ip_fw2.c revision 161456
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 161456 2006-08-18 22:36:05Z julian $ 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->sa_family != AF_INET) 469 continue; 470 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 471 (ia->ifa_addr))->sin_addr.s_addr) 472 return(1); /* match */ 473 } 474 } 475 return(0); /* no match, fail ... */ 476} 477 478/* 479 * The verify_path function checks if a route to the src exists and 480 * if it is reachable via ifp (when provided). 481 * 482 * The 'verrevpath' option checks that the interface that an IP packet 483 * arrives on is the same interface that traffic destined for the 484 * packet's source address would be routed out of. The 'versrcreach' 485 * option just checks that the source address is reachable via any route 486 * (except default) in the routing table. These two are a measure to block 487 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 488 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 489 * is purposely reminiscent of the Cisco IOS command, 490 * 491 * ip verify unicast reverse-path 492 * ip verify unicast source reachable-via any 493 * 494 * which implements the same functionality. But note that syntax is 495 * misleading. The check may be performed on all IP packets whether unicast, 496 * multicast, or broadcast. 497 */ 498static int 499verify_path(struct in_addr src, struct ifnet *ifp) 500{ 501 struct route ro; 502 struct sockaddr_in *dst; 503 504 bzero(&ro, sizeof(ro)); 505 506 dst = (struct sockaddr_in *)&(ro.ro_dst); 507 dst->sin_family = AF_INET; 508 dst->sin_len = sizeof(*dst); 509 dst->sin_addr = src; 510 rtalloc_ign(&ro, RTF_CLONING); 511 512 if (ro.ro_rt == NULL) 513 return 0; 514 515 /* 516 * If ifp is provided, check for equality with rtentry. 517 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 518 * in order to pass packets injected back by if_simloop(): 519 * if useloopback == 1 routing entry (via lo0) for our own address 520 * may exist, so we need to handle routing assymetry. 521 */ 522 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 523 RTFREE(ro.ro_rt); 524 return 0; 525 } 526 527 /* if no ifp provided, check if rtentry is not default route */ 528 if (ifp == NULL && 529 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 530 RTFREE(ro.ro_rt); 531 return 0; 532 } 533 534 /* or if this is a blackhole/reject route */ 535 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 536 RTFREE(ro.ro_rt); 537 return 0; 538 } 539 540 /* found valid route */ 541 RTFREE(ro.ro_rt); 542 return 1; 543} 544 545#ifdef INET6 546/* 547 * ipv6 specific rules here... 548 */ 549static __inline int 550icmp6type_match (int type, ipfw_insn_u32 *cmd) 551{ 552 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 553} 554 555static int 556flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 557{ 558 int i; 559 for (i=0; i <= cmd->o.arg1; ++i ) 560 if (curr_flow == cmd->d[i] ) 561 return 1; 562 return 0; 563} 564 565/* support for IP6_*_ME opcodes */ 566static int 567search_ip6_addr_net (struct in6_addr * ip6_addr) 568{ 569 struct ifnet *mdc; 570 struct ifaddr *mdc2; 571 struct in6_ifaddr *fdm; 572 struct in6_addr copia; 573 574 TAILQ_FOREACH(mdc, &ifnet, if_link) 575 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 576 if (mdc2->ifa_addr->sa_family == AF_INET6) { 577 fdm = (struct in6_ifaddr *)mdc2; 578 copia = fdm->ia_addr.sin6_addr; 579 /* need for leaving scope_id in the sock_addr */ 580 in6_clearscope(&copia); 581 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 582 return 1; 583 } 584 } 585 return 0; 586} 587 588static int 589verify_path6(struct in6_addr *src, struct ifnet *ifp) 590{ 591 struct route_in6 ro; 592 struct sockaddr_in6 *dst; 593 594 bzero(&ro, sizeof(ro)); 595 596 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 597 dst->sin6_family = AF_INET6; 598 dst->sin6_len = sizeof(*dst); 599 dst->sin6_addr = *src; 600 rtalloc_ign((struct route *)&ro, RTF_CLONING); 601 602 if (ro.ro_rt == NULL) 603 return 0; 604 605 /* 606 * if ifp is provided, check for equality with rtentry 607 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 608 * to support the case of sending packets to an address of our own. 609 * (where the former interface is the first argument of if_simloop() 610 * (=ifp), the latter is lo0) 611 */ 612 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 613 RTFREE(ro.ro_rt); 614 return 0; 615 } 616 617 /* if no ifp provided, check if rtentry is not default route */ 618 if (ifp == NULL && 619 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 620 RTFREE(ro.ro_rt); 621 return 0; 622 } 623 624 /* or if this is a blackhole/reject route */ 625 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 626 RTFREE(ro.ro_rt); 627 return 0; 628 } 629 630 /* found valid route */ 631 RTFREE(ro.ro_rt); 632 return 1; 633 634} 635static __inline int 636hash_packet6(struct ipfw_flow_id *id) 637{ 638 u_int32_t i; 639 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 640 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 641 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 642 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 643 (id->dst_port) ^ (id->src_port); 644 return i; 645} 646 647static int 648is_icmp6_query(int icmp6_type) 649{ 650 if ((icmp6_type <= ICMP6_MAXTYPE) && 651 (icmp6_type == ICMP6_ECHO_REQUEST || 652 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 653 icmp6_type == ICMP6_WRUREQUEST || 654 icmp6_type == ICMP6_FQDN_QUERY || 655 icmp6_type == ICMP6_NI_QUERY)) 656 return (1); 657 658 return (0); 659} 660 661static void 662send_reject6(struct ip_fw_args *args, int code, u_int hlen) 663{ 664 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 665 struct ip6_hdr *ip6; 666 struct tcphdr *tcp; 667 tcp_seq ack, seq; 668 int flags; 669 struct { 670 struct ip6_hdr ip6; 671 struct tcphdr th; 672 } ti; 673 674 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) { 675 args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr)); 676 if (args->m == NULL) 677 return; 678 } 679 680 ip6 = mtod(args->m, struct ip6_hdr *); 681 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 682 683 if ((tcp->th_flags & TH_RST) != 0) { 684 m_freem(args->m); 685 return; 686 } 687 688 ti.ip6 = *ip6; 689 ti.th = *tcp; 690 ti.th.th_seq = ntohl(ti.th.th_seq); 691 ti.th.th_ack = ntohl(ti.th.th_ack); 692 ti.ip6.ip6_nxt = IPPROTO_TCP; 693 694 if (ti.th.th_flags & TH_ACK) { 695 ack = 0; 696 seq = ti.th.th_ack; 697 flags = TH_RST; 698 } else { 699 ack = ti.th.th_seq; 700 if (((args->m)->m_flags & M_PKTHDR) != 0) { 701 ack += (args->m)->m_pkthdr.len - hlen 702 - (ti.th.th_off << 2); 703 } else if (ip6->ip6_plen) { 704 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) 705 - hlen - (ti.th.th_off << 2); 706 } else { 707 m_freem(args->m); 708 return; 709 } 710 if (tcp->th_flags & TH_SYN) 711 ack++; 712 seq = 0; 713 flags = TH_RST|TH_ACK; 714 } 715 bcopy(&ti, ip6, sizeof(ti)); 716 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1), 717 args->m, ack, seq, flags); 718 719 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 720 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0); 721 722 } else 723 m_freem(args->m); 724 725 args->m = NULL; 726} 727 728#endif /* INET6 */ 729 730static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 731 732#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 733#define SNP(buf) buf, sizeof(buf) 734 735/* 736 * We enter here when we have a rule with O_LOG. 737 * XXX this function alone takes about 2Kbytes of code! 738 */ 739static void 740ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 741 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg) 742{ 743 struct ether_header *eh = args->eh; 744 char *action; 745 int limit_reached = 0; 746 char action2[40], proto[128], fragment[32]; 747 748 fragment[0] = '\0'; 749 proto[0] = '\0'; 750 751 if (f == NULL) { /* bogus pkt */ 752 if (verbose_limit != 0 && norule_counter >= verbose_limit) 753 return; 754 norule_counter++; 755 if (norule_counter == verbose_limit) 756 limit_reached = verbose_limit; 757 action = "Refuse"; 758 } else { /* O_LOG is the first action, find the real one */ 759 ipfw_insn *cmd = ACTION_PTR(f); 760 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 761 762 if (l->max_log != 0 && l->log_left == 0) 763 return; 764 l->log_left--; 765 if (l->log_left == 0) 766 limit_reached = l->max_log; 767 cmd += F_LEN(cmd); /* point to first action */ 768 if (cmd->opcode == O_ALTQ) { 769 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 770 771 snprintf(SNPARGS(action2, 0), "Altq %d", 772 altq->qid); 773 cmd += F_LEN(cmd); 774 } 775 if (cmd->opcode == O_PROB) 776 cmd += F_LEN(cmd); 777 778 if (cmd->opcode == O_TAG) 779 cmd += F_LEN(cmd); 780 781 action = action2; 782 switch (cmd->opcode) { 783 case O_DENY: 784 action = "Deny"; 785 break; 786 787 case O_REJECT: 788 if (cmd->arg1==ICMP_REJECT_RST) 789 action = "Reset"; 790 else if (cmd->arg1==ICMP_UNREACH_HOST) 791 action = "Reject"; 792 else 793 snprintf(SNPARGS(action2, 0), "Unreach %d", 794 cmd->arg1); 795 break; 796 797 case O_UNREACH6: 798 if (cmd->arg1==ICMP6_UNREACH_RST) 799 action = "Reset"; 800 else 801 snprintf(SNPARGS(action2, 0), "Unreach %d", 802 cmd->arg1); 803 break; 804 805 case O_ACCEPT: 806 action = "Accept"; 807 break; 808 case O_COUNT: 809 action = "Count"; 810 break; 811 case O_DIVERT: 812 snprintf(SNPARGS(action2, 0), "Divert %d", 813 cmd->arg1); 814 break; 815 case O_TEE: 816 snprintf(SNPARGS(action2, 0), "Tee %d", 817 cmd->arg1); 818 break; 819 case O_SKIPTO: 820 snprintf(SNPARGS(action2, 0), "SkipTo %d", 821 cmd->arg1); 822 break; 823 case O_PIPE: 824 snprintf(SNPARGS(action2, 0), "Pipe %d", 825 cmd->arg1); 826 break; 827 case O_QUEUE: 828 snprintf(SNPARGS(action2, 0), "Queue %d", 829 cmd->arg1); 830 break; 831 case O_FORWARD_IP: { 832 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 833 int len; 834 struct in_addr dummyaddr; 835 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 836 dummyaddr.s_addr = htonl(tablearg); 837 else 838 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 839 840 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 841 inet_ntoa(dummyaddr)); 842 843 if (sa->sa.sin_port) 844 snprintf(SNPARGS(action2, len), ":%d", 845 sa->sa.sin_port); 846 } 847 break; 848 case O_NETGRAPH: 849 snprintf(SNPARGS(action2, 0), "Netgraph %d", 850 cmd->arg1); 851 break; 852 case O_NGTEE: 853 snprintf(SNPARGS(action2, 0), "Ngtee %d", 854 cmd->arg1); 855 break; 856 default: 857 action = "UNKNOWN"; 858 break; 859 } 860 } 861 862 if (hlen == 0) { /* non-ip */ 863 snprintf(SNPARGS(proto, 0), "MAC"); 864 865 } else { 866 int len; 867 char src[48], dst[48]; 868 struct icmphdr *icmp; 869 struct tcphdr *tcp; 870 struct udphdr *udp; 871 /* Initialize to make compiler happy. */ 872 struct ip *ip = NULL; 873#ifdef INET6 874 struct ip6_hdr *ip6 = NULL; 875 struct icmp6_hdr *icmp6; 876#endif 877 src[0] = '\0'; 878 dst[0] = '\0'; 879#ifdef INET6 880 if (args->f_id.addr_type == 6) { 881 snprintf(src, sizeof(src), "[%s]", 882 ip6_sprintf(&args->f_id.src_ip6)); 883 snprintf(dst, sizeof(dst), "[%s]", 884 ip6_sprintf(&args->f_id.dst_ip6)); 885 886 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 887 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 888 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 889 } else 890#endif 891 { 892 ip = mtod(m, struct ip *); 893 tcp = L3HDR(struct tcphdr, ip); 894 udp = L3HDR(struct udphdr, ip); 895 896 inet_ntoa_r(ip->ip_src, src); 897 inet_ntoa_r(ip->ip_dst, dst); 898 } 899 900 switch (args->f_id.proto) { 901 case IPPROTO_TCP: 902 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 903 if (offset == 0) 904 snprintf(SNPARGS(proto, len), ":%d %s:%d", 905 ntohs(tcp->th_sport), 906 dst, 907 ntohs(tcp->th_dport)); 908 else 909 snprintf(SNPARGS(proto, len), " %s", dst); 910 break; 911 912 case IPPROTO_UDP: 913 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 914 if (offset == 0) 915 snprintf(SNPARGS(proto, len), ":%d %s:%d", 916 ntohs(udp->uh_sport), 917 dst, 918 ntohs(udp->uh_dport)); 919 else 920 snprintf(SNPARGS(proto, len), " %s", dst); 921 break; 922 923 case IPPROTO_ICMP: 924 icmp = L3HDR(struct icmphdr, ip); 925 if (offset == 0) 926 len = snprintf(SNPARGS(proto, 0), 927 "ICMP:%u.%u ", 928 icmp->icmp_type, icmp->icmp_code); 929 else 930 len = snprintf(SNPARGS(proto, 0), "ICMP "); 931 len += snprintf(SNPARGS(proto, len), "%s", src); 932 snprintf(SNPARGS(proto, len), " %s", dst); 933 break; 934#ifdef INET6 935 case IPPROTO_ICMPV6: 936 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 937 if (offset == 0) 938 len = snprintf(SNPARGS(proto, 0), 939 "ICMPv6:%u.%u ", 940 icmp6->icmp6_type, icmp6->icmp6_code); 941 else 942 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 943 len += snprintf(SNPARGS(proto, len), "%s", src); 944 snprintf(SNPARGS(proto, len), " %s", dst); 945 break; 946#endif 947 default: 948 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 949 args->f_id.proto, src); 950 snprintf(SNPARGS(proto, len), " %s", dst); 951 break; 952 } 953 954#ifdef INET6 955 if (args->f_id.addr_type == 6) { 956 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 957 snprintf(SNPARGS(fragment, 0), 958 " (frag %08x:%d@%d%s)", 959 args->f_id.frag_id6, 960 ntohs(ip6->ip6_plen) - hlen, 961 ntohs(offset & IP6F_OFF_MASK) << 3, 962 (offset & IP6F_MORE_FRAG) ? "+" : ""); 963 } else 964#endif 965 { 966 int ip_off, ip_len; 967 if (eh != NULL) { /* layer 2 packets are as on the wire */ 968 ip_off = ntohs(ip->ip_off); 969 ip_len = ntohs(ip->ip_len); 970 } else { 971 ip_off = ip->ip_off; 972 ip_len = ip->ip_len; 973 } 974 if (ip_off & (IP_MF | IP_OFFMASK)) 975 snprintf(SNPARGS(fragment, 0), 976 " (frag %d:%d@%d%s)", 977 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 978 offset << 3, 979 (ip_off & IP_MF) ? "+" : ""); 980 } 981 } 982 if (oif || m->m_pkthdr.rcvif) 983 log(LOG_SECURITY | LOG_INFO, 984 "ipfw: %d %s %s %s via %s%s\n", 985 f ? f->rulenum : -1, 986 action, proto, oif ? "out" : "in", 987 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 988 fragment); 989 else 990 log(LOG_SECURITY | LOG_INFO, 991 "ipfw: %d %s %s [no if info]%s\n", 992 f ? f->rulenum : -1, 993 action, proto, fragment); 994 if (limit_reached) 995 log(LOG_SECURITY | LOG_NOTICE, 996 "ipfw: limit %d reached on entry %d\n", 997 limit_reached, f ? f->rulenum : -1); 998} 999 1000/* 1001 * IMPORTANT: the hash function for dynamic rules must be commutative 1002 * in source and destination (ip,port), because rules are bidirectional 1003 * and we want to find both in the same bucket. 1004 */ 1005static __inline int 1006hash_packet(struct ipfw_flow_id *id) 1007{ 1008 u_int32_t i; 1009 1010#ifdef INET6 1011 if (IS_IP6_FLOW_ID(id)) 1012 i = hash_packet6(id); 1013 else 1014#endif /* INET6 */ 1015 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1016 i &= (curr_dyn_buckets - 1); 1017 return i; 1018} 1019 1020/** 1021 * unlink a dynamic rule from a chain. prev is a pointer to 1022 * the previous one, q is a pointer to the rule to delete, 1023 * head is a pointer to the head of the queue. 1024 * Modifies q and potentially also head. 1025 */ 1026#define UNLINK_DYN_RULE(prev, head, q) { \ 1027 ipfw_dyn_rule *old_q = q; \ 1028 \ 1029 /* remove a refcount to the parent */ \ 1030 if (q->dyn_type == O_LIMIT) \ 1031 q->parent->count--; \ 1032 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1033 (q->id.src_ip), (q->id.src_port), \ 1034 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1035 if (prev != NULL) \ 1036 prev->next = q = q->next; \ 1037 else \ 1038 head = q = q->next; \ 1039 dyn_count--; \ 1040 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1041 1042#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1043 1044/** 1045 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1046 * 1047 * If keep_me == NULL, rules are deleted even if not expired, 1048 * otherwise only expired rules are removed. 1049 * 1050 * The value of the second parameter is also used to point to identify 1051 * a rule we absolutely do not want to remove (e.g. because we are 1052 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1053 * rules). The pointer is only used for comparison, so any non-null 1054 * value will do. 1055 */ 1056static void 1057remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1058{ 1059 static u_int32_t last_remove = 0; 1060 1061#define FORCE (keep_me == NULL) 1062 1063 ipfw_dyn_rule *prev, *q; 1064 int i, pass = 0, max_pass = 0; 1065 1066 IPFW_DYN_LOCK_ASSERT(); 1067 1068 if (ipfw_dyn_v == NULL || dyn_count == 0) 1069 return; 1070 /* do not expire more than once per second, it is useless */ 1071 if (!FORCE && last_remove == time_uptime) 1072 return; 1073 last_remove = time_uptime; 1074 1075 /* 1076 * because O_LIMIT refer to parent rules, during the first pass only 1077 * remove child and mark any pending LIMIT_PARENT, and remove 1078 * them in a second pass. 1079 */ 1080next_pass: 1081 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1082 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1083 /* 1084 * Logic can become complex here, so we split tests. 1085 */ 1086 if (q == keep_me) 1087 goto next; 1088 if (rule != NULL && rule != q->rule) 1089 goto next; /* not the one we are looking for */ 1090 if (q->dyn_type == O_LIMIT_PARENT) { 1091 /* 1092 * handle parent in the second pass, 1093 * record we need one. 1094 */ 1095 max_pass = 1; 1096 if (pass == 0) 1097 goto next; 1098 if (FORCE && q->count != 0 ) { 1099 /* XXX should not happen! */ 1100 printf("ipfw: OUCH! cannot remove rule," 1101 " count %d\n", q->count); 1102 } 1103 } else { 1104 if (!FORCE && 1105 !TIME_LEQ( q->expire, time_uptime )) 1106 goto next; 1107 } 1108 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1109 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1110 continue; 1111 } 1112next: 1113 prev=q; 1114 q=q->next; 1115 } 1116 } 1117 if (pass++ < max_pass) 1118 goto next_pass; 1119} 1120 1121 1122/** 1123 * lookup a dynamic rule. 1124 */ 1125static ipfw_dyn_rule * 1126lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1127 struct tcphdr *tcp) 1128{ 1129 /* 1130 * stateful ipfw extensions. 1131 * Lookup into dynamic session queue 1132 */ 1133#define MATCH_REVERSE 0 1134#define MATCH_FORWARD 1 1135#define MATCH_NONE 2 1136#define MATCH_UNKNOWN 3 1137 int i, dir = MATCH_NONE; 1138 ipfw_dyn_rule *prev, *q=NULL; 1139 1140 IPFW_DYN_LOCK_ASSERT(); 1141 1142 if (ipfw_dyn_v == NULL) 1143 goto done; /* not found */ 1144 i = hash_packet( pkt ); 1145 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1146 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1147 goto next; 1148 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1149 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1150 continue; 1151 } 1152 if (pkt->proto == q->id.proto && 1153 q->dyn_type != O_LIMIT_PARENT) { 1154 if (IS_IP6_FLOW_ID(pkt)) { 1155 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1156 &(q->id.src_ip6)) && 1157 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1158 &(q->id.dst_ip6)) && 1159 pkt->src_port == q->id.src_port && 1160 pkt->dst_port == q->id.dst_port ) { 1161 dir = MATCH_FORWARD; 1162 break; 1163 } 1164 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1165 &(q->id.dst_ip6)) && 1166 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1167 &(q->id.src_ip6)) && 1168 pkt->src_port == q->id.dst_port && 1169 pkt->dst_port == q->id.src_port ) { 1170 dir = MATCH_REVERSE; 1171 break; 1172 } 1173 } else { 1174 if (pkt->src_ip == q->id.src_ip && 1175 pkt->dst_ip == q->id.dst_ip && 1176 pkt->src_port == q->id.src_port && 1177 pkt->dst_port == q->id.dst_port ) { 1178 dir = MATCH_FORWARD; 1179 break; 1180 } 1181 if (pkt->src_ip == q->id.dst_ip && 1182 pkt->dst_ip == q->id.src_ip && 1183 pkt->src_port == q->id.dst_port && 1184 pkt->dst_port == q->id.src_port ) { 1185 dir = MATCH_REVERSE; 1186 break; 1187 } 1188 } 1189 } 1190next: 1191 prev = q; 1192 q = q->next; 1193 } 1194 if (q == NULL) 1195 goto done; /* q = NULL, not found */ 1196 1197 if ( prev != NULL) { /* found and not in front */ 1198 prev->next = q->next; 1199 q->next = ipfw_dyn_v[i]; 1200 ipfw_dyn_v[i] = q; 1201 } 1202 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1203 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1204 1205#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1206#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1207 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1208 switch (q->state) { 1209 case TH_SYN: /* opening */ 1210 q->expire = time_uptime + dyn_syn_lifetime; 1211 break; 1212 1213 case BOTH_SYN: /* move to established */ 1214 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1215 case BOTH_SYN | (TH_FIN << 8) : 1216 if (tcp) { 1217#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1218 u_int32_t ack = ntohl(tcp->th_ack); 1219 if (dir == MATCH_FORWARD) { 1220 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1221 q->ack_fwd = ack; 1222 else { /* ignore out-of-sequence */ 1223 break; 1224 } 1225 } else { 1226 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1227 q->ack_rev = ack; 1228 else { /* ignore out-of-sequence */ 1229 break; 1230 } 1231 } 1232 } 1233 q->expire = time_uptime + dyn_ack_lifetime; 1234 break; 1235 1236 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1237 if (dyn_fin_lifetime >= dyn_keepalive_period) 1238 dyn_fin_lifetime = dyn_keepalive_period - 1; 1239 q->expire = time_uptime + dyn_fin_lifetime; 1240 break; 1241 1242 default: 1243#if 0 1244 /* 1245 * reset or some invalid combination, but can also 1246 * occur if we use keep-state the wrong way. 1247 */ 1248 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1249 printf("invalid state: 0x%x\n", q->state); 1250#endif 1251 if (dyn_rst_lifetime >= dyn_keepalive_period) 1252 dyn_rst_lifetime = dyn_keepalive_period - 1; 1253 q->expire = time_uptime + dyn_rst_lifetime; 1254 break; 1255 } 1256 } else if (pkt->proto == IPPROTO_UDP) { 1257 q->expire = time_uptime + dyn_udp_lifetime; 1258 } else { 1259 /* other protocols */ 1260 q->expire = time_uptime + dyn_short_lifetime; 1261 } 1262done: 1263 if (match_direction) 1264 *match_direction = dir; 1265 return q; 1266} 1267 1268static ipfw_dyn_rule * 1269lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1270 struct tcphdr *tcp) 1271{ 1272 ipfw_dyn_rule *q; 1273 1274 IPFW_DYN_LOCK(); 1275 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1276 if (q == NULL) 1277 IPFW_DYN_UNLOCK(); 1278 /* NB: return table locked when q is not NULL */ 1279 return q; 1280} 1281 1282static void 1283realloc_dynamic_table(void) 1284{ 1285 IPFW_DYN_LOCK_ASSERT(); 1286 1287 /* 1288 * Try reallocation, make sure we have a power of 2 and do 1289 * not allow more than 64k entries. In case of overflow, 1290 * default to 1024. 1291 */ 1292 1293 if (dyn_buckets > 65536) 1294 dyn_buckets = 1024; 1295 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1296 dyn_buckets = curr_dyn_buckets; /* reset */ 1297 return; 1298 } 1299 curr_dyn_buckets = dyn_buckets; 1300 if (ipfw_dyn_v != NULL) 1301 free(ipfw_dyn_v, M_IPFW); 1302 for (;;) { 1303 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1304 M_IPFW, M_NOWAIT | M_ZERO); 1305 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1306 break; 1307 curr_dyn_buckets /= 2; 1308 } 1309} 1310 1311/** 1312 * Install state of type 'type' for a dynamic session. 1313 * The hash table contains two type of rules: 1314 * - regular rules (O_KEEP_STATE) 1315 * - rules for sessions with limited number of sess per user 1316 * (O_LIMIT). When they are created, the parent is 1317 * increased by 1, and decreased on delete. In this case, 1318 * the third parameter is the parent rule and not the chain. 1319 * - "parent" rules for the above (O_LIMIT_PARENT). 1320 */ 1321static ipfw_dyn_rule * 1322add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1323{ 1324 ipfw_dyn_rule *r; 1325 int i; 1326 1327 IPFW_DYN_LOCK_ASSERT(); 1328 1329 if (ipfw_dyn_v == NULL || 1330 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1331 realloc_dynamic_table(); 1332 if (ipfw_dyn_v == NULL) 1333 return NULL; /* failed ! */ 1334 } 1335 i = hash_packet(id); 1336 1337 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1338 if (r == NULL) { 1339 printf ("ipfw: sorry cannot allocate state\n"); 1340 return NULL; 1341 } 1342 1343 /* increase refcount on parent, and set pointer */ 1344 if (dyn_type == O_LIMIT) { 1345 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1346 if ( parent->dyn_type != O_LIMIT_PARENT) 1347 panic("invalid parent"); 1348 parent->count++; 1349 r->parent = parent; 1350 rule = parent->rule; 1351 } 1352 1353 r->id = *id; 1354 r->expire = time_uptime + dyn_syn_lifetime; 1355 r->rule = rule; 1356 r->dyn_type = dyn_type; 1357 r->pcnt = r->bcnt = 0; 1358 r->count = 0; 1359 1360 r->bucket = i; 1361 r->next = ipfw_dyn_v[i]; 1362 ipfw_dyn_v[i] = r; 1363 dyn_count++; 1364 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1365 dyn_type, 1366 (r->id.src_ip), (r->id.src_port), 1367 (r->id.dst_ip), (r->id.dst_port), 1368 dyn_count ); ) 1369 return r; 1370} 1371 1372/** 1373 * lookup dynamic parent rule using pkt and rule as search keys. 1374 * If the lookup fails, then install one. 1375 */ 1376static ipfw_dyn_rule * 1377lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1378{ 1379 ipfw_dyn_rule *q; 1380 int i; 1381 1382 IPFW_DYN_LOCK_ASSERT(); 1383 1384 if (ipfw_dyn_v) { 1385 int is_v6 = IS_IP6_FLOW_ID(pkt); 1386 i = hash_packet( pkt ); 1387 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1388 if (q->dyn_type == O_LIMIT_PARENT && 1389 rule== q->rule && 1390 pkt->proto == q->id.proto && 1391 pkt->src_port == q->id.src_port && 1392 pkt->dst_port == q->id.dst_port && 1393 ( 1394 (is_v6 && 1395 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1396 &(q->id.src_ip6)) && 1397 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1398 &(q->id.dst_ip6))) || 1399 (!is_v6 && 1400 pkt->src_ip == q->id.src_ip && 1401 pkt->dst_ip == q->id.dst_ip) 1402 ) 1403 ) { 1404 q->expire = time_uptime + dyn_short_lifetime; 1405 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1406 return q; 1407 } 1408 } 1409 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1410} 1411 1412/** 1413 * Install dynamic state for rule type cmd->o.opcode 1414 * 1415 * Returns 1 (failure) if state is not installed because of errors or because 1416 * session limitations are enforced. 1417 */ 1418static int 1419install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1420 struct ip_fw_args *args, uint32_t tablearg) 1421{ 1422 static int last_log; 1423 1424 ipfw_dyn_rule *q; 1425 1426 DEB( 1427 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1428 __func__, cmd->o.opcode, 1429 (args->f_id.src_ip), (args->f_id.src_port), 1430 (args->f_id.dst_ip), (args->f_id.dst_port)); 1431 ) 1432 1433 IPFW_DYN_LOCK(); 1434 1435 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1436 1437 if (q != NULL) { /* should never occur */ 1438 if (last_log != time_uptime) { 1439 last_log = time_uptime; 1440 printf("ipfw: %s: entry already present, done\n", 1441 __func__); 1442 } 1443 IPFW_DYN_UNLOCK(); 1444 return (0); 1445 } 1446 1447 if (dyn_count >= dyn_max) 1448 /* Run out of slots, try to remove any expired rule. */ 1449 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1450 1451 if (dyn_count >= dyn_max) { 1452 if (last_log != time_uptime) { 1453 last_log = time_uptime; 1454 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1455 } 1456 IPFW_DYN_UNLOCK(); 1457 return (1); /* cannot install, notify caller */ 1458 } 1459 1460 switch (cmd->o.opcode) { 1461 case O_KEEP_STATE: /* bidir rule */ 1462 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1463 break; 1464 1465 case O_LIMIT: { /* limit number of sessions */ 1466 struct ipfw_flow_id id; 1467 ipfw_dyn_rule *parent; 1468 uint32_t conn_limit; 1469 uint16_t limit_mask = cmd->limit_mask; 1470 1471 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1472 tablearg : cmd->conn_limit; 1473 1474 DEB( 1475 if (cmd->conn_limit == IP_FW_TABLEARG) 1476 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1477 "(tablearg)\n", __func__, conn_limit); 1478 else 1479 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1480 __func__, conn_limit); 1481 ) 1482 1483 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1484 id.proto = args->f_id.proto; 1485 id.addr_type = args->f_id.addr_type; 1486 1487 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1488 if (limit_mask & DYN_SRC_ADDR) 1489 id.src_ip6 = args->f_id.src_ip6; 1490 if (limit_mask & DYN_DST_ADDR) 1491 id.dst_ip6 = args->f_id.dst_ip6; 1492 } else { 1493 if (limit_mask & DYN_SRC_ADDR) 1494 id.src_ip = args->f_id.src_ip; 1495 if (limit_mask & DYN_DST_ADDR) 1496 id.dst_ip = args->f_id.dst_ip; 1497 } 1498 if (limit_mask & DYN_SRC_PORT) 1499 id.src_port = args->f_id.src_port; 1500 if (limit_mask & DYN_DST_PORT) 1501 id.dst_port = args->f_id.dst_port; 1502 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1503 printf("ipfw: %s: add parent failed\n", __func__); 1504 IPFW_DYN_UNLOCK(); 1505 return (1); 1506 } 1507 1508 if (parent->count >= conn_limit) { 1509 /* See if we can remove some expired rule. */ 1510 remove_dyn_rule(rule, parent); 1511 if (parent->count >= conn_limit) { 1512 if (fw_verbose && last_log != time_uptime) { 1513 last_log = time_uptime; 1514 log(LOG_SECURITY | LOG_DEBUG, 1515 "drop session, too many entries\n"); 1516 } 1517 IPFW_DYN_UNLOCK(); 1518 return (1); 1519 } 1520 } 1521 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1522 break; 1523 } 1524 default: 1525 printf("ipfw: %s: unknown dynamic rule type %u\n", 1526 __func__, cmd->o.opcode); 1527 IPFW_DYN_UNLOCK(); 1528 return (1); 1529 } 1530 1531 /* XXX just set lifetime */ 1532 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1533 1534 IPFW_DYN_UNLOCK(); 1535 return (0); 1536} 1537 1538/* 1539 * Generate a TCP packet, containing either a RST or a keepalive. 1540 * When flags & TH_RST, we are sending a RST packet, because of a 1541 * "reset" action matched the packet. 1542 * Otherwise we are sending a keepalive, and flags & TH_ 1543 */ 1544static struct mbuf * 1545send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1546{ 1547 struct mbuf *m; 1548 struct ip *ip; 1549 struct tcphdr *tcp; 1550 1551 MGETHDR(m, M_DONTWAIT, MT_DATA); 1552 if (m == 0) 1553 return (NULL); 1554 m->m_pkthdr.rcvif = (struct ifnet *)0; 1555 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1556 m->m_data += max_linkhdr; 1557 1558 ip = mtod(m, struct ip *); 1559 bzero(ip, m->m_len); 1560 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1561 ip->ip_p = IPPROTO_TCP; 1562 tcp->th_off = 5; 1563 /* 1564 * Assume we are sending a RST (or a keepalive in the reverse 1565 * direction), swap src and destination addresses and ports. 1566 */ 1567 ip->ip_src.s_addr = htonl(id->dst_ip); 1568 ip->ip_dst.s_addr = htonl(id->src_ip); 1569 tcp->th_sport = htons(id->dst_port); 1570 tcp->th_dport = htons(id->src_port); 1571 if (flags & TH_RST) { /* we are sending a RST */ 1572 if (flags & TH_ACK) { 1573 tcp->th_seq = htonl(ack); 1574 tcp->th_ack = htonl(0); 1575 tcp->th_flags = TH_RST; 1576 } else { 1577 if (flags & TH_SYN) 1578 seq++; 1579 tcp->th_seq = htonl(0); 1580 tcp->th_ack = htonl(seq); 1581 tcp->th_flags = TH_RST | TH_ACK; 1582 } 1583 } else { 1584 /* 1585 * We are sending a keepalive. flags & TH_SYN determines 1586 * the direction, forward if set, reverse if clear. 1587 * NOTE: seq and ack are always assumed to be correct 1588 * as set by the caller. This may be confusing... 1589 */ 1590 if (flags & TH_SYN) { 1591 /* 1592 * we have to rewrite the correct addresses! 1593 */ 1594 ip->ip_dst.s_addr = htonl(id->dst_ip); 1595 ip->ip_src.s_addr = htonl(id->src_ip); 1596 tcp->th_dport = htons(id->dst_port); 1597 tcp->th_sport = htons(id->src_port); 1598 } 1599 tcp->th_seq = htonl(seq); 1600 tcp->th_ack = htonl(ack); 1601 tcp->th_flags = TH_ACK; 1602 } 1603 /* 1604 * set ip_len to the payload size so we can compute 1605 * the tcp checksum on the pseudoheader 1606 * XXX check this, could save a couple of words ? 1607 */ 1608 ip->ip_len = htons(sizeof(struct tcphdr)); 1609 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1610 /* 1611 * now fill fields left out earlier 1612 */ 1613 ip->ip_ttl = ip_defttl; 1614 ip->ip_len = m->m_pkthdr.len; 1615 m->m_flags |= M_SKIP_FIREWALL; 1616 return (m); 1617} 1618 1619/* 1620 * sends a reject message, consuming the mbuf passed as an argument. 1621 */ 1622static void 1623send_reject(struct ip_fw_args *args, int code, int ip_len) 1624{ 1625 1626 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1627 /* We need the IP header in host order for icmp_error(). */ 1628 if (args->eh != NULL) { 1629 struct ip *ip = mtod(args->m, struct ip *); 1630 ip->ip_len = ntohs(ip->ip_len); 1631 ip->ip_off = ntohs(ip->ip_off); 1632 } 1633 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1634 } else if (args->f_id.proto == IPPROTO_TCP) { 1635 struct tcphdr *const tcp = 1636 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1637 if ( (tcp->th_flags & TH_RST) == 0) { 1638 struct mbuf *m; 1639 m = send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1640 ntohl(tcp->th_ack), 1641 tcp->th_flags | TH_RST); 1642 if (m != NULL) 1643 ip_output(m, NULL, NULL, 0, NULL, NULL); 1644 } 1645 m_freem(args->m); 1646 } else 1647 m_freem(args->m); 1648 args->m = NULL; 1649} 1650 1651/** 1652 * 1653 * Given an ip_fw *, lookup_next_rule will return a pointer 1654 * to the next rule, which can be either the jump 1655 * target (for skipto instructions) or the next one in the list (in 1656 * all other cases including a missing jump target). 1657 * The result is also written in the "next_rule" field of the rule. 1658 * Backward jumps are not allowed, so start looking from the next 1659 * rule... 1660 * 1661 * This never returns NULL -- in case we do not have an exact match, 1662 * the next rule is returned. When the ruleset is changed, 1663 * pointers are flushed so we are always correct. 1664 */ 1665 1666static struct ip_fw * 1667lookup_next_rule(struct ip_fw *me) 1668{ 1669 struct ip_fw *rule = NULL; 1670 ipfw_insn *cmd; 1671 1672 /* look for action, in case it is a skipto */ 1673 cmd = ACTION_PTR(me); 1674 if (cmd->opcode == O_LOG) 1675 cmd += F_LEN(cmd); 1676 if (cmd->opcode == O_ALTQ) 1677 cmd += F_LEN(cmd); 1678 if (cmd->opcode == O_TAG) 1679 cmd += F_LEN(cmd); 1680 if ( cmd->opcode == O_SKIPTO ) 1681 for (rule = me->next; rule ; rule = rule->next) 1682 if (rule->rulenum >= cmd->arg1) 1683 break; 1684 if (rule == NULL) /* failure or not a skipto */ 1685 rule = me->next; 1686 me->next_rule = rule; 1687 return rule; 1688} 1689 1690static int 1691add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1692 uint8_t mlen, uint32_t value) 1693{ 1694 struct radix_node_head *rnh; 1695 struct table_entry *ent; 1696 1697 if (tbl >= IPFW_TABLES_MAX) 1698 return (EINVAL); 1699 rnh = ch->tables[tbl]; 1700 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1701 if (ent == NULL) 1702 return (ENOMEM); 1703 ent->value = value; 1704 ent->addr.sin_len = ent->mask.sin_len = 8; 1705 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1706 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1707 IPFW_WLOCK(&layer3_chain); 1708 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1709 NULL) { 1710 IPFW_WUNLOCK(&layer3_chain); 1711 free(ent, M_IPFW_TBL); 1712 return (EEXIST); 1713 } 1714 IPFW_WUNLOCK(&layer3_chain); 1715 return (0); 1716} 1717 1718static int 1719del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1720 uint8_t mlen) 1721{ 1722 struct radix_node_head *rnh; 1723 struct table_entry *ent; 1724 struct sockaddr_in sa, mask; 1725 1726 if (tbl >= IPFW_TABLES_MAX) 1727 return (EINVAL); 1728 rnh = ch->tables[tbl]; 1729 sa.sin_len = mask.sin_len = 8; 1730 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1731 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1732 IPFW_WLOCK(ch); 1733 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1734 if (ent == NULL) { 1735 IPFW_WUNLOCK(ch); 1736 return (ESRCH); 1737 } 1738 IPFW_WUNLOCK(ch); 1739 free(ent, M_IPFW_TBL); 1740 return (0); 1741} 1742 1743static int 1744flush_table_entry(struct radix_node *rn, void *arg) 1745{ 1746 struct radix_node_head * const rnh = arg; 1747 struct table_entry *ent; 1748 1749 ent = (struct table_entry *) 1750 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1751 if (ent != NULL) 1752 free(ent, M_IPFW_TBL); 1753 return (0); 1754} 1755 1756static int 1757flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1758{ 1759 struct radix_node_head *rnh; 1760 1761 IPFW_WLOCK_ASSERT(ch); 1762 1763 if (tbl >= IPFW_TABLES_MAX) 1764 return (EINVAL); 1765 rnh = ch->tables[tbl]; 1766 KASSERT(rnh != NULL, ("NULL IPFW table")); 1767 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1768 return (0); 1769} 1770 1771static void 1772flush_tables(struct ip_fw_chain *ch) 1773{ 1774 uint16_t tbl; 1775 1776 IPFW_WLOCK_ASSERT(ch); 1777 1778 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1779 flush_table(ch, tbl); 1780} 1781 1782static int 1783init_tables(struct ip_fw_chain *ch) 1784{ 1785 int i; 1786 uint16_t j; 1787 1788 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1789 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1790 for (j = 0; j < i; j++) { 1791 (void) flush_table(ch, j); 1792 } 1793 return (ENOMEM); 1794 } 1795 } 1796 return (0); 1797} 1798 1799static int 1800lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1801 uint32_t *val) 1802{ 1803 struct radix_node_head *rnh; 1804 struct table_entry *ent; 1805 struct sockaddr_in sa; 1806 1807 if (tbl >= IPFW_TABLES_MAX) 1808 return (0); 1809 rnh = ch->tables[tbl]; 1810 sa.sin_len = 8; 1811 sa.sin_addr.s_addr = addr; 1812 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1813 if (ent != NULL) { 1814 *val = ent->value; 1815 return (1); 1816 } 1817 return (0); 1818} 1819 1820static int 1821count_table_entry(struct radix_node *rn, void *arg) 1822{ 1823 u_int32_t * const cnt = arg; 1824 1825 (*cnt)++; 1826 return (0); 1827} 1828 1829static int 1830count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1831{ 1832 struct radix_node_head *rnh; 1833 1834 if (tbl >= IPFW_TABLES_MAX) 1835 return (EINVAL); 1836 rnh = ch->tables[tbl]; 1837 *cnt = 0; 1838 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1839 return (0); 1840} 1841 1842static int 1843dump_table_entry(struct radix_node *rn, void *arg) 1844{ 1845 struct table_entry * const n = (struct table_entry *)rn; 1846 ipfw_table * const tbl = arg; 1847 ipfw_table_entry *ent; 1848 1849 if (tbl->cnt == tbl->size) 1850 return (1); 1851 ent = &tbl->ent[tbl->cnt]; 1852 ent->tbl = tbl->tbl; 1853 if (in_nullhost(n->mask.sin_addr)) 1854 ent->masklen = 0; 1855 else 1856 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1857 ent->addr = n->addr.sin_addr.s_addr; 1858 ent->value = n->value; 1859 tbl->cnt++; 1860 return (0); 1861} 1862 1863static int 1864dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1865{ 1866 struct radix_node_head *rnh; 1867 1868 if (tbl->tbl >= IPFW_TABLES_MAX) 1869 return (EINVAL); 1870 rnh = ch->tables[tbl->tbl]; 1871 tbl->cnt = 0; 1872 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1873 return (0); 1874} 1875 1876static void 1877fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1878{ 1879 struct ucred *cr; 1880 1881 if (inp->inp_socket != NULL) { 1882 cr = inp->inp_socket->so_cred; 1883 ugp->fw_prid = jailed(cr) ? 1884 cr->cr_prison->pr_id : -1; 1885 ugp->fw_uid = cr->cr_uid; 1886 ugp->fw_ngroups = cr->cr_ngroups; 1887 bcopy(cr->cr_groups, ugp->fw_groups, 1888 sizeof(ugp->fw_groups)); 1889 } 1890} 1891 1892static int 1893check_uidgid(ipfw_insn_u32 *insn, 1894 int proto, struct ifnet *oif, 1895 struct in_addr dst_ip, u_int16_t dst_port, 1896 struct in_addr src_ip, u_int16_t src_port, 1897 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1898{ 1899 struct inpcbinfo *pi; 1900 int wildcard; 1901 struct inpcb *pcb; 1902 int match; 1903 gid_t *gp; 1904 1905 /* 1906 * Check to see if the UDP or TCP stack supplied us with 1907 * the PCB. If so, rather then holding a lock and looking 1908 * up the PCB, we can use the one that was supplied. 1909 */ 1910 if (inp && *lookup == 0) { 1911 INP_LOCK_ASSERT(inp); 1912 if (inp->inp_socket != NULL) { 1913 fill_ugid_cache(inp, ugp); 1914 *lookup = 1; 1915 } 1916 } 1917 /* 1918 * If we have already been here and the packet has no 1919 * PCB entry associated with it, then we can safely 1920 * assume that this is a no match. 1921 */ 1922 if (*lookup == -1) 1923 return (0); 1924 if (proto == IPPROTO_TCP) { 1925 wildcard = 0; 1926 pi = &tcbinfo; 1927 } else if (proto == IPPROTO_UDP) { 1928 wildcard = INPLOOKUP_WILDCARD; 1929 pi = &udbinfo; 1930 } else 1931 return 0; 1932 match = 0; 1933 if (*lookup == 0) { 1934 INP_INFO_RLOCK(pi); 1935 pcb = (oif) ? 1936 in_pcblookup_hash(pi, 1937 dst_ip, htons(dst_port), 1938 src_ip, htons(src_port), 1939 wildcard, oif) : 1940 in_pcblookup_hash(pi, 1941 src_ip, htons(src_port), 1942 dst_ip, htons(dst_port), 1943 wildcard, NULL); 1944 if (pcb != NULL) { 1945 INP_LOCK(pcb); 1946 if (pcb->inp_socket != NULL) { 1947 fill_ugid_cache(pcb, ugp); 1948 *lookup = 1; 1949 } 1950 INP_UNLOCK(pcb); 1951 } 1952 INP_INFO_RUNLOCK(pi); 1953 if (*lookup == 0) { 1954 /* 1955 * If the lookup did not yield any results, there 1956 * is no sense in coming back and trying again. So 1957 * we can set lookup to -1 and ensure that we wont 1958 * bother the pcb system again. 1959 */ 1960 *lookup = -1; 1961 return (0); 1962 } 1963 } 1964 if (insn->o.opcode == O_UID) 1965 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1966 else if (insn->o.opcode == O_GID) { 1967 for (gp = ugp->fw_groups; 1968 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1969 if (*gp == (gid_t)insn->d[0]) { 1970 match = 1; 1971 break; 1972 } 1973 } else if (insn->o.opcode == O_JAIL) 1974 match = (ugp->fw_prid == (int)insn->d[0]); 1975 return match; 1976} 1977 1978/* 1979 * The main check routine for the firewall. 1980 * 1981 * All arguments are in args so we can modify them and return them 1982 * back to the caller. 1983 * 1984 * Parameters: 1985 * 1986 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1987 * Starts with the IP header. 1988 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1989 * args->oif Outgoing interface, or NULL if packet is incoming. 1990 * The incoming interface is in the mbuf. (in) 1991 * args->divert_rule (in/out) 1992 * Skip up to the first rule past this rule number; 1993 * upon return, non-zero port number for divert or tee. 1994 * 1995 * args->rule Pointer to the last matching rule (in/out) 1996 * args->next_hop Socket we are forwarding to (out). 1997 * args->f_id Addresses grabbed from the packet (out) 1998 * args->cookie a cookie depending on rule action 1999 * 2000 * Return value: 2001 * 2002 * IP_FW_PASS the packet must be accepted 2003 * IP_FW_DENY the packet must be dropped 2004 * IP_FW_DIVERT divert packet, port in m_tag 2005 * IP_FW_TEE tee packet, port in m_tag 2006 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2007 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2008 * 2009 */ 2010 2011int 2012ipfw_chk(struct ip_fw_args *args) 2013{ 2014 /* 2015 * Local variables hold state during the processing of a packet. 2016 * 2017 * IMPORTANT NOTE: to speed up the processing of rules, there 2018 * are some assumption on the values of the variables, which 2019 * are documented here. Should you change them, please check 2020 * the implementation of the various instructions to make sure 2021 * that they still work. 2022 * 2023 * args->eh The MAC header. It is non-null for a layer2 2024 * packet, it is NULL for a layer-3 packet. 2025 * 2026 * m | args->m Pointer to the mbuf, as received from the caller. 2027 * It may change if ipfw_chk() does an m_pullup, or if it 2028 * consumes the packet because it calls send_reject(). 2029 * XXX This has to change, so that ipfw_chk() never modifies 2030 * or consumes the buffer. 2031 * ip is simply an alias of the value of m, and it is kept 2032 * in sync with it (the packet is supposed to start with 2033 * the ip header). 2034 */ 2035 struct mbuf *m = args->m; 2036 struct ip *ip = mtod(m, struct ip *); 2037 2038 /* 2039 * For rules which contain uid/gid or jail constraints, cache 2040 * a copy of the users credentials after the pcb lookup has been 2041 * executed. This will speed up the processing of rules with 2042 * these types of constraints, as well as decrease contention 2043 * on pcb related locks. 2044 */ 2045 struct ip_fw_ugid fw_ugid_cache; 2046 int ugid_lookup = 0; 2047 2048 /* 2049 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2050 * associated with a packet input on a divert socket. This 2051 * will allow to distinguish traffic and its direction when 2052 * it originates from a divert socket. 2053 */ 2054 u_int divinput_flags = 0; 2055 2056 /* 2057 * oif | args->oif If NULL, ipfw_chk has been called on the 2058 * inbound path (ether_input, ip_input). 2059 * If non-NULL, ipfw_chk has been called on the outbound path 2060 * (ether_output, ip_output). 2061 */ 2062 struct ifnet *oif = args->oif; 2063 2064 struct ip_fw *f = NULL; /* matching rule */ 2065 int retval = 0; 2066 2067 /* 2068 * hlen The length of the IP header. 2069 */ 2070 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2071 2072 /* 2073 * offset The offset of a fragment. offset != 0 means that 2074 * we have a fragment at this offset of an IPv4 packet. 2075 * offset == 0 means that (if this is an IPv4 packet) 2076 * this is the first or only fragment. 2077 * For IPv6 offset == 0 means there is no Fragment Header. 2078 * If offset != 0 for IPv6 always use correct mask to 2079 * get the correct offset because we add IP6F_MORE_FRAG 2080 * to be able to dectect the first fragment which would 2081 * otherwise have offset = 0. 2082 */ 2083 u_short offset = 0; 2084 2085 /* 2086 * Local copies of addresses. They are only valid if we have 2087 * an IP packet. 2088 * 2089 * proto The protocol. Set to 0 for non-ip packets, 2090 * or to the protocol read from the packet otherwise. 2091 * proto != 0 means that we have an IPv4 packet. 2092 * 2093 * src_port, dst_port port numbers, in HOST format. Only 2094 * valid for TCP and UDP packets. 2095 * 2096 * src_ip, dst_ip ip addresses, in NETWORK format. 2097 * Only valid for IPv4 packets. 2098 */ 2099 u_int8_t proto; 2100 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2101 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2102 u_int16_t ip_len=0; 2103 int pktlen; 2104 2105 /* 2106 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2107 * MATCH_NONE when checked and not matched (q = NULL), 2108 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2109 */ 2110 int dyn_dir = MATCH_UNKNOWN; 2111 ipfw_dyn_rule *q = NULL; 2112 struct ip_fw_chain *chain = &layer3_chain; 2113 struct m_tag *mtag; 2114 2115 /* 2116 * We store in ulp a pointer to the upper layer protocol header. 2117 * In the ipv4 case this is easy to determine from the header, 2118 * but for ipv6 we might have some additional headers in the middle. 2119 * ulp is NULL if not found. 2120 */ 2121 void *ulp = NULL; /* upper layer protocol pointer. */ 2122 /* XXX ipv6 variables */ 2123 int is_ipv6 = 0; 2124 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2125 /* end of ipv6 variables */ 2126 int is_ipv4 = 0; 2127 2128 if (m->m_flags & M_SKIP_FIREWALL) 2129 return (IP_FW_PASS); /* accept */ 2130 2131 pktlen = m->m_pkthdr.len; 2132 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2133 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2134 2135/* 2136 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2137 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2138 * pointer might become stale after other pullups (but we never use it 2139 * this way). 2140 */ 2141#define PULLUP_TO(len, p, T) \ 2142do { \ 2143 int x = (len) + sizeof(T); \ 2144 if ((m)->m_len < x) { \ 2145 args->m = m = m_pullup(m, x); \ 2146 if (m == NULL) \ 2147 goto pullup_failed; \ 2148 } \ 2149 p = (mtod(m, char *) + (len)); \ 2150} while (0) 2151 2152 /* Identify IP packets and fill up variables. */ 2153 if (pktlen >= sizeof(struct ip6_hdr) && 2154 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2155 mtod(m, struct ip *)->ip_v == 6) { 2156 is_ipv6 = 1; 2157 args->f_id.addr_type = 6; 2158 hlen = sizeof(struct ip6_hdr); 2159 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2160 2161 /* Search extension headers to find upper layer protocols */ 2162 while (ulp == NULL) { 2163 switch (proto) { 2164 case IPPROTO_ICMPV6: 2165 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2166 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2167 break; 2168 2169 case IPPROTO_TCP: 2170 PULLUP_TO(hlen, ulp, struct tcphdr); 2171 dst_port = TCP(ulp)->th_dport; 2172 src_port = TCP(ulp)->th_sport; 2173 args->f_id.flags = TCP(ulp)->th_flags; 2174 break; 2175 2176 case IPPROTO_UDP: 2177 PULLUP_TO(hlen, ulp, struct udphdr); 2178 dst_port = UDP(ulp)->uh_dport; 2179 src_port = UDP(ulp)->uh_sport; 2180 break; 2181 2182 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2183 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2184 ext_hd |= EXT_HOPOPTS; 2185 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2186 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2187 ulp = NULL; 2188 break; 2189 2190 case IPPROTO_ROUTING: /* RFC 2460 */ 2191 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2192 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2193 case 0: 2194 break; 2195 default: 2196 printf("IPFW2: IPV6 - Unknown Routing " 2197 "Header type(%d)\n", 2198 ((struct ip6_rthdr *)ulp)->ip6r_type); 2199 if (fw_deny_unknown_exthdrs) 2200 return (IP_FW_DENY); 2201 break; 2202 } 2203 ext_hd |= EXT_ROUTING; 2204 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2205 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2206 ulp = NULL; 2207 break; 2208 2209 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2210 PULLUP_TO(hlen, ulp, struct ip6_frag); 2211 ext_hd |= EXT_FRAGMENT; 2212 hlen += sizeof (struct ip6_frag); 2213 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2214 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2215 IP6F_OFF_MASK; 2216 /* Add IP6F_MORE_FRAG for offset of first 2217 * fragment to be != 0. */ 2218 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2219 IP6F_MORE_FRAG; 2220 if (offset == 0) { 2221 printf("IPFW2: IPV6 - Invalid Fragment " 2222 "Header\n"); 2223 if (fw_deny_unknown_exthdrs) 2224 return (IP_FW_DENY); 2225 break; 2226 } 2227 args->f_id.frag_id6 = 2228 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2229 ulp = NULL; 2230 break; 2231 2232 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2233 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2234 ext_hd |= EXT_DSTOPTS; 2235 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2236 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2237 ulp = NULL; 2238 break; 2239 2240 case IPPROTO_AH: /* RFC 2402 */ 2241 PULLUP_TO(hlen, ulp, struct ip6_ext); 2242 ext_hd |= EXT_AH; 2243 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2244 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2245 ulp = NULL; 2246 break; 2247 2248 case IPPROTO_ESP: /* RFC 2406 */ 2249 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2250 /* Anything past Seq# is variable length and 2251 * data past this ext. header is encrypted. */ 2252 ext_hd |= EXT_ESP; 2253 break; 2254 2255 case IPPROTO_NONE: /* RFC 2460 */ 2256 PULLUP_TO(hlen, ulp, struct ip6_ext); 2257 /* Packet ends here. if ip6e_len!=0 octets 2258 * must be ignored. */ 2259 break; 2260 2261 case IPPROTO_OSPFIGP: 2262 /* XXX OSPF header check? */ 2263 PULLUP_TO(hlen, ulp, struct ip6_ext); 2264 break; 2265 2266 case IPPROTO_IPV6: /* RFC 2893 */ 2267 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2268 break; 2269 2270 case IPPROTO_IPV4: /* RFC 2893 */ 2271 PULLUP_TO(hlen, ulp, struct ip); 2272 break; 2273 2274 default: 2275 printf("IPFW2: IPV6 - Unknown Extension " 2276 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2277 if (fw_deny_unknown_exthdrs) 2278 return (IP_FW_DENY); 2279 PULLUP_TO(hlen, ulp, struct ip6_ext); 2280 break; 2281 } /*switch */ 2282 } 2283 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2284 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2285 args->f_id.src_ip = 0; 2286 args->f_id.dst_ip = 0; 2287 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2288 } else if (pktlen >= sizeof(struct ip) && 2289 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2290 mtod(m, struct ip *)->ip_v == 4) { 2291 is_ipv4 = 1; 2292 ip = mtod(m, struct ip *); 2293 hlen = ip->ip_hl << 2; 2294 args->f_id.addr_type = 4; 2295 2296 /* 2297 * Collect parameters into local variables for faster matching. 2298 */ 2299 proto = ip->ip_p; 2300 src_ip = ip->ip_src; 2301 dst_ip = ip->ip_dst; 2302 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2303 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2304 ip_len = ntohs(ip->ip_len); 2305 } else { 2306 offset = ip->ip_off & IP_OFFMASK; 2307 ip_len = ip->ip_len; 2308 } 2309 pktlen = ip_len < pktlen ? ip_len : pktlen; 2310 2311 if (offset == 0) { 2312 switch (proto) { 2313 case IPPROTO_TCP: 2314 PULLUP_TO(hlen, ulp, struct tcphdr); 2315 dst_port = TCP(ulp)->th_dport; 2316 src_port = TCP(ulp)->th_sport; 2317 args->f_id.flags = TCP(ulp)->th_flags; 2318 break; 2319 2320 case IPPROTO_UDP: 2321 PULLUP_TO(hlen, ulp, struct udphdr); 2322 dst_port = UDP(ulp)->uh_dport; 2323 src_port = UDP(ulp)->uh_sport; 2324 break; 2325 2326 case IPPROTO_ICMP: 2327 PULLUP_TO(hlen, ulp, struct icmphdr); 2328 args->f_id.flags = ICMP(ulp)->icmp_type; 2329 break; 2330 2331 default: 2332 break; 2333 } 2334 } 2335 2336 args->f_id.src_ip = ntohl(src_ip.s_addr); 2337 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2338 } 2339#undef PULLUP_TO 2340 if (proto) { /* we may have port numbers, store them */ 2341 args->f_id.proto = proto; 2342 args->f_id.src_port = src_port = ntohs(src_port); 2343 args->f_id.dst_port = dst_port = ntohs(dst_port); 2344 } 2345 2346 IPFW_RLOCK(chain); 2347 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2348 if (args->rule) { 2349 /* 2350 * Packet has already been tagged. Look for the next rule 2351 * to restart processing. 2352 * 2353 * If fw_one_pass != 0 then just accept it. 2354 * XXX should not happen here, but optimized out in 2355 * the caller. 2356 */ 2357 if (fw_one_pass) { 2358 IPFW_RUNLOCK(chain); 2359 return (IP_FW_PASS); 2360 } 2361 2362 f = args->rule->next_rule; 2363 if (f == NULL) 2364 f = lookup_next_rule(args->rule); 2365 } else { 2366 /* 2367 * Find the starting rule. It can be either the first 2368 * one, or the one after divert_rule if asked so. 2369 */ 2370 int skipto = mtag ? divert_cookie(mtag) : 0; 2371 2372 f = chain->rules; 2373 if (args->eh == NULL && skipto != 0) { 2374 if (skipto >= IPFW_DEFAULT_RULE) { 2375 IPFW_RUNLOCK(chain); 2376 return (IP_FW_DENY); /* invalid */ 2377 } 2378 while (f && f->rulenum <= skipto) 2379 f = f->next; 2380 if (f == NULL) { /* drop packet */ 2381 IPFW_RUNLOCK(chain); 2382 return (IP_FW_DENY); 2383 } 2384 } 2385 } 2386 /* reset divert rule to avoid confusion later */ 2387 if (mtag) { 2388 divinput_flags = divert_info(mtag) & 2389 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2390 m_tag_delete(m, mtag); 2391 } 2392 2393 /* 2394 * Now scan the rules, and parse microinstructions for each rule. 2395 */ 2396 for (; f; f = f->next) { 2397 ipfw_insn *cmd; 2398 uint32_t tablearg = 0; 2399 int l, cmdlen, skip_or; /* skip rest of OR block */ 2400 2401again: 2402 if (set_disable & (1 << f->set) ) 2403 continue; 2404 2405 skip_or = 0; 2406 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2407 l -= cmdlen, cmd += cmdlen) { 2408 int match; 2409 2410 /* 2411 * check_body is a jump target used when we find a 2412 * CHECK_STATE, and need to jump to the body of 2413 * the target rule. 2414 */ 2415 2416check_body: 2417 cmdlen = F_LEN(cmd); 2418 /* 2419 * An OR block (insn_1 || .. || insn_n) has the 2420 * F_OR bit set in all but the last instruction. 2421 * The first match will set "skip_or", and cause 2422 * the following instructions to be skipped until 2423 * past the one with the F_OR bit clear. 2424 */ 2425 if (skip_or) { /* skip this instruction */ 2426 if ((cmd->len & F_OR) == 0) 2427 skip_or = 0; /* next one is good */ 2428 continue; 2429 } 2430 match = 0; /* set to 1 if we succeed */ 2431 2432 switch (cmd->opcode) { 2433 /* 2434 * The first set of opcodes compares the packet's 2435 * fields with some pattern, setting 'match' if a 2436 * match is found. At the end of the loop there is 2437 * logic to deal with F_NOT and F_OR flags associated 2438 * with the opcode. 2439 */ 2440 case O_NOP: 2441 match = 1; 2442 break; 2443 2444 case O_FORWARD_MAC: 2445 printf("ipfw: opcode %d unimplemented\n", 2446 cmd->opcode); 2447 break; 2448 2449 case O_GID: 2450 case O_UID: 2451 case O_JAIL: 2452 /* 2453 * We only check offset == 0 && proto != 0, 2454 * as this ensures that we have a 2455 * packet with the ports info. 2456 */ 2457 if (offset!=0) 2458 break; 2459 if (is_ipv6) /* XXX to be fixed later */ 2460 break; 2461 if (proto == IPPROTO_TCP || 2462 proto == IPPROTO_UDP) 2463 match = check_uidgid( 2464 (ipfw_insn_u32 *)cmd, 2465 proto, oif, 2466 dst_ip, dst_port, 2467 src_ip, src_port, &fw_ugid_cache, 2468 &ugid_lookup, args->inp); 2469 break; 2470 2471 case O_RECV: 2472 match = iface_match(m->m_pkthdr.rcvif, 2473 (ipfw_insn_if *)cmd); 2474 break; 2475 2476 case O_XMIT: 2477 match = iface_match(oif, (ipfw_insn_if *)cmd); 2478 break; 2479 2480 case O_VIA: 2481 match = iface_match(oif ? oif : 2482 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2483 break; 2484 2485 case O_MACADDR2: 2486 if (args->eh != NULL) { /* have MAC header */ 2487 u_int32_t *want = (u_int32_t *) 2488 ((ipfw_insn_mac *)cmd)->addr; 2489 u_int32_t *mask = (u_int32_t *) 2490 ((ipfw_insn_mac *)cmd)->mask; 2491 u_int32_t *hdr = (u_int32_t *)args->eh; 2492 2493 match = 2494 ( want[0] == (hdr[0] & mask[0]) && 2495 want[1] == (hdr[1] & mask[1]) && 2496 want[2] == (hdr[2] & mask[2]) ); 2497 } 2498 break; 2499 2500 case O_MAC_TYPE: 2501 if (args->eh != NULL) { 2502 u_int16_t t = 2503 ntohs(args->eh->ether_type); 2504 u_int16_t *p = 2505 ((ipfw_insn_u16 *)cmd)->ports; 2506 int i; 2507 2508 for (i = cmdlen - 1; !match && i>0; 2509 i--, p += 2) 2510 match = (t>=p[0] && t<=p[1]); 2511 } 2512 break; 2513 2514 case O_FRAG: 2515 match = (offset != 0); 2516 break; 2517 2518 case O_IN: /* "out" is "not in" */ 2519 match = (oif == NULL); 2520 break; 2521 2522 case O_LAYER2: 2523 match = (args->eh != NULL); 2524 break; 2525 2526 case O_DIVERTED: 2527 match = (cmd->arg1 & 1 && divinput_flags & 2528 IP_FW_DIVERT_LOOPBACK_FLAG) || 2529 (cmd->arg1 & 2 && divinput_flags & 2530 IP_FW_DIVERT_OUTPUT_FLAG); 2531 break; 2532 2533 case O_PROTO: 2534 /* 2535 * We do not allow an arg of 0 so the 2536 * check of "proto" only suffices. 2537 */ 2538 match = (proto == cmd->arg1); 2539 break; 2540 2541 case O_IP_SRC: 2542 match = is_ipv4 && 2543 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2544 src_ip.s_addr); 2545 break; 2546 2547 case O_IP_SRC_LOOKUP: 2548 case O_IP_DST_LOOKUP: 2549 if (is_ipv4) { 2550 uint32_t a = 2551 (cmd->opcode == O_IP_DST_LOOKUP) ? 2552 dst_ip.s_addr : src_ip.s_addr; 2553 uint32_t v; 2554 2555 match = lookup_table(chain, cmd->arg1, a, 2556 &v); 2557 if (!match) 2558 break; 2559 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2560 match = 2561 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2562 else 2563 tablearg = v; 2564 } 2565 break; 2566 2567 case O_IP_SRC_MASK: 2568 case O_IP_DST_MASK: 2569 if (is_ipv4) { 2570 uint32_t a = 2571 (cmd->opcode == O_IP_DST_MASK) ? 2572 dst_ip.s_addr : src_ip.s_addr; 2573 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2574 int i = cmdlen-1; 2575 2576 for (; !match && i>0; i-= 2, p+= 2) 2577 match = (p[0] == (a & p[1])); 2578 } 2579 break; 2580 2581 case O_IP_SRC_ME: 2582 if (is_ipv4) { 2583 struct ifnet *tif; 2584 2585 INADDR_TO_IFP(src_ip, tif); 2586 match = (tif != NULL); 2587 } 2588 break; 2589 2590 case O_IP_DST_SET: 2591 case O_IP_SRC_SET: 2592 if (is_ipv4) { 2593 u_int32_t *d = (u_int32_t *)(cmd+1); 2594 u_int32_t addr = 2595 cmd->opcode == O_IP_DST_SET ? 2596 args->f_id.dst_ip : 2597 args->f_id.src_ip; 2598 2599 if (addr < d[0]) 2600 break; 2601 addr -= d[0]; /* subtract base */ 2602 match = (addr < cmd->arg1) && 2603 ( d[ 1 + (addr>>5)] & 2604 (1<<(addr & 0x1f)) ); 2605 } 2606 break; 2607 2608 case O_IP_DST: 2609 match = is_ipv4 && 2610 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2611 dst_ip.s_addr); 2612 break; 2613 2614 case O_IP_DST_ME: 2615 if (is_ipv4) { 2616 struct ifnet *tif; 2617 2618 INADDR_TO_IFP(dst_ip, tif); 2619 match = (tif != NULL); 2620 } 2621 break; 2622 2623 case O_IP_SRCPORT: 2624 case O_IP_DSTPORT: 2625 /* 2626 * offset == 0 && proto != 0 is enough 2627 * to guarantee that we have a 2628 * packet with port info. 2629 */ 2630 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2631 && offset == 0) { 2632 u_int16_t x = 2633 (cmd->opcode == O_IP_SRCPORT) ? 2634 src_port : dst_port ; 2635 u_int16_t *p = 2636 ((ipfw_insn_u16 *)cmd)->ports; 2637 int i; 2638 2639 for (i = cmdlen - 1; !match && i>0; 2640 i--, p += 2) 2641 match = (x>=p[0] && x<=p[1]); 2642 } 2643 break; 2644 2645 case O_ICMPTYPE: 2646 match = (offset == 0 && proto==IPPROTO_ICMP && 2647 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2648 break; 2649 2650#ifdef INET6 2651 case O_ICMP6TYPE: 2652 match = is_ipv6 && offset == 0 && 2653 proto==IPPROTO_ICMPV6 && 2654 icmp6type_match( 2655 ICMP6(ulp)->icmp6_type, 2656 (ipfw_insn_u32 *)cmd); 2657 break; 2658#endif /* INET6 */ 2659 2660 case O_IPOPT: 2661 match = (is_ipv4 && 2662 ipopts_match(mtod(m, struct ip *), cmd) ); 2663 break; 2664 2665 case O_IPVER: 2666 match = (is_ipv4 && 2667 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2668 break; 2669 2670 case O_IPID: 2671 case O_IPLEN: 2672 case O_IPTTL: 2673 if (is_ipv4) { /* only for IP packets */ 2674 uint16_t x; 2675 uint16_t *p; 2676 int i; 2677 2678 if (cmd->opcode == O_IPLEN) 2679 x = ip_len; 2680 else if (cmd->opcode == O_IPTTL) 2681 x = mtod(m, struct ip *)->ip_ttl; 2682 else /* must be IPID */ 2683 x = ntohs(mtod(m, struct ip *)->ip_id); 2684 if (cmdlen == 1) { 2685 match = (cmd->arg1 == x); 2686 break; 2687 } 2688 /* otherwise we have ranges */ 2689 p = ((ipfw_insn_u16 *)cmd)->ports; 2690 i = cmdlen - 1; 2691 for (; !match && i>0; i--, p += 2) 2692 match = (x >= p[0] && x <= p[1]); 2693 } 2694 break; 2695 2696 case O_IPPRECEDENCE: 2697 match = (is_ipv4 && 2698 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2699 break; 2700 2701 case O_IPTOS: 2702 match = (is_ipv4 && 2703 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2704 break; 2705 2706 case O_TCPDATALEN: 2707 if (proto == IPPROTO_TCP && offset == 0) { 2708 struct tcphdr *tcp; 2709 uint16_t x; 2710 uint16_t *p; 2711 int i; 2712 2713 tcp = TCP(ulp); 2714 x = ip_len - 2715 ((ip->ip_hl + tcp->th_off) << 2); 2716 if (cmdlen == 1) { 2717 match = (cmd->arg1 == x); 2718 break; 2719 } 2720 /* otherwise we have ranges */ 2721 p = ((ipfw_insn_u16 *)cmd)->ports; 2722 i = cmdlen - 1; 2723 for (; !match && i>0; i--, p += 2) 2724 match = (x >= p[0] && x <= p[1]); 2725 } 2726 break; 2727 2728 case O_TCPFLAGS: 2729 match = (proto == IPPROTO_TCP && offset == 0 && 2730 flags_match(cmd, TCP(ulp)->th_flags)); 2731 break; 2732 2733 case O_TCPOPTS: 2734 match = (proto == IPPROTO_TCP && offset == 0 && 2735 tcpopts_match(TCP(ulp), cmd)); 2736 break; 2737 2738 case O_TCPSEQ: 2739 match = (proto == IPPROTO_TCP && offset == 0 && 2740 ((ipfw_insn_u32 *)cmd)->d[0] == 2741 TCP(ulp)->th_seq); 2742 break; 2743 2744 case O_TCPACK: 2745 match = (proto == IPPROTO_TCP && offset == 0 && 2746 ((ipfw_insn_u32 *)cmd)->d[0] == 2747 TCP(ulp)->th_ack); 2748 break; 2749 2750 case O_TCPWIN: 2751 match = (proto == IPPROTO_TCP && offset == 0 && 2752 cmd->arg1 == TCP(ulp)->th_win); 2753 break; 2754 2755 case O_ESTAB: 2756 /* reject packets which have SYN only */ 2757 /* XXX should i also check for TH_ACK ? */ 2758 match = (proto == IPPROTO_TCP && offset == 0 && 2759 (TCP(ulp)->th_flags & 2760 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2761 break; 2762 2763 case O_ALTQ: { 2764 struct altq_tag *at; 2765 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2766 2767 match = 1; 2768 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2769 if (mtag != NULL) 2770 break; 2771 mtag = m_tag_get(PACKET_TAG_PF_QID, 2772 sizeof(struct altq_tag), 2773 M_NOWAIT); 2774 if (mtag == NULL) { 2775 /* 2776 * Let the packet fall back to the 2777 * default ALTQ. 2778 */ 2779 break; 2780 } 2781 at = (struct altq_tag *)(mtag+1); 2782 at->qid = altq->qid; 2783 if (is_ipv4) 2784 at->af = AF_INET; 2785 else 2786 at->af = AF_LINK; 2787 at->hdr = ip; 2788 m_tag_prepend(m, mtag); 2789 break; 2790 } 2791 2792 case O_LOG: 2793 if (fw_verbose) 2794 ipfw_log(f, hlen, args, m, 2795 oif, offset, tablearg); 2796 match = 1; 2797 break; 2798 2799 case O_PROB: 2800 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2801 break; 2802 2803 case O_VERREVPATH: 2804 /* Outgoing packets automatically pass/match */ 2805 match = ((oif != NULL) || 2806 (m->m_pkthdr.rcvif == NULL) || 2807 ( 2808#ifdef INET6 2809 is_ipv6 ? 2810 verify_path6(&(args->f_id.src_ip6), 2811 m->m_pkthdr.rcvif) : 2812#endif 2813 verify_path(src_ip, m->m_pkthdr.rcvif))); 2814 break; 2815 2816 case O_VERSRCREACH: 2817 /* Outgoing packets automatically pass/match */ 2818 match = (hlen > 0 && ((oif != NULL) || 2819#ifdef INET6 2820 is_ipv6 ? 2821 verify_path6(&(args->f_id.src_ip6), 2822 NULL) : 2823#endif 2824 verify_path(src_ip, NULL))); 2825 break; 2826 2827 case O_ANTISPOOF: 2828 /* Outgoing packets automatically pass/match */ 2829 if (oif == NULL && hlen > 0 && 2830 ( (is_ipv4 && in_localaddr(src_ip)) 2831#ifdef INET6 2832 || (is_ipv6 && 2833 in6_localaddr(&(args->f_id.src_ip6))) 2834#endif 2835 )) 2836 match = 2837#ifdef INET6 2838 is_ipv6 ? verify_path6( 2839 &(args->f_id.src_ip6), 2840 m->m_pkthdr.rcvif) : 2841#endif 2842 verify_path(src_ip, 2843 m->m_pkthdr.rcvif); 2844 else 2845 match = 1; 2846 break; 2847 2848 case O_IPSEC: 2849#ifdef FAST_IPSEC 2850 match = (m_tag_find(m, 2851 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2852#endif 2853#ifdef IPSEC 2854 match = (ipsec_getnhist(m) != 0); 2855#endif 2856 /* otherwise no match */ 2857 break; 2858 2859#ifdef INET6 2860 case O_IP6_SRC: 2861 match = is_ipv6 && 2862 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2863 &((ipfw_insn_ip6 *)cmd)->addr6); 2864 break; 2865 2866 case O_IP6_DST: 2867 match = is_ipv6 && 2868 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2869 &((ipfw_insn_ip6 *)cmd)->addr6); 2870 break; 2871 case O_IP6_SRC_MASK: 2872 if (is_ipv6) { 2873 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2874 struct in6_addr p = args->f_id.src_ip6; 2875 2876 APPLY_MASK(&p, &te->mask6); 2877 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2878 } 2879 break; 2880 2881 case O_IP6_DST_MASK: 2882 if (is_ipv6) { 2883 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2884 struct in6_addr p = args->f_id.dst_ip6; 2885 2886 APPLY_MASK(&p, &te->mask6); 2887 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2888 } 2889 break; 2890 2891 case O_IP6_SRC_ME: 2892 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2893 break; 2894 2895 case O_IP6_DST_ME: 2896 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2897 break; 2898 2899 case O_FLOW6ID: 2900 match = is_ipv6 && 2901 flow6id_match(args->f_id.flow_id6, 2902 (ipfw_insn_u32 *) cmd); 2903 break; 2904 2905 case O_EXT_HDR: 2906 match = is_ipv6 && 2907 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2908 break; 2909 2910 case O_IP6: 2911 match = is_ipv6; 2912 break; 2913#endif 2914 2915 case O_IP4: 2916 match = is_ipv4; 2917 break; 2918 2919 case O_TAG: { 2920 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2921 tablearg : cmd->arg1; 2922 2923 /* Packet is already tagged with this tag? */ 2924 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2925 2926 /* We have `untag' action when F_NOT flag is 2927 * present. And we must remove this mtag from 2928 * mbuf and reset `match' to zero (`match' will 2929 * be inversed later). 2930 * Otherwise we should allocate new mtag and 2931 * push it into mbuf. 2932 */ 2933 if (cmd->len & F_NOT) { /* `untag' action */ 2934 if (mtag != NULL) 2935 m_tag_delete(m, mtag); 2936 } else if (mtag == NULL) { 2937 if ((mtag = m_tag_alloc(MTAG_IPFW, 2938 tag, 0, M_NOWAIT)) != NULL) 2939 m_tag_prepend(m, mtag); 2940 } 2941 match = (cmd->len & F_NOT) ? 0: 1; 2942 break; 2943 } 2944 2945 case O_TAGGED: { 2946 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2947 tablearg : cmd->arg1; 2948 2949 if (cmdlen == 1) { 2950 match = m_tag_locate(m, MTAG_IPFW, 2951 tag, NULL) != NULL; 2952 break; 2953 } 2954 2955 /* we have ranges */ 2956 for (mtag = m_tag_first(m); 2957 mtag != NULL && !match; 2958 mtag = m_tag_next(m, mtag)) { 2959 uint16_t *p; 2960 int i; 2961 2962 if (mtag->m_tag_cookie != MTAG_IPFW) 2963 continue; 2964 2965 p = ((ipfw_insn_u16 *)cmd)->ports; 2966 i = cmdlen - 1; 2967 for(; !match && i > 0; i--, p += 2) 2968 match = 2969 mtag->m_tag_id >= p[0] && 2970 mtag->m_tag_id <= p[1]; 2971 } 2972 break; 2973 } 2974 2975 /* 2976 * The second set of opcodes represents 'actions', 2977 * i.e. the terminal part of a rule once the packet 2978 * matches all previous patterns. 2979 * Typically there is only one action for each rule, 2980 * and the opcode is stored at the end of the rule 2981 * (but there are exceptions -- see below). 2982 * 2983 * In general, here we set retval and terminate the 2984 * outer loop (would be a 'break 3' in some language, 2985 * but we need to do a 'goto done'). 2986 * 2987 * Exceptions: 2988 * O_COUNT and O_SKIPTO actions: 2989 * instead of terminating, we jump to the next rule 2990 * ('goto next_rule', equivalent to a 'break 2'), 2991 * or to the SKIPTO target ('goto again' after 2992 * having set f, cmd and l), respectively. 2993 * 2994 * O_TAG, O_LOG and O_ALTQ action parameters: 2995 * perform some action and set match = 1; 2996 * 2997 * O_LIMIT and O_KEEP_STATE: these opcodes are 2998 * not real 'actions', and are stored right 2999 * before the 'action' part of the rule. 3000 * These opcodes try to install an entry in the 3001 * state tables; if successful, we continue with 3002 * the next opcode (match=1; break;), otherwise 3003 * the packet * must be dropped 3004 * ('goto done' after setting retval); 3005 * 3006 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3007 * cause a lookup of the state table, and a jump 3008 * to the 'action' part of the parent rule 3009 * ('goto check_body') if an entry is found, or 3010 * (CHECK_STATE only) a jump to the next rule if 3011 * the entry is not found ('goto next_rule'). 3012 * The result of the lookup is cached to make 3013 * further instances of these opcodes are 3014 * effectively NOPs. 3015 */ 3016 case O_LIMIT: 3017 case O_KEEP_STATE: 3018 if (install_state(f, 3019 (ipfw_insn_limit *)cmd, args, tablearg)) { 3020 retval = IP_FW_DENY; 3021 goto done; /* error/limit violation */ 3022 } 3023 match = 1; 3024 break; 3025 3026 case O_PROBE_STATE: 3027 case O_CHECK_STATE: 3028 /* 3029 * dynamic rules are checked at the first 3030 * keep-state or check-state occurrence, 3031 * with the result being stored in dyn_dir. 3032 * The compiler introduces a PROBE_STATE 3033 * instruction for us when we have a 3034 * KEEP_STATE (because PROBE_STATE needs 3035 * to be run first). 3036 */ 3037 if (dyn_dir == MATCH_UNKNOWN && 3038 (q = lookup_dyn_rule(&args->f_id, 3039 &dyn_dir, proto == IPPROTO_TCP ? 3040 TCP(ulp) : NULL)) 3041 != NULL) { 3042 /* 3043 * Found dynamic entry, update stats 3044 * and jump to the 'action' part of 3045 * the parent rule. 3046 */ 3047 q->pcnt++; 3048 q->bcnt += pktlen; 3049 f = q->rule; 3050 cmd = ACTION_PTR(f); 3051 l = f->cmd_len - f->act_ofs; 3052 IPFW_DYN_UNLOCK(); 3053 goto check_body; 3054 } 3055 /* 3056 * Dynamic entry not found. If CHECK_STATE, 3057 * skip to next rule, if PROBE_STATE just 3058 * ignore and continue with next opcode. 3059 */ 3060 if (cmd->opcode == O_CHECK_STATE) 3061 goto next_rule; 3062 match = 1; 3063 break; 3064 3065 case O_ACCEPT: 3066 retval = 0; /* accept */ 3067 goto done; 3068 3069 case O_PIPE: 3070 case O_QUEUE: 3071 args->rule = f; /* report matching rule */ 3072 if (cmd->arg1 == IP_FW_TABLEARG) 3073 args->cookie = tablearg; 3074 else 3075 args->cookie = cmd->arg1; 3076 retval = IP_FW_DUMMYNET; 3077 goto done; 3078 3079 case O_DIVERT: 3080 case O_TEE: { 3081 struct divert_tag *dt; 3082 3083 if (args->eh) /* not on layer 2 */ 3084 break; 3085 mtag = m_tag_get(PACKET_TAG_DIVERT, 3086 sizeof(struct divert_tag), 3087 M_NOWAIT); 3088 if (mtag == NULL) { 3089 /* XXX statistic */ 3090 /* drop packet */ 3091 IPFW_RUNLOCK(chain); 3092 return (IP_FW_DENY); 3093 } 3094 dt = (struct divert_tag *)(mtag+1); 3095 dt->cookie = f->rulenum; 3096 if (cmd->arg1 == IP_FW_TABLEARG) 3097 dt->info = tablearg; 3098 else 3099 dt->info = cmd->arg1; 3100 m_tag_prepend(m, mtag); 3101 retval = (cmd->opcode == O_DIVERT) ? 3102 IP_FW_DIVERT : IP_FW_TEE; 3103 goto done; 3104 } 3105 3106 case O_COUNT: 3107 case O_SKIPTO: 3108 f->pcnt++; /* update stats */ 3109 f->bcnt += pktlen; 3110 f->timestamp = time_uptime; 3111 if (cmd->opcode == O_COUNT) 3112 goto next_rule; 3113 /* handle skipto */ 3114 if (f->next_rule == NULL) 3115 lookup_next_rule(f); 3116 f = f->next_rule; 3117 goto again; 3118 3119 case O_REJECT: 3120 /* 3121 * Drop the packet and send a reject notice 3122 * if the packet is not ICMP (or is an ICMP 3123 * query), and it is not multicast/broadcast. 3124 */ 3125 if (hlen > 0 && is_ipv4 && offset == 0 && 3126 (proto != IPPROTO_ICMP || 3127 is_icmp_query(ICMP(ulp))) && 3128 !(m->m_flags & (M_BCAST|M_MCAST)) && 3129 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3130 send_reject(args, cmd->arg1, ip_len); 3131 m = args->m; 3132 } 3133 /* FALLTHROUGH */ 3134#ifdef INET6 3135 case O_UNREACH6: 3136 if (hlen > 0 && is_ipv6 && 3137 ((offset & IP6F_OFF_MASK) == 0) && 3138 (proto != IPPROTO_ICMPV6 || 3139 (is_icmp6_query(args->f_id.flags) == 1)) && 3140 !(m->m_flags & (M_BCAST|M_MCAST)) && 3141 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3142 send_reject6(args, cmd->arg1, hlen); 3143 m = args->m; 3144 } 3145 /* FALLTHROUGH */ 3146#endif 3147 case O_DENY: 3148 retval = IP_FW_DENY; 3149 goto done; 3150 3151 case O_FORWARD_IP: { 3152 struct sockaddr_in *sa; 3153 sa = &(((ipfw_insn_sa *)cmd)->sa); 3154 if (args->eh) /* not valid on layer2 pkts */ 3155 break; 3156 if (!q || dyn_dir == MATCH_FORWARD) { 3157 if (sa->sin_addr.s_addr == INADDR_ANY) { 3158 bcopy(sa, &args->hopstore, 3159 sizeof(*sa)); 3160 args->hopstore.sin_addr.s_addr = 3161 htonl(tablearg); 3162 args->next_hop = 3163 &args->hopstore; 3164 } else { 3165 args->next_hop = sa; 3166 } 3167 } 3168 retval = IP_FW_PASS; 3169 } 3170 goto done; 3171 3172 case O_NETGRAPH: 3173 case O_NGTEE: 3174 args->rule = f; /* report matching rule */ 3175 if (cmd->arg1 == IP_FW_TABLEARG) 3176 args->cookie = tablearg; 3177 else 3178 args->cookie = cmd->arg1; 3179 retval = (cmd->opcode == O_NETGRAPH) ? 3180 IP_FW_NETGRAPH : IP_FW_NGTEE; 3181 goto done; 3182 3183 default: 3184 panic("-- unknown opcode %d\n", cmd->opcode); 3185 } /* end of switch() on opcodes */ 3186 3187 if (cmd->len & F_NOT) 3188 match = !match; 3189 3190 if (match) { 3191 if (cmd->len & F_OR) 3192 skip_or = 1; 3193 } else { 3194 if (!(cmd->len & F_OR)) /* not an OR block, */ 3195 break; /* try next rule */ 3196 } 3197 3198 } /* end of inner for, scan opcodes */ 3199 3200next_rule:; /* try next rule */ 3201 3202 } /* end of outer for, scan rules */ 3203 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3204 IPFW_RUNLOCK(chain); 3205 return (IP_FW_DENY); 3206 3207done: 3208 /* Update statistics */ 3209 f->pcnt++; 3210 f->bcnt += pktlen; 3211 f->timestamp = time_uptime; 3212 IPFW_RUNLOCK(chain); 3213 return (retval); 3214 3215pullup_failed: 3216 if (fw_verbose) 3217 printf("ipfw: pullup failed\n"); 3218 return (IP_FW_DENY); 3219} 3220 3221/* 3222 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3223 * These will be reconstructed on the fly as packets are matched. 3224 */ 3225static void 3226flush_rule_ptrs(struct ip_fw_chain *chain) 3227{ 3228 struct ip_fw *rule; 3229 3230 IPFW_WLOCK_ASSERT(chain); 3231 3232 for (rule = chain->rules; rule; rule = rule->next) 3233 rule->next_rule = NULL; 3234} 3235 3236/* 3237 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3238 * possibly create a rule number and add the rule to the list. 3239 * Update the rule_number in the input struct so the caller knows it as well. 3240 */ 3241static int 3242add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3243{ 3244 struct ip_fw *rule, *f, *prev; 3245 int l = RULESIZE(input_rule); 3246 3247 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3248 return (EINVAL); 3249 3250 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3251 if (rule == NULL) 3252 return (ENOSPC); 3253 3254 bcopy(input_rule, rule, l); 3255 3256 rule->next = NULL; 3257 rule->next_rule = NULL; 3258 3259 rule->pcnt = 0; 3260 rule->bcnt = 0; 3261 rule->timestamp = 0; 3262 3263 IPFW_WLOCK(chain); 3264 3265 if (chain->rules == NULL) { /* default rule */ 3266 chain->rules = rule; 3267 goto done; 3268 } 3269 3270 /* 3271 * If rulenum is 0, find highest numbered rule before the 3272 * default rule, and add autoinc_step 3273 */ 3274 if (autoinc_step < 1) 3275 autoinc_step = 1; 3276 else if (autoinc_step > 1000) 3277 autoinc_step = 1000; 3278 if (rule->rulenum == 0) { 3279 /* 3280 * locate the highest numbered rule before default 3281 */ 3282 for (f = chain->rules; f; f = f->next) { 3283 if (f->rulenum == IPFW_DEFAULT_RULE) 3284 break; 3285 rule->rulenum = f->rulenum; 3286 } 3287 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3288 rule->rulenum += autoinc_step; 3289 input_rule->rulenum = rule->rulenum; 3290 } 3291 3292 /* 3293 * Now insert the new rule in the right place in the sorted list. 3294 */ 3295 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3296 if (f->rulenum > rule->rulenum) { /* found the location */ 3297 if (prev) { 3298 rule->next = f; 3299 prev->next = rule; 3300 } else { /* head insert */ 3301 rule->next = chain->rules; 3302 chain->rules = rule; 3303 } 3304 break; 3305 } 3306 } 3307 flush_rule_ptrs(chain); 3308done: 3309 static_count++; 3310 static_len += l; 3311 IPFW_WUNLOCK(chain); 3312 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3313 rule->rulenum, static_count);) 3314 return (0); 3315} 3316 3317/** 3318 * Remove a static rule (including derived * dynamic rules) 3319 * and place it on the ``reap list'' for later reclamation. 3320 * The caller is in charge of clearing rule pointers to avoid 3321 * dangling pointers. 3322 * @return a pointer to the next entry. 3323 * Arguments are not checked, so they better be correct. 3324 */ 3325static struct ip_fw * 3326remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3327{ 3328 struct ip_fw *n; 3329 int l = RULESIZE(rule); 3330 3331 IPFW_WLOCK_ASSERT(chain); 3332 3333 n = rule->next; 3334 IPFW_DYN_LOCK(); 3335 remove_dyn_rule(rule, NULL /* force removal */); 3336 IPFW_DYN_UNLOCK(); 3337 if (prev == NULL) 3338 chain->rules = n; 3339 else 3340 prev->next = n; 3341 static_count--; 3342 static_len -= l; 3343 3344 rule->next = chain->reap; 3345 chain->reap = rule; 3346 3347 return n; 3348} 3349 3350/** 3351 * Reclaim storage associated with a list of rules. This is 3352 * typically the list created using remove_rule. 3353 */ 3354static void 3355reap_rules(struct ip_fw *head) 3356{ 3357 struct ip_fw *rule; 3358 3359 while ((rule = head) != NULL) { 3360 head = head->next; 3361 if (DUMMYNET_LOADED) 3362 ip_dn_ruledel_ptr(rule); 3363 free(rule, M_IPFW); 3364 } 3365} 3366 3367/* 3368 * Remove all rules from a chain (except rules in set RESVD_SET 3369 * unless kill_default = 1). The caller is responsible for 3370 * reclaiming storage for the rules left in chain->reap. 3371 */ 3372static void 3373free_chain(struct ip_fw_chain *chain, int kill_default) 3374{ 3375 struct ip_fw *prev, *rule; 3376 3377 IPFW_WLOCK_ASSERT(chain); 3378 3379 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3380 for (prev = NULL, rule = chain->rules; rule ; ) 3381 if (kill_default || rule->set != RESVD_SET) 3382 rule = remove_rule(chain, rule, prev); 3383 else { 3384 prev = rule; 3385 rule = rule->next; 3386 } 3387} 3388 3389/** 3390 * Remove all rules with given number, and also do set manipulation. 3391 * Assumes chain != NULL && *chain != NULL. 3392 * 3393 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3394 * the next 8 bits are the new set, the top 8 bits are the command: 3395 * 3396 * 0 delete rules with given number 3397 * 1 delete rules with given set number 3398 * 2 move rules with given number to new set 3399 * 3 move rules with given set number to new set 3400 * 4 swap sets with given numbers 3401 */ 3402static int 3403del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3404{ 3405 struct ip_fw *prev = NULL, *rule; 3406 u_int16_t rulenum; /* rule or old_set */ 3407 u_int8_t cmd, new_set; 3408 3409 rulenum = arg & 0xffff; 3410 cmd = (arg >> 24) & 0xff; 3411 new_set = (arg >> 16) & 0xff; 3412 3413 if (cmd > 4) 3414 return EINVAL; 3415 if (new_set > RESVD_SET) 3416 return EINVAL; 3417 if (cmd == 0 || cmd == 2) { 3418 if (rulenum >= IPFW_DEFAULT_RULE) 3419 return EINVAL; 3420 } else { 3421 if (rulenum > RESVD_SET) /* old_set */ 3422 return EINVAL; 3423 } 3424 3425 IPFW_WLOCK(chain); 3426 rule = chain->rules; 3427 chain->reap = NULL; 3428 switch (cmd) { 3429 case 0: /* delete rules with given number */ 3430 /* 3431 * locate first rule to delete 3432 */ 3433 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3434 ; 3435 if (rule->rulenum != rulenum) { 3436 IPFW_WUNLOCK(chain); 3437 return EINVAL; 3438 } 3439 3440 /* 3441 * flush pointers outside the loop, then delete all matching 3442 * rules. prev remains the same throughout the cycle. 3443 */ 3444 flush_rule_ptrs(chain); 3445 while (rule->rulenum == rulenum) 3446 rule = remove_rule(chain, rule, prev); 3447 break; 3448 3449 case 1: /* delete all rules with given set number */ 3450 flush_rule_ptrs(chain); 3451 rule = chain->rules; 3452 while (rule->rulenum < IPFW_DEFAULT_RULE) 3453 if (rule->set == rulenum) 3454 rule = remove_rule(chain, rule, prev); 3455 else { 3456 prev = rule; 3457 rule = rule->next; 3458 } 3459 break; 3460 3461 case 2: /* move rules with given number to new set */ 3462 rule = chain->rules; 3463 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3464 if (rule->rulenum == rulenum) 3465 rule->set = new_set; 3466 break; 3467 3468 case 3: /* move rules with given set number to new set */ 3469 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3470 if (rule->set == rulenum) 3471 rule->set = new_set; 3472 break; 3473 3474 case 4: /* swap two sets */ 3475 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3476 if (rule->set == rulenum) 3477 rule->set = new_set; 3478 else if (rule->set == new_set) 3479 rule->set = rulenum; 3480 break; 3481 } 3482 /* 3483 * Look for rules to reclaim. We grab the list before 3484 * releasing the lock then reclaim them w/o the lock to 3485 * avoid a LOR with dummynet. 3486 */ 3487 rule = chain->reap; 3488 chain->reap = NULL; 3489 IPFW_WUNLOCK(chain); 3490 if (rule) 3491 reap_rules(rule); 3492 return 0; 3493} 3494 3495/* 3496 * Clear counters for a specific rule. 3497 * The enclosing "table" is assumed locked. 3498 */ 3499static void 3500clear_counters(struct ip_fw *rule, int log_only) 3501{ 3502 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3503 3504 if (log_only == 0) { 3505 rule->bcnt = rule->pcnt = 0; 3506 rule->timestamp = 0; 3507 } 3508 if (l->o.opcode == O_LOG) 3509 l->log_left = l->max_log; 3510} 3511 3512/** 3513 * Reset some or all counters on firewall rules. 3514 * @arg frwl is null to clear all entries, or contains a specific 3515 * rule number. 3516 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3517 */ 3518static int 3519zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3520{ 3521 struct ip_fw *rule; 3522 char *msg; 3523 3524 IPFW_WLOCK(chain); 3525 if (rulenum == 0) { 3526 norule_counter = 0; 3527 for (rule = chain->rules; rule; rule = rule->next) 3528 clear_counters(rule, log_only); 3529 msg = log_only ? "ipfw: All logging counts reset.\n" : 3530 "ipfw: Accounting cleared.\n"; 3531 } else { 3532 int cleared = 0; 3533 /* 3534 * We can have multiple rules with the same number, so we 3535 * need to clear them all. 3536 */ 3537 for (rule = chain->rules; rule; rule = rule->next) 3538 if (rule->rulenum == rulenum) { 3539 while (rule && rule->rulenum == rulenum) { 3540 clear_counters(rule, log_only); 3541 rule = rule->next; 3542 } 3543 cleared = 1; 3544 break; 3545 } 3546 if (!cleared) { /* we did not find any matching rules */ 3547 IPFW_WUNLOCK(chain); 3548 return (EINVAL); 3549 } 3550 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3551 "ipfw: Entry %d cleared.\n"; 3552 } 3553 IPFW_WUNLOCK(chain); 3554 3555 if (fw_verbose) 3556 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3557 return (0); 3558} 3559 3560/* 3561 * Check validity of the structure before insert. 3562 * Fortunately rules are simple, so this mostly need to check rule sizes. 3563 */ 3564static int 3565check_ipfw_struct(struct ip_fw *rule, int size) 3566{ 3567 int l, cmdlen = 0; 3568 int have_action=0; 3569 ipfw_insn *cmd; 3570 3571 if (size < sizeof(*rule)) { 3572 printf("ipfw: rule too short\n"); 3573 return (EINVAL); 3574 } 3575 /* first, check for valid size */ 3576 l = RULESIZE(rule); 3577 if (l != size) { 3578 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3579 return (EINVAL); 3580 } 3581 if (rule->act_ofs >= rule->cmd_len) { 3582 printf("ipfw: bogus action offset (%u > %u)\n", 3583 rule->act_ofs, rule->cmd_len - 1); 3584 return (EINVAL); 3585 } 3586 /* 3587 * Now go for the individual checks. Very simple ones, basically only 3588 * instruction sizes. 3589 */ 3590 for (l = rule->cmd_len, cmd = rule->cmd ; 3591 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3592 cmdlen = F_LEN(cmd); 3593 if (cmdlen > l) { 3594 printf("ipfw: opcode %d size truncated\n", 3595 cmd->opcode); 3596 return EINVAL; 3597 } 3598 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3599 switch (cmd->opcode) { 3600 case O_PROBE_STATE: 3601 case O_KEEP_STATE: 3602 case O_PROTO: 3603 case O_IP_SRC_ME: 3604 case O_IP_DST_ME: 3605 case O_LAYER2: 3606 case O_IN: 3607 case O_FRAG: 3608 case O_DIVERTED: 3609 case O_IPOPT: 3610 case O_IPTOS: 3611 case O_IPPRECEDENCE: 3612 case O_IPVER: 3613 case O_TCPWIN: 3614 case O_TCPFLAGS: 3615 case O_TCPOPTS: 3616 case O_ESTAB: 3617 case O_VERREVPATH: 3618 case O_VERSRCREACH: 3619 case O_ANTISPOOF: 3620 case O_IPSEC: 3621#ifdef INET6 3622 case O_IP6_SRC_ME: 3623 case O_IP6_DST_ME: 3624 case O_EXT_HDR: 3625 case O_IP6: 3626#endif 3627 case O_IP4: 3628 case O_TAG: 3629 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3630 goto bad_size; 3631 break; 3632 3633 case O_UID: 3634 case O_GID: 3635 case O_JAIL: 3636 case O_IP_SRC: 3637 case O_IP_DST: 3638 case O_TCPSEQ: 3639 case O_TCPACK: 3640 case O_PROB: 3641 case O_ICMPTYPE: 3642 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3643 goto bad_size; 3644 break; 3645 3646 case O_LIMIT: 3647 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3648 goto bad_size; 3649 break; 3650 3651 case O_LOG: 3652 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3653 goto bad_size; 3654 3655 ((ipfw_insn_log *)cmd)->log_left = 3656 ((ipfw_insn_log *)cmd)->max_log; 3657 3658 break; 3659 3660 case O_IP_SRC_MASK: 3661 case O_IP_DST_MASK: 3662 /* only odd command lengths */ 3663 if ( !(cmdlen & 1) || cmdlen > 31) 3664 goto bad_size; 3665 break; 3666 3667 case O_IP_SRC_SET: 3668 case O_IP_DST_SET: 3669 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3670 printf("ipfw: invalid set size %d\n", 3671 cmd->arg1); 3672 return EINVAL; 3673 } 3674 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3675 (cmd->arg1+31)/32 ) 3676 goto bad_size; 3677 break; 3678 3679 case O_IP_SRC_LOOKUP: 3680 case O_IP_DST_LOOKUP: 3681 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3682 printf("ipfw: invalid table number %d\n", 3683 cmd->arg1); 3684 return (EINVAL); 3685 } 3686 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3687 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3688 goto bad_size; 3689 break; 3690 3691 case O_MACADDR2: 3692 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3693 goto bad_size; 3694 break; 3695 3696 case O_NOP: 3697 case O_IPID: 3698 case O_IPTTL: 3699 case O_IPLEN: 3700 case O_TCPDATALEN: 3701 case O_TAGGED: 3702 if (cmdlen < 1 || cmdlen > 31) 3703 goto bad_size; 3704 break; 3705 3706 case O_MAC_TYPE: 3707 case O_IP_SRCPORT: 3708 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3709 if (cmdlen < 2 || cmdlen > 31) 3710 goto bad_size; 3711 break; 3712 3713 case O_RECV: 3714 case O_XMIT: 3715 case O_VIA: 3716 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3717 goto bad_size; 3718 break; 3719 3720 case O_ALTQ: 3721 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3722 goto bad_size; 3723 break; 3724 3725 case O_PIPE: 3726 case O_QUEUE: 3727 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3728 goto bad_size; 3729 goto check_action; 3730 3731 case O_FORWARD_IP: 3732#ifdef IPFIREWALL_FORWARD 3733 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3734 goto bad_size; 3735 goto check_action; 3736#else 3737 return EINVAL; 3738#endif 3739 3740 case O_DIVERT: 3741 case O_TEE: 3742 if (ip_divert_ptr == NULL) 3743 return EINVAL; 3744 else 3745 goto check_size; 3746 case O_NETGRAPH: 3747 case O_NGTEE: 3748 if (!NG_IPFW_LOADED) 3749 return EINVAL; 3750 else 3751 goto check_size; 3752 case O_FORWARD_MAC: /* XXX not implemented yet */ 3753 case O_CHECK_STATE: 3754 case O_COUNT: 3755 case O_ACCEPT: 3756 case O_DENY: 3757 case O_REJECT: 3758#ifdef INET6 3759 case O_UNREACH6: 3760#endif 3761 case O_SKIPTO: 3762check_size: 3763 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3764 goto bad_size; 3765check_action: 3766 if (have_action) { 3767 printf("ipfw: opcode %d, multiple actions" 3768 " not allowed\n", 3769 cmd->opcode); 3770 return EINVAL; 3771 } 3772 have_action = 1; 3773 if (l != cmdlen) { 3774 printf("ipfw: opcode %d, action must be" 3775 " last opcode\n", 3776 cmd->opcode); 3777 return EINVAL; 3778 } 3779 break; 3780#ifdef INET6 3781 case O_IP6_SRC: 3782 case O_IP6_DST: 3783 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3784 F_INSN_SIZE(ipfw_insn)) 3785 goto bad_size; 3786 break; 3787 3788 case O_FLOW6ID: 3789 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3790 ((ipfw_insn_u32 *)cmd)->o.arg1) 3791 goto bad_size; 3792 break; 3793 3794 case O_IP6_SRC_MASK: 3795 case O_IP6_DST_MASK: 3796 if ( !(cmdlen & 1) || cmdlen > 127) 3797 goto bad_size; 3798 break; 3799 case O_ICMP6TYPE: 3800 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3801 goto bad_size; 3802 break; 3803#endif 3804 3805 default: 3806 switch (cmd->opcode) { 3807#ifndef INET6 3808 case O_IP6_SRC_ME: 3809 case O_IP6_DST_ME: 3810 case O_EXT_HDR: 3811 case O_IP6: 3812 case O_UNREACH6: 3813 case O_IP6_SRC: 3814 case O_IP6_DST: 3815 case O_FLOW6ID: 3816 case O_IP6_SRC_MASK: 3817 case O_IP6_DST_MASK: 3818 case O_ICMP6TYPE: 3819 printf("ipfw: no IPv6 support in kernel\n"); 3820 return EPROTONOSUPPORT; 3821#endif 3822 default: 3823 printf("ipfw: opcode %d, unknown opcode\n", 3824 cmd->opcode); 3825 return EINVAL; 3826 } 3827 } 3828 } 3829 if (have_action == 0) { 3830 printf("ipfw: missing action\n"); 3831 return EINVAL; 3832 } 3833 return 0; 3834 3835bad_size: 3836 printf("ipfw: opcode %d size %d wrong\n", 3837 cmd->opcode, cmdlen); 3838 return EINVAL; 3839} 3840 3841/* 3842 * Copy the static and dynamic rules to the supplied buffer 3843 * and return the amount of space actually used. 3844 */ 3845static size_t 3846ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3847{ 3848 char *bp = buf; 3849 char *ep = bp + space; 3850 struct ip_fw *rule; 3851 int i; 3852 3853 /* XXX this can take a long time and locking will block packet flow */ 3854 IPFW_RLOCK(chain); 3855 for (rule = chain->rules; rule ; rule = rule->next) { 3856 /* 3857 * Verify the entry fits in the buffer in case the 3858 * rules changed between calculating buffer space and 3859 * now. This would be better done using a generation 3860 * number but should suffice for now. 3861 */ 3862 i = RULESIZE(rule); 3863 if (bp + i <= ep) { 3864 bcopy(rule, bp, i); 3865 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3866 sizeof(set_disable)); 3867 bp += i; 3868 } 3869 } 3870 IPFW_RUNLOCK(chain); 3871 if (ipfw_dyn_v) { 3872 ipfw_dyn_rule *p, *last = NULL; 3873 3874 IPFW_DYN_LOCK(); 3875 for (i = 0 ; i < curr_dyn_buckets; i++) 3876 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3877 if (bp + sizeof *p <= ep) { 3878 ipfw_dyn_rule *dst = 3879 (ipfw_dyn_rule *)bp; 3880 bcopy(p, dst, sizeof *p); 3881 bcopy(&(p->rule->rulenum), &(dst->rule), 3882 sizeof(p->rule->rulenum)); 3883 /* 3884 * store a non-null value in "next". 3885 * The userland code will interpret a 3886 * NULL here as a marker 3887 * for the last dynamic rule. 3888 */ 3889 bcopy(&dst, &dst->next, sizeof(dst)); 3890 last = dst; 3891 dst->expire = 3892 TIME_LEQ(dst->expire, time_uptime) ? 3893 0 : dst->expire - time_uptime ; 3894 bp += sizeof(ipfw_dyn_rule); 3895 } 3896 } 3897 IPFW_DYN_UNLOCK(); 3898 if (last != NULL) /* mark last dynamic rule */ 3899 bzero(&last->next, sizeof(last)); 3900 } 3901 return (bp - (char *)buf); 3902} 3903 3904 3905/** 3906 * {set|get}sockopt parser. 3907 */ 3908static int 3909ipfw_ctl(struct sockopt *sopt) 3910{ 3911#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3912 int error, rule_num; 3913 size_t size; 3914 struct ip_fw *buf, *rule; 3915 u_int32_t rulenum[2]; 3916 3917 error = suser(sopt->sopt_td); 3918 if (error) 3919 return (error); 3920 3921 /* 3922 * Disallow modifications in really-really secure mode, but still allow 3923 * the logging counters to be reset. 3924 */ 3925 if (sopt->sopt_name == IP_FW_ADD || 3926 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3927 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3928 if (error) 3929 return (error); 3930 } 3931 3932 error = 0; 3933 3934 switch (sopt->sopt_name) { 3935 case IP_FW_GET: 3936 /* 3937 * pass up a copy of the current rules. Static rules 3938 * come first (the last of which has number IPFW_DEFAULT_RULE), 3939 * followed by a possibly empty list of dynamic rule. 3940 * The last dynamic rule has NULL in the "next" field. 3941 * 3942 * Note that the calculated size is used to bound the 3943 * amount of data returned to the user. The rule set may 3944 * change between calculating the size and returning the 3945 * data in which case we'll just return what fits. 3946 */ 3947 size = static_len; /* size of static rules */ 3948 if (ipfw_dyn_v) /* add size of dyn.rules */ 3949 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3950 3951 /* 3952 * XXX todo: if the user passes a short length just to know 3953 * how much room is needed, do not bother filling up the 3954 * buffer, just jump to the sooptcopyout. 3955 */ 3956 buf = malloc(size, M_TEMP, M_WAITOK); 3957 error = sooptcopyout(sopt, buf, 3958 ipfw_getrules(&layer3_chain, buf, size)); 3959 free(buf, M_TEMP); 3960 break; 3961 3962 case IP_FW_FLUSH: 3963 /* 3964 * Normally we cannot release the lock on each iteration. 3965 * We could do it here only because we start from the head all 3966 * the times so there is no risk of missing some entries. 3967 * On the other hand, the risk is that we end up with 3968 * a very inconsistent ruleset, so better keep the lock 3969 * around the whole cycle. 3970 * 3971 * XXX this code can be improved by resetting the head of 3972 * the list to point to the default rule, and then freeing 3973 * the old list without the need for a lock. 3974 */ 3975 3976 IPFW_WLOCK(&layer3_chain); 3977 layer3_chain.reap = NULL; 3978 free_chain(&layer3_chain, 0 /* keep default rule */); 3979 rule = layer3_chain.reap; 3980 layer3_chain.reap = NULL; 3981 IPFW_WUNLOCK(&layer3_chain); 3982 if (rule != NULL) 3983 reap_rules(rule); 3984 break; 3985 3986 case IP_FW_ADD: 3987 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3988 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3989 sizeof(struct ip_fw) ); 3990 if (error == 0) 3991 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3992 if (error == 0) { 3993 error = add_rule(&layer3_chain, rule); 3994 size = RULESIZE(rule); 3995 if (!error && sopt->sopt_dir == SOPT_GET) 3996 error = sooptcopyout(sopt, rule, size); 3997 } 3998 free(rule, M_TEMP); 3999 break; 4000 4001 case IP_FW_DEL: 4002 /* 4003 * IP_FW_DEL is used for deleting single rules or sets, 4004 * and (ab)used to atomically manipulate sets. Argument size 4005 * is used to distinguish between the two: 4006 * sizeof(u_int32_t) 4007 * delete single rule or set of rules, 4008 * or reassign rules (or sets) to a different set. 4009 * 2*sizeof(u_int32_t) 4010 * atomic disable/enable sets. 4011 * first u_int32_t contains sets to be disabled, 4012 * second u_int32_t contains sets to be enabled. 4013 */ 4014 error = sooptcopyin(sopt, rulenum, 4015 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4016 if (error) 4017 break; 4018 size = sopt->sopt_valsize; 4019 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4020 error = del_entry(&layer3_chain, rulenum[0]); 4021 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4022 set_disable = 4023 (set_disable | rulenum[0]) & ~rulenum[1] & 4024 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4025 else 4026 error = EINVAL; 4027 break; 4028 4029 case IP_FW_ZERO: 4030 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 4031 rule_num = 0; 4032 if (sopt->sopt_val != 0) { 4033 error = sooptcopyin(sopt, &rule_num, 4034 sizeof(int), sizeof(int)); 4035 if (error) 4036 break; 4037 } 4038 error = zero_entry(&layer3_chain, rule_num, 4039 sopt->sopt_name == IP_FW_RESETLOG); 4040 break; 4041 4042 case IP_FW_TABLE_ADD: 4043 { 4044 ipfw_table_entry ent; 4045 4046 error = sooptcopyin(sopt, &ent, 4047 sizeof(ent), sizeof(ent)); 4048 if (error) 4049 break; 4050 error = add_table_entry(&layer3_chain, ent.tbl, 4051 ent.addr, ent.masklen, ent.value); 4052 } 4053 break; 4054 4055 case IP_FW_TABLE_DEL: 4056 { 4057 ipfw_table_entry ent; 4058 4059 error = sooptcopyin(sopt, &ent, 4060 sizeof(ent), sizeof(ent)); 4061 if (error) 4062 break; 4063 error = del_table_entry(&layer3_chain, ent.tbl, 4064 ent.addr, ent.masklen); 4065 } 4066 break; 4067 4068 case IP_FW_TABLE_FLUSH: 4069 { 4070 u_int16_t tbl; 4071 4072 error = sooptcopyin(sopt, &tbl, 4073 sizeof(tbl), sizeof(tbl)); 4074 if (error) 4075 break; 4076 IPFW_WLOCK(&layer3_chain); 4077 error = flush_table(&layer3_chain, tbl); 4078 IPFW_WUNLOCK(&layer3_chain); 4079 } 4080 break; 4081 4082 case IP_FW_TABLE_GETSIZE: 4083 { 4084 u_int32_t tbl, cnt; 4085 4086 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4087 sizeof(tbl)))) 4088 break; 4089 IPFW_RLOCK(&layer3_chain); 4090 error = count_table(&layer3_chain, tbl, &cnt); 4091 IPFW_RUNLOCK(&layer3_chain); 4092 if (error) 4093 break; 4094 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4095 } 4096 break; 4097 4098 case IP_FW_TABLE_LIST: 4099 { 4100 ipfw_table *tbl; 4101 4102 if (sopt->sopt_valsize < sizeof(*tbl)) { 4103 error = EINVAL; 4104 break; 4105 } 4106 size = sopt->sopt_valsize; 4107 tbl = malloc(size, M_TEMP, M_WAITOK); 4108 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4109 if (error) { 4110 free(tbl, M_TEMP); 4111 break; 4112 } 4113 tbl->size = (size - sizeof(*tbl)) / 4114 sizeof(ipfw_table_entry); 4115 IPFW_RLOCK(&layer3_chain); 4116 error = dump_table(&layer3_chain, tbl); 4117 IPFW_RUNLOCK(&layer3_chain); 4118 if (error) { 4119 free(tbl, M_TEMP); 4120 break; 4121 } 4122 error = sooptcopyout(sopt, tbl, size); 4123 free(tbl, M_TEMP); 4124 } 4125 break; 4126 4127 default: 4128 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4129 error = EINVAL; 4130 } 4131 4132 return (error); 4133#undef RULE_MAXSIZE 4134} 4135 4136/** 4137 * dummynet needs a reference to the default rule, because rules can be 4138 * deleted while packets hold a reference to them. When this happens, 4139 * dummynet changes the reference to the default rule (it could well be a 4140 * NULL pointer, but this way we do not need to check for the special 4141 * case, plus here he have info on the default behaviour). 4142 */ 4143struct ip_fw *ip_fw_default_rule; 4144 4145/* 4146 * This procedure is only used to handle keepalives. It is invoked 4147 * every dyn_keepalive_period 4148 */ 4149static void 4150ipfw_tick(void * __unused unused) 4151{ 4152 struct mbuf *m0, *m, *mnext, **mtailp; 4153 int i; 4154 ipfw_dyn_rule *q; 4155 4156 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4157 goto done; 4158 4159 /* 4160 * We make a chain of packets to go out here -- not deferring 4161 * until after we drop the IPFW dynamic rule lock would result 4162 * in a lock order reversal with the normal packet input -> ipfw 4163 * call stack. 4164 */ 4165 m0 = NULL; 4166 mtailp = &m0; 4167 IPFW_DYN_LOCK(); 4168 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4169 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4170 if (q->dyn_type == O_LIMIT_PARENT) 4171 continue; 4172 if (q->id.proto != IPPROTO_TCP) 4173 continue; 4174 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4175 continue; 4176 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4177 q->expire)) 4178 continue; /* too early */ 4179 if (TIME_LEQ(q->expire, time_uptime)) 4180 continue; /* too late, rule expired */ 4181 4182 *mtailp = send_pkt(&(q->id), q->ack_rev - 1, 4183 q->ack_fwd, TH_SYN); 4184 if (*mtailp != NULL) 4185 mtailp = &(*mtailp)->m_nextpkt; 4186 *mtailp = send_pkt(&(q->id), q->ack_fwd - 1, 4187 q->ack_rev, 0); 4188 if (*mtailp != NULL) 4189 mtailp = &(*mtailp)->m_nextpkt; 4190 } 4191 } 4192 IPFW_DYN_UNLOCK(); 4193 for (m = mnext = m0; m != NULL; m = mnext) { 4194 mnext = m->m_nextpkt; 4195 m->m_nextpkt = NULL; 4196 ip_output(m, NULL, NULL, 0, NULL, NULL); 4197 } 4198done: 4199 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4200} 4201 4202int 4203ipfw_init(void) 4204{ 4205 struct ip_fw default_rule; 4206 int error; 4207 4208#ifdef INET6 4209 /* Setup IPv6 fw sysctl tree. */ 4210 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4211 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4212 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4213 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4214 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4215 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4216 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4217 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4218 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4219 &fw_deny_unknown_exthdrs, 0, 4220 "Deny packets with unknown IPv6 Extension Headers"); 4221#endif 4222 4223 layer3_chain.rules = NULL; 4224 IPFW_LOCK_INIT(&layer3_chain); 4225 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4226 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4227 UMA_ALIGN_PTR, 0); 4228 IPFW_DYN_LOCK_INIT(); 4229 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4230 4231 bzero(&default_rule, sizeof default_rule); 4232 4233 default_rule.act_ofs = 0; 4234 default_rule.rulenum = IPFW_DEFAULT_RULE; 4235 default_rule.cmd_len = 1; 4236 default_rule.set = RESVD_SET; 4237 4238 default_rule.cmd[0].len = 1; 4239 default_rule.cmd[0].opcode = 4240#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4241 1 ? O_ACCEPT : 4242#endif 4243 O_DENY; 4244 4245 error = add_rule(&layer3_chain, &default_rule); 4246 if (error != 0) { 4247 printf("ipfw2: error %u initializing default rule " 4248 "(support disabled)\n", error); 4249 IPFW_DYN_LOCK_DESTROY(); 4250 IPFW_LOCK_DESTROY(&layer3_chain); 4251 uma_zdestroy(ipfw_dyn_rule_zone); 4252 return (error); 4253 } 4254 4255 ip_fw_default_rule = layer3_chain.rules; 4256 printf("ipfw2 " 4257#ifdef INET6 4258 "(+ipv6) " 4259#endif 4260 "initialized, divert %s, " 4261 "rule-based forwarding " 4262#ifdef IPFIREWALL_FORWARD 4263 "enabled, " 4264#else 4265 "disabled, " 4266#endif 4267 "default to %s, logging ", 4268#ifdef IPDIVERT 4269 "enabled", 4270#else 4271 "loadable", 4272#endif 4273 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4274 4275#ifdef IPFIREWALL_VERBOSE 4276 fw_verbose = 1; 4277#endif 4278#ifdef IPFIREWALL_VERBOSE_LIMIT 4279 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4280#endif 4281 if (fw_verbose == 0) 4282 printf("disabled\n"); 4283 else if (verbose_limit == 0) 4284 printf("unlimited\n"); 4285 else 4286 printf("limited to %d packets/entry by default\n", 4287 verbose_limit); 4288 4289 error = init_tables(&layer3_chain); 4290 if (error) { 4291 IPFW_DYN_LOCK_DESTROY(); 4292 IPFW_LOCK_DESTROY(&layer3_chain); 4293 uma_zdestroy(ipfw_dyn_rule_zone); 4294 return (error); 4295 } 4296 ip_fw_ctl_ptr = ipfw_ctl; 4297 ip_fw_chk_ptr = ipfw_chk; 4298 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4299 4300 return (0); 4301} 4302 4303void 4304ipfw_destroy(void) 4305{ 4306 struct ip_fw *reap; 4307 4308 ip_fw_chk_ptr = NULL; 4309 ip_fw_ctl_ptr = NULL; 4310 callout_drain(&ipfw_timeout); 4311 IPFW_WLOCK(&layer3_chain); 4312 flush_tables(&layer3_chain); 4313 layer3_chain.reap = NULL; 4314 free_chain(&layer3_chain, 1 /* kill default rule */); 4315 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4316 IPFW_WUNLOCK(&layer3_chain); 4317 if (reap != NULL) 4318 reap_rules(reap); 4319 IPFW_DYN_LOCK_DESTROY(); 4320 uma_zdestroy(ipfw_dyn_rule_zone); 4321 IPFW_LOCK_DESTROY(&layer3_chain); 4322 4323#ifdef INET6 4324 /* Free IPv6 fw sysctl tree. */ 4325 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4326#endif 4327 4328 printf("IP firewall unloaded\n"); 4329} 4330