ip_fw2.c revision 152928
1/*- 2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions 6 * are met: 7 * 1. Redistributions of source code must retain the above copyright 8 * notice, this list of conditions and the following disclaimer. 9 * 2. Redistributions in binary form must reproduce the above copyright 10 * notice, this list of conditions and the following disclaimer in the 11 * documentation and/or other materials provided with the distribution. 12 * 13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23 * SUCH DAMAGE. 24 * 25 * $FreeBSD: head/sys/netinet/ip_fw2.c 152928 2005-11-29 17:56:11Z ume $ 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_ipdn.h" 39#include "opt_inet.h" 40#ifndef INET 41#error IPFIREWALL requires INET. 42#endif /* INET */ 43#endif 44#include "opt_inet6.h" 45#include "opt_ipsec.h" 46 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/condvar.h> 50#include <sys/malloc.h> 51#include <sys/mbuf.h> 52#include <sys/kernel.h> 53#include <sys/jail.h> 54#include <sys/module.h> 55#include <sys/proc.h> 56#include <sys/socket.h> 57#include <sys/socketvar.h> 58#include <sys/sysctl.h> 59#include <sys/syslog.h> 60#include <sys/ucred.h> 61#include <net/if.h> 62#include <net/radix.h> 63#include <net/route.h> 64#include <netinet/in.h> 65#include <netinet/in_systm.h> 66#include <netinet/in_var.h> 67#include <netinet/in_pcb.h> 68#include <netinet/ip.h> 69#include <netinet/ip_var.h> 70#include <netinet/ip_icmp.h> 71#include <netinet/ip_fw.h> 72#include <netinet/ip_divert.h> 73#include <netinet/ip_dummynet.h> 74#include <netinet/tcp.h> 75#include <netinet/tcp_timer.h> 76#include <netinet/tcp_var.h> 77#include <netinet/tcpip.h> 78#include <netinet/udp.h> 79#include <netinet/udp_var.h> 80 81#include <netgraph/ng_ipfw.h> 82 83#include <altq/if_altq.h> 84 85#ifdef IPSEC 86#include <netinet6/ipsec.h> 87#endif 88 89#include <netinet/ip6.h> 90#include <netinet/icmp6.h> 91#ifdef INET6 92#include <netinet6/scope6_var.h> 93#endif 94 95#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 96 97#include <machine/in_cksum.h> /* XXX for in_cksum */ 98 99/* 100 * set_disable contains one bit per set value (0..31). 101 * If the bit is set, all rules with the corresponding set 102 * are disabled. Set RESVD_SET(31) is reserved for the default rule 103 * and rules that are not deleted by the flush command, 104 * and CANNOT be disabled. 105 * Rules in set RESVD_SET can only be deleted explicitly. 106 */ 107static u_int32_t set_disable; 108 109static int fw_verbose; 110static int verbose_limit; 111 112static struct callout ipfw_timeout; 113static uma_zone_t ipfw_dyn_rule_zone; 114#define IPFW_DEFAULT_RULE 65535 115 116/* 117 * Data structure to cache our ucred related 118 * information. This structure only gets used if 119 * the user specified UID/GID based constraints in 120 * a firewall rule. 121 */ 122struct ip_fw_ugid { 123 gid_t fw_groups[NGROUPS]; 124 int fw_ngroups; 125 uid_t fw_uid; 126 int fw_prid; 127}; 128 129struct ip_fw_chain { 130 struct ip_fw *rules; /* list of rules */ 131 struct ip_fw *reap; /* list of rules to reap */ 132 struct mtx mtx; /* lock guarding rule list */ 133 int busy_count; /* busy count for rw locks */ 134 int want_write; 135 struct cv cv; 136}; 137#define IPFW_LOCK_INIT(_chain) \ 138 mtx_init(&(_chain)->mtx, "IPFW static rules", NULL, \ 139 MTX_DEF | MTX_RECURSE) 140#define IPFW_LOCK_DESTROY(_chain) mtx_destroy(&(_chain)->mtx) 141#define IPFW_WLOCK_ASSERT(_chain) do { \ 142 mtx_assert(&(_chain)->mtx, MA_OWNED); \ 143 NET_ASSERT_GIANT(); \ 144} while (0) 145 146static __inline void 147IPFW_RLOCK(struct ip_fw_chain *chain) 148{ 149 mtx_lock(&chain->mtx); 150 chain->busy_count++; 151 mtx_unlock(&chain->mtx); 152} 153 154static __inline void 155IPFW_RUNLOCK(struct ip_fw_chain *chain) 156{ 157 mtx_lock(&chain->mtx); 158 chain->busy_count--; 159 if (chain->busy_count == 0 && chain->want_write) 160 cv_signal(&chain->cv); 161 mtx_unlock(&chain->mtx); 162} 163 164static __inline void 165IPFW_WLOCK(struct ip_fw_chain *chain) 166{ 167 mtx_lock(&chain->mtx); 168 chain->want_write++; 169 while (chain->busy_count > 0) 170 cv_wait(&chain->cv, &chain->mtx); 171} 172 173static __inline void 174IPFW_WUNLOCK(struct ip_fw_chain *chain) 175{ 176 chain->want_write--; 177 cv_signal(&chain->cv); 178 mtx_unlock(&chain->mtx); 179} 180 181/* 182 * list of rules for layer 3 183 */ 184static struct ip_fw_chain layer3_chain; 185 186MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 187MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 188 189struct table_entry { 190 struct radix_node rn[2]; 191 struct sockaddr_in addr, mask; 192 u_int32_t value; 193}; 194 195#define IPFW_TABLES_MAX 128 196static struct ip_fw_table { 197 struct radix_node_head *rnh; 198 int modified; 199 in_addr_t last_addr; 200 int last_match; 201 u_int32_t last_value; 202} ipfw_tables[IPFW_TABLES_MAX]; 203 204static int fw_debug = 1; 205static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 206 207#ifdef SYSCTL_NODE 208SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 209SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 210 CTLFLAG_RW | CTLFLAG_SECURE3, 211 &fw_enable, 0, "Enable ipfw"); 212SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 213 &autoinc_step, 0, "Rule number autincrement step"); 214SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 215 CTLFLAG_RW | CTLFLAG_SECURE3, 216 &fw_one_pass, 0, 217 "Only do a single pass through ipfw when using dummynet(4)"); 218SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 219 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 220SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 221 CTLFLAG_RW | CTLFLAG_SECURE3, 222 &fw_verbose, 0, "Log matches to ipfw rules"); 223SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 224 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 225 226/* 227 * Description of dynamic rules. 228 * 229 * Dynamic rules are stored in lists accessed through a hash table 230 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 231 * be modified through the sysctl variable dyn_buckets which is 232 * updated when the table becomes empty. 233 * 234 * XXX currently there is only one list, ipfw_dyn. 235 * 236 * When a packet is received, its address fields are first masked 237 * with the mask defined for the rule, then hashed, then matched 238 * against the entries in the corresponding list. 239 * Dynamic rules can be used for different purposes: 240 * + stateful rules; 241 * + enforcing limits on the number of sessions; 242 * + in-kernel NAT (not implemented yet) 243 * 244 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 245 * measured in seconds and depending on the flags. 246 * 247 * The total number of dynamic rules is stored in dyn_count. 248 * The max number of dynamic rules is dyn_max. When we reach 249 * the maximum number of rules we do not create anymore. This is 250 * done to avoid consuming too much memory, but also too much 251 * time when searching on each packet (ideally, we should try instead 252 * to put a limit on the length of the list on each bucket...). 253 * 254 * Each dynamic rule holds a pointer to the parent ipfw rule so 255 * we know what action to perform. Dynamic rules are removed when 256 * the parent rule is deleted. XXX we should make them survive. 257 * 258 * There are some limitations with dynamic rules -- we do not 259 * obey the 'randomized match', and we do not do multiple 260 * passes through the firewall. XXX check the latter!!! 261 */ 262static ipfw_dyn_rule **ipfw_dyn_v = NULL; 263static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 264static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 265 266static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 267#define IPFW_DYN_LOCK_INIT() \ 268 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 269#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 270#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 271#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 272#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 273 274/* 275 * Timeouts for various events in handing dynamic rules. 276 */ 277static u_int32_t dyn_ack_lifetime = 300; 278static u_int32_t dyn_syn_lifetime = 20; 279static u_int32_t dyn_fin_lifetime = 1; 280static u_int32_t dyn_rst_lifetime = 1; 281static u_int32_t dyn_udp_lifetime = 10; 282static u_int32_t dyn_short_lifetime = 5; 283 284/* 285 * Keepalives are sent if dyn_keepalive is set. They are sent every 286 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 287 * seconds of lifetime of a rule. 288 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 289 * than dyn_keepalive_period. 290 */ 291 292static u_int32_t dyn_keepalive_interval = 20; 293static u_int32_t dyn_keepalive_period = 5; 294static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 295 296static u_int32_t static_count; /* # of static rules */ 297static u_int32_t static_len; /* size in bytes of static rules */ 298static u_int32_t dyn_count; /* # of dynamic rules */ 299static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 300 301SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 302 &dyn_buckets, 0, "Number of dyn. buckets"); 303SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 304 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 305SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 306 &dyn_count, 0, "Number of dyn. rules"); 307SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 308 &dyn_max, 0, "Max number of dyn. rules"); 309SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 310 &static_count, 0, "Number of static rules"); 311SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 312 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 313SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 314 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 315SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 316 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 317SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 318 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 319SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 320 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 321SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 322 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 323SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 324 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 325 326#ifdef INET6 327/* 328 * IPv6 specific variables 329 */ 330SYSCTL_DECL(_net_inet6_ip6); 331 332static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 333static struct sysctl_oid *ip6_fw_sysctl_tree; 334#endif /* INET6 */ 335#endif /* SYSCTL_NODE */ 336 337static int fw_deny_unknown_exthdrs = 1; 338 339 340/* 341 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 342 * Other macros just cast void * into the appropriate type 343 */ 344#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 345#define TCP(p) ((struct tcphdr *)(p)) 346#define UDP(p) ((struct udphdr *)(p)) 347#define ICMP(p) ((struct icmphdr *)(p)) 348#define ICMP6(p) ((struct icmp6_hdr *)(p)) 349 350static __inline int 351icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 352{ 353 int type = icmp->icmp_type; 354 355 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 356} 357 358#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 359 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 360 361static int 362is_icmp_query(struct icmphdr *icmp) 363{ 364 int type = icmp->icmp_type; 365 366 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 367} 368#undef TT 369 370/* 371 * The following checks use two arrays of 8 or 16 bits to store the 372 * bits that we want set or clear, respectively. They are in the 373 * low and high half of cmd->arg1 or cmd->d[0]. 374 * 375 * We scan options and store the bits we find set. We succeed if 376 * 377 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 378 * 379 * The code is sometimes optimized not to store additional variables. 380 */ 381 382static int 383flags_match(ipfw_insn *cmd, u_int8_t bits) 384{ 385 u_char want_clear; 386 bits = ~bits; 387 388 if ( ((cmd->arg1 & 0xff) & bits) != 0) 389 return 0; /* some bits we want set were clear */ 390 want_clear = (cmd->arg1 >> 8) & 0xff; 391 if ( (want_clear & bits) != want_clear) 392 return 0; /* some bits we want clear were set */ 393 return 1; 394} 395 396static int 397ipopts_match(struct ip *ip, ipfw_insn *cmd) 398{ 399 int optlen, bits = 0; 400 u_char *cp = (u_char *)(ip + 1); 401 int x = (ip->ip_hl << 2) - sizeof (struct ip); 402 403 for (; x > 0; x -= optlen, cp += optlen) { 404 int opt = cp[IPOPT_OPTVAL]; 405 406 if (opt == IPOPT_EOL) 407 break; 408 if (opt == IPOPT_NOP) 409 optlen = 1; 410 else { 411 optlen = cp[IPOPT_OLEN]; 412 if (optlen <= 0 || optlen > x) 413 return 0; /* invalid or truncated */ 414 } 415 switch (opt) { 416 417 default: 418 break; 419 420 case IPOPT_LSRR: 421 bits |= IP_FW_IPOPT_LSRR; 422 break; 423 424 case IPOPT_SSRR: 425 bits |= IP_FW_IPOPT_SSRR; 426 break; 427 428 case IPOPT_RR: 429 bits |= IP_FW_IPOPT_RR; 430 break; 431 432 case IPOPT_TS: 433 bits |= IP_FW_IPOPT_TS; 434 break; 435 } 436 } 437 return (flags_match(cmd, bits)); 438} 439 440static int 441tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 442{ 443 int optlen, bits = 0; 444 u_char *cp = (u_char *)(tcp + 1); 445 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 446 447 for (; x > 0; x -= optlen, cp += optlen) { 448 int opt = cp[0]; 449 if (opt == TCPOPT_EOL) 450 break; 451 if (opt == TCPOPT_NOP) 452 optlen = 1; 453 else { 454 optlen = cp[1]; 455 if (optlen <= 0) 456 break; 457 } 458 459 switch (opt) { 460 461 default: 462 break; 463 464 case TCPOPT_MAXSEG: 465 bits |= IP_FW_TCPOPT_MSS; 466 break; 467 468 case TCPOPT_WINDOW: 469 bits |= IP_FW_TCPOPT_WINDOW; 470 break; 471 472 case TCPOPT_SACK_PERMITTED: 473 case TCPOPT_SACK: 474 bits |= IP_FW_TCPOPT_SACK; 475 break; 476 477 case TCPOPT_TIMESTAMP: 478 bits |= IP_FW_TCPOPT_TS; 479 break; 480 481 } 482 } 483 return (flags_match(cmd, bits)); 484} 485 486static int 487iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 488{ 489 if (ifp == NULL) /* no iface with this packet, match fails */ 490 return 0; 491 /* Check by name or by IP address */ 492 if (cmd->name[0] != '\0') { /* match by name */ 493 /* Check name */ 494 if (cmd->p.glob) { 495 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 496 return(1); 497 } else { 498 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 499 return(1); 500 } 501 } else { 502 struct ifaddr *ia; 503 504 /* XXX lock? */ 505 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 506 if (ia->ifa_addr == NULL) 507 continue; 508 if (ia->ifa_addr->sa_family != AF_INET) 509 continue; 510 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 511 (ia->ifa_addr))->sin_addr.s_addr) 512 return(1); /* match */ 513 } 514 } 515 return(0); /* no match, fail ... */ 516} 517 518/* 519 * The verify_path function checks if a route to the src exists and 520 * if it is reachable via ifp (when provided). 521 * 522 * The 'verrevpath' option checks that the interface that an IP packet 523 * arrives on is the same interface that traffic destined for the 524 * packet's source address would be routed out of. The 'versrcreach' 525 * option just checks that the source address is reachable via any route 526 * (except default) in the routing table. These two are a measure to block 527 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 528 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 529 * is purposely reminiscent of the Cisco IOS command, 530 * 531 * ip verify unicast reverse-path 532 * ip verify unicast source reachable-via any 533 * 534 * which implements the same functionality. But note that syntax is 535 * misleading. The check may be performed on all IP packets whether unicast, 536 * multicast, or broadcast. 537 */ 538static int 539verify_path(struct in_addr src, struct ifnet *ifp) 540{ 541 struct route ro; 542 struct sockaddr_in *dst; 543 544 bzero(&ro, sizeof(ro)); 545 546 dst = (struct sockaddr_in *)&(ro.ro_dst); 547 dst->sin_family = AF_INET; 548 dst->sin_len = sizeof(*dst); 549 dst->sin_addr = src; 550 rtalloc_ign(&ro, RTF_CLONING); 551 552 if (ro.ro_rt == NULL) 553 return 0; 554 555 /* if ifp is provided, check for equality with rtentry */ 556 if (ifp != NULL && ro.ro_rt->rt_ifp != ifp) { 557 RTFREE(ro.ro_rt); 558 return 0; 559 } 560 561 /* if no ifp provided, check if rtentry is not default route */ 562 if (ifp == NULL && 563 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 564 RTFREE(ro.ro_rt); 565 return 0; 566 } 567 568 /* or if this is a blackhole/reject route */ 569 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 570 RTFREE(ro.ro_rt); 571 return 0; 572 } 573 574 /* found valid route */ 575 RTFREE(ro.ro_rt); 576 return 1; 577} 578 579#ifdef INET6 580/* 581 * ipv6 specific rules here... 582 */ 583static __inline int 584icmp6type_match (int type, ipfw_insn_u32 *cmd) 585{ 586 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 587} 588 589static int 590flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 591{ 592 int i; 593 for (i=0; i <= cmd->o.arg1; ++i ) 594 if (curr_flow == cmd->d[i] ) 595 return 1; 596 return 0; 597} 598 599/* support for IP6_*_ME opcodes */ 600static int 601search_ip6_addr_net (struct in6_addr * ip6_addr) 602{ 603 struct ifnet *mdc; 604 struct ifaddr *mdc2; 605 struct in6_ifaddr *fdm; 606 struct in6_addr copia; 607 608 TAILQ_FOREACH(mdc, &ifnet, if_link) 609 for (mdc2 = mdc->if_addrlist.tqh_first; mdc2; 610 mdc2 = mdc2->ifa_list.tqe_next) { 611 if (!mdc2->ifa_addr) 612 continue; 613 if (mdc2->ifa_addr->sa_family == AF_INET6) { 614 fdm = (struct in6_ifaddr *)mdc2; 615 copia = fdm->ia_addr.sin6_addr; 616 /* need for leaving scope_id in the sock_addr */ 617 in6_clearscope(&copia); 618 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 619 return 1; 620 } 621 } 622 return 0; 623} 624 625static int 626verify_path6(struct in6_addr *src, struct ifnet *ifp) 627{ 628 struct route_in6 ro; 629 struct sockaddr_in6 *dst; 630 631 bzero(&ro, sizeof(ro)); 632 633 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 634 dst->sin6_family = AF_INET6; 635 dst->sin6_len = sizeof(*dst); 636 dst->sin6_addr = *src; 637 rtalloc_ign((struct route *)&ro, RTF_CLONING); 638 639 if (ro.ro_rt == NULL) 640 return 0; 641 642 /* 643 * if ifp is provided, check for equality with rtentry 644 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 645 * to support the case of sending packets to an address of our own. 646 * (where the former interface is the first argument of if_simloop() 647 * (=ifp), the latter is lo0) 648 */ 649 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 650 RTFREE(ro.ro_rt); 651 return 0; 652 } 653 654 /* if no ifp provided, check if rtentry is not default route */ 655 if (ifp == NULL && 656 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 657 RTFREE(ro.ro_rt); 658 return 0; 659 } 660 661 /* or if this is a blackhole/reject route */ 662 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 663 RTFREE(ro.ro_rt); 664 return 0; 665 } 666 667 /* found valid route */ 668 RTFREE(ro.ro_rt); 669 return 1; 670 671} 672static __inline int 673hash_packet6(struct ipfw_flow_id *id) 674{ 675 u_int32_t i; 676 i = (id->dst_ip6.__u6_addr.__u6_addr32[0]) ^ 677 (id->dst_ip6.__u6_addr.__u6_addr32[1]) ^ 678 (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 679 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 680 (id->dst_port) ^ (id->src_port) ^ (id->flow_id6); 681 return i; 682} 683 684static int 685is_icmp6_query(int icmp6_type) 686{ 687 if ((icmp6_type <= ICMP6_MAXTYPE) && 688 (icmp6_type == ICMP6_ECHO_REQUEST || 689 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 690 icmp6_type == ICMP6_WRUREQUEST || 691 icmp6_type == ICMP6_FQDN_QUERY || 692 icmp6_type == ICMP6_NI_QUERY)) 693 return (1); 694 695 return (0); 696} 697 698static void 699send_reject6(struct ip_fw_args *args, int code, u_short offset, u_int hlen) 700{ 701 if (code == ICMP6_UNREACH_RST && offset == 0 && 702 args->f_id.proto == IPPROTO_TCP) { 703 struct ip6_hdr *ip6; 704 struct tcphdr *tcp; 705 tcp_seq ack, seq; 706 int flags; 707 struct { 708 struct ip6_hdr ip6; 709 struct tcphdr th; 710 } ti; 711 712 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) { 713 args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr)); 714 if (args->m == NULL) 715 return; 716 } 717 718 ip6 = mtod(args->m, struct ip6_hdr *); 719 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 720 721 if ((tcp->th_flags & TH_RST) != 0) { 722 m_freem(args->m); 723 return; 724 } 725 726 ti.ip6 = *ip6; 727 ti.th = *tcp; 728 ti.th.th_seq = ntohl(ti.th.th_seq); 729 ti.th.th_ack = ntohl(ti.th.th_ack); 730 ti.ip6.ip6_nxt = IPPROTO_TCP; 731 732 if (ti.th.th_flags & TH_ACK) { 733 ack = 0; 734 seq = ti.th.th_ack; 735 flags = TH_RST; 736 } else { 737 ack = ti.th.th_seq; 738 if (((args->m)->m_flags & M_PKTHDR) != 0) { 739 ack += (args->m)->m_pkthdr.len - hlen 740 - (ti.th.th_off << 2); 741 } else if (ip6->ip6_plen) { 742 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) 743 - hlen - (ti.th.th_off << 2); 744 } else { 745 m_freem(args->m); 746 return; 747 } 748 if (tcp->th_flags & TH_SYN) 749 ack++; 750 seq = 0; 751 flags = TH_RST|TH_ACK; 752 } 753 bcopy(&ti, ip6, sizeof(ti)); 754 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1), 755 args->m, ack, seq, flags); 756 757 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 758 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0); 759 760 } else 761 m_freem(args->m); 762 763 args->m = NULL; 764} 765 766#endif /* INET6 */ 767 768static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 769 770#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 771#define SNP(buf) buf, sizeof(buf) 772 773/* 774 * We enter here when we have a rule with O_LOG. 775 * XXX this function alone takes about 2Kbytes of code! 776 */ 777static void 778ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 779 struct mbuf *m, struct ifnet *oif, u_short offset) 780{ 781 struct ether_header *eh = args->eh; 782 char *action; 783 int limit_reached = 0; 784 char action2[40], proto[128], fragment[32]; 785 786 fragment[0] = '\0'; 787 proto[0] = '\0'; 788 789 if (f == NULL) { /* bogus pkt */ 790 if (verbose_limit != 0 && norule_counter >= verbose_limit) 791 return; 792 norule_counter++; 793 if (norule_counter == verbose_limit) 794 limit_reached = verbose_limit; 795 action = "Refuse"; 796 } else { /* O_LOG is the first action, find the real one */ 797 ipfw_insn *cmd = ACTION_PTR(f); 798 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 799 800 if (l->max_log != 0 && l->log_left == 0) 801 return; 802 l->log_left--; 803 if (l->log_left == 0) 804 limit_reached = l->max_log; 805 cmd += F_LEN(cmd); /* point to first action */ 806 if (cmd->opcode == O_ALTQ) { 807 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 808 809 snprintf(SNPARGS(action2, 0), "Altq %d", 810 altq->qid); 811 cmd += F_LEN(cmd); 812 } 813 if (cmd->opcode == O_PROB) 814 cmd += F_LEN(cmd); 815 816 action = action2; 817 switch (cmd->opcode) { 818 case O_DENY: 819 action = "Deny"; 820 break; 821 822 case O_REJECT: 823 if (cmd->arg1==ICMP_REJECT_RST) 824 action = "Reset"; 825 else if (cmd->arg1==ICMP_UNREACH_HOST) 826 action = "Reject"; 827 else 828 snprintf(SNPARGS(action2, 0), "Unreach %d", 829 cmd->arg1); 830 break; 831 832 case O_UNREACH6: 833 if (cmd->arg1==ICMP6_UNREACH_RST) 834 action = "Reset"; 835 else 836 snprintf(SNPARGS(action2, 0), "Unreach %d", 837 cmd->arg1); 838 break; 839 840 case O_ACCEPT: 841 action = "Accept"; 842 break; 843 case O_COUNT: 844 action = "Count"; 845 break; 846 case O_DIVERT: 847 snprintf(SNPARGS(action2, 0), "Divert %d", 848 cmd->arg1); 849 break; 850 case O_TEE: 851 snprintf(SNPARGS(action2, 0), "Tee %d", 852 cmd->arg1); 853 break; 854 case O_SKIPTO: 855 snprintf(SNPARGS(action2, 0), "SkipTo %d", 856 cmd->arg1); 857 break; 858 case O_PIPE: 859 snprintf(SNPARGS(action2, 0), "Pipe %d", 860 cmd->arg1); 861 break; 862 case O_QUEUE: 863 snprintf(SNPARGS(action2, 0), "Queue %d", 864 cmd->arg1); 865 break; 866 case O_FORWARD_IP: { 867 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 868 int len; 869 870 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 871 inet_ntoa(sa->sa.sin_addr)); 872 if (sa->sa.sin_port) 873 snprintf(SNPARGS(action2, len), ":%d", 874 sa->sa.sin_port); 875 } 876 break; 877 case O_NETGRAPH: 878 snprintf(SNPARGS(action2, 0), "Netgraph %d", 879 cmd->arg1); 880 break; 881 case O_NGTEE: 882 snprintf(SNPARGS(action2, 0), "Ngtee %d", 883 cmd->arg1); 884 break; 885 default: 886 action = "UNKNOWN"; 887 break; 888 } 889 } 890 891 if (hlen == 0) { /* non-ip */ 892 snprintf(SNPARGS(proto, 0), "MAC"); 893 894 } else { 895 int len; 896 char src[48], dst[48]; 897 struct icmphdr *icmp; 898 struct tcphdr *tcp; 899 struct udphdr *udp; 900 /* Initialize to make compiler happy. */ 901 struct ip *ip = NULL; 902#ifdef INET6 903 struct ip6_hdr *ip6 = NULL; 904 struct icmp6_hdr *icmp6; 905#endif 906 src[0] = '\0'; 907 dst[0] = '\0'; 908#ifdef INET6 909 if (args->f_id.addr_type == 6) { 910 snprintf(src, sizeof(src), "[%s]", 911 ip6_sprintf(&args->f_id.src_ip6)); 912 snprintf(dst, sizeof(dst), "[%s]", 913 ip6_sprintf(&args->f_id.dst_ip6)); 914 915 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 916 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 917 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 918 } else 919#endif 920 { 921 ip = mtod(m, struct ip *); 922 tcp = L3HDR(struct tcphdr, ip); 923 udp = L3HDR(struct udphdr, ip); 924 925 inet_ntoa_r(ip->ip_src, src); 926 inet_ntoa_r(ip->ip_dst, dst); 927 } 928 929 switch (args->f_id.proto) { 930 case IPPROTO_TCP: 931 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 932 if (offset == 0) 933 snprintf(SNPARGS(proto, len), ":%d %s:%d", 934 ntohs(tcp->th_sport), 935 dst, 936 ntohs(tcp->th_dport)); 937 else 938 snprintf(SNPARGS(proto, len), " %s", dst); 939 break; 940 941 case IPPROTO_UDP: 942 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 943 if (offset == 0) 944 snprintf(SNPARGS(proto, len), ":%d %s:%d", 945 ntohs(udp->uh_sport), 946 dst, 947 ntohs(udp->uh_dport)); 948 else 949 snprintf(SNPARGS(proto, len), " %s", dst); 950 break; 951 952 case IPPROTO_ICMP: 953 icmp = L3HDR(struct icmphdr, ip); 954 if (offset == 0) 955 len = snprintf(SNPARGS(proto, 0), 956 "ICMP:%u.%u ", 957 icmp->icmp_type, icmp->icmp_code); 958 else 959 len = snprintf(SNPARGS(proto, 0), "ICMP "); 960 len += snprintf(SNPARGS(proto, len), "%s", src); 961 snprintf(SNPARGS(proto, len), " %s", dst); 962 break; 963#ifdef INET6 964 case IPPROTO_ICMPV6: 965 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 966 if (offset == 0) 967 len = snprintf(SNPARGS(proto, 0), 968 "ICMPv6:%u.%u ", 969 icmp6->icmp6_type, icmp6->icmp6_code); 970 else 971 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 972 len += snprintf(SNPARGS(proto, len), "%s", src); 973 snprintf(SNPARGS(proto, len), " %s", dst); 974 break; 975#endif 976 default: 977 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 978 args->f_id.proto, src); 979 snprintf(SNPARGS(proto, len), " %s", dst); 980 break; 981 } 982 983#ifdef INET6 984 if (args->f_id.addr_type == 6) { 985 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 986 snprintf(SNPARGS(fragment, 0), 987 " (frag %08x:%d@%d%s)", 988 args->f_id.frag_id6, 989 ntohs(ip6->ip6_plen) - hlen, 990 ntohs(offset & IP6F_OFF_MASK) << 3, 991 (offset & IP6F_MORE_FRAG) ? "+" : ""); 992 } else 993#endif 994 { 995 int ip_off, ip_len; 996 if (eh != NULL) { /* layer 2 packets are as on the wire */ 997 ip_off = ntohs(ip->ip_off); 998 ip_len = ntohs(ip->ip_len); 999 } else { 1000 ip_off = ip->ip_off; 1001 ip_len = ip->ip_len; 1002 } 1003 if (ip_off & (IP_MF | IP_OFFMASK)) 1004 snprintf(SNPARGS(fragment, 0), 1005 " (frag %d:%d@%d%s)", 1006 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1007 offset << 3, 1008 (ip_off & IP_MF) ? "+" : ""); 1009 } 1010 } 1011 if (oif || m->m_pkthdr.rcvif) 1012 log(LOG_SECURITY | LOG_INFO, 1013 "ipfw: %d %s %s %s via %s%s\n", 1014 f ? f->rulenum : -1, 1015 action, proto, oif ? "out" : "in", 1016 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1017 fragment); 1018 else 1019 log(LOG_SECURITY | LOG_INFO, 1020 "ipfw: %d %s %s [no if info]%s\n", 1021 f ? f->rulenum : -1, 1022 action, proto, fragment); 1023 if (limit_reached) 1024 log(LOG_SECURITY | LOG_NOTICE, 1025 "ipfw: limit %d reached on entry %d\n", 1026 limit_reached, f ? f->rulenum : -1); 1027} 1028 1029/* 1030 * IMPORTANT: the hash function for dynamic rules must be commutative 1031 * in source and destination (ip,port), because rules are bidirectional 1032 * and we want to find both in the same bucket. 1033 */ 1034static __inline int 1035hash_packet(struct ipfw_flow_id *id) 1036{ 1037 u_int32_t i; 1038 1039#ifdef INET6 1040 if (IS_IP6_FLOW_ID(id)) 1041 i = hash_packet6(id); 1042 else 1043#endif /* INET6 */ 1044 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1045 i &= (curr_dyn_buckets - 1); 1046 return i; 1047} 1048 1049/** 1050 * unlink a dynamic rule from a chain. prev is a pointer to 1051 * the previous one, q is a pointer to the rule to delete, 1052 * head is a pointer to the head of the queue. 1053 * Modifies q and potentially also head. 1054 */ 1055#define UNLINK_DYN_RULE(prev, head, q) { \ 1056 ipfw_dyn_rule *old_q = q; \ 1057 \ 1058 /* remove a refcount to the parent */ \ 1059 if (q->dyn_type == O_LIMIT) \ 1060 q->parent->count--; \ 1061 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1062 (q->id.src_ip), (q->id.src_port), \ 1063 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1064 if (prev != NULL) \ 1065 prev->next = q = q->next; \ 1066 else \ 1067 head = q = q->next; \ 1068 dyn_count--; \ 1069 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1070 1071#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1072 1073/** 1074 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1075 * 1076 * If keep_me == NULL, rules are deleted even if not expired, 1077 * otherwise only expired rules are removed. 1078 * 1079 * The value of the second parameter is also used to point to identify 1080 * a rule we absolutely do not want to remove (e.g. because we are 1081 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1082 * rules). The pointer is only used for comparison, so any non-null 1083 * value will do. 1084 */ 1085static void 1086remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1087{ 1088 static u_int32_t last_remove = 0; 1089 1090#define FORCE (keep_me == NULL) 1091 1092 ipfw_dyn_rule *prev, *q; 1093 int i, pass = 0, max_pass = 0; 1094 1095 IPFW_DYN_LOCK_ASSERT(); 1096 1097 if (ipfw_dyn_v == NULL || dyn_count == 0) 1098 return; 1099 /* do not expire more than once per second, it is useless */ 1100 if (!FORCE && last_remove == time_uptime) 1101 return; 1102 last_remove = time_uptime; 1103 1104 /* 1105 * because O_LIMIT refer to parent rules, during the first pass only 1106 * remove child and mark any pending LIMIT_PARENT, and remove 1107 * them in a second pass. 1108 */ 1109next_pass: 1110 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1111 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1112 /* 1113 * Logic can become complex here, so we split tests. 1114 */ 1115 if (q == keep_me) 1116 goto next; 1117 if (rule != NULL && rule != q->rule) 1118 goto next; /* not the one we are looking for */ 1119 if (q->dyn_type == O_LIMIT_PARENT) { 1120 /* 1121 * handle parent in the second pass, 1122 * record we need one. 1123 */ 1124 max_pass = 1; 1125 if (pass == 0) 1126 goto next; 1127 if (FORCE && q->count != 0 ) { 1128 /* XXX should not happen! */ 1129 printf("ipfw: OUCH! cannot remove rule," 1130 " count %d\n", q->count); 1131 } 1132 } else { 1133 if (!FORCE && 1134 !TIME_LEQ( q->expire, time_uptime )) 1135 goto next; 1136 } 1137 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1138 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1139 continue; 1140 } 1141next: 1142 prev=q; 1143 q=q->next; 1144 } 1145 } 1146 if (pass++ < max_pass) 1147 goto next_pass; 1148} 1149 1150 1151/** 1152 * lookup a dynamic rule. 1153 */ 1154static ipfw_dyn_rule * 1155lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1156 struct tcphdr *tcp) 1157{ 1158 /* 1159 * stateful ipfw extensions. 1160 * Lookup into dynamic session queue 1161 */ 1162#define MATCH_REVERSE 0 1163#define MATCH_FORWARD 1 1164#define MATCH_NONE 2 1165#define MATCH_UNKNOWN 3 1166 int i, dir = MATCH_NONE; 1167 ipfw_dyn_rule *prev, *q=NULL; 1168 1169 IPFW_DYN_LOCK_ASSERT(); 1170 1171 if (ipfw_dyn_v == NULL) 1172 goto done; /* not found */ 1173 i = hash_packet( pkt ); 1174 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1175 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1176 goto next; 1177 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1178 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1179 continue; 1180 } 1181 if (pkt->proto == q->id.proto && 1182 q->dyn_type != O_LIMIT_PARENT) { 1183 if (IS_IP6_FLOW_ID(pkt)) { 1184 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1185 &(q->id.src_ip6)) && 1186 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1187 &(q->id.dst_ip6)) && 1188 pkt->src_port == q->id.src_port && 1189 pkt->dst_port == q->id.dst_port ) { 1190 dir = MATCH_FORWARD; 1191 break; 1192 } 1193 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1194 &(q->id.dst_ip6)) && 1195 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1196 &(q->id.src_ip6)) && 1197 pkt->src_port == q->id.dst_port && 1198 pkt->dst_port == q->id.src_port ) { 1199 dir = MATCH_REVERSE; 1200 break; 1201 } 1202 } else { 1203 if (pkt->src_ip == q->id.src_ip && 1204 pkt->dst_ip == q->id.dst_ip && 1205 pkt->src_port == q->id.src_port && 1206 pkt->dst_port == q->id.dst_port ) { 1207 dir = MATCH_FORWARD; 1208 break; 1209 } 1210 if (pkt->src_ip == q->id.dst_ip && 1211 pkt->dst_ip == q->id.src_ip && 1212 pkt->src_port == q->id.dst_port && 1213 pkt->dst_port == q->id.src_port ) { 1214 dir = MATCH_REVERSE; 1215 break; 1216 } 1217 } 1218 } 1219next: 1220 prev = q; 1221 q = q->next; 1222 } 1223 if (q == NULL) 1224 goto done; /* q = NULL, not found */ 1225 1226 if ( prev != NULL) { /* found and not in front */ 1227 prev->next = q->next; 1228 q->next = ipfw_dyn_v[i]; 1229 ipfw_dyn_v[i] = q; 1230 } 1231 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1232 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1233 1234#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1235#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1236 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1237 switch (q->state) { 1238 case TH_SYN: /* opening */ 1239 q->expire = time_uptime + dyn_syn_lifetime; 1240 break; 1241 1242 case BOTH_SYN: /* move to established */ 1243 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1244 case BOTH_SYN | (TH_FIN << 8) : 1245 if (tcp) { 1246#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1247 u_int32_t ack = ntohl(tcp->th_ack); 1248 if (dir == MATCH_FORWARD) { 1249 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1250 q->ack_fwd = ack; 1251 else { /* ignore out-of-sequence */ 1252 break; 1253 } 1254 } else { 1255 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1256 q->ack_rev = ack; 1257 else { /* ignore out-of-sequence */ 1258 break; 1259 } 1260 } 1261 } 1262 q->expire = time_uptime + dyn_ack_lifetime; 1263 break; 1264 1265 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1266 if (dyn_fin_lifetime >= dyn_keepalive_period) 1267 dyn_fin_lifetime = dyn_keepalive_period - 1; 1268 q->expire = time_uptime + dyn_fin_lifetime; 1269 break; 1270 1271 default: 1272#if 0 1273 /* 1274 * reset or some invalid combination, but can also 1275 * occur if we use keep-state the wrong way. 1276 */ 1277 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1278 printf("invalid state: 0x%x\n", q->state); 1279#endif 1280 if (dyn_rst_lifetime >= dyn_keepalive_period) 1281 dyn_rst_lifetime = dyn_keepalive_period - 1; 1282 q->expire = time_uptime + dyn_rst_lifetime; 1283 break; 1284 } 1285 } else if (pkt->proto == IPPROTO_UDP) { 1286 q->expire = time_uptime + dyn_udp_lifetime; 1287 } else { 1288 /* other protocols */ 1289 q->expire = time_uptime + dyn_short_lifetime; 1290 } 1291done: 1292 if (match_direction) 1293 *match_direction = dir; 1294 return q; 1295} 1296 1297static ipfw_dyn_rule * 1298lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1299 struct tcphdr *tcp) 1300{ 1301 ipfw_dyn_rule *q; 1302 1303 IPFW_DYN_LOCK(); 1304 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1305 if (q == NULL) 1306 IPFW_DYN_UNLOCK(); 1307 /* NB: return table locked when q is not NULL */ 1308 return q; 1309} 1310 1311static void 1312realloc_dynamic_table(void) 1313{ 1314 IPFW_DYN_LOCK_ASSERT(); 1315 1316 /* 1317 * Try reallocation, make sure we have a power of 2 and do 1318 * not allow more than 64k entries. In case of overflow, 1319 * default to 1024. 1320 */ 1321 1322 if (dyn_buckets > 65536) 1323 dyn_buckets = 1024; 1324 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1325 dyn_buckets = curr_dyn_buckets; /* reset */ 1326 return; 1327 } 1328 curr_dyn_buckets = dyn_buckets; 1329 if (ipfw_dyn_v != NULL) 1330 free(ipfw_dyn_v, M_IPFW); 1331 for (;;) { 1332 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1333 M_IPFW, M_NOWAIT | M_ZERO); 1334 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1335 break; 1336 curr_dyn_buckets /= 2; 1337 } 1338} 1339 1340/** 1341 * Install state of type 'type' for a dynamic session. 1342 * The hash table contains two type of rules: 1343 * - regular rules (O_KEEP_STATE) 1344 * - rules for sessions with limited number of sess per user 1345 * (O_LIMIT). When they are created, the parent is 1346 * increased by 1, and decreased on delete. In this case, 1347 * the third parameter is the parent rule and not the chain. 1348 * - "parent" rules for the above (O_LIMIT_PARENT). 1349 */ 1350static ipfw_dyn_rule * 1351add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1352{ 1353 ipfw_dyn_rule *r; 1354 int i; 1355 1356 IPFW_DYN_LOCK_ASSERT(); 1357 1358 if (ipfw_dyn_v == NULL || 1359 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1360 realloc_dynamic_table(); 1361 if (ipfw_dyn_v == NULL) 1362 return NULL; /* failed ! */ 1363 } 1364 i = hash_packet(id); 1365 1366 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1367 if (r == NULL) { 1368 printf ("ipfw: sorry cannot allocate state\n"); 1369 return NULL; 1370 } 1371 1372 /* increase refcount on parent, and set pointer */ 1373 if (dyn_type == O_LIMIT) { 1374 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1375 if ( parent->dyn_type != O_LIMIT_PARENT) 1376 panic("invalid parent"); 1377 parent->count++; 1378 r->parent = parent; 1379 rule = parent->rule; 1380 } 1381 1382 r->id = *id; 1383 r->expire = time_uptime + dyn_syn_lifetime; 1384 r->rule = rule; 1385 r->dyn_type = dyn_type; 1386 r->pcnt = r->bcnt = 0; 1387 r->count = 0; 1388 1389 r->bucket = i; 1390 r->next = ipfw_dyn_v[i]; 1391 ipfw_dyn_v[i] = r; 1392 dyn_count++; 1393 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1394 dyn_type, 1395 (r->id.src_ip), (r->id.src_port), 1396 (r->id.dst_ip), (r->id.dst_port), 1397 dyn_count ); ) 1398 return r; 1399} 1400 1401/** 1402 * lookup dynamic parent rule using pkt and rule as search keys. 1403 * If the lookup fails, then install one. 1404 */ 1405static ipfw_dyn_rule * 1406lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1407{ 1408 ipfw_dyn_rule *q; 1409 int i; 1410 1411 IPFW_DYN_LOCK_ASSERT(); 1412 1413 if (ipfw_dyn_v) { 1414 int is_v6 = IS_IP6_FLOW_ID(pkt); 1415 i = hash_packet( pkt ); 1416 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1417 if (q->dyn_type == O_LIMIT_PARENT && 1418 rule== q->rule && 1419 pkt->proto == q->id.proto && 1420 pkt->src_port == q->id.src_port && 1421 pkt->dst_port == q->id.dst_port && 1422 ( 1423 (is_v6 && 1424 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1425 &(q->id.src_ip6)) && 1426 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1427 &(q->id.dst_ip6))) || 1428 (!is_v6 && 1429 pkt->src_ip == q->id.src_ip && 1430 pkt->dst_ip == q->id.dst_ip) 1431 ) 1432 ) { 1433 q->expire = time_uptime + dyn_short_lifetime; 1434 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1435 return q; 1436 } 1437 } 1438 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1439} 1440 1441/** 1442 * Install dynamic state for rule type cmd->o.opcode 1443 * 1444 * Returns 1 (failure) if state is not installed because of errors or because 1445 * session limitations are enforced. 1446 */ 1447static int 1448install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1449 struct ip_fw_args *args) 1450{ 1451 static int last_log; 1452 1453 ipfw_dyn_rule *q; 1454 1455 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 1456 cmd->o.opcode, 1457 (args->f_id.src_ip), (args->f_id.src_port), 1458 (args->f_id.dst_ip), (args->f_id.dst_port) );) 1459 1460 IPFW_DYN_LOCK(); 1461 1462 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1463 1464 if (q != NULL) { /* should never occur */ 1465 if (last_log != time_uptime) { 1466 last_log = time_uptime; 1467 printf("ipfw: install_state: entry already present, done\n"); 1468 } 1469 IPFW_DYN_UNLOCK(); 1470 return 0; 1471 } 1472 1473 if (dyn_count >= dyn_max) 1474 /* 1475 * Run out of slots, try to remove any expired rule. 1476 */ 1477 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1478 1479 if (dyn_count >= dyn_max) { 1480 if (last_log != time_uptime) { 1481 last_log = time_uptime; 1482 printf("ipfw: install_state: Too many dynamic rules\n"); 1483 } 1484 IPFW_DYN_UNLOCK(); 1485 return 1; /* cannot install, notify caller */ 1486 } 1487 1488 switch (cmd->o.opcode) { 1489 case O_KEEP_STATE: /* bidir rule */ 1490 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1491 break; 1492 1493 case O_LIMIT: /* limit number of sessions */ 1494 { 1495 u_int16_t limit_mask = cmd->limit_mask; 1496 struct ipfw_flow_id id; 1497 ipfw_dyn_rule *parent; 1498 1499 DEB(printf("ipfw: installing dyn-limit rule %d\n", 1500 cmd->conn_limit);) 1501 1502 id.dst_ip = id.src_ip = 0; 1503 id.dst_port = id.src_port = 0; 1504 id.proto = args->f_id.proto; 1505 1506 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1507 if (limit_mask & DYN_SRC_ADDR) 1508 id.src_ip6 = args->f_id.src_ip6; 1509 if (limit_mask & DYN_DST_ADDR) 1510 id.dst_ip6 = args->f_id.dst_ip6; 1511 } else { 1512 if (limit_mask & DYN_SRC_ADDR) 1513 id.src_ip = args->f_id.src_ip; 1514 if (limit_mask & DYN_DST_ADDR) 1515 id.dst_ip = args->f_id.dst_ip; 1516 } 1517 if (limit_mask & DYN_SRC_PORT) 1518 id.src_port = args->f_id.src_port; 1519 if (limit_mask & DYN_DST_PORT) 1520 id.dst_port = args->f_id.dst_port; 1521 parent = lookup_dyn_parent(&id, rule); 1522 if (parent == NULL) { 1523 printf("ipfw: add parent failed\n"); 1524 IPFW_DYN_UNLOCK(); 1525 return 1; 1526 } 1527 if (parent->count >= cmd->conn_limit) { 1528 /* 1529 * See if we can remove some expired rule. 1530 */ 1531 remove_dyn_rule(rule, parent); 1532 if (parent->count >= cmd->conn_limit) { 1533 if (fw_verbose && last_log != time_uptime) { 1534 last_log = time_uptime; 1535 log(LOG_SECURITY | LOG_DEBUG, 1536 "drop session, too many entries\n"); 1537 } 1538 IPFW_DYN_UNLOCK(); 1539 return 1; 1540 } 1541 } 1542 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1543 } 1544 break; 1545 default: 1546 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 1547 IPFW_DYN_UNLOCK(); 1548 return 1; 1549 } 1550 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1551 IPFW_DYN_UNLOCK(); 1552 return 0; 1553} 1554 1555/* 1556 * Generate a TCP packet, containing either a RST or a keepalive. 1557 * When flags & TH_RST, we are sending a RST packet, because of a 1558 * "reset" action matched the packet. 1559 * Otherwise we are sending a keepalive, and flags & TH_ 1560 */ 1561static struct mbuf * 1562send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1563{ 1564 struct mbuf *m; 1565 struct ip *ip; 1566 struct tcphdr *tcp; 1567 1568 MGETHDR(m, M_DONTWAIT, MT_DATA); 1569 if (m == 0) 1570 return (NULL); 1571 m->m_pkthdr.rcvif = (struct ifnet *)0; 1572 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1573 m->m_data += max_linkhdr; 1574 1575 ip = mtod(m, struct ip *); 1576 bzero(ip, m->m_len); 1577 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1578 ip->ip_p = IPPROTO_TCP; 1579 tcp->th_off = 5; 1580 /* 1581 * Assume we are sending a RST (or a keepalive in the reverse 1582 * direction), swap src and destination addresses and ports. 1583 */ 1584 ip->ip_src.s_addr = htonl(id->dst_ip); 1585 ip->ip_dst.s_addr = htonl(id->src_ip); 1586 tcp->th_sport = htons(id->dst_port); 1587 tcp->th_dport = htons(id->src_port); 1588 if (flags & TH_RST) { /* we are sending a RST */ 1589 if (flags & TH_ACK) { 1590 tcp->th_seq = htonl(ack); 1591 tcp->th_ack = htonl(0); 1592 tcp->th_flags = TH_RST; 1593 } else { 1594 if (flags & TH_SYN) 1595 seq++; 1596 tcp->th_seq = htonl(0); 1597 tcp->th_ack = htonl(seq); 1598 tcp->th_flags = TH_RST | TH_ACK; 1599 } 1600 } else { 1601 /* 1602 * We are sending a keepalive. flags & TH_SYN determines 1603 * the direction, forward if set, reverse if clear. 1604 * NOTE: seq and ack are always assumed to be correct 1605 * as set by the caller. This may be confusing... 1606 */ 1607 if (flags & TH_SYN) { 1608 /* 1609 * we have to rewrite the correct addresses! 1610 */ 1611 ip->ip_dst.s_addr = htonl(id->dst_ip); 1612 ip->ip_src.s_addr = htonl(id->src_ip); 1613 tcp->th_dport = htons(id->dst_port); 1614 tcp->th_sport = htons(id->src_port); 1615 } 1616 tcp->th_seq = htonl(seq); 1617 tcp->th_ack = htonl(ack); 1618 tcp->th_flags = TH_ACK; 1619 } 1620 /* 1621 * set ip_len to the payload size so we can compute 1622 * the tcp checksum on the pseudoheader 1623 * XXX check this, could save a couple of words ? 1624 */ 1625 ip->ip_len = htons(sizeof(struct tcphdr)); 1626 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1627 /* 1628 * now fill fields left out earlier 1629 */ 1630 ip->ip_ttl = ip_defttl; 1631 ip->ip_len = m->m_pkthdr.len; 1632 m->m_flags |= M_SKIP_FIREWALL; 1633 return (m); 1634} 1635 1636/* 1637 * sends a reject message, consuming the mbuf passed as an argument. 1638 */ 1639static void 1640send_reject(struct ip_fw_args *args, int code, u_short offset, int ip_len) 1641{ 1642 1643 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1644 /* We need the IP header in host order for icmp_error(). */ 1645 if (args->eh != NULL) { 1646 struct ip *ip = mtod(args->m, struct ip *); 1647 ip->ip_len = ntohs(ip->ip_len); 1648 ip->ip_off = ntohs(ip->ip_off); 1649 } 1650 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1651 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1652 struct tcphdr *const tcp = 1653 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1654 if ( (tcp->th_flags & TH_RST) == 0) { 1655 struct mbuf *m; 1656 m = send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1657 ntohl(tcp->th_ack), 1658 tcp->th_flags | TH_RST); 1659 if (m != NULL) 1660 ip_output(m, NULL, NULL, 0, NULL, NULL); 1661 } 1662 m_freem(args->m); 1663 } else 1664 m_freem(args->m); 1665 args->m = NULL; 1666} 1667 1668/** 1669 * 1670 * Given an ip_fw *, lookup_next_rule will return a pointer 1671 * to the next rule, which can be either the jump 1672 * target (for skipto instructions) or the next one in the list (in 1673 * all other cases including a missing jump target). 1674 * The result is also written in the "next_rule" field of the rule. 1675 * Backward jumps are not allowed, so start looking from the next 1676 * rule... 1677 * 1678 * This never returns NULL -- in case we do not have an exact match, 1679 * the next rule is returned. When the ruleset is changed, 1680 * pointers are flushed so we are always correct. 1681 */ 1682 1683static struct ip_fw * 1684lookup_next_rule(struct ip_fw *me) 1685{ 1686 struct ip_fw *rule = NULL; 1687 ipfw_insn *cmd; 1688 1689 /* look for action, in case it is a skipto */ 1690 cmd = ACTION_PTR(me); 1691 if (cmd->opcode == O_LOG) 1692 cmd += F_LEN(cmd); 1693 if (cmd->opcode == O_ALTQ) 1694 cmd += F_LEN(cmd); 1695 if ( cmd->opcode == O_SKIPTO ) 1696 for (rule = me->next; rule ; rule = rule->next) 1697 if (rule->rulenum >= cmd->arg1) 1698 break; 1699 if (rule == NULL) /* failure or not a skipto */ 1700 rule = me->next; 1701 me->next_rule = rule; 1702 return rule; 1703} 1704 1705static void 1706init_tables(void) 1707{ 1708 int i; 1709 1710 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1711 rn_inithead((void **)&ipfw_tables[i].rnh, 32); 1712 ipfw_tables[i].modified = 1; 1713 } 1714} 1715 1716static int 1717add_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen, u_int32_t value) 1718{ 1719 struct radix_node_head *rnh; 1720 struct table_entry *ent; 1721 1722 if (tbl >= IPFW_TABLES_MAX) 1723 return (EINVAL); 1724 rnh = ipfw_tables[tbl].rnh; 1725 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1726 if (ent == NULL) 1727 return (ENOMEM); 1728 ent->value = value; 1729 ent->addr.sin_len = ent->mask.sin_len = 8; 1730 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1731 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1732 RADIX_NODE_HEAD_LOCK(rnh); 1733 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1734 NULL) { 1735 RADIX_NODE_HEAD_UNLOCK(rnh); 1736 free(ent, M_IPFW_TBL); 1737 return (EEXIST); 1738 } 1739 ipfw_tables[tbl].modified = 1; 1740 RADIX_NODE_HEAD_UNLOCK(rnh); 1741 return (0); 1742} 1743 1744static int 1745del_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen) 1746{ 1747 struct radix_node_head *rnh; 1748 struct table_entry *ent; 1749 struct sockaddr_in sa, mask; 1750 1751 if (tbl >= IPFW_TABLES_MAX) 1752 return (EINVAL); 1753 rnh = ipfw_tables[tbl].rnh; 1754 sa.sin_len = mask.sin_len = 8; 1755 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1756 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1757 RADIX_NODE_HEAD_LOCK(rnh); 1758 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1759 if (ent == NULL) { 1760 RADIX_NODE_HEAD_UNLOCK(rnh); 1761 return (ESRCH); 1762 } 1763 ipfw_tables[tbl].modified = 1; 1764 RADIX_NODE_HEAD_UNLOCK(rnh); 1765 free(ent, M_IPFW_TBL); 1766 return (0); 1767} 1768 1769static int 1770flush_table_entry(struct radix_node *rn, void *arg) 1771{ 1772 struct radix_node_head * const rnh = arg; 1773 struct table_entry *ent; 1774 1775 ent = (struct table_entry *) 1776 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1777 if (ent != NULL) 1778 free(ent, M_IPFW_TBL); 1779 return (0); 1780} 1781 1782static int 1783flush_table(u_int16_t tbl) 1784{ 1785 struct radix_node_head *rnh; 1786 1787 if (tbl >= IPFW_TABLES_MAX) 1788 return (EINVAL); 1789 rnh = ipfw_tables[tbl].rnh; 1790 RADIX_NODE_HEAD_LOCK(rnh); 1791 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1792 ipfw_tables[tbl].modified = 1; 1793 RADIX_NODE_HEAD_UNLOCK(rnh); 1794 return (0); 1795} 1796 1797static void 1798flush_tables(void) 1799{ 1800 u_int16_t tbl; 1801 1802 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1803 flush_table(tbl); 1804} 1805 1806static int 1807lookup_table(u_int16_t tbl, in_addr_t addr, u_int32_t *val) 1808{ 1809 struct radix_node_head *rnh; 1810 struct ip_fw_table *table; 1811 struct table_entry *ent; 1812 struct sockaddr_in sa; 1813 int last_match; 1814 1815 if (tbl >= IPFW_TABLES_MAX) 1816 return (0); 1817 table = &ipfw_tables[tbl]; 1818 rnh = table->rnh; 1819 RADIX_NODE_HEAD_LOCK(rnh); 1820 if (addr == table->last_addr && !table->modified) { 1821 last_match = table->last_match; 1822 if (last_match) 1823 *val = table->last_value; 1824 RADIX_NODE_HEAD_UNLOCK(rnh); 1825 return (last_match); 1826 } 1827 table->modified = 0; 1828 sa.sin_len = 8; 1829 sa.sin_addr.s_addr = addr; 1830 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1831 table->last_addr = addr; 1832 if (ent != NULL) { 1833 table->last_value = *val = ent->value; 1834 table->last_match = 1; 1835 RADIX_NODE_HEAD_UNLOCK(rnh); 1836 return (1); 1837 } 1838 table->last_match = 0; 1839 RADIX_NODE_HEAD_UNLOCK(rnh); 1840 return (0); 1841} 1842 1843static int 1844count_table_entry(struct radix_node *rn, void *arg) 1845{ 1846 u_int32_t * const cnt = arg; 1847 1848 (*cnt)++; 1849 return (0); 1850} 1851 1852static int 1853count_table(u_int32_t tbl, u_int32_t *cnt) 1854{ 1855 struct radix_node_head *rnh; 1856 1857 if (tbl >= IPFW_TABLES_MAX) 1858 return (EINVAL); 1859 rnh = ipfw_tables[tbl].rnh; 1860 *cnt = 0; 1861 RADIX_NODE_HEAD_LOCK(rnh); 1862 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1863 RADIX_NODE_HEAD_UNLOCK(rnh); 1864 return (0); 1865} 1866 1867static int 1868dump_table_entry(struct radix_node *rn, void *arg) 1869{ 1870 struct table_entry * const n = (struct table_entry *)rn; 1871 ipfw_table * const tbl = arg; 1872 ipfw_table_entry *ent; 1873 1874 if (tbl->cnt == tbl->size) 1875 return (1); 1876 ent = &tbl->ent[tbl->cnt]; 1877 ent->tbl = tbl->tbl; 1878 if (in_nullhost(n->mask.sin_addr)) 1879 ent->masklen = 0; 1880 else 1881 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1882 ent->addr = n->addr.sin_addr.s_addr; 1883 ent->value = n->value; 1884 tbl->cnt++; 1885 return (0); 1886} 1887 1888static int 1889dump_table(ipfw_table *tbl) 1890{ 1891 struct radix_node_head *rnh; 1892 1893 if (tbl->tbl >= IPFW_TABLES_MAX) 1894 return (EINVAL); 1895 rnh = ipfw_tables[tbl->tbl].rnh; 1896 tbl->cnt = 0; 1897 RADIX_NODE_HEAD_LOCK(rnh); 1898 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1899 RADIX_NODE_HEAD_UNLOCK(rnh); 1900 return (0); 1901} 1902 1903static void 1904fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1905{ 1906 struct ucred *cr; 1907 1908 if (inp->inp_socket != NULL) { 1909 cr = inp->inp_socket->so_cred; 1910 ugp->fw_prid = jailed(cr) ? 1911 cr->cr_prison->pr_id : -1; 1912 ugp->fw_uid = cr->cr_uid; 1913 ugp->fw_ngroups = cr->cr_ngroups; 1914 bcopy(cr->cr_groups, ugp->fw_groups, 1915 sizeof(ugp->fw_groups)); 1916 } 1917} 1918 1919static int 1920check_uidgid(ipfw_insn_u32 *insn, 1921 int proto, struct ifnet *oif, 1922 struct in_addr dst_ip, u_int16_t dst_port, 1923 struct in_addr src_ip, u_int16_t src_port, 1924 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1925{ 1926 struct inpcbinfo *pi; 1927 int wildcard; 1928 struct inpcb *pcb; 1929 int match; 1930 gid_t *gp; 1931 1932 /* 1933 * Check to see if the UDP or TCP stack supplied us with 1934 * the PCB. If so, rather then holding a lock and looking 1935 * up the PCB, we can use the one that was supplied. 1936 */ 1937 if (inp && *lookup == 0) { 1938 INP_LOCK_ASSERT(inp); 1939 if (inp->inp_socket != NULL) { 1940 fill_ugid_cache(inp, ugp); 1941 *lookup = 1; 1942 } 1943 } 1944 /* 1945 * If we have already been here and the packet has no 1946 * PCB entry associated with it, then we can safely 1947 * assume that this is a no match. 1948 */ 1949 if (*lookup == -1) 1950 return (0); 1951 if (proto == IPPROTO_TCP) { 1952 wildcard = 0; 1953 pi = &tcbinfo; 1954 } else if (proto == IPPROTO_UDP) { 1955 wildcard = 1; 1956 pi = &udbinfo; 1957 } else 1958 return 0; 1959 match = 0; 1960 if (*lookup == 0) { 1961 INP_INFO_RLOCK(pi); 1962 pcb = (oif) ? 1963 in_pcblookup_hash(pi, 1964 dst_ip, htons(dst_port), 1965 src_ip, htons(src_port), 1966 wildcard, oif) : 1967 in_pcblookup_hash(pi, 1968 src_ip, htons(src_port), 1969 dst_ip, htons(dst_port), 1970 wildcard, NULL); 1971 if (pcb != NULL) { 1972 INP_LOCK(pcb); 1973 if (pcb->inp_socket != NULL) { 1974 fill_ugid_cache(pcb, ugp); 1975 *lookup = 1; 1976 } 1977 INP_UNLOCK(pcb); 1978 } 1979 INP_INFO_RUNLOCK(pi); 1980 if (*lookup == 0) { 1981 /* 1982 * If the lookup did not yield any results, there 1983 * is no sense in coming back and trying again. So 1984 * we can set lookup to -1 and ensure that we wont 1985 * bother the pcb system again. 1986 */ 1987 *lookup = -1; 1988 return (0); 1989 } 1990 } 1991 if (insn->o.opcode == O_UID) 1992 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1993 else if (insn->o.opcode == O_GID) { 1994 for (gp = ugp->fw_groups; 1995 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1996 if (*gp == (gid_t)insn->d[0]) { 1997 match = 1; 1998 break; 1999 } 2000 } else if (insn->o.opcode == O_JAIL) 2001 match = (ugp->fw_prid == (int)insn->d[0]); 2002 return match; 2003} 2004 2005/* 2006 * The main check routine for the firewall. 2007 * 2008 * All arguments are in args so we can modify them and return them 2009 * back to the caller. 2010 * 2011 * Parameters: 2012 * 2013 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2014 * Starts with the IP header. 2015 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2016 * args->oif Outgoing interface, or NULL if packet is incoming. 2017 * The incoming interface is in the mbuf. (in) 2018 * args->divert_rule (in/out) 2019 * Skip up to the first rule past this rule number; 2020 * upon return, non-zero port number for divert or tee. 2021 * 2022 * args->rule Pointer to the last matching rule (in/out) 2023 * args->next_hop Socket we are forwarding to (out). 2024 * args->f_id Addresses grabbed from the packet (out) 2025 * args->cookie a cookie depending on rule action 2026 * 2027 * Return value: 2028 * 2029 * IP_FW_PASS the packet must be accepted 2030 * IP_FW_DENY the packet must be dropped 2031 * IP_FW_DIVERT divert packet, port in m_tag 2032 * IP_FW_TEE tee packet, port in m_tag 2033 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2034 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2035 * 2036 */ 2037 2038int 2039ipfw_chk(struct ip_fw_args *args) 2040{ 2041 /* 2042 * Local variables hold state during the processing of a packet. 2043 * 2044 * IMPORTANT NOTE: to speed up the processing of rules, there 2045 * are some assumption on the values of the variables, which 2046 * are documented here. Should you change them, please check 2047 * the implementation of the various instructions to make sure 2048 * that they still work. 2049 * 2050 * args->eh The MAC header. It is non-null for a layer2 2051 * packet, it is NULL for a layer-3 packet. 2052 * 2053 * m | args->m Pointer to the mbuf, as received from the caller. 2054 * It may change if ipfw_chk() does an m_pullup, or if it 2055 * consumes the packet because it calls send_reject(). 2056 * XXX This has to change, so that ipfw_chk() never modifies 2057 * or consumes the buffer. 2058 * ip is simply an alias of the value of m, and it is kept 2059 * in sync with it (the packet is supposed to start with 2060 * the ip header). 2061 */ 2062 struct mbuf *m = args->m; 2063 struct ip *ip = mtod(m, struct ip *); 2064 2065 /* 2066 * For rules which contain uid/gid or jail constraints, cache 2067 * a copy of the users credentials after the pcb lookup has been 2068 * executed. This will speed up the processing of rules with 2069 * these types of constraints, as well as decrease contention 2070 * on pcb related locks. 2071 */ 2072 struct ip_fw_ugid fw_ugid_cache; 2073 int ugid_lookup = 0; 2074 2075 /* 2076 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2077 * associated with a packet input on a divert socket. This 2078 * will allow to distinguish traffic and its direction when 2079 * it originates from a divert socket. 2080 */ 2081 u_int divinput_flags = 0; 2082 2083 /* 2084 * oif | args->oif If NULL, ipfw_chk has been called on the 2085 * inbound path (ether_input, ip_input). 2086 * If non-NULL, ipfw_chk has been called on the outbound path 2087 * (ether_output, ip_output). 2088 */ 2089 struct ifnet *oif = args->oif; 2090 2091 struct ip_fw *f = NULL; /* matching rule */ 2092 int retval = 0; 2093 2094 /* 2095 * hlen The length of the IP header. 2096 */ 2097 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2098 2099 /* 2100 * offset The offset of a fragment. offset != 0 means that 2101 * we have a fragment at this offset of an IPv4 packet. 2102 * offset == 0 means that (if this is an IPv4 packet) 2103 * this is the first or only fragment. 2104 * For IPv6 offset == 0 means there is no Fragment Header. 2105 * If offset != 0 for IPv6 always use correct mask to 2106 * get the correct offset because we add IP6F_MORE_FRAG 2107 * to be able to dectect the first fragment which would 2108 * otherwise have offset = 0. 2109 */ 2110 u_short offset = 0; 2111 2112 /* 2113 * Local copies of addresses. They are only valid if we have 2114 * an IP packet. 2115 * 2116 * proto The protocol. Set to 0 for non-ip packets, 2117 * or to the protocol read from the packet otherwise. 2118 * proto != 0 means that we have an IPv4 packet. 2119 * 2120 * src_port, dst_port port numbers, in HOST format. Only 2121 * valid for TCP and UDP packets. 2122 * 2123 * src_ip, dst_ip ip addresses, in NETWORK format. 2124 * Only valid for IPv4 packets. 2125 */ 2126 u_int8_t proto; 2127 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2128 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2129 u_int16_t ip_len=0; 2130 int pktlen; 2131 2132 /* 2133 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2134 * MATCH_NONE when checked and not matched (q = NULL), 2135 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2136 */ 2137 int dyn_dir = MATCH_UNKNOWN; 2138 ipfw_dyn_rule *q = NULL; 2139 struct ip_fw_chain *chain = &layer3_chain; 2140 struct m_tag *mtag; 2141 2142 /* 2143 * We store in ulp a pointer to the upper layer protocol header. 2144 * In the ipv4 case this is easy to determine from the header, 2145 * but for ipv6 we might have some additional headers in the middle. 2146 * ulp is NULL if not found. 2147 */ 2148 void *ulp = NULL; /* upper layer protocol pointer. */ 2149 /* XXX ipv6 variables */ 2150 int is_ipv6 = 0; 2151 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2152 /* end of ipv6 variables */ 2153 int is_ipv4 = 0; 2154 2155 if (m->m_flags & M_SKIP_FIREWALL) 2156 return (IP_FW_PASS); /* accept */ 2157 2158 pktlen = m->m_pkthdr.len; 2159 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2160 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2161 2162/* 2163 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2164 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2165 * pointer might become stale after other pullups (but we never use it 2166 * this way). 2167 */ 2168#define PULLUP_TO(len, p, T) \ 2169do { \ 2170 int x = (len) + sizeof(T); \ 2171 if ((m)->m_len < x) { \ 2172 args->m = m = m_pullup(m, x); \ 2173 if (m == NULL) \ 2174 goto pullup_failed; \ 2175 } \ 2176 p = (mtod(m, char *) + (len)); \ 2177} while (0) 2178 2179 /* Identify IP packets and fill up variables. */ 2180 if (pktlen >= sizeof(struct ip6_hdr) && 2181 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2182 mtod(m, struct ip *)->ip_v == 6) { 2183 is_ipv6 = 1; 2184 args->f_id.addr_type = 6; 2185 hlen = sizeof(struct ip6_hdr); 2186 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2187 2188 /* Search extension headers to find upper layer protocols */ 2189 while (ulp == NULL) { 2190 switch (proto) { 2191 case IPPROTO_ICMPV6: 2192 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2193 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2194 break; 2195 2196 case IPPROTO_TCP: 2197 PULLUP_TO(hlen, ulp, struct tcphdr); 2198 dst_port = TCP(ulp)->th_dport; 2199 src_port = TCP(ulp)->th_sport; 2200 args->f_id.flags = TCP(ulp)->th_flags; 2201 break; 2202 2203 case IPPROTO_UDP: 2204 PULLUP_TO(hlen, ulp, struct udphdr); 2205 dst_port = UDP(ulp)->uh_dport; 2206 src_port = UDP(ulp)->uh_sport; 2207 break; 2208 2209 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2210 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2211 ext_hd |= EXT_HOPOPTS; 2212 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2213 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2214 ulp = NULL; 2215 break; 2216 2217 case IPPROTO_ROUTING: /* RFC 2460 */ 2218 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2219 if (((struct ip6_rthdr *)ulp)->ip6r_type != 0) { 2220 printf("IPFW2: IPV6 - Unknown Routing " 2221 "Header type(%d)\n", 2222 ((struct ip6_rthdr *)ulp)->ip6r_type); 2223 if (fw_deny_unknown_exthdrs) 2224 return (IP_FW_DENY); 2225 break; 2226 } 2227 ext_hd |= EXT_ROUTING; 2228 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2229 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2230 ulp = NULL; 2231 break; 2232 2233 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2234 PULLUP_TO(hlen, ulp, struct ip6_frag); 2235 ext_hd |= EXT_FRAGMENT; 2236 hlen += sizeof (struct ip6_frag); 2237 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2238 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2239 IP6F_OFF_MASK; 2240 /* Add IP6F_MORE_FRAG for offset of first 2241 * fragment to be != 0. */ 2242 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2243 IP6F_MORE_FRAG; 2244 if (offset == 0) { 2245 printf("IPFW2: IPV6 - Invalid Fragment " 2246 "Header\n"); 2247 if (fw_deny_unknown_exthdrs) 2248 return (IP_FW_DENY); 2249 break; 2250 } 2251 args->f_id.frag_id6 = 2252 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2253 ulp = NULL; 2254 break; 2255 2256 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2257 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2258 ext_hd |= EXT_DSTOPTS; 2259 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2260 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2261 ulp = NULL; 2262 break; 2263 2264 case IPPROTO_AH: /* RFC 2402 */ 2265 PULLUP_TO(hlen, ulp, struct ip6_ext); 2266 ext_hd |= EXT_AH; 2267 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2268 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2269 ulp = NULL; 2270 break; 2271 2272 case IPPROTO_ESP: /* RFC 2406 */ 2273 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2274 /* Anything past Seq# is variable length and 2275 * data past this ext. header is encrypted. */ 2276 ext_hd |= EXT_ESP; 2277 break; 2278 2279 case IPPROTO_NONE: /* RFC 2460 */ 2280 PULLUP_TO(hlen, ulp, struct ip6_ext); 2281 /* Packet ends here. if ip6e_len!=0 octets 2282 * must be ignored. */ 2283 break; 2284 2285 case IPPROTO_OSPFIGP: 2286 /* XXX OSPF header check? */ 2287 PULLUP_TO(hlen, ulp, struct ip6_ext); 2288 break; 2289 2290 default: 2291 printf("IPFW2: IPV6 - Unknown Extension " 2292 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2293 if (fw_deny_unknown_exthdrs) 2294 return (IP_FW_DENY); 2295 break; 2296 } /*switch */ 2297 } 2298 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2299 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2300 args->f_id.src_ip = 0; 2301 args->f_id.dst_ip = 0; 2302 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2303 } else if (pktlen >= sizeof(struct ip) && 2304 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2305 mtod(m, struct ip *)->ip_v == 4) { 2306 is_ipv4 = 1; 2307 ip = mtod(m, struct ip *); 2308 hlen = ip->ip_hl << 2; 2309 args->f_id.addr_type = 4; 2310 2311 /* 2312 * Collect parameters into local variables for faster matching. 2313 */ 2314 proto = ip->ip_p; 2315 src_ip = ip->ip_src; 2316 dst_ip = ip->ip_dst; 2317 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2318 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2319 ip_len = ntohs(ip->ip_len); 2320 } else { 2321 offset = ip->ip_off & IP_OFFMASK; 2322 ip_len = ip->ip_len; 2323 } 2324 pktlen = ip_len < pktlen ? ip_len : pktlen; 2325 2326 if (offset == 0) { 2327 switch (proto) { 2328 case IPPROTO_TCP: 2329 PULLUP_TO(hlen, ulp, struct tcphdr); 2330 dst_port = TCP(ulp)->th_dport; 2331 src_port = TCP(ulp)->th_sport; 2332 args->f_id.flags = TCP(ulp)->th_flags; 2333 break; 2334 2335 case IPPROTO_UDP: 2336 PULLUP_TO(hlen, ulp, struct udphdr); 2337 dst_port = UDP(ulp)->uh_dport; 2338 src_port = UDP(ulp)->uh_sport; 2339 break; 2340 2341 case IPPROTO_ICMP: 2342 PULLUP_TO(hlen, ulp, struct icmphdr); 2343 args->f_id.flags = ICMP(ulp)->icmp_type; 2344 break; 2345 2346 default: 2347 break; 2348 } 2349 } 2350 2351 args->f_id.src_ip = ntohl(src_ip.s_addr); 2352 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2353 } 2354#undef PULLUP_TO 2355 if (proto) { /* we may have port numbers, store them */ 2356 args->f_id.proto = proto; 2357 args->f_id.src_port = src_port = ntohs(src_port); 2358 args->f_id.dst_port = dst_port = ntohs(dst_port); 2359 } 2360 2361 IPFW_RLOCK(chain); 2362 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2363 if (args->rule) { 2364 /* 2365 * Packet has already been tagged. Look for the next rule 2366 * to restart processing. 2367 * 2368 * If fw_one_pass != 0 then just accept it. 2369 * XXX should not happen here, but optimized out in 2370 * the caller. 2371 */ 2372 if (fw_one_pass) { 2373 IPFW_RUNLOCK(chain); 2374 return (IP_FW_PASS); 2375 } 2376 2377 f = args->rule->next_rule; 2378 if (f == NULL) 2379 f = lookup_next_rule(args->rule); 2380 } else { 2381 /* 2382 * Find the starting rule. It can be either the first 2383 * one, or the one after divert_rule if asked so. 2384 */ 2385 int skipto = mtag ? divert_cookie(mtag) : 0; 2386 2387 f = chain->rules; 2388 if (args->eh == NULL && skipto != 0) { 2389 if (skipto >= IPFW_DEFAULT_RULE) { 2390 IPFW_RUNLOCK(chain); 2391 return (IP_FW_DENY); /* invalid */ 2392 } 2393 while (f && f->rulenum <= skipto) 2394 f = f->next; 2395 if (f == NULL) { /* drop packet */ 2396 IPFW_RUNLOCK(chain); 2397 return (IP_FW_DENY); 2398 } 2399 } 2400 } 2401 /* reset divert rule to avoid confusion later */ 2402 if (mtag) { 2403 divinput_flags = divert_info(mtag) & 2404 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2405 m_tag_delete(m, mtag); 2406 } 2407 2408 /* 2409 * Now scan the rules, and parse microinstructions for each rule. 2410 */ 2411 for (; f; f = f->next) { 2412 int l, cmdlen; 2413 ipfw_insn *cmd; 2414 int skip_or; /* skip rest of OR block */ 2415 2416again: 2417 if (set_disable & (1 << f->set) ) 2418 continue; 2419 2420 skip_or = 0; 2421 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2422 l -= cmdlen, cmd += cmdlen) { 2423 int match; 2424 2425 /* 2426 * check_body is a jump target used when we find a 2427 * CHECK_STATE, and need to jump to the body of 2428 * the target rule. 2429 */ 2430 2431check_body: 2432 cmdlen = F_LEN(cmd); 2433 /* 2434 * An OR block (insn_1 || .. || insn_n) has the 2435 * F_OR bit set in all but the last instruction. 2436 * The first match will set "skip_or", and cause 2437 * the following instructions to be skipped until 2438 * past the one with the F_OR bit clear. 2439 */ 2440 if (skip_or) { /* skip this instruction */ 2441 if ((cmd->len & F_OR) == 0) 2442 skip_or = 0; /* next one is good */ 2443 continue; 2444 } 2445 match = 0; /* set to 1 if we succeed */ 2446 2447 switch (cmd->opcode) { 2448 /* 2449 * The first set of opcodes compares the packet's 2450 * fields with some pattern, setting 'match' if a 2451 * match is found. At the end of the loop there is 2452 * logic to deal with F_NOT and F_OR flags associated 2453 * with the opcode. 2454 */ 2455 case O_NOP: 2456 match = 1; 2457 break; 2458 2459 case O_FORWARD_MAC: 2460 printf("ipfw: opcode %d unimplemented\n", 2461 cmd->opcode); 2462 break; 2463 2464 case O_GID: 2465 case O_UID: 2466 case O_JAIL: 2467 /* 2468 * We only check offset == 0 && proto != 0, 2469 * as this ensures that we have a 2470 * packet with the ports info. 2471 */ 2472 if (offset!=0) 2473 break; 2474 if (is_ipv6) /* XXX to be fixed later */ 2475 break; 2476 if (proto == IPPROTO_TCP || 2477 proto == IPPROTO_UDP) 2478 match = check_uidgid( 2479 (ipfw_insn_u32 *)cmd, 2480 proto, oif, 2481 dst_ip, dst_port, 2482 src_ip, src_port, &fw_ugid_cache, 2483 &ugid_lookup, args->inp); 2484 break; 2485 2486 case O_RECV: 2487 match = iface_match(m->m_pkthdr.rcvif, 2488 (ipfw_insn_if *)cmd); 2489 break; 2490 2491 case O_XMIT: 2492 match = iface_match(oif, (ipfw_insn_if *)cmd); 2493 break; 2494 2495 case O_VIA: 2496 match = iface_match(oif ? oif : 2497 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2498 break; 2499 2500 case O_MACADDR2: 2501 if (args->eh != NULL) { /* have MAC header */ 2502 u_int32_t *want = (u_int32_t *) 2503 ((ipfw_insn_mac *)cmd)->addr; 2504 u_int32_t *mask = (u_int32_t *) 2505 ((ipfw_insn_mac *)cmd)->mask; 2506 u_int32_t *hdr = (u_int32_t *)args->eh; 2507 2508 match = 2509 ( want[0] == (hdr[0] & mask[0]) && 2510 want[1] == (hdr[1] & mask[1]) && 2511 want[2] == (hdr[2] & mask[2]) ); 2512 } 2513 break; 2514 2515 case O_MAC_TYPE: 2516 if (args->eh != NULL) { 2517 u_int16_t t = 2518 ntohs(args->eh->ether_type); 2519 u_int16_t *p = 2520 ((ipfw_insn_u16 *)cmd)->ports; 2521 int i; 2522 2523 for (i = cmdlen - 1; !match && i>0; 2524 i--, p += 2) 2525 match = (t>=p[0] && t<=p[1]); 2526 } 2527 break; 2528 2529 case O_FRAG: 2530 match = (offset != 0); 2531 break; 2532 2533 case O_IN: /* "out" is "not in" */ 2534 match = (oif == NULL); 2535 break; 2536 2537 case O_LAYER2: 2538 match = (args->eh != NULL); 2539 break; 2540 2541 case O_DIVERTED: 2542 match = (cmd->arg1 & 1 && divinput_flags & 2543 IP_FW_DIVERT_LOOPBACK_FLAG) || 2544 (cmd->arg1 & 2 && divinput_flags & 2545 IP_FW_DIVERT_OUTPUT_FLAG); 2546 break; 2547 2548 case O_PROTO: 2549 /* 2550 * We do not allow an arg of 0 so the 2551 * check of "proto" only suffices. 2552 */ 2553 match = (proto == cmd->arg1); 2554 break; 2555 2556 case O_IP_SRC: 2557 match = is_ipv4 && 2558 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2559 src_ip.s_addr); 2560 break; 2561 2562 case O_IP_SRC_LOOKUP: 2563 case O_IP_DST_LOOKUP: 2564 if (is_ipv4) { 2565 uint32_t a = 2566 (cmd->opcode == O_IP_DST_LOOKUP) ? 2567 dst_ip.s_addr : src_ip.s_addr; 2568 uint32_t v; 2569 2570 match = lookup_table(cmd->arg1, a, &v); 2571 if (!match) 2572 break; 2573 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2574 match = 2575 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2576 } 2577 break; 2578 2579 case O_IP_SRC_MASK: 2580 case O_IP_DST_MASK: 2581 if (is_ipv4) { 2582 uint32_t a = 2583 (cmd->opcode == O_IP_DST_MASK) ? 2584 dst_ip.s_addr : src_ip.s_addr; 2585 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2586 int i = cmdlen-1; 2587 2588 for (; !match && i>0; i-= 2, p+= 2) 2589 match = (p[0] == (a & p[1])); 2590 } 2591 break; 2592 2593 case O_IP_SRC_ME: 2594 if (is_ipv4) { 2595 struct ifnet *tif; 2596 2597 INADDR_TO_IFP(src_ip, tif); 2598 match = (tif != NULL); 2599 } 2600 break; 2601 2602 case O_IP_DST_SET: 2603 case O_IP_SRC_SET: 2604 if (is_ipv4) { 2605 u_int32_t *d = (u_int32_t *)(cmd+1); 2606 u_int32_t addr = 2607 cmd->opcode == O_IP_DST_SET ? 2608 args->f_id.dst_ip : 2609 args->f_id.src_ip; 2610 2611 if (addr < d[0]) 2612 break; 2613 addr -= d[0]; /* subtract base */ 2614 match = (addr < cmd->arg1) && 2615 ( d[ 1 + (addr>>5)] & 2616 (1<<(addr & 0x1f)) ); 2617 } 2618 break; 2619 2620 case O_IP_DST: 2621 match = is_ipv4 && 2622 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2623 dst_ip.s_addr); 2624 break; 2625 2626 case O_IP_DST_ME: 2627 if (is_ipv4) { 2628 struct ifnet *tif; 2629 2630 INADDR_TO_IFP(dst_ip, tif); 2631 match = (tif != NULL); 2632 } 2633 break; 2634 2635 case O_IP_SRCPORT: 2636 case O_IP_DSTPORT: 2637 /* 2638 * offset == 0 && proto != 0 is enough 2639 * to guarantee that we have a 2640 * packet with port info. 2641 */ 2642 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2643 && offset == 0) { 2644 u_int16_t x = 2645 (cmd->opcode == O_IP_SRCPORT) ? 2646 src_port : dst_port ; 2647 u_int16_t *p = 2648 ((ipfw_insn_u16 *)cmd)->ports; 2649 int i; 2650 2651 for (i = cmdlen - 1; !match && i>0; 2652 i--, p += 2) 2653 match = (x>=p[0] && x<=p[1]); 2654 } 2655 break; 2656 2657 case O_ICMPTYPE: 2658 match = (offset == 0 && proto==IPPROTO_ICMP && 2659 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2660 break; 2661 2662#ifdef INET6 2663 case O_ICMP6TYPE: 2664 match = is_ipv6 && offset == 0 && 2665 proto==IPPROTO_ICMPV6 && 2666 icmp6type_match( 2667 ICMP6(ulp)->icmp6_type, 2668 (ipfw_insn_u32 *)cmd); 2669 break; 2670#endif /* INET6 */ 2671 2672 case O_IPOPT: 2673 match = (is_ipv4 && 2674 ipopts_match(mtod(m, struct ip *), cmd) ); 2675 break; 2676 2677 case O_IPVER: 2678 match = (is_ipv4 && 2679 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2680 break; 2681 2682 case O_IPID: 2683 case O_IPLEN: 2684 case O_IPTTL: 2685 if (is_ipv4) { /* only for IP packets */ 2686 uint16_t x; 2687 uint16_t *p; 2688 int i; 2689 2690 if (cmd->opcode == O_IPLEN) 2691 x = ip_len; 2692 else if (cmd->opcode == O_IPTTL) 2693 x = mtod(m, struct ip *)->ip_ttl; 2694 else /* must be IPID */ 2695 x = ntohs(mtod(m, struct ip *)->ip_id); 2696 if (cmdlen == 1) { 2697 match = (cmd->arg1 == x); 2698 break; 2699 } 2700 /* otherwise we have ranges */ 2701 p = ((ipfw_insn_u16 *)cmd)->ports; 2702 i = cmdlen - 1; 2703 for (; !match && i>0; i--, p += 2) 2704 match = (x >= p[0] && x <= p[1]); 2705 } 2706 break; 2707 2708 case O_IPPRECEDENCE: 2709 match = (is_ipv4 && 2710 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2711 break; 2712 2713 case O_IPTOS: 2714 match = (is_ipv4 && 2715 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2716 break; 2717 2718 case O_TCPDATALEN: 2719 if (proto == IPPROTO_TCP && offset == 0) { 2720 struct tcphdr *tcp; 2721 uint16_t x; 2722 uint16_t *p; 2723 int i; 2724 2725 tcp = TCP(ulp); 2726 x = ip_len - 2727 ((ip->ip_hl + tcp->th_off) << 2); 2728 if (cmdlen == 1) { 2729 match = (cmd->arg1 == x); 2730 break; 2731 } 2732 /* otherwise we have ranges */ 2733 p = ((ipfw_insn_u16 *)cmd)->ports; 2734 i = cmdlen - 1; 2735 for (; !match && i>0; i--, p += 2) 2736 match = (x >= p[0] && x <= p[1]); 2737 } 2738 break; 2739 2740 case O_TCPFLAGS: 2741 match = (proto == IPPROTO_TCP && offset == 0 && 2742 flags_match(cmd, TCP(ulp)->th_flags)); 2743 break; 2744 2745 case O_TCPOPTS: 2746 match = (proto == IPPROTO_TCP && offset == 0 && 2747 tcpopts_match(TCP(ulp), cmd)); 2748 break; 2749 2750 case O_TCPSEQ: 2751 match = (proto == IPPROTO_TCP && offset == 0 && 2752 ((ipfw_insn_u32 *)cmd)->d[0] == 2753 TCP(ulp)->th_seq); 2754 break; 2755 2756 case O_TCPACK: 2757 match = (proto == IPPROTO_TCP && offset == 0 && 2758 ((ipfw_insn_u32 *)cmd)->d[0] == 2759 TCP(ulp)->th_ack); 2760 break; 2761 2762 case O_TCPWIN: 2763 match = (proto == IPPROTO_TCP && offset == 0 && 2764 cmd->arg1 == TCP(ulp)->th_win); 2765 break; 2766 2767 case O_ESTAB: 2768 /* reject packets which have SYN only */ 2769 /* XXX should i also check for TH_ACK ? */ 2770 match = (proto == IPPROTO_TCP && offset == 0 && 2771 (TCP(ulp)->th_flags & 2772 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2773 break; 2774 2775 case O_ALTQ: { 2776 struct altq_tag *at; 2777 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2778 2779 match = 1; 2780 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2781 if (mtag != NULL) 2782 break; 2783 mtag = m_tag_get(PACKET_TAG_PF_QID, 2784 sizeof(struct altq_tag), 2785 M_NOWAIT); 2786 if (mtag == NULL) { 2787 /* 2788 * Let the packet fall back to the 2789 * default ALTQ. 2790 */ 2791 break; 2792 } 2793 at = (struct altq_tag *)(mtag+1); 2794 at->qid = altq->qid; 2795 if (is_ipv4) 2796 at->af = AF_INET; 2797 else 2798 at->af = AF_LINK; 2799 at->hdr = ip; 2800 m_tag_prepend(m, mtag); 2801 break; 2802 } 2803 2804 case O_LOG: 2805 if (fw_verbose) 2806 ipfw_log(f, hlen, args, m, oif, offset); 2807 match = 1; 2808 break; 2809 2810 case O_PROB: 2811 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2812 break; 2813 2814 case O_VERREVPATH: 2815 /* Outgoing packets automatically pass/match */ 2816 match = ((oif != NULL) || 2817 (m->m_pkthdr.rcvif == NULL) || 2818 ( 2819#ifdef INET6 2820 is_ipv6 ? 2821 verify_path6(&(args->f_id.src_ip6), 2822 m->m_pkthdr.rcvif) : 2823#endif 2824 verify_path(src_ip, m->m_pkthdr.rcvif))); 2825 break; 2826 2827 case O_VERSRCREACH: 2828 /* Outgoing packets automatically pass/match */ 2829 match = (hlen > 0 && ((oif != NULL) || 2830#ifdef INET6 2831 is_ipv6 ? 2832 verify_path6(&(args->f_id.src_ip6), 2833 NULL) : 2834#endif 2835 verify_path(src_ip, NULL))); 2836 break; 2837 2838 case O_ANTISPOOF: 2839 /* Outgoing packets automatically pass/match */ 2840 if (oif == NULL && hlen > 0 && 2841 ( (is_ipv4 && in_localaddr(src_ip)) 2842#ifdef INET6 2843 || (is_ipv6 && 2844 in6_localaddr(&(args->f_id.src_ip6))) 2845#endif 2846 )) 2847 match = 2848#ifdef INET6 2849 is_ipv6 ? verify_path6( 2850 &(args->f_id.src_ip6), 2851 m->m_pkthdr.rcvif) : 2852#endif 2853 verify_path(src_ip, 2854 m->m_pkthdr.rcvif); 2855 else 2856 match = 1; 2857 break; 2858 2859 case O_IPSEC: 2860#ifdef FAST_IPSEC 2861 match = (m_tag_find(m, 2862 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2863#endif 2864#ifdef IPSEC 2865 match = (ipsec_getnhist(m) != 0); 2866#endif 2867 /* otherwise no match */ 2868 break; 2869 2870#ifdef INET6 2871 case O_IP6_SRC: 2872 match = is_ipv6 && 2873 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2874 &((ipfw_insn_ip6 *)cmd)->addr6); 2875 break; 2876 2877 case O_IP6_DST: 2878 match = is_ipv6 && 2879 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2880 &((ipfw_insn_ip6 *)cmd)->addr6); 2881 break; 2882 case O_IP6_SRC_MASK: 2883 if (is_ipv6) { 2884 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2885 struct in6_addr p = args->f_id.src_ip6; 2886 2887 APPLY_MASK(&p, &te->mask6); 2888 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2889 } 2890 break; 2891 2892 case O_IP6_DST_MASK: 2893 if (is_ipv6) { 2894 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2895 struct in6_addr p = args->f_id.dst_ip6; 2896 2897 APPLY_MASK(&p, &te->mask6); 2898 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2899 } 2900 break; 2901 2902 case O_IP6_SRC_ME: 2903 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2904 break; 2905 2906 case O_IP6_DST_ME: 2907 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2908 break; 2909 2910 case O_FLOW6ID: 2911 match = is_ipv6 && 2912 flow6id_match(args->f_id.flow_id6, 2913 (ipfw_insn_u32 *) cmd); 2914 break; 2915 2916 case O_EXT_HDR: 2917 match = is_ipv6 && 2918 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2919 break; 2920 2921 case O_IP6: 2922 match = is_ipv6; 2923 break; 2924#endif 2925 2926 case O_IP4: 2927 match = is_ipv4; 2928 break; 2929 2930 /* 2931 * The second set of opcodes represents 'actions', 2932 * i.e. the terminal part of a rule once the packet 2933 * matches all previous patterns. 2934 * Typically there is only one action for each rule, 2935 * and the opcode is stored at the end of the rule 2936 * (but there are exceptions -- see below). 2937 * 2938 * In general, here we set retval and terminate the 2939 * outer loop (would be a 'break 3' in some language, 2940 * but we need to do a 'goto done'). 2941 * 2942 * Exceptions: 2943 * O_COUNT and O_SKIPTO actions: 2944 * instead of terminating, we jump to the next rule 2945 * ('goto next_rule', equivalent to a 'break 2'), 2946 * or to the SKIPTO target ('goto again' after 2947 * having set f, cmd and l), respectively. 2948 * 2949 * O_LOG and O_ALTQ action parameters: 2950 * perform some action and set match = 1; 2951 * 2952 * O_LIMIT and O_KEEP_STATE: these opcodes are 2953 * not real 'actions', and are stored right 2954 * before the 'action' part of the rule. 2955 * These opcodes try to install an entry in the 2956 * state tables; if successful, we continue with 2957 * the next opcode (match=1; break;), otherwise 2958 * the packet * must be dropped 2959 * ('goto done' after setting retval); 2960 * 2961 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2962 * cause a lookup of the state table, and a jump 2963 * to the 'action' part of the parent rule 2964 * ('goto check_body') if an entry is found, or 2965 * (CHECK_STATE only) a jump to the next rule if 2966 * the entry is not found ('goto next_rule'). 2967 * The result of the lookup is cached to make 2968 * further instances of these opcodes are 2969 * effectively NOPs. 2970 */ 2971 case O_LIMIT: 2972 case O_KEEP_STATE: 2973 if (install_state(f, 2974 (ipfw_insn_limit *)cmd, args)) { 2975 retval = IP_FW_DENY; 2976 goto done; /* error/limit violation */ 2977 } 2978 match = 1; 2979 break; 2980 2981 case O_PROBE_STATE: 2982 case O_CHECK_STATE: 2983 /* 2984 * dynamic rules are checked at the first 2985 * keep-state or check-state occurrence, 2986 * with the result being stored in dyn_dir. 2987 * The compiler introduces a PROBE_STATE 2988 * instruction for us when we have a 2989 * KEEP_STATE (because PROBE_STATE needs 2990 * to be run first). 2991 */ 2992 if (dyn_dir == MATCH_UNKNOWN && 2993 (q = lookup_dyn_rule(&args->f_id, 2994 &dyn_dir, proto == IPPROTO_TCP ? 2995 TCP(ulp) : NULL)) 2996 != NULL) { 2997 /* 2998 * Found dynamic entry, update stats 2999 * and jump to the 'action' part of 3000 * the parent rule. 3001 */ 3002 q->pcnt++; 3003 q->bcnt += pktlen; 3004 f = q->rule; 3005 cmd = ACTION_PTR(f); 3006 l = f->cmd_len - f->act_ofs; 3007 IPFW_DYN_UNLOCK(); 3008 goto check_body; 3009 } 3010 /* 3011 * Dynamic entry not found. If CHECK_STATE, 3012 * skip to next rule, if PROBE_STATE just 3013 * ignore and continue with next opcode. 3014 */ 3015 if (cmd->opcode == O_CHECK_STATE) 3016 goto next_rule; 3017 match = 1; 3018 break; 3019 3020 case O_ACCEPT: 3021 retval = 0; /* accept */ 3022 goto done; 3023 3024 case O_PIPE: 3025 case O_QUEUE: 3026 args->rule = f; /* report matching rule */ 3027 args->cookie = cmd->arg1; 3028 retval = IP_FW_DUMMYNET; 3029 goto done; 3030 3031 case O_DIVERT: 3032 case O_TEE: { 3033 struct divert_tag *dt; 3034 3035 if (args->eh) /* not on layer 2 */ 3036 break; 3037 mtag = m_tag_get(PACKET_TAG_DIVERT, 3038 sizeof(struct divert_tag), 3039 M_NOWAIT); 3040 if (mtag == NULL) { 3041 /* XXX statistic */ 3042 /* drop packet */ 3043 IPFW_RUNLOCK(chain); 3044 return (IP_FW_DENY); 3045 } 3046 dt = (struct divert_tag *)(mtag+1); 3047 dt->cookie = f->rulenum; 3048 dt->info = cmd->arg1; 3049 m_tag_prepend(m, mtag); 3050 retval = (cmd->opcode == O_DIVERT) ? 3051 IP_FW_DIVERT : IP_FW_TEE; 3052 goto done; 3053 } 3054 3055 case O_COUNT: 3056 case O_SKIPTO: 3057 f->pcnt++; /* update stats */ 3058 f->bcnt += pktlen; 3059 f->timestamp = time_uptime; 3060 if (cmd->opcode == O_COUNT) 3061 goto next_rule; 3062 /* handle skipto */ 3063 if (f->next_rule == NULL) 3064 lookup_next_rule(f); 3065 f = f->next_rule; 3066 goto again; 3067 3068 case O_REJECT: 3069 /* 3070 * Drop the packet and send a reject notice 3071 * if the packet is not ICMP (or is an ICMP 3072 * query), and it is not multicast/broadcast. 3073 */ 3074 if (hlen > 0 && is_ipv4 && 3075 (proto != IPPROTO_ICMP || 3076 is_icmp_query(ICMP(ulp))) && 3077 !(m->m_flags & (M_BCAST|M_MCAST)) && 3078 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3079 send_reject(args, cmd->arg1, 3080 offset,ip_len); 3081 m = args->m; 3082 } 3083 /* FALLTHROUGH */ 3084#ifdef INET6 3085 case O_UNREACH6: 3086 if (hlen > 0 && is_ipv6 && 3087 (proto != IPPROTO_ICMPV6 || 3088 (is_icmp6_query(args->f_id.flags) == 1)) && 3089 !(m->m_flags & (M_BCAST|M_MCAST)) && 3090 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3091 send_reject6(args, cmd->arg1, 3092 offset, hlen); 3093 m = args->m; 3094 } 3095 /* FALLTHROUGH */ 3096#endif 3097 case O_DENY: 3098 retval = IP_FW_DENY; 3099 goto done; 3100 3101 case O_FORWARD_IP: 3102 if (args->eh) /* not valid on layer2 pkts */ 3103 break; 3104 if (!q || dyn_dir == MATCH_FORWARD) 3105 args->next_hop = 3106 &((ipfw_insn_sa *)cmd)->sa; 3107 retval = IP_FW_PASS; 3108 goto done; 3109 3110 case O_NETGRAPH: 3111 case O_NGTEE: 3112 args->rule = f; /* report matching rule */ 3113 args->cookie = cmd->arg1; 3114 retval = (cmd->opcode == O_NETGRAPH) ? 3115 IP_FW_NETGRAPH : IP_FW_NGTEE; 3116 goto done; 3117 3118 default: 3119 panic("-- unknown opcode %d\n", cmd->opcode); 3120 } /* end of switch() on opcodes */ 3121 3122 if (cmd->len & F_NOT) 3123 match = !match; 3124 3125 if (match) { 3126 if (cmd->len & F_OR) 3127 skip_or = 1; 3128 } else { 3129 if (!(cmd->len & F_OR)) /* not an OR block, */ 3130 break; /* try next rule */ 3131 } 3132 3133 } /* end of inner for, scan opcodes */ 3134 3135next_rule:; /* try next rule */ 3136 3137 } /* end of outer for, scan rules */ 3138 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3139 IPFW_RUNLOCK(chain); 3140 return (IP_FW_DENY); 3141 3142done: 3143 /* Update statistics */ 3144 f->pcnt++; 3145 f->bcnt += pktlen; 3146 f->timestamp = time_uptime; 3147 IPFW_RUNLOCK(chain); 3148 return (retval); 3149 3150pullup_failed: 3151 if (fw_verbose) 3152 printf("ipfw: pullup failed\n"); 3153 return (IP_FW_DENY); 3154} 3155 3156/* 3157 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3158 * These will be reconstructed on the fly as packets are matched. 3159 */ 3160static void 3161flush_rule_ptrs(struct ip_fw_chain *chain) 3162{ 3163 struct ip_fw *rule; 3164 3165 IPFW_WLOCK_ASSERT(chain); 3166 3167 for (rule = chain->rules; rule; rule = rule->next) 3168 rule->next_rule = NULL; 3169} 3170 3171/* 3172 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3173 * possibly create a rule number and add the rule to the list. 3174 * Update the rule_number in the input struct so the caller knows it as well. 3175 */ 3176static int 3177add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3178{ 3179 struct ip_fw *rule, *f, *prev; 3180 int l = RULESIZE(input_rule); 3181 3182 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3183 return (EINVAL); 3184 3185 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3186 if (rule == NULL) 3187 return (ENOSPC); 3188 3189 bcopy(input_rule, rule, l); 3190 3191 rule->next = NULL; 3192 rule->next_rule = NULL; 3193 3194 rule->pcnt = 0; 3195 rule->bcnt = 0; 3196 rule->timestamp = 0; 3197 3198 IPFW_WLOCK(chain); 3199 3200 if (chain->rules == NULL) { /* default rule */ 3201 chain->rules = rule; 3202 goto done; 3203 } 3204 3205 /* 3206 * If rulenum is 0, find highest numbered rule before the 3207 * default rule, and add autoinc_step 3208 */ 3209 if (autoinc_step < 1) 3210 autoinc_step = 1; 3211 else if (autoinc_step > 1000) 3212 autoinc_step = 1000; 3213 if (rule->rulenum == 0) { 3214 /* 3215 * locate the highest numbered rule before default 3216 */ 3217 for (f = chain->rules; f; f = f->next) { 3218 if (f->rulenum == IPFW_DEFAULT_RULE) 3219 break; 3220 rule->rulenum = f->rulenum; 3221 } 3222 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3223 rule->rulenum += autoinc_step; 3224 input_rule->rulenum = rule->rulenum; 3225 } 3226 3227 /* 3228 * Now insert the new rule in the right place in the sorted list. 3229 */ 3230 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3231 if (f->rulenum > rule->rulenum) { /* found the location */ 3232 if (prev) { 3233 rule->next = f; 3234 prev->next = rule; 3235 } else { /* head insert */ 3236 rule->next = chain->rules; 3237 chain->rules = rule; 3238 } 3239 break; 3240 } 3241 } 3242 flush_rule_ptrs(chain); 3243done: 3244 static_count++; 3245 static_len += l; 3246 IPFW_WUNLOCK(chain); 3247 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3248 rule->rulenum, static_count);) 3249 return (0); 3250} 3251 3252/** 3253 * Remove a static rule (including derived * dynamic rules) 3254 * and place it on the ``reap list'' for later reclamation. 3255 * The caller is in charge of clearing rule pointers to avoid 3256 * dangling pointers. 3257 * @return a pointer to the next entry. 3258 * Arguments are not checked, so they better be correct. 3259 */ 3260static struct ip_fw * 3261remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3262{ 3263 struct ip_fw *n; 3264 int l = RULESIZE(rule); 3265 3266 IPFW_WLOCK_ASSERT(chain); 3267 3268 n = rule->next; 3269 IPFW_DYN_LOCK(); 3270 remove_dyn_rule(rule, NULL /* force removal */); 3271 IPFW_DYN_UNLOCK(); 3272 if (prev == NULL) 3273 chain->rules = n; 3274 else 3275 prev->next = n; 3276 static_count--; 3277 static_len -= l; 3278 3279 rule->next = chain->reap; 3280 chain->reap = rule; 3281 3282 return n; 3283} 3284 3285/** 3286 * Reclaim storage associated with a list of rules. This is 3287 * typically the list created using remove_rule. 3288 */ 3289static void 3290reap_rules(struct ip_fw *head) 3291{ 3292 struct ip_fw *rule; 3293 3294 while ((rule = head) != NULL) { 3295 head = head->next; 3296 if (DUMMYNET_LOADED) 3297 ip_dn_ruledel_ptr(rule); 3298 free(rule, M_IPFW); 3299 } 3300} 3301 3302/* 3303 * Remove all rules from a chain (except rules in set RESVD_SET 3304 * unless kill_default = 1). The caller is responsible for 3305 * reclaiming storage for the rules left in chain->reap. 3306 */ 3307static void 3308free_chain(struct ip_fw_chain *chain, int kill_default) 3309{ 3310 struct ip_fw *prev, *rule; 3311 3312 IPFW_WLOCK_ASSERT(chain); 3313 3314 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3315 for (prev = NULL, rule = chain->rules; rule ; ) 3316 if (kill_default || rule->set != RESVD_SET) 3317 rule = remove_rule(chain, rule, prev); 3318 else { 3319 prev = rule; 3320 rule = rule->next; 3321 } 3322} 3323 3324/** 3325 * Remove all rules with given number, and also do set manipulation. 3326 * Assumes chain != NULL && *chain != NULL. 3327 * 3328 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3329 * the next 8 bits are the new set, the top 8 bits are the command: 3330 * 3331 * 0 delete rules with given number 3332 * 1 delete rules with given set number 3333 * 2 move rules with given number to new set 3334 * 3 move rules with given set number to new set 3335 * 4 swap sets with given numbers 3336 */ 3337static int 3338del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3339{ 3340 struct ip_fw *prev = NULL, *rule; 3341 u_int16_t rulenum; /* rule or old_set */ 3342 u_int8_t cmd, new_set; 3343 3344 rulenum = arg & 0xffff; 3345 cmd = (arg >> 24) & 0xff; 3346 new_set = (arg >> 16) & 0xff; 3347 3348 if (cmd > 4) 3349 return EINVAL; 3350 if (new_set > RESVD_SET) 3351 return EINVAL; 3352 if (cmd == 0 || cmd == 2) { 3353 if (rulenum >= IPFW_DEFAULT_RULE) 3354 return EINVAL; 3355 } else { 3356 if (rulenum > RESVD_SET) /* old_set */ 3357 return EINVAL; 3358 } 3359 3360 IPFW_WLOCK(chain); 3361 rule = chain->rules; 3362 chain->reap = NULL; 3363 switch (cmd) { 3364 case 0: /* delete rules with given number */ 3365 /* 3366 * locate first rule to delete 3367 */ 3368 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3369 ; 3370 if (rule->rulenum != rulenum) { 3371 IPFW_WUNLOCK(chain); 3372 return EINVAL; 3373 } 3374 3375 /* 3376 * flush pointers outside the loop, then delete all matching 3377 * rules. prev remains the same throughout the cycle. 3378 */ 3379 flush_rule_ptrs(chain); 3380 while (rule->rulenum == rulenum) 3381 rule = remove_rule(chain, rule, prev); 3382 break; 3383 3384 case 1: /* delete all rules with given set number */ 3385 flush_rule_ptrs(chain); 3386 rule = chain->rules; 3387 while (rule->rulenum < IPFW_DEFAULT_RULE) 3388 if (rule->set == rulenum) 3389 rule = remove_rule(chain, rule, prev); 3390 else { 3391 prev = rule; 3392 rule = rule->next; 3393 } 3394 break; 3395 3396 case 2: /* move rules with given number to new set */ 3397 rule = chain->rules; 3398 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3399 if (rule->rulenum == rulenum) 3400 rule->set = new_set; 3401 break; 3402 3403 case 3: /* move rules with given set number to new set */ 3404 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3405 if (rule->set == rulenum) 3406 rule->set = new_set; 3407 break; 3408 3409 case 4: /* swap two sets */ 3410 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3411 if (rule->set == rulenum) 3412 rule->set = new_set; 3413 else if (rule->set == new_set) 3414 rule->set = rulenum; 3415 break; 3416 } 3417 /* 3418 * Look for rules to reclaim. We grab the list before 3419 * releasing the lock then reclaim them w/o the lock to 3420 * avoid a LOR with dummynet. 3421 */ 3422 rule = chain->reap; 3423 chain->reap = NULL; 3424 IPFW_WUNLOCK(chain); 3425 if (rule) 3426 reap_rules(rule); 3427 return 0; 3428} 3429 3430/* 3431 * Clear counters for a specific rule. 3432 * The enclosing "table" is assumed locked. 3433 */ 3434static void 3435clear_counters(struct ip_fw *rule, int log_only) 3436{ 3437 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3438 3439 if (log_only == 0) { 3440 rule->bcnt = rule->pcnt = 0; 3441 rule->timestamp = 0; 3442 } 3443 if (l->o.opcode == O_LOG) 3444 l->log_left = l->max_log; 3445} 3446 3447/** 3448 * Reset some or all counters on firewall rules. 3449 * @arg frwl is null to clear all entries, or contains a specific 3450 * rule number. 3451 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3452 */ 3453static int 3454zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3455{ 3456 struct ip_fw *rule; 3457 char *msg; 3458 3459 IPFW_WLOCK(chain); 3460 if (rulenum == 0) { 3461 norule_counter = 0; 3462 for (rule = chain->rules; rule; rule = rule->next) 3463 clear_counters(rule, log_only); 3464 msg = log_only ? "ipfw: All logging counts reset.\n" : 3465 "ipfw: Accounting cleared.\n"; 3466 } else { 3467 int cleared = 0; 3468 /* 3469 * We can have multiple rules with the same number, so we 3470 * need to clear them all. 3471 */ 3472 for (rule = chain->rules; rule; rule = rule->next) 3473 if (rule->rulenum == rulenum) { 3474 while (rule && rule->rulenum == rulenum) { 3475 clear_counters(rule, log_only); 3476 rule = rule->next; 3477 } 3478 cleared = 1; 3479 break; 3480 } 3481 if (!cleared) { /* we did not find any matching rules */ 3482 IPFW_WUNLOCK(chain); 3483 return (EINVAL); 3484 } 3485 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3486 "ipfw: Entry %d cleared.\n"; 3487 } 3488 IPFW_WUNLOCK(chain); 3489 3490 if (fw_verbose) 3491 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3492 return (0); 3493} 3494 3495/* 3496 * Check validity of the structure before insert. 3497 * Fortunately rules are simple, so this mostly need to check rule sizes. 3498 */ 3499static int 3500check_ipfw_struct(struct ip_fw *rule, int size) 3501{ 3502 int l, cmdlen = 0; 3503 int have_action=0; 3504 ipfw_insn *cmd; 3505 3506 if (size < sizeof(*rule)) { 3507 printf("ipfw: rule too short\n"); 3508 return (EINVAL); 3509 } 3510 /* first, check for valid size */ 3511 l = RULESIZE(rule); 3512 if (l != size) { 3513 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3514 return (EINVAL); 3515 } 3516 if (rule->act_ofs >= rule->cmd_len) { 3517 printf("ipfw: bogus action offset (%u > %u)\n", 3518 rule->act_ofs, rule->cmd_len - 1); 3519 return (EINVAL); 3520 } 3521 /* 3522 * Now go for the individual checks. Very simple ones, basically only 3523 * instruction sizes. 3524 */ 3525 for (l = rule->cmd_len, cmd = rule->cmd ; 3526 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3527 cmdlen = F_LEN(cmd); 3528 if (cmdlen > l) { 3529 printf("ipfw: opcode %d size truncated\n", 3530 cmd->opcode); 3531 return EINVAL; 3532 } 3533 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3534 switch (cmd->opcode) { 3535 case O_PROBE_STATE: 3536 case O_KEEP_STATE: 3537 case O_PROTO: 3538 case O_IP_SRC_ME: 3539 case O_IP_DST_ME: 3540 case O_LAYER2: 3541 case O_IN: 3542 case O_FRAG: 3543 case O_DIVERTED: 3544 case O_IPOPT: 3545 case O_IPTOS: 3546 case O_IPPRECEDENCE: 3547 case O_IPVER: 3548 case O_TCPWIN: 3549 case O_TCPFLAGS: 3550 case O_TCPOPTS: 3551 case O_ESTAB: 3552 case O_VERREVPATH: 3553 case O_VERSRCREACH: 3554 case O_ANTISPOOF: 3555 case O_IPSEC: 3556#ifdef INET6 3557 case O_IP6_SRC_ME: 3558 case O_IP6_DST_ME: 3559 case O_EXT_HDR: 3560 case O_IP6: 3561#endif 3562 case O_IP4: 3563 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3564 goto bad_size; 3565 break; 3566 3567 case O_UID: 3568 case O_GID: 3569 case O_JAIL: 3570 case O_IP_SRC: 3571 case O_IP_DST: 3572 case O_TCPSEQ: 3573 case O_TCPACK: 3574 case O_PROB: 3575 case O_ICMPTYPE: 3576 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3577 goto bad_size; 3578 break; 3579 3580 case O_LIMIT: 3581 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3582 goto bad_size; 3583 break; 3584 3585 case O_LOG: 3586 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3587 goto bad_size; 3588 3589 ((ipfw_insn_log *)cmd)->log_left = 3590 ((ipfw_insn_log *)cmd)->max_log; 3591 3592 break; 3593 3594 case O_IP_SRC_MASK: 3595 case O_IP_DST_MASK: 3596 /* only odd command lengths */ 3597 if ( !(cmdlen & 1) || cmdlen > 31) 3598 goto bad_size; 3599 break; 3600 3601 case O_IP_SRC_SET: 3602 case O_IP_DST_SET: 3603 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3604 printf("ipfw: invalid set size %d\n", 3605 cmd->arg1); 3606 return EINVAL; 3607 } 3608 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3609 (cmd->arg1+31)/32 ) 3610 goto bad_size; 3611 break; 3612 3613 case O_IP_SRC_LOOKUP: 3614 case O_IP_DST_LOOKUP: 3615 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3616 printf("ipfw: invalid table number %d\n", 3617 cmd->arg1); 3618 return (EINVAL); 3619 } 3620 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3621 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3622 goto bad_size; 3623 break; 3624 3625 case O_MACADDR2: 3626 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3627 goto bad_size; 3628 break; 3629 3630 case O_NOP: 3631 case O_IPID: 3632 case O_IPTTL: 3633 case O_IPLEN: 3634 case O_TCPDATALEN: 3635 if (cmdlen < 1 || cmdlen > 31) 3636 goto bad_size; 3637 break; 3638 3639 case O_MAC_TYPE: 3640 case O_IP_SRCPORT: 3641 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3642 if (cmdlen < 2 || cmdlen > 31) 3643 goto bad_size; 3644 break; 3645 3646 case O_RECV: 3647 case O_XMIT: 3648 case O_VIA: 3649 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3650 goto bad_size; 3651 break; 3652 3653 case O_ALTQ: 3654 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3655 goto bad_size; 3656 break; 3657 3658 case O_PIPE: 3659 case O_QUEUE: 3660 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3661 goto bad_size; 3662 goto check_action; 3663 3664 case O_FORWARD_IP: 3665#ifdef IPFIREWALL_FORWARD 3666 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3667 goto bad_size; 3668 goto check_action; 3669#else 3670 return EINVAL; 3671#endif 3672 3673 case O_DIVERT: 3674 case O_TEE: 3675 if (ip_divert_ptr == NULL) 3676 return EINVAL; 3677 else 3678 goto check_size; 3679 case O_NETGRAPH: 3680 case O_NGTEE: 3681 if (!NG_IPFW_LOADED) 3682 return EINVAL; 3683 else 3684 goto check_size; 3685 case O_FORWARD_MAC: /* XXX not implemented yet */ 3686 case O_CHECK_STATE: 3687 case O_COUNT: 3688 case O_ACCEPT: 3689 case O_DENY: 3690 case O_REJECT: 3691#ifdef INET6 3692 case O_UNREACH6: 3693#endif 3694 case O_SKIPTO: 3695check_size: 3696 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3697 goto bad_size; 3698check_action: 3699 if (have_action) { 3700 printf("ipfw: opcode %d, multiple actions" 3701 " not allowed\n", 3702 cmd->opcode); 3703 return EINVAL; 3704 } 3705 have_action = 1; 3706 if (l != cmdlen) { 3707 printf("ipfw: opcode %d, action must be" 3708 " last opcode\n", 3709 cmd->opcode); 3710 return EINVAL; 3711 } 3712 break; 3713#ifdef INET6 3714 case O_IP6_SRC: 3715 case O_IP6_DST: 3716 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3717 F_INSN_SIZE(ipfw_insn)) 3718 goto bad_size; 3719 break; 3720 3721 case O_FLOW6ID: 3722 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3723 ((ipfw_insn_u32 *)cmd)->o.arg1) 3724 goto bad_size; 3725 break; 3726 3727 case O_IP6_SRC_MASK: 3728 case O_IP6_DST_MASK: 3729 if ( !(cmdlen & 1) || cmdlen > 127) 3730 goto bad_size; 3731 break; 3732 case O_ICMP6TYPE: 3733 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3734 goto bad_size; 3735 break; 3736#endif 3737 3738 default: 3739 switch (cmd->opcode) { 3740#ifndef INET6 3741 case O_IP6_SRC_ME: 3742 case O_IP6_DST_ME: 3743 case O_EXT_HDR: 3744 case O_IP6: 3745 case O_UNREACH6: 3746 case O_IP6_SRC: 3747 case O_IP6_DST: 3748 case O_FLOW6ID: 3749 case O_IP6_SRC_MASK: 3750 case O_IP6_DST_MASK: 3751 case O_ICMP6TYPE: 3752 printf("ipfw: no IPv6 support in kernel\n"); 3753 return EPROTONOSUPPORT; 3754#endif 3755 default: 3756 printf("ipfw: opcode %d, unknown opcode\n", 3757 cmd->opcode); 3758 return EINVAL; 3759 } 3760 } 3761 } 3762 if (have_action == 0) { 3763 printf("ipfw: missing action\n"); 3764 return EINVAL; 3765 } 3766 return 0; 3767 3768bad_size: 3769 printf("ipfw: opcode %d size %d wrong\n", 3770 cmd->opcode, cmdlen); 3771 return EINVAL; 3772} 3773 3774/* 3775 * Copy the static and dynamic rules to the supplied buffer 3776 * and return the amount of space actually used. 3777 */ 3778static size_t 3779ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3780{ 3781 char *bp = buf; 3782 char *ep = bp + space; 3783 struct ip_fw *rule; 3784 int i; 3785 3786 /* XXX this can take a long time and locking will block packet flow */ 3787 IPFW_RLOCK(chain); 3788 for (rule = chain->rules; rule ; rule = rule->next) { 3789 /* 3790 * Verify the entry fits in the buffer in case the 3791 * rules changed between calculating buffer space and 3792 * now. This would be better done using a generation 3793 * number but should suffice for now. 3794 */ 3795 i = RULESIZE(rule); 3796 if (bp + i <= ep) { 3797 bcopy(rule, bp, i); 3798 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3799 sizeof(set_disable)); 3800 bp += i; 3801 } 3802 } 3803 IPFW_RUNLOCK(chain); 3804 if (ipfw_dyn_v) { 3805 ipfw_dyn_rule *p, *last = NULL; 3806 3807 IPFW_DYN_LOCK(); 3808 for (i = 0 ; i < curr_dyn_buckets; i++) 3809 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3810 if (bp + sizeof *p <= ep) { 3811 ipfw_dyn_rule *dst = 3812 (ipfw_dyn_rule *)bp; 3813 bcopy(p, dst, sizeof *p); 3814 bcopy(&(p->rule->rulenum), &(dst->rule), 3815 sizeof(p->rule->rulenum)); 3816 /* 3817 * store a non-null value in "next". 3818 * The userland code will interpret a 3819 * NULL here as a marker 3820 * for the last dynamic rule. 3821 */ 3822 bcopy(&dst, &dst->next, sizeof(dst)); 3823 last = dst; 3824 dst->expire = 3825 TIME_LEQ(dst->expire, time_uptime) ? 3826 0 : dst->expire - time_uptime ; 3827 bp += sizeof(ipfw_dyn_rule); 3828 } 3829 } 3830 IPFW_DYN_UNLOCK(); 3831 if (last != NULL) /* mark last dynamic rule */ 3832 bzero(&last->next, sizeof(last)); 3833 } 3834 return (bp - (char *)buf); 3835} 3836 3837 3838/** 3839 * {set|get}sockopt parser. 3840 */ 3841static int 3842ipfw_ctl(struct sockopt *sopt) 3843{ 3844#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3845 int error, rule_num; 3846 size_t size; 3847 struct ip_fw *buf, *rule; 3848 u_int32_t rulenum[2]; 3849 3850 error = suser(sopt->sopt_td); 3851 if (error) 3852 return (error); 3853 3854 /* 3855 * Disallow modifications in really-really secure mode, but still allow 3856 * the logging counters to be reset. 3857 */ 3858 if (sopt->sopt_name == IP_FW_ADD || 3859 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3860#if __FreeBSD_version >= 500034 3861 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3862 if (error) 3863 return (error); 3864#else /* FreeBSD 4.x */ 3865 if (securelevel >= 3) 3866 return (EPERM); 3867#endif 3868 } 3869 3870 error = 0; 3871 3872 switch (sopt->sopt_name) { 3873 case IP_FW_GET: 3874 /* 3875 * pass up a copy of the current rules. Static rules 3876 * come first (the last of which has number IPFW_DEFAULT_RULE), 3877 * followed by a possibly empty list of dynamic rule. 3878 * The last dynamic rule has NULL in the "next" field. 3879 * 3880 * Note that the calculated size is used to bound the 3881 * amount of data returned to the user. The rule set may 3882 * change between calculating the size and returning the 3883 * data in which case we'll just return what fits. 3884 */ 3885 size = static_len; /* size of static rules */ 3886 if (ipfw_dyn_v) /* add size of dyn.rules */ 3887 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3888 3889 /* 3890 * XXX todo: if the user passes a short length just to know 3891 * how much room is needed, do not bother filling up the 3892 * buffer, just jump to the sooptcopyout. 3893 */ 3894 buf = malloc(size, M_TEMP, M_WAITOK); 3895 error = sooptcopyout(sopt, buf, 3896 ipfw_getrules(&layer3_chain, buf, size)); 3897 free(buf, M_TEMP); 3898 break; 3899 3900 case IP_FW_FLUSH: 3901 /* 3902 * Normally we cannot release the lock on each iteration. 3903 * We could do it here only because we start from the head all 3904 * the times so there is no risk of missing some entries. 3905 * On the other hand, the risk is that we end up with 3906 * a very inconsistent ruleset, so better keep the lock 3907 * around the whole cycle. 3908 * 3909 * XXX this code can be improved by resetting the head of 3910 * the list to point to the default rule, and then freeing 3911 * the old list without the need for a lock. 3912 */ 3913 3914 IPFW_WLOCK(&layer3_chain); 3915 layer3_chain.reap = NULL; 3916 free_chain(&layer3_chain, 0 /* keep default rule */); 3917 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3918 IPFW_WUNLOCK(&layer3_chain); 3919 if (layer3_chain.reap != NULL) 3920 reap_rules(rule); 3921 break; 3922 3923 case IP_FW_ADD: 3924 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3925 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3926 sizeof(struct ip_fw) ); 3927 if (error == 0) 3928 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3929 if (error == 0) { 3930 error = add_rule(&layer3_chain, rule); 3931 size = RULESIZE(rule); 3932 if (!error && sopt->sopt_dir == SOPT_GET) 3933 error = sooptcopyout(sopt, rule, size); 3934 } 3935 free(rule, M_TEMP); 3936 break; 3937 3938 case IP_FW_DEL: 3939 /* 3940 * IP_FW_DEL is used for deleting single rules or sets, 3941 * and (ab)used to atomically manipulate sets. Argument size 3942 * is used to distinguish between the two: 3943 * sizeof(u_int32_t) 3944 * delete single rule or set of rules, 3945 * or reassign rules (or sets) to a different set. 3946 * 2*sizeof(u_int32_t) 3947 * atomic disable/enable sets. 3948 * first u_int32_t contains sets to be disabled, 3949 * second u_int32_t contains sets to be enabled. 3950 */ 3951 error = sooptcopyin(sopt, rulenum, 3952 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3953 if (error) 3954 break; 3955 size = sopt->sopt_valsize; 3956 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3957 error = del_entry(&layer3_chain, rulenum[0]); 3958 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3959 set_disable = 3960 (set_disable | rulenum[0]) & ~rulenum[1] & 3961 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3962 else 3963 error = EINVAL; 3964 break; 3965 3966 case IP_FW_ZERO: 3967 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3968 rule_num = 0; 3969 if (sopt->sopt_val != 0) { 3970 error = sooptcopyin(sopt, &rule_num, 3971 sizeof(int), sizeof(int)); 3972 if (error) 3973 break; 3974 } 3975 error = zero_entry(&layer3_chain, rule_num, 3976 sopt->sopt_name == IP_FW_RESETLOG); 3977 break; 3978 3979 case IP_FW_TABLE_ADD: 3980 { 3981 ipfw_table_entry ent; 3982 3983 error = sooptcopyin(sopt, &ent, 3984 sizeof(ent), sizeof(ent)); 3985 if (error) 3986 break; 3987 error = add_table_entry(ent.tbl, ent.addr, 3988 ent.masklen, ent.value); 3989 } 3990 break; 3991 3992 case IP_FW_TABLE_DEL: 3993 { 3994 ipfw_table_entry ent; 3995 3996 error = sooptcopyin(sopt, &ent, 3997 sizeof(ent), sizeof(ent)); 3998 if (error) 3999 break; 4000 error = del_table_entry(ent.tbl, ent.addr, ent.masklen); 4001 } 4002 break; 4003 4004 case IP_FW_TABLE_FLUSH: 4005 { 4006 u_int16_t tbl; 4007 4008 error = sooptcopyin(sopt, &tbl, 4009 sizeof(tbl), sizeof(tbl)); 4010 if (error) 4011 break; 4012 error = flush_table(tbl); 4013 } 4014 break; 4015 4016 case IP_FW_TABLE_GETSIZE: 4017 { 4018 u_int32_t tbl, cnt; 4019 4020 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4021 sizeof(tbl)))) 4022 break; 4023 if ((error = count_table(tbl, &cnt))) 4024 break; 4025 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4026 } 4027 break; 4028 4029 case IP_FW_TABLE_LIST: 4030 { 4031 ipfw_table *tbl; 4032 4033 if (sopt->sopt_valsize < sizeof(*tbl)) { 4034 error = EINVAL; 4035 break; 4036 } 4037 size = sopt->sopt_valsize; 4038 tbl = malloc(size, M_TEMP, M_WAITOK); 4039 if (tbl == NULL) { 4040 error = ENOMEM; 4041 break; 4042 } 4043 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4044 if (error) { 4045 free(tbl, M_TEMP); 4046 break; 4047 } 4048 tbl->size = (size - sizeof(*tbl)) / 4049 sizeof(ipfw_table_entry); 4050 error = dump_table(tbl); 4051 if (error) { 4052 free(tbl, M_TEMP); 4053 break; 4054 } 4055 error = sooptcopyout(sopt, tbl, size); 4056 free(tbl, M_TEMP); 4057 } 4058 break; 4059 4060 default: 4061 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4062 error = EINVAL; 4063 } 4064 4065 return (error); 4066#undef RULE_MAXSIZE 4067} 4068 4069/** 4070 * dummynet needs a reference to the default rule, because rules can be 4071 * deleted while packets hold a reference to them. When this happens, 4072 * dummynet changes the reference to the default rule (it could well be a 4073 * NULL pointer, but this way we do not need to check for the special 4074 * case, plus here he have info on the default behaviour). 4075 */ 4076struct ip_fw *ip_fw_default_rule; 4077 4078/* 4079 * This procedure is only used to handle keepalives. It is invoked 4080 * every dyn_keepalive_period 4081 */ 4082static void 4083ipfw_tick(void * __unused unused) 4084{ 4085 struct mbuf *m0, *m, *mnext, **mtailp; 4086 int i; 4087 ipfw_dyn_rule *q; 4088 4089 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4090 goto done; 4091 4092 /* 4093 * We make a chain of packets to go out here -- not deferring 4094 * until after we drop the IPFW dynamic rule lock would result 4095 * in a lock order reversal with the normal packet input -> ipfw 4096 * call stack. 4097 */ 4098 m0 = NULL; 4099 mtailp = &m0; 4100 IPFW_DYN_LOCK(); 4101 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4102 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4103 if (q->dyn_type == O_LIMIT_PARENT) 4104 continue; 4105 if (q->id.proto != IPPROTO_TCP) 4106 continue; 4107 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4108 continue; 4109 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4110 q->expire)) 4111 continue; /* too early */ 4112 if (TIME_LEQ(q->expire, time_uptime)) 4113 continue; /* too late, rule expired */ 4114 4115 *mtailp = send_pkt(&(q->id), q->ack_rev - 1, 4116 q->ack_fwd, TH_SYN); 4117 if (*mtailp != NULL) 4118 mtailp = &(*mtailp)->m_nextpkt; 4119 *mtailp = send_pkt(&(q->id), q->ack_fwd - 1, 4120 q->ack_rev, 0); 4121 if (*mtailp != NULL) 4122 mtailp = &(*mtailp)->m_nextpkt; 4123 } 4124 } 4125 IPFW_DYN_UNLOCK(); 4126 for (m = mnext = m0; m != NULL; m = mnext) { 4127 mnext = m->m_nextpkt; 4128 m->m_nextpkt = NULL; 4129 ip_output(m, NULL, NULL, 0, NULL, NULL); 4130 } 4131done: 4132 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4133} 4134 4135int 4136ipfw_init(void) 4137{ 4138 struct ip_fw default_rule; 4139 int error; 4140 4141#ifdef INET6 4142 /* Setup IPv6 fw sysctl tree. */ 4143 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4144 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4145 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4146 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4147 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4148 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4149 &fw_deny_unknown_exthdrs, 0, 4150 "Deny packets with unknown IPv6 Extension Headers"); 4151#endif 4152 4153 layer3_chain.rules = NULL; 4154 layer3_chain.want_write = 0; 4155 layer3_chain.busy_count = 0; 4156 cv_init(&layer3_chain.cv, "Condition variable for IPFW rw locks"); 4157 IPFW_LOCK_INIT(&layer3_chain); 4158 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4159 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4160 UMA_ALIGN_PTR, 0); 4161 IPFW_DYN_LOCK_INIT(); 4162 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4163 4164 bzero(&default_rule, sizeof default_rule); 4165 4166 default_rule.act_ofs = 0; 4167 default_rule.rulenum = IPFW_DEFAULT_RULE; 4168 default_rule.cmd_len = 1; 4169 default_rule.set = RESVD_SET; 4170 4171 default_rule.cmd[0].len = 1; 4172 default_rule.cmd[0].opcode = 4173#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4174 1 ? O_ACCEPT : 4175#endif 4176 O_DENY; 4177 4178 error = add_rule(&layer3_chain, &default_rule); 4179 if (error != 0) { 4180 printf("ipfw2: error %u initializing default rule " 4181 "(support disabled)\n", error); 4182 IPFW_DYN_LOCK_DESTROY(); 4183 IPFW_LOCK_DESTROY(&layer3_chain); 4184 return (error); 4185 } 4186 4187 ip_fw_default_rule = layer3_chain.rules; 4188 printf("ipfw2 (+ipv6) initialized, divert %s, " 4189 "rule-based forwarding " 4190#ifdef IPFIREWALL_FORWARD 4191 "enabled, " 4192#else 4193 "disabled, " 4194#endif 4195 "default to %s, logging ", 4196#ifdef IPDIVERT 4197 "enabled", 4198#else 4199 "loadable", 4200#endif 4201 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4202 4203#ifdef IPFIREWALL_VERBOSE 4204 fw_verbose = 1; 4205#endif 4206#ifdef IPFIREWALL_VERBOSE_LIMIT 4207 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4208#endif 4209 if (fw_verbose == 0) 4210 printf("disabled\n"); 4211 else if (verbose_limit == 0) 4212 printf("unlimited\n"); 4213 else 4214 printf("limited to %d packets/entry by default\n", 4215 verbose_limit); 4216 4217 init_tables(); 4218 ip_fw_ctl_ptr = ipfw_ctl; 4219 ip_fw_chk_ptr = ipfw_chk; 4220 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4221 4222 return (0); 4223} 4224 4225void 4226ipfw_destroy(void) 4227{ 4228 struct ip_fw *reap; 4229 4230 ip_fw_chk_ptr = NULL; 4231 ip_fw_ctl_ptr = NULL; 4232 callout_drain(&ipfw_timeout); 4233 IPFW_WLOCK(&layer3_chain); 4234 layer3_chain.reap = NULL; 4235 free_chain(&layer3_chain, 1 /* kill default rule */); 4236 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4237 IPFW_WUNLOCK(&layer3_chain); 4238 if (reap != NULL) 4239 reap_rules(reap); 4240 flush_tables(); 4241 IPFW_DYN_LOCK_DESTROY(); 4242 uma_zdestroy(ipfw_dyn_rule_zone); 4243 IPFW_LOCK_DESTROY(&layer3_chain); 4244 4245#ifdef INET6 4246 /* Free IPv6 fw sysctl tree. */ 4247 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4248#endif 4249 4250 printf("IP firewall unloaded\n"); 4251} 4252