ip_fw2.c revision 183606
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 183606 2008-10-04 15:06:34Z bz $"); 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 <sys/vimage.h> 68#include <net/if.h> 69#include <net/radix.h> 70#include <net/route.h> 71#include <net/pf_mtag.h> 72 73#define IPFW_INTERNAL /* Access to protected data structures in ip_fw.h. */ 74 75#include <netinet/in.h> 76#include <netinet/in_systm.h> 77#include <netinet/in_var.h> 78#include <netinet/in_pcb.h> 79#include <netinet/ip.h> 80#include <netinet/ip_var.h> 81#include <netinet/ip_icmp.h> 82#include <netinet/ip_fw.h> 83#include <netinet/ip_divert.h> 84#include <netinet/ip_dummynet.h> 85#include <netinet/ip_carp.h> 86#include <netinet/pim.h> 87#include <netinet/tcp.h> 88#include <netinet/tcp_timer.h> 89#include <netinet/tcp_var.h> 90#include <netinet/tcpip.h> 91#include <netinet/udp.h> 92#include <netinet/udp_var.h> 93#include <netinet/sctp.h> 94#include <netgraph/ng_ipfw.h> 95 96#include <altq/if_altq.h> 97 98#include <netinet/ip6.h> 99#include <netinet/icmp6.h> 100#ifdef INET6 101#include <netinet6/scope6_var.h> 102#endif 103 104#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 105 106#include <machine/in_cksum.h> /* XXX for in_cksum */ 107 108#include <security/mac/mac_framework.h> 109 110/* 111 * set_disable contains one bit per set value (0..31). 112 * If the bit is set, all rules with the corresponding set 113 * are disabled. Set RESVD_SET(31) is reserved for the default rule 114 * and rules that are not deleted by the flush command, 115 * and CANNOT be disabled. 116 * Rules in set RESVD_SET can only be deleted explicitly. 117 */ 118static u_int32_t set_disable; 119 120static int fw_verbose; 121static int verbose_limit; 122 123static struct callout ipfw_timeout; 124static uma_zone_t ipfw_dyn_rule_zone; 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_V_PROC(V_NET, vnet_ipfw, _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_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, autoinc_step, 170 CTLFLAG_RW, autoinc_step, 0, "Rule number autincrement step"); 171SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, one_pass, 172 CTLFLAG_RW | CTLFLAG_SECURE3, fw_one_pass, 0, 173 "Only do a single pass through ipfw when using dummynet(4)"); 174SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 175 fw_debug, 0, "Enable printing of debug ip_fw statements"); 176SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, verbose, 177 CTLFLAG_RW | CTLFLAG_SECURE3, 178 fw_verbose, 0, "Log matches to ipfw rules"); 179SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 180 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 181SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 182 NULL, IPFW_DEFAULT_RULE, "The default/max possible rule number."); 183SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 184 NULL, IPFW_TABLES_MAX, "The maximum number of tables."); 185 186/* 187 * Description of dynamic rules. 188 * 189 * Dynamic rules are stored in lists accessed through a hash table 190 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 191 * be modified through the sysctl variable dyn_buckets which is 192 * updated when the table becomes empty. 193 * 194 * XXX currently there is only one list, ipfw_dyn. 195 * 196 * When a packet is received, its address fields are first masked 197 * with the mask defined for the rule, then hashed, then matched 198 * against the entries in the corresponding list. 199 * Dynamic rules can be used for different purposes: 200 * + stateful rules; 201 * + enforcing limits on the number of sessions; 202 * + in-kernel NAT (not implemented yet) 203 * 204 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 205 * measured in seconds and depending on the flags. 206 * 207 * The total number of dynamic rules is stored in dyn_count. 208 * The max number of dynamic rules is dyn_max. When we reach 209 * the maximum number of rules we do not create anymore. This is 210 * done to avoid consuming too much memory, but also too much 211 * time when searching on each packet (ideally, we should try instead 212 * to put a limit on the length of the list on each bucket...). 213 * 214 * Each dynamic rule holds a pointer to the parent ipfw rule so 215 * we know what action to perform. Dynamic rules are removed when 216 * the parent rule is deleted. XXX we should make them survive. 217 * 218 * There are some limitations with dynamic rules -- we do not 219 * obey the 'randomized match', and we do not do multiple 220 * passes through the firewall. XXX check the latter!!! 221 */ 222static ipfw_dyn_rule **ipfw_dyn_v = NULL; 223static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 224static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 225 226static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 227#define IPFW_DYN_LOCK_INIT() \ 228 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 229#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 230#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 231#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 232#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 233 234/* 235 * Timeouts for various events in handing dynamic rules. 236 */ 237static u_int32_t dyn_ack_lifetime = 300; 238static u_int32_t dyn_syn_lifetime = 20; 239static u_int32_t dyn_fin_lifetime = 1; 240static u_int32_t dyn_rst_lifetime = 1; 241static u_int32_t dyn_udp_lifetime = 10; 242static u_int32_t dyn_short_lifetime = 5; 243 244/* 245 * Keepalives are sent if dyn_keepalive is set. They are sent every 246 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 247 * seconds of lifetime of a rule. 248 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 249 * than dyn_keepalive_period. 250 */ 251 252static u_int32_t dyn_keepalive_interval = 20; 253static u_int32_t dyn_keepalive_period = 5; 254static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 255 256static u_int32_t static_count; /* # of static rules */ 257static u_int32_t static_len; /* size in bytes of static rules */ 258static u_int32_t dyn_count; /* # of dynamic rules */ 259static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 260 261SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_buckets, 262 CTLFLAG_RW, dyn_buckets, 0, "Number of dyn. buckets"); 263SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, 264 CTLFLAG_RD, curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 265SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_count, 266 CTLFLAG_RD, dyn_count, 0, "Number of dyn. rules"); 267SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_max, 268 CTLFLAG_RW, dyn_max, 0, "Max number of dyn. rules"); 269SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, static_count, 270 CTLFLAG_RD, static_count, 0, "Number of static rules"); 271SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, 272 CTLFLAG_RW, dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 273SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, 274 CTLFLAG_RW, dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 275SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, 276 CTLFLAG_RW, dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 277SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, 278 CTLFLAG_RW, dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 279SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, 280 CTLFLAG_RW, dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 281SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, 282 CTLFLAG_RW, dyn_short_lifetime, 0, 283 "Lifetime of dyn. rules for other situations"); 284SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_keepalive, 285 CTLFLAG_RW, dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 286 287 288#ifdef INET6 289/* 290 * IPv6 specific variables 291 */ 292SYSCTL_DECL(_net_inet6_ip6); 293 294static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 295static struct sysctl_oid *ip6_fw_sysctl_tree; 296#endif /* INET6 */ 297#endif /* SYSCTL_NODE */ 298 299static int fw_deny_unknown_exthdrs = 1; 300 301 302/* 303 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 304 * Other macros just cast void * into the appropriate type 305 */ 306#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 307#define TCP(p) ((struct tcphdr *)(p)) 308#define SCTP(p) ((struct sctphdr *)(p)) 309#define UDP(p) ((struct udphdr *)(p)) 310#define ICMP(p) ((struct icmphdr *)(p)) 311#define ICMP6(p) ((struct icmp6_hdr *)(p)) 312 313static __inline int 314icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 315{ 316 int type = icmp->icmp_type; 317 318 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 319} 320 321#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 322 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 323 324static int 325is_icmp_query(struct icmphdr *icmp) 326{ 327 int type = icmp->icmp_type; 328 329 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 330} 331#undef TT 332 333/* 334 * The following checks use two arrays of 8 or 16 bits to store the 335 * bits that we want set or clear, respectively. They are in the 336 * low and high half of cmd->arg1 or cmd->d[0]. 337 * 338 * We scan options and store the bits we find set. We succeed if 339 * 340 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 341 * 342 * The code is sometimes optimized not to store additional variables. 343 */ 344 345static int 346flags_match(ipfw_insn *cmd, u_int8_t bits) 347{ 348 u_char want_clear; 349 bits = ~bits; 350 351 if ( ((cmd->arg1 & 0xff) & bits) != 0) 352 return 0; /* some bits we want set were clear */ 353 want_clear = (cmd->arg1 >> 8) & 0xff; 354 if ( (want_clear & bits) != want_clear) 355 return 0; /* some bits we want clear were set */ 356 return 1; 357} 358 359static int 360ipopts_match(struct ip *ip, ipfw_insn *cmd) 361{ 362 int optlen, bits = 0; 363 u_char *cp = (u_char *)(ip + 1); 364 int x = (ip->ip_hl << 2) - sizeof (struct ip); 365 366 for (; x > 0; x -= optlen, cp += optlen) { 367 int opt = cp[IPOPT_OPTVAL]; 368 369 if (opt == IPOPT_EOL) 370 break; 371 if (opt == IPOPT_NOP) 372 optlen = 1; 373 else { 374 optlen = cp[IPOPT_OLEN]; 375 if (optlen <= 0 || optlen > x) 376 return 0; /* invalid or truncated */ 377 } 378 switch (opt) { 379 380 default: 381 break; 382 383 case IPOPT_LSRR: 384 bits |= IP_FW_IPOPT_LSRR; 385 break; 386 387 case IPOPT_SSRR: 388 bits |= IP_FW_IPOPT_SSRR; 389 break; 390 391 case IPOPT_RR: 392 bits |= IP_FW_IPOPT_RR; 393 break; 394 395 case IPOPT_TS: 396 bits |= IP_FW_IPOPT_TS; 397 break; 398 } 399 } 400 return (flags_match(cmd, bits)); 401} 402 403static int 404tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 405{ 406 int optlen, bits = 0; 407 u_char *cp = (u_char *)(tcp + 1); 408 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 409 410 for (; x > 0; x -= optlen, cp += optlen) { 411 int opt = cp[0]; 412 if (opt == TCPOPT_EOL) 413 break; 414 if (opt == TCPOPT_NOP) 415 optlen = 1; 416 else { 417 optlen = cp[1]; 418 if (optlen <= 0) 419 break; 420 } 421 422 switch (opt) { 423 424 default: 425 break; 426 427 case TCPOPT_MAXSEG: 428 bits |= IP_FW_TCPOPT_MSS; 429 break; 430 431 case TCPOPT_WINDOW: 432 bits |= IP_FW_TCPOPT_WINDOW; 433 break; 434 435 case TCPOPT_SACK_PERMITTED: 436 case TCPOPT_SACK: 437 bits |= IP_FW_TCPOPT_SACK; 438 break; 439 440 case TCPOPT_TIMESTAMP: 441 bits |= IP_FW_TCPOPT_TS; 442 break; 443 444 } 445 } 446 return (flags_match(cmd, bits)); 447} 448 449static int 450iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 451{ 452 if (ifp == NULL) /* no iface with this packet, match fails */ 453 return 0; 454 /* Check by name or by IP address */ 455 if (cmd->name[0] != '\0') { /* match by name */ 456 /* Check name */ 457 if (cmd->p.glob) { 458 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 459 return(1); 460 } else { 461 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 462 return(1); 463 } 464 } else { 465 struct ifaddr *ia; 466 467 /* XXX lock? */ 468 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 469 if (ia->ifa_addr->sa_family != AF_INET) 470 continue; 471 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 472 (ia->ifa_addr))->sin_addr.s_addr) 473 return(1); /* match */ 474 } 475 } 476 return(0); /* no match, fail ... */ 477} 478 479/* 480 * The verify_path function checks if a route to the src exists and 481 * if it is reachable via ifp (when provided). 482 * 483 * The 'verrevpath' option checks that the interface that an IP packet 484 * arrives on is the same interface that traffic destined for the 485 * packet's source address would be routed out of. The 'versrcreach' 486 * option just checks that the source address is reachable via any route 487 * (except default) in the routing table. These two are a measure to block 488 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 489 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 490 * is purposely reminiscent of the Cisco IOS command, 491 * 492 * ip verify unicast reverse-path 493 * ip verify unicast source reachable-via any 494 * 495 * which implements the same functionality. But note that syntax is 496 * misleading. The check may be performed on all IP packets whether unicast, 497 * multicast, or broadcast. 498 */ 499static int 500verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 501{ 502 struct route ro; 503 struct sockaddr_in *dst; 504 505 bzero(&ro, sizeof(ro)); 506 507 dst = (struct sockaddr_in *)&(ro.ro_dst); 508 dst->sin_family = AF_INET; 509 dst->sin_len = sizeof(*dst); 510 dst->sin_addr = src; 511 in_rtalloc_ign(&ro, RTF_CLONING, fib); 512 513 if (ro.ro_rt == NULL) 514 return 0; 515 516 /* 517 * If ifp is provided, check for equality with rtentry. 518 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 519 * in order to pass packets injected back by if_simloop(): 520 * if useloopback == 1 routing entry (via lo0) for our own address 521 * may exist, so we need to handle routing assymetry. 522 */ 523 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 524 RTFREE(ro.ro_rt); 525 return 0; 526 } 527 528 /* if no ifp provided, check if rtentry is not default route */ 529 if (ifp == NULL && 530 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 531 RTFREE(ro.ro_rt); 532 return 0; 533 } 534 535 /* or if this is a blackhole/reject route */ 536 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 537 RTFREE(ro.ro_rt); 538 return 0; 539 } 540 541 /* found valid route */ 542 RTFREE(ro.ro_rt); 543 return 1; 544} 545 546#ifdef INET6 547/* 548 * ipv6 specific rules here... 549 */ 550static __inline int 551icmp6type_match (int type, ipfw_insn_u32 *cmd) 552{ 553 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 554} 555 556static int 557flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 558{ 559 int i; 560 for (i=0; i <= cmd->o.arg1; ++i ) 561 if (curr_flow == cmd->d[i] ) 562 return 1; 563 return 0; 564} 565 566/* support for IP6_*_ME opcodes */ 567static int 568search_ip6_addr_net (struct in6_addr * ip6_addr) 569{ 570 INIT_VNET_NET(curvnet); 571 struct ifnet *mdc; 572 struct ifaddr *mdc2; 573 struct in6_ifaddr *fdm; 574 struct in6_addr copia; 575 576 TAILQ_FOREACH(mdc, &V_ifnet, if_link) 577 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 578 if (mdc2->ifa_addr->sa_family == AF_INET6) { 579 fdm = (struct in6_ifaddr *)mdc2; 580 copia = fdm->ia_addr.sin6_addr; 581 /* need for leaving scope_id in the sock_addr */ 582 in6_clearscope(&copia); 583 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 584 return 1; 585 } 586 } 587 return 0; 588} 589 590static int 591verify_path6(struct in6_addr *src, struct ifnet *ifp) 592{ 593 struct route_in6 ro; 594 struct sockaddr_in6 *dst; 595 596 bzero(&ro, sizeof(ro)); 597 598 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 599 dst->sin6_family = AF_INET6; 600 dst->sin6_len = sizeof(*dst); 601 dst->sin6_addr = *src; 602 /* XXX MRT 0 for ipv6 at this time */ 603 rtalloc_ign((struct route *)&ro, RTF_CLONING); 604 605 if (ro.ro_rt == NULL) 606 return 0; 607 608 /* 609 * if ifp is provided, check for equality with rtentry 610 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 611 * to support the case of sending packets to an address of our own. 612 * (where the former interface is the first argument of if_simloop() 613 * (=ifp), the latter is lo0) 614 */ 615 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 616 RTFREE(ro.ro_rt); 617 return 0; 618 } 619 620 /* if no ifp provided, check if rtentry is not default route */ 621 if (ifp == NULL && 622 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 623 RTFREE(ro.ro_rt); 624 return 0; 625 } 626 627 /* or if this is a blackhole/reject route */ 628 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 629 RTFREE(ro.ro_rt); 630 return 0; 631 } 632 633 /* found valid route */ 634 RTFREE(ro.ro_rt); 635 return 1; 636 637} 638static __inline int 639hash_packet6(struct ipfw_flow_id *id) 640{ 641 u_int32_t i; 642 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 643 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 644 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 645 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 646 (id->dst_port) ^ (id->src_port); 647 return i; 648} 649 650static int 651is_icmp6_query(int icmp6_type) 652{ 653 if ((icmp6_type <= ICMP6_MAXTYPE) && 654 (icmp6_type == ICMP6_ECHO_REQUEST || 655 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 656 icmp6_type == ICMP6_WRUREQUEST || 657 icmp6_type == ICMP6_FQDN_QUERY || 658 icmp6_type == ICMP6_NI_QUERY)) 659 return (1); 660 661 return (0); 662} 663 664static void 665send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 666{ 667 struct mbuf *m; 668 669 m = args->m; 670 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 671 struct tcphdr *tcp; 672 tcp_seq ack, seq; 673 int flags; 674 struct { 675 struct ip6_hdr ip6; 676 struct tcphdr th; 677 } ti; 678 tcp = (struct tcphdr *)((char *)ip6 + hlen); 679 680 if ((tcp->th_flags & TH_RST) != 0) { 681 m_freem(m); 682 args->m = NULL; 683 return; 684 } 685 686 ti.ip6 = *ip6; 687 ti.th = *tcp; 688 ti.th.th_seq = ntohl(ti.th.th_seq); 689 ti.th.th_ack = ntohl(ti.th.th_ack); 690 ti.ip6.ip6_nxt = IPPROTO_TCP; 691 692 if (ti.th.th_flags & TH_ACK) { 693 ack = 0; 694 seq = ti.th.th_ack; 695 flags = TH_RST; 696 } else { 697 ack = ti.th.th_seq; 698 if ((m->m_flags & M_PKTHDR) != 0) { 699 /* 700 * total new data to ACK is: 701 * total packet length, 702 * minus the header length, 703 * minus the tcp header length. 704 */ 705 ack += m->m_pkthdr.len - hlen 706 - (ti.th.th_off << 2); 707 } else if (ip6->ip6_plen) { 708 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 709 hlen - (ti.th.th_off << 2); 710 } else { 711 m_freem(m); 712 return; 713 } 714 if (tcp->th_flags & TH_SYN) 715 ack++; 716 seq = 0; 717 flags = TH_RST|TH_ACK; 718 } 719 bcopy(&ti, ip6, sizeof(ti)); 720 /* 721 * m is only used to recycle the mbuf 722 * The data in it is never read so we don't need 723 * to correct the offsets or anything 724 */ 725 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 726 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 727#if 0 728 /* 729 * Unlike above, the mbufs need to line up with the ip6 hdr, 730 * as the contents are read. We need to m_adj() the 731 * needed amount. 732 * The mbuf will however be thrown away so we can adjust it. 733 * Remember we did an m_pullup on it already so we 734 * can make some assumptions about contiguousness. 735 */ 736 if (args->L3offset) 737 m_adj(m, args->L3offset); 738#endif 739 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 740 } else 741 m_freem(m); 742 743 args->m = NULL; 744} 745 746#endif /* INET6 */ 747 748static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 749 750#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 751#define SNP(buf) buf, sizeof(buf) 752 753/* 754 * We enter here when we have a rule with O_LOG. 755 * XXX this function alone takes about 2Kbytes of code! 756 */ 757static void 758ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 759 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 760 struct ip *ip) 761{ 762 INIT_VNET_IPFW(curvnet); 763 struct ether_header *eh = args->eh; 764 char *action; 765 int limit_reached = 0; 766 char action2[40], proto[128], fragment[32]; 767 768 fragment[0] = '\0'; 769 proto[0] = '\0'; 770 771 if (f == NULL) { /* bogus pkt */ 772 if (V_verbose_limit != 0 && V_norule_counter >= V_verbose_limit) 773 return; 774 V_norule_counter++; 775 if (V_norule_counter == V_verbose_limit) 776 limit_reached = V_verbose_limit; 777 action = "Refuse"; 778 } else { /* O_LOG is the first action, find the real one */ 779 ipfw_insn *cmd = ACTION_PTR(f); 780 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 781 782 if (l->max_log != 0 && l->log_left == 0) 783 return; 784 l->log_left--; 785 if (l->log_left == 0) 786 limit_reached = l->max_log; 787 cmd += F_LEN(cmd); /* point to first action */ 788 if (cmd->opcode == O_ALTQ) { 789 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 790 791 snprintf(SNPARGS(action2, 0), "Altq %d", 792 altq->qid); 793 cmd += F_LEN(cmd); 794 } 795 if (cmd->opcode == O_PROB) 796 cmd += F_LEN(cmd); 797 798 if (cmd->opcode == O_TAG) 799 cmd += F_LEN(cmd); 800 801 action = action2; 802 switch (cmd->opcode) { 803 case O_DENY: 804 action = "Deny"; 805 break; 806 807 case O_REJECT: 808 if (cmd->arg1==ICMP_REJECT_RST) 809 action = "Reset"; 810 else if (cmd->arg1==ICMP_UNREACH_HOST) 811 action = "Reject"; 812 else 813 snprintf(SNPARGS(action2, 0), "Unreach %d", 814 cmd->arg1); 815 break; 816 817 case O_UNREACH6: 818 if (cmd->arg1==ICMP6_UNREACH_RST) 819 action = "Reset"; 820 else 821 snprintf(SNPARGS(action2, 0), "Unreach %d", 822 cmd->arg1); 823 break; 824 825 case O_ACCEPT: 826 action = "Accept"; 827 break; 828 case O_COUNT: 829 action = "Count"; 830 break; 831 case O_DIVERT: 832 snprintf(SNPARGS(action2, 0), "Divert %d", 833 cmd->arg1); 834 break; 835 case O_TEE: 836 snprintf(SNPARGS(action2, 0), "Tee %d", 837 cmd->arg1); 838 break; 839 case O_SETFIB: 840 snprintf(SNPARGS(action2, 0), "SetFib %d", 841 cmd->arg1); 842 break; 843 case O_SKIPTO: 844 snprintf(SNPARGS(action2, 0), "SkipTo %d", 845 cmd->arg1); 846 break; 847 case O_PIPE: 848 snprintf(SNPARGS(action2, 0), "Pipe %d", 849 cmd->arg1); 850 break; 851 case O_QUEUE: 852 snprintf(SNPARGS(action2, 0), "Queue %d", 853 cmd->arg1); 854 break; 855 case O_FORWARD_IP: { 856 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 857 int len; 858 struct in_addr dummyaddr; 859 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 860 dummyaddr.s_addr = htonl(tablearg); 861 else 862 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 863 864 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 865 inet_ntoa(dummyaddr)); 866 867 if (sa->sa.sin_port) 868 snprintf(SNPARGS(action2, len), ":%d", 869 sa->sa.sin_port); 870 } 871 break; 872 case O_NETGRAPH: 873 snprintf(SNPARGS(action2, 0), "Netgraph %d", 874 cmd->arg1); 875 break; 876 case O_NGTEE: 877 snprintf(SNPARGS(action2, 0), "Ngtee %d", 878 cmd->arg1); 879 break; 880 case O_NAT: 881 action = "Nat"; 882 break; 883 default: 884 action = "UNKNOWN"; 885 break; 886 } 887 } 888 889 if (hlen == 0) { /* non-ip */ 890 snprintf(SNPARGS(proto, 0), "MAC"); 891 892 } else { 893 int len; 894 char src[48], dst[48]; 895 struct icmphdr *icmp; 896 struct tcphdr *tcp; 897 struct udphdr *udp; 898#ifdef INET6 899 struct ip6_hdr *ip6 = NULL; 900 struct icmp6_hdr *icmp6; 901#endif 902 src[0] = '\0'; 903 dst[0] = '\0'; 904#ifdef INET6 905 if (IS_IP6_FLOW_ID(&(args->f_id))) { 906 char ip6buf[INET6_ADDRSTRLEN]; 907 snprintf(src, sizeof(src), "[%s]", 908 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 909 snprintf(dst, sizeof(dst), "[%s]", 910 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 911 912 ip6 = (struct ip6_hdr *)ip; 913 tcp = (struct tcphdr *)(((char *)ip) + hlen); 914 udp = (struct udphdr *)(((char *)ip) + hlen); 915 } else 916#endif 917 { 918 tcp = L3HDR(struct tcphdr, ip); 919 udp = L3HDR(struct udphdr, ip); 920 921 inet_ntoa_r(ip->ip_src, src); 922 inet_ntoa_r(ip->ip_dst, dst); 923 } 924 925 switch (args->f_id.proto) { 926 case IPPROTO_TCP: 927 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 928 if (offset == 0) 929 snprintf(SNPARGS(proto, len), ":%d %s:%d", 930 ntohs(tcp->th_sport), 931 dst, 932 ntohs(tcp->th_dport)); 933 else 934 snprintf(SNPARGS(proto, len), " %s", dst); 935 break; 936 937 case IPPROTO_UDP: 938 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 939 if (offset == 0) 940 snprintf(SNPARGS(proto, len), ":%d %s:%d", 941 ntohs(udp->uh_sport), 942 dst, 943 ntohs(udp->uh_dport)); 944 else 945 snprintf(SNPARGS(proto, len), " %s", dst); 946 break; 947 948 case IPPROTO_ICMP: 949 icmp = L3HDR(struct icmphdr, ip); 950 if (offset == 0) 951 len = snprintf(SNPARGS(proto, 0), 952 "ICMP:%u.%u ", 953 icmp->icmp_type, icmp->icmp_code); 954 else 955 len = snprintf(SNPARGS(proto, 0), "ICMP "); 956 len += snprintf(SNPARGS(proto, len), "%s", src); 957 snprintf(SNPARGS(proto, len), " %s", dst); 958 break; 959#ifdef INET6 960 case IPPROTO_ICMPV6: 961 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 962 if (offset == 0) 963 len = snprintf(SNPARGS(proto, 0), 964 "ICMPv6:%u.%u ", 965 icmp6->icmp6_type, icmp6->icmp6_code); 966 else 967 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 968 len += snprintf(SNPARGS(proto, len), "%s", src); 969 snprintf(SNPARGS(proto, len), " %s", dst); 970 break; 971#endif 972 default: 973 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 974 args->f_id.proto, src); 975 snprintf(SNPARGS(proto, len), " %s", dst); 976 break; 977 } 978 979#ifdef INET6 980 if (IS_IP6_FLOW_ID(&(args->f_id))) { 981 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 982 snprintf(SNPARGS(fragment, 0), 983 " (frag %08x:%d@%d%s)", 984 args->f_id.frag_id6, 985 ntohs(ip6->ip6_plen) - hlen, 986 ntohs(offset & IP6F_OFF_MASK) << 3, 987 (offset & IP6F_MORE_FRAG) ? "+" : ""); 988 } else 989#endif 990 { 991 int ip_off, ip_len; 992 if (eh != NULL) { /* layer 2 packets are as on the wire */ 993 ip_off = ntohs(ip->ip_off); 994 ip_len = ntohs(ip->ip_len); 995 } else { 996 ip_off = ip->ip_off; 997 ip_len = ip->ip_len; 998 } 999 if (ip_off & (IP_MF | IP_OFFMASK)) 1000 snprintf(SNPARGS(fragment, 0), 1001 " (frag %d:%d@%d%s)", 1002 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1003 offset << 3, 1004 (ip_off & IP_MF) ? "+" : ""); 1005 } 1006 } 1007 if (oif || m->m_pkthdr.rcvif) 1008 log(LOG_SECURITY | LOG_INFO, 1009 "ipfw: %d %s %s %s via %s%s\n", 1010 f ? f->rulenum : -1, 1011 action, proto, oif ? "out" : "in", 1012 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1013 fragment); 1014 else 1015 log(LOG_SECURITY | LOG_INFO, 1016 "ipfw: %d %s %s [no if info]%s\n", 1017 f ? f->rulenum : -1, 1018 action, proto, fragment); 1019 if (limit_reached) 1020 log(LOG_SECURITY | LOG_NOTICE, 1021 "ipfw: limit %d reached on entry %d\n", 1022 limit_reached, f ? f->rulenum : -1); 1023} 1024 1025/* 1026 * IMPORTANT: the hash function for dynamic rules must be commutative 1027 * in source and destination (ip,port), because rules are bidirectional 1028 * and we want to find both in the same bucket. 1029 */ 1030static __inline int 1031hash_packet(struct ipfw_flow_id *id) 1032{ 1033 INIT_VNET_IPFW(curvnet); 1034 u_int32_t i; 1035 1036#ifdef INET6 1037 if (IS_IP6_FLOW_ID(id)) 1038 i = hash_packet6(id); 1039 else 1040#endif /* INET6 */ 1041 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1042 i &= (V_curr_dyn_buckets - 1); 1043 return i; 1044} 1045 1046/** 1047 * unlink a dynamic rule from a chain. prev is a pointer to 1048 * the previous one, q is a pointer to the rule to delete, 1049 * head is a pointer to the head of the queue. 1050 * Modifies q and potentially also head. 1051 */ 1052#define UNLINK_DYN_RULE(prev, head, q) { \ 1053 ipfw_dyn_rule *old_q = q; \ 1054 \ 1055 /* remove a refcount to the parent */ \ 1056 if (q->dyn_type == O_LIMIT) \ 1057 q->parent->count--; \ 1058 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1059 (q->id.src_ip), (q->id.src_port), \ 1060 (q->id.dst_ip), (q->id.dst_port), V_dyn_count-1 ); ) \ 1061 if (prev != NULL) \ 1062 prev->next = q = q->next; \ 1063 else \ 1064 head = q = q->next; \ 1065 V_dyn_count--; \ 1066 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1067 1068#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1069 1070/** 1071 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1072 * 1073 * If keep_me == NULL, rules are deleted even if not expired, 1074 * otherwise only expired rules are removed. 1075 * 1076 * The value of the second parameter is also used to point to identify 1077 * a rule we absolutely do not want to remove (e.g. because we are 1078 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1079 * rules). The pointer is only used for comparison, so any non-null 1080 * value will do. 1081 */ 1082static void 1083remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1084{ 1085 INIT_VNET_IPFW(curvnet); 1086 static u_int32_t last_remove = 0; 1087 1088#define FORCE (keep_me == NULL) 1089 1090 ipfw_dyn_rule *prev, *q; 1091 int i, pass = 0, max_pass = 0; 1092 1093 IPFW_DYN_LOCK_ASSERT(); 1094 1095 if (V_ipfw_dyn_v == NULL || V_dyn_count == 0) 1096 return; 1097 /* do not expire more than once per second, it is useless */ 1098 if (!FORCE && last_remove == time_uptime) 1099 return; 1100 last_remove = time_uptime; 1101 1102 /* 1103 * because O_LIMIT refer to parent rules, during the first pass only 1104 * remove child and mark any pending LIMIT_PARENT, and remove 1105 * them in a second pass. 1106 */ 1107next_pass: 1108 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 1109 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q ; ) { 1110 /* 1111 * Logic can become complex here, so we split tests. 1112 */ 1113 if (q == keep_me) 1114 goto next; 1115 if (rule != NULL && rule != q->rule) 1116 goto next; /* not the one we are looking for */ 1117 if (q->dyn_type == O_LIMIT_PARENT) { 1118 /* 1119 * handle parent in the second pass, 1120 * record we need one. 1121 */ 1122 max_pass = 1; 1123 if (pass == 0) 1124 goto next; 1125 if (FORCE && q->count != 0 ) { 1126 /* XXX should not happen! */ 1127 printf("ipfw: OUCH! cannot remove rule," 1128 " count %d\n", q->count); 1129 } 1130 } else { 1131 if (!FORCE && 1132 !TIME_LEQ( q->expire, time_uptime )) 1133 goto next; 1134 } 1135 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1136 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1137 continue; 1138 } 1139next: 1140 prev=q; 1141 q=q->next; 1142 } 1143 } 1144 if (pass++ < max_pass) 1145 goto next_pass; 1146} 1147 1148 1149/** 1150 * lookup a dynamic rule. 1151 */ 1152static ipfw_dyn_rule * 1153lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1154 struct tcphdr *tcp) 1155{ 1156 INIT_VNET_IPFW(curvnet); 1157 /* 1158 * stateful ipfw extensions. 1159 * Lookup into dynamic session queue 1160 */ 1161#define MATCH_REVERSE 0 1162#define MATCH_FORWARD 1 1163#define MATCH_NONE 2 1164#define MATCH_UNKNOWN 3 1165 int i, dir = MATCH_NONE; 1166 ipfw_dyn_rule *prev, *q=NULL; 1167 1168 IPFW_DYN_LOCK_ASSERT(); 1169 1170 if (V_ipfw_dyn_v == NULL) 1171 goto done; /* not found */ 1172 i = hash_packet( pkt ); 1173 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q != NULL ; ) { 1174 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1175 goto next; 1176 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1177 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1178 continue; 1179 } 1180 if (pkt->proto == q->id.proto && 1181 q->dyn_type != O_LIMIT_PARENT) { 1182 if (IS_IP6_FLOW_ID(pkt)) { 1183 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1184 &(q->id.src_ip6)) && 1185 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1186 &(q->id.dst_ip6)) && 1187 pkt->src_port == q->id.src_port && 1188 pkt->dst_port == q->id.dst_port ) { 1189 dir = MATCH_FORWARD; 1190 break; 1191 } 1192 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1193 &(q->id.dst_ip6)) && 1194 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1195 &(q->id.src_ip6)) && 1196 pkt->src_port == q->id.dst_port && 1197 pkt->dst_port == q->id.src_port ) { 1198 dir = MATCH_REVERSE; 1199 break; 1200 } 1201 } else { 1202 if (pkt->src_ip == q->id.src_ip && 1203 pkt->dst_ip == q->id.dst_ip && 1204 pkt->src_port == q->id.src_port && 1205 pkt->dst_port == q->id.dst_port ) { 1206 dir = MATCH_FORWARD; 1207 break; 1208 } 1209 if (pkt->src_ip == q->id.dst_ip && 1210 pkt->dst_ip == q->id.src_ip && 1211 pkt->src_port == q->id.dst_port && 1212 pkt->dst_port == q->id.src_port ) { 1213 dir = MATCH_REVERSE; 1214 break; 1215 } 1216 } 1217 } 1218next: 1219 prev = q; 1220 q = q->next; 1221 } 1222 if (q == NULL) 1223 goto done; /* q = NULL, not found */ 1224 1225 if ( prev != NULL) { /* found and not in front */ 1226 prev->next = q->next; 1227 q->next = V_ipfw_dyn_v[i]; 1228 V_ipfw_dyn_v[i] = q; 1229 } 1230 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1231 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1232 1233#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1234#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1235 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1236 switch (q->state) { 1237 case TH_SYN: /* opening */ 1238 q->expire = time_uptime + V_dyn_syn_lifetime; 1239 break; 1240 1241 case BOTH_SYN: /* move to established */ 1242 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1243 case BOTH_SYN | (TH_FIN << 8) : 1244 if (tcp) { 1245#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1246 u_int32_t ack = ntohl(tcp->th_ack); 1247 if (dir == MATCH_FORWARD) { 1248 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1249 q->ack_fwd = ack; 1250 else { /* ignore out-of-sequence */ 1251 break; 1252 } 1253 } else { 1254 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1255 q->ack_rev = ack; 1256 else { /* ignore out-of-sequence */ 1257 break; 1258 } 1259 } 1260 } 1261 q->expire = time_uptime + V_dyn_ack_lifetime; 1262 break; 1263 1264 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1265 if (V_dyn_fin_lifetime >= V_dyn_keepalive_period) 1266 V_dyn_fin_lifetime = V_dyn_keepalive_period - 1; 1267 q->expire = time_uptime + V_dyn_fin_lifetime; 1268 break; 1269 1270 default: 1271#if 0 1272 /* 1273 * reset or some invalid combination, but can also 1274 * occur if we use keep-state the wrong way. 1275 */ 1276 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1277 printf("invalid state: 0x%x\n", q->state); 1278#endif 1279 if (V_dyn_rst_lifetime >= V_dyn_keepalive_period) 1280 V_dyn_rst_lifetime = V_dyn_keepalive_period - 1; 1281 q->expire = time_uptime + V_dyn_rst_lifetime; 1282 break; 1283 } 1284 } else if (pkt->proto == IPPROTO_UDP) { 1285 q->expire = time_uptime + V_dyn_udp_lifetime; 1286 } else { 1287 /* other protocols */ 1288 q->expire = time_uptime + V_dyn_short_lifetime; 1289 } 1290done: 1291 if (match_direction) 1292 *match_direction = dir; 1293 return q; 1294} 1295 1296static ipfw_dyn_rule * 1297lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1298 struct tcphdr *tcp) 1299{ 1300 ipfw_dyn_rule *q; 1301 1302 IPFW_DYN_LOCK(); 1303 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1304 if (q == NULL) 1305 IPFW_DYN_UNLOCK(); 1306 /* NB: return table locked when q is not NULL */ 1307 return q; 1308} 1309 1310static void 1311realloc_dynamic_table(void) 1312{ 1313 INIT_VNET_IPFW(curvnet); 1314 IPFW_DYN_LOCK_ASSERT(); 1315 1316 /* 1317 * Try reallocation, make sure we have a power of 2 and do 1318 * not allow more than 64k entries. In case of overflow, 1319 * default to 1024. 1320 */ 1321 1322 if (V_dyn_buckets > 65536) 1323 V_dyn_buckets = 1024; 1324 if ((V_dyn_buckets & (V_dyn_buckets-1)) != 0) { /* not a power of 2 */ 1325 V_dyn_buckets = V_curr_dyn_buckets; /* reset */ 1326 return; 1327 } 1328 V_curr_dyn_buckets = V_dyn_buckets; 1329 if (V_ipfw_dyn_v != NULL) 1330 free(V_ipfw_dyn_v, M_IPFW); 1331 for (;;) { 1332 V_ipfw_dyn_v = malloc(V_curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1333 M_IPFW, M_NOWAIT | M_ZERO); 1334 if (V_ipfw_dyn_v != NULL || V_curr_dyn_buckets <= 2) 1335 break; 1336 V_curr_dyn_buckets /= 2; 1337 } 1338} 1339 1340/** 1341 * Install state of type 'type' for a dynamic session. 1342 * The hash table contains two type of rules: 1343 * - regular rules (O_KEEP_STATE) 1344 * - rules for sessions with limited number of sess per user 1345 * (O_LIMIT). When they are created, the parent is 1346 * increased by 1, and decreased on delete. In this case, 1347 * the third parameter is the parent rule and not the chain. 1348 * - "parent" rules for the above (O_LIMIT_PARENT). 1349 */ 1350static ipfw_dyn_rule * 1351add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1352{ 1353 INIT_VNET_IPFW(curvnet); 1354 ipfw_dyn_rule *r; 1355 int i; 1356 1357 IPFW_DYN_LOCK_ASSERT(); 1358 1359 if (V_ipfw_dyn_v == NULL || 1360 (V_dyn_count == 0 && V_dyn_buckets != V_curr_dyn_buckets)) { 1361 realloc_dynamic_table(); 1362 if (V_ipfw_dyn_v == NULL) 1363 return NULL; /* failed ! */ 1364 } 1365 i = hash_packet(id); 1366 1367 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1368 if (r == NULL) { 1369 printf ("ipfw: sorry cannot allocate state\n"); 1370 return NULL; 1371 } 1372 1373 /* increase refcount on parent, and set pointer */ 1374 if (dyn_type == O_LIMIT) { 1375 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1376 if ( parent->dyn_type != O_LIMIT_PARENT) 1377 panic("invalid parent"); 1378 parent->count++; 1379 r->parent = parent; 1380 rule = parent->rule; 1381 } 1382 1383 r->id = *id; 1384 r->expire = time_uptime + V_dyn_syn_lifetime; 1385 r->rule = rule; 1386 r->dyn_type = dyn_type; 1387 r->pcnt = r->bcnt = 0; 1388 r->count = 0; 1389 1390 r->bucket = i; 1391 r->next = V_ipfw_dyn_v[i]; 1392 V_ipfw_dyn_v[i] = r; 1393 V_dyn_count++; 1394 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1395 dyn_type, 1396 (r->id.src_ip), (r->id.src_port), 1397 (r->id.dst_ip), (r->id.dst_port), 1398 V_dyn_count ); ) 1399 return r; 1400} 1401 1402/** 1403 * lookup dynamic parent rule using pkt and rule as search keys. 1404 * If the lookup fails, then install one. 1405 */ 1406static ipfw_dyn_rule * 1407lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1408{ 1409 INIT_VNET_IPFW(curvnet); 1410 ipfw_dyn_rule *q; 1411 int i; 1412 1413 IPFW_DYN_LOCK_ASSERT(); 1414 1415 if (V_ipfw_dyn_v) { 1416 int is_v6 = IS_IP6_FLOW_ID(pkt); 1417 i = hash_packet( pkt ); 1418 for (q = V_ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1419 if (q->dyn_type == O_LIMIT_PARENT && 1420 rule== q->rule && 1421 pkt->proto == q->id.proto && 1422 pkt->src_port == q->id.src_port && 1423 pkt->dst_port == q->id.dst_port && 1424 ( 1425 (is_v6 && 1426 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1427 &(q->id.src_ip6)) && 1428 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1429 &(q->id.dst_ip6))) || 1430 (!is_v6 && 1431 pkt->src_ip == q->id.src_ip && 1432 pkt->dst_ip == q->id.dst_ip) 1433 ) 1434 ) { 1435 q->expire = time_uptime + V_dyn_short_lifetime; 1436 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1437 return q; 1438 } 1439 } 1440 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1441} 1442 1443/** 1444 * Install dynamic state for rule type cmd->o.opcode 1445 * 1446 * Returns 1 (failure) if state is not installed because of errors or because 1447 * session limitations are enforced. 1448 */ 1449static int 1450install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1451 struct ip_fw_args *args, uint32_t tablearg) 1452{ 1453 INIT_VNET_IPFW(curvnet); 1454 static int last_log; 1455 ipfw_dyn_rule *q; 1456 struct in_addr da; 1457 char src[48], dst[48]; 1458 1459 src[0] = '\0'; 1460 dst[0] = '\0'; 1461 1462 DEB( 1463 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1464 __func__, cmd->o.opcode, 1465 (args->f_id.src_ip), (args->f_id.src_port), 1466 (args->f_id.dst_ip), (args->f_id.dst_port)); 1467 ) 1468 1469 IPFW_DYN_LOCK(); 1470 1471 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1472 1473 if (q != NULL) { /* should never occur */ 1474 if (last_log != time_uptime) { 1475 last_log = time_uptime; 1476 printf("ipfw: %s: entry already present, done\n", 1477 __func__); 1478 } 1479 IPFW_DYN_UNLOCK(); 1480 return (0); 1481 } 1482 1483 if (V_dyn_count >= V_dyn_max) 1484 /* Run out of slots, try to remove any expired rule. */ 1485 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1486 1487 if (V_dyn_count >= V_dyn_max) { 1488 if (last_log != time_uptime) { 1489 last_log = time_uptime; 1490 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1491 } 1492 IPFW_DYN_UNLOCK(); 1493 return (1); /* cannot install, notify caller */ 1494 } 1495 1496 switch (cmd->o.opcode) { 1497 case O_KEEP_STATE: /* bidir rule */ 1498 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1499 break; 1500 1501 case O_LIMIT: { /* limit number of sessions */ 1502 struct ipfw_flow_id id; 1503 ipfw_dyn_rule *parent; 1504 uint32_t conn_limit; 1505 uint16_t limit_mask = cmd->limit_mask; 1506 1507 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1508 tablearg : cmd->conn_limit; 1509 1510 DEB( 1511 if (cmd->conn_limit == IP_FW_TABLEARG) 1512 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1513 "(tablearg)\n", __func__, conn_limit); 1514 else 1515 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1516 __func__, conn_limit); 1517 ) 1518 1519 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1520 id.proto = args->f_id.proto; 1521 id.addr_type = args->f_id.addr_type; 1522 id.fib = M_GETFIB(args->m); 1523 1524 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1525 if (limit_mask & DYN_SRC_ADDR) 1526 id.src_ip6 = args->f_id.src_ip6; 1527 if (limit_mask & DYN_DST_ADDR) 1528 id.dst_ip6 = args->f_id.dst_ip6; 1529 } else { 1530 if (limit_mask & DYN_SRC_ADDR) 1531 id.src_ip = args->f_id.src_ip; 1532 if (limit_mask & DYN_DST_ADDR) 1533 id.dst_ip = args->f_id.dst_ip; 1534 } 1535 if (limit_mask & DYN_SRC_PORT) 1536 id.src_port = args->f_id.src_port; 1537 if (limit_mask & DYN_DST_PORT) 1538 id.dst_port = args->f_id.dst_port; 1539 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1540 printf("ipfw: %s: add parent failed\n", __func__); 1541 IPFW_DYN_UNLOCK(); 1542 return (1); 1543 } 1544 1545 if (parent->count >= conn_limit) { 1546 /* See if we can remove some expired rule. */ 1547 remove_dyn_rule(rule, parent); 1548 if (parent->count >= conn_limit) { 1549 if (V_fw_verbose && last_log != time_uptime) { 1550 last_log = time_uptime; 1551#ifdef INET6 1552 /* 1553 * XXX IPv6 flows are not 1554 * supported yet. 1555 */ 1556 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1557 char ip6buf[INET6_ADDRSTRLEN]; 1558 snprintf(src, sizeof(src), 1559 "[%s]", ip6_sprintf(ip6buf, 1560 &args->f_id.src_ip6)); 1561 snprintf(dst, sizeof(dst), 1562 "[%s]", ip6_sprintf(ip6buf, 1563 &args->f_id.dst_ip6)); 1564 } else 1565#endif 1566 { 1567 da.s_addr = 1568 htonl(args->f_id.src_ip); 1569 inet_ntoa_r(da, src); 1570 da.s_addr = 1571 htonl(args->f_id.dst_ip); 1572 inet_ntoa_r(da, dst); 1573 } 1574 log(LOG_SECURITY | LOG_DEBUG, 1575 "ipfw: %d %s %s:%u -> %s:%u, %s\n", 1576 parent->rule->rulenum, 1577 "drop session", 1578 src, (args->f_id.src_port), 1579 dst, (args->f_id.dst_port), 1580 "too many entries"); 1581 } 1582 IPFW_DYN_UNLOCK(); 1583 return (1); 1584 } 1585 } 1586 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1587 break; 1588 } 1589 default: 1590 printf("ipfw: %s: unknown dynamic rule type %u\n", 1591 __func__, cmd->o.opcode); 1592 IPFW_DYN_UNLOCK(); 1593 return (1); 1594 } 1595 1596 /* XXX just set lifetime */ 1597 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1598 1599 IPFW_DYN_UNLOCK(); 1600 return (0); 1601} 1602 1603/* 1604 * Generate a TCP packet, containing either a RST or a keepalive. 1605 * When flags & TH_RST, we are sending a RST packet, because of a 1606 * "reset" action matched the packet. 1607 * Otherwise we are sending a keepalive, and flags & TH_ 1608 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1609 * so that MAC can label the reply appropriately. 1610 */ 1611static struct mbuf * 1612send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1613 u_int32_t ack, int flags) 1614{ 1615 INIT_VNET_INET(curvnet); 1616 struct mbuf *m; 1617 struct ip *ip; 1618 struct tcphdr *tcp; 1619 1620 MGETHDR(m, M_DONTWAIT, MT_DATA); 1621 if (m == 0) 1622 return (NULL); 1623 m->m_pkthdr.rcvif = (struct ifnet *)0; 1624 1625 M_SETFIB(m, id->fib); 1626#ifdef MAC 1627 if (replyto != NULL) 1628 mac_netinet_firewall_reply(replyto, m); 1629 else 1630 mac_netinet_firewall_send(m); 1631#else 1632 (void)replyto; /* don't warn about unused arg */ 1633#endif 1634 1635 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1636 m->m_data += max_linkhdr; 1637 1638 ip = mtod(m, struct ip *); 1639 bzero(ip, m->m_len); 1640 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1641 ip->ip_p = IPPROTO_TCP; 1642 tcp->th_off = 5; 1643 /* 1644 * Assume we are sending a RST (or a keepalive in the reverse 1645 * direction), swap src and destination addresses and ports. 1646 */ 1647 ip->ip_src.s_addr = htonl(id->dst_ip); 1648 ip->ip_dst.s_addr = htonl(id->src_ip); 1649 tcp->th_sport = htons(id->dst_port); 1650 tcp->th_dport = htons(id->src_port); 1651 if (flags & TH_RST) { /* we are sending a RST */ 1652 if (flags & TH_ACK) { 1653 tcp->th_seq = htonl(ack); 1654 tcp->th_ack = htonl(0); 1655 tcp->th_flags = TH_RST; 1656 } else { 1657 if (flags & TH_SYN) 1658 seq++; 1659 tcp->th_seq = htonl(0); 1660 tcp->th_ack = htonl(seq); 1661 tcp->th_flags = TH_RST | TH_ACK; 1662 } 1663 } else { 1664 /* 1665 * We are sending a keepalive. flags & TH_SYN determines 1666 * the direction, forward if set, reverse if clear. 1667 * NOTE: seq and ack are always assumed to be correct 1668 * as set by the caller. This may be confusing... 1669 */ 1670 if (flags & TH_SYN) { 1671 /* 1672 * we have to rewrite the correct addresses! 1673 */ 1674 ip->ip_dst.s_addr = htonl(id->dst_ip); 1675 ip->ip_src.s_addr = htonl(id->src_ip); 1676 tcp->th_dport = htons(id->dst_port); 1677 tcp->th_sport = htons(id->src_port); 1678 } 1679 tcp->th_seq = htonl(seq); 1680 tcp->th_ack = htonl(ack); 1681 tcp->th_flags = TH_ACK; 1682 } 1683 /* 1684 * set ip_len to the payload size so we can compute 1685 * the tcp checksum on the pseudoheader 1686 * XXX check this, could save a couple of words ? 1687 */ 1688 ip->ip_len = htons(sizeof(struct tcphdr)); 1689 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1690 /* 1691 * now fill fields left out earlier 1692 */ 1693 ip->ip_ttl = V_ip_defttl; 1694 ip->ip_len = m->m_pkthdr.len; 1695 m->m_flags |= M_SKIP_FIREWALL; 1696 return (m); 1697} 1698 1699/* 1700 * sends a reject message, consuming the mbuf passed as an argument. 1701 */ 1702static void 1703send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1704{ 1705 1706#if 0 1707 /* XXX When ip is not guaranteed to be at mtod() we will 1708 * need to account for this */ 1709 * The mbuf will however be thrown away so we can adjust it. 1710 * Remember we did an m_pullup on it already so we 1711 * can make some assumptions about contiguousness. 1712 */ 1713 if (args->L3offset) 1714 m_adj(m, args->L3offset); 1715#endif 1716 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1717 /* We need the IP header in host order for icmp_error(). */ 1718 if (args->eh != NULL) { 1719 ip->ip_len = ntohs(ip->ip_len); 1720 ip->ip_off = ntohs(ip->ip_off); 1721 } 1722 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1723 } else if (args->f_id.proto == IPPROTO_TCP) { 1724 struct tcphdr *const tcp = 1725 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1726 if ( (tcp->th_flags & TH_RST) == 0) { 1727 struct mbuf *m; 1728 m = send_pkt(args->m, &(args->f_id), 1729 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1730 tcp->th_flags | TH_RST); 1731 if (m != NULL) 1732 ip_output(m, NULL, NULL, 0, NULL, NULL); 1733 } 1734 m_freem(args->m); 1735 } else 1736 m_freem(args->m); 1737 args->m = NULL; 1738} 1739 1740/** 1741 * 1742 * Given an ip_fw *, lookup_next_rule will return a pointer 1743 * to the next rule, which can be either the jump 1744 * target (for skipto instructions) or the next one in the list (in 1745 * all other cases including a missing jump target). 1746 * The result is also written in the "next_rule" field of the rule. 1747 * Backward jumps are not allowed, so start looking from the next 1748 * rule... 1749 * 1750 * This never returns NULL -- in case we do not have an exact match, 1751 * the next rule is returned. When the ruleset is changed, 1752 * pointers are flushed so we are always correct. 1753 */ 1754 1755static struct ip_fw * 1756lookup_next_rule(struct ip_fw *me, u_int32_t tablearg) 1757{ 1758 struct ip_fw *rule = NULL; 1759 ipfw_insn *cmd; 1760 u_int16_t rulenum; 1761 1762 /* look for action, in case it is a skipto */ 1763 cmd = ACTION_PTR(me); 1764 if (cmd->opcode == O_LOG) 1765 cmd += F_LEN(cmd); 1766 if (cmd->opcode == O_ALTQ) 1767 cmd += F_LEN(cmd); 1768 if (cmd->opcode == O_TAG) 1769 cmd += F_LEN(cmd); 1770 if (cmd->opcode == O_SKIPTO ) { 1771 if (tablearg != 0) { 1772 rulenum = (u_int16_t)tablearg; 1773 } else { 1774 rulenum = cmd->arg1; 1775 } 1776 for (rule = me->next; rule ; rule = rule->next) { 1777 if (rule->rulenum >= rulenum) { 1778 break; 1779 } 1780 } 1781 } 1782 if (rule == NULL) /* failure or not a skipto */ 1783 rule = me->next; 1784 me->next_rule = rule; 1785 return rule; 1786} 1787 1788static int 1789add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1790 uint8_t mlen, uint32_t value) 1791{ 1792 INIT_VNET_IPFW(curvnet); 1793 struct radix_node_head *rnh; 1794 struct table_entry *ent; 1795 1796 if (tbl >= IPFW_TABLES_MAX) 1797 return (EINVAL); 1798 rnh = ch->tables[tbl]; 1799 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1800 if (ent == NULL) 1801 return (ENOMEM); 1802 ent->value = value; 1803 ent->addr.sin_len = ent->mask.sin_len = 8; 1804 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1805 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1806 IPFW_WLOCK(&V_layer3_chain); 1807 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1808 NULL) { 1809 IPFW_WUNLOCK(&V_layer3_chain); 1810 free(ent, M_IPFW_TBL); 1811 return (EEXIST); 1812 } 1813 IPFW_WUNLOCK(&V_layer3_chain); 1814 return (0); 1815} 1816 1817static int 1818del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1819 uint8_t mlen) 1820{ 1821 struct radix_node_head *rnh; 1822 struct table_entry *ent; 1823 struct sockaddr_in sa, mask; 1824 1825 if (tbl >= IPFW_TABLES_MAX) 1826 return (EINVAL); 1827 rnh = ch->tables[tbl]; 1828 sa.sin_len = mask.sin_len = 8; 1829 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1830 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1831 IPFW_WLOCK(ch); 1832 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1833 if (ent == NULL) { 1834 IPFW_WUNLOCK(ch); 1835 return (ESRCH); 1836 } 1837 IPFW_WUNLOCK(ch); 1838 free(ent, M_IPFW_TBL); 1839 return (0); 1840} 1841 1842static int 1843flush_table_entry(struct radix_node *rn, void *arg) 1844{ 1845 struct radix_node_head * const rnh = arg; 1846 struct table_entry *ent; 1847 1848 ent = (struct table_entry *) 1849 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1850 if (ent != NULL) 1851 free(ent, M_IPFW_TBL); 1852 return (0); 1853} 1854 1855static int 1856flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1857{ 1858 struct radix_node_head *rnh; 1859 1860 IPFW_WLOCK_ASSERT(ch); 1861 1862 if (tbl >= IPFW_TABLES_MAX) 1863 return (EINVAL); 1864 rnh = ch->tables[tbl]; 1865 KASSERT(rnh != NULL, ("NULL IPFW table")); 1866 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1867 return (0); 1868} 1869 1870static void 1871flush_tables(struct ip_fw_chain *ch) 1872{ 1873 uint16_t tbl; 1874 1875 IPFW_WLOCK_ASSERT(ch); 1876 1877 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1878 flush_table(ch, tbl); 1879} 1880 1881static int 1882init_tables(struct ip_fw_chain *ch) 1883{ 1884 int i; 1885 uint16_t j; 1886 1887 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1888 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1889 for (j = 0; j < i; j++) { 1890 (void) flush_table(ch, j); 1891 } 1892 return (ENOMEM); 1893 } 1894 } 1895 return (0); 1896} 1897 1898static int 1899lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1900 uint32_t *val) 1901{ 1902 struct radix_node_head *rnh; 1903 struct table_entry *ent; 1904 struct sockaddr_in sa; 1905 1906 if (tbl >= IPFW_TABLES_MAX) 1907 return (0); 1908 rnh = ch->tables[tbl]; 1909 sa.sin_len = 8; 1910 sa.sin_addr.s_addr = addr; 1911 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1912 if (ent != NULL) { 1913 *val = ent->value; 1914 return (1); 1915 } 1916 return (0); 1917} 1918 1919static int 1920count_table_entry(struct radix_node *rn, void *arg) 1921{ 1922 u_int32_t * const cnt = arg; 1923 1924 (*cnt)++; 1925 return (0); 1926} 1927 1928static int 1929count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1930{ 1931 struct radix_node_head *rnh; 1932 1933 if (tbl >= IPFW_TABLES_MAX) 1934 return (EINVAL); 1935 rnh = ch->tables[tbl]; 1936 *cnt = 0; 1937 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1938 return (0); 1939} 1940 1941static int 1942dump_table_entry(struct radix_node *rn, void *arg) 1943{ 1944 struct table_entry * const n = (struct table_entry *)rn; 1945 ipfw_table * const tbl = arg; 1946 ipfw_table_entry *ent; 1947 1948 if (tbl->cnt == tbl->size) 1949 return (1); 1950 ent = &tbl->ent[tbl->cnt]; 1951 ent->tbl = tbl->tbl; 1952 if (in_nullhost(n->mask.sin_addr)) 1953 ent->masklen = 0; 1954 else 1955 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1956 ent->addr = n->addr.sin_addr.s_addr; 1957 ent->value = n->value; 1958 tbl->cnt++; 1959 return (0); 1960} 1961 1962static int 1963dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1964{ 1965 struct radix_node_head *rnh; 1966 1967 if (tbl->tbl >= IPFW_TABLES_MAX) 1968 return (EINVAL); 1969 rnh = ch->tables[tbl->tbl]; 1970 tbl->cnt = 0; 1971 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1972 return (0); 1973} 1974 1975static void 1976fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1977{ 1978 struct ucred *cr; 1979 1980 cr = inp->inp_cred; 1981 ugp->fw_prid = jailed(cr) ? cr->cr_prison->pr_id : -1; 1982 ugp->fw_uid = cr->cr_uid; 1983 ugp->fw_ngroups = cr->cr_ngroups; 1984 bcopy(cr->cr_groups, ugp->fw_groups, sizeof(ugp->fw_groups)); 1985} 1986 1987static int 1988check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 1989 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 1990 u_int16_t src_port, struct ip_fw_ugid *ugp, int *ugid_lookupp, 1991 struct inpcb *inp) 1992{ 1993 INIT_VNET_INET(curvnet); 1994 struct inpcbinfo *pi; 1995 int wildcard; 1996 struct inpcb *pcb; 1997 int match; 1998 gid_t *gp; 1999 2000 /* 2001 * Check to see if the UDP or TCP stack supplied us with 2002 * the PCB. If so, rather then holding a lock and looking 2003 * up the PCB, we can use the one that was supplied. 2004 */ 2005 if (inp && *ugid_lookupp == 0) { 2006 INP_LOCK_ASSERT(inp); 2007 if (inp->inp_socket != NULL) { 2008 fill_ugid_cache(inp, ugp); 2009 *ugid_lookupp = 1; 2010 } else 2011 *ugid_lookupp = -1; 2012 } 2013 /* 2014 * If we have already been here and the packet has no 2015 * PCB entry associated with it, then we can safely 2016 * assume that this is a no match. 2017 */ 2018 if (*ugid_lookupp == -1) 2019 return (0); 2020 if (proto == IPPROTO_TCP) { 2021 wildcard = 0; 2022 pi = &V_tcbinfo; 2023 } else if (proto == IPPROTO_UDP) { 2024 wildcard = INPLOOKUP_WILDCARD; 2025 pi = &V_udbinfo; 2026 } else 2027 return 0; 2028 match = 0; 2029 if (*ugid_lookupp == 0) { 2030 INP_INFO_RLOCK(pi); 2031 pcb = (oif) ? 2032 in_pcblookup_hash(pi, 2033 dst_ip, htons(dst_port), 2034 src_ip, htons(src_port), 2035 wildcard, oif) : 2036 in_pcblookup_hash(pi, 2037 src_ip, htons(src_port), 2038 dst_ip, htons(dst_port), 2039 wildcard, NULL); 2040 if (pcb != NULL) { 2041 fill_ugid_cache(pcb, ugp); 2042 *ugid_lookupp = 1; 2043 } 2044 INP_INFO_RUNLOCK(pi); 2045 if (*ugid_lookupp == 0) { 2046 /* 2047 * If the lookup did not yield any results, there 2048 * is no sense in coming back and trying again. So 2049 * we can set lookup to -1 and ensure that we wont 2050 * bother the pcb system again. 2051 */ 2052 *ugid_lookupp = -1; 2053 return (0); 2054 } 2055 } 2056 if (insn->o.opcode == O_UID) 2057 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2058 else if (insn->o.opcode == O_GID) { 2059 for (gp = ugp->fw_groups; 2060 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2061 if (*gp == (gid_t)insn->d[0]) { 2062 match = 1; 2063 break; 2064 } 2065 } else if (insn->o.opcode == O_JAIL) 2066 match = (ugp->fw_prid == (int)insn->d[0]); 2067 return match; 2068} 2069 2070/* 2071 * The main check routine for the firewall. 2072 * 2073 * All arguments are in args so we can modify them and return them 2074 * back to the caller. 2075 * 2076 * Parameters: 2077 * 2078 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2079 * Starts with the IP header. 2080 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2081 * args->L3offset Number of bytes bypassed if we came from L2. 2082 * e.g. often sizeof(eh) ** NOTYET ** 2083 * args->oif Outgoing interface, or NULL if packet is incoming. 2084 * The incoming interface is in the mbuf. (in) 2085 * args->divert_rule (in/out) 2086 * Skip up to the first rule past this rule number; 2087 * upon return, non-zero port number for divert or tee. 2088 * 2089 * args->rule Pointer to the last matching rule (in/out) 2090 * args->next_hop Socket we are forwarding to (out). 2091 * args->f_id Addresses grabbed from the packet (out) 2092 * args->cookie a cookie depending on rule action 2093 * 2094 * Return value: 2095 * 2096 * IP_FW_PASS the packet must be accepted 2097 * IP_FW_DENY the packet must be dropped 2098 * IP_FW_DIVERT divert packet, port in m_tag 2099 * IP_FW_TEE tee packet, port in m_tag 2100 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2101 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2102 * 2103 */ 2104int 2105ipfw_chk(struct ip_fw_args *args) 2106{ 2107 INIT_VNET_INET(curvnet); 2108 INIT_VNET_IPFW(curvnet); 2109 2110 /* 2111 * Local variables holding state during the processing of a packet: 2112 * 2113 * IMPORTANT NOTE: to speed up the processing of rules, there 2114 * are some assumption on the values of the variables, which 2115 * are documented here. Should you change them, please check 2116 * the implementation of the various instructions to make sure 2117 * that they still work. 2118 * 2119 * args->eh The MAC header. It is non-null for a layer2 2120 * packet, it is NULL for a layer-3 packet. 2121 * **notyet** 2122 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2123 * 2124 * m | args->m Pointer to the mbuf, as received from the caller. 2125 * It may change if ipfw_chk() does an m_pullup, or if it 2126 * consumes the packet because it calls send_reject(). 2127 * XXX This has to change, so that ipfw_chk() never modifies 2128 * or consumes the buffer. 2129 * ip is the beginning of the ip(4 or 6) header. 2130 * Calculated by adding the L3offset to the start of data. 2131 * (Until we start using L3offset, the packet is 2132 * supposed to start with the ip header). 2133 */ 2134 struct mbuf *m = args->m; 2135 struct ip *ip = mtod(m, struct ip *); 2136 2137 /* 2138 * For rules which contain uid/gid or jail constraints, cache 2139 * a copy of the users credentials after the pcb lookup has been 2140 * executed. This will speed up the processing of rules with 2141 * these types of constraints, as well as decrease contention 2142 * on pcb related locks. 2143 */ 2144 struct ip_fw_ugid fw_ugid_cache; 2145 int ugid_lookup = 0; 2146 2147 /* 2148 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2149 * associated with a packet input on a divert socket. This 2150 * will allow to distinguish traffic and its direction when 2151 * it originates from a divert socket. 2152 */ 2153 u_int divinput_flags = 0; 2154 2155 /* 2156 * oif | args->oif If NULL, ipfw_chk has been called on the 2157 * inbound path (ether_input, ip_input). 2158 * If non-NULL, ipfw_chk has been called on the outbound path 2159 * (ether_output, ip_output). 2160 */ 2161 struct ifnet *oif = args->oif; 2162 2163 struct ip_fw *f = NULL; /* matching rule */ 2164 int retval = 0; 2165 2166 /* 2167 * hlen The length of the IP header. 2168 */ 2169 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2170 2171 /* 2172 * offset The offset of a fragment. offset != 0 means that 2173 * we have a fragment at this offset of an IPv4 packet. 2174 * offset == 0 means that (if this is an IPv4 packet) 2175 * this is the first or only fragment. 2176 * For IPv6 offset == 0 means there is no Fragment Header. 2177 * If offset != 0 for IPv6 always use correct mask to 2178 * get the correct offset because we add IP6F_MORE_FRAG 2179 * to be able to dectect the first fragment which would 2180 * otherwise have offset = 0. 2181 */ 2182 u_short offset = 0; 2183 2184 /* 2185 * Local copies of addresses. They are only valid if we have 2186 * an IP packet. 2187 * 2188 * proto The protocol. Set to 0 for non-ip packets, 2189 * or to the protocol read from the packet otherwise. 2190 * proto != 0 means that we have an IPv4 packet. 2191 * 2192 * src_port, dst_port port numbers, in HOST format. Only 2193 * valid for TCP and UDP packets. 2194 * 2195 * src_ip, dst_ip ip addresses, in NETWORK format. 2196 * Only valid for IPv4 packets. 2197 */ 2198 u_int8_t proto; 2199 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2200 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2201 u_int16_t ip_len=0; 2202 int pktlen; 2203 u_int16_t etype = 0; /* Host order stored ether type */ 2204 2205 /* 2206 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2207 * MATCH_NONE when checked and not matched (q = NULL), 2208 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2209 */ 2210 int dyn_dir = MATCH_UNKNOWN; 2211 ipfw_dyn_rule *q = NULL; 2212 struct ip_fw_chain *chain = &V_layer3_chain; 2213 struct m_tag *mtag; 2214 2215 /* 2216 * We store in ulp a pointer to the upper layer protocol header. 2217 * In the ipv4 case this is easy to determine from the header, 2218 * but for ipv6 we might have some additional headers in the middle. 2219 * ulp is NULL if not found. 2220 */ 2221 void *ulp = NULL; /* upper layer protocol pointer. */ 2222 /* XXX ipv6 variables */ 2223 int is_ipv6 = 0; 2224 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2225 /* end of ipv6 variables */ 2226 int is_ipv4 = 0; 2227 2228 if (m->m_flags & M_SKIP_FIREWALL) 2229 return (IP_FW_PASS); /* accept */ 2230 2231 pktlen = m->m_pkthdr.len; 2232 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 2233 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2234 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2235 2236/* 2237 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2238 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2239 * pointer might become stale after other pullups (but we never use it 2240 * this way). 2241 */ 2242#define PULLUP_TO(len, p, T) \ 2243do { \ 2244 int x = (len) + sizeof(T); \ 2245 if ((m)->m_len < x) { \ 2246 args->m = m = m_pullup(m, x); \ 2247 if (m == NULL) \ 2248 goto pullup_failed; \ 2249 } \ 2250 p = (mtod(m, char *) + (len)); \ 2251} while (0) 2252 2253 /* 2254 * if we have an ether header, 2255 */ 2256 if (args->eh) 2257 etype = ntohs(args->eh->ether_type); 2258 2259 /* Identify IP packets and fill up variables. */ 2260 if (pktlen >= sizeof(struct ip6_hdr) && 2261 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2262 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2263 is_ipv6 = 1; 2264 args->f_id.addr_type = 6; 2265 hlen = sizeof(struct ip6_hdr); 2266 proto = ip6->ip6_nxt; 2267 2268 /* Search extension headers to find upper layer protocols */ 2269 while (ulp == NULL) { 2270 switch (proto) { 2271 case IPPROTO_ICMPV6: 2272 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2273 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2274 break; 2275 2276 case IPPROTO_TCP: 2277 PULLUP_TO(hlen, ulp, struct tcphdr); 2278 dst_port = TCP(ulp)->th_dport; 2279 src_port = TCP(ulp)->th_sport; 2280 args->f_id.flags = TCP(ulp)->th_flags; 2281 break; 2282 2283 case IPPROTO_SCTP: 2284 PULLUP_TO(hlen, ulp, struct sctphdr); 2285 src_port = SCTP(ulp)->src_port; 2286 dst_port = SCTP(ulp)->dest_port; 2287 break; 2288 2289 case IPPROTO_UDP: 2290 PULLUP_TO(hlen, ulp, struct udphdr); 2291 dst_port = UDP(ulp)->uh_dport; 2292 src_port = UDP(ulp)->uh_sport; 2293 break; 2294 2295 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2296 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2297 ext_hd |= EXT_HOPOPTS; 2298 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2299 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2300 ulp = NULL; 2301 break; 2302 2303 case IPPROTO_ROUTING: /* RFC 2460 */ 2304 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2305 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2306 case 0: 2307 ext_hd |= EXT_RTHDR0; 2308 break; 2309 case 2: 2310 ext_hd |= EXT_RTHDR2; 2311 break; 2312 default: 2313 printf("IPFW2: IPV6 - Unknown Routing " 2314 "Header type(%d)\n", 2315 ((struct ip6_rthdr *)ulp)->ip6r_type); 2316 if (V_fw_deny_unknown_exthdrs) 2317 return (IP_FW_DENY); 2318 break; 2319 } 2320 ext_hd |= EXT_ROUTING; 2321 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2322 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2323 ulp = NULL; 2324 break; 2325 2326 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2327 PULLUP_TO(hlen, ulp, struct ip6_frag); 2328 ext_hd |= EXT_FRAGMENT; 2329 hlen += sizeof (struct ip6_frag); 2330 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2331 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2332 IP6F_OFF_MASK; 2333 /* Add IP6F_MORE_FRAG for offset of first 2334 * fragment to be != 0. */ 2335 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2336 IP6F_MORE_FRAG; 2337 if (offset == 0) { 2338 printf("IPFW2: IPV6 - Invalid Fragment " 2339 "Header\n"); 2340 if (V_fw_deny_unknown_exthdrs) 2341 return (IP_FW_DENY); 2342 break; 2343 } 2344 args->f_id.frag_id6 = 2345 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2346 ulp = NULL; 2347 break; 2348 2349 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2350 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2351 ext_hd |= EXT_DSTOPTS; 2352 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2353 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2354 ulp = NULL; 2355 break; 2356 2357 case IPPROTO_AH: /* RFC 2402 */ 2358 PULLUP_TO(hlen, ulp, struct ip6_ext); 2359 ext_hd |= EXT_AH; 2360 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2361 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2362 ulp = NULL; 2363 break; 2364 2365 case IPPROTO_ESP: /* RFC 2406 */ 2366 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2367 /* Anything past Seq# is variable length and 2368 * data past this ext. header is encrypted. */ 2369 ext_hd |= EXT_ESP; 2370 break; 2371 2372 case IPPROTO_NONE: /* RFC 2460 */ 2373 /* 2374 * Packet ends here, and IPv6 header has 2375 * already been pulled up. If ip6e_len!=0 2376 * then octets must be ignored. 2377 */ 2378 ulp = ip; /* non-NULL to get out of loop. */ 2379 break; 2380 2381 case IPPROTO_OSPFIGP: 2382 /* XXX OSPF header check? */ 2383 PULLUP_TO(hlen, ulp, struct ip6_ext); 2384 break; 2385 2386 case IPPROTO_PIM: 2387 /* XXX PIM header check? */ 2388 PULLUP_TO(hlen, ulp, struct pim); 2389 break; 2390 2391 case IPPROTO_CARP: 2392 PULLUP_TO(hlen, ulp, struct carp_header); 2393 if (((struct carp_header *)ulp)->carp_version != 2394 CARP_VERSION) 2395 return (IP_FW_DENY); 2396 if (((struct carp_header *)ulp)->carp_type != 2397 CARP_ADVERTISEMENT) 2398 return (IP_FW_DENY); 2399 break; 2400 2401 case IPPROTO_IPV6: /* RFC 2893 */ 2402 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2403 break; 2404 2405 case IPPROTO_IPV4: /* RFC 2893 */ 2406 PULLUP_TO(hlen, ulp, struct ip); 2407 break; 2408 2409 default: 2410 printf("IPFW2: IPV6 - Unknown Extension " 2411 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2412 if (V_fw_deny_unknown_exthdrs) 2413 return (IP_FW_DENY); 2414 PULLUP_TO(hlen, ulp, struct ip6_ext); 2415 break; 2416 } /*switch */ 2417 } 2418 ip = mtod(m, struct ip *); 2419 ip6 = (struct ip6_hdr *)ip; 2420 args->f_id.src_ip6 = ip6->ip6_src; 2421 args->f_id.dst_ip6 = ip6->ip6_dst; 2422 args->f_id.src_ip = 0; 2423 args->f_id.dst_ip = 0; 2424 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2425 } else if (pktlen >= sizeof(struct ip) && 2426 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2427 is_ipv4 = 1; 2428 hlen = ip->ip_hl << 2; 2429 args->f_id.addr_type = 4; 2430 2431 /* 2432 * Collect parameters into local variables for faster matching. 2433 */ 2434 proto = ip->ip_p; 2435 src_ip = ip->ip_src; 2436 dst_ip = ip->ip_dst; 2437 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2438 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2439 ip_len = ntohs(ip->ip_len); 2440 } else { 2441 offset = ip->ip_off & IP_OFFMASK; 2442 ip_len = ip->ip_len; 2443 } 2444 pktlen = ip_len < pktlen ? ip_len : pktlen; 2445 2446 if (offset == 0) { 2447 switch (proto) { 2448 case IPPROTO_TCP: 2449 PULLUP_TO(hlen, ulp, struct tcphdr); 2450 dst_port = TCP(ulp)->th_dport; 2451 src_port = TCP(ulp)->th_sport; 2452 args->f_id.flags = TCP(ulp)->th_flags; 2453 break; 2454 2455 case IPPROTO_UDP: 2456 PULLUP_TO(hlen, ulp, struct udphdr); 2457 dst_port = UDP(ulp)->uh_dport; 2458 src_port = UDP(ulp)->uh_sport; 2459 break; 2460 2461 case IPPROTO_ICMP: 2462 PULLUP_TO(hlen, ulp, struct icmphdr); 2463 args->f_id.flags = ICMP(ulp)->icmp_type; 2464 break; 2465 2466 default: 2467 break; 2468 } 2469 } 2470 2471 ip = mtod(m, struct ip *); 2472 args->f_id.src_ip = ntohl(src_ip.s_addr); 2473 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2474 } 2475#undef PULLUP_TO 2476 if (proto) { /* we may have port numbers, store them */ 2477 args->f_id.proto = proto; 2478 args->f_id.src_port = src_port = ntohs(src_port); 2479 args->f_id.dst_port = dst_port = ntohs(dst_port); 2480 } 2481 2482 IPFW_RLOCK(chain); 2483 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2484 if (args->rule) { 2485 /* 2486 * Packet has already been tagged. Look for the next rule 2487 * to restart processing. 2488 * 2489 * If fw_one_pass != 0 then just accept it. 2490 * XXX should not happen here, but optimized out in 2491 * the caller. 2492 */ 2493 if (V_fw_one_pass) { 2494 IPFW_RUNLOCK(chain); 2495 return (IP_FW_PASS); 2496 } 2497 2498 f = args->rule->next_rule; 2499 if (f == NULL) 2500 f = lookup_next_rule(args->rule, 0); 2501 } else { 2502 /* 2503 * Find the starting rule. It can be either the first 2504 * one, or the one after divert_rule if asked so. 2505 */ 2506 int skipto = mtag ? divert_cookie(mtag) : 0; 2507 2508 f = chain->rules; 2509 if (args->eh == NULL && skipto != 0) { 2510 if (skipto >= IPFW_DEFAULT_RULE) { 2511 IPFW_RUNLOCK(chain); 2512 return (IP_FW_DENY); /* invalid */ 2513 } 2514 while (f && f->rulenum <= skipto) 2515 f = f->next; 2516 if (f == NULL) { /* drop packet */ 2517 IPFW_RUNLOCK(chain); 2518 return (IP_FW_DENY); 2519 } 2520 } 2521 } 2522 /* reset divert rule to avoid confusion later */ 2523 if (mtag) { 2524 divinput_flags = divert_info(mtag) & 2525 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2526 m_tag_delete(m, mtag); 2527 } 2528 2529 /* 2530 * Now scan the rules, and parse microinstructions for each rule. 2531 */ 2532 for (; f; f = f->next) { 2533 ipfw_insn *cmd; 2534 uint32_t tablearg = 0; 2535 int l, cmdlen, skip_or; /* skip rest of OR block */ 2536 2537again: 2538 if (V_set_disable & (1 << f->set) ) 2539 continue; 2540 2541 skip_or = 0; 2542 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2543 l -= cmdlen, cmd += cmdlen) { 2544 int match; 2545 2546 /* 2547 * check_body is a jump target used when we find a 2548 * CHECK_STATE, and need to jump to the body of 2549 * the target rule. 2550 */ 2551 2552check_body: 2553 cmdlen = F_LEN(cmd); 2554 /* 2555 * An OR block (insn_1 || .. || insn_n) has the 2556 * F_OR bit set in all but the last instruction. 2557 * The first match will set "skip_or", and cause 2558 * the following instructions to be skipped until 2559 * past the one with the F_OR bit clear. 2560 */ 2561 if (skip_or) { /* skip this instruction */ 2562 if ((cmd->len & F_OR) == 0) 2563 skip_or = 0; /* next one is good */ 2564 continue; 2565 } 2566 match = 0; /* set to 1 if we succeed */ 2567 2568 switch (cmd->opcode) { 2569 /* 2570 * The first set of opcodes compares the packet's 2571 * fields with some pattern, setting 'match' if a 2572 * match is found. At the end of the loop there is 2573 * logic to deal with F_NOT and F_OR flags associated 2574 * with the opcode. 2575 */ 2576 case O_NOP: 2577 match = 1; 2578 break; 2579 2580 case O_FORWARD_MAC: 2581 printf("ipfw: opcode %d unimplemented\n", 2582 cmd->opcode); 2583 break; 2584 2585 case O_GID: 2586 case O_UID: 2587 case O_JAIL: 2588 /* 2589 * We only check offset == 0 && proto != 0, 2590 * as this ensures that we have a 2591 * packet with the ports info. 2592 */ 2593 if (offset!=0) 2594 break; 2595 if (is_ipv6) /* XXX to be fixed later */ 2596 break; 2597 if (proto == IPPROTO_TCP || 2598 proto == IPPROTO_UDP) 2599 match = check_uidgid( 2600 (ipfw_insn_u32 *)cmd, 2601 proto, oif, 2602 dst_ip, dst_port, 2603 src_ip, src_port, &fw_ugid_cache, 2604 &ugid_lookup, args->inp); 2605 break; 2606 2607 case O_RECV: 2608 match = iface_match(m->m_pkthdr.rcvif, 2609 (ipfw_insn_if *)cmd); 2610 break; 2611 2612 case O_XMIT: 2613 match = iface_match(oif, (ipfw_insn_if *)cmd); 2614 break; 2615 2616 case O_VIA: 2617 match = iface_match(oif ? oif : 2618 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2619 break; 2620 2621 case O_MACADDR2: 2622 if (args->eh != NULL) { /* have MAC header */ 2623 u_int32_t *want = (u_int32_t *) 2624 ((ipfw_insn_mac *)cmd)->addr; 2625 u_int32_t *mask = (u_int32_t *) 2626 ((ipfw_insn_mac *)cmd)->mask; 2627 u_int32_t *hdr = (u_int32_t *)args->eh; 2628 2629 match = 2630 ( want[0] == (hdr[0] & mask[0]) && 2631 want[1] == (hdr[1] & mask[1]) && 2632 want[2] == (hdr[2] & mask[2]) ); 2633 } 2634 break; 2635 2636 case O_MAC_TYPE: 2637 if (args->eh != NULL) { 2638 u_int16_t *p = 2639 ((ipfw_insn_u16 *)cmd)->ports; 2640 int i; 2641 2642 for (i = cmdlen - 1; !match && i>0; 2643 i--, p += 2) 2644 match = (etype >= p[0] && 2645 etype <= p[1]); 2646 } 2647 break; 2648 2649 case O_FRAG: 2650 match = (offset != 0); 2651 break; 2652 2653 case O_IN: /* "out" is "not in" */ 2654 match = (oif == NULL); 2655 break; 2656 2657 case O_LAYER2: 2658 match = (args->eh != NULL); 2659 break; 2660 2661 case O_DIVERTED: 2662 match = (cmd->arg1 & 1 && divinput_flags & 2663 IP_FW_DIVERT_LOOPBACK_FLAG) || 2664 (cmd->arg1 & 2 && divinput_flags & 2665 IP_FW_DIVERT_OUTPUT_FLAG); 2666 break; 2667 2668 case O_PROTO: 2669 /* 2670 * We do not allow an arg of 0 so the 2671 * check of "proto" only suffices. 2672 */ 2673 match = (proto == cmd->arg1); 2674 break; 2675 2676 case O_IP_SRC: 2677 match = is_ipv4 && 2678 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2679 src_ip.s_addr); 2680 break; 2681 2682 case O_IP_SRC_LOOKUP: 2683 case O_IP_DST_LOOKUP: 2684 if (is_ipv4) { 2685 uint32_t a = 2686 (cmd->opcode == O_IP_DST_LOOKUP) ? 2687 dst_ip.s_addr : src_ip.s_addr; 2688 uint32_t v; 2689 2690 match = lookup_table(chain, cmd->arg1, a, 2691 &v); 2692 if (!match) 2693 break; 2694 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2695 match = 2696 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2697 else 2698 tablearg = v; 2699 } 2700 break; 2701 2702 case O_IP_SRC_MASK: 2703 case O_IP_DST_MASK: 2704 if (is_ipv4) { 2705 uint32_t a = 2706 (cmd->opcode == O_IP_DST_MASK) ? 2707 dst_ip.s_addr : src_ip.s_addr; 2708 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2709 int i = cmdlen-1; 2710 2711 for (; !match && i>0; i-= 2, p+= 2) 2712 match = (p[0] == (a & p[1])); 2713 } 2714 break; 2715 2716 case O_IP_SRC_ME: 2717 if (is_ipv4) { 2718 struct ifnet *tif; 2719 2720 INADDR_TO_IFP(src_ip, tif); 2721 match = (tif != NULL); 2722 } 2723 break; 2724 2725 case O_IP_DST_SET: 2726 case O_IP_SRC_SET: 2727 if (is_ipv4) { 2728 u_int32_t *d = (u_int32_t *)(cmd+1); 2729 u_int32_t addr = 2730 cmd->opcode == O_IP_DST_SET ? 2731 args->f_id.dst_ip : 2732 args->f_id.src_ip; 2733 2734 if (addr < d[0]) 2735 break; 2736 addr -= d[0]; /* subtract base */ 2737 match = (addr < cmd->arg1) && 2738 ( d[ 1 + (addr>>5)] & 2739 (1<<(addr & 0x1f)) ); 2740 } 2741 break; 2742 2743 case O_IP_DST: 2744 match = is_ipv4 && 2745 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2746 dst_ip.s_addr); 2747 break; 2748 2749 case O_IP_DST_ME: 2750 if (is_ipv4) { 2751 struct ifnet *tif; 2752 2753 INADDR_TO_IFP(dst_ip, tif); 2754 match = (tif != NULL); 2755 } 2756 break; 2757 2758 case O_IP_SRCPORT: 2759 case O_IP_DSTPORT: 2760 /* 2761 * offset == 0 && proto != 0 is enough 2762 * to guarantee that we have a 2763 * packet with port info. 2764 */ 2765 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2766 && offset == 0) { 2767 u_int16_t x = 2768 (cmd->opcode == O_IP_SRCPORT) ? 2769 src_port : dst_port ; 2770 u_int16_t *p = 2771 ((ipfw_insn_u16 *)cmd)->ports; 2772 int i; 2773 2774 for (i = cmdlen - 1; !match && i>0; 2775 i--, p += 2) 2776 match = (x>=p[0] && x<=p[1]); 2777 } 2778 break; 2779 2780 case O_ICMPTYPE: 2781 match = (offset == 0 && proto==IPPROTO_ICMP && 2782 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2783 break; 2784 2785#ifdef INET6 2786 case O_ICMP6TYPE: 2787 match = is_ipv6 && offset == 0 && 2788 proto==IPPROTO_ICMPV6 && 2789 icmp6type_match( 2790 ICMP6(ulp)->icmp6_type, 2791 (ipfw_insn_u32 *)cmd); 2792 break; 2793#endif /* INET6 */ 2794 2795 case O_IPOPT: 2796 match = (is_ipv4 && 2797 ipopts_match(ip, cmd) ); 2798 break; 2799 2800 case O_IPVER: 2801 match = (is_ipv4 && 2802 cmd->arg1 == ip->ip_v); 2803 break; 2804 2805 case O_IPID: 2806 case O_IPLEN: 2807 case O_IPTTL: 2808 if (is_ipv4) { /* only for IP packets */ 2809 uint16_t x; 2810 uint16_t *p; 2811 int i; 2812 2813 if (cmd->opcode == O_IPLEN) 2814 x = ip_len; 2815 else if (cmd->opcode == O_IPTTL) 2816 x = ip->ip_ttl; 2817 else /* must be IPID */ 2818 x = ntohs(ip->ip_id); 2819 if (cmdlen == 1) { 2820 match = (cmd->arg1 == x); 2821 break; 2822 } 2823 /* otherwise we have ranges */ 2824 p = ((ipfw_insn_u16 *)cmd)->ports; 2825 i = cmdlen - 1; 2826 for (; !match && i>0; i--, p += 2) 2827 match = (x >= p[0] && x <= p[1]); 2828 } 2829 break; 2830 2831 case O_IPPRECEDENCE: 2832 match = (is_ipv4 && 2833 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2834 break; 2835 2836 case O_IPTOS: 2837 match = (is_ipv4 && 2838 flags_match(cmd, ip->ip_tos)); 2839 break; 2840 2841 case O_TCPDATALEN: 2842 if (proto == IPPROTO_TCP && offset == 0) { 2843 struct tcphdr *tcp; 2844 uint16_t x; 2845 uint16_t *p; 2846 int i; 2847 2848 tcp = TCP(ulp); 2849 x = ip_len - 2850 ((ip->ip_hl + tcp->th_off) << 2); 2851 if (cmdlen == 1) { 2852 match = (cmd->arg1 == x); 2853 break; 2854 } 2855 /* otherwise we have ranges */ 2856 p = ((ipfw_insn_u16 *)cmd)->ports; 2857 i = cmdlen - 1; 2858 for (; !match && i>0; i--, p += 2) 2859 match = (x >= p[0] && x <= p[1]); 2860 } 2861 break; 2862 2863 case O_TCPFLAGS: 2864 match = (proto == IPPROTO_TCP && offset == 0 && 2865 flags_match(cmd, TCP(ulp)->th_flags)); 2866 break; 2867 2868 case O_TCPOPTS: 2869 match = (proto == IPPROTO_TCP && offset == 0 && 2870 tcpopts_match(TCP(ulp), cmd)); 2871 break; 2872 2873 case O_TCPSEQ: 2874 match = (proto == IPPROTO_TCP && offset == 0 && 2875 ((ipfw_insn_u32 *)cmd)->d[0] == 2876 TCP(ulp)->th_seq); 2877 break; 2878 2879 case O_TCPACK: 2880 match = (proto == IPPROTO_TCP && offset == 0 && 2881 ((ipfw_insn_u32 *)cmd)->d[0] == 2882 TCP(ulp)->th_ack); 2883 break; 2884 2885 case O_TCPWIN: 2886 match = (proto == IPPROTO_TCP && offset == 0 && 2887 cmd->arg1 == TCP(ulp)->th_win); 2888 break; 2889 2890 case O_ESTAB: 2891 /* reject packets which have SYN only */ 2892 /* XXX should i also check for TH_ACK ? */ 2893 match = (proto == IPPROTO_TCP && offset == 0 && 2894 (TCP(ulp)->th_flags & 2895 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2896 break; 2897 2898 case O_ALTQ: { 2899 struct pf_mtag *at; 2900 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2901 2902 match = 1; 2903 at = pf_find_mtag(m); 2904 if (at != NULL && at->qid != 0) 2905 break; 2906 at = pf_get_mtag(m); 2907 if (at == NULL) { 2908 /* 2909 * Let the packet fall back to the 2910 * default ALTQ. 2911 */ 2912 break; 2913 } 2914 at->qid = altq->qid; 2915 if (is_ipv4) 2916 at->af = AF_INET; 2917 else 2918 at->af = AF_LINK; 2919 at->hdr = ip; 2920 break; 2921 } 2922 2923 case O_LOG: 2924 if (V_fw_verbose) 2925 ipfw_log(f, hlen, args, m, 2926 oif, offset, tablearg, ip); 2927 match = 1; 2928 break; 2929 2930 case O_PROB: 2931 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2932 break; 2933 2934 case O_VERREVPATH: 2935 /* Outgoing packets automatically pass/match */ 2936 match = ((oif != NULL) || 2937 (m->m_pkthdr.rcvif == NULL) || 2938 ( 2939#ifdef INET6 2940 is_ipv6 ? 2941 verify_path6(&(args->f_id.src_ip6), 2942 m->m_pkthdr.rcvif) : 2943#endif 2944 verify_path(src_ip, m->m_pkthdr.rcvif, 2945 args->f_id.fib))); 2946 break; 2947 2948 case O_VERSRCREACH: 2949 /* Outgoing packets automatically pass/match */ 2950 match = (hlen > 0 && ((oif != NULL) || 2951#ifdef INET6 2952 is_ipv6 ? 2953 verify_path6(&(args->f_id.src_ip6), 2954 NULL) : 2955#endif 2956 verify_path(src_ip, NULL, args->f_id.fib))); 2957 break; 2958 2959 case O_ANTISPOOF: 2960 /* Outgoing packets automatically pass/match */ 2961 if (oif == NULL && hlen > 0 && 2962 ( (is_ipv4 && in_localaddr(src_ip)) 2963#ifdef INET6 2964 || (is_ipv6 && 2965 in6_localaddr(&(args->f_id.src_ip6))) 2966#endif 2967 )) 2968 match = 2969#ifdef INET6 2970 is_ipv6 ? verify_path6( 2971 &(args->f_id.src_ip6), 2972 m->m_pkthdr.rcvif) : 2973#endif 2974 verify_path(src_ip, 2975 m->m_pkthdr.rcvif, 2976 args->f_id.fib); 2977 else 2978 match = 1; 2979 break; 2980 2981 case O_IPSEC: 2982#ifdef IPSEC 2983 match = (m_tag_find(m, 2984 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2985#endif 2986 /* otherwise no match */ 2987 break; 2988 2989#ifdef INET6 2990 case O_IP6_SRC: 2991 match = is_ipv6 && 2992 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2993 &((ipfw_insn_ip6 *)cmd)->addr6); 2994 break; 2995 2996 case O_IP6_DST: 2997 match = is_ipv6 && 2998 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2999 &((ipfw_insn_ip6 *)cmd)->addr6); 3000 break; 3001 case O_IP6_SRC_MASK: 3002 case O_IP6_DST_MASK: 3003 if (is_ipv6) { 3004 int i = cmdlen - 1; 3005 struct in6_addr p; 3006 struct in6_addr *d = 3007 &((ipfw_insn_ip6 *)cmd)->addr6; 3008 3009 for (; !match && i > 0; d += 2, 3010 i -= F_INSN_SIZE(struct in6_addr) 3011 * 2) { 3012 p = (cmd->opcode == 3013 O_IP6_SRC_MASK) ? 3014 args->f_id.src_ip6: 3015 args->f_id.dst_ip6; 3016 APPLY_MASK(&p, &d[1]); 3017 match = 3018 IN6_ARE_ADDR_EQUAL(&d[0], 3019 &p); 3020 } 3021 } 3022 break; 3023 3024 case O_IP6_SRC_ME: 3025 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3026 break; 3027 3028 case O_IP6_DST_ME: 3029 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3030 break; 3031 3032 case O_FLOW6ID: 3033 match = is_ipv6 && 3034 flow6id_match(args->f_id.flow_id6, 3035 (ipfw_insn_u32 *) cmd); 3036 break; 3037 3038 case O_EXT_HDR: 3039 match = is_ipv6 && 3040 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3041 break; 3042 3043 case O_IP6: 3044 match = is_ipv6; 3045 break; 3046#endif 3047 3048 case O_IP4: 3049 match = is_ipv4; 3050 break; 3051 3052 case O_TAG: { 3053 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3054 tablearg : cmd->arg1; 3055 3056 /* Packet is already tagged with this tag? */ 3057 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3058 3059 /* We have `untag' action when F_NOT flag is 3060 * present. And we must remove this mtag from 3061 * mbuf and reset `match' to zero (`match' will 3062 * be inversed later). 3063 * Otherwise we should allocate new mtag and 3064 * push it into mbuf. 3065 */ 3066 if (cmd->len & F_NOT) { /* `untag' action */ 3067 if (mtag != NULL) 3068 m_tag_delete(m, mtag); 3069 } else if (mtag == NULL) { 3070 if ((mtag = m_tag_alloc(MTAG_IPFW, 3071 tag, 0, M_NOWAIT)) != NULL) 3072 m_tag_prepend(m, mtag); 3073 } 3074 match = (cmd->len & F_NOT) ? 0: 1; 3075 break; 3076 } 3077 3078 case O_FIB: /* try match the specified fib */ 3079 if (args->f_id.fib == cmd->arg1) 3080 match = 1; 3081 break; 3082 3083 case O_TAGGED: { 3084 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3085 tablearg : cmd->arg1; 3086 3087 if (cmdlen == 1) { 3088 match = m_tag_locate(m, MTAG_IPFW, 3089 tag, NULL) != NULL; 3090 break; 3091 } 3092 3093 /* we have ranges */ 3094 for (mtag = m_tag_first(m); 3095 mtag != NULL && !match; 3096 mtag = m_tag_next(m, mtag)) { 3097 uint16_t *p; 3098 int i; 3099 3100 if (mtag->m_tag_cookie != MTAG_IPFW) 3101 continue; 3102 3103 p = ((ipfw_insn_u16 *)cmd)->ports; 3104 i = cmdlen - 1; 3105 for(; !match && i > 0; i--, p += 2) 3106 match = 3107 mtag->m_tag_id >= p[0] && 3108 mtag->m_tag_id <= p[1]; 3109 } 3110 break; 3111 } 3112 3113 /* 3114 * The second set of opcodes represents 'actions', 3115 * i.e. the terminal part of a rule once the packet 3116 * matches all previous patterns. 3117 * Typically there is only one action for each rule, 3118 * and the opcode is stored at the end of the rule 3119 * (but there are exceptions -- see below). 3120 * 3121 * In general, here we set retval and terminate the 3122 * outer loop (would be a 'break 3' in some language, 3123 * but we need to do a 'goto done'). 3124 * 3125 * Exceptions: 3126 * O_COUNT and O_SKIPTO actions: 3127 * instead of terminating, we jump to the next rule 3128 * ('goto next_rule', equivalent to a 'break 2'), 3129 * or to the SKIPTO target ('goto again' after 3130 * having set f, cmd and l), respectively. 3131 * 3132 * O_TAG, O_LOG and O_ALTQ action parameters: 3133 * perform some action and set match = 1; 3134 * 3135 * O_LIMIT and O_KEEP_STATE: these opcodes are 3136 * not real 'actions', and are stored right 3137 * before the 'action' part of the rule. 3138 * These opcodes try to install an entry in the 3139 * state tables; if successful, we continue with 3140 * the next opcode (match=1; break;), otherwise 3141 * the packet * must be dropped 3142 * ('goto done' after setting retval); 3143 * 3144 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3145 * cause a lookup of the state table, and a jump 3146 * to the 'action' part of the parent rule 3147 * ('goto check_body') if an entry is found, or 3148 * (CHECK_STATE only) a jump to the next rule if 3149 * the entry is not found ('goto next_rule'). 3150 * The result of the lookup is cached to make 3151 * further instances of these opcodes are 3152 * effectively NOPs. 3153 */ 3154 case O_LIMIT: 3155 case O_KEEP_STATE: 3156 if (install_state(f, 3157 (ipfw_insn_limit *)cmd, args, tablearg)) { 3158 retval = IP_FW_DENY; 3159 goto done; /* error/limit violation */ 3160 } 3161 match = 1; 3162 break; 3163 3164 case O_PROBE_STATE: 3165 case O_CHECK_STATE: 3166 /* 3167 * dynamic rules are checked at the first 3168 * keep-state or check-state occurrence, 3169 * with the result being stored in dyn_dir. 3170 * The compiler introduces a PROBE_STATE 3171 * instruction for us when we have a 3172 * KEEP_STATE (because PROBE_STATE needs 3173 * to be run first). 3174 */ 3175 if (dyn_dir == MATCH_UNKNOWN && 3176 (q = lookup_dyn_rule(&args->f_id, 3177 &dyn_dir, proto == IPPROTO_TCP ? 3178 TCP(ulp) : NULL)) 3179 != NULL) { 3180 /* 3181 * Found dynamic entry, update stats 3182 * and jump to the 'action' part of 3183 * the parent rule. 3184 */ 3185 q->pcnt++; 3186 q->bcnt += pktlen; 3187 f = q->rule; 3188 cmd = ACTION_PTR(f); 3189 l = f->cmd_len - f->act_ofs; 3190 IPFW_DYN_UNLOCK(); 3191 goto check_body; 3192 } 3193 /* 3194 * Dynamic entry not found. If CHECK_STATE, 3195 * skip to next rule, if PROBE_STATE just 3196 * ignore and continue with next opcode. 3197 */ 3198 if (cmd->opcode == O_CHECK_STATE) 3199 goto next_rule; 3200 match = 1; 3201 break; 3202 3203 case O_ACCEPT: 3204 retval = 0; /* accept */ 3205 goto done; 3206 3207 case O_PIPE: 3208 case O_QUEUE: 3209 args->rule = f; /* report matching rule */ 3210 if (cmd->arg1 == IP_FW_TABLEARG) 3211 args->cookie = tablearg; 3212 else 3213 args->cookie = cmd->arg1; 3214 retval = IP_FW_DUMMYNET; 3215 goto done; 3216 3217 case O_DIVERT: 3218 case O_TEE: { 3219 struct divert_tag *dt; 3220 3221 if (args->eh) /* not on layer 2 */ 3222 break; 3223 mtag = m_tag_get(PACKET_TAG_DIVERT, 3224 sizeof(struct divert_tag), 3225 M_NOWAIT); 3226 if (mtag == NULL) { 3227 /* XXX statistic */ 3228 /* drop packet */ 3229 IPFW_RUNLOCK(chain); 3230 return (IP_FW_DENY); 3231 } 3232 dt = (struct divert_tag *)(mtag+1); 3233 dt->cookie = f->rulenum; 3234 if (cmd->arg1 == IP_FW_TABLEARG) 3235 dt->info = tablearg; 3236 else 3237 dt->info = cmd->arg1; 3238 m_tag_prepend(m, mtag); 3239 retval = (cmd->opcode == O_DIVERT) ? 3240 IP_FW_DIVERT : IP_FW_TEE; 3241 goto done; 3242 } 3243 case O_COUNT: 3244 case O_SKIPTO: 3245 f->pcnt++; /* update stats */ 3246 f->bcnt += pktlen; 3247 f->timestamp = time_uptime; 3248 if (cmd->opcode == O_COUNT) 3249 goto next_rule; 3250 /* handle skipto */ 3251 if (cmd->arg1 == IP_FW_TABLEARG) { 3252 f = lookup_next_rule(f, tablearg); 3253 } else { 3254 if (f->next_rule == NULL) 3255 lookup_next_rule(f, 0); 3256 f = f->next_rule; 3257 } 3258 goto again; 3259 3260 case O_REJECT: 3261 /* 3262 * Drop the packet and send a reject notice 3263 * if the packet is not ICMP (or is an ICMP 3264 * query), and it is not multicast/broadcast. 3265 */ 3266 if (hlen > 0 && is_ipv4 && offset == 0 && 3267 (proto != IPPROTO_ICMP || 3268 is_icmp_query(ICMP(ulp))) && 3269 !(m->m_flags & (M_BCAST|M_MCAST)) && 3270 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3271 send_reject(args, cmd->arg1, ip_len, ip); 3272 m = args->m; 3273 } 3274 /* FALLTHROUGH */ 3275#ifdef INET6 3276 case O_UNREACH6: 3277 if (hlen > 0 && is_ipv6 && 3278 ((offset & IP6F_OFF_MASK) == 0) && 3279 (proto != IPPROTO_ICMPV6 || 3280 (is_icmp6_query(args->f_id.flags) == 1)) && 3281 !(m->m_flags & (M_BCAST|M_MCAST)) && 3282 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3283 send_reject6( 3284 args, cmd->arg1, hlen, 3285 (struct ip6_hdr *)ip); 3286 m = args->m; 3287 } 3288 /* FALLTHROUGH */ 3289#endif 3290 case O_DENY: 3291 retval = IP_FW_DENY; 3292 goto done; 3293 3294 case O_FORWARD_IP: { 3295 struct sockaddr_in *sa; 3296 sa = &(((ipfw_insn_sa *)cmd)->sa); 3297 if (args->eh) /* not valid on layer2 pkts */ 3298 break; 3299 if (!q || dyn_dir == MATCH_FORWARD) { 3300 if (sa->sin_addr.s_addr == INADDR_ANY) { 3301 bcopy(sa, &args->hopstore, 3302 sizeof(*sa)); 3303 args->hopstore.sin_addr.s_addr = 3304 htonl(tablearg); 3305 args->next_hop = 3306 &args->hopstore; 3307 } else { 3308 args->next_hop = sa; 3309 } 3310 } 3311 retval = IP_FW_PASS; 3312 } 3313 goto done; 3314 3315 case O_NETGRAPH: 3316 case O_NGTEE: 3317 args->rule = f; /* report matching rule */ 3318 if (cmd->arg1 == IP_FW_TABLEARG) 3319 args->cookie = tablearg; 3320 else 3321 args->cookie = cmd->arg1; 3322 retval = (cmd->opcode == O_NETGRAPH) ? 3323 IP_FW_NETGRAPH : IP_FW_NGTEE; 3324 goto done; 3325 3326 case O_SETFIB: 3327 f->pcnt++; /* update stats */ 3328 f->bcnt += pktlen; 3329 f->timestamp = time_uptime; 3330 M_SETFIB(m, cmd->arg1); 3331 args->f_id.fib = cmd->arg1; 3332 goto next_rule; 3333 3334 case O_NAT: { 3335 struct cfg_nat *t; 3336 int nat_id; 3337 3338 if (IPFW_NAT_LOADED) { 3339 args->rule = f; /* Report matching rule. */ 3340 t = ((ipfw_insn_nat *)cmd)->nat; 3341 if (t == NULL) { 3342 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 3343 tablearg : cmd->arg1; 3344 LOOKUP_NAT(V_layer3_chain, nat_id, t); 3345 if (t == NULL) { 3346 retval = IP_FW_DENY; 3347 goto done; 3348 } 3349 if (cmd->arg1 != IP_FW_TABLEARG) 3350 ((ipfw_insn_nat *)cmd)->nat = t; 3351 } 3352 retval = ipfw_nat_ptr(args, t, m); 3353 } else 3354 retval = IP_FW_DENY; 3355 goto done; 3356 } 3357 3358 default: 3359 panic("-- unknown opcode %d\n", cmd->opcode); 3360 } /* end of switch() on opcodes */ 3361 3362 if (cmd->len & F_NOT) 3363 match = !match; 3364 3365 if (match) { 3366 if (cmd->len & F_OR) 3367 skip_or = 1; 3368 } else { 3369 if (!(cmd->len & F_OR)) /* not an OR block, */ 3370 break; /* try next rule */ 3371 } 3372 3373 } /* end of inner for, scan opcodes */ 3374 3375next_rule:; /* try next rule */ 3376 3377 } /* end of outer for, scan rules */ 3378 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3379 IPFW_RUNLOCK(chain); 3380 return (IP_FW_DENY); 3381 3382done: 3383 /* Update statistics */ 3384 f->pcnt++; 3385 f->bcnt += pktlen; 3386 f->timestamp = time_uptime; 3387 IPFW_RUNLOCK(chain); 3388 return (retval); 3389 3390pullup_failed: 3391 if (V_fw_verbose) 3392 printf("ipfw: pullup failed\n"); 3393 return (IP_FW_DENY); 3394} 3395 3396/* 3397 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3398 * These will be reconstructed on the fly as packets are matched. 3399 */ 3400static void 3401flush_rule_ptrs(struct ip_fw_chain *chain) 3402{ 3403 struct ip_fw *rule; 3404 3405 IPFW_WLOCK_ASSERT(chain); 3406 3407 for (rule = chain->rules; rule; rule = rule->next) 3408 rule->next_rule = NULL; 3409} 3410 3411/* 3412 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3413 * possibly create a rule number and add the rule to the list. 3414 * Update the rule_number in the input struct so the caller knows it as well. 3415 */ 3416static int 3417add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3418{ 3419 INIT_VNET_IPFW(curvnet); 3420 struct ip_fw *rule, *f, *prev; 3421 int l = RULESIZE(input_rule); 3422 3423 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3424 return (EINVAL); 3425 3426 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3427 if (rule == NULL) 3428 return (ENOSPC); 3429 3430 bcopy(input_rule, rule, l); 3431 3432 rule->next = NULL; 3433 rule->next_rule = NULL; 3434 3435 rule->pcnt = 0; 3436 rule->bcnt = 0; 3437 rule->timestamp = 0; 3438 3439 IPFW_WLOCK(chain); 3440 3441 if (chain->rules == NULL) { /* default rule */ 3442 chain->rules = rule; 3443 goto done; 3444 } 3445 3446 /* 3447 * If rulenum is 0, find highest numbered rule before the 3448 * default rule, and add autoinc_step 3449 */ 3450 if (V_autoinc_step < 1) 3451 V_autoinc_step = 1; 3452 else if (V_autoinc_step > 1000) 3453 V_autoinc_step = 1000; 3454 if (rule->rulenum == 0) { 3455 /* 3456 * locate the highest numbered rule before default 3457 */ 3458 for (f = chain->rules; f; f = f->next) { 3459 if (f->rulenum == IPFW_DEFAULT_RULE) 3460 break; 3461 rule->rulenum = f->rulenum; 3462 } 3463 if (rule->rulenum < IPFW_DEFAULT_RULE - V_autoinc_step) 3464 rule->rulenum += V_autoinc_step; 3465 input_rule->rulenum = rule->rulenum; 3466 } 3467 3468 /* 3469 * Now insert the new rule in the right place in the sorted list. 3470 */ 3471 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3472 if (f->rulenum > rule->rulenum) { /* found the location */ 3473 if (prev) { 3474 rule->next = f; 3475 prev->next = rule; 3476 } else { /* head insert */ 3477 rule->next = chain->rules; 3478 chain->rules = rule; 3479 } 3480 break; 3481 } 3482 } 3483 flush_rule_ptrs(chain); 3484done: 3485 V_static_count++; 3486 V_static_len += l; 3487 IPFW_WUNLOCK(chain); 3488 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3489 rule->rulenum, V_static_count);) 3490 return (0); 3491} 3492 3493/** 3494 * Remove a static rule (including derived * dynamic rules) 3495 * and place it on the ``reap list'' for later reclamation. 3496 * The caller is in charge of clearing rule pointers to avoid 3497 * dangling pointers. 3498 * @return a pointer to the next entry. 3499 * Arguments are not checked, so they better be correct. 3500 */ 3501static struct ip_fw * 3502remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 3503 struct ip_fw *prev) 3504{ 3505 INIT_VNET_IPFW(curvnet); 3506 struct ip_fw *n; 3507 int l = RULESIZE(rule); 3508 3509 IPFW_WLOCK_ASSERT(chain); 3510 3511 n = rule->next; 3512 IPFW_DYN_LOCK(); 3513 remove_dyn_rule(rule, NULL /* force removal */); 3514 IPFW_DYN_UNLOCK(); 3515 if (prev == NULL) 3516 chain->rules = n; 3517 else 3518 prev->next = n; 3519 V_static_count--; 3520 V_static_len -= l; 3521 3522 rule->next = chain->reap; 3523 chain->reap = rule; 3524 3525 return n; 3526} 3527 3528/** 3529 * Reclaim storage associated with a list of rules. This is 3530 * typically the list created using remove_rule. 3531 */ 3532static void 3533reap_rules(struct ip_fw *head) 3534{ 3535 struct ip_fw *rule; 3536 3537 while ((rule = head) != NULL) { 3538 head = head->next; 3539 if (DUMMYNET_LOADED) 3540 ip_dn_ruledel_ptr(rule); 3541 free(rule, M_IPFW); 3542 } 3543} 3544 3545/* 3546 * Remove all rules from a chain (except rules in set RESVD_SET 3547 * unless kill_default = 1). The caller is responsible for 3548 * reclaiming storage for the rules left in chain->reap. 3549 */ 3550static void 3551free_chain(struct ip_fw_chain *chain, int kill_default) 3552{ 3553 struct ip_fw *prev, *rule; 3554 3555 IPFW_WLOCK_ASSERT(chain); 3556 3557 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3558 for (prev = NULL, rule = chain->rules; rule ; ) 3559 if (kill_default || rule->set != RESVD_SET) 3560 rule = remove_rule(chain, rule, prev); 3561 else { 3562 prev = rule; 3563 rule = rule->next; 3564 } 3565} 3566 3567/** 3568 * Remove all rules with given number, and also do set manipulation. 3569 * Assumes chain != NULL && *chain != NULL. 3570 * 3571 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3572 * the next 8 bits are the new set, the top 8 bits are the command: 3573 * 3574 * 0 delete rules with given number 3575 * 1 delete rules with given set number 3576 * 2 move rules with given number to new set 3577 * 3 move rules with given set number to new set 3578 * 4 swap sets with given numbers 3579 * 5 delete rules with given number and with given set number 3580 */ 3581static int 3582del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3583{ 3584 struct ip_fw *prev = NULL, *rule; 3585 u_int16_t rulenum; /* rule or old_set */ 3586 u_int8_t cmd, new_set; 3587 3588 rulenum = arg & 0xffff; 3589 cmd = (arg >> 24) & 0xff; 3590 new_set = (arg >> 16) & 0xff; 3591 3592 if (cmd > 5 || new_set > RESVD_SET) 3593 return EINVAL; 3594 if (cmd == 0 || cmd == 2 || cmd == 5) { 3595 if (rulenum >= IPFW_DEFAULT_RULE) 3596 return EINVAL; 3597 } else { 3598 if (rulenum > RESVD_SET) /* old_set */ 3599 return EINVAL; 3600 } 3601 3602 IPFW_WLOCK(chain); 3603 rule = chain->rules; 3604 chain->reap = NULL; 3605 switch (cmd) { 3606 case 0: /* delete rules with given number */ 3607 /* 3608 * locate first rule to delete 3609 */ 3610 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3611 ; 3612 if (rule->rulenum != rulenum) { 3613 IPFW_WUNLOCK(chain); 3614 return EINVAL; 3615 } 3616 3617 /* 3618 * flush pointers outside the loop, then delete all matching 3619 * rules. prev remains the same throughout the cycle. 3620 */ 3621 flush_rule_ptrs(chain); 3622 while (rule->rulenum == rulenum) 3623 rule = remove_rule(chain, rule, prev); 3624 break; 3625 3626 case 1: /* delete all rules with given set number */ 3627 flush_rule_ptrs(chain); 3628 rule = chain->rules; 3629 while (rule->rulenum < IPFW_DEFAULT_RULE) 3630 if (rule->set == rulenum) 3631 rule = remove_rule(chain, rule, prev); 3632 else { 3633 prev = rule; 3634 rule = rule->next; 3635 } 3636 break; 3637 3638 case 2: /* move rules with given number to new set */ 3639 rule = chain->rules; 3640 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3641 if (rule->rulenum == rulenum) 3642 rule->set = new_set; 3643 break; 3644 3645 case 3: /* move rules with given set number to new set */ 3646 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3647 if (rule->set == rulenum) 3648 rule->set = new_set; 3649 break; 3650 3651 case 4: /* swap two sets */ 3652 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3653 if (rule->set == rulenum) 3654 rule->set = new_set; 3655 else if (rule->set == new_set) 3656 rule->set = rulenum; 3657 break; 3658 case 5: /* delete rules with given number and with given set number. 3659 * rulenum - given rule number; 3660 * new_set - given set number. 3661 */ 3662 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3663 ; 3664 if (rule->rulenum != rulenum) { 3665 IPFW_WUNLOCK(chain); 3666 return (EINVAL); 3667 } 3668 flush_rule_ptrs(chain); 3669 while (rule->rulenum == rulenum) { 3670 if (rule->set == new_set) 3671 rule = remove_rule(chain, rule, prev); 3672 else { 3673 prev = rule; 3674 rule = rule->next; 3675 } 3676 } 3677 } 3678 /* 3679 * Look for rules to reclaim. We grab the list before 3680 * releasing the lock then reclaim them w/o the lock to 3681 * avoid a LOR with dummynet. 3682 */ 3683 rule = chain->reap; 3684 chain->reap = NULL; 3685 IPFW_WUNLOCK(chain); 3686 if (rule) 3687 reap_rules(rule); 3688 return 0; 3689} 3690 3691/* 3692 * Clear counters for a specific rule. 3693 * The enclosing "table" is assumed locked. 3694 */ 3695static void 3696clear_counters(struct ip_fw *rule, int log_only) 3697{ 3698 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3699 3700 if (log_only == 0) { 3701 rule->bcnt = rule->pcnt = 0; 3702 rule->timestamp = 0; 3703 } 3704 if (l->o.opcode == O_LOG) 3705 l->log_left = l->max_log; 3706} 3707 3708/** 3709 * Reset some or all counters on firewall rules. 3710 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 3711 * the next 8 bits are the set number, the top 8 bits are the command: 3712 * 0 work with rules from all set's; 3713 * 1 work with rules only from specified set. 3714 * Specified rule number is zero if we want to clear all entries. 3715 * log_only is 1 if we only want to reset logs, zero otherwise. 3716 */ 3717static int 3718zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 3719{ 3720 INIT_VNET_IPFW(curvnet); 3721 struct ip_fw *rule; 3722 char *msg; 3723 3724 uint16_t rulenum = arg & 0xffff; 3725 uint8_t set = (arg >> 16) & 0xff; 3726 uint8_t cmd = (arg >> 24) & 0xff; 3727 3728 if (cmd > 1) 3729 return (EINVAL); 3730 if (cmd == 1 && set > RESVD_SET) 3731 return (EINVAL); 3732 3733 IPFW_WLOCK(chain); 3734 if (rulenum == 0) { 3735 V_norule_counter = 0; 3736 for (rule = chain->rules; rule; rule = rule->next) { 3737 /* Skip rules from another set. */ 3738 if (cmd == 1 && rule->set != set) 3739 continue; 3740 clear_counters(rule, log_only); 3741 } 3742 msg = log_only ? "ipfw: All logging counts reset.\n" : 3743 "ipfw: Accounting cleared.\n"; 3744 } else { 3745 int cleared = 0; 3746 /* 3747 * We can have multiple rules with the same number, so we 3748 * need to clear them all. 3749 */ 3750 for (rule = chain->rules; rule; rule = rule->next) 3751 if (rule->rulenum == rulenum) { 3752 while (rule && rule->rulenum == rulenum) { 3753 if (cmd == 0 || rule->set == set) 3754 clear_counters(rule, log_only); 3755 rule = rule->next; 3756 } 3757 cleared = 1; 3758 break; 3759 } 3760 if (!cleared) { /* we did not find any matching rules */ 3761 IPFW_WUNLOCK(chain); 3762 return (EINVAL); 3763 } 3764 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3765 "ipfw: Entry %d cleared.\n"; 3766 } 3767 IPFW_WUNLOCK(chain); 3768 3769 if (V_fw_verbose) 3770 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3771 return (0); 3772} 3773 3774/* 3775 * Check validity of the structure before insert. 3776 * Fortunately rules are simple, so this mostly need to check rule sizes. 3777 */ 3778static int 3779check_ipfw_struct(struct ip_fw *rule, int size) 3780{ 3781 int l, cmdlen = 0; 3782 int have_action=0; 3783 ipfw_insn *cmd; 3784 3785 if (size < sizeof(*rule)) { 3786 printf("ipfw: rule too short\n"); 3787 return (EINVAL); 3788 } 3789 /* first, check for valid size */ 3790 l = RULESIZE(rule); 3791 if (l != size) { 3792 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3793 return (EINVAL); 3794 } 3795 if (rule->act_ofs >= rule->cmd_len) { 3796 printf("ipfw: bogus action offset (%u > %u)\n", 3797 rule->act_ofs, rule->cmd_len - 1); 3798 return (EINVAL); 3799 } 3800 /* 3801 * Now go for the individual checks. Very simple ones, basically only 3802 * instruction sizes. 3803 */ 3804 for (l = rule->cmd_len, cmd = rule->cmd ; 3805 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3806 cmdlen = F_LEN(cmd); 3807 if (cmdlen > l) { 3808 printf("ipfw: opcode %d size truncated\n", 3809 cmd->opcode); 3810 return EINVAL; 3811 } 3812 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3813 switch (cmd->opcode) { 3814 case O_PROBE_STATE: 3815 case O_KEEP_STATE: 3816 case O_PROTO: 3817 case O_IP_SRC_ME: 3818 case O_IP_DST_ME: 3819 case O_LAYER2: 3820 case O_IN: 3821 case O_FRAG: 3822 case O_DIVERTED: 3823 case O_IPOPT: 3824 case O_IPTOS: 3825 case O_IPPRECEDENCE: 3826 case O_IPVER: 3827 case O_TCPWIN: 3828 case O_TCPFLAGS: 3829 case O_TCPOPTS: 3830 case O_ESTAB: 3831 case O_VERREVPATH: 3832 case O_VERSRCREACH: 3833 case O_ANTISPOOF: 3834 case O_IPSEC: 3835#ifdef INET6 3836 case O_IP6_SRC_ME: 3837 case O_IP6_DST_ME: 3838 case O_EXT_HDR: 3839 case O_IP6: 3840#endif 3841 case O_IP4: 3842 case O_TAG: 3843 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3844 goto bad_size; 3845 break; 3846 3847 case O_FIB: 3848 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3849 goto bad_size; 3850 if (cmd->arg1 >= rt_numfibs) { 3851 printf("ipfw: invalid fib number %d\n", 3852 cmd->arg1); 3853 return EINVAL; 3854 } 3855 break; 3856 3857 case O_SETFIB: 3858 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3859 goto bad_size; 3860 if (cmd->arg1 >= rt_numfibs) { 3861 printf("ipfw: invalid fib number %d\n", 3862 cmd->arg1); 3863 return EINVAL; 3864 } 3865 goto check_action; 3866 3867 case O_UID: 3868 case O_GID: 3869 case O_JAIL: 3870 case O_IP_SRC: 3871 case O_IP_DST: 3872 case O_TCPSEQ: 3873 case O_TCPACK: 3874 case O_PROB: 3875 case O_ICMPTYPE: 3876 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3877 goto bad_size; 3878 break; 3879 3880 case O_LIMIT: 3881 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3882 goto bad_size; 3883 break; 3884 3885 case O_LOG: 3886 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3887 goto bad_size; 3888 3889 ((ipfw_insn_log *)cmd)->log_left = 3890 ((ipfw_insn_log *)cmd)->max_log; 3891 3892 break; 3893 3894 case O_IP_SRC_MASK: 3895 case O_IP_DST_MASK: 3896 /* only odd command lengths */ 3897 if ( !(cmdlen & 1) || cmdlen > 31) 3898 goto bad_size; 3899 break; 3900 3901 case O_IP_SRC_SET: 3902 case O_IP_DST_SET: 3903 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3904 printf("ipfw: invalid set size %d\n", 3905 cmd->arg1); 3906 return EINVAL; 3907 } 3908 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3909 (cmd->arg1+31)/32 ) 3910 goto bad_size; 3911 break; 3912 3913 case O_IP_SRC_LOOKUP: 3914 case O_IP_DST_LOOKUP: 3915 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3916 printf("ipfw: invalid table number %d\n", 3917 cmd->arg1); 3918 return (EINVAL); 3919 } 3920 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3921 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3922 goto bad_size; 3923 break; 3924 3925 case O_MACADDR2: 3926 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3927 goto bad_size; 3928 break; 3929 3930 case O_NOP: 3931 case O_IPID: 3932 case O_IPTTL: 3933 case O_IPLEN: 3934 case O_TCPDATALEN: 3935 case O_TAGGED: 3936 if (cmdlen < 1 || cmdlen > 31) 3937 goto bad_size; 3938 break; 3939 3940 case O_MAC_TYPE: 3941 case O_IP_SRCPORT: 3942 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3943 if (cmdlen < 2 || cmdlen > 31) 3944 goto bad_size; 3945 break; 3946 3947 case O_RECV: 3948 case O_XMIT: 3949 case O_VIA: 3950 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3951 goto bad_size; 3952 break; 3953 3954 case O_ALTQ: 3955 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3956 goto bad_size; 3957 break; 3958 3959 case O_PIPE: 3960 case O_QUEUE: 3961 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3962 goto bad_size; 3963 goto check_action; 3964 3965 case O_FORWARD_IP: 3966#ifdef IPFIREWALL_FORWARD 3967 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3968 goto bad_size; 3969 goto check_action; 3970#else 3971 return EINVAL; 3972#endif 3973 3974 case O_DIVERT: 3975 case O_TEE: 3976 if (ip_divert_ptr == NULL) 3977 return EINVAL; 3978 else 3979 goto check_size; 3980 case O_NETGRAPH: 3981 case O_NGTEE: 3982 if (!NG_IPFW_LOADED) 3983 return EINVAL; 3984 else 3985 goto check_size; 3986 case O_NAT: 3987 if (!IPFW_NAT_LOADED) 3988 return EINVAL; 3989 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 3990 goto bad_size; 3991 goto check_action; 3992 case O_FORWARD_MAC: /* XXX not implemented yet */ 3993 case O_CHECK_STATE: 3994 case O_COUNT: 3995 case O_ACCEPT: 3996 case O_DENY: 3997 case O_REJECT: 3998#ifdef INET6 3999 case O_UNREACH6: 4000#endif 4001 case O_SKIPTO: 4002check_size: 4003 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4004 goto bad_size; 4005check_action: 4006 if (have_action) { 4007 printf("ipfw: opcode %d, multiple actions" 4008 " not allowed\n", 4009 cmd->opcode); 4010 return EINVAL; 4011 } 4012 have_action = 1; 4013 if (l != cmdlen) { 4014 printf("ipfw: opcode %d, action must be" 4015 " last opcode\n", 4016 cmd->opcode); 4017 return EINVAL; 4018 } 4019 break; 4020#ifdef INET6 4021 case O_IP6_SRC: 4022 case O_IP6_DST: 4023 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 4024 F_INSN_SIZE(ipfw_insn)) 4025 goto bad_size; 4026 break; 4027 4028 case O_FLOW6ID: 4029 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4030 ((ipfw_insn_u32 *)cmd)->o.arg1) 4031 goto bad_size; 4032 break; 4033 4034 case O_IP6_SRC_MASK: 4035 case O_IP6_DST_MASK: 4036 if ( !(cmdlen & 1) || cmdlen > 127) 4037 goto bad_size; 4038 break; 4039 case O_ICMP6TYPE: 4040 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4041 goto bad_size; 4042 break; 4043#endif 4044 4045 default: 4046 switch (cmd->opcode) { 4047#ifndef INET6 4048 case O_IP6_SRC_ME: 4049 case O_IP6_DST_ME: 4050 case O_EXT_HDR: 4051 case O_IP6: 4052 case O_UNREACH6: 4053 case O_IP6_SRC: 4054 case O_IP6_DST: 4055 case O_FLOW6ID: 4056 case O_IP6_SRC_MASK: 4057 case O_IP6_DST_MASK: 4058 case O_ICMP6TYPE: 4059 printf("ipfw: no IPv6 support in kernel\n"); 4060 return EPROTONOSUPPORT; 4061#endif 4062 default: 4063 printf("ipfw: opcode %d, unknown opcode\n", 4064 cmd->opcode); 4065 return EINVAL; 4066 } 4067 } 4068 } 4069 if (have_action == 0) { 4070 printf("ipfw: missing action\n"); 4071 return EINVAL; 4072 } 4073 return 0; 4074 4075bad_size: 4076 printf("ipfw: opcode %d size %d wrong\n", 4077 cmd->opcode, cmdlen); 4078 return EINVAL; 4079} 4080 4081/* 4082 * Copy the static and dynamic rules to the supplied buffer 4083 * and return the amount of space actually used. 4084 */ 4085static size_t 4086ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4087{ 4088 INIT_VNET_IPFW(curvnet); 4089 char *bp = buf; 4090 char *ep = bp + space; 4091 struct ip_fw *rule; 4092 int i; 4093 time_t boot_seconds; 4094 4095 boot_seconds = boottime.tv_sec; 4096 /* XXX this can take a long time and locking will block packet flow */ 4097 IPFW_RLOCK(chain); 4098 for (rule = chain->rules; rule ; rule = rule->next) { 4099 /* 4100 * Verify the entry fits in the buffer in case the 4101 * rules changed between calculating buffer space and 4102 * now. This would be better done using a generation 4103 * number but should suffice for now. 4104 */ 4105 i = RULESIZE(rule); 4106 if (bp + i <= ep) { 4107 bcopy(rule, bp, i); 4108 /* 4109 * XXX HACK. Store the disable mask in the "next" 4110 * pointer in a wild attempt to keep the ABI the same. 4111 * Why do we do this on EVERY rule? 4112 */ 4113 bcopy(&V_set_disable, 4114 &(((struct ip_fw *)bp)->next_rule), 4115 sizeof(V_set_disable)); 4116 if (((struct ip_fw *)bp)->timestamp) 4117 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4118 bp += i; 4119 } 4120 } 4121 IPFW_RUNLOCK(chain); 4122 if (V_ipfw_dyn_v) { 4123 ipfw_dyn_rule *p, *last = NULL; 4124 4125 IPFW_DYN_LOCK(); 4126 for (i = 0 ; i < V_curr_dyn_buckets; i++) 4127 for (p = V_ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4128 if (bp + sizeof *p <= ep) { 4129 ipfw_dyn_rule *dst = 4130 (ipfw_dyn_rule *)bp; 4131 bcopy(p, dst, sizeof *p); 4132 bcopy(&(p->rule->rulenum), &(dst->rule), 4133 sizeof(p->rule->rulenum)); 4134 /* 4135 * store set number into high word of 4136 * dst->rule pointer. 4137 */ 4138 bcopy(&(p->rule->set), 4139 (char *)&dst->rule + 4140 sizeof(p->rule->rulenum), 4141 sizeof(p->rule->set)); 4142 /* 4143 * store a non-null value in "next". 4144 * The userland code will interpret a 4145 * NULL here as a marker 4146 * for the last dynamic rule. 4147 */ 4148 bcopy(&dst, &dst->next, sizeof(dst)); 4149 last = dst; 4150 dst->expire = 4151 TIME_LEQ(dst->expire, time_uptime) ? 4152 0 : dst->expire - time_uptime ; 4153 bp += sizeof(ipfw_dyn_rule); 4154 } 4155 } 4156 IPFW_DYN_UNLOCK(); 4157 if (last != NULL) /* mark last dynamic rule */ 4158 bzero(&last->next, sizeof(last)); 4159 } 4160 return (bp - (char *)buf); 4161} 4162 4163 4164/** 4165 * {set|get}sockopt parser. 4166 */ 4167static int 4168ipfw_ctl(struct sockopt *sopt) 4169{ 4170#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4171 INIT_VNET_IPFW(curvnet); 4172 int error; 4173 size_t size; 4174 struct ip_fw *buf, *rule; 4175 u_int32_t rulenum[2]; 4176 4177 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4178 if (error) 4179 return (error); 4180 4181 /* 4182 * Disallow modifications in really-really secure mode, but still allow 4183 * the logging counters to be reset. 4184 */ 4185 if (sopt->sopt_name == IP_FW_ADD || 4186 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4187 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4188 if (error) 4189 return (error); 4190 } 4191 4192 error = 0; 4193 4194 switch (sopt->sopt_name) { 4195 case IP_FW_GET: 4196 /* 4197 * pass up a copy of the current rules. Static rules 4198 * come first (the last of which has number IPFW_DEFAULT_RULE), 4199 * followed by a possibly empty list of dynamic rule. 4200 * The last dynamic rule has NULL in the "next" field. 4201 * 4202 * Note that the calculated size is used to bound the 4203 * amount of data returned to the user. The rule set may 4204 * change between calculating the size and returning the 4205 * data in which case we'll just return what fits. 4206 */ 4207 size = V_static_len; /* size of static rules */ 4208 if (V_ipfw_dyn_v) /* add size of dyn.rules */ 4209 size += (V_dyn_count * sizeof(ipfw_dyn_rule)); 4210 4211 /* 4212 * XXX todo: if the user passes a short length just to know 4213 * how much room is needed, do not bother filling up the 4214 * buffer, just jump to the sooptcopyout. 4215 */ 4216 buf = malloc(size, M_TEMP, M_WAITOK); 4217 error = sooptcopyout(sopt, buf, 4218 ipfw_getrules(&V_layer3_chain, buf, size)); 4219 free(buf, M_TEMP); 4220 break; 4221 4222 case IP_FW_FLUSH: 4223 /* 4224 * Normally we cannot release the lock on each iteration. 4225 * We could do it here only because we start from the head all 4226 * the times so there is no risk of missing some entries. 4227 * On the other hand, the risk is that we end up with 4228 * a very inconsistent ruleset, so better keep the lock 4229 * around the whole cycle. 4230 * 4231 * XXX this code can be improved by resetting the head of 4232 * the list to point to the default rule, and then freeing 4233 * the old list without the need for a lock. 4234 */ 4235 4236 IPFW_WLOCK(&V_layer3_chain); 4237 V_layer3_chain.reap = NULL; 4238 free_chain(&V_layer3_chain, 0 /* keep default rule */); 4239 rule = V_layer3_chain.reap; 4240 V_layer3_chain.reap = NULL; 4241 IPFW_WUNLOCK(&V_layer3_chain); 4242 if (rule != NULL) 4243 reap_rules(rule); 4244 break; 4245 4246 case IP_FW_ADD: 4247 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4248 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4249 sizeof(struct ip_fw) ); 4250 if (error == 0) 4251 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4252 if (error == 0) { 4253 error = add_rule(&V_layer3_chain, rule); 4254 size = RULESIZE(rule); 4255 if (!error && sopt->sopt_dir == SOPT_GET) 4256 error = sooptcopyout(sopt, rule, size); 4257 } 4258 free(rule, M_TEMP); 4259 break; 4260 4261 case IP_FW_DEL: 4262 /* 4263 * IP_FW_DEL is used for deleting single rules or sets, 4264 * and (ab)used to atomically manipulate sets. Argument size 4265 * is used to distinguish between the two: 4266 * sizeof(u_int32_t) 4267 * delete single rule or set of rules, 4268 * or reassign rules (or sets) to a different set. 4269 * 2*sizeof(u_int32_t) 4270 * atomic disable/enable sets. 4271 * first u_int32_t contains sets to be disabled, 4272 * second u_int32_t contains sets to be enabled. 4273 */ 4274 error = sooptcopyin(sopt, rulenum, 4275 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4276 if (error) 4277 break; 4278 size = sopt->sopt_valsize; 4279 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4280 error = del_entry(&V_layer3_chain, rulenum[0]); 4281 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4282 V_set_disable = 4283 (V_set_disable | rulenum[0]) & ~rulenum[1] & 4284 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4285 else 4286 error = EINVAL; 4287 break; 4288 4289 case IP_FW_ZERO: 4290 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 4291 rulenum[0] = 0; 4292 if (sopt->sopt_val != 0) { 4293 error = sooptcopyin(sopt, rulenum, 4294 sizeof(u_int32_t), sizeof(u_int32_t)); 4295 if (error) 4296 break; 4297 } 4298 error = zero_entry(&V_layer3_chain, rulenum[0], 4299 sopt->sopt_name == IP_FW_RESETLOG); 4300 break; 4301 4302 case IP_FW_TABLE_ADD: 4303 { 4304 ipfw_table_entry ent; 4305 4306 error = sooptcopyin(sopt, &ent, 4307 sizeof(ent), sizeof(ent)); 4308 if (error) 4309 break; 4310 error = add_table_entry(&V_layer3_chain, ent.tbl, 4311 ent.addr, ent.masklen, ent.value); 4312 } 4313 break; 4314 4315 case IP_FW_TABLE_DEL: 4316 { 4317 ipfw_table_entry ent; 4318 4319 error = sooptcopyin(sopt, &ent, 4320 sizeof(ent), sizeof(ent)); 4321 if (error) 4322 break; 4323 error = del_table_entry(&V_layer3_chain, ent.tbl, 4324 ent.addr, ent.masklen); 4325 } 4326 break; 4327 4328 case IP_FW_TABLE_FLUSH: 4329 { 4330 u_int16_t tbl; 4331 4332 error = sooptcopyin(sopt, &tbl, 4333 sizeof(tbl), sizeof(tbl)); 4334 if (error) 4335 break; 4336 IPFW_WLOCK(&V_layer3_chain); 4337 error = flush_table(&V_layer3_chain, tbl); 4338 IPFW_WUNLOCK(&V_layer3_chain); 4339 } 4340 break; 4341 4342 case IP_FW_TABLE_GETSIZE: 4343 { 4344 u_int32_t tbl, cnt; 4345 4346 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4347 sizeof(tbl)))) 4348 break; 4349 IPFW_RLOCK(&V_layer3_chain); 4350 error = count_table(&V_layer3_chain, tbl, &cnt); 4351 IPFW_RUNLOCK(&V_layer3_chain); 4352 if (error) 4353 break; 4354 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4355 } 4356 break; 4357 4358 case IP_FW_TABLE_LIST: 4359 { 4360 ipfw_table *tbl; 4361 4362 if (sopt->sopt_valsize < sizeof(*tbl)) { 4363 error = EINVAL; 4364 break; 4365 } 4366 size = sopt->sopt_valsize; 4367 tbl = malloc(size, M_TEMP, M_WAITOK); 4368 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4369 if (error) { 4370 free(tbl, M_TEMP); 4371 break; 4372 } 4373 tbl->size = (size - sizeof(*tbl)) / 4374 sizeof(ipfw_table_entry); 4375 IPFW_RLOCK(&V_layer3_chain); 4376 error = dump_table(&V_layer3_chain, tbl); 4377 IPFW_RUNLOCK(&V_layer3_chain); 4378 if (error) { 4379 free(tbl, M_TEMP); 4380 break; 4381 } 4382 error = sooptcopyout(sopt, tbl, size); 4383 free(tbl, M_TEMP); 4384 } 4385 break; 4386 4387 case IP_FW_NAT_CFG: 4388 { 4389 if (IPFW_NAT_LOADED) 4390 error = ipfw_nat_cfg_ptr(sopt); 4391 else { 4392 printf("IP_FW_NAT_CFG: ipfw_nat not present, please load it.\n"); 4393 error = EINVAL; 4394 } 4395 } 4396 break; 4397 4398 case IP_FW_NAT_DEL: 4399 { 4400 if (IPFW_NAT_LOADED) 4401 error = ipfw_nat_del_ptr(sopt); 4402 else { 4403 printf("IP_FW_NAT_DEL: ipfw_nat not present, please load it.\n"); 4404 printf("ipfw_nat not loaded: %d\n", sopt->sopt_name); 4405 error = EINVAL; 4406 } 4407 } 4408 break; 4409 4410 case IP_FW_NAT_GET_CONFIG: 4411 { 4412 if (IPFW_NAT_LOADED) 4413 error = ipfw_nat_get_cfg_ptr(sopt); 4414 else { 4415 printf("IP_FW_NAT_GET_CFG: ipfw_nat not present, please load it.\n"); 4416 error = EINVAL; 4417 } 4418 } 4419 break; 4420 4421 case IP_FW_NAT_GET_LOG: 4422 { 4423 if (IPFW_NAT_LOADED) 4424 error = ipfw_nat_get_log_ptr(sopt); 4425 else { 4426 printf("IP_FW_NAT_GET_LOG: ipfw_nat not present, please load it.\n"); 4427 error = EINVAL; 4428 } 4429 } 4430 break; 4431 4432 default: 4433 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4434 error = EINVAL; 4435 } 4436 4437 return (error); 4438#undef RULE_MAXSIZE 4439} 4440 4441/** 4442 * dummynet needs a reference to the default rule, because rules can be 4443 * deleted while packets hold a reference to them. When this happens, 4444 * dummynet changes the reference to the default rule (it could well be a 4445 * NULL pointer, but this way we do not need to check for the special 4446 * case, plus here he have info on the default behaviour). 4447 */ 4448struct ip_fw *ip_fw_default_rule; 4449 4450/* 4451 * This procedure is only used to handle keepalives. It is invoked 4452 * every dyn_keepalive_period 4453 */ 4454static void 4455ipfw_tick(void * __unused unused) 4456{ 4457 struct mbuf *m0, *m, *mnext, **mtailp; 4458 int i; 4459 ipfw_dyn_rule *q; 4460 4461 if (V_dyn_keepalive == 0 || V_ipfw_dyn_v == NULL || V_dyn_count == 0) 4462 goto done; 4463 4464 /* 4465 * We make a chain of packets to go out here -- not deferring 4466 * until after we drop the IPFW dynamic rule lock would result 4467 * in a lock order reversal with the normal packet input -> ipfw 4468 * call stack. 4469 */ 4470 m0 = NULL; 4471 mtailp = &m0; 4472 IPFW_DYN_LOCK(); 4473 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 4474 for (q = V_ipfw_dyn_v[i] ; q ; q = q->next ) { 4475 if (q->dyn_type == O_LIMIT_PARENT) 4476 continue; 4477 if (q->id.proto != IPPROTO_TCP) 4478 continue; 4479 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4480 continue; 4481 if (TIME_LEQ(time_uptime + V_dyn_keepalive_interval, 4482 q->expire)) 4483 continue; /* too early */ 4484 if (TIME_LEQ(q->expire, time_uptime)) 4485 continue; /* too late, rule expired */ 4486 4487 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4488 q->ack_fwd, TH_SYN); 4489 if (*mtailp != NULL) 4490 mtailp = &(*mtailp)->m_nextpkt; 4491 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4492 q->ack_rev, 0); 4493 if (*mtailp != NULL) 4494 mtailp = &(*mtailp)->m_nextpkt; 4495 } 4496 } 4497 IPFW_DYN_UNLOCK(); 4498 for (m = mnext = m0; m != NULL; m = mnext) { 4499 mnext = m->m_nextpkt; 4500 m->m_nextpkt = NULL; 4501 ip_output(m, NULL, NULL, 0, NULL, NULL); 4502 } 4503done: 4504 callout_reset(&V_ipfw_timeout, V_dyn_keepalive_period * hz, 4505 ipfw_tick, NULL); 4506} 4507 4508int 4509ipfw_init(void) 4510{ 4511 INIT_VNET_IPFW(curvnet); 4512 struct ip_fw default_rule; 4513 int error; 4514 4515#ifdef INET6 4516 /* Setup IPv6 fw sysctl tree. */ 4517 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4518 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4519 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4520 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4521 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4522 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4523 &V_fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4524 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4525 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4526 &V_fw_deny_unknown_exthdrs, 0, 4527 "Deny packets with unknown IPv6 Extension Headers"); 4528#endif 4529 4530 V_layer3_chain.rules = NULL; 4531 IPFW_LOCK_INIT(&V_layer3_chain); 4532 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4533 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4534 UMA_ALIGN_PTR, 0); 4535 IPFW_DYN_LOCK_INIT(); 4536 callout_init(&V_ipfw_timeout, CALLOUT_MPSAFE); 4537 4538 bzero(&default_rule, sizeof default_rule); 4539 4540 default_rule.act_ofs = 0; 4541 default_rule.rulenum = IPFW_DEFAULT_RULE; 4542 default_rule.cmd_len = 1; 4543 default_rule.set = RESVD_SET; 4544 4545 default_rule.cmd[0].len = 1; 4546 default_rule.cmd[0].opcode = 4547#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4548 1 ? O_ACCEPT : 4549#endif 4550 O_DENY; 4551 4552 error = add_rule(&V_layer3_chain, &default_rule); 4553 if (error != 0) { 4554 printf("ipfw2: error %u initializing default rule " 4555 "(support disabled)\n", error); 4556 IPFW_DYN_LOCK_DESTROY(); 4557 IPFW_LOCK_DESTROY(&V_layer3_chain); 4558 uma_zdestroy(ipfw_dyn_rule_zone); 4559 return (error); 4560 } 4561 4562 ip_fw_default_rule = V_layer3_chain.rules; 4563 printf("ipfw2 " 4564#ifdef INET6 4565 "(+ipv6) " 4566#endif 4567 "initialized, divert %s, nat %s, " 4568 "rule-based forwarding " 4569#ifdef IPFIREWALL_FORWARD 4570 "enabled, " 4571#else 4572 "disabled, " 4573#endif 4574 "default to %s, logging ", 4575#ifdef IPDIVERT 4576 "enabled", 4577#else 4578 "loadable", 4579#endif 4580#ifdef IPFIREWALL_NAT 4581 "enabled", 4582#else 4583 "loadable", 4584#endif 4585 4586 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4587 4588#ifdef IPFIREWALL_VERBOSE 4589 V_fw_verbose = 1; 4590#endif 4591#ifdef IPFIREWALL_VERBOSE_LIMIT 4592 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4593#endif 4594 if (V_fw_verbose == 0) 4595 printf("disabled\n"); 4596 else if (V_verbose_limit == 0) 4597 printf("unlimited\n"); 4598 else 4599 printf("limited to %d packets/entry by default\n", 4600 V_verbose_limit); 4601 4602 error = init_tables(&V_layer3_chain); 4603 if (error) { 4604 IPFW_DYN_LOCK_DESTROY(); 4605 IPFW_LOCK_DESTROY(&V_layer3_chain); 4606 uma_zdestroy(ipfw_dyn_rule_zone); 4607 return (error); 4608 } 4609 ip_fw_ctl_ptr = ipfw_ctl; 4610 ip_fw_chk_ptr = ipfw_chk; 4611 callout_reset(&V_ipfw_timeout, hz, ipfw_tick, NULL); 4612 LIST_INIT(&V_layer3_chain.nat); 4613 return (0); 4614} 4615 4616void 4617ipfw_destroy(void) 4618{ 4619 struct ip_fw *reap; 4620 4621 ip_fw_chk_ptr = NULL; 4622 ip_fw_ctl_ptr = NULL; 4623 callout_drain(&V_ipfw_timeout); 4624 IPFW_WLOCK(&V_layer3_chain); 4625 flush_tables(&V_layer3_chain); 4626 V_layer3_chain.reap = NULL; 4627 free_chain(&V_layer3_chain, 1 /* kill default rule */); 4628 reap = V_layer3_chain.reap, V_layer3_chain.reap = NULL; 4629 IPFW_WUNLOCK(&V_layer3_chain); 4630 if (reap != NULL) 4631 reap_rules(reap); 4632 IPFW_DYN_LOCK_DESTROY(); 4633 uma_zdestroy(ipfw_dyn_rule_zone); 4634 if (V_ipfw_dyn_v != NULL) 4635 free(V_ipfw_dyn_v, M_IPFW); 4636 IPFW_LOCK_DESTROY(&V_layer3_chain); 4637 4638#ifdef INET6 4639 /* Free IPv6 fw sysctl tree. */ 4640 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4641#endif 4642 4643 printf("IP firewall unloaded\n"); 4644} 4645