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 * $FreeBSD: releng/6.3/sys/netinet/ip_fw2.c 170398 2007-06-07 09:50:53Z bz $ 26 */ 27 28 #define DEB(x) 29 #define DDB(x) x 30 31 /* 32 * Implement IP packet firewall (new version) 33 */ 34 35 #if !defined(KLD_MODULE) 36 #include "opt_ipfw.h" 37 #include "opt_ip6fw.h" 38 #include "opt_ipdivert.h" 39 #include "opt_ipdn.h" 40 #include "opt_inet.h" 41 #ifndef INET 42 #error IPFIREWALL requires INET. 43 #endif /* INET */ 44 #endif 45 #include "opt_inet6.h" 46 #include "opt_ipsec.h" 47 #include "opt_mac.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/condvar.h> 52 #include <sys/malloc.h> 53 #include <sys/mbuf.h> 54 #include <sys/kernel.h> 55 #include <sys/jail.h> 56 #include <sys/mac.h> 57 #include <sys/module.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/if.h> 65 #include <net/radix.h> 66 #include <net/route.h> 67 #include <netinet/in.h> 68 #include <netinet/in_systm.h> 69 #include <netinet/in_var.h> 70 #include <netinet/in_pcb.h> 71 #include <netinet/ip.h> 72 #include <netinet/ip_var.h> 73 #include <netinet/ip_icmp.h> 74 #include <netinet/ip_fw.h> 75 #include <netinet/ip_divert.h> 76 #include <netinet/ip_dummynet.h> 77 #include <netinet/ip_carp.h> 78 #include <netinet/pim.h> 79 #include <netinet/tcp.h> 80 #include <netinet/tcp_timer.h> 81 #include <netinet/tcp_var.h> 82 #include <netinet/tcpip.h> 83 #include <netinet/udp.h> 84 #include <netinet/udp_var.h> 85 86 #include <netgraph/ng_ipfw.h> 87 88 #include <altq/if_altq.h> 89 90 #ifdef IPSEC 91 #include <netinet6/ipsec.h> 92 #endif 93 94 #include <netinet/ip6.h> 95 #include <netinet/icmp6.h> 96 #ifdef INET6 97 #include <netinet6/scope6_var.h> 98 #endif 99 100 #include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 101 102 #include <machine/in_cksum.h> /* XXX for in_cksum */ 103 104 /* 105 * set_disable contains one bit per set value (0..31). 106 * If the bit is set, all rules with the corresponding set 107 * are disabled. Set RESVD_SET(31) is reserved for the default rule 108 * and rules that are not deleted by the flush command, 109 * and CANNOT be disabled. 110 * Rules in set RESVD_SET can only be deleted explicitly. 111 */ 112 static u_int32_t set_disable; 113 114 static int fw_verbose; 115 static int verbose_limit; 116 117 static struct callout ipfw_timeout; 118 static uma_zone_t ipfw_dyn_rule_zone; 119 #define IPFW_DEFAULT_RULE 65535 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 #define IPFW_TABLES_MAX 128 135 struct ip_fw_chain { 136 struct ip_fw *rules; /* list of rules */ 137 struct ip_fw *reap; /* list of rules to reap */ 138 struct radix_node_head *tables[IPFW_TABLES_MAX]; 139 struct mtx mtx; /* lock guarding rule list */ 140 int busy_count; /* busy count for rw locks */ 141 int want_write; 142 struct cv cv; 143 }; 144 #define IPFW_LOCK_INIT(_chain) \ 145 mtx_init(&(_chain)->mtx, "IPFW static rules", NULL, \ 146 MTX_DEF | MTX_RECURSE) 147 #define IPFW_LOCK_DESTROY(_chain) mtx_destroy(&(_chain)->mtx) 148 #define IPFW_WLOCK_ASSERT(_chain) do { \ 149 mtx_assert(&(_chain)->mtx, MA_OWNED); \ 150 NET_ASSERT_GIANT(); \ 151 } while (0) 152 153 static __inline void 154 IPFW_RLOCK(struct ip_fw_chain *chain) 155 { 156 mtx_lock(&chain->mtx); 157 chain->busy_count++; 158 mtx_unlock(&chain->mtx); 159 } 160 161 static __inline void 162 IPFW_RUNLOCK(struct ip_fw_chain *chain) 163 { 164 mtx_lock(&chain->mtx); 165 chain->busy_count--; 166 if (chain->busy_count == 0 && chain->want_write) 167 cv_signal(&chain->cv); 168 mtx_unlock(&chain->mtx); 169 } 170 171 static __inline void 172 IPFW_WLOCK(struct ip_fw_chain *chain) 173 { 174 mtx_lock(&chain->mtx); 175 chain->want_write++; 176 while (chain->busy_count > 0) 177 cv_wait(&chain->cv, &chain->mtx); 178 } 179 180 static __inline void 181 IPFW_WUNLOCK(struct ip_fw_chain *chain) 182 { 183 chain->want_write--; 184 cv_signal(&chain->cv); 185 mtx_unlock(&chain->mtx); 186 } 187 188 /* 189 * list of rules for layer 3 190 */ 191 static struct ip_fw_chain layer3_chain; 192 193 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 194 MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 195 196 struct table_entry { 197 struct radix_node rn[2]; 198 struct sockaddr_in addr, mask; 199 u_int32_t value; 200 }; 201 202 static int fw_debug = 1; 203 static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 204 205 #ifdef SYSCTL_NODE 206 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 207 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 208 CTLFLAG_RW | CTLFLAG_SECURE3, 209 &fw_enable, 0, "Enable ipfw"); 210 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 211 &autoinc_step, 0, "Rule number autincrement step"); 212 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 213 CTLFLAG_RW | CTLFLAG_SECURE3, 214 &fw_one_pass, 0, 215 "Only do a single pass through ipfw when using dummynet(4)"); 216 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 217 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 218 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 219 CTLFLAG_RW | CTLFLAG_SECURE3, 220 &fw_verbose, 0, "Log matches to ipfw rules"); 221 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 222 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 223 224 /* 225 * Description of dynamic rules. 226 * 227 * Dynamic rules are stored in lists accessed through a hash table 228 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 229 * be modified through the sysctl variable dyn_buckets which is 230 * updated when the table becomes empty. 231 * 232 * XXX currently there is only one list, ipfw_dyn. 233 * 234 * When a packet is received, its address fields are first masked 235 * with the mask defined for the rule, then hashed, then matched 236 * against the entries in the corresponding list. 237 * Dynamic rules can be used for different purposes: 238 * + stateful rules; 239 * + enforcing limits on the number of sessions; 240 * + in-kernel NAT (not implemented yet) 241 * 242 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 243 * measured in seconds and depending on the flags. 244 * 245 * The total number of dynamic rules is stored in dyn_count. 246 * The max number of dynamic rules is dyn_max. When we reach 247 * the maximum number of rules we do not create anymore. This is 248 * done to avoid consuming too much memory, but also too much 249 * time when searching on each packet (ideally, we should try instead 250 * to put a limit on the length of the list on each bucket...). 251 * 252 * Each dynamic rule holds a pointer to the parent ipfw rule so 253 * we know what action to perform. Dynamic rules are removed when 254 * the parent rule is deleted. XXX we should make them survive. 255 * 256 * There are some limitations with dynamic rules -- we do not 257 * obey the 'randomized match', and we do not do multiple 258 * passes through the firewall. XXX check the latter!!! 259 */ 260 static ipfw_dyn_rule **ipfw_dyn_v = NULL; 261 static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 262 static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 263 264 static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 265 #define IPFW_DYN_LOCK_INIT() \ 266 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 267 #define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 268 #define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 269 #define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 270 #define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 271 272 /* 273 * Timeouts for various events in handing dynamic rules. 274 */ 275 static u_int32_t dyn_ack_lifetime = 300; 276 static u_int32_t dyn_syn_lifetime = 20; 277 static u_int32_t dyn_fin_lifetime = 1; 278 static u_int32_t dyn_rst_lifetime = 1; 279 static u_int32_t dyn_udp_lifetime = 10; 280 static u_int32_t dyn_short_lifetime = 5; 281 282 /* 283 * Keepalives are sent if dyn_keepalive is set. They are sent every 284 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 285 * seconds of lifetime of a rule. 286 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 287 * than dyn_keepalive_period. 288 */ 289 290 static u_int32_t dyn_keepalive_interval = 20; 291 static u_int32_t dyn_keepalive_period = 5; 292 static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 293 294 static u_int32_t static_count; /* # of static rules */ 295 static u_int32_t static_len; /* size in bytes of static rules */ 296 static u_int32_t dyn_count; /* # of dynamic rules */ 297 static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 298 299 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 300 &dyn_buckets, 0, "Number of dyn. buckets"); 301 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 302 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 303 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 304 &dyn_count, 0, "Number of dyn. rules"); 305 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 306 &dyn_max, 0, "Max number of dyn. rules"); 307 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 308 &static_count, 0, "Number of static rules"); 309 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 310 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 311 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 312 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 313 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 314 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 315 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 316 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 317 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 318 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 319 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 320 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 321 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 322 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 323 324 #ifdef INET6 325 /* 326 * IPv6 specific variables 327 */ 328 SYSCTL_DECL(_net_inet6_ip6); 329 330 static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 331 static struct sysctl_oid *ip6_fw_sysctl_tree; 332 #endif /* INET6 */ 333 #endif /* SYSCTL_NODE */ 334 335 static int fw_deny_unknown_exthdrs = 1; 336 337 338 /* 339 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 340 * Other macros just cast void * into the appropriate type 341 */ 342 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 343 #define TCP(p) ((struct tcphdr *)(p)) 344 #define UDP(p) ((struct udphdr *)(p)) 345 #define ICMP(p) ((struct icmphdr *)(p)) 346 #define ICMP6(p) ((struct icmp6_hdr *)(p)) 347 348 static __inline int 349 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 350 { 351 int type = icmp->icmp_type; 352 353 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 354 } 355 356 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 357 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 358 359 static int 360 is_icmp_query(struct icmphdr *icmp) 361 { 362 int type = icmp->icmp_type; 363 364 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 365 } 366 #undef TT 367 368 /* 369 * The following checks use two arrays of 8 or 16 bits to store the 370 * bits that we want set or clear, respectively. They are in the 371 * low and high half of cmd->arg1 or cmd->d[0]. 372 * 373 * We scan options and store the bits we find set. We succeed if 374 * 375 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 376 * 377 * The code is sometimes optimized not to store additional variables. 378 */ 379 380 static int 381 flags_match(ipfw_insn *cmd, u_int8_t bits) 382 { 383 u_char want_clear; 384 bits = ~bits; 385 386 if ( ((cmd->arg1 & 0xff) & bits) != 0) 387 return 0; /* some bits we want set were clear */ 388 want_clear = (cmd->arg1 >> 8) & 0xff; 389 if ( (want_clear & bits) != want_clear) 390 return 0; /* some bits we want clear were set */ 391 return 1; 392 } 393 394 static int 395 ipopts_match(struct ip *ip, ipfw_insn *cmd) 396 { 397 int optlen, bits = 0; 398 u_char *cp = (u_char *)(ip + 1); 399 int x = (ip->ip_hl << 2) - sizeof (struct ip); 400 401 for (; x > 0; x -= optlen, cp += optlen) { 402 int opt = cp[IPOPT_OPTVAL]; 403 404 if (opt == IPOPT_EOL) 405 break; 406 if (opt == IPOPT_NOP) 407 optlen = 1; 408 else { 409 optlen = cp[IPOPT_OLEN]; 410 if (optlen <= 0 || optlen > x) 411 return 0; /* invalid or truncated */ 412 } 413 switch (opt) { 414 415 default: 416 break; 417 418 case IPOPT_LSRR: 419 bits |= IP_FW_IPOPT_LSRR; 420 break; 421 422 case IPOPT_SSRR: 423 bits |= IP_FW_IPOPT_SSRR; 424 break; 425 426 case IPOPT_RR: 427 bits |= IP_FW_IPOPT_RR; 428 break; 429 430 case IPOPT_TS: 431 bits |= IP_FW_IPOPT_TS; 432 break; 433 } 434 } 435 return (flags_match(cmd, bits)); 436 } 437 438 static int 439 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 440 { 441 int optlen, bits = 0; 442 u_char *cp = (u_char *)(tcp + 1); 443 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 444 445 for (; x > 0; x -= optlen, cp += optlen) { 446 int opt = cp[0]; 447 if (opt == TCPOPT_EOL) 448 break; 449 if (opt == TCPOPT_NOP) 450 optlen = 1; 451 else { 452 optlen = cp[1]; 453 if (optlen <= 0) 454 break; 455 } 456 457 switch (opt) { 458 459 default: 460 break; 461 462 case TCPOPT_MAXSEG: 463 bits |= IP_FW_TCPOPT_MSS; 464 break; 465 466 case TCPOPT_WINDOW: 467 bits |= IP_FW_TCPOPT_WINDOW; 468 break; 469 470 case TCPOPT_SACK_PERMITTED: 471 case TCPOPT_SACK: 472 bits |= IP_FW_TCPOPT_SACK; 473 break; 474 475 case TCPOPT_TIMESTAMP: 476 bits |= IP_FW_TCPOPT_TS; 477 break; 478 479 } 480 } 481 return (flags_match(cmd, bits)); 482 } 483 484 static int 485 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 486 { 487 if (ifp == NULL) /* no iface with this packet, match fails */ 488 return 0; 489 /* Check by name or by IP address */ 490 if (cmd->name[0] != '\0') { /* match by name */ 491 /* Check name */ 492 if (cmd->p.glob) { 493 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 494 return(1); 495 } else { 496 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 497 return(1); 498 } 499 } else { 500 struct ifaddr *ia; 501 502 /* XXX lock? */ 503 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 504 if (ia->ifa_addr == NULL) 505 continue; 506 if (ia->ifa_addr->sa_family != AF_INET) 507 continue; 508 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 509 (ia->ifa_addr))->sin_addr.s_addr) 510 return(1); /* match */ 511 } 512 } 513 return(0); /* no match, fail ... */ 514 } 515 516 /* 517 * The verify_path function checks if a route to the src exists and 518 * if it is reachable via ifp (when provided). 519 * 520 * The 'verrevpath' option checks that the interface that an IP packet 521 * arrives on is the same interface that traffic destined for the 522 * packet's source address would be routed out of. The 'versrcreach' 523 * option just checks that the source address is reachable via any route 524 * (except default) in the routing table. These two are a measure to block 525 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 526 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 527 * is purposely reminiscent of the Cisco IOS command, 528 * 529 * ip verify unicast reverse-path 530 * ip verify unicast source reachable-via any 531 * 532 * which implements the same functionality. But note that syntax is 533 * misleading. The check may be performed on all IP packets whether unicast, 534 * multicast, or broadcast. 535 */ 536 static int 537 verify_path(struct in_addr src, struct ifnet *ifp) 538 { 539 struct route ro; 540 struct sockaddr_in *dst; 541 542 bzero(&ro, sizeof(ro)); 543 544 dst = (struct sockaddr_in *)&(ro.ro_dst); 545 dst->sin_family = AF_INET; 546 dst->sin_len = sizeof(*dst); 547 dst->sin_addr = src; 548 rtalloc_ign(&ro, RTF_CLONING); 549 550 if (ro.ro_rt == NULL) 551 return 0; 552 553 /* 554 * If ifp is provided, check for equality with rtentry. 555 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 556 * in order to pass packets injected back by if_simloop(): 557 * if useloopback == 1 routing entry (via lo0) for our own address 558 * may exist, so we need to handle routing assymetry. 559 */ 560 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 561 RTFREE(ro.ro_rt); 562 return 0; 563 } 564 565 /* if no ifp provided, check if rtentry is not default route */ 566 if (ifp == NULL && 567 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 568 RTFREE(ro.ro_rt); 569 return 0; 570 } 571 572 /* or if this is a blackhole/reject route */ 573 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 574 RTFREE(ro.ro_rt); 575 return 0; 576 } 577 578 /* found valid route */ 579 RTFREE(ro.ro_rt); 580 return 1; 581 } 582 583 #ifdef INET6 584 /* 585 * ipv6 specific rules here... 586 */ 587 static __inline int 588 icmp6type_match (int type, ipfw_insn_u32 *cmd) 589 { 590 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 591 } 592 593 static int 594 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 595 { 596 int i; 597 for (i=0; i <= cmd->o.arg1; ++i ) 598 if (curr_flow == cmd->d[i] ) 599 return 1; 600 return 0; 601 } 602 603 /* support for IP6_*_ME opcodes */ 604 static int 605 search_ip6_addr_net (struct in6_addr * ip6_addr) 606 { 607 struct ifnet *mdc; 608 struct ifaddr *mdc2; 609 struct in6_ifaddr *fdm; 610 struct in6_addr copia; 611 612 TAILQ_FOREACH(mdc, &ifnet, if_link) 613 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 614 if (!mdc2->ifa_addr) 615 continue; 616 if (mdc2->ifa_addr->sa_family == AF_INET6) { 617 fdm = (struct in6_ifaddr *)mdc2; 618 copia = fdm->ia_addr.sin6_addr; 619 /* need for leaving scope_id in the sock_addr */ 620 in6_clearscope(&copia); 621 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 622 return 1; 623 } 624 } 625 return 0; 626 } 627 628 static int 629 verify_path6(struct in6_addr *src, struct ifnet *ifp) 630 { 631 struct route_in6 ro; 632 struct sockaddr_in6 *dst; 633 634 bzero(&ro, sizeof(ro)); 635 636 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 637 dst->sin6_family = AF_INET6; 638 dst->sin6_len = sizeof(*dst); 639 dst->sin6_addr = *src; 640 rtalloc_ign((struct route *)&ro, RTF_CLONING); 641 642 if (ro.ro_rt == NULL) 643 return 0; 644 645 /* 646 * if ifp is provided, check for equality with rtentry 647 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 648 * to support the case of sending packets to an address of our own. 649 * (where the former interface is the first argument of if_simloop() 650 * (=ifp), the latter is lo0) 651 */ 652 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 653 RTFREE(ro.ro_rt); 654 return 0; 655 } 656 657 /* if no ifp provided, check if rtentry is not default route */ 658 if (ifp == NULL && 659 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 660 RTFREE(ro.ro_rt); 661 return 0; 662 } 663 664 /* or if this is a blackhole/reject route */ 665 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 666 RTFREE(ro.ro_rt); 667 return 0; 668 } 669 670 /* found valid route */ 671 RTFREE(ro.ro_rt); 672 return 1; 673 674 } 675 static __inline int 676 hash_packet6(struct ipfw_flow_id *id) 677 { 678 u_int32_t i; 679 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 680 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 681 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 682 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 683 (id->dst_port) ^ (id->src_port); 684 return i; 685 } 686 687 static int 688 is_icmp6_query(int icmp6_type) 689 { 690 if ((icmp6_type <= ICMP6_MAXTYPE) && 691 (icmp6_type == ICMP6_ECHO_REQUEST || 692 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 693 icmp6_type == ICMP6_WRUREQUEST || 694 icmp6_type == ICMP6_FQDN_QUERY || 695 icmp6_type == ICMP6_NI_QUERY)) 696 return (1); 697 698 return (0); 699 } 700 701 static void 702 send_reject6(struct ip_fw_args *args, int code, u_short offset, u_int hlen, struct ip6_hdr *ip6) 703 { 704 struct mbuf *m; 705 706 m = args->m; 707 if (code == ICMP6_UNREACH_RST && offset == 0 && 708 args->f_id.proto == IPPROTO_TCP) { 709 struct tcphdr *tcp; 710 tcp_seq ack, seq; 711 int flags; 712 struct { 713 struct ip6_hdr ip6; 714 struct tcphdr th; 715 } ti; 716 tcp = (struct tcphdr *)((char *)ip6 + hlen); 717 718 if ((tcp->th_flags & TH_RST) != 0) { 719 m_freem(m); 720 args->m = NULL; 721 return; 722 } 723 724 ti.ip6 = *ip6; 725 ti.th = *tcp; 726 ti.th.th_seq = ntohl(ti.th.th_seq); 727 ti.th.th_ack = ntohl(ti.th.th_ack); 728 ti.ip6.ip6_nxt = IPPROTO_TCP; 729 730 if (ti.th.th_flags & TH_ACK) { 731 ack = 0; 732 seq = ti.th.th_ack; 733 flags = TH_RST; 734 } else { 735 ack = ti.th.th_seq; 736 if ((m->m_flags & M_PKTHDR) != 0) { 737 /* 738 * total new data to ACK is: 739 * total packet length, 740 * minus the header length, 741 * minus the tcp header length. 742 */ 743 ack += m->m_pkthdr.len - hlen 744 - (ti.th.th_off << 2); 745 } else if (ip6->ip6_plen) { 746 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 747 hlen - (ti.th.th_off << 2); 748 } else { 749 m_freem(m); 750 return; 751 } 752 if (tcp->th_flags & TH_SYN) 753 ack++; 754 seq = 0; 755 flags = TH_RST|TH_ACK; 756 } 757 bcopy(&ti, ip6, sizeof(ti)); 758 /* 759 * m is only used to recycle the mbuf 760 * The data in it is never read so we don't need 761 * to correct the offsets or anything 762 */ 763 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 764 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 765 #if 0 766 /* 767 * Unlike above, the mbufs need to line up with the ip6 hdr, 768 * as the contents are read. We need to m_adj() the 769 * needed amount. 770 * The mbuf will however be thrown away so we can adjust it. 771 * Remember we did an m_pullup on it already so we 772 * can make some assumptions about contiguousness. 773 */ 774 if (args->L3offset) 775 m_adj(m, args->L3offset); 776 #endif 777 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 778 } else 779 m_freem(m); 780 781 args->m = NULL; 782 } 783 784 #endif /* INET6 */ 785 786 static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 787 788 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 789 #define SNP(buf) buf, sizeof(buf) 790 791 /* 792 * We enter here when we have a rule with O_LOG. 793 * XXX this function alone takes about 2Kbytes of code! 794 */ 795 static void 796 ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 797 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, struct ip *ip) 798 { 799 struct ether_header *eh = args->eh; 800 char *action; 801 int limit_reached = 0; 802 char action2[40], proto[128], fragment[32]; 803 804 fragment[0] = '\0'; 805 proto[0] = '\0'; 806 807 if (f == NULL) { /* bogus pkt */ 808 if (verbose_limit != 0 && norule_counter >= verbose_limit) 809 return; 810 norule_counter++; 811 if (norule_counter == verbose_limit) 812 limit_reached = verbose_limit; 813 action = "Refuse"; 814 } else { /* O_LOG is the first action, find the real one */ 815 ipfw_insn *cmd = ACTION_PTR(f); 816 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 817 818 if (l->max_log != 0 && l->log_left == 0) 819 return; 820 l->log_left--; 821 if (l->log_left == 0) 822 limit_reached = l->max_log; 823 cmd += F_LEN(cmd); /* point to first action */ 824 if (cmd->opcode == O_ALTQ) { 825 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 826 827 snprintf(SNPARGS(action2, 0), "Altq %d", 828 altq->qid); 829 cmd += F_LEN(cmd); 830 } 831 if (cmd->opcode == O_PROB) 832 cmd += F_LEN(cmd); 833 834 if (cmd->opcode == O_TAG) 835 cmd += F_LEN(cmd); 836 837 action = action2; 838 switch (cmd->opcode) { 839 case O_DENY: 840 action = "Deny"; 841 break; 842 843 case O_REJECT: 844 if (cmd->arg1==ICMP_REJECT_RST) 845 action = "Reset"; 846 else if (cmd->arg1==ICMP_UNREACH_HOST) 847 action = "Reject"; 848 else 849 snprintf(SNPARGS(action2, 0), "Unreach %d", 850 cmd->arg1); 851 break; 852 853 case O_UNREACH6: 854 if (cmd->arg1==ICMP6_UNREACH_RST) 855 action = "Reset"; 856 else 857 snprintf(SNPARGS(action2, 0), "Unreach %d", 858 cmd->arg1); 859 break; 860 861 case O_ACCEPT: 862 action = "Accept"; 863 break; 864 case O_COUNT: 865 action = "Count"; 866 break; 867 case O_DIVERT: 868 snprintf(SNPARGS(action2, 0), "Divert %d", 869 cmd->arg1); 870 break; 871 case O_TEE: 872 snprintf(SNPARGS(action2, 0), "Tee %d", 873 cmd->arg1); 874 break; 875 case O_SKIPTO: 876 snprintf(SNPARGS(action2, 0), "SkipTo %d", 877 cmd->arg1); 878 break; 879 case O_PIPE: 880 snprintf(SNPARGS(action2, 0), "Pipe %d", 881 cmd->arg1); 882 break; 883 case O_QUEUE: 884 snprintf(SNPARGS(action2, 0), "Queue %d", 885 cmd->arg1); 886 break; 887 case O_FORWARD_IP: { 888 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 889 int len; 890 struct in_addr dummyaddr; 891 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 892 dummyaddr.s_addr = htonl(tablearg); 893 else 894 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 895 896 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 897 inet_ntoa(dummyaddr)); 898 899 if (sa->sa.sin_port) 900 snprintf(SNPARGS(action2, len), ":%d", 901 sa->sa.sin_port); 902 } 903 break; 904 case O_NETGRAPH: 905 snprintf(SNPARGS(action2, 0), "Netgraph %d", 906 cmd->arg1); 907 break; 908 case O_NGTEE: 909 snprintf(SNPARGS(action2, 0), "Ngtee %d", 910 cmd->arg1); 911 break; 912 default: 913 action = "UNKNOWN"; 914 break; 915 } 916 } 917 918 if (hlen == 0) { /* non-ip */ 919 snprintf(SNPARGS(proto, 0), "MAC"); 920 921 } else { 922 int len; 923 char src[48], dst[48]; 924 struct icmphdr *icmp; 925 struct tcphdr *tcp; 926 struct udphdr *udp; 927 #ifdef INET6 928 struct ip6_hdr *ip6 = NULL; 929 struct icmp6_hdr *icmp6; 930 #endif 931 src[0] = '\0'; 932 dst[0] = '\0'; 933 #ifdef INET6 934 if (IS_IP6_FLOW_ID(&(args->f_id))) { 935 snprintf(src, sizeof(src), "[%s]", 936 ip6_sprintf(&args->f_id.src_ip6)); 937 snprintf(dst, sizeof(dst), "[%s]", 938 ip6_sprintf(&args->f_id.dst_ip6)); 939 940 ip6 = (struct ip6_hdr *)ip; 941 tcp = (struct tcphdr *)(((char *)ip) + hlen); 942 udp = (struct udphdr *)(((char *)ip) + hlen); 943 } else 944 #endif 945 { 946 tcp = L3HDR(struct tcphdr, ip); 947 udp = L3HDR(struct udphdr, ip); 948 949 inet_ntoa_r(ip->ip_src, src); 950 inet_ntoa_r(ip->ip_dst, dst); 951 } 952 953 switch (args->f_id.proto) { 954 case IPPROTO_TCP: 955 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 956 if (offset == 0) 957 snprintf(SNPARGS(proto, len), ":%d %s:%d", 958 ntohs(tcp->th_sport), 959 dst, 960 ntohs(tcp->th_dport)); 961 else 962 snprintf(SNPARGS(proto, len), " %s", dst); 963 break; 964 965 case IPPROTO_UDP: 966 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 967 if (offset == 0) 968 snprintf(SNPARGS(proto, len), ":%d %s:%d", 969 ntohs(udp->uh_sport), 970 dst, 971 ntohs(udp->uh_dport)); 972 else 973 snprintf(SNPARGS(proto, len), " %s", dst); 974 break; 975 976 case IPPROTO_ICMP: 977 icmp = L3HDR(struct icmphdr, ip); 978 if (offset == 0) 979 len = snprintf(SNPARGS(proto, 0), 980 "ICMP:%u.%u ", 981 icmp->icmp_type, icmp->icmp_code); 982 else 983 len = snprintf(SNPARGS(proto, 0), "ICMP "); 984 len += snprintf(SNPARGS(proto, len), "%s", src); 985 snprintf(SNPARGS(proto, len), " %s", dst); 986 break; 987 #ifdef INET6 988 case IPPROTO_ICMPV6: 989 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 990 if (offset == 0) 991 len = snprintf(SNPARGS(proto, 0), 992 "ICMPv6:%u.%u ", 993 icmp6->icmp6_type, icmp6->icmp6_code); 994 else 995 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 996 len += snprintf(SNPARGS(proto, len), "%s", src); 997 snprintf(SNPARGS(proto, len), " %s", dst); 998 break; 999 #endif 1000 default: 1001 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 1002 args->f_id.proto, src); 1003 snprintf(SNPARGS(proto, len), " %s", dst); 1004 break; 1005 } 1006 1007 #ifdef INET6 1008 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1009 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 1010 snprintf(SNPARGS(fragment, 0), 1011 " (frag %08x:%d@%d%s)", 1012 args->f_id.frag_id6, 1013 ntohs(ip6->ip6_plen) - hlen, 1014 ntohs(offset & IP6F_OFF_MASK) << 3, 1015 (offset & IP6F_MORE_FRAG) ? "+" : ""); 1016 } else 1017 #endif 1018 { 1019 int ip_off, ip_len; 1020 if (eh != NULL) { /* layer 2 packets are as on the wire */ 1021 ip_off = ntohs(ip->ip_off); 1022 ip_len = ntohs(ip->ip_len); 1023 } else { 1024 ip_off = ip->ip_off; 1025 ip_len = ip->ip_len; 1026 } 1027 if (ip_off & (IP_MF | IP_OFFMASK)) 1028 snprintf(SNPARGS(fragment, 0), 1029 " (frag %d:%d@%d%s)", 1030 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1031 offset << 3, 1032 (ip_off & IP_MF) ? "+" : ""); 1033 } 1034 } 1035 if (oif || m->m_pkthdr.rcvif) 1036 log(LOG_SECURITY | LOG_INFO, 1037 "ipfw: %d %s %s %s via %s%s\n", 1038 f ? f->rulenum : -1, 1039 action, proto, oif ? "out" : "in", 1040 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1041 fragment); 1042 else 1043 log(LOG_SECURITY | LOG_INFO, 1044 "ipfw: %d %s %s [no if info]%s\n", 1045 f ? f->rulenum : -1, 1046 action, proto, fragment); 1047 if (limit_reached) 1048 log(LOG_SECURITY | LOG_NOTICE, 1049 "ipfw: limit %d reached on entry %d\n", 1050 limit_reached, f ? f->rulenum : -1); 1051 } 1052 1053 /* 1054 * IMPORTANT: the hash function for dynamic rules must be commutative 1055 * in source and destination (ip,port), because rules are bidirectional 1056 * and we want to find both in the same bucket. 1057 */ 1058 static __inline int 1059 hash_packet(struct ipfw_flow_id *id) 1060 { 1061 u_int32_t i; 1062 1063 #ifdef INET6 1064 if (IS_IP6_FLOW_ID(id)) 1065 i = hash_packet6(id); 1066 else 1067 #endif /* INET6 */ 1068 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1069 i &= (curr_dyn_buckets - 1); 1070 return i; 1071 } 1072 1073 /** 1074 * unlink a dynamic rule from a chain. prev is a pointer to 1075 * the previous one, q is a pointer to the rule to delete, 1076 * head is a pointer to the head of the queue. 1077 * Modifies q and potentially also head. 1078 */ 1079 #define UNLINK_DYN_RULE(prev, head, q) { \ 1080 ipfw_dyn_rule *old_q = q; \ 1081 \ 1082 /* remove a refcount to the parent */ \ 1083 if (q->dyn_type == O_LIMIT) \ 1084 q->parent->count--; \ 1085 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1086 (q->id.src_ip), (q->id.src_port), \ 1087 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1088 if (prev != NULL) \ 1089 prev->next = q = q->next; \ 1090 else \ 1091 head = q = q->next; \ 1092 dyn_count--; \ 1093 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1094 1095 #define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1096 1097 /** 1098 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1099 * 1100 * If keep_me == NULL, rules are deleted even if not expired, 1101 * otherwise only expired rules are removed. 1102 * 1103 * The value of the second parameter is also used to point to identify 1104 * a rule we absolutely do not want to remove (e.g. because we are 1105 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1106 * rules). The pointer is only used for comparison, so any non-null 1107 * value will do. 1108 */ 1109 static void 1110 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1111 { 1112 static u_int32_t last_remove = 0; 1113 1114 #define FORCE (keep_me == NULL) 1115 1116 ipfw_dyn_rule *prev, *q; 1117 int i, pass = 0, max_pass = 0; 1118 1119 IPFW_DYN_LOCK_ASSERT(); 1120 1121 if (ipfw_dyn_v == NULL || dyn_count == 0) 1122 return; 1123 /* do not expire more than once per second, it is useless */ 1124 if (!FORCE && last_remove == time_uptime) 1125 return; 1126 last_remove = time_uptime; 1127 1128 /* 1129 * because O_LIMIT refer to parent rules, during the first pass only 1130 * remove child and mark any pending LIMIT_PARENT, and remove 1131 * them in a second pass. 1132 */ 1133 next_pass: 1134 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1135 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1136 /* 1137 * Logic can become complex here, so we split tests. 1138 */ 1139 if (q == keep_me) 1140 goto next; 1141 if (rule != NULL && rule != q->rule) 1142 goto next; /* not the one we are looking for */ 1143 if (q->dyn_type == O_LIMIT_PARENT) { 1144 /* 1145 * handle parent in the second pass, 1146 * record we need one. 1147 */ 1148 max_pass = 1; 1149 if (pass == 0) 1150 goto next; 1151 if (FORCE && q->count != 0 ) { 1152 /* XXX should not happen! */ 1153 printf("ipfw: OUCH! cannot remove rule," 1154 " count %d\n", q->count); 1155 } 1156 } else { 1157 if (!FORCE && 1158 !TIME_LEQ( q->expire, time_uptime )) 1159 goto next; 1160 } 1161 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1162 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1163 continue; 1164 } 1165 next: 1166 prev=q; 1167 q=q->next; 1168 } 1169 } 1170 if (pass++ < max_pass) 1171 goto next_pass; 1172 } 1173 1174 1175 /** 1176 * lookup a dynamic rule. 1177 */ 1178 static ipfw_dyn_rule * 1179 lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1180 struct tcphdr *tcp) 1181 { 1182 /* 1183 * stateful ipfw extensions. 1184 * Lookup into dynamic session queue 1185 */ 1186 #define MATCH_REVERSE 0 1187 #define MATCH_FORWARD 1 1188 #define MATCH_NONE 2 1189 #define MATCH_UNKNOWN 3 1190 int i, dir = MATCH_NONE; 1191 ipfw_dyn_rule *prev, *q=NULL; 1192 1193 IPFW_DYN_LOCK_ASSERT(); 1194 1195 if (ipfw_dyn_v == NULL) 1196 goto done; /* not found */ 1197 i = hash_packet( pkt ); 1198 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1199 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1200 goto next; 1201 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1202 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1203 continue; 1204 } 1205 if (pkt->proto == q->id.proto && 1206 q->dyn_type != O_LIMIT_PARENT) { 1207 if (IS_IP6_FLOW_ID(pkt)) { 1208 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1209 &(q->id.src_ip6)) && 1210 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1211 &(q->id.dst_ip6)) && 1212 pkt->src_port == q->id.src_port && 1213 pkt->dst_port == q->id.dst_port ) { 1214 dir = MATCH_FORWARD; 1215 break; 1216 } 1217 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1218 &(q->id.dst_ip6)) && 1219 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1220 &(q->id.src_ip6)) && 1221 pkt->src_port == q->id.dst_port && 1222 pkt->dst_port == q->id.src_port ) { 1223 dir = MATCH_REVERSE; 1224 break; 1225 } 1226 } else { 1227 if (pkt->src_ip == q->id.src_ip && 1228 pkt->dst_ip == q->id.dst_ip && 1229 pkt->src_port == q->id.src_port && 1230 pkt->dst_port == q->id.dst_port ) { 1231 dir = MATCH_FORWARD; 1232 break; 1233 } 1234 if (pkt->src_ip == q->id.dst_ip && 1235 pkt->dst_ip == q->id.src_ip && 1236 pkt->src_port == q->id.dst_port && 1237 pkt->dst_port == q->id.src_port ) { 1238 dir = MATCH_REVERSE; 1239 break; 1240 } 1241 } 1242 } 1243 next: 1244 prev = q; 1245 q = q->next; 1246 } 1247 if (q == NULL) 1248 goto done; /* q = NULL, not found */ 1249 1250 if ( prev != NULL) { /* found and not in front */ 1251 prev->next = q->next; 1252 q->next = ipfw_dyn_v[i]; 1253 ipfw_dyn_v[i] = q; 1254 } 1255 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1256 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1257 1258 #define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1259 #define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1260 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1261 switch (q->state) { 1262 case TH_SYN: /* opening */ 1263 q->expire = time_uptime + dyn_syn_lifetime; 1264 break; 1265 1266 case BOTH_SYN: /* move to established */ 1267 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1268 case BOTH_SYN | (TH_FIN << 8) : 1269 if (tcp) { 1270 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1271 u_int32_t ack = ntohl(tcp->th_ack); 1272 if (dir == MATCH_FORWARD) { 1273 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1274 q->ack_fwd = ack; 1275 else { /* ignore out-of-sequence */ 1276 break; 1277 } 1278 } else { 1279 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1280 q->ack_rev = ack; 1281 else { /* ignore out-of-sequence */ 1282 break; 1283 } 1284 } 1285 } 1286 q->expire = time_uptime + dyn_ack_lifetime; 1287 break; 1288 1289 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1290 if (dyn_fin_lifetime >= dyn_keepalive_period) 1291 dyn_fin_lifetime = dyn_keepalive_period - 1; 1292 q->expire = time_uptime + dyn_fin_lifetime; 1293 break; 1294 1295 default: 1296 #if 0 1297 /* 1298 * reset or some invalid combination, but can also 1299 * occur if we use keep-state the wrong way. 1300 */ 1301 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1302 printf("invalid state: 0x%x\n", q->state); 1303 #endif 1304 if (dyn_rst_lifetime >= dyn_keepalive_period) 1305 dyn_rst_lifetime = dyn_keepalive_period - 1; 1306 q->expire = time_uptime + dyn_rst_lifetime; 1307 break; 1308 } 1309 } else if (pkt->proto == IPPROTO_UDP) { 1310 q->expire = time_uptime + dyn_udp_lifetime; 1311 } else { 1312 /* other protocols */ 1313 q->expire = time_uptime + dyn_short_lifetime; 1314 } 1315 done: 1316 if (match_direction) 1317 *match_direction = dir; 1318 return q; 1319 } 1320 1321 static ipfw_dyn_rule * 1322 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1323 struct tcphdr *tcp) 1324 { 1325 ipfw_dyn_rule *q; 1326 1327 IPFW_DYN_LOCK(); 1328 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1329 if (q == NULL) 1330 IPFW_DYN_UNLOCK(); 1331 /* NB: return table locked when q is not NULL */ 1332 return q; 1333 } 1334 1335 static void 1336 realloc_dynamic_table(void) 1337 { 1338 IPFW_DYN_LOCK_ASSERT(); 1339 1340 /* 1341 * Try reallocation, make sure we have a power of 2 and do 1342 * not allow more than 64k entries. In case of overflow, 1343 * default to 1024. 1344 */ 1345 1346 if (dyn_buckets > 65536) 1347 dyn_buckets = 1024; 1348 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1349 dyn_buckets = curr_dyn_buckets; /* reset */ 1350 return; 1351 } 1352 curr_dyn_buckets = dyn_buckets; 1353 if (ipfw_dyn_v != NULL) 1354 free(ipfw_dyn_v, M_IPFW); 1355 for (;;) { 1356 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1357 M_IPFW, M_NOWAIT | M_ZERO); 1358 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1359 break; 1360 curr_dyn_buckets /= 2; 1361 } 1362 } 1363 1364 /** 1365 * Install state of type 'type' for a dynamic session. 1366 * The hash table contains two type of rules: 1367 * - regular rules (O_KEEP_STATE) 1368 * - rules for sessions with limited number of sess per user 1369 * (O_LIMIT). When they are created, the parent is 1370 * increased by 1, and decreased on delete. In this case, 1371 * the third parameter is the parent rule and not the chain. 1372 * - "parent" rules for the above (O_LIMIT_PARENT). 1373 */ 1374 static ipfw_dyn_rule * 1375 add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1376 { 1377 ipfw_dyn_rule *r; 1378 int i; 1379 1380 IPFW_DYN_LOCK_ASSERT(); 1381 1382 if (ipfw_dyn_v == NULL || 1383 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1384 realloc_dynamic_table(); 1385 if (ipfw_dyn_v == NULL) 1386 return NULL; /* failed ! */ 1387 } 1388 i = hash_packet(id); 1389 1390 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1391 if (r == NULL) { 1392 printf ("ipfw: sorry cannot allocate state\n"); 1393 return NULL; 1394 } 1395 1396 /* increase refcount on parent, and set pointer */ 1397 if (dyn_type == O_LIMIT) { 1398 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1399 if ( parent->dyn_type != O_LIMIT_PARENT) 1400 panic("invalid parent"); 1401 parent->count++; 1402 r->parent = parent; 1403 rule = parent->rule; 1404 } 1405 1406 r->id = *id; 1407 r->expire = time_uptime + dyn_syn_lifetime; 1408 r->rule = rule; 1409 r->dyn_type = dyn_type; 1410 r->pcnt = r->bcnt = 0; 1411 r->count = 0; 1412 1413 r->bucket = i; 1414 r->next = ipfw_dyn_v[i]; 1415 ipfw_dyn_v[i] = r; 1416 dyn_count++; 1417 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1418 dyn_type, 1419 (r->id.src_ip), (r->id.src_port), 1420 (r->id.dst_ip), (r->id.dst_port), 1421 dyn_count ); ) 1422 return r; 1423 } 1424 1425 /** 1426 * lookup dynamic parent rule using pkt and rule as search keys. 1427 * If the lookup fails, then install one. 1428 */ 1429 static ipfw_dyn_rule * 1430 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1431 { 1432 ipfw_dyn_rule *q; 1433 int i; 1434 1435 IPFW_DYN_LOCK_ASSERT(); 1436 1437 if (ipfw_dyn_v) { 1438 int is_v6 = IS_IP6_FLOW_ID(pkt); 1439 i = hash_packet( pkt ); 1440 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1441 if (q->dyn_type == O_LIMIT_PARENT && 1442 rule== q->rule && 1443 pkt->proto == q->id.proto && 1444 pkt->src_port == q->id.src_port && 1445 pkt->dst_port == q->id.dst_port && 1446 ( 1447 (is_v6 && 1448 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1449 &(q->id.src_ip6)) && 1450 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1451 &(q->id.dst_ip6))) || 1452 (!is_v6 && 1453 pkt->src_ip == q->id.src_ip && 1454 pkt->dst_ip == q->id.dst_ip) 1455 ) 1456 ) { 1457 q->expire = time_uptime + dyn_short_lifetime; 1458 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1459 return q; 1460 } 1461 } 1462 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1463 } 1464 1465 /** 1466 * Install dynamic state for rule type cmd->o.opcode 1467 * 1468 * Returns 1 (failure) if state is not installed because of errors or because 1469 * session limitations are enforced. 1470 */ 1471 static int 1472 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1473 struct ip_fw_args *args, uint32_t tablearg) 1474 { 1475 static int last_log; 1476 ipfw_dyn_rule *q; 1477 struct in_addr da; 1478 char src[48], dst[48]; 1479 1480 src[0] = '\0'; 1481 dst[0] = '\0'; 1482 1483 DEB( 1484 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1485 __func__, cmd->o.opcode, 1486 (args->f_id.src_ip), (args->f_id.src_port), 1487 (args->f_id.dst_ip), (args->f_id.dst_port)); 1488 ) 1489 1490 IPFW_DYN_LOCK(); 1491 1492 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1493 1494 if (q != NULL) { /* should never occur */ 1495 if (last_log != time_uptime) { 1496 last_log = time_uptime; 1497 printf("ipfw: %s: entry already present, done\n", 1498 __func__); 1499 } 1500 IPFW_DYN_UNLOCK(); 1501 return (0); 1502 } 1503 1504 if (dyn_count >= dyn_max) 1505 /* Run out of slots, try to remove any expired rule. */ 1506 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1507 1508 if (dyn_count >= dyn_max) { 1509 if (last_log != time_uptime) { 1510 last_log = time_uptime; 1511 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1512 } 1513 IPFW_DYN_UNLOCK(); 1514 return (1); /* cannot install, notify caller */ 1515 } 1516 1517 switch (cmd->o.opcode) { 1518 case O_KEEP_STATE: /* bidir rule */ 1519 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1520 break; 1521 1522 case O_LIMIT: { /* limit number of sessions */ 1523 struct ipfw_flow_id id; 1524 ipfw_dyn_rule *parent; 1525 uint32_t conn_limit; 1526 uint16_t limit_mask = cmd->limit_mask; 1527 1528 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1529 tablearg : cmd->conn_limit; 1530 1531 DEB( 1532 if (cmd->conn_limit == IP_FW_TABLEARG) 1533 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1534 "(tablearg)\n", __func__, conn_limit); 1535 else 1536 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1537 __func__, conn_limit); 1538 ) 1539 1540 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1541 id.proto = args->f_id.proto; 1542 id.addr_type = args->f_id.addr_type; 1543 1544 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1545 if (limit_mask & DYN_SRC_ADDR) 1546 id.src_ip6 = args->f_id.src_ip6; 1547 if (limit_mask & DYN_DST_ADDR) 1548 id.dst_ip6 = args->f_id.dst_ip6; 1549 } else { 1550 if (limit_mask & DYN_SRC_ADDR) 1551 id.src_ip = args->f_id.src_ip; 1552 if (limit_mask & DYN_DST_ADDR) 1553 id.dst_ip = args->f_id.dst_ip; 1554 } 1555 if (limit_mask & DYN_SRC_PORT) 1556 id.src_port = args->f_id.src_port; 1557 if (limit_mask & DYN_DST_PORT) 1558 id.dst_port = args->f_id.dst_port; 1559 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1560 printf("ipfw: %s: add parent failed\n", __func__); 1561 IPFW_DYN_UNLOCK(); 1562 return (1); 1563 } 1564 1565 if (parent->count >= conn_limit) { 1566 /* See if we can remove some expired rule. */ 1567 remove_dyn_rule(rule, parent); 1568 if (parent->count >= conn_limit) { 1569 if (fw_verbose && last_log != time_uptime) { 1570 last_log = time_uptime; 1571 #ifdef INET6 1572 /* 1573 * XXX IPv6 flows are not 1574 * supported yet. 1575 * */ 1576 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1577 snprintf(src, sizeof(src), 1578 "[%s]", ip6_sprintf( 1579 &args->f_id.src_ip6)); 1580 snprintf(dst, sizeof(dst), 1581 "[%s]", ip6_sprintf( 1582 &args->f_id.dst_ip6)); 1583 } else 1584 #endif 1585 { 1586 da.s_addr = 1587 htonl(args->f_id.src_ip); 1588 inet_ntoa_r(da, src); 1589 da.s_addr = 1590 htonl(args->f_id.dst_ip); 1591 inet_ntoa_r(da, dst); 1592 } 1593 log(LOG_SECURITY | LOG_DEBUG, 1594 "%s %s:%u -> %s:%u, %s\n", 1595 "drop session", 1596 src, (args->f_id.src_port), 1597 dst, (args->f_id.dst_port), 1598 "too many entries"); 1599 } 1600 IPFW_DYN_UNLOCK(); 1601 return (1); 1602 } 1603 } 1604 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1605 break; 1606 } 1607 default: 1608 printf("ipfw: %s: unknown dynamic rule type %u\n", 1609 __func__, cmd->o.opcode); 1610 IPFW_DYN_UNLOCK(); 1611 return (1); 1612 } 1613 1614 /* XXX just set lifetime */ 1615 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1616 1617 IPFW_DYN_UNLOCK(); 1618 return (0); 1619 } 1620 1621 /* 1622 * Generate a TCP packet, containing either a RST or a keepalive. 1623 * When flags & TH_RST, we are sending a RST packet, because of a 1624 * "reset" action matched the packet. 1625 * Otherwise we are sending a keepalive, and flags & TH_ 1626 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1627 * so that MAC can label the reply appropriately. 1628 */ 1629 static struct mbuf * 1630 send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1631 u_int32_t ack, int flags) 1632 { 1633 struct mbuf *m; 1634 struct ip *ip; 1635 struct tcphdr *tcp; 1636 1637 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1638 if (m == 0) 1639 return (NULL); 1640 m->m_pkthdr.rcvif = (struct ifnet *)0; 1641 1642 #ifdef MAC 1643 if (replyto != NULL) 1644 mac_create_mbuf_netlayer(replyto, m); 1645 else 1646 mac_create_mbuf_from_firewall(m); 1647 #else 1648 (void)replyto; /* don't warn about unused arg */ 1649 #endif 1650 1651 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1652 m->m_data += max_linkhdr; 1653 1654 ip = mtod(m, struct ip *); 1655 bzero(ip, m->m_len); 1656 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1657 ip->ip_p = IPPROTO_TCP; 1658 tcp->th_off = 5; 1659 /* 1660 * Assume we are sending a RST (or a keepalive in the reverse 1661 * direction), swap src and destination addresses and ports. 1662 */ 1663 ip->ip_src.s_addr = htonl(id->dst_ip); 1664 ip->ip_dst.s_addr = htonl(id->src_ip); 1665 tcp->th_sport = htons(id->dst_port); 1666 tcp->th_dport = htons(id->src_port); 1667 if (flags & TH_RST) { /* we are sending a RST */ 1668 if (flags & TH_ACK) { 1669 tcp->th_seq = htonl(ack); 1670 tcp->th_ack = htonl(0); 1671 tcp->th_flags = TH_RST; 1672 } else { 1673 if (flags & TH_SYN) 1674 seq++; 1675 tcp->th_seq = htonl(0); 1676 tcp->th_ack = htonl(seq); 1677 tcp->th_flags = TH_RST | TH_ACK; 1678 } 1679 } else { 1680 /* 1681 * We are sending a keepalive. flags & TH_SYN determines 1682 * the direction, forward if set, reverse if clear. 1683 * NOTE: seq and ack are always assumed to be correct 1684 * as set by the caller. This may be confusing... 1685 */ 1686 if (flags & TH_SYN) { 1687 /* 1688 * we have to rewrite the correct addresses! 1689 */ 1690 ip->ip_dst.s_addr = htonl(id->dst_ip); 1691 ip->ip_src.s_addr = htonl(id->src_ip); 1692 tcp->th_dport = htons(id->dst_port); 1693 tcp->th_sport = htons(id->src_port); 1694 } 1695 tcp->th_seq = htonl(seq); 1696 tcp->th_ack = htonl(ack); 1697 tcp->th_flags = TH_ACK; 1698 } 1699 /* 1700 * set ip_len to the payload size so we can compute 1701 * the tcp checksum on the pseudoheader 1702 * XXX check this, could save a couple of words ? 1703 */ 1704 ip->ip_len = htons(sizeof(struct tcphdr)); 1705 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1706 /* 1707 * now fill fields left out earlier 1708 */ 1709 ip->ip_ttl = ip_defttl; 1710 ip->ip_len = m->m_pkthdr.len; 1711 m->m_flags |= M_SKIP_FIREWALL; 1712 return (m); 1713 } 1714 1715 /* 1716 * sends a reject message, consuming the mbuf passed as an argument. 1717 */ 1718 static void 1719 send_reject(struct ip_fw_args *args, int code, u_short offset, int ip_len, struct ip *ip) 1720 { 1721 1722 #if 0 1723 /* XXX When ip is not guaranteed to be at mtod() we will 1724 * need to account for this */ 1725 * The mbuf will however be thrown away so we can adjust it. 1726 * Remember we did an m_pullup on it already so we 1727 * can make some assumptions about contiguousness. 1728 */ 1729 if (args->L3offset) 1730 m_adj(m, args->L3offset); 1731 #endif 1732 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1733 /* We need the IP header in host order for icmp_error(). */ 1734 if (args->eh != NULL) { 1735 ip->ip_len = ntohs(ip->ip_len); 1736 ip->ip_off = ntohs(ip->ip_off); 1737 } 1738 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1739 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1740 struct tcphdr *const tcp = 1741 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1742 if ( (tcp->th_flags & TH_RST) == 0) { 1743 struct mbuf *m; 1744 m = send_pkt(args->m, &(args->f_id), 1745 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1746 tcp->th_flags | TH_RST); 1747 if (m != NULL) 1748 ip_output(m, NULL, NULL, 0, NULL, NULL); 1749 } 1750 m_freem(args->m); 1751 } else 1752 m_freem(args->m); 1753 args->m = NULL; 1754 } 1755 1756 /** 1757 * 1758 * Given an ip_fw *, lookup_next_rule will return a pointer 1759 * to the next rule, which can be either the jump 1760 * target (for skipto instructions) or the next one in the list (in 1761 * all other cases including a missing jump target). 1762 * The result is also written in the "next_rule" field of the rule. 1763 * Backward jumps are not allowed, so start looking from the next 1764 * rule... 1765 * 1766 * This never returns NULL -- in case we do not have an exact match, 1767 * the next rule is returned. When the ruleset is changed, 1768 * pointers are flushed so we are always correct. 1769 */ 1770 1771 static struct ip_fw * 1772 lookup_next_rule(struct ip_fw *me) 1773 { 1774 struct ip_fw *rule = NULL; 1775 ipfw_insn *cmd; 1776 1777 /* look for action, in case it is a skipto */ 1778 cmd = ACTION_PTR(me); 1779 if (cmd->opcode == O_LOG) 1780 cmd += F_LEN(cmd); 1781 if (cmd->opcode == O_ALTQ) 1782 cmd += F_LEN(cmd); 1783 if (cmd->opcode == O_TAG) 1784 cmd += F_LEN(cmd); 1785 if ( cmd->opcode == O_SKIPTO ) 1786 for (rule = me->next; rule ; rule = rule->next) 1787 if (rule->rulenum >= cmd->arg1) 1788 break; 1789 if (rule == NULL) /* failure or not a skipto */ 1790 rule = me->next; 1791 me->next_rule = rule; 1792 return rule; 1793 } 1794 1795 static int 1796 add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1797 uint8_t mlen, uint32_t value) 1798 { 1799 struct radix_node_head *rnh; 1800 struct table_entry *ent; 1801 1802 if (tbl >= IPFW_TABLES_MAX) 1803 return (EINVAL); 1804 rnh = ch->tables[tbl]; 1805 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1806 if (ent == NULL) 1807 return (ENOMEM); 1808 ent->value = value; 1809 ent->addr.sin_len = ent->mask.sin_len = 8; 1810 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1811 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1812 IPFW_WLOCK(&layer3_chain); 1813 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1814 NULL) { 1815 IPFW_WUNLOCK(&layer3_chain); 1816 free(ent, M_IPFW_TBL); 1817 return (EEXIST); 1818 } 1819 IPFW_WUNLOCK(&layer3_chain); 1820 return (0); 1821 } 1822 1823 static int 1824 del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1825 uint8_t mlen) 1826 { 1827 struct radix_node_head *rnh; 1828 struct table_entry *ent; 1829 struct sockaddr_in sa, mask; 1830 1831 if (tbl >= IPFW_TABLES_MAX) 1832 return (EINVAL); 1833 rnh = ch->tables[tbl]; 1834 sa.sin_len = mask.sin_len = 8; 1835 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1836 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1837 IPFW_WLOCK(ch); 1838 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1839 if (ent == NULL) { 1840 IPFW_WUNLOCK(ch); 1841 return (ESRCH); 1842 } 1843 IPFW_WUNLOCK(ch); 1844 free(ent, M_IPFW_TBL); 1845 return (0); 1846 } 1847 1848 static int 1849 flush_table_entry(struct radix_node *rn, void *arg) 1850 { 1851 struct radix_node_head * const rnh = arg; 1852 struct table_entry *ent; 1853 1854 ent = (struct table_entry *) 1855 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1856 if (ent != NULL) 1857 free(ent, M_IPFW_TBL); 1858 return (0); 1859 } 1860 1861 static int 1862 flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1863 { 1864 struct radix_node_head *rnh; 1865 1866 IPFW_WLOCK_ASSERT(ch); 1867 1868 if (tbl >= IPFW_TABLES_MAX) 1869 return (EINVAL); 1870 rnh = ch->tables[tbl]; 1871 KASSERT(rnh != NULL, ("NULL IPFW table")); 1872 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1873 return (0); 1874 } 1875 1876 static void 1877 flush_tables(struct ip_fw_chain *ch) 1878 { 1879 uint16_t tbl; 1880 1881 IPFW_WLOCK_ASSERT(ch); 1882 1883 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1884 flush_table(ch, tbl); 1885 } 1886 1887 static int 1888 init_tables(struct ip_fw_chain *ch) 1889 { 1890 int i; 1891 uint16_t j; 1892 1893 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1894 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1895 for (j = 0; j < i; j++) { 1896 (void) flush_table(ch, j); 1897 } 1898 return (ENOMEM); 1899 } 1900 } 1901 return (0); 1902 } 1903 1904 static int 1905 lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1906 uint32_t *val) 1907 { 1908 struct radix_node_head *rnh; 1909 struct table_entry *ent; 1910 struct sockaddr_in sa; 1911 1912 if (tbl >= IPFW_TABLES_MAX) 1913 return (0); 1914 rnh = ch->tables[tbl]; 1915 sa.sin_len = 8; 1916 sa.sin_addr.s_addr = addr; 1917 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1918 if (ent != NULL) { 1919 *val = ent->value; 1920 return (1); 1921 } 1922 return (0); 1923 } 1924 1925 static int 1926 count_table_entry(struct radix_node *rn, void *arg) 1927 { 1928 u_int32_t * const cnt = arg; 1929 1930 (*cnt)++; 1931 return (0); 1932 } 1933 1934 static int 1935 count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1936 { 1937 struct radix_node_head *rnh; 1938 1939 if (tbl >= IPFW_TABLES_MAX) 1940 return (EINVAL); 1941 rnh = ch->tables[tbl]; 1942 *cnt = 0; 1943 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1944 return (0); 1945 } 1946 1947 static int 1948 dump_table_entry(struct radix_node *rn, void *arg) 1949 { 1950 struct table_entry * const n = (struct table_entry *)rn; 1951 ipfw_table * const tbl = arg; 1952 ipfw_table_entry *ent; 1953 1954 if (tbl->cnt == tbl->size) 1955 return (1); 1956 ent = &tbl->ent[tbl->cnt]; 1957 ent->tbl = tbl->tbl; 1958 if (in_nullhost(n->mask.sin_addr)) 1959 ent->masklen = 0; 1960 else 1961 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1962 ent->addr = n->addr.sin_addr.s_addr; 1963 ent->value = n->value; 1964 tbl->cnt++; 1965 return (0); 1966 } 1967 1968 static int 1969 dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1970 { 1971 struct radix_node_head *rnh; 1972 1973 if (tbl->tbl >= IPFW_TABLES_MAX) 1974 return (EINVAL); 1975 rnh = ch->tables[tbl->tbl]; 1976 tbl->cnt = 0; 1977 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1978 return (0); 1979 } 1980 1981 static void 1982 fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1983 { 1984 struct ucred *cr; 1985 1986 if (inp->inp_socket != NULL) { 1987 cr = inp->inp_socket->so_cred; 1988 ugp->fw_prid = jailed(cr) ? 1989 cr->cr_prison->pr_id : -1; 1990 ugp->fw_uid = cr->cr_uid; 1991 ugp->fw_ngroups = cr->cr_ngroups; 1992 bcopy(cr->cr_groups, ugp->fw_groups, 1993 sizeof(ugp->fw_groups)); 1994 } 1995 } 1996 1997 static int 1998 check_uidgid(ipfw_insn_u32 *insn, 1999 int proto, struct ifnet *oif, 2000 struct in_addr dst_ip, u_int16_t dst_port, 2001 struct in_addr src_ip, u_int16_t src_port, 2002 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 2003 { 2004 struct inpcbinfo *pi; 2005 int wildcard; 2006 struct inpcb *pcb; 2007 int match; 2008 gid_t *gp; 2009 2010 /* 2011 * Check to see if the UDP or TCP stack supplied us with 2012 * the PCB. If so, rather then holding a lock and looking 2013 * up the PCB, we can use the one that was supplied. 2014 */ 2015 if (inp && *lookup == 0) { 2016 INP_LOCK_ASSERT(inp); 2017 if (inp->inp_socket != NULL) { 2018 fill_ugid_cache(inp, ugp); 2019 *lookup = 1; 2020 } 2021 } 2022 /* 2023 * If we have already been here and the packet has no 2024 * PCB entry associated with it, then we can safely 2025 * assume that this is a no match. 2026 */ 2027 if (*lookup == -1) 2028 return (0); 2029 if (proto == IPPROTO_TCP) { 2030 wildcard = 0; 2031 pi = &tcbinfo; 2032 } else if (proto == IPPROTO_UDP) { 2033 wildcard = INPLOOKUP_WILDCARD; 2034 pi = &udbinfo; 2035 } else 2036 return 0; 2037 match = 0; 2038 if (*lookup == 0) { 2039 INP_INFO_RLOCK(pi); 2040 pcb = (oif) ? 2041 in_pcblookup_hash(pi, 2042 dst_ip, htons(dst_port), 2043 src_ip, htons(src_port), 2044 wildcard, oif) : 2045 in_pcblookup_hash(pi, 2046 src_ip, htons(src_port), 2047 dst_ip, htons(dst_port), 2048 wildcard, NULL); 2049 if (pcb != NULL) { 2050 INP_LOCK(pcb); 2051 if (pcb->inp_socket != NULL) { 2052 fill_ugid_cache(pcb, ugp); 2053 *lookup = 1; 2054 } 2055 INP_UNLOCK(pcb); 2056 } 2057 INP_INFO_RUNLOCK(pi); 2058 if (*lookup == 0) { 2059 /* 2060 * If the lookup did not yield any results, there 2061 * is no sense in coming back and trying again. So 2062 * we can set lookup to -1 and ensure that we wont 2063 * bother the pcb system again. 2064 */ 2065 *lookup = -1; 2066 return (0); 2067 } 2068 } 2069 if (insn->o.opcode == O_UID) 2070 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2071 else if (insn->o.opcode == O_GID) { 2072 for (gp = ugp->fw_groups; 2073 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2074 if (*gp == (gid_t)insn->d[0]) { 2075 match = 1; 2076 break; 2077 } 2078 } else if (insn->o.opcode == O_JAIL) 2079 match = (ugp->fw_prid == (int)insn->d[0]); 2080 return match; 2081 } 2082 2083 /* 2084 * The main check routine for the firewall. 2085 * 2086 * All arguments are in args so we can modify them and return them 2087 * back to the caller. 2088 * 2089 * Parameters: 2090 * 2091 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2092 * Starts with the IP header. 2093 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2094 * args->L3offset Number of bytes bypassed if we came from L2. 2095 * e.g. often sizeof(eh) ** NOTYET ** 2096 * args->oif Outgoing interface, or NULL if packet is incoming. 2097 * The incoming interface is in the mbuf. (in) 2098 * args->divert_rule (in/out) 2099 * Skip up to the first rule past this rule number; 2100 * upon return, non-zero port number for divert or tee. 2101 * 2102 * args->rule Pointer to the last matching rule (in/out) 2103 * args->next_hop Socket we are forwarding to (out). 2104 * args->f_id Addresses grabbed from the packet (out) 2105 * args->cookie a cookie depending on rule action 2106 * 2107 * Return value: 2108 * 2109 * IP_FW_PASS the packet must be accepted 2110 * IP_FW_DENY the packet must be dropped 2111 * IP_FW_DIVERT divert packet, port in m_tag 2112 * IP_FW_TEE tee packet, port in m_tag 2113 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2114 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2115 * 2116 */ 2117 int 2118 ipfw_chk(struct ip_fw_args *args) 2119 { 2120 /* 2121 * Local variables holding state during the processing of a packet: 2122 * 2123 * IMPORTANT NOTE: to speed up the processing of rules, there 2124 * are some assumption on the values of the variables, which 2125 * are documented here. Should you change them, please check 2126 * the implementation of the various instructions to make sure 2127 * that they still work. 2128 * 2129 * args->eh The MAC header. It is non-null for a layer2 2130 * packet, it is NULL for a layer-3 packet. 2131 * **notyet** 2132 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2133 * 2134 * m | args->m Pointer to the mbuf, as received from the caller. 2135 * It may change if ipfw_chk() does an m_pullup, or if it 2136 * consumes the packet because it calls send_reject(). 2137 * XXX This has to change, so that ipfw_chk() never modifies 2138 * or consumes the buffer. 2139 * ip is the beginning of the ip(4 or 6) header. 2140 * Calculated by adding the L3offset to the start of data. 2141 * (Until we start using L3offset, the packet is 2142 * supposed to start with the ip header). 2143 */ 2144 struct mbuf *m = args->m; 2145 struct ip *ip = mtod(m, struct ip *); 2146 2147 /* 2148 * For rules which contain uid/gid or jail constraints, cache 2149 * a copy of the users credentials after the pcb lookup has been 2150 * executed. This will speed up the processing of rules with 2151 * these types of constraints, as well as decrease contention 2152 * on pcb related locks. 2153 */ 2154 struct ip_fw_ugid fw_ugid_cache; 2155 int ugid_lookup = 0; 2156 2157 /* 2158 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2159 * associated with a packet input on a divert socket. This 2160 * will allow to distinguish traffic and its direction when 2161 * it originates from a divert socket. 2162 */ 2163 u_int divinput_flags = 0; 2164 2165 /* 2166 * oif | args->oif If NULL, ipfw_chk has been called on the 2167 * inbound path (ether_input, bdg_forward, ip_input). 2168 * If non-NULL, ipfw_chk has been called on the outbound path 2169 * (ether_output, ip_output). 2170 */ 2171 struct ifnet *oif = args->oif; 2172 2173 struct ip_fw *f = NULL; /* matching rule */ 2174 int retval = 0; 2175 2176 /* 2177 * hlen The length of the IP header. 2178 */ 2179 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2180 2181 /* 2182 * offset The offset of a fragment. offset != 0 means that 2183 * we have a fragment at this offset of an IPv4 packet. 2184 * offset == 0 means that (if this is an IPv4 packet) 2185 * this is the first or only fragment. 2186 * For IPv6 offset == 0 means there is no Fragment Header. 2187 * If offset != 0 for IPv6 always use correct mask to 2188 * get the correct offset because we add IP6F_MORE_FRAG 2189 * to be able to dectect the first fragment which would 2190 * otherwise have offset = 0. 2191 */ 2192 u_short offset = 0; 2193 2194 /* 2195 * Local copies of addresses. They are only valid if we have 2196 * an IP packet. 2197 * 2198 * proto The protocol. Set to 0 for non-ip packets, 2199 * or to the protocol read from the packet otherwise. 2200 * proto != 0 means that we have an IPv4 packet. 2201 * 2202 * src_port, dst_port port numbers, in HOST format. Only 2203 * valid for TCP and UDP packets. 2204 * 2205 * src_ip, dst_ip ip addresses, in NETWORK format. 2206 * Only valid for IPv4 packets. 2207 */ 2208 u_int8_t proto; 2209 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2210 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2211 u_int16_t ip_len=0; 2212 int pktlen; 2213 u_int16_t etype = 0; /* Host order stored ether type */ 2214 2215 /* 2216 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2217 * MATCH_NONE when checked and not matched (q = NULL), 2218 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2219 */ 2220 int dyn_dir = MATCH_UNKNOWN; 2221 ipfw_dyn_rule *q = NULL; 2222 struct ip_fw_chain *chain = &layer3_chain; 2223 struct m_tag *mtag; 2224 2225 /* 2226 * We store in ulp a pointer to the upper layer protocol header. 2227 * In the ipv4 case this is easy to determine from the header, 2228 * but for ipv6 we might have some additional headers in the middle. 2229 * ulp is NULL if not found. 2230 */ 2231 void *ulp = NULL; /* upper layer protocol pointer. */ 2232 /* XXX ipv6 variables */ 2233 int is_ipv6 = 0; 2234 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2235 /* end of ipv6 variables */ 2236 int is_ipv4 = 0; 2237 2238 if (m->m_flags & M_SKIP_FIREWALL) 2239 return (IP_FW_PASS); /* accept */ 2240 2241 pktlen = m->m_pkthdr.len; 2242 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2243 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2244 2245 /* 2246 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2247 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2248 * pointer might become stale after other pullups (but we never use it 2249 * this way). 2250 */ 2251 #define PULLUP_TO(len, p, T) \ 2252 do { \ 2253 int x = (len) + sizeof(T); \ 2254 if ((m)->m_len < x) { \ 2255 args->m = m = m_pullup(m, x); \ 2256 if (m == NULL) \ 2257 goto pullup_failed; \ 2258 } \ 2259 p = (mtod(m, char *) + (len)); \ 2260 } while (0) 2261 2262 /* 2263 * if we have an ether header, 2264 */ 2265 if (args->eh) 2266 etype = ntohs(args->eh->ether_type); 2267 2268 /* Identify IP packets and fill up variables. */ 2269 if (pktlen >= sizeof(struct ip6_hdr) && 2270 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2271 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2272 is_ipv6 = 1; 2273 args->f_id.addr_type = 6; 2274 hlen = sizeof(struct ip6_hdr); 2275 proto = ip6->ip6_nxt; 2276 2277 /* Search extension headers to find upper layer protocols */ 2278 while (ulp == NULL) { 2279 switch (proto) { 2280 case IPPROTO_ICMPV6: 2281 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2282 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2283 break; 2284 2285 case IPPROTO_TCP: 2286 PULLUP_TO(hlen, ulp, struct tcphdr); 2287 dst_port = TCP(ulp)->th_dport; 2288 src_port = TCP(ulp)->th_sport; 2289 args->f_id.flags = TCP(ulp)->th_flags; 2290 break; 2291 2292 case IPPROTO_UDP: 2293 PULLUP_TO(hlen, ulp, struct udphdr); 2294 dst_port = UDP(ulp)->uh_dport; 2295 src_port = UDP(ulp)->uh_sport; 2296 break; 2297 2298 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2299 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2300 ext_hd |= EXT_HOPOPTS; 2301 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2302 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2303 ulp = NULL; 2304 break; 2305 2306 case IPPROTO_ROUTING: /* RFC 2460 */ 2307 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2308 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2309 case 0: 2310 ext_hd |= EXT_RTHDR0; 2311 break; 2312 case 2: 2313 ext_hd |= EXT_RTHDR2; 2314 break; 2315 default: 2316 printf("IPFW2: IPV6 - Unknown Routing " 2317 "Header type(%d)\n", 2318 ((struct ip6_rthdr *)ulp)->ip6r_type); 2319 if (fw_deny_unknown_exthdrs) 2320 return (IP_FW_DENY); 2321 break; 2322 } 2323 ext_hd |= EXT_ROUTING; 2324 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2325 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2326 ulp = NULL; 2327 break; 2328 2329 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2330 PULLUP_TO(hlen, ulp, struct ip6_frag); 2331 ext_hd |= EXT_FRAGMENT; 2332 hlen += sizeof (struct ip6_frag); 2333 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2334 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2335 IP6F_OFF_MASK; 2336 /* Add IP6F_MORE_FRAG for offset of first 2337 * fragment to be != 0. */ 2338 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2339 IP6F_MORE_FRAG; 2340 if (offset == 0) { 2341 printf("IPFW2: IPV6 - Invalid Fragment " 2342 "Header\n"); 2343 if (fw_deny_unknown_exthdrs) 2344 return (IP_FW_DENY); 2345 break; 2346 } 2347 args->f_id.frag_id6 = 2348 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2349 ulp = NULL; 2350 break; 2351 2352 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2353 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2354 ext_hd |= EXT_DSTOPTS; 2355 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2356 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2357 ulp = NULL; 2358 break; 2359 2360 case IPPROTO_AH: /* RFC 2402 */ 2361 PULLUP_TO(hlen, ulp, struct ip6_ext); 2362 ext_hd |= EXT_AH; 2363 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2364 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2365 ulp = NULL; 2366 break; 2367 2368 case IPPROTO_ESP: /* RFC 2406 */ 2369 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2370 /* Anything past Seq# is variable length and 2371 * data past this ext. header is encrypted. */ 2372 ext_hd |= EXT_ESP; 2373 break; 2374 2375 case IPPROTO_NONE: /* RFC 2460 */ 2376 PULLUP_TO(hlen, ulp, struct ip6_ext); 2377 /* Packet ends here. if ip6e_len!=0 octets 2378 * must be ignored. */ 2379 break; 2380 2381 case IPPROTO_OSPFIGP: 2382 /* XXX OSPF header check? */ 2383 PULLUP_TO(hlen, ulp, struct ip6_ext); 2384 break; 2385 2386 case IPPROTO_PIM: 2387 /* XXX PIM header check? */ 2388 PULLUP_TO(hlen, ulp, struct pim); 2389 break; 2390 2391 case IPPROTO_CARP: 2392 PULLUP_TO(hlen, ulp, struct carp_header); 2393 if (((struct carp_header *)ulp)->carp_version != 2394 CARP_VERSION) 2395 return (IP_FW_DENY); 2396 if (((struct carp_header *)ulp)->carp_type != 2397 CARP_ADVERTISEMENT) 2398 return (IP_FW_DENY); 2399 break; 2400 2401 case IPPROTO_IPV6: /* RFC 2893 */ 2402 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2403 break; 2404 2405 case IPPROTO_IPV4: /* RFC 2893 */ 2406 PULLUP_TO(hlen, ulp, struct ip); 2407 break; 2408 2409 default: 2410 printf("IPFW2: IPV6 - Unknown Extension " 2411 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2412 if (fw_deny_unknown_exthdrs) 2413 return (IP_FW_DENY); 2414 PULLUP_TO(hlen, ulp, struct ip6_ext); 2415 break; 2416 } /*switch */ 2417 } 2418 ip = mtod(m, struct ip *); 2419 ip6 = (struct ip6_hdr *)ip; 2420 args->f_id.src_ip6 = ip6->ip6_src; 2421 args->f_id.dst_ip6 = ip6->ip6_dst; 2422 args->f_id.src_ip = 0; 2423 args->f_id.dst_ip = 0; 2424 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2425 } else if (pktlen >= sizeof(struct ip) && 2426 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2427 is_ipv4 = 1; 2428 hlen = ip->ip_hl << 2; 2429 args->f_id.addr_type = 4; 2430 2431 /* 2432 * Collect parameters into local variables for faster matching. 2433 */ 2434 proto = ip->ip_p; 2435 src_ip = ip->ip_src; 2436 dst_ip = ip->ip_dst; 2437 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2438 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2439 ip_len = ntohs(ip->ip_len); 2440 } else { 2441 offset = ip->ip_off & IP_OFFMASK; 2442 ip_len = ip->ip_len; 2443 } 2444 pktlen = ip_len < pktlen ? ip_len : pktlen; 2445 2446 if (offset == 0) { 2447 switch (proto) { 2448 case IPPROTO_TCP: 2449 PULLUP_TO(hlen, ulp, struct tcphdr); 2450 dst_port = TCP(ulp)->th_dport; 2451 src_port = TCP(ulp)->th_sport; 2452 args->f_id.flags = TCP(ulp)->th_flags; 2453 break; 2454 2455 case IPPROTO_UDP: 2456 PULLUP_TO(hlen, ulp, struct udphdr); 2457 dst_port = UDP(ulp)->uh_dport; 2458 src_port = UDP(ulp)->uh_sport; 2459 break; 2460 2461 case IPPROTO_ICMP: 2462 PULLUP_TO(hlen, ulp, struct icmphdr); 2463 args->f_id.flags = ICMP(ulp)->icmp_type; 2464 break; 2465 2466 default: 2467 break; 2468 } 2469 } 2470 2471 ip = mtod(m, struct ip *); 2472 args->f_id.src_ip = ntohl(src_ip.s_addr); 2473 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2474 } 2475 #undef PULLUP_TO 2476 if (proto) { /* we may have port numbers, store them */ 2477 args->f_id.proto = proto; 2478 args->f_id.src_port = src_port = ntohs(src_port); 2479 args->f_id.dst_port = dst_port = ntohs(dst_port); 2480 } 2481 2482 IPFW_RLOCK(chain); 2483 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2484 if (args->rule) { 2485 /* 2486 * Packet has already been tagged. Look for the next rule 2487 * to restart processing. 2488 * 2489 * If fw_one_pass != 0 then just accept it. 2490 * XXX should not happen here, but optimized out in 2491 * the caller. 2492 */ 2493 if (fw_one_pass) { 2494 IPFW_RUNLOCK(chain); 2495 return (IP_FW_PASS); 2496 } 2497 2498 f = args->rule->next_rule; 2499 if (f == NULL) 2500 f = lookup_next_rule(args->rule); 2501 } else { 2502 /* 2503 * Find the starting rule. It can be either the first 2504 * one, or the one after divert_rule if asked so. 2505 */ 2506 int skipto = mtag ? divert_cookie(mtag) : 0; 2507 2508 f = chain->rules; 2509 if (args->eh == NULL && skipto != 0) { 2510 if (skipto >= IPFW_DEFAULT_RULE) { 2511 IPFW_RUNLOCK(chain); 2512 return (IP_FW_DENY); /* invalid */ 2513 } 2514 while (f && f->rulenum <= skipto) 2515 f = f->next; 2516 if (f == NULL) { /* drop packet */ 2517 IPFW_RUNLOCK(chain); 2518 return (IP_FW_DENY); 2519 } 2520 } 2521 } 2522 /* reset divert rule to avoid confusion later */ 2523 if (mtag) { 2524 divinput_flags = divert_info(mtag) & 2525 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2526 m_tag_delete(m, mtag); 2527 } 2528 2529 /* 2530 * Now scan the rules, and parse microinstructions for each rule. 2531 */ 2532 for (; f; f = f->next) { 2533 ipfw_insn *cmd; 2534 uint32_t tablearg = 0; 2535 int l, cmdlen, skip_or; /* skip rest of OR block */ 2536 2537 again: 2538 if (set_disable & (1 << f->set) ) 2539 continue; 2540 2541 skip_or = 0; 2542 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2543 l -= cmdlen, cmd += cmdlen) { 2544 int match; 2545 2546 /* 2547 * check_body is a jump target used when we find a 2548 * CHECK_STATE, and need to jump to the body of 2549 * the target rule. 2550 */ 2551 2552 check_body: 2553 cmdlen = F_LEN(cmd); 2554 /* 2555 * An OR block (insn_1 || .. || insn_n) has the 2556 * F_OR bit set in all but the last instruction. 2557 * The first match will set "skip_or", and cause 2558 * the following instructions to be skipped until 2559 * past the one with the F_OR bit clear. 2560 */ 2561 if (skip_or) { /* skip this instruction */ 2562 if ((cmd->len & F_OR) == 0) 2563 skip_or = 0; /* next one is good */ 2564 continue; 2565 } 2566 match = 0; /* set to 1 if we succeed */ 2567 2568 switch (cmd->opcode) { 2569 /* 2570 * The first set of opcodes compares the packet's 2571 * fields with some pattern, setting 'match' if a 2572 * match is found. At the end of the loop there is 2573 * logic to deal with F_NOT and F_OR flags associated 2574 * with the opcode. 2575 */ 2576 case O_NOP: 2577 match = 1; 2578 break; 2579 2580 case O_FORWARD_MAC: 2581 printf("ipfw: opcode %d unimplemented\n", 2582 cmd->opcode); 2583 break; 2584 2585 case O_GID: 2586 case O_UID: 2587 case O_JAIL: 2588 /* 2589 * We only check offset == 0 && proto != 0, 2590 * as this ensures that we have a 2591 * packet with the ports info. 2592 */ 2593 if (offset!=0) 2594 break; 2595 if (is_ipv6) /* XXX to be fixed later */ 2596 break; 2597 if (proto == IPPROTO_TCP || 2598 proto == IPPROTO_UDP) 2599 match = check_uidgid( 2600 (ipfw_insn_u32 *)cmd, 2601 proto, oif, 2602 dst_ip, dst_port, 2603 src_ip, src_port, &fw_ugid_cache, 2604 &ugid_lookup, args->inp); 2605 break; 2606 2607 case O_RECV: 2608 match = iface_match(m->m_pkthdr.rcvif, 2609 (ipfw_insn_if *)cmd); 2610 break; 2611 2612 case O_XMIT: 2613 match = iface_match(oif, (ipfw_insn_if *)cmd); 2614 break; 2615 2616 case O_VIA: 2617 match = iface_match(oif ? oif : 2618 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2619 break; 2620 2621 case O_MACADDR2: 2622 if (args->eh != NULL) { /* have MAC header */ 2623 u_int32_t *want = (u_int32_t *) 2624 ((ipfw_insn_mac *)cmd)->addr; 2625 u_int32_t *mask = (u_int32_t *) 2626 ((ipfw_insn_mac *)cmd)->mask; 2627 u_int32_t *hdr = (u_int32_t *)args->eh; 2628 2629 match = 2630 ( want[0] == (hdr[0] & mask[0]) && 2631 want[1] == (hdr[1] & mask[1]) && 2632 want[2] == (hdr[2] & mask[2]) ); 2633 } 2634 break; 2635 2636 case O_MAC_TYPE: 2637 if (args->eh != NULL) { 2638 u_int16_t *p = 2639 ((ipfw_insn_u16 *)cmd)->ports; 2640 int i; 2641 2642 for (i = cmdlen - 1; !match && i>0; 2643 i--, p += 2) 2644 match = (etype >= p[0] && 2645 etype <= p[1]); 2646 } 2647 break; 2648 2649 case O_FRAG: 2650 match = (offset != 0); 2651 break; 2652 2653 case O_IN: /* "out" is "not in" */ 2654 match = (oif == NULL); 2655 break; 2656 2657 case O_LAYER2: 2658 match = (args->eh != NULL); 2659 break; 2660 2661 case O_DIVERTED: 2662 match = (cmd->arg1 & 1 && divinput_flags & 2663 IP_FW_DIVERT_LOOPBACK_FLAG) || 2664 (cmd->arg1 & 2 && divinput_flags & 2665 IP_FW_DIVERT_OUTPUT_FLAG); 2666 break; 2667 2668 case O_PROTO: 2669 /* 2670 * We do not allow an arg of 0 so the 2671 * check of "proto" only suffices. 2672 */ 2673 match = (proto == cmd->arg1); 2674 break; 2675 2676 case O_IP_SRC: 2677 match = is_ipv4 && 2678 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2679 src_ip.s_addr); 2680 break; 2681 2682 case O_IP_SRC_LOOKUP: 2683 case O_IP_DST_LOOKUP: 2684 if (is_ipv4) { 2685 uint32_t a = 2686 (cmd->opcode == O_IP_DST_LOOKUP) ? 2687 dst_ip.s_addr : src_ip.s_addr; 2688 uint32_t v; 2689 2690 match = lookup_table(chain, cmd->arg1, a, 2691 &v); 2692 if (!match) 2693 break; 2694 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2695 match = 2696 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2697 else 2698 tablearg = v; 2699 } 2700 break; 2701 2702 case O_IP_SRC_MASK: 2703 case O_IP_DST_MASK: 2704 if (is_ipv4) { 2705 uint32_t a = 2706 (cmd->opcode == O_IP_DST_MASK) ? 2707 dst_ip.s_addr : src_ip.s_addr; 2708 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2709 int i = cmdlen-1; 2710 2711 for (; !match && i>0; i-= 2, p+= 2) 2712 match = (p[0] == (a & p[1])); 2713 } 2714 break; 2715 2716 case O_IP_SRC_ME: 2717 if (is_ipv4) { 2718 struct ifnet *tif; 2719 2720 INADDR_TO_IFP(src_ip, tif); 2721 match = (tif != NULL); 2722 } 2723 break; 2724 2725 case O_IP_DST_SET: 2726 case O_IP_SRC_SET: 2727 if (is_ipv4) { 2728 u_int32_t *d = (u_int32_t *)(cmd+1); 2729 u_int32_t addr = 2730 cmd->opcode == O_IP_DST_SET ? 2731 args->f_id.dst_ip : 2732 args->f_id.src_ip; 2733 2734 if (addr < d[0]) 2735 break; 2736 addr -= d[0]; /* subtract base */ 2737 match = (addr < cmd->arg1) && 2738 ( d[ 1 + (addr>>5)] & 2739 (1<<(addr & 0x1f)) ); 2740 } 2741 break; 2742 2743 case O_IP_DST: 2744 match = is_ipv4 && 2745 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2746 dst_ip.s_addr); 2747 break; 2748 2749 case O_IP_DST_ME: 2750 if (is_ipv4) { 2751 struct ifnet *tif; 2752 2753 INADDR_TO_IFP(dst_ip, tif); 2754 match = (tif != NULL); 2755 } 2756 break; 2757 2758 case O_IP_SRCPORT: 2759 case O_IP_DSTPORT: 2760 /* 2761 * offset == 0 && proto != 0 is enough 2762 * to guarantee that we have a 2763 * packet with port info. 2764 */ 2765 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2766 && offset == 0) { 2767 u_int16_t x = 2768 (cmd->opcode == O_IP_SRCPORT) ? 2769 src_port : dst_port ; 2770 u_int16_t *p = 2771 ((ipfw_insn_u16 *)cmd)->ports; 2772 int i; 2773 2774 for (i = cmdlen - 1; !match && i>0; 2775 i--, p += 2) 2776 match = (x>=p[0] && x<=p[1]); 2777 } 2778 break; 2779 2780 case O_ICMPTYPE: 2781 match = (offset == 0 && proto==IPPROTO_ICMP && 2782 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2783 break; 2784 2785 #ifdef INET6 2786 case O_ICMP6TYPE: 2787 match = is_ipv6 && offset == 0 && 2788 proto==IPPROTO_ICMPV6 && 2789 icmp6type_match( 2790 ICMP6(ulp)->icmp6_type, 2791 (ipfw_insn_u32 *)cmd); 2792 break; 2793 #endif /* INET6 */ 2794 2795 case O_IPOPT: 2796 match = (is_ipv4 && 2797 ipopts_match(ip, cmd) ); 2798 break; 2799 2800 case O_IPVER: 2801 match = (is_ipv4 && 2802 cmd->arg1 == ip->ip_v); 2803 break; 2804 2805 case O_IPID: 2806 case O_IPLEN: 2807 case O_IPTTL: 2808 if (is_ipv4) { /* only for IP packets */ 2809 uint16_t x; 2810 uint16_t *p; 2811 int i; 2812 2813 if (cmd->opcode == O_IPLEN) 2814 x = ip_len; 2815 else if (cmd->opcode == O_IPTTL) 2816 x = ip->ip_ttl; 2817 else /* must be IPID */ 2818 x = ntohs(ip->ip_id); 2819 if (cmdlen == 1) { 2820 match = (cmd->arg1 == x); 2821 break; 2822 } 2823 /* otherwise we have ranges */ 2824 p = ((ipfw_insn_u16 *)cmd)->ports; 2825 i = cmdlen - 1; 2826 for (; !match && i>0; i--, p += 2) 2827 match = (x >= p[0] && x <= p[1]); 2828 } 2829 break; 2830 2831 case O_IPPRECEDENCE: 2832 match = (is_ipv4 && 2833 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2834 break; 2835 2836 case O_IPTOS: 2837 match = (is_ipv4 && 2838 flags_match(cmd, ip->ip_tos)); 2839 break; 2840 2841 case O_TCPDATALEN: 2842 if (proto == IPPROTO_TCP && offset == 0) { 2843 struct tcphdr *tcp; 2844 uint16_t x; 2845 uint16_t *p; 2846 int i; 2847 2848 tcp = TCP(ulp); 2849 x = ip_len - 2850 ((ip->ip_hl + tcp->th_off) << 2); 2851 if (cmdlen == 1) { 2852 match = (cmd->arg1 == x); 2853 break; 2854 } 2855 /* otherwise we have ranges */ 2856 p = ((ipfw_insn_u16 *)cmd)->ports; 2857 i = cmdlen - 1; 2858 for (; !match && i>0; i--, p += 2) 2859 match = (x >= p[0] && x <= p[1]); 2860 } 2861 break; 2862 2863 case O_TCPFLAGS: 2864 match = (proto == IPPROTO_TCP && offset == 0 && 2865 flags_match(cmd, TCP(ulp)->th_flags)); 2866 break; 2867 2868 case O_TCPOPTS: 2869 match = (proto == IPPROTO_TCP && offset == 0 && 2870 tcpopts_match(TCP(ulp), cmd)); 2871 break; 2872 2873 case O_TCPSEQ: 2874 match = (proto == IPPROTO_TCP && offset == 0 && 2875 ((ipfw_insn_u32 *)cmd)->d[0] == 2876 TCP(ulp)->th_seq); 2877 break; 2878 2879 case O_TCPACK: 2880 match = (proto == IPPROTO_TCP && offset == 0 && 2881 ((ipfw_insn_u32 *)cmd)->d[0] == 2882 TCP(ulp)->th_ack); 2883 break; 2884 2885 case O_TCPWIN: 2886 match = (proto == IPPROTO_TCP && offset == 0 && 2887 cmd->arg1 == TCP(ulp)->th_win); 2888 break; 2889 2890 case O_ESTAB: 2891 /* reject packets which have SYN only */ 2892 /* XXX should i also check for TH_ACK ? */ 2893 match = (proto == IPPROTO_TCP && offset == 0 && 2894 (TCP(ulp)->th_flags & 2895 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2896 break; 2897 2898 case O_ALTQ: { 2899 struct altq_tag *at; 2900 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2901 2902 match = 1; 2903 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2904 if (mtag != NULL) 2905 break; 2906 mtag = m_tag_get(PACKET_TAG_PF_QID, 2907 sizeof(struct altq_tag), 2908 M_NOWAIT); 2909 if (mtag == NULL) { 2910 /* 2911 * Let the packet fall back to the 2912 * default ALTQ. 2913 */ 2914 break; 2915 } 2916 at = (struct altq_tag *)(mtag+1); 2917 at->qid = altq->qid; 2918 if (is_ipv4) 2919 at->af = AF_INET; 2920 else 2921 at->af = AF_LINK; 2922 at->hdr = ip; 2923 m_tag_prepend(m, mtag); 2924 break; 2925 } 2926 2927 case O_LOG: 2928 if (fw_verbose) 2929 ipfw_log(f, hlen, args, m, 2930 oif, offset, tablearg, ip); 2931 match = 1; 2932 break; 2933 2934 case O_PROB: 2935 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2936 break; 2937 2938 case O_VERREVPATH: 2939 /* Outgoing packets automatically pass/match */ 2940 match = ((oif != NULL) || 2941 (m->m_pkthdr.rcvif == NULL) || 2942 ( 2943 #ifdef INET6 2944 is_ipv6 ? 2945 verify_path6(&(args->f_id.src_ip6), 2946 m->m_pkthdr.rcvif) : 2947 #endif 2948 verify_path(src_ip, m->m_pkthdr.rcvif))); 2949 break; 2950 2951 case O_VERSRCREACH: 2952 /* Outgoing packets automatically pass/match */ 2953 match = (hlen > 0 && ((oif != NULL) || 2954 #ifdef INET6 2955 is_ipv6 ? 2956 verify_path6(&(args->f_id.src_ip6), 2957 NULL) : 2958 #endif 2959 verify_path(src_ip, NULL))); 2960 break; 2961 2962 case O_ANTISPOOF: 2963 /* Outgoing packets automatically pass/match */ 2964 if (oif == NULL && hlen > 0 && 2965 ( (is_ipv4 && in_localaddr(src_ip)) 2966 #ifdef INET6 2967 || (is_ipv6 && 2968 in6_localaddr(&(args->f_id.src_ip6))) 2969 #endif 2970 )) 2971 match = 2972 #ifdef INET6 2973 is_ipv6 ? verify_path6( 2974 &(args->f_id.src_ip6), 2975 m->m_pkthdr.rcvif) : 2976 #endif 2977 verify_path(src_ip, 2978 m->m_pkthdr.rcvif); 2979 else 2980 match = 1; 2981 break; 2982 2983 case O_IPSEC: 2984 #ifdef FAST_IPSEC 2985 match = (m_tag_find(m, 2986 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2987 #endif 2988 #ifdef IPSEC 2989 match = (ipsec_getnhist(m) != 0); 2990 #endif 2991 /* otherwise no match */ 2992 break; 2993 2994 #ifdef INET6 2995 case O_IP6_SRC: 2996 match = is_ipv6 && 2997 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2998 &((ipfw_insn_ip6 *)cmd)->addr6); 2999 break; 3000 3001 case O_IP6_DST: 3002 match = is_ipv6 && 3003 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 3004 &((ipfw_insn_ip6 *)cmd)->addr6); 3005 break; 3006 case O_IP6_SRC_MASK: 3007 case O_IP6_DST_MASK: 3008 if (is_ipv6) { 3009 int i = cmdlen - 1; 3010 struct in6_addr p; 3011 struct in6_addr *d = 3012 &((ipfw_insn_ip6 *)cmd)->addr6; 3013 3014 for (; !match && i > 0; d += 2, 3015 i -= F_INSN_SIZE(struct in6_addr) 3016 * 2) { 3017 p = (cmd->opcode == 3018 O_IP6_SRC_MASK) ? 3019 args->f_id.src_ip6: 3020 args->f_id.dst_ip6; 3021 APPLY_MASK(&p, &d[1]); 3022 match = 3023 IN6_ARE_ADDR_EQUAL(&d[0], 3024 &p); 3025 } 3026 } 3027 break; 3028 3029 case O_IP6_SRC_ME: 3030 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3031 break; 3032 3033 case O_IP6_DST_ME: 3034 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3035 break; 3036 3037 case O_FLOW6ID: 3038 match = is_ipv6 && 3039 flow6id_match(args->f_id.flow_id6, 3040 (ipfw_insn_u32 *) cmd); 3041 break; 3042 3043 case O_EXT_HDR: 3044 match = is_ipv6 && 3045 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3046 break; 3047 3048 case O_IP6: 3049 match = is_ipv6; 3050 break; 3051 #endif 3052 3053 case O_IP4: 3054 match = is_ipv4; 3055 break; 3056 3057 case O_TAG: { 3058 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3059 tablearg : cmd->arg1; 3060 3061 /* Packet is already tagged with this tag? */ 3062 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3063 3064 /* We have `untag' action when F_NOT flag is 3065 * present. And we must remove this mtag from 3066 * mbuf and reset `match' to zero (`match' will 3067 * be inversed later). 3068 * Otherwise we should allocate new mtag and 3069 * push it into mbuf. 3070 */ 3071 if (cmd->len & F_NOT) { /* `untag' action */ 3072 if (mtag != NULL) 3073 m_tag_delete(m, mtag); 3074 } else if (mtag == NULL) { 3075 if ((mtag = m_tag_alloc(MTAG_IPFW, 3076 tag, 0, M_NOWAIT)) != NULL) 3077 m_tag_prepend(m, mtag); 3078 } 3079 match = (cmd->len & F_NOT) ? 0: 1; 3080 break; 3081 } 3082 3083 case O_TAGGED: { 3084 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3085 tablearg : cmd->arg1; 3086 3087 if (cmdlen == 1) { 3088 match = m_tag_locate(m, MTAG_IPFW, 3089 tag, NULL) != NULL; 3090 break; 3091 } 3092 3093 /* we have ranges */ 3094 for (mtag = m_tag_first(m); 3095 mtag != NULL && !match; 3096 mtag = m_tag_next(m, mtag)) { 3097 uint16_t *p; 3098 int i; 3099 3100 if (mtag->m_tag_cookie != MTAG_IPFW) 3101 continue; 3102 3103 p = ((ipfw_insn_u16 *)cmd)->ports; 3104 i = cmdlen - 1; 3105 for(; !match && i > 0; i--, p += 2) 3106 match = 3107 mtag->m_tag_id >= p[0] && 3108 mtag->m_tag_id <= p[1]; 3109 } 3110 break; 3111 } 3112 3113 /* 3114 * The second set of opcodes represents 'actions', 3115 * i.e. the terminal part of a rule once the packet 3116 * matches all previous patterns. 3117 * Typically there is only one action for each rule, 3118 * and the opcode is stored at the end of the rule 3119 * (but there are exceptions -- see below). 3120 * 3121 * In general, here we set retval and terminate the 3122 * outer loop (would be a 'break 3' in some language, 3123 * but we need to do a 'goto done'). 3124 * 3125 * Exceptions: 3126 * O_COUNT and O_SKIPTO actions: 3127 * instead of terminating, we jump to the next rule 3128 * ('goto next_rule', equivalent to a 'break 2'), 3129 * or to the SKIPTO target ('goto again' after 3130 * having set f, cmd and l), respectively. 3131 * 3132 * O_TAG, O_LOG and O_ALTQ action parameters: 3133 * perform some action and set match = 1; 3134 * 3135 * O_LIMIT and O_KEEP_STATE: these opcodes are 3136 * not real 'actions', and are stored right 3137 * before the 'action' part of the rule. 3138 * These opcodes try to install an entry in the 3139 * state tables; if successful, we continue with 3140 * the next opcode (match=1; break;), otherwise 3141 * the packet * must be dropped 3142 * ('goto done' after setting retval); 3143 * 3144 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3145 * cause a lookup of the state table, and a jump 3146 * to the 'action' part of the parent rule 3147 * ('goto check_body') if an entry is found, or 3148 * (CHECK_STATE only) a jump to the next rule if 3149 * the entry is not found ('goto next_rule'). 3150 * The result of the lookup is cached to make 3151 * further instances of these opcodes are 3152 * effectively NOPs. 3153 */ 3154 case O_LIMIT: 3155 case O_KEEP_STATE: 3156 if (install_state(f, 3157 (ipfw_insn_limit *)cmd, args, tablearg)) { 3158 retval = IP_FW_DENY; 3159 goto done; /* error/limit violation */ 3160 } 3161 match = 1; 3162 break; 3163 3164 case O_PROBE_STATE: 3165 case O_CHECK_STATE: 3166 /* 3167 * dynamic rules are checked at the first 3168 * keep-state or check-state occurrence, 3169 * with the result being stored in dyn_dir. 3170 * The compiler introduces a PROBE_STATE 3171 * instruction for us when we have a 3172 * KEEP_STATE (because PROBE_STATE needs 3173 * to be run first). 3174 */ 3175 if (dyn_dir == MATCH_UNKNOWN && 3176 (q = lookup_dyn_rule(&args->f_id, 3177 &dyn_dir, proto == IPPROTO_TCP ? 3178 TCP(ulp) : NULL)) 3179 != NULL) { 3180 /* 3181 * Found dynamic entry, update stats 3182 * and jump to the 'action' part of 3183 * the parent rule. 3184 */ 3185 q->pcnt++; 3186 q->bcnt += pktlen; 3187 f = q->rule; 3188 cmd = ACTION_PTR(f); 3189 l = f->cmd_len - f->act_ofs; 3190 IPFW_DYN_UNLOCK(); 3191 goto check_body; 3192 } 3193 /* 3194 * Dynamic entry not found. If CHECK_STATE, 3195 * skip to next rule, if PROBE_STATE just 3196 * ignore and continue with next opcode. 3197 */ 3198 if (cmd->opcode == O_CHECK_STATE) 3199 goto next_rule; 3200 match = 1; 3201 break; 3202 3203 case O_ACCEPT: 3204 retval = 0; /* accept */ 3205 goto done; 3206 3207 case O_PIPE: 3208 case O_QUEUE: 3209 args->rule = f; /* report matching rule */ 3210 if (cmd->arg1 == IP_FW_TABLEARG) 3211 args->cookie = tablearg; 3212 else 3213 args->cookie = cmd->arg1; 3214 retval = IP_FW_DUMMYNET; 3215 goto done; 3216 3217 case O_DIVERT: 3218 case O_TEE: { 3219 struct divert_tag *dt; 3220 3221 if (args->eh) /* not on layer 2 */ 3222 break; 3223 mtag = m_tag_get(PACKET_TAG_DIVERT, 3224 sizeof(struct divert_tag), 3225 M_NOWAIT); 3226 if (mtag == NULL) { 3227 /* XXX statistic */ 3228 /* drop packet */ 3229 IPFW_RUNLOCK(chain); 3230 return (IP_FW_DENY); 3231 } 3232 dt = (struct divert_tag *)(mtag+1); 3233 dt->cookie = f->rulenum; 3234 if (cmd->arg1 == IP_FW_TABLEARG) 3235 dt->info = tablearg; 3236 else 3237 dt->info = cmd->arg1; 3238 m_tag_prepend(m, mtag); 3239 retval = (cmd->opcode == O_DIVERT) ? 3240 IP_FW_DIVERT : IP_FW_TEE; 3241 goto done; 3242 } 3243 3244 case O_COUNT: 3245 case O_SKIPTO: 3246 f->pcnt++; /* update stats */ 3247 f->bcnt += pktlen; 3248 f->timestamp = time_uptime; 3249 if (cmd->opcode == O_COUNT) 3250 goto next_rule; 3251 /* handle skipto */ 3252 if (f->next_rule == NULL) 3253 lookup_next_rule(f); 3254 f = f->next_rule; 3255 goto again; 3256 3257 case O_REJECT: 3258 /* 3259 * Drop the packet and send a reject notice 3260 * if the packet is not ICMP (or is an ICMP 3261 * query), and it is not multicast/broadcast. 3262 */ 3263 if (hlen > 0 && is_ipv4 && offset == 0 && 3264 (proto != IPPROTO_ICMP || 3265 is_icmp_query(ICMP(ulp))) && 3266 !(m->m_flags & (M_BCAST|M_MCAST)) && 3267 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3268 send_reject(args, cmd->arg1, 3269 offset,ip_len, ip); 3270 m = args->m; 3271 } 3272 /* FALLTHROUGH */ 3273 #ifdef INET6 3274 case O_UNREACH6: 3275 if (hlen > 0 && is_ipv6 && 3276 (proto != IPPROTO_ICMPV6 || 3277 (is_icmp6_query(args->f_id.flags) == 1)) && 3278 !(m->m_flags & (M_BCAST|M_MCAST)) && 3279 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3280 send_reject6(args, cmd->arg1, 3281 offset, hlen, (struct ip6_hdr *)ip); 3282 m = args->m; 3283 } 3284 /* FALLTHROUGH */ 3285 #endif 3286 case O_DENY: 3287 retval = IP_FW_DENY; 3288 goto done; 3289 3290 case O_FORWARD_IP: { 3291 struct sockaddr_in *sa; 3292 sa = &(((ipfw_insn_sa *)cmd)->sa); 3293 if (args->eh) /* not valid on layer2 pkts */ 3294 break; 3295 if (!q || dyn_dir == MATCH_FORWARD) { 3296 if (sa->sin_addr.s_addr == INADDR_ANY) { 3297 bcopy(sa, &args->hopstore, 3298 sizeof(*sa)); 3299 args->hopstore.sin_addr.s_addr = 3300 htonl(tablearg); 3301 args->next_hop = 3302 &args->hopstore; 3303 } else { 3304 args->next_hop = sa; 3305 } 3306 } 3307 retval = IP_FW_PASS; 3308 } 3309 goto done; 3310 3311 case O_NETGRAPH: 3312 case O_NGTEE: 3313 args->rule = f; /* report matching rule */ 3314 if (cmd->arg1 == IP_FW_TABLEARG) 3315 args->cookie = tablearg; 3316 else 3317 args->cookie = cmd->arg1; 3318 retval = (cmd->opcode == O_NETGRAPH) ? 3319 IP_FW_NETGRAPH : IP_FW_NGTEE; 3320 goto done; 3321 3322 default: 3323 panic("-- unknown opcode %d\n", cmd->opcode); 3324 } /* end of switch() on opcodes */ 3325 3326 if (cmd->len & F_NOT) 3327 match = !match; 3328 3329 if (match) { 3330 if (cmd->len & F_OR) 3331 skip_or = 1; 3332 } else { 3333 if (!(cmd->len & F_OR)) /* not an OR block, */ 3334 break; /* try next rule */ 3335 } 3336 3337 } /* end of inner for, scan opcodes */ 3338 3339 next_rule:; /* try next rule */ 3340 3341 } /* end of outer for, scan rules */ 3342 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3343 IPFW_RUNLOCK(chain); 3344 return (IP_FW_DENY); 3345 3346 done: 3347 /* Update statistics */ 3348 f->pcnt++; 3349 f->bcnt += pktlen; 3350 f->timestamp = time_uptime; 3351 IPFW_RUNLOCK(chain); 3352 return (retval); 3353 3354 pullup_failed: 3355 if (fw_verbose) 3356 printf("ipfw: pullup failed\n"); 3357 return (IP_FW_DENY); 3358 } 3359 3360 /* 3361 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3362 * These will be reconstructed on the fly as packets are matched. 3363 */ 3364 static void 3365 flush_rule_ptrs(struct ip_fw_chain *chain) 3366 { 3367 struct ip_fw *rule; 3368 3369 IPFW_WLOCK_ASSERT(chain); 3370 3371 for (rule = chain->rules; rule; rule = rule->next) 3372 rule->next_rule = NULL; 3373 } 3374 3375 /* 3376 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3377 * possibly create a rule number and add the rule to the list. 3378 * Update the rule_number in the input struct so the caller knows it as well. 3379 */ 3380 static int 3381 add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3382 { 3383 struct ip_fw *rule, *f, *prev; 3384 int l = RULESIZE(input_rule); 3385 3386 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3387 return (EINVAL); 3388 3389 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3390 if (rule == NULL) 3391 return (ENOSPC); 3392 3393 bcopy(input_rule, rule, l); 3394 3395 rule->next = NULL; 3396 rule->next_rule = NULL; 3397 3398 rule->pcnt = 0; 3399 rule->bcnt = 0; 3400 rule->timestamp = 0; 3401 3402 IPFW_WLOCK(chain); 3403 3404 if (chain->rules == NULL) { /* default rule */ 3405 chain->rules = rule; 3406 goto done; 3407 } 3408 3409 /* 3410 * If rulenum is 0, find highest numbered rule before the 3411 * default rule, and add autoinc_step 3412 */ 3413 if (autoinc_step < 1) 3414 autoinc_step = 1; 3415 else if (autoinc_step > 1000) 3416 autoinc_step = 1000; 3417 if (rule->rulenum == 0) { 3418 /* 3419 * locate the highest numbered rule before default 3420 */ 3421 for (f = chain->rules; f; f = f->next) { 3422 if (f->rulenum == IPFW_DEFAULT_RULE) 3423 break; 3424 rule->rulenum = f->rulenum; 3425 } 3426 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3427 rule->rulenum += autoinc_step; 3428 input_rule->rulenum = rule->rulenum; 3429 } 3430 3431 /* 3432 * Now insert the new rule in the right place in the sorted list. 3433 */ 3434 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3435 if (f->rulenum > rule->rulenum) { /* found the location */ 3436 if (prev) { 3437 rule->next = f; 3438 prev->next = rule; 3439 } else { /* head insert */ 3440 rule->next = chain->rules; 3441 chain->rules = rule; 3442 } 3443 break; 3444 } 3445 } 3446 flush_rule_ptrs(chain); 3447 done: 3448 static_count++; 3449 static_len += l; 3450 IPFW_WUNLOCK(chain); 3451 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3452 rule->rulenum, static_count);) 3453 return (0); 3454 } 3455 3456 /** 3457 * Remove a static rule (including derived * dynamic rules) 3458 * and place it on the ``reap list'' for later reclamation. 3459 * The caller is in charge of clearing rule pointers to avoid 3460 * dangling pointers. 3461 * @return a pointer to the next entry. 3462 * Arguments are not checked, so they better be correct. 3463 */ 3464 static struct ip_fw * 3465 remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3466 { 3467 struct ip_fw *n; 3468 int l = RULESIZE(rule); 3469 3470 IPFW_WLOCK_ASSERT(chain); 3471 3472 n = rule->next; 3473 IPFW_DYN_LOCK(); 3474 remove_dyn_rule(rule, NULL /* force removal */); 3475 IPFW_DYN_UNLOCK(); 3476 if (prev == NULL) 3477 chain->rules = n; 3478 else 3479 prev->next = n; 3480 static_count--; 3481 static_len -= l; 3482 3483 rule->next = chain->reap; 3484 chain->reap = rule; 3485 3486 return n; 3487 } 3488 3489 /** 3490 * Reclaim storage associated with a list of rules. This is 3491 * typically the list created using remove_rule. 3492 */ 3493 static void 3494 reap_rules(struct ip_fw *head) 3495 { 3496 struct ip_fw *rule; 3497 3498 while ((rule = head) != NULL) { 3499 head = head->next; 3500 if (DUMMYNET_LOADED) 3501 ip_dn_ruledel_ptr(rule); 3502 free(rule, M_IPFW); 3503 } 3504 } 3505 3506 /* 3507 * Remove all rules from a chain (except rules in set RESVD_SET 3508 * unless kill_default = 1). The caller is responsible for 3509 * reclaiming storage for the rules left in chain->reap. 3510 */ 3511 static void 3512 free_chain(struct ip_fw_chain *chain, int kill_default) 3513 { 3514 struct ip_fw *prev, *rule; 3515 3516 IPFW_WLOCK_ASSERT(chain); 3517 3518 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3519 for (prev = NULL, rule = chain->rules; rule ; ) 3520 if (kill_default || rule->set != RESVD_SET) 3521 rule = remove_rule(chain, rule, prev); 3522 else { 3523 prev = rule; 3524 rule = rule->next; 3525 } 3526 } 3527 3528 /** 3529 * Remove all rules with given number, and also do set manipulation. 3530 * Assumes chain != NULL && *chain != NULL. 3531 * 3532 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3533 * the next 8 bits are the new set, the top 8 bits are the command: 3534 * 3535 * 0 delete rules with given number 3536 * 1 delete rules with given set number 3537 * 2 move rules with given number to new set 3538 * 3 move rules with given set number to new set 3539 * 4 swap sets with given numbers 3540 */ 3541 static int 3542 del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3543 { 3544 struct ip_fw *prev = NULL, *rule; 3545 u_int16_t rulenum; /* rule or old_set */ 3546 u_int8_t cmd, new_set; 3547 3548 rulenum = arg & 0xffff; 3549 cmd = (arg >> 24) & 0xff; 3550 new_set = (arg >> 16) & 0xff; 3551 3552 if (cmd > 4) 3553 return EINVAL; 3554 if (new_set > RESVD_SET) 3555 return EINVAL; 3556 if (cmd == 0 || cmd == 2) { 3557 if (rulenum >= IPFW_DEFAULT_RULE) 3558 return EINVAL; 3559 } else { 3560 if (rulenum > RESVD_SET) /* old_set */ 3561 return EINVAL; 3562 } 3563 3564 IPFW_WLOCK(chain); 3565 rule = chain->rules; 3566 chain->reap = NULL; 3567 switch (cmd) { 3568 case 0: /* delete rules with given number */ 3569 /* 3570 * locate first rule to delete 3571 */ 3572 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3573 ; 3574 if (rule->rulenum != rulenum) { 3575 IPFW_WUNLOCK(chain); 3576 return EINVAL; 3577 } 3578 3579 /* 3580 * flush pointers outside the loop, then delete all matching 3581 * rules. prev remains the same throughout the cycle. 3582 */ 3583 flush_rule_ptrs(chain); 3584 while (rule->rulenum == rulenum) 3585 rule = remove_rule(chain, rule, prev); 3586 break; 3587 3588 case 1: /* delete all rules with given set number */ 3589 flush_rule_ptrs(chain); 3590 rule = chain->rules; 3591 while (rule->rulenum < IPFW_DEFAULT_RULE) 3592 if (rule->set == rulenum) 3593 rule = remove_rule(chain, rule, prev); 3594 else { 3595 prev = rule; 3596 rule = rule->next; 3597 } 3598 break; 3599 3600 case 2: /* move rules with given number to new set */ 3601 rule = chain->rules; 3602 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3603 if (rule->rulenum == rulenum) 3604 rule->set = new_set; 3605 break; 3606 3607 case 3: /* move rules with given set number to new set */ 3608 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3609 if (rule->set == rulenum) 3610 rule->set = new_set; 3611 break; 3612 3613 case 4: /* swap two sets */ 3614 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3615 if (rule->set == rulenum) 3616 rule->set = new_set; 3617 else if (rule->set == new_set) 3618 rule->set = rulenum; 3619 break; 3620 } 3621 /* 3622 * Look for rules to reclaim. We grab the list before 3623 * releasing the lock then reclaim them w/o the lock to 3624 * avoid a LOR with dummynet. 3625 */ 3626 rule = chain->reap; 3627 chain->reap = NULL; 3628 IPFW_WUNLOCK(chain); 3629 if (rule) 3630 reap_rules(rule); 3631 return 0; 3632 } 3633 3634 /* 3635 * Clear counters for a specific rule. 3636 * The enclosing "table" is assumed locked. 3637 */ 3638 static void 3639 clear_counters(struct ip_fw *rule, int log_only) 3640 { 3641 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3642 3643 if (log_only == 0) { 3644 rule->bcnt = rule->pcnt = 0; 3645 rule->timestamp = 0; 3646 } 3647 if (l->o.opcode == O_LOG) 3648 l->log_left = l->max_log; 3649 } 3650 3651 /** 3652 * Reset some or all counters on firewall rules. 3653 * @arg frwl is null to clear all entries, or contains a specific 3654 * rule number. 3655 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3656 */ 3657 static int 3658 zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3659 { 3660 struct ip_fw *rule; 3661 char *msg; 3662 3663 IPFW_WLOCK(chain); 3664 if (rulenum == 0) { 3665 norule_counter = 0; 3666 for (rule = chain->rules; rule; rule = rule->next) 3667 clear_counters(rule, log_only); 3668 msg = log_only ? "ipfw: All logging counts reset.\n" : 3669 "ipfw: Accounting cleared.\n"; 3670 } else { 3671 int cleared = 0; 3672 /* 3673 * We can have multiple rules with the same number, so we 3674 * need to clear them all. 3675 */ 3676 for (rule = chain->rules; rule; rule = rule->next) 3677 if (rule->rulenum == rulenum) { 3678 while (rule && rule->rulenum == rulenum) { 3679 clear_counters(rule, log_only); 3680 rule = rule->next; 3681 } 3682 cleared = 1; 3683 break; 3684 } 3685 if (!cleared) { /* we did not find any matching rules */ 3686 IPFW_WUNLOCK(chain); 3687 return (EINVAL); 3688 } 3689 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3690 "ipfw: Entry %d cleared.\n"; 3691 } 3692 IPFW_WUNLOCK(chain); 3693 3694 if (fw_verbose) 3695 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3696 return (0); 3697 } 3698 3699 /* 3700 * Check validity of the structure before insert. 3701 * Fortunately rules are simple, so this mostly need to check rule sizes. 3702 */ 3703 static int 3704 check_ipfw_struct(struct ip_fw *rule, int size) 3705 { 3706 int l, cmdlen = 0; 3707 int have_action=0; 3708 ipfw_insn *cmd; 3709 3710 if (size < sizeof(*rule)) { 3711 printf("ipfw: rule too short\n"); 3712 return (EINVAL); 3713 } 3714 /* first, check for valid size */ 3715 l = RULESIZE(rule); 3716 if (l != size) { 3717 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3718 return (EINVAL); 3719 } 3720 if (rule->act_ofs >= rule->cmd_len) { 3721 printf("ipfw: bogus action offset (%u > %u)\n", 3722 rule->act_ofs, rule->cmd_len - 1); 3723 return (EINVAL); 3724 } 3725 /* 3726 * Now go for the individual checks. Very simple ones, basically only 3727 * instruction sizes. 3728 */ 3729 for (l = rule->cmd_len, cmd = rule->cmd ; 3730 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3731 cmdlen = F_LEN(cmd); 3732 if (cmdlen > l) { 3733 printf("ipfw: opcode %d size truncated\n", 3734 cmd->opcode); 3735 return EINVAL; 3736 } 3737 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3738 switch (cmd->opcode) { 3739 case O_PROBE_STATE: 3740 case O_KEEP_STATE: 3741 case O_PROTO: 3742 case O_IP_SRC_ME: 3743 case O_IP_DST_ME: 3744 case O_LAYER2: 3745 case O_IN: 3746 case O_FRAG: 3747 case O_DIVERTED: 3748 case O_IPOPT: 3749 case O_IPTOS: 3750 case O_IPPRECEDENCE: 3751 case O_IPVER: 3752 case O_TCPWIN: 3753 case O_TCPFLAGS: 3754 case O_TCPOPTS: 3755 case O_ESTAB: 3756 case O_VERREVPATH: 3757 case O_VERSRCREACH: 3758 case O_ANTISPOOF: 3759 case O_IPSEC: 3760 #ifdef INET6 3761 case O_IP6_SRC_ME: 3762 case O_IP6_DST_ME: 3763 case O_EXT_HDR: 3764 case O_IP6: 3765 #endif 3766 case O_IP4: 3767 case O_TAG: 3768 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3769 goto bad_size; 3770 break; 3771 3772 case O_UID: 3773 case O_GID: 3774 case O_JAIL: 3775 case O_IP_SRC: 3776 case O_IP_DST: 3777 case O_TCPSEQ: 3778 case O_TCPACK: 3779 case O_PROB: 3780 case O_ICMPTYPE: 3781 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3782 goto bad_size; 3783 break; 3784 3785 case O_LIMIT: 3786 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3787 goto bad_size; 3788 break; 3789 3790 case O_LOG: 3791 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3792 goto bad_size; 3793 3794 ((ipfw_insn_log *)cmd)->log_left = 3795 ((ipfw_insn_log *)cmd)->max_log; 3796 3797 break; 3798 3799 case O_IP_SRC_MASK: 3800 case O_IP_DST_MASK: 3801 /* only odd command lengths */ 3802 if ( !(cmdlen & 1) || cmdlen > 31) 3803 goto bad_size; 3804 break; 3805 3806 case O_IP_SRC_SET: 3807 case O_IP_DST_SET: 3808 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3809 printf("ipfw: invalid set size %d\n", 3810 cmd->arg1); 3811 return EINVAL; 3812 } 3813 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3814 (cmd->arg1+31)/32 ) 3815 goto bad_size; 3816 break; 3817 3818 case O_IP_SRC_LOOKUP: 3819 case O_IP_DST_LOOKUP: 3820 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3821 printf("ipfw: invalid table number %d\n", 3822 cmd->arg1); 3823 return (EINVAL); 3824 } 3825 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3826 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3827 goto bad_size; 3828 break; 3829 3830 case O_MACADDR2: 3831 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3832 goto bad_size; 3833 break; 3834 3835 case O_NOP: 3836 case O_IPID: 3837 case O_IPTTL: 3838 case O_IPLEN: 3839 case O_TCPDATALEN: 3840 case O_TAGGED: 3841 if (cmdlen < 1 || cmdlen > 31) 3842 goto bad_size; 3843 break; 3844 3845 case O_MAC_TYPE: 3846 case O_IP_SRCPORT: 3847 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3848 if (cmdlen < 2 || cmdlen > 31) 3849 goto bad_size; 3850 break; 3851 3852 case O_RECV: 3853 case O_XMIT: 3854 case O_VIA: 3855 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3856 goto bad_size; 3857 break; 3858 3859 case O_ALTQ: 3860 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3861 goto bad_size; 3862 break; 3863 3864 case O_PIPE: 3865 case O_QUEUE: 3866 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 3867 goto bad_size; 3868 goto check_action; 3869 3870 case O_FORWARD_IP: 3871 #ifdef IPFIREWALL_FORWARD 3872 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3873 goto bad_size; 3874 goto check_action; 3875 #else 3876 return EINVAL; 3877 #endif 3878 3879 case O_DIVERT: 3880 case O_TEE: 3881 if (ip_divert_ptr == NULL) 3882 return EINVAL; 3883 else 3884 goto check_size; 3885 case O_NETGRAPH: 3886 case O_NGTEE: 3887 if (!NG_IPFW_LOADED) 3888 return EINVAL; 3889 else 3890 goto check_size; 3891 case O_FORWARD_MAC: /* XXX not implemented yet */ 3892 case O_CHECK_STATE: 3893 case O_COUNT: 3894 case O_ACCEPT: 3895 case O_DENY: 3896 case O_REJECT: 3897 #ifdef INET6 3898 case O_UNREACH6: 3899 #endif 3900 case O_SKIPTO: 3901 check_size: 3902 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3903 goto bad_size; 3904 check_action: 3905 if (have_action) { 3906 printf("ipfw: opcode %d, multiple actions" 3907 " not allowed\n", 3908 cmd->opcode); 3909 return EINVAL; 3910 } 3911 have_action = 1; 3912 if (l != cmdlen) { 3913 printf("ipfw: opcode %d, action must be" 3914 " last opcode\n", 3915 cmd->opcode); 3916 return EINVAL; 3917 } 3918 break; 3919 #ifdef INET6 3920 case O_IP6_SRC: 3921 case O_IP6_DST: 3922 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3923 F_INSN_SIZE(ipfw_insn)) 3924 goto bad_size; 3925 break; 3926 3927 case O_FLOW6ID: 3928 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3929 ((ipfw_insn_u32 *)cmd)->o.arg1) 3930 goto bad_size; 3931 break; 3932 3933 case O_IP6_SRC_MASK: 3934 case O_IP6_DST_MASK: 3935 if ( !(cmdlen & 1) || cmdlen > 127) 3936 goto bad_size; 3937 break; 3938 case O_ICMP6TYPE: 3939 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3940 goto bad_size; 3941 break; 3942 #endif 3943 3944 default: 3945 switch (cmd->opcode) { 3946 #ifndef INET6 3947 case O_IP6_SRC_ME: 3948 case O_IP6_DST_ME: 3949 case O_EXT_HDR: 3950 case O_IP6: 3951 case O_UNREACH6: 3952 case O_IP6_SRC: 3953 case O_IP6_DST: 3954 case O_FLOW6ID: 3955 case O_IP6_SRC_MASK: 3956 case O_IP6_DST_MASK: 3957 case O_ICMP6TYPE: 3958 printf("ipfw: no IPv6 support in kernel\n"); 3959 return EPROTONOSUPPORT; 3960 #endif 3961 default: 3962 printf("ipfw: opcode %d, unknown opcode\n", 3963 cmd->opcode); 3964 return EINVAL; 3965 } 3966 } 3967 } 3968 if (have_action == 0) { 3969 printf("ipfw: missing action\n"); 3970 return EINVAL; 3971 } 3972 return 0; 3973 3974 bad_size: 3975 printf("ipfw: opcode %d size %d wrong\n", 3976 cmd->opcode, cmdlen); 3977 return EINVAL; 3978 } 3979 3980 /* 3981 * Copy the static and dynamic rules to the supplied buffer 3982 * and return the amount of space actually used. 3983 */ 3984 static size_t 3985 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3986 { 3987 char *bp = buf; 3988 char *ep = bp + space; 3989 struct ip_fw *rule; 3990 int i; 3991 time_t boot_seconds; 3992 3993 boot_seconds = boottime.tv_sec; 3994 /* XXX this can take a long time and locking will block packet flow */ 3995 IPFW_RLOCK(chain); 3996 for (rule = chain->rules; rule ; rule = rule->next) { 3997 /* 3998 * Verify the entry fits in the buffer in case the 3999 * rules changed between calculating buffer space and 4000 * now. This would be better done using a generation 4001 * number but should suffice for now. 4002 */ 4003 i = RULESIZE(rule); 4004 if (bp + i <= ep) { 4005 bcopy(rule, bp, i); 4006 /* 4007 * XXX HACK. Store the disable mask in the "next" pointer 4008 * in a wild attempt to keep the ABI the same. 4009 * Why do we do this on EVERY rule? 4010 */ 4011 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 4012 sizeof(set_disable)); 4013 if (((struct ip_fw *)bp)->timestamp) 4014 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4015 bp += i; 4016 } 4017 } 4018 IPFW_RUNLOCK(chain); 4019 if (ipfw_dyn_v) { 4020 ipfw_dyn_rule *p, *last = NULL; 4021 4022 IPFW_DYN_LOCK(); 4023 for (i = 0 ; i < curr_dyn_buckets; i++) 4024 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4025 if (bp + sizeof *p <= ep) { 4026 ipfw_dyn_rule *dst = 4027 (ipfw_dyn_rule *)bp; 4028 bcopy(p, dst, sizeof *p); 4029 bcopy(&(p->rule->rulenum), &(dst->rule), 4030 sizeof(p->rule->rulenum)); 4031 /* 4032 * store a non-null value in "next". 4033 * The userland code will interpret a 4034 * NULL here as a marker 4035 * for the last dynamic rule. 4036 */ 4037 bcopy(&dst, &dst->next, sizeof(dst)); 4038 last = dst; 4039 dst->expire = 4040 TIME_LEQ(dst->expire, time_uptime) ? 4041 0 : dst->expire - time_uptime ; 4042 bp += sizeof(ipfw_dyn_rule); 4043 } 4044 } 4045 IPFW_DYN_UNLOCK(); 4046 if (last != NULL) /* mark last dynamic rule */ 4047 bzero(&last->next, sizeof(last)); 4048 } 4049 return (bp - (char *)buf); 4050 } 4051 4052 4053 /** 4054 * {set|get}sockopt parser. 4055 */ 4056 static int 4057 ipfw_ctl(struct sockopt *sopt) 4058 { 4059 #define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4060 int error, rule_num; 4061 size_t size; 4062 struct ip_fw *buf, *rule; 4063 u_int32_t rulenum[2]; 4064 4065 error = suser(sopt->sopt_td); 4066 if (error) 4067 return (error); 4068 4069 /* 4070 * Disallow modifications in really-really secure mode, but still allow 4071 * the logging counters to be reset. 4072 */ 4073 if (sopt->sopt_name == IP_FW_ADD || 4074 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4075 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4076 if (error) 4077 return (error); 4078 } 4079 4080 error = 0; 4081 4082 switch (sopt->sopt_name) { 4083 case IP_FW_GET: 4084 /* 4085 * pass up a copy of the current rules. Static rules 4086 * come first (the last of which has number IPFW_DEFAULT_RULE), 4087 * followed by a possibly empty list of dynamic rule. 4088 * The last dynamic rule has NULL in the "next" field. 4089 * 4090 * Note that the calculated size is used to bound the 4091 * amount of data returned to the user. The rule set may 4092 * change between calculating the size and returning the 4093 * data in which case we'll just return what fits. 4094 */ 4095 size = static_len; /* size of static rules */ 4096 if (ipfw_dyn_v) /* add size of dyn.rules */ 4097 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4098 4099 /* 4100 * XXX todo: if the user passes a short length just to know 4101 * how much room is needed, do not bother filling up the 4102 * buffer, just jump to the sooptcopyout. 4103 */ 4104 buf = malloc(size, M_TEMP, M_WAITOK); 4105 error = sooptcopyout(sopt, buf, 4106 ipfw_getrules(&layer3_chain, buf, size)); 4107 free(buf, M_TEMP); 4108 break; 4109 4110 case IP_FW_FLUSH: 4111 /* 4112 * Normally we cannot release the lock on each iteration. 4113 * We could do it here only because we start from the head all 4114 * the times so there is no risk of missing some entries. 4115 * On the other hand, the risk is that we end up with 4116 * a very inconsistent ruleset, so better keep the lock 4117 * around the whole cycle. 4118 * 4119 * XXX this code can be improved by resetting the head of 4120 * the list to point to the default rule, and then freeing 4121 * the old list without the need for a lock. 4122 */ 4123 4124 IPFW_WLOCK(&layer3_chain); 4125 layer3_chain.reap = NULL; 4126 free_chain(&layer3_chain, 0 /* keep default rule */); 4127 rule = layer3_chain.reap; 4128 layer3_chain.reap = NULL; 4129 IPFW_WUNLOCK(&layer3_chain); 4130 if (rule != NULL) 4131 reap_rules(rule); 4132 break; 4133 4134 case IP_FW_ADD: 4135 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4136 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4137 sizeof(struct ip_fw) ); 4138 if (error == 0) 4139 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4140 if (error == 0) { 4141 error = add_rule(&layer3_chain, rule); 4142 size = RULESIZE(rule); 4143 if (!error && sopt->sopt_dir == SOPT_GET) 4144 error = sooptcopyout(sopt, rule, size); 4145 } 4146 free(rule, M_TEMP); 4147 break; 4148 4149 case IP_FW_DEL: 4150 /* 4151 * IP_FW_DEL is used for deleting single rules or sets, 4152 * and (ab)used to atomically manipulate sets. Argument size 4153 * is used to distinguish between the two: 4154 * sizeof(u_int32_t) 4155 * delete single rule or set of rules, 4156 * or reassign rules (or sets) to a different set. 4157 * 2*sizeof(u_int32_t) 4158 * atomic disable/enable sets. 4159 * first u_int32_t contains sets to be disabled, 4160 * second u_int32_t contains sets to be enabled. 4161 */ 4162 error = sooptcopyin(sopt, rulenum, 4163 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4164 if (error) 4165 break; 4166 size = sopt->sopt_valsize; 4167 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4168 error = del_entry(&layer3_chain, rulenum[0]); 4169 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4170 set_disable = 4171 (set_disable | rulenum[0]) & ~rulenum[1] & 4172 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4173 else 4174 error = EINVAL; 4175 break; 4176 4177 case IP_FW_ZERO: 4178 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 4179 rule_num = 0; 4180 if (sopt->sopt_val != 0) { 4181 error = sooptcopyin(sopt, &rule_num, 4182 sizeof(int), sizeof(int)); 4183 if (error) 4184 break; 4185 } 4186 error = zero_entry(&layer3_chain, rule_num, 4187 sopt->sopt_name == IP_FW_RESETLOG); 4188 break; 4189 4190 case IP_FW_TABLE_ADD: 4191 { 4192 ipfw_table_entry ent; 4193 4194 error = sooptcopyin(sopt, &ent, 4195 sizeof(ent), sizeof(ent)); 4196 if (error) 4197 break; 4198 error = add_table_entry(&layer3_chain, ent.tbl, 4199 ent.addr, ent.masklen, ent.value); 4200 } 4201 break; 4202 4203 case IP_FW_TABLE_DEL: 4204 { 4205 ipfw_table_entry ent; 4206 4207 error = sooptcopyin(sopt, &ent, 4208 sizeof(ent), sizeof(ent)); 4209 if (error) 4210 break; 4211 error = del_table_entry(&layer3_chain, ent.tbl, 4212 ent.addr, ent.masklen); 4213 } 4214 break; 4215 4216 case IP_FW_TABLE_FLUSH: 4217 { 4218 u_int16_t tbl; 4219 4220 error = sooptcopyin(sopt, &tbl, 4221 sizeof(tbl), sizeof(tbl)); 4222 if (error) 4223 break; 4224 IPFW_WLOCK(&layer3_chain); 4225 error = flush_table(&layer3_chain, tbl); 4226 IPFW_WUNLOCK(&layer3_chain); 4227 } 4228 break; 4229 4230 case IP_FW_TABLE_GETSIZE: 4231 { 4232 u_int32_t tbl, cnt; 4233 4234 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4235 sizeof(tbl)))) 4236 break; 4237 IPFW_RLOCK(&layer3_chain); 4238 error = count_table(&layer3_chain, tbl, &cnt); 4239 IPFW_RUNLOCK(&layer3_chain); 4240 if (error) 4241 break; 4242 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4243 } 4244 break; 4245 4246 case IP_FW_TABLE_LIST: 4247 { 4248 ipfw_table *tbl; 4249 4250 if (sopt->sopt_valsize < sizeof(*tbl)) { 4251 error = EINVAL; 4252 break; 4253 } 4254 size = sopt->sopt_valsize; 4255 tbl = malloc(size, M_TEMP, M_WAITOK); 4256 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4257 if (error) { 4258 free(tbl, M_TEMP); 4259 break; 4260 } 4261 tbl->size = (size - sizeof(*tbl)) / 4262 sizeof(ipfw_table_entry); 4263 IPFW_RLOCK(&layer3_chain); 4264 error = dump_table(&layer3_chain, tbl); 4265 IPFW_RUNLOCK(&layer3_chain); 4266 if (error) { 4267 free(tbl, M_TEMP); 4268 break; 4269 } 4270 error = sooptcopyout(sopt, tbl, size); 4271 free(tbl, M_TEMP); 4272 } 4273 break; 4274 4275 default: 4276 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4277 error = EINVAL; 4278 } 4279 4280 return (error); 4281 #undef RULE_MAXSIZE 4282 } 4283 4284 /** 4285 * dummynet needs a reference to the default rule, because rules can be 4286 * deleted while packets hold a reference to them. When this happens, 4287 * dummynet changes the reference to the default rule (it could well be a 4288 * NULL pointer, but this way we do not need to check for the special 4289 * case, plus here he have info on the default behaviour). 4290 */ 4291 struct ip_fw *ip_fw_default_rule; 4292 4293 /* 4294 * This procedure is only used to handle keepalives. It is invoked 4295 * every dyn_keepalive_period 4296 */ 4297 static void 4298 ipfw_tick(void * __unused unused) 4299 { 4300 struct mbuf *m0, *m, *mnext, **mtailp; 4301 int i; 4302 ipfw_dyn_rule *q; 4303 4304 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4305 goto done; 4306 4307 /* 4308 * We make a chain of packets to go out here -- not deferring 4309 * until after we drop the IPFW dynamic rule lock would result 4310 * in a lock order reversal with the normal packet input -> ipfw 4311 * call stack. 4312 */ 4313 m0 = NULL; 4314 mtailp = &m0; 4315 IPFW_DYN_LOCK(); 4316 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4317 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4318 if (q->dyn_type == O_LIMIT_PARENT) 4319 continue; 4320 if (q->id.proto != IPPROTO_TCP) 4321 continue; 4322 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4323 continue; 4324 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4325 q->expire)) 4326 continue; /* too early */ 4327 if (TIME_LEQ(q->expire, time_uptime)) 4328 continue; /* too late, rule expired */ 4329 4330 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4331 q->ack_fwd, TH_SYN); 4332 if (*mtailp != NULL) 4333 mtailp = &(*mtailp)->m_nextpkt; 4334 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4335 q->ack_rev, 0); 4336 if (*mtailp != NULL) 4337 mtailp = &(*mtailp)->m_nextpkt; 4338 } 4339 } 4340 IPFW_DYN_UNLOCK(); 4341 for (m = mnext = m0; m != NULL; m = mnext) { 4342 mnext = m->m_nextpkt; 4343 m->m_nextpkt = NULL; 4344 ip_output(m, NULL, NULL, 0, NULL, NULL); 4345 } 4346 done: 4347 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4348 } 4349 4350 int 4351 ipfw_init(void) 4352 { 4353 struct ip_fw default_rule; 4354 int error; 4355 4356 #ifdef INET6 4357 /* Setup IPv6 fw sysctl tree. */ 4358 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4359 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4360 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4361 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4362 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4363 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4364 &fw_deny_unknown_exthdrs, 0, 4365 "Deny packets with unknown IPv6 Extension Headers"); 4366 #endif 4367 4368 layer3_chain.rules = NULL; 4369 layer3_chain.want_write = 0; 4370 layer3_chain.busy_count = 0; 4371 cv_init(&layer3_chain.cv, "Condition variable for IPFW rw locks"); 4372 IPFW_LOCK_INIT(&layer3_chain); 4373 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4374 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4375 UMA_ALIGN_PTR, 0); 4376 IPFW_DYN_LOCK_INIT(); 4377 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4378 4379 bzero(&default_rule, sizeof default_rule); 4380 4381 default_rule.act_ofs = 0; 4382 default_rule.rulenum = IPFW_DEFAULT_RULE; 4383 default_rule.cmd_len = 1; 4384 default_rule.set = RESVD_SET; 4385 4386 default_rule.cmd[0].len = 1; 4387 default_rule.cmd[0].opcode = 4388 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4389 1 ? O_ACCEPT : 4390 #endif 4391 O_DENY; 4392 4393 error = add_rule(&layer3_chain, &default_rule); 4394 if (error != 0) { 4395 printf("ipfw2: error %u initializing default rule " 4396 "(support disabled)\n", error); 4397 IPFW_DYN_LOCK_DESTROY(); 4398 IPFW_LOCK_DESTROY(&layer3_chain); 4399 uma_zdestroy(ipfw_dyn_rule_zone); 4400 return (error); 4401 } 4402 4403 ip_fw_default_rule = layer3_chain.rules; 4404 printf("ipfw2 " 4405 #ifdef INET6 4406 "(+ipv6) " 4407 #endif 4408 "initialized, divert %s, " 4409 "rule-based forwarding " 4410 #ifdef IPFIREWALL_FORWARD 4411 "enabled, " 4412 #else 4413 "disabled, " 4414 #endif 4415 "default to %s, logging ", 4416 #ifdef IPDIVERT 4417 "enabled", 4418 #else 4419 "loadable", 4420 #endif 4421 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4422 4423 #ifdef IPFIREWALL_VERBOSE 4424 fw_verbose = 1; 4425 #endif 4426 #ifdef IPFIREWALL_VERBOSE_LIMIT 4427 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4428 #endif 4429 if (fw_verbose == 0) 4430 printf("disabled\n"); 4431 else if (verbose_limit == 0) 4432 printf("unlimited\n"); 4433 else 4434 printf("limited to %d packets/entry by default\n", 4435 verbose_limit); 4436 4437 error = init_tables(&layer3_chain); 4438 if (error) { 4439 IPFW_DYN_LOCK_DESTROY(); 4440 IPFW_LOCK_DESTROY(&layer3_chain); 4441 uma_zdestroy(ipfw_dyn_rule_zone); 4442 return (error); 4443 } 4444 ip_fw_ctl_ptr = ipfw_ctl; 4445 ip_fw_chk_ptr = ipfw_chk; 4446 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4447 4448 return (0); 4449 } 4450 4451 void 4452 ipfw_destroy(void) 4453 { 4454 struct ip_fw *reap; 4455 4456 ip_fw_chk_ptr = NULL; 4457 ip_fw_ctl_ptr = NULL; 4458 callout_drain(&ipfw_timeout); 4459 IPFW_WLOCK(&layer3_chain); 4460 flush_tables(&layer3_chain); 4461 layer3_chain.reap = NULL; 4462 free_chain(&layer3_chain, 1 /* kill default rule */); 4463 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4464 IPFW_WUNLOCK(&layer3_chain); 4465 if (reap != NULL) 4466 reap_rules(reap); 4467 IPFW_DYN_LOCK_DESTROY(); 4468 uma_zdestroy(ipfw_dyn_rule_zone); 4469 IPFW_LOCK_DESTROY(&layer3_chain); 4470 4471 #ifdef INET6 4472 /* Free IPv6 fw sysctl tree. */ 4473 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4474 #endif 4475 4476 printf("IP firewall unloaded\n"); 4477 }
Cache object: 64236273adf349472472b5fd6752f5c3
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]