ip_fw2.c revision 183015
1/*- 2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions 6 * are met: 7 * 1. Redistributions of source code must retain the above copyright 8 * notice, this list of conditions and the following disclaimer. 9 * 2. Redistributions in binary form must reproduce the above copyright 10 * notice, this list of conditions and the following disclaimer in the 11 * documentation and/or other materials provided with the distribution. 12 * 13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23 * SUCH DAMAGE. 24 */ 25 26#include <sys/cdefs.h> 27__FBSDID("$FreeBSD: head/sys/netinet/ip_fw2.c 183015 2008-09-14 09:24:12Z rik $"); 28 29#define DEB(x) 30#define DDB(x) x 31 32/* 33 * Implement IP packet firewall (new version) 34 */ 35 36#if !defined(KLD_MODULE) 37#include "opt_ipfw.h" 38#include "opt_ipdivert.h" 39#include "opt_ipdn.h" 40#include "opt_inet.h" 41#ifndef INET 42#error IPFIREWALL requires INET. 43#endif /* INET */ 44#endif 45#include "opt_inet6.h" 46#include "opt_ipsec.h" 47#include "opt_mac.h" 48 49#include <sys/param.h> 50#include <sys/systm.h> 51#include <sys/condvar.h> 52#include <sys/eventhandler.h> 53#include <sys/malloc.h> 54#include <sys/mbuf.h> 55#include <sys/kernel.h> 56#include <sys/lock.h> 57#include <sys/jail.h> 58#include <sys/module.h> 59#include <sys/priv.h> 60#include <sys/proc.h> 61#include <sys/rwlock.h> 62#include <sys/socket.h> 63#include <sys/socketvar.h> 64#include <sys/sysctl.h> 65#include <sys/syslog.h> 66#include <sys/ucred.h> 67#include <sys/vimage.h> 68#include <net/if.h> 69#include <net/radix.h> 70#include <net/route.h> 71#include <net/pf_mtag.h> 72 73#define IPFW_INTERNAL /* Access to protected data structures in ip_fw.h. */ 74 75#include <netinet/in.h> 76#include <netinet/in_systm.h> 77#include <netinet/in_var.h> 78#include <netinet/in_pcb.h> 79#include <netinet/ip.h> 80#include <netinet/ip_var.h> 81#include <netinet/ip_icmp.h> 82#include <netinet/ip_fw.h> 83#include <netinet/ip_divert.h> 84#include <netinet/ip_dummynet.h> 85#include <netinet/ip_carp.h> 86#include <netinet/pim.h> 87#include <netinet/tcp.h> 88#include <netinet/tcp_timer.h> 89#include <netinet/tcp_var.h> 90#include <netinet/tcpip.h> 91#include <netinet/udp.h> 92#include <netinet/udp_var.h> 93#include <netinet/sctp.h> 94#include <netgraph/ng_ipfw.h> 95 96#include <altq/if_altq.h> 97 98#include <netinet/ip6.h> 99#include <netinet/icmp6.h> 100#ifdef INET6 101#include <netinet6/scope6_var.h> 102#endif 103 104#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 105 106#include <machine/in_cksum.h> /* XXX for in_cksum */ 107 108#include <security/mac/mac_framework.h> 109 110/* 111 * set_disable contains one bit per set value (0..31). 112 * If the bit is set, all rules with the corresponding set 113 * are disabled. Set RESVD_SET(31) is reserved for the default rule 114 * and rules that are not deleted by the flush command, 115 * and CANNOT be disabled. 116 * Rules in set RESVD_SET can only be deleted explicitly. 117 */ 118static u_int32_t set_disable; 119 120static int fw_verbose; 121static int verbose_limit; 122 123static struct callout ipfw_timeout; 124static uma_zone_t ipfw_dyn_rule_zone; 125 126/* 127 * Data structure to cache our ucred related 128 * information. This structure only gets used if 129 * the user specified UID/GID based constraints in 130 * a firewall rule. 131 */ 132struct ip_fw_ugid { 133 gid_t fw_groups[NGROUPS]; 134 int fw_ngroups; 135 uid_t fw_uid; 136 int fw_prid; 137}; 138 139/* 140 * list of rules for layer 3 141 */ 142struct ip_fw_chain layer3_chain; 143 144MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 145MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 146#define IPFW_NAT_LOADED (ipfw_nat_ptr != NULL) 147ipfw_nat_t *ipfw_nat_ptr = NULL; 148ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 149ipfw_nat_cfg_t *ipfw_nat_del_ptr; 150ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 151ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 152 153struct table_entry { 154 struct radix_node rn[2]; 155 struct sockaddr_in addr, mask; 156 u_int32_t value; 157}; 158 159static int fw_debug = 1; 160static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 161 162extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 163 164#ifdef SYSCTL_NODE 165SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 166SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 167 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 168 ipfw_chg_hook, "I", "Enable ipfw"); 169SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 170 &autoinc_step, 0, "Rule number autincrement step"); 171SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 172 CTLFLAG_RW | CTLFLAG_SECURE3, 173 &fw_one_pass, 0, 174 "Only do a single pass through ipfw when using dummynet(4)"); 175SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 176 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 177SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 178 CTLFLAG_RW | CTLFLAG_SECURE3, 179 &fw_verbose, 0, "Log matches to ipfw rules"); 180SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 181 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 182SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 183 NULL, IPFW_DEFAULT_RULE, "The default/max possible rule number."); 184SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 185 NULL, IPFW_TABLES_MAX, "The maximum number of tables."); 186 187/* 188 * Description of dynamic rules. 189 * 190 * Dynamic rules are stored in lists accessed through a hash table 191 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 192 * be modified through the sysctl variable dyn_buckets which is 193 * updated when the table becomes empty. 194 * 195 * XXX currently there is only one list, ipfw_dyn. 196 * 197 * When a packet is received, its address fields are first masked 198 * with the mask defined for the rule, then hashed, then matched 199 * against the entries in the corresponding list. 200 * Dynamic rules can be used for different purposes: 201 * + stateful rules; 202 * + enforcing limits on the number of sessions; 203 * + in-kernel NAT (not implemented yet) 204 * 205 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 206 * measured in seconds and depending on the flags. 207 * 208 * The total number of dynamic rules is stored in dyn_count. 209 * The max number of dynamic rules is dyn_max. When we reach 210 * the maximum number of rules we do not create anymore. This is 211 * done to avoid consuming too much memory, but also too much 212 * time when searching on each packet (ideally, we should try instead 213 * to put a limit on the length of the list on each bucket...). 214 * 215 * Each dynamic rule holds a pointer to the parent ipfw rule so 216 * we know what action to perform. Dynamic rules are removed when 217 * the parent rule is deleted. XXX we should make them survive. 218 * 219 * There are some limitations with dynamic rules -- we do not 220 * obey the 'randomized match', and we do not do multiple 221 * passes through the firewall. XXX check the latter!!! 222 */ 223static ipfw_dyn_rule **ipfw_dyn_v = NULL; 224static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 225static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 226 227static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 228#define IPFW_DYN_LOCK_INIT() \ 229 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 230#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 231#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 232#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 233#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 234 235/* 236 * Timeouts for various events in handing dynamic rules. 237 */ 238static u_int32_t dyn_ack_lifetime = 300; 239static u_int32_t dyn_syn_lifetime = 20; 240static u_int32_t dyn_fin_lifetime = 1; 241static u_int32_t dyn_rst_lifetime = 1; 242static u_int32_t dyn_udp_lifetime = 10; 243static u_int32_t dyn_short_lifetime = 5; 244 245/* 246 * Keepalives are sent if dyn_keepalive is set. They are sent every 247 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 248 * seconds of lifetime of a rule. 249 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 250 * than dyn_keepalive_period. 251 */ 252 253static u_int32_t dyn_keepalive_interval = 20; 254static u_int32_t dyn_keepalive_period = 5; 255static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 256 257static u_int32_t static_count; /* # of static rules */ 258static u_int32_t static_len; /* size in bytes of static rules */ 259static u_int32_t dyn_count; /* # of dynamic rules */ 260static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 261 262SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 263 &dyn_buckets, 0, "Number of dyn. buckets"); 264SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 265 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 266SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 267 &dyn_count, 0, "Number of dyn. rules"); 268SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 269 &dyn_max, 0, "Max number of dyn. rules"); 270SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 271 &static_count, 0, "Number of static rules"); 272SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 273 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 274SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 275 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 276SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 277 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 278SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 279 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 280SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 281 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 282SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 283 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 284SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 285 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 286 287#ifdef INET6 288/* 289 * IPv6 specific variables 290 */ 291SYSCTL_DECL(_net_inet6_ip6); 292 293static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 294static struct sysctl_oid *ip6_fw_sysctl_tree; 295#endif /* INET6 */ 296#endif /* SYSCTL_NODE */ 297 298static int fw_deny_unknown_exthdrs = 1; 299 300 301/* 302 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 303 * Other macros just cast void * into the appropriate type 304 */ 305#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 306#define TCP(p) ((struct tcphdr *)(p)) 307#define SCTP(p) ((struct sctphdr *)(p)) 308#define UDP(p) ((struct udphdr *)(p)) 309#define ICMP(p) ((struct icmphdr *)(p)) 310#define ICMP6(p) ((struct icmp6_hdr *)(p)) 311 312static __inline int 313icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 314{ 315 int type = icmp->icmp_type; 316 317 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 318} 319 320#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 321 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 322 323static int 324is_icmp_query(struct icmphdr *icmp) 325{ 326 int type = icmp->icmp_type; 327 328 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 329} 330#undef TT 331 332/* 333 * The following checks use two arrays of 8 or 16 bits to store the 334 * bits that we want set or clear, respectively. They are in the 335 * low and high half of cmd->arg1 or cmd->d[0]. 336 * 337 * We scan options and store the bits we find set. We succeed if 338 * 339 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 340 * 341 * The code is sometimes optimized not to store additional variables. 342 */ 343 344static int 345flags_match(ipfw_insn *cmd, u_int8_t bits) 346{ 347 u_char want_clear; 348 bits = ~bits; 349 350 if ( ((cmd->arg1 & 0xff) & bits) != 0) 351 return 0; /* some bits we want set were clear */ 352 want_clear = (cmd->arg1 >> 8) & 0xff; 353 if ( (want_clear & bits) != want_clear) 354 return 0; /* some bits we want clear were set */ 355 return 1; 356} 357 358static int 359ipopts_match(struct ip *ip, ipfw_insn *cmd) 360{ 361 int optlen, bits = 0; 362 u_char *cp = (u_char *)(ip + 1); 363 int x = (ip->ip_hl << 2) - sizeof (struct ip); 364 365 for (; x > 0; x -= optlen, cp += optlen) { 366 int opt = cp[IPOPT_OPTVAL]; 367 368 if (opt == IPOPT_EOL) 369 break; 370 if (opt == IPOPT_NOP) 371 optlen = 1; 372 else { 373 optlen = cp[IPOPT_OLEN]; 374 if (optlen <= 0 || optlen > x) 375 return 0; /* invalid or truncated */ 376 } 377 switch (opt) { 378 379 default: 380 break; 381 382 case IPOPT_LSRR: 383 bits |= IP_FW_IPOPT_LSRR; 384 break; 385 386 case IPOPT_SSRR: 387 bits |= IP_FW_IPOPT_SSRR; 388 break; 389 390 case IPOPT_RR: 391 bits |= IP_FW_IPOPT_RR; 392 break; 393 394 case IPOPT_TS: 395 bits |= IP_FW_IPOPT_TS; 396 break; 397 } 398 } 399 return (flags_match(cmd, bits)); 400} 401 402static int 403tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 404{ 405 int optlen, bits = 0; 406 u_char *cp = (u_char *)(tcp + 1); 407 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 408 409 for (; x > 0; x -= optlen, cp += optlen) { 410 int opt = cp[0]; 411 if (opt == TCPOPT_EOL) 412 break; 413 if (opt == TCPOPT_NOP) 414 optlen = 1; 415 else { 416 optlen = cp[1]; 417 if (optlen <= 0) 418 break; 419 } 420 421 switch (opt) { 422 423 default: 424 break; 425 426 case TCPOPT_MAXSEG: 427 bits |= IP_FW_TCPOPT_MSS; 428 break; 429 430 case TCPOPT_WINDOW: 431 bits |= IP_FW_TCPOPT_WINDOW; 432 break; 433 434 case TCPOPT_SACK_PERMITTED: 435 case TCPOPT_SACK: 436 bits |= IP_FW_TCPOPT_SACK; 437 break; 438 439 case TCPOPT_TIMESTAMP: 440 bits |= IP_FW_TCPOPT_TS; 441 break; 442 443 } 444 } 445 return (flags_match(cmd, bits)); 446} 447 448static int 449iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 450{ 451 if (ifp == NULL) /* no iface with this packet, match fails */ 452 return 0; 453 /* Check by name or by IP address */ 454 if (cmd->name[0] != '\0') { /* match by name */ 455 /* Check name */ 456 if (cmd->p.glob) { 457 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 458 return(1); 459 } else { 460 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 461 return(1); 462 } 463 } else { 464 struct ifaddr *ia; 465 466 /* XXX lock? */ 467 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 468 if (ia->ifa_addr->sa_family != AF_INET) 469 continue; 470 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 471 (ia->ifa_addr))->sin_addr.s_addr) 472 return(1); /* match */ 473 } 474 } 475 return(0); /* no match, fail ... */ 476} 477 478/* 479 * The verify_path function checks if a route to the src exists and 480 * if it is reachable via ifp (when provided). 481 * 482 * The 'verrevpath' option checks that the interface that an IP packet 483 * arrives on is the same interface that traffic destined for the 484 * packet's source address would be routed out of. The 'versrcreach' 485 * option just checks that the source address is reachable via any route 486 * (except default) in the routing table. These two are a measure to block 487 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 488 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 489 * is purposely reminiscent of the Cisco IOS command, 490 * 491 * ip verify unicast reverse-path 492 * ip verify unicast source reachable-via any 493 * 494 * which implements the same functionality. But note that syntax is 495 * misleading. The check may be performed on all IP packets whether unicast, 496 * multicast, or broadcast. 497 */ 498static int 499verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 500{ 501 struct route ro; 502 struct sockaddr_in *dst; 503 504 bzero(&ro, sizeof(ro)); 505 506 dst = (struct sockaddr_in *)&(ro.ro_dst); 507 dst->sin_family = AF_INET; 508 dst->sin_len = sizeof(*dst); 509 dst->sin_addr = src; 510 in_rtalloc_ign(&ro, RTF_CLONING, fib); 511 512 if (ro.ro_rt == NULL) 513 return 0; 514 515 /* 516 * If ifp is provided, check for equality with rtentry. 517 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 518 * in order to pass packets injected back by if_simloop(): 519 * if useloopback == 1 routing entry (via lo0) for our own address 520 * may exist, so we need to handle routing assymetry. 521 */ 522 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 523 RTFREE(ro.ro_rt); 524 return 0; 525 } 526 527 /* if no ifp provided, check if rtentry is not default route */ 528 if (ifp == NULL && 529 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 530 RTFREE(ro.ro_rt); 531 return 0; 532 } 533 534 /* or if this is a blackhole/reject route */ 535 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 536 RTFREE(ro.ro_rt); 537 return 0; 538 } 539 540 /* found valid route */ 541 RTFREE(ro.ro_rt); 542 return 1; 543} 544 545#ifdef INET6 546/* 547 * ipv6 specific rules here... 548 */ 549static __inline int 550icmp6type_match (int type, ipfw_insn_u32 *cmd) 551{ 552 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 553} 554 555static int 556flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 557{ 558 int i; 559 for (i=0; i <= cmd->o.arg1; ++i ) 560 if (curr_flow == cmd->d[i] ) 561 return 1; 562 return 0; 563} 564 565/* support for IP6_*_ME opcodes */ 566static int 567search_ip6_addr_net (struct in6_addr * ip6_addr) 568{ 569 struct ifnet *mdc; 570 struct ifaddr *mdc2; 571 struct in6_ifaddr *fdm; 572 struct in6_addr copia; 573 574 TAILQ_FOREACH(mdc, &V_ifnet, if_link) 575 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 576 if (mdc2->ifa_addr->sa_family == AF_INET6) { 577 fdm = (struct in6_ifaddr *)mdc2; 578 copia = fdm->ia_addr.sin6_addr; 579 /* need for leaving scope_id in the sock_addr */ 580 in6_clearscope(&copia); 581 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 582 return 1; 583 } 584 } 585 return 0; 586} 587 588static int 589verify_path6(struct in6_addr *src, struct ifnet *ifp) 590{ 591 struct route_in6 ro; 592 struct sockaddr_in6 *dst; 593 594 bzero(&ro, sizeof(ro)); 595 596 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 597 dst->sin6_family = AF_INET6; 598 dst->sin6_len = sizeof(*dst); 599 dst->sin6_addr = *src; 600 /* XXX MRT 0 for ipv6 at this time */ 601 rtalloc_ign((struct route *)&ro, RTF_CLONING); 602 603 if (ro.ro_rt == NULL) 604 return 0; 605 606 /* 607 * if ifp is provided, check for equality with rtentry 608 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 609 * to support the case of sending packets to an address of our own. 610 * (where the former interface is the first argument of if_simloop() 611 * (=ifp), the latter is lo0) 612 */ 613 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 614 RTFREE(ro.ro_rt); 615 return 0; 616 } 617 618 /* if no ifp provided, check if rtentry is not default route */ 619 if (ifp == NULL && 620 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 621 RTFREE(ro.ro_rt); 622 return 0; 623 } 624 625 /* or if this is a blackhole/reject route */ 626 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 627 RTFREE(ro.ro_rt); 628 return 0; 629 } 630 631 /* found valid route */ 632 RTFREE(ro.ro_rt); 633 return 1; 634 635} 636static __inline int 637hash_packet6(struct ipfw_flow_id *id) 638{ 639 u_int32_t i; 640 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 641 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 642 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 643 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 644 (id->dst_port) ^ (id->src_port); 645 return i; 646} 647 648static int 649is_icmp6_query(int icmp6_type) 650{ 651 if ((icmp6_type <= ICMP6_MAXTYPE) && 652 (icmp6_type == ICMP6_ECHO_REQUEST || 653 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 654 icmp6_type == ICMP6_WRUREQUEST || 655 icmp6_type == ICMP6_FQDN_QUERY || 656 icmp6_type == ICMP6_NI_QUERY)) 657 return (1); 658 659 return (0); 660} 661 662static void 663send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 664{ 665 struct mbuf *m; 666 667 m = args->m; 668 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 669 struct tcphdr *tcp; 670 tcp_seq ack, seq; 671 int flags; 672 struct { 673 struct ip6_hdr ip6; 674 struct tcphdr th; 675 } ti; 676 tcp = (struct tcphdr *)((char *)ip6 + hlen); 677 678 if ((tcp->th_flags & TH_RST) != 0) { 679 m_freem(m); 680 args->m = NULL; 681 return; 682 } 683 684 ti.ip6 = *ip6; 685 ti.th = *tcp; 686 ti.th.th_seq = ntohl(ti.th.th_seq); 687 ti.th.th_ack = ntohl(ti.th.th_ack); 688 ti.ip6.ip6_nxt = IPPROTO_TCP; 689 690 if (ti.th.th_flags & TH_ACK) { 691 ack = 0; 692 seq = ti.th.th_ack; 693 flags = TH_RST; 694 } else { 695 ack = ti.th.th_seq; 696 if ((m->m_flags & M_PKTHDR) != 0) { 697 /* 698 * total new data to ACK is: 699 * total packet length, 700 * minus the header length, 701 * minus the tcp header length. 702 */ 703 ack += m->m_pkthdr.len - hlen 704 - (ti.th.th_off << 2); 705 } else if (ip6->ip6_plen) { 706 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 707 hlen - (ti.th.th_off << 2); 708 } else { 709 m_freem(m); 710 return; 711 } 712 if (tcp->th_flags & TH_SYN) 713 ack++; 714 seq = 0; 715 flags = TH_RST|TH_ACK; 716 } 717 bcopy(&ti, ip6, sizeof(ti)); 718 /* 719 * m is only used to recycle the mbuf 720 * The data in it is never read so we don't need 721 * to correct the offsets or anything 722 */ 723 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 724 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 725#if 0 726 /* 727 * Unlike above, the mbufs need to line up with the ip6 hdr, 728 * as the contents are read. We need to m_adj() the 729 * needed amount. 730 * The mbuf will however be thrown away so we can adjust it. 731 * Remember we did an m_pullup on it already so we 732 * can make some assumptions about contiguousness. 733 */ 734 if (args->L3offset) 735 m_adj(m, args->L3offset); 736#endif 737 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 738 } else 739 m_freem(m); 740 741 args->m = NULL; 742} 743 744#endif /* INET6 */ 745 746static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 747 748#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 749#define SNP(buf) buf, sizeof(buf) 750 751/* 752 * We enter here when we have a rule with O_LOG. 753 * XXX this function alone takes about 2Kbytes of code! 754 */ 755static void 756ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 757 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 758 struct ip *ip) 759{ 760 struct ether_header *eh = args->eh; 761 char *action; 762 int limit_reached = 0; 763 char action2[40], proto[128], fragment[32]; 764 765 fragment[0] = '\0'; 766 proto[0] = '\0'; 767 768 if (f == NULL) { /* bogus pkt */ 769 if (V_verbose_limit != 0 && V_norule_counter >= V_verbose_limit) 770 return; 771 V_norule_counter++; 772 if (V_norule_counter == V_verbose_limit) 773 limit_reached = V_verbose_limit; 774 action = "Refuse"; 775 } else { /* O_LOG is the first action, find the real one */ 776 ipfw_insn *cmd = ACTION_PTR(f); 777 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 778 779 if (l->max_log != 0 && l->log_left == 0) 780 return; 781 l->log_left--; 782 if (l->log_left == 0) 783 limit_reached = l->max_log; 784 cmd += F_LEN(cmd); /* point to first action */ 785 if (cmd->opcode == O_ALTQ) { 786 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 787 788 snprintf(SNPARGS(action2, 0), "Altq %d", 789 altq->qid); 790 cmd += F_LEN(cmd); 791 } 792 if (cmd->opcode == O_PROB) 793 cmd += F_LEN(cmd); 794 795 if (cmd->opcode == O_TAG) 796 cmd += F_LEN(cmd); 797 798 action = action2; 799 switch (cmd->opcode) { 800 case O_DENY: 801 action = "Deny"; 802 break; 803 804 case O_REJECT: 805 if (cmd->arg1==ICMP_REJECT_RST) 806 action = "Reset"; 807 else if (cmd->arg1==ICMP_UNREACH_HOST) 808 action = "Reject"; 809 else 810 snprintf(SNPARGS(action2, 0), "Unreach %d", 811 cmd->arg1); 812 break; 813 814 case O_UNREACH6: 815 if (cmd->arg1==ICMP6_UNREACH_RST) 816 action = "Reset"; 817 else 818 snprintf(SNPARGS(action2, 0), "Unreach %d", 819 cmd->arg1); 820 break; 821 822 case O_ACCEPT: 823 action = "Accept"; 824 break; 825 case O_COUNT: 826 action = "Count"; 827 break; 828 case O_DIVERT: 829 snprintf(SNPARGS(action2, 0), "Divert %d", 830 cmd->arg1); 831 break; 832 case O_TEE: 833 snprintf(SNPARGS(action2, 0), "Tee %d", 834 cmd->arg1); 835 break; 836 case O_SETFIB: 837 snprintf(SNPARGS(action2, 0), "SetFib %d", 838 cmd->arg1); 839 break; 840 case O_SKIPTO: 841 snprintf(SNPARGS(action2, 0), "SkipTo %d", 842 cmd->arg1); 843 break; 844 case O_PIPE: 845 snprintf(SNPARGS(action2, 0), "Pipe %d", 846 cmd->arg1); 847 break; 848 case O_QUEUE: 849 snprintf(SNPARGS(action2, 0), "Queue %d", 850 cmd->arg1); 851 break; 852 case O_FORWARD_IP: { 853 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 854 int len; 855 struct in_addr dummyaddr; 856 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 857 dummyaddr.s_addr = htonl(tablearg); 858 else 859 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 860 861 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 862 inet_ntoa(dummyaddr)); 863 864 if (sa->sa.sin_port) 865 snprintf(SNPARGS(action2, len), ":%d", 866 sa->sa.sin_port); 867 } 868 break; 869 case O_NETGRAPH: 870 snprintf(SNPARGS(action2, 0), "Netgraph %d", 871 cmd->arg1); 872 break; 873 case O_NGTEE: 874 snprintf(SNPARGS(action2, 0), "Ngtee %d", 875 cmd->arg1); 876 break; 877 case O_NAT: 878 action = "Nat"; 879 break; 880 default: 881 action = "UNKNOWN"; 882 break; 883 } 884 } 885 886 if (hlen == 0) { /* non-ip */ 887 snprintf(SNPARGS(proto, 0), "MAC"); 888 889 } else { 890 int len; 891 char src[48], dst[48]; 892 struct icmphdr *icmp; 893 struct tcphdr *tcp; 894 struct udphdr *udp; 895#ifdef INET6 896 struct ip6_hdr *ip6 = NULL; 897 struct icmp6_hdr *icmp6; 898#endif 899 src[0] = '\0'; 900 dst[0] = '\0'; 901#ifdef INET6 902 if (IS_IP6_FLOW_ID(&(args->f_id))) { 903 char ip6buf[INET6_ADDRSTRLEN]; 904 snprintf(src, sizeof(src), "[%s]", 905 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 906 snprintf(dst, sizeof(dst), "[%s]", 907 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 908 909 ip6 = (struct ip6_hdr *)ip; 910 tcp = (struct tcphdr *)(((char *)ip) + hlen); 911 udp = (struct udphdr *)(((char *)ip) + hlen); 912 } else 913#endif 914 { 915 tcp = L3HDR(struct tcphdr, ip); 916 udp = L3HDR(struct udphdr, ip); 917 918 inet_ntoa_r(ip->ip_src, src); 919 inet_ntoa_r(ip->ip_dst, dst); 920 } 921 922 switch (args->f_id.proto) { 923 case IPPROTO_TCP: 924 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 925 if (offset == 0) 926 snprintf(SNPARGS(proto, len), ":%d %s:%d", 927 ntohs(tcp->th_sport), 928 dst, 929 ntohs(tcp->th_dport)); 930 else 931 snprintf(SNPARGS(proto, len), " %s", dst); 932 break; 933 934 case IPPROTO_UDP: 935 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 936 if (offset == 0) 937 snprintf(SNPARGS(proto, len), ":%d %s:%d", 938 ntohs(udp->uh_sport), 939 dst, 940 ntohs(udp->uh_dport)); 941 else 942 snprintf(SNPARGS(proto, len), " %s", dst); 943 break; 944 945 case IPPROTO_ICMP: 946 icmp = L3HDR(struct icmphdr, ip); 947 if (offset == 0) 948 len = snprintf(SNPARGS(proto, 0), 949 "ICMP:%u.%u ", 950 icmp->icmp_type, icmp->icmp_code); 951 else 952 len = snprintf(SNPARGS(proto, 0), "ICMP "); 953 len += snprintf(SNPARGS(proto, len), "%s", src); 954 snprintf(SNPARGS(proto, len), " %s", dst); 955 break; 956#ifdef INET6 957 case IPPROTO_ICMPV6: 958 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 959 if (offset == 0) 960 len = snprintf(SNPARGS(proto, 0), 961 "ICMPv6:%u.%u ", 962 icmp6->icmp6_type, icmp6->icmp6_code); 963 else 964 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 965 len += snprintf(SNPARGS(proto, len), "%s", src); 966 snprintf(SNPARGS(proto, len), " %s", dst); 967 break; 968#endif 969 default: 970 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 971 args->f_id.proto, src); 972 snprintf(SNPARGS(proto, len), " %s", dst); 973 break; 974 } 975 976#ifdef INET6 977 if (IS_IP6_FLOW_ID(&(args->f_id))) { 978 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 979 snprintf(SNPARGS(fragment, 0), 980 " (frag %08x:%d@%d%s)", 981 args->f_id.frag_id6, 982 ntohs(ip6->ip6_plen) - hlen, 983 ntohs(offset & IP6F_OFF_MASK) << 3, 984 (offset & IP6F_MORE_FRAG) ? "+" : ""); 985 } else 986#endif 987 { 988 int ip_off, ip_len; 989 if (eh != NULL) { /* layer 2 packets are as on the wire */ 990 ip_off = ntohs(ip->ip_off); 991 ip_len = ntohs(ip->ip_len); 992 } else { 993 ip_off = ip->ip_off; 994 ip_len = ip->ip_len; 995 } 996 if (ip_off & (IP_MF | IP_OFFMASK)) 997 snprintf(SNPARGS(fragment, 0), 998 " (frag %d:%d@%d%s)", 999 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1000 offset << 3, 1001 (ip_off & IP_MF) ? "+" : ""); 1002 } 1003 } 1004 if (oif || m->m_pkthdr.rcvif) 1005 log(LOG_SECURITY | LOG_INFO, 1006 "ipfw: %d %s %s %s via %s%s\n", 1007 f ? f->rulenum : -1, 1008 action, proto, oif ? "out" : "in", 1009 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1010 fragment); 1011 else 1012 log(LOG_SECURITY | LOG_INFO, 1013 "ipfw: %d %s %s [no if info]%s\n", 1014 f ? f->rulenum : -1, 1015 action, proto, fragment); 1016 if (limit_reached) 1017 log(LOG_SECURITY | LOG_NOTICE, 1018 "ipfw: limit %d reached on entry %d\n", 1019 limit_reached, f ? f->rulenum : -1); 1020} 1021 1022/* 1023 * IMPORTANT: the hash function for dynamic rules must be commutative 1024 * in source and destination (ip,port), because rules are bidirectional 1025 * and we want to find both in the same bucket. 1026 */ 1027static __inline int 1028hash_packet(struct ipfw_flow_id *id) 1029{ 1030 u_int32_t i; 1031 1032#ifdef INET6 1033 if (IS_IP6_FLOW_ID(id)) 1034 i = hash_packet6(id); 1035 else 1036#endif /* INET6 */ 1037 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1038 i &= (V_curr_dyn_buckets - 1); 1039 return i; 1040} 1041 1042/** 1043 * unlink a dynamic rule from a chain. prev is a pointer to 1044 * the previous one, q is a pointer to the rule to delete, 1045 * head is a pointer to the head of the queue. 1046 * Modifies q and potentially also head. 1047 */ 1048#define UNLINK_DYN_RULE(prev, head, q) { \ 1049 ipfw_dyn_rule *old_q = q; \ 1050 \ 1051 /* remove a refcount to the parent */ \ 1052 if (q->dyn_type == O_LIMIT) \ 1053 q->parent->count--; \ 1054 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1055 (q->id.src_ip), (q->id.src_port), \ 1056 (q->id.dst_ip), (q->id.dst_port), V_dyn_count-1 ); ) \ 1057 if (prev != NULL) \ 1058 prev->next = q = q->next; \ 1059 else \ 1060 head = q = q->next; \ 1061 V_dyn_count--; \ 1062 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1063 1064#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1065 1066/** 1067 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1068 * 1069 * If keep_me == NULL, rules are deleted even if not expired, 1070 * otherwise only expired rules are removed. 1071 * 1072 * The value of the second parameter is also used to point to identify 1073 * a rule we absolutely do not want to remove (e.g. because we are 1074 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1075 * rules). The pointer is only used for comparison, so any non-null 1076 * value will do. 1077 */ 1078static void 1079remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1080{ 1081 static u_int32_t last_remove = 0; 1082 1083#define FORCE (keep_me == NULL) 1084 1085 ipfw_dyn_rule *prev, *q; 1086 int i, pass = 0, max_pass = 0; 1087 1088 IPFW_DYN_LOCK_ASSERT(); 1089 1090 if (V_ipfw_dyn_v == NULL || V_dyn_count == 0) 1091 return; 1092 /* do not expire more than once per second, it is useless */ 1093 if (!FORCE && last_remove == time_uptime) 1094 return; 1095 last_remove = time_uptime; 1096 1097 /* 1098 * because O_LIMIT refer to parent rules, during the first pass only 1099 * remove child and mark any pending LIMIT_PARENT, and remove 1100 * them in a second pass. 1101 */ 1102next_pass: 1103 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 1104 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q ; ) { 1105 /* 1106 * Logic can become complex here, so we split tests. 1107 */ 1108 if (q == keep_me) 1109 goto next; 1110 if (rule != NULL && rule != q->rule) 1111 goto next; /* not the one we are looking for */ 1112 if (q->dyn_type == O_LIMIT_PARENT) { 1113 /* 1114 * handle parent in the second pass, 1115 * record we need one. 1116 */ 1117 max_pass = 1; 1118 if (pass == 0) 1119 goto next; 1120 if (FORCE && q->count != 0 ) { 1121 /* XXX should not happen! */ 1122 printf("ipfw: OUCH! cannot remove rule," 1123 " count %d\n", q->count); 1124 } 1125 } else { 1126 if (!FORCE && 1127 !TIME_LEQ( q->expire, time_uptime )) 1128 goto next; 1129 } 1130 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1131 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1132 continue; 1133 } 1134next: 1135 prev=q; 1136 q=q->next; 1137 } 1138 } 1139 if (pass++ < max_pass) 1140 goto next_pass; 1141} 1142 1143 1144/** 1145 * lookup a dynamic rule. 1146 */ 1147static ipfw_dyn_rule * 1148lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1149 struct tcphdr *tcp) 1150{ 1151 /* 1152 * stateful ipfw extensions. 1153 * Lookup into dynamic session queue 1154 */ 1155#define MATCH_REVERSE 0 1156#define MATCH_FORWARD 1 1157#define MATCH_NONE 2 1158#define MATCH_UNKNOWN 3 1159 int i, dir = MATCH_NONE; 1160 ipfw_dyn_rule *prev, *q=NULL; 1161 1162 IPFW_DYN_LOCK_ASSERT(); 1163 1164 if (V_ipfw_dyn_v == NULL) 1165 goto done; /* not found */ 1166 i = hash_packet( pkt ); 1167 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q != NULL ; ) { 1168 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1169 goto next; 1170 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1171 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1172 continue; 1173 } 1174 if (pkt->proto == q->id.proto && 1175 q->dyn_type != O_LIMIT_PARENT) { 1176 if (IS_IP6_FLOW_ID(pkt)) { 1177 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1178 &(q->id.src_ip6)) && 1179 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1180 &(q->id.dst_ip6)) && 1181 pkt->src_port == q->id.src_port && 1182 pkt->dst_port == q->id.dst_port ) { 1183 dir = MATCH_FORWARD; 1184 break; 1185 } 1186 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1187 &(q->id.dst_ip6)) && 1188 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1189 &(q->id.src_ip6)) && 1190 pkt->src_port == q->id.dst_port && 1191 pkt->dst_port == q->id.src_port ) { 1192 dir = MATCH_REVERSE; 1193 break; 1194 } 1195 } else { 1196 if (pkt->src_ip == q->id.src_ip && 1197 pkt->dst_ip == q->id.dst_ip && 1198 pkt->src_port == q->id.src_port && 1199 pkt->dst_port == q->id.dst_port ) { 1200 dir = MATCH_FORWARD; 1201 break; 1202 } 1203 if (pkt->src_ip == q->id.dst_ip && 1204 pkt->dst_ip == q->id.src_ip && 1205 pkt->src_port == q->id.dst_port && 1206 pkt->dst_port == q->id.src_port ) { 1207 dir = MATCH_REVERSE; 1208 break; 1209 } 1210 } 1211 } 1212next: 1213 prev = q; 1214 q = q->next; 1215 } 1216 if (q == NULL) 1217 goto done; /* q = NULL, not found */ 1218 1219 if ( prev != NULL) { /* found and not in front */ 1220 prev->next = q->next; 1221 q->next = V_ipfw_dyn_v[i]; 1222 V_ipfw_dyn_v[i] = q; 1223 } 1224 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1225 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1226 1227#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1228#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1229 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1230 switch (q->state) { 1231 case TH_SYN: /* opening */ 1232 q->expire = time_uptime + V_dyn_syn_lifetime; 1233 break; 1234 1235 case BOTH_SYN: /* move to established */ 1236 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1237 case BOTH_SYN | (TH_FIN << 8) : 1238 if (tcp) { 1239#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1240 u_int32_t ack = ntohl(tcp->th_ack); 1241 if (dir == MATCH_FORWARD) { 1242 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1243 q->ack_fwd = ack; 1244 else { /* ignore out-of-sequence */ 1245 break; 1246 } 1247 } else { 1248 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1249 q->ack_rev = ack; 1250 else { /* ignore out-of-sequence */ 1251 break; 1252 } 1253 } 1254 } 1255 q->expire = time_uptime + V_dyn_ack_lifetime; 1256 break; 1257 1258 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1259 if (V_dyn_fin_lifetime >= V_dyn_keepalive_period) 1260 V_dyn_fin_lifetime = V_dyn_keepalive_period - 1; 1261 q->expire = time_uptime + V_dyn_fin_lifetime; 1262 break; 1263 1264 default: 1265#if 0 1266 /* 1267 * reset or some invalid combination, but can also 1268 * occur if we use keep-state the wrong way. 1269 */ 1270 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1271 printf("invalid state: 0x%x\n", q->state); 1272#endif 1273 if (V_dyn_rst_lifetime >= V_dyn_keepalive_period) 1274 V_dyn_rst_lifetime = V_dyn_keepalive_period - 1; 1275 q->expire = time_uptime + V_dyn_rst_lifetime; 1276 break; 1277 } 1278 } else if (pkt->proto == IPPROTO_UDP) { 1279 q->expire = time_uptime + V_dyn_udp_lifetime; 1280 } else { 1281 /* other protocols */ 1282 q->expire = time_uptime + V_dyn_short_lifetime; 1283 } 1284done: 1285 if (match_direction) 1286 *match_direction = dir; 1287 return q; 1288} 1289 1290static ipfw_dyn_rule * 1291lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1292 struct tcphdr *tcp) 1293{ 1294 ipfw_dyn_rule *q; 1295 1296 IPFW_DYN_LOCK(); 1297 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1298 if (q == NULL) 1299 IPFW_DYN_UNLOCK(); 1300 /* NB: return table locked when q is not NULL */ 1301 return q; 1302} 1303 1304static void 1305realloc_dynamic_table(void) 1306{ 1307 IPFW_DYN_LOCK_ASSERT(); 1308 1309 /* 1310 * Try reallocation, make sure we have a power of 2 and do 1311 * not allow more than 64k entries. In case of overflow, 1312 * default to 1024. 1313 */ 1314 1315 if (V_dyn_buckets > 65536) 1316 V_dyn_buckets = 1024; 1317 if ((V_dyn_buckets & (V_dyn_buckets-1)) != 0) { /* not a power of 2 */ 1318 V_dyn_buckets = V_curr_dyn_buckets; /* reset */ 1319 return; 1320 } 1321 V_curr_dyn_buckets = V_dyn_buckets; 1322 if (V_ipfw_dyn_v != NULL) 1323 free(V_ipfw_dyn_v, M_IPFW); 1324 for (;;) { 1325 V_ipfw_dyn_v = malloc(V_curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1326 M_IPFW, M_NOWAIT | M_ZERO); 1327 if (V_ipfw_dyn_v != NULL || V_curr_dyn_buckets <= 2) 1328 break; 1329 V_curr_dyn_buckets /= 2; 1330 } 1331} 1332 1333/** 1334 * Install state of type 'type' for a dynamic session. 1335 * The hash table contains two type of rules: 1336 * - regular rules (O_KEEP_STATE) 1337 * - rules for sessions with limited number of sess per user 1338 * (O_LIMIT). When they are created, the parent is 1339 * increased by 1, and decreased on delete. In this case, 1340 * the third parameter is the parent rule and not the chain. 1341 * - "parent" rules for the above (O_LIMIT_PARENT). 1342 */ 1343static ipfw_dyn_rule * 1344add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1345{ 1346 ipfw_dyn_rule *r; 1347 int i; 1348 1349 IPFW_DYN_LOCK_ASSERT(); 1350 1351 if (V_ipfw_dyn_v == NULL || 1352 (V_dyn_count == 0 && V_dyn_buckets != V_curr_dyn_buckets)) { 1353 realloc_dynamic_table(); 1354 if (V_ipfw_dyn_v == NULL) 1355 return NULL; /* failed ! */ 1356 } 1357 i = hash_packet(id); 1358 1359 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1360 if (r == NULL) { 1361 printf ("ipfw: sorry cannot allocate state\n"); 1362 return NULL; 1363 } 1364 1365 /* increase refcount on parent, and set pointer */ 1366 if (dyn_type == O_LIMIT) { 1367 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1368 if ( parent->dyn_type != O_LIMIT_PARENT) 1369 panic("invalid parent"); 1370 parent->count++; 1371 r->parent = parent; 1372 rule = parent->rule; 1373 } 1374 1375 r->id = *id; 1376 r->expire = time_uptime + V_dyn_syn_lifetime; 1377 r->rule = rule; 1378 r->dyn_type = dyn_type; 1379 r->pcnt = r->bcnt = 0; 1380 r->count = 0; 1381 1382 r->bucket = i; 1383 r->next = V_ipfw_dyn_v[i]; 1384 V_ipfw_dyn_v[i] = r; 1385 V_dyn_count++; 1386 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1387 dyn_type, 1388 (r->id.src_ip), (r->id.src_port), 1389 (r->id.dst_ip), (r->id.dst_port), 1390 V_dyn_count ); ) 1391 return r; 1392} 1393 1394/** 1395 * lookup dynamic parent rule using pkt and rule as search keys. 1396 * If the lookup fails, then install one. 1397 */ 1398static ipfw_dyn_rule * 1399lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1400{ 1401 ipfw_dyn_rule *q; 1402 int i; 1403 1404 IPFW_DYN_LOCK_ASSERT(); 1405 1406 if (V_ipfw_dyn_v) { 1407 int is_v6 = IS_IP6_FLOW_ID(pkt); 1408 i = hash_packet( pkt ); 1409 for (q = V_ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1410 if (q->dyn_type == O_LIMIT_PARENT && 1411 rule== q->rule && 1412 pkt->proto == q->id.proto && 1413 pkt->src_port == q->id.src_port && 1414 pkt->dst_port == q->id.dst_port && 1415 ( 1416 (is_v6 && 1417 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1418 &(q->id.src_ip6)) && 1419 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1420 &(q->id.dst_ip6))) || 1421 (!is_v6 && 1422 pkt->src_ip == q->id.src_ip && 1423 pkt->dst_ip == q->id.dst_ip) 1424 ) 1425 ) { 1426 q->expire = time_uptime + V_dyn_short_lifetime; 1427 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1428 return q; 1429 } 1430 } 1431 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1432} 1433 1434/** 1435 * Install dynamic state for rule type cmd->o.opcode 1436 * 1437 * Returns 1 (failure) if state is not installed because of errors or because 1438 * session limitations are enforced. 1439 */ 1440static int 1441install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1442 struct ip_fw_args *args, uint32_t tablearg) 1443{ 1444 static int last_log; 1445 ipfw_dyn_rule *q; 1446 struct in_addr da; 1447 char src[48], dst[48]; 1448 1449 src[0] = '\0'; 1450 dst[0] = '\0'; 1451 1452 DEB( 1453 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1454 __func__, cmd->o.opcode, 1455 (args->f_id.src_ip), (args->f_id.src_port), 1456 (args->f_id.dst_ip), (args->f_id.dst_port)); 1457 ) 1458 1459 IPFW_DYN_LOCK(); 1460 1461 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1462 1463 if (q != NULL) { /* should never occur */ 1464 if (last_log != time_uptime) { 1465 last_log = time_uptime; 1466 printf("ipfw: %s: entry already present, done\n", 1467 __func__); 1468 } 1469 IPFW_DYN_UNLOCK(); 1470 return (0); 1471 } 1472 1473 if (V_dyn_count >= V_dyn_max) 1474 /* Run out of slots, try to remove any expired rule. */ 1475 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1476 1477 if (V_dyn_count >= V_dyn_max) { 1478 if (last_log != time_uptime) { 1479 last_log = time_uptime; 1480 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1481 } 1482 IPFW_DYN_UNLOCK(); 1483 return (1); /* cannot install, notify caller */ 1484 } 1485 1486 switch (cmd->o.opcode) { 1487 case O_KEEP_STATE: /* bidir rule */ 1488 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1489 break; 1490 1491 case O_LIMIT: { /* limit number of sessions */ 1492 struct ipfw_flow_id id; 1493 ipfw_dyn_rule *parent; 1494 uint32_t conn_limit; 1495 uint16_t limit_mask = cmd->limit_mask; 1496 1497 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1498 tablearg : cmd->conn_limit; 1499 1500 DEB( 1501 if (cmd->conn_limit == IP_FW_TABLEARG) 1502 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1503 "(tablearg)\n", __func__, conn_limit); 1504 else 1505 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1506 __func__, conn_limit); 1507 ) 1508 1509 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1510 id.proto = args->f_id.proto; 1511 id.addr_type = args->f_id.addr_type; 1512 id.fib = M_GETFIB(args->m); 1513 1514 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1515 if (limit_mask & DYN_SRC_ADDR) 1516 id.src_ip6 = args->f_id.src_ip6; 1517 if (limit_mask & DYN_DST_ADDR) 1518 id.dst_ip6 = args->f_id.dst_ip6; 1519 } else { 1520 if (limit_mask & DYN_SRC_ADDR) 1521 id.src_ip = args->f_id.src_ip; 1522 if (limit_mask & DYN_DST_ADDR) 1523 id.dst_ip = args->f_id.dst_ip; 1524 } 1525 if (limit_mask & DYN_SRC_PORT) 1526 id.src_port = args->f_id.src_port; 1527 if (limit_mask & DYN_DST_PORT) 1528 id.dst_port = args->f_id.dst_port; 1529 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1530 printf("ipfw: %s: add parent failed\n", __func__); 1531 IPFW_DYN_UNLOCK(); 1532 return (1); 1533 } 1534 1535 if (parent->count >= conn_limit) { 1536 /* See if we can remove some expired rule. */ 1537 remove_dyn_rule(rule, parent); 1538 if (parent->count >= conn_limit) { 1539 if (V_fw_verbose && last_log != time_uptime) { 1540 last_log = time_uptime; 1541#ifdef INET6 1542 /* 1543 * XXX IPv6 flows are not 1544 * supported yet. 1545 */ 1546 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1547 char ip6buf[INET6_ADDRSTRLEN]; 1548 snprintf(src, sizeof(src), 1549 "[%s]", ip6_sprintf(ip6buf, 1550 &args->f_id.src_ip6)); 1551 snprintf(dst, sizeof(dst), 1552 "[%s]", ip6_sprintf(ip6buf, 1553 &args->f_id.dst_ip6)); 1554 } else 1555#endif 1556 { 1557 da.s_addr = 1558 htonl(args->f_id.src_ip); 1559 inet_ntoa_r(da, src); 1560 da.s_addr = 1561 htonl(args->f_id.dst_ip); 1562 inet_ntoa_r(da, dst); 1563 } 1564 log(LOG_SECURITY | LOG_DEBUG, 1565 "ipfw: %d %s %s:%u -> %s:%u, %s\n", 1566 parent->rule->rulenum, 1567 "drop session", 1568 src, (args->f_id.src_port), 1569 dst, (args->f_id.dst_port), 1570 "too many entries"); 1571 } 1572 IPFW_DYN_UNLOCK(); 1573 return (1); 1574 } 1575 } 1576 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1577 break; 1578 } 1579 default: 1580 printf("ipfw: %s: unknown dynamic rule type %u\n", 1581 __func__, cmd->o.opcode); 1582 IPFW_DYN_UNLOCK(); 1583 return (1); 1584 } 1585 1586 /* XXX just set lifetime */ 1587 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1588 1589 IPFW_DYN_UNLOCK(); 1590 return (0); 1591} 1592 1593/* 1594 * Generate a TCP packet, containing either a RST or a keepalive. 1595 * When flags & TH_RST, we are sending a RST packet, because of a 1596 * "reset" action matched the packet. 1597 * Otherwise we are sending a keepalive, and flags & TH_ 1598 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1599 * so that MAC can label the reply appropriately. 1600 */ 1601static struct mbuf * 1602send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1603 u_int32_t ack, int flags) 1604{ 1605 struct mbuf *m; 1606 struct ip *ip; 1607 struct tcphdr *tcp; 1608 1609 MGETHDR(m, M_DONTWAIT, MT_DATA); 1610 if (m == 0) 1611 return (NULL); 1612 m->m_pkthdr.rcvif = (struct ifnet *)0; 1613 1614 M_SETFIB(m, id->fib); 1615#ifdef MAC 1616 if (replyto != NULL) 1617 mac_netinet_firewall_reply(replyto, m); 1618 else 1619 mac_netinet_firewall_send(m); 1620#else 1621 (void)replyto; /* don't warn about unused arg */ 1622#endif 1623 1624 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1625 m->m_data += max_linkhdr; 1626 1627 ip = mtod(m, struct ip *); 1628 bzero(ip, m->m_len); 1629 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1630 ip->ip_p = IPPROTO_TCP; 1631 tcp->th_off = 5; 1632 /* 1633 * Assume we are sending a RST (or a keepalive in the reverse 1634 * direction), swap src and destination addresses and ports. 1635 */ 1636 ip->ip_src.s_addr = htonl(id->dst_ip); 1637 ip->ip_dst.s_addr = htonl(id->src_ip); 1638 tcp->th_sport = htons(id->dst_port); 1639 tcp->th_dport = htons(id->src_port); 1640 if (flags & TH_RST) { /* we are sending a RST */ 1641 if (flags & TH_ACK) { 1642 tcp->th_seq = htonl(ack); 1643 tcp->th_ack = htonl(0); 1644 tcp->th_flags = TH_RST; 1645 } else { 1646 if (flags & TH_SYN) 1647 seq++; 1648 tcp->th_seq = htonl(0); 1649 tcp->th_ack = htonl(seq); 1650 tcp->th_flags = TH_RST | TH_ACK; 1651 } 1652 } else { 1653 /* 1654 * We are sending a keepalive. flags & TH_SYN determines 1655 * the direction, forward if set, reverse if clear. 1656 * NOTE: seq and ack are always assumed to be correct 1657 * as set by the caller. This may be confusing... 1658 */ 1659 if (flags & TH_SYN) { 1660 /* 1661 * we have to rewrite the correct addresses! 1662 */ 1663 ip->ip_dst.s_addr = htonl(id->dst_ip); 1664 ip->ip_src.s_addr = htonl(id->src_ip); 1665 tcp->th_dport = htons(id->dst_port); 1666 tcp->th_sport = htons(id->src_port); 1667 } 1668 tcp->th_seq = htonl(seq); 1669 tcp->th_ack = htonl(ack); 1670 tcp->th_flags = TH_ACK; 1671 } 1672 /* 1673 * set ip_len to the payload size so we can compute 1674 * the tcp checksum on the pseudoheader 1675 * XXX check this, could save a couple of words ? 1676 */ 1677 ip->ip_len = htons(sizeof(struct tcphdr)); 1678 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1679 /* 1680 * now fill fields left out earlier 1681 */ 1682 ip->ip_ttl = V_ip_defttl; 1683 ip->ip_len = m->m_pkthdr.len; 1684 m->m_flags |= M_SKIP_FIREWALL; 1685 return (m); 1686} 1687 1688/* 1689 * sends a reject message, consuming the mbuf passed as an argument. 1690 */ 1691static void 1692send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1693{ 1694 1695#if 0 1696 /* XXX When ip is not guaranteed to be at mtod() we will 1697 * need to account for this */ 1698 * The mbuf will however be thrown away so we can adjust it. 1699 * Remember we did an m_pullup on it already so we 1700 * can make some assumptions about contiguousness. 1701 */ 1702 if (args->L3offset) 1703 m_adj(m, args->L3offset); 1704#endif 1705 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1706 /* We need the IP header in host order for icmp_error(). */ 1707 if (args->eh != NULL) { 1708 ip->ip_len = ntohs(ip->ip_len); 1709 ip->ip_off = ntohs(ip->ip_off); 1710 } 1711 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1712 } else if (args->f_id.proto == IPPROTO_TCP) { 1713 struct tcphdr *const tcp = 1714 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1715 if ( (tcp->th_flags & TH_RST) == 0) { 1716 struct mbuf *m; 1717 m = send_pkt(args->m, &(args->f_id), 1718 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1719 tcp->th_flags | TH_RST); 1720 if (m != NULL) 1721 ip_output(m, NULL, NULL, 0, NULL, NULL); 1722 } 1723 m_freem(args->m); 1724 } else 1725 m_freem(args->m); 1726 args->m = NULL; 1727} 1728 1729/** 1730 * 1731 * Given an ip_fw *, lookup_next_rule will return a pointer 1732 * to the next rule, which can be either the jump 1733 * target (for skipto instructions) or the next one in the list (in 1734 * all other cases including a missing jump target). 1735 * The result is also written in the "next_rule" field of the rule. 1736 * Backward jumps are not allowed, so start looking from the next 1737 * rule... 1738 * 1739 * This never returns NULL -- in case we do not have an exact match, 1740 * the next rule is returned. When the ruleset is changed, 1741 * pointers are flushed so we are always correct. 1742 */ 1743 1744static struct ip_fw * 1745lookup_next_rule(struct ip_fw *me, u_int32_t tablearg) 1746{ 1747 struct ip_fw *rule = NULL; 1748 ipfw_insn *cmd; 1749 u_int16_t rulenum; 1750 1751 /* look for action, in case it is a skipto */ 1752 cmd = ACTION_PTR(me); 1753 if (cmd->opcode == O_LOG) 1754 cmd += F_LEN(cmd); 1755 if (cmd->opcode == O_ALTQ) 1756 cmd += F_LEN(cmd); 1757 if (cmd->opcode == O_TAG) 1758 cmd += F_LEN(cmd); 1759 if (cmd->opcode == O_SKIPTO ) { 1760 if (tablearg != 0) { 1761 rulenum = (u_int16_t)tablearg; 1762 } else { 1763 rulenum = cmd->arg1; 1764 } 1765 for (rule = me->next; rule ; rule = rule->next) { 1766 if (rule->rulenum >= rulenum) { 1767 break; 1768 } 1769 } 1770 } 1771 if (rule == NULL) /* failure or not a skipto */ 1772 rule = me->next; 1773 me->next_rule = rule; 1774 return rule; 1775} 1776 1777static int 1778add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1779 uint8_t mlen, uint32_t value) 1780{ 1781 struct radix_node_head *rnh; 1782 struct table_entry *ent; 1783 1784 if (tbl >= IPFW_TABLES_MAX) 1785 return (EINVAL); 1786 rnh = ch->tables[tbl]; 1787 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1788 if (ent == NULL) 1789 return (ENOMEM); 1790 ent->value = value; 1791 ent->addr.sin_len = ent->mask.sin_len = 8; 1792 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1793 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1794 IPFW_WLOCK(&V_layer3_chain); 1795 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1796 NULL) { 1797 IPFW_WUNLOCK(&V_layer3_chain); 1798 free(ent, M_IPFW_TBL); 1799 return (EEXIST); 1800 } 1801 IPFW_WUNLOCK(&V_layer3_chain); 1802 return (0); 1803} 1804 1805static int 1806del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1807 uint8_t mlen) 1808{ 1809 struct radix_node_head *rnh; 1810 struct table_entry *ent; 1811 struct sockaddr_in sa, mask; 1812 1813 if (tbl >= IPFW_TABLES_MAX) 1814 return (EINVAL); 1815 rnh = ch->tables[tbl]; 1816 sa.sin_len = mask.sin_len = 8; 1817 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1818 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1819 IPFW_WLOCK(ch); 1820 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1821 if (ent == NULL) { 1822 IPFW_WUNLOCK(ch); 1823 return (ESRCH); 1824 } 1825 IPFW_WUNLOCK(ch); 1826 free(ent, M_IPFW_TBL); 1827 return (0); 1828} 1829 1830static int 1831flush_table_entry(struct radix_node *rn, void *arg) 1832{ 1833 struct radix_node_head * const rnh = arg; 1834 struct table_entry *ent; 1835 1836 ent = (struct table_entry *) 1837 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1838 if (ent != NULL) 1839 free(ent, M_IPFW_TBL); 1840 return (0); 1841} 1842 1843static int 1844flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1845{ 1846 struct radix_node_head *rnh; 1847 1848 IPFW_WLOCK_ASSERT(ch); 1849 1850 if (tbl >= IPFW_TABLES_MAX) 1851 return (EINVAL); 1852 rnh = ch->tables[tbl]; 1853 KASSERT(rnh != NULL, ("NULL IPFW table")); 1854 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1855 return (0); 1856} 1857 1858static void 1859flush_tables(struct ip_fw_chain *ch) 1860{ 1861 uint16_t tbl; 1862 1863 IPFW_WLOCK_ASSERT(ch); 1864 1865 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1866 flush_table(ch, tbl); 1867} 1868 1869static int 1870init_tables(struct ip_fw_chain *ch) 1871{ 1872 int i; 1873 uint16_t j; 1874 1875 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1876 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1877 for (j = 0; j < i; j++) { 1878 (void) flush_table(ch, j); 1879 } 1880 return (ENOMEM); 1881 } 1882 } 1883 return (0); 1884} 1885 1886static int 1887lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1888 uint32_t *val) 1889{ 1890 struct radix_node_head *rnh; 1891 struct table_entry *ent; 1892 struct sockaddr_in sa; 1893 1894 if (tbl >= IPFW_TABLES_MAX) 1895 return (0); 1896 rnh = ch->tables[tbl]; 1897 sa.sin_len = 8; 1898 sa.sin_addr.s_addr = addr; 1899 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1900 if (ent != NULL) { 1901 *val = ent->value; 1902 return (1); 1903 } 1904 return (0); 1905} 1906 1907static int 1908count_table_entry(struct radix_node *rn, void *arg) 1909{ 1910 u_int32_t * const cnt = arg; 1911 1912 (*cnt)++; 1913 return (0); 1914} 1915 1916static int 1917count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1918{ 1919 struct radix_node_head *rnh; 1920 1921 if (tbl >= IPFW_TABLES_MAX) 1922 return (EINVAL); 1923 rnh = ch->tables[tbl]; 1924 *cnt = 0; 1925 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1926 return (0); 1927} 1928 1929static int 1930dump_table_entry(struct radix_node *rn, void *arg) 1931{ 1932 struct table_entry * const n = (struct table_entry *)rn; 1933 ipfw_table * const tbl = arg; 1934 ipfw_table_entry *ent; 1935 1936 if (tbl->cnt == tbl->size) 1937 return (1); 1938 ent = &tbl->ent[tbl->cnt]; 1939 ent->tbl = tbl->tbl; 1940 if (in_nullhost(n->mask.sin_addr)) 1941 ent->masklen = 0; 1942 else 1943 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1944 ent->addr = n->addr.sin_addr.s_addr; 1945 ent->value = n->value; 1946 tbl->cnt++; 1947 return (0); 1948} 1949 1950static int 1951dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1952{ 1953 struct radix_node_head *rnh; 1954 1955 if (tbl->tbl >= IPFW_TABLES_MAX) 1956 return (EINVAL); 1957 rnh = ch->tables[tbl->tbl]; 1958 tbl->cnt = 0; 1959 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1960 return (0); 1961} 1962 1963static void 1964fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1965{ 1966 struct ucred *cr; 1967 1968 if (inp->inp_socket != NULL) { 1969 cr = inp->inp_socket->so_cred; 1970 ugp->fw_prid = jailed(cr) ? 1971 cr->cr_prison->pr_id : -1; 1972 ugp->fw_uid = cr->cr_uid; 1973 ugp->fw_ngroups = cr->cr_ngroups; 1974 bcopy(cr->cr_groups, ugp->fw_groups, 1975 sizeof(ugp->fw_groups)); 1976 } 1977} 1978 1979static int 1980check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 1981 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 1982 u_int16_t src_port, struct ip_fw_ugid *ugp, int *lookup, 1983 struct inpcb *inp) 1984{ 1985 struct inpcbinfo *pi; 1986 int wildcard; 1987 struct inpcb *pcb; 1988 int match; 1989 gid_t *gp; 1990 1991 /* 1992 * Check to see if the UDP or TCP stack supplied us with 1993 * the PCB. If so, rather then holding a lock and looking 1994 * up the PCB, we can use the one that was supplied. 1995 */ 1996 if (inp && *lookup == 0) { 1997 INP_LOCK_ASSERT(inp); 1998 if (inp->inp_socket != NULL) { 1999 fill_ugid_cache(inp, ugp); 2000 *lookup = 1; 2001 } 2002 } 2003 /* 2004 * If we have already been here and the packet has no 2005 * PCB entry associated with it, then we can safely 2006 * assume that this is a no match. 2007 */ 2008 if (*lookup == -1) 2009 return (0); 2010 if (proto == IPPROTO_TCP) { 2011 wildcard = 0; 2012 pi = &V_tcbinfo; 2013 } else if (proto == IPPROTO_UDP) { 2014 wildcard = INPLOOKUP_WILDCARD; 2015 pi = &V_udbinfo; 2016 } else 2017 return 0; 2018 match = 0; 2019 if (*lookup == 0) { 2020 INP_INFO_RLOCK(pi); 2021 pcb = (oif) ? 2022 in_pcblookup_hash(pi, 2023 dst_ip, htons(dst_port), 2024 src_ip, htons(src_port), 2025 wildcard, oif) : 2026 in_pcblookup_hash(pi, 2027 src_ip, htons(src_port), 2028 dst_ip, htons(dst_port), 2029 wildcard, NULL); 2030 if (pcb != NULL) { 2031 INP_RLOCK(pcb); 2032 if (pcb->inp_socket != NULL) { 2033 fill_ugid_cache(pcb, ugp); 2034 *lookup = 1; 2035 } 2036 INP_RUNLOCK(pcb); 2037 } 2038 INP_INFO_RUNLOCK(pi); 2039 if (*lookup == 0) { 2040 /* 2041 * If the lookup did not yield any results, there 2042 * is no sense in coming back and trying again. So 2043 * we can set lookup to -1 and ensure that we wont 2044 * bother the pcb system again. 2045 */ 2046 *lookup = -1; 2047 return (0); 2048 } 2049 } 2050 if (insn->o.opcode == O_UID) 2051 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2052 else if (insn->o.opcode == O_GID) { 2053 for (gp = ugp->fw_groups; 2054 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2055 if (*gp == (gid_t)insn->d[0]) { 2056 match = 1; 2057 break; 2058 } 2059 } else if (insn->o.opcode == O_JAIL) 2060 match = (ugp->fw_prid == (int)insn->d[0]); 2061 return match; 2062} 2063 2064/* 2065 * The main check routine for the firewall. 2066 * 2067 * All arguments are in args so we can modify them and return them 2068 * back to the caller. 2069 * 2070 * Parameters: 2071 * 2072 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2073 * Starts with the IP header. 2074 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2075 * args->L3offset Number of bytes bypassed if we came from L2. 2076 * e.g. often sizeof(eh) ** NOTYET ** 2077 * args->oif Outgoing interface, or NULL if packet is incoming. 2078 * The incoming interface is in the mbuf. (in) 2079 * args->divert_rule (in/out) 2080 * Skip up to the first rule past this rule number; 2081 * upon return, non-zero port number for divert or tee. 2082 * 2083 * args->rule Pointer to the last matching rule (in/out) 2084 * args->next_hop Socket we are forwarding to (out). 2085 * args->f_id Addresses grabbed from the packet (out) 2086 * args->cookie a cookie depending on rule action 2087 * 2088 * Return value: 2089 * 2090 * IP_FW_PASS the packet must be accepted 2091 * IP_FW_DENY the packet must be dropped 2092 * IP_FW_DIVERT divert packet, port in m_tag 2093 * IP_FW_TEE tee packet, port in m_tag 2094 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2095 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2096 * 2097 */ 2098int 2099ipfw_chk(struct ip_fw_args *args) 2100{ 2101 /* 2102 * Local variables holding state during the processing of a packet: 2103 * 2104 * IMPORTANT NOTE: to speed up the processing of rules, there 2105 * are some assumption on the values of the variables, which 2106 * are documented here. Should you change them, please check 2107 * the implementation of the various instructions to make sure 2108 * that they still work. 2109 * 2110 * args->eh The MAC header. It is non-null for a layer2 2111 * packet, it is NULL for a layer-3 packet. 2112 * **notyet** 2113 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2114 * 2115 * m | args->m Pointer to the mbuf, as received from the caller. 2116 * It may change if ipfw_chk() does an m_pullup, or if it 2117 * consumes the packet because it calls send_reject(). 2118 * XXX This has to change, so that ipfw_chk() never modifies 2119 * or consumes the buffer. 2120 * ip is the beginning of the ip(4 or 6) header. 2121 * Calculated by adding the L3offset to the start of data. 2122 * (Until we start using L3offset, the packet is 2123 * supposed to start with the ip header). 2124 */ 2125 struct mbuf *m = args->m; 2126 struct ip *ip = mtod(m, struct ip *); 2127 2128 /* 2129 * For rules which contain uid/gid or jail constraints, cache 2130 * a copy of the users credentials after the pcb lookup has been 2131 * executed. This will speed up the processing of rules with 2132 * these types of constraints, as well as decrease contention 2133 * on pcb related locks. 2134 */ 2135 struct ip_fw_ugid fw_ugid_cache; 2136 int ugid_lookup = 0; 2137 2138 /* 2139 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2140 * associated with a packet input on a divert socket. This 2141 * will allow to distinguish traffic and its direction when 2142 * it originates from a divert socket. 2143 */ 2144 u_int divinput_flags = 0; 2145 2146 /* 2147 * oif | args->oif If NULL, ipfw_chk has been called on the 2148 * inbound path (ether_input, ip_input). 2149 * If non-NULL, ipfw_chk has been called on the outbound path 2150 * (ether_output, ip_output). 2151 */ 2152 struct ifnet *oif = args->oif; 2153 2154 struct ip_fw *f = NULL; /* matching rule */ 2155 int retval = 0; 2156 2157 /* 2158 * hlen The length of the IP header. 2159 */ 2160 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2161 2162 /* 2163 * offset The offset of a fragment. offset != 0 means that 2164 * we have a fragment at this offset of an IPv4 packet. 2165 * offset == 0 means that (if this is an IPv4 packet) 2166 * this is the first or only fragment. 2167 * For IPv6 offset == 0 means there is no Fragment Header. 2168 * If offset != 0 for IPv6 always use correct mask to 2169 * get the correct offset because we add IP6F_MORE_FRAG 2170 * to be able to dectect the first fragment which would 2171 * otherwise have offset = 0. 2172 */ 2173 u_short offset = 0; 2174 2175 /* 2176 * Local copies of addresses. They are only valid if we have 2177 * an IP packet. 2178 * 2179 * proto The protocol. Set to 0 for non-ip packets, 2180 * or to the protocol read from the packet otherwise. 2181 * proto != 0 means that we have an IPv4 packet. 2182 * 2183 * src_port, dst_port port numbers, in HOST format. Only 2184 * valid for TCP and UDP packets. 2185 * 2186 * src_ip, dst_ip ip addresses, in NETWORK format. 2187 * Only valid for IPv4 packets. 2188 */ 2189 u_int8_t proto; 2190 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2191 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2192 u_int16_t ip_len=0; 2193 int pktlen; 2194 u_int16_t etype = 0; /* Host order stored ether type */ 2195 2196 /* 2197 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2198 * MATCH_NONE when checked and not matched (q = NULL), 2199 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2200 */ 2201 int dyn_dir = MATCH_UNKNOWN; 2202 ipfw_dyn_rule *q = NULL; 2203 struct ip_fw_chain *chain = &V_layer3_chain; 2204 struct m_tag *mtag; 2205 2206 /* 2207 * We store in ulp a pointer to the upper layer protocol header. 2208 * In the ipv4 case this is easy to determine from the header, 2209 * but for ipv6 we might have some additional headers in the middle. 2210 * ulp is NULL if not found. 2211 */ 2212 void *ulp = NULL; /* upper layer protocol pointer. */ 2213 /* XXX ipv6 variables */ 2214 int is_ipv6 = 0; 2215 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2216 /* end of ipv6 variables */ 2217 int is_ipv4 = 0; 2218 2219 if (m->m_flags & M_SKIP_FIREWALL) 2220 return (IP_FW_PASS); /* accept */ 2221 2222 pktlen = m->m_pkthdr.len; 2223 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 2224 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2225 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2226 2227/* 2228 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2229 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2230 * pointer might become stale after other pullups (but we never use it 2231 * this way). 2232 */ 2233#define PULLUP_TO(len, p, T) \ 2234do { \ 2235 int x = (len) + sizeof(T); \ 2236 if ((m)->m_len < x) { \ 2237 args->m = m = m_pullup(m, x); \ 2238 if (m == NULL) \ 2239 goto pullup_failed; \ 2240 } \ 2241 p = (mtod(m, char *) + (len)); \ 2242} while (0) 2243 2244 /* 2245 * if we have an ether header, 2246 */ 2247 if (args->eh) 2248 etype = ntohs(args->eh->ether_type); 2249 2250 /* Identify IP packets and fill up variables. */ 2251 if (pktlen >= sizeof(struct ip6_hdr) && 2252 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2253 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2254 is_ipv6 = 1; 2255 args->f_id.addr_type = 6; 2256 hlen = sizeof(struct ip6_hdr); 2257 proto = ip6->ip6_nxt; 2258 2259 /* Search extension headers to find upper layer protocols */ 2260 while (ulp == NULL) { 2261 switch (proto) { 2262 case IPPROTO_ICMPV6: 2263 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2264 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2265 break; 2266 2267 case IPPROTO_TCP: 2268 PULLUP_TO(hlen, ulp, struct tcphdr); 2269 dst_port = TCP(ulp)->th_dport; 2270 src_port = TCP(ulp)->th_sport; 2271 args->f_id.flags = TCP(ulp)->th_flags; 2272 break; 2273 2274 case IPPROTO_SCTP: 2275 PULLUP_TO(hlen, ulp, struct sctphdr); 2276 src_port = SCTP(ulp)->src_port; 2277 dst_port = SCTP(ulp)->dest_port; 2278 break; 2279 2280 case IPPROTO_UDP: 2281 PULLUP_TO(hlen, ulp, struct udphdr); 2282 dst_port = UDP(ulp)->uh_dport; 2283 src_port = UDP(ulp)->uh_sport; 2284 break; 2285 2286 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2287 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2288 ext_hd |= EXT_HOPOPTS; 2289 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2290 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2291 ulp = NULL; 2292 break; 2293 2294 case IPPROTO_ROUTING: /* RFC 2460 */ 2295 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2296 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2297 case 0: 2298 ext_hd |= EXT_RTHDR0; 2299 break; 2300 case 2: 2301 ext_hd |= EXT_RTHDR2; 2302 break; 2303 default: 2304 printf("IPFW2: IPV6 - Unknown Routing " 2305 "Header type(%d)\n", 2306 ((struct ip6_rthdr *)ulp)->ip6r_type); 2307 if (V_fw_deny_unknown_exthdrs) 2308 return (IP_FW_DENY); 2309 break; 2310 } 2311 ext_hd |= EXT_ROUTING; 2312 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2313 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2314 ulp = NULL; 2315 break; 2316 2317 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2318 PULLUP_TO(hlen, ulp, struct ip6_frag); 2319 ext_hd |= EXT_FRAGMENT; 2320 hlen += sizeof (struct ip6_frag); 2321 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2322 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2323 IP6F_OFF_MASK; 2324 /* Add IP6F_MORE_FRAG for offset of first 2325 * fragment to be != 0. */ 2326 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2327 IP6F_MORE_FRAG; 2328 if (offset == 0) { 2329 printf("IPFW2: IPV6 - Invalid Fragment " 2330 "Header\n"); 2331 if (V_fw_deny_unknown_exthdrs) 2332 return (IP_FW_DENY); 2333 break; 2334 } 2335 args->f_id.frag_id6 = 2336 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2337 ulp = NULL; 2338 break; 2339 2340 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2341 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2342 ext_hd |= EXT_DSTOPTS; 2343 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2344 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2345 ulp = NULL; 2346 break; 2347 2348 case IPPROTO_AH: /* RFC 2402 */ 2349 PULLUP_TO(hlen, ulp, struct ip6_ext); 2350 ext_hd |= EXT_AH; 2351 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2352 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2353 ulp = NULL; 2354 break; 2355 2356 case IPPROTO_ESP: /* RFC 2406 */ 2357 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2358 /* Anything past Seq# is variable length and 2359 * data past this ext. header is encrypted. */ 2360 ext_hd |= EXT_ESP; 2361 break; 2362 2363 case IPPROTO_NONE: /* RFC 2460 */ 2364 /* 2365 * Packet ends here, and IPv6 header has 2366 * already been pulled up. If ip6e_len!=0 2367 * then octets must be ignored. 2368 */ 2369 ulp = ip; /* non-NULL to get out of loop. */ 2370 break; 2371 2372 case IPPROTO_OSPFIGP: 2373 /* XXX OSPF header check? */ 2374 PULLUP_TO(hlen, ulp, struct ip6_ext); 2375 break; 2376 2377 case IPPROTO_PIM: 2378 /* XXX PIM header check? */ 2379 PULLUP_TO(hlen, ulp, struct pim); 2380 break; 2381 2382 case IPPROTO_CARP: 2383 PULLUP_TO(hlen, ulp, struct carp_header); 2384 if (((struct carp_header *)ulp)->carp_version != 2385 CARP_VERSION) 2386 return (IP_FW_DENY); 2387 if (((struct carp_header *)ulp)->carp_type != 2388 CARP_ADVERTISEMENT) 2389 return (IP_FW_DENY); 2390 break; 2391 2392 case IPPROTO_IPV6: /* RFC 2893 */ 2393 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2394 break; 2395 2396 case IPPROTO_IPV4: /* RFC 2893 */ 2397 PULLUP_TO(hlen, ulp, struct ip); 2398 break; 2399 2400 default: 2401 printf("IPFW2: IPV6 - Unknown Extension " 2402 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2403 if (V_fw_deny_unknown_exthdrs) 2404 return (IP_FW_DENY); 2405 PULLUP_TO(hlen, ulp, struct ip6_ext); 2406 break; 2407 } /*switch */ 2408 } 2409 ip = mtod(m, struct ip *); 2410 ip6 = (struct ip6_hdr *)ip; 2411 args->f_id.src_ip6 = ip6->ip6_src; 2412 args->f_id.dst_ip6 = ip6->ip6_dst; 2413 args->f_id.src_ip = 0; 2414 args->f_id.dst_ip = 0; 2415 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2416 } else if (pktlen >= sizeof(struct ip) && 2417 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2418 is_ipv4 = 1; 2419 hlen = ip->ip_hl << 2; 2420 args->f_id.addr_type = 4; 2421 2422 /* 2423 * Collect parameters into local variables for faster matching. 2424 */ 2425 proto = ip->ip_p; 2426 src_ip = ip->ip_src; 2427 dst_ip = ip->ip_dst; 2428 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2429 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2430 ip_len = ntohs(ip->ip_len); 2431 } else { 2432 offset = ip->ip_off & IP_OFFMASK; 2433 ip_len = ip->ip_len; 2434 } 2435 pktlen = ip_len < pktlen ? ip_len : pktlen; 2436 2437 if (offset == 0) { 2438 switch (proto) { 2439 case IPPROTO_TCP: 2440 PULLUP_TO(hlen, ulp, struct tcphdr); 2441 dst_port = TCP(ulp)->th_dport; 2442 src_port = TCP(ulp)->th_sport; 2443 args->f_id.flags = TCP(ulp)->th_flags; 2444 break; 2445 2446 case IPPROTO_UDP: 2447 PULLUP_TO(hlen, ulp, struct udphdr); 2448 dst_port = UDP(ulp)->uh_dport; 2449 src_port = UDP(ulp)->uh_sport; 2450 break; 2451 2452 case IPPROTO_ICMP: 2453 PULLUP_TO(hlen, ulp, struct icmphdr); 2454 args->f_id.flags = ICMP(ulp)->icmp_type; 2455 break; 2456 2457 default: 2458 break; 2459 } 2460 } 2461 2462 ip = mtod(m, struct ip *); 2463 args->f_id.src_ip = ntohl(src_ip.s_addr); 2464 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2465 } 2466#undef PULLUP_TO 2467 if (proto) { /* we may have port numbers, store them */ 2468 args->f_id.proto = proto; 2469 args->f_id.src_port = src_port = ntohs(src_port); 2470 args->f_id.dst_port = dst_port = ntohs(dst_port); 2471 } 2472 2473 IPFW_RLOCK(chain); 2474 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2475 if (args->rule) { 2476 /* 2477 * Packet has already been tagged. Look for the next rule 2478 * to restart processing. 2479 * 2480 * If fw_one_pass != 0 then just accept it. 2481 * XXX should not happen here, but optimized out in 2482 * the caller. 2483 */ 2484 if (V_fw_one_pass) { 2485 IPFW_RUNLOCK(chain); 2486 return (IP_FW_PASS); 2487 } 2488 2489 f = args->rule->next_rule; 2490 if (f == NULL) 2491 f = lookup_next_rule(args->rule, 0); 2492 } else { 2493 /* 2494 * Find the starting rule. It can be either the first 2495 * one, or the one after divert_rule if asked so. 2496 */ 2497 int skipto = mtag ? divert_cookie(mtag) : 0; 2498 2499 f = chain->rules; 2500 if (args->eh == NULL && skipto != 0) { 2501 if (skipto >= IPFW_DEFAULT_RULE) { 2502 IPFW_RUNLOCK(chain); 2503 return (IP_FW_DENY); /* invalid */ 2504 } 2505 while (f && f->rulenum <= skipto) 2506 f = f->next; 2507 if (f == NULL) { /* drop packet */ 2508 IPFW_RUNLOCK(chain); 2509 return (IP_FW_DENY); 2510 } 2511 } 2512 } 2513 /* reset divert rule to avoid confusion later */ 2514 if (mtag) { 2515 divinput_flags = divert_info(mtag) & 2516 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2517 m_tag_delete(m, mtag); 2518 } 2519 2520 /* 2521 * Now scan the rules, and parse microinstructions for each rule. 2522 */ 2523 for (; f; f = f->next) { 2524 ipfw_insn *cmd; 2525 uint32_t tablearg = 0; 2526 int l, cmdlen, skip_or; /* skip rest of OR block */ 2527 2528again: 2529 if (V_set_disable & (1 << f->set) ) 2530 continue; 2531 2532 skip_or = 0; 2533 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2534 l -= cmdlen, cmd += cmdlen) { 2535 int match; 2536 2537 /* 2538 * check_body is a jump target used when we find a 2539 * CHECK_STATE, and need to jump to the body of 2540 * the target rule. 2541 */ 2542 2543check_body: 2544 cmdlen = F_LEN(cmd); 2545 /* 2546 * An OR block (insn_1 || .. || insn_n) has the 2547 * F_OR bit set in all but the last instruction. 2548 * The first match will set "skip_or", and cause 2549 * the following instructions to be skipped until 2550 * past the one with the F_OR bit clear. 2551 */ 2552 if (skip_or) { /* skip this instruction */ 2553 if ((cmd->len & F_OR) == 0) 2554 skip_or = 0; /* next one is good */ 2555 continue; 2556 } 2557 match = 0; /* set to 1 if we succeed */ 2558 2559 switch (cmd->opcode) { 2560 /* 2561 * The first set of opcodes compares the packet's 2562 * fields with some pattern, setting 'match' if a 2563 * match is found. At the end of the loop there is 2564 * logic to deal with F_NOT and F_OR flags associated 2565 * with the opcode. 2566 */ 2567 case O_NOP: 2568 match = 1; 2569 break; 2570 2571 case O_FORWARD_MAC: 2572 printf("ipfw: opcode %d unimplemented\n", 2573 cmd->opcode); 2574 break; 2575 2576 case O_GID: 2577 case O_UID: 2578 case O_JAIL: 2579 /* 2580 * We only check offset == 0 && proto != 0, 2581 * as this ensures that we have a 2582 * packet with the ports info. 2583 */ 2584 if (offset!=0) 2585 break; 2586 if (is_ipv6) /* XXX to be fixed later */ 2587 break; 2588 if (proto == IPPROTO_TCP || 2589 proto == IPPROTO_UDP) 2590 match = check_uidgid( 2591 (ipfw_insn_u32 *)cmd, 2592 proto, oif, 2593 dst_ip, dst_port, 2594 src_ip, src_port, &fw_ugid_cache, 2595 &ugid_lookup, args->inp); 2596 break; 2597 2598 case O_RECV: 2599 match = iface_match(m->m_pkthdr.rcvif, 2600 (ipfw_insn_if *)cmd); 2601 break; 2602 2603 case O_XMIT: 2604 match = iface_match(oif, (ipfw_insn_if *)cmd); 2605 break; 2606 2607 case O_VIA: 2608 match = iface_match(oif ? oif : 2609 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2610 break; 2611 2612 case O_MACADDR2: 2613 if (args->eh != NULL) { /* have MAC header */ 2614 u_int32_t *want = (u_int32_t *) 2615 ((ipfw_insn_mac *)cmd)->addr; 2616 u_int32_t *mask = (u_int32_t *) 2617 ((ipfw_insn_mac *)cmd)->mask; 2618 u_int32_t *hdr = (u_int32_t *)args->eh; 2619 2620 match = 2621 ( want[0] == (hdr[0] & mask[0]) && 2622 want[1] == (hdr[1] & mask[1]) && 2623 want[2] == (hdr[2] & mask[2]) ); 2624 } 2625 break; 2626 2627 case O_MAC_TYPE: 2628 if (args->eh != NULL) { 2629 u_int16_t *p = 2630 ((ipfw_insn_u16 *)cmd)->ports; 2631 int i; 2632 2633 for (i = cmdlen - 1; !match && i>0; 2634 i--, p += 2) 2635 match = (etype >= p[0] && 2636 etype <= p[1]); 2637 } 2638 break; 2639 2640 case O_FRAG: 2641 match = (offset != 0); 2642 break; 2643 2644 case O_IN: /* "out" is "not in" */ 2645 match = (oif == NULL); 2646 break; 2647 2648 case O_LAYER2: 2649 match = (args->eh != NULL); 2650 break; 2651 2652 case O_DIVERTED: 2653 match = (cmd->arg1 & 1 && divinput_flags & 2654 IP_FW_DIVERT_LOOPBACK_FLAG) || 2655 (cmd->arg1 & 2 && divinput_flags & 2656 IP_FW_DIVERT_OUTPUT_FLAG); 2657 break; 2658 2659 case O_PROTO: 2660 /* 2661 * We do not allow an arg of 0 so the 2662 * check of "proto" only suffices. 2663 */ 2664 match = (proto == cmd->arg1); 2665 break; 2666 2667 case O_IP_SRC: 2668 match = is_ipv4 && 2669 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2670 src_ip.s_addr); 2671 break; 2672 2673 case O_IP_SRC_LOOKUP: 2674 case O_IP_DST_LOOKUP: 2675 if (is_ipv4) { 2676 uint32_t a = 2677 (cmd->opcode == O_IP_DST_LOOKUP) ? 2678 dst_ip.s_addr : src_ip.s_addr; 2679 uint32_t v; 2680 2681 match = lookup_table(chain, cmd->arg1, a, 2682 &v); 2683 if (!match) 2684 break; 2685 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2686 match = 2687 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2688 else 2689 tablearg = v; 2690 } 2691 break; 2692 2693 case O_IP_SRC_MASK: 2694 case O_IP_DST_MASK: 2695 if (is_ipv4) { 2696 uint32_t a = 2697 (cmd->opcode == O_IP_DST_MASK) ? 2698 dst_ip.s_addr : src_ip.s_addr; 2699 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2700 int i = cmdlen-1; 2701 2702 for (; !match && i>0; i-= 2, p+= 2) 2703 match = (p[0] == (a & p[1])); 2704 } 2705 break; 2706 2707 case O_IP_SRC_ME: 2708 if (is_ipv4) { 2709 struct ifnet *tif; 2710 2711 INADDR_TO_IFP(src_ip, tif); 2712 match = (tif != NULL); 2713 } 2714 break; 2715 2716 case O_IP_DST_SET: 2717 case O_IP_SRC_SET: 2718 if (is_ipv4) { 2719 u_int32_t *d = (u_int32_t *)(cmd+1); 2720 u_int32_t addr = 2721 cmd->opcode == O_IP_DST_SET ? 2722 args->f_id.dst_ip : 2723 args->f_id.src_ip; 2724 2725 if (addr < d[0]) 2726 break; 2727 addr -= d[0]; /* subtract base */ 2728 match = (addr < cmd->arg1) && 2729 ( d[ 1 + (addr>>5)] & 2730 (1<<(addr & 0x1f)) ); 2731 } 2732 break; 2733 2734 case O_IP_DST: 2735 match = is_ipv4 && 2736 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2737 dst_ip.s_addr); 2738 break; 2739 2740 case O_IP_DST_ME: 2741 if (is_ipv4) { 2742 struct ifnet *tif; 2743 2744 INADDR_TO_IFP(dst_ip, tif); 2745 match = (tif != NULL); 2746 } 2747 break; 2748 2749 case O_IP_SRCPORT: 2750 case O_IP_DSTPORT: 2751 /* 2752 * offset == 0 && proto != 0 is enough 2753 * to guarantee that we have a 2754 * packet with port info. 2755 */ 2756 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2757 && offset == 0) { 2758 u_int16_t x = 2759 (cmd->opcode == O_IP_SRCPORT) ? 2760 src_port : dst_port ; 2761 u_int16_t *p = 2762 ((ipfw_insn_u16 *)cmd)->ports; 2763 int i; 2764 2765 for (i = cmdlen - 1; !match && i>0; 2766 i--, p += 2) 2767 match = (x>=p[0] && x<=p[1]); 2768 } 2769 break; 2770 2771 case O_ICMPTYPE: 2772 match = (offset == 0 && proto==IPPROTO_ICMP && 2773 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2774 break; 2775 2776#ifdef INET6 2777 case O_ICMP6TYPE: 2778 match = is_ipv6 && offset == 0 && 2779 proto==IPPROTO_ICMPV6 && 2780 icmp6type_match( 2781 ICMP6(ulp)->icmp6_type, 2782 (ipfw_insn_u32 *)cmd); 2783 break; 2784#endif /* INET6 */ 2785 2786 case O_IPOPT: 2787 match = (is_ipv4 && 2788 ipopts_match(ip, cmd) ); 2789 break; 2790 2791 case O_IPVER: 2792 match = (is_ipv4 && 2793 cmd->arg1 == ip->ip_v); 2794 break; 2795 2796 case O_IPID: 2797 case O_IPLEN: 2798 case O_IPTTL: 2799 if (is_ipv4) { /* only for IP packets */ 2800 uint16_t x; 2801 uint16_t *p; 2802 int i; 2803 2804 if (cmd->opcode == O_IPLEN) 2805 x = ip_len; 2806 else if (cmd->opcode == O_IPTTL) 2807 x = ip->ip_ttl; 2808 else /* must be IPID */ 2809 x = ntohs(ip->ip_id); 2810 if (cmdlen == 1) { 2811 match = (cmd->arg1 == x); 2812 break; 2813 } 2814 /* otherwise we have ranges */ 2815 p = ((ipfw_insn_u16 *)cmd)->ports; 2816 i = cmdlen - 1; 2817 for (; !match && i>0; i--, p += 2) 2818 match = (x >= p[0] && x <= p[1]); 2819 } 2820 break; 2821 2822 case O_IPPRECEDENCE: 2823 match = (is_ipv4 && 2824 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2825 break; 2826 2827 case O_IPTOS: 2828 match = (is_ipv4 && 2829 flags_match(cmd, ip->ip_tos)); 2830 break; 2831 2832 case O_TCPDATALEN: 2833 if (proto == IPPROTO_TCP && offset == 0) { 2834 struct tcphdr *tcp; 2835 uint16_t x; 2836 uint16_t *p; 2837 int i; 2838 2839 tcp = TCP(ulp); 2840 x = ip_len - 2841 ((ip->ip_hl + tcp->th_off) << 2); 2842 if (cmdlen == 1) { 2843 match = (cmd->arg1 == x); 2844 break; 2845 } 2846 /* otherwise we have ranges */ 2847 p = ((ipfw_insn_u16 *)cmd)->ports; 2848 i = cmdlen - 1; 2849 for (; !match && i>0; i--, p += 2) 2850 match = (x >= p[0] && x <= p[1]); 2851 } 2852 break; 2853 2854 case O_TCPFLAGS: 2855 match = (proto == IPPROTO_TCP && offset == 0 && 2856 flags_match(cmd, TCP(ulp)->th_flags)); 2857 break; 2858 2859 case O_TCPOPTS: 2860 match = (proto == IPPROTO_TCP && offset == 0 && 2861 tcpopts_match(TCP(ulp), cmd)); 2862 break; 2863 2864 case O_TCPSEQ: 2865 match = (proto == IPPROTO_TCP && offset == 0 && 2866 ((ipfw_insn_u32 *)cmd)->d[0] == 2867 TCP(ulp)->th_seq); 2868 break; 2869 2870 case O_TCPACK: 2871 match = (proto == IPPROTO_TCP && offset == 0 && 2872 ((ipfw_insn_u32 *)cmd)->d[0] == 2873 TCP(ulp)->th_ack); 2874 break; 2875 2876 case O_TCPWIN: 2877 match = (proto == IPPROTO_TCP && offset == 0 && 2878 cmd->arg1 == TCP(ulp)->th_win); 2879 break; 2880 2881 case O_ESTAB: 2882 /* reject packets which have SYN only */ 2883 /* XXX should i also check for TH_ACK ? */ 2884 match = (proto == IPPROTO_TCP && offset == 0 && 2885 (TCP(ulp)->th_flags & 2886 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2887 break; 2888 2889 case O_ALTQ: { 2890 struct pf_mtag *at; 2891 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2892 2893 match = 1; 2894 at = pf_find_mtag(m); 2895 if (at != NULL && at->qid != 0) 2896 break; 2897 at = pf_get_mtag(m); 2898 if (at == NULL) { 2899 /* 2900 * Let the packet fall back to the 2901 * default ALTQ. 2902 */ 2903 break; 2904 } 2905 at->qid = altq->qid; 2906 if (is_ipv4) 2907 at->af = AF_INET; 2908 else 2909 at->af = AF_LINK; 2910 at->hdr = ip; 2911 break; 2912 } 2913 2914 case O_LOG: 2915 if (V_fw_verbose) 2916 ipfw_log(f, hlen, args, m, 2917 oif, offset, tablearg, ip); 2918 match = 1; 2919 break; 2920 2921 case O_PROB: 2922 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2923 break; 2924 2925 case O_VERREVPATH: 2926 /* Outgoing packets automatically pass/match */ 2927 match = ((oif != NULL) || 2928 (m->m_pkthdr.rcvif == NULL) || 2929 ( 2930#ifdef INET6 2931 is_ipv6 ? 2932 verify_path6(&(args->f_id.src_ip6), 2933 m->m_pkthdr.rcvif) : 2934#endif 2935 verify_path(src_ip, m->m_pkthdr.rcvif, 2936 args->f_id.fib))); 2937 break; 2938 2939 case O_VERSRCREACH: 2940 /* Outgoing packets automatically pass/match */ 2941 match = (hlen > 0 && ((oif != NULL) || 2942#ifdef INET6 2943 is_ipv6 ? 2944 verify_path6(&(args->f_id.src_ip6), 2945 NULL) : 2946#endif 2947 verify_path(src_ip, NULL, args->f_id.fib))); 2948 break; 2949 2950 case O_ANTISPOOF: 2951 /* Outgoing packets automatically pass/match */ 2952 if (oif == NULL && hlen > 0 && 2953 ( (is_ipv4 && in_localaddr(src_ip)) 2954#ifdef INET6 2955 || (is_ipv6 && 2956 in6_localaddr(&(args->f_id.src_ip6))) 2957#endif 2958 )) 2959 match = 2960#ifdef INET6 2961 is_ipv6 ? verify_path6( 2962 &(args->f_id.src_ip6), 2963 m->m_pkthdr.rcvif) : 2964#endif 2965 verify_path(src_ip, 2966 m->m_pkthdr.rcvif, 2967 args->f_id.fib); 2968 else 2969 match = 1; 2970 break; 2971 2972 case O_IPSEC: 2973#ifdef IPSEC 2974 match = (m_tag_find(m, 2975 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2976#endif 2977 /* otherwise no match */ 2978 break; 2979 2980#ifdef INET6 2981 case O_IP6_SRC: 2982 match = is_ipv6 && 2983 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2984 &((ipfw_insn_ip6 *)cmd)->addr6); 2985 break; 2986 2987 case O_IP6_DST: 2988 match = is_ipv6 && 2989 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2990 &((ipfw_insn_ip6 *)cmd)->addr6); 2991 break; 2992 case O_IP6_SRC_MASK: 2993 case O_IP6_DST_MASK: 2994 if (is_ipv6) { 2995 int i = cmdlen - 1; 2996 struct in6_addr p; 2997 struct in6_addr *d = 2998 &((ipfw_insn_ip6 *)cmd)->addr6; 2999 3000 for (; !match && i > 0; d += 2, 3001 i -= F_INSN_SIZE(struct in6_addr) 3002 * 2) { 3003 p = (cmd->opcode == 3004 O_IP6_SRC_MASK) ? 3005 args->f_id.src_ip6: 3006 args->f_id.dst_ip6; 3007 APPLY_MASK(&p, &d[1]); 3008 match = 3009 IN6_ARE_ADDR_EQUAL(&d[0], 3010 &p); 3011 } 3012 } 3013 break; 3014 3015 case O_IP6_SRC_ME: 3016 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3017 break; 3018 3019 case O_IP6_DST_ME: 3020 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3021 break; 3022 3023 case O_FLOW6ID: 3024 match = is_ipv6 && 3025 flow6id_match(args->f_id.flow_id6, 3026 (ipfw_insn_u32 *) cmd); 3027 break; 3028 3029 case O_EXT_HDR: 3030 match = is_ipv6 && 3031 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3032 break; 3033 3034 case O_IP6: 3035 match = is_ipv6; 3036 break; 3037#endif 3038 3039 case O_IP4: 3040 match = is_ipv4; 3041 break; 3042 3043 case O_TAG: { 3044 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3045 tablearg : cmd->arg1; 3046 3047 /* Packet is already tagged with this tag? */ 3048 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3049 3050 /* We have `untag' action when F_NOT flag is 3051 * present. And we must remove this mtag from 3052 * mbuf and reset `match' to zero (`match' will 3053 * be inversed later). 3054 * Otherwise we should allocate new mtag and 3055 * push it into mbuf. 3056 */ 3057 if (cmd->len & F_NOT) { /* `untag' action */ 3058 if (mtag != NULL) 3059 m_tag_delete(m, mtag); 3060 } else if (mtag == NULL) { 3061 if ((mtag = m_tag_alloc(MTAG_IPFW, 3062 tag, 0, M_NOWAIT)) != NULL) 3063 m_tag_prepend(m, mtag); 3064 } 3065 match = (cmd->len & F_NOT) ? 0: 1; 3066 break; 3067 } 3068 3069 case O_FIB: /* try match the specified fib */ 3070 if (args->f_id.fib == cmd->arg1) 3071 match = 1; 3072 break; 3073 3074 case O_TAGGED: { 3075 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3076 tablearg : cmd->arg1; 3077 3078 if (cmdlen == 1) { 3079 match = m_tag_locate(m, MTAG_IPFW, 3080 tag, NULL) != NULL; 3081 break; 3082 } 3083 3084 /* we have ranges */ 3085 for (mtag = m_tag_first(m); 3086 mtag != NULL && !match; 3087 mtag = m_tag_next(m, mtag)) { 3088 uint16_t *p; 3089 int i; 3090 3091 if (mtag->m_tag_cookie != MTAG_IPFW) 3092 continue; 3093 3094 p = ((ipfw_insn_u16 *)cmd)->ports; 3095 i = cmdlen - 1; 3096 for(; !match && i > 0; i--, p += 2) 3097 match = 3098 mtag->m_tag_id >= p[0] && 3099 mtag->m_tag_id <= p[1]; 3100 } 3101 break; 3102 } 3103 3104 /* 3105 * The second set of opcodes represents 'actions', 3106 * i.e. the terminal part of a rule once the packet 3107 * matches all previous patterns. 3108 * Typically there is only one action for each rule, 3109 * and the opcode is stored at the end of the rule 3110 * (but there are exceptions -- see below). 3111 * 3112 * In general, here we set retval and terminate the 3113 * outer loop (would be a 'break 3' in some language, 3114 * but we need to do a 'goto done'). 3115 * 3116 * Exceptions: 3117 * O_COUNT and O_SKIPTO actions: 3118 * instead of terminating, we jump to the next rule 3119 * ('goto next_rule', equivalent to a 'break 2'), 3120 * or to the SKIPTO target ('goto again' after 3121 * having set f, cmd and l), respectively. 3122 * 3123 * O_TAG, O_LOG and O_ALTQ action parameters: 3124 * perform some action and set match = 1; 3125 * 3126 * O_LIMIT and O_KEEP_STATE: these opcodes are 3127 * not real 'actions', and are stored right 3128 * before the 'action' part of the rule. 3129 * These opcodes try to install an entry in the 3130 * state tables; if successful, we continue with 3131 * the next opcode (match=1; break;), otherwise 3132 * the packet * must be dropped 3133 * ('goto done' after setting retval); 3134 * 3135 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3136 * cause a lookup of the state table, and a jump 3137 * to the 'action' part of the parent rule 3138 * ('goto check_body') if an entry is found, or 3139 * (CHECK_STATE only) a jump to the next rule if 3140 * the entry is not found ('goto next_rule'). 3141 * The result of the lookup is cached to make 3142 * further instances of these opcodes are 3143 * effectively NOPs. 3144 */ 3145 case O_LIMIT: 3146 case O_KEEP_STATE: 3147 if (install_state(f, 3148 (ipfw_insn_limit *)cmd, args, tablearg)) { 3149 retval = IP_FW_DENY; 3150 goto done; /* error/limit violation */ 3151 } 3152 match = 1; 3153 break; 3154 3155 case O_PROBE_STATE: 3156 case O_CHECK_STATE: 3157 /* 3158 * dynamic rules are checked at the first 3159 * keep-state or check-state occurrence, 3160 * with the result being stored in dyn_dir. 3161 * The compiler introduces a PROBE_STATE 3162 * instruction for us when we have a 3163 * KEEP_STATE (because PROBE_STATE needs 3164 * to be run first). 3165 */ 3166 if (dyn_dir == MATCH_UNKNOWN && 3167 (q = lookup_dyn_rule(&args->f_id, 3168 &dyn_dir, proto == IPPROTO_TCP ? 3169 TCP(ulp) : NULL)) 3170 != NULL) { 3171 /* 3172 * Found dynamic entry, update stats 3173 * and jump to the 'action' part of 3174 * the parent rule. 3175 */ 3176 q->pcnt++; 3177 q->bcnt += pktlen; 3178 f = q->rule; 3179 cmd = ACTION_PTR(f); 3180 l = f->cmd_len - f->act_ofs; 3181 IPFW_DYN_UNLOCK(); 3182 goto check_body; 3183 } 3184 /* 3185 * Dynamic entry not found. If CHECK_STATE, 3186 * skip to next rule, if PROBE_STATE just 3187 * ignore and continue with next opcode. 3188 */ 3189 if (cmd->opcode == O_CHECK_STATE) 3190 goto next_rule; 3191 match = 1; 3192 break; 3193 3194 case O_ACCEPT: 3195 retval = 0; /* accept */ 3196 goto done; 3197 3198 case O_PIPE: 3199 case O_QUEUE: 3200 args->rule = f; /* report matching rule */ 3201 if (cmd->arg1 == IP_FW_TABLEARG) 3202 args->cookie = tablearg; 3203 else 3204 args->cookie = cmd->arg1; 3205 retval = IP_FW_DUMMYNET; 3206 goto done; 3207 3208 case O_DIVERT: 3209 case O_TEE: { 3210 struct divert_tag *dt; 3211 3212 if (args->eh) /* not on layer 2 */ 3213 break; 3214 mtag = m_tag_get(PACKET_TAG_DIVERT, 3215 sizeof(struct divert_tag), 3216 M_NOWAIT); 3217 if (mtag == NULL) { 3218 /* XXX statistic */ 3219 /* drop packet */ 3220 IPFW_RUNLOCK(chain); 3221 return (IP_FW_DENY); 3222 } 3223 dt = (struct divert_tag *)(mtag+1); 3224 dt->cookie = f->rulenum; 3225 if (cmd->arg1 == IP_FW_TABLEARG) 3226 dt->info = tablearg; 3227 else 3228 dt->info = cmd->arg1; 3229 m_tag_prepend(m, mtag); 3230 retval = (cmd->opcode == O_DIVERT) ? 3231 IP_FW_DIVERT : IP_FW_TEE; 3232 goto done; 3233 } 3234 case O_COUNT: 3235 case O_SKIPTO: 3236 f->pcnt++; /* update stats */ 3237 f->bcnt += pktlen; 3238 f->timestamp = time_uptime; 3239 if (cmd->opcode == O_COUNT) 3240 goto next_rule; 3241 /* handle skipto */ 3242 if (cmd->arg1 == IP_FW_TABLEARG) { 3243 f = lookup_next_rule(f, tablearg); 3244 } else { 3245 if (f->next_rule == NULL) 3246 lookup_next_rule(f, 0); 3247 f = f->next_rule; 3248 } 3249 goto again; 3250 3251 case O_REJECT: 3252 /* 3253 * Drop the packet and send a reject notice 3254 * if the packet is not ICMP (or is an ICMP 3255 * query), and it is not multicast/broadcast. 3256 */ 3257 if (hlen > 0 && is_ipv4 && offset == 0 && 3258 (proto != IPPROTO_ICMP || 3259 is_icmp_query(ICMP(ulp))) && 3260 !(m->m_flags & (M_BCAST|M_MCAST)) && 3261 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3262 send_reject(args, cmd->arg1, ip_len, ip); 3263 m = args->m; 3264 } 3265 /* FALLTHROUGH */ 3266#ifdef INET6 3267 case O_UNREACH6: 3268 if (hlen > 0 && is_ipv6 && 3269 ((offset & IP6F_OFF_MASK) == 0) && 3270 (proto != IPPROTO_ICMPV6 || 3271 (is_icmp6_query(args->f_id.flags) == 1)) && 3272 !(m->m_flags & (M_BCAST|M_MCAST)) && 3273 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3274 send_reject6( 3275 args, cmd->arg1, hlen, 3276 (struct ip6_hdr *)ip); 3277 m = args->m; 3278 } 3279 /* FALLTHROUGH */ 3280#endif 3281 case O_DENY: 3282 retval = IP_FW_DENY; 3283 goto done; 3284 3285 case O_FORWARD_IP: { 3286 struct sockaddr_in *sa; 3287 sa = &(((ipfw_insn_sa *)cmd)->sa); 3288 if (args->eh) /* not valid on layer2 pkts */ 3289 break; 3290 if (!q || dyn_dir == MATCH_FORWARD) { 3291 if (sa->sin_addr.s_addr == INADDR_ANY) { 3292 bcopy(sa, &args->hopstore, 3293 sizeof(*sa)); 3294 args->hopstore.sin_addr.s_addr = 3295 htonl(tablearg); 3296 args->next_hop = 3297 &args->hopstore; 3298 } else { 3299 args->next_hop = sa; 3300 } 3301 } 3302 retval = IP_FW_PASS; 3303 } 3304 goto done; 3305 3306 case O_NETGRAPH: 3307 case O_NGTEE: 3308 args->rule = f; /* report matching rule */ 3309 if (cmd->arg1 == IP_FW_TABLEARG) 3310 args->cookie = tablearg; 3311 else 3312 args->cookie = cmd->arg1; 3313 retval = (cmd->opcode == O_NETGRAPH) ? 3314 IP_FW_NETGRAPH : IP_FW_NGTEE; 3315 goto done; 3316 3317 case O_SETFIB: 3318 f->pcnt++; /* update stats */ 3319 f->bcnt += pktlen; 3320 f->timestamp = time_uptime; 3321 M_SETFIB(m, cmd->arg1); 3322 args->f_id.fib = cmd->arg1; 3323 goto next_rule; 3324 3325 case O_NAT: { 3326 struct cfg_nat *t; 3327 int nat_id; 3328 3329 if (IPFW_NAT_LOADED) { 3330 args->rule = f; /* Report matching rule. */ 3331 t = ((ipfw_insn_nat *)cmd)->nat; 3332 if (t == NULL) { 3333 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 3334 tablearg : cmd->arg1; 3335 LOOKUP_NAT(V_layer3_chain, nat_id, t); 3336 if (t == NULL) { 3337 retval = IP_FW_DENY; 3338 goto done; 3339 } 3340 if (cmd->arg1 != IP_FW_TABLEARG) 3341 ((ipfw_insn_nat *)cmd)->nat = t; 3342 } 3343 retval = ipfw_nat_ptr(args, t, m); 3344 } else 3345 retval = IP_FW_DENY; 3346 goto done; 3347 } 3348 3349 default: 3350 panic("-- unknown opcode %d\n", cmd->opcode); 3351 } /* end of switch() on opcodes */ 3352 3353 if (cmd->len & F_NOT) 3354 match = !match; 3355 3356 if (match) { 3357 if (cmd->len & F_OR) 3358 skip_or = 1; 3359 } else { 3360 if (!(cmd->len & F_OR)) /* not an OR block, */ 3361 break; /* try next rule */ 3362 } 3363 3364 } /* end of inner for, scan opcodes */ 3365 3366next_rule:; /* try next rule */ 3367 3368 } /* end of outer for, scan rules */ 3369 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3370 IPFW_RUNLOCK(chain); 3371 return (IP_FW_DENY); 3372 3373done: 3374 /* Update statistics */ 3375 f->pcnt++; 3376 f->bcnt += pktlen; 3377 f->timestamp = time_uptime; 3378 IPFW_RUNLOCK(chain); 3379 return (retval); 3380 3381pullup_failed: 3382 if (V_fw_verbose) 3383 printf("ipfw: pullup failed\n"); 3384 return (IP_FW_DENY); 3385} 3386 3387/* 3388 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3389 * These will be reconstructed on the fly as packets are matched. 3390 */ 3391static void 3392flush_rule_ptrs(struct ip_fw_chain *chain) 3393{ 3394 struct ip_fw *rule; 3395 3396 IPFW_WLOCK_ASSERT(chain); 3397 3398 for (rule = chain->rules; rule; rule = rule->next) 3399 rule->next_rule = NULL; 3400} 3401 3402/* 3403 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3404 * possibly create a rule number and add the rule to the list. 3405 * Update the rule_number in the input struct so the caller knows it as well. 3406 */ 3407static int 3408add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3409{ 3410 struct ip_fw *rule, *f, *prev; 3411 int l = RULESIZE(input_rule); 3412 3413 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3414 return (EINVAL); 3415 3416 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3417 if (rule == NULL) 3418 return (ENOSPC); 3419 3420 bcopy(input_rule, rule, l); 3421 3422 rule->next = NULL; 3423 rule->next_rule = NULL; 3424 3425 rule->pcnt = 0; 3426 rule->bcnt = 0; 3427 rule->timestamp = 0; 3428 3429 IPFW_WLOCK(chain); 3430 3431 if (chain->rules == NULL) { /* default rule */ 3432 chain->rules = rule; 3433 goto done; 3434 } 3435 3436 /* 3437 * If rulenum is 0, find highest numbered rule before the 3438 * default rule, and add autoinc_step 3439 */ 3440 if (V_autoinc_step < 1) 3441 V_autoinc_step = 1; 3442 else if (V_autoinc_step > 1000) 3443 V_autoinc_step = 1000; 3444 if (rule->rulenum == 0) { 3445 /* 3446 * locate the highest numbered rule before default 3447 */ 3448 for (f = chain->rules; f; f = f->next) { 3449 if (f->rulenum == IPFW_DEFAULT_RULE) 3450 break; 3451 rule->rulenum = f->rulenum; 3452 } 3453 if (rule->rulenum < IPFW_DEFAULT_RULE - V_autoinc_step) 3454 rule->rulenum += V_autoinc_step; 3455 input_rule->rulenum = rule->rulenum; 3456 } 3457 3458 /* 3459 * Now insert the new rule in the right place in the sorted list. 3460 */ 3461 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3462 if (f->rulenum > rule->rulenum) { /* found the location */ 3463 if (prev) { 3464 rule->next = f; 3465 prev->next = rule; 3466 } else { /* head insert */ 3467 rule->next = chain->rules; 3468 chain->rules = rule; 3469 } 3470 break; 3471 } 3472 } 3473 flush_rule_ptrs(chain); 3474done: 3475 V_static_count++; 3476 V_static_len += l; 3477 IPFW_WUNLOCK(chain); 3478 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3479 rule->rulenum, V_static_count);) 3480 return (0); 3481} 3482 3483/** 3484 * Remove a static rule (including derived * dynamic rules) 3485 * and place it on the ``reap list'' for later reclamation. 3486 * The caller is in charge of clearing rule pointers to avoid 3487 * dangling pointers. 3488 * @return a pointer to the next entry. 3489 * Arguments are not checked, so they better be correct. 3490 */ 3491static struct ip_fw * 3492remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 3493 struct ip_fw *prev) 3494{ 3495 struct ip_fw *n; 3496 int l = RULESIZE(rule); 3497 3498 IPFW_WLOCK_ASSERT(chain); 3499 3500 n = rule->next; 3501 IPFW_DYN_LOCK(); 3502 remove_dyn_rule(rule, NULL /* force removal */); 3503 IPFW_DYN_UNLOCK(); 3504 if (prev == NULL) 3505 chain->rules = n; 3506 else 3507 prev->next = n; 3508 V_static_count--; 3509 V_static_len -= l; 3510 3511 rule->next = chain->reap; 3512 chain->reap = rule; 3513 3514 return n; 3515} 3516 3517/** 3518 * Reclaim storage associated with a list of rules. This is 3519 * typically the list created using remove_rule. 3520 */ 3521static void 3522reap_rules(struct ip_fw *head) 3523{ 3524 struct ip_fw *rule; 3525 3526 while ((rule = head) != NULL) { 3527 head = head->next; 3528 if (DUMMYNET_LOADED) 3529 ip_dn_ruledel_ptr(rule); 3530 free(rule, M_IPFW); 3531 } 3532} 3533 3534/* 3535 * Remove all rules from a chain (except rules in set RESVD_SET 3536 * unless kill_default = 1). The caller is responsible for 3537 * reclaiming storage for the rules left in chain->reap. 3538 */ 3539static void 3540free_chain(struct ip_fw_chain *chain, int kill_default) 3541{ 3542 struct ip_fw *prev, *rule; 3543 3544 IPFW_WLOCK_ASSERT(chain); 3545 3546 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3547 for (prev = NULL, rule = chain->rules; rule ; ) 3548 if (kill_default || rule->set != RESVD_SET) 3549 rule = remove_rule(chain, rule, prev); 3550 else { 3551 prev = rule; 3552 rule = rule->next; 3553 } 3554} 3555 3556/** 3557 * Remove all rules with given number, and also do set manipulation. 3558 * Assumes chain != NULL && *chain != NULL. 3559 * 3560 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3561 * the next 8 bits are the new set, the top 8 bits are the command: 3562 * 3563 * 0 delete rules with given number 3564 * 1 delete rules with given set number 3565 * 2 move rules with given number to new set 3566 * 3 move rules with given set number to new set 3567 * 4 swap sets with given numbers 3568 * 5 delete rules with given number and with given set number 3569 */ 3570static int 3571del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3572{ 3573 struct ip_fw *prev = NULL, *rule; 3574 u_int16_t rulenum; /* rule or old_set */ 3575 u_int8_t cmd, new_set; 3576 3577 rulenum = arg & 0xffff; 3578 cmd = (arg >> 24) & 0xff; 3579 new_set = (arg >> 16) & 0xff; 3580 3581 if (cmd > 5 || new_set > RESVD_SET) 3582 return EINVAL; 3583 if (cmd == 0 || cmd == 2 || cmd == 5) { 3584 if (rulenum >= IPFW_DEFAULT_RULE) 3585 return EINVAL; 3586 } else { 3587 if (rulenum > RESVD_SET) /* old_set */ 3588 return EINVAL; 3589 } 3590 3591 IPFW_WLOCK(chain); 3592 rule = chain->rules; 3593 chain->reap = NULL; 3594 switch (cmd) { 3595 case 0: /* delete rules with given number */ 3596 /* 3597 * locate first rule to delete 3598 */ 3599 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3600 ; 3601 if (rule->rulenum != rulenum) { 3602 IPFW_WUNLOCK(chain); 3603 return EINVAL; 3604 } 3605 3606 /* 3607 * flush pointers outside the loop, then delete all matching 3608 * rules. prev remains the same throughout the cycle. 3609 */ 3610 flush_rule_ptrs(chain); 3611 while (rule->rulenum == rulenum) 3612 rule = remove_rule(chain, rule, prev); 3613 break; 3614 3615 case 1: /* delete all rules with given set number */ 3616 flush_rule_ptrs(chain); 3617 rule = chain->rules; 3618 while (rule->rulenum < IPFW_DEFAULT_RULE) 3619 if (rule->set == rulenum) 3620 rule = remove_rule(chain, rule, prev); 3621 else { 3622 prev = rule; 3623 rule = rule->next; 3624 } 3625 break; 3626 3627 case 2: /* move rules with given number to new set */ 3628 rule = chain->rules; 3629 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3630 if (rule->rulenum == rulenum) 3631 rule->set = new_set; 3632 break; 3633 3634 case 3: /* move rules with given set number to new set */ 3635 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3636 if (rule->set == rulenum) 3637 rule->set = new_set; 3638 break; 3639 3640 case 4: /* swap two sets */ 3641 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3642 if (rule->set == rulenum) 3643 rule->set = new_set; 3644 else if (rule->set == new_set) 3645 rule->set = rulenum; 3646 break; 3647 case 5: /* delete rules with given number and with given set number. 3648 * rulenum - given rule number; 3649 * new_set - given set number. 3650 */ 3651 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3652 ; 3653 if (rule->rulenum != rulenum) { 3654 IPFW_WUNLOCK(chain); 3655 return (EINVAL); 3656 } 3657 flush_rule_ptrs(chain); 3658 while (rule->rulenum == rulenum) { 3659 if (rule->set == new_set) 3660 rule = remove_rule(chain, rule, prev); 3661 else { 3662 prev = rule; 3663 rule = rule->next; 3664 } 3665 } 3666 } 3667 /* 3668 * Look for rules to reclaim. We grab the list before 3669 * releasing the lock then reclaim them w/o the lock to 3670 * avoid a LOR with dummynet. 3671 */ 3672 rule = chain->reap; 3673 chain->reap = NULL; 3674 IPFW_WUNLOCK(chain); 3675 if (rule) 3676 reap_rules(rule); 3677 return 0; 3678} 3679 3680/* 3681 * Clear counters for a specific rule. 3682 * The enclosing "table" is assumed locked. 3683 */ 3684static void 3685clear_counters(struct ip_fw *rule, int log_only) 3686{ 3687 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3688 3689 if (log_only == 0) { 3690 rule->bcnt = rule->pcnt = 0; 3691 rule->timestamp = 0; 3692 } 3693 if (l->o.opcode == O_LOG) 3694 l->log_left = l->max_log; 3695} 3696 3697/** 3698 * Reset some or all counters on firewall rules. 3699 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 3700 * the next 8 bits are the set number, the top 8 bits are the command: 3701 * 0 work with rules from all set's; 3702 * 1 work with rules only from specified set. 3703 * Specified rule number is zero if we want to clear all entries. 3704 * log_only is 1 if we only want to reset logs, zero otherwise. 3705 */ 3706static int 3707zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 3708{ 3709 struct ip_fw *rule; 3710 char *msg; 3711 3712 uint16_t rulenum = arg & 0xffff; 3713 uint8_t set = (arg >> 16) & 0xff; 3714 uint8_t cmd = (arg >> 24) & 0xff; 3715 3716 if (cmd > 1) 3717 return (EINVAL); 3718 if (cmd == 1 && set > RESVD_SET) 3719 return (EINVAL); 3720 3721 IPFW_WLOCK(chain); 3722 if (rulenum == 0) { 3723 V_norule_counter = 0; 3724 for (rule = chain->rules; rule; rule = rule->next) { 3725 /* Skip rules from another set. */ 3726 if (cmd == 1 && rule->set != set) 3727 continue; 3728 clear_counters(rule, log_only); 3729 } 3730 msg = log_only ? "ipfw: All logging counts reset.\n" : 3731 "ipfw: Accounting cleared.\n"; 3732 } else { 3733 int cleared = 0; 3734 /* 3735 * We can have multiple rules with the same number, so we 3736 * need to clear them all. 3737 */ 3738 for (rule = chain->rules; rule; rule = rule->next) 3739 if (rule->rulenum == rulenum) { 3740 while (rule && rule->rulenum == rulenum) { 3741 if (cmd == 0 || rule->set == set) 3742 clear_counters(rule, log_only); 3743 rule = rule->next; 3744 } 3745 cleared = 1; 3746 break; 3747 } 3748 if (!cleared) { /* we did not find any matching rules */ 3749 IPFW_WUNLOCK(chain); 3750 return (EINVAL); 3751 } 3752 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3753 "ipfw: Entry %d cleared.\n"; 3754 } 3755 IPFW_WUNLOCK(chain); 3756 3757 if (V_fw_verbose) 3758 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3759 return (0); 3760} 3761 3762/* 3763 * Check validity of the structure before insert. 3764 * Fortunately rules are simple, so this mostly need to check rule sizes. 3765 */ 3766static int 3767check_ipfw_struct(struct ip_fw *rule, int size) 3768{ 3769 int l, cmdlen = 0; 3770 int have_action=0; 3771 ipfw_insn *cmd; 3772 3773 if (size < sizeof(*rule)) { 3774 printf("ipfw: rule too short\n"); 3775 return (EINVAL); 3776 } 3777 /* first, check for valid size */ 3778 l = RULESIZE(rule); 3779 if (l != size) { 3780 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3781 return (EINVAL); 3782 } 3783 if (rule->act_ofs >= rule->cmd_len) { 3784 printf("ipfw: bogus action offset (%u > %u)\n", 3785 rule->act_ofs, rule->cmd_len - 1); 3786 return (EINVAL); 3787 } 3788 /* 3789 * Now go for the individual checks. Very simple ones, basically only 3790 * instruction sizes. 3791 */ 3792 for (l = rule->cmd_len, cmd = rule->cmd ; 3793 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3794 cmdlen = F_LEN(cmd); 3795 if (cmdlen > l) { 3796 printf("ipfw: opcode %d size truncated\n", 3797 cmd->opcode); 3798 return EINVAL; 3799 } 3800 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3801 switch (cmd->opcode) { 3802 case O_PROBE_STATE: 3803 case O_KEEP_STATE: 3804 case O_PROTO: 3805 case O_IP_SRC_ME: 3806 case O_IP_DST_ME: 3807 case O_LAYER2: 3808 case O_IN: 3809 case O_FRAG: 3810 case O_DIVERTED: 3811 case O_IPOPT: 3812 case O_IPTOS: 3813 case O_IPPRECEDENCE: 3814 case O_IPVER: 3815 case O_TCPWIN: 3816 case O_TCPFLAGS: 3817 case O_TCPOPTS: 3818 case O_ESTAB: 3819 case O_VERREVPATH: 3820 case O_VERSRCREACH: 3821 case O_ANTISPOOF: 3822 case O_IPSEC: 3823#ifdef INET6 3824 case O_IP6_SRC_ME: 3825 case O_IP6_DST_ME: 3826 case O_EXT_HDR: 3827 case O_IP6: 3828#endif 3829 case O_IP4: 3830 case O_TAG: 3831 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3832 goto bad_size; 3833 break; 3834 3835 case O_FIB: 3836 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3837 goto bad_size; 3838 if (cmd->arg1 >= rt_numfibs) { 3839 printf("ipfw: invalid fib number %d\n", 3840 cmd->arg1); 3841 return EINVAL; 3842 } 3843 break; 3844 3845 case O_SETFIB: 3846 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3847 goto bad_size; 3848 if (cmd->arg1 >= rt_numfibs) { 3849 printf("ipfw: invalid fib number %d\n", 3850 cmd->arg1); 3851 return EINVAL; 3852 } 3853 goto check_action; 3854 3855 case O_UID: 3856 case O_GID: 3857 case O_JAIL: 3858 case O_IP_SRC: 3859 case O_IP_DST: 3860 case O_TCPSEQ: 3861 case O_TCPACK: 3862 case O_PROB: 3863 case O_ICMPTYPE: 3864 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3865 goto bad_size; 3866 break; 3867 3868 case O_LIMIT: 3869 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3870 goto bad_size; 3871 break; 3872 3873 case O_LOG: 3874 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3875 goto bad_size; 3876 3877 ((ipfw_insn_log *)cmd)->log_left = 3878 ((ipfw_insn_log *)cmd)->max_log; 3879 3880 break; 3881 3882 case O_IP_SRC_MASK: 3883 case O_IP_DST_MASK: 3884 /* only odd command lengths */ 3885 if ( !(cmdlen & 1) || cmdlen > 31) 3886 goto bad_size; 3887 break; 3888 3889 case O_IP_SRC_SET: 3890 case O_IP_DST_SET: 3891 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3892 printf("ipfw: invalid set size %d\n", 3893 cmd->arg1); 3894 return EINVAL; 3895 } 3896 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3897 (cmd->arg1+31)/32 ) 3898 goto bad_size; 3899 break; 3900 3901 case O_IP_SRC_LOOKUP: 3902 case O_IP_DST_LOOKUP: 3903 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3904 printf("ipfw: invalid table number %d\n", 3905 cmd->arg1); 3906 return (EINVAL); 3907 } 3908 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3909 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3910 goto bad_size; 3911 break; 3912 3913 case O_MACADDR2: 3914 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3915 goto bad_size; 3916 break; 3917 3918 case O_NOP: 3919 case O_IPID: 3920 case O_IPTTL: 3921 case O_IPLEN: 3922 case O_TCPDATALEN: 3923 case O_TAGGED: 3924 if (cmdlen < 1 || cmdlen > 31) 3925 goto bad_size; 3926 break; 3927 3928 case O_MAC_TYPE: 3929 case O_IP_SRCPORT: 3930 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3931 if (cmdlen < 2 || cmdlen > 31) 3932 goto bad_size; 3933 break; 3934 3935 case O_RECV: 3936 case O_XMIT: 3937 case O_VIA: 3938 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3939 goto bad_size; 3940 break; 3941 3942 case O_ALTQ: 3943 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3944 goto bad_size; 3945 break; 3946 3947 case O_PIPE: 3948 case O_QUEUE: 3949 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3950 goto bad_size; 3951 goto check_action; 3952 3953 case O_FORWARD_IP: 3954#ifdef IPFIREWALL_FORWARD 3955 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3956 goto bad_size; 3957 goto check_action; 3958#else 3959 return EINVAL; 3960#endif 3961 3962 case O_DIVERT: 3963 case O_TEE: 3964 if (ip_divert_ptr == NULL) 3965 return EINVAL; 3966 else 3967 goto check_size; 3968 case O_NETGRAPH: 3969 case O_NGTEE: 3970 if (!NG_IPFW_LOADED) 3971 return EINVAL; 3972 else 3973 goto check_size; 3974 case O_NAT: 3975 if (!IPFW_NAT_LOADED) 3976 return EINVAL; 3977 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 3978 goto bad_size; 3979 goto check_action; 3980 case O_FORWARD_MAC: /* XXX not implemented yet */ 3981 case O_CHECK_STATE: 3982 case O_COUNT: 3983 case O_ACCEPT: 3984 case O_DENY: 3985 case O_REJECT: 3986#ifdef INET6 3987 case O_UNREACH6: 3988#endif 3989 case O_SKIPTO: 3990check_size: 3991 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3992 goto bad_size; 3993check_action: 3994 if (have_action) { 3995 printf("ipfw: opcode %d, multiple actions" 3996 " not allowed\n", 3997 cmd->opcode); 3998 return EINVAL; 3999 } 4000 have_action = 1; 4001 if (l != cmdlen) { 4002 printf("ipfw: opcode %d, action must be" 4003 " last opcode\n", 4004 cmd->opcode); 4005 return EINVAL; 4006 } 4007 break; 4008#ifdef INET6 4009 case O_IP6_SRC: 4010 case O_IP6_DST: 4011 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 4012 F_INSN_SIZE(ipfw_insn)) 4013 goto bad_size; 4014 break; 4015 4016 case O_FLOW6ID: 4017 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4018 ((ipfw_insn_u32 *)cmd)->o.arg1) 4019 goto bad_size; 4020 break; 4021 4022 case O_IP6_SRC_MASK: 4023 case O_IP6_DST_MASK: 4024 if ( !(cmdlen & 1) || cmdlen > 127) 4025 goto bad_size; 4026 break; 4027 case O_ICMP6TYPE: 4028 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4029 goto bad_size; 4030 break; 4031#endif 4032 4033 default: 4034 switch (cmd->opcode) { 4035#ifndef INET6 4036 case O_IP6_SRC_ME: 4037 case O_IP6_DST_ME: 4038 case O_EXT_HDR: 4039 case O_IP6: 4040 case O_UNREACH6: 4041 case O_IP6_SRC: 4042 case O_IP6_DST: 4043 case O_FLOW6ID: 4044 case O_IP6_SRC_MASK: 4045 case O_IP6_DST_MASK: 4046 case O_ICMP6TYPE: 4047 printf("ipfw: no IPv6 support in kernel\n"); 4048 return EPROTONOSUPPORT; 4049#endif 4050 default: 4051 printf("ipfw: opcode %d, unknown opcode\n", 4052 cmd->opcode); 4053 return EINVAL; 4054 } 4055 } 4056 } 4057 if (have_action == 0) { 4058 printf("ipfw: missing action\n"); 4059 return EINVAL; 4060 } 4061 return 0; 4062 4063bad_size: 4064 printf("ipfw: opcode %d size %d wrong\n", 4065 cmd->opcode, cmdlen); 4066 return EINVAL; 4067} 4068 4069/* 4070 * Copy the static and dynamic rules to the supplied buffer 4071 * and return the amount of space actually used. 4072 */ 4073static size_t 4074ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4075{ 4076 char *bp = buf; 4077 char *ep = bp + space; 4078 struct ip_fw *rule; 4079 int i; 4080 time_t boot_seconds; 4081 4082 boot_seconds = boottime.tv_sec; 4083 /* XXX this can take a long time and locking will block packet flow */ 4084 IPFW_RLOCK(chain); 4085 for (rule = chain->rules; rule ; rule = rule->next) { 4086 /* 4087 * Verify the entry fits in the buffer in case the 4088 * rules changed between calculating buffer space and 4089 * now. This would be better done using a generation 4090 * number but should suffice for now. 4091 */ 4092 i = RULESIZE(rule); 4093 if (bp + i <= ep) { 4094 bcopy(rule, bp, i); 4095 /* 4096 * XXX HACK. Store the disable mask in the "next" 4097 * pointer in a wild attempt to keep the ABI the same. 4098 * Why do we do this on EVERY rule? 4099 */ 4100 bcopy(&V_set_disable, 4101 &(((struct ip_fw *)bp)->next_rule), 4102 sizeof(V_set_disable)); 4103 if (((struct ip_fw *)bp)->timestamp) 4104 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4105 bp += i; 4106 } 4107 } 4108 IPFW_RUNLOCK(chain); 4109 if (V_ipfw_dyn_v) { 4110 ipfw_dyn_rule *p, *last = NULL; 4111 4112 IPFW_DYN_LOCK(); 4113 for (i = 0 ; i < V_curr_dyn_buckets; i++) 4114 for (p = V_ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4115 if (bp + sizeof *p <= ep) { 4116 ipfw_dyn_rule *dst = 4117 (ipfw_dyn_rule *)bp; 4118 bcopy(p, dst, sizeof *p); 4119 bcopy(&(p->rule->rulenum), &(dst->rule), 4120 sizeof(p->rule->rulenum)); 4121 /* 4122 * store set number into high word of 4123 * dst->rule pointer. 4124 */ 4125 bcopy(&(p->rule->set), 4126 (char *)&dst->rule + 4127 sizeof(p->rule->rulenum), 4128 sizeof(p->rule->set)); 4129 /* 4130 * store a non-null value in "next". 4131 * The userland code will interpret a 4132 * NULL here as a marker 4133 * for the last dynamic rule. 4134 */ 4135 bcopy(&dst, &dst->next, sizeof(dst)); 4136 last = dst; 4137 dst->expire = 4138 TIME_LEQ(dst->expire, time_uptime) ? 4139 0 : dst->expire - time_uptime ; 4140 bp += sizeof(ipfw_dyn_rule); 4141 } 4142 } 4143 IPFW_DYN_UNLOCK(); 4144 if (last != NULL) /* mark last dynamic rule */ 4145 bzero(&last->next, sizeof(last)); 4146 } 4147 return (bp - (char *)buf); 4148} 4149 4150 4151/** 4152 * {set|get}sockopt parser. 4153 */ 4154static int 4155ipfw_ctl(struct sockopt *sopt) 4156{ 4157#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4158 int error; 4159 size_t size; 4160 struct ip_fw *buf, *rule; 4161 u_int32_t rulenum[2]; 4162 4163 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4164 if (error) 4165 return (error); 4166 4167 /* 4168 * Disallow modifications in really-really secure mode, but still allow 4169 * the logging counters to be reset. 4170 */ 4171 if (sopt->sopt_name == IP_FW_ADD || 4172 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4173 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4174 if (error) 4175 return (error); 4176 } 4177 4178 error = 0; 4179 4180 switch (sopt->sopt_name) { 4181 case IP_FW_GET: 4182 /* 4183 * pass up a copy of the current rules. Static rules 4184 * come first (the last of which has number IPFW_DEFAULT_RULE), 4185 * followed by a possibly empty list of dynamic rule. 4186 * The last dynamic rule has NULL in the "next" field. 4187 * 4188 * Note that the calculated size is used to bound the 4189 * amount of data returned to the user. The rule set may 4190 * change between calculating the size and returning the 4191 * data in which case we'll just return what fits. 4192 */ 4193 size = V_static_len; /* size of static rules */ 4194 if (V_ipfw_dyn_v) /* add size of dyn.rules */ 4195 size += (V_dyn_count * sizeof(ipfw_dyn_rule)); 4196 4197 /* 4198 * XXX todo: if the user passes a short length just to know 4199 * how much room is needed, do not bother filling up the 4200 * buffer, just jump to the sooptcopyout. 4201 */ 4202 buf = malloc(size, M_TEMP, M_WAITOK); 4203 error = sooptcopyout(sopt, buf, 4204 ipfw_getrules(&V_layer3_chain, buf, size)); 4205 free(buf, M_TEMP); 4206 break; 4207 4208 case IP_FW_FLUSH: 4209 /* 4210 * Normally we cannot release the lock on each iteration. 4211 * We could do it here only because we start from the head all 4212 * the times so there is no risk of missing some entries. 4213 * On the other hand, the risk is that we end up with 4214 * a very inconsistent ruleset, so better keep the lock 4215 * around the whole cycle. 4216 * 4217 * XXX this code can be improved by resetting the head of 4218 * the list to point to the default rule, and then freeing 4219 * the old list without the need for a lock. 4220 */ 4221 4222 IPFW_WLOCK(&V_layer3_chain); 4223 V_layer3_chain.reap = NULL; 4224 free_chain(&V_layer3_chain, 0 /* keep default rule */); 4225 rule = V_layer3_chain.reap; 4226 V_layer3_chain.reap = NULL; 4227 IPFW_WUNLOCK(&V_layer3_chain); 4228 if (rule != NULL) 4229 reap_rules(rule); 4230 break; 4231 4232 case IP_FW_ADD: 4233 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4234 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4235 sizeof(struct ip_fw) ); 4236 if (error == 0) 4237 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4238 if (error == 0) { 4239 error = add_rule(&V_layer3_chain, rule); 4240 size = RULESIZE(rule); 4241 if (!error && sopt->sopt_dir == SOPT_GET) 4242 error = sooptcopyout(sopt, rule, size); 4243 } 4244 free(rule, M_TEMP); 4245 break; 4246 4247 case IP_FW_DEL: 4248 /* 4249 * IP_FW_DEL is used for deleting single rules or sets, 4250 * and (ab)used to atomically manipulate sets. Argument size 4251 * is used to distinguish between the two: 4252 * sizeof(u_int32_t) 4253 * delete single rule or set of rules, 4254 * or reassign rules (or sets) to a different set. 4255 * 2*sizeof(u_int32_t) 4256 * atomic disable/enable sets. 4257 * first u_int32_t contains sets to be disabled, 4258 * second u_int32_t contains sets to be enabled. 4259 */ 4260 error = sooptcopyin(sopt, rulenum, 4261 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4262 if (error) 4263 break; 4264 size = sopt->sopt_valsize; 4265 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4266 error = del_entry(&V_layer3_chain, rulenum[0]); 4267 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4268 V_set_disable = 4269 (V_set_disable | rulenum[0]) & ~rulenum[1] & 4270 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4271 else 4272 error = EINVAL; 4273 break; 4274 4275 case IP_FW_ZERO: 4276 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 4277 rulenum[0] = 0; 4278 if (sopt->sopt_val != 0) { 4279 error = sooptcopyin(sopt, rulenum, 4280 sizeof(u_int32_t), sizeof(u_int32_t)); 4281 if (error) 4282 break; 4283 } 4284 error = zero_entry(&V_layer3_chain, rulenum[0], 4285 sopt->sopt_name == IP_FW_RESETLOG); 4286 break; 4287 4288 case IP_FW_TABLE_ADD: 4289 { 4290 ipfw_table_entry ent; 4291 4292 error = sooptcopyin(sopt, &ent, 4293 sizeof(ent), sizeof(ent)); 4294 if (error) 4295 break; 4296 error = add_table_entry(&V_layer3_chain, ent.tbl, 4297 ent.addr, ent.masklen, ent.value); 4298 } 4299 break; 4300 4301 case IP_FW_TABLE_DEL: 4302 { 4303 ipfw_table_entry ent; 4304 4305 error = sooptcopyin(sopt, &ent, 4306 sizeof(ent), sizeof(ent)); 4307 if (error) 4308 break; 4309 error = del_table_entry(&V_layer3_chain, ent.tbl, 4310 ent.addr, ent.masklen); 4311 } 4312 break; 4313 4314 case IP_FW_TABLE_FLUSH: 4315 { 4316 u_int16_t tbl; 4317 4318 error = sooptcopyin(sopt, &tbl, 4319 sizeof(tbl), sizeof(tbl)); 4320 if (error) 4321 break; 4322 IPFW_WLOCK(&V_layer3_chain); 4323 error = flush_table(&V_layer3_chain, tbl); 4324 IPFW_WUNLOCK(&V_layer3_chain); 4325 } 4326 break; 4327 4328 case IP_FW_TABLE_GETSIZE: 4329 { 4330 u_int32_t tbl, cnt; 4331 4332 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4333 sizeof(tbl)))) 4334 break; 4335 IPFW_RLOCK(&V_layer3_chain); 4336 error = count_table(&V_layer3_chain, tbl, &cnt); 4337 IPFW_RUNLOCK(&V_layer3_chain); 4338 if (error) 4339 break; 4340 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4341 } 4342 break; 4343 4344 case IP_FW_TABLE_LIST: 4345 { 4346 ipfw_table *tbl; 4347 4348 if (sopt->sopt_valsize < sizeof(*tbl)) { 4349 error = EINVAL; 4350 break; 4351 } 4352 size = sopt->sopt_valsize; 4353 tbl = malloc(size, M_TEMP, M_WAITOK); 4354 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4355 if (error) { 4356 free(tbl, M_TEMP); 4357 break; 4358 } 4359 tbl->size = (size - sizeof(*tbl)) / 4360 sizeof(ipfw_table_entry); 4361 IPFW_RLOCK(&V_layer3_chain); 4362 error = dump_table(&V_layer3_chain, tbl); 4363 IPFW_RUNLOCK(&V_layer3_chain); 4364 if (error) { 4365 free(tbl, M_TEMP); 4366 break; 4367 } 4368 error = sooptcopyout(sopt, tbl, size); 4369 free(tbl, M_TEMP); 4370 } 4371 break; 4372 4373 case IP_FW_NAT_CFG: 4374 { 4375 if (IPFW_NAT_LOADED) 4376 error = ipfw_nat_cfg_ptr(sopt); 4377 else { 4378 printf("IP_FW_NAT_CFG: ipfw_nat not present, please load it.\n"); 4379 error = EINVAL; 4380 } 4381 } 4382 break; 4383 4384 case IP_FW_NAT_DEL: 4385 { 4386 if (IPFW_NAT_LOADED) 4387 error = ipfw_nat_del_ptr(sopt); 4388 else { 4389 printf("IP_FW_NAT_DEL: ipfw_nat not present, please load it.\n"); 4390 printf("ipfw_nat not loaded: %d\n", sopt->sopt_name); 4391 error = EINVAL; 4392 } 4393 } 4394 break; 4395 4396 case IP_FW_NAT_GET_CONFIG: 4397 { 4398 if (IPFW_NAT_LOADED) 4399 error = ipfw_nat_get_cfg_ptr(sopt); 4400 else { 4401 printf("IP_FW_NAT_GET_CFG: ipfw_nat not present, please load it.\n"); 4402 error = EINVAL; 4403 } 4404 } 4405 break; 4406 4407 case IP_FW_NAT_GET_LOG: 4408 { 4409 if (IPFW_NAT_LOADED) 4410 error = ipfw_nat_get_log_ptr(sopt); 4411 else { 4412 printf("IP_FW_NAT_GET_LOG: ipfw_nat not present, please load it.\n"); 4413 error = EINVAL; 4414 } 4415 } 4416 break; 4417 4418 default: 4419 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4420 error = EINVAL; 4421 } 4422 4423 return (error); 4424#undef RULE_MAXSIZE 4425} 4426 4427/** 4428 * dummynet needs a reference to the default rule, because rules can be 4429 * deleted while packets hold a reference to them. When this happens, 4430 * dummynet changes the reference to the default rule (it could well be a 4431 * NULL pointer, but this way we do not need to check for the special 4432 * case, plus here he have info on the default behaviour). 4433 */ 4434struct ip_fw *ip_fw_default_rule; 4435 4436/* 4437 * This procedure is only used to handle keepalives. It is invoked 4438 * every dyn_keepalive_period 4439 */ 4440static void 4441ipfw_tick(void * __unused unused) 4442{ 4443 struct mbuf *m0, *m, *mnext, **mtailp; 4444 int i; 4445 ipfw_dyn_rule *q; 4446 4447 if (V_dyn_keepalive == 0 || V_ipfw_dyn_v == NULL || V_dyn_count == 0) 4448 goto done; 4449 4450 /* 4451 * We make a chain of packets to go out here -- not deferring 4452 * until after we drop the IPFW dynamic rule lock would result 4453 * in a lock order reversal with the normal packet input -> ipfw 4454 * call stack. 4455 */ 4456 m0 = NULL; 4457 mtailp = &m0; 4458 IPFW_DYN_LOCK(); 4459 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 4460 for (q = V_ipfw_dyn_v[i] ; q ; q = q->next ) { 4461 if (q->dyn_type == O_LIMIT_PARENT) 4462 continue; 4463 if (q->id.proto != IPPROTO_TCP) 4464 continue; 4465 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4466 continue; 4467 if (TIME_LEQ(time_uptime + V_dyn_keepalive_interval, 4468 q->expire)) 4469 continue; /* too early */ 4470 if (TIME_LEQ(q->expire, time_uptime)) 4471 continue; /* too late, rule expired */ 4472 4473 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4474 q->ack_fwd, TH_SYN); 4475 if (*mtailp != NULL) 4476 mtailp = &(*mtailp)->m_nextpkt; 4477 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4478 q->ack_rev, 0); 4479 if (*mtailp != NULL) 4480 mtailp = &(*mtailp)->m_nextpkt; 4481 } 4482 } 4483 IPFW_DYN_UNLOCK(); 4484 for (m = mnext = m0; m != NULL; m = mnext) { 4485 mnext = m->m_nextpkt; 4486 m->m_nextpkt = NULL; 4487 ip_output(m, NULL, NULL, 0, NULL, NULL); 4488 } 4489done: 4490 callout_reset(&V_ipfw_timeout, V_dyn_keepalive_period * hz, 4491 ipfw_tick, NULL); 4492} 4493 4494int 4495ipfw_init(void) 4496{ 4497 struct ip_fw default_rule; 4498 int error; 4499 4500#ifdef INET6 4501 /* Setup IPv6 fw sysctl tree. */ 4502 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4503 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4504 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4505 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4506 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4507 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4508 &V_fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4509 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4510 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4511 &V_fw_deny_unknown_exthdrs, 0, 4512 "Deny packets with unknown IPv6 Extension Headers"); 4513#endif 4514 4515 V_layer3_chain.rules = NULL; 4516 IPFW_LOCK_INIT(&V_layer3_chain); 4517 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4518 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4519 UMA_ALIGN_PTR, 0); 4520 IPFW_DYN_LOCK_INIT(); 4521 callout_init(&V_ipfw_timeout, CALLOUT_MPSAFE); 4522 4523 bzero(&default_rule, sizeof default_rule); 4524 4525 default_rule.act_ofs = 0; 4526 default_rule.rulenum = IPFW_DEFAULT_RULE; 4527 default_rule.cmd_len = 1; 4528 default_rule.set = RESVD_SET; 4529 4530 default_rule.cmd[0].len = 1; 4531 default_rule.cmd[0].opcode = 4532#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4533 1 ? O_ACCEPT : 4534#endif 4535 O_DENY; 4536 4537 error = add_rule(&V_layer3_chain, &default_rule); 4538 if (error != 0) { 4539 printf("ipfw2: error %u initializing default rule " 4540 "(support disabled)\n", error); 4541 IPFW_DYN_LOCK_DESTROY(); 4542 IPFW_LOCK_DESTROY(&V_layer3_chain); 4543 uma_zdestroy(ipfw_dyn_rule_zone); 4544 return (error); 4545 } 4546 4547 ip_fw_default_rule = V_layer3_chain.rules; 4548 printf("ipfw2 " 4549#ifdef INET6 4550 "(+ipv6) " 4551#endif 4552 "initialized, divert %s, nat %s, " 4553 "rule-based forwarding " 4554#ifdef IPFIREWALL_FORWARD 4555 "enabled, " 4556#else 4557 "disabled, " 4558#endif 4559 "default to %s, logging ", 4560#ifdef IPDIVERT 4561 "enabled", 4562#else 4563 "loadable", 4564#endif 4565#ifdef IPFIREWALL_NAT 4566 "enabled", 4567#else 4568 "loadable", 4569#endif 4570 4571 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4572 4573#ifdef IPFIREWALL_VERBOSE 4574 V_fw_verbose = 1; 4575#endif 4576#ifdef IPFIREWALL_VERBOSE_LIMIT 4577 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4578#endif 4579 if (V_fw_verbose == 0) 4580 printf("disabled\n"); 4581 else if (V_verbose_limit == 0) 4582 printf("unlimited\n"); 4583 else 4584 printf("limited to %d packets/entry by default\n", 4585 V_verbose_limit); 4586 4587 error = init_tables(&V_layer3_chain); 4588 if (error) { 4589 IPFW_DYN_LOCK_DESTROY(); 4590 IPFW_LOCK_DESTROY(&V_layer3_chain); 4591 uma_zdestroy(ipfw_dyn_rule_zone); 4592 return (error); 4593 } 4594 ip_fw_ctl_ptr = ipfw_ctl; 4595 ip_fw_chk_ptr = ipfw_chk; 4596 callout_reset(&V_ipfw_timeout, hz, ipfw_tick, NULL); 4597 LIST_INIT(&V_layer3_chain.nat); 4598 return (0); 4599} 4600 4601void 4602ipfw_destroy(void) 4603{ 4604 struct ip_fw *reap; 4605 4606 ip_fw_chk_ptr = NULL; 4607 ip_fw_ctl_ptr = NULL; 4608 callout_drain(&V_ipfw_timeout); 4609 IPFW_WLOCK(&V_layer3_chain); 4610 flush_tables(&V_layer3_chain); 4611 V_layer3_chain.reap = NULL; 4612 free_chain(&V_layer3_chain, 1 /* kill default rule */); 4613 reap = V_layer3_chain.reap, V_layer3_chain.reap = NULL; 4614 IPFW_WUNLOCK(&V_layer3_chain); 4615 if (reap != NULL) 4616 reap_rules(reap); 4617 IPFW_DYN_LOCK_DESTROY(); 4618 uma_zdestroy(ipfw_dyn_rule_zone); 4619 if (V_ipfw_dyn_v != NULL) 4620 free(V_ipfw_dyn_v, M_IPFW); 4621 IPFW_LOCK_DESTROY(&V_layer3_chain); 4622 4623#ifdef INET6 4624 /* Free IPv6 fw sysctl tree. */ 4625 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4626#endif 4627 4628 printf("IP firewall unloaded\n"); 4629} 4630