ip_fw2.c revision 160025
1/*- 2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions 6 * are met: 7 * 1. Redistributions of source code must retain the above copyright 8 * notice, this list of conditions and the following disclaimer. 9 * 2. Redistributions in binary form must reproduce the above copyright 10 * notice, this list of conditions and the following disclaimer in the 11 * documentation and/or other materials provided with the distribution. 12 * 13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23 * SUCH DAMAGE. 24 * 25 * $FreeBSD: head/sys/netinet/ip_fw2.c 160025 2006-06-29 11:17:16Z bz $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdn.h" 38#include "opt_inet.h" 39#ifndef INET 40#error IPFIREWALL requires INET. 41#endif /* INET */ 42#endif 43#include "opt_inet6.h" 44#include "opt_ipsec.h" 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/condvar.h> 49#include <sys/malloc.h> 50#include <sys/mbuf.h> 51#include <sys/kernel.h> 52#include <sys/lock.h> 53#include <sys/jail.h> 54#include <sys/module.h> 55#include <sys/proc.h> 56#include <sys/rwlock.h> 57#include <sys/socket.h> 58#include <sys/socketvar.h> 59#include <sys/sysctl.h> 60#include <sys/syslog.h> 61#include <sys/ucred.h> 62#include <net/if.h> 63#include <net/radix.h> 64#include <net/route.h> 65#include <netinet/in.h> 66#include <netinet/in_systm.h> 67#include <netinet/in_var.h> 68#include <netinet/in_pcb.h> 69#include <netinet/ip.h> 70#include <netinet/ip_var.h> 71#include <netinet/ip_icmp.h> 72#include <netinet/ip_fw.h> 73#include <netinet/ip_divert.h> 74#include <netinet/ip_dummynet.h> 75#include <netinet/tcp.h> 76#include <netinet/tcp_timer.h> 77#include <netinet/tcp_var.h> 78#include <netinet/tcpip.h> 79#include <netinet/udp.h> 80#include <netinet/udp_var.h> 81 82#include <netgraph/ng_ipfw.h> 83 84#include <altq/if_altq.h> 85 86#ifdef IPSEC 87#include <netinet6/ipsec.h> 88#endif 89 90#include <netinet/ip6.h> 91#include <netinet/icmp6.h> 92#ifdef INET6 93#include <netinet6/scope6_var.h> 94#endif 95 96#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 97 98#include <machine/in_cksum.h> /* XXX for in_cksum */ 99 100/* 101 * set_disable contains one bit per set value (0..31). 102 * If the bit is set, all rules with the corresponding set 103 * are disabled. Set RESVD_SET(31) is reserved for the default rule 104 * and rules that are not deleted by the flush command, 105 * and CANNOT be disabled. 106 * Rules in set RESVD_SET can only be deleted explicitly. 107 */ 108static u_int32_t set_disable; 109 110static int fw_verbose; 111static int verbose_limit; 112 113static struct callout ipfw_timeout; 114static uma_zone_t ipfw_dyn_rule_zone; 115#define IPFW_DEFAULT_RULE 65535 116 117/* 118 * Data structure to cache our ucred related 119 * information. This structure only gets used if 120 * the user specified UID/GID based constraints in 121 * a firewall rule. 122 */ 123struct ip_fw_ugid { 124 gid_t fw_groups[NGROUPS]; 125 int fw_ngroups; 126 uid_t fw_uid; 127 int fw_prid; 128}; 129 130#define IPFW_TABLES_MAX 128 131struct ip_fw_chain { 132 struct ip_fw *rules; /* list of rules */ 133 struct ip_fw *reap; /* list of rules to reap */ 134 struct radix_node_head *tables[IPFW_TABLES_MAX]; 135 struct rwlock rwmtx; 136}; 137#define IPFW_LOCK_INIT(_chain) \ 138 rw_init(&(_chain)->rwmtx, "IPFW static rules") 139#define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx) 140#define IPFW_WLOCK_ASSERT(_chain) do { \ 141 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \ 142 NET_ASSERT_GIANT(); \ 143} while (0) 144 145#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 146#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 147#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 148#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 149 150/* 151 * list of rules for layer 3 152 */ 153static struct ip_fw_chain layer3_chain; 154 155MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 156MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 157 158struct table_entry { 159 struct radix_node rn[2]; 160 struct sockaddr_in addr, mask; 161 u_int32_t value; 162}; 163 164static int fw_debug = 1; 165static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 166 167extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 168 169#ifdef SYSCTL_NODE 170SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 171SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 172 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 173 ipfw_chg_hook, "I", "Enable ipfw"); 174SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 175 &autoinc_step, 0, "Rule number autincrement step"); 176SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 177 CTLFLAG_RW | CTLFLAG_SECURE3, 178 &fw_one_pass, 0, 179 "Only do a single pass through ipfw when using dummynet(4)"); 180SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 181 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 182SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 183 CTLFLAG_RW | CTLFLAG_SECURE3, 184 &fw_verbose, 0, "Log matches to ipfw rules"); 185SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 186 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 187 188/* 189 * Description of dynamic rules. 190 * 191 * Dynamic rules are stored in lists accessed through a hash table 192 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 193 * be modified through the sysctl variable dyn_buckets which is 194 * updated when the table becomes empty. 195 * 196 * XXX currently there is only one list, ipfw_dyn. 197 * 198 * When a packet is received, its address fields are first masked 199 * with the mask defined for the rule, then hashed, then matched 200 * against the entries in the corresponding list. 201 * Dynamic rules can be used for different purposes: 202 * + stateful rules; 203 * + enforcing limits on the number of sessions; 204 * + in-kernel NAT (not implemented yet) 205 * 206 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 207 * measured in seconds and depending on the flags. 208 * 209 * The total number of dynamic rules is stored in dyn_count. 210 * The max number of dynamic rules is dyn_max. When we reach 211 * the maximum number of rules we do not create anymore. This is 212 * done to avoid consuming too much memory, but also too much 213 * time when searching on each packet (ideally, we should try instead 214 * to put a limit on the length of the list on each bucket...). 215 * 216 * Each dynamic rule holds a pointer to the parent ipfw rule so 217 * we know what action to perform. Dynamic rules are removed when 218 * the parent rule is deleted. XXX we should make them survive. 219 * 220 * There are some limitations with dynamic rules -- we do not 221 * obey the 'randomized match', and we do not do multiple 222 * passes through the firewall. XXX check the latter!!! 223 */ 224static ipfw_dyn_rule **ipfw_dyn_v = NULL; 225static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 226static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 227 228static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 229#define IPFW_DYN_LOCK_INIT() \ 230 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 231#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 232#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 233#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 234#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 235 236/* 237 * Timeouts for various events in handing dynamic rules. 238 */ 239static u_int32_t dyn_ack_lifetime = 300; 240static u_int32_t dyn_syn_lifetime = 20; 241static u_int32_t dyn_fin_lifetime = 1; 242static u_int32_t dyn_rst_lifetime = 1; 243static u_int32_t dyn_udp_lifetime = 10; 244static u_int32_t dyn_short_lifetime = 5; 245 246/* 247 * Keepalives are sent if dyn_keepalive is set. They are sent every 248 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 249 * seconds of lifetime of a rule. 250 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 251 * than dyn_keepalive_period. 252 */ 253 254static u_int32_t dyn_keepalive_interval = 20; 255static u_int32_t dyn_keepalive_period = 5; 256static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 257 258static u_int32_t static_count; /* # of static rules */ 259static u_int32_t static_len; /* size in bytes of static rules */ 260static u_int32_t dyn_count; /* # of dynamic rules */ 261static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 262 263SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 264 &dyn_buckets, 0, "Number of dyn. buckets"); 265SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 266 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 267SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 268 &dyn_count, 0, "Number of dyn. rules"); 269SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 270 &dyn_max, 0, "Max number of dyn. rules"); 271SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 272 &static_count, 0, "Number of static rules"); 273SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 274 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 275SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 276 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 277SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 278 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 279SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 280 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 281SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 282 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 283SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 284 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 285SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 286 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 287 288#ifdef INET6 289/* 290 * IPv6 specific variables 291 */ 292SYSCTL_DECL(_net_inet6_ip6); 293 294static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 295static struct sysctl_oid *ip6_fw_sysctl_tree; 296#endif /* INET6 */ 297#endif /* SYSCTL_NODE */ 298 299static int fw_deny_unknown_exthdrs = 1; 300 301 302/* 303 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 304 * Other macros just cast void * into the appropriate type 305 */ 306#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 307#define TCP(p) ((struct tcphdr *)(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 == NULL) 469 continue; 470 if (ia->ifa_addr->sa_family != AF_INET) 471 continue; 472 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 473 (ia->ifa_addr))->sin_addr.s_addr) 474 return(1); /* match */ 475 } 476 } 477 return(0); /* no match, fail ... */ 478} 479 480/* 481 * The verify_path function checks if a route to the src exists and 482 * if it is reachable via ifp (when provided). 483 * 484 * The 'verrevpath' option checks that the interface that an IP packet 485 * arrives on is the same interface that traffic destined for the 486 * packet's source address would be routed out of. The 'versrcreach' 487 * option just checks that the source address is reachable via any route 488 * (except default) in the routing table. These two are a measure to block 489 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 490 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 491 * is purposely reminiscent of the Cisco IOS command, 492 * 493 * ip verify unicast reverse-path 494 * ip verify unicast source reachable-via any 495 * 496 * which implements the same functionality. But note that syntax is 497 * misleading. The check may be performed on all IP packets whether unicast, 498 * multicast, or broadcast. 499 */ 500static int 501verify_path(struct in_addr src, struct ifnet *ifp) 502{ 503 struct route ro; 504 struct sockaddr_in *dst; 505 506 bzero(&ro, sizeof(ro)); 507 508 dst = (struct sockaddr_in *)&(ro.ro_dst); 509 dst->sin_family = AF_INET; 510 dst->sin_len = sizeof(*dst); 511 dst->sin_addr = src; 512 rtalloc_ign(&ro, RTF_CLONING); 513 514 if (ro.ro_rt == NULL) 515 return 0; 516 517 /* 518 * If ifp is provided, check for equality with rtentry. 519 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 520 * in order to pass packets injected back by if_simloop(): 521 * if useloopback == 1 routing entry (via lo0) for our own address 522 * may exist, so we need to handle routing assymetry. 523 */ 524 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 525 RTFREE(ro.ro_rt); 526 return 0; 527 } 528 529 /* if no ifp provided, check if rtentry is not default route */ 530 if (ifp == NULL && 531 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 532 RTFREE(ro.ro_rt); 533 return 0; 534 } 535 536 /* or if this is a blackhole/reject route */ 537 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 538 RTFREE(ro.ro_rt); 539 return 0; 540 } 541 542 /* found valid route */ 543 RTFREE(ro.ro_rt); 544 return 1; 545} 546 547#ifdef INET6 548/* 549 * ipv6 specific rules here... 550 */ 551static __inline int 552icmp6type_match (int type, ipfw_insn_u32 *cmd) 553{ 554 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 555} 556 557static int 558flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 559{ 560 int i; 561 for (i=0; i <= cmd->o.arg1; ++i ) 562 if (curr_flow == cmd->d[i] ) 563 return 1; 564 return 0; 565} 566 567/* support for IP6_*_ME opcodes */ 568static int 569search_ip6_addr_net (struct in6_addr * ip6_addr) 570{ 571 struct ifnet *mdc; 572 struct ifaddr *mdc2; 573 struct in6_ifaddr *fdm; 574 struct in6_addr copia; 575 576 TAILQ_FOREACH(mdc, &ifnet, if_link) 577 for (mdc2 = mdc->if_addrlist.tqh_first; mdc2; 578 mdc2 = mdc2->ifa_list.tqe_next) { 579 if (!mdc2->ifa_addr) 580 continue; 581 if (mdc2->ifa_addr->sa_family == AF_INET6) { 582 fdm = (struct in6_ifaddr *)mdc2; 583 copia = fdm->ia_addr.sin6_addr; 584 /* need for leaving scope_id in the sock_addr */ 585 in6_clearscope(&copia); 586 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 587 return 1; 588 } 589 } 590 return 0; 591} 592 593static int 594verify_path6(struct in6_addr *src, struct ifnet *ifp) 595{ 596 struct route_in6 ro; 597 struct sockaddr_in6 *dst; 598 599 bzero(&ro, sizeof(ro)); 600 601 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 602 dst->sin6_family = AF_INET6; 603 dst->sin6_len = sizeof(*dst); 604 dst->sin6_addr = *src; 605 rtalloc_ign((struct route *)&ro, RTF_CLONING); 606 607 if (ro.ro_rt == NULL) 608 return 0; 609 610 /* 611 * if ifp is provided, check for equality with rtentry 612 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 613 * to support the case of sending packets to an address of our own. 614 * (where the former interface is the first argument of if_simloop() 615 * (=ifp), the latter is lo0) 616 */ 617 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 618 RTFREE(ro.ro_rt); 619 return 0; 620 } 621 622 /* if no ifp provided, check if rtentry is not default route */ 623 if (ifp == NULL && 624 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 625 RTFREE(ro.ro_rt); 626 return 0; 627 } 628 629 /* or if this is a blackhole/reject route */ 630 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 631 RTFREE(ro.ro_rt); 632 return 0; 633 } 634 635 /* found valid route */ 636 RTFREE(ro.ro_rt); 637 return 1; 638 639} 640static __inline int 641hash_packet6(struct ipfw_flow_id *id) 642{ 643 u_int32_t i; 644 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 645 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 646 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 647 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 648 (id->dst_port) ^ (id->src_port); 649 return i; 650} 651 652static int 653is_icmp6_query(int icmp6_type) 654{ 655 if ((icmp6_type <= ICMP6_MAXTYPE) && 656 (icmp6_type == ICMP6_ECHO_REQUEST || 657 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 658 icmp6_type == ICMP6_WRUREQUEST || 659 icmp6_type == ICMP6_FQDN_QUERY || 660 icmp6_type == ICMP6_NI_QUERY)) 661 return (1); 662 663 return (0); 664} 665 666static void 667send_reject6(struct ip_fw_args *args, int code, u_int hlen) 668{ 669 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 670 struct ip6_hdr *ip6; 671 struct tcphdr *tcp; 672 tcp_seq ack, seq; 673 int flags; 674 struct { 675 struct ip6_hdr ip6; 676 struct tcphdr th; 677 } ti; 678 679 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) { 680 args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr)); 681 if (args->m == NULL) 682 return; 683 } 684 685 ip6 = mtod(args->m, struct ip6_hdr *); 686 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 687 688 if ((tcp->th_flags & TH_RST) != 0) { 689 m_freem(args->m); 690 return; 691 } 692 693 ti.ip6 = *ip6; 694 ti.th = *tcp; 695 ti.th.th_seq = ntohl(ti.th.th_seq); 696 ti.th.th_ack = ntohl(ti.th.th_ack); 697 ti.ip6.ip6_nxt = IPPROTO_TCP; 698 699 if (ti.th.th_flags & TH_ACK) { 700 ack = 0; 701 seq = ti.th.th_ack; 702 flags = TH_RST; 703 } else { 704 ack = ti.th.th_seq; 705 if (((args->m)->m_flags & M_PKTHDR) != 0) { 706 ack += (args->m)->m_pkthdr.len - hlen 707 - (ti.th.th_off << 2); 708 } else if (ip6->ip6_plen) { 709 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) 710 - hlen - (ti.th.th_off << 2); 711 } else { 712 m_freem(args->m); 713 return; 714 } 715 if (tcp->th_flags & TH_SYN) 716 ack++; 717 seq = 0; 718 flags = TH_RST|TH_ACK; 719 } 720 bcopy(&ti, ip6, sizeof(ti)); 721 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1), 722 args->m, ack, seq, flags); 723 724 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 725 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0); 726 727 } else 728 m_freem(args->m); 729 730 args->m = NULL; 731} 732 733#endif /* INET6 */ 734 735static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 736 737#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 738#define SNP(buf) buf, sizeof(buf) 739 740/* 741 * We enter here when we have a rule with O_LOG. 742 * XXX this function alone takes about 2Kbytes of code! 743 */ 744static void 745ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 746 struct mbuf *m, struct ifnet *oif, u_short offset) 747{ 748 struct ether_header *eh = args->eh; 749 char *action; 750 int limit_reached = 0; 751 char action2[40], proto[128], fragment[32]; 752 753 fragment[0] = '\0'; 754 proto[0] = '\0'; 755 756 if (f == NULL) { /* bogus pkt */ 757 if (verbose_limit != 0 && norule_counter >= verbose_limit) 758 return; 759 norule_counter++; 760 if (norule_counter == verbose_limit) 761 limit_reached = verbose_limit; 762 action = "Refuse"; 763 } else { /* O_LOG is the first action, find the real one */ 764 ipfw_insn *cmd = ACTION_PTR(f); 765 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 766 767 if (l->max_log != 0 && l->log_left == 0) 768 return; 769 l->log_left--; 770 if (l->log_left == 0) 771 limit_reached = l->max_log; 772 cmd += F_LEN(cmd); /* point to first action */ 773 if (cmd->opcode == O_ALTQ) { 774 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 775 776 snprintf(SNPARGS(action2, 0), "Altq %d", 777 altq->qid); 778 cmd += F_LEN(cmd); 779 } 780 if (cmd->opcode == O_PROB) 781 cmd += F_LEN(cmd); 782 783 if (cmd->opcode == O_TAG) 784 cmd += F_LEN(cmd); 785 786 action = action2; 787 switch (cmd->opcode) { 788 case O_DENY: 789 action = "Deny"; 790 break; 791 792 case O_REJECT: 793 if (cmd->arg1==ICMP_REJECT_RST) 794 action = "Reset"; 795 else if (cmd->arg1==ICMP_UNREACH_HOST) 796 action = "Reject"; 797 else 798 snprintf(SNPARGS(action2, 0), "Unreach %d", 799 cmd->arg1); 800 break; 801 802 case O_UNREACH6: 803 if (cmd->arg1==ICMP6_UNREACH_RST) 804 action = "Reset"; 805 else 806 snprintf(SNPARGS(action2, 0), "Unreach %d", 807 cmd->arg1); 808 break; 809 810 case O_ACCEPT: 811 action = "Accept"; 812 break; 813 case O_COUNT: 814 action = "Count"; 815 break; 816 case O_DIVERT: 817 snprintf(SNPARGS(action2, 0), "Divert %d", 818 cmd->arg1); 819 break; 820 case O_TEE: 821 snprintf(SNPARGS(action2, 0), "Tee %d", 822 cmd->arg1); 823 break; 824 case O_SKIPTO: 825 snprintf(SNPARGS(action2, 0), "SkipTo %d", 826 cmd->arg1); 827 break; 828 case O_PIPE: 829 snprintf(SNPARGS(action2, 0), "Pipe %d", 830 cmd->arg1); 831 break; 832 case O_QUEUE: 833 snprintf(SNPARGS(action2, 0), "Queue %d", 834 cmd->arg1); 835 break; 836 case O_FORWARD_IP: { 837 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 838 int len; 839 840 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 841 inet_ntoa(sa->sa.sin_addr)); 842 if (sa->sa.sin_port) 843 snprintf(SNPARGS(action2, len), ":%d", 844 sa->sa.sin_port); 845 } 846 break; 847 case O_NETGRAPH: 848 snprintf(SNPARGS(action2, 0), "Netgraph %d", 849 cmd->arg1); 850 break; 851 case O_NGTEE: 852 snprintf(SNPARGS(action2, 0), "Ngtee %d", 853 cmd->arg1); 854 break; 855 default: 856 action = "UNKNOWN"; 857 break; 858 } 859 } 860 861 if (hlen == 0) { /* non-ip */ 862 snprintf(SNPARGS(proto, 0), "MAC"); 863 864 } else { 865 int len; 866 char src[48], dst[48]; 867 struct icmphdr *icmp; 868 struct tcphdr *tcp; 869 struct udphdr *udp; 870 /* Initialize to make compiler happy. */ 871 struct ip *ip = NULL; 872#ifdef INET6 873 struct ip6_hdr *ip6 = NULL; 874 struct icmp6_hdr *icmp6; 875#endif 876 src[0] = '\0'; 877 dst[0] = '\0'; 878#ifdef INET6 879 if (args->f_id.addr_type == 6) { 880 snprintf(src, sizeof(src), "[%s]", 881 ip6_sprintf(&args->f_id.src_ip6)); 882 snprintf(dst, sizeof(dst), "[%s]", 883 ip6_sprintf(&args->f_id.dst_ip6)); 884 885 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 886 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 887 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 888 } else 889#endif 890 { 891 ip = mtod(m, struct ip *); 892 tcp = L3HDR(struct tcphdr, ip); 893 udp = L3HDR(struct udphdr, ip); 894 895 inet_ntoa_r(ip->ip_src, src); 896 inet_ntoa_r(ip->ip_dst, dst); 897 } 898 899 switch (args->f_id.proto) { 900 case IPPROTO_TCP: 901 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 902 if (offset == 0) 903 snprintf(SNPARGS(proto, len), ":%d %s:%d", 904 ntohs(tcp->th_sport), 905 dst, 906 ntohs(tcp->th_dport)); 907 else 908 snprintf(SNPARGS(proto, len), " %s", dst); 909 break; 910 911 case IPPROTO_UDP: 912 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 913 if (offset == 0) 914 snprintf(SNPARGS(proto, len), ":%d %s:%d", 915 ntohs(udp->uh_sport), 916 dst, 917 ntohs(udp->uh_dport)); 918 else 919 snprintf(SNPARGS(proto, len), " %s", dst); 920 break; 921 922 case IPPROTO_ICMP: 923 icmp = L3HDR(struct icmphdr, ip); 924 if (offset == 0) 925 len = snprintf(SNPARGS(proto, 0), 926 "ICMP:%u.%u ", 927 icmp->icmp_type, icmp->icmp_code); 928 else 929 len = snprintf(SNPARGS(proto, 0), "ICMP "); 930 len += snprintf(SNPARGS(proto, len), "%s", src); 931 snprintf(SNPARGS(proto, len), " %s", dst); 932 break; 933#ifdef INET6 934 case IPPROTO_ICMPV6: 935 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 936 if (offset == 0) 937 len = snprintf(SNPARGS(proto, 0), 938 "ICMPv6:%u.%u ", 939 icmp6->icmp6_type, icmp6->icmp6_code); 940 else 941 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 942 len += snprintf(SNPARGS(proto, len), "%s", src); 943 snprintf(SNPARGS(proto, len), " %s", dst); 944 break; 945#endif 946 default: 947 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 948 args->f_id.proto, src); 949 snprintf(SNPARGS(proto, len), " %s", dst); 950 break; 951 } 952 953#ifdef INET6 954 if (args->f_id.addr_type == 6) { 955 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 956 snprintf(SNPARGS(fragment, 0), 957 " (frag %08x:%d@%d%s)", 958 args->f_id.frag_id6, 959 ntohs(ip6->ip6_plen) - hlen, 960 ntohs(offset & IP6F_OFF_MASK) << 3, 961 (offset & IP6F_MORE_FRAG) ? "+" : ""); 962 } else 963#endif 964 { 965 int ip_off, ip_len; 966 if (eh != NULL) { /* layer 2 packets are as on the wire */ 967 ip_off = ntohs(ip->ip_off); 968 ip_len = ntohs(ip->ip_len); 969 } else { 970 ip_off = ip->ip_off; 971 ip_len = ip->ip_len; 972 } 973 if (ip_off & (IP_MF | IP_OFFMASK)) 974 snprintf(SNPARGS(fragment, 0), 975 " (frag %d:%d@%d%s)", 976 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 977 offset << 3, 978 (ip_off & IP_MF) ? "+" : ""); 979 } 980 } 981 if (oif || m->m_pkthdr.rcvif) 982 log(LOG_SECURITY | LOG_INFO, 983 "ipfw: %d %s %s %s via %s%s\n", 984 f ? f->rulenum : -1, 985 action, proto, oif ? "out" : "in", 986 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 987 fragment); 988 else 989 log(LOG_SECURITY | LOG_INFO, 990 "ipfw: %d %s %s [no if info]%s\n", 991 f ? f->rulenum : -1, 992 action, proto, fragment); 993 if (limit_reached) 994 log(LOG_SECURITY | LOG_NOTICE, 995 "ipfw: limit %d reached on entry %d\n", 996 limit_reached, f ? f->rulenum : -1); 997} 998 999/* 1000 * IMPORTANT: the hash function for dynamic rules must be commutative 1001 * in source and destination (ip,port), because rules are bidirectional 1002 * and we want to find both in the same bucket. 1003 */ 1004static __inline int 1005hash_packet(struct ipfw_flow_id *id) 1006{ 1007 u_int32_t i; 1008 1009#ifdef INET6 1010 if (IS_IP6_FLOW_ID(id)) 1011 i = hash_packet6(id); 1012 else 1013#endif /* INET6 */ 1014 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1015 i &= (curr_dyn_buckets - 1); 1016 return i; 1017} 1018 1019/** 1020 * unlink a dynamic rule from a chain. prev is a pointer to 1021 * the previous one, q is a pointer to the rule to delete, 1022 * head is a pointer to the head of the queue. 1023 * Modifies q and potentially also head. 1024 */ 1025#define UNLINK_DYN_RULE(prev, head, q) { \ 1026 ipfw_dyn_rule *old_q = q; \ 1027 \ 1028 /* remove a refcount to the parent */ \ 1029 if (q->dyn_type == O_LIMIT) \ 1030 q->parent->count--; \ 1031 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1032 (q->id.src_ip), (q->id.src_port), \ 1033 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1034 if (prev != NULL) \ 1035 prev->next = q = q->next; \ 1036 else \ 1037 head = q = q->next; \ 1038 dyn_count--; \ 1039 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1040 1041#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1042 1043/** 1044 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1045 * 1046 * If keep_me == NULL, rules are deleted even if not expired, 1047 * otherwise only expired rules are removed. 1048 * 1049 * The value of the second parameter is also used to point to identify 1050 * a rule we absolutely do not want to remove (e.g. because we are 1051 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1052 * rules). The pointer is only used for comparison, so any non-null 1053 * value will do. 1054 */ 1055static void 1056remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1057{ 1058 static u_int32_t last_remove = 0; 1059 1060#define FORCE (keep_me == NULL) 1061 1062 ipfw_dyn_rule *prev, *q; 1063 int i, pass = 0, max_pass = 0; 1064 1065 IPFW_DYN_LOCK_ASSERT(); 1066 1067 if (ipfw_dyn_v == NULL || dyn_count == 0) 1068 return; 1069 /* do not expire more than once per second, it is useless */ 1070 if (!FORCE && last_remove == time_uptime) 1071 return; 1072 last_remove = time_uptime; 1073 1074 /* 1075 * because O_LIMIT refer to parent rules, during the first pass only 1076 * remove child and mark any pending LIMIT_PARENT, and remove 1077 * them in a second pass. 1078 */ 1079next_pass: 1080 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1081 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1082 /* 1083 * Logic can become complex here, so we split tests. 1084 */ 1085 if (q == keep_me) 1086 goto next; 1087 if (rule != NULL && rule != q->rule) 1088 goto next; /* not the one we are looking for */ 1089 if (q->dyn_type == O_LIMIT_PARENT) { 1090 /* 1091 * handle parent in the second pass, 1092 * record we need one. 1093 */ 1094 max_pass = 1; 1095 if (pass == 0) 1096 goto next; 1097 if (FORCE && q->count != 0 ) { 1098 /* XXX should not happen! */ 1099 printf("ipfw: OUCH! cannot remove rule," 1100 " count %d\n", q->count); 1101 } 1102 } else { 1103 if (!FORCE && 1104 !TIME_LEQ( q->expire, time_uptime )) 1105 goto next; 1106 } 1107 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1108 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1109 continue; 1110 } 1111next: 1112 prev=q; 1113 q=q->next; 1114 } 1115 } 1116 if (pass++ < max_pass) 1117 goto next_pass; 1118} 1119 1120 1121/** 1122 * lookup a dynamic rule. 1123 */ 1124static ipfw_dyn_rule * 1125lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1126 struct tcphdr *tcp) 1127{ 1128 /* 1129 * stateful ipfw extensions. 1130 * Lookup into dynamic session queue 1131 */ 1132#define MATCH_REVERSE 0 1133#define MATCH_FORWARD 1 1134#define MATCH_NONE 2 1135#define MATCH_UNKNOWN 3 1136 int i, dir = MATCH_NONE; 1137 ipfw_dyn_rule *prev, *q=NULL; 1138 1139 IPFW_DYN_LOCK_ASSERT(); 1140 1141 if (ipfw_dyn_v == NULL) 1142 goto done; /* not found */ 1143 i = hash_packet( pkt ); 1144 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1145 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1146 goto next; 1147 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1148 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1149 continue; 1150 } 1151 if (pkt->proto == q->id.proto && 1152 q->dyn_type != O_LIMIT_PARENT) { 1153 if (IS_IP6_FLOW_ID(pkt)) { 1154 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1155 &(q->id.src_ip6)) && 1156 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1157 &(q->id.dst_ip6)) && 1158 pkt->src_port == q->id.src_port && 1159 pkt->dst_port == q->id.dst_port ) { 1160 dir = MATCH_FORWARD; 1161 break; 1162 } 1163 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1164 &(q->id.dst_ip6)) && 1165 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1166 &(q->id.src_ip6)) && 1167 pkt->src_port == q->id.dst_port && 1168 pkt->dst_port == q->id.src_port ) { 1169 dir = MATCH_REVERSE; 1170 break; 1171 } 1172 } else { 1173 if (pkt->src_ip == q->id.src_ip && 1174 pkt->dst_ip == q->id.dst_ip && 1175 pkt->src_port == q->id.src_port && 1176 pkt->dst_port == q->id.dst_port ) { 1177 dir = MATCH_FORWARD; 1178 break; 1179 } 1180 if (pkt->src_ip == q->id.dst_ip && 1181 pkt->dst_ip == q->id.src_ip && 1182 pkt->src_port == q->id.dst_port && 1183 pkt->dst_port == q->id.src_port ) { 1184 dir = MATCH_REVERSE; 1185 break; 1186 } 1187 } 1188 } 1189next: 1190 prev = q; 1191 q = q->next; 1192 } 1193 if (q == NULL) 1194 goto done; /* q = NULL, not found */ 1195 1196 if ( prev != NULL) { /* found and not in front */ 1197 prev->next = q->next; 1198 q->next = ipfw_dyn_v[i]; 1199 ipfw_dyn_v[i] = q; 1200 } 1201 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1202 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1203 1204#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1205#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1206 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1207 switch (q->state) { 1208 case TH_SYN: /* opening */ 1209 q->expire = time_uptime + dyn_syn_lifetime; 1210 break; 1211 1212 case BOTH_SYN: /* move to established */ 1213 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1214 case BOTH_SYN | (TH_FIN << 8) : 1215 if (tcp) { 1216#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1217 u_int32_t ack = ntohl(tcp->th_ack); 1218 if (dir == MATCH_FORWARD) { 1219 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1220 q->ack_fwd = ack; 1221 else { /* ignore out-of-sequence */ 1222 break; 1223 } 1224 } else { 1225 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1226 q->ack_rev = ack; 1227 else { /* ignore out-of-sequence */ 1228 break; 1229 } 1230 } 1231 } 1232 q->expire = time_uptime + dyn_ack_lifetime; 1233 break; 1234 1235 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1236 if (dyn_fin_lifetime >= dyn_keepalive_period) 1237 dyn_fin_lifetime = dyn_keepalive_period - 1; 1238 q->expire = time_uptime + dyn_fin_lifetime; 1239 break; 1240 1241 default: 1242#if 0 1243 /* 1244 * reset or some invalid combination, but can also 1245 * occur if we use keep-state the wrong way. 1246 */ 1247 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1248 printf("invalid state: 0x%x\n", q->state); 1249#endif 1250 if (dyn_rst_lifetime >= dyn_keepalive_period) 1251 dyn_rst_lifetime = dyn_keepalive_period - 1; 1252 q->expire = time_uptime + dyn_rst_lifetime; 1253 break; 1254 } 1255 } else if (pkt->proto == IPPROTO_UDP) { 1256 q->expire = time_uptime + dyn_udp_lifetime; 1257 } else { 1258 /* other protocols */ 1259 q->expire = time_uptime + dyn_short_lifetime; 1260 } 1261done: 1262 if (match_direction) 1263 *match_direction = dir; 1264 return q; 1265} 1266 1267static ipfw_dyn_rule * 1268lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1269 struct tcphdr *tcp) 1270{ 1271 ipfw_dyn_rule *q; 1272 1273 IPFW_DYN_LOCK(); 1274 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1275 if (q == NULL) 1276 IPFW_DYN_UNLOCK(); 1277 /* NB: return table locked when q is not NULL */ 1278 return q; 1279} 1280 1281static void 1282realloc_dynamic_table(void) 1283{ 1284 IPFW_DYN_LOCK_ASSERT(); 1285 1286 /* 1287 * Try reallocation, make sure we have a power of 2 and do 1288 * not allow more than 64k entries. In case of overflow, 1289 * default to 1024. 1290 */ 1291 1292 if (dyn_buckets > 65536) 1293 dyn_buckets = 1024; 1294 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1295 dyn_buckets = curr_dyn_buckets; /* reset */ 1296 return; 1297 } 1298 curr_dyn_buckets = dyn_buckets; 1299 if (ipfw_dyn_v != NULL) 1300 free(ipfw_dyn_v, M_IPFW); 1301 for (;;) { 1302 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1303 M_IPFW, M_NOWAIT | M_ZERO); 1304 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1305 break; 1306 curr_dyn_buckets /= 2; 1307 } 1308} 1309 1310/** 1311 * Install state of type 'type' for a dynamic session. 1312 * The hash table contains two type of rules: 1313 * - regular rules (O_KEEP_STATE) 1314 * - rules for sessions with limited number of sess per user 1315 * (O_LIMIT). When they are created, the parent is 1316 * increased by 1, and decreased on delete. In this case, 1317 * the third parameter is the parent rule and not the chain. 1318 * - "parent" rules for the above (O_LIMIT_PARENT). 1319 */ 1320static ipfw_dyn_rule * 1321add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1322{ 1323 ipfw_dyn_rule *r; 1324 int i; 1325 1326 IPFW_DYN_LOCK_ASSERT(); 1327 1328 if (ipfw_dyn_v == NULL || 1329 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1330 realloc_dynamic_table(); 1331 if (ipfw_dyn_v == NULL) 1332 return NULL; /* failed ! */ 1333 } 1334 i = hash_packet(id); 1335 1336 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1337 if (r == NULL) { 1338 printf ("ipfw: sorry cannot allocate state\n"); 1339 return NULL; 1340 } 1341 1342 /* increase refcount on parent, and set pointer */ 1343 if (dyn_type == O_LIMIT) { 1344 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1345 if ( parent->dyn_type != O_LIMIT_PARENT) 1346 panic("invalid parent"); 1347 parent->count++; 1348 r->parent = parent; 1349 rule = parent->rule; 1350 } 1351 1352 r->id = *id; 1353 r->expire = time_uptime + dyn_syn_lifetime; 1354 r->rule = rule; 1355 r->dyn_type = dyn_type; 1356 r->pcnt = r->bcnt = 0; 1357 r->count = 0; 1358 1359 r->bucket = i; 1360 r->next = ipfw_dyn_v[i]; 1361 ipfw_dyn_v[i] = r; 1362 dyn_count++; 1363 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1364 dyn_type, 1365 (r->id.src_ip), (r->id.src_port), 1366 (r->id.dst_ip), (r->id.dst_port), 1367 dyn_count ); ) 1368 return r; 1369} 1370 1371/** 1372 * lookup dynamic parent rule using pkt and rule as search keys. 1373 * If the lookup fails, then install one. 1374 */ 1375static ipfw_dyn_rule * 1376lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1377{ 1378 ipfw_dyn_rule *q; 1379 int i; 1380 1381 IPFW_DYN_LOCK_ASSERT(); 1382 1383 if (ipfw_dyn_v) { 1384 int is_v6 = IS_IP6_FLOW_ID(pkt); 1385 i = hash_packet( pkt ); 1386 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1387 if (q->dyn_type == O_LIMIT_PARENT && 1388 rule== q->rule && 1389 pkt->proto == q->id.proto && 1390 pkt->src_port == q->id.src_port && 1391 pkt->dst_port == q->id.dst_port && 1392 ( 1393 (is_v6 && 1394 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1395 &(q->id.src_ip6)) && 1396 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1397 &(q->id.dst_ip6))) || 1398 (!is_v6 && 1399 pkt->src_ip == q->id.src_ip && 1400 pkt->dst_ip == q->id.dst_ip) 1401 ) 1402 ) { 1403 q->expire = time_uptime + dyn_short_lifetime; 1404 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1405 return q; 1406 } 1407 } 1408 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1409} 1410 1411/** 1412 * Install dynamic state for rule type cmd->o.opcode 1413 * 1414 * Returns 1 (failure) if state is not installed because of errors or because 1415 * session limitations are enforced. 1416 */ 1417static int 1418install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1419 struct ip_fw_args *args, uint32_t tablearg) 1420{ 1421 static int last_log; 1422 1423 ipfw_dyn_rule *q; 1424 1425 DEB( 1426 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1427 __func__, cmd->o.opcode, 1428 (args->f_id.src_ip), (args->f_id.src_port), 1429 (args->f_id.dst_ip), (args->f_id.dst_port)); 1430 ) 1431 1432 IPFW_DYN_LOCK(); 1433 1434 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1435 1436 if (q != NULL) { /* should never occur */ 1437 if (last_log != time_uptime) { 1438 last_log = time_uptime; 1439 printf("ipfw: %s: entry already present, done\n", 1440 __func__); 1441 } 1442 IPFW_DYN_UNLOCK(); 1443 return (0); 1444 } 1445 1446 if (dyn_count >= dyn_max) 1447 /* Run out of slots, try to remove any expired rule. */ 1448 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1449 1450 if (dyn_count >= dyn_max) { 1451 if (last_log != time_uptime) { 1452 last_log = time_uptime; 1453 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1454 } 1455 IPFW_DYN_UNLOCK(); 1456 return (1); /* cannot install, notify caller */ 1457 } 1458 1459 switch (cmd->o.opcode) { 1460 case O_KEEP_STATE: /* bidir rule */ 1461 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1462 break; 1463 1464 case O_LIMIT: { /* limit number of sessions */ 1465 struct ipfw_flow_id id; 1466 ipfw_dyn_rule *parent; 1467 uint32_t conn_limit; 1468 uint16_t limit_mask = cmd->limit_mask; 1469 1470 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1471 tablearg : cmd->conn_limit; 1472 1473 DEB( 1474 if (cmd->conn_limit == IP_FW_TABLEARG) 1475 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1476 "(tablearg)\n", __func__, conn_limit); 1477 else 1478 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1479 __func__, conn_limit); 1480 ) 1481 1482 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1483 id.proto = args->f_id.proto; 1484 id.addr_type = args->f_id.addr_type; 1485 1486 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1487 if (limit_mask & DYN_SRC_ADDR) 1488 id.src_ip6 = args->f_id.src_ip6; 1489 if (limit_mask & DYN_DST_ADDR) 1490 id.dst_ip6 = args->f_id.dst_ip6; 1491 } else { 1492 if (limit_mask & DYN_SRC_ADDR) 1493 id.src_ip = args->f_id.src_ip; 1494 if (limit_mask & DYN_DST_ADDR) 1495 id.dst_ip = args->f_id.dst_ip; 1496 } 1497 if (limit_mask & DYN_SRC_PORT) 1498 id.src_port = args->f_id.src_port; 1499 if (limit_mask & DYN_DST_PORT) 1500 id.dst_port = args->f_id.dst_port; 1501 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1502 printf("ipfw: %s: add parent failed\n", __func__); 1503 IPFW_DYN_UNLOCK(); 1504 return (1); 1505 } 1506 1507 if (parent->count >= conn_limit) { 1508 /* See if we can remove some expired rule. */ 1509 remove_dyn_rule(rule, parent); 1510 if (parent->count >= conn_limit) { 1511 if (fw_verbose && last_log != time_uptime) { 1512 last_log = time_uptime; 1513 log(LOG_SECURITY | LOG_DEBUG, 1514 "drop session, too many entries\n"); 1515 } 1516 IPFW_DYN_UNLOCK(); 1517 return (1); 1518 } 1519 } 1520 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1521 break; 1522 } 1523 default: 1524 printf("ipfw: %s: unknown dynamic rule type %u\n", 1525 __func__, cmd->o.opcode); 1526 IPFW_DYN_UNLOCK(); 1527 return (1); 1528 } 1529 1530 /* XXX just set lifetime */ 1531 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1532 1533 IPFW_DYN_UNLOCK(); 1534 return (0); 1535} 1536 1537/* 1538 * Generate a TCP packet, containing either a RST or a keepalive. 1539 * When flags & TH_RST, we are sending a RST packet, because of a 1540 * "reset" action matched the packet. 1541 * Otherwise we are sending a keepalive, and flags & TH_ 1542 */ 1543static struct mbuf * 1544send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1545{ 1546 struct mbuf *m; 1547 struct ip *ip; 1548 struct tcphdr *tcp; 1549 1550 MGETHDR(m, M_DONTWAIT, MT_DATA); 1551 if (m == 0) 1552 return (NULL); 1553 m->m_pkthdr.rcvif = (struct ifnet *)0; 1554 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1555 m->m_data += max_linkhdr; 1556 1557 ip = mtod(m, struct ip *); 1558 bzero(ip, m->m_len); 1559 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1560 ip->ip_p = IPPROTO_TCP; 1561 tcp->th_off = 5; 1562 /* 1563 * Assume we are sending a RST (or a keepalive in the reverse 1564 * direction), swap src and destination addresses and ports. 1565 */ 1566 ip->ip_src.s_addr = htonl(id->dst_ip); 1567 ip->ip_dst.s_addr = htonl(id->src_ip); 1568 tcp->th_sport = htons(id->dst_port); 1569 tcp->th_dport = htons(id->src_port); 1570 if (flags & TH_RST) { /* we are sending a RST */ 1571 if (flags & TH_ACK) { 1572 tcp->th_seq = htonl(ack); 1573 tcp->th_ack = htonl(0); 1574 tcp->th_flags = TH_RST; 1575 } else { 1576 if (flags & TH_SYN) 1577 seq++; 1578 tcp->th_seq = htonl(0); 1579 tcp->th_ack = htonl(seq); 1580 tcp->th_flags = TH_RST | TH_ACK; 1581 } 1582 } else { 1583 /* 1584 * We are sending a keepalive. flags & TH_SYN determines 1585 * the direction, forward if set, reverse if clear. 1586 * NOTE: seq and ack are always assumed to be correct 1587 * as set by the caller. This may be confusing... 1588 */ 1589 if (flags & TH_SYN) { 1590 /* 1591 * we have to rewrite the correct addresses! 1592 */ 1593 ip->ip_dst.s_addr = htonl(id->dst_ip); 1594 ip->ip_src.s_addr = htonl(id->src_ip); 1595 tcp->th_dport = htons(id->dst_port); 1596 tcp->th_sport = htons(id->src_port); 1597 } 1598 tcp->th_seq = htonl(seq); 1599 tcp->th_ack = htonl(ack); 1600 tcp->th_flags = TH_ACK; 1601 } 1602 /* 1603 * set ip_len to the payload size so we can compute 1604 * the tcp checksum on the pseudoheader 1605 * XXX check this, could save a couple of words ? 1606 */ 1607 ip->ip_len = htons(sizeof(struct tcphdr)); 1608 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1609 /* 1610 * now fill fields left out earlier 1611 */ 1612 ip->ip_ttl = ip_defttl; 1613 ip->ip_len = m->m_pkthdr.len; 1614 m->m_flags |= M_SKIP_FIREWALL; 1615 return (m); 1616} 1617 1618/* 1619 * sends a reject message, consuming the mbuf passed as an argument. 1620 */ 1621static void 1622send_reject(struct ip_fw_args *args, int code, int ip_len) 1623{ 1624 1625 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1626 /* We need the IP header in host order for icmp_error(). */ 1627 if (args->eh != NULL) { 1628 struct ip *ip = mtod(args->m, struct ip *); 1629 ip->ip_len = ntohs(ip->ip_len); 1630 ip->ip_off = ntohs(ip->ip_off); 1631 } 1632 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1633 } else if (args->f_id.proto == IPPROTO_TCP) { 1634 struct tcphdr *const tcp = 1635 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1636 if ( (tcp->th_flags & TH_RST) == 0) { 1637 struct mbuf *m; 1638 m = send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1639 ntohl(tcp->th_ack), 1640 tcp->th_flags | TH_RST); 1641 if (m != NULL) 1642 ip_output(m, NULL, NULL, 0, NULL, NULL); 1643 } 1644 m_freem(args->m); 1645 } else 1646 m_freem(args->m); 1647 args->m = NULL; 1648} 1649 1650/** 1651 * 1652 * Given an ip_fw *, lookup_next_rule will return a pointer 1653 * to the next rule, which can be either the jump 1654 * target (for skipto instructions) or the next one in the list (in 1655 * all other cases including a missing jump target). 1656 * The result is also written in the "next_rule" field of the rule. 1657 * Backward jumps are not allowed, so start looking from the next 1658 * rule... 1659 * 1660 * This never returns NULL -- in case we do not have an exact match, 1661 * the next rule is returned. When the ruleset is changed, 1662 * pointers are flushed so we are always correct. 1663 */ 1664 1665static struct ip_fw * 1666lookup_next_rule(struct ip_fw *me) 1667{ 1668 struct ip_fw *rule = NULL; 1669 ipfw_insn *cmd; 1670 1671 /* look for action, in case it is a skipto */ 1672 cmd = ACTION_PTR(me); 1673 if (cmd->opcode == O_LOG) 1674 cmd += F_LEN(cmd); 1675 if (cmd->opcode == O_ALTQ) 1676 cmd += F_LEN(cmd); 1677 if (cmd->opcode == O_TAG) 1678 cmd += F_LEN(cmd); 1679 if ( cmd->opcode == O_SKIPTO ) 1680 for (rule = me->next; rule ; rule = rule->next) 1681 if (rule->rulenum >= cmd->arg1) 1682 break; 1683 if (rule == NULL) /* failure or not a skipto */ 1684 rule = me->next; 1685 me->next_rule = rule; 1686 return rule; 1687} 1688 1689static int 1690add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1691 uint8_t mlen, uint32_t value) 1692{ 1693 struct radix_node_head *rnh; 1694 struct table_entry *ent; 1695 1696 if (tbl >= IPFW_TABLES_MAX) 1697 return (EINVAL); 1698 rnh = ch->tables[tbl]; 1699 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1700 if (ent == NULL) 1701 return (ENOMEM); 1702 ent->value = value; 1703 ent->addr.sin_len = ent->mask.sin_len = 8; 1704 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1705 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1706 IPFW_WLOCK(&layer3_chain); 1707 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1708 NULL) { 1709 IPFW_WUNLOCK(&layer3_chain); 1710 free(ent, M_IPFW_TBL); 1711 return (EEXIST); 1712 } 1713 IPFW_WUNLOCK(&layer3_chain); 1714 return (0); 1715} 1716 1717static int 1718del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1719 uint8_t mlen) 1720{ 1721 struct radix_node_head *rnh; 1722 struct table_entry *ent; 1723 struct sockaddr_in sa, mask; 1724 1725 if (tbl >= IPFW_TABLES_MAX) 1726 return (EINVAL); 1727 rnh = ch->tables[tbl]; 1728 sa.sin_len = mask.sin_len = 8; 1729 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1730 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1731 IPFW_WLOCK(ch); 1732 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1733 if (ent == NULL) { 1734 IPFW_WUNLOCK(ch); 1735 return (ESRCH); 1736 } 1737 IPFW_WUNLOCK(ch); 1738 free(ent, M_IPFW_TBL); 1739 return (0); 1740} 1741 1742static int 1743flush_table_entry(struct radix_node *rn, void *arg) 1744{ 1745 struct radix_node_head * const rnh = arg; 1746 struct table_entry *ent; 1747 1748 ent = (struct table_entry *) 1749 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1750 if (ent != NULL) 1751 free(ent, M_IPFW_TBL); 1752 return (0); 1753} 1754 1755static int 1756flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1757{ 1758 struct radix_node_head *rnh; 1759 1760 IPFW_WLOCK_ASSERT(ch); 1761 1762 if (tbl >= IPFW_TABLES_MAX) 1763 return (EINVAL); 1764 rnh = ch->tables[tbl]; 1765 KASSERT(rnh != NULL, ("NULL IPFW table")); 1766 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1767 return (0); 1768} 1769 1770static void 1771flush_tables(struct ip_fw_chain *ch) 1772{ 1773 uint16_t tbl; 1774 1775 IPFW_WLOCK_ASSERT(ch); 1776 1777 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1778 flush_table(ch, tbl); 1779} 1780 1781static int 1782init_tables(struct ip_fw_chain *ch) 1783{ 1784 int i; 1785 uint16_t j; 1786 1787 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1788 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1789 for (j = 0; j < i; j++) { 1790 (void) flush_table(ch, j); 1791 } 1792 return (ENOMEM); 1793 } 1794 } 1795 return (0); 1796} 1797 1798static int 1799lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1800 uint32_t *val) 1801{ 1802 struct radix_node_head *rnh; 1803 struct table_entry *ent; 1804 struct sockaddr_in sa; 1805 1806 if (tbl >= IPFW_TABLES_MAX) 1807 return (0); 1808 rnh = ch->tables[tbl]; 1809 sa.sin_len = 8; 1810 sa.sin_addr.s_addr = addr; 1811 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1812 if (ent != NULL) { 1813 *val = ent->value; 1814 return (1); 1815 } 1816 return (0); 1817} 1818 1819static int 1820count_table_entry(struct radix_node *rn, void *arg) 1821{ 1822 u_int32_t * const cnt = arg; 1823 1824 (*cnt)++; 1825 return (0); 1826} 1827 1828static int 1829count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1830{ 1831 struct radix_node_head *rnh; 1832 1833 if (tbl >= IPFW_TABLES_MAX) 1834 return (EINVAL); 1835 rnh = ch->tables[tbl]; 1836 *cnt = 0; 1837 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1838 return (0); 1839} 1840 1841static int 1842dump_table_entry(struct radix_node *rn, void *arg) 1843{ 1844 struct table_entry * const n = (struct table_entry *)rn; 1845 ipfw_table * const tbl = arg; 1846 ipfw_table_entry *ent; 1847 1848 if (tbl->cnt == tbl->size) 1849 return (1); 1850 ent = &tbl->ent[tbl->cnt]; 1851 ent->tbl = tbl->tbl; 1852 if (in_nullhost(n->mask.sin_addr)) 1853 ent->masklen = 0; 1854 else 1855 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1856 ent->addr = n->addr.sin_addr.s_addr; 1857 ent->value = n->value; 1858 tbl->cnt++; 1859 return (0); 1860} 1861 1862static int 1863dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1864{ 1865 struct radix_node_head *rnh; 1866 1867 if (tbl->tbl >= IPFW_TABLES_MAX) 1868 return (EINVAL); 1869 rnh = ch->tables[tbl->tbl]; 1870 tbl->cnt = 0; 1871 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1872 return (0); 1873} 1874 1875static void 1876fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1877{ 1878 struct ucred *cr; 1879 1880 if (inp->inp_socket != NULL) { 1881 cr = inp->inp_socket->so_cred; 1882 ugp->fw_prid = jailed(cr) ? 1883 cr->cr_prison->pr_id : -1; 1884 ugp->fw_uid = cr->cr_uid; 1885 ugp->fw_ngroups = cr->cr_ngroups; 1886 bcopy(cr->cr_groups, ugp->fw_groups, 1887 sizeof(ugp->fw_groups)); 1888 } 1889} 1890 1891static int 1892check_uidgid(ipfw_insn_u32 *insn, 1893 int proto, struct ifnet *oif, 1894 struct in_addr dst_ip, u_int16_t dst_port, 1895 struct in_addr src_ip, u_int16_t src_port, 1896 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1897{ 1898 struct inpcbinfo *pi; 1899 int wildcard; 1900 struct inpcb *pcb; 1901 int match; 1902 gid_t *gp; 1903 1904 /* 1905 * Check to see if the UDP or TCP stack supplied us with 1906 * the PCB. If so, rather then holding a lock and looking 1907 * up the PCB, we can use the one that was supplied. 1908 */ 1909 if (inp && *lookup == 0) { 1910 INP_LOCK_ASSERT(inp); 1911 if (inp->inp_socket != NULL) { 1912 fill_ugid_cache(inp, ugp); 1913 *lookup = 1; 1914 } 1915 } 1916 /* 1917 * If we have already been here and the packet has no 1918 * PCB entry associated with it, then we can safely 1919 * assume that this is a no match. 1920 */ 1921 if (*lookup == -1) 1922 return (0); 1923 if (proto == IPPROTO_TCP) { 1924 wildcard = 0; 1925 pi = &tcbinfo; 1926 } else if (proto == IPPROTO_UDP) { 1927 wildcard = INPLOOKUP_WILDCARD; 1928 pi = &udbinfo; 1929 } else 1930 return 0; 1931 match = 0; 1932 if (*lookup == 0) { 1933 INP_INFO_RLOCK(pi); 1934 pcb = (oif) ? 1935 in_pcblookup_hash(pi, 1936 dst_ip, htons(dst_port), 1937 src_ip, htons(src_port), 1938 wildcard, oif) : 1939 in_pcblookup_hash(pi, 1940 src_ip, htons(src_port), 1941 dst_ip, htons(dst_port), 1942 wildcard, NULL); 1943 if (pcb != NULL) { 1944 INP_LOCK(pcb); 1945 if (pcb->inp_socket != NULL) { 1946 fill_ugid_cache(pcb, ugp); 1947 *lookup = 1; 1948 } 1949 INP_UNLOCK(pcb); 1950 } 1951 INP_INFO_RUNLOCK(pi); 1952 if (*lookup == 0) { 1953 /* 1954 * If the lookup did not yield any results, there 1955 * is no sense in coming back and trying again. So 1956 * we can set lookup to -1 and ensure that we wont 1957 * bother the pcb system again. 1958 */ 1959 *lookup = -1; 1960 return (0); 1961 } 1962 } 1963 if (insn->o.opcode == O_UID) 1964 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1965 else if (insn->o.opcode == O_GID) { 1966 for (gp = ugp->fw_groups; 1967 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1968 if (*gp == (gid_t)insn->d[0]) { 1969 match = 1; 1970 break; 1971 } 1972 } else if (insn->o.opcode == O_JAIL) 1973 match = (ugp->fw_prid == (int)insn->d[0]); 1974 return match; 1975} 1976 1977/* 1978 * The main check routine for the firewall. 1979 * 1980 * All arguments are in args so we can modify them and return them 1981 * back to the caller. 1982 * 1983 * Parameters: 1984 * 1985 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1986 * Starts with the IP header. 1987 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1988 * args->oif Outgoing interface, or NULL if packet is incoming. 1989 * The incoming interface is in the mbuf. (in) 1990 * args->divert_rule (in/out) 1991 * Skip up to the first rule past this rule number; 1992 * upon return, non-zero port number for divert or tee. 1993 * 1994 * args->rule Pointer to the last matching rule (in/out) 1995 * args->next_hop Socket we are forwarding to (out). 1996 * args->f_id Addresses grabbed from the packet (out) 1997 * args->cookie a cookie depending on rule action 1998 * 1999 * Return value: 2000 * 2001 * IP_FW_PASS the packet must be accepted 2002 * IP_FW_DENY the packet must be dropped 2003 * IP_FW_DIVERT divert packet, port in m_tag 2004 * IP_FW_TEE tee packet, port in m_tag 2005 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2006 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2007 * 2008 */ 2009 2010int 2011ipfw_chk(struct ip_fw_args *args) 2012{ 2013 /* 2014 * Local variables hold state during the processing of a packet. 2015 * 2016 * IMPORTANT NOTE: to speed up the processing of rules, there 2017 * are some assumption on the values of the variables, which 2018 * are documented here. Should you change them, please check 2019 * the implementation of the various instructions to make sure 2020 * that they still work. 2021 * 2022 * args->eh The MAC header. It is non-null for a layer2 2023 * packet, it is NULL for a layer-3 packet. 2024 * 2025 * m | args->m Pointer to the mbuf, as received from the caller. 2026 * It may change if ipfw_chk() does an m_pullup, or if it 2027 * consumes the packet because it calls send_reject(). 2028 * XXX This has to change, so that ipfw_chk() never modifies 2029 * or consumes the buffer. 2030 * ip is simply an alias of the value of m, and it is kept 2031 * in sync with it (the packet is supposed to start with 2032 * the ip header). 2033 */ 2034 struct mbuf *m = args->m; 2035 struct ip *ip = mtod(m, struct ip *); 2036 2037 /* 2038 * For rules which contain uid/gid or jail constraints, cache 2039 * a copy of the users credentials after the pcb lookup has been 2040 * executed. This will speed up the processing of rules with 2041 * these types of constraints, as well as decrease contention 2042 * on pcb related locks. 2043 */ 2044 struct ip_fw_ugid fw_ugid_cache; 2045 int ugid_lookup = 0; 2046 2047 /* 2048 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2049 * associated with a packet input on a divert socket. This 2050 * will allow to distinguish traffic and its direction when 2051 * it originates from a divert socket. 2052 */ 2053 u_int divinput_flags = 0; 2054 2055 /* 2056 * oif | args->oif If NULL, ipfw_chk has been called on the 2057 * inbound path (ether_input, ip_input). 2058 * If non-NULL, ipfw_chk has been called on the outbound path 2059 * (ether_output, ip_output). 2060 */ 2061 struct ifnet *oif = args->oif; 2062 2063 struct ip_fw *f = NULL; /* matching rule */ 2064 int retval = 0; 2065 2066 /* 2067 * hlen The length of the IP header. 2068 */ 2069 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2070 2071 /* 2072 * offset The offset of a fragment. offset != 0 means that 2073 * we have a fragment at this offset of an IPv4 packet. 2074 * offset == 0 means that (if this is an IPv4 packet) 2075 * this is the first or only fragment. 2076 * For IPv6 offset == 0 means there is no Fragment Header. 2077 * If offset != 0 for IPv6 always use correct mask to 2078 * get the correct offset because we add IP6F_MORE_FRAG 2079 * to be able to dectect the first fragment which would 2080 * otherwise have offset = 0. 2081 */ 2082 u_short offset = 0; 2083 2084 /* 2085 * Local copies of addresses. They are only valid if we have 2086 * an IP packet. 2087 * 2088 * proto The protocol. Set to 0 for non-ip packets, 2089 * or to the protocol read from the packet otherwise. 2090 * proto != 0 means that we have an IPv4 packet. 2091 * 2092 * src_port, dst_port port numbers, in HOST format. Only 2093 * valid for TCP and UDP packets. 2094 * 2095 * src_ip, dst_ip ip addresses, in NETWORK format. 2096 * Only valid for IPv4 packets. 2097 */ 2098 u_int8_t proto; 2099 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2100 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2101 u_int16_t ip_len=0; 2102 int pktlen; 2103 2104 /* 2105 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2106 * MATCH_NONE when checked and not matched (q = NULL), 2107 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2108 */ 2109 int dyn_dir = MATCH_UNKNOWN; 2110 ipfw_dyn_rule *q = NULL; 2111 struct ip_fw_chain *chain = &layer3_chain; 2112 struct m_tag *mtag; 2113 2114 /* 2115 * We store in ulp a pointer to the upper layer protocol header. 2116 * In the ipv4 case this is easy to determine from the header, 2117 * but for ipv6 we might have some additional headers in the middle. 2118 * ulp is NULL if not found. 2119 */ 2120 void *ulp = NULL; /* upper layer protocol pointer. */ 2121 /* XXX ipv6 variables */ 2122 int is_ipv6 = 0; 2123 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2124 /* end of ipv6 variables */ 2125 int is_ipv4 = 0; 2126 2127 if (m->m_flags & M_SKIP_FIREWALL) 2128 return (IP_FW_PASS); /* accept */ 2129 2130 pktlen = m->m_pkthdr.len; 2131 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2132 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2133 2134/* 2135 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2136 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2137 * pointer might become stale after other pullups (but we never use it 2138 * this way). 2139 */ 2140#define PULLUP_TO(len, p, T) \ 2141do { \ 2142 int x = (len) + sizeof(T); \ 2143 if ((m)->m_len < x) { \ 2144 args->m = m = m_pullup(m, x); \ 2145 if (m == NULL) \ 2146 goto pullup_failed; \ 2147 } \ 2148 p = (mtod(m, char *) + (len)); \ 2149} while (0) 2150 2151 /* Identify IP packets and fill up variables. */ 2152 if (pktlen >= sizeof(struct ip6_hdr) && 2153 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2154 mtod(m, struct ip *)->ip_v == 6) { 2155 is_ipv6 = 1; 2156 args->f_id.addr_type = 6; 2157 hlen = sizeof(struct ip6_hdr); 2158 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2159 2160 /* Search extension headers to find upper layer protocols */ 2161 while (ulp == NULL) { 2162 switch (proto) { 2163 case IPPROTO_ICMPV6: 2164 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2165 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2166 break; 2167 2168 case IPPROTO_TCP: 2169 PULLUP_TO(hlen, ulp, struct tcphdr); 2170 dst_port = TCP(ulp)->th_dport; 2171 src_port = TCP(ulp)->th_sport; 2172 args->f_id.flags = TCP(ulp)->th_flags; 2173 break; 2174 2175 case IPPROTO_UDP: 2176 PULLUP_TO(hlen, ulp, struct udphdr); 2177 dst_port = UDP(ulp)->uh_dport; 2178 src_port = UDP(ulp)->uh_sport; 2179 break; 2180 2181 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2182 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2183 ext_hd |= EXT_HOPOPTS; 2184 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2185 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2186 ulp = NULL; 2187 break; 2188 2189 case IPPROTO_ROUTING: /* RFC 2460 */ 2190 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2191 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2192 case 0: 2193 break; 2194 default: 2195 printf("IPFW2: IPV6 - Unknown Routing " 2196 "Header type(%d)\n", 2197 ((struct ip6_rthdr *)ulp)->ip6r_type); 2198 if (fw_deny_unknown_exthdrs) 2199 return (IP_FW_DENY); 2200 break; 2201 } 2202 ext_hd |= EXT_ROUTING; 2203 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2204 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2205 ulp = NULL; 2206 break; 2207 2208 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2209 PULLUP_TO(hlen, ulp, struct ip6_frag); 2210 ext_hd |= EXT_FRAGMENT; 2211 hlen += sizeof (struct ip6_frag); 2212 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2213 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2214 IP6F_OFF_MASK; 2215 /* Add IP6F_MORE_FRAG for offset of first 2216 * fragment to be != 0. */ 2217 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2218 IP6F_MORE_FRAG; 2219 if (offset == 0) { 2220 printf("IPFW2: IPV6 - Invalid Fragment " 2221 "Header\n"); 2222 if (fw_deny_unknown_exthdrs) 2223 return (IP_FW_DENY); 2224 break; 2225 } 2226 args->f_id.frag_id6 = 2227 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2228 ulp = NULL; 2229 break; 2230 2231 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2232 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2233 ext_hd |= EXT_DSTOPTS; 2234 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2235 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2236 ulp = NULL; 2237 break; 2238 2239 case IPPROTO_AH: /* RFC 2402 */ 2240 PULLUP_TO(hlen, ulp, struct ip6_ext); 2241 ext_hd |= EXT_AH; 2242 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2243 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2244 ulp = NULL; 2245 break; 2246 2247 case IPPROTO_ESP: /* RFC 2406 */ 2248 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2249 /* Anything past Seq# is variable length and 2250 * data past this ext. header is encrypted. */ 2251 ext_hd |= EXT_ESP; 2252 break; 2253 2254 case IPPROTO_NONE: /* RFC 2460 */ 2255 PULLUP_TO(hlen, ulp, struct ip6_ext); 2256 /* Packet ends here. if ip6e_len!=0 octets 2257 * must be ignored. */ 2258 break; 2259 2260 case IPPROTO_OSPFIGP: 2261 /* XXX OSPF header check? */ 2262 PULLUP_TO(hlen, ulp, struct ip6_ext); 2263 break; 2264 2265 case IPPROTO_IPV6: /* RFC 2893 */ 2266 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2267 break; 2268 2269 case IPPROTO_IPV4: /* RFC 2893 */ 2270 PULLUP_TO(hlen, ulp, struct ip); 2271 break; 2272 2273 default: 2274 printf("IPFW2: IPV6 - Unknown Extension " 2275 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2276 if (fw_deny_unknown_exthdrs) 2277 return (IP_FW_DENY); 2278 PULLUP_TO(hlen, ulp, struct ip6_ext); 2279 break; 2280 } /*switch */ 2281 } 2282 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2283 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2284 args->f_id.src_ip = 0; 2285 args->f_id.dst_ip = 0; 2286 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2287 } else if (pktlen >= sizeof(struct ip) && 2288 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2289 mtod(m, struct ip *)->ip_v == 4) { 2290 is_ipv4 = 1; 2291 ip = mtod(m, struct ip *); 2292 hlen = ip->ip_hl << 2; 2293 args->f_id.addr_type = 4; 2294 2295 /* 2296 * Collect parameters into local variables for faster matching. 2297 */ 2298 proto = ip->ip_p; 2299 src_ip = ip->ip_src; 2300 dst_ip = ip->ip_dst; 2301 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2302 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2303 ip_len = ntohs(ip->ip_len); 2304 } else { 2305 offset = ip->ip_off & IP_OFFMASK; 2306 ip_len = ip->ip_len; 2307 } 2308 pktlen = ip_len < pktlen ? ip_len : pktlen; 2309 2310 if (offset == 0) { 2311 switch (proto) { 2312 case IPPROTO_TCP: 2313 PULLUP_TO(hlen, ulp, struct tcphdr); 2314 dst_port = TCP(ulp)->th_dport; 2315 src_port = TCP(ulp)->th_sport; 2316 args->f_id.flags = TCP(ulp)->th_flags; 2317 break; 2318 2319 case IPPROTO_UDP: 2320 PULLUP_TO(hlen, ulp, struct udphdr); 2321 dst_port = UDP(ulp)->uh_dport; 2322 src_port = UDP(ulp)->uh_sport; 2323 break; 2324 2325 case IPPROTO_ICMP: 2326 PULLUP_TO(hlen, ulp, struct icmphdr); 2327 args->f_id.flags = ICMP(ulp)->icmp_type; 2328 break; 2329 2330 default: 2331 break; 2332 } 2333 } 2334 2335 args->f_id.src_ip = ntohl(src_ip.s_addr); 2336 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2337 } 2338#undef PULLUP_TO 2339 if (proto) { /* we may have port numbers, store them */ 2340 args->f_id.proto = proto; 2341 args->f_id.src_port = src_port = ntohs(src_port); 2342 args->f_id.dst_port = dst_port = ntohs(dst_port); 2343 } 2344 2345 IPFW_RLOCK(chain); 2346 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2347 if (args->rule) { 2348 /* 2349 * Packet has already been tagged. Look for the next rule 2350 * to restart processing. 2351 * 2352 * If fw_one_pass != 0 then just accept it. 2353 * XXX should not happen here, but optimized out in 2354 * the caller. 2355 */ 2356 if (fw_one_pass) { 2357 IPFW_RUNLOCK(chain); 2358 return (IP_FW_PASS); 2359 } 2360 2361 f = args->rule->next_rule; 2362 if (f == NULL) 2363 f = lookup_next_rule(args->rule); 2364 } else { 2365 /* 2366 * Find the starting rule. It can be either the first 2367 * one, or the one after divert_rule if asked so. 2368 */ 2369 int skipto = mtag ? divert_cookie(mtag) : 0; 2370 2371 f = chain->rules; 2372 if (args->eh == NULL && skipto != 0) { 2373 if (skipto >= IPFW_DEFAULT_RULE) { 2374 IPFW_RUNLOCK(chain); 2375 return (IP_FW_DENY); /* invalid */ 2376 } 2377 while (f && f->rulenum <= skipto) 2378 f = f->next; 2379 if (f == NULL) { /* drop packet */ 2380 IPFW_RUNLOCK(chain); 2381 return (IP_FW_DENY); 2382 } 2383 } 2384 } 2385 /* reset divert rule to avoid confusion later */ 2386 if (mtag) { 2387 divinput_flags = divert_info(mtag) & 2388 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2389 m_tag_delete(m, mtag); 2390 } 2391 2392 /* 2393 * Now scan the rules, and parse microinstructions for each rule. 2394 */ 2395 for (; f; f = f->next) { 2396 ipfw_insn *cmd; 2397 uint32_t tablearg = 0; 2398 int l, cmdlen, skip_or; /* skip rest of OR block */ 2399 2400again: 2401 if (set_disable & (1 << f->set) ) 2402 continue; 2403 2404 skip_or = 0; 2405 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2406 l -= cmdlen, cmd += cmdlen) { 2407 int match; 2408 2409 /* 2410 * check_body is a jump target used when we find a 2411 * CHECK_STATE, and need to jump to the body of 2412 * the target rule. 2413 */ 2414 2415check_body: 2416 cmdlen = F_LEN(cmd); 2417 /* 2418 * An OR block (insn_1 || .. || insn_n) has the 2419 * F_OR bit set in all but the last instruction. 2420 * The first match will set "skip_or", and cause 2421 * the following instructions to be skipped until 2422 * past the one with the F_OR bit clear. 2423 */ 2424 if (skip_or) { /* skip this instruction */ 2425 if ((cmd->len & F_OR) == 0) 2426 skip_or = 0; /* next one is good */ 2427 continue; 2428 } 2429 match = 0; /* set to 1 if we succeed */ 2430 2431 switch (cmd->opcode) { 2432 /* 2433 * The first set of opcodes compares the packet's 2434 * fields with some pattern, setting 'match' if a 2435 * match is found. At the end of the loop there is 2436 * logic to deal with F_NOT and F_OR flags associated 2437 * with the opcode. 2438 */ 2439 case O_NOP: 2440 match = 1; 2441 break; 2442 2443 case O_FORWARD_MAC: 2444 printf("ipfw: opcode %d unimplemented\n", 2445 cmd->opcode); 2446 break; 2447 2448 case O_GID: 2449 case O_UID: 2450 case O_JAIL: 2451 /* 2452 * We only check offset == 0 && proto != 0, 2453 * as this ensures that we have a 2454 * packet with the ports info. 2455 */ 2456 if (offset!=0) 2457 break; 2458 if (is_ipv6) /* XXX to be fixed later */ 2459 break; 2460 if (proto == IPPROTO_TCP || 2461 proto == IPPROTO_UDP) 2462 match = check_uidgid( 2463 (ipfw_insn_u32 *)cmd, 2464 proto, oif, 2465 dst_ip, dst_port, 2466 src_ip, src_port, &fw_ugid_cache, 2467 &ugid_lookup, args->inp); 2468 break; 2469 2470 case O_RECV: 2471 match = iface_match(m->m_pkthdr.rcvif, 2472 (ipfw_insn_if *)cmd); 2473 break; 2474 2475 case O_XMIT: 2476 match = iface_match(oif, (ipfw_insn_if *)cmd); 2477 break; 2478 2479 case O_VIA: 2480 match = iface_match(oif ? oif : 2481 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2482 break; 2483 2484 case O_MACADDR2: 2485 if (args->eh != NULL) { /* have MAC header */ 2486 u_int32_t *want = (u_int32_t *) 2487 ((ipfw_insn_mac *)cmd)->addr; 2488 u_int32_t *mask = (u_int32_t *) 2489 ((ipfw_insn_mac *)cmd)->mask; 2490 u_int32_t *hdr = (u_int32_t *)args->eh; 2491 2492 match = 2493 ( want[0] == (hdr[0] & mask[0]) && 2494 want[1] == (hdr[1] & mask[1]) && 2495 want[2] == (hdr[2] & mask[2]) ); 2496 } 2497 break; 2498 2499 case O_MAC_TYPE: 2500 if (args->eh != NULL) { 2501 u_int16_t t = 2502 ntohs(args->eh->ether_type); 2503 u_int16_t *p = 2504 ((ipfw_insn_u16 *)cmd)->ports; 2505 int i; 2506 2507 for (i = cmdlen - 1; !match && i>0; 2508 i--, p += 2) 2509 match = (t>=p[0] && t<=p[1]); 2510 } 2511 break; 2512 2513 case O_FRAG: 2514 match = (offset != 0); 2515 break; 2516 2517 case O_IN: /* "out" is "not in" */ 2518 match = (oif == NULL); 2519 break; 2520 2521 case O_LAYER2: 2522 match = (args->eh != NULL); 2523 break; 2524 2525 case O_DIVERTED: 2526 match = (cmd->arg1 & 1 && divinput_flags & 2527 IP_FW_DIVERT_LOOPBACK_FLAG) || 2528 (cmd->arg1 & 2 && divinput_flags & 2529 IP_FW_DIVERT_OUTPUT_FLAG); 2530 break; 2531 2532 case O_PROTO: 2533 /* 2534 * We do not allow an arg of 0 so the 2535 * check of "proto" only suffices. 2536 */ 2537 match = (proto == cmd->arg1); 2538 break; 2539 2540 case O_IP_SRC: 2541 match = is_ipv4 && 2542 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2543 src_ip.s_addr); 2544 break; 2545 2546 case O_IP_SRC_LOOKUP: 2547 case O_IP_DST_LOOKUP: 2548 if (is_ipv4) { 2549 uint32_t a = 2550 (cmd->opcode == O_IP_DST_LOOKUP) ? 2551 dst_ip.s_addr : src_ip.s_addr; 2552 uint32_t v; 2553 2554 match = lookup_table(chain, cmd->arg1, a, 2555 &v); 2556 if (!match) 2557 break; 2558 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2559 match = 2560 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2561 else 2562 tablearg = v; 2563 } 2564 break; 2565 2566 case O_IP_SRC_MASK: 2567 case O_IP_DST_MASK: 2568 if (is_ipv4) { 2569 uint32_t a = 2570 (cmd->opcode == O_IP_DST_MASK) ? 2571 dst_ip.s_addr : src_ip.s_addr; 2572 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2573 int i = cmdlen-1; 2574 2575 for (; !match && i>0; i-= 2, p+= 2) 2576 match = (p[0] == (a & p[1])); 2577 } 2578 break; 2579 2580 case O_IP_SRC_ME: 2581 if (is_ipv4) { 2582 struct ifnet *tif; 2583 2584 INADDR_TO_IFP(src_ip, tif); 2585 match = (tif != NULL); 2586 } 2587 break; 2588 2589 case O_IP_DST_SET: 2590 case O_IP_SRC_SET: 2591 if (is_ipv4) { 2592 u_int32_t *d = (u_int32_t *)(cmd+1); 2593 u_int32_t addr = 2594 cmd->opcode == O_IP_DST_SET ? 2595 args->f_id.dst_ip : 2596 args->f_id.src_ip; 2597 2598 if (addr < d[0]) 2599 break; 2600 addr -= d[0]; /* subtract base */ 2601 match = (addr < cmd->arg1) && 2602 ( d[ 1 + (addr>>5)] & 2603 (1<<(addr & 0x1f)) ); 2604 } 2605 break; 2606 2607 case O_IP_DST: 2608 match = is_ipv4 && 2609 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2610 dst_ip.s_addr); 2611 break; 2612 2613 case O_IP_DST_ME: 2614 if (is_ipv4) { 2615 struct ifnet *tif; 2616 2617 INADDR_TO_IFP(dst_ip, tif); 2618 match = (tif != NULL); 2619 } 2620 break; 2621 2622 case O_IP_SRCPORT: 2623 case O_IP_DSTPORT: 2624 /* 2625 * offset == 0 && proto != 0 is enough 2626 * to guarantee that we have a 2627 * packet with port info. 2628 */ 2629 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2630 && offset == 0) { 2631 u_int16_t x = 2632 (cmd->opcode == O_IP_SRCPORT) ? 2633 src_port : dst_port ; 2634 u_int16_t *p = 2635 ((ipfw_insn_u16 *)cmd)->ports; 2636 int i; 2637 2638 for (i = cmdlen - 1; !match && i>0; 2639 i--, p += 2) 2640 match = (x>=p[0] && x<=p[1]); 2641 } 2642 break; 2643 2644 case O_ICMPTYPE: 2645 match = (offset == 0 && proto==IPPROTO_ICMP && 2646 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2647 break; 2648 2649#ifdef INET6 2650 case O_ICMP6TYPE: 2651 match = is_ipv6 && offset == 0 && 2652 proto==IPPROTO_ICMPV6 && 2653 icmp6type_match( 2654 ICMP6(ulp)->icmp6_type, 2655 (ipfw_insn_u32 *)cmd); 2656 break; 2657#endif /* INET6 */ 2658 2659 case O_IPOPT: 2660 match = (is_ipv4 && 2661 ipopts_match(mtod(m, struct ip *), cmd) ); 2662 break; 2663 2664 case O_IPVER: 2665 match = (is_ipv4 && 2666 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2667 break; 2668 2669 case O_IPID: 2670 case O_IPLEN: 2671 case O_IPTTL: 2672 if (is_ipv4) { /* only for IP packets */ 2673 uint16_t x; 2674 uint16_t *p; 2675 int i; 2676 2677 if (cmd->opcode == O_IPLEN) 2678 x = ip_len; 2679 else if (cmd->opcode == O_IPTTL) 2680 x = mtod(m, struct ip *)->ip_ttl; 2681 else /* must be IPID */ 2682 x = ntohs(mtod(m, struct ip *)->ip_id); 2683 if (cmdlen == 1) { 2684 match = (cmd->arg1 == x); 2685 break; 2686 } 2687 /* otherwise we have ranges */ 2688 p = ((ipfw_insn_u16 *)cmd)->ports; 2689 i = cmdlen - 1; 2690 for (; !match && i>0; i--, p += 2) 2691 match = (x >= p[0] && x <= p[1]); 2692 } 2693 break; 2694 2695 case O_IPPRECEDENCE: 2696 match = (is_ipv4 && 2697 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2698 break; 2699 2700 case O_IPTOS: 2701 match = (is_ipv4 && 2702 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2703 break; 2704 2705 case O_TCPDATALEN: 2706 if (proto == IPPROTO_TCP && offset == 0) { 2707 struct tcphdr *tcp; 2708 uint16_t x; 2709 uint16_t *p; 2710 int i; 2711 2712 tcp = TCP(ulp); 2713 x = ip_len - 2714 ((ip->ip_hl + tcp->th_off) << 2); 2715 if (cmdlen == 1) { 2716 match = (cmd->arg1 == x); 2717 break; 2718 } 2719 /* otherwise we have ranges */ 2720 p = ((ipfw_insn_u16 *)cmd)->ports; 2721 i = cmdlen - 1; 2722 for (; !match && i>0; i--, p += 2) 2723 match = (x >= p[0] && x <= p[1]); 2724 } 2725 break; 2726 2727 case O_TCPFLAGS: 2728 match = (proto == IPPROTO_TCP && offset == 0 && 2729 flags_match(cmd, TCP(ulp)->th_flags)); 2730 break; 2731 2732 case O_TCPOPTS: 2733 match = (proto == IPPROTO_TCP && offset == 0 && 2734 tcpopts_match(TCP(ulp), cmd)); 2735 break; 2736 2737 case O_TCPSEQ: 2738 match = (proto == IPPROTO_TCP && offset == 0 && 2739 ((ipfw_insn_u32 *)cmd)->d[0] == 2740 TCP(ulp)->th_seq); 2741 break; 2742 2743 case O_TCPACK: 2744 match = (proto == IPPROTO_TCP && offset == 0 && 2745 ((ipfw_insn_u32 *)cmd)->d[0] == 2746 TCP(ulp)->th_ack); 2747 break; 2748 2749 case O_TCPWIN: 2750 match = (proto == IPPROTO_TCP && offset == 0 && 2751 cmd->arg1 == TCP(ulp)->th_win); 2752 break; 2753 2754 case O_ESTAB: 2755 /* reject packets which have SYN only */ 2756 /* XXX should i also check for TH_ACK ? */ 2757 match = (proto == IPPROTO_TCP && offset == 0 && 2758 (TCP(ulp)->th_flags & 2759 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2760 break; 2761 2762 case O_ALTQ: { 2763 struct altq_tag *at; 2764 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2765 2766 match = 1; 2767 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2768 if (mtag != NULL) 2769 break; 2770 mtag = m_tag_get(PACKET_TAG_PF_QID, 2771 sizeof(struct altq_tag), 2772 M_NOWAIT); 2773 if (mtag == NULL) { 2774 /* 2775 * Let the packet fall back to the 2776 * default ALTQ. 2777 */ 2778 break; 2779 } 2780 at = (struct altq_tag *)(mtag+1); 2781 at->qid = altq->qid; 2782 if (is_ipv4) 2783 at->af = AF_INET; 2784 else 2785 at->af = AF_LINK; 2786 at->hdr = ip; 2787 m_tag_prepend(m, mtag); 2788 break; 2789 } 2790 2791 case O_LOG: 2792 if (fw_verbose) 2793 ipfw_log(f, hlen, args, m, oif, offset); 2794 match = 1; 2795 break; 2796 2797 case O_PROB: 2798 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2799 break; 2800 2801 case O_VERREVPATH: 2802 /* Outgoing packets automatically pass/match */ 2803 match = ((oif != NULL) || 2804 (m->m_pkthdr.rcvif == NULL) || 2805 ( 2806#ifdef INET6 2807 is_ipv6 ? 2808 verify_path6(&(args->f_id.src_ip6), 2809 m->m_pkthdr.rcvif) : 2810#endif 2811 verify_path(src_ip, m->m_pkthdr.rcvif))); 2812 break; 2813 2814 case O_VERSRCREACH: 2815 /* Outgoing packets automatically pass/match */ 2816 match = (hlen > 0 && ((oif != NULL) || 2817#ifdef INET6 2818 is_ipv6 ? 2819 verify_path6(&(args->f_id.src_ip6), 2820 NULL) : 2821#endif 2822 verify_path(src_ip, NULL))); 2823 break; 2824 2825 case O_ANTISPOOF: 2826 /* Outgoing packets automatically pass/match */ 2827 if (oif == NULL && hlen > 0 && 2828 ( (is_ipv4 && in_localaddr(src_ip)) 2829#ifdef INET6 2830 || (is_ipv6 && 2831 in6_localaddr(&(args->f_id.src_ip6))) 2832#endif 2833 )) 2834 match = 2835#ifdef INET6 2836 is_ipv6 ? verify_path6( 2837 &(args->f_id.src_ip6), 2838 m->m_pkthdr.rcvif) : 2839#endif 2840 verify_path(src_ip, 2841 m->m_pkthdr.rcvif); 2842 else 2843 match = 1; 2844 break; 2845 2846 case O_IPSEC: 2847#ifdef FAST_IPSEC 2848 match = (m_tag_find(m, 2849 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2850#endif 2851#ifdef IPSEC 2852 match = (ipsec_getnhist(m) != 0); 2853#endif 2854 /* otherwise no match */ 2855 break; 2856 2857#ifdef INET6 2858 case O_IP6_SRC: 2859 match = is_ipv6 && 2860 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2861 &((ipfw_insn_ip6 *)cmd)->addr6); 2862 break; 2863 2864 case O_IP6_DST: 2865 match = is_ipv6 && 2866 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2867 &((ipfw_insn_ip6 *)cmd)->addr6); 2868 break; 2869 case O_IP6_SRC_MASK: 2870 if (is_ipv6) { 2871 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2872 struct in6_addr p = args->f_id.src_ip6; 2873 2874 APPLY_MASK(&p, &te->mask6); 2875 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2876 } 2877 break; 2878 2879 case O_IP6_DST_MASK: 2880 if (is_ipv6) { 2881 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2882 struct in6_addr p = args->f_id.dst_ip6; 2883 2884 APPLY_MASK(&p, &te->mask6); 2885 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2886 } 2887 break; 2888 2889 case O_IP6_SRC_ME: 2890 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2891 break; 2892 2893 case O_IP6_DST_ME: 2894 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2895 break; 2896 2897 case O_FLOW6ID: 2898 match = is_ipv6 && 2899 flow6id_match(args->f_id.flow_id6, 2900 (ipfw_insn_u32 *) cmd); 2901 break; 2902 2903 case O_EXT_HDR: 2904 match = is_ipv6 && 2905 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2906 break; 2907 2908 case O_IP6: 2909 match = is_ipv6; 2910 break; 2911#endif 2912 2913 case O_IP4: 2914 match = is_ipv4; 2915 break; 2916 2917 case O_TAG: { 2918 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2919 tablearg : cmd->arg1; 2920 2921 /* Packet is already tagged with this tag? */ 2922 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2923 2924 /* We have `untag' action when F_NOT flag is 2925 * present. And we must remove this mtag from 2926 * mbuf and reset `match' to zero (`match' will 2927 * be inversed later). 2928 * Otherwise we should allocate new mtag and 2929 * push it into mbuf. 2930 */ 2931 if (cmd->len & F_NOT) { /* `untag' action */ 2932 if (mtag != NULL) 2933 m_tag_delete(m, mtag); 2934 } else if (mtag == NULL) { 2935 if ((mtag = m_tag_alloc(MTAG_IPFW, 2936 tag, 0, M_NOWAIT)) != NULL) 2937 m_tag_prepend(m, mtag); 2938 } 2939 match = (cmd->len & F_NOT) ? 0: 1; 2940 break; 2941 } 2942 2943 case O_TAGGED: { 2944 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2945 tablearg : cmd->arg1; 2946 2947 if (cmdlen == 1) { 2948 match = m_tag_locate(m, MTAG_IPFW, 2949 tag, NULL) != NULL; 2950 break; 2951 } 2952 2953 /* we have ranges */ 2954 for (mtag = m_tag_first(m); 2955 mtag != NULL && !match; 2956 mtag = m_tag_next(m, mtag)) { 2957 uint16_t *p; 2958 int i; 2959 2960 if (mtag->m_tag_cookie != MTAG_IPFW) 2961 continue; 2962 2963 p = ((ipfw_insn_u16 *)cmd)->ports; 2964 i = cmdlen - 1; 2965 for(; !match && i > 0; i--, p += 2) 2966 match = 2967 mtag->m_tag_id >= p[0] && 2968 mtag->m_tag_id <= p[1]; 2969 } 2970 break; 2971 } 2972 2973 /* 2974 * The second set of opcodes represents 'actions', 2975 * i.e. the terminal part of a rule once the packet 2976 * matches all previous patterns. 2977 * Typically there is only one action for each rule, 2978 * and the opcode is stored at the end of the rule 2979 * (but there are exceptions -- see below). 2980 * 2981 * In general, here we set retval and terminate the 2982 * outer loop (would be a 'break 3' in some language, 2983 * but we need to do a 'goto done'). 2984 * 2985 * Exceptions: 2986 * O_COUNT and O_SKIPTO actions: 2987 * instead of terminating, we jump to the next rule 2988 * ('goto next_rule', equivalent to a 'break 2'), 2989 * or to the SKIPTO target ('goto again' after 2990 * having set f, cmd and l), respectively. 2991 * 2992 * O_TAG, O_LOG and O_ALTQ action parameters: 2993 * perform some action and set match = 1; 2994 * 2995 * O_LIMIT and O_KEEP_STATE: these opcodes are 2996 * not real 'actions', and are stored right 2997 * before the 'action' part of the rule. 2998 * These opcodes try to install an entry in the 2999 * state tables; if successful, we continue with 3000 * the next opcode (match=1; break;), otherwise 3001 * the packet * must be dropped 3002 * ('goto done' after setting retval); 3003 * 3004 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3005 * cause a lookup of the state table, and a jump 3006 * to the 'action' part of the parent rule 3007 * ('goto check_body') if an entry is found, or 3008 * (CHECK_STATE only) a jump to the next rule if 3009 * the entry is not found ('goto next_rule'). 3010 * The result of the lookup is cached to make 3011 * further instances of these opcodes are 3012 * effectively NOPs. 3013 */ 3014 case O_LIMIT: 3015 case O_KEEP_STATE: 3016 if (install_state(f, 3017 (ipfw_insn_limit *)cmd, args, tablearg)) { 3018 retval = IP_FW_DENY; 3019 goto done; /* error/limit violation */ 3020 } 3021 match = 1; 3022 break; 3023 3024 case O_PROBE_STATE: 3025 case O_CHECK_STATE: 3026 /* 3027 * dynamic rules are checked at the first 3028 * keep-state or check-state occurrence, 3029 * with the result being stored in dyn_dir. 3030 * The compiler introduces a PROBE_STATE 3031 * instruction for us when we have a 3032 * KEEP_STATE (because PROBE_STATE needs 3033 * to be run first). 3034 */ 3035 if (dyn_dir == MATCH_UNKNOWN && 3036 (q = lookup_dyn_rule(&args->f_id, 3037 &dyn_dir, proto == IPPROTO_TCP ? 3038 TCP(ulp) : NULL)) 3039 != NULL) { 3040 /* 3041 * Found dynamic entry, update stats 3042 * and jump to the 'action' part of 3043 * the parent rule. 3044 */ 3045 q->pcnt++; 3046 q->bcnt += pktlen; 3047 f = q->rule; 3048 cmd = ACTION_PTR(f); 3049 l = f->cmd_len - f->act_ofs; 3050 IPFW_DYN_UNLOCK(); 3051 goto check_body; 3052 } 3053 /* 3054 * Dynamic entry not found. If CHECK_STATE, 3055 * skip to next rule, if PROBE_STATE just 3056 * ignore and continue with next opcode. 3057 */ 3058 if (cmd->opcode == O_CHECK_STATE) 3059 goto next_rule; 3060 match = 1; 3061 break; 3062 3063 case O_ACCEPT: 3064 retval = 0; /* accept */ 3065 goto done; 3066 3067 case O_PIPE: 3068 case O_QUEUE: 3069 args->rule = f; /* report matching rule */ 3070 if (cmd->arg1 == IP_FW_TABLEARG) 3071 args->cookie = tablearg; 3072 else 3073 args->cookie = cmd->arg1; 3074 retval = IP_FW_DUMMYNET; 3075 goto done; 3076 3077 case O_DIVERT: 3078 case O_TEE: { 3079 struct divert_tag *dt; 3080 3081 if (args->eh) /* not on layer 2 */ 3082 break; 3083 mtag = m_tag_get(PACKET_TAG_DIVERT, 3084 sizeof(struct divert_tag), 3085 M_NOWAIT); 3086 if (mtag == NULL) { 3087 /* XXX statistic */ 3088 /* drop packet */ 3089 IPFW_RUNLOCK(chain); 3090 return (IP_FW_DENY); 3091 } 3092 dt = (struct divert_tag *)(mtag+1); 3093 dt->cookie = f->rulenum; 3094 if (cmd->arg1 == IP_FW_TABLEARG) 3095 dt->info = tablearg; 3096 else 3097 dt->info = cmd->arg1; 3098 m_tag_prepend(m, mtag); 3099 retval = (cmd->opcode == O_DIVERT) ? 3100 IP_FW_DIVERT : IP_FW_TEE; 3101 goto done; 3102 } 3103 3104 case O_COUNT: 3105 case O_SKIPTO: 3106 f->pcnt++; /* update stats */ 3107 f->bcnt += pktlen; 3108 f->timestamp = time_uptime; 3109 if (cmd->opcode == O_COUNT) 3110 goto next_rule; 3111 /* handle skipto */ 3112 if (f->next_rule == NULL) 3113 lookup_next_rule(f); 3114 f = f->next_rule; 3115 goto again; 3116 3117 case O_REJECT: 3118 /* 3119 * Drop the packet and send a reject notice 3120 * if the packet is not ICMP (or is an ICMP 3121 * query), and it is not multicast/broadcast. 3122 */ 3123 if (hlen > 0 && is_ipv4 && offset == 0 && 3124 (proto != IPPROTO_ICMP || 3125 is_icmp_query(ICMP(ulp))) && 3126 !(m->m_flags & (M_BCAST|M_MCAST)) && 3127 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3128 send_reject(args, cmd->arg1, ip_len); 3129 m = args->m; 3130 } 3131 /* FALLTHROUGH */ 3132#ifdef INET6 3133 case O_UNREACH6: 3134 if (hlen > 0 && is_ipv6 && 3135 ((offset & IP6F_OFF_MASK) == 0) && 3136 (proto != IPPROTO_ICMPV6 || 3137 (is_icmp6_query(args->f_id.flags) == 1)) && 3138 !(m->m_flags & (M_BCAST|M_MCAST)) && 3139 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3140 send_reject6(args, cmd->arg1, hlen); 3141 m = args->m; 3142 } 3143 /* FALLTHROUGH */ 3144#endif 3145 case O_DENY: 3146 retval = IP_FW_DENY; 3147 goto done; 3148 3149 case O_FORWARD_IP: 3150 if (args->eh) /* not valid on layer2 pkts */ 3151 break; 3152 if (!q || dyn_dir == MATCH_FORWARD) 3153 args->next_hop = 3154 &((ipfw_insn_sa *)cmd)->sa; 3155 retval = IP_FW_PASS; 3156 goto done; 3157 3158 case O_NETGRAPH: 3159 case O_NGTEE: 3160 args->rule = f; /* report matching rule */ 3161 if (cmd->arg1 == IP_FW_TABLEARG) 3162 args->cookie = tablearg; 3163 else 3164 args->cookie = cmd->arg1; 3165 retval = (cmd->opcode == O_NETGRAPH) ? 3166 IP_FW_NETGRAPH : IP_FW_NGTEE; 3167 goto done; 3168 3169 default: 3170 panic("-- unknown opcode %d\n", cmd->opcode); 3171 } /* end of switch() on opcodes */ 3172 3173 if (cmd->len & F_NOT) 3174 match = !match; 3175 3176 if (match) { 3177 if (cmd->len & F_OR) 3178 skip_or = 1; 3179 } else { 3180 if (!(cmd->len & F_OR)) /* not an OR block, */ 3181 break; /* try next rule */ 3182 } 3183 3184 } /* end of inner for, scan opcodes */ 3185 3186next_rule:; /* try next rule */ 3187 3188 } /* end of outer for, scan rules */ 3189 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3190 IPFW_RUNLOCK(chain); 3191 return (IP_FW_DENY); 3192 3193done: 3194 /* Update statistics */ 3195 f->pcnt++; 3196 f->bcnt += pktlen; 3197 f->timestamp = time_uptime; 3198 IPFW_RUNLOCK(chain); 3199 return (retval); 3200 3201pullup_failed: 3202 if (fw_verbose) 3203 printf("ipfw: pullup failed\n"); 3204 return (IP_FW_DENY); 3205} 3206 3207/* 3208 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3209 * These will be reconstructed on the fly as packets are matched. 3210 */ 3211static void 3212flush_rule_ptrs(struct ip_fw_chain *chain) 3213{ 3214 struct ip_fw *rule; 3215 3216 IPFW_WLOCK_ASSERT(chain); 3217 3218 for (rule = chain->rules; rule; rule = rule->next) 3219 rule->next_rule = NULL; 3220} 3221 3222/* 3223 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3224 * possibly create a rule number and add the rule to the list. 3225 * Update the rule_number in the input struct so the caller knows it as well. 3226 */ 3227static int 3228add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3229{ 3230 struct ip_fw *rule, *f, *prev; 3231 int l = RULESIZE(input_rule); 3232 3233 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3234 return (EINVAL); 3235 3236 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3237 if (rule == NULL) 3238 return (ENOSPC); 3239 3240 bcopy(input_rule, rule, l); 3241 3242 rule->next = NULL; 3243 rule->next_rule = NULL; 3244 3245 rule->pcnt = 0; 3246 rule->bcnt = 0; 3247 rule->timestamp = 0; 3248 3249 IPFW_WLOCK(chain); 3250 3251 if (chain->rules == NULL) { /* default rule */ 3252 chain->rules = rule; 3253 goto done; 3254 } 3255 3256 /* 3257 * If rulenum is 0, find highest numbered rule before the 3258 * default rule, and add autoinc_step 3259 */ 3260 if (autoinc_step < 1) 3261 autoinc_step = 1; 3262 else if (autoinc_step > 1000) 3263 autoinc_step = 1000; 3264 if (rule->rulenum == 0) { 3265 /* 3266 * locate the highest numbered rule before default 3267 */ 3268 for (f = chain->rules; f; f = f->next) { 3269 if (f->rulenum == IPFW_DEFAULT_RULE) 3270 break; 3271 rule->rulenum = f->rulenum; 3272 } 3273 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3274 rule->rulenum += autoinc_step; 3275 input_rule->rulenum = rule->rulenum; 3276 } 3277 3278 /* 3279 * Now insert the new rule in the right place in the sorted list. 3280 */ 3281 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3282 if (f->rulenum > rule->rulenum) { /* found the location */ 3283 if (prev) { 3284 rule->next = f; 3285 prev->next = rule; 3286 } else { /* head insert */ 3287 rule->next = chain->rules; 3288 chain->rules = rule; 3289 } 3290 break; 3291 } 3292 } 3293 flush_rule_ptrs(chain); 3294done: 3295 static_count++; 3296 static_len += l; 3297 IPFW_WUNLOCK(chain); 3298 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3299 rule->rulenum, static_count);) 3300 return (0); 3301} 3302 3303/** 3304 * Remove a static rule (including derived * dynamic rules) 3305 * and place it on the ``reap list'' for later reclamation. 3306 * The caller is in charge of clearing rule pointers to avoid 3307 * dangling pointers. 3308 * @return a pointer to the next entry. 3309 * Arguments are not checked, so they better be correct. 3310 */ 3311static struct ip_fw * 3312remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3313{ 3314 struct ip_fw *n; 3315 int l = RULESIZE(rule); 3316 3317 IPFW_WLOCK_ASSERT(chain); 3318 3319 n = rule->next; 3320 IPFW_DYN_LOCK(); 3321 remove_dyn_rule(rule, NULL /* force removal */); 3322 IPFW_DYN_UNLOCK(); 3323 if (prev == NULL) 3324 chain->rules = n; 3325 else 3326 prev->next = n; 3327 static_count--; 3328 static_len -= l; 3329 3330 rule->next = chain->reap; 3331 chain->reap = rule; 3332 3333 return n; 3334} 3335 3336/** 3337 * Reclaim storage associated with a list of rules. This is 3338 * typically the list created using remove_rule. 3339 */ 3340static void 3341reap_rules(struct ip_fw *head) 3342{ 3343 struct ip_fw *rule; 3344 3345 while ((rule = head) != NULL) { 3346 head = head->next; 3347 if (DUMMYNET_LOADED) 3348 ip_dn_ruledel_ptr(rule); 3349 free(rule, M_IPFW); 3350 } 3351} 3352 3353/* 3354 * Remove all rules from a chain (except rules in set RESVD_SET 3355 * unless kill_default = 1). The caller is responsible for 3356 * reclaiming storage for the rules left in chain->reap. 3357 */ 3358static void 3359free_chain(struct ip_fw_chain *chain, int kill_default) 3360{ 3361 struct ip_fw *prev, *rule; 3362 3363 IPFW_WLOCK_ASSERT(chain); 3364 3365 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3366 for (prev = NULL, rule = chain->rules; rule ; ) 3367 if (kill_default || rule->set != RESVD_SET) 3368 rule = remove_rule(chain, rule, prev); 3369 else { 3370 prev = rule; 3371 rule = rule->next; 3372 } 3373} 3374 3375/** 3376 * Remove all rules with given number, and also do set manipulation. 3377 * Assumes chain != NULL && *chain != NULL. 3378 * 3379 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3380 * the next 8 bits are the new set, the top 8 bits are the command: 3381 * 3382 * 0 delete rules with given number 3383 * 1 delete rules with given set number 3384 * 2 move rules with given number to new set 3385 * 3 move rules with given set number to new set 3386 * 4 swap sets with given numbers 3387 */ 3388static int 3389del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3390{ 3391 struct ip_fw *prev = NULL, *rule; 3392 u_int16_t rulenum; /* rule or old_set */ 3393 u_int8_t cmd, new_set; 3394 3395 rulenum = arg & 0xffff; 3396 cmd = (arg >> 24) & 0xff; 3397 new_set = (arg >> 16) & 0xff; 3398 3399 if (cmd > 4) 3400 return EINVAL; 3401 if (new_set > RESVD_SET) 3402 return EINVAL; 3403 if (cmd == 0 || cmd == 2) { 3404 if (rulenum >= IPFW_DEFAULT_RULE) 3405 return EINVAL; 3406 } else { 3407 if (rulenum > RESVD_SET) /* old_set */ 3408 return EINVAL; 3409 } 3410 3411 IPFW_WLOCK(chain); 3412 rule = chain->rules; 3413 chain->reap = NULL; 3414 switch (cmd) { 3415 case 0: /* delete rules with given number */ 3416 /* 3417 * locate first rule to delete 3418 */ 3419 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3420 ; 3421 if (rule->rulenum != rulenum) { 3422 IPFW_WUNLOCK(chain); 3423 return EINVAL; 3424 } 3425 3426 /* 3427 * flush pointers outside the loop, then delete all matching 3428 * rules. prev remains the same throughout the cycle. 3429 */ 3430 flush_rule_ptrs(chain); 3431 while (rule->rulenum == rulenum) 3432 rule = remove_rule(chain, rule, prev); 3433 break; 3434 3435 case 1: /* delete all rules with given set number */ 3436 flush_rule_ptrs(chain); 3437 rule = chain->rules; 3438 while (rule->rulenum < IPFW_DEFAULT_RULE) 3439 if (rule->set == rulenum) 3440 rule = remove_rule(chain, rule, prev); 3441 else { 3442 prev = rule; 3443 rule = rule->next; 3444 } 3445 break; 3446 3447 case 2: /* move rules with given number to new set */ 3448 rule = chain->rules; 3449 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3450 if (rule->rulenum == rulenum) 3451 rule->set = new_set; 3452 break; 3453 3454 case 3: /* move rules with given set number to new set */ 3455 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3456 if (rule->set == rulenum) 3457 rule->set = new_set; 3458 break; 3459 3460 case 4: /* swap two sets */ 3461 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3462 if (rule->set == rulenum) 3463 rule->set = new_set; 3464 else if (rule->set == new_set) 3465 rule->set = rulenum; 3466 break; 3467 } 3468 /* 3469 * Look for rules to reclaim. We grab the list before 3470 * releasing the lock then reclaim them w/o the lock to 3471 * avoid a LOR with dummynet. 3472 */ 3473 rule = chain->reap; 3474 chain->reap = NULL; 3475 IPFW_WUNLOCK(chain); 3476 if (rule) 3477 reap_rules(rule); 3478 return 0; 3479} 3480 3481/* 3482 * Clear counters for a specific rule. 3483 * The enclosing "table" is assumed locked. 3484 */ 3485static void 3486clear_counters(struct ip_fw *rule, int log_only) 3487{ 3488 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3489 3490 if (log_only == 0) { 3491 rule->bcnt = rule->pcnt = 0; 3492 rule->timestamp = 0; 3493 } 3494 if (l->o.opcode == O_LOG) 3495 l->log_left = l->max_log; 3496} 3497 3498/** 3499 * Reset some or all counters on firewall rules. 3500 * @arg frwl is null to clear all entries, or contains a specific 3501 * rule number. 3502 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3503 */ 3504static int 3505zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3506{ 3507 struct ip_fw *rule; 3508 char *msg; 3509 3510 IPFW_WLOCK(chain); 3511 if (rulenum == 0) { 3512 norule_counter = 0; 3513 for (rule = chain->rules; rule; rule = rule->next) 3514 clear_counters(rule, log_only); 3515 msg = log_only ? "ipfw: All logging counts reset.\n" : 3516 "ipfw: Accounting cleared.\n"; 3517 } else { 3518 int cleared = 0; 3519 /* 3520 * We can have multiple rules with the same number, so we 3521 * need to clear them all. 3522 */ 3523 for (rule = chain->rules; rule; rule = rule->next) 3524 if (rule->rulenum == rulenum) { 3525 while (rule && rule->rulenum == rulenum) { 3526 clear_counters(rule, log_only); 3527 rule = rule->next; 3528 } 3529 cleared = 1; 3530 break; 3531 } 3532 if (!cleared) { /* we did not find any matching rules */ 3533 IPFW_WUNLOCK(chain); 3534 return (EINVAL); 3535 } 3536 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3537 "ipfw: Entry %d cleared.\n"; 3538 } 3539 IPFW_WUNLOCK(chain); 3540 3541 if (fw_verbose) 3542 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3543 return (0); 3544} 3545 3546/* 3547 * Check validity of the structure before insert. 3548 * Fortunately rules are simple, so this mostly need to check rule sizes. 3549 */ 3550static int 3551check_ipfw_struct(struct ip_fw *rule, int size) 3552{ 3553 int l, cmdlen = 0; 3554 int have_action=0; 3555 ipfw_insn *cmd; 3556 3557 if (size < sizeof(*rule)) { 3558 printf("ipfw: rule too short\n"); 3559 return (EINVAL); 3560 } 3561 /* first, check for valid size */ 3562 l = RULESIZE(rule); 3563 if (l != size) { 3564 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3565 return (EINVAL); 3566 } 3567 if (rule->act_ofs >= rule->cmd_len) { 3568 printf("ipfw: bogus action offset (%u > %u)\n", 3569 rule->act_ofs, rule->cmd_len - 1); 3570 return (EINVAL); 3571 } 3572 /* 3573 * Now go for the individual checks. Very simple ones, basically only 3574 * instruction sizes. 3575 */ 3576 for (l = rule->cmd_len, cmd = rule->cmd ; 3577 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3578 cmdlen = F_LEN(cmd); 3579 if (cmdlen > l) { 3580 printf("ipfw: opcode %d size truncated\n", 3581 cmd->opcode); 3582 return EINVAL; 3583 } 3584 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3585 switch (cmd->opcode) { 3586 case O_PROBE_STATE: 3587 case O_KEEP_STATE: 3588 case O_PROTO: 3589 case O_IP_SRC_ME: 3590 case O_IP_DST_ME: 3591 case O_LAYER2: 3592 case O_IN: 3593 case O_FRAG: 3594 case O_DIVERTED: 3595 case O_IPOPT: 3596 case O_IPTOS: 3597 case O_IPPRECEDENCE: 3598 case O_IPVER: 3599 case O_TCPWIN: 3600 case O_TCPFLAGS: 3601 case O_TCPOPTS: 3602 case O_ESTAB: 3603 case O_VERREVPATH: 3604 case O_VERSRCREACH: 3605 case O_ANTISPOOF: 3606 case O_IPSEC: 3607#ifdef INET6 3608 case O_IP6_SRC_ME: 3609 case O_IP6_DST_ME: 3610 case O_EXT_HDR: 3611 case O_IP6: 3612#endif 3613 case O_IP4: 3614 case O_TAG: 3615 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3616 goto bad_size; 3617 break; 3618 3619 case O_UID: 3620 case O_GID: 3621 case O_JAIL: 3622 case O_IP_SRC: 3623 case O_IP_DST: 3624 case O_TCPSEQ: 3625 case O_TCPACK: 3626 case O_PROB: 3627 case O_ICMPTYPE: 3628 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3629 goto bad_size; 3630 break; 3631 3632 case O_LIMIT: 3633 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3634 goto bad_size; 3635 break; 3636 3637 case O_LOG: 3638 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3639 goto bad_size; 3640 3641 ((ipfw_insn_log *)cmd)->log_left = 3642 ((ipfw_insn_log *)cmd)->max_log; 3643 3644 break; 3645 3646 case O_IP_SRC_MASK: 3647 case O_IP_DST_MASK: 3648 /* only odd command lengths */ 3649 if ( !(cmdlen & 1) || cmdlen > 31) 3650 goto bad_size; 3651 break; 3652 3653 case O_IP_SRC_SET: 3654 case O_IP_DST_SET: 3655 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3656 printf("ipfw: invalid set size %d\n", 3657 cmd->arg1); 3658 return EINVAL; 3659 } 3660 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3661 (cmd->arg1+31)/32 ) 3662 goto bad_size; 3663 break; 3664 3665 case O_IP_SRC_LOOKUP: 3666 case O_IP_DST_LOOKUP: 3667 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3668 printf("ipfw: invalid table number %d\n", 3669 cmd->arg1); 3670 return (EINVAL); 3671 } 3672 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3673 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3674 goto bad_size; 3675 break; 3676 3677 case O_MACADDR2: 3678 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3679 goto bad_size; 3680 break; 3681 3682 case O_NOP: 3683 case O_IPID: 3684 case O_IPTTL: 3685 case O_IPLEN: 3686 case O_TCPDATALEN: 3687 case O_TAGGED: 3688 if (cmdlen < 1 || cmdlen > 31) 3689 goto bad_size; 3690 break; 3691 3692 case O_MAC_TYPE: 3693 case O_IP_SRCPORT: 3694 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3695 if (cmdlen < 2 || cmdlen > 31) 3696 goto bad_size; 3697 break; 3698 3699 case O_RECV: 3700 case O_XMIT: 3701 case O_VIA: 3702 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3703 goto bad_size; 3704 break; 3705 3706 case O_ALTQ: 3707 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3708 goto bad_size; 3709 break; 3710 3711 case O_PIPE: 3712 case O_QUEUE: 3713 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3714 goto bad_size; 3715 goto check_action; 3716 3717 case O_FORWARD_IP: 3718#ifdef IPFIREWALL_FORWARD 3719 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3720 goto bad_size; 3721 goto check_action; 3722#else 3723 return EINVAL; 3724#endif 3725 3726 case O_DIVERT: 3727 case O_TEE: 3728 if (ip_divert_ptr == NULL) 3729 return EINVAL; 3730 else 3731 goto check_size; 3732 case O_NETGRAPH: 3733 case O_NGTEE: 3734 if (!NG_IPFW_LOADED) 3735 return EINVAL; 3736 else 3737 goto check_size; 3738 case O_FORWARD_MAC: /* XXX not implemented yet */ 3739 case O_CHECK_STATE: 3740 case O_COUNT: 3741 case O_ACCEPT: 3742 case O_DENY: 3743 case O_REJECT: 3744#ifdef INET6 3745 case O_UNREACH6: 3746#endif 3747 case O_SKIPTO: 3748check_size: 3749 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3750 goto bad_size; 3751check_action: 3752 if (have_action) { 3753 printf("ipfw: opcode %d, multiple actions" 3754 " not allowed\n", 3755 cmd->opcode); 3756 return EINVAL; 3757 } 3758 have_action = 1; 3759 if (l != cmdlen) { 3760 printf("ipfw: opcode %d, action must be" 3761 " last opcode\n", 3762 cmd->opcode); 3763 return EINVAL; 3764 } 3765 break; 3766#ifdef INET6 3767 case O_IP6_SRC: 3768 case O_IP6_DST: 3769 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3770 F_INSN_SIZE(ipfw_insn)) 3771 goto bad_size; 3772 break; 3773 3774 case O_FLOW6ID: 3775 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3776 ((ipfw_insn_u32 *)cmd)->o.arg1) 3777 goto bad_size; 3778 break; 3779 3780 case O_IP6_SRC_MASK: 3781 case O_IP6_DST_MASK: 3782 if ( !(cmdlen & 1) || cmdlen > 127) 3783 goto bad_size; 3784 break; 3785 case O_ICMP6TYPE: 3786 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3787 goto bad_size; 3788 break; 3789#endif 3790 3791 default: 3792 switch (cmd->opcode) { 3793#ifndef INET6 3794 case O_IP6_SRC_ME: 3795 case O_IP6_DST_ME: 3796 case O_EXT_HDR: 3797 case O_IP6: 3798 case O_UNREACH6: 3799 case O_IP6_SRC: 3800 case O_IP6_DST: 3801 case O_FLOW6ID: 3802 case O_IP6_SRC_MASK: 3803 case O_IP6_DST_MASK: 3804 case O_ICMP6TYPE: 3805 printf("ipfw: no IPv6 support in kernel\n"); 3806 return EPROTONOSUPPORT; 3807#endif 3808 default: 3809 printf("ipfw: opcode %d, unknown opcode\n", 3810 cmd->opcode); 3811 return EINVAL; 3812 } 3813 } 3814 } 3815 if (have_action == 0) { 3816 printf("ipfw: missing action\n"); 3817 return EINVAL; 3818 } 3819 return 0; 3820 3821bad_size: 3822 printf("ipfw: opcode %d size %d wrong\n", 3823 cmd->opcode, cmdlen); 3824 return EINVAL; 3825} 3826 3827/* 3828 * Copy the static and dynamic rules to the supplied buffer 3829 * and return the amount of space actually used. 3830 */ 3831static size_t 3832ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3833{ 3834 char *bp = buf; 3835 char *ep = bp + space; 3836 struct ip_fw *rule; 3837 int i; 3838 3839 /* XXX this can take a long time and locking will block packet flow */ 3840 IPFW_RLOCK(chain); 3841 for (rule = chain->rules; rule ; rule = rule->next) { 3842 /* 3843 * Verify the entry fits in the buffer in case the 3844 * rules changed between calculating buffer space and 3845 * now. This would be better done using a generation 3846 * number but should suffice for now. 3847 */ 3848 i = RULESIZE(rule); 3849 if (bp + i <= ep) { 3850 bcopy(rule, bp, i); 3851 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3852 sizeof(set_disable)); 3853 bp += i; 3854 } 3855 } 3856 IPFW_RUNLOCK(chain); 3857 if (ipfw_dyn_v) { 3858 ipfw_dyn_rule *p, *last = NULL; 3859 3860 IPFW_DYN_LOCK(); 3861 for (i = 0 ; i < curr_dyn_buckets; i++) 3862 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3863 if (bp + sizeof *p <= ep) { 3864 ipfw_dyn_rule *dst = 3865 (ipfw_dyn_rule *)bp; 3866 bcopy(p, dst, sizeof *p); 3867 bcopy(&(p->rule->rulenum), &(dst->rule), 3868 sizeof(p->rule->rulenum)); 3869 /* 3870 * store a non-null value in "next". 3871 * The userland code will interpret a 3872 * NULL here as a marker 3873 * for the last dynamic rule. 3874 */ 3875 bcopy(&dst, &dst->next, sizeof(dst)); 3876 last = dst; 3877 dst->expire = 3878 TIME_LEQ(dst->expire, time_uptime) ? 3879 0 : dst->expire - time_uptime ; 3880 bp += sizeof(ipfw_dyn_rule); 3881 } 3882 } 3883 IPFW_DYN_UNLOCK(); 3884 if (last != NULL) /* mark last dynamic rule */ 3885 bzero(&last->next, sizeof(last)); 3886 } 3887 return (bp - (char *)buf); 3888} 3889 3890 3891/** 3892 * {set|get}sockopt parser. 3893 */ 3894static int 3895ipfw_ctl(struct sockopt *sopt) 3896{ 3897#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3898 int error, rule_num; 3899 size_t size; 3900 struct ip_fw *buf, *rule; 3901 u_int32_t rulenum[2]; 3902 3903 error = suser(sopt->sopt_td); 3904 if (error) 3905 return (error); 3906 3907 /* 3908 * Disallow modifications in really-really secure mode, but still allow 3909 * the logging counters to be reset. 3910 */ 3911 if (sopt->sopt_name == IP_FW_ADD || 3912 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3913 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3914 if (error) 3915 return (error); 3916 } 3917 3918 error = 0; 3919 3920 switch (sopt->sopt_name) { 3921 case IP_FW_GET: 3922 /* 3923 * pass up a copy of the current rules. Static rules 3924 * come first (the last of which has number IPFW_DEFAULT_RULE), 3925 * followed by a possibly empty list of dynamic rule. 3926 * The last dynamic rule has NULL in the "next" field. 3927 * 3928 * Note that the calculated size is used to bound the 3929 * amount of data returned to the user. The rule set may 3930 * change between calculating the size and returning the 3931 * data in which case we'll just return what fits. 3932 */ 3933 size = static_len; /* size of static rules */ 3934 if (ipfw_dyn_v) /* add size of dyn.rules */ 3935 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3936 3937 /* 3938 * XXX todo: if the user passes a short length just to know 3939 * how much room is needed, do not bother filling up the 3940 * buffer, just jump to the sooptcopyout. 3941 */ 3942 buf = malloc(size, M_TEMP, M_WAITOK); 3943 error = sooptcopyout(sopt, buf, 3944 ipfw_getrules(&layer3_chain, buf, size)); 3945 free(buf, M_TEMP); 3946 break; 3947 3948 case IP_FW_FLUSH: 3949 /* 3950 * Normally we cannot release the lock on each iteration. 3951 * We could do it here only because we start from the head all 3952 * the times so there is no risk of missing some entries. 3953 * On the other hand, the risk is that we end up with 3954 * a very inconsistent ruleset, so better keep the lock 3955 * around the whole cycle. 3956 * 3957 * XXX this code can be improved by resetting the head of 3958 * the list to point to the default rule, and then freeing 3959 * the old list without the need for a lock. 3960 */ 3961 3962 IPFW_WLOCK(&layer3_chain); 3963 layer3_chain.reap = NULL; 3964 free_chain(&layer3_chain, 0 /* keep default rule */); 3965 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3966 IPFW_WUNLOCK(&layer3_chain); 3967 if (layer3_chain.reap != NULL) 3968 reap_rules(rule); 3969 break; 3970 3971 case IP_FW_ADD: 3972 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3973 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3974 sizeof(struct ip_fw) ); 3975 if (error == 0) 3976 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3977 if (error == 0) { 3978 error = add_rule(&layer3_chain, rule); 3979 size = RULESIZE(rule); 3980 if (!error && sopt->sopt_dir == SOPT_GET) 3981 error = sooptcopyout(sopt, rule, size); 3982 } 3983 free(rule, M_TEMP); 3984 break; 3985 3986 case IP_FW_DEL: 3987 /* 3988 * IP_FW_DEL is used for deleting single rules or sets, 3989 * and (ab)used to atomically manipulate sets. Argument size 3990 * is used to distinguish between the two: 3991 * sizeof(u_int32_t) 3992 * delete single rule or set of rules, 3993 * or reassign rules (or sets) to a different set. 3994 * 2*sizeof(u_int32_t) 3995 * atomic disable/enable sets. 3996 * first u_int32_t contains sets to be disabled, 3997 * second u_int32_t contains sets to be enabled. 3998 */ 3999 error = sooptcopyin(sopt, rulenum, 4000 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4001 if (error) 4002 break; 4003 size = sopt->sopt_valsize; 4004 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4005 error = del_entry(&layer3_chain, rulenum[0]); 4006 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4007 set_disable = 4008 (set_disable | rulenum[0]) & ~rulenum[1] & 4009 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4010 else 4011 error = EINVAL; 4012 break; 4013 4014 case IP_FW_ZERO: 4015 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 4016 rule_num = 0; 4017 if (sopt->sopt_val != 0) { 4018 error = sooptcopyin(sopt, &rule_num, 4019 sizeof(int), sizeof(int)); 4020 if (error) 4021 break; 4022 } 4023 error = zero_entry(&layer3_chain, rule_num, 4024 sopt->sopt_name == IP_FW_RESETLOG); 4025 break; 4026 4027 case IP_FW_TABLE_ADD: 4028 { 4029 ipfw_table_entry ent; 4030 4031 error = sooptcopyin(sopt, &ent, 4032 sizeof(ent), sizeof(ent)); 4033 if (error) 4034 break; 4035 error = add_table_entry(&layer3_chain, ent.tbl, 4036 ent.addr, ent.masklen, ent.value); 4037 } 4038 break; 4039 4040 case IP_FW_TABLE_DEL: 4041 { 4042 ipfw_table_entry ent; 4043 4044 error = sooptcopyin(sopt, &ent, 4045 sizeof(ent), sizeof(ent)); 4046 if (error) 4047 break; 4048 error = del_table_entry(&layer3_chain, ent.tbl, 4049 ent.addr, ent.masklen); 4050 } 4051 break; 4052 4053 case IP_FW_TABLE_FLUSH: 4054 { 4055 u_int16_t tbl; 4056 4057 error = sooptcopyin(sopt, &tbl, 4058 sizeof(tbl), sizeof(tbl)); 4059 if (error) 4060 break; 4061 IPFW_WLOCK(&layer3_chain); 4062 error = flush_table(&layer3_chain, tbl); 4063 IPFW_WUNLOCK(&layer3_chain); 4064 } 4065 break; 4066 4067 case IP_FW_TABLE_GETSIZE: 4068 { 4069 u_int32_t tbl, cnt; 4070 4071 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4072 sizeof(tbl)))) 4073 break; 4074 IPFW_RLOCK(&layer3_chain); 4075 error = count_table(&layer3_chain, tbl, &cnt); 4076 IPFW_RUNLOCK(&layer3_chain); 4077 if (error) 4078 break; 4079 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4080 } 4081 break; 4082 4083 case IP_FW_TABLE_LIST: 4084 { 4085 ipfw_table *tbl; 4086 4087 if (sopt->sopt_valsize < sizeof(*tbl)) { 4088 error = EINVAL; 4089 break; 4090 } 4091 size = sopt->sopt_valsize; 4092 tbl = malloc(size, M_TEMP, M_WAITOK); 4093 if (tbl == NULL) { 4094 error = ENOMEM; 4095 break; 4096 } 4097 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4098 if (error) { 4099 free(tbl, M_TEMP); 4100 break; 4101 } 4102 tbl->size = (size - sizeof(*tbl)) / 4103 sizeof(ipfw_table_entry); 4104 IPFW_RLOCK(&layer3_chain); 4105 error = dump_table(&layer3_chain, tbl); 4106 IPFW_RUNLOCK(&layer3_chain); 4107 if (error) { 4108 free(tbl, M_TEMP); 4109 break; 4110 } 4111 error = sooptcopyout(sopt, tbl, size); 4112 free(tbl, M_TEMP); 4113 } 4114 break; 4115 4116 default: 4117 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4118 error = EINVAL; 4119 } 4120 4121 return (error); 4122#undef RULE_MAXSIZE 4123} 4124 4125/** 4126 * dummynet needs a reference to the default rule, because rules can be 4127 * deleted while packets hold a reference to them. When this happens, 4128 * dummynet changes the reference to the default rule (it could well be a 4129 * NULL pointer, but this way we do not need to check for the special 4130 * case, plus here he have info on the default behaviour). 4131 */ 4132struct ip_fw *ip_fw_default_rule; 4133 4134/* 4135 * This procedure is only used to handle keepalives. It is invoked 4136 * every dyn_keepalive_period 4137 */ 4138static void 4139ipfw_tick(void * __unused unused) 4140{ 4141 struct mbuf *m0, *m, *mnext, **mtailp; 4142 int i; 4143 ipfw_dyn_rule *q; 4144 4145 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4146 goto done; 4147 4148 /* 4149 * We make a chain of packets to go out here -- not deferring 4150 * until after we drop the IPFW dynamic rule lock would result 4151 * in a lock order reversal with the normal packet input -> ipfw 4152 * call stack. 4153 */ 4154 m0 = NULL; 4155 mtailp = &m0; 4156 IPFW_DYN_LOCK(); 4157 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4158 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4159 if (q->dyn_type == O_LIMIT_PARENT) 4160 continue; 4161 if (q->id.proto != IPPROTO_TCP) 4162 continue; 4163 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4164 continue; 4165 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4166 q->expire)) 4167 continue; /* too early */ 4168 if (TIME_LEQ(q->expire, time_uptime)) 4169 continue; /* too late, rule expired */ 4170 4171 *mtailp = send_pkt(&(q->id), q->ack_rev - 1, 4172 q->ack_fwd, TH_SYN); 4173 if (*mtailp != NULL) 4174 mtailp = &(*mtailp)->m_nextpkt; 4175 *mtailp = send_pkt(&(q->id), q->ack_fwd - 1, 4176 q->ack_rev, 0); 4177 if (*mtailp != NULL) 4178 mtailp = &(*mtailp)->m_nextpkt; 4179 } 4180 } 4181 IPFW_DYN_UNLOCK(); 4182 for (m = mnext = m0; m != NULL; m = mnext) { 4183 mnext = m->m_nextpkt; 4184 m->m_nextpkt = NULL; 4185 ip_output(m, NULL, NULL, 0, NULL, NULL); 4186 } 4187done: 4188 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4189} 4190 4191int 4192ipfw_init(void) 4193{ 4194 struct ip_fw default_rule; 4195 int error; 4196 4197#ifdef INET6 4198 /* Setup IPv6 fw sysctl tree. */ 4199 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4200 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4201 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4202 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4203 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4204 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4205 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4206 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4207 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4208 &fw_deny_unknown_exthdrs, 0, 4209 "Deny packets with unknown IPv6 Extension Headers"); 4210#endif 4211 4212 layer3_chain.rules = NULL; 4213 IPFW_LOCK_INIT(&layer3_chain); 4214 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4215 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4216 UMA_ALIGN_PTR, 0); 4217 IPFW_DYN_LOCK_INIT(); 4218 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4219 4220 bzero(&default_rule, sizeof default_rule); 4221 4222 default_rule.act_ofs = 0; 4223 default_rule.rulenum = IPFW_DEFAULT_RULE; 4224 default_rule.cmd_len = 1; 4225 default_rule.set = RESVD_SET; 4226 4227 default_rule.cmd[0].len = 1; 4228 default_rule.cmd[0].opcode = 4229#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4230 1 ? O_ACCEPT : 4231#endif 4232 O_DENY; 4233 4234 error = add_rule(&layer3_chain, &default_rule); 4235 if (error != 0) { 4236 printf("ipfw2: error %u initializing default rule " 4237 "(support disabled)\n", error); 4238 IPFW_DYN_LOCK_DESTROY(); 4239 IPFW_LOCK_DESTROY(&layer3_chain); 4240 uma_zdestroy(ipfw_dyn_rule_zone); 4241 return (error); 4242 } 4243 4244 ip_fw_default_rule = layer3_chain.rules; 4245 printf("ipfw2 " 4246#ifdef INET6 4247 "(+ipv6) " 4248#endif 4249 "initialized, divert %s, " 4250 "rule-based forwarding " 4251#ifdef IPFIREWALL_FORWARD 4252 "enabled, " 4253#else 4254 "disabled, " 4255#endif 4256 "default to %s, logging ", 4257#ifdef IPDIVERT 4258 "enabled", 4259#else 4260 "loadable", 4261#endif 4262 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4263 4264#ifdef IPFIREWALL_VERBOSE 4265 fw_verbose = 1; 4266#endif 4267#ifdef IPFIREWALL_VERBOSE_LIMIT 4268 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4269#endif 4270 if (fw_verbose == 0) 4271 printf("disabled\n"); 4272 else if (verbose_limit == 0) 4273 printf("unlimited\n"); 4274 else 4275 printf("limited to %d packets/entry by default\n", 4276 verbose_limit); 4277 4278 error = init_tables(&layer3_chain); 4279 if (error) { 4280 IPFW_DYN_LOCK_DESTROY(); 4281 IPFW_LOCK_DESTROY(&layer3_chain); 4282 uma_zdestroy(ipfw_dyn_rule_zone); 4283 return (error); 4284 } 4285 ip_fw_ctl_ptr = ipfw_ctl; 4286 ip_fw_chk_ptr = ipfw_chk; 4287 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4288 4289 return (0); 4290} 4291 4292void 4293ipfw_destroy(void) 4294{ 4295 struct ip_fw *reap; 4296 4297 ip_fw_chk_ptr = NULL; 4298 ip_fw_ctl_ptr = NULL; 4299 callout_drain(&ipfw_timeout); 4300 IPFW_WLOCK(&layer3_chain); 4301 flush_tables(&layer3_chain); 4302 layer3_chain.reap = NULL; 4303 free_chain(&layer3_chain, 1 /* kill default rule */); 4304 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4305 IPFW_WUNLOCK(&layer3_chain); 4306 if (reap != NULL) 4307 reap_rules(reap); 4308 IPFW_DYN_LOCK_DESTROY(); 4309 uma_zdestroy(ipfw_dyn_rule_zone); 4310 IPFW_LOCK_DESTROY(&layer3_chain); 4311 4312#ifdef INET6 4313 /* Free IPv6 fw sysctl tree. */ 4314 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4315#endif 4316 4317 printf("IP firewall unloaded\n"); 4318} 4319