ip_fw2.c revision 149052
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 149052 2005-08-14 18:20:33Z bz $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ip6fw.h" 38#include "opt_ipdn.h" 39#include "opt_inet.h" 40#include "opt_inet6.h" 41#include "opt_ipsec.h" 42#ifndef INET 43#error IPFIREWALL requires INET. 44#endif /* INET */ 45#endif 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 /* if ifp is provided, check for equality with rtentry */ 643 if (ifp != NULL && ro.ro_rt->rt_ifp != ifp) { 644 RTFREE(ro.ro_rt); 645 return 0; 646 } 647 648 /* if no ifp provided, check if rtentry is not default route */ 649 if (ifp == NULL && 650 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 651 RTFREE(ro.ro_rt); 652 return 0; 653 } 654 655 /* or if this is a blackhole/reject route */ 656 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 657 RTFREE(ro.ro_rt); 658 return 0; 659 } 660 661 /* found valid route */ 662 RTFREE(ro.ro_rt); 663 return 1; 664 665} 666static __inline int 667hash_packet6(struct ipfw_flow_id *id) 668{ 669 u_int32_t i; 670 i = (id->dst_ip6.__u6_addr.__u6_addr32[0]) ^ 671 (id->dst_ip6.__u6_addr.__u6_addr32[1]) ^ 672 (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 673 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 674 (id->dst_port) ^ (id->src_port) ^ (id->flow_id6); 675 return i; 676} 677 678static int 679is_icmp6_query(int icmp6_type) 680{ 681 if ((icmp6_type <= ICMP6_MAXTYPE) && 682 (icmp6_type == ICMP6_ECHO_REQUEST || 683 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 684 icmp6_type == ICMP6_WRUREQUEST || 685 icmp6_type == ICMP6_FQDN_QUERY || 686 icmp6_type == ICMP6_NI_QUERY)) 687 return (1); 688 689 return (0); 690} 691 692static void 693send_reject6(struct ip_fw_args *args, int code, u_short offset, u_int hlen) 694{ 695 if (code == ICMP6_UNREACH_RST && offset == 0 && 696 args->f_id.proto == IPPROTO_TCP) { 697 struct ip6_hdr *ip6; 698 struct tcphdr *tcp; 699 tcp_seq ack, seq; 700 int flags; 701 struct { 702 struct ip6_hdr ip6; 703 struct tcphdr th; 704 } ti; 705 706 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) { 707 args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr)); 708 if (args->m == NULL) 709 return; 710 } 711 712 ip6 = mtod(args->m, struct ip6_hdr *); 713 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 714 715 if ((tcp->th_flags & TH_RST) != 0) { 716 m_freem(args->m); 717 return; 718 } 719 720 ti.ip6 = *ip6; 721 ti.th = *tcp; 722 ti.th.th_seq = ntohl(ti.th.th_seq); 723 ti.th.th_ack = ntohl(ti.th.th_ack); 724 ti.ip6.ip6_nxt = IPPROTO_TCP; 725 726 if (ti.th.th_flags & TH_ACK) { 727 ack = 0; 728 seq = ti.th.th_ack; 729 flags = TH_RST; 730 } else { 731 ack = ti.th.th_seq; 732 if (((args->m)->m_flags & M_PKTHDR) != 0) { 733 ack += (args->m)->m_pkthdr.len - hlen 734 - (ti.th.th_off << 2); 735 } else if (ip6->ip6_plen) { 736 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) 737 - hlen - (ti.th.th_off << 2); 738 } else { 739 m_freem(args->m); 740 return; 741 } 742 if (tcp->th_flags & TH_SYN) 743 ack++; 744 seq = 0; 745 flags = TH_RST|TH_ACK; 746 } 747 bcopy(&ti, ip6, sizeof(ti)); 748 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1), 749 args->m, ack, seq, flags); 750 751 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 752 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0); 753 754 } else 755 m_freem(args->m); 756 757 args->m = NULL; 758} 759 760#endif /* INET6 */ 761 762static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 763 764#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 765#define SNP(buf) buf, sizeof(buf) 766 767/* 768 * We enter here when we have a rule with O_LOG. 769 * XXX this function alone takes about 2Kbytes of code! 770 */ 771static void 772ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 773 struct mbuf *m, struct ifnet *oif, u_short offset) 774{ 775 struct ether_header *eh = args->eh; 776 char *action; 777 int limit_reached = 0; 778 char action2[40], proto[128], fragment[32]; 779 780 fragment[0] = '\0'; 781 proto[0] = '\0'; 782 783 if (f == NULL) { /* bogus pkt */ 784 if (verbose_limit != 0 && norule_counter >= verbose_limit) 785 return; 786 norule_counter++; 787 if (norule_counter == verbose_limit) 788 limit_reached = verbose_limit; 789 action = "Refuse"; 790 } else { /* O_LOG is the first action, find the real one */ 791 ipfw_insn *cmd = ACTION_PTR(f); 792 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 793 794 if (l->max_log != 0 && l->log_left == 0) 795 return; 796 l->log_left--; 797 if (l->log_left == 0) 798 limit_reached = l->max_log; 799 cmd += F_LEN(cmd); /* point to first action */ 800 if (cmd->opcode == O_ALTQ) { 801 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 802 803 snprintf(SNPARGS(action2, 0), "Altq %d", 804 altq->qid); 805 cmd += F_LEN(cmd); 806 } 807 if (cmd->opcode == O_PROB) 808 cmd += F_LEN(cmd); 809 810 action = action2; 811 switch (cmd->opcode) { 812 case O_DENY: 813 action = "Deny"; 814 break; 815 816 case O_REJECT: 817 if (cmd->arg1==ICMP_REJECT_RST) 818 action = "Reset"; 819 else if (cmd->arg1==ICMP_UNREACH_HOST) 820 action = "Reject"; 821 else 822 snprintf(SNPARGS(action2, 0), "Unreach %d", 823 cmd->arg1); 824 break; 825 826 case O_UNREACH6: 827 if (cmd->arg1==ICMP6_UNREACH_RST) 828 action = "Reset"; 829 else 830 snprintf(SNPARGS(action2, 0), "Unreach %d", 831 cmd->arg1); 832 break; 833 834 case O_ACCEPT: 835 action = "Accept"; 836 break; 837 case O_COUNT: 838 action = "Count"; 839 break; 840 case O_DIVERT: 841 snprintf(SNPARGS(action2, 0), "Divert %d", 842 cmd->arg1); 843 break; 844 case O_TEE: 845 snprintf(SNPARGS(action2, 0), "Tee %d", 846 cmd->arg1); 847 break; 848 case O_SKIPTO: 849 snprintf(SNPARGS(action2, 0), "SkipTo %d", 850 cmd->arg1); 851 break; 852 case O_PIPE: 853 snprintf(SNPARGS(action2, 0), "Pipe %d", 854 cmd->arg1); 855 break; 856 case O_QUEUE: 857 snprintf(SNPARGS(action2, 0), "Queue %d", 858 cmd->arg1); 859 break; 860 case O_FORWARD_IP: { 861 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 862 int len; 863 864 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 865 inet_ntoa(sa->sa.sin_addr)); 866 if (sa->sa.sin_port) 867 snprintf(SNPARGS(action2, len), ":%d", 868 sa->sa.sin_port); 869 } 870 break; 871 case O_NETGRAPH: 872 snprintf(SNPARGS(action2, 0), "Netgraph %d", 873 cmd->arg1); 874 break; 875 case O_NGTEE: 876 snprintf(SNPARGS(action2, 0), "Ngtee %d", 877 cmd->arg1); 878 break; 879 default: 880 action = "UNKNOWN"; 881 break; 882 } 883 } 884 885 if (hlen == 0) { /* non-ip */ 886 snprintf(SNPARGS(proto, 0), "MAC"); 887 888 } else { 889 int len; 890 char src[48], dst[48]; 891 struct icmphdr *icmp; 892 struct tcphdr *tcp; 893 struct udphdr *udp; 894 /* Initialize to make compiler happy. */ 895 struct ip *ip = NULL; 896#ifdef INET6 897 struct ip6_hdr *ip6 = NULL; 898 struct icmp6_hdr *icmp6; 899#endif 900 src[0] = '\0'; 901 dst[0] = '\0'; 902#ifdef INET6 903 if (args->f_id.addr_type == 6) { 904 snprintf(src, sizeof(src), "[%s]", 905 ip6_sprintf(&args->f_id.src_ip6)); 906 snprintf(dst, sizeof(dst), "[%s]", 907 ip6_sprintf(&args->f_id.dst_ip6)); 908 909 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 910 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 911 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 912 } else 913#endif 914 { 915 ip = mtod(m, struct ip *); 916 tcp = L3HDR(struct tcphdr, ip); 917 udp = L3HDR(struct udphdr, ip); 918 919 inet_ntoa_r(ip->ip_src, src); 920 inet_ntoa_r(ip->ip_dst, dst); 921 } 922 923 switch (args->f_id.proto) { 924 case IPPROTO_TCP: 925 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 926 if (offset == 0) 927 snprintf(SNPARGS(proto, len), ":%d %s:%d", 928 ntohs(tcp->th_sport), 929 dst, 930 ntohs(tcp->th_dport)); 931 else 932 snprintf(SNPARGS(proto, len), " %s", dst); 933 break; 934 935 case IPPROTO_UDP: 936 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 937 if (offset == 0) 938 snprintf(SNPARGS(proto, len), ":%d %s:%d", 939 ntohs(udp->uh_sport), 940 dst, 941 ntohs(udp->uh_dport)); 942 else 943 snprintf(SNPARGS(proto, len), " %s", dst); 944 break; 945 946 case IPPROTO_ICMP: 947 icmp = L3HDR(struct icmphdr, ip); 948 if (offset == 0) 949 len = snprintf(SNPARGS(proto, 0), 950 "ICMP:%u.%u ", 951 icmp->icmp_type, icmp->icmp_code); 952 else 953 len = snprintf(SNPARGS(proto, 0), "ICMP "); 954 len += snprintf(SNPARGS(proto, len), "%s", src); 955 snprintf(SNPARGS(proto, len), " %s", dst); 956 break; 957#ifdef INET6 958 case IPPROTO_ICMPV6: 959 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 960 if (offset == 0) 961 len = snprintf(SNPARGS(proto, 0), 962 "ICMPv6:%u.%u ", 963 icmp6->icmp6_type, icmp6->icmp6_code); 964 else 965 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 966 len += snprintf(SNPARGS(proto, len), "%s", src); 967 snprintf(SNPARGS(proto, len), " %s", dst); 968 break; 969#endif 970 default: 971 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 972 args->f_id.proto, src); 973 snprintf(SNPARGS(proto, len), " %s", dst); 974 break; 975 } 976 977#ifdef INET6 978 if (args->f_id.addr_type == 6) { 979 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 980 snprintf(SNPARGS(fragment, 0), 981 " (frag %08x:%d@%d%s)", 982 args->f_id.frag_id6, 983 ntohs(ip6->ip6_plen) - hlen, 984 ntohs(offset & IP6F_OFF_MASK) << 3, 985 (offset & IP6F_MORE_FRAG) ? "+" : ""); 986 } else 987#endif 988 { 989 int ip_off, ip_len; 990 if (eh != NULL) { /* layer 2 packets are as on the wire */ 991 ip_off = ntohs(ip->ip_off); 992 ip_len = ntohs(ip->ip_len); 993 } else { 994 ip_off = ip->ip_off; 995 ip_len = ip->ip_len; 996 } 997 if (ip_off & (IP_MF | IP_OFFMASK)) 998 snprintf(SNPARGS(fragment, 0), 999 " (frag %d:%d@%d%s)", 1000 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1001 offset << 3, 1002 (ip_off & IP_MF) ? "+" : ""); 1003 } 1004 } 1005 if (oif || m->m_pkthdr.rcvif) 1006 log(LOG_SECURITY | LOG_INFO, 1007 "ipfw: %d %s %s %s via %s%s\n", 1008 f ? f->rulenum : -1, 1009 action, proto, oif ? "out" : "in", 1010 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1011 fragment); 1012 else 1013 log(LOG_SECURITY | LOG_INFO, 1014 "ipfw: %d %s %s [no if info]%s\n", 1015 f ? f->rulenum : -1, 1016 action, proto, fragment); 1017 if (limit_reached) 1018 log(LOG_SECURITY | LOG_NOTICE, 1019 "ipfw: limit %d reached on entry %d\n", 1020 limit_reached, f ? f->rulenum : -1); 1021} 1022 1023/* 1024 * IMPORTANT: the hash function for dynamic rules must be commutative 1025 * in source and destination (ip,port), because rules are bidirectional 1026 * and we want to find both in the same bucket. 1027 */ 1028static __inline int 1029hash_packet(struct ipfw_flow_id *id) 1030{ 1031 u_int32_t i; 1032 1033#ifdef INET6 1034 if (IS_IP6_FLOW_ID(id)) 1035 i = hash_packet6(id); 1036 else 1037#endif /* INET6 */ 1038 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1039 i &= (curr_dyn_buckets - 1); 1040 return i; 1041} 1042 1043/** 1044 * unlink a dynamic rule from a chain. prev is a pointer to 1045 * the previous one, q is a pointer to the rule to delete, 1046 * head is a pointer to the head of the queue. 1047 * Modifies q and potentially also head. 1048 */ 1049#define UNLINK_DYN_RULE(prev, head, q) { \ 1050 ipfw_dyn_rule *old_q = q; \ 1051 \ 1052 /* remove a refcount to the parent */ \ 1053 if (q->dyn_type == O_LIMIT) \ 1054 q->parent->count--; \ 1055 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1056 (q->id.src_ip), (q->id.src_port), \ 1057 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1058 if (prev != NULL) \ 1059 prev->next = q = q->next; \ 1060 else \ 1061 head = q = q->next; \ 1062 dyn_count--; \ 1063 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1064 1065#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1066 1067/** 1068 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1069 * 1070 * If keep_me == NULL, rules are deleted even if not expired, 1071 * otherwise only expired rules are removed. 1072 * 1073 * The value of the second parameter is also used to point to identify 1074 * a rule we absolutely do not want to remove (e.g. because we are 1075 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1076 * rules). The pointer is only used for comparison, so any non-null 1077 * value will do. 1078 */ 1079static void 1080remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1081{ 1082 static u_int32_t last_remove = 0; 1083 1084#define FORCE (keep_me == NULL) 1085 1086 ipfw_dyn_rule *prev, *q; 1087 int i, pass = 0, max_pass = 0; 1088 1089 IPFW_DYN_LOCK_ASSERT(); 1090 1091 if (ipfw_dyn_v == NULL || dyn_count == 0) 1092 return; 1093 /* do not expire more than once per second, it is useless */ 1094 if (!FORCE && last_remove == time_second) 1095 return; 1096 last_remove = time_second; 1097 1098 /* 1099 * because O_LIMIT refer to parent rules, during the first pass only 1100 * remove child and mark any pending LIMIT_PARENT, and remove 1101 * them in a second pass. 1102 */ 1103next_pass: 1104 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1105 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1106 /* 1107 * Logic can become complex here, so we split tests. 1108 */ 1109 if (q == keep_me) 1110 goto next; 1111 if (rule != NULL && rule != q->rule) 1112 goto next; /* not the one we are looking for */ 1113 if (q->dyn_type == O_LIMIT_PARENT) { 1114 /* 1115 * handle parent in the second pass, 1116 * record we need one. 1117 */ 1118 max_pass = 1; 1119 if (pass == 0) 1120 goto next; 1121 if (FORCE && q->count != 0 ) { 1122 /* XXX should not happen! */ 1123 printf("ipfw: OUCH! cannot remove rule," 1124 " count %d\n", q->count); 1125 } 1126 } else { 1127 if (!FORCE && 1128 !TIME_LEQ( q->expire, time_second )) 1129 goto next; 1130 } 1131 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1132 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1133 continue; 1134 } 1135next: 1136 prev=q; 1137 q=q->next; 1138 } 1139 } 1140 if (pass++ < max_pass) 1141 goto next_pass; 1142} 1143 1144 1145/** 1146 * lookup a dynamic rule. 1147 */ 1148static ipfw_dyn_rule * 1149lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1150 struct tcphdr *tcp) 1151{ 1152 /* 1153 * stateful ipfw extensions. 1154 * Lookup into dynamic session queue 1155 */ 1156#define MATCH_REVERSE 0 1157#define MATCH_FORWARD 1 1158#define MATCH_NONE 2 1159#define MATCH_UNKNOWN 3 1160 int i, dir = MATCH_NONE; 1161 ipfw_dyn_rule *prev, *q=NULL; 1162 1163 IPFW_DYN_LOCK_ASSERT(); 1164 1165 if (ipfw_dyn_v == NULL) 1166 goto done; /* not found */ 1167 i = hash_packet( pkt ); 1168 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1169 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1170 goto next; 1171 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 1172 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1173 continue; 1174 } 1175 if (pkt->proto == q->id.proto && 1176 q->dyn_type != O_LIMIT_PARENT) { 1177 if (IS_IP6_FLOW_ID(pkt)) { 1178 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1179 &(q->id.src_ip6)) && 1180 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1181 &(q->id.dst_ip6)) && 1182 pkt->src_port == q->id.src_port && 1183 pkt->dst_port == q->id.dst_port ) { 1184 dir = MATCH_FORWARD; 1185 break; 1186 } 1187 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1188 &(q->id.dst_ip6)) && 1189 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1190 &(q->id.src_ip6)) && 1191 pkt->src_port == q->id.dst_port && 1192 pkt->dst_port == q->id.src_port ) { 1193 dir = MATCH_REVERSE; 1194 break; 1195 } 1196 } else { 1197 if (pkt->src_ip == q->id.src_ip && 1198 pkt->dst_ip == q->id.dst_ip && 1199 pkt->src_port == q->id.src_port && 1200 pkt->dst_port == q->id.dst_port ) { 1201 dir = MATCH_FORWARD; 1202 break; 1203 } 1204 if (pkt->src_ip == q->id.dst_ip && 1205 pkt->dst_ip == q->id.src_ip && 1206 pkt->src_port == q->id.dst_port && 1207 pkt->dst_port == q->id.src_port ) { 1208 dir = MATCH_REVERSE; 1209 break; 1210 } 1211 } 1212 } 1213next: 1214 prev = q; 1215 q = q->next; 1216 } 1217 if (q == NULL) 1218 goto done; /* q = NULL, not found */ 1219 1220 if ( prev != NULL) { /* found and not in front */ 1221 prev->next = q->next; 1222 q->next = ipfw_dyn_v[i]; 1223 ipfw_dyn_v[i] = q; 1224 } 1225 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1226 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1227 1228#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1229#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1230 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1231 switch (q->state) { 1232 case TH_SYN: /* opening */ 1233 q->expire = time_second + dyn_syn_lifetime; 1234 break; 1235 1236 case BOTH_SYN: /* move to established */ 1237 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1238 case BOTH_SYN | (TH_FIN << 8) : 1239 if (tcp) { 1240#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1241 u_int32_t ack = ntohl(tcp->th_ack); 1242 if (dir == MATCH_FORWARD) { 1243 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1244 q->ack_fwd = ack; 1245 else { /* ignore out-of-sequence */ 1246 break; 1247 } 1248 } else { 1249 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1250 q->ack_rev = ack; 1251 else { /* ignore out-of-sequence */ 1252 break; 1253 } 1254 } 1255 } 1256 q->expire = time_second + dyn_ack_lifetime; 1257 break; 1258 1259 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1260 if (dyn_fin_lifetime >= dyn_keepalive_period) 1261 dyn_fin_lifetime = dyn_keepalive_period - 1; 1262 q->expire = time_second + dyn_fin_lifetime; 1263 break; 1264 1265 default: 1266#if 0 1267 /* 1268 * reset or some invalid combination, but can also 1269 * occur if we use keep-state the wrong way. 1270 */ 1271 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1272 printf("invalid state: 0x%x\n", q->state); 1273#endif 1274 if (dyn_rst_lifetime >= dyn_keepalive_period) 1275 dyn_rst_lifetime = dyn_keepalive_period - 1; 1276 q->expire = time_second + dyn_rst_lifetime; 1277 break; 1278 } 1279 } else if (pkt->proto == IPPROTO_UDP) { 1280 q->expire = time_second + dyn_udp_lifetime; 1281 } else { 1282 /* other protocols */ 1283 q->expire = time_second + dyn_short_lifetime; 1284 } 1285done: 1286 if (match_direction) 1287 *match_direction = dir; 1288 return q; 1289} 1290 1291static ipfw_dyn_rule * 1292lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1293 struct tcphdr *tcp) 1294{ 1295 ipfw_dyn_rule *q; 1296 1297 IPFW_DYN_LOCK(); 1298 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1299 if (q == NULL) 1300 IPFW_DYN_UNLOCK(); 1301 /* NB: return table locked when q is not NULL */ 1302 return q; 1303} 1304 1305static void 1306realloc_dynamic_table(void) 1307{ 1308 IPFW_DYN_LOCK_ASSERT(); 1309 1310 /* 1311 * Try reallocation, make sure we have a power of 2 and do 1312 * not allow more than 64k entries. In case of overflow, 1313 * default to 1024. 1314 */ 1315 1316 if (dyn_buckets > 65536) 1317 dyn_buckets = 1024; 1318 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1319 dyn_buckets = curr_dyn_buckets; /* reset */ 1320 return; 1321 } 1322 curr_dyn_buckets = dyn_buckets; 1323 if (ipfw_dyn_v != NULL) 1324 free(ipfw_dyn_v, M_IPFW); 1325 for (;;) { 1326 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1327 M_IPFW, M_NOWAIT | M_ZERO); 1328 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1329 break; 1330 curr_dyn_buckets /= 2; 1331 } 1332} 1333 1334/** 1335 * Install state of type 'type' for a dynamic session. 1336 * The hash table contains two type of rules: 1337 * - regular rules (O_KEEP_STATE) 1338 * - rules for sessions with limited number of sess per user 1339 * (O_LIMIT). When they are created, the parent is 1340 * increased by 1, and decreased on delete. In this case, 1341 * the third parameter is the parent rule and not the chain. 1342 * - "parent" rules for the above (O_LIMIT_PARENT). 1343 */ 1344static ipfw_dyn_rule * 1345add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1346{ 1347 ipfw_dyn_rule *r; 1348 int i; 1349 1350 IPFW_DYN_LOCK_ASSERT(); 1351 1352 if (ipfw_dyn_v == NULL || 1353 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1354 realloc_dynamic_table(); 1355 if (ipfw_dyn_v == NULL) 1356 return NULL; /* failed ! */ 1357 } 1358 i = hash_packet(id); 1359 1360 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1361 if (r == NULL) { 1362 printf ("ipfw: sorry cannot allocate state\n"); 1363 return NULL; 1364 } 1365 1366 /* increase refcount on parent, and set pointer */ 1367 if (dyn_type == O_LIMIT) { 1368 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1369 if ( parent->dyn_type != O_LIMIT_PARENT) 1370 panic("invalid parent"); 1371 parent->count++; 1372 r->parent = parent; 1373 rule = parent->rule; 1374 } 1375 1376 r->id = *id; 1377 r->expire = time_second + dyn_syn_lifetime; 1378 r->rule = rule; 1379 r->dyn_type = dyn_type; 1380 r->pcnt = r->bcnt = 0; 1381 r->count = 0; 1382 1383 r->bucket = i; 1384 r->next = ipfw_dyn_v[i]; 1385 ipfw_dyn_v[i] = r; 1386 dyn_count++; 1387 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1388 dyn_type, 1389 (r->id.src_ip), (r->id.src_port), 1390 (r->id.dst_ip), (r->id.dst_port), 1391 dyn_count ); ) 1392 return r; 1393} 1394 1395/** 1396 * lookup dynamic parent rule using pkt and rule as search keys. 1397 * If the lookup fails, then install one. 1398 */ 1399static ipfw_dyn_rule * 1400lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1401{ 1402 ipfw_dyn_rule *q; 1403 int i; 1404 1405 IPFW_DYN_LOCK_ASSERT(); 1406 1407 if (ipfw_dyn_v) { 1408 int is_v6 = IS_IP6_FLOW_ID(pkt); 1409 i = hash_packet( pkt ); 1410 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1411 if (q->dyn_type == O_LIMIT_PARENT && 1412 rule== q->rule && 1413 pkt->proto == q->id.proto && 1414 pkt->src_port == q->id.src_port && 1415 pkt->dst_port == q->id.dst_port && 1416 ( 1417 (is_v6 && 1418 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1419 &(q->id.src_ip6)) && 1420 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1421 &(q->id.dst_ip6))) || 1422 (!is_v6 && 1423 pkt->src_ip == q->id.src_ip && 1424 pkt->dst_ip == q->id.dst_ip) 1425 ) 1426 ) { 1427 q->expire = time_second + dyn_short_lifetime; 1428 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1429 return q; 1430 } 1431 } 1432 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1433} 1434 1435/** 1436 * Install dynamic state for rule type cmd->o.opcode 1437 * 1438 * Returns 1 (failure) if state is not installed because of errors or because 1439 * session limitations are enforced. 1440 */ 1441static int 1442install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1443 struct ip_fw_args *args) 1444{ 1445 static int last_log; 1446 1447 ipfw_dyn_rule *q; 1448 1449 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 1450 cmd->o.opcode, 1451 (args->f_id.src_ip), (args->f_id.src_port), 1452 (args->f_id.dst_ip), (args->f_id.dst_port) );) 1453 1454 IPFW_DYN_LOCK(); 1455 1456 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1457 1458 if (q != NULL) { /* should never occur */ 1459 if (last_log != time_second) { 1460 last_log = time_second; 1461 printf("ipfw: install_state: entry already present, done\n"); 1462 } 1463 IPFW_DYN_UNLOCK(); 1464 return 0; 1465 } 1466 1467 if (dyn_count >= dyn_max) 1468 /* 1469 * Run out of slots, try to remove any expired rule. 1470 */ 1471 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1472 1473 if (dyn_count >= dyn_max) { 1474 if (last_log != time_second) { 1475 last_log = time_second; 1476 printf("ipfw: install_state: Too many dynamic rules\n"); 1477 } 1478 IPFW_DYN_UNLOCK(); 1479 return 1; /* cannot install, notify caller */ 1480 } 1481 1482 switch (cmd->o.opcode) { 1483 case O_KEEP_STATE: /* bidir rule */ 1484 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1485 break; 1486 1487 case O_LIMIT: /* limit number of sessions */ 1488 { 1489 u_int16_t limit_mask = cmd->limit_mask; 1490 struct ipfw_flow_id id; 1491 ipfw_dyn_rule *parent; 1492 1493 DEB(printf("ipfw: installing dyn-limit rule %d\n", 1494 cmd->conn_limit);) 1495 1496 id.dst_ip = id.src_ip = 0; 1497 id.dst_port = id.src_port = 0; 1498 id.proto = args->f_id.proto; 1499 1500 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1501 if (limit_mask & DYN_SRC_ADDR) 1502 id.src_ip6 = args->f_id.src_ip6; 1503 if (limit_mask & DYN_DST_ADDR) 1504 id.dst_ip6 = args->f_id.dst_ip6; 1505 } else { 1506 if (limit_mask & DYN_SRC_ADDR) 1507 id.src_ip = args->f_id.src_ip; 1508 if (limit_mask & DYN_DST_ADDR) 1509 id.dst_ip = args->f_id.dst_ip; 1510 } 1511 if (limit_mask & DYN_SRC_PORT) 1512 id.src_port = args->f_id.src_port; 1513 if (limit_mask & DYN_DST_PORT) 1514 id.dst_port = args->f_id.dst_port; 1515 parent = lookup_dyn_parent(&id, rule); 1516 if (parent == NULL) { 1517 printf("ipfw: add parent failed\n"); 1518 return 1; 1519 } 1520 if (parent->count >= cmd->conn_limit) { 1521 /* 1522 * See if we can remove some expired rule. 1523 */ 1524 remove_dyn_rule(rule, parent); 1525 if (parent->count >= cmd->conn_limit) { 1526 if (fw_verbose && last_log != time_second) { 1527 last_log = time_second; 1528 log(LOG_SECURITY | LOG_DEBUG, 1529 "drop session, too many entries\n"); 1530 } 1531 IPFW_DYN_UNLOCK(); 1532 return 1; 1533 } 1534 } 1535 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1536 } 1537 break; 1538 default: 1539 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 1540 IPFW_DYN_UNLOCK(); 1541 return 1; 1542 } 1543 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1544 IPFW_DYN_UNLOCK(); 1545 return 0; 1546} 1547 1548/* 1549 * Generate a TCP packet, containing either a RST or a keepalive. 1550 * When flags & TH_RST, we are sending a RST packet, because of a 1551 * "reset" action matched the packet. 1552 * Otherwise we are sending a keepalive, and flags & TH_ 1553 */ 1554static struct mbuf * 1555send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1556{ 1557 struct mbuf *m; 1558 struct ip *ip; 1559 struct tcphdr *tcp; 1560 1561 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1562 if (m == 0) 1563 return (NULL); 1564 m->m_pkthdr.rcvif = (struct ifnet *)0; 1565 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1566 m->m_data += max_linkhdr; 1567 1568 ip = mtod(m, struct ip *); 1569 bzero(ip, m->m_len); 1570 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1571 ip->ip_p = IPPROTO_TCP; 1572 tcp->th_off = 5; 1573 /* 1574 * Assume we are sending a RST (or a keepalive in the reverse 1575 * direction), swap src and destination addresses and ports. 1576 */ 1577 ip->ip_src.s_addr = htonl(id->dst_ip); 1578 ip->ip_dst.s_addr = htonl(id->src_ip); 1579 tcp->th_sport = htons(id->dst_port); 1580 tcp->th_dport = htons(id->src_port); 1581 if (flags & TH_RST) { /* we are sending a RST */ 1582 if (flags & TH_ACK) { 1583 tcp->th_seq = htonl(ack); 1584 tcp->th_ack = htonl(0); 1585 tcp->th_flags = TH_RST; 1586 } else { 1587 if (flags & TH_SYN) 1588 seq++; 1589 tcp->th_seq = htonl(0); 1590 tcp->th_ack = htonl(seq); 1591 tcp->th_flags = TH_RST | TH_ACK; 1592 } 1593 } else { 1594 /* 1595 * We are sending a keepalive. flags & TH_SYN determines 1596 * the direction, forward if set, reverse if clear. 1597 * NOTE: seq and ack are always assumed to be correct 1598 * as set by the caller. This may be confusing... 1599 */ 1600 if (flags & TH_SYN) { 1601 /* 1602 * we have to rewrite the correct addresses! 1603 */ 1604 ip->ip_dst.s_addr = htonl(id->dst_ip); 1605 ip->ip_src.s_addr = htonl(id->src_ip); 1606 tcp->th_dport = htons(id->dst_port); 1607 tcp->th_sport = htons(id->src_port); 1608 } 1609 tcp->th_seq = htonl(seq); 1610 tcp->th_ack = htonl(ack); 1611 tcp->th_flags = TH_ACK; 1612 } 1613 /* 1614 * set ip_len to the payload size so we can compute 1615 * the tcp checksum on the pseudoheader 1616 * XXX check this, could save a couple of words ? 1617 */ 1618 ip->ip_len = htons(sizeof(struct tcphdr)); 1619 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1620 /* 1621 * now fill fields left out earlier 1622 */ 1623 ip->ip_ttl = ip_defttl; 1624 ip->ip_len = m->m_pkthdr.len; 1625 m->m_flags |= M_SKIP_FIREWALL; 1626 return (m); 1627} 1628 1629/* 1630 * sends a reject message, consuming the mbuf passed as an argument. 1631 */ 1632static void 1633send_reject(struct ip_fw_args *args, int code, u_short offset, int ip_len) 1634{ 1635 1636 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1637 /* We need the IP header in host order for icmp_error(). */ 1638 if (args->eh != NULL) { 1639 struct ip *ip = mtod(args->m, struct ip *); 1640 ip->ip_len = ntohs(ip->ip_len); 1641 ip->ip_off = ntohs(ip->ip_off); 1642 } 1643 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1644 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1645 struct tcphdr *const tcp = 1646 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1647 if ( (tcp->th_flags & TH_RST) == 0) { 1648 struct mbuf *m; 1649 m = send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1650 ntohl(tcp->th_ack), 1651 tcp->th_flags | TH_RST); 1652 if (m != NULL) 1653 ip_output(m, NULL, NULL, 0, NULL, NULL); 1654 } 1655 m_freem(args->m); 1656 } else 1657 m_freem(args->m); 1658 args->m = NULL; 1659} 1660 1661/** 1662 * 1663 * Given an ip_fw *, lookup_next_rule will return a pointer 1664 * to the next rule, which can be either the jump 1665 * target (for skipto instructions) or the next one in the list (in 1666 * all other cases including a missing jump target). 1667 * The result is also written in the "next_rule" field of the rule. 1668 * Backward jumps are not allowed, so start looking from the next 1669 * rule... 1670 * 1671 * This never returns NULL -- in case we do not have an exact match, 1672 * the next rule is returned. When the ruleset is changed, 1673 * pointers are flushed so we are always correct. 1674 */ 1675 1676static struct ip_fw * 1677lookup_next_rule(struct ip_fw *me) 1678{ 1679 struct ip_fw *rule = NULL; 1680 ipfw_insn *cmd; 1681 1682 /* look for action, in case it is a skipto */ 1683 cmd = ACTION_PTR(me); 1684 if (cmd->opcode == O_LOG) 1685 cmd += F_LEN(cmd); 1686 if (cmd->opcode == O_ALTQ) 1687 cmd += F_LEN(cmd); 1688 if ( cmd->opcode == O_SKIPTO ) 1689 for (rule = me->next; rule ; rule = rule->next) 1690 if (rule->rulenum >= cmd->arg1) 1691 break; 1692 if (rule == NULL) /* failure or not a skipto */ 1693 rule = me->next; 1694 me->next_rule = rule; 1695 return rule; 1696} 1697 1698static void 1699init_tables(void) 1700{ 1701 int i; 1702 1703 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1704 rn_inithead((void **)&ipfw_tables[i].rnh, 32); 1705 ipfw_tables[i].modified = 1; 1706 } 1707} 1708 1709static int 1710add_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen, u_int32_t value) 1711{ 1712 struct radix_node_head *rnh; 1713 struct table_entry *ent; 1714 1715 if (tbl >= IPFW_TABLES_MAX) 1716 return (EINVAL); 1717 rnh = ipfw_tables[tbl].rnh; 1718 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1719 if (ent == NULL) 1720 return (ENOMEM); 1721 ent->value = value; 1722 ent->addr.sin_len = ent->mask.sin_len = 8; 1723 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1724 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1725 RADIX_NODE_HEAD_LOCK(rnh); 1726 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1727 NULL) { 1728 RADIX_NODE_HEAD_UNLOCK(rnh); 1729 free(ent, M_IPFW_TBL); 1730 return (EEXIST); 1731 } 1732 ipfw_tables[tbl].modified = 1; 1733 RADIX_NODE_HEAD_UNLOCK(rnh); 1734 return (0); 1735} 1736 1737static int 1738del_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen) 1739{ 1740 struct radix_node_head *rnh; 1741 struct table_entry *ent; 1742 struct sockaddr_in sa, mask; 1743 1744 if (tbl >= IPFW_TABLES_MAX) 1745 return (EINVAL); 1746 rnh = ipfw_tables[tbl].rnh; 1747 sa.sin_len = mask.sin_len = 8; 1748 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1749 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1750 RADIX_NODE_HEAD_LOCK(rnh); 1751 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1752 if (ent == NULL) { 1753 RADIX_NODE_HEAD_UNLOCK(rnh); 1754 return (ESRCH); 1755 } 1756 ipfw_tables[tbl].modified = 1; 1757 RADIX_NODE_HEAD_UNLOCK(rnh); 1758 free(ent, M_IPFW_TBL); 1759 return (0); 1760} 1761 1762static int 1763flush_table_entry(struct radix_node *rn, void *arg) 1764{ 1765 struct radix_node_head * const rnh = arg; 1766 struct table_entry *ent; 1767 1768 ent = (struct table_entry *) 1769 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1770 if (ent != NULL) 1771 free(ent, M_IPFW_TBL); 1772 return (0); 1773} 1774 1775static int 1776flush_table(u_int16_t tbl) 1777{ 1778 struct radix_node_head *rnh; 1779 1780 if (tbl >= IPFW_TABLES_MAX) 1781 return (EINVAL); 1782 rnh = ipfw_tables[tbl].rnh; 1783 RADIX_NODE_HEAD_LOCK(rnh); 1784 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1785 ipfw_tables[tbl].modified = 1; 1786 RADIX_NODE_HEAD_UNLOCK(rnh); 1787 return (0); 1788} 1789 1790static void 1791flush_tables(void) 1792{ 1793 u_int16_t tbl; 1794 1795 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1796 flush_table(tbl); 1797} 1798 1799static int 1800lookup_table(u_int16_t tbl, in_addr_t addr, u_int32_t *val) 1801{ 1802 struct radix_node_head *rnh; 1803 struct ip_fw_table *table; 1804 struct table_entry *ent; 1805 struct sockaddr_in sa; 1806 int last_match; 1807 1808 if (tbl >= IPFW_TABLES_MAX) 1809 return (0); 1810 table = &ipfw_tables[tbl]; 1811 rnh = table->rnh; 1812 RADIX_NODE_HEAD_LOCK(rnh); 1813 if (addr == table->last_addr && !table->modified) { 1814 last_match = table->last_match; 1815 if (last_match) 1816 *val = table->last_value; 1817 RADIX_NODE_HEAD_UNLOCK(rnh); 1818 return (last_match); 1819 } 1820 table->modified = 0; 1821 sa.sin_len = 8; 1822 sa.sin_addr.s_addr = addr; 1823 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1824 table->last_addr = addr; 1825 if (ent != NULL) { 1826 table->last_value = *val = ent->value; 1827 table->last_match = 1; 1828 RADIX_NODE_HEAD_UNLOCK(rnh); 1829 return (1); 1830 } 1831 table->last_match = 0; 1832 RADIX_NODE_HEAD_UNLOCK(rnh); 1833 return (0); 1834} 1835 1836static int 1837count_table_entry(struct radix_node *rn, void *arg) 1838{ 1839 u_int32_t * const cnt = arg; 1840 1841 (*cnt)++; 1842 return (0); 1843} 1844 1845static int 1846count_table(u_int32_t tbl, u_int32_t *cnt) 1847{ 1848 struct radix_node_head *rnh; 1849 1850 if (tbl >= IPFW_TABLES_MAX) 1851 return (EINVAL); 1852 rnh = ipfw_tables[tbl].rnh; 1853 *cnt = 0; 1854 RADIX_NODE_HEAD_LOCK(rnh); 1855 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1856 RADIX_NODE_HEAD_UNLOCK(rnh); 1857 return (0); 1858} 1859 1860static int 1861dump_table_entry(struct radix_node *rn, void *arg) 1862{ 1863 struct table_entry * const n = (struct table_entry *)rn; 1864 ipfw_table * const tbl = arg; 1865 ipfw_table_entry *ent; 1866 1867 if (tbl->cnt == tbl->size) 1868 return (1); 1869 ent = &tbl->ent[tbl->cnt]; 1870 ent->tbl = tbl->tbl; 1871 if (in_nullhost(n->mask.sin_addr)) 1872 ent->masklen = 0; 1873 else 1874 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1875 ent->addr = n->addr.sin_addr.s_addr; 1876 ent->value = n->value; 1877 tbl->cnt++; 1878 return (0); 1879} 1880 1881static int 1882dump_table(ipfw_table *tbl) 1883{ 1884 struct radix_node_head *rnh; 1885 1886 if (tbl->tbl >= IPFW_TABLES_MAX) 1887 return (EINVAL); 1888 rnh = ipfw_tables[tbl->tbl].rnh; 1889 tbl->cnt = 0; 1890 RADIX_NODE_HEAD_LOCK(rnh); 1891 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1892 RADIX_NODE_HEAD_UNLOCK(rnh); 1893 return (0); 1894} 1895 1896static void 1897fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1898{ 1899 struct ucred *cr; 1900 1901 if (inp->inp_socket != NULL) { 1902 cr = inp->inp_socket->so_cred; 1903 ugp->fw_prid = jailed(cr) ? 1904 cr->cr_prison->pr_id : -1; 1905 ugp->fw_uid = cr->cr_uid; 1906 ugp->fw_ngroups = cr->cr_ngroups; 1907 bcopy(cr->cr_groups, ugp->fw_groups, 1908 sizeof(ugp->fw_groups)); 1909 } 1910} 1911 1912static int 1913check_uidgid(ipfw_insn_u32 *insn, 1914 int proto, struct ifnet *oif, 1915 struct in_addr dst_ip, u_int16_t dst_port, 1916 struct in_addr src_ip, u_int16_t src_port, 1917 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1918{ 1919 struct inpcbinfo *pi; 1920 int wildcard; 1921 struct inpcb *pcb; 1922 int match; 1923 gid_t *gp; 1924 1925 /* 1926 * Check to see if the UDP or TCP stack supplied us with 1927 * the PCB. If so, rather then holding a lock and looking 1928 * up the PCB, we can use the one that was supplied. 1929 */ 1930 if (inp && *lookup == 0) { 1931 INP_LOCK_ASSERT(inp); 1932 if (inp->inp_socket != NULL) { 1933 fill_ugid_cache(inp, ugp); 1934 *lookup = 1; 1935 } 1936 } 1937 /* 1938 * If we have already been here and the packet has no 1939 * PCB entry associated with it, then we can safely 1940 * assume that this is a no match. 1941 */ 1942 if (*lookup == -1) 1943 return (0); 1944 if (proto == IPPROTO_TCP) { 1945 wildcard = 0; 1946 pi = &tcbinfo; 1947 } else if (proto == IPPROTO_UDP) { 1948 wildcard = 1; 1949 pi = &udbinfo; 1950 } else 1951 return 0; 1952 match = 0; 1953 if (*lookup == 0) { 1954 INP_INFO_RLOCK(pi); 1955 pcb = (oif) ? 1956 in_pcblookup_hash(pi, 1957 dst_ip, htons(dst_port), 1958 src_ip, htons(src_port), 1959 wildcard, oif) : 1960 in_pcblookup_hash(pi, 1961 src_ip, htons(src_port), 1962 dst_ip, htons(dst_port), 1963 wildcard, NULL); 1964 if (pcb != NULL) { 1965 INP_LOCK(pcb); 1966 if (pcb->inp_socket != NULL) { 1967 fill_ugid_cache(pcb, ugp); 1968 *lookup = 1; 1969 } 1970 INP_UNLOCK(pcb); 1971 } 1972 INP_INFO_RUNLOCK(pi); 1973 if (*lookup == 0) { 1974 /* 1975 * If the lookup did not yield any results, there 1976 * is no sense in coming back and trying again. So 1977 * we can set lookup to -1 and ensure that we wont 1978 * bother the pcb system again. 1979 */ 1980 *lookup = -1; 1981 return (0); 1982 } 1983 } 1984 if (insn->o.opcode == O_UID) 1985 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1986 else if (insn->o.opcode == O_GID) { 1987 for (gp = ugp->fw_groups; 1988 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1989 if (*gp == (gid_t)insn->d[0]) { 1990 match = 1; 1991 break; 1992 } 1993 } else if (insn->o.opcode == O_JAIL) 1994 match = (ugp->fw_prid == (int)insn->d[0]); 1995 return match; 1996} 1997 1998/* 1999 * The main check routine for the firewall. 2000 * 2001 * All arguments are in args so we can modify them and return them 2002 * back to the caller. 2003 * 2004 * Parameters: 2005 * 2006 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2007 * Starts with the IP header. 2008 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2009 * args->oif Outgoing interface, or NULL if packet is incoming. 2010 * The incoming interface is in the mbuf. (in) 2011 * args->divert_rule (in/out) 2012 * Skip up to the first rule past this rule number; 2013 * upon return, non-zero port number for divert or tee. 2014 * 2015 * args->rule Pointer to the last matching rule (in/out) 2016 * args->next_hop Socket we are forwarding to (out). 2017 * args->f_id Addresses grabbed from the packet (out) 2018 * args->cookie a cookie depending on rule action 2019 * 2020 * Return value: 2021 * 2022 * IP_FW_PASS the packet must be accepted 2023 * IP_FW_DENY the packet must be dropped 2024 * IP_FW_DIVERT divert packet, port in m_tag 2025 * IP_FW_TEE tee packet, port in m_tag 2026 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2027 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2028 * 2029 */ 2030 2031int 2032ipfw_chk(struct ip_fw_args *args) 2033{ 2034 /* 2035 * Local variables hold state during the processing of a packet. 2036 * 2037 * IMPORTANT NOTE: to speed up the processing of rules, there 2038 * are some assumption on the values of the variables, which 2039 * are documented here. Should you change them, please check 2040 * the implementation of the various instructions to make sure 2041 * that they still work. 2042 * 2043 * args->eh The MAC header. It is non-null for a layer2 2044 * packet, it is NULL for a layer-3 packet. 2045 * 2046 * m | args->m Pointer to the mbuf, as received from the caller. 2047 * It may change if ipfw_chk() does an m_pullup, or if it 2048 * consumes the packet because it calls send_reject(). 2049 * XXX This has to change, so that ipfw_chk() never modifies 2050 * or consumes the buffer. 2051 * ip is simply an alias of the value of m, and it is kept 2052 * in sync with it (the packet is supposed to start with 2053 * the ip header). 2054 */ 2055 struct mbuf *m = args->m; 2056 struct ip *ip = mtod(m, struct ip *); 2057 2058 /* 2059 * For rules which contain uid/gid or jail constraints, cache 2060 * a copy of the users credentials after the pcb lookup has been 2061 * executed. This will speed up the processing of rules with 2062 * these types of constraints, as well as decrease contention 2063 * on pcb related locks. 2064 */ 2065 struct ip_fw_ugid fw_ugid_cache; 2066 int ugid_lookup = 0; 2067 2068 /* 2069 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2070 * associated with a packet input on a divert socket. This 2071 * will allow to distinguish traffic and its direction when 2072 * it originates from a divert socket. 2073 */ 2074 u_int divinput_flags = 0; 2075 2076 /* 2077 * oif | args->oif If NULL, ipfw_chk has been called on the 2078 * inbound path (ether_input, bdg_forward, ip_input). 2079 * If non-NULL, ipfw_chk has been called on the outbound path 2080 * (ether_output, ip_output). 2081 */ 2082 struct ifnet *oif = args->oif; 2083 2084 struct ip_fw *f = NULL; /* matching rule */ 2085 int retval = 0; 2086 2087 /* 2088 * hlen The length of the IP header. 2089 */ 2090 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2091 2092 /* 2093 * offset The offset of a fragment. offset != 0 means that 2094 * we have a fragment at this offset of an IPv4 packet. 2095 * offset == 0 means that (if this is an IPv4 packet) 2096 * this is the first or only fragment. 2097 * For IPv6 offset == 0 means there is no Fragment Header. 2098 * If offset != 0 for IPv6 always use correct mask to 2099 * get the correct offset because we add IP6F_MORE_FRAG 2100 * to be able to dectect the first fragment which would 2101 * otherwise have offset = 0. 2102 */ 2103 u_short offset = 0; 2104 2105 /* 2106 * Local copies of addresses. They are only valid if we have 2107 * an IP packet. 2108 * 2109 * proto The protocol. Set to 0 for non-ip packets, 2110 * or to the protocol read from the packet otherwise. 2111 * proto != 0 means that we have an IPv4 packet. 2112 * 2113 * src_port, dst_port port numbers, in HOST format. Only 2114 * valid for TCP and UDP packets. 2115 * 2116 * src_ip, dst_ip ip addresses, in NETWORK format. 2117 * Only valid for IPv4 packets. 2118 */ 2119 u_int8_t proto; 2120 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2121 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2122 u_int16_t ip_len=0; 2123 int pktlen; 2124 2125 /* 2126 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2127 * MATCH_NONE when checked and not matched (q = NULL), 2128 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2129 */ 2130 int dyn_dir = MATCH_UNKNOWN; 2131 ipfw_dyn_rule *q = NULL; 2132 struct ip_fw_chain *chain = &layer3_chain; 2133 struct m_tag *mtag; 2134 2135 /* 2136 * We store in ulp a pointer to the upper layer protocol header. 2137 * In the ipv4 case this is easy to determine from the header, 2138 * but for ipv6 we might have some additional headers in the middle. 2139 * ulp is NULL if not found. 2140 */ 2141 void *ulp = NULL; /* upper layer protocol pointer. */ 2142 /* XXX ipv6 variables */ 2143 int is_ipv6 = 0; 2144 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2145 /* end of ipv6 variables */ 2146 int is_ipv4 = 0; 2147 2148 if (m->m_flags & M_SKIP_FIREWALL) 2149 return (IP_FW_PASS); /* accept */ 2150 2151 pktlen = m->m_pkthdr.len; 2152 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2153 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2154 2155/* 2156 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2157 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2158 * pointer might become stale after other pullups (but we never use it 2159 * this way). 2160 */ 2161#define PULLUP_TO(len, p, T) \ 2162do { \ 2163 int x = (len) + sizeof(T); \ 2164 if ((m)->m_len < x) { \ 2165 args->m = m = m_pullup(m, x); \ 2166 if (m == NULL) \ 2167 goto pullup_failed; \ 2168 } \ 2169 p = (mtod(m, char *) + (len)); \ 2170} while (0) 2171 2172 /* Identify IP packets and fill up variables. */ 2173 if (pktlen >= sizeof(struct ip6_hdr) && 2174 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2175 mtod(m, struct ip *)->ip_v == 6) { 2176 is_ipv6 = 1; 2177 args->f_id.addr_type = 6; 2178 hlen = sizeof(struct ip6_hdr); 2179 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2180 2181 /* Search extension headers to find upper layer protocols */ 2182 while (ulp == NULL) { 2183 switch (proto) { 2184 case IPPROTO_ICMPV6: 2185 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2186 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2187 break; 2188 2189 case IPPROTO_TCP: 2190 PULLUP_TO(hlen, ulp, struct tcphdr); 2191 dst_port = TCP(ulp)->th_dport; 2192 src_port = TCP(ulp)->th_sport; 2193 args->f_id.flags = TCP(ulp)->th_flags; 2194 break; 2195 2196 case IPPROTO_UDP: 2197 PULLUP_TO(hlen, ulp, struct udphdr); 2198 dst_port = UDP(ulp)->uh_dport; 2199 src_port = UDP(ulp)->uh_sport; 2200 break; 2201 2202 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2203 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2204 ext_hd |= EXT_HOPOPTS; 2205 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2206 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2207 ulp = NULL; 2208 break; 2209 2210 case IPPROTO_ROUTING: /* RFC 2460 */ 2211 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2212 if (((struct ip6_rthdr *)ulp)->ip6r_type != 0) { 2213 printf("IPFW2: IPV6 - Unknown Routing " 2214 "Header type(%d)\n", 2215 ((struct ip6_rthdr *)ulp)->ip6r_type); 2216 if (fw_deny_unknown_exthdrs) 2217 return (IP_FW_DENY); 2218 break; 2219 } 2220 ext_hd |= EXT_ROUTING; 2221 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2222 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2223 ulp = NULL; 2224 break; 2225 2226 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2227 PULLUP_TO(hlen, ulp, struct ip6_frag); 2228 ext_hd |= EXT_FRAGMENT; 2229 hlen += sizeof (struct ip6_frag); 2230 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2231 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2232 IP6F_OFF_MASK; 2233 /* Add IP6F_MORE_FRAG for offset of first 2234 * fragment to be != 0. */ 2235 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2236 IP6F_MORE_FRAG; 2237 if (offset == 0) { 2238 printf("IPFW2: IPV6 - Invalid Fragment " 2239 "Header\n"); 2240 if (fw_deny_unknown_exthdrs) 2241 return (IP_FW_DENY); 2242 break; 2243 } 2244 args->f_id.frag_id6 = 2245 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2246 ulp = NULL; 2247 break; 2248 2249 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2250 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2251 ext_hd |= EXT_DSTOPTS; 2252 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2253 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2254 ulp = NULL; 2255 break; 2256 2257 case IPPROTO_AH: /* RFC 2402 */ 2258 PULLUP_TO(hlen, ulp, struct ip6_ext); 2259 ext_hd |= EXT_AH; 2260 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2261 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2262 ulp = NULL; 2263 break; 2264 2265 case IPPROTO_ESP: /* RFC 2406 */ 2266 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2267 /* Anything past Seq# is variable length and 2268 * data past this ext. header is encrypted. */ 2269 ext_hd |= EXT_ESP; 2270 break; 2271 2272 case IPPROTO_NONE: /* RFC 2460 */ 2273 PULLUP_TO(hlen, ulp, struct ip6_ext); 2274 /* Packet ends here. if ip6e_len!=0 octets 2275 * must be ignored. */ 2276 break; 2277 2278 case IPPROTO_OSPFIGP: 2279 /* XXX OSPF header check? */ 2280 PULLUP_TO(hlen, ulp, struct ip6_ext); 2281 break; 2282 2283 default: 2284 printf("IPFW2: IPV6 - Unknown Extension " 2285 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2286 if (fw_deny_unknown_exthdrs) 2287 return (IP_FW_DENY); 2288 break; 2289 } /*switch */ 2290 } 2291 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2292 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2293 args->f_id.src_ip = 0; 2294 args->f_id.dst_ip = 0; 2295 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2296 } else if (pktlen >= sizeof(struct ip) && 2297 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2298 mtod(m, struct ip *)->ip_v == 4) { 2299 is_ipv4 = 1; 2300 ip = mtod(m, struct ip *); 2301 hlen = ip->ip_hl << 2; 2302 args->f_id.addr_type = 4; 2303 2304 /* 2305 * Collect parameters into local variables for faster matching. 2306 */ 2307 proto = ip->ip_p; 2308 src_ip = ip->ip_src; 2309 dst_ip = ip->ip_dst; 2310 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2311 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2312 ip_len = ntohs(ip->ip_len); 2313 } else { 2314 offset = ip->ip_off & IP_OFFMASK; 2315 ip_len = ip->ip_len; 2316 } 2317 pktlen = ip_len < pktlen ? ip_len : pktlen; 2318 2319 if (offset == 0) { 2320 switch (proto) { 2321 case IPPROTO_TCP: 2322 PULLUP_TO(hlen, ulp, struct tcphdr); 2323 dst_port = TCP(ulp)->th_dport; 2324 src_port = TCP(ulp)->th_sport; 2325 args->f_id.flags = TCP(ulp)->th_flags; 2326 break; 2327 2328 case IPPROTO_UDP: 2329 PULLUP_TO(hlen, ulp, struct udphdr); 2330 dst_port = UDP(ulp)->uh_dport; 2331 src_port = UDP(ulp)->uh_sport; 2332 break; 2333 2334 case IPPROTO_ICMP: 2335 PULLUP_TO(hlen, ulp, struct icmphdr); 2336 args->f_id.flags = ICMP(ulp)->icmp_type; 2337 break; 2338 2339 default: 2340 break; 2341 } 2342 } 2343 2344 args->f_id.src_ip = ntohl(src_ip.s_addr); 2345 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2346 } 2347#undef PULLUP_TO 2348 if (proto) { /* we may have port numbers, store them */ 2349 args->f_id.proto = proto; 2350 args->f_id.src_port = src_port = ntohs(src_port); 2351 args->f_id.dst_port = dst_port = ntohs(dst_port); 2352 } 2353 2354 IPFW_RLOCK(chain); 2355 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2356 if (args->rule) { 2357 /* 2358 * Packet has already been tagged. Look for the next rule 2359 * to restart processing. 2360 * 2361 * If fw_one_pass != 0 then just accept it. 2362 * XXX should not happen here, but optimized out in 2363 * the caller. 2364 */ 2365 if (fw_one_pass) { 2366 IPFW_RUNLOCK(chain); 2367 return (IP_FW_PASS); 2368 } 2369 2370 f = args->rule->next_rule; 2371 if (f == NULL) 2372 f = lookup_next_rule(args->rule); 2373 } else { 2374 /* 2375 * Find the starting rule. It can be either the first 2376 * one, or the one after divert_rule if asked so. 2377 */ 2378 int skipto = mtag ? divert_cookie(mtag) : 0; 2379 2380 f = chain->rules; 2381 if (args->eh == NULL && skipto != 0) { 2382 if (skipto >= IPFW_DEFAULT_RULE) { 2383 IPFW_RUNLOCK(chain); 2384 return (IP_FW_DENY); /* invalid */ 2385 } 2386 while (f && f->rulenum <= skipto) 2387 f = f->next; 2388 if (f == NULL) { /* drop packet */ 2389 IPFW_RUNLOCK(chain); 2390 return (IP_FW_DENY); 2391 } 2392 } 2393 } 2394 /* reset divert rule to avoid confusion later */ 2395 if (mtag) { 2396 divinput_flags = divert_info(mtag) & 2397 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2398 m_tag_delete(m, mtag); 2399 } 2400 2401 /* 2402 * Now scan the rules, and parse microinstructions for each rule. 2403 */ 2404 for (; f; f = f->next) { 2405 int l, cmdlen; 2406 ipfw_insn *cmd; 2407 int skip_or; /* skip rest of OR block */ 2408 2409again: 2410 if (set_disable & (1 << f->set) ) 2411 continue; 2412 2413 skip_or = 0; 2414 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2415 l -= cmdlen, cmd += cmdlen) { 2416 int match; 2417 2418 /* 2419 * check_body is a jump target used when we find a 2420 * CHECK_STATE, and need to jump to the body of 2421 * the target rule. 2422 */ 2423 2424check_body: 2425 cmdlen = F_LEN(cmd); 2426 /* 2427 * An OR block (insn_1 || .. || insn_n) has the 2428 * F_OR bit set in all but the last instruction. 2429 * The first match will set "skip_or", and cause 2430 * the following instructions to be skipped until 2431 * past the one with the F_OR bit clear. 2432 */ 2433 if (skip_or) { /* skip this instruction */ 2434 if ((cmd->len & F_OR) == 0) 2435 skip_or = 0; /* next one is good */ 2436 continue; 2437 } 2438 match = 0; /* set to 1 if we succeed */ 2439 2440 switch (cmd->opcode) { 2441 /* 2442 * The first set of opcodes compares the packet's 2443 * fields with some pattern, setting 'match' if a 2444 * match is found. At the end of the loop there is 2445 * logic to deal with F_NOT and F_OR flags associated 2446 * with the opcode. 2447 */ 2448 case O_NOP: 2449 match = 1; 2450 break; 2451 2452 case O_FORWARD_MAC: 2453 printf("ipfw: opcode %d unimplemented\n", 2454 cmd->opcode); 2455 break; 2456 2457 case O_GID: 2458 case O_UID: 2459 case O_JAIL: 2460 /* 2461 * We only check offset == 0 && proto != 0, 2462 * as this ensures that we have a 2463 * packet with the ports info. 2464 */ 2465 if (offset!=0) 2466 break; 2467 if (is_ipv6) /* XXX to be fixed later */ 2468 break; 2469 if (proto == IPPROTO_TCP || 2470 proto == IPPROTO_UDP) 2471 match = check_uidgid( 2472 (ipfw_insn_u32 *)cmd, 2473 proto, oif, 2474 dst_ip, dst_port, 2475 src_ip, src_port, &fw_ugid_cache, 2476 &ugid_lookup, args->inp); 2477 break; 2478 2479 case O_RECV: 2480 match = iface_match(m->m_pkthdr.rcvif, 2481 (ipfw_insn_if *)cmd); 2482 break; 2483 2484 case O_XMIT: 2485 match = iface_match(oif, (ipfw_insn_if *)cmd); 2486 break; 2487 2488 case O_VIA: 2489 match = iface_match(oif ? oif : 2490 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2491 break; 2492 2493 case O_MACADDR2: 2494 if (args->eh != NULL) { /* have MAC header */ 2495 u_int32_t *want = (u_int32_t *) 2496 ((ipfw_insn_mac *)cmd)->addr; 2497 u_int32_t *mask = (u_int32_t *) 2498 ((ipfw_insn_mac *)cmd)->mask; 2499 u_int32_t *hdr = (u_int32_t *)args->eh; 2500 2501 match = 2502 ( want[0] == (hdr[0] & mask[0]) && 2503 want[1] == (hdr[1] & mask[1]) && 2504 want[2] == (hdr[2] & mask[2]) ); 2505 } 2506 break; 2507 2508 case O_MAC_TYPE: 2509 if (args->eh != NULL) { 2510 u_int16_t t = 2511 ntohs(args->eh->ether_type); 2512 u_int16_t *p = 2513 ((ipfw_insn_u16 *)cmd)->ports; 2514 int i; 2515 2516 for (i = cmdlen - 1; !match && i>0; 2517 i--, p += 2) 2518 match = (t>=p[0] && t<=p[1]); 2519 } 2520 break; 2521 2522 case O_FRAG: 2523 match = (offset != 0); 2524 break; 2525 2526 case O_IN: /* "out" is "not in" */ 2527 match = (oif == NULL); 2528 break; 2529 2530 case O_LAYER2: 2531 match = (args->eh != NULL); 2532 break; 2533 2534 case O_DIVERTED: 2535 match = (cmd->arg1 & 1 && divinput_flags & 2536 IP_FW_DIVERT_LOOPBACK_FLAG) || 2537 (cmd->arg1 & 2 && divinput_flags & 2538 IP_FW_DIVERT_OUTPUT_FLAG); 2539 break; 2540 2541 case O_PROTO: 2542 /* 2543 * We do not allow an arg of 0 so the 2544 * check of "proto" only suffices. 2545 */ 2546 match = (proto == cmd->arg1); 2547 break; 2548 2549 case O_IP_SRC: 2550 match = is_ipv4 && 2551 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2552 src_ip.s_addr); 2553 break; 2554 2555 case O_IP_SRC_LOOKUP: 2556 case O_IP_DST_LOOKUP: 2557 if (is_ipv4) { 2558 uint32_t a = 2559 (cmd->opcode == O_IP_DST_LOOKUP) ? 2560 dst_ip.s_addr : src_ip.s_addr; 2561 uint32_t v; 2562 2563 match = lookup_table(cmd->arg1, a, &v); 2564 if (!match) 2565 break; 2566 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2567 match = 2568 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2569 } 2570 break; 2571 2572 case O_IP_SRC_MASK: 2573 case O_IP_DST_MASK: 2574 if (is_ipv4) { 2575 uint32_t a = 2576 (cmd->opcode == O_IP_DST_MASK) ? 2577 dst_ip.s_addr : src_ip.s_addr; 2578 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2579 int i = cmdlen-1; 2580 2581 for (; !match && i>0; i-= 2, p+= 2) 2582 match = (p[0] == (a & p[1])); 2583 } 2584 break; 2585 2586 case O_IP_SRC_ME: 2587 if (is_ipv4) { 2588 struct ifnet *tif; 2589 2590 INADDR_TO_IFP(src_ip, tif); 2591 match = (tif != NULL); 2592 } 2593 break; 2594 2595 case O_IP_DST_SET: 2596 case O_IP_SRC_SET: 2597 if (is_ipv4) { 2598 u_int32_t *d = (u_int32_t *)(cmd+1); 2599 u_int32_t addr = 2600 cmd->opcode == O_IP_DST_SET ? 2601 args->f_id.dst_ip : 2602 args->f_id.src_ip; 2603 2604 if (addr < d[0]) 2605 break; 2606 addr -= d[0]; /* subtract base */ 2607 match = (addr < cmd->arg1) && 2608 ( d[ 1 + (addr>>5)] & 2609 (1<<(addr & 0x1f)) ); 2610 } 2611 break; 2612 2613 case O_IP_DST: 2614 match = is_ipv4 && 2615 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2616 dst_ip.s_addr); 2617 break; 2618 2619 case O_IP_DST_ME: 2620 if (is_ipv4) { 2621 struct ifnet *tif; 2622 2623 INADDR_TO_IFP(dst_ip, tif); 2624 match = (tif != NULL); 2625 } 2626 break; 2627 2628 case O_IP_SRCPORT: 2629 case O_IP_DSTPORT: 2630 /* 2631 * offset == 0 && proto != 0 is enough 2632 * to guarantee that we have a 2633 * packet with port info. 2634 */ 2635 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2636 && offset == 0) { 2637 u_int16_t x = 2638 (cmd->opcode == O_IP_SRCPORT) ? 2639 src_port : dst_port ; 2640 u_int16_t *p = 2641 ((ipfw_insn_u16 *)cmd)->ports; 2642 int i; 2643 2644 for (i = cmdlen - 1; !match && i>0; 2645 i--, p += 2) 2646 match = (x>=p[0] && x<=p[1]); 2647 } 2648 break; 2649 2650 case O_ICMPTYPE: 2651 match = (offset == 0 && proto==IPPROTO_ICMP && 2652 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2653 break; 2654 2655#ifdef INET6 2656 case O_ICMP6TYPE: 2657 match = is_ipv6 && offset == 0 && 2658 proto==IPPROTO_ICMPV6 && 2659 icmp6type_match( 2660 ICMP6(ulp)->icmp6_type, 2661 (ipfw_insn_u32 *)cmd); 2662 break; 2663#endif /* INET6 */ 2664 2665 case O_IPOPT: 2666 match = (is_ipv4 && 2667 ipopts_match(mtod(m, struct ip *), cmd) ); 2668 break; 2669 2670 case O_IPVER: 2671 match = (is_ipv4 && 2672 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2673 break; 2674 2675 case O_IPID: 2676 case O_IPLEN: 2677 case O_IPTTL: 2678 if (is_ipv4) { /* only for IP packets */ 2679 uint16_t x; 2680 uint16_t *p; 2681 int i; 2682 2683 if (cmd->opcode == O_IPLEN) 2684 x = ip_len; 2685 else if (cmd->opcode == O_IPTTL) 2686 x = mtod(m, struct ip *)->ip_ttl; 2687 else /* must be IPID */ 2688 x = ntohs(mtod(m, struct ip *)->ip_id); 2689 if (cmdlen == 1) { 2690 match = (cmd->arg1 == x); 2691 break; 2692 } 2693 /* otherwise we have ranges */ 2694 p = ((ipfw_insn_u16 *)cmd)->ports; 2695 i = cmdlen - 1; 2696 for (; !match && i>0; i--, p += 2) 2697 match = (x >= p[0] && x <= p[1]); 2698 } 2699 break; 2700 2701 case O_IPPRECEDENCE: 2702 match = (is_ipv4 && 2703 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2704 break; 2705 2706 case O_IPTOS: 2707 match = (is_ipv4 && 2708 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2709 break; 2710 2711 case O_TCPDATALEN: 2712 if (proto == IPPROTO_TCP && offset == 0) { 2713 struct tcphdr *tcp; 2714 uint16_t x; 2715 uint16_t *p; 2716 int i; 2717 2718 tcp = TCP(ulp); 2719 x = ip_len - 2720 ((ip->ip_hl + tcp->th_off) << 2); 2721 if (cmdlen == 1) { 2722 match = (cmd->arg1 == x); 2723 break; 2724 } 2725 /* otherwise we have ranges */ 2726 p = ((ipfw_insn_u16 *)cmd)->ports; 2727 i = cmdlen - 1; 2728 for (; !match && i>0; i--, p += 2) 2729 match = (x >= p[0] && x <= p[1]); 2730 } 2731 break; 2732 2733 case O_TCPFLAGS: 2734 match = (proto == IPPROTO_TCP && offset == 0 && 2735 flags_match(cmd, TCP(ulp)->th_flags)); 2736 break; 2737 2738 case O_TCPOPTS: 2739 match = (proto == IPPROTO_TCP && offset == 0 && 2740 tcpopts_match(TCP(ulp), cmd)); 2741 break; 2742 2743 case O_TCPSEQ: 2744 match = (proto == IPPROTO_TCP && offset == 0 && 2745 ((ipfw_insn_u32 *)cmd)->d[0] == 2746 TCP(ulp)->th_seq); 2747 break; 2748 2749 case O_TCPACK: 2750 match = (proto == IPPROTO_TCP && offset == 0 && 2751 ((ipfw_insn_u32 *)cmd)->d[0] == 2752 TCP(ulp)->th_ack); 2753 break; 2754 2755 case O_TCPWIN: 2756 match = (proto == IPPROTO_TCP && offset == 0 && 2757 cmd->arg1 == TCP(ulp)->th_win); 2758 break; 2759 2760 case O_ESTAB: 2761 /* reject packets which have SYN only */ 2762 /* XXX should i also check for TH_ACK ? */ 2763 match = (proto == IPPROTO_TCP && offset == 0 && 2764 (TCP(ulp)->th_flags & 2765 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2766 break; 2767 2768 case O_ALTQ: { 2769 struct altq_tag *at; 2770 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2771 2772 match = 1; 2773 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2774 if (mtag != NULL) 2775 break; 2776 mtag = m_tag_get(PACKET_TAG_PF_QID, 2777 sizeof(struct altq_tag), 2778 M_NOWAIT); 2779 if (mtag == NULL) { 2780 /* 2781 * Let the packet fall back to the 2782 * default ALTQ. 2783 */ 2784 break; 2785 } 2786 at = (struct altq_tag *)(mtag+1); 2787 at->qid = altq->qid; 2788 if (is_ipv4) 2789 at->af = AF_INET; 2790 else 2791 at->af = AF_LINK; 2792 at->hdr = ip; 2793 m_tag_prepend(m, mtag); 2794 break; 2795 } 2796 2797 case O_LOG: 2798 if (fw_verbose) 2799 ipfw_log(f, hlen, args, m, oif, offset); 2800 match = 1; 2801 break; 2802 2803 case O_PROB: 2804 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2805 break; 2806 2807 case O_VERREVPATH: 2808 /* Outgoing packets automatically pass/match */ 2809 match = ((oif != NULL) || 2810 (m->m_pkthdr.rcvif == NULL) || 2811 ( 2812#ifdef INET6 2813 is_ipv6 ? 2814 verify_path6(&(args->f_id.src_ip6), 2815 m->m_pkthdr.rcvif) : 2816#endif 2817 verify_path(src_ip, m->m_pkthdr.rcvif))); 2818 break; 2819 2820 case O_VERSRCREACH: 2821 /* Outgoing packets automatically pass/match */ 2822 match = (hlen > 0 && ((oif != NULL) || 2823#ifdef INET6 2824 is_ipv6 ? 2825 verify_path6(&(args->f_id.src_ip6), 2826 NULL) : 2827#endif 2828 verify_path(src_ip, NULL))); 2829 break; 2830 2831 case O_ANTISPOOF: 2832 /* Outgoing packets automatically pass/match */ 2833 if (oif == NULL && hlen > 0 && 2834 ( (is_ipv4 && in_localaddr(src_ip)) 2835#ifdef INET6 2836 || (is_ipv6 && 2837 in6_localaddr(&(args->f_id.src_ip6))) 2838#endif 2839 )) 2840 match = 2841#ifdef INET6 2842 is_ipv6 ? verify_path6( 2843 &(args->f_id.src_ip6), 2844 m->m_pkthdr.rcvif) : 2845#endif 2846 verify_path(src_ip, 2847 m->m_pkthdr.rcvif); 2848 else 2849 match = 1; 2850 break; 2851 2852 case O_IPSEC: 2853#ifdef FAST_IPSEC 2854 match = (m_tag_find(m, 2855 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2856#endif 2857#ifdef IPSEC 2858 match = (ipsec_getnhist(m) != 0); 2859#endif 2860 /* otherwise no match */ 2861 break; 2862 2863 case O_IP6_SRC: 2864 match = is_ipv6 && 2865 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2866 &((ipfw_insn_ip6 *)cmd)->addr6); 2867 break; 2868 2869 case O_IP6_DST: 2870 match = is_ipv6 && 2871 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2872 &((ipfw_insn_ip6 *)cmd)->addr6); 2873 break; 2874 case O_IP6_SRC_MASK: 2875 if (is_ipv6) { 2876 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2877 struct in6_addr p = args->f_id.src_ip6; 2878 2879 APPLY_MASK(&p, &te->mask6); 2880 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2881 } 2882 break; 2883 2884 case O_IP6_DST_MASK: 2885 if (is_ipv6) { 2886 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2887 struct in6_addr p = args->f_id.dst_ip6; 2888 2889 APPLY_MASK(&p, &te->mask6); 2890 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2891 } 2892 break; 2893 2894#ifdef INET6 2895 case O_IP6_SRC_ME: 2896 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2897 break; 2898 2899 case O_IP6_DST_ME: 2900 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2901 break; 2902 2903 case O_FLOW6ID: 2904 match = is_ipv6 && 2905 flow6id_match(args->f_id.flow_id6, 2906 (ipfw_insn_u32 *) cmd); 2907 break; 2908 2909 case O_EXT_HDR: 2910 match = is_ipv6 && 2911 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2912 break; 2913 2914 case O_IP6: 2915 match = is_ipv6; 2916 break; 2917#endif 2918 2919 case O_IP4: 2920 match = is_ipv4; 2921 break; 2922 2923 /* 2924 * The second set of opcodes represents 'actions', 2925 * i.e. the terminal part of a rule once the packet 2926 * matches all previous patterns. 2927 * Typically there is only one action for each rule, 2928 * and the opcode is stored at the end of the rule 2929 * (but there are exceptions -- see below). 2930 * 2931 * In general, here we set retval and terminate the 2932 * outer loop (would be a 'break 3' in some language, 2933 * but we need to do a 'goto done'). 2934 * 2935 * Exceptions: 2936 * O_COUNT and O_SKIPTO actions: 2937 * instead of terminating, we jump to the next rule 2938 * ('goto next_rule', equivalent to a 'break 2'), 2939 * or to the SKIPTO target ('goto again' after 2940 * having set f, cmd and l), respectively. 2941 * 2942 * O_LOG and O_ALTQ action parameters: 2943 * perform some action and set match = 1; 2944 * 2945 * O_LIMIT and O_KEEP_STATE: these opcodes are 2946 * not real 'actions', and are stored right 2947 * before the 'action' part of the rule. 2948 * These opcodes try to install an entry in the 2949 * state tables; if successful, we continue with 2950 * the next opcode (match=1; break;), otherwise 2951 * the packet * must be dropped 2952 * ('goto done' after setting retval); 2953 * 2954 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2955 * cause a lookup of the state table, and a jump 2956 * to the 'action' part of the parent rule 2957 * ('goto check_body') if an entry is found, or 2958 * (CHECK_STATE only) a jump to the next rule if 2959 * the entry is not found ('goto next_rule'). 2960 * The result of the lookup is cached to make 2961 * further instances of these opcodes are 2962 * effectively NOPs. 2963 */ 2964 case O_LIMIT: 2965 case O_KEEP_STATE: 2966 if (install_state(f, 2967 (ipfw_insn_limit *)cmd, args)) { 2968 retval = IP_FW_DENY; 2969 goto done; /* error/limit violation */ 2970 } 2971 match = 1; 2972 break; 2973 2974 case O_PROBE_STATE: 2975 case O_CHECK_STATE: 2976 /* 2977 * dynamic rules are checked at the first 2978 * keep-state or check-state occurrence, 2979 * with the result being stored in dyn_dir. 2980 * The compiler introduces a PROBE_STATE 2981 * instruction for us when we have a 2982 * KEEP_STATE (because PROBE_STATE needs 2983 * to be run first). 2984 */ 2985 if (dyn_dir == MATCH_UNKNOWN && 2986 (q = lookup_dyn_rule(&args->f_id, 2987 &dyn_dir, proto == IPPROTO_TCP ? 2988 TCP(ulp) : NULL)) 2989 != NULL) { 2990 /* 2991 * Found dynamic entry, update stats 2992 * and jump to the 'action' part of 2993 * the parent rule. 2994 */ 2995 q->pcnt++; 2996 q->bcnt += pktlen; 2997 f = q->rule; 2998 cmd = ACTION_PTR(f); 2999 l = f->cmd_len - f->act_ofs; 3000 IPFW_DYN_UNLOCK(); 3001 goto check_body; 3002 } 3003 /* 3004 * Dynamic entry not found. If CHECK_STATE, 3005 * skip to next rule, if PROBE_STATE just 3006 * ignore and continue with next opcode. 3007 */ 3008 if (cmd->opcode == O_CHECK_STATE) 3009 goto next_rule; 3010 match = 1; 3011 break; 3012 3013 case O_ACCEPT: 3014 retval = 0; /* accept */ 3015 goto done; 3016 3017 case O_PIPE: 3018 case O_QUEUE: 3019 args->rule = f; /* report matching rule */ 3020 args->cookie = cmd->arg1; 3021 retval = IP_FW_DUMMYNET; 3022 goto done; 3023 3024 case O_DIVERT: 3025 case O_TEE: { 3026 struct divert_tag *dt; 3027 3028 if (args->eh) /* not on layer 2 */ 3029 break; 3030 mtag = m_tag_get(PACKET_TAG_DIVERT, 3031 sizeof(struct divert_tag), 3032 M_NOWAIT); 3033 if (mtag == NULL) { 3034 /* XXX statistic */ 3035 /* drop packet */ 3036 IPFW_RUNLOCK(chain); 3037 return (IP_FW_DENY); 3038 } 3039 dt = (struct divert_tag *)(mtag+1); 3040 dt->cookie = f->rulenum; 3041 dt->info = cmd->arg1; 3042 m_tag_prepend(m, mtag); 3043 retval = (cmd->opcode == O_DIVERT) ? 3044 IP_FW_DIVERT : IP_FW_TEE; 3045 goto done; 3046 } 3047 3048 case O_COUNT: 3049 case O_SKIPTO: 3050 f->pcnt++; /* update stats */ 3051 f->bcnt += pktlen; 3052 f->timestamp = time_second; 3053 if (cmd->opcode == O_COUNT) 3054 goto next_rule; 3055 /* handle skipto */ 3056 if (f->next_rule == NULL) 3057 lookup_next_rule(f); 3058 f = f->next_rule; 3059 goto again; 3060 3061 case O_REJECT: 3062 /* 3063 * Drop the packet and send a reject notice 3064 * if the packet is not ICMP (or is an ICMP 3065 * query), and it is not multicast/broadcast. 3066 */ 3067 if (hlen > 0 && is_ipv4 && 3068 (proto != IPPROTO_ICMP || 3069 is_icmp_query(ICMP(ulp))) && 3070 !(m->m_flags & (M_BCAST|M_MCAST)) && 3071 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3072 send_reject(args, cmd->arg1, 3073 offset,ip_len); 3074 m = args->m; 3075 } 3076 /* FALLTHROUGH */ 3077#ifdef INET6 3078 case O_UNREACH6: 3079 if (hlen > 0 && is_ipv6 && 3080 (proto != IPPROTO_ICMPV6 || 3081 (is_icmp6_query(args->f_id.flags) == 1)) && 3082 !(m->m_flags & (M_BCAST|M_MCAST)) && 3083 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3084 send_reject6(args, cmd->arg1, 3085 offset, hlen); 3086 m = args->m; 3087 } 3088 /* FALLTHROUGH */ 3089#endif 3090 case O_DENY: 3091 retval = IP_FW_DENY; 3092 goto done; 3093 3094 case O_FORWARD_IP: 3095 if (args->eh) /* not valid on layer2 pkts */ 3096 break; 3097 if (!q || dyn_dir == MATCH_FORWARD) 3098 args->next_hop = 3099 &((ipfw_insn_sa *)cmd)->sa; 3100 retval = IP_FW_PASS; 3101 goto done; 3102 3103 case O_NETGRAPH: 3104 case O_NGTEE: 3105 args->rule = f; /* report matching rule */ 3106 args->cookie = cmd->arg1; 3107 retval = (cmd->opcode == O_NETGRAPH) ? 3108 IP_FW_NETGRAPH : IP_FW_NGTEE; 3109 goto done; 3110 3111 default: 3112 panic("-- unknown opcode %d\n", cmd->opcode); 3113 } /* end of switch() on opcodes */ 3114 3115 if (cmd->len & F_NOT) 3116 match = !match; 3117 3118 if (match) { 3119 if (cmd->len & F_OR) 3120 skip_or = 1; 3121 } else { 3122 if (!(cmd->len & F_OR)) /* not an OR block, */ 3123 break; /* try next rule */ 3124 } 3125 3126 } /* end of inner for, scan opcodes */ 3127 3128next_rule:; /* try next rule */ 3129 3130 } /* end of outer for, scan rules */ 3131 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3132 IPFW_RUNLOCK(chain); 3133 return (IP_FW_DENY); 3134 3135done: 3136 /* Update statistics */ 3137 f->pcnt++; 3138 f->bcnt += pktlen; 3139 f->timestamp = time_second; 3140 IPFW_RUNLOCK(chain); 3141 return (retval); 3142 3143pullup_failed: 3144 if (fw_verbose) 3145 printf("ipfw: pullup failed\n"); 3146 return (IP_FW_DENY); 3147} 3148 3149/* 3150 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3151 * These will be reconstructed on the fly as packets are matched. 3152 */ 3153static void 3154flush_rule_ptrs(struct ip_fw_chain *chain) 3155{ 3156 struct ip_fw *rule; 3157 3158 IPFW_WLOCK_ASSERT(chain); 3159 3160 for (rule = chain->rules; rule; rule = rule->next) 3161 rule->next_rule = NULL; 3162} 3163 3164/* 3165 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 3166 * pipe/queue, or to all of them (match == NULL). 3167 */ 3168void 3169flush_pipe_ptrs(struct dn_flow_set *match) 3170{ 3171 struct ip_fw *rule; 3172 3173 IPFW_WLOCK(&layer3_chain); 3174 for (rule = layer3_chain.rules; rule; rule = rule->next) { 3175 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 3176 3177 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 3178 continue; 3179 /* 3180 * XXX Use bcmp/bzero to handle pipe_ptr to overcome 3181 * possible alignment problems on 64-bit architectures. 3182 * This code is seldom used so we do not worry too 3183 * much about efficiency. 3184 */ 3185 if (match == NULL || 3186 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) ) 3187 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr)); 3188 } 3189 IPFW_WUNLOCK(&layer3_chain); 3190} 3191 3192/* 3193 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3194 * possibly create a rule number and add the rule to the list. 3195 * Update the rule_number in the input struct so the caller knows it as well. 3196 */ 3197static int 3198add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3199{ 3200 struct ip_fw *rule, *f, *prev; 3201 int l = RULESIZE(input_rule); 3202 3203 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3204 return (EINVAL); 3205 3206 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3207 if (rule == NULL) 3208 return (ENOSPC); 3209 3210 bcopy(input_rule, rule, l); 3211 3212 rule->next = NULL; 3213 rule->next_rule = NULL; 3214 3215 rule->pcnt = 0; 3216 rule->bcnt = 0; 3217 rule->timestamp = 0; 3218 3219 IPFW_WLOCK(chain); 3220 3221 if (chain->rules == NULL) { /* default rule */ 3222 chain->rules = rule; 3223 goto done; 3224 } 3225 3226 /* 3227 * If rulenum is 0, find highest numbered rule before the 3228 * default rule, and add autoinc_step 3229 */ 3230 if (autoinc_step < 1) 3231 autoinc_step = 1; 3232 else if (autoinc_step > 1000) 3233 autoinc_step = 1000; 3234 if (rule->rulenum == 0) { 3235 /* 3236 * locate the highest numbered rule before default 3237 */ 3238 for (f = chain->rules; f; f = f->next) { 3239 if (f->rulenum == IPFW_DEFAULT_RULE) 3240 break; 3241 rule->rulenum = f->rulenum; 3242 } 3243 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3244 rule->rulenum += autoinc_step; 3245 input_rule->rulenum = rule->rulenum; 3246 } 3247 3248 /* 3249 * Now insert the new rule in the right place in the sorted list. 3250 */ 3251 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3252 if (f->rulenum > rule->rulenum) { /* found the location */ 3253 if (prev) { 3254 rule->next = f; 3255 prev->next = rule; 3256 } else { /* head insert */ 3257 rule->next = chain->rules; 3258 chain->rules = rule; 3259 } 3260 break; 3261 } 3262 } 3263 flush_rule_ptrs(chain); 3264done: 3265 static_count++; 3266 static_len += l; 3267 IPFW_WUNLOCK(chain); 3268 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3269 rule->rulenum, static_count);) 3270 return (0); 3271} 3272 3273/** 3274 * Remove a static rule (including derived * dynamic rules) 3275 * and place it on the ``reap list'' for later reclamation. 3276 * The caller is in charge of clearing rule pointers to avoid 3277 * dangling pointers. 3278 * @return a pointer to the next entry. 3279 * Arguments are not checked, so they better be correct. 3280 */ 3281static struct ip_fw * 3282remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3283{ 3284 struct ip_fw *n; 3285 int l = RULESIZE(rule); 3286 3287 IPFW_WLOCK_ASSERT(chain); 3288 3289 n = rule->next; 3290 IPFW_DYN_LOCK(); 3291 remove_dyn_rule(rule, NULL /* force removal */); 3292 IPFW_DYN_UNLOCK(); 3293 if (prev == NULL) 3294 chain->rules = n; 3295 else 3296 prev->next = n; 3297 static_count--; 3298 static_len -= l; 3299 3300 rule->next = chain->reap; 3301 chain->reap = rule; 3302 3303 return n; 3304} 3305 3306/** 3307 * Reclaim storage associated with a list of rules. This is 3308 * typically the list created using remove_rule. 3309 */ 3310static void 3311reap_rules(struct ip_fw *head) 3312{ 3313 struct ip_fw *rule; 3314 3315 while ((rule = head) != NULL) { 3316 head = head->next; 3317 if (DUMMYNET_LOADED) 3318 ip_dn_ruledel_ptr(rule); 3319 free(rule, M_IPFW); 3320 } 3321} 3322 3323/* 3324 * Remove all rules from a chain (except rules in set RESVD_SET 3325 * unless kill_default = 1). The caller is responsible for 3326 * reclaiming storage for the rules left in chain->reap. 3327 */ 3328static void 3329free_chain(struct ip_fw_chain *chain, int kill_default) 3330{ 3331 struct ip_fw *prev, *rule; 3332 3333 IPFW_WLOCK_ASSERT(chain); 3334 3335 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3336 for (prev = NULL, rule = chain->rules; rule ; ) 3337 if (kill_default || rule->set != RESVD_SET) 3338 rule = remove_rule(chain, rule, prev); 3339 else { 3340 prev = rule; 3341 rule = rule->next; 3342 } 3343} 3344 3345/** 3346 * Remove all rules with given number, and also do set manipulation. 3347 * Assumes chain != NULL && *chain != NULL. 3348 * 3349 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3350 * the next 8 bits are the new set, the top 8 bits are the command: 3351 * 3352 * 0 delete rules with given number 3353 * 1 delete rules with given set number 3354 * 2 move rules with given number to new set 3355 * 3 move rules with given set number to new set 3356 * 4 swap sets with given numbers 3357 */ 3358static int 3359del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3360{ 3361 struct ip_fw *prev = NULL, *rule; 3362 u_int16_t rulenum; /* rule or old_set */ 3363 u_int8_t cmd, new_set; 3364 3365 rulenum = arg & 0xffff; 3366 cmd = (arg >> 24) & 0xff; 3367 new_set = (arg >> 16) & 0xff; 3368 3369 if (cmd > 4) 3370 return EINVAL; 3371 if (new_set > RESVD_SET) 3372 return EINVAL; 3373 if (cmd == 0 || cmd == 2) { 3374 if (rulenum >= IPFW_DEFAULT_RULE) 3375 return EINVAL; 3376 } else { 3377 if (rulenum > RESVD_SET) /* old_set */ 3378 return EINVAL; 3379 } 3380 3381 IPFW_WLOCK(chain); 3382 rule = chain->rules; 3383 chain->reap = NULL; 3384 switch (cmd) { 3385 case 0: /* delete rules with given number */ 3386 /* 3387 * locate first rule to delete 3388 */ 3389 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3390 ; 3391 if (rule->rulenum != rulenum) { 3392 IPFW_WUNLOCK(chain); 3393 return EINVAL; 3394 } 3395 3396 /* 3397 * flush pointers outside the loop, then delete all matching 3398 * rules. prev remains the same throughout the cycle. 3399 */ 3400 flush_rule_ptrs(chain); 3401 while (rule->rulenum == rulenum) 3402 rule = remove_rule(chain, rule, prev); 3403 break; 3404 3405 case 1: /* delete all rules with given set number */ 3406 flush_rule_ptrs(chain); 3407 rule = chain->rules; 3408 while (rule->rulenum < IPFW_DEFAULT_RULE) 3409 if (rule->set == rulenum) 3410 rule = remove_rule(chain, rule, prev); 3411 else { 3412 prev = rule; 3413 rule = rule->next; 3414 } 3415 break; 3416 3417 case 2: /* move rules with given number to new set */ 3418 rule = chain->rules; 3419 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3420 if (rule->rulenum == rulenum) 3421 rule->set = new_set; 3422 break; 3423 3424 case 3: /* move rules with given set number to new set */ 3425 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3426 if (rule->set == rulenum) 3427 rule->set = new_set; 3428 break; 3429 3430 case 4: /* swap two sets */ 3431 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3432 if (rule->set == rulenum) 3433 rule->set = new_set; 3434 else if (rule->set == new_set) 3435 rule->set = rulenum; 3436 break; 3437 } 3438 /* 3439 * Look for rules to reclaim. We grab the list before 3440 * releasing the lock then reclaim them w/o the lock to 3441 * avoid a LOR with dummynet. 3442 */ 3443 rule = chain->reap; 3444 chain->reap = NULL; 3445 IPFW_WUNLOCK(chain); 3446 if (rule) 3447 reap_rules(rule); 3448 return 0; 3449} 3450 3451/* 3452 * Clear counters for a specific rule. 3453 * The enclosing "table" is assumed locked. 3454 */ 3455static void 3456clear_counters(struct ip_fw *rule, int log_only) 3457{ 3458 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3459 3460 if (log_only == 0) { 3461 rule->bcnt = rule->pcnt = 0; 3462 rule->timestamp = 0; 3463 } 3464 if (l->o.opcode == O_LOG) 3465 l->log_left = l->max_log; 3466} 3467 3468/** 3469 * Reset some or all counters on firewall rules. 3470 * @arg frwl is null to clear all entries, or contains a specific 3471 * rule number. 3472 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3473 */ 3474static int 3475zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3476{ 3477 struct ip_fw *rule; 3478 char *msg; 3479 3480 IPFW_WLOCK(chain); 3481 if (rulenum == 0) { 3482 norule_counter = 0; 3483 for (rule = chain->rules; rule; rule = rule->next) 3484 clear_counters(rule, log_only); 3485 msg = log_only ? "ipfw: All logging counts reset.\n" : 3486 "ipfw: Accounting cleared.\n"; 3487 } else { 3488 int cleared = 0; 3489 /* 3490 * We can have multiple rules with the same number, so we 3491 * need to clear them all. 3492 */ 3493 for (rule = chain->rules; rule; rule = rule->next) 3494 if (rule->rulenum == rulenum) { 3495 while (rule && rule->rulenum == rulenum) { 3496 clear_counters(rule, log_only); 3497 rule = rule->next; 3498 } 3499 cleared = 1; 3500 break; 3501 } 3502 if (!cleared) { /* we did not find any matching rules */ 3503 IPFW_WUNLOCK(chain); 3504 return (EINVAL); 3505 } 3506 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3507 "ipfw: Entry %d cleared.\n"; 3508 } 3509 IPFW_WUNLOCK(chain); 3510 3511 if (fw_verbose) 3512 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3513 return (0); 3514} 3515 3516/* 3517 * Check validity of the structure before insert. 3518 * Fortunately rules are simple, so this mostly need to check rule sizes. 3519 */ 3520static int 3521check_ipfw_struct(struct ip_fw *rule, int size) 3522{ 3523 int l, cmdlen = 0; 3524 int have_action=0; 3525 ipfw_insn *cmd; 3526 3527 if (size < sizeof(*rule)) { 3528 printf("ipfw: rule too short\n"); 3529 return (EINVAL); 3530 } 3531 /* first, check for valid size */ 3532 l = RULESIZE(rule); 3533 if (l != size) { 3534 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3535 return (EINVAL); 3536 } 3537 if (rule->act_ofs >= rule->cmd_len) { 3538 printf("ipfw: bogus action offset (%u > %u)\n", 3539 rule->act_ofs, rule->cmd_len - 1); 3540 return (EINVAL); 3541 } 3542 /* 3543 * Now go for the individual checks. Very simple ones, basically only 3544 * instruction sizes. 3545 */ 3546 for (l = rule->cmd_len, cmd = rule->cmd ; 3547 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3548 cmdlen = F_LEN(cmd); 3549 if (cmdlen > l) { 3550 printf("ipfw: opcode %d size truncated\n", 3551 cmd->opcode); 3552 return EINVAL; 3553 } 3554 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3555 switch (cmd->opcode) { 3556 case O_PROBE_STATE: 3557 case O_KEEP_STATE: 3558 case O_PROTO: 3559 case O_IP_SRC_ME: 3560 case O_IP_DST_ME: 3561 case O_LAYER2: 3562 case O_IN: 3563 case O_FRAG: 3564 case O_DIVERTED: 3565 case O_IPOPT: 3566 case O_IPTOS: 3567 case O_IPPRECEDENCE: 3568 case O_IPVER: 3569 case O_TCPWIN: 3570 case O_TCPFLAGS: 3571 case O_TCPOPTS: 3572 case O_ESTAB: 3573 case O_VERREVPATH: 3574 case O_VERSRCREACH: 3575 case O_ANTISPOOF: 3576 case O_IPSEC: 3577 case O_IP6_SRC_ME: 3578 case O_IP6_DST_ME: 3579 case O_EXT_HDR: 3580 case O_IP6: 3581 case O_IP4: 3582 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3583 goto bad_size; 3584 break; 3585 3586 case O_UID: 3587 case O_GID: 3588 case O_JAIL: 3589 case O_IP_SRC: 3590 case O_IP_DST: 3591 case O_TCPSEQ: 3592 case O_TCPACK: 3593 case O_PROB: 3594 case O_ICMPTYPE: 3595 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3596 goto bad_size; 3597 break; 3598 3599 case O_LIMIT: 3600 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3601 goto bad_size; 3602 break; 3603 3604 case O_LOG: 3605 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3606 goto bad_size; 3607 3608 ((ipfw_insn_log *)cmd)->log_left = 3609 ((ipfw_insn_log *)cmd)->max_log; 3610 3611 break; 3612 3613 case O_IP_SRC_MASK: 3614 case O_IP_DST_MASK: 3615 /* only odd command lengths */ 3616 if ( !(cmdlen & 1) || cmdlen > 31) 3617 goto bad_size; 3618 break; 3619 3620 case O_IP_SRC_SET: 3621 case O_IP_DST_SET: 3622 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3623 printf("ipfw: invalid set size %d\n", 3624 cmd->arg1); 3625 return EINVAL; 3626 } 3627 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3628 (cmd->arg1+31)/32 ) 3629 goto bad_size; 3630 break; 3631 3632 case O_IP_SRC_LOOKUP: 3633 case O_IP_DST_LOOKUP: 3634 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3635 printf("ipfw: invalid table number %d\n", 3636 cmd->arg1); 3637 return (EINVAL); 3638 } 3639 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3640 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3641 goto bad_size; 3642 break; 3643 3644 case O_MACADDR2: 3645 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3646 goto bad_size; 3647 break; 3648 3649 case O_NOP: 3650 case O_IPID: 3651 case O_IPTTL: 3652 case O_IPLEN: 3653 case O_TCPDATALEN: 3654 if (cmdlen < 1 || cmdlen > 31) 3655 goto bad_size; 3656 break; 3657 3658 case O_MAC_TYPE: 3659 case O_IP_SRCPORT: 3660 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3661 if (cmdlen < 2 || cmdlen > 31) 3662 goto bad_size; 3663 break; 3664 3665 case O_RECV: 3666 case O_XMIT: 3667 case O_VIA: 3668 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3669 goto bad_size; 3670 break; 3671 3672 case O_ALTQ: 3673 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3674 goto bad_size; 3675 break; 3676 3677 case O_PIPE: 3678 case O_QUEUE: 3679 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 3680 goto bad_size; 3681 goto check_action; 3682 3683 case O_FORWARD_IP: 3684#ifdef IPFIREWALL_FORWARD 3685 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3686 goto bad_size; 3687 goto check_action; 3688#else 3689 return EINVAL; 3690#endif 3691 3692 case O_DIVERT: 3693 case O_TEE: 3694 if (ip_divert_ptr == NULL) 3695 return EINVAL; 3696 else 3697 goto check_size; 3698 case O_NETGRAPH: 3699 case O_NGTEE: 3700 if (!NG_IPFW_LOADED) 3701 return EINVAL; 3702 else 3703 goto check_size; 3704 case O_FORWARD_MAC: /* XXX not implemented yet */ 3705 case O_CHECK_STATE: 3706 case O_COUNT: 3707 case O_ACCEPT: 3708 case O_DENY: 3709 case O_REJECT: 3710 case O_UNREACH6: 3711 case O_SKIPTO: 3712check_size: 3713 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3714 goto bad_size; 3715check_action: 3716 if (have_action) { 3717 printf("ipfw: opcode %d, multiple actions" 3718 " not allowed\n", 3719 cmd->opcode); 3720 return EINVAL; 3721 } 3722 have_action = 1; 3723 if (l != cmdlen) { 3724 printf("ipfw: opcode %d, action must be" 3725 " last opcode\n", 3726 cmd->opcode); 3727 return EINVAL; 3728 } 3729 break; 3730 case O_IP6_SRC: 3731 case O_IP6_DST: 3732 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3733 F_INSN_SIZE(ipfw_insn)) 3734 goto bad_size; 3735 break; 3736 3737 case O_FLOW6ID: 3738 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3739 ((ipfw_insn_u32 *)cmd)->o.arg1) 3740 goto bad_size; 3741 break; 3742 3743 case O_IP6_SRC_MASK: 3744 case O_IP6_DST_MASK: 3745 if ( !(cmdlen & 1) || cmdlen > 127) 3746 goto bad_size; 3747 break; 3748 case O_ICMP6TYPE: 3749 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3750 goto bad_size; 3751 break; 3752 3753 default: 3754 printf("ipfw: opcode %d, unknown opcode\n", 3755 cmd->opcode); 3756 return EINVAL; 3757 } 3758 } 3759 if (have_action == 0) { 3760 printf("ipfw: missing action\n"); 3761 return EINVAL; 3762 } 3763 return 0; 3764 3765bad_size: 3766 printf("ipfw: opcode %d size %d wrong\n", 3767 cmd->opcode, cmdlen); 3768 return EINVAL; 3769} 3770 3771/* 3772 * Copy the static and dynamic rules to the supplied buffer 3773 * and return the amount of space actually used. 3774 */ 3775static size_t 3776ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3777{ 3778 char *bp = buf; 3779 char *ep = bp + space; 3780 struct ip_fw *rule; 3781 int i; 3782 3783 /* XXX this can take a long time and locking will block packet flow */ 3784 IPFW_RLOCK(chain); 3785 for (rule = chain->rules; rule ; rule = rule->next) { 3786 /* 3787 * Verify the entry fits in the buffer in case the 3788 * rules changed between calculating buffer space and 3789 * now. This would be better done using a generation 3790 * number but should suffice for now. 3791 */ 3792 i = RULESIZE(rule); 3793 if (bp + i <= ep) { 3794 bcopy(rule, bp, i); 3795 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3796 sizeof(set_disable)); 3797 bp += i; 3798 } 3799 } 3800 IPFW_RUNLOCK(chain); 3801 if (ipfw_dyn_v) { 3802 ipfw_dyn_rule *p, *last = NULL; 3803 3804 IPFW_DYN_LOCK(); 3805 for (i = 0 ; i < curr_dyn_buckets; i++) 3806 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3807 if (bp + sizeof *p <= ep) { 3808 ipfw_dyn_rule *dst = 3809 (ipfw_dyn_rule *)bp; 3810 bcopy(p, dst, sizeof *p); 3811 bcopy(&(p->rule->rulenum), &(dst->rule), 3812 sizeof(p->rule->rulenum)); 3813 /* 3814 * store a non-null value in "next". 3815 * The userland code will interpret a 3816 * NULL here as a marker 3817 * for the last dynamic rule. 3818 */ 3819 bcopy(&dst, &dst->next, sizeof(dst)); 3820 last = dst; 3821 dst->expire = 3822 TIME_LEQ(dst->expire, time_second) ? 3823 0 : dst->expire - time_second ; 3824 bp += sizeof(ipfw_dyn_rule); 3825 } 3826 } 3827 IPFW_DYN_UNLOCK(); 3828 if (last != NULL) /* mark last dynamic rule */ 3829 bzero(&last->next, sizeof(last)); 3830 } 3831 return (bp - (char *)buf); 3832} 3833 3834 3835/** 3836 * {set|get}sockopt parser. 3837 */ 3838static int 3839ipfw_ctl(struct sockopt *sopt) 3840{ 3841#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3842 int error, rule_num; 3843 size_t size; 3844 struct ip_fw *buf, *rule; 3845 u_int32_t rulenum[2]; 3846 3847 error = suser(sopt->sopt_td); 3848 if (error) 3849 return (error); 3850 3851 /* 3852 * Disallow modifications in really-really secure mode, but still allow 3853 * the logging counters to be reset. 3854 */ 3855 if (sopt->sopt_name == IP_FW_ADD || 3856 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3857#if __FreeBSD_version >= 500034 3858 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3859 if (error) 3860 return (error); 3861#else /* FreeBSD 4.x */ 3862 if (securelevel >= 3) 3863 return (EPERM); 3864#endif 3865 } 3866 3867 error = 0; 3868 3869 switch (sopt->sopt_name) { 3870 case IP_FW_GET: 3871 /* 3872 * pass up a copy of the current rules. Static rules 3873 * come first (the last of which has number IPFW_DEFAULT_RULE), 3874 * followed by a possibly empty list of dynamic rule. 3875 * The last dynamic rule has NULL in the "next" field. 3876 * 3877 * Note that the calculated size is used to bound the 3878 * amount of data returned to the user. The rule set may 3879 * change between calculating the size and returning the 3880 * data in which case we'll just return what fits. 3881 */ 3882 size = static_len; /* size of static rules */ 3883 if (ipfw_dyn_v) /* add size of dyn.rules */ 3884 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3885 3886 /* 3887 * XXX todo: if the user passes a short length just to know 3888 * how much room is needed, do not bother filling up the 3889 * buffer, just jump to the sooptcopyout. 3890 */ 3891 buf = malloc(size, M_TEMP, M_WAITOK); 3892 error = sooptcopyout(sopt, buf, 3893 ipfw_getrules(&layer3_chain, buf, size)); 3894 free(buf, M_TEMP); 3895 break; 3896 3897 case IP_FW_FLUSH: 3898 /* 3899 * Normally we cannot release the lock on each iteration. 3900 * We could do it here only because we start from the head all 3901 * the times so there is no risk of missing some entries. 3902 * On the other hand, the risk is that we end up with 3903 * a very inconsistent ruleset, so better keep the lock 3904 * around the whole cycle. 3905 * 3906 * XXX this code can be improved by resetting the head of 3907 * the list to point to the default rule, and then freeing 3908 * the old list without the need for a lock. 3909 */ 3910 3911 IPFW_WLOCK(&layer3_chain); 3912 layer3_chain.reap = NULL; 3913 free_chain(&layer3_chain, 0 /* keep default rule */); 3914 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3915 IPFW_WUNLOCK(&layer3_chain); 3916 if (layer3_chain.reap != NULL) 3917 reap_rules(rule); 3918 break; 3919 3920 case IP_FW_ADD: 3921 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3922 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3923 sizeof(struct ip_fw) ); 3924 if (error == 0) 3925 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3926 if (error == 0) { 3927 error = add_rule(&layer3_chain, rule); 3928 size = RULESIZE(rule); 3929 if (!error && sopt->sopt_dir == SOPT_GET) 3930 error = sooptcopyout(sopt, rule, size); 3931 } 3932 free(rule, M_TEMP); 3933 break; 3934 3935 case IP_FW_DEL: 3936 /* 3937 * IP_FW_DEL is used for deleting single rules or sets, 3938 * and (ab)used to atomically manipulate sets. Argument size 3939 * is used to distinguish between the two: 3940 * sizeof(u_int32_t) 3941 * delete single rule or set of rules, 3942 * or reassign rules (or sets) to a different set. 3943 * 2*sizeof(u_int32_t) 3944 * atomic disable/enable sets. 3945 * first u_int32_t contains sets to be disabled, 3946 * second u_int32_t contains sets to be enabled. 3947 */ 3948 error = sooptcopyin(sopt, rulenum, 3949 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3950 if (error) 3951 break; 3952 size = sopt->sopt_valsize; 3953 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3954 error = del_entry(&layer3_chain, rulenum[0]); 3955 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3956 set_disable = 3957 (set_disable | rulenum[0]) & ~rulenum[1] & 3958 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3959 else 3960 error = EINVAL; 3961 break; 3962 3963 case IP_FW_ZERO: 3964 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3965 rule_num = 0; 3966 if (sopt->sopt_val != 0) { 3967 error = sooptcopyin(sopt, &rule_num, 3968 sizeof(int), sizeof(int)); 3969 if (error) 3970 break; 3971 } 3972 error = zero_entry(&layer3_chain, rule_num, 3973 sopt->sopt_name == IP_FW_RESETLOG); 3974 break; 3975 3976 case IP_FW_TABLE_ADD: 3977 { 3978 ipfw_table_entry ent; 3979 3980 error = sooptcopyin(sopt, &ent, 3981 sizeof(ent), sizeof(ent)); 3982 if (error) 3983 break; 3984 error = add_table_entry(ent.tbl, ent.addr, 3985 ent.masklen, ent.value); 3986 } 3987 break; 3988 3989 case IP_FW_TABLE_DEL: 3990 { 3991 ipfw_table_entry ent; 3992 3993 error = sooptcopyin(sopt, &ent, 3994 sizeof(ent), sizeof(ent)); 3995 if (error) 3996 break; 3997 error = del_table_entry(ent.tbl, ent.addr, ent.masklen); 3998 } 3999 break; 4000 4001 case IP_FW_TABLE_FLUSH: 4002 { 4003 u_int16_t tbl; 4004 4005 error = sooptcopyin(sopt, &tbl, 4006 sizeof(tbl), sizeof(tbl)); 4007 if (error) 4008 break; 4009 error = flush_table(tbl); 4010 } 4011 break; 4012 4013 case IP_FW_TABLE_GETSIZE: 4014 { 4015 u_int32_t tbl, cnt; 4016 4017 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4018 sizeof(tbl)))) 4019 break; 4020 if ((error = count_table(tbl, &cnt))) 4021 break; 4022 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4023 } 4024 break; 4025 4026 case IP_FW_TABLE_LIST: 4027 { 4028 ipfw_table *tbl; 4029 4030 if (sopt->sopt_valsize < sizeof(*tbl)) { 4031 error = EINVAL; 4032 break; 4033 } 4034 size = sopt->sopt_valsize; 4035 tbl = malloc(size, M_TEMP, M_WAITOK); 4036 if (tbl == NULL) { 4037 error = ENOMEM; 4038 break; 4039 } 4040 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4041 if (error) { 4042 free(tbl, M_TEMP); 4043 break; 4044 } 4045 tbl->size = (size - sizeof(*tbl)) / 4046 sizeof(ipfw_table_entry); 4047 error = dump_table(tbl); 4048 if (error) { 4049 free(tbl, M_TEMP); 4050 break; 4051 } 4052 error = sooptcopyout(sopt, tbl, size); 4053 free(tbl, M_TEMP); 4054 } 4055 break; 4056 4057 default: 4058 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4059 error = EINVAL; 4060 } 4061 4062 return (error); 4063#undef RULE_MAXSIZE 4064} 4065 4066/** 4067 * dummynet needs a reference to the default rule, because rules can be 4068 * deleted while packets hold a reference to them. When this happens, 4069 * dummynet changes the reference to the default rule (it could well be a 4070 * NULL pointer, but this way we do not need to check for the special 4071 * case, plus here he have info on the default behaviour). 4072 */ 4073struct ip_fw *ip_fw_default_rule; 4074 4075/* 4076 * This procedure is only used to handle keepalives. It is invoked 4077 * every dyn_keepalive_period 4078 */ 4079static void 4080ipfw_tick(void * __unused unused) 4081{ 4082 struct mbuf *m0, *m, *mnext, **mtailp; 4083 int i; 4084 ipfw_dyn_rule *q; 4085 4086 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4087 goto done; 4088 4089 /* 4090 * We make a chain of packets to go out here -- not deferring 4091 * until after we drop the IPFW dynamic rule lock would result 4092 * in a lock order reversal with the normal packet input -> ipfw 4093 * call stack. 4094 */ 4095 m0 = NULL; 4096 mtailp = &m0; 4097 IPFW_DYN_LOCK(); 4098 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4099 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4100 if (q->dyn_type == O_LIMIT_PARENT) 4101 continue; 4102 if (q->id.proto != IPPROTO_TCP) 4103 continue; 4104 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4105 continue; 4106 if (TIME_LEQ( time_second+dyn_keepalive_interval, 4107 q->expire)) 4108 continue; /* too early */ 4109 if (TIME_LEQ(q->expire, time_second)) 4110 continue; /* too late, rule expired */ 4111 4112 *mtailp = send_pkt(&(q->id), q->ack_rev - 1, 4113 q->ack_fwd, TH_SYN); 4114 if (*mtailp != NULL) 4115 mtailp = &(*mtailp)->m_nextpkt; 4116 *mtailp = send_pkt(&(q->id), q->ack_fwd - 1, 4117 q->ack_rev, 0); 4118 if (*mtailp != NULL) 4119 mtailp = &(*mtailp)->m_nextpkt; 4120 } 4121 } 4122 IPFW_DYN_UNLOCK(); 4123 for (m = mnext = m0; m != NULL; m = mnext) { 4124 mnext = m->m_nextpkt; 4125 m->m_nextpkt = NULL; 4126 ip_output(m, NULL, NULL, 0, NULL, NULL); 4127 } 4128done: 4129 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4130} 4131 4132int 4133ipfw_init(void) 4134{ 4135 struct ip_fw default_rule; 4136 int error; 4137 4138#ifdef INET6 4139 /* Setup IPv6 fw sysctl tree. */ 4140 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4141 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4142 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4143 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4144 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4145 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4146 &fw_deny_unknown_exthdrs, 0, 4147 "Deny packets with unknown IPv6 Extension Headers"); 4148#endif 4149 4150 layer3_chain.rules = NULL; 4151 layer3_chain.want_write = 0; 4152 layer3_chain.busy_count = 0; 4153 cv_init(&layer3_chain.cv, "Condition variable for IPFW rw locks"); 4154 IPFW_LOCK_INIT(&layer3_chain); 4155 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4156 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4157 UMA_ALIGN_PTR, 0); 4158 IPFW_DYN_LOCK_INIT(); 4159 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4160 4161 bzero(&default_rule, sizeof default_rule); 4162 4163 default_rule.act_ofs = 0; 4164 default_rule.rulenum = IPFW_DEFAULT_RULE; 4165 default_rule.cmd_len = 1; 4166 default_rule.set = RESVD_SET; 4167 4168 default_rule.cmd[0].len = 1; 4169 default_rule.cmd[0].opcode = 4170#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4171 1 ? O_ACCEPT : 4172#endif 4173 O_DENY; 4174 4175 error = add_rule(&layer3_chain, &default_rule); 4176 if (error != 0) { 4177 printf("ipfw2: error %u initializing default rule " 4178 "(support disabled)\n", error); 4179 IPFW_DYN_LOCK_DESTROY(); 4180 IPFW_LOCK_DESTROY(&layer3_chain); 4181 return (error); 4182 } 4183 4184 ip_fw_default_rule = layer3_chain.rules; 4185 printf("ipfw2 (+ipv6) initialized, divert %s, " 4186 "rule-based forwarding " 4187#ifdef IPFIREWALL_FORWARD 4188 "enabled, " 4189#else 4190 "disabled, " 4191#endif 4192 "default to %s, logging ", 4193#ifdef IPDIVERT 4194 "enabled", 4195#else 4196 "loadable", 4197#endif 4198 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4199 4200#ifdef IPFIREWALL_VERBOSE 4201 fw_verbose = 1; 4202#endif 4203#ifdef IPFIREWALL_VERBOSE_LIMIT 4204 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4205#endif 4206 if (fw_verbose == 0) 4207 printf("disabled\n"); 4208 else if (verbose_limit == 0) 4209 printf("unlimited\n"); 4210 else 4211 printf("limited to %d packets/entry by default\n", 4212 verbose_limit); 4213 4214 init_tables(); 4215 ip_fw_ctl_ptr = ipfw_ctl; 4216 ip_fw_chk_ptr = ipfw_chk; 4217 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4218 4219 return (0); 4220} 4221 4222void 4223ipfw_destroy(void) 4224{ 4225 struct ip_fw *reap; 4226 4227 ip_fw_chk_ptr = NULL; 4228 ip_fw_ctl_ptr = NULL; 4229 callout_drain(&ipfw_timeout); 4230 IPFW_WLOCK(&layer3_chain); 4231 layer3_chain.reap = NULL; 4232 free_chain(&layer3_chain, 1 /* kill default rule */); 4233 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4234 IPFW_WUNLOCK(&layer3_chain); 4235 if (reap != NULL) 4236 reap_rules(reap); 4237 flush_tables(); 4238 IPFW_DYN_LOCK_DESTROY(); 4239 uma_zdestroy(ipfw_dyn_rule_zone); 4240 IPFW_LOCK_DESTROY(&layer3_chain); 4241 4242#ifdef INET6 4243 /* Free IPv6 fw sysctl tree. */ 4244 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4245#endif 4246 4247 printf("IP firewall unloaded\n"); 4248} 4249