Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/netinet/ip_fw2.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
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: releng/7.4/sys/netinet/ip_fw2.c 192336 2009-05-18 20:23:16Z jhb $"); 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/malloc.h> 52 #include <sys/mbuf.h> 53 #include <sys/kernel.h> 54 #include <sys/lock.h> 55 #include <sys/jail.h> 56 #include <sys/module.h> 57 #include <sys/priv.h> 58 #include <sys/proc.h> 59 #include <sys/socket.h> 60 #include <sys/socketvar.h> 61 #include <sys/sysctl.h> 62 #include <sys/syslog.h> 63 #include <sys/ucred.h> 64 #include <net/ethernet.h> /* for ETHERTYPE_IP */ 65 #include <net/if.h> 66 #include <net/radix.h> 67 #include <net/route.h> 68 #include <net/pf_mtag.h> 69 70 #define IPFW_INTERNAL /* Access to protected data structures in ip_fw.h. */ 71 72 #include <netinet/in.h> 73 #include <netinet/in_var.h> 74 #include <netinet/in_pcb.h> 75 #include <netinet/ip.h> 76 #include <netinet/ip_var.h> 77 #include <netinet/ip_icmp.h> 78 #include <netinet/ip_fw.h> 79 #include <netinet/ip_divert.h> 80 #include <netinet/ip_dummynet.h> 81 #include <netinet/ip_carp.h> 82 #include <netinet/pim.h> 83 #include <netinet/tcp_var.h> 84 #include <netinet/udp.h> 85 #include <netinet/udp_var.h> 86 #include <netinet/sctp.h> 87 #include <netgraph/ng_ipfw.h> 88 89 #include <netinet/ip6.h> 90 #include <netinet/icmp6.h> 91 #ifdef INET6 92 #include <netinet6/scope6_var.h> 93 #endif 94 95 #include <machine/in_cksum.h> /* XXX for in_cksum */ 96 97 #ifdef MAC 98 #include <security/mac/mac_framework.h> 99 #endif 100 101 /* 102 * set_disable contains one bit per set value (0..31). 103 * If the bit is set, all rules with the corresponding set 104 * are disabled. Set RESVD_SET(31) is reserved for the default rule 105 * and rules that are not deleted by the flush command, 106 * and CANNOT be disabled. 107 * Rules in set RESVD_SET can only be deleted explicitly. 108 */ 109 static u_int32_t set_disable; 110 static int fw_verbose; 111 static struct callout ipfw_timeout; 112 static int verbose_limit; 113 114 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 115 static int default_to_accept = 1; 116 #else 117 static int default_to_accept; 118 #endif 119 static uma_zone_t ipfw_dyn_rule_zone; 120 121 /* 122 * Data structure to cache our ucred related 123 * information. This structure only gets used if 124 * the user specified UID/GID based constraints in 125 * a firewall rule. 126 */ 127 struct ip_fw_ugid { 128 gid_t fw_groups[NGROUPS]; 129 int fw_ngroups; 130 uid_t fw_uid; 131 int fw_prid; 132 }; 133 134 /* 135 * list of rules for layer 3 136 */ 137 struct ip_fw_chain layer3_chain; 138 139 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 140 MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 141 #define IPFW_NAT_LOADED (ipfw_nat_ptr != NULL) 142 ipfw_nat_t *ipfw_nat_ptr = NULL; 143 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 144 ipfw_nat_cfg_t *ipfw_nat_del_ptr; 145 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 146 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 147 148 struct table_entry { 149 struct radix_node rn[2]; 150 struct sockaddr_in addr, mask; 151 u_int32_t value; 152 }; 153 154 static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 155 156 extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 157 158 #ifdef SYSCTL_NODE 159 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 160 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 161 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 162 ipfw_chg_hook, "I", "Enable ipfw"); 163 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 164 &autoinc_step, 0, "Rule number autincrement step"); 165 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 166 CTLFLAG_RW | CTLFLAG_SECURE3, 167 &fw_one_pass, 0, 168 "Only do a single pass through ipfw when using dummynet(4)"); 169 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 170 CTLFLAG_RW | CTLFLAG_SECURE3, 171 &fw_verbose, 0, "Log matches to ipfw rules"); 172 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 173 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 174 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 175 NULL, IPFW_DEFAULT_RULE, "The default/max possible rule number."); 176 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 177 NULL, IPFW_TABLES_MAX, "The maximum number of tables."); 178 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 179 &default_to_accept, 0, "Make the default rule accept all packets."); 180 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept); 181 #endif /* SYSCTL_NODE */ 182 183 /* 184 * Description of dynamic rules. 185 * 186 * Dynamic rules are stored in lists accessed through a hash table 187 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 188 * be modified through the sysctl variable dyn_buckets which is 189 * updated when the table becomes empty. 190 * 191 * XXX currently there is only one list, ipfw_dyn. 192 * 193 * When a packet is received, its address fields are first masked 194 * with the mask defined for the rule, then hashed, then matched 195 * against the entries in the corresponding list. 196 * Dynamic rules can be used for different purposes: 197 * + stateful rules; 198 * + enforcing limits on the number of sessions; 199 * + in-kernel NAT (not implemented yet) 200 * 201 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 202 * measured in seconds and depending on the flags. 203 * 204 * The total number of dynamic rules is stored in dyn_count. 205 * The max number of dynamic rules is dyn_max. When we reach 206 * the maximum number of rules we do not create anymore. This is 207 * done to avoid consuming too much memory, but also too much 208 * time when searching on each packet (ideally, we should try instead 209 * to put a limit on the length of the list on each bucket...). 210 * 211 * Each dynamic rule holds a pointer to the parent ipfw rule so 212 * we know what action to perform. Dynamic rules are removed when 213 * the parent rule is deleted. XXX we should make them survive. 214 * 215 * There are some limitations with dynamic rules -- we do not 216 * obey the 'randomized match', and we do not do multiple 217 * passes through the firewall. XXX check the latter!!! 218 */ 219 static ipfw_dyn_rule **ipfw_dyn_v = NULL; 220 static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 221 static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 222 223 static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 224 #define IPFW_DYN_LOCK_INIT() \ 225 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 226 #define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 227 #define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 228 #define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 229 #define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 230 231 /* 232 * Timeouts for various events in handing dynamic rules. 233 */ 234 static u_int32_t dyn_ack_lifetime = 300; 235 static u_int32_t dyn_syn_lifetime = 20; 236 static u_int32_t dyn_fin_lifetime = 1; 237 static u_int32_t dyn_rst_lifetime = 1; 238 static u_int32_t dyn_udp_lifetime = 10; 239 static u_int32_t dyn_short_lifetime = 5; 240 241 /* 242 * Keepalives are sent if dyn_keepalive is set. They are sent every 243 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 244 * seconds of lifetime of a rule. 245 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 246 * than dyn_keepalive_period. 247 */ 248 249 static u_int32_t dyn_keepalive_interval = 20; 250 static u_int32_t dyn_keepalive_period = 5; 251 static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 252 253 static u_int32_t static_count; /* # of static rules */ 254 static u_int32_t static_len; /* size in bytes of static rules */ 255 static u_int32_t dyn_count; /* # of dynamic rules */ 256 static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 257 258 #ifdef SYSCTL_NODE 259 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 260 &dyn_buckets, 0, "Number of dyn. buckets"); 261 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 262 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 263 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 264 &dyn_count, 0, "Number of dyn. rules"); 265 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 266 &dyn_max, 0, "Max number of dyn. rules"); 267 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 268 &static_count, 0, "Number of static rules"); 269 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 270 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 271 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 272 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 273 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 274 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 275 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 276 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 277 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 278 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 279 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 280 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 281 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 282 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 283 #endif /* SYSCTL_NODE */ 284 285 #ifdef INET6 286 /* 287 * IPv6 specific variables 288 */ 289 #ifdef SYSCTL_NODE 290 SYSCTL_DECL(_net_inet6_ip6); 291 #endif /* SYSCTL_NODE */ 292 293 static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 294 static struct sysctl_oid *ip6_fw_sysctl_tree; 295 #endif /* INET6 */ 296 297 static int fw_deny_unknown_exthdrs = 1; 298 299 300 /* 301 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 302 * Other macros just cast void * into the appropriate type 303 */ 304 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 305 #define TCP(p) ((struct tcphdr *)(p)) 306 #define SCTP(p) ((struct sctphdr *)(p)) 307 #define UDP(p) ((struct udphdr *)(p)) 308 #define ICMP(p) ((struct icmphdr *)(p)) 309 #define ICMP6(p) ((struct icmp6_hdr *)(p)) 310 311 static __inline int 312 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 313 { 314 int type = icmp->icmp_type; 315 316 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 317 } 318 319 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 320 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 321 322 static int 323 is_icmp_query(struct icmphdr *icmp) 324 { 325 int type = icmp->icmp_type; 326 327 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 328 } 329 #undef TT 330 331 /* 332 * The following checks use two arrays of 8 or 16 bits to store the 333 * bits that we want set or clear, respectively. They are in the 334 * low and high half of cmd->arg1 or cmd->d[0]. 335 * 336 * We scan options and store the bits we find set. We succeed if 337 * 338 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 339 * 340 * The code is sometimes optimized not to store additional variables. 341 */ 342 343 static int 344 flags_match(ipfw_insn *cmd, u_int8_t bits) 345 { 346 u_char want_clear; 347 bits = ~bits; 348 349 if ( ((cmd->arg1 & 0xff) & bits) != 0) 350 return 0; /* some bits we want set were clear */ 351 want_clear = (cmd->arg1 >> 8) & 0xff; 352 if ( (want_clear & bits) != want_clear) 353 return 0; /* some bits we want clear were set */ 354 return 1; 355 } 356 357 static int 358 ipopts_match(struct ip *ip, ipfw_insn *cmd) 359 { 360 int optlen, bits = 0; 361 u_char *cp = (u_char *)(ip + 1); 362 int x = (ip->ip_hl << 2) - sizeof (struct ip); 363 364 for (; x > 0; x -= optlen, cp += optlen) { 365 int opt = cp[IPOPT_OPTVAL]; 366 367 if (opt == IPOPT_EOL) 368 break; 369 if (opt == IPOPT_NOP) 370 optlen = 1; 371 else { 372 optlen = cp[IPOPT_OLEN]; 373 if (optlen <= 0 || optlen > x) 374 return 0; /* invalid or truncated */ 375 } 376 switch (opt) { 377 378 default: 379 break; 380 381 case IPOPT_LSRR: 382 bits |= IP_FW_IPOPT_LSRR; 383 break; 384 385 case IPOPT_SSRR: 386 bits |= IP_FW_IPOPT_SSRR; 387 break; 388 389 case IPOPT_RR: 390 bits |= IP_FW_IPOPT_RR; 391 break; 392 393 case IPOPT_TS: 394 bits |= IP_FW_IPOPT_TS; 395 break; 396 } 397 } 398 return (flags_match(cmd, bits)); 399 } 400 401 static int 402 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 403 { 404 int optlen, bits = 0; 405 u_char *cp = (u_char *)(tcp + 1); 406 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 407 408 for (; x > 0; x -= optlen, cp += optlen) { 409 int opt = cp[0]; 410 if (opt == TCPOPT_EOL) 411 break; 412 if (opt == TCPOPT_NOP) 413 optlen = 1; 414 else { 415 optlen = cp[1]; 416 if (optlen <= 0) 417 break; 418 } 419 420 switch (opt) { 421 422 default: 423 break; 424 425 case TCPOPT_MAXSEG: 426 bits |= IP_FW_TCPOPT_MSS; 427 break; 428 429 case TCPOPT_WINDOW: 430 bits |= IP_FW_TCPOPT_WINDOW; 431 break; 432 433 case TCPOPT_SACK_PERMITTED: 434 case TCPOPT_SACK: 435 bits |= IP_FW_TCPOPT_SACK; 436 break; 437 438 case TCPOPT_TIMESTAMP: 439 bits |= IP_FW_TCPOPT_TS; 440 break; 441 442 } 443 } 444 return (flags_match(cmd, bits)); 445 } 446 447 static int 448 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 449 { 450 if (ifp == NULL) /* no iface with this packet, match fails */ 451 return 0; 452 /* Check by name or by IP address */ 453 if (cmd->name[0] != '\0') { /* match by name */ 454 /* Check name */ 455 if (cmd->p.glob) { 456 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 457 return(1); 458 } else { 459 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 460 return(1); 461 } 462 } else { 463 struct ifaddr *ia; 464 465 /* XXX lock? */ 466 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 467 if (ia->ifa_addr->sa_family != AF_INET) 468 continue; 469 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 470 (ia->ifa_addr))->sin_addr.s_addr) 471 return(1); /* match */ 472 } 473 } 474 return(0); /* no match, fail ... */ 475 } 476 477 /* 478 * The verify_path function checks if a route to the src exists and 479 * if it is reachable via ifp (when provided). 480 * 481 * The 'verrevpath' option checks that the interface that an IP packet 482 * arrives on is the same interface that traffic destined for the 483 * packet's source address would be routed out of. The 'versrcreach' 484 * option just checks that the source address is reachable via any route 485 * (except default) in the routing table. These two are a measure to block 486 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 487 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 488 * is purposely reminiscent of the Cisco IOS command, 489 * 490 * ip verify unicast reverse-path 491 * ip verify unicast source reachable-via any 492 * 493 * which implements the same functionality. But note that syntax is 494 * misleading. The check may be performed on all IP packets whether unicast, 495 * multicast, or broadcast. 496 */ 497 static int 498 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 499 { 500 struct route ro; 501 struct sockaddr_in *dst; 502 503 bzero(&ro, sizeof(ro)); 504 505 dst = (struct sockaddr_in *)&(ro.ro_dst); 506 dst->sin_family = AF_INET; 507 dst->sin_len = sizeof(*dst); 508 dst->sin_addr = src; 509 in_rtalloc_ign(&ro, RTF_CLONING, fib); 510 511 if (ro.ro_rt == NULL) 512 return 0; 513 514 /* 515 * If ifp is provided, check for equality with rtentry. 516 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 517 * in order to pass packets injected back by if_simloop(): 518 * if useloopback == 1 routing entry (via lo0) for our own address 519 * may exist, so we need to handle routing assymetry. 520 */ 521 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 522 RTFREE(ro.ro_rt); 523 return 0; 524 } 525 526 /* if no ifp provided, check if rtentry is not default route */ 527 if (ifp == NULL && 528 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 529 RTFREE(ro.ro_rt); 530 return 0; 531 } 532 533 /* or if this is a blackhole/reject route */ 534 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 535 RTFREE(ro.ro_rt); 536 return 0; 537 } 538 539 /* found valid route */ 540 RTFREE(ro.ro_rt); 541 return 1; 542 } 543 544 #ifdef INET6 545 /* 546 * ipv6 specific rules here... 547 */ 548 static __inline int 549 icmp6type_match (int type, ipfw_insn_u32 *cmd) 550 { 551 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 552 } 553 554 static int 555 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 556 { 557 int i; 558 for (i=0; i <= cmd->o.arg1; ++i ) 559 if (curr_flow == cmd->d[i] ) 560 return 1; 561 return 0; 562 } 563 564 /* support for IP6_*_ME opcodes */ 565 static int 566 search_ip6_addr_net (struct in6_addr * ip6_addr) 567 { 568 struct ifnet *mdc; 569 struct ifaddr *mdc2; 570 struct in6_ifaddr *fdm; 571 struct in6_addr copia; 572 573 TAILQ_FOREACH(mdc, &ifnet, if_link) 574 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 575 if (mdc2->ifa_addr->sa_family == AF_INET6) { 576 fdm = (struct in6_ifaddr *)mdc2; 577 copia = fdm->ia_addr.sin6_addr; 578 /* need for leaving scope_id in the sock_addr */ 579 in6_clearscope(&copia); 580 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 581 return 1; 582 } 583 } 584 return 0; 585 } 586 587 static int 588 verify_path6(struct in6_addr *src, struct ifnet *ifp) 589 { 590 struct route_in6 ro; 591 struct sockaddr_in6 *dst; 592 593 bzero(&ro, sizeof(ro)); 594 595 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 596 dst->sin6_family = AF_INET6; 597 dst->sin6_len = sizeof(*dst); 598 dst->sin6_addr = *src; 599 /* XXX MRT 0 for ipv6 at this time */ 600 rtalloc_ign((struct route *)&ro, RTF_CLONING); 601 602 if (ro.ro_rt == NULL) 603 return 0; 604 605 /* 606 * if ifp is provided, check for equality with rtentry 607 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 608 * to support the case of sending packets to an address of our own. 609 * (where the former interface is the first argument of if_simloop() 610 * (=ifp), the latter is lo0) 611 */ 612 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 613 RTFREE(ro.ro_rt); 614 return 0; 615 } 616 617 /* if no ifp provided, check if rtentry is not default route */ 618 if (ifp == NULL && 619 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 620 RTFREE(ro.ro_rt); 621 return 0; 622 } 623 624 /* or if this is a blackhole/reject route */ 625 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 626 RTFREE(ro.ro_rt); 627 return 0; 628 } 629 630 /* found valid route */ 631 RTFREE(ro.ro_rt); 632 return 1; 633 634 } 635 static __inline int 636 hash_packet6(struct ipfw_flow_id *id) 637 { 638 u_int32_t i; 639 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 640 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 641 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 642 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 643 (id->dst_port) ^ (id->src_port); 644 return i; 645 } 646 647 static int 648 is_icmp6_query(int icmp6_type) 649 { 650 if ((icmp6_type <= ICMP6_MAXTYPE) && 651 (icmp6_type == ICMP6_ECHO_REQUEST || 652 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 653 icmp6_type == ICMP6_WRUREQUEST || 654 icmp6_type == ICMP6_FQDN_QUERY || 655 icmp6_type == ICMP6_NI_QUERY)) 656 return (1); 657 658 return (0); 659 } 660 661 static void 662 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 663 { 664 struct mbuf *m; 665 666 m = args->m; 667 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 668 struct tcphdr *tcp; 669 tcp_seq ack, seq; 670 int flags; 671 struct { 672 struct ip6_hdr ip6; 673 struct tcphdr th; 674 } ti; 675 tcp = (struct tcphdr *)((char *)ip6 + hlen); 676 677 if ((tcp->th_flags & TH_RST) != 0) { 678 m_freem(m); 679 args->m = NULL; 680 return; 681 } 682 683 ti.ip6 = *ip6; 684 ti.th = *tcp; 685 ti.th.th_seq = ntohl(ti.th.th_seq); 686 ti.th.th_ack = ntohl(ti.th.th_ack); 687 ti.ip6.ip6_nxt = IPPROTO_TCP; 688 689 if (ti.th.th_flags & TH_ACK) { 690 ack = 0; 691 seq = ti.th.th_ack; 692 flags = TH_RST; 693 } else { 694 ack = ti.th.th_seq; 695 if ((m->m_flags & M_PKTHDR) != 0) { 696 /* 697 * total new data to ACK is: 698 * total packet length, 699 * minus the header length, 700 * minus the tcp header length. 701 */ 702 ack += m->m_pkthdr.len - hlen 703 - (ti.th.th_off << 2); 704 } else if (ip6->ip6_plen) { 705 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 706 hlen - (ti.th.th_off << 2); 707 } else { 708 m_freem(m); 709 return; 710 } 711 if (tcp->th_flags & TH_SYN) 712 ack++; 713 seq = 0; 714 flags = TH_RST|TH_ACK; 715 } 716 bcopy(&ti, ip6, sizeof(ti)); 717 /* 718 * m is only used to recycle the mbuf 719 * The data in it is never read so we don't need 720 * to correct the offsets or anything 721 */ 722 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 723 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 724 #if 0 725 /* 726 * Unlike above, the mbufs need to line up with the ip6 hdr, 727 * as the contents are read. We need to m_adj() the 728 * needed amount. 729 * The mbuf will however be thrown away so we can adjust it. 730 * Remember we did an m_pullup on it already so we 731 * can make some assumptions about contiguousness. 732 */ 733 if (args->L3offset) 734 m_adj(m, args->L3offset); 735 #endif 736 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 737 } else 738 m_freem(m); 739 740 args->m = NULL; 741 } 742 743 #endif /* INET6 */ 744 745 static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 746 747 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 748 #define SNP(buf) buf, sizeof(buf) 749 750 /* 751 * We enter here when we have a rule with O_LOG. 752 * XXX this function alone takes about 2Kbytes of code! 753 */ 754 static void 755 ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 756 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 757 struct ip *ip) 758 { 759 struct ether_header *eh = args->eh; 760 char *action; 761 int limit_reached = 0; 762 char action2[40], proto[128], fragment[32]; 763 764 fragment[0] = '\0'; 765 proto[0] = '\0'; 766 767 if (f == NULL) { /* bogus pkt */ 768 if (verbose_limit != 0 && norule_counter >= verbose_limit) 769 return; 770 norule_counter++; 771 if (norule_counter == verbose_limit) 772 limit_reached = verbose_limit; 773 action = "Refuse"; 774 } else { /* O_LOG is the first action, find the real one */ 775 ipfw_insn *cmd = ACTION_PTR(f); 776 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 777 778 if (l->max_log != 0 && l->log_left == 0) 779 return; 780 l->log_left--; 781 if (l->log_left == 0) 782 limit_reached = l->max_log; 783 cmd += F_LEN(cmd); /* point to first action */ 784 if (cmd->opcode == O_ALTQ) { 785 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 786 787 snprintf(SNPARGS(action2, 0), "Altq %d", 788 altq->qid); 789 cmd += F_LEN(cmd); 790 } 791 if (cmd->opcode == O_PROB) 792 cmd += F_LEN(cmd); 793 794 if (cmd->opcode == O_TAG) 795 cmd += F_LEN(cmd); 796 797 action = action2; 798 switch (cmd->opcode) { 799 case O_DENY: 800 action = "Deny"; 801 break; 802 803 case O_REJECT: 804 if (cmd->arg1==ICMP_REJECT_RST) 805 action = "Reset"; 806 else if (cmd->arg1==ICMP_UNREACH_HOST) 807 action = "Reject"; 808 else 809 snprintf(SNPARGS(action2, 0), "Unreach %d", 810 cmd->arg1); 811 break; 812 813 case O_UNREACH6: 814 if (cmd->arg1==ICMP6_UNREACH_RST) 815 action = "Reset"; 816 else 817 snprintf(SNPARGS(action2, 0), "Unreach %d", 818 cmd->arg1); 819 break; 820 821 case O_ACCEPT: 822 action = "Accept"; 823 break; 824 case O_COUNT: 825 action = "Count"; 826 break; 827 case O_DIVERT: 828 snprintf(SNPARGS(action2, 0), "Divert %d", 829 cmd->arg1); 830 break; 831 case O_TEE: 832 snprintf(SNPARGS(action2, 0), "Tee %d", 833 cmd->arg1); 834 break; 835 case O_SETFIB: 836 snprintf(SNPARGS(action2, 0), "SetFib %d", 837 cmd->arg1); 838 break; 839 case O_SKIPTO: 840 snprintf(SNPARGS(action2, 0), "SkipTo %d", 841 cmd->arg1); 842 break; 843 case O_PIPE: 844 snprintf(SNPARGS(action2, 0), "Pipe %d", 845 cmd->arg1); 846 break; 847 case O_QUEUE: 848 snprintf(SNPARGS(action2, 0), "Queue %d", 849 cmd->arg1); 850 break; 851 case O_FORWARD_IP: { 852 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 853 int len; 854 struct in_addr dummyaddr; 855 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 856 dummyaddr.s_addr = htonl(tablearg); 857 else 858 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 859 860 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 861 inet_ntoa(dummyaddr)); 862 863 if (sa->sa.sin_port) 864 snprintf(SNPARGS(action2, len), ":%d", 865 sa->sa.sin_port); 866 } 867 break; 868 case O_NETGRAPH: 869 snprintf(SNPARGS(action2, 0), "Netgraph %d", 870 cmd->arg1); 871 break; 872 case O_NGTEE: 873 snprintf(SNPARGS(action2, 0), "Ngtee %d", 874 cmd->arg1); 875 break; 876 case O_NAT: 877 action = "Nat"; 878 break; 879 default: 880 action = "UNKNOWN"; 881 break; 882 } 883 } 884 885 if (hlen == 0) { /* non-ip */ 886 snprintf(SNPARGS(proto, 0), "MAC"); 887 888 } else { 889 int len; 890 char src[48], dst[48]; 891 struct icmphdr *icmp; 892 struct tcphdr *tcp; 893 struct udphdr *udp; 894 #ifdef INET6 895 struct ip6_hdr *ip6 = NULL; 896 struct icmp6_hdr *icmp6; 897 #endif 898 src[0] = '\0'; 899 dst[0] = '\0'; 900 #ifdef INET6 901 if (IS_IP6_FLOW_ID(&(args->f_id))) { 902 char ip6buf[INET6_ADDRSTRLEN]; 903 snprintf(src, sizeof(src), "[%s]", 904 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 905 snprintf(dst, sizeof(dst), "[%s]", 906 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 907 908 ip6 = (struct ip6_hdr *)ip; 909 tcp = (struct tcphdr *)(((char *)ip) + hlen); 910 udp = (struct udphdr *)(((char *)ip) + hlen); 911 } else 912 #endif 913 { 914 tcp = L3HDR(struct tcphdr, ip); 915 udp = L3HDR(struct udphdr, ip); 916 917 inet_ntoa_r(ip->ip_src, src); 918 inet_ntoa_r(ip->ip_dst, dst); 919 } 920 921 switch (args->f_id.proto) { 922 case IPPROTO_TCP: 923 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 924 if (offset == 0) 925 snprintf(SNPARGS(proto, len), ":%d %s:%d", 926 ntohs(tcp->th_sport), 927 dst, 928 ntohs(tcp->th_dport)); 929 else 930 snprintf(SNPARGS(proto, len), " %s", dst); 931 break; 932 933 case IPPROTO_UDP: 934 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 935 if (offset == 0) 936 snprintf(SNPARGS(proto, len), ":%d %s:%d", 937 ntohs(udp->uh_sport), 938 dst, 939 ntohs(udp->uh_dport)); 940 else 941 snprintf(SNPARGS(proto, len), " %s", dst); 942 break; 943 944 case IPPROTO_ICMP: 945 icmp = L3HDR(struct icmphdr, ip); 946 if (offset == 0) 947 len = snprintf(SNPARGS(proto, 0), 948 "ICMP:%u.%u ", 949 icmp->icmp_type, icmp->icmp_code); 950 else 951 len = snprintf(SNPARGS(proto, 0), "ICMP "); 952 len += snprintf(SNPARGS(proto, len), "%s", src); 953 snprintf(SNPARGS(proto, len), " %s", dst); 954 break; 955 #ifdef INET6 956 case IPPROTO_ICMPV6: 957 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 958 if (offset == 0) 959 len = snprintf(SNPARGS(proto, 0), 960 "ICMPv6:%u.%u ", 961 icmp6->icmp6_type, icmp6->icmp6_code); 962 else 963 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 964 len += snprintf(SNPARGS(proto, len), "%s", src); 965 snprintf(SNPARGS(proto, len), " %s", dst); 966 break; 967 #endif 968 default: 969 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 970 args->f_id.proto, src); 971 snprintf(SNPARGS(proto, len), " %s", dst); 972 break; 973 } 974 975 #ifdef INET6 976 if (IS_IP6_FLOW_ID(&(args->f_id))) { 977 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 978 snprintf(SNPARGS(fragment, 0), 979 " (frag %08x:%d@%d%s)", 980 args->f_id.frag_id6, 981 ntohs(ip6->ip6_plen) - hlen, 982 ntohs(offset & IP6F_OFF_MASK) << 3, 983 (offset & IP6F_MORE_FRAG) ? "+" : ""); 984 } else 985 #endif 986 { 987 int ip_off, ip_len; 988 if (eh != NULL) { /* layer 2 packets are as on the wire */ 989 ip_off = ntohs(ip->ip_off); 990 ip_len = ntohs(ip->ip_len); 991 } else { 992 ip_off = ip->ip_off; 993 ip_len = ip->ip_len; 994 } 995 if (ip_off & (IP_MF | IP_OFFMASK)) 996 snprintf(SNPARGS(fragment, 0), 997 " (frag %d:%d@%d%s)", 998 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 999 offset << 3, 1000 (ip_off & IP_MF) ? "+" : ""); 1001 } 1002 } 1003 if (oif || m->m_pkthdr.rcvif) 1004 log(LOG_SECURITY | LOG_INFO, 1005 "ipfw: %d %s %s %s via %s%s\n", 1006 f ? f->rulenum : -1, 1007 action, proto, oif ? "out" : "in", 1008 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1009 fragment); 1010 else 1011 log(LOG_SECURITY | LOG_INFO, 1012 "ipfw: %d %s %s [no if info]%s\n", 1013 f ? f->rulenum : -1, 1014 action, proto, fragment); 1015 if (limit_reached) 1016 log(LOG_SECURITY | LOG_NOTICE, 1017 "ipfw: limit %d reached on entry %d\n", 1018 limit_reached, f ? f->rulenum : -1); 1019 } 1020 1021 /* 1022 * IMPORTANT: the hash function for dynamic rules must be commutative 1023 * in source and destination (ip,port), because rules are bidirectional 1024 * and we want to find both in the same bucket. 1025 */ 1026 static __inline int 1027 hash_packet(struct ipfw_flow_id *id) 1028 { 1029 u_int32_t i; 1030 1031 #ifdef INET6 1032 if (IS_IP6_FLOW_ID(id)) 1033 i = hash_packet6(id); 1034 else 1035 #endif /* INET6 */ 1036 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1037 i &= (curr_dyn_buckets - 1); 1038 return i; 1039 } 1040 1041 /** 1042 * unlink a dynamic rule from a chain. prev is a pointer to 1043 * the previous one, q is a pointer to the rule to delete, 1044 * head is a pointer to the head of the queue. 1045 * Modifies q and potentially also head. 1046 */ 1047 #define UNLINK_DYN_RULE(prev, head, q) { \ 1048 ipfw_dyn_rule *old_q = q; \ 1049 \ 1050 /* remove a refcount to the parent */ \ 1051 if (q->dyn_type == O_LIMIT) \ 1052 q->parent->count--; \ 1053 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1054 (q->id.src_ip), (q->id.src_port), \ 1055 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1056 if (prev != NULL) \ 1057 prev->next = q = q->next; \ 1058 else \ 1059 head = q = q->next; \ 1060 dyn_count--; \ 1061 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1062 1063 #define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1064 1065 /** 1066 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1067 * 1068 * If keep_me == NULL, rules are deleted even if not expired, 1069 * otherwise only expired rules are removed. 1070 * 1071 * The value of the second parameter is also used to point to identify 1072 * a rule we absolutely do not want to remove (e.g. because we are 1073 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1074 * rules). The pointer is only used for comparison, so any non-null 1075 * value will do. 1076 */ 1077 static void 1078 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1079 { 1080 static u_int32_t last_remove = 0; 1081 1082 #define FORCE (keep_me == NULL) 1083 1084 ipfw_dyn_rule *prev, *q; 1085 int i, pass = 0, max_pass = 0; 1086 1087 IPFW_DYN_LOCK_ASSERT(); 1088 1089 if (ipfw_dyn_v == NULL || dyn_count == 0) 1090 return; 1091 /* do not expire more than once per second, it is useless */ 1092 if (!FORCE && last_remove == time_uptime) 1093 return; 1094 last_remove = time_uptime; 1095 1096 /* 1097 * because O_LIMIT refer to parent rules, during the first pass only 1098 * remove child and mark any pending LIMIT_PARENT, and remove 1099 * them in a second pass. 1100 */ 1101 next_pass: 1102 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1103 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1104 /* 1105 * Logic can become complex here, so we split tests. 1106 */ 1107 if (q == keep_me) 1108 goto next; 1109 if (rule != NULL && rule != q->rule) 1110 goto next; /* not the one we are looking for */ 1111 if (q->dyn_type == O_LIMIT_PARENT) { 1112 /* 1113 * handle parent in the second pass, 1114 * record we need one. 1115 */ 1116 max_pass = 1; 1117 if (pass == 0) 1118 goto next; 1119 if (FORCE && q->count != 0 ) { 1120 /* XXX should not happen! */ 1121 printf("ipfw: OUCH! cannot remove rule," 1122 " count %d\n", q->count); 1123 } 1124 } else { 1125 if (!FORCE && 1126 !TIME_LEQ( q->expire, time_uptime )) 1127 goto next; 1128 } 1129 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1130 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1131 continue; 1132 } 1133 next: 1134 prev=q; 1135 q=q->next; 1136 } 1137 } 1138 if (pass++ < max_pass) 1139 goto next_pass; 1140 } 1141 1142 1143 /** 1144 * lookup a dynamic rule. 1145 */ 1146 static ipfw_dyn_rule * 1147 lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1148 struct tcphdr *tcp) 1149 { 1150 /* 1151 * stateful ipfw extensions. 1152 * Lookup into dynamic session queue 1153 */ 1154 #define MATCH_REVERSE 0 1155 #define MATCH_FORWARD 1 1156 #define MATCH_NONE 2 1157 #define MATCH_UNKNOWN 3 1158 int i, dir = MATCH_NONE; 1159 ipfw_dyn_rule *prev, *q=NULL; 1160 1161 IPFW_DYN_LOCK_ASSERT(); 1162 1163 if (ipfw_dyn_v == NULL) 1164 goto done; /* not found */ 1165 i = hash_packet( pkt ); 1166 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1167 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1168 goto next; 1169 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1170 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1171 continue; 1172 } 1173 if (pkt->proto == q->id.proto && 1174 q->dyn_type != O_LIMIT_PARENT) { 1175 if (IS_IP6_FLOW_ID(pkt)) { 1176 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1177 &(q->id.src_ip6)) && 1178 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1179 &(q->id.dst_ip6)) && 1180 pkt->src_port == q->id.src_port && 1181 pkt->dst_port == q->id.dst_port ) { 1182 dir = MATCH_FORWARD; 1183 break; 1184 } 1185 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1186 &(q->id.dst_ip6)) && 1187 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1188 &(q->id.src_ip6)) && 1189 pkt->src_port == q->id.dst_port && 1190 pkt->dst_port == q->id.src_port ) { 1191 dir = MATCH_REVERSE; 1192 break; 1193 } 1194 } else { 1195 if (pkt->src_ip == q->id.src_ip && 1196 pkt->dst_ip == q->id.dst_ip && 1197 pkt->src_port == q->id.src_port && 1198 pkt->dst_port == q->id.dst_port ) { 1199 dir = MATCH_FORWARD; 1200 break; 1201 } 1202 if (pkt->src_ip == q->id.dst_ip && 1203 pkt->dst_ip == q->id.src_ip && 1204 pkt->src_port == q->id.dst_port && 1205 pkt->dst_port == q->id.src_port ) { 1206 dir = MATCH_REVERSE; 1207 break; 1208 } 1209 } 1210 } 1211 next: 1212 prev = q; 1213 q = q->next; 1214 } 1215 if (q == NULL) 1216 goto done; /* q = NULL, not found */ 1217 1218 if ( prev != NULL) { /* found and not in front */ 1219 prev->next = q->next; 1220 q->next = ipfw_dyn_v[i]; 1221 ipfw_dyn_v[i] = q; 1222 } 1223 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1224 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1225 1226 #define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1227 #define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1228 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1229 switch (q->state) { 1230 case TH_SYN: /* opening */ 1231 q->expire = time_uptime + dyn_syn_lifetime; 1232 break; 1233 1234 case BOTH_SYN: /* move to established */ 1235 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1236 case BOTH_SYN | (TH_FIN << 8) : 1237 if (tcp) { 1238 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1239 u_int32_t ack = ntohl(tcp->th_ack); 1240 if (dir == MATCH_FORWARD) { 1241 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1242 q->ack_fwd = ack; 1243 else { /* ignore out-of-sequence */ 1244 break; 1245 } 1246 } else { 1247 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1248 q->ack_rev = ack; 1249 else { /* ignore out-of-sequence */ 1250 break; 1251 } 1252 } 1253 } 1254 q->expire = time_uptime + dyn_ack_lifetime; 1255 break; 1256 1257 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1258 if (dyn_fin_lifetime >= dyn_keepalive_period) 1259 dyn_fin_lifetime = dyn_keepalive_period - 1; 1260 q->expire = time_uptime + dyn_fin_lifetime; 1261 break; 1262 1263 default: 1264 #if 0 1265 /* 1266 * reset or some invalid combination, but can also 1267 * occur if we use keep-state the wrong way. 1268 */ 1269 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1270 printf("invalid state: 0x%x\n", q->state); 1271 #endif 1272 if (dyn_rst_lifetime >= dyn_keepalive_period) 1273 dyn_rst_lifetime = dyn_keepalive_period - 1; 1274 q->expire = time_uptime + dyn_rst_lifetime; 1275 break; 1276 } 1277 } else if (pkt->proto == IPPROTO_UDP) { 1278 q->expire = time_uptime + dyn_udp_lifetime; 1279 } else { 1280 /* other protocols */ 1281 q->expire = time_uptime + dyn_short_lifetime; 1282 } 1283 done: 1284 if (match_direction) 1285 *match_direction = dir; 1286 return q; 1287 } 1288 1289 static ipfw_dyn_rule * 1290 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1291 struct tcphdr *tcp) 1292 { 1293 ipfw_dyn_rule *q; 1294 1295 IPFW_DYN_LOCK(); 1296 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1297 if (q == NULL) 1298 IPFW_DYN_UNLOCK(); 1299 /* NB: return table locked when q is not NULL */ 1300 return q; 1301 } 1302 1303 static void 1304 realloc_dynamic_table(void) 1305 { 1306 IPFW_DYN_LOCK_ASSERT(); 1307 1308 /* 1309 * Try reallocation, make sure we have a power of 2 and do 1310 * not allow more than 64k entries. In case of overflow, 1311 * default to 1024. 1312 */ 1313 1314 if (dyn_buckets > 65536) 1315 dyn_buckets = 1024; 1316 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1317 dyn_buckets = curr_dyn_buckets; /* reset */ 1318 return; 1319 } 1320 curr_dyn_buckets = dyn_buckets; 1321 if (ipfw_dyn_v != NULL) 1322 free(ipfw_dyn_v, M_IPFW); 1323 for (;;) { 1324 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1325 M_IPFW, M_NOWAIT | M_ZERO); 1326 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1327 break; 1328 curr_dyn_buckets /= 2; 1329 } 1330 } 1331 1332 /** 1333 * Install state of type 'type' for a dynamic session. 1334 * The hash table contains two type of rules: 1335 * - regular rules (O_KEEP_STATE) 1336 * - rules for sessions with limited number of sess per user 1337 * (O_LIMIT). When they are created, the parent is 1338 * increased by 1, and decreased on delete. In this case, 1339 * the third parameter is the parent rule and not the chain. 1340 * - "parent" rules for the above (O_LIMIT_PARENT). 1341 */ 1342 static ipfw_dyn_rule * 1343 add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1344 { 1345 ipfw_dyn_rule *r; 1346 int i; 1347 1348 IPFW_DYN_LOCK_ASSERT(); 1349 1350 if (ipfw_dyn_v == NULL || 1351 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1352 realloc_dynamic_table(); 1353 if (ipfw_dyn_v == NULL) 1354 return NULL; /* failed ! */ 1355 } 1356 i = hash_packet(id); 1357 1358 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1359 if (r == NULL) { 1360 printf ("ipfw: sorry cannot allocate state\n"); 1361 return NULL; 1362 } 1363 1364 /* increase refcount on parent, and set pointer */ 1365 if (dyn_type == O_LIMIT) { 1366 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1367 if ( parent->dyn_type != O_LIMIT_PARENT) 1368 panic("invalid parent"); 1369 parent->count++; 1370 r->parent = parent; 1371 rule = parent->rule; 1372 } 1373 1374 r->id = *id; 1375 r->expire = time_uptime + dyn_syn_lifetime; 1376 r->rule = rule; 1377 r->dyn_type = dyn_type; 1378 r->pcnt = r->bcnt = 0; 1379 r->count = 0; 1380 1381 r->bucket = i; 1382 r->next = ipfw_dyn_v[i]; 1383 ipfw_dyn_v[i] = r; 1384 dyn_count++; 1385 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1386 dyn_type, 1387 (r->id.src_ip), (r->id.src_port), 1388 (r->id.dst_ip), (r->id.dst_port), 1389 dyn_count ); ) 1390 return r; 1391 } 1392 1393 /** 1394 * lookup dynamic parent rule using pkt and rule as search keys. 1395 * If the lookup fails, then install one. 1396 */ 1397 static ipfw_dyn_rule * 1398 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1399 { 1400 ipfw_dyn_rule *q; 1401 int i; 1402 1403 IPFW_DYN_LOCK_ASSERT(); 1404 1405 if (ipfw_dyn_v) { 1406 int is_v6 = IS_IP6_FLOW_ID(pkt); 1407 i = hash_packet( pkt ); 1408 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1409 if (q->dyn_type == O_LIMIT_PARENT && 1410 rule== q->rule && 1411 pkt->proto == q->id.proto && 1412 pkt->src_port == q->id.src_port && 1413 pkt->dst_port == q->id.dst_port && 1414 ( 1415 (is_v6 && 1416 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1417 &(q->id.src_ip6)) && 1418 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1419 &(q->id.dst_ip6))) || 1420 (!is_v6 && 1421 pkt->src_ip == q->id.src_ip && 1422 pkt->dst_ip == q->id.dst_ip) 1423 ) 1424 ) { 1425 q->expire = time_uptime + dyn_short_lifetime; 1426 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1427 return q; 1428 } 1429 } 1430 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1431 } 1432 1433 /** 1434 * Install dynamic state for rule type cmd->o.opcode 1435 * 1436 * Returns 1 (failure) if state is not installed because of errors or because 1437 * session limitations are enforced. 1438 */ 1439 static int 1440 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1441 struct ip_fw_args *args, uint32_t tablearg) 1442 { 1443 static int last_log; 1444 ipfw_dyn_rule *q; 1445 struct in_addr da; 1446 char src[48], dst[48]; 1447 1448 src[0] = '\0'; 1449 dst[0] = '\0'; 1450 1451 DEB( 1452 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1453 __func__, cmd->o.opcode, 1454 (args->f_id.src_ip), (args->f_id.src_port), 1455 (args->f_id.dst_ip), (args->f_id.dst_port)); 1456 ) 1457 1458 IPFW_DYN_LOCK(); 1459 1460 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1461 1462 if (q != NULL) { /* should never occur */ 1463 if (last_log != time_uptime) { 1464 last_log = time_uptime; 1465 printf("ipfw: %s: entry already present, done\n", 1466 __func__); 1467 } 1468 IPFW_DYN_UNLOCK(); 1469 return (0); 1470 } 1471 1472 if (dyn_count >= dyn_max) 1473 /* Run out of slots, try to remove any expired rule. */ 1474 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1475 1476 if (dyn_count >= dyn_max) { 1477 if (last_log != time_uptime) { 1478 last_log = time_uptime; 1479 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1480 } 1481 IPFW_DYN_UNLOCK(); 1482 return (1); /* cannot install, notify caller */ 1483 } 1484 1485 switch (cmd->o.opcode) { 1486 case O_KEEP_STATE: /* bidir rule */ 1487 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1488 break; 1489 1490 case O_LIMIT: { /* limit number of sessions */ 1491 struct ipfw_flow_id id; 1492 ipfw_dyn_rule *parent; 1493 uint32_t conn_limit; 1494 uint16_t limit_mask = cmd->limit_mask; 1495 1496 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1497 tablearg : cmd->conn_limit; 1498 1499 DEB( 1500 if (cmd->conn_limit == IP_FW_TABLEARG) 1501 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1502 "(tablearg)\n", __func__, conn_limit); 1503 else 1504 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1505 __func__, conn_limit); 1506 ) 1507 1508 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1509 id.proto = args->f_id.proto; 1510 id.addr_type = args->f_id.addr_type; 1511 id.fib = M_GETFIB(args->m); 1512 1513 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1514 if (limit_mask & DYN_SRC_ADDR) 1515 id.src_ip6 = args->f_id.src_ip6; 1516 if (limit_mask & DYN_DST_ADDR) 1517 id.dst_ip6 = args->f_id.dst_ip6; 1518 } else { 1519 if (limit_mask & DYN_SRC_ADDR) 1520 id.src_ip = args->f_id.src_ip; 1521 if (limit_mask & DYN_DST_ADDR) 1522 id.dst_ip = args->f_id.dst_ip; 1523 } 1524 if (limit_mask & DYN_SRC_PORT) 1525 id.src_port = args->f_id.src_port; 1526 if (limit_mask & DYN_DST_PORT) 1527 id.dst_port = args->f_id.dst_port; 1528 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1529 printf("ipfw: %s: add parent failed\n", __func__); 1530 IPFW_DYN_UNLOCK(); 1531 return (1); 1532 } 1533 1534 if (parent->count >= conn_limit) { 1535 /* See if we can remove some expired rule. */ 1536 remove_dyn_rule(rule, parent); 1537 if (parent->count >= conn_limit) { 1538 if (fw_verbose && last_log != time_uptime) { 1539 last_log = time_uptime; 1540 #ifdef INET6 1541 /* 1542 * XXX IPv6 flows are not 1543 * supported yet. 1544 */ 1545 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1546 char ip6buf[INET6_ADDRSTRLEN]; 1547 snprintf(src, sizeof(src), 1548 "[%s]", ip6_sprintf(ip6buf, 1549 &args->f_id.src_ip6)); 1550 snprintf(dst, sizeof(dst), 1551 "[%s]", ip6_sprintf(ip6buf, 1552 &args->f_id.dst_ip6)); 1553 } else 1554 #endif 1555 { 1556 da.s_addr = 1557 htonl(args->f_id.src_ip); 1558 inet_ntoa_r(da, src); 1559 da.s_addr = 1560 htonl(args->f_id.dst_ip); 1561 inet_ntoa_r(da, dst); 1562 } 1563 log(LOG_SECURITY | LOG_DEBUG, 1564 "ipfw: %d %s %s:%u -> %s:%u, %s\n", 1565 parent->rule->rulenum, 1566 "drop session", 1567 src, (args->f_id.src_port), 1568 dst, (args->f_id.dst_port), 1569 "too many entries"); 1570 } 1571 IPFW_DYN_UNLOCK(); 1572 return (1); 1573 } 1574 } 1575 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1576 break; 1577 } 1578 default: 1579 printf("ipfw: %s: unknown dynamic rule type %u\n", 1580 __func__, cmd->o.opcode); 1581 IPFW_DYN_UNLOCK(); 1582 return (1); 1583 } 1584 1585 /* XXX just set lifetime */ 1586 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1587 1588 IPFW_DYN_UNLOCK(); 1589 return (0); 1590 } 1591 1592 /* 1593 * Generate a TCP packet, containing either a RST or a keepalive. 1594 * When flags & TH_RST, we are sending a RST packet, because of a 1595 * "reset" action matched the packet. 1596 * Otherwise we are sending a keepalive, and flags & TH_ 1597 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1598 * so that MAC can label the reply appropriately. 1599 */ 1600 static struct mbuf * 1601 send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1602 u_int32_t ack, int flags) 1603 { 1604 struct mbuf *m; 1605 struct ip *ip; 1606 struct tcphdr *tcp; 1607 1608 MGETHDR(m, M_DONTWAIT, MT_DATA); 1609 if (m == 0) 1610 return (NULL); 1611 m->m_pkthdr.rcvif = (struct ifnet *)0; 1612 1613 M_SETFIB(m, id->fib); 1614 #ifdef MAC 1615 if (replyto != NULL) 1616 mac_create_mbuf_netlayer(replyto, m); 1617 else 1618 mac_create_mbuf_from_firewall(m); 1619 #else 1620 (void)replyto; /* don't warn about unused arg */ 1621 #endif 1622 1623 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1624 m->m_data += max_linkhdr; 1625 1626 ip = mtod(m, struct ip *); 1627 bzero(ip, m->m_len); 1628 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1629 ip->ip_p = IPPROTO_TCP; 1630 tcp->th_off = 5; 1631 /* 1632 * Assume we are sending a RST (or a keepalive in the reverse 1633 * direction), swap src and destination addresses and ports. 1634 */ 1635 ip->ip_src.s_addr = htonl(id->dst_ip); 1636 ip->ip_dst.s_addr = htonl(id->src_ip); 1637 tcp->th_sport = htons(id->dst_port); 1638 tcp->th_dport = htons(id->src_port); 1639 if (flags & TH_RST) { /* we are sending a RST */ 1640 if (flags & TH_ACK) { 1641 tcp->th_seq = htonl(ack); 1642 tcp->th_ack = htonl(0); 1643 tcp->th_flags = TH_RST; 1644 } else { 1645 if (flags & TH_SYN) 1646 seq++; 1647 tcp->th_seq = htonl(0); 1648 tcp->th_ack = htonl(seq); 1649 tcp->th_flags = TH_RST | TH_ACK; 1650 } 1651 } else { 1652 /* 1653 * We are sending a keepalive. flags & TH_SYN determines 1654 * the direction, forward if set, reverse if clear. 1655 * NOTE: seq and ack are always assumed to be correct 1656 * as set by the caller. This may be confusing... 1657 */ 1658 if (flags & TH_SYN) { 1659 /* 1660 * we have to rewrite the correct addresses! 1661 */ 1662 ip->ip_dst.s_addr = htonl(id->dst_ip); 1663 ip->ip_src.s_addr = htonl(id->src_ip); 1664 tcp->th_dport = htons(id->dst_port); 1665 tcp->th_sport = htons(id->src_port); 1666 } 1667 tcp->th_seq = htonl(seq); 1668 tcp->th_ack = htonl(ack); 1669 tcp->th_flags = TH_ACK; 1670 } 1671 /* 1672 * set ip_len to the payload size so we can compute 1673 * the tcp checksum on the pseudoheader 1674 * XXX check this, could save a couple of words ? 1675 */ 1676 ip->ip_len = htons(sizeof(struct tcphdr)); 1677 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1678 /* 1679 * now fill fields left out earlier 1680 */ 1681 ip->ip_ttl = ip_defttl; 1682 ip->ip_len = m->m_pkthdr.len; 1683 m->m_flags |= M_SKIP_FIREWALL; 1684 return (m); 1685 } 1686 1687 /* 1688 * sends a reject message, consuming the mbuf passed as an argument. 1689 */ 1690 static void 1691 send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1692 { 1693 1694 #if 0 1695 /* XXX When ip is not guaranteed to be at mtod() we will 1696 * need to account for this */ 1697 * The mbuf will however be thrown away so we can adjust it. 1698 * Remember we did an m_pullup on it already so we 1699 * can make some assumptions about contiguousness. 1700 */ 1701 if (args->L3offset) 1702 m_adj(m, args->L3offset); 1703 #endif 1704 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1705 /* We need the IP header in host order for icmp_error(). */ 1706 if (args->eh != NULL) { 1707 ip->ip_len = ntohs(ip->ip_len); 1708 ip->ip_off = ntohs(ip->ip_off); 1709 } 1710 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1711 } else if (args->f_id.proto == IPPROTO_TCP) { 1712 struct tcphdr *const tcp = 1713 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1714 if ( (tcp->th_flags & TH_RST) == 0) { 1715 struct mbuf *m; 1716 m = send_pkt(args->m, &(args->f_id), 1717 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1718 tcp->th_flags | TH_RST); 1719 if (m != NULL) 1720 ip_output(m, NULL, NULL, 0, NULL, NULL); 1721 } 1722 m_freem(args->m); 1723 } else 1724 m_freem(args->m); 1725 args->m = NULL; 1726 } 1727 1728 /** 1729 * 1730 * Given an ip_fw *, lookup_next_rule will return a pointer 1731 * to the next rule, which can be either the jump 1732 * target (for skipto instructions) or the next one in the list (in 1733 * all other cases including a missing jump target). 1734 * The result is also written in the "next_rule" field of the rule. 1735 * Backward jumps are not allowed, so start looking from the next 1736 * rule... 1737 * 1738 * This never returns NULL -- in case we do not have an exact match, 1739 * the next rule is returned. When the ruleset is changed, 1740 * pointers are flushed so we are always correct. 1741 */ 1742 1743 static struct ip_fw * 1744 lookup_next_rule(struct ip_fw *me, u_int32_t tablearg) 1745 { 1746 struct ip_fw *rule = NULL; 1747 ipfw_insn *cmd; 1748 u_int16_t rulenum; 1749 1750 /* look for action, in case it is a skipto */ 1751 cmd = ACTION_PTR(me); 1752 if (cmd->opcode == O_LOG) 1753 cmd += F_LEN(cmd); 1754 if (cmd->opcode == O_ALTQ) 1755 cmd += F_LEN(cmd); 1756 if (cmd->opcode == O_TAG) 1757 cmd += F_LEN(cmd); 1758 if (cmd->opcode == O_SKIPTO ) { 1759 if (tablearg != 0) { 1760 rulenum = (u_int16_t)tablearg; 1761 } else { 1762 rulenum = cmd->arg1; 1763 } 1764 for (rule = me->next; rule ; rule = rule->next) { 1765 if (rule->rulenum >= rulenum) { 1766 break; 1767 } 1768 } 1769 } 1770 if (rule == NULL) /* failure or not a skipto */ 1771 rule = me->next; 1772 me->next_rule = rule; 1773 return rule; 1774 } 1775 1776 static int 1777 add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1778 uint8_t mlen, uint32_t value) 1779 { 1780 struct radix_node_head *rnh; 1781 struct table_entry *ent; 1782 struct radix_node *rn; 1783 1784 if (tbl >= IPFW_TABLES_MAX) 1785 return (EINVAL); 1786 rnh = ch->tables[tbl]; 1787 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1788 if (ent == NULL) 1789 return (ENOMEM); 1790 ent->value = value; 1791 ent->addr.sin_len = ent->mask.sin_len = 8; 1792 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1793 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1794 IPFW_WLOCK(ch); 1795 rn = rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent); 1796 if (rn == NULL) { 1797 IPFW_WUNLOCK(ch); 1798 free(ent, M_IPFW_TBL); 1799 return (EEXIST); 1800 } 1801 IPFW_WUNLOCK(ch); 1802 return (0); 1803 } 1804 1805 static int 1806 del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1807 uint8_t mlen) 1808 { 1809 struct radix_node_head *rnh; 1810 struct table_entry *ent; 1811 struct sockaddr_in sa, mask; 1812 1813 if (tbl >= IPFW_TABLES_MAX) 1814 return (EINVAL); 1815 rnh = ch->tables[tbl]; 1816 sa.sin_len = mask.sin_len = 8; 1817 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1818 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1819 IPFW_WLOCK(ch); 1820 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1821 if (ent == NULL) { 1822 IPFW_WUNLOCK(ch); 1823 return (ESRCH); 1824 } 1825 IPFW_WUNLOCK(ch); 1826 free(ent, M_IPFW_TBL); 1827 return (0); 1828 } 1829 1830 static int 1831 flush_table_entry(struct radix_node *rn, void *arg) 1832 { 1833 struct radix_node_head * const rnh = arg; 1834 struct table_entry *ent; 1835 1836 ent = (struct table_entry *) 1837 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1838 if (ent != NULL) 1839 free(ent, M_IPFW_TBL); 1840 return (0); 1841 } 1842 1843 static int 1844 flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1845 { 1846 struct radix_node_head *rnh; 1847 1848 IPFW_WLOCK_ASSERT(ch); 1849 1850 if (tbl >= IPFW_TABLES_MAX) 1851 return (EINVAL); 1852 rnh = ch->tables[tbl]; 1853 KASSERT(rnh != NULL, ("NULL IPFW table")); 1854 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1855 return (0); 1856 } 1857 1858 static void 1859 flush_tables(struct ip_fw_chain *ch) 1860 { 1861 uint16_t tbl; 1862 1863 IPFW_WLOCK_ASSERT(ch); 1864 1865 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1866 flush_table(ch, tbl); 1867 } 1868 1869 static int 1870 init_tables(struct ip_fw_chain *ch) 1871 { 1872 int i; 1873 uint16_t j; 1874 1875 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1876 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1877 for (j = 0; j < i; j++) { 1878 (void) flush_table(ch, j); 1879 } 1880 return (ENOMEM); 1881 } 1882 } 1883 return (0); 1884 } 1885 1886 static int 1887 lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1888 uint32_t *val) 1889 { 1890 struct radix_node_head *rnh; 1891 struct table_entry *ent; 1892 struct sockaddr_in sa; 1893 1894 if (tbl >= IPFW_TABLES_MAX) 1895 return (0); 1896 rnh = ch->tables[tbl]; 1897 sa.sin_len = 8; 1898 sa.sin_addr.s_addr = addr; 1899 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1900 if (ent != NULL) { 1901 *val = ent->value; 1902 return (1); 1903 } 1904 return (0); 1905 } 1906 1907 static int 1908 count_table_entry(struct radix_node *rn, void *arg) 1909 { 1910 u_int32_t * const cnt = arg; 1911 1912 (*cnt)++; 1913 return (0); 1914 } 1915 1916 static int 1917 count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1918 { 1919 struct radix_node_head *rnh; 1920 1921 if (tbl >= IPFW_TABLES_MAX) 1922 return (EINVAL); 1923 rnh = ch->tables[tbl]; 1924 *cnt = 0; 1925 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1926 return (0); 1927 } 1928 1929 static int 1930 dump_table_entry(struct radix_node *rn, void *arg) 1931 { 1932 struct table_entry * const n = (struct table_entry *)rn; 1933 ipfw_table * const tbl = arg; 1934 ipfw_table_entry *ent; 1935 1936 if (tbl->cnt == tbl->size) 1937 return (1); 1938 ent = &tbl->ent[tbl->cnt]; 1939 ent->tbl = tbl->tbl; 1940 if (in_nullhost(n->mask.sin_addr)) 1941 ent->masklen = 0; 1942 else 1943 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1944 ent->addr = n->addr.sin_addr.s_addr; 1945 ent->value = n->value; 1946 tbl->cnt++; 1947 return (0); 1948 } 1949 1950 static int 1951 dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1952 { 1953 struct radix_node_head *rnh; 1954 1955 if (tbl->tbl >= IPFW_TABLES_MAX) 1956 return (EINVAL); 1957 rnh = ch->tables[tbl->tbl]; 1958 tbl->cnt = 0; 1959 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1960 return (0); 1961 } 1962 1963 static void 1964 fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1965 { 1966 struct ucred *cr; 1967 1968 cr = inp->inp_cred; 1969 ugp->fw_prid = jailed(cr) ? cr->cr_prison->pr_id : -1; 1970 ugp->fw_uid = cr->cr_uid; 1971 ugp->fw_ngroups = cr->cr_ngroups; 1972 bcopy(cr->cr_groups, ugp->fw_groups, sizeof(ugp->fw_groups)); 1973 } 1974 1975 static int 1976 check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 1977 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 1978 u_int16_t src_port, struct ip_fw_ugid *ugp, int *ugid_lookupp, 1979 struct inpcb *inp) 1980 { 1981 struct inpcbinfo *pi; 1982 int wildcard; 1983 struct inpcb *pcb; 1984 int match; 1985 gid_t *gp; 1986 1987 /* 1988 * Check to see if the UDP or TCP stack supplied us with 1989 * the PCB. If so, rather then holding a lock and looking 1990 * up the PCB, we can use the one that was supplied. 1991 */ 1992 if (inp && *ugid_lookupp == 0) { 1993 INP_LOCK_ASSERT(inp); 1994 if (inp->inp_socket != NULL) { 1995 fill_ugid_cache(inp, ugp); 1996 *ugid_lookupp = 1; 1997 } else 1998 *ugid_lookupp = -1; 1999 } 2000 /* 2001 * If we have already been here and the packet has no 2002 * PCB entry associated with it, then we can safely 2003 * assume that this is a no match. 2004 */ 2005 if (*ugid_lookupp == -1) 2006 return (0); 2007 if (proto == IPPROTO_TCP) { 2008 wildcard = 0; 2009 pi = &tcbinfo; 2010 } else if (proto == IPPROTO_UDP) { 2011 wildcard = INPLOOKUP_WILDCARD; 2012 pi = &udbinfo; 2013 } else 2014 return 0; 2015 match = 0; 2016 if (*ugid_lookupp == 0) { 2017 INP_INFO_RLOCK(pi); 2018 pcb = (oif) ? 2019 in_pcblookup_hash(pi, 2020 dst_ip, htons(dst_port), 2021 src_ip, htons(src_port), 2022 wildcard, oif) : 2023 in_pcblookup_hash(pi, 2024 src_ip, htons(src_port), 2025 dst_ip, htons(dst_port), 2026 wildcard, NULL); 2027 if (pcb != NULL) { 2028 fill_ugid_cache(pcb, ugp); 2029 *ugid_lookupp = 1; 2030 } 2031 INP_INFO_RUNLOCK(pi); 2032 if (*ugid_lookupp == 0) { 2033 /* 2034 * If the lookup did not yield any results, there 2035 * is no sense in coming back and trying again. So 2036 * we can set lookup to -1 and ensure that we wont 2037 * bother the pcb system again. 2038 */ 2039 *ugid_lookupp = -1; 2040 return (0); 2041 } 2042 } 2043 if (insn->o.opcode == O_UID) 2044 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2045 else if (insn->o.opcode == O_GID) { 2046 for (gp = ugp->fw_groups; 2047 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2048 if (*gp == (gid_t)insn->d[0]) { 2049 match = 1; 2050 break; 2051 } 2052 } else if (insn->o.opcode == O_JAIL) 2053 match = (ugp->fw_prid == (int)insn->d[0]); 2054 return match; 2055 } 2056 2057 /* 2058 * The main check routine for the firewall. 2059 * 2060 * All arguments are in args so we can modify them and return them 2061 * back to the caller. 2062 * 2063 * Parameters: 2064 * 2065 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2066 * Starts with the IP header. 2067 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2068 * args->L3offset Number of bytes bypassed if we came from L2. 2069 * e.g. often sizeof(eh) ** NOTYET ** 2070 * args->oif Outgoing interface, or NULL if packet is incoming. 2071 * The incoming interface is in the mbuf. (in) 2072 * args->divert_rule (in/out) 2073 * Skip up to the first rule past this rule number; 2074 * upon return, non-zero port number for divert or tee. 2075 * 2076 * args->rule Pointer to the last matching rule (in/out) 2077 * args->next_hop Socket we are forwarding to (out). 2078 * args->f_id Addresses grabbed from the packet (out) 2079 * args->cookie a cookie depending on rule action 2080 * 2081 * Return value: 2082 * 2083 * IP_FW_PASS the packet must be accepted 2084 * IP_FW_DENY the packet must be dropped 2085 * IP_FW_DIVERT divert packet, port in m_tag 2086 * IP_FW_TEE tee packet, port in m_tag 2087 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2088 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2089 * 2090 */ 2091 int 2092 ipfw_chk(struct ip_fw_args *args) 2093 { 2094 /* 2095 * Local variables holding state during the processing of a packet: 2096 * 2097 * IMPORTANT NOTE: to speed up the processing of rules, there 2098 * are some assumption on the values of the variables, which 2099 * are documented here. Should you change them, please check 2100 * the implementation of the various instructions to make sure 2101 * that they still work. 2102 * 2103 * args->eh The MAC header. It is non-null for a layer2 2104 * packet, it is NULL for a layer-3 packet. 2105 * **notyet** 2106 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2107 * 2108 * m | args->m Pointer to the mbuf, as received from the caller. 2109 * It may change if ipfw_chk() does an m_pullup, or if it 2110 * consumes the packet because it calls send_reject(). 2111 * XXX This has to change, so that ipfw_chk() never modifies 2112 * or consumes the buffer. 2113 * ip is the beginning of the ip(4 or 6) header. 2114 * Calculated by adding the L3offset to the start of data. 2115 * (Until we start using L3offset, the packet is 2116 * supposed to start with the ip header). 2117 */ 2118 struct mbuf *m = args->m; 2119 struct ip *ip = mtod(m, struct ip *); 2120 2121 /* 2122 * For rules which contain uid/gid or jail constraints, cache 2123 * a copy of the users credentials after the pcb lookup has been 2124 * executed. This will speed up the processing of rules with 2125 * these types of constraints, as well as decrease contention 2126 * on pcb related locks. 2127 */ 2128 struct ip_fw_ugid fw_ugid_cache; 2129 int ugid_lookup = 0; 2130 2131 /* 2132 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2133 * associated with a packet input on a divert socket. This 2134 * will allow to distinguish traffic and its direction when 2135 * it originates from a divert socket. 2136 */ 2137 u_int divinput_flags = 0; 2138 2139 /* 2140 * oif | args->oif If NULL, ipfw_chk has been called on the 2141 * inbound path (ether_input, ip_input). 2142 * If non-NULL, ipfw_chk has been called on the outbound path 2143 * (ether_output, ip_output). 2144 */ 2145 struct ifnet *oif = args->oif; 2146 2147 struct ip_fw *f = NULL; /* matching rule */ 2148 int retval = 0; 2149 2150 /* 2151 * hlen The length of the IP header. 2152 */ 2153 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2154 2155 /* 2156 * offset The offset of a fragment. offset != 0 means that 2157 * we have a fragment at this offset of an IPv4 packet. 2158 * offset == 0 means that (if this is an IPv4 packet) 2159 * this is the first or only fragment. 2160 * For IPv6 offset == 0 means there is no Fragment Header. 2161 * If offset != 0 for IPv6 always use correct mask to 2162 * get the correct offset because we add IP6F_MORE_FRAG 2163 * to be able to dectect the first fragment which would 2164 * otherwise have offset = 0. 2165 */ 2166 u_short offset = 0; 2167 2168 /* 2169 * Local copies of addresses. They are only valid if we have 2170 * an IP packet. 2171 * 2172 * proto The protocol. Set to 0 for non-ip packets, 2173 * or to the protocol read from the packet otherwise. 2174 * proto != 0 means that we have an IPv4 packet. 2175 * 2176 * src_port, dst_port port numbers, in HOST format. Only 2177 * valid for TCP and UDP packets. 2178 * 2179 * src_ip, dst_ip ip addresses, in NETWORK format. 2180 * Only valid for IPv4 packets. 2181 */ 2182 u_int8_t proto; 2183 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2184 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2185 u_int16_t ip_len=0; 2186 int pktlen; 2187 u_int16_t etype = 0; /* Host order stored ether type */ 2188 2189 /* 2190 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2191 * MATCH_NONE when checked and not matched (q = NULL), 2192 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2193 */ 2194 int dyn_dir = MATCH_UNKNOWN; 2195 ipfw_dyn_rule *q = NULL; 2196 struct ip_fw_chain *chain = &layer3_chain; 2197 struct m_tag *mtag; 2198 2199 /* 2200 * We store in ulp a pointer to the upper layer protocol header. 2201 * In the ipv4 case this is easy to determine from the header, 2202 * but for ipv6 we might have some additional headers in the middle. 2203 * ulp is NULL if not found. 2204 */ 2205 void *ulp = NULL; /* upper layer protocol pointer. */ 2206 /* XXX ipv6 variables */ 2207 int is_ipv6 = 0; 2208 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2209 /* end of ipv6 variables */ 2210 int is_ipv4 = 0; 2211 2212 if (m->m_flags & M_SKIP_FIREWALL) 2213 return (IP_FW_PASS); /* accept */ 2214 2215 dst_ip.s_addr = 0; /* make sure it is initialized */ 2216 src_ip.s_addr = 0; /* make sure it is initialized */ 2217 pktlen = m->m_pkthdr.len; 2218 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 2219 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2220 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2221 2222 /* 2223 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2224 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2225 * pointer might become stale after other pullups (but we never use it 2226 * this way). 2227 */ 2228 #define PULLUP_TO(_len, p, T) \ 2229 do { \ 2230 int x = (_len) + sizeof(T); \ 2231 if ((m)->m_len < x) { \ 2232 args->m = m = m_pullup(m, x); \ 2233 if (m == NULL) \ 2234 goto pullup_failed; \ 2235 } \ 2236 p = (mtod(m, char *) + (_len)); \ 2237 } while (0) 2238 2239 /* 2240 * if we have an ether header, 2241 */ 2242 if (args->eh) 2243 etype = ntohs(args->eh->ether_type); 2244 2245 /* Identify IP packets and fill up variables. */ 2246 if (pktlen >= sizeof(struct ip6_hdr) && 2247 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2248 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2249 is_ipv6 = 1; 2250 args->f_id.addr_type = 6; 2251 hlen = sizeof(struct ip6_hdr); 2252 proto = ip6->ip6_nxt; 2253 2254 /* Search extension headers to find upper layer protocols */ 2255 while (ulp == NULL) { 2256 switch (proto) { 2257 case IPPROTO_ICMPV6: 2258 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2259 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2260 break; 2261 2262 case IPPROTO_TCP: 2263 PULLUP_TO(hlen, ulp, struct tcphdr); 2264 dst_port = TCP(ulp)->th_dport; 2265 src_port = TCP(ulp)->th_sport; 2266 args->f_id.flags = TCP(ulp)->th_flags; 2267 break; 2268 2269 case IPPROTO_SCTP: 2270 PULLUP_TO(hlen, ulp, struct sctphdr); 2271 src_port = SCTP(ulp)->src_port; 2272 dst_port = SCTP(ulp)->dest_port; 2273 break; 2274 2275 case IPPROTO_UDP: 2276 PULLUP_TO(hlen, ulp, struct udphdr); 2277 dst_port = UDP(ulp)->uh_dport; 2278 src_port = UDP(ulp)->uh_sport; 2279 break; 2280 2281 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2282 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2283 ext_hd |= EXT_HOPOPTS; 2284 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2285 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2286 ulp = NULL; 2287 break; 2288 2289 case IPPROTO_ROUTING: /* RFC 2460 */ 2290 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2291 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2292 case 0: 2293 ext_hd |= EXT_RTHDR0; 2294 break; 2295 case 2: 2296 ext_hd |= EXT_RTHDR2; 2297 break; 2298 default: 2299 printf("IPFW2: IPV6 - Unknown Routing " 2300 "Header type(%d)\n", 2301 ((struct ip6_rthdr *)ulp)->ip6r_type); 2302 if (fw_deny_unknown_exthdrs) 2303 return (IP_FW_DENY); 2304 break; 2305 } 2306 ext_hd |= EXT_ROUTING; 2307 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2308 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2309 ulp = NULL; 2310 break; 2311 2312 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2313 PULLUP_TO(hlen, ulp, struct ip6_frag); 2314 ext_hd |= EXT_FRAGMENT; 2315 hlen += sizeof (struct ip6_frag); 2316 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2317 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2318 IP6F_OFF_MASK; 2319 /* Add IP6F_MORE_FRAG for offset of first 2320 * fragment to be != 0. */ 2321 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2322 IP6F_MORE_FRAG; 2323 if (offset == 0) { 2324 printf("IPFW2: IPV6 - Invalid Fragment " 2325 "Header\n"); 2326 if (fw_deny_unknown_exthdrs) 2327 return (IP_FW_DENY); 2328 break; 2329 } 2330 args->f_id.frag_id6 = 2331 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2332 ulp = NULL; 2333 break; 2334 2335 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2336 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2337 ext_hd |= EXT_DSTOPTS; 2338 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2339 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2340 ulp = NULL; 2341 break; 2342 2343 case IPPROTO_AH: /* RFC 2402 */ 2344 PULLUP_TO(hlen, ulp, struct ip6_ext); 2345 ext_hd |= EXT_AH; 2346 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2347 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2348 ulp = NULL; 2349 break; 2350 2351 case IPPROTO_ESP: /* RFC 2406 */ 2352 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2353 /* Anything past Seq# is variable length and 2354 * data past this ext. header is encrypted. */ 2355 ext_hd |= EXT_ESP; 2356 break; 2357 2358 case IPPROTO_NONE: /* RFC 2460 */ 2359 /* 2360 * Packet ends here, and IPv6 header has 2361 * already been pulled up. If ip6e_len!=0 2362 * then octets must be ignored. 2363 */ 2364 ulp = ip; /* non-NULL to get out of loop. */ 2365 break; 2366 2367 case IPPROTO_OSPFIGP: 2368 /* XXX OSPF header check? */ 2369 PULLUP_TO(hlen, ulp, struct ip6_ext); 2370 break; 2371 2372 case IPPROTO_PIM: 2373 /* XXX PIM header check? */ 2374 PULLUP_TO(hlen, ulp, struct pim); 2375 break; 2376 2377 case IPPROTO_CARP: 2378 PULLUP_TO(hlen, ulp, struct carp_header); 2379 if (((struct carp_header *)ulp)->carp_version != 2380 CARP_VERSION) 2381 return (IP_FW_DENY); 2382 if (((struct carp_header *)ulp)->carp_type != 2383 CARP_ADVERTISEMENT) 2384 return (IP_FW_DENY); 2385 break; 2386 2387 case IPPROTO_IPV6: /* RFC 2893 */ 2388 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2389 break; 2390 2391 case IPPROTO_IPV4: /* RFC 2893 */ 2392 PULLUP_TO(hlen, ulp, struct ip); 2393 break; 2394 2395 default: 2396 printf("IPFW2: IPV6 - Unknown Extension " 2397 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2398 if (fw_deny_unknown_exthdrs) 2399 return (IP_FW_DENY); 2400 PULLUP_TO(hlen, ulp, struct ip6_ext); 2401 break; 2402 } /*switch */ 2403 } 2404 ip = mtod(m, struct ip *); 2405 ip6 = (struct ip6_hdr *)ip; 2406 args->f_id.src_ip6 = ip6->ip6_src; 2407 args->f_id.dst_ip6 = ip6->ip6_dst; 2408 args->f_id.src_ip = 0; 2409 args->f_id.dst_ip = 0; 2410 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2411 } else if (pktlen >= sizeof(struct ip) && 2412 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2413 is_ipv4 = 1; 2414 hlen = ip->ip_hl << 2; 2415 args->f_id.addr_type = 4; 2416 2417 /* 2418 * Collect parameters into local variables for faster matching. 2419 */ 2420 proto = ip->ip_p; 2421 src_ip = ip->ip_src; 2422 dst_ip = ip->ip_dst; 2423 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2424 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2425 ip_len = ntohs(ip->ip_len); 2426 } else { 2427 offset = ip->ip_off & IP_OFFMASK; 2428 ip_len = ip->ip_len; 2429 } 2430 pktlen = ip_len < pktlen ? ip_len : pktlen; 2431 2432 if (offset == 0) { 2433 switch (proto) { 2434 case IPPROTO_TCP: 2435 PULLUP_TO(hlen, ulp, struct tcphdr); 2436 dst_port = TCP(ulp)->th_dport; 2437 src_port = TCP(ulp)->th_sport; 2438 args->f_id.flags = TCP(ulp)->th_flags; 2439 break; 2440 2441 case IPPROTO_UDP: 2442 PULLUP_TO(hlen, ulp, struct udphdr); 2443 dst_port = UDP(ulp)->uh_dport; 2444 src_port = UDP(ulp)->uh_sport; 2445 break; 2446 2447 case IPPROTO_ICMP: 2448 PULLUP_TO(hlen, ulp, struct icmphdr); 2449 args->f_id.flags = ICMP(ulp)->icmp_type; 2450 break; 2451 2452 default: 2453 break; 2454 } 2455 } 2456 2457 ip = mtod(m, struct ip *); 2458 args->f_id.src_ip = ntohl(src_ip.s_addr); 2459 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2460 } 2461 #undef PULLUP_TO 2462 if (proto) { /* we may have port numbers, store them */ 2463 args->f_id.proto = proto; 2464 args->f_id.src_port = src_port = ntohs(src_port); 2465 args->f_id.dst_port = dst_port = ntohs(dst_port); 2466 } 2467 2468 IPFW_RLOCK(chain); 2469 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2470 if (args->rule) { 2471 /* 2472 * Packet has already been tagged. Look for the next rule 2473 * to restart processing. 2474 * 2475 * If fw_one_pass != 0 then just accept it. 2476 * XXX should not happen here, but optimized out in 2477 * the caller. 2478 */ 2479 if (fw_one_pass) { 2480 IPFW_RUNLOCK(chain); 2481 return (IP_FW_PASS); 2482 } 2483 2484 f = args->rule->next_rule; 2485 if (f == NULL) 2486 f = lookup_next_rule(args->rule, 0); 2487 } else { 2488 /* 2489 * Find the starting rule. It can be either the first 2490 * one, or the one after divert_rule if asked so. 2491 */ 2492 int skipto = mtag ? divert_cookie(mtag) : 0; 2493 2494 f = chain->rules; 2495 if (args->eh == NULL && skipto != 0) { 2496 if (skipto >= IPFW_DEFAULT_RULE) { 2497 IPFW_RUNLOCK(chain); 2498 return (IP_FW_DENY); /* invalid */ 2499 } 2500 while (f && f->rulenum <= skipto) 2501 f = f->next; 2502 if (f == NULL) { /* drop packet */ 2503 IPFW_RUNLOCK(chain); 2504 return (IP_FW_DENY); 2505 } 2506 } 2507 } 2508 /* reset divert rule to avoid confusion later */ 2509 if (mtag) { 2510 divinput_flags = divert_info(mtag) & 2511 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2512 m_tag_delete(m, mtag); 2513 } 2514 2515 /* 2516 * Now scan the rules, and parse microinstructions for each rule. 2517 */ 2518 for (; f; f = f->next) { 2519 ipfw_insn *cmd; 2520 uint32_t tablearg = 0; 2521 int l, cmdlen, skip_or; /* skip rest of OR block */ 2522 2523 again: 2524 if (set_disable & (1 << f->set) ) 2525 continue; 2526 2527 skip_or = 0; 2528 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2529 l -= cmdlen, cmd += cmdlen) { 2530 int match; 2531 2532 /* 2533 * check_body is a jump target used when we find a 2534 * CHECK_STATE, and need to jump to the body of 2535 * the target rule. 2536 */ 2537 2538 check_body: 2539 cmdlen = F_LEN(cmd); 2540 /* 2541 * An OR block (insn_1 || .. || insn_n) has the 2542 * F_OR bit set in all but the last instruction. 2543 * The first match will set "skip_or", and cause 2544 * the following instructions to be skipped until 2545 * past the one with the F_OR bit clear. 2546 */ 2547 if (skip_or) { /* skip this instruction */ 2548 if ((cmd->len & F_OR) == 0) 2549 skip_or = 0; /* next one is good */ 2550 continue; 2551 } 2552 match = 0; /* set to 1 if we succeed */ 2553 2554 switch (cmd->opcode) { 2555 /* 2556 * The first set of opcodes compares the packet's 2557 * fields with some pattern, setting 'match' if a 2558 * match is found. At the end of the loop there is 2559 * logic to deal with F_NOT and F_OR flags associated 2560 * with the opcode. 2561 */ 2562 case O_NOP: 2563 match = 1; 2564 break; 2565 2566 case O_FORWARD_MAC: 2567 printf("ipfw: opcode %d unimplemented\n", 2568 cmd->opcode); 2569 break; 2570 2571 case O_GID: 2572 case O_UID: 2573 case O_JAIL: 2574 /* 2575 * We only check offset == 0 && proto != 0, 2576 * as this ensures that we have a 2577 * packet with the ports info. 2578 */ 2579 if (offset!=0) 2580 break; 2581 if (is_ipv6) /* XXX to be fixed later */ 2582 break; 2583 if (proto == IPPROTO_TCP || 2584 proto == IPPROTO_UDP) 2585 match = check_uidgid( 2586 (ipfw_insn_u32 *)cmd, 2587 proto, oif, 2588 dst_ip, dst_port, 2589 src_ip, src_port, &fw_ugid_cache, 2590 &ugid_lookup, args->inp); 2591 break; 2592 2593 case O_RECV: 2594 match = iface_match(m->m_pkthdr.rcvif, 2595 (ipfw_insn_if *)cmd); 2596 break; 2597 2598 case O_XMIT: 2599 match = iface_match(oif, (ipfw_insn_if *)cmd); 2600 break; 2601 2602 case O_VIA: 2603 match = iface_match(oif ? oif : 2604 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2605 break; 2606 2607 case O_MACADDR2: 2608 if (args->eh != NULL) { /* have MAC header */ 2609 u_int32_t *want = (u_int32_t *) 2610 ((ipfw_insn_mac *)cmd)->addr; 2611 u_int32_t *mask = (u_int32_t *) 2612 ((ipfw_insn_mac *)cmd)->mask; 2613 u_int32_t *hdr = (u_int32_t *)args->eh; 2614 2615 match = 2616 ( want[0] == (hdr[0] & mask[0]) && 2617 want[1] == (hdr[1] & mask[1]) && 2618 want[2] == (hdr[2] & mask[2]) ); 2619 } 2620 break; 2621 2622 case O_MAC_TYPE: 2623 if (args->eh != NULL) { 2624 u_int16_t *p = 2625 ((ipfw_insn_u16 *)cmd)->ports; 2626 int i; 2627 2628 for (i = cmdlen - 1; !match && i>0; 2629 i--, p += 2) 2630 match = (etype >= p[0] && 2631 etype <= p[1]); 2632 } 2633 break; 2634 2635 case O_FRAG: 2636 match = (offset != 0); 2637 break; 2638 2639 case O_IN: /* "out" is "not in" */ 2640 match = (oif == NULL); 2641 break; 2642 2643 case O_LAYER2: 2644 match = (args->eh != NULL); 2645 break; 2646 2647 case O_DIVERTED: 2648 match = (cmd->arg1 & 1 && divinput_flags & 2649 IP_FW_DIVERT_LOOPBACK_FLAG) || 2650 (cmd->arg1 & 2 && divinput_flags & 2651 IP_FW_DIVERT_OUTPUT_FLAG); 2652 break; 2653 2654 case O_PROTO: 2655 /* 2656 * We do not allow an arg of 0 so the 2657 * check of "proto" only suffices. 2658 */ 2659 match = (proto == cmd->arg1); 2660 break; 2661 2662 case O_IP_SRC: 2663 match = is_ipv4 && 2664 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2665 src_ip.s_addr); 2666 break; 2667 2668 case O_IP_SRC_LOOKUP: 2669 case O_IP_DST_LOOKUP: 2670 if (is_ipv4) { 2671 uint32_t a = 2672 (cmd->opcode == O_IP_DST_LOOKUP) ? 2673 dst_ip.s_addr : src_ip.s_addr; 2674 uint32_t v = 0; 2675 2676 match = lookup_table(chain, cmd->arg1, a, 2677 &v); 2678 if (!match) 2679 break; 2680 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2681 match = 2682 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2683 else 2684 tablearg = v; 2685 } 2686 break; 2687 2688 case O_IP_SRC_MASK: 2689 case O_IP_DST_MASK: 2690 if (is_ipv4) { 2691 uint32_t a = 2692 (cmd->opcode == O_IP_DST_MASK) ? 2693 dst_ip.s_addr : src_ip.s_addr; 2694 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2695 int i = cmdlen-1; 2696 2697 for (; !match && i>0; i-= 2, p+= 2) 2698 match = (p[0] == (a & p[1])); 2699 } 2700 break; 2701 2702 case O_IP_SRC_ME: 2703 if (is_ipv4) { 2704 struct ifnet *tif; 2705 2706 INADDR_TO_IFP(src_ip, tif); 2707 match = (tif != NULL); 2708 } 2709 break; 2710 2711 case O_IP_DST_SET: 2712 case O_IP_SRC_SET: 2713 if (is_ipv4) { 2714 u_int32_t *d = (u_int32_t *)(cmd+1); 2715 u_int32_t addr = 2716 cmd->opcode == O_IP_DST_SET ? 2717 args->f_id.dst_ip : 2718 args->f_id.src_ip; 2719 2720 if (addr < d[0]) 2721 break; 2722 addr -= d[0]; /* subtract base */ 2723 match = (addr < cmd->arg1) && 2724 ( d[ 1 + (addr>>5)] & 2725 (1<<(addr & 0x1f)) ); 2726 } 2727 break; 2728 2729 case O_IP_DST: 2730 match = is_ipv4 && 2731 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2732 dst_ip.s_addr); 2733 break; 2734 2735 case O_IP_DST_ME: 2736 if (is_ipv4) { 2737 struct ifnet *tif; 2738 2739 INADDR_TO_IFP(dst_ip, tif); 2740 match = (tif != NULL); 2741 } 2742 break; 2743 2744 case O_IP_SRCPORT: 2745 case O_IP_DSTPORT: 2746 /* 2747 * offset == 0 && proto != 0 is enough 2748 * to guarantee that we have a 2749 * packet with port info. 2750 */ 2751 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2752 && offset == 0) { 2753 u_int16_t x = 2754 (cmd->opcode == O_IP_SRCPORT) ? 2755 src_port : dst_port ; 2756 u_int16_t *p = 2757 ((ipfw_insn_u16 *)cmd)->ports; 2758 int i; 2759 2760 for (i = cmdlen - 1; !match && i>0; 2761 i--, p += 2) 2762 match = (x>=p[0] && x<=p[1]); 2763 } 2764 break; 2765 2766 case O_ICMPTYPE: 2767 match = (offset == 0 && proto==IPPROTO_ICMP && 2768 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2769 break; 2770 2771 #ifdef INET6 2772 case O_ICMP6TYPE: 2773 match = is_ipv6 && offset == 0 && 2774 proto==IPPROTO_ICMPV6 && 2775 icmp6type_match( 2776 ICMP6(ulp)->icmp6_type, 2777 (ipfw_insn_u32 *)cmd); 2778 break; 2779 #endif /* INET6 */ 2780 2781 case O_IPOPT: 2782 match = (is_ipv4 && 2783 ipopts_match(ip, cmd) ); 2784 break; 2785 2786 case O_IPVER: 2787 match = (is_ipv4 && 2788 cmd->arg1 == ip->ip_v); 2789 break; 2790 2791 case O_IPID: 2792 case O_IPLEN: 2793 case O_IPTTL: 2794 if (is_ipv4) { /* only for IP packets */ 2795 uint16_t x; 2796 uint16_t *p; 2797 int i; 2798 2799 if (cmd->opcode == O_IPLEN) 2800 x = ip_len; 2801 else if (cmd->opcode == O_IPTTL) 2802 x = ip->ip_ttl; 2803 else /* must be IPID */ 2804 x = ntohs(ip->ip_id); 2805 if (cmdlen == 1) { 2806 match = (cmd->arg1 == x); 2807 break; 2808 } 2809 /* otherwise we have ranges */ 2810 p = ((ipfw_insn_u16 *)cmd)->ports; 2811 i = cmdlen - 1; 2812 for (; !match && i>0; i--, p += 2) 2813 match = (x >= p[0] && x <= p[1]); 2814 } 2815 break; 2816 2817 case O_IPPRECEDENCE: 2818 match = (is_ipv4 && 2819 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2820 break; 2821 2822 case O_IPTOS: 2823 match = (is_ipv4 && 2824 flags_match(cmd, ip->ip_tos)); 2825 break; 2826 2827 case O_TCPDATALEN: 2828 if (proto == IPPROTO_TCP && offset == 0) { 2829 struct tcphdr *tcp; 2830 uint16_t x; 2831 uint16_t *p; 2832 int i; 2833 2834 tcp = TCP(ulp); 2835 x = ip_len - 2836 ((ip->ip_hl + tcp->th_off) << 2); 2837 if (cmdlen == 1) { 2838 match = (cmd->arg1 == x); 2839 break; 2840 } 2841 /* otherwise we have ranges */ 2842 p = ((ipfw_insn_u16 *)cmd)->ports; 2843 i = cmdlen - 1; 2844 for (; !match && i>0; i--, p += 2) 2845 match = (x >= p[0] && x <= p[1]); 2846 } 2847 break; 2848 2849 case O_TCPFLAGS: 2850 match = (proto == IPPROTO_TCP && offset == 0 && 2851 flags_match(cmd, TCP(ulp)->th_flags)); 2852 break; 2853 2854 case O_TCPOPTS: 2855 match = (proto == IPPROTO_TCP && offset == 0 && 2856 tcpopts_match(TCP(ulp), cmd)); 2857 break; 2858 2859 case O_TCPSEQ: 2860 match = (proto == IPPROTO_TCP && offset == 0 && 2861 ((ipfw_insn_u32 *)cmd)->d[0] == 2862 TCP(ulp)->th_seq); 2863 break; 2864 2865 case O_TCPACK: 2866 match = (proto == IPPROTO_TCP && offset == 0 && 2867 ((ipfw_insn_u32 *)cmd)->d[0] == 2868 TCP(ulp)->th_ack); 2869 break; 2870 2871 case O_TCPWIN: 2872 match = (proto == IPPROTO_TCP && offset == 0 && 2873 cmd->arg1 == TCP(ulp)->th_win); 2874 break; 2875 2876 case O_ESTAB: 2877 /* reject packets which have SYN only */ 2878 /* XXX should i also check for TH_ACK ? */ 2879 match = (proto == IPPROTO_TCP && offset == 0 && 2880 (TCP(ulp)->th_flags & 2881 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2882 break; 2883 2884 case O_ALTQ: { 2885 struct pf_mtag *at; 2886 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2887 2888 match = 1; 2889 at = pf_find_mtag(m); 2890 if (at != NULL && at->qid != 0) 2891 break; 2892 at = pf_get_mtag(m); 2893 if (at == NULL) { 2894 /* 2895 * Let the packet fall back to the 2896 * default ALTQ. 2897 */ 2898 break; 2899 } 2900 at->qid = altq->qid; 2901 if (is_ipv4) 2902 at->af = AF_INET; 2903 else 2904 at->af = AF_LINK; 2905 at->hdr = ip; 2906 break; 2907 } 2908 2909 case O_LOG: 2910 if (fw_verbose) 2911 ipfw_log(f, hlen, args, m, 2912 oif, offset, tablearg, ip); 2913 match = 1; 2914 break; 2915 2916 case O_PROB: 2917 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2918 break; 2919 2920 case O_VERREVPATH: 2921 /* Outgoing packets automatically pass/match */ 2922 match = ((oif != NULL) || 2923 (m->m_pkthdr.rcvif == NULL) || 2924 ( 2925 #ifdef INET6 2926 is_ipv6 ? 2927 verify_path6(&(args->f_id.src_ip6), 2928 m->m_pkthdr.rcvif) : 2929 #endif 2930 verify_path(src_ip, m->m_pkthdr.rcvif, 2931 args->f_id.fib))); 2932 break; 2933 2934 case O_VERSRCREACH: 2935 /* Outgoing packets automatically pass/match */ 2936 match = (hlen > 0 && ((oif != NULL) || 2937 #ifdef INET6 2938 is_ipv6 ? 2939 verify_path6(&(args->f_id.src_ip6), 2940 NULL) : 2941 #endif 2942 verify_path(src_ip, NULL, args->f_id.fib))); 2943 break; 2944 2945 case O_ANTISPOOF: 2946 /* Outgoing packets automatically pass/match */ 2947 if (oif == NULL && hlen > 0 && 2948 ( (is_ipv4 && in_localaddr(src_ip)) 2949 #ifdef INET6 2950 || (is_ipv6 && 2951 in6_localaddr(&(args->f_id.src_ip6))) 2952 #endif 2953 )) 2954 match = 2955 #ifdef INET6 2956 is_ipv6 ? verify_path6( 2957 &(args->f_id.src_ip6), 2958 m->m_pkthdr.rcvif) : 2959 #endif 2960 verify_path(src_ip, 2961 m->m_pkthdr.rcvif, 2962 args->f_id.fib); 2963 else 2964 match = 1; 2965 break; 2966 2967 case O_IPSEC: 2968 #ifdef IPSEC 2969 match = (m_tag_find(m, 2970 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2971 #endif 2972 /* otherwise no match */ 2973 break; 2974 2975 #ifdef INET6 2976 case O_IP6_SRC: 2977 match = is_ipv6 && 2978 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2979 &((ipfw_insn_ip6 *)cmd)->addr6); 2980 break; 2981 2982 case O_IP6_DST: 2983 match = is_ipv6 && 2984 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2985 &((ipfw_insn_ip6 *)cmd)->addr6); 2986 break; 2987 case O_IP6_SRC_MASK: 2988 case O_IP6_DST_MASK: 2989 if (is_ipv6) { 2990 int i = cmdlen - 1; 2991 struct in6_addr p; 2992 struct in6_addr *d = 2993 &((ipfw_insn_ip6 *)cmd)->addr6; 2994 2995 for (; !match && i > 0; d += 2, 2996 i -= F_INSN_SIZE(struct in6_addr) 2997 * 2) { 2998 p = (cmd->opcode == 2999 O_IP6_SRC_MASK) ? 3000 args->f_id.src_ip6: 3001 args->f_id.dst_ip6; 3002 APPLY_MASK(&p, &d[1]); 3003 match = 3004 IN6_ARE_ADDR_EQUAL(&d[0], 3005 &p); 3006 } 3007 } 3008 break; 3009 3010 case O_IP6_SRC_ME: 3011 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3012 break; 3013 3014 case O_IP6_DST_ME: 3015 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3016 break; 3017 3018 case O_FLOW6ID: 3019 match = is_ipv6 && 3020 flow6id_match(args->f_id.flow_id6, 3021 (ipfw_insn_u32 *) cmd); 3022 break; 3023 3024 case O_EXT_HDR: 3025 match = is_ipv6 && 3026 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3027 break; 3028 3029 case O_IP6: 3030 match = is_ipv6; 3031 break; 3032 #endif 3033 3034 case O_IP4: 3035 match = is_ipv4; 3036 break; 3037 3038 case O_TAG: { 3039 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3040 tablearg : cmd->arg1; 3041 3042 /* Packet is already tagged with this tag? */ 3043 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3044 3045 /* We have `untag' action when F_NOT flag is 3046 * present. And we must remove this mtag from 3047 * mbuf and reset `match' to zero (`match' will 3048 * be inversed later). 3049 * Otherwise we should allocate new mtag and 3050 * push it into mbuf. 3051 */ 3052 if (cmd->len & F_NOT) { /* `untag' action */ 3053 if (mtag != NULL) 3054 m_tag_delete(m, mtag); 3055 } else if (mtag == NULL) { 3056 if ((mtag = m_tag_alloc(MTAG_IPFW, 3057 tag, 0, M_NOWAIT)) != NULL) 3058 m_tag_prepend(m, mtag); 3059 } 3060 match = (cmd->len & F_NOT) ? 0: 1; 3061 break; 3062 } 3063 3064 case O_FIB: /* try match the specified fib */ 3065 if (args->f_id.fib == cmd->arg1) 3066 match = 1; 3067 break; 3068 3069 case O_TAGGED: { 3070 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3071 tablearg : cmd->arg1; 3072 3073 if (cmdlen == 1) { 3074 match = m_tag_locate(m, MTAG_IPFW, 3075 tag, NULL) != NULL; 3076 break; 3077 } 3078 3079 /* we have ranges */ 3080 for (mtag = m_tag_first(m); 3081 mtag != NULL && !match; 3082 mtag = m_tag_next(m, mtag)) { 3083 uint16_t *p; 3084 int i; 3085 3086 if (mtag->m_tag_cookie != MTAG_IPFW) 3087 continue; 3088 3089 p = ((ipfw_insn_u16 *)cmd)->ports; 3090 i = cmdlen - 1; 3091 for(; !match && i > 0; i--, p += 2) 3092 match = 3093 mtag->m_tag_id >= p[0] && 3094 mtag->m_tag_id <= p[1]; 3095 } 3096 break; 3097 } 3098 3099 /* 3100 * The second set of opcodes represents 'actions', 3101 * i.e. the terminal part of a rule once the packet 3102 * matches all previous patterns. 3103 * Typically there is only one action for each rule, 3104 * and the opcode is stored at the end of the rule 3105 * (but there are exceptions -- see below). 3106 * 3107 * In general, here we set retval and terminate the 3108 * outer loop (would be a 'break 3' in some language, 3109 * but we need to do a 'goto done'). 3110 * 3111 * Exceptions: 3112 * O_COUNT and O_SKIPTO actions: 3113 * instead of terminating, we jump to the next rule 3114 * ('goto next_rule', equivalent to a 'break 2'), 3115 * or to the SKIPTO target ('goto again' after 3116 * having set f, cmd and l), respectively. 3117 * 3118 * O_TAG, O_LOG and O_ALTQ action parameters: 3119 * perform some action and set match = 1; 3120 * 3121 * O_LIMIT and O_KEEP_STATE: these opcodes are 3122 * not real 'actions', and are stored right 3123 * before the 'action' part of the rule. 3124 * These opcodes try to install an entry in the 3125 * state tables; if successful, we continue with 3126 * the next opcode (match=1; break;), otherwise 3127 * the packet * must be dropped 3128 * ('goto done' after setting retval); 3129 * 3130 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3131 * cause a lookup of the state table, and a jump 3132 * to the 'action' part of the parent rule 3133 * ('goto check_body') if an entry is found, or 3134 * (CHECK_STATE only) a jump to the next rule if 3135 * the entry is not found ('goto next_rule'). 3136 * The result of the lookup is cached to make 3137 * further instances of these opcodes are 3138 * effectively NOPs. 3139 */ 3140 case O_LIMIT: 3141 case O_KEEP_STATE: 3142 if (install_state(f, 3143 (ipfw_insn_limit *)cmd, args, tablearg)) { 3144 retval = IP_FW_DENY; 3145 goto done; /* error/limit violation */ 3146 } 3147 match = 1; 3148 break; 3149 3150 case O_PROBE_STATE: 3151 case O_CHECK_STATE: 3152 /* 3153 * dynamic rules are checked at the first 3154 * keep-state or check-state occurrence, 3155 * with the result being stored in dyn_dir. 3156 * The compiler introduces a PROBE_STATE 3157 * instruction for us when we have a 3158 * KEEP_STATE (because PROBE_STATE needs 3159 * to be run first). 3160 */ 3161 if (dyn_dir == MATCH_UNKNOWN && 3162 (q = lookup_dyn_rule(&args->f_id, 3163 &dyn_dir, proto == IPPROTO_TCP ? 3164 TCP(ulp) : NULL)) 3165 != NULL) { 3166 /* 3167 * Found dynamic entry, update stats 3168 * and jump to the 'action' part of 3169 * the parent rule. 3170 */ 3171 q->pcnt++; 3172 q->bcnt += pktlen; 3173 f = q->rule; 3174 cmd = ACTION_PTR(f); 3175 l = f->cmd_len - f->act_ofs; 3176 IPFW_DYN_UNLOCK(); 3177 goto check_body; 3178 } 3179 /* 3180 * Dynamic entry not found. If CHECK_STATE, 3181 * skip to next rule, if PROBE_STATE just 3182 * ignore and continue with next opcode. 3183 */ 3184 if (cmd->opcode == O_CHECK_STATE) 3185 goto next_rule; 3186 match = 1; 3187 break; 3188 3189 case O_ACCEPT: 3190 retval = 0; /* accept */ 3191 goto done; 3192 3193 case O_PIPE: 3194 case O_QUEUE: 3195 args->rule = f; /* report matching rule */ 3196 if (cmd->arg1 == IP_FW_TABLEARG) 3197 args->cookie = tablearg; 3198 else 3199 args->cookie = cmd->arg1; 3200 retval = IP_FW_DUMMYNET; 3201 goto done; 3202 3203 case O_DIVERT: 3204 case O_TEE: { 3205 struct divert_tag *dt; 3206 3207 if (args->eh) /* not on layer 2 */ 3208 break; 3209 mtag = m_tag_get(PACKET_TAG_DIVERT, 3210 sizeof(struct divert_tag), 3211 M_NOWAIT); 3212 if (mtag == NULL) { 3213 /* XXX statistic */ 3214 /* drop packet */ 3215 IPFW_RUNLOCK(chain); 3216 return (IP_FW_DENY); 3217 } 3218 dt = (struct divert_tag *)(mtag+1); 3219 dt->cookie = f->rulenum; 3220 if (cmd->arg1 == IP_FW_TABLEARG) 3221 dt->info = tablearg; 3222 else 3223 dt->info = cmd->arg1; 3224 m_tag_prepend(m, mtag); 3225 retval = (cmd->opcode == O_DIVERT) ? 3226 IP_FW_DIVERT : IP_FW_TEE; 3227 goto done; 3228 } 3229 3230 case O_COUNT: 3231 case O_SKIPTO: 3232 f->pcnt++; /* update stats */ 3233 f->bcnt += pktlen; 3234 f->timestamp = time_uptime; 3235 if (cmd->opcode == O_COUNT) 3236 goto next_rule; 3237 /* handle skipto */ 3238 if (cmd->arg1 == IP_FW_TABLEARG) { 3239 f = lookup_next_rule(f, tablearg); 3240 } else { 3241 if (f->next_rule == NULL) 3242 lookup_next_rule(f, 0); 3243 f = f->next_rule; 3244 } 3245 goto again; 3246 3247 case O_REJECT: 3248 /* 3249 * Drop the packet and send a reject notice 3250 * if the packet is not ICMP (or is an ICMP 3251 * query), and it is not multicast/broadcast. 3252 */ 3253 if (hlen > 0 && is_ipv4 && offset == 0 && 3254 (proto != IPPROTO_ICMP || 3255 is_icmp_query(ICMP(ulp))) && 3256 !(m->m_flags & (M_BCAST|M_MCAST)) && 3257 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3258 send_reject(args, cmd->arg1, ip_len, ip); 3259 m = args->m; 3260 } 3261 /* FALLTHROUGH */ 3262 #ifdef INET6 3263 case O_UNREACH6: 3264 if (hlen > 0 && is_ipv6 && 3265 ((offset & IP6F_OFF_MASK) == 0) && 3266 (proto != IPPROTO_ICMPV6 || 3267 (is_icmp6_query(args->f_id.flags) == 1)) && 3268 !(m->m_flags & (M_BCAST|M_MCAST)) && 3269 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3270 send_reject6( 3271 args, cmd->arg1, hlen, 3272 (struct ip6_hdr *)ip); 3273 m = args->m; 3274 } 3275 /* FALLTHROUGH */ 3276 #endif 3277 case O_DENY: 3278 retval = IP_FW_DENY; 3279 goto done; 3280 3281 case O_FORWARD_IP: { 3282 struct sockaddr_in *sa; 3283 sa = &(((ipfw_insn_sa *)cmd)->sa); 3284 if (args->eh) /* not valid on layer2 pkts */ 3285 break; 3286 if (!q || dyn_dir == MATCH_FORWARD) { 3287 if (sa->sin_addr.s_addr == INADDR_ANY) { 3288 bcopy(sa, &args->hopstore, 3289 sizeof(*sa)); 3290 args->hopstore.sin_addr.s_addr = 3291 htonl(tablearg); 3292 args->next_hop = 3293 &args->hopstore; 3294 } else { 3295 args->next_hop = sa; 3296 } 3297 } 3298 retval = IP_FW_PASS; 3299 } 3300 goto done; 3301 3302 case O_NETGRAPH: 3303 case O_NGTEE: 3304 args->rule = f; /* report matching rule */ 3305 if (cmd->arg1 == IP_FW_TABLEARG) 3306 args->cookie = tablearg; 3307 else 3308 args->cookie = cmd->arg1; 3309 retval = (cmd->opcode == O_NETGRAPH) ? 3310 IP_FW_NETGRAPH : IP_FW_NGTEE; 3311 goto done; 3312 3313 case O_SETFIB: 3314 f->pcnt++; /* update stats */ 3315 f->bcnt += pktlen; 3316 f->timestamp = time_uptime; 3317 M_SETFIB(m, cmd->arg1); 3318 args->f_id.fib = cmd->arg1; 3319 goto next_rule; 3320 3321 case O_NAT: { 3322 struct cfg_nat *t; 3323 int nat_id; 3324 3325 if (IPFW_NAT_LOADED) { 3326 args->rule = f; /* Report matching rule. */ 3327 t = ((ipfw_insn_nat *)cmd)->nat; 3328 if (t == NULL) { 3329 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 3330 tablearg : cmd->arg1; 3331 LOOKUP_NAT(layer3_chain, nat_id, t); 3332 if (t == NULL) { 3333 retval = IP_FW_DENY; 3334 goto done; 3335 } 3336 if (cmd->arg1 != IP_FW_TABLEARG) 3337 ((ipfw_insn_nat *)cmd)->nat = t; 3338 } 3339 retval = ipfw_nat_ptr(args, t, m); 3340 } else 3341 retval = IP_FW_DENY; 3342 goto done; 3343 } 3344 3345 default: 3346 panic("-- unknown opcode %d\n", cmd->opcode); 3347 } /* end of switch() on opcodes */ 3348 3349 if (cmd->len & F_NOT) 3350 match = !match; 3351 3352 if (match) { 3353 if (cmd->len & F_OR) 3354 skip_or = 1; 3355 } else { 3356 if (!(cmd->len & F_OR)) /* not an OR block, */ 3357 break; /* try next rule */ 3358 } 3359 3360 } /* end of inner for, scan opcodes */ 3361 3362 next_rule:; /* try next rule */ 3363 3364 } /* end of outer for, scan rules */ 3365 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3366 IPFW_RUNLOCK(chain); 3367 return (IP_FW_DENY); 3368 3369 done: 3370 /* Update statistics */ 3371 f->pcnt++; 3372 f->bcnt += pktlen; 3373 f->timestamp = time_uptime; 3374 IPFW_RUNLOCK(chain); 3375 return (retval); 3376 3377 pullup_failed: 3378 if (fw_verbose) 3379 printf("ipfw: pullup failed\n"); 3380 return (IP_FW_DENY); 3381 } 3382 3383 /* 3384 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3385 * These will be reconstructed on the fly as packets are matched. 3386 */ 3387 static void 3388 flush_rule_ptrs(struct ip_fw_chain *chain) 3389 { 3390 struct ip_fw *rule; 3391 3392 IPFW_WLOCK_ASSERT(chain); 3393 3394 for (rule = chain->rules; rule; rule = rule->next) 3395 rule->next_rule = NULL; 3396 } 3397 3398 /* 3399 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3400 * possibly create a rule number and add the rule to the list. 3401 * Update the rule_number in the input struct so the caller knows it as well. 3402 */ 3403 static int 3404 add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3405 { 3406 struct ip_fw *rule, *f, *prev; 3407 int l = RULESIZE(input_rule); 3408 3409 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3410 return (EINVAL); 3411 3412 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3413 if (rule == NULL) 3414 return (ENOSPC); 3415 3416 bcopy(input_rule, rule, l); 3417 3418 rule->next = NULL; 3419 rule->next_rule = NULL; 3420 3421 rule->pcnt = 0; 3422 rule->bcnt = 0; 3423 rule->timestamp = 0; 3424 3425 IPFW_WLOCK(chain); 3426 3427 if (chain->rules == NULL) { /* default rule */ 3428 chain->rules = rule; 3429 goto done; 3430 } 3431 3432 /* 3433 * If rulenum is 0, find highest numbered rule before the 3434 * default rule, and add autoinc_step 3435 */ 3436 if (autoinc_step < 1) 3437 autoinc_step = 1; 3438 else if (autoinc_step > 1000) 3439 autoinc_step = 1000; 3440 if (rule->rulenum == 0) { 3441 /* 3442 * locate the highest numbered rule before default 3443 */ 3444 for (f = chain->rules; f; f = f->next) { 3445 if (f->rulenum == IPFW_DEFAULT_RULE) 3446 break; 3447 rule->rulenum = f->rulenum; 3448 } 3449 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3450 rule->rulenum += autoinc_step; 3451 input_rule->rulenum = rule->rulenum; 3452 } 3453 3454 /* 3455 * Now insert the new rule in the right place in the sorted list. 3456 */ 3457 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3458 if (f->rulenum > rule->rulenum) { /* found the location */ 3459 if (prev) { 3460 rule->next = f; 3461 prev->next = rule; 3462 } else { /* head insert */ 3463 rule->next = chain->rules; 3464 chain->rules = rule; 3465 } 3466 break; 3467 } 3468 } 3469 flush_rule_ptrs(chain); 3470 done: 3471 static_count++; 3472 static_len += l; 3473 IPFW_WUNLOCK(chain); 3474 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3475 rule->rulenum, static_count);) 3476 return (0); 3477 } 3478 3479 /** 3480 * Remove a static rule (including derived * dynamic rules) 3481 * and place it on the ``reap list'' for later reclamation. 3482 * The caller is in charge of clearing rule pointers to avoid 3483 * dangling pointers. 3484 * @return a pointer to the next entry. 3485 * Arguments are not checked, so they better be correct. 3486 */ 3487 static struct ip_fw * 3488 remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 3489 struct ip_fw *prev) 3490 { 3491 struct ip_fw *n; 3492 int l = RULESIZE(rule); 3493 3494 IPFW_WLOCK_ASSERT(chain); 3495 3496 n = rule->next; 3497 IPFW_DYN_LOCK(); 3498 remove_dyn_rule(rule, NULL /* force removal */); 3499 IPFW_DYN_UNLOCK(); 3500 if (prev == NULL) 3501 chain->rules = n; 3502 else 3503 prev->next = n; 3504 static_count--; 3505 static_len -= l; 3506 3507 rule->next = chain->reap; 3508 chain->reap = rule; 3509 3510 return n; 3511 } 3512 3513 /** 3514 * Reclaim storage associated with a list of rules. This is 3515 * typically the list created using remove_rule. 3516 */ 3517 static void 3518 reap_rules(struct ip_fw *head) 3519 { 3520 struct ip_fw *rule; 3521 3522 while ((rule = head) != NULL) { 3523 head = head->next; 3524 if (DUMMYNET_LOADED) 3525 ip_dn_ruledel_ptr(rule); 3526 free(rule, M_IPFW); 3527 } 3528 } 3529 3530 /* 3531 * Remove all rules from a chain (except rules in set RESVD_SET 3532 * unless kill_default = 1). The caller is responsible for 3533 * reclaiming storage for the rules left in chain->reap. 3534 */ 3535 static void 3536 free_chain(struct ip_fw_chain *chain, int kill_default) 3537 { 3538 struct ip_fw *prev, *rule; 3539 3540 IPFW_WLOCK_ASSERT(chain); 3541 3542 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3543 for (prev = NULL, rule = chain->rules; rule ; ) 3544 if (kill_default || rule->set != RESVD_SET) 3545 rule = remove_rule(chain, rule, prev); 3546 else { 3547 prev = rule; 3548 rule = rule->next; 3549 } 3550 } 3551 3552 /** 3553 * Remove all rules with given number, and also do set manipulation. 3554 * Assumes chain != NULL && *chain != NULL. 3555 * 3556 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3557 * the next 8 bits are the new set, the top 8 bits are the command: 3558 * 3559 * 0 delete rules with given number 3560 * 1 delete rules with given set number 3561 * 2 move rules with given number to new set 3562 * 3 move rules with given set number to new set 3563 * 4 swap sets with given numbers 3564 * 5 delete rules with given number and with given set number 3565 */ 3566 static int 3567 del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3568 { 3569 struct ip_fw *prev = NULL, *rule; 3570 u_int16_t rulenum; /* rule or old_set */ 3571 u_int8_t cmd, new_set; 3572 3573 rulenum = arg & 0xffff; 3574 cmd = (arg >> 24) & 0xff; 3575 new_set = (arg >> 16) & 0xff; 3576 3577 if (cmd > 5 || new_set > RESVD_SET) 3578 return EINVAL; 3579 if (cmd == 0 || cmd == 2 || cmd == 5) { 3580 if (rulenum >= IPFW_DEFAULT_RULE) 3581 return EINVAL; 3582 } else { 3583 if (rulenum > RESVD_SET) /* old_set */ 3584 return EINVAL; 3585 } 3586 3587 IPFW_WLOCK(chain); 3588 rule = chain->rules; 3589 chain->reap = NULL; 3590 switch (cmd) { 3591 case 0: /* delete rules with given number */ 3592 /* 3593 * locate first rule to delete 3594 */ 3595 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3596 ; 3597 if (rule->rulenum != rulenum) { 3598 IPFW_WUNLOCK(chain); 3599 return EINVAL; 3600 } 3601 3602 /* 3603 * flush pointers outside the loop, then delete all matching 3604 * rules. prev remains the same throughout the cycle. 3605 */ 3606 flush_rule_ptrs(chain); 3607 while (rule->rulenum == rulenum) 3608 rule = remove_rule(chain, rule, prev); 3609 break; 3610 3611 case 1: /* delete all rules with given set number */ 3612 flush_rule_ptrs(chain); 3613 rule = chain->rules; 3614 while (rule->rulenum < IPFW_DEFAULT_RULE) 3615 if (rule->set == rulenum) 3616 rule = remove_rule(chain, rule, prev); 3617 else { 3618 prev = rule; 3619 rule = rule->next; 3620 } 3621 break; 3622 3623 case 2: /* move rules with given number to new set */ 3624 rule = chain->rules; 3625 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3626 if (rule->rulenum == rulenum) 3627 rule->set = new_set; 3628 break; 3629 3630 case 3: /* move rules with given set number to new set */ 3631 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3632 if (rule->set == rulenum) 3633 rule->set = new_set; 3634 break; 3635 3636 case 4: /* swap two sets */ 3637 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3638 if (rule->set == rulenum) 3639 rule->set = new_set; 3640 else if (rule->set == new_set) 3641 rule->set = rulenum; 3642 break; 3643 case 5: /* delete rules with given number and with given set number. 3644 * rulenum - given rule number; 3645 * new_set - given set number. 3646 */ 3647 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3648 ; 3649 if (rule->rulenum != rulenum) { 3650 IPFW_WUNLOCK(chain); 3651 return (EINVAL); 3652 } 3653 flush_rule_ptrs(chain); 3654 while (rule->rulenum == rulenum) { 3655 if (rule->set == new_set) 3656 rule = remove_rule(chain, rule, prev); 3657 else { 3658 prev = rule; 3659 rule = rule->next; 3660 } 3661 } 3662 } 3663 /* 3664 * Look for rules to reclaim. We grab the list before 3665 * releasing the lock then reclaim them w/o the lock to 3666 * avoid a LOR with dummynet. 3667 */ 3668 rule = chain->reap; 3669 chain->reap = NULL; 3670 IPFW_WUNLOCK(chain); 3671 if (rule) 3672 reap_rules(rule); 3673 return 0; 3674 } 3675 3676 /* 3677 * Clear counters for a specific rule. 3678 * The enclosing "table" is assumed locked. 3679 */ 3680 static void 3681 clear_counters(struct ip_fw *rule, int log_only) 3682 { 3683 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3684 3685 if (log_only == 0) { 3686 rule->bcnt = rule->pcnt = 0; 3687 rule->timestamp = 0; 3688 } 3689 if (l->o.opcode == O_LOG) 3690 l->log_left = l->max_log; 3691 } 3692 3693 /** 3694 * Reset some or all counters on firewall rules. 3695 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 3696 * the next 8 bits are the set number, the top 8 bits are the command: 3697 * 0 work with rules from all set's; 3698 * 1 work with rules only from specified set. 3699 * Specified rule number is zero if we want to clear all entries. 3700 * log_only is 1 if we only want to reset logs, zero otherwise. 3701 */ 3702 static int 3703 zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 3704 { 3705 struct ip_fw *rule; 3706 char *msg; 3707 3708 uint16_t rulenum = arg & 0xffff; 3709 uint8_t set = (arg >> 16) & 0xff; 3710 uint8_t cmd = (arg >> 24) & 0xff; 3711 3712 if (cmd > 1) 3713 return (EINVAL); 3714 if (cmd == 1 && set > RESVD_SET) 3715 return (EINVAL); 3716 3717 IPFW_WLOCK(chain); 3718 if (rulenum == 0) { 3719 norule_counter = 0; 3720 for (rule = chain->rules; rule; rule = rule->next) { 3721 /* Skip rules from another set. */ 3722 if (cmd == 1 && rule->set != set) 3723 continue; 3724 clear_counters(rule, log_only); 3725 } 3726 msg = log_only ? "All logging counts reset" : 3727 "Accounting cleared"; 3728 } else { 3729 int cleared = 0; 3730 /* 3731 * We can have multiple rules with the same number, so we 3732 * need to clear them all. 3733 */ 3734 for (rule = chain->rules; rule; rule = rule->next) 3735 if (rule->rulenum == rulenum) { 3736 while (rule && rule->rulenum == rulenum) { 3737 if (cmd == 0 || rule->set == set) 3738 clear_counters(rule, log_only); 3739 rule = rule->next; 3740 } 3741 cleared = 1; 3742 break; 3743 } 3744 if (!cleared) { /* we did not find any matching rules */ 3745 IPFW_WUNLOCK(chain); 3746 return (EINVAL); 3747 } 3748 msg = log_only ? "logging count reset" : "cleared"; 3749 } 3750 IPFW_WUNLOCK(chain); 3751 3752 if (fw_verbose) { 3753 int lev = LOG_SECURITY | LOG_NOTICE; 3754 3755 if (rulenum) 3756 log(lev, "ipfw: Entry %d %s.\n", rulenum, msg); 3757 else 3758 log(lev, "ipfw: %s.\n", msg); 3759 } 3760 return (0); 3761 } 3762 3763 /* 3764 * Check validity of the structure before insert. 3765 * Fortunately rules are simple, so this mostly need to check rule sizes. 3766 */ 3767 static int 3768 check_ipfw_struct(struct ip_fw *rule, int size) 3769 { 3770 int l, cmdlen = 0; 3771 int have_action=0; 3772 ipfw_insn *cmd; 3773 3774 if (size < sizeof(*rule)) { 3775 printf("ipfw: rule too short\n"); 3776 return (EINVAL); 3777 } 3778 /* first, check for valid size */ 3779 l = RULESIZE(rule); 3780 if (l != size) { 3781 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3782 return (EINVAL); 3783 } 3784 if (rule->act_ofs >= rule->cmd_len) { 3785 printf("ipfw: bogus action offset (%u > %u)\n", 3786 rule->act_ofs, rule->cmd_len - 1); 3787 return (EINVAL); 3788 } 3789 /* 3790 * Now go for the individual checks. Very simple ones, basically only 3791 * instruction sizes. 3792 */ 3793 for (l = rule->cmd_len, cmd = rule->cmd ; 3794 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3795 cmdlen = F_LEN(cmd); 3796 if (cmdlen > l) { 3797 printf("ipfw: opcode %d size truncated\n", 3798 cmd->opcode); 3799 return EINVAL; 3800 } 3801 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3802 switch (cmd->opcode) { 3803 case O_PROBE_STATE: 3804 case O_KEEP_STATE: 3805 case O_PROTO: 3806 case O_IP_SRC_ME: 3807 case O_IP_DST_ME: 3808 case O_LAYER2: 3809 case O_IN: 3810 case O_FRAG: 3811 case O_DIVERTED: 3812 case O_IPOPT: 3813 case O_IPTOS: 3814 case O_IPPRECEDENCE: 3815 case O_IPVER: 3816 case O_TCPWIN: 3817 case O_TCPFLAGS: 3818 case O_TCPOPTS: 3819 case O_ESTAB: 3820 case O_VERREVPATH: 3821 case O_VERSRCREACH: 3822 case O_ANTISPOOF: 3823 case O_IPSEC: 3824 #ifdef INET6 3825 case O_IP6_SRC_ME: 3826 case O_IP6_DST_ME: 3827 case O_EXT_HDR: 3828 case O_IP6: 3829 #endif 3830 case O_IP4: 3831 case O_TAG: 3832 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3833 goto bad_size; 3834 break; 3835 3836 case O_FIB: 3837 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3838 goto bad_size; 3839 if (cmd->arg1 >= rt_numfibs) { 3840 printf("ipfw: invalid fib number %d\n", 3841 cmd->arg1); 3842 return EINVAL; 3843 } 3844 break; 3845 3846 case O_SETFIB: 3847 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3848 goto bad_size; 3849 if (cmd->arg1 >= rt_numfibs) { 3850 printf("ipfw: invalid fib number %d\n", 3851 cmd->arg1); 3852 return EINVAL; 3853 } 3854 goto check_action; 3855 3856 case O_UID: 3857 case O_GID: 3858 case O_JAIL: 3859 case O_IP_SRC: 3860 case O_IP_DST: 3861 case O_TCPSEQ: 3862 case O_TCPACK: 3863 case O_PROB: 3864 case O_ICMPTYPE: 3865 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3866 goto bad_size; 3867 break; 3868 3869 case O_LIMIT: 3870 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3871 goto bad_size; 3872 break; 3873 3874 case O_LOG: 3875 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3876 goto bad_size; 3877 3878 ((ipfw_insn_log *)cmd)->log_left = 3879 ((ipfw_insn_log *)cmd)->max_log; 3880 3881 break; 3882 3883 case O_IP_SRC_MASK: 3884 case O_IP_DST_MASK: 3885 /* only odd command lengths */ 3886 if ( !(cmdlen & 1) || cmdlen > 31) 3887 goto bad_size; 3888 break; 3889 3890 case O_IP_SRC_SET: 3891 case O_IP_DST_SET: 3892 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3893 printf("ipfw: invalid set size %d\n", 3894 cmd->arg1); 3895 return EINVAL; 3896 } 3897 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3898 (cmd->arg1+31)/32 ) 3899 goto bad_size; 3900 break; 3901 3902 case O_IP_SRC_LOOKUP: 3903 case O_IP_DST_LOOKUP: 3904 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3905 printf("ipfw: invalid table number %d\n", 3906 cmd->arg1); 3907 return (EINVAL); 3908 } 3909 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3910 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3911 goto bad_size; 3912 break; 3913 3914 case O_MACADDR2: 3915 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3916 goto bad_size; 3917 break; 3918 3919 case O_NOP: 3920 case O_IPID: 3921 case O_IPTTL: 3922 case O_IPLEN: 3923 case O_TCPDATALEN: 3924 case O_TAGGED: 3925 if (cmdlen < 1 || cmdlen > 31) 3926 goto bad_size; 3927 break; 3928 3929 case O_MAC_TYPE: 3930 case O_IP_SRCPORT: 3931 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3932 if (cmdlen < 2 || cmdlen > 31) 3933 goto bad_size; 3934 break; 3935 3936 case O_RECV: 3937 case O_XMIT: 3938 case O_VIA: 3939 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3940 goto bad_size; 3941 break; 3942 3943 case O_ALTQ: 3944 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3945 goto bad_size; 3946 break; 3947 3948 case O_PIPE: 3949 case O_QUEUE: 3950 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3951 goto bad_size; 3952 goto check_action; 3953 3954 case O_FORWARD_IP: 3955 #ifdef IPFIREWALL_FORWARD 3956 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3957 goto bad_size; 3958 goto check_action; 3959 #else 3960 return EINVAL; 3961 #endif 3962 3963 case O_DIVERT: 3964 case O_TEE: 3965 if (ip_divert_ptr == NULL) 3966 return EINVAL; 3967 else 3968 goto check_size; 3969 case O_NETGRAPH: 3970 case O_NGTEE: 3971 if (!NG_IPFW_LOADED) 3972 return EINVAL; 3973 else 3974 goto check_size; 3975 case O_NAT: 3976 if (!IPFW_NAT_LOADED) 3977 return EINVAL; 3978 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 3979 goto bad_size; 3980 goto check_action; 3981 case O_FORWARD_MAC: /* XXX not implemented yet */ 3982 case O_CHECK_STATE: 3983 case O_COUNT: 3984 case O_ACCEPT: 3985 case O_DENY: 3986 case O_REJECT: 3987 #ifdef INET6 3988 case O_UNREACH6: 3989 #endif 3990 case O_SKIPTO: 3991 check_size: 3992 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3993 goto bad_size; 3994 check_action: 3995 if (have_action) { 3996 printf("ipfw: opcode %d, multiple actions" 3997 " not allowed\n", 3998 cmd->opcode); 3999 return EINVAL; 4000 } 4001 have_action = 1; 4002 if (l != cmdlen) { 4003 printf("ipfw: opcode %d, action must be" 4004 " last opcode\n", 4005 cmd->opcode); 4006 return EINVAL; 4007 } 4008 break; 4009 #ifdef INET6 4010 case O_IP6_SRC: 4011 case O_IP6_DST: 4012 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 4013 F_INSN_SIZE(ipfw_insn)) 4014 goto bad_size; 4015 break; 4016 4017 case O_FLOW6ID: 4018 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4019 ((ipfw_insn_u32 *)cmd)->o.arg1) 4020 goto bad_size; 4021 break; 4022 4023 case O_IP6_SRC_MASK: 4024 case O_IP6_DST_MASK: 4025 if ( !(cmdlen & 1) || cmdlen > 127) 4026 goto bad_size; 4027 break; 4028 case O_ICMP6TYPE: 4029 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4030 goto bad_size; 4031 break; 4032 #endif 4033 4034 default: 4035 switch (cmd->opcode) { 4036 #ifndef INET6 4037 case O_IP6_SRC_ME: 4038 case O_IP6_DST_ME: 4039 case O_EXT_HDR: 4040 case O_IP6: 4041 case O_UNREACH6: 4042 case O_IP6_SRC: 4043 case O_IP6_DST: 4044 case O_FLOW6ID: 4045 case O_IP6_SRC_MASK: 4046 case O_IP6_DST_MASK: 4047 case O_ICMP6TYPE: 4048 printf("ipfw: no IPv6 support in kernel\n"); 4049 return EPROTONOSUPPORT; 4050 #endif 4051 default: 4052 printf("ipfw: opcode %d, unknown opcode\n", 4053 cmd->opcode); 4054 return EINVAL; 4055 } 4056 } 4057 } 4058 if (have_action == 0) { 4059 printf("ipfw: missing action\n"); 4060 return EINVAL; 4061 } 4062 return 0; 4063 4064 bad_size: 4065 printf("ipfw: opcode %d size %d wrong\n", 4066 cmd->opcode, cmdlen); 4067 return EINVAL; 4068 } 4069 4070 /* 4071 * Copy the static and dynamic rules to the supplied buffer 4072 * and return the amount of space actually used. 4073 */ 4074 static size_t 4075 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4076 { 4077 char *bp = buf; 4078 char *ep = bp + space; 4079 struct ip_fw *rule; 4080 int i; 4081 time_t boot_seconds; 4082 4083 boot_seconds = boottime.tv_sec; 4084 /* XXX this can take a long time and locking will block packet flow */ 4085 IPFW_RLOCK(chain); 4086 for (rule = chain->rules; rule ; rule = rule->next) { 4087 /* 4088 * Verify the entry fits in the buffer in case the 4089 * rules changed between calculating buffer space and 4090 * now. This would be better done using a generation 4091 * number but should suffice for now. 4092 */ 4093 i = RULESIZE(rule); 4094 if (bp + i <= ep) { 4095 bcopy(rule, bp, i); 4096 /* 4097 * XXX HACK. Store the disable mask in the "next" pointer 4098 * in a wild attempt to keep the ABI the same. 4099 * Why do we do this on EVERY rule? 4100 */ 4101 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 4102 sizeof(set_disable)); 4103 if (((struct ip_fw *)bp)->timestamp) 4104 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4105 bp += i; 4106 } 4107 } 4108 IPFW_RUNLOCK(chain); 4109 if (ipfw_dyn_v) { 4110 ipfw_dyn_rule *p, *last = NULL; 4111 4112 IPFW_DYN_LOCK(); 4113 for (i = 0 ; i < curr_dyn_buckets; i++) 4114 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4115 if (bp + sizeof *p <= ep) { 4116 ipfw_dyn_rule *dst = 4117 (ipfw_dyn_rule *)bp; 4118 bcopy(p, dst, sizeof *p); 4119 bcopy(&(p->rule->rulenum), &(dst->rule), 4120 sizeof(p->rule->rulenum)); 4121 /* 4122 * store set number into high word of 4123 * dst->rule pointer. 4124 */ 4125 bcopy(&(p->rule->set), 4126 (char *)&dst->rule + 4127 sizeof(p->rule->rulenum), 4128 sizeof(p->rule->set)); 4129 /* 4130 * store a non-null value in "next". 4131 * The userland code will interpret a 4132 * NULL here as a marker 4133 * for the last dynamic rule. 4134 */ 4135 bcopy(&dst, &dst->next, sizeof(dst)); 4136 last = dst; 4137 dst->expire = 4138 TIME_LEQ(dst->expire, time_uptime) ? 4139 0 : dst->expire - time_uptime ; 4140 bp += sizeof(ipfw_dyn_rule); 4141 } 4142 } 4143 IPFW_DYN_UNLOCK(); 4144 if (last != NULL) /* mark last dynamic rule */ 4145 bzero(&last->next, sizeof(last)); 4146 } 4147 return (bp - (char *)buf); 4148 } 4149 4150 4151 /** 4152 * {set|get}sockopt parser. 4153 */ 4154 static int 4155 ipfw_ctl(struct sockopt *sopt) 4156 { 4157 #define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4158 int error; 4159 size_t size; 4160 struct ip_fw *buf, *rule; 4161 u_int32_t rulenum[2]; 4162 4163 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4164 if (error) 4165 return (error); 4166 4167 /* 4168 * Disallow modifications in really-really secure mode, but still allow 4169 * the logging counters to be reset. 4170 */ 4171 if (sopt->sopt_name == IP_FW_ADD || 4172 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4173 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4174 if (error) 4175 return (error); 4176 } 4177 4178 error = 0; 4179 4180 switch (sopt->sopt_name) { 4181 case IP_FW_GET: 4182 /* 4183 * pass up a copy of the current rules. Static rules 4184 * come first (the last of which has number IPFW_DEFAULT_RULE), 4185 * followed by a possibly empty list of dynamic rule. 4186 * The last dynamic rule has NULL in the "next" field. 4187 * 4188 * Note that the calculated size is used to bound the 4189 * amount of data returned to the user. The rule set may 4190 * change between calculating the size and returning the 4191 * data in which case we'll just return what fits. 4192 */ 4193 size = static_len; /* size of static rules */ 4194 if (ipfw_dyn_v) /* add size of dyn.rules */ 4195 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4196 4197 /* 4198 * XXX todo: if the user passes a short length just to know 4199 * how much room is needed, do not bother filling up the 4200 * buffer, just jump to the sooptcopyout. 4201 */ 4202 buf = malloc(size, M_TEMP, M_WAITOK); 4203 error = sooptcopyout(sopt, buf, 4204 ipfw_getrules(&layer3_chain, buf, size)); 4205 free(buf, M_TEMP); 4206 break; 4207 4208 case IP_FW_FLUSH: 4209 /* 4210 * Normally we cannot release the lock on each iteration. 4211 * We could do it here only because we start from the head all 4212 * the times so there is no risk of missing some entries. 4213 * On the other hand, the risk is that we end up with 4214 * a very inconsistent ruleset, so better keep the lock 4215 * around the whole cycle. 4216 * 4217 * XXX this code can be improved by resetting the head of 4218 * the list to point to the default rule, and then freeing 4219 * the old list without the need for a lock. 4220 */ 4221 4222 IPFW_WLOCK(&layer3_chain); 4223 layer3_chain.reap = NULL; 4224 free_chain(&layer3_chain, 0 /* keep default rule */); 4225 rule = layer3_chain.reap; 4226 layer3_chain.reap = NULL; 4227 IPFW_WUNLOCK(&layer3_chain); 4228 if (rule != NULL) 4229 reap_rules(rule); 4230 break; 4231 4232 case IP_FW_ADD: 4233 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4234 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4235 sizeof(struct ip_fw) ); 4236 if (error == 0) 4237 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4238 if (error == 0) { 4239 error = add_rule(&layer3_chain, rule); 4240 size = RULESIZE(rule); 4241 if (!error && sopt->sopt_dir == SOPT_GET) 4242 error = sooptcopyout(sopt, rule, size); 4243 } 4244 free(rule, M_TEMP); 4245 break; 4246 4247 case IP_FW_DEL: 4248 /* 4249 * IP_FW_DEL is used for deleting single rules or sets, 4250 * and (ab)used to atomically manipulate sets. Argument size 4251 * is used to distinguish between the two: 4252 * sizeof(u_int32_t) 4253 * delete single rule or set of rules, 4254 * or reassign rules (or sets) to a different set. 4255 * 2*sizeof(u_int32_t) 4256 * atomic disable/enable sets. 4257 * first u_int32_t contains sets to be disabled, 4258 * second u_int32_t contains sets to be enabled. 4259 */ 4260 error = sooptcopyin(sopt, rulenum, 4261 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4262 if (error) 4263 break; 4264 size = sopt->sopt_valsize; 4265 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4266 error = del_entry(&layer3_chain, rulenum[0]); 4267 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4268 set_disable = 4269 (set_disable | rulenum[0]) & ~rulenum[1] & 4270 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4271 else 4272 error = EINVAL; 4273 break; 4274 4275 case IP_FW_ZERO: 4276 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 4277 rulenum[0] = 0; 4278 if (sopt->sopt_val != 0) { 4279 error = sooptcopyin(sopt, rulenum, 4280 sizeof(u_int32_t), sizeof(u_int32_t)); 4281 if (error) 4282 break; 4283 } 4284 error = zero_entry(&layer3_chain, rulenum[0], 4285 sopt->sopt_name == IP_FW_RESETLOG); 4286 break; 4287 4288 case IP_FW_TABLE_ADD: 4289 { 4290 ipfw_table_entry ent; 4291 4292 error = sooptcopyin(sopt, &ent, 4293 sizeof(ent), sizeof(ent)); 4294 if (error) 4295 break; 4296 error = add_table_entry(&layer3_chain, ent.tbl, 4297 ent.addr, ent.masklen, ent.value); 4298 } 4299 break; 4300 4301 case IP_FW_TABLE_DEL: 4302 { 4303 ipfw_table_entry ent; 4304 4305 error = sooptcopyin(sopt, &ent, 4306 sizeof(ent), sizeof(ent)); 4307 if (error) 4308 break; 4309 error = del_table_entry(&layer3_chain, ent.tbl, 4310 ent.addr, ent.masklen); 4311 } 4312 break; 4313 4314 case IP_FW_TABLE_FLUSH: 4315 { 4316 u_int16_t tbl; 4317 4318 error = sooptcopyin(sopt, &tbl, 4319 sizeof(tbl), sizeof(tbl)); 4320 if (error) 4321 break; 4322 IPFW_WLOCK(&layer3_chain); 4323 error = flush_table(&layer3_chain, tbl); 4324 IPFW_WUNLOCK(&layer3_chain); 4325 } 4326 break; 4327 4328 case IP_FW_TABLE_GETSIZE: 4329 { 4330 u_int32_t tbl, cnt; 4331 4332 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4333 sizeof(tbl)))) 4334 break; 4335 IPFW_RLOCK(&layer3_chain); 4336 error = count_table(&layer3_chain, tbl, &cnt); 4337 IPFW_RUNLOCK(&layer3_chain); 4338 if (error) 4339 break; 4340 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4341 } 4342 break; 4343 4344 case IP_FW_TABLE_LIST: 4345 { 4346 ipfw_table *tbl; 4347 4348 if (sopt->sopt_valsize < sizeof(*tbl)) { 4349 error = EINVAL; 4350 break; 4351 } 4352 size = sopt->sopt_valsize; 4353 tbl = malloc(size, M_TEMP, M_WAITOK); 4354 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4355 if (error) { 4356 free(tbl, M_TEMP); 4357 break; 4358 } 4359 tbl->size = (size - sizeof(*tbl)) / 4360 sizeof(ipfw_table_entry); 4361 IPFW_RLOCK(&layer3_chain); 4362 error = dump_table(&layer3_chain, tbl); 4363 IPFW_RUNLOCK(&layer3_chain); 4364 if (error) { 4365 free(tbl, M_TEMP); 4366 break; 4367 } 4368 error = sooptcopyout(sopt, tbl, size); 4369 free(tbl, M_TEMP); 4370 } 4371 break; 4372 4373 case IP_FW_NAT_CFG: 4374 if (IPFW_NAT_LOADED) 4375 error = ipfw_nat_cfg_ptr(sopt); 4376 else { 4377 printf("IP_FW_NAT_CFG: %s\n", 4378 "ipfw_nat not present, please load it"); 4379 error = EINVAL; 4380 } 4381 break; 4382 4383 case IP_FW_NAT_DEL: 4384 if (IPFW_NAT_LOADED) 4385 error = ipfw_nat_del_ptr(sopt); 4386 else { 4387 printf("IP_FW_NAT_DEL: %s\n", 4388 "ipfw_nat not present, please load it"); 4389 error = EINVAL; 4390 } 4391 break; 4392 4393 case IP_FW_NAT_GET_CONFIG: 4394 if (IPFW_NAT_LOADED) 4395 error = ipfw_nat_get_cfg_ptr(sopt); 4396 else { 4397 printf("IP_FW_NAT_GET_CFG: %s\n", 4398 "ipfw_nat not present, please load it"); 4399 error = EINVAL; 4400 } 4401 break; 4402 4403 case IP_FW_NAT_GET_LOG: 4404 if (IPFW_NAT_LOADED) 4405 error = ipfw_nat_get_log_ptr(sopt); 4406 else { 4407 printf("IP_FW_NAT_GET_LOG: %s\n", 4408 "ipfw_nat not present, please load it"); 4409 error = EINVAL; 4410 } 4411 break; 4412 4413 default: 4414 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4415 error = EINVAL; 4416 } 4417 4418 return (error); 4419 #undef RULE_MAXSIZE 4420 } 4421 4422 /** 4423 * dummynet needs a reference to the default rule, because rules can be 4424 * deleted while packets hold a reference to them. When this happens, 4425 * dummynet changes the reference to the default rule (it could well be a 4426 * NULL pointer, but this way we do not need to check for the special 4427 * case, plus here he have info on the default behaviour). 4428 */ 4429 struct ip_fw *ip_fw_default_rule; 4430 4431 /* 4432 * This procedure is only used to handle keepalives. It is invoked 4433 * every dyn_keepalive_period 4434 */ 4435 static void 4436 ipfw_tick(void * __unused unused) 4437 { 4438 struct mbuf *m0, *m, *mnext, **mtailp; 4439 int i; 4440 ipfw_dyn_rule *q; 4441 4442 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4443 goto done; 4444 4445 /* 4446 * We make a chain of packets to go out here -- not deferring 4447 * until after we drop the IPFW dynamic rule lock would result 4448 * in a lock order reversal with the normal packet input -> ipfw 4449 * call stack. 4450 */ 4451 m0 = NULL; 4452 mtailp = &m0; 4453 IPFW_DYN_LOCK(); 4454 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4455 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4456 if (q->dyn_type == O_LIMIT_PARENT) 4457 continue; 4458 if (q->id.proto != IPPROTO_TCP) 4459 continue; 4460 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4461 continue; 4462 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4463 q->expire)) 4464 continue; /* too early */ 4465 if (TIME_LEQ(q->expire, time_uptime)) 4466 continue; /* too late, rule expired */ 4467 4468 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4469 q->ack_fwd, TH_SYN); 4470 if (*mtailp != NULL) 4471 mtailp = &(*mtailp)->m_nextpkt; 4472 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4473 q->ack_rev, 0); 4474 if (*mtailp != NULL) 4475 mtailp = &(*mtailp)->m_nextpkt; 4476 } 4477 } 4478 IPFW_DYN_UNLOCK(); 4479 for (m = mnext = m0; m != NULL; m = mnext) { 4480 mnext = m->m_nextpkt; 4481 m->m_nextpkt = NULL; 4482 ip_output(m, NULL, NULL, 0, NULL, NULL); 4483 } 4484 done: 4485 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4486 } 4487 4488 int 4489 ipfw_init(void) 4490 { 4491 struct ip_fw default_rule; 4492 int error; 4493 4494 #ifdef INET6 4495 /* Setup IPv6 fw sysctl tree. */ 4496 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4497 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4498 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4499 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4500 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4501 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4502 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4503 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4504 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4505 &fw_deny_unknown_exthdrs, 0, 4506 "Deny packets with unknown IPv6 Extension Headers"); 4507 #endif 4508 4509 layer3_chain.rules = NULL; 4510 IPFW_LOCK_INIT(&layer3_chain); 4511 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4512 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4513 UMA_ALIGN_PTR, 0); 4514 IPFW_DYN_LOCK_INIT(); 4515 callout_init(&ipfw_timeout, CALLOUT_MPSAFE); 4516 4517 bzero(&default_rule, sizeof default_rule); 4518 4519 default_rule.act_ofs = 0; 4520 default_rule.rulenum = IPFW_DEFAULT_RULE; 4521 default_rule.cmd_len = 1; 4522 default_rule.set = RESVD_SET; 4523 4524 default_rule.cmd[0].len = 1; 4525 default_rule.cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 4526 4527 error = add_rule(&layer3_chain, &default_rule); 4528 if (error != 0) { 4529 printf("ipfw2: error %u initializing default rule " 4530 "(support disabled)\n", error); 4531 IPFW_DYN_LOCK_DESTROY(); 4532 IPFW_LOCK_DESTROY(&layer3_chain); 4533 uma_zdestroy(ipfw_dyn_rule_zone); 4534 return (error); 4535 } 4536 4537 ip_fw_default_rule = layer3_chain.rules; 4538 printf("ipfw2 " 4539 #ifdef INET6 4540 "(+ipv6) " 4541 #endif 4542 "initialized, divert %s, nat %s, " 4543 "rule-based forwarding " 4544 #ifdef IPFIREWALL_FORWARD 4545 "enabled, " 4546 #else 4547 "disabled, " 4548 #endif 4549 "default to %s, logging ", 4550 #ifdef IPDIVERT 4551 "enabled", 4552 #else 4553 "loadable", 4554 #endif 4555 #ifdef IPFIREWALL_NAT 4556 "enabled", 4557 #else 4558 "loadable", 4559 #endif 4560 4561 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4562 4563 #ifdef IPFIREWALL_VERBOSE 4564 fw_verbose = 1; 4565 #endif 4566 #ifdef IPFIREWALL_VERBOSE_LIMIT 4567 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4568 #endif 4569 if (fw_verbose == 0) 4570 printf("disabled\n"); 4571 else if (verbose_limit == 0) 4572 printf("unlimited\n"); 4573 else 4574 printf("limited to %d packets/entry by default\n", 4575 verbose_limit); 4576 4577 error = init_tables(&layer3_chain); 4578 if (error) { 4579 IPFW_DYN_LOCK_DESTROY(); 4580 IPFW_LOCK_DESTROY(&layer3_chain); 4581 uma_zdestroy(ipfw_dyn_rule_zone); 4582 return (error); 4583 } 4584 ip_fw_ctl_ptr = ipfw_ctl; 4585 ip_fw_chk_ptr = ipfw_chk; 4586 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4587 LIST_INIT(&layer3_chain.nat); 4588 return (0); 4589 } 4590 4591 void 4592 ipfw_destroy(void) 4593 { 4594 struct ip_fw *reap; 4595 4596 ip_fw_chk_ptr = NULL; 4597 ip_fw_ctl_ptr = NULL; 4598 callout_drain(&ipfw_timeout); 4599 IPFW_WLOCK(&layer3_chain); 4600 flush_tables(&layer3_chain); 4601 layer3_chain.reap = NULL; 4602 free_chain(&layer3_chain, 1 /* kill default rule */); 4603 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4604 IPFW_WUNLOCK(&layer3_chain); 4605 if (reap != NULL) 4606 reap_rules(reap); 4607 IPFW_DYN_LOCK_DESTROY(); 4608 uma_zdestroy(ipfw_dyn_rule_zone); 4609 if (ipfw_dyn_v != NULL) 4610 free(ipfw_dyn_v, M_IPFW); 4611 IPFW_LOCK_DESTROY(&layer3_chain); 4612 4613 #ifdef INET6 4614 /* Free IPv6 fw sysctl tree. */ 4615 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4616 #endif 4617 4618 printf("IP firewall unloaded\n"); 4619 }
Cache object: 4da7f8f9ee17fa4e08b296514c3ba53c
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]