSTOPTS; 2330 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2331 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2332 ulp = NULL; 2333 break; 2334 2335 case IPPROTO_AH: /* RFC 2402 */ 2336 PULLUP_TO(hlen, ulp, struct ip6_ext); 2337 ext_hd |= EXT_AH; 2338 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2339 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2340 ulp = NULL; 2341 break; 2342 2343 case IPPROTO_ESP: /* RFC 2406 */ 2344 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2345 /* Anything past Seq# is variable length and 2346 * data past this ext. header is encrypted. */ 2347 ext_hd |= EXT_ESP; 2348 break; 2349 2350 case IPPROTO_NONE: /* RFC 2460 */ 2351 /* 2352 * Packet ends here, and IPv6 header has 2353 * already been pulled up. If ip6e_len!=0 2354 * then octets must be ignored. 2355 */ 2356 ulp = ip; /* non-NULL to get out of loop. */ 2357 break; 2358 2359 case IPPROTO_OSPFIGP: 2360 /* XXX OSPF header check? */ 2361 PULLUP_TO(hlen, ulp, struct ip6_ext); 2362 break; 2363 2364 case IPPROTO_PIM: 2365 /* XXX PIM header check? */ 2366 PULLUP_TO(hlen, ulp, struct pim); 2367 break; 2368 2369 case IPPROTO_CARP: 2370 PULLUP_TO(hlen, ulp, struct carp_header); 2371 if (((struct carp_header *)ulp)->carp_version != 2372 CARP_VERSION) 2373 return (IP_FW_DENY); 2374 if (((struct carp_header *)ulp)->carp_type != 2375 CARP_ADVERTISEMENT) 2376 return (IP_FW_DENY); 2377 break; 2378 2379 case IPPROTO_IPV6: /* RFC 2893 */ 2380 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2381 break; 2382 2383 case IPPROTO_IPV4: /* RFC 2893 */ 2384 PULLUP_TO(hlen, ulp, struct ip); 2385 break; 2386 2387 default: 2388 printf("IPFW2: IPV6 - Unknown Extension " 2389 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2390 if (fw_deny_unknown_exthdrs) 2391 return (IP_FW_DENY); 2392 PULLUP_TO(hlen, ulp, struct ip6_ext); 2393 break; 2394 } /*switch */ 2395 } 2396 ip = mtod(m, struct ip *); 2397 ip6 = (struct ip6_hdr *)ip; 2398 args->f_id.src_ip6 = ip6->ip6_src; 2399 args->f_id.dst_ip6 = ip6->ip6_dst; 2400 args->f_id.src_ip = 0; 2401 args->f_id.dst_ip = 0; 2402 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2403 } else if (pktlen >= sizeof(struct ip) && 2404 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2405 is_ipv4 = 1; 2406 hlen = ip->ip_hl << 2; 2407 args->f_id.addr_type = 4; 2408 2409 /* 2410 * Collect parameters into local variables for faster matching. 2411 */ 2412 proto = ip->ip_p; 2413 src_ip = ip->ip_src; 2414 dst_ip = ip->ip_dst; 2415 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2416 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2417 ip_len = ntohs(ip->ip_len); 2418 } else { 2419 offset = ip->ip_off & IP_OFFMASK; 2420 ip_len = ip->ip_len; 2421 } 2422 pktlen = ip_len < pktlen ? ip_len : pktlen; 2423 2424 if (offset == 0) { 2425 switch (proto) { 2426 case IPPROTO_TCP: 2427 PULLUP_TO(hlen, ulp, struct tcphdr); 2428 dst_port = TCP(ulp)->th_dport; 2429 src_port = TCP(ulp)->th_sport; 2430 args->f_id.flags = TCP(ulp)->th_flags; 2431 break; 2432 2433 case IPPROTO_UDP: 2434 PULLUP_TO(hlen, ulp, struct udphdr); 2435 dst_port = UDP(ulp)->uh_dport; 2436 src_port = UDP(ulp)->uh_sport; 2437 break; 2438 2439 case IPPROTO_ICMP: 2440 PULLUP_TO(hlen, ulp, struct icmphdr); 2441 args->f_id.flags = ICMP(ulp)->icmp_type; 2442 break; 2443 2444 default: 2445 break; 2446 } 2447 } 2448 2449 ip = mtod(m, struct ip *); 2450 args->f_id.src_ip = ntohl(src_ip.s_addr); 2451 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2452 } 2453#undef PULLUP_TO 2454 if (proto) { /* we may have port numbers, store them */ 2455 args->f_id.proto = proto; 2456 args->f_id.src_port = src_port = ntohs(src_port); 2457 args->f_id.dst_port = dst_port = ntohs(dst_port); 2458 } 2459 2460 IPFW_RLOCK(chain); 2461 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2462 if (args->rule) { 2463 /* 2464 * Packet has already been tagged. Look for the next rule 2465 * to restart processing. 2466 * 2467 * If fw_one_pass != 0 then just accept it. 2468 * XXX should not happen here, but optimized out in 2469 * the caller. 2470 */ 2471 if (fw_one_pass) { 2472 IPFW_RUNLOCK(chain); 2473 return (IP_FW_PASS); 2474 } 2475 2476 f = args->rule->next_rule; 2477 if (f == NULL) 2478 f = lookup_next_rule(args->rule); 2479 } else { 2480 /* 2481 * Find the starting rule. It can be either the first 2482 * one, or the one after divert_rule if asked so. 2483 */ 2484 int skipto = mtag ? divert_cookie(mtag) : 0; 2485 2486 f = chain->rules; 2487 if (args->eh == NULL && skipto != 0) { 2488 if (skipto >= IPFW_DEFAULT_RULE) { 2489 IPFW_RUNLOCK(chain); 2490 return (IP_FW_DENY); /* invalid */ 2491 } 2492 while (f && f->rulenum <= skipto) 2493 f = f->next; 2494 if (f == NULL) { /* drop packet */ 2495 IPFW_RUNLOCK(chain); 2496 return (IP_FW_DENY); 2497 } 2498 } 2499 } 2500 /* reset divert rule to avoid confusion later */ 2501 if (mtag) { 2502 divinput_flags = divert_info(mtag) & 2503 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2504 m_tag_delete(m, mtag); 2505 } 2506 2507 /* 2508 * Now scan the rules, and parse microinstructions for each rule. 2509 */ 2510 for (; f; f = f->next) { 2511 ipfw_insn *cmd; 2512 uint32_t tablearg = 0; 2513 int l, cmdlen, skip_or; /* skip rest of OR block */ 2514 2515again: 2516 if (set_disable & (1 << f->set) ) 2517 continue; 2518 2519 skip_or = 0; 2520 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2521 l -= cmdlen, cmd += cmdlen) { 2522 int match; 2523 2524 /* 2525 * check_body is a jump target used when we find a 2526 * CHECK_STATE, and need to jump to the body of 2527 * the target rule. 2528 */ 2529 2530check_body: 2531 cmdlen = F_LEN(cmd); 2532 /* 2533 * An OR block (insn_1 || .. || insn_n) has the 2534 * F_OR bit set in all but the last instruction. 2535 * The first match will set "skip_or", and cause 2536 * the following instructions to be skipped until 2537 * past the one with the F_OR bit clear. 2538 */ 2539 if (skip_or) { /* skip this instruction */ 2540 if ((cmd->len & F_OR) == 0) 2541 skip_or = 0; /* next one is good */ 2542 continue; 2543 } 2544 match = 0; /* set to 1 if we succeed */ 2545 2546 switch (cmd->opcode) { 2547 /* 2548 * The first set of opcodes compares the packet's 2549 * fields with some pattern, setting 'match' if a 2550 * match is found. At the end of the loop there is 2551 * logic to deal with F_NOT and F_OR flags associated 2552 * with the opcode. 2553 */ 2554 case O_NOP: 2555 match = 1; 2556 break; 2557 2558 case O_FORWARD_MAC: 2559 printf("ipfw: opcode %d unimplemented\n", 2560 cmd->opcode); 2561 break; 2562 2563 case O_GID: 2564 case O_UID: 2565 case O_JAIL: 2566 /* 2567 * We only check offset == 0 && proto != 0, 2568 * as this ensures that we have a 2569 * packet with the ports info. 2570 */ 2571 if (offset!=0) 2572 break; 2573 if (is_ipv6) /* XXX to be fixed later */ 2574 break; 2575 if (proto == IPPROTO_TCP || 2576 proto == IPPROTO_UDP) 2577 match = check_uidgid( 2578 (ipfw_insn_u32 *)cmd, 2579 proto, oif, 2580 dst_ip, dst_port, 2581 src_ip, src_port, &fw_ugid_cache, 2582 &ugid_lookup, args->inp); 2583 break; 2584 2585 case O_RECV: 2586 match = iface_match(m->m_pkthdr.rcvif, 2587 (ipfw_insn_if *)cmd); 2588 break; 2589 2590 case O_XMIT: 2591 match = iface_match(oif, (ipfw_insn_if *)cmd); 2592 break; 2593 2594 case O_VIA: 2595 match = iface_match(oif ? oif : 2596 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2597 break; 2598 2599 case O_MACADDR2: 2600 if (args->eh != NULL) { /* have MAC header */ 2601 u_int32_t *want = (u_int32_t *) 2602 ((ipfw_insn_mac *)cmd)->addr; 2603 u_int32_t *mask = (u_int32_t *) 2604 ((ipfw_insn_mac *)cmd)->mask; 2605 u_int32_t *hdr = (u_int32_t *)args->eh; 2606 2607 match = 2608 ( want[0] == (hdr[0] & mask[0]) && 2609 want[1] == (hdr[1] & mask[1]) && 2610 want[2] == (hdr[2] & mask[2]) ); 2611 } 2612 break; 2613 2614 case O_MAC_TYPE: 2615 if (args->eh != NULL) { 2616 u_int16_t *p = 2617 ((ipfw_insn_u16 *)cmd)->ports; 2618 int i; 2619 2620 for (i = cmdlen - 1; !match && i>0; 2621 i--, p += 2) 2622 match = (etype >= p[0] && 2623 etype <= p[1]); 2624 } 2625 break; 2626 2627 case O_FRAG: 2628 match = (offset != 0); 2629 break; 2630 2631 case O_IN: /* "out" is "not in" */ 2632 match = (oif == NULL); 2633 break; 2634 2635 case O_LAYER2: 2636 match = (args->eh != NULL); 2637 break; 2638 2639 case O_DIVERTED: 2640 match = (cmd->arg1 & 1 && divinput_flags & 2641 IP_FW_DIVERT_LOOPBACK_FLAG) || 2642 (cmd->arg1 & 2 && divinput_flags & 2643 IP_FW_DIVERT_OUTPUT_FLAG); 2644 break; 2645 2646 case O_PROTO: 2647 /* 2648 * We do not allow an arg of 0 so the 2649 * check of "proto" only suffices. 2650 */ 2651 match = (proto == cmd->arg1); 2652 break; 2653 2654 case O_IP_SRC: 2655 match = is_ipv4 && 2656 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2657 src_ip.s_addr); 2658 break; 2659 2660 case O_IP_SRC_LOOKUP: 2661 case O_IP_DST_LOOKUP: 2662 if (is_ipv4) { 2663 uint32_t a = 2664 (cmd->opcode == O_IP_DST_LOOKUP) ? 2665 dst_ip.s_addr : src_ip.s_addr; 2666 uint32_t v; 2667 2668 match = lookup_table(chain, cmd->arg1, a, 2669 &v); 2670 if (!match) 2671 break; 2672 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2673 match = 2674 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2675 else 2676 tablearg = v; 2677 } 2678 break; 2679 2680 case O_IP_SRC_MASK: 2681 case O_IP_DST_MASK: 2682 if (is_ipv4) { 2683 uint32_t a = 2684 (cmd->opcode == O_IP_DST_MASK) ? 2685 dst_ip.s_addr : src_ip.s_addr; 2686 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2687 int i = cmdlen-1; 2688 2689 for (; !match && i>0; i-= 2, p+= 2) 2690 match = (p[0] == (a & p[1])); 2691 } 2692 break; 2693 2694 case O_IP_SRC_ME: 2695 if (is_ipv4) { 2696 struct ifnet *tif; 2697 2698 INADDR_TO_IFP(src_ip, tif); 2699 match = (tif != NULL); 2700 } 2701 break; 2702 2703 case O_IP_DST_SET: 2704 case O_IP_SRC_SET: 2705 if (is_ipv4) { 2706 u_int32_t *d = (u_int32_t *)(cmd+1); 2707 u_int32_t addr = 2708 cmd->opcode == O_IP_DST_SET ? 2709 args->f_id.dst_ip : 2710 args->f_id.src_ip; 2711 2712 if (addr < d[0]) 2713 break; 2714 addr -= d[0]; /* subtract base */ 2715 match = (addr < cmd->arg1) && 2716 ( d[ 1 + (addr>>5)] & 2717 (1<<(addr & 0x1f)) ); 2718 } 2719 break; 2720 2721 case O_IP_DST: 2722 match = is_ipv4 && 2723 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2724 dst_ip.s_addr); 2725 break; 2726 2727 case O_IP_DST_ME: 2728 if (is_ipv4) { 2729 struct ifnet *tif; 2730 2731 INADDR_TO_IFP(dst_ip, tif); 2732 match = (tif != NULL); 2733 } 2734 break; 2735 2736 case O_IP_SRCPORT: 2737 case O_IP_DSTPORT: 2738 /* 2739 * offset == 0 && proto != 0 is enough 2740 * to guarantee that we have a 2741 * packet with port info. 2742 */ 2743 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2744 && offset == 0) { 2745 u_int16_t x = 2746 (cmd->opcode == O_IP_SRCPORT) ? 2747 src_port : dst_port ; 2748 u_int16_t *p = 2749 ((ipfw_insn_u16 *)cmd)->ports; 2750 int i; 2751 2752 for (i = cmdlen - 1; !match && i>0; 2753 i--, p += 2) 2754 match = (x>=p[0] && x<=p[1]); 2755 } 2756 break; 2757 2758 case O_ICMPTYPE: 2759 match = (offset == 0 && proto==IPPROTO_ICMP && 2760 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2761 break; 2762 2763#ifdef INET6 2764 case O_ICMP6TYPE: 2765 match = is_ipv6 && offset == 0 && 2766 proto==IPPROTO_ICMPV6 && 2767 icmp6type_match( 2768 ICMP6(ulp)->icmp6_type, 2769 (ipfw_insn_u32 *)cmd); 2770 break; 2771#endif /* INET6 */ 2772 2773 case O_IPOPT: 2774 match = (is_ipv4 && 2775 ipopts_match(ip, cmd) ); 2776 break; 2777 2778 case O_IPVER: 2779 match = (is_ipv4 && 2780 cmd->arg1 == ip->ip_v); 2781 break; 2782 2783 case O_IPID: 2784 case O_IPLEN: 2785 case O_IPTTL: 2786 if (is_ipv4) { /* only for IP packets */ 2787 uint16_t x; 2788 uint16_t *p; 2789 int i; 2790 2791 if (cmd->opcode == O_IPLEN) 2792 x = ip_len; 2793 else if (cmd->opcode == O_IPTTL) 2794 x = ip->ip_ttl; 2795 else /* must be IPID */ 2796 x = ntohs(ip->ip_id); 2797 if (cmdlen == 1) { 2798 match = (cmd->arg1 == x); 2799 break; 2800 } 2801 /* otherwise we have ranges */ 2802 p = ((ipfw_insn_u16 *)cmd)->ports; 2803 i = cmdlen - 1; 2804 for (; !match && i>0; i--, p += 2) 2805 match = (x >= p[0] && x <= p[1]); 2806 } 2807 break; 2808 2809 case O_IPPRECEDENCE: 2810 match = (is_ipv4 && 2811 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2812 break; 2813 2814 case O_IPTOS: 2815 match = (is_ipv4 && 2816 flags_match(cmd, ip->ip_tos)); 2817 break; 2818 2819 case O_TCPDATALEN: 2820 if (proto == IPPROTO_TCP && offset == 0) { 2821 struct tcphdr *tcp; 2822 uint16_t x; 2823 uint16_t *p; 2824 int i; 2825 2826 tcp = TCP(ulp); 2827 x = ip_len - 2828 ((ip->ip_hl + tcp->th_off) << 2); 2829 if (cmdlen == 1) { 2830 match = (cmd->arg1 == x); 2831 break; 2832 } 2833 /* otherwise we have ranges */ 2834 p = ((ipfw_insn_u16 *)cmd)->ports; 2835 i = cmdlen - 1; 2836 for (; !match && i>0; i--, p += 2) 2837 match = (x >= p[0] && x <= p[1]); 2838 } 2839 break; 2840 2841 case O_TCPFLAGS: 2842 match = (proto == IPPROTO_TCP && offset == 0 && 2843 flags_match(cmd, TCP(ulp)->th_flags)); 2844 break; 2845 2846 case O_TCPOPTS: 2847 match = (proto == IPPROTO_TCP && offset == 0 && 2848 tcpopts_match(TCP(ulp), cmd)); 2849 break; 2850 2851 case O_TCPSEQ: 2852 match = (proto == IPPROTO_TCP && offset == 0 && 2853 ((ipfw_insn_u32 *)cmd)->d[0] == 2854 TCP(ulp)->th_seq); 2855 break; 2856 2857 case O_TCPACK: 2858 match = (proto == IPPROTO_TCP && offset == 0 && 2859 ((ipfw_insn_u32 *)cmd)->d[0] == 2860 TCP(ulp)->th_ack); 2861 break; 2862 2863 case O_TCPWIN: 2864 match = (proto == IPPROTO_TCP && offset == 0 && 2865 cmd->arg1 == TCP(ulp)->th_win); 2866 break; 2867 2868 case O_ESTAB: 2869 /* reject packets which have SYN only */ 2870 /* XXX should i also check for TH_ACK ? */ 2871 match = (proto == IPPROTO_TCP && offset == 0 && 2872 (TCP(ulp)->th_flags & 2873 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2874 break; 2875 2876 case O_ALTQ: { 2877 struct pf_mtag *at; 2878 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2879 2880 match = 1; 2881 at = pf_find_mtag(m); 2882 if (at != NULL && at->qid != 0) 2883 break; 2884 at = pf_get_mtag(m); 2885 if (at == NULL) { 2886 /* 2887 * Let the packet fall back to the 2888 * default ALTQ. 2889 */ 2890 break; 2891 } 2892 at->qid = altq->qid; 2893 if (is_ipv4) 2894 at->af = AF_INET; 2895 else 2896 at->af = AF_LINK; 2897 at->hdr = ip; 2898 break; 2899 } 2900 2901 case O_LOG: 2902 if (fw_verbose) 2903 ipfw_log(f, hlen, args, m, 2904 oif, offset, tablearg, ip); 2905 match = 1; 2906 break; 2907 2908 case O_PROB: 2909 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2910 break; 2911 2912 case O_VERREVPATH: 2913 /* Outgoing packets automatically pass/match */ 2914 match = ((oif != NULL) || 2915 (m->m_pkthdr.rcvif == NULL) || 2916 ( 2917#ifdef INET6 2918 is_ipv6 ? 2919 verify_path6(&(args->f_id.src_ip6), 2920 m->m_pkthdr.rcvif) : 2921#endif 2922 verify_path(src_ip, m->m_pkthdr.rcvif, 2923 args->f_id.fib))); 2924 break; 2925 2926 case O_VERSRCREACH: 2927 /* Outgoing packets automatically pass/match */ 2928 match = (hlen > 0 && ((oif != NULL) || 2929#ifdef INET6 2930 is_ipv6 ? 2931 verify_path6(&(args->f_id.src_ip6), 2932 NULL) : 2933#endif 2934 verify_path(src_ip, NULL, args->f_id.fib))); 2935 break; 2936 2937 case O_ANTISPOOF: 2938 /* Outgoing packets automatically pass/match */ 2939 if (oif == NULL && hlen > 0 && 2940 ( (is_ipv4 && in_localaddr(src_ip)) 2941#ifdef INET6 2942 || (is_ipv6 && 2943 in6_localaddr(&(args->f_id.src_ip6))) 2944#endif 2945 )) 2946 match = 2947#ifdef INET6 2948 is_ipv6 ? verify_path6( 2949 &(args->f_id.src_ip6), 2950 m->m_pkthdr.rcvif) : 2951#endif 2952 verify_path(src_ip, 2953 m->m_pkthdr.rcvif, 2954 args->f_id.fib); 2955 else 2956 match = 1; 2957 break; 2958 2959 case O_IPSEC: 2960#ifdef IPSEC 2961 match = (m_tag_find(m, 2962 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2963#endif 2964 /* otherwise no match */ 2965 break; 2966 2967#ifdef INET6 2968 case O_IP6_SRC: 2969 match = is_ipv6 && 2970 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2971 &((ipfw_insn_ip6 *)cmd)->addr6); 2972 break; 2973 2974 case O_IP6_DST: 2975 match = is_ipv6 && 2976 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2977 &((ipfw_insn_ip6 *)cmd)->addr6); 2978 break; 2979 case O_IP6_SRC_MASK: 2980 case O_IP6_DST_MASK: 2981 if (is_ipv6) { 2982 int i = cmdlen - 1; 2983 struct in6_addr p; 2984 struct in6_addr *d = 2985 &((ipfw_insn_ip6 *)cmd)->addr6; 2986 2987 for (; !match && i > 0; d += 2, 2988 i -= F_INSN_SIZE(struct in6_addr) 2989 * 2) { 2990 p = (cmd->opcode == 2991 O_IP6_SRC_MASK) ? 2992 args->f_id.src_ip6: 2993 args->f_id.dst_ip6; 2994 APPLY_MASK(&p, &d[1]); 2995 match = 2996 IN6_ARE_ADDR_EQUAL(&d[0], 2997 &p); 2998 } 2999 } 3000 break; 3001 3002 case O_IP6_SRC_ME: 3003 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3004 break; 3005 3006 case O_IP6_DST_ME: 3007 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3008 break; 3009 3010 case O_FLOW6ID: 3011 match = is_ipv6 && 3012 flow6id_match(args->f_id.flow_id6, 3013 (ipfw_insn_u32 *) cmd); 3014 break; 3015 3016 case O_EXT_HDR: 3017 match = is_ipv6 && 3018 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3019 break; 3020 3021 case O_IP6: 3022 match = is_ipv6; 3023 break; 3024#endif 3025 3026 case O_IP4: 3027 match = is_ipv4; 3028 break; 3029 3030 case O_TAG: { 3031 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3032 tablearg : cmd->arg1; 3033 3034 /* Packet is already tagged with this tag? */ 3035 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3036 3037 /* We have `untag' action when F_NOT flag is 3038 * present. And we must remove this mtag from 3039 * mbuf and reset `match' to zero (`match' will 3040 * be inversed later). 3041 * Otherwise we should allocate new mtag and 3042 * push it into mbuf. 3043 */ 3044 if (cmd->len & F_NOT) { /* `untag' action */ 3045 if (mtag != NULL) 3046 m_tag_delete(m, mtag); 3047 } else if (mtag == NULL) { 3048 if ((mtag = m_tag_alloc(MTAG_IPFW, 3049 tag, 0, M_NOWAIT)) != NULL) 3050 m_tag_prepend(m, mtag); 3051 } 3052 match = (cmd->len & F_NOT) ? 0: 1; 3053 break; 3054 } 3055 3056 case O_FIB: /* try match the specified fib */ 3057 if (args->f_id.fib == cmd->arg1) 3058 match = 1; 3059 break; 3060 3061 case O_TAGGED: { 3062 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3063 tablearg : cmd->arg1; 3064 3065 if (cmdlen == 1) { 3066 match = m_tag_locate(m, MTAG_IPFW, 3067 tag, NULL) != NULL; 3068 break; 3069 } 3070 3071 /* we have ranges */ 3072 for (mtag = m_tag_first(m); 3073 mtag != NULL && !match; 3074 mtag = m_tag_next(m, mtag)) { 3075 uint16_t *p; 3076 int i; 3077 3078 if (mtag->m_tag_cookie != MTAG_IPFW) 3079 continue; 3080 3081 p = ((ipfw_insn_u16 *)cmd)->ports; 3082 i = cmdlen - 1; 3083 for(; !match && i > 0; i--, p += 2) 3084 match = 3085 mtag->m_tag_id >= p[0] && 3086 mtag->m_tag_id <= p[1]; 3087 } 3088 break; 3089 } 3090 3091 /* 3092 * The second set of opcodes represents 'actions', 3093 * i.e. the terminal part of a rule once the packet 3094 * matches all previous patterns. 3095 * Typically there is only one action for each rule, 3096 * and the opcode is stored at the end of the rule 3097 * (but there are exceptions -- see below). 3098 * 3099 * In general, here we set retval and terminate the 3100 * outer loop (would be a 'break 3' in some language, 3101 * but we need to do a 'goto done'). 3102 * 3103 * Exceptions: 3104 * O_COUNT and O_SKIPTO actions: 3105 * instead of terminating, we jump to the next rule 3106 * ('goto next_rule', equivalent to a 'break 2'), 3107 * or to the SKIPTO target ('goto again' after 3108 * having set f, cmd and l), respectively. 3109 * 3110 * O_TAG, O_LOG and O_ALTQ action parameters: 3111 * perform some action and set match = 1; 3112 * 3113 * O_LIMIT and O_KEEP_STATE: these opcodes are 3114 * not real 'actions', and are stored right 3115 * before the 'action' part of the rule. 3116 * These opcodes try to install an entry in the 3117 * state tables; if successful, we continue with 3118 * the next opcode (match=1; break;), otherwise 3119 * the packet * must be dropped 3120 * ('goto done' after setting retval); 3121 * 3122 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3123 * cause a lookup of the state table, and a jump 3124 * to the 'action' part of the parent rule 3125 * ('goto check_body') if an entry is found, or 3126 * (CHECK_STATE only) a jump to the next rule if 3127 * the entry is not found ('goto next_rule'). 3128 * The result of the lookup is cached to make 3129 * further instances of these opcodes are 3130 * effectively NOPs. 3131 */ 3132 case O_LIMIT: 3133 case O_KEEP_STATE: 3134 if (install_state(f, 3135 (ipfw_insn_limit *)cmd, args, tablearg)) { 3136 retval = IP_FW_DENY; 3137 goto done; /* error/limit violation */ 3138 } 3139 match = 1; 3140 break; 3141 3142 case O_PROBE_STATE: 3143 case O_CHECK_STATE: 3144 /* 3145 * dynamic rules are checked at the first 3146 * keep-state or check-state occurrence, 3147 * with the result being stored in dyn_dir. 3148 * The compiler introduces a PROBE_STATE 3149 * instruction for us when we have a 3150 * KEEP_STATE (because PROBE_STATE needs 3151 * to be run first). 3152 */ 3153 if (dyn_dir == MATCH_UNKNOWN && 3154 (q = lookup_dyn_rule(&args->f_id, 3155 &dyn_dir, proto == IPPROTO_TCP ? 3156 TCP(ulp) : NULL)) 3157 != NULL) { 3158 /* 3159 * Found dynamic entry, update stats 3160 * and jump to the 'action' part of 3161 * the parent rule. 3162 */ 3163 q->pcnt++; 3164 q->bcnt += pktlen; 3165 f = q->rule; 3166 cmd = ACTION_PTR(f); 3167 l = f->cmd_len - f->act_ofs; 3168 IPFW_DYN_UNLOCK(); 3169 goto check_body; 3170 } 3171 /* 3172 * Dynamic entry not found. If CHECK_STATE, 3173 * skip to next rule, if PROBE_STATE just 3174 * ignore and continue with next opcode. 3175 */ 3176 if (cmd->opcode == O_CHECK_STATE) 3177 goto next_rule; 3178 match = 1; 3179 break; 3180 3181 case O_ACCEPT: 3182 retval = 0; /* accept */ 3183 goto done; 3184 3185 case O_PIPE: 3186 case O_QUEUE: 3187 args->rule = f; /* report matching rule */ 3188 if (cmd->arg1 == IP_FW_TABLEARG) 3189 args->cookie = tablearg; 3190 else 3191 args->cookie = cmd->arg1; 3192 retval = IP_FW_DUMMYNET; 3193 goto done; 3194 3195 case O_DIVERT: 3196 case O_TEE: { 3197 struct divert_tag *dt; 3198 3199 if (args->eh) /* not on layer 2 */ 3200 break; 3201 mtag = m_tag_get(PACKET_TAG_DIVERT, 3202 sizeof(struct divert_tag), 3203 M_NOWAIT); 3204 if (mtag == NULL) { 3205 /* XXX statistic */ 3206 /* drop packet */ 3207 IPFW_RUNLOCK(chain); 3208 return (IP_FW_DENY); 3209 } 3210 dt = (struct divert_tag *)(mtag+1); 3211 dt->cookie = f->rulenum; 3212 if (cmd->arg1 == IP_FW_TABLEARG) 3213 dt->info = tablearg; 3214 else 3215 dt->info = cmd->arg1; 3216 m_tag_prepend(m, mtag); 3217 retval = (cmd->opcode == O_DIVERT) ? 3218 IP_FW_DIVERT : IP_FW_TEE; 3219 goto done; 3220 } 3221 case O_COUNT: 3222 case O_SKIPTO: 3223 f->pcnt++; /* update stats */ 3224 f->bcnt += pktlen; 3225 f->timestamp = time_uptime; 3226 if (cmd->opcode == O_COUNT) 3227 goto next_rule; 3228 /* handle skipto */ 3229 if (f->next_rule == NULL) 3230 lookup_next_rule(f); 3231 f = f->next_rule; 3232 goto again; 3233 3234 case O_REJECT: 3235 /* 3236 * Drop the packet and send a reject notice 3237 * if the packet is not ICMP (or is an ICMP 3238 * query), and it is not multicast/broadcast. 3239 */ 3240 if (hlen > 0 && is_ipv4 && offset == 0 && 3241 (proto != IPPROTO_ICMP || 3242 is_icmp_query(ICMP(ulp))) && 3243 !(m->m_flags & (M_BCAST|M_MCAST)) && 3244 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3245 send_reject(args, cmd->arg1, ip_len, ip); 3246 m = args->m; 3247 } 3248 /* FALLTHROUGH */ 3249#ifdef INET6 3250 case O_UNREACH6: 3251 if (hlen > 0 && is_ipv6 && 3252 ((offset & IP6F_OFF_MASK) == 0) && 3253 (proto != IPPROTO_ICMPV6 || 3254 (is_icmp6_query(args->f_id.flags) == 1)) && 3255 !(m->m_flags & (M_BCAST|M_MCAST)) && 3256 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3257 send_reject6( 3258 args, cmd->arg1, hlen, 3259 (struct ip6_hdr *)ip); 3260 m = args->m; 3261 } 3262 /* FALLTHROUGH */ 3263#endif 3264 case O_DENY: 3265 retval = IP_FW_DENY; 3266 goto done; 3267 3268 case O_FORWARD_IP: { 3269 struct sockaddr_in *sa; 3270 sa = &(((ipfw_insn_sa *)cmd)->sa); 3271 if (args->eh) /* not valid on layer2 pkts */ 3272 break; 3273 if (!q || dyn_dir == MATCH_FORWARD) { 3274 if (sa->sin_addr.s_addr == INADDR_ANY) { 3275 bcopy(sa, &args->hopstore, 3276 sizeof(*sa)); 3277 args->hopstore.sin_addr.s_addr = 3278 htonl(tablearg); 3279 args->next_hop = 3280 &args->hopstore; 3281 } else { 3282 args->next_hop = sa; 3283 } 3284 } 3285 retval = IP_FW_PASS; 3286 } 3287 goto done; 3288 3289 case O_NETGRAPH: 3290 case O_NGTEE: 3291 args->rule = f; /* report matching rule */ 3292 if (cmd->arg1 == IP_FW_TABLEARG) 3293 args->cookie = tablearg; 3294 else 3295 args->cookie = cmd->arg1; 3296 retval = (cmd->opcode == O_NETGRAPH) ? 3297 IP_FW_NETGRAPH : IP_FW_NGTEE; 3298 goto done; 3299 3300 case O_SETFIB: 3301 f->pcnt++; /* update stats */ 3302 f->bcnt += pktlen; 3303 f->timestamp = time_uptime; 3304 M_SETFIB(m, cmd->arg1); 3305 args->f_id.fib = cmd->arg1; 3306 goto next_rule; 3307 3308 case O_NAT: { 3309 struct cfg_nat *t; 3310 int nat_id; 3311 3312 if (IPFW_NAT_LOADED) { 3313 args->rule = f; /* Report matching rule. */ 3314 t = ((ipfw_insn_nat *)cmd)->nat; 3315 if (t == NULL) { 3316 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 3317 tablearg : cmd->arg1; 3318 LOOKUP_NAT(layer3_chain, nat_id, t); 3319 if (t == NULL) { 3320 retval = IP_FW_DENY; 3321 goto done; 3322 } 3323 if (cmd->arg1 != IP_FW_TABLEARG) 3324 ((ipfw_insn_nat *)cmd)->nat = t; 3325 } 3326 retval = ipfw_nat_ptr(args, t, m); 3327 } else 3328 retval = IP_FW_DENY; 3329 goto done; 3330 } 3331 3332 default: 3333 panic("-- unknown opcode %d\n", cmd->opcode); 3334 } /* end of switch() on opcodes */ 3335 3336 if (cmd->len & F_NOT) 3337 match = !match; 3338 3339 if (match) { 3340 if (cmd->len & F_OR) 3341 skip_or = 1; 3342 } else { 3343 if (!(cmd->len & F_OR)) /* not an OR block, */ 3344 break; /* try next rule */ 3345 } 3346 3347 } /* end of inner for, scan opcodes */ 3348 3349next_rule:; /* try next rule */ 3350 3351 } /* end of outer for, scan rules */ 3352 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3353 IPFW_RUNLOCK(chain); 3354 return (IP_FW_DENY); 3355 3356done: 3357 /* Update statistics */ 3358 f->pcnt++; 3359 f->bcnt += pktlen; 3360 f->timestamp = time_uptime; 3361 IPFW_RUNLOCK(chain); 3362 return (retval); 3363 3364pullup_failed: 3365 if (fw_verbose) 3366 printf("ipfw: pullup failed\n"); 3367 return (IP_FW_DENY); 3368} 3369 3370/* 3371 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3372 * These will be reconstructed on the fly as packets are matched. 3373 */ 3374static void 3375flush_rule_ptrs(struct ip_fw_chain *chain) 3376{ 3377 struct ip_fw *rule; 3378 3379 IPFW_WLOCK_ASSERT(chain); 3380 3381 for (rule = chain->rules; rule; rule = rule->next) 3382 rule->next_rule = NULL; 3383} 3384 3385/* 3386 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3387 * possibly create a rule number and add the rule to the list. 3388 * Update the rule_number in the input struct so the caller knows it as well. 3389 */ 3390static int 3391add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3392{ 3393 struct ip_fw *rule, *f, *prev; 3394 int l = RULESIZE(input_rule); 3395 3396 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3397 return (EINVAL); 3398 3399 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3400 if (rule == NULL) 3401 return (ENOSPC); 3402 3403 bcopy(input_rule, rule, l); 3404 3405 rule->next = NULL; 3406 rule->next_rule = NULL; 3407 3408 rule->pcnt = 0; 3409 rule->bcnt = 0; 3410 rule->timestamp = 0; 3411 3412 IPFW_WLOCK(chain); 3413 3414 if (chain->rules == NULL) { /* default rule */ 3415 chain->rules = rule; 3416 goto done; 3417 } 3418 3419 /* 3420 * If rulenum is 0, find highest numbered rule before the 3421 * default rule, and add autoinc_step 3422 */ 3423 if (autoinc_step < 1) 3424 autoinc_step = 1; 3425 else if (autoinc_step > 1000) 3426 autoinc_step = 1000; 3427 if (rule->rulenum == 0) { 3428 /* 3429 * locate the highest numbered rule before default 3430 */ 3431 for (f = chain->rules; f; f = f->next) { 3432 if (f->rulenum == IPFW_DEFAULT_RULE) 3433 break; 3434 rule->rulenum = f->rulenum; 3435 } 3436 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3437 rule->rulenum += autoinc_step; 3438 input_rule->rulenum = rule->rulenum; 3439 } 3440 3441 /* 3442 * Now insert the new rule in the right place in the sorted list. 3443 */ 3444 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3445 if (f->rulenum > rule->rulenum) { /* found the location */ 3446 if (prev) { 3447 rule->next = f; 3448 prev->next = rule; 3449 } else { /* head insert */ 3450 rule->next = chain->rules; 3451 chain->rules = rule; 3452 } 3453 break; 3454 } 3455 } 3456 flush_rule_ptrs(chain); 3457done: 3458 static_count++; 3459 static_len += l; 3460 IPFW_WUNLOCK(chain); 3461 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3462 rule->rulenum, static_count);) 3463 return (0); 3464} 3465 3466/** 3467 * Remove a static rule (including derived * dynamic rules) 3468 * and place it on the ``reap list'' for later reclamation. 3469 * The caller is in charge of clearing rule pointers to avoid 3470 * dangling pointers. 3471 * @return a pointer to the next entry. 3472 * Arguments are not checked, so they better be correct. 3473 */ 3474static struct ip_fw * 3475remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 3476 struct ip_fw *prev) 3477{ 3478 struct ip_fw *n; 3479 int l = RULESIZE(rule); 3480 3481 IPFW_WLOCK_ASSERT(chain); 3482 3483 n = rule->next; 3484 IPFW_DYN_LOCK(); 3485 remove_dyn_rule(rule, NULL /* force removal */); 3486 IPFW_DYN_UNLOCK(); 3487 if (prev == NULL) 3488 chain->rules = n; 3489 else 3490 prev->next = n; 3491 static_count--; 3492 static_len -= l; 3493 3494 rule->next = chain->reap; 3495 chain->reap = rule; 3496 3497 return n; 3498} 3499 3500/** 3501 * Reclaim storage associated with a list of rules. This is 3502 * typically the list created using remove_rule. 3503 */ 3504static void 3505reap_rules(struct ip_fw *head) 3506{ 3507 struct ip_fw *rule; 3508 3509 while ((rule = head) != NULL) { 3510 head = head->next; 3511 if (DUMMYNET_LOADED) 3512 ip_dn_ruledel_ptr(rule); 3513 free(rule, M_IPFW); 3514 } 3515} 3516 3517/* 3518 * Remove all rules from a chain (except rules in set RESVD_SET 3519 * unless kill_default = 1). The caller is responsible for 3520 * reclaiming storage for the rules left in chain->reap. 3521 */ 3522static void 3523free_chain(struct ip_fw_chain *chain, int kill_default) 3524{ 3525 struct ip_fw *prev, *rule; 3526 3527 IPFW_WLOCK_ASSERT(chain); 3528 3529 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3530 for (prev = NULL, rule = chain->rules; rule ; ) 3531 if (kill_default || rule->set != RESVD_SET) 3532 rule = remove_rule(chain, rule, prev); 3533 else { 3534 prev = rule; 3535 rule = rule->next; 3536 } 3537} 3538 3539/** 3540 * Remove all rules with given number, and also do set manipulation. 3541 * Assumes chain != NULL && *chain != NULL. 3542 * 3543 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3544 * the next 8 bits are the new set, the top 8 bits are the command: 3545 * 3546 * 0 delete rules with given number 3547 * 1 delete rules with given set number 3548 * 2 move rules with given number to new set 3549 * 3 move rules with given set number to new set 3550 * 4 swap sets with given numbers 3551 * 5 delete rules with given number and with given set number 3552 */ 3553static int 3554del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3555{ 3556 struct ip_fw *prev = NULL, *rule; 3557 u_int16_t rulenum; /* rule or old_set */ 3558 u_int8_t cmd, new_set; 3559 3560 rulenum = arg & 0xffff; 3561 cmd = (arg >> 24) & 0xff; 3562 new_set = (arg >> 16) & 0xff; 3563 3564 if (cmd > 5 || new_set > RESVD_SET) 3565 return EINVAL; 3566 if (cmd == 0 || cmd == 2 || cmd == 5) { 3567 if (rulenum >= IPFW_DEFAULT_RULE) 3568 return EINVAL; 3569 } else { 3570 if (rulenum > RESVD_SET) /* old_set */ 3571 return EINVAL; 3572 } 3573 3574 IPFW_WLOCK(chain); 3575 rule = chain->rules; 3576 chain->reap = NULL; 3577 switch (cmd) { 3578 case 0: /* delete rules with given number */ 3579 /* 3580 * locate first rule to delete 3581 */ 3582 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3583 ; 3584 if (rule->rulenum != rulenum) { 3585 IPFW_WUNLOCK(chain); 3586 return EINVAL; 3587 } 3588 3589 /* 3590 * flush pointers outside the loop, then delete all matching 3591 * rules. prev remains the same throughout the cycle. 3592 */ 3593 flush_rule_ptrs(chain); 3594 while (rule->rulenum == rulenum) 3595 rule = remove_rule(chain, rule, prev); 3596 break; 3597 3598 case 1: /* delete all rules with given set number */ 3599 flush_rule_ptrs(chain); 3600 rule = chain->rules; 3601 while (rule->rulenum < IPFW_DEFAULT_RULE) 3602 if (rule->set == rulenum) 3603 rule = remove_rule(chain, rule, prev); 3604 else { 3605 prev = rule; 3606 rule = rule->next; 3607 } 3608 break; 3609 3610 case 2: /* move rules with given number to new set */ 3611 rule = chain->rules; 3612 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3613 if (rule->rulenum == rulenum) 3614 rule->set = new_set; 3615 break; 3616 3617 case 3: /* move rules with given set number to new set */ 3618 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3619 if (rule->set == rulenum) 3620 rule->set = new_set; 3621 break; 3622 3623 case 4: /* swap two sets */ 3624 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3625 if (rule->set == rulenum) 3626 rule->set = new_set; 3627 else if (rule->set == new_set) 3628 rule->set = rulenum; 3629 break; 3630 case 5: /* delete rules with given number and with given set number. 3631 * rulenum - given rule number; 3632 * new_set - given set number. 3633 */ 3634 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3635 ; 3636 if (rule->rulenum != rulenum) { 3637 IPFW_WUNLOCK(chain); 3638 return (EINVAL); 3639 } 3640 flush_rule_ptrs(chain); 3641 while (rule->rulenum == rulenum) { 3642 if (rule->set == new_set) 3643 rule = remove_rule(chain, rule, prev); 3644 else { 3645 prev = rule; 3646 rule = rule->next; 3647 } 3648 } 3649 } 3650 /* 3651 * Look for rules to reclaim. We grab the list before 3652 * releasing the lock then reclaim them w/o the lock to 3653 * avoid a LOR with dummynet. 3654 */ 3655 rule = chain->reap; 3656 chain->reap = NULL; 3657 IPFW_WUNLOCK(chain); 3658 if (rule) 3659 reap_rules(rule); 3660 return 0; 3661} 3662 3663/* 3664 * Clear counters for a specific rule. 3665 * The enclosing "table" is assumed locked. 3666 */ 3667static void 3668clear_counters(struct ip_fw *rule, int log_only) 3669{ 3670 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3671 3672 if (log_only == 0) { 3673 rule->bcnt = rule->pcnt = 0; 3674 rule->timestamp = 0; 3675 } 3676 if (l->o.opcode == O_LOG) 3677 l->log_left = l->max_log; 3678} 3679 3680/** 3681 * Reset some or all counters on firewall rules. 3682 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 3683 * the next 8 bits are the set number, the top 8 bits are the command: 3684 * 0 work with rules from all set's; 3685 * 1 work with rules only from specified set. 3686 * Specified rule number is zero if we want to clear all entries. 3687 * log_only is 1 if we only want to reset logs, zero otherwise. 3688 */ 3689static int 3690zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 3691{ 3692 struct ip_fw *rule; 3693 char *msg; 3694 3695 uint16_t rulenum = arg & 0xffff; 3696 uint8_t set = (arg >> 16) & 0xff; 3697 uint8_t cmd = (arg >> 24) & 0xff; 3698 3699 if (cmd > 1) 3700 return (EINVAL); 3701 if (cmd == 1 && set > RESVD_SET) 3702 return (EINVAL); 3703 3704 IPFW_WLOCK(chain); 3705 if (rulenum == 0) { 3706 norule_counter = 0; 3707 for (rule = chain->rules; rule; rule = rule->next) { 3708 /* Skip rules from another set. */ 3709 if (cmd == 1 && rule->set != set) 3710 continue; 3711 clear_counters(rule, log_only); 3712 } 3713 msg = log_only ? "ipfw: All logging counts reset.\n" : 3714 "ipfw: Accounting cleared.\n"; 3715 } else { 3716 int cleared = 0; 3717 /* 3718 * We can have multiple rules with the same number, so we 3719 * need to clear them all. 3720 */ 3721 for (rule = chain->rules; rule; rule = rule->next) 3722 if (rule->rulenum == rulenum) { 3723 while (rule && rule->rulenum == rulenum) { 3724 if (cmd == 0 || rule->set == set) 3725 clear_counters(rule, log_only); 3726 rule = rule->next; 3727 } 3728 cleared = 1; 3729 break; 3730 } 3731 if (!cleared) { /* we did not find any matching rules */ 3732 IPFW_WUNLOCK(chain); 3733 return (EINVAL); 3734 } 3735 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3736 "ipfw: Entry %d cleared.\n"; 3737 } 3738 IPFW_WUNLOCK(chain); 3739 3740 if (fw_verbose) 3741 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3742 return (0); 3743} 3744 3745/* 3746 * Check validity of the structure before insert. 3747 * Fortunately rules are simple, so this mostly need to check rule sizes. 3748 */ 3749static int 3750check_ipfw_struct(struct ip_fw *rule, int size) 3751{ 3752 int l, cmdlen = 0; 3753 int have_action=0; 3754 ipfw_insn *cmd; 3755 3756 if (size < sizeof(*rule)) { 3757 printf("ipfw: rule too short\n"); 3758 return (EINVAL); 3759 } 3760 /* first, check for valid size */ 3761 l = RULESIZE(rule); 3762 if (l != size) { 3763 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3764 return (EINVAL); 3765 } 3766 if (rule->act_ofs >= rule->cmd_len) { 3767 printf("ipfw: bogus action offset (%u > %u)\n", 3768 rule->act_ofs, rule->cmd_len - 1); 3769 return (EINVAL); 3770 } 3771 /* 3772 * Now go for the individual checks. Very simple ones, basically only 3773 * instruction sizes. 3774 */ 3775 for (l = rule->cmd_len, cmd = rule->cmd ; 3776 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3777 cmdlen = F_LEN(cmd); 3778 if (cmdlen > l) { 3779 printf("ipfw: opcode %d size truncated\n", 3780 cmd->opcode); 3781 return EINVAL; 3782 } 3783 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3784 switch (cmd->opcode) { 3785 case O_PROBE_STATE: 3786 case O_KEEP_STATE: 3787 case O_PROTO: 3788 case O_IP_SRC_ME: 3789 case O_IP_DST_ME: 3790 case O_LAYER2: 3791 case O_IN: 3792 case O_FRAG: 3793 case O_DIVERTED: 3794 case O_IPOPT: 3795 case O_IPTOS: 3796 case O_IPPRECEDENCE: 3797 case O_IPVER: 3798 case O_TCPWIN: 3799 case O_TCPFLAGS: 3800 case O_TCPOPTS: 3801 case O_ESTAB: 3802 case O_VERREVPATH: 3803 case O_VERSRCREACH: 3804 case O_ANTISPOOF: 3805 case O_IPSEC: 3806#ifdef INET6 3807 case O_IP6_SRC_ME: 3808 case O_IP6_DST_ME: 3809 case O_EXT_HDR: 3810 case O_IP6: 3811#endif 3812 case O_IP4: 3813 case O_TAG: 3814 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3815 goto bad_size; 3816 break; 3817 3818 case O_FIB: 3819 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3820 goto bad_size; 3821 if (cmd->arg1 >= rt_numfibs) { 3822 printf("ipfw: invalid fib number %d\n", 3823 cmd->arg1); 3824 return EINVAL; 3825 } 3826 break; 3827 3828 case O_SETFIB: 3829 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3830 goto bad_size; 3831 if (cmd->arg1 >= rt_numfibs) { 3832 printf("ipfw: invalid fib number %d\n", 3833 cmd->arg1); 3834 return EINVAL; 3835 } 3836 goto check_action; 3837 3838 case O_UID: 3839 case O_GID: 3840 case O_JAIL: 3841 case O_IP_SRC: 3842 case O_IP_DST: 3843 case O_TCPSEQ: 3844 case O_TCPACK: 3845 case O_PROB: 3846 case O_ICMPTYPE: 3847 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3848 goto bad_size; 3849 break; 3850 3851 case O_LIMIT: 3852 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3853 goto bad_size; 3854 break; 3855 3856 case O_LOG: 3857 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3858 goto bad_size; 3859 3860 ((ipfw_insn_log *)cmd)->log_left = 3861 ((ipfw_insn_log *)cmd)->max_log; 3862 3863 break; 3864 3865 case O_IP_SRC_MASK: 3866 case O_IP_DST_MASK: 3867 /* only odd command lengths */ 3868 if ( !(cmdlen & 1) || cmdlen > 31) 3869 goto bad_size; 3870 break; 3871 3872 case O_IP_SRC_SET: 3873 case O_IP_DST_SET: 3874 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3875 printf("ipfw: invalid set size %d\n", 3876 cmd->arg1); 3877 return EINVAL; 3878 } 3879 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3880 (cmd->arg1+31)/32 ) 3881 goto bad_size; 3882 break; 3883 3884 case O_IP_SRC_LOOKUP: 3885 case O_IP_DST_LOOKUP: 3886 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3887 printf("ipfw: invalid table number %d\n", 3888 cmd->arg1); 3889 return (EINVAL); 3890 } 3891 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3892 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3893 goto bad_size; 3894 break; 3895 3896 case O_MACADDR2: 3897 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3898 goto bad_size; 3899 break; 3900 3901 case O_NOP: 3902 case O_IPID: 3903 case O_IPTTL: 3904 case O_IPLEN: 3905 case O_TCPDATALEN: 3906 case O_TAGGED: 3907 if (cmdlen < 1 || cmdlen > 31) 3908 goto bad_size; 3909 break; 3910 3911 case O_MAC_TYPE: 3912 case O_IP_SRCPORT: 3913 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3914 if (cmdlen < 2 || cmdlen > 31) 3915 goto bad_size; 3916 break; 3917 3918 case O_RECV: 3919 case O_XMIT: 3920 case O_VIA: 3921 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3922 goto bad_size; 3923 break; 3924 3925 case O_ALTQ: 3926 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3927 goto bad_size; 3928 break; 3929 3930 case O_PIPE: 3931 case O_QUEUE: 3932 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3933 goto bad_size; 3934 goto check_action; 3935 3936 case O_FORWARD_IP: 3937#ifdef IPFIREWALL_FORWARD 3938 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3939 goto bad_size; 3940 goto check_action; 3941#else 3942 return EINVAL; 3943#endif 3944 3945 case O_DIVERT: 3946 case O_TEE: 3947 if (ip_divert_ptr == NULL) 3948 return EINVAL; 3949 else 3950 goto check_size; 3951 case O_NETGRAPH: 3952 case O_NGTEE: 3953 if (!NG_IPFW_LOADED) 3954 return EINVAL; 3955 else 3956 goto check_size; 3957 case O_NAT: 3958 if (!IPFW_NAT_LOADED) 3959 return EINVAL; 3960 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 3961 goto bad_size; 3962 goto check_action; 3963 case O_FORWARD_MAC: /* XXX not implemented yet */ 3964 case O_CHECK_STATE: 3965 case O_COUNT: 3966 case O_ACCEPT: 3967 case O_DENY: 3968 case O_REJECT: 3969#ifdef INET6 3970 case O_UNREACH6: 3971#endif 3972 case O_SKIPTO: 3973check_size: 3974 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3975 goto bad_size; 3976check_action: 3977 if (have_action) { 3978 printf("ipfw: opcode %d, multiple actions" 3979 " not allowed\n", 3980 cmd->opcode); 3981 return EINVAL; 3982 } 3983 have_action = 1; 3984 if (l != cmdlen) { 3985 printf("ipfw: opcode %d, action must be" 3986 " last opcode\n", 3987 cmd->opcode); 3988 return EINVAL; 3989 } 3990 break; 3991#ifdef INET6 3992 case O_IP6_SRC: 3993 case O_IP6_DST: 3994 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3995 F_INSN_SIZE(ipfw_insn)) 3996 goto bad_size; 3997 break; 3998 3999 case O_FLOW6ID: 4000 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4001 ((ipfw_insn_u32 *)cmd)->o.arg1) 4002 goto bad_size; 4003 break; 4004 4005 case O_IP6_SRC_MASK: 4006 case O_IP6_DST_MASK: 4007 if ( !(cmdlen & 1) || cmdlen > 127) 4008 goto bad_size; 4009 break; 4010 case O_ICMP6TYPE: 4011 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4012 goto bad_size; 4013 break; 4014#endif 4015 4016 default: 4017 switch (cmd->opcode) { 4018#ifndef INET6 4019 case O_IP6_SRC_ME: 4020 case O_IP6_DST_ME: 4021 case O_EXT_HDR: 4022 case O_IP6: 4023 case O_UNREACH6: 4024 case O_IP6_SRC: 4025 case O_IP6_DST: 4026 case O_FLOW6ID: 4027 case O_IP6_SRC_MASK: 4028 case O_IP6_DST_MASK: 4029 case O_ICMP6TYPE: 4030 printf("ipfw: no IPv6 support in kernel\n"); 4031 return EPROTONOSUPPORT; 4032#endif 4033 default: 4034 printf("ipfw: opcode %d, unknown opcode\n", 4035 cmd->opcode); 4036 return EINVAL; 4037 } 4038 } 4039 } 4040 if (have_action == 0) { 4041 printf("ipfw: missing action\n"); 4042 return EINVAL; 4043 } 4044 return 0; 4045 4046bad_size: 4047 printf("ipfw: opcode %d size %d wrong\n", 4048 cmd->opcode, cmdlen); 4049 return EINVAL; 4050} 4051 4052/* 4053 * Copy the static and dynamic rules to the supplied buffer 4054 * and return the amount of space actually used. 4055 */ 4056static size_t 4057ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4058{ 4059 char *bp = buf; 4060 char *ep = bp + space; 4061 struct ip_fw *rule; 4062 int i; 4063 time_t boot_seconds; 4064 4065 boot_seconds = boottime.tv_sec; 4066 /* XXX this can take a long time and locking will block packet flow */ 4067 IPFW_RLOCK(chain); 4068 for (rule = chain->rules; rule ; rule = rule->next) { 4069 /* 4070 * Verify the entry fits in the buffer in case the 4071 * rules changed between calculating buffer space and 4072 * now. This would be better done using a generation 4073 * number but should suffice for now. 4074 */ 4075 i = RULESIZE(rule); 4076 if (bp + i <= ep) { 4077 bcopy(rule, bp, i); 4078 /* 4079 * XXX HACK. Store the disable mask in the "next" pointer 4080 * in a wild attempt to keep the ABI the same. 4081 * Why do we do this on EVERY rule? 4082 */ 4083 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 4084 sizeof(set_disable)); 4085 if (((struct ip_fw *)bp)->timestamp) 4086 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4087 bp += i; 4088 } 4089 } 4090 IPFW_RUNLOCK(chain); 4091 if (ipfw_dyn_v) { 4092 ipfw_dyn_rule *p, *last = NULL; 4093 4094 IPFW_DYN_LOCK(); 4095 for (i = 0 ; i < curr_dyn_buckets; i++) 4096 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4097 if (bp + sizeof *p <= ep) { 4098 ipfw_dyn_rule *dst = 4099 (ipfw_dyn_rule *)bp; 4100 bcopy(p, dst, sizeof *p); 4101 bcopy(&(p->rule->rulenum), &(dst->rule), 4102 sizeof(p->rule->rulenum)); 4103 /* 4104 * store set number into high word of 4105 * dst->rule pointer. 4106 */ 4107 bcopy(&(p->rule->set), 4108 (char *)&dst->rule + 4109 sizeof(p->rule->rulenum), 4110 sizeof(p->rule->set)); 4111 /* 4112 * store a non-null value in "next". 4113 * The userland code will interpret a 4114 * NULL here as a marker 4115 * for the last dynamic rule. 4116 */ 4117 bcopy(&dst, &dst->next, sizeof(dst)); 4118 last = dst; 4119 dst->expire = 4120 TIME_LEQ(dst->expire, time_uptime) ? 4121 0 : dst->expire - time_uptime ; 4122 bp += sizeof(ipfw_dyn_rule); 4123 } 4124 } 4125 IPFW_DYN_UNLOCK(); 4126 if (last != NULL) /* mark last dynamic rule */ 4127 bzero(&last->next, sizeof(last)); 4128 } 4129 return (bp - (char *)buf); 4130} 4131 4132 4133/** 4134 * {set|get}sockopt parser. 4135 */ 4136static int 4137ipfw_ctl(struct sockopt *sopt) 4138{ 4139#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4140 int error; 4141 size_t size; 4142 struct ip_fw *buf, *rule; 4143 u_int32_t rulenum[2]; 4144 4145 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4146 if (error) 4147 return (error); 4148 4149 /* 4150 * Disallow modifications in really-really secure mode, but still allow 4151 * the logging counters to be reset. 4152 */ 4153 if (sopt->sopt_name == IP_FW_ADD || 4154 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4155 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4156 if (error) 4157 return (error); 4158 } 4159 4160 error = 0; 4161 4162 switch (sopt->sopt_name) { 4163 case IP_FW_GET: 4164 /* 4165 * pass up a copy of the current rules. Static rules 4166 * come first (the last of which has number IPFW_DEFAULT_RULE), 4167 * followed by a possibly empty list of dynamic rule. 4168 * The last dynamic rule has NULL in the "next" field. 4169 * 4170 * Note that the calculated size is used to bound the 4171 * amount of data returned to the user. The rule set may 4172 * change between calculating the size and returning the 4173 * data in which case we'll just return what fits. 4174 */ 4175 size = static_len; /* size of static rules */ 4176 if (ipfw_dyn_v) /* add size of dyn.rules */ 4177 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4178 4179 /* 4180 * XXX todo: if the user passes a short length just to know 4181 * how much room is needed, do not bother filling up the 4182 * buffer, just jump to the sooptcopyout. 4183 */ 4184 buf = malloc(size, M_TEMP, M_WAITOK); 4185 error = sooptcopyout(sopt, buf, 4186 ipfw_getrules(&layer3_chain, buf, size)); 4187 free(buf, M_TEMP); 4188 break; 4189 4190 case IP_FW_FLUSH: 4191 /* 4192 * Normally we cannot release the lock on each iteration. 4193 * We could do it here only because we start from the head all 4194 * the times so there is no risk of missing some entries. 4195 * On the other hand, the risk is that we end up with 4196 * a very inconsistent ruleset, so better keep the lock 4197 * around the whole cycle. 4198 * 4199 * XXX this code can be improved by resetting the head of 4200 * the list to point to the default rule, and then freeing 4201 * the old list without the need for a lock. 4202 */ 4203 4204 IPFW_WLOCK(&layer3_chain); 4205 layer3_chain.reap = NULL; 4206 free_chain(&layer3_chain, 0 /* keep default rule */); 4207 rule = layer3_chain.reap; 4208 layer3_chain.reap = NULL; 4209 IPFW_WUNLOCK(&layer3_chain); 4210 if (rule != NULL) 4211 reap_rules(rule); 4212 break; 4213 4214 case IP_FW_ADD: 4215 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4216 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4217 sizeof(struct ip_fw) ); 4218 if (error == 0) 4219 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4220 if (error == 0) { 4221 error = add_rule(&layer3_chain, rule); 4222 size = RULESIZE(rule); 4223 if (!error && sopt->sopt_dir == SOPT_GET) 4224 error = sooptcopyout(sopt, rule, size); 4225 } 4226 free(rule, M_TEMP); 4227 break; 4228 4229 case IP_FW_DEL: 4230 /* 4231 * IP_FW_DEL is used for deleting single rules or sets, 4232 * and (ab)used to atomically manipulate sets. Argument size 4233 * is used to distinguish between the two: 4234 * sizeof(u_int32_t) 4235 * delete single rule or set of rules, 4236 * or reassign rules (or sets) to a different set. 4237 * 2*sizeof(u_int32_t) 4238 * atomic disable/enable sets. 4239 * first u_int32_t contains sets to be disabled, 4240 * second u_int32_t contains sets to be enabled. 4241 */ 4242 error = sooptcopyin(sopt, rulenum, 4243 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4244 if (error) 4245 break; 4246 size = sopt->sopt_valsize; 4247 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4248 error = del_entry(&layer3_chain, rulenum[0]); 4249 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4250 set_disable = 4251 (set_disable | rulenum[0]) & ~rulenum[1] & 4252 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4253 else 4254 error = EINVAL; 4255 break; 4256 4257 case IP_FW_ZERO: 4258 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 4259 rulenum[0] = 0; 4260 if (sopt->sopt_val != 0) { 4261 error = sooptcopyin(sopt, rulenum, 4262 sizeof(u_int32_t), sizeof(u_int32_t)); 4263 if (error) 4264 break; 4265 } 4266 error = zero_entry(&layer3_chain, rulenum[0], 4267 sopt->sopt_name == IP_FW_RESETLOG); 4268 break; 4269 4270 case IP_FW_TABLE_ADD: 4271 { 4272 ipfw_table_entry ent; 4273 4274 error = sooptcopyin(sopt, &ent, 4275 sizeof(ent), sizeof(ent)); 4276 if (error) 4277 break; 4278 error = add_table_entry(&layer3_chain, ent.tbl, 4279 ent.addr, ent.masklen, ent.value); 4280 } 4281 break; 4282 4283 case IP_FW_TABLE_DEL: 4284 { 4285 ipfw_table_entry ent; 4286 4287 error = sooptcopyin(sopt, &ent, 4288 sizeof(ent), sizeof(ent)); 4289 if (error) 4290 break; 4291 error = del_table_entry(&layer3_chain, ent.tbl, 4292 ent.addr, ent.masklen); 4293 } 4294 break; 4295 4296 case IP_FW_TABLE_FLUSH: 4297 { 4298 u_int16_t tbl; 4299 4300 error = sooptcopyin(sopt, &tbl, 4301 sizeof(tbl), sizeof(tbl)); 4302 if (error) 4303 break; 4304 IPFW_WLOCK(&layer3_chain); 4305 error = flush_table(&layer3_chain, tbl); 4306 IPFW_WUNLOCK(&layer3_chain); 4307 } 4308 break; 4309 4310 case IP_FW_TABLE_GETSIZE: 4311 { 4312 u_int32_t tbl, cnt; 4313 4314 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4315 sizeof(tbl)))) 4316 break; 4317 IPFW_RLOCK(&layer3_chain); 4318 error = count_table(&layer3_chain, tbl, &cnt); 4319 IPFW_RUNLOCK(&layer3_chain); 4320 if (error) 4321 break; 4322 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4323 } 4324 break; 4325 4326 case IP_FW_TABLE_LIST: 4327 { 4328 ipfw_table *tbl; 4329 4330 if (sopt->sopt_valsize < sizeof(*tbl)) { 4331 error = EINVAL; 4332 break; 4333 } 4334 size = sopt->sopt_valsize; 4335 tbl = malloc(size, M_TEMP, M_WAITOK); 4336 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4337 if (error) { 4338 free(tbl, M_TEMP); 4339 break; 4340 } 4341 tbl->size = (size - sizeof(*tbl)) / 4342 sizeof(ipfw_table_entry); 4343 IPFW_RLOCK(&layer3_chain); 4344 error = dump_table(&layer3_chain, tbl); 4345 IPFW_RUNLOCK(&layer3_chain); 4346 if (error) { 4347 free(tbl, M_TEMP); 4348 break; 4349 } 4350 error = sooptcopyout(sopt, tbl, size); 4351 free(tbl, M_TEMP); 4352 } 4353 break; 4354 4355 case IP_FW_NAT_CFG: 4356 { 4357 if (IPFW_NAT_LOADED) 4358 error = ipfw_nat_cfg_ptr(sopt); 4359 else { 4360 printf("IP_FW_NAT_CFG: ipfw_nat not present, please load it.\n"); 4361 error = EINVAL; 4362 } 4363 } 4364 break; 4365 4366 case IP_FW_NAT_DEL: 4367 { 4368 if (IPFW_NAT_LOADED) 4369 error = ipfw_nat_del_ptr(sopt); 4370 else { 4371 printf("IP_FW_NAT_DEL: ipfw_nat not present, please load it.\n"); 4372 printf("ipfw_nat not loaded: %d\n", sopt->sopt_name); 4373 error = EINVAL; 4374 } 4375 } 4376 break; 4377 4378 case IP_FW_NAT_GET_CONFIG: 4379 { 4380 if (IPFW_NAT_LOADED) 4381 error = ipfw_nat_get_cfg_ptr(sopt); 4382 else { 4383 printf("IP_FW_NAT_GET_CFG: ipfw_nat not present, please load it.\n"); 4384 error = EINVAL; 4385 } 4386 } 4387 break; 4388 4389 case IP_FW_NAT_GET_LOG: 4390 { 4391 if (IPFW_NAT_LOADED) 4392 error = ipfw_nat_get_log_ptr(sopt); 4393 else { 4394 printf("IP_FW_NAT_GET_LOG: ipfw_nat not present, please load it.\n"); 4395 error = EINVAL; 4396 } 4397 } 4398 break; 4399 4400 default: 4401 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4402 error = EINVAL; 4403 } 4404 4405 return (error); 4406#undef RULE_MAXSIZE 4407} 4408 4409/** 4410 * dummynet needs a reference to the default rule, because rules can be 4411 * deleted while packets hold a reference to them. When this happens, 4412 * dummynet changes the reference to the default rule (it could well be a 4413 * NULL pointer, but this way we do not need to check for the special 4414 * case, plus here he have info on the default behaviour). 4415 */ 4416struct ip_fw *ip_fw_default_rule; 4417 4418/* 4419 * This procedure is only used to handle keepalives. It is invoked 4420 * every dyn_keepalive_period 4421 */ 4422static void 4423ipfw_tick(void * __unused unused) 4424{ 4425 struct mbuf *m0, *m, *mnext, **mtailp; 4426 int i; 4427 ipfw_dyn_rule *q; 4428 4429 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4430 goto done; 4431 4432 /* 4433 * We make a chain of packets to go out here -- not deferring 4434 * until after we drop the IPFW dynamic rule lock would result 4435 * in a lock order reversal with the normal packet input -> ipfw 4436 * call stack. 4437 */ 4438 m0 = NULL; 4439 mtailp = &m0; 4440 IPFW_DYN_LOCK(); 4441 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4442 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4443 if (q->dyn_type == O_LIMIT_PARENT) 4444 continue; 4445 if (q->id.proto != IPPROTO_TCP) 4446 continue; 4447 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4448 continue; 4449 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4450 q->expire)) 4451 continue; /* too early */ 4452 if (TIME_LEQ(q->expire, time_uptime)) 4453 continue; /* too late, rule expired */ 4454 4455 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4456 q->ack_fwd, TH_SYN); 4457 if (*mtailp != NULL) 4458 mtailp = &(*mtailp)->m_nextpkt; 4459 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4460 q->ack_rev, 0); 4461 if (*mtailp != NULL) 4462 mtailp = &(*mtailp)->m_nextpkt; 4463 } 4464 } 4465 IPFW_DYN_UNLOCK(); 4466 for (m = mnext = m0; m != NULL; m = mnext) { 4467 mnext = m->m_nextpkt; 4468 m->m_nextpkt = NULL; 4469 ip_output(m, NULL, NULL, 0, NULL, NULL); 4470 } 4471done: 4472 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4473} 4474 4475int 4476ipfw_init(void) 4477{ 4478 struct ip_fw default_rule; 4479 int error; 4480 4481#ifdef INET6 4482 /* Setup IPv6 fw sysctl tree. */ 4483 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4484 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4485 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4486 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4487 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4488 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4489 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4490 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4491 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4492 &fw_deny_unknown_exthdrs, 0, 4493 "Deny packets with unknown IPv6 Extension Headers"); 4494#endif 4495 4496 layer3_chain.rules = NULL; 4497 IPFW_LOCK_INIT(&layer3_chain); 4498 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4499 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4500 UMA_ALIGN_PTR, 0); 4501 IPFW_DYN_LOCK_INIT(); 4502 callout_init(&ipfw_timeout, CALLOUT_MPSAFE); 4503 4504 bzero(&default_rule, sizeof default_rule); 4505 4506 default_rule.act_ofs = 0; 4507 default_rule.rulenum = IPFW_DEFAULT_RULE; 4508 default_rule.cmd_len = 1; 4509 default_rule.set = RESVD_SET; 4510 4511 default_rule.cmd[0].len = 1; 4512 default_rule.cmd[0].opcode = 4513#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4514 1 ? O_ACCEPT : 4515#endif 4516 O_DENY; 4517 4518 error = add_rule(&layer3_chain, &default_rule); 4519 if (error != 0) { 4520 printf("ipfw2: error %u initializing default rule " 4521 "(support disabled)\n", error); 4522 IPFW_DYN_LOCK_DESTROY(); 4523 IPFW_LOCK_DESTROY(&layer3_chain); 4524 uma_zdestroy(ipfw_dyn_rule_zone); 4525 return (error); 4526 } 4527 4528 ip_fw_default_rule = layer3_chain.rules; 4529 printf("ipfw2 " 4530#ifdef INET6 4531 "(+ipv6) " 4532#endif 4533 "initialized, divert %s, nat %s, " 4534 "rule-based forwarding " 4535#ifdef IPFIREWALL_FORWARD 4536 "enabled, " 4537#else 4538 "disabled, " 4539#endif 4540 "default to %s, logging ", 4541#ifdef IPDIVERT 4542 "enabled", 4543#else 4544 "loadable", 4545#endif 4546#ifdef IPFIREWALL_NAT 4547 "enabled", 4548#else 4549 "loadable", 4550#endif 4551 4552 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4553 4554#ifdef IPFIREWALL_VERBOSE 4555 fw_verbose = 1; 4556#endif 4557#ifdef IPFIREWALL_VERBOSE_LIMIT 4558 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4559#endif 4560 if (fw_verbose == 0) 4561 printf("disabled\n"); 4562 else if (verbose_limit == 0) 4563 printf("unlimited\n"); 4564 else 4565 printf("limited to %d packets/entry by default\n", 4566 verbose_limit); 4567 4568 error = init_tables(&layer3_chain); 4569 if (error) { 4570 IPFW_DYN_LOCK_DESTROY(); 4571 IPFW_LOCK_DESTROY(&layer3_chain); 4572 uma_zdestroy(ipfw_dyn_rule_zone); 4573 return (error); 4574 } 4575 ip_fw_ctl_ptr = ipfw_ctl; 4576 ip_fw_chk_ptr = ipfw_chk; 4577 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4578 LIST_INIT(&layer3_chain.nat); 4579 return (0); 4580} 4581 4582void 4583ipfw_destroy(void) 4584{ 4585 struct ip_fw *reap; 4586 4587 ip_fw_chk_ptr = NULL; 4588 ip_fw_ctl_ptr = NULL; 4589 callout_drain(&ipfw_timeout); 4590 IPFW_WLOCK(&layer3_chain); 4591 flush_tables(&layer3_chain); 4592 layer3_chain.reap = NULL; 4593 free_chain(&layer3_chain, 1 /* kill default rule */); 4594 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4595 IPFW_WUNLOCK(&layer3_chain); 4596 if (reap != NULL) 4597 reap_rules(reap); 4598 IPFW_DYN_LOCK_DESTROY(); 4599 uma_zdestroy(ipfw_dyn_rule_zone); 4600 if (ipfw_dyn_v != NULL) 4601 free(ipfw_dyn_v, M_IPFW); 4602 IPFW_LOCK_DESTROY(&layer3_chain); 4603 4604#ifdef INET6 4605 /* Free IPv6 fw sysctl tree. */ 4606 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4607#endif 4608 4609 printf("IP firewall unloaded\n"); 4610} 4611