ip_fw2.c revision 160032
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 160032 2006-06-29 17:09:47Z yar $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdn.h" 38#include "opt_inet.h" 39#ifndef INET 40#error IPFIREWALL requires INET. 41#endif /* INET */ 42#endif 43#include "opt_inet6.h" 44#include "opt_ipsec.h" 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/condvar.h> 49#include <sys/malloc.h> 50#include <sys/mbuf.h> 51#include <sys/kernel.h> 52#include <sys/lock.h> 53#include <sys/jail.h> 54#include <sys/module.h> 55#include <sys/proc.h> 56#include <sys/rwlock.h> 57#include <sys/socket.h> 58#include <sys/socketvar.h> 59#include <sys/sysctl.h> 60#include <sys/syslog.h> 61#include <sys/ucred.h> 62#include <net/if.h> 63#include <net/radix.h> 64#include <net/route.h> 65#include <netinet/in.h> 66#include <netinet/in_systm.h> 67#include <netinet/in_var.h> 68#include <netinet/in_pcb.h> 69#include <netinet/ip.h> 70#include <netinet/ip_var.h> 71#include <netinet/ip_icmp.h> 72#include <netinet/ip_fw.h> 73#include <netinet/ip_divert.h> 74#include <netinet/ip_dummynet.h> 75#include <netinet/tcp.h> 76#include <netinet/tcp_timer.h> 77#include <netinet/tcp_var.h> 78#include <netinet/tcpip.h> 79#include <netinet/udp.h> 80#include <netinet/udp_var.h> 81 82#include <netgraph/ng_ipfw.h> 83 84#include <altq/if_altq.h> 85 86#ifdef IPSEC 87#include <netinet6/ipsec.h> 88#endif 89 90#include <netinet/ip6.h> 91#include <netinet/icmp6.h> 92#ifdef INET6 93#include <netinet6/scope6_var.h> 94#endif 95 96#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 97 98#include <machine/in_cksum.h> /* XXX for in_cksum */ 99 100/* 101 * set_disable contains one bit per set value (0..31). 102 * If the bit is set, all rules with the corresponding set 103 * are disabled. Set RESVD_SET(31) is reserved for the default rule 104 * and rules that are not deleted by the flush command, 105 * and CANNOT be disabled. 106 * Rules in set RESVD_SET can only be deleted explicitly. 107 */ 108static u_int32_t set_disable; 109 110static int fw_verbose; 111static int verbose_limit; 112 113static struct callout ipfw_timeout; 114static uma_zone_t ipfw_dyn_rule_zone; 115#define IPFW_DEFAULT_RULE 65535 116 117/* 118 * Data structure to cache our ucred related 119 * information. This structure only gets used if 120 * the user specified UID/GID based constraints in 121 * a firewall rule. 122 */ 123struct ip_fw_ugid { 124 gid_t fw_groups[NGROUPS]; 125 int fw_ngroups; 126 uid_t fw_uid; 127 int fw_prid; 128}; 129 130#define IPFW_TABLES_MAX 128 131struct ip_fw_chain { 132 struct ip_fw *rules; /* list of rules */ 133 struct ip_fw *reap; /* list of rules to reap */ 134 struct radix_node_head *tables[IPFW_TABLES_MAX]; 135 struct rwlock rwmtx; 136}; 137#define IPFW_LOCK_INIT(_chain) \ 138 rw_init(&(_chain)->rwmtx, "IPFW static rules") 139#define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx) 140#define IPFW_WLOCK_ASSERT(_chain) do { \ 141 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \ 142 NET_ASSERT_GIANT(); \ 143} while (0) 144 145#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 146#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 147#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 148#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 149 150/* 151 * list of rules for layer 3 152 */ 153static struct ip_fw_chain layer3_chain; 154 155MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 156MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 157 158struct table_entry { 159 struct radix_node rn[2]; 160 struct sockaddr_in addr, mask; 161 u_int32_t value; 162}; 163 164static int fw_debug = 1; 165static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 166 167extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 168 169#ifdef SYSCTL_NODE 170SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 171SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 172 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 173 ipfw_chg_hook, "I", "Enable ipfw"); 174SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 175 &autoinc_step, 0, "Rule number autincrement step"); 176SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 177 CTLFLAG_RW | CTLFLAG_SECURE3, 178 &fw_one_pass, 0, 179 "Only do a single pass through ipfw when using dummynet(4)"); 180SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 181 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 182SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 183 CTLFLAG_RW | CTLFLAG_SECURE3, 184 &fw_verbose, 0, "Log matches to ipfw rules"); 185SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 186 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 187 188/* 189 * Description of dynamic rules. 190 * 191 * Dynamic rules are stored in lists accessed through a hash table 192 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 193 * be modified through the sysctl variable dyn_buckets which is 194 * updated when the table becomes empty. 195 * 196 * XXX currently there is only one list, ipfw_dyn. 197 * 198 * When a packet is received, its address fields are first masked 199 * with the mask defined for the rule, then hashed, then matched 200 * against the entries in the corresponding list. 201 * Dynamic rules can be used for different purposes: 202 * + stateful rules; 203 * + enforcing limits on the number of sessions; 204 * + in-kernel NAT (not implemented yet) 205 * 206 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 207 * measured in seconds and depending on the flags. 208 * 209 * The total number of dynamic rules is stored in dyn_count. 210 * The max number of dynamic rules is dyn_max. When we reach 211 * the maximum number of rules we do not create anymore. This is 212 * done to avoid consuming too much memory, but also too much 213 * time when searching on each packet (ideally, we should try instead 214 * to put a limit on the length of the list on each bucket...). 215 * 216 * Each dynamic rule holds a pointer to the parent ipfw rule so 217 * we know what action to perform. Dynamic rules are removed when 218 * the parent rule is deleted. XXX we should make them survive. 219 * 220 * There are some limitations with dynamic rules -- we do not 221 * obey the 'randomized match', and we do not do multiple 222 * passes through the firewall. XXX check the latter!!! 223 */ 224static ipfw_dyn_rule **ipfw_dyn_v = NULL; 225static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 226static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 227 228static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 229#define IPFW_DYN_LOCK_INIT() \ 230 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 231#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 232#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 233#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 234#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 235 236/* 237 * Timeouts for various events in handing dynamic rules. 238 */ 239static u_int32_t dyn_ack_lifetime = 300; 240static u_int32_t dyn_syn_lifetime = 20; 241static u_int32_t dyn_fin_lifetime = 1; 242static u_int32_t dyn_rst_lifetime = 1; 243static u_int32_t dyn_udp_lifetime = 10; 244static u_int32_t dyn_short_lifetime = 5; 245 246/* 247 * Keepalives are sent if dyn_keepalive is set. They are sent every 248 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 249 * seconds of lifetime of a rule. 250 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 251 * than dyn_keepalive_period. 252 */ 253 254static u_int32_t dyn_keepalive_interval = 20; 255static u_int32_t dyn_keepalive_period = 5; 256static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 257 258static u_int32_t static_count; /* # of static rules */ 259static u_int32_t static_len; /* size in bytes of static rules */ 260static u_int32_t dyn_count; /* # of dynamic rules */ 261static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 262 263SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 264 &dyn_buckets, 0, "Number of dyn. buckets"); 265SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 266 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 267SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 268 &dyn_count, 0, "Number of dyn. rules"); 269SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 270 &dyn_max, 0, "Max number of dyn. rules"); 271SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 272 &static_count, 0, "Number of static rules"); 273SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 274 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 275SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 276 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 277SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 278 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 279SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 280 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 281SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 282 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 283SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 284 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 285SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 286 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 287 288#ifdef INET6 289/* 290 * IPv6 specific variables 291 */ 292SYSCTL_DECL(_net_inet6_ip6); 293 294static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 295static struct sysctl_oid *ip6_fw_sysctl_tree; 296#endif /* INET6 */ 297#endif /* SYSCTL_NODE */ 298 299static int fw_deny_unknown_exthdrs = 1; 300 301 302/* 303 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 304 * Other macros just cast void * into the appropriate type 305 */ 306#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 307#define TCP(p) ((struct tcphdr *)(p)) 308#define UDP(p) ((struct udphdr *)(p)) 309#define ICMP(p) ((struct icmphdr *)(p)) 310#define ICMP6(p) ((struct icmp6_hdr *)(p)) 311 312static __inline int 313icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 314{ 315 int type = icmp->icmp_type; 316 317 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 318} 319 320#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 321 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 322 323static int 324is_icmp_query(struct icmphdr *icmp) 325{ 326 int type = icmp->icmp_type; 327 328 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 329} 330#undef TT 331 332/* 333 * The following checks use two arrays of 8 or 16 bits to store the 334 * bits that we want set or clear, respectively. They are in the 335 * low and high half of cmd->arg1 or cmd->d[0]. 336 * 337 * We scan options and store the bits we find set. We succeed if 338 * 339 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 340 * 341 * The code is sometimes optimized not to store additional variables. 342 */ 343 344static int 345flags_match(ipfw_insn *cmd, u_int8_t bits) 346{ 347 u_char want_clear; 348 bits = ~bits; 349 350 if ( ((cmd->arg1 & 0xff) & bits) != 0) 351 return 0; /* some bits we want set were clear */ 352 want_clear = (cmd->arg1 >> 8) & 0xff; 353 if ( (want_clear & bits) != want_clear) 354 return 0; /* some bits we want clear were set */ 355 return 1; 356} 357 358static int 359ipopts_match(struct ip *ip, ipfw_insn *cmd) 360{ 361 int optlen, bits = 0; 362 u_char *cp = (u_char *)(ip + 1); 363 int x = (ip->ip_hl << 2) - sizeof (struct ip); 364 365 for (; x > 0; x -= optlen, cp += optlen) { 366 int opt = cp[IPOPT_OPTVAL]; 367 368 if (opt == IPOPT_EOL) 369 break; 370 if (opt == IPOPT_NOP) 371 optlen = 1; 372 else { 373 optlen = cp[IPOPT_OLEN]; 374 if (optlen <= 0 || optlen > x) 375 return 0; /* invalid or truncated */ 376 } 377 switch (opt) { 378 379 default: 380 break; 381 382 case IPOPT_LSRR: 383 bits |= IP_FW_IPOPT_LSRR; 384 break; 385 386 case IPOPT_SSRR: 387 bits |= IP_FW_IPOPT_SSRR; 388 break; 389 390 case IPOPT_RR: 391 bits |= IP_FW_IPOPT_RR; 392 break; 393 394 case IPOPT_TS: 395 bits |= IP_FW_IPOPT_TS; 396 break; 397 } 398 } 399 return (flags_match(cmd, bits)); 400} 401 402static int 403tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 404{ 405 int optlen, bits = 0; 406 u_char *cp = (u_char *)(tcp + 1); 407 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 408 409 for (; x > 0; x -= optlen, cp += optlen) { 410 int opt = cp[0]; 411 if (opt == TCPOPT_EOL) 412 break; 413 if (opt == TCPOPT_NOP) 414 optlen = 1; 415 else { 416 optlen = cp[1]; 417 if (optlen <= 0) 418 break; 419 } 420 421 switch (opt) { 422 423 default: 424 break; 425 426 case TCPOPT_MAXSEG: 427 bits |= IP_FW_TCPOPT_MSS; 428 break; 429 430 case TCPOPT_WINDOW: 431 bits |= IP_FW_TCPOPT_WINDOW; 432 break; 433 434 case TCPOPT_SACK_PERMITTED: 435 case TCPOPT_SACK: 436 bits |= IP_FW_TCPOPT_SACK; 437 break; 438 439 case TCPOPT_TIMESTAMP: 440 bits |= IP_FW_TCPOPT_TS; 441 break; 442 443 } 444 } 445 return (flags_match(cmd, bits)); 446} 447 448static int 449iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 450{ 451 if (ifp == NULL) /* no iface with this packet, match fails */ 452 return 0; 453 /* Check by name or by IP address */ 454 if (cmd->name[0] != '\0') { /* match by name */ 455 /* Check name */ 456 if (cmd->p.glob) { 457 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 458 return(1); 459 } else { 460 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 461 return(1); 462 } 463 } else { 464 struct ifaddr *ia; 465 466 /* XXX lock? */ 467 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 468 if (ia->ifa_addr == NULL) 469 continue; 470 if (ia->ifa_addr->sa_family != AF_INET) 471 continue; 472 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 473 (ia->ifa_addr))->sin_addr.s_addr) 474 return(1); /* match */ 475 } 476 } 477 return(0); /* no match, fail ... */ 478} 479 480/* 481 * The verify_path function checks if a route to the src exists and 482 * if it is reachable via ifp (when provided). 483 * 484 * The 'verrevpath' option checks that the interface that an IP packet 485 * arrives on is the same interface that traffic destined for the 486 * packet's source address would be routed out of. The 'versrcreach' 487 * option just checks that the source address is reachable via any route 488 * (except default) in the routing table. These two are a measure to block 489 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 490 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 491 * is purposely reminiscent of the Cisco IOS command, 492 * 493 * ip verify unicast reverse-path 494 * ip verify unicast source reachable-via any 495 * 496 * which implements the same functionality. But note that syntax is 497 * misleading. The check may be performed on all IP packets whether unicast, 498 * multicast, or broadcast. 499 */ 500static int 501verify_path(struct in_addr src, struct ifnet *ifp) 502{ 503 struct route ro; 504 struct sockaddr_in *dst; 505 506 bzero(&ro, sizeof(ro)); 507 508 dst = (struct sockaddr_in *)&(ro.ro_dst); 509 dst->sin_family = AF_INET; 510 dst->sin_len = sizeof(*dst); 511 dst->sin_addr = src; 512 rtalloc_ign(&ro, RTF_CLONING); 513 514 if (ro.ro_rt == NULL) 515 return 0; 516 517 /* 518 * If ifp is provided, check for equality with rtentry. 519 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 520 * in order to pass packets injected back by if_simloop(): 521 * if useloopback == 1 routing entry (via lo0) for our own address 522 * may exist, so we need to handle routing assymetry. 523 */ 524 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 525 RTFREE(ro.ro_rt); 526 return 0; 527 } 528 529 /* if no ifp provided, check if rtentry is not default route */ 530 if (ifp == NULL && 531 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 532 RTFREE(ro.ro_rt); 533 return 0; 534 } 535 536 /* or if this is a blackhole/reject route */ 537 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 538 RTFREE(ro.ro_rt); 539 return 0; 540 } 541 542 /* found valid route */ 543 RTFREE(ro.ro_rt); 544 return 1; 545} 546 547#ifdef INET6 548/* 549 * ipv6 specific rules here... 550 */ 551static __inline int 552icmp6type_match (int type, ipfw_insn_u32 *cmd) 553{ 554 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 555} 556 557static int 558flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 559{ 560 int i; 561 for (i=0; i <= cmd->o.arg1; ++i ) 562 if (curr_flow == cmd->d[i] ) 563 return 1; 564 return 0; 565} 566 567/* support for IP6_*_ME opcodes */ 568static int 569search_ip6_addr_net (struct in6_addr * ip6_addr) 570{ 571 struct ifnet *mdc; 572 struct ifaddr *mdc2; 573 struct in6_ifaddr *fdm; 574 struct in6_addr copia; 575 576 TAILQ_FOREACH(mdc, &ifnet, if_link) 577 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 578 if (!mdc2->ifa_addr) 579 continue; 580 if (mdc2->ifa_addr->sa_family == AF_INET6) { 581 fdm = (struct in6_ifaddr *)mdc2; 582 copia = fdm->ia_addr.sin6_addr; 583 /* need for leaving scope_id in the sock_addr */ 584 in6_clearscope(&copia); 585 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 586 return 1; 587 } 588 } 589 return 0; 590} 591 592static int 593verify_path6(struct in6_addr *src, struct ifnet *ifp) 594{ 595 struct route_in6 ro; 596 struct sockaddr_in6 *dst; 597 598 bzero(&ro, sizeof(ro)); 599 600 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 601 dst->sin6_family = AF_INET6; 602 dst->sin6_len = sizeof(*dst); 603 dst->sin6_addr = *src; 604 rtalloc_ign((struct route *)&ro, RTF_CLONING); 605 606 if (ro.ro_rt == NULL) 607 return 0; 608 609 /* 610 * if ifp is provided, check for equality with rtentry 611 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 612 * to support the case of sending packets to an address of our own. 613 * (where the former interface is the first argument of if_simloop() 614 * (=ifp), the latter is lo0) 615 */ 616 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 617 RTFREE(ro.ro_rt); 618 return 0; 619 } 620 621 /* if no ifp provided, check if rtentry is not default route */ 622 if (ifp == NULL && 623 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 624 RTFREE(ro.ro_rt); 625 return 0; 626 } 627 628 /* or if this is a blackhole/reject route */ 629 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 630 RTFREE(ro.ro_rt); 631 return 0; 632 } 633 634 /* found valid route */ 635 RTFREE(ro.ro_rt); 636 return 1; 637 638} 639static __inline int 640hash_packet6(struct ipfw_flow_id *id) 641{ 642 u_int32_t i; 643 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 644 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 645 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 646 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 647 (id->dst_port) ^ (id->src_port); 648 return i; 649} 650 651static int 652is_icmp6_query(int icmp6_type) 653{ 654 if ((icmp6_type <= ICMP6_MAXTYPE) && 655 (icmp6_type == ICMP6_ECHO_REQUEST || 656 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 657 icmp6_type == ICMP6_WRUREQUEST || 658 icmp6_type == ICMP6_FQDN_QUERY || 659 icmp6_type == ICMP6_NI_QUERY)) 660 return (1); 661 662 return (0); 663} 664 665static void 666send_reject6(struct ip_fw_args *args, int code, u_int hlen) 667{ 668 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 669 struct ip6_hdr *ip6; 670 struct tcphdr *tcp; 671 tcp_seq ack, seq; 672 int flags; 673 struct { 674 struct ip6_hdr ip6; 675 struct tcphdr th; 676 } ti; 677 678 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) { 679 args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr)); 680 if (args->m == NULL) 681 return; 682 } 683 684 ip6 = mtod(args->m, struct ip6_hdr *); 685 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 686 687 if ((tcp->th_flags & TH_RST) != 0) { 688 m_freem(args->m); 689 return; 690 } 691 692 ti.ip6 = *ip6; 693 ti.th = *tcp; 694 ti.th.th_seq = ntohl(ti.th.th_seq); 695 ti.th.th_ack = ntohl(ti.th.th_ack); 696 ti.ip6.ip6_nxt = IPPROTO_TCP; 697 698 if (ti.th.th_flags & TH_ACK) { 699 ack = 0; 700 seq = ti.th.th_ack; 701 flags = TH_RST; 702 } else { 703 ack = ti.th.th_seq; 704 if (((args->m)->m_flags & M_PKTHDR) != 0) { 705 ack += (args->m)->m_pkthdr.len - hlen 706 - (ti.th.th_off << 2); 707 } else if (ip6->ip6_plen) { 708 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) 709 - hlen - (ti.th.th_off << 2); 710 } else { 711 m_freem(args->m); 712 return; 713 } 714 if (tcp->th_flags & TH_SYN) 715 ack++; 716 seq = 0; 717 flags = TH_RST|TH_ACK; 718 } 719 bcopy(&ti, ip6, sizeof(ti)); 720 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1), 721 args->m, ack, seq, flags); 722 723 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 724 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0); 725 726 } else 727 m_freem(args->m); 728 729 args->m = NULL; 730} 731 732#endif /* INET6 */ 733 734static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 735 736#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 737#define SNP(buf) buf, sizeof(buf) 738 739/* 740 * We enter here when we have a rule with O_LOG. 741 * XXX this function alone takes about 2Kbytes of code! 742 */ 743static void 744ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 745 struct mbuf *m, struct ifnet *oif, u_short offset) 746{ 747 struct ether_header *eh = args->eh; 748 char *action; 749 int limit_reached = 0; 750 char action2[40], proto[128], fragment[32]; 751 752 fragment[0] = '\0'; 753 proto[0] = '\0'; 754 755 if (f == NULL) { /* bogus pkt */ 756 if (verbose_limit != 0 && norule_counter >= verbose_limit) 757 return; 758 norule_counter++; 759 if (norule_counter == verbose_limit) 760 limit_reached = verbose_limit; 761 action = "Refuse"; 762 } else { /* O_LOG is the first action, find the real one */ 763 ipfw_insn *cmd = ACTION_PTR(f); 764 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 765 766 if (l->max_log != 0 && l->log_left == 0) 767 return; 768 l->log_left--; 769 if (l->log_left == 0) 770 limit_reached = l->max_log; 771 cmd += F_LEN(cmd); /* point to first action */ 772 if (cmd->opcode == O_ALTQ) { 773 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 774 775 snprintf(SNPARGS(action2, 0), "Altq %d", 776 altq->qid); 777 cmd += F_LEN(cmd); 778 } 779 if (cmd->opcode == O_PROB) 780 cmd += F_LEN(cmd); 781 782 if (cmd->opcode == O_TAG) 783 cmd += F_LEN(cmd); 784 785 action = action2; 786 switch (cmd->opcode) { 787 case O_DENY: 788 action = "Deny"; 789 break; 790 791 case O_REJECT: 792 if (cmd->arg1==ICMP_REJECT_RST) 793 action = "Reset"; 794 else if (cmd->arg1==ICMP_UNREACH_HOST) 795 action = "Reject"; 796 else 797 snprintf(SNPARGS(action2, 0), "Unreach %d", 798 cmd->arg1); 799 break; 800 801 case O_UNREACH6: 802 if (cmd->arg1==ICMP6_UNREACH_RST) 803 action = "Reset"; 804 else 805 snprintf(SNPARGS(action2, 0), "Unreach %d", 806 cmd->arg1); 807 break; 808 809 case O_ACCEPT: 810 action = "Accept"; 811 break; 812 case O_COUNT: 813 action = "Count"; 814 break; 815 case O_DIVERT: 816 snprintf(SNPARGS(action2, 0), "Divert %d", 817 cmd->arg1); 818 break; 819 case O_TEE: 820 snprintf(SNPARGS(action2, 0), "Tee %d", 821 cmd->arg1); 822 break; 823 case O_SKIPTO: 824 snprintf(SNPARGS(action2, 0), "SkipTo %d", 825 cmd->arg1); 826 break; 827 case O_PIPE: 828 snprintf(SNPARGS(action2, 0), "Pipe %d", 829 cmd->arg1); 830 break; 831 case O_QUEUE: 832 snprintf(SNPARGS(action2, 0), "Queue %d", 833 cmd->arg1); 834 break; 835 case O_FORWARD_IP: { 836 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 837 int len; 838 839 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 840 inet_ntoa(sa->sa.sin_addr)); 841 if (sa->sa.sin_port) 842 snprintf(SNPARGS(action2, len), ":%d", 843 sa->sa.sin_port); 844 } 845 break; 846 case O_NETGRAPH: 847 snprintf(SNPARGS(action2, 0), "Netgraph %d", 848 cmd->arg1); 849 break; 850 case O_NGTEE: 851 snprintf(SNPARGS(action2, 0), "Ngtee %d", 852 cmd->arg1); 853 break; 854 default: 855 action = "UNKNOWN"; 856 break; 857 } 858 } 859 860 if (hlen == 0) { /* non-ip */ 861 snprintf(SNPARGS(proto, 0), "MAC"); 862 863 } else { 864 int len; 865 char src[48], dst[48]; 866 struct icmphdr *icmp; 867 struct tcphdr *tcp; 868 struct udphdr *udp; 869 /* Initialize to make compiler happy. */ 870 struct ip *ip = NULL; 871#ifdef INET6 872 struct ip6_hdr *ip6 = NULL; 873 struct icmp6_hdr *icmp6; 874#endif 875 src[0] = '\0'; 876 dst[0] = '\0'; 877#ifdef INET6 878 if (args->f_id.addr_type == 6) { 879 snprintf(src, sizeof(src), "[%s]", 880 ip6_sprintf(&args->f_id.src_ip6)); 881 snprintf(dst, sizeof(dst), "[%s]", 882 ip6_sprintf(&args->f_id.dst_ip6)); 883 884 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 885 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 886 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 887 } else 888#endif 889 { 890 ip = mtod(m, struct ip *); 891 tcp = L3HDR(struct tcphdr, ip); 892 udp = L3HDR(struct udphdr, ip); 893 894 inet_ntoa_r(ip->ip_src, src); 895 inet_ntoa_r(ip->ip_dst, dst); 896 } 897 898 switch (args->f_id.proto) { 899 case IPPROTO_TCP: 900 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 901 if (offset == 0) 902 snprintf(SNPARGS(proto, len), ":%d %s:%d", 903 ntohs(tcp->th_sport), 904 dst, 905 ntohs(tcp->th_dport)); 906 else 907 snprintf(SNPARGS(proto, len), " %s", dst); 908 break; 909 910 case IPPROTO_UDP: 911 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 912 if (offset == 0) 913 snprintf(SNPARGS(proto, len), ":%d %s:%d", 914 ntohs(udp->uh_sport), 915 dst, 916 ntohs(udp->uh_dport)); 917 else 918 snprintf(SNPARGS(proto, len), " %s", dst); 919 break; 920 921 case IPPROTO_ICMP: 922 icmp = L3HDR(struct icmphdr, ip); 923 if (offset == 0) 924 len = snprintf(SNPARGS(proto, 0), 925 "ICMP:%u.%u ", 926 icmp->icmp_type, icmp->icmp_code); 927 else 928 len = snprintf(SNPARGS(proto, 0), "ICMP "); 929 len += snprintf(SNPARGS(proto, len), "%s", src); 930 snprintf(SNPARGS(proto, len), " %s", dst); 931 break; 932#ifdef INET6 933 case IPPROTO_ICMPV6: 934 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 935 if (offset == 0) 936 len = snprintf(SNPARGS(proto, 0), 937 "ICMPv6:%u.%u ", 938 icmp6->icmp6_type, icmp6->icmp6_code); 939 else 940 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 941 len += snprintf(SNPARGS(proto, len), "%s", src); 942 snprintf(SNPARGS(proto, len), " %s", dst); 943 break; 944#endif 945 default: 946 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 947 args->f_id.proto, src); 948 snprintf(SNPARGS(proto, len), " %s", dst); 949 break; 950 } 951 952#ifdef INET6 953 if (args->f_id.addr_type == 6) { 954 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 955 snprintf(SNPARGS(fragment, 0), 956 " (frag %08x:%d@%d%s)", 957 args->f_id.frag_id6, 958 ntohs(ip6->ip6_plen) - hlen, 959 ntohs(offset & IP6F_OFF_MASK) << 3, 960 (offset & IP6F_MORE_FRAG) ? "+" : ""); 961 } else 962#endif 963 { 964 int ip_off, ip_len; 965 if (eh != NULL) { /* layer 2 packets are as on the wire */ 966 ip_off = ntohs(ip->ip_off); 967 ip_len = ntohs(ip->ip_len); 968 } else { 969 ip_off = ip->ip_off; 970 ip_len = ip->ip_len; 971 } 972 if (ip_off & (IP_MF | IP_OFFMASK)) 973 snprintf(SNPARGS(fragment, 0), 974 " (frag %d:%d@%d%s)", 975 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 976 offset << 3, 977 (ip_off & IP_MF) ? "+" : ""); 978 } 979 } 980 if (oif || m->m_pkthdr.rcvif) 981 log(LOG_SECURITY | LOG_INFO, 982 "ipfw: %d %s %s %s via %s%s\n", 983 f ? f->rulenum : -1, 984 action, proto, oif ? "out" : "in", 985 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 986 fragment); 987 else 988 log(LOG_SECURITY | LOG_INFO, 989 "ipfw: %d %s %s [no if info]%s\n", 990 f ? f->rulenum : -1, 991 action, proto, fragment); 992 if (limit_reached) 993 log(LOG_SECURITY | LOG_NOTICE, 994 "ipfw: limit %d reached on entry %d\n", 995 limit_reached, f ? f->rulenum : -1); 996} 997 998/* 999 * IMPORTANT: the hash function for dynamic rules must be commutative 1000 * in source and destination (ip,port), because rules are bidirectional 1001 * and we want to find both in the same bucket. 1002 */ 1003static __inline int 1004hash_packet(struct ipfw_flow_id *id) 1005{ 1006 u_int32_t i; 1007 1008#ifdef INET6 1009 if (IS_IP6_FLOW_ID(id)) 1010 i = hash_packet6(id); 1011 else 1012#endif /* INET6 */ 1013 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1014 i &= (curr_dyn_buckets - 1); 1015 return i; 1016} 1017 1018/** 1019 * unlink a dynamic rule from a chain. prev is a pointer to 1020 * the previous one, q is a pointer to the rule to delete, 1021 * head is a pointer to the head of the queue. 1022 * Modifies q and potentially also head. 1023 */ 1024#define UNLINK_DYN_RULE(prev, head, q) { \ 1025 ipfw_dyn_rule *old_q = q; \ 1026 \ 1027 /* remove a refcount to the parent */ \ 1028 if (q->dyn_type == O_LIMIT) \ 1029 q->parent->count--; \ 1030 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1031 (q->id.src_ip), (q->id.src_port), \ 1032 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1033 if (prev != NULL) \ 1034 prev->next = q = q->next; \ 1035 else \ 1036 head = q = q->next; \ 1037 dyn_count--; \ 1038 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1039 1040#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1041 1042/** 1043 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1044 * 1045 * If keep_me == NULL, rules are deleted even if not expired, 1046 * otherwise only expired rules are removed. 1047 * 1048 * The value of the second parameter is also used to point to identify 1049 * a rule we absolutely do not want to remove (e.g. because we are 1050 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1051 * rules). The pointer is only used for comparison, so any non-null 1052 * value will do. 1053 */ 1054static void 1055remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1056{ 1057 static u_int32_t last_remove = 0; 1058 1059#define FORCE (keep_me == NULL) 1060 1061 ipfw_dyn_rule *prev, *q; 1062 int i, pass = 0, max_pass = 0; 1063 1064 IPFW_DYN_LOCK_ASSERT(); 1065 1066 if (ipfw_dyn_v == NULL || dyn_count == 0) 1067 return; 1068 /* do not expire more than once per second, it is useless */ 1069 if (!FORCE && last_remove == time_uptime) 1070 return; 1071 last_remove = time_uptime; 1072 1073 /* 1074 * because O_LIMIT refer to parent rules, during the first pass only 1075 * remove child and mark any pending LIMIT_PARENT, and remove 1076 * them in a second pass. 1077 */ 1078next_pass: 1079 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1080 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1081 /* 1082 * Logic can become complex here, so we split tests. 1083 */ 1084 if (q == keep_me) 1085 goto next; 1086 if (rule != NULL && rule != q->rule) 1087 goto next; /* not the one we are looking for */ 1088 if (q->dyn_type == O_LIMIT_PARENT) { 1089 /* 1090 * handle parent in the second pass, 1091 * record we need one. 1092 */ 1093 max_pass = 1; 1094 if (pass == 0) 1095 goto next; 1096 if (FORCE && q->count != 0 ) { 1097 /* XXX should not happen! */ 1098 printf("ipfw: OUCH! cannot remove rule," 1099 " count %d\n", q->count); 1100 } 1101 } else { 1102 if (!FORCE && 1103 !TIME_LEQ( q->expire, time_uptime )) 1104 goto next; 1105 } 1106 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1107 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1108 continue; 1109 } 1110next: 1111 prev=q; 1112 q=q->next; 1113 } 1114 } 1115 if (pass++ < max_pass) 1116 goto next_pass; 1117} 1118 1119 1120/** 1121 * lookup a dynamic rule. 1122 */ 1123static ipfw_dyn_rule * 1124lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1125 struct tcphdr *tcp) 1126{ 1127 /* 1128 * stateful ipfw extensions. 1129 * Lookup into dynamic session queue 1130 */ 1131#define MATCH_REVERSE 0 1132#define MATCH_FORWARD 1 1133#define MATCH_NONE 2 1134#define MATCH_UNKNOWN 3 1135 int i, dir = MATCH_NONE; 1136 ipfw_dyn_rule *prev, *q=NULL; 1137 1138 IPFW_DYN_LOCK_ASSERT(); 1139 1140 if (ipfw_dyn_v == NULL) 1141 goto done; /* not found */ 1142 i = hash_packet( pkt ); 1143 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1144 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1145 goto next; 1146 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1147 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1148 continue; 1149 } 1150 if (pkt->proto == q->id.proto && 1151 q->dyn_type != O_LIMIT_PARENT) { 1152 if (IS_IP6_FLOW_ID(pkt)) { 1153 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1154 &(q->id.src_ip6)) && 1155 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1156 &(q->id.dst_ip6)) && 1157 pkt->src_port == q->id.src_port && 1158 pkt->dst_port == q->id.dst_port ) { 1159 dir = MATCH_FORWARD; 1160 break; 1161 } 1162 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1163 &(q->id.dst_ip6)) && 1164 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1165 &(q->id.src_ip6)) && 1166 pkt->src_port == q->id.dst_port && 1167 pkt->dst_port == q->id.src_port ) { 1168 dir = MATCH_REVERSE; 1169 break; 1170 } 1171 } else { 1172 if (pkt->src_ip == q->id.src_ip && 1173 pkt->dst_ip == q->id.dst_ip && 1174 pkt->src_port == q->id.src_port && 1175 pkt->dst_port == q->id.dst_port ) { 1176 dir = MATCH_FORWARD; 1177 break; 1178 } 1179 if (pkt->src_ip == q->id.dst_ip && 1180 pkt->dst_ip == q->id.src_ip && 1181 pkt->src_port == q->id.dst_port && 1182 pkt->dst_port == q->id.src_port ) { 1183 dir = MATCH_REVERSE; 1184 break; 1185 } 1186 } 1187 } 1188next: 1189 prev = q; 1190 q = q->next; 1191 } 1192 if (q == NULL) 1193 goto done; /* q = NULL, not found */ 1194 1195 if ( prev != NULL) { /* found and not in front */ 1196 prev->next = q->next; 1197 q->next = ipfw_dyn_v[i]; 1198 ipfw_dyn_v[i] = q; 1199 } 1200 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1201 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1202 1203#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1204#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1205 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1206 switch (q->state) { 1207 case TH_SYN: /* opening */ 1208 q->expire = time_uptime + dyn_syn_lifetime; 1209 break; 1210 1211 case BOTH_SYN: /* move to established */ 1212 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1213 case BOTH_SYN | (TH_FIN << 8) : 1214 if (tcp) { 1215#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1216 u_int32_t ack = ntohl(tcp->th_ack); 1217 if (dir == MATCH_FORWARD) { 1218 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1219 q->ack_fwd = ack; 1220 else { /* ignore out-of-sequence */ 1221 break; 1222 } 1223 } else { 1224 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1225 q->ack_rev = ack; 1226 else { /* ignore out-of-sequence */ 1227 break; 1228 } 1229 } 1230 } 1231 q->expire = time_uptime + dyn_ack_lifetime; 1232 break; 1233 1234 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1235 if (dyn_fin_lifetime >= dyn_keepalive_period) 1236 dyn_fin_lifetime = dyn_keepalive_period - 1; 1237 q->expire = time_uptime + dyn_fin_lifetime; 1238 break; 1239 1240 default: 1241#if 0 1242 /* 1243 * reset or some invalid combination, but can also 1244 * occur if we use keep-state the wrong way. 1245 */ 1246 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1247 printf("invalid state: 0x%x\n", q->state); 1248#endif 1249 if (dyn_rst_lifetime >= dyn_keepalive_period) 1250 dyn_rst_lifetime = dyn_keepalive_period - 1; 1251 q->expire = time_uptime + dyn_rst_lifetime; 1252 break; 1253 } 1254 } else if (pkt->proto == IPPROTO_UDP) { 1255 q->expire = time_uptime + dyn_udp_lifetime; 1256 } else { 1257 /* other protocols */ 1258 q->expire = time_uptime + dyn_short_lifetime; 1259 } 1260done: 1261 if (match_direction) 1262 *match_direction = dir; 1263 return q; 1264} 1265 1266static ipfw_dyn_rule * 1267lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1268 struct tcphdr *tcp) 1269{ 1270 ipfw_dyn_rule *q; 1271 1272 IPFW_DYN_LOCK(); 1273 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1274 if (q == NULL) 1275 IPFW_DYN_UNLOCK(); 1276 /* NB: return table locked when q is not NULL */ 1277 return q; 1278} 1279 1280static void 1281realloc_dynamic_table(void) 1282{ 1283 IPFW_DYN_LOCK_ASSERT(); 1284 1285 /* 1286 * Try reallocation, make sure we have a power of 2 and do 1287 * not allow more than 64k entries. In case of overflow, 1288 * default to 1024. 1289 */ 1290 1291 if (dyn_buckets > 65536) 1292 dyn_buckets = 1024; 1293 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1294 dyn_buckets = curr_dyn_buckets; /* reset */ 1295 return; 1296 } 1297 curr_dyn_buckets = dyn_buckets; 1298 if (ipfw_dyn_v != NULL) 1299 free(ipfw_dyn_v, M_IPFW); 1300 for (;;) { 1301 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1302 M_IPFW, M_NOWAIT | M_ZERO); 1303 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1304 break; 1305 curr_dyn_buckets /= 2; 1306 } 1307} 1308 1309/** 1310 * Install state of type 'type' for a dynamic session. 1311 * The hash table contains two type of rules: 1312 * - regular rules (O_KEEP_STATE) 1313 * - rules for sessions with limited number of sess per user 1314 * (O_LIMIT). When they are created, the parent is 1315 * increased by 1, and decreased on delete. In this case, 1316 * the third parameter is the parent rule and not the chain. 1317 * - "parent" rules for the above (O_LIMIT_PARENT). 1318 */ 1319static ipfw_dyn_rule * 1320add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1321{ 1322 ipfw_dyn_rule *r; 1323 int i; 1324 1325 IPFW_DYN_LOCK_ASSERT(); 1326 1327 if (ipfw_dyn_v == NULL || 1328 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1329 realloc_dynamic_table(); 1330 if (ipfw_dyn_v == NULL) 1331 return NULL; /* failed ! */ 1332 } 1333 i = hash_packet(id); 1334 1335 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1336 if (r == NULL) { 1337 printf ("ipfw: sorry cannot allocate state\n"); 1338 return NULL; 1339 } 1340 1341 /* increase refcount on parent, and set pointer */ 1342 if (dyn_type == O_LIMIT) { 1343 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1344 if ( parent->dyn_type != O_LIMIT_PARENT) 1345 panic("invalid parent"); 1346 parent->count++; 1347 r->parent = parent; 1348 rule = parent->rule; 1349 } 1350 1351 r->id = *id; 1352 r->expire = time_uptime + dyn_syn_lifetime; 1353 r->rule = rule; 1354 r->dyn_type = dyn_type; 1355 r->pcnt = r->bcnt = 0; 1356 r->count = 0; 1357 1358 r->bucket = i; 1359 r->next = ipfw_dyn_v[i]; 1360 ipfw_dyn_v[i] = r; 1361 dyn_count++; 1362 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1363 dyn_type, 1364 (r->id.src_ip), (r->id.src_port), 1365 (r->id.dst_ip), (r->id.dst_port), 1366 dyn_count ); ) 1367 return r; 1368} 1369 1370/** 1371 * lookup dynamic parent rule using pkt and rule as search keys. 1372 * If the lookup fails, then install one. 1373 */ 1374static ipfw_dyn_rule * 1375lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1376{ 1377 ipfw_dyn_rule *q; 1378 int i; 1379 1380 IPFW_DYN_LOCK_ASSERT(); 1381 1382 if (ipfw_dyn_v) { 1383 int is_v6 = IS_IP6_FLOW_ID(pkt); 1384 i = hash_packet( pkt ); 1385 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1386 if (q->dyn_type == O_LIMIT_PARENT && 1387 rule== q->rule && 1388 pkt->proto == q->id.proto && 1389 pkt->src_port == q->id.src_port && 1390 pkt->dst_port == q->id.dst_port && 1391 ( 1392 (is_v6 && 1393 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1394 &(q->id.src_ip6)) && 1395 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1396 &(q->id.dst_ip6))) || 1397 (!is_v6 && 1398 pkt->src_ip == q->id.src_ip && 1399 pkt->dst_ip == q->id.dst_ip) 1400 ) 1401 ) { 1402 q->expire = time_uptime + dyn_short_lifetime; 1403 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1404 return q; 1405 } 1406 } 1407 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1408} 1409 1410/** 1411 * Install dynamic state for rule type cmd->o.opcode 1412 * 1413 * Returns 1 (failure) if state is not installed because of errors or because 1414 * session limitations are enforced. 1415 */ 1416static int 1417install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1418 struct ip_fw_args *args, uint32_t tablearg) 1419{ 1420 static int last_log; 1421 1422 ipfw_dyn_rule *q; 1423 1424 DEB( 1425 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1426 __func__, cmd->o.opcode, 1427 (args->f_id.src_ip), (args->f_id.src_port), 1428 (args->f_id.dst_ip), (args->f_id.dst_port)); 1429 ) 1430 1431 IPFW_DYN_LOCK(); 1432 1433 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1434 1435 if (q != NULL) { /* should never occur */ 1436 if (last_log != time_uptime) { 1437 last_log = time_uptime; 1438 printf("ipfw: %s: entry already present, done\n", 1439 __func__); 1440 } 1441 IPFW_DYN_UNLOCK(); 1442 return (0); 1443 } 1444 1445 if (dyn_count >= dyn_max) 1446 /* Run out of slots, try to remove any expired rule. */ 1447 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1448 1449 if (dyn_count >= dyn_max) { 1450 if (last_log != time_uptime) { 1451 last_log = time_uptime; 1452 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1453 } 1454 IPFW_DYN_UNLOCK(); 1455 return (1); /* cannot install, notify caller */ 1456 } 1457 1458 switch (cmd->o.opcode) { 1459 case O_KEEP_STATE: /* bidir rule */ 1460 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1461 break; 1462 1463 case O_LIMIT: { /* limit number of sessions */ 1464 struct ipfw_flow_id id; 1465 ipfw_dyn_rule *parent; 1466 uint32_t conn_limit; 1467 uint16_t limit_mask = cmd->limit_mask; 1468 1469 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1470 tablearg : cmd->conn_limit; 1471 1472 DEB( 1473 if (cmd->conn_limit == IP_FW_TABLEARG) 1474 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1475 "(tablearg)\n", __func__, conn_limit); 1476 else 1477 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1478 __func__, conn_limit); 1479 ) 1480 1481 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1482 id.proto = args->f_id.proto; 1483 id.addr_type = args->f_id.addr_type; 1484 1485 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1486 if (limit_mask & DYN_SRC_ADDR) 1487 id.src_ip6 = args->f_id.src_ip6; 1488 if (limit_mask & DYN_DST_ADDR) 1489 id.dst_ip6 = args->f_id.dst_ip6; 1490 } else { 1491 if (limit_mask & DYN_SRC_ADDR) 1492 id.src_ip = args->f_id.src_ip; 1493 if (limit_mask & DYN_DST_ADDR) 1494 id.dst_ip = args->f_id.dst_ip; 1495 } 1496 if (limit_mask & DYN_SRC_PORT) 1497 id.src_port = args->f_id.src_port; 1498 if (limit_mask & DYN_DST_PORT) 1499 id.dst_port = args->f_id.dst_port; 1500 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1501 printf("ipfw: %s: add parent failed\n", __func__); 1502 IPFW_DYN_UNLOCK(); 1503 return (1); 1504 } 1505 1506 if (parent->count >= conn_limit) { 1507 /* See if we can remove some expired rule. */ 1508 remove_dyn_rule(rule, parent); 1509 if (parent->count >= conn_limit) { 1510 if (fw_verbose && last_log != time_uptime) { 1511 last_log = time_uptime; 1512 log(LOG_SECURITY | LOG_DEBUG, 1513 "drop session, too many entries\n"); 1514 } 1515 IPFW_DYN_UNLOCK(); 1516 return (1); 1517 } 1518 } 1519 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1520 break; 1521 } 1522 default: 1523 printf("ipfw: %s: unknown dynamic rule type %u\n", 1524 __func__, cmd->o.opcode); 1525 IPFW_DYN_UNLOCK(); 1526 return (1); 1527 } 1528 1529 /* XXX just set lifetime */ 1530 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1531 1532 IPFW_DYN_UNLOCK(); 1533 return (0); 1534} 1535 1536/* 1537 * Generate a TCP packet, containing either a RST or a keepalive. 1538 * When flags & TH_RST, we are sending a RST packet, because of a 1539 * "reset" action matched the packet. 1540 * Otherwise we are sending a keepalive, and flags & TH_ 1541 */ 1542static struct mbuf * 1543send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1544{ 1545 struct mbuf *m; 1546 struct ip *ip; 1547 struct tcphdr *tcp; 1548 1549 MGETHDR(m, M_DONTWAIT, MT_DATA); 1550 if (m == 0) 1551 return (NULL); 1552 m->m_pkthdr.rcvif = (struct ifnet *)0; 1553 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1554 m->m_data += max_linkhdr; 1555 1556 ip = mtod(m, struct ip *); 1557 bzero(ip, m->m_len); 1558 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1559 ip->ip_p = IPPROTO_TCP; 1560 tcp->th_off = 5; 1561 /* 1562 * Assume we are sending a RST (or a keepalive in the reverse 1563 * direction), swap src and destination addresses and ports. 1564 */ 1565 ip->ip_src.s_addr = htonl(id->dst_ip); 1566 ip->ip_dst.s_addr = htonl(id->src_ip); 1567 tcp->th_sport = htons(id->dst_port); 1568 tcp->th_dport = htons(id->src_port); 1569 if (flags & TH_RST) { /* we are sending a RST */ 1570 if (flags & TH_ACK) { 1571 tcp->th_seq = htonl(ack); 1572 tcp->th_ack = htonl(0); 1573 tcp->th_flags = TH_RST; 1574 } else { 1575 if (flags & TH_SYN) 1576 seq++; 1577 tcp->th_seq = htonl(0); 1578 tcp->th_ack = htonl(seq); 1579 tcp->th_flags = TH_RST | TH_ACK; 1580 } 1581 } else { 1582 /* 1583 * We are sending a keepalive. flags & TH_SYN determines 1584 * the direction, forward if set, reverse if clear. 1585 * NOTE: seq and ack are always assumed to be correct 1586 * as set by the caller. This may be confusing... 1587 */ 1588 if (flags & TH_SYN) { 1589 /* 1590 * we have to rewrite the correct addresses! 1591 */ 1592 ip->ip_dst.s_addr = htonl(id->dst_ip); 1593 ip->ip_src.s_addr = htonl(id->src_ip); 1594 tcp->th_dport = htons(id->dst_port); 1595 tcp->th_sport = htons(id->src_port); 1596 } 1597 tcp->th_seq = htonl(seq); 1598 tcp->th_ack = htonl(ack); 1599 tcp->th_flags = TH_ACK; 1600 } 1601 /* 1602 * set ip_len to the payload size so we can compute 1603 * the tcp checksum on the pseudoheader 1604 * XXX check this, could save a couple of words ? 1605 */ 1606 ip->ip_len = htons(sizeof(struct tcphdr)); 1607 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1608 /* 1609 * now fill fields left out earlier 1610 */ 1611 ip->ip_ttl = ip_defttl; 1612 ip->ip_len = m->m_pkthdr.len; 1613 m->m_flags |= M_SKIP_FIREWALL; 1614 return (m); 1615} 1616 1617/* 1618 * sends a reject message, consuming the mbuf passed as an argument. 1619 */ 1620static void 1621send_reject(struct ip_fw_args *args, int code, int ip_len) 1622{ 1623 1624 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1625 /* We need the IP header in host order for icmp_error(). */ 1626 if (args->eh != NULL) { 1627 struct ip *ip = mtod(args->m, struct ip *); 1628 ip->ip_len = ntohs(ip->ip_len); 1629 ip->ip_off = ntohs(ip->ip_off); 1630 } 1631 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1632 } else if (args->f_id.proto == IPPROTO_TCP) { 1633 struct tcphdr *const tcp = 1634 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1635 if ( (tcp->th_flags & TH_RST) == 0) { 1636 struct mbuf *m; 1637 m = send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1638 ntohl(tcp->th_ack), 1639 tcp->th_flags | TH_RST); 1640 if (m != NULL) 1641 ip_output(m, NULL, NULL, 0, NULL, NULL); 1642 } 1643 m_freem(args->m); 1644 } else 1645 m_freem(args->m); 1646 args->m = NULL; 1647} 1648 1649/** 1650 * 1651 * Given an ip_fw *, lookup_next_rule will return a pointer 1652 * to the next rule, which can be either the jump 1653 * target (for skipto instructions) or the next one in the list (in 1654 * all other cases including a missing jump target). 1655 * The result is also written in the "next_rule" field of the rule. 1656 * Backward jumps are not allowed, so start looking from the next 1657 * rule... 1658 * 1659 * This never returns NULL -- in case we do not have an exact match, 1660 * the next rule is returned. When the ruleset is changed, 1661 * pointers are flushed so we are always correct. 1662 */ 1663 1664static struct ip_fw * 1665lookup_next_rule(struct ip_fw *me) 1666{ 1667 struct ip_fw *rule = NULL; 1668 ipfw_insn *cmd; 1669 1670 /* look for action, in case it is a skipto */ 1671 cmd = ACTION_PTR(me); 1672 if (cmd->opcode == O_LOG) 1673 cmd += F_LEN(cmd); 1674 if (cmd->opcode == O_ALTQ) 1675 cmd += F_LEN(cmd); 1676 if (cmd->opcode == O_TAG) 1677 cmd += F_LEN(cmd); 1678 if ( cmd->opcode == O_SKIPTO ) 1679 for (rule = me->next; rule ; rule = rule->next) 1680 if (rule->rulenum >= cmd->arg1) 1681 break; 1682 if (rule == NULL) /* failure or not a skipto */ 1683 rule = me->next; 1684 me->next_rule = rule; 1685 return rule; 1686} 1687 1688static int 1689add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1690 uint8_t mlen, uint32_t value) 1691{ 1692 struct radix_node_head *rnh; 1693 struct table_entry *ent; 1694 1695 if (tbl >= IPFW_TABLES_MAX) 1696 return (EINVAL); 1697 rnh = ch->tables[tbl]; 1698 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1699 if (ent == NULL) 1700 return (ENOMEM); 1701 ent->value = value; 1702 ent->addr.sin_len = ent->mask.sin_len = 8; 1703 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1704 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1705 IPFW_WLOCK(&layer3_chain); 1706 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1707 NULL) { 1708 IPFW_WUNLOCK(&layer3_chain); 1709 free(ent, M_IPFW_TBL); 1710 return (EEXIST); 1711 } 1712 IPFW_WUNLOCK(&layer3_chain); 1713 return (0); 1714} 1715 1716static int 1717del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1718 uint8_t mlen) 1719{ 1720 struct radix_node_head *rnh; 1721 struct table_entry *ent; 1722 struct sockaddr_in sa, mask; 1723 1724 if (tbl >= IPFW_TABLES_MAX) 1725 return (EINVAL); 1726 rnh = ch->tables[tbl]; 1727 sa.sin_len = mask.sin_len = 8; 1728 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1729 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1730 IPFW_WLOCK(ch); 1731 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1732 if (ent == NULL) { 1733 IPFW_WUNLOCK(ch); 1734 return (ESRCH); 1735 } 1736 IPFW_WUNLOCK(ch); 1737 free(ent, M_IPFW_TBL); 1738 return (0); 1739} 1740 1741static int 1742flush_table_entry(struct radix_node *rn, void *arg) 1743{ 1744 struct radix_node_head * const rnh = arg; 1745 struct table_entry *ent; 1746 1747 ent = (struct table_entry *) 1748 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1749 if (ent != NULL) 1750 free(ent, M_IPFW_TBL); 1751 return (0); 1752} 1753 1754static int 1755flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1756{ 1757 struct radix_node_head *rnh; 1758 1759 IPFW_WLOCK_ASSERT(ch); 1760 1761 if (tbl >= IPFW_TABLES_MAX) 1762 return (EINVAL); 1763 rnh = ch->tables[tbl]; 1764 KASSERT(rnh != NULL, ("NULL IPFW table")); 1765 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1766 return (0); 1767} 1768 1769static void 1770flush_tables(struct ip_fw_chain *ch) 1771{ 1772 uint16_t tbl; 1773 1774 IPFW_WLOCK_ASSERT(ch); 1775 1776 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1777 flush_table(ch, tbl); 1778} 1779 1780static int 1781init_tables(struct ip_fw_chain *ch) 1782{ 1783 int i; 1784 uint16_t j; 1785 1786 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1787 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1788 for (j = 0; j < i; j++) { 1789 (void) flush_table(ch, j); 1790 } 1791 return (ENOMEM); 1792 } 1793 } 1794 return (0); 1795} 1796 1797static int 1798lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1799 uint32_t *val) 1800{ 1801 struct radix_node_head *rnh; 1802 struct table_entry *ent; 1803 struct sockaddr_in sa; 1804 1805 if (tbl >= IPFW_TABLES_MAX) 1806 return (0); 1807 rnh = ch->tables[tbl]; 1808 sa.sin_len = 8; 1809 sa.sin_addr.s_addr = addr; 1810 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1811 if (ent != NULL) { 1812 *val = ent->value; 1813 return (1); 1814 } 1815 return (0); 1816} 1817 1818static int 1819count_table_entry(struct radix_node *rn, void *arg) 1820{ 1821 u_int32_t * const cnt = arg; 1822 1823 (*cnt)++; 1824 return (0); 1825} 1826 1827static int 1828count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1829{ 1830 struct radix_node_head *rnh; 1831 1832 if (tbl >= IPFW_TABLES_MAX) 1833 return (EINVAL); 1834 rnh = ch->tables[tbl]; 1835 *cnt = 0; 1836 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1837 return (0); 1838} 1839 1840static int 1841dump_table_entry(struct radix_node *rn, void *arg) 1842{ 1843 struct table_entry * const n = (struct table_entry *)rn; 1844 ipfw_table * const tbl = arg; 1845 ipfw_table_entry *ent; 1846 1847 if (tbl->cnt == tbl->size) 1848 return (1); 1849 ent = &tbl->ent[tbl->cnt]; 1850 ent->tbl = tbl->tbl; 1851 if (in_nullhost(n->mask.sin_addr)) 1852 ent->masklen = 0; 1853 else 1854 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1855 ent->addr = n->addr.sin_addr.s_addr; 1856 ent->value = n->value; 1857 tbl->cnt++; 1858 return (0); 1859} 1860 1861static int 1862dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1863{ 1864 struct radix_node_head *rnh; 1865 1866 if (tbl->tbl >= IPFW_TABLES_MAX) 1867 return (EINVAL); 1868 rnh = ch->tables[tbl->tbl]; 1869 tbl->cnt = 0; 1870 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1871 return (0); 1872} 1873 1874static void 1875fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1876{ 1877 struct ucred *cr; 1878 1879 if (inp->inp_socket != NULL) { 1880 cr = inp->inp_socket->so_cred; 1881 ugp->fw_prid = jailed(cr) ? 1882 cr->cr_prison->pr_id : -1; 1883 ugp->fw_uid = cr->cr_uid; 1884 ugp->fw_ngroups = cr->cr_ngroups; 1885 bcopy(cr->cr_groups, ugp->fw_groups, 1886 sizeof(ugp->fw_groups)); 1887 } 1888} 1889 1890static int 1891check_uidgid(ipfw_insn_u32 *insn, 1892 int proto, struct ifnet *oif, 1893 struct in_addr dst_ip, u_int16_t dst_port, 1894 struct in_addr src_ip, u_int16_t src_port, 1895 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1896{ 1897 struct inpcbinfo *pi; 1898 int wildcard; 1899 struct inpcb *pcb; 1900 int match; 1901 gid_t *gp; 1902 1903 /* 1904 * Check to see if the UDP or TCP stack supplied us with 1905 * the PCB. If so, rather then holding a lock and looking 1906 * up the PCB, we can use the one that was supplied. 1907 */ 1908 if (inp && *lookup == 0) { 1909 INP_LOCK_ASSERT(inp); 1910 if (inp->inp_socket != NULL) { 1911 fill_ugid_cache(inp, ugp); 1912 *lookup = 1; 1913 } 1914 } 1915 /* 1916 * If we have already been here and the packet has no 1917 * PCB entry associated with it, then we can safely 1918 * assume that this is a no match. 1919 */ 1920 if (*lookup == -1) 1921 return (0); 1922 if (proto == IPPROTO_TCP) { 1923 wildcard = 0; 1924 pi = &tcbinfo; 1925 } else if (proto == IPPROTO_UDP) { 1926 wildcard = INPLOOKUP_WILDCARD; 1927 pi = &udbinfo; 1928 } else 1929 return 0; 1930 match = 0; 1931 if (*lookup == 0) { 1932 INP_INFO_RLOCK(pi); 1933 pcb = (oif) ? 1934 in_pcblookup_hash(pi, 1935 dst_ip, htons(dst_port), 1936 src_ip, htons(src_port), 1937 wildcard, oif) : 1938 in_pcblookup_hash(pi, 1939 src_ip, htons(src_port), 1940 dst_ip, htons(dst_port), 1941 wildcard, NULL); 1942 if (pcb != NULL) { 1943 INP_LOCK(pcb); 1944 if (pcb->inp_socket != NULL) { 1945 fill_ugid_cache(pcb, ugp); 1946 *lookup = 1; 1947 } 1948 INP_UNLOCK(pcb); 1949 } 1950 INP_INFO_RUNLOCK(pi); 1951 if (*lookup == 0) { 1952 /* 1953 * If the lookup did not yield any results, there 1954 * is no sense in coming back and trying again. So 1955 * we can set lookup to -1 and ensure that we wont 1956 * bother the pcb system again. 1957 */ 1958 *lookup = -1; 1959 return (0); 1960 } 1961 } 1962 if (insn->o.opcode == O_UID) 1963 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1964 else if (insn->o.opcode == O_GID) { 1965 for (gp = ugp->fw_groups; 1966 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1967 if (*gp == (gid_t)insn->d[0]) { 1968 match = 1; 1969 break; 1970 } 1971 } else if (insn->o.opcode == O_JAIL) 1972 match = (ugp->fw_prid == (int)insn->d[0]); 1973 return match; 1974} 1975 1976/* 1977 * The main check routine for the firewall. 1978 * 1979 * All arguments are in args so we can modify them and return them 1980 * back to the caller. 1981 * 1982 * Parameters: 1983 * 1984 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1985 * Starts with the IP header. 1986 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1987 * args->oif Outgoing interface, or NULL if packet is incoming. 1988 * The incoming interface is in the mbuf. (in) 1989 * args->divert_rule (in/out) 1990 * Skip up to the first rule past this rule number; 1991 * upon return, non-zero port number for divert or tee. 1992 * 1993 * args->rule Pointer to the last matching rule (in/out) 1994 * args->next_hop Socket we are forwarding to (out). 1995 * args->f_id Addresses grabbed from the packet (out) 1996 * args->cookie a cookie depending on rule action 1997 * 1998 * Return value: 1999 * 2000 * IP_FW_PASS the packet must be accepted 2001 * IP_FW_DENY the packet must be dropped 2002 * IP_FW_DIVERT divert packet, port in m_tag 2003 * IP_FW_TEE tee packet, port in m_tag 2004 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2005 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2006 * 2007 */ 2008 2009int 2010ipfw_chk(struct ip_fw_args *args) 2011{ 2012 /* 2013 * Local variables hold state during the processing of a packet. 2014 * 2015 * IMPORTANT NOTE: to speed up the processing of rules, there 2016 * are some assumption on the values of the variables, which 2017 * are documented here. Should you change them, please check 2018 * the implementation of the various instructions to make sure 2019 * that they still work. 2020 * 2021 * args->eh The MAC header. It is non-null for a layer2 2022 * packet, it is NULL for a layer-3 packet. 2023 * 2024 * m | args->m Pointer to the mbuf, as received from the caller. 2025 * It may change if ipfw_chk() does an m_pullup, or if it 2026 * consumes the packet because it calls send_reject(). 2027 * XXX This has to change, so that ipfw_chk() never modifies 2028 * or consumes the buffer. 2029 * ip is simply an alias of the value of m, and it is kept 2030 * in sync with it (the packet is supposed to start with 2031 * the ip header). 2032 */ 2033 struct mbuf *m = args->m; 2034 struct ip *ip = mtod(m, struct ip *); 2035 2036 /* 2037 * For rules which contain uid/gid or jail constraints, cache 2038 * a copy of the users credentials after the pcb lookup has been 2039 * executed. This will speed up the processing of rules with 2040 * these types of constraints, as well as decrease contention 2041 * on pcb related locks. 2042 */ 2043 struct ip_fw_ugid fw_ugid_cache; 2044 int ugid_lookup = 0; 2045 2046 /* 2047 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2048 * associated with a packet input on a divert socket. This 2049 * will allow to distinguish traffic and its direction when 2050 * it originates from a divert socket. 2051 */ 2052 u_int divinput_flags = 0; 2053 2054 /* 2055 * oif | args->oif If NULL, ipfw_chk has been called on the 2056 * inbound path (ether_input, ip_input). 2057 * If non-NULL, ipfw_chk has been called on the outbound path 2058 * (ether_output, ip_output). 2059 */ 2060 struct ifnet *oif = args->oif; 2061 2062 struct ip_fw *f = NULL; /* matching rule */ 2063 int retval = 0; 2064 2065 /* 2066 * hlen The length of the IP header. 2067 */ 2068 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2069 2070 /* 2071 * offset The offset of a fragment. offset != 0 means that 2072 * we have a fragment at this offset of an IPv4 packet. 2073 * offset == 0 means that (if this is an IPv4 packet) 2074 * this is the first or only fragment. 2075 * For IPv6 offset == 0 means there is no Fragment Header. 2076 * If offset != 0 for IPv6 always use correct mask to 2077 * get the correct offset because we add IP6F_MORE_FRAG 2078 * to be able to dectect the first fragment which would 2079 * otherwise have offset = 0. 2080 */ 2081 u_short offset = 0; 2082 2083 /* 2084 * Local copies of addresses. They are only valid if we have 2085 * an IP packet. 2086 * 2087 * proto The protocol. Set to 0 for non-ip packets, 2088 * or to the protocol read from the packet otherwise. 2089 * proto != 0 means that we have an IPv4 packet. 2090 * 2091 * src_port, dst_port port numbers, in HOST format. Only 2092 * valid for TCP and UDP packets. 2093 * 2094 * src_ip, dst_ip ip addresses, in NETWORK format. 2095 * Only valid for IPv4 packets. 2096 */ 2097 u_int8_t proto; 2098 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2099 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2100 u_int16_t ip_len=0; 2101 int pktlen; 2102 2103 /* 2104 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2105 * MATCH_NONE when checked and not matched (q = NULL), 2106 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2107 */ 2108 int dyn_dir = MATCH_UNKNOWN; 2109 ipfw_dyn_rule *q = NULL; 2110 struct ip_fw_chain *chain = &layer3_chain; 2111 struct m_tag *mtag; 2112 2113 /* 2114 * We store in ulp a pointer to the upper layer protocol header. 2115 * In the ipv4 case this is easy to determine from the header, 2116 * but for ipv6 we might have some additional headers in the middle. 2117 * ulp is NULL if not found. 2118 */ 2119 void *ulp = NULL; /* upper layer protocol pointer. */ 2120 /* XXX ipv6 variables */ 2121 int is_ipv6 = 0; 2122 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2123 /* end of ipv6 variables */ 2124 int is_ipv4 = 0; 2125 2126 if (m->m_flags & M_SKIP_FIREWALL) 2127 return (IP_FW_PASS); /* accept */ 2128 2129 pktlen = m->m_pkthdr.len; 2130 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2131 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2132 2133/* 2134 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2135 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2136 * pointer might become stale after other pullups (but we never use it 2137 * this way). 2138 */ 2139#define PULLUP_TO(len, p, T) \ 2140do { \ 2141 int x = (len) + sizeof(T); \ 2142 if ((m)->m_len < x) { \ 2143 args->m = m = m_pullup(m, x); \ 2144 if (m == NULL) \ 2145 goto pullup_failed; \ 2146 } \ 2147 p = (mtod(m, char *) + (len)); \ 2148} while (0) 2149 2150 /* Identify IP packets and fill up variables. */ 2151 if (pktlen >= sizeof(struct ip6_hdr) && 2152 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2153 mtod(m, struct ip *)->ip_v == 6) { 2154 is_ipv6 = 1; 2155 args->f_id.addr_type = 6; 2156 hlen = sizeof(struct ip6_hdr); 2157 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2158 2159 /* Search extension headers to find upper layer protocols */ 2160 while (ulp == NULL) { 2161 switch (proto) { 2162 case IPPROTO_ICMPV6: 2163 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2164 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2165 break; 2166 2167 case IPPROTO_TCP: 2168 PULLUP_TO(hlen, ulp, struct tcphdr); 2169 dst_port = TCP(ulp)->th_dport; 2170 src_port = TCP(ulp)->th_sport; 2171 args->f_id.flags = TCP(ulp)->th_flags; 2172 break; 2173 2174 case IPPROTO_UDP: 2175 PULLUP_TO(hlen, ulp, struct udphdr); 2176 dst_port = UDP(ulp)->uh_dport; 2177 src_port = UDP(ulp)->uh_sport; 2178 break; 2179 2180 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2181 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2182 ext_hd |= EXT_HOPOPTS; 2183 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2184 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2185 ulp = NULL; 2186 break; 2187 2188 case IPPROTO_ROUTING: /* RFC 2460 */ 2189 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2190 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2191 case 0: 2192 break; 2193 default: 2194 printf("IPFW2: IPV6 - Unknown Routing " 2195 "Header type(%d)\n", 2196 ((struct ip6_rthdr *)ulp)->ip6r_type); 2197 if (fw_deny_unknown_exthdrs) 2198 return (IP_FW_DENY); 2199 break; 2200 } 2201 ext_hd |= EXT_ROUTING; 2202 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2203 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2204 ulp = NULL; 2205 break; 2206 2207 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2208 PULLUP_TO(hlen, ulp, struct ip6_frag); 2209 ext_hd |= EXT_FRAGMENT; 2210 hlen += sizeof (struct ip6_frag); 2211 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2212 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2213 IP6F_OFF_MASK; 2214 /* Add IP6F_MORE_FRAG for offset of first 2215 * fragment to be != 0. */ 2216 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2217 IP6F_MORE_FRAG; 2218 if (offset == 0) { 2219 printf("IPFW2: IPV6 - Invalid Fragment " 2220 "Header\n"); 2221 if (fw_deny_unknown_exthdrs) 2222 return (IP_FW_DENY); 2223 break; 2224 } 2225 args->f_id.frag_id6 = 2226 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2227 ulp = NULL; 2228 break; 2229 2230 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2231 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2232 ext_hd |= EXT_DSTOPTS; 2233 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2234 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2235 ulp = NULL; 2236 break; 2237 2238 case IPPROTO_AH: /* RFC 2402 */ 2239 PULLUP_TO(hlen, ulp, struct ip6_ext); 2240 ext_hd |= EXT_AH; 2241 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2242 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2243 ulp = NULL; 2244 break; 2245 2246 case IPPROTO_ESP: /* RFC 2406 */ 2247 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2248 /* Anything past Seq# is variable length and 2249 * data past this ext. header is encrypted. */ 2250 ext_hd |= EXT_ESP; 2251 break; 2252 2253 case IPPROTO_NONE: /* RFC 2460 */ 2254 PULLUP_TO(hlen, ulp, struct ip6_ext); 2255 /* Packet ends here. if ip6e_len!=0 octets 2256 * must be ignored. */ 2257 break; 2258 2259 case IPPROTO_OSPFIGP: 2260 /* XXX OSPF header check? */ 2261 PULLUP_TO(hlen, ulp, struct ip6_ext); 2262 break; 2263 2264 case IPPROTO_IPV6: /* RFC 2893 */ 2265 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2266 break; 2267 2268 case IPPROTO_IPV4: /* RFC 2893 */ 2269 PULLUP_TO(hlen, ulp, struct ip); 2270 break; 2271 2272 default: 2273 printf("IPFW2: IPV6 - Unknown Extension " 2274 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2275 if (fw_deny_unknown_exthdrs) 2276 return (IP_FW_DENY); 2277 PULLUP_TO(hlen, ulp, struct ip6_ext); 2278 break; 2279 } /*switch */ 2280 } 2281 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2282 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2283 args->f_id.src_ip = 0; 2284 args->f_id.dst_ip = 0; 2285 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2286 } else if (pktlen >= sizeof(struct ip) && 2287 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2288 mtod(m, struct ip *)->ip_v == 4) { 2289 is_ipv4 = 1; 2290 ip = mtod(m, struct ip *); 2291 hlen = ip->ip_hl << 2; 2292 args->f_id.addr_type = 4; 2293 2294 /* 2295 * Collect parameters into local variables for faster matching. 2296 */ 2297 proto = ip->ip_p; 2298 src_ip = ip->ip_src; 2299 dst_ip = ip->ip_dst; 2300 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2301 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2302 ip_len = ntohs(ip->ip_len); 2303 } else { 2304 offset = ip->ip_off & IP_OFFMASK; 2305 ip_len = ip->ip_len; 2306 } 2307 pktlen = ip_len < pktlen ? ip_len : pktlen; 2308 2309 if (offset == 0) { 2310 switch (proto) { 2311 case IPPROTO_TCP: 2312 PULLUP_TO(hlen, ulp, struct tcphdr); 2313 dst_port = TCP(ulp)->th_dport; 2314 src_port = TCP(ulp)->th_sport; 2315 args->f_id.flags = TCP(ulp)->th_flags; 2316 break; 2317 2318 case IPPROTO_UDP: 2319 PULLUP_TO(hlen, ulp, struct udphdr); 2320 dst_port = UDP(ulp)->uh_dport; 2321 src_port = UDP(ulp)->uh_sport; 2322 break; 2323 2324 case IPPROTO_ICMP: 2325 PULLUP_TO(hlen, ulp, struct icmphdr); 2326 args->f_id.flags = ICMP(ulp)->icmp_type; 2327 break; 2328 2329 default: 2330 break; 2331 } 2332 } 2333 2334 args->f_id.src_ip = ntohl(src_ip.s_addr); 2335 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2336 } 2337#undef PULLUP_TO 2338 if (proto) { /* we may have port numbers, store them */ 2339 args->f_id.proto = proto; 2340 args->f_id.src_port = src_port = ntohs(src_port); 2341 args->f_id.dst_port = dst_port = ntohs(dst_port); 2342 } 2343 2344 IPFW_RLOCK(chain); 2345 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2346 if (args->rule) { 2347 /* 2348 * Packet has already been tagged. Look for the next rule 2349 * to restart processing. 2350 * 2351 * If fw_one_pass != 0 then just accept it. 2352 * XXX should not happen here, but optimized out in 2353 * the caller. 2354 */ 2355 if (fw_one_pass) { 2356 IPFW_RUNLOCK(chain); 2357 return (IP_FW_PASS); 2358 } 2359 2360 f = args->rule->next_rule; 2361 if (f == NULL) 2362 f = lookup_next_rule(args->rule); 2363 } else { 2364 /* 2365 * Find the starting rule. It can be either the first 2366 * one, or the one after divert_rule if asked so. 2367 */ 2368 int skipto = mtag ? divert_cookie(mtag) : 0; 2369 2370 f = chain->rules; 2371 if (args->eh == NULL && skipto != 0) { 2372 if (skipto >= IPFW_DEFAULT_RULE) { 2373 IPFW_RUNLOCK(chain); 2374 return (IP_FW_DENY); /* invalid */ 2375 } 2376 while (f && f->rulenum <= skipto) 2377 f = f->next; 2378 if (f == NULL) { /* drop packet */ 2379 IPFW_RUNLOCK(chain); 2380 return (IP_FW_DENY); 2381 } 2382 } 2383 } 2384 /* reset divert rule to avoid confusion later */ 2385 if (mtag) { 2386 divinput_flags = divert_info(mtag) & 2387 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2388 m_tag_delete(m, mtag); 2389 } 2390 2391 /* 2392 * Now scan the rules, and parse microinstructions for each rule. 2393 */ 2394 for (; f; f = f->next) { 2395 ipfw_insn *cmd; 2396 uint32_t tablearg = 0; 2397 int l, cmdlen, skip_or; /* skip rest of OR block */ 2398 2399again: 2400 if (set_disable & (1 << f->set) ) 2401 continue; 2402 2403 skip_or = 0; 2404 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2405 l -= cmdlen, cmd += cmdlen) { 2406 int match; 2407 2408 /* 2409 * check_body is a jump target used when we find a 2410 * CHECK_STATE, and need to jump to the body of 2411 * the target rule. 2412 */ 2413 2414check_body: 2415 cmdlen = F_LEN(cmd); 2416 /* 2417 * An OR block (insn_1 || .. || insn_n) has the 2418 * F_OR bit set in all but the last instruction. 2419 * The first match will set "skip_or", and cause 2420 * the following instructions to be skipped until 2421 * past the one with the F_OR bit clear. 2422 */ 2423 if (skip_or) { /* skip this instruction */ 2424 if ((cmd->len & F_OR) == 0) 2425 skip_or = 0; /* next one is good */ 2426 continue; 2427 } 2428 match = 0; /* set to 1 if we succeed */ 2429 2430 switch (cmd->opcode) { 2431 /* 2432 * The first set of opcodes compares the packet's 2433 * fields with some pattern, setting 'match' if a 2434 * match is found. At the end of the loop there is 2435 * logic to deal with F_NOT and F_OR flags associated 2436 * with the opcode. 2437 */ 2438 case O_NOP: 2439 match = 1; 2440 break; 2441 2442 case O_FORWARD_MAC: 2443 printf("ipfw: opcode %d unimplemented\n", 2444 cmd->opcode); 2445 break; 2446 2447 case O_GID: 2448 case O_UID: 2449 case O_JAIL: 2450 /* 2451 * We only check offset == 0 && proto != 0, 2452 * as this ensures that we have a 2453 * packet with the ports info. 2454 */ 2455 if (offset!=0) 2456 break; 2457 if (is_ipv6) /* XXX to be fixed later */ 2458 break; 2459 if (proto == IPPROTO_TCP || 2460 proto == IPPROTO_UDP) 2461 match = check_uidgid( 2462 (ipfw_insn_u32 *)cmd, 2463 proto, oif, 2464 dst_ip, dst_port, 2465 src_ip, src_port, &fw_ugid_cache, 2466 &ugid_lookup, args->inp); 2467 break; 2468 2469 case O_RECV: 2470 match = iface_match(m->m_pkthdr.rcvif, 2471 (ipfw_insn_if *)cmd); 2472 break; 2473 2474 case O_XMIT: 2475 match = iface_match(oif, (ipfw_insn_if *)cmd); 2476 break; 2477 2478 case O_VIA: 2479 match = iface_match(oif ? oif : 2480 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2481 break; 2482 2483 case O_MACADDR2: 2484 if (args->eh != NULL) { /* have MAC header */ 2485 u_int32_t *want = (u_int32_t *) 2486 ((ipfw_insn_mac *)cmd)->addr; 2487 u_int32_t *mask = (u_int32_t *) 2488 ((ipfw_insn_mac *)cmd)->mask; 2489 u_int32_t *hdr = (u_int32_t *)args->eh; 2490 2491 match = 2492 ( want[0] == (hdr[0] & mask[0]) && 2493 want[1] == (hdr[1] & mask[1]) && 2494 want[2] == (hdr[2] & mask[2]) ); 2495 } 2496 break; 2497 2498 case O_MAC_TYPE: 2499 if (args->eh != NULL) { 2500 u_int16_t t = 2501 ntohs(args->eh->ether_type); 2502 u_int16_t *p = 2503 ((ipfw_insn_u16 *)cmd)->ports; 2504 int i; 2505 2506 for (i = cmdlen - 1; !match && i>0; 2507 i--, p += 2) 2508 match = (t>=p[0] && t<=p[1]); 2509 } 2510 break; 2511 2512 case O_FRAG: 2513 match = (offset != 0); 2514 break; 2515 2516 case O_IN: /* "out" is "not in" */ 2517 match = (oif == NULL); 2518 break; 2519 2520 case O_LAYER2: 2521 match = (args->eh != NULL); 2522 break; 2523 2524 case O_DIVERTED: 2525 match = (cmd->arg1 & 1 && divinput_flags & 2526 IP_FW_DIVERT_LOOPBACK_FLAG) || 2527 (cmd->arg1 & 2 && divinput_flags & 2528 IP_FW_DIVERT_OUTPUT_FLAG); 2529 break; 2530 2531 case O_PROTO: 2532 /* 2533 * We do not allow an arg of 0 so the 2534 * check of "proto" only suffices. 2535 */ 2536 match = (proto == cmd->arg1); 2537 break; 2538 2539 case O_IP_SRC: 2540 match = is_ipv4 && 2541 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2542 src_ip.s_addr); 2543 break; 2544 2545 case O_IP_SRC_LOOKUP: 2546 case O_IP_DST_LOOKUP: 2547 if (is_ipv4) { 2548 uint32_t a = 2549 (cmd->opcode == O_IP_DST_LOOKUP) ? 2550 dst_ip.s_addr : src_ip.s_addr; 2551 uint32_t v; 2552 2553 match = lookup_table(chain, cmd->arg1, a, 2554 &v); 2555 if (!match) 2556 break; 2557 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2558 match = 2559 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2560 else 2561 tablearg = v; 2562 } 2563 break; 2564 2565 case O_IP_SRC_MASK: 2566 case O_IP_DST_MASK: 2567 if (is_ipv4) { 2568 uint32_t a = 2569 (cmd->opcode == O_IP_DST_MASK) ? 2570 dst_ip.s_addr : src_ip.s_addr; 2571 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2572 int i = cmdlen-1; 2573 2574 for (; !match && i>0; i-= 2, p+= 2) 2575 match = (p[0] == (a & p[1])); 2576 } 2577 break; 2578 2579 case O_IP_SRC_ME: 2580 if (is_ipv4) { 2581 struct ifnet *tif; 2582 2583 INADDR_TO_IFP(src_ip, tif); 2584 match = (tif != NULL); 2585 } 2586 break; 2587 2588 case O_IP_DST_SET: 2589 case O_IP_SRC_SET: 2590 if (is_ipv4) { 2591 u_int32_t *d = (u_int32_t *)(cmd+1); 2592 u_int32_t addr = 2593 cmd->opcode == O_IP_DST_SET ? 2594 args->f_id.dst_ip : 2595 args->f_id.src_ip; 2596 2597 if (addr < d[0]) 2598 break; 2599 addr -= d[0]; /* subtract base */ 2600 match = (addr < cmd->arg1) && 2601 ( d[ 1 + (addr>>5)] & 2602 (1<<(addr & 0x1f)) ); 2603 } 2604 break; 2605 2606 case O_IP_DST: 2607 match = is_ipv4 && 2608 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2609 dst_ip.s_addr); 2610 break; 2611 2612 case O_IP_DST_ME: 2613 if (is_ipv4) { 2614 struct ifnet *tif; 2615 2616 INADDR_TO_IFP(dst_ip, tif); 2617 match = (tif != NULL); 2618 } 2619 break; 2620 2621 case O_IP_SRCPORT: 2622 case O_IP_DSTPORT: 2623 /* 2624 * offset == 0 && proto != 0 is enough 2625 * to guarantee that we have a 2626 * packet with port info. 2627 */ 2628 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2629 && offset == 0) { 2630 u_int16_t x = 2631 (cmd->opcode == O_IP_SRCPORT) ? 2632 src_port : dst_port ; 2633 u_int16_t *p = 2634 ((ipfw_insn_u16 *)cmd)->ports; 2635 int i; 2636 2637 for (i = cmdlen - 1; !match && i>0; 2638 i--, p += 2) 2639 match = (x>=p[0] && x<=p[1]); 2640 } 2641 break; 2642 2643 case O_ICMPTYPE: 2644 match = (offset == 0 && proto==IPPROTO_ICMP && 2645 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2646 break; 2647 2648#ifdef INET6 2649 case O_ICMP6TYPE: 2650 match = is_ipv6 && offset == 0 && 2651 proto==IPPROTO_ICMPV6 && 2652 icmp6type_match( 2653 ICMP6(ulp)->icmp6_type, 2654 (ipfw_insn_u32 *)cmd); 2655 break; 2656#endif /* INET6 */ 2657 2658 case O_IPOPT: 2659 match = (is_ipv4 && 2660 ipopts_match(mtod(m, struct ip *), cmd) ); 2661 break; 2662 2663 case O_IPVER: 2664 match = (is_ipv4 && 2665 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2666 break; 2667 2668 case O_IPID: 2669 case O_IPLEN: 2670 case O_IPTTL: 2671 if (is_ipv4) { /* only for IP packets */ 2672 uint16_t x; 2673 uint16_t *p; 2674 int i; 2675 2676 if (cmd->opcode == O_IPLEN) 2677 x = ip_len; 2678 else if (cmd->opcode == O_IPTTL) 2679 x = mtod(m, struct ip *)->ip_ttl; 2680 else /* must be IPID */ 2681 x = ntohs(mtod(m, struct ip *)->ip_id); 2682 if (cmdlen == 1) { 2683 match = (cmd->arg1 == x); 2684 break; 2685 } 2686 /* otherwise we have ranges */ 2687 p = ((ipfw_insn_u16 *)cmd)->ports; 2688 i = cmdlen - 1; 2689 for (; !match && i>0; i--, p += 2) 2690 match = (x >= p[0] && x <= p[1]); 2691 } 2692 break; 2693 2694 case O_IPPRECEDENCE: 2695 match = (is_ipv4 && 2696 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2697 break; 2698 2699 case O_IPTOS: 2700 match = (is_ipv4 && 2701 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2702 break; 2703 2704 case O_TCPDATALEN: 2705 if (proto == IPPROTO_TCP && offset == 0) { 2706 struct tcphdr *tcp; 2707 uint16_t x; 2708 uint16_t *p; 2709 int i; 2710 2711 tcp = TCP(ulp); 2712 x = ip_len - 2713 ((ip->ip_hl + tcp->th_off) << 2); 2714 if (cmdlen == 1) { 2715 match = (cmd->arg1 == x); 2716 break; 2717 } 2718 /* otherwise we have ranges */ 2719 p = ((ipfw_insn_u16 *)cmd)->ports; 2720 i = cmdlen - 1; 2721 for (; !match && i>0; i--, p += 2) 2722 match = (x >= p[0] && x <= p[1]); 2723 } 2724 break; 2725 2726 case O_TCPFLAGS: 2727 match = (proto == IPPROTO_TCP && offset == 0 && 2728 flags_match(cmd, TCP(ulp)->th_flags)); 2729 break; 2730 2731 case O_TCPOPTS: 2732 match = (proto == IPPROTO_TCP && offset == 0 && 2733 tcpopts_match(TCP(ulp), cmd)); 2734 break; 2735 2736 case O_TCPSEQ: 2737 match = (proto == IPPROTO_TCP && offset == 0 && 2738 ((ipfw_insn_u32 *)cmd)->d[0] == 2739 TCP(ulp)->th_seq); 2740 break; 2741 2742 case O_TCPACK: 2743 match = (proto == IPPROTO_TCP && offset == 0 && 2744 ((ipfw_insn_u32 *)cmd)->d[0] == 2745 TCP(ulp)->th_ack); 2746 break; 2747 2748 case O_TCPWIN: 2749 match = (proto == IPPROTO_TCP && offset == 0 && 2750 cmd->arg1 == TCP(ulp)->th_win); 2751 break; 2752 2753 case O_ESTAB: 2754 /* reject packets which have SYN only */ 2755 /* XXX should i also check for TH_ACK ? */ 2756 match = (proto == IPPROTO_TCP && offset == 0 && 2757 (TCP(ulp)->th_flags & 2758 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2759 break; 2760 2761 case O_ALTQ: { 2762 struct altq_tag *at; 2763 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2764 2765 match = 1; 2766 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2767 if (mtag != NULL) 2768 break; 2769 mtag = m_tag_get(PACKET_TAG_PF_QID, 2770 sizeof(struct altq_tag), 2771 M_NOWAIT); 2772 if (mtag == NULL) { 2773 /* 2774 * Let the packet fall back to the 2775 * default ALTQ. 2776 */ 2777 break; 2778 } 2779 at = (struct altq_tag *)(mtag+1); 2780 at->qid = altq->qid; 2781 if (is_ipv4) 2782 at->af = AF_INET; 2783 else 2784 at->af = AF_LINK; 2785 at->hdr = ip; 2786 m_tag_prepend(m, mtag); 2787 break; 2788 } 2789 2790 case O_LOG: 2791 if (fw_verbose) 2792 ipfw_log(f, hlen, args, m, oif, offset); 2793 match = 1; 2794 break; 2795 2796 case O_PROB: 2797 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2798 break; 2799 2800 case O_VERREVPATH: 2801 /* Outgoing packets automatically pass/match */ 2802 match = ((oif != NULL) || 2803 (m->m_pkthdr.rcvif == NULL) || 2804 ( 2805#ifdef INET6 2806 is_ipv6 ? 2807 verify_path6(&(args->f_id.src_ip6), 2808 m->m_pkthdr.rcvif) : 2809#endif 2810 verify_path(src_ip, m->m_pkthdr.rcvif))); 2811 break; 2812 2813 case O_VERSRCREACH: 2814 /* Outgoing packets automatically pass/match */ 2815 match = (hlen > 0 && ((oif != NULL) || 2816#ifdef INET6 2817 is_ipv6 ? 2818 verify_path6(&(args->f_id.src_ip6), 2819 NULL) : 2820#endif 2821 verify_path(src_ip, NULL))); 2822 break; 2823 2824 case O_ANTISPOOF: 2825 /* Outgoing packets automatically pass/match */ 2826 if (oif == NULL && hlen > 0 && 2827 ( (is_ipv4 && in_localaddr(src_ip)) 2828#ifdef INET6 2829 || (is_ipv6 && 2830 in6_localaddr(&(args->f_id.src_ip6))) 2831#endif 2832 )) 2833 match = 2834#ifdef INET6 2835 is_ipv6 ? verify_path6( 2836 &(args->f_id.src_ip6), 2837 m->m_pkthdr.rcvif) : 2838#endif 2839 verify_path(src_ip, 2840 m->m_pkthdr.rcvif); 2841 else 2842 match = 1; 2843 break; 2844 2845 case O_IPSEC: 2846#ifdef FAST_IPSEC 2847 match = (m_tag_find(m, 2848 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2849#endif 2850#ifdef IPSEC 2851 match = (ipsec_getnhist(m) != 0); 2852#endif 2853 /* otherwise no match */ 2854 break; 2855 2856#ifdef INET6 2857 case O_IP6_SRC: 2858 match = is_ipv6 && 2859 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2860 &((ipfw_insn_ip6 *)cmd)->addr6); 2861 break; 2862 2863 case O_IP6_DST: 2864 match = is_ipv6 && 2865 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2866 &((ipfw_insn_ip6 *)cmd)->addr6); 2867 break; 2868 case O_IP6_SRC_MASK: 2869 if (is_ipv6) { 2870 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2871 struct in6_addr p = args->f_id.src_ip6; 2872 2873 APPLY_MASK(&p, &te->mask6); 2874 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2875 } 2876 break; 2877 2878 case O_IP6_DST_MASK: 2879 if (is_ipv6) { 2880 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2881 struct in6_addr p = args->f_id.dst_ip6; 2882 2883 APPLY_MASK(&p, &te->mask6); 2884 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2885 } 2886 break; 2887 2888 case O_IP6_SRC_ME: 2889 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2890 break; 2891 2892 case O_IP6_DST_ME: 2893 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2894 break; 2895 2896 case O_FLOW6ID: 2897 match = is_ipv6 && 2898 flow6id_match(args->f_id.flow_id6, 2899 (ipfw_insn_u32 *) cmd); 2900 break; 2901 2902 case O_EXT_HDR: 2903 match = is_ipv6 && 2904 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2905 break; 2906 2907 case O_IP6: 2908 match = is_ipv6; 2909 break; 2910#endif 2911 2912 case O_IP4: 2913 match = is_ipv4; 2914 break; 2915 2916 case O_TAG: { 2917 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2918 tablearg : cmd->arg1; 2919 2920 /* Packet is already tagged with this tag? */ 2921 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2922 2923 /* We have `untag' action when F_NOT flag is 2924 * present. And we must remove this mtag from 2925 * mbuf and reset `match' to zero (`match' will 2926 * be inversed later). 2927 * Otherwise we should allocate new mtag and 2928 * push it into mbuf. 2929 */ 2930 if (cmd->len & F_NOT) { /* `untag' action */ 2931 if (mtag != NULL) 2932 m_tag_delete(m, mtag); 2933 } else if (mtag == NULL) { 2934 if ((mtag = m_tag_alloc(MTAG_IPFW, 2935 tag, 0, M_NOWAIT)) != NULL) 2936 m_tag_prepend(m, mtag); 2937 } 2938 match = (cmd->len & F_NOT) ? 0: 1; 2939 break; 2940 } 2941 2942 case O_TAGGED: { 2943 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2944 tablearg : cmd->arg1; 2945 2946 if (cmdlen == 1) { 2947 match = m_tag_locate(m, MTAG_IPFW, 2948 tag, NULL) != NULL; 2949 break; 2950 } 2951 2952 /* we have ranges */ 2953 for (mtag = m_tag_first(m); 2954 mtag != NULL && !match; 2955 mtag = m_tag_next(m, mtag)) { 2956 uint16_t *p; 2957 int i; 2958 2959 if (mtag->m_tag_cookie != MTAG_IPFW) 2960 continue; 2961 2962 p = ((ipfw_insn_u16 *)cmd)->ports; 2963 i = cmdlen - 1; 2964 for(; !match && i > 0; i--, p += 2) 2965 match = 2966 mtag->m_tag_id >= p[0] && 2967 mtag->m_tag_id <= p[1]; 2968 } 2969 break; 2970 } 2971 2972 /* 2973 * The second set of opcodes represents 'actions', 2974 * i.e. the terminal part of a rule once the packet 2975 * matches all previous patterns. 2976 * Typically there is only one action for each rule, 2977 * and the opcode is stored at the end of the rule 2978 * (but there are exceptions -- see below). 2979 * 2980 * In general, here we set retval and terminate the 2981 * outer loop (would be a 'break 3' in some language, 2982 * but we need to do a 'goto done'). 2983 * 2984 * Exceptions: 2985 * O_COUNT and O_SKIPTO actions: 2986 * instead of terminating, we jump to the next rule 2987 * ('goto next_rule', equivalent to a 'break 2'), 2988 * or to the SKIPTO target ('goto again' after 2989 * having set f, cmd and l), respectively. 2990 * 2991 * O_TAG, O_LOG and O_ALTQ action parameters: 2992 * perform some action and set match = 1; 2993 * 2994 * O_LIMIT and O_KEEP_STATE: these opcodes are 2995 * not real 'actions', and are stored right 2996 * before the 'action' part of the rule. 2997 * These opcodes try to install an entry in the 2998 * state tables; if successful, we continue with 2999 * the next opcode (match=1; break;), otherwise 3000 * the packet * must be dropped 3001 * ('goto done' after setting retval); 3002 * 3003 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3004 * cause a lookup of the state table, and a jump 3005 * to the 'action' part of the parent rule 3006 * ('goto check_body') if an entry is found, or 3007 * (CHECK_STATE only) a jump to the next rule if 3008 * the entry is not found ('goto next_rule'). 3009 * The result of the lookup is cached to make 3010 * further instances of these opcodes are 3011 * effectively NOPs. 3012 */ 3013 case O_LIMIT: 3014 case O_KEEP_STATE: 3015 if (install_state(f, 3016 (ipfw_insn_limit *)cmd, args, tablearg)) { 3017 retval = IP_FW_DENY; 3018 goto done; /* error/limit violation */ 3019 } 3020 match = 1; 3021 break; 3022 3023 case O_PROBE_STATE: 3024 case O_CHECK_STATE: 3025 /* 3026 * dynamic rules are checked at the first 3027 * keep-state or check-state occurrence, 3028 * with the result being stored in dyn_dir. 3029 * The compiler introduces a PROBE_STATE 3030 * instruction for us when we have a 3031 * KEEP_STATE (because PROBE_STATE needs 3032 * to be run first). 3033 */ 3034 if (dyn_dir == MATCH_UNKNOWN && 3035 (q = lookup_dyn_rule(&args->f_id, 3036 &dyn_dir, proto == IPPROTO_TCP ? 3037 TCP(ulp) : NULL)) 3038 != NULL) { 3039 /* 3040 * Found dynamic entry, update stats 3041 * and jump to the 'action' part of 3042 * the parent rule. 3043 */ 3044 q->pcnt++; 3045 q->bcnt += pktlen; 3046 f = q->rule; 3047 cmd = ACTION_PTR(f); 3048 l = f->cmd_len - f->act_ofs; 3049 IPFW_DYN_UNLOCK(); 3050 goto check_body; 3051 } 3052 /* 3053 * Dynamic entry not found. If CHECK_STATE, 3054 * skip to next rule, if PROBE_STATE just 3055 * ignore and continue with next opcode. 3056 */ 3057 if (cmd->opcode == O_CHECK_STATE) 3058 goto next_rule; 3059 match = 1; 3060 break; 3061 3062 case O_ACCEPT: 3063 retval = 0; /* accept */ 3064 goto done; 3065 3066 case O_PIPE: 3067 case O_QUEUE: 3068 args->rule = f; /* report matching rule */ 3069 if (cmd->arg1 == IP_FW_TABLEARG) 3070 args->cookie = tablearg; 3071 else 3072 args->cookie = cmd->arg1; 3073 retval = IP_FW_DUMMYNET; 3074 goto done; 3075 3076 case O_DIVERT: 3077 case O_TEE: { 3078 struct divert_tag *dt; 3079 3080 if (args->eh) /* not on layer 2 */ 3081 break; 3082 mtag = m_tag_get(PACKET_TAG_DIVERT, 3083 sizeof(struct divert_tag), 3084 M_NOWAIT); 3085 if (mtag == NULL) { 3086 /* XXX statistic */ 3087 /* drop packet */ 3088 IPFW_RUNLOCK(chain); 3089 return (IP_FW_DENY); 3090 } 3091 dt = (struct divert_tag *)(mtag+1); 3092 dt->cookie = f->rulenum; 3093 if (cmd->arg1 == IP_FW_TABLEARG) 3094 dt->info = tablearg; 3095 else 3096 dt->info = cmd->arg1; 3097 m_tag_prepend(m, mtag); 3098 retval = (cmd->opcode == O_DIVERT) ? 3099 IP_FW_DIVERT : IP_FW_TEE; 3100 goto done; 3101 } 3102 3103 case O_COUNT: 3104 case O_SKIPTO: 3105 f->pcnt++; /* update stats */ 3106 f->bcnt += pktlen; 3107 f->timestamp = time_uptime; 3108 if (cmd->opcode == O_COUNT) 3109 goto next_rule; 3110 /* handle skipto */ 3111 if (f->next_rule == NULL) 3112 lookup_next_rule(f); 3113 f = f->next_rule; 3114 goto again; 3115 3116 case O_REJECT: 3117 /* 3118 * Drop the packet and send a reject notice 3119 * if the packet is not ICMP (or is an ICMP 3120 * query), and it is not multicast/broadcast. 3121 */ 3122 if (hlen > 0 && is_ipv4 && offset == 0 && 3123 (proto != IPPROTO_ICMP || 3124 is_icmp_query(ICMP(ulp))) && 3125 !(m->m_flags & (M_BCAST|M_MCAST)) && 3126 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3127 send_reject(args, cmd->arg1, ip_len); 3128 m = args->m; 3129 } 3130 /* FALLTHROUGH */ 3131#ifdef INET6 3132 case O_UNREACH6: 3133 if (hlen > 0 && is_ipv6 && 3134 ((offset & IP6F_OFF_MASK) == 0) && 3135 (proto != IPPROTO_ICMPV6 || 3136 (is_icmp6_query(args->f_id.flags) == 1)) && 3137 !(m->m_flags & (M_BCAST|M_MCAST)) && 3138 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3139 send_reject6(args, cmd->arg1, hlen); 3140 m = args->m; 3141 } 3142 /* FALLTHROUGH */ 3143#endif 3144 case O_DENY: 3145 retval = IP_FW_DENY; 3146 goto done; 3147 3148 case O_FORWARD_IP: 3149 if (args->eh) /* not valid on layer2 pkts */ 3150 break; 3151 if (!q || dyn_dir == MATCH_FORWARD) 3152 args->next_hop = 3153 &((ipfw_insn_sa *)cmd)->sa; 3154 retval = IP_FW_PASS; 3155 goto done; 3156 3157 case O_NETGRAPH: 3158 case O_NGTEE: 3159 args->rule = f; /* report matching rule */ 3160 if (cmd->arg1 == IP_FW_TABLEARG) 3161 args->cookie = tablearg; 3162 else 3163 args->cookie = cmd->arg1; 3164 retval = (cmd->opcode == O_NETGRAPH) ? 3165 IP_FW_NETGRAPH : IP_FW_NGTEE; 3166 goto done; 3167 3168 default: 3169 panic("-- unknown opcode %d\n", cmd->opcode); 3170 } /* end of switch() on opcodes */ 3171 3172 if (cmd->len & F_NOT) 3173 match = !match; 3174 3175 if (match) { 3176 if (cmd->len & F_OR) 3177 skip_or = 1; 3178 } else { 3179 if (!(cmd->len & F_OR)) /* not an OR block, */ 3180 break; /* try next rule */ 3181 } 3182 3183 } /* end of inner for, scan opcodes */ 3184 3185next_rule:; /* try next rule */ 3186 3187 } /* end of outer for, scan rules */ 3188 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3189 IPFW_RUNLOCK(chain); 3190 return (IP_FW_DENY); 3191 3192done: 3193 /* Update statistics */ 3194 f->pcnt++; 3195 f->bcnt += pktlen; 3196 f->timestamp = time_uptime; 3197 IPFW_RUNLOCK(chain); 3198 return (retval); 3199 3200pullup_failed: 3201 if (fw_verbose) 3202 printf("ipfw: pullup failed\n"); 3203 return (IP_FW_DENY); 3204} 3205 3206/* 3207 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3208 * These will be reconstructed on the fly as packets are matched. 3209 */ 3210static void 3211flush_rule_ptrs(struct ip_fw_chain *chain) 3212{ 3213 struct ip_fw *rule; 3214 3215 IPFW_WLOCK_ASSERT(chain); 3216 3217 for (rule = chain->rules; rule; rule = rule->next) 3218 rule->next_rule = NULL; 3219} 3220 3221/* 3222 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3223 * possibly create a rule number and add the rule to the list. 3224 * Update the rule_number in the input struct so the caller knows it as well. 3225 */ 3226static int 3227add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3228{ 3229 struct ip_fw *rule, *f, *prev; 3230 int l = RULESIZE(input_rule); 3231 3232 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3233 return (EINVAL); 3234 3235 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3236 if (rule == NULL) 3237 return (ENOSPC); 3238 3239 bcopy(input_rule, rule, l); 3240 3241 rule->next = NULL; 3242 rule->next_rule = NULL; 3243 3244 rule->pcnt = 0; 3245 rule->bcnt = 0; 3246 rule->timestamp = 0; 3247 3248 IPFW_WLOCK(chain); 3249 3250 if (chain->rules == NULL) { /* default rule */ 3251 chain->rules = rule; 3252 goto done; 3253 } 3254 3255 /* 3256 * If rulenum is 0, find highest numbered rule before the 3257 * default rule, and add autoinc_step 3258 */ 3259 if (autoinc_step < 1) 3260 autoinc_step = 1; 3261 else if (autoinc_step > 1000) 3262 autoinc_step = 1000; 3263 if (rule->rulenum == 0) { 3264 /* 3265 * locate the highest numbered rule before default 3266 */ 3267 for (f = chain->rules; f; f = f->next) { 3268 if (f->rulenum == IPFW_DEFAULT_RULE) 3269 break; 3270 rule->rulenum = f->rulenum; 3271 } 3272 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3273 rule->rulenum += autoinc_step; 3274 input_rule->rulenum = rule->rulenum; 3275 } 3276 3277 /* 3278 * Now insert the new rule in the right place in the sorted list. 3279 */ 3280 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3281 if (f->rulenum > rule->rulenum) { /* found the location */ 3282 if (prev) { 3283 rule->next = f; 3284 prev->next = rule; 3285 } else { /* head insert */ 3286 rule->next = chain->rules; 3287 chain->rules = rule; 3288 } 3289 break; 3290 } 3291 } 3292 flush_rule_ptrs(chain); 3293done: 3294 static_count++; 3295 static_len += l; 3296 IPFW_WUNLOCK(chain); 3297 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3298 rule->rulenum, static_count);) 3299 return (0); 3300} 3301 3302/** 3303 * Remove a static rule (including derived * dynamic rules) 3304 * and place it on the ``reap list'' for later reclamation. 3305 * The caller is in charge of clearing rule pointers to avoid 3306 * dangling pointers. 3307 * @return a pointer to the next entry. 3308 * Arguments are not checked, so they better be correct. 3309 */ 3310static struct ip_fw * 3311remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3312{ 3313 struct ip_fw *n; 3314 int l = RULESIZE(rule); 3315 3316 IPFW_WLOCK_ASSERT(chain); 3317 3318 n = rule->next; 3319 IPFW_DYN_LOCK(); 3320 remove_dyn_rule(rule, NULL /* force removal */); 3321 IPFW_DYN_UNLOCK(); 3322 if (prev == NULL) 3323 chain->rules = n; 3324 else 3325 prev->next = n; 3326 static_count--; 3327 static_len -= l; 3328 3329 rule->next = chain->reap; 3330 chain->reap = rule; 3331 3332 return n; 3333} 3334 3335/** 3336 * Reclaim storage associated with a list of rules. This is 3337 * typically the list created using remove_rule. 3338 */ 3339static void 3340reap_rules(struct ip_fw *head) 3341{ 3342 struct ip_fw *rule; 3343 3344 while ((rule = head) != NULL) { 3345 head = head->next; 3346 if (DUMMYNET_LOADED) 3347 ip_dn_ruledel_ptr(rule); 3348 free(rule, M_IPFW); 3349 } 3350} 3351 3352/* 3353 * Remove all rules from a chain (except rules in set RESVD_SET 3354 * unless kill_default = 1). The caller is responsible for 3355 * reclaiming storage for the rules left in chain->reap. 3356 */ 3357static void 3358free_chain(struct ip_fw_chain *chain, int kill_default) 3359{ 3360 struct ip_fw *prev, *rule; 3361 3362 IPFW_WLOCK_ASSERT(chain); 3363 3364 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3365 for (prev = NULL, rule = chain->rules; rule ; ) 3366 if (kill_default || rule->set != RESVD_SET) 3367 rule = remove_rule(chain, rule, prev); 3368 else { 3369 prev = rule; 3370 rule = rule->next; 3371 } 3372} 3373 3374/** 3375 * Remove all rules with given number, and also do set manipulation. 3376 * Assumes chain != NULL && *chain != NULL. 3377 * 3378 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3379 * the next 8 bits are the new set, the top 8 bits are the command: 3380 * 3381 * 0 delete rules with given number 3382 * 1 delete rules with given set number 3383 * 2 move rules with given number to new set 3384 * 3 move rules with given set number to new set 3385 * 4 swap sets with given numbers 3386 */ 3387static int 3388del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3389{ 3390 struct ip_fw *prev = NULL, *rule; 3391 u_int16_t rulenum; /* rule or old_set */ 3392 u_int8_t cmd, new_set; 3393 3394 rulenum = arg & 0xffff; 3395 cmd = (arg >> 24) & 0xff; 3396 new_set = (arg >> 16) & 0xff; 3397 3398 if (cmd > 4) 3399 return EINVAL; 3400 if (new_set > RESVD_SET) 3401 return EINVAL; 3402 if (cmd == 0 || cmd == 2) { 3403 if (rulenum >= IPFW_DEFAULT_RULE) 3404 return EINVAL; 3405 } else { 3406 if (rulenum > RESVD_SET) /* old_set */ 3407 return EINVAL; 3408 } 3409 3410 IPFW_WLOCK(chain); 3411 rule = chain->rules; 3412 chain->reap = NULL; 3413 switch (cmd) { 3414 case 0: /* delete rules with given number */ 3415 /* 3416 * locate first rule to delete 3417 */ 3418 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3419 ; 3420 if (rule->rulenum != rulenum) { 3421 IPFW_WUNLOCK(chain); 3422 return EINVAL; 3423 } 3424 3425 /* 3426 * flush pointers outside the loop, then delete all matching 3427 * rules. prev remains the same throughout the cycle. 3428 */ 3429 flush_rule_ptrs(chain); 3430 while (rule->rulenum == rulenum) 3431 rule = remove_rule(chain, rule, prev); 3432 break; 3433 3434 case 1: /* delete all rules with given set number */ 3435 flush_rule_ptrs(chain); 3436 rule = chain->rules; 3437 while (rule->rulenum < IPFW_DEFAULT_RULE) 3438 if (rule->set == rulenum) 3439 rule = remove_rule(chain, rule, prev); 3440 else { 3441 prev = rule; 3442 rule = rule->next; 3443 } 3444 break; 3445 3446 case 2: /* move rules with given number to new set */ 3447 rule = chain->rules; 3448 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3449 if (rule->rulenum == rulenum) 3450 rule->set = new_set; 3451 break; 3452 3453 case 3: /* move rules with given set number to new set */ 3454 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3455 if (rule->set == rulenum) 3456 rule->set = new_set; 3457 break; 3458 3459 case 4: /* swap two sets */ 3460 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3461 if (rule->set == rulenum) 3462 rule->set = new_set; 3463 else if (rule->set == new_set) 3464 rule->set = rulenum; 3465 break; 3466 } 3467 /* 3468 * Look for rules to reclaim. We grab the list before 3469 * releasing the lock then reclaim them w/o the lock to 3470 * avoid a LOR with dummynet. 3471 */ 3472 rule = chain->reap; 3473 chain->reap = NULL; 3474 IPFW_WUNLOCK(chain); 3475 if (rule) 3476 reap_rules(rule); 3477 return 0; 3478} 3479 3480/* 3481 * Clear counters for a specific rule. 3482 * The enclosing "table" is assumed locked. 3483 */ 3484static void 3485clear_counters(struct ip_fw *rule, int log_only) 3486{ 3487 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3488 3489 if (log_only == 0) { 3490 rule->bcnt = rule->pcnt = 0; 3491 rule->timestamp = 0; 3492 } 3493 if (l->o.opcode == O_LOG) 3494 l->log_left = l->max_log; 3495} 3496 3497/** 3498 * Reset some or all counters on firewall rules. 3499 * @arg frwl is null to clear all entries, or contains a specific 3500 * rule number. 3501 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3502 */ 3503static int 3504zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3505{ 3506 struct ip_fw *rule; 3507 char *msg; 3508 3509 IPFW_WLOCK(chain); 3510 if (rulenum == 0) { 3511 norule_counter = 0; 3512 for (rule = chain->rules; rule; rule = rule->next) 3513 clear_counters(rule, log_only); 3514 msg = log_only ? "ipfw: All logging counts reset.\n" : 3515 "ipfw: Accounting cleared.\n"; 3516 } else { 3517 int cleared = 0; 3518 /* 3519 * We can have multiple rules with the same number, so we 3520 * need to clear them all. 3521 */ 3522 for (rule = chain->rules; rule; rule = rule->next) 3523 if (rule->rulenum == rulenum) { 3524 while (rule && rule->rulenum == rulenum) { 3525 clear_counters(rule, log_only); 3526 rule = rule->next; 3527 } 3528 cleared = 1; 3529 break; 3530 } 3531 if (!cleared) { /* we did not find any matching rules */ 3532 IPFW_WUNLOCK(chain); 3533 return (EINVAL); 3534 } 3535 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3536 "ipfw: Entry %d cleared.\n"; 3537 } 3538 IPFW_WUNLOCK(chain); 3539 3540 if (fw_verbose) 3541 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3542 return (0); 3543} 3544 3545/* 3546 * Check validity of the structure before insert. 3547 * Fortunately rules are simple, so this mostly need to check rule sizes. 3548 */ 3549static int 3550check_ipfw_struct(struct ip_fw *rule, int size) 3551{ 3552 int l, cmdlen = 0; 3553 int have_action=0; 3554 ipfw_insn *cmd; 3555 3556 if (size < sizeof(*rule)) { 3557 printf("ipfw: rule too short\n"); 3558 return (EINVAL); 3559 } 3560 /* first, check for valid size */ 3561 l = RULESIZE(rule); 3562 if (l != size) { 3563 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3564 return (EINVAL); 3565 } 3566 if (rule->act_ofs >= rule->cmd_len) { 3567 printf("ipfw: bogus action offset (%u > %u)\n", 3568 rule->act_ofs, rule->cmd_len - 1); 3569 return (EINVAL); 3570 } 3571 /* 3572 * Now go for the individual checks. Very simple ones, basically only 3573 * instruction sizes. 3574 */ 3575 for (l = rule->cmd_len, cmd = rule->cmd ; 3576 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3577 cmdlen = F_LEN(cmd); 3578 if (cmdlen > l) { 3579 printf("ipfw: opcode %d size truncated\n", 3580 cmd->opcode); 3581 return EINVAL; 3582 } 3583 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3584 switch (cmd->opcode) { 3585 case O_PROBE_STATE: 3586 case O_KEEP_STATE: 3587 case O_PROTO: 3588 case O_IP_SRC_ME: 3589 case O_IP_DST_ME: 3590 case O_LAYER2: 3591 case O_IN: 3592 case O_FRAG: 3593 case O_DIVERTED: 3594 case O_IPOPT: 3595 case O_IPTOS: 3596 case O_IPPRECEDENCE: 3597 case O_IPVER: 3598 case O_TCPWIN: 3599 case O_TCPFLAGS: 3600 case O_TCPOPTS: 3601 case O_ESTAB: 3602 case O_VERREVPATH: 3603 case O_VERSRCREACH: 3604 case O_ANTISPOOF: 3605 case O_IPSEC: 3606#ifdef INET6 3607 case O_IP6_SRC_ME: 3608 case O_IP6_DST_ME: 3609 case O_EXT_HDR: 3610 case O_IP6: 3611#endif 3612 case O_IP4: 3613 case O_TAG: 3614 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3615 goto bad_size; 3616 break; 3617 3618 case O_UID: 3619 case O_GID: 3620 case O_JAIL: 3621 case O_IP_SRC: 3622 case O_IP_DST: 3623 case O_TCPSEQ: 3624 case O_TCPACK: 3625 case O_PROB: 3626 case O_ICMPTYPE: 3627 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3628 goto bad_size; 3629 break; 3630 3631 case O_LIMIT: 3632 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3633 goto bad_size; 3634 break; 3635 3636 case O_LOG: 3637 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3638 goto bad_size; 3639 3640 ((ipfw_insn_log *)cmd)->log_left = 3641 ((ipfw_insn_log *)cmd)->max_log; 3642 3643 break; 3644 3645 case O_IP_SRC_MASK: 3646 case O_IP_DST_MASK: 3647 /* only odd command lengths */ 3648 if ( !(cmdlen & 1) || cmdlen > 31) 3649 goto bad_size; 3650 break; 3651 3652 case O_IP_SRC_SET: 3653 case O_IP_DST_SET: 3654 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3655 printf("ipfw: invalid set size %d\n", 3656 cmd->arg1); 3657 return EINVAL; 3658 } 3659 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3660 (cmd->arg1+31)/32 ) 3661 goto bad_size; 3662 break; 3663 3664 case O_IP_SRC_LOOKUP: 3665 case O_IP_DST_LOOKUP: 3666 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3667 printf("ipfw: invalid table number %d\n", 3668 cmd->arg1); 3669 return (EINVAL); 3670 } 3671 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3672 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3673 goto bad_size; 3674 break; 3675 3676 case O_MACADDR2: 3677 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3678 goto bad_size; 3679 break; 3680 3681 case O_NOP: 3682 case O_IPID: 3683 case O_IPTTL: 3684 case O_IPLEN: 3685 case O_TCPDATALEN: 3686 case O_TAGGED: 3687 if (cmdlen < 1 || cmdlen > 31) 3688 goto bad_size; 3689 break; 3690 3691 case O_MAC_TYPE: 3692 case O_IP_SRCPORT: 3693 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3694 if (cmdlen < 2 || cmdlen > 31) 3695 goto bad_size; 3696 break; 3697 3698 case O_RECV: 3699 case O_XMIT: 3700 case O_VIA: 3701 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3702 goto bad_size; 3703 break; 3704 3705 case O_ALTQ: 3706 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3707 goto bad_size; 3708 break; 3709 3710 case O_PIPE: 3711 case O_QUEUE: 3712 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3713 goto bad_size; 3714 goto check_action; 3715 3716 case O_FORWARD_IP: 3717#ifdef IPFIREWALL_FORWARD 3718 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3719 goto bad_size; 3720 goto check_action; 3721#else 3722 return EINVAL; 3723#endif 3724 3725 case O_DIVERT: 3726 case O_TEE: 3727 if (ip_divert_ptr == NULL) 3728 return EINVAL; 3729 else 3730 goto check_size; 3731 case O_NETGRAPH: 3732 case O_NGTEE: 3733 if (!NG_IPFW_LOADED) 3734 return EINVAL; 3735 else 3736 goto check_size; 3737 case O_FORWARD_MAC: /* XXX not implemented yet */ 3738 case O_CHECK_STATE: 3739 case O_COUNT: 3740 case O_ACCEPT: 3741 case O_DENY: 3742 case O_REJECT: 3743#ifdef INET6 3744 case O_UNREACH6: 3745#endif 3746 case O_SKIPTO: 3747check_size: 3748 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3749 goto bad_size; 3750check_action: 3751 if (have_action) { 3752 printf("ipfw: opcode %d, multiple actions" 3753 " not allowed\n", 3754 cmd->opcode); 3755 return EINVAL; 3756 } 3757 have_action = 1; 3758 if (l != cmdlen) { 3759 printf("ipfw: opcode %d, action must be" 3760 " last opcode\n", 3761 cmd->opcode); 3762 return EINVAL; 3763 } 3764 break; 3765#ifdef INET6 3766 case O_IP6_SRC: 3767 case O_IP6_DST: 3768 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3769 F_INSN_SIZE(ipfw_insn)) 3770 goto bad_size; 3771 break; 3772 3773 case O_FLOW6ID: 3774 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3775 ((ipfw_insn_u32 *)cmd)->o.arg1) 3776 goto bad_size; 3777 break; 3778 3779 case O_IP6_SRC_MASK: 3780 case O_IP6_DST_MASK: 3781 if ( !(cmdlen & 1) || cmdlen > 127) 3782 goto bad_size; 3783 break; 3784 case O_ICMP6TYPE: 3785 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3786 goto bad_size; 3787 break; 3788#endif 3789 3790 default: 3791 switch (cmd->opcode) { 3792#ifndef INET6 3793 case O_IP6_SRC_ME: 3794 case O_IP6_DST_ME: 3795 case O_EXT_HDR: 3796 case O_IP6: 3797 case O_UNREACH6: 3798 case O_IP6_SRC: 3799 case O_IP6_DST: 3800 case O_FLOW6ID: 3801 case O_IP6_SRC_MASK: 3802 case O_IP6_DST_MASK: 3803 case O_ICMP6TYPE: 3804 printf("ipfw: no IPv6 support in kernel\n"); 3805 return EPROTONOSUPPORT; 3806#endif 3807 default: 3808 printf("ipfw: opcode %d, unknown opcode\n", 3809 cmd->opcode); 3810 return EINVAL; 3811 } 3812 } 3813 } 3814 if (have_action == 0) { 3815 printf("ipfw: missing action\n"); 3816 return EINVAL; 3817 } 3818 return 0; 3819 3820bad_size: 3821 printf("ipfw: opcode %d size %d wrong\n", 3822 cmd->opcode, cmdlen); 3823 return EINVAL; 3824} 3825 3826/* 3827 * Copy the static and dynamic rules to the supplied buffer 3828 * and return the amount of space actually used. 3829 */ 3830static size_t 3831ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3832{ 3833 char *bp = buf; 3834 char *ep = bp + space; 3835 struct ip_fw *rule; 3836 int i; 3837 3838 /* XXX this can take a long time and locking will block packet flow */ 3839 IPFW_RLOCK(chain); 3840 for (rule = chain->rules; rule ; rule = rule->next) { 3841 /* 3842 * Verify the entry fits in the buffer in case the 3843 * rules changed between calculating buffer space and 3844 * now. This would be better done using a generation 3845 * number but should suffice for now. 3846 */ 3847 i = RULESIZE(rule); 3848 if (bp + i <= ep) { 3849 bcopy(rule, bp, i); 3850 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3851 sizeof(set_disable)); 3852 bp += i; 3853 } 3854 } 3855 IPFW_RUNLOCK(chain); 3856 if (ipfw_dyn_v) { 3857 ipfw_dyn_rule *p, *last = NULL; 3858 3859 IPFW_DYN_LOCK(); 3860 for (i = 0 ; i < curr_dyn_buckets; i++) 3861 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3862 if (bp + sizeof *p <= ep) { 3863 ipfw_dyn_rule *dst = 3864 (ipfw_dyn_rule *)bp; 3865 bcopy(p, dst, sizeof *p); 3866 bcopy(&(p->rule->rulenum), &(dst->rule), 3867 sizeof(p->rule->rulenum)); 3868 /* 3869 * store a non-null value in "next". 3870 * The userland code will interpret a 3871 * NULL here as a marker 3872 * for the last dynamic rule. 3873 */ 3874 bcopy(&dst, &dst->next, sizeof(dst)); 3875 last = dst; 3876 dst->expire = 3877 TIME_LEQ(dst->expire, time_uptime) ? 3878 0 : dst->expire - time_uptime ; 3879 bp += sizeof(ipfw_dyn_rule); 3880 } 3881 } 3882 IPFW_DYN_UNLOCK(); 3883 if (last != NULL) /* mark last dynamic rule */ 3884 bzero(&last->next, sizeof(last)); 3885 } 3886 return (bp - (char *)buf); 3887} 3888 3889 3890/** 3891 * {set|get}sockopt parser. 3892 */ 3893static int 3894ipfw_ctl(struct sockopt *sopt) 3895{ 3896#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3897 int error, rule_num; 3898 size_t size; 3899 struct ip_fw *buf, *rule; 3900 u_int32_t rulenum[2]; 3901 3902 error = suser(sopt->sopt_td); 3903 if (error) 3904 return (error); 3905 3906 /* 3907 * Disallow modifications in really-really secure mode, but still allow 3908 * the logging counters to be reset. 3909 */ 3910 if (sopt->sopt_name == IP_FW_ADD || 3911 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3912 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3913 if (error) 3914 return (error); 3915 } 3916 3917 error = 0; 3918 3919 switch (sopt->sopt_name) { 3920 case IP_FW_GET: 3921 /* 3922 * pass up a copy of the current rules. Static rules 3923 * come first (the last of which has number IPFW_DEFAULT_RULE), 3924 * followed by a possibly empty list of dynamic rule. 3925 * The last dynamic rule has NULL in the "next" field. 3926 * 3927 * Note that the calculated size is used to bound the 3928 * amount of data returned to the user. The rule set may 3929 * change between calculating the size and returning the 3930 * data in which case we'll just return what fits. 3931 */ 3932 size = static_len; /* size of static rules */ 3933 if (ipfw_dyn_v) /* add size of dyn.rules */ 3934 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3935 3936 /* 3937 * XXX todo: if the user passes a short length just to know 3938 * how much room is needed, do not bother filling up the 3939 * buffer, just jump to the sooptcopyout. 3940 */ 3941 buf = malloc(size, M_TEMP, M_WAITOK); 3942 error = sooptcopyout(sopt, buf, 3943 ipfw_getrules(&layer3_chain, buf, size)); 3944 free(buf, M_TEMP); 3945 break; 3946 3947 case IP_FW_FLUSH: 3948 /* 3949 * Normally we cannot release the lock on each iteration. 3950 * We could do it here only because we start from the head all 3951 * the times so there is no risk of missing some entries. 3952 * On the other hand, the risk is that we end up with 3953 * a very inconsistent ruleset, so better keep the lock 3954 * around the whole cycle. 3955 * 3956 * XXX this code can be improved by resetting the head of 3957 * the list to point to the default rule, and then freeing 3958 * the old list without the need for a lock. 3959 */ 3960 3961 IPFW_WLOCK(&layer3_chain); 3962 layer3_chain.reap = NULL; 3963 free_chain(&layer3_chain, 0 /* keep default rule */); 3964 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3965 IPFW_WUNLOCK(&layer3_chain); 3966 if (layer3_chain.reap != NULL) 3967 reap_rules(rule); 3968 break; 3969 3970 case IP_FW_ADD: 3971 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3972 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3973 sizeof(struct ip_fw) ); 3974 if (error == 0) 3975 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3976 if (error == 0) { 3977 error = add_rule(&layer3_chain, rule); 3978 size = RULESIZE(rule); 3979 if (!error && sopt->sopt_dir == SOPT_GET) 3980 error = sooptcopyout(sopt, rule, size); 3981 } 3982 free(rule, M_TEMP); 3983 break; 3984 3985 case IP_FW_DEL: 3986 /* 3987 * IP_FW_DEL is used for deleting single rules or sets, 3988 * and (ab)used to atomically manipulate sets. Argument size 3989 * is used to distinguish between the two: 3990 * sizeof(u_int32_t) 3991 * delete single rule or set of rules, 3992 * or reassign rules (or sets) to a different set. 3993 * 2*sizeof(u_int32_t) 3994 * atomic disable/enable sets. 3995 * first u_int32_t contains sets to be disabled, 3996 * second u_int32_t contains sets to be enabled. 3997 */ 3998 error = sooptcopyin(sopt, rulenum, 3999 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4000 if (error) 4001 break; 4002 size = sopt->sopt_valsize; 4003 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4004 error = del_entry(&layer3_chain, rulenum[0]); 4005 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4006 set_disable = 4007 (set_disable | rulenum[0]) & ~rulenum[1] & 4008 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4009 else 4010 error = EINVAL; 4011 break; 4012 4013 case IP_FW_ZERO: 4014 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 4015 rule_num = 0; 4016 if (sopt->sopt_val != 0) { 4017 error = sooptcopyin(sopt, &rule_num, 4018 sizeof(int), sizeof(int)); 4019 if (error) 4020 break; 4021 } 4022 error = zero_entry(&layer3_chain, rule_num, 4023 sopt->sopt_name == IP_FW_RESETLOG); 4024 break; 4025 4026 case IP_FW_TABLE_ADD: 4027 { 4028 ipfw_table_entry ent; 4029 4030 error = sooptcopyin(sopt, &ent, 4031 sizeof(ent), sizeof(ent)); 4032 if (error) 4033 break; 4034 error = add_table_entry(&layer3_chain, ent.tbl, 4035 ent.addr, ent.masklen, ent.value); 4036 } 4037 break; 4038 4039 case IP_FW_TABLE_DEL: 4040 { 4041 ipfw_table_entry ent; 4042 4043 error = sooptcopyin(sopt, &ent, 4044 sizeof(ent), sizeof(ent)); 4045 if (error) 4046 break; 4047 error = del_table_entry(&layer3_chain, ent.tbl, 4048 ent.addr, ent.masklen); 4049 } 4050 break; 4051 4052 case IP_FW_TABLE_FLUSH: 4053 { 4054 u_int16_t tbl; 4055 4056 error = sooptcopyin(sopt, &tbl, 4057 sizeof(tbl), sizeof(tbl)); 4058 if (error) 4059 break; 4060 IPFW_WLOCK(&layer3_chain); 4061 error = flush_table(&layer3_chain, tbl); 4062 IPFW_WUNLOCK(&layer3_chain); 4063 } 4064 break; 4065 4066 case IP_FW_TABLE_GETSIZE: 4067 { 4068 u_int32_t tbl, cnt; 4069 4070 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4071 sizeof(tbl)))) 4072 break; 4073 IPFW_RLOCK(&layer3_chain); 4074 error = count_table(&layer3_chain, tbl, &cnt); 4075 IPFW_RUNLOCK(&layer3_chain); 4076 if (error) 4077 break; 4078 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4079 } 4080 break; 4081 4082 case IP_FW_TABLE_LIST: 4083 { 4084 ipfw_table *tbl; 4085 4086 if (sopt->sopt_valsize < sizeof(*tbl)) { 4087 error = EINVAL; 4088 break; 4089 } 4090 size = sopt->sopt_valsize; 4091 tbl = malloc(size, M_TEMP, M_WAITOK); 4092 if (tbl == NULL) { 4093 error = ENOMEM; 4094 break; 4095 } 4096 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4097 if (error) { 4098 free(tbl, M_TEMP); 4099 break; 4100 } 4101 tbl->size = (size - sizeof(*tbl)) / 4102 sizeof(ipfw_table_entry); 4103 IPFW_RLOCK(&layer3_chain); 4104 error = dump_table(&layer3_chain, tbl); 4105 IPFW_RUNLOCK(&layer3_chain); 4106 if (error) { 4107 free(tbl, M_TEMP); 4108 break; 4109 } 4110 error = sooptcopyout(sopt, tbl, size); 4111 free(tbl, M_TEMP); 4112 } 4113 break; 4114 4115 default: 4116 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4117 error = EINVAL; 4118 } 4119 4120 return (error); 4121#undef RULE_MAXSIZE 4122} 4123 4124/** 4125 * dummynet needs a reference to the default rule, because rules can be 4126 * deleted while packets hold a reference to them. When this happens, 4127 * dummynet changes the reference to the default rule (it could well be a 4128 * NULL pointer, but this way we do not need to check for the special 4129 * case, plus here he have info on the default behaviour). 4130 */ 4131struct ip_fw *ip_fw_default_rule; 4132 4133/* 4134 * This procedure is only used to handle keepalives. It is invoked 4135 * every dyn_keepalive_period 4136 */ 4137static void 4138ipfw_tick(void * __unused unused) 4139{ 4140 struct mbuf *m0, *m, *mnext, **mtailp; 4141 int i; 4142 ipfw_dyn_rule *q; 4143 4144 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4145 goto done; 4146 4147 /* 4148 * We make a chain of packets to go out here -- not deferring 4149 * until after we drop the IPFW dynamic rule lock would result 4150 * in a lock order reversal with the normal packet input -> ipfw 4151 * call stack. 4152 */ 4153 m0 = NULL; 4154 mtailp = &m0; 4155 IPFW_DYN_LOCK(); 4156 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4157 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4158 if (q->dyn_type == O_LIMIT_PARENT) 4159 continue; 4160 if (q->id.proto != IPPROTO_TCP) 4161 continue; 4162 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4163 continue; 4164 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4165 q->expire)) 4166 continue; /* too early */ 4167 if (TIME_LEQ(q->expire, time_uptime)) 4168 continue; /* too late, rule expired */ 4169 4170 *mtailp = send_pkt(&(q->id), q->ack_rev - 1, 4171 q->ack_fwd, TH_SYN); 4172 if (*mtailp != NULL) 4173 mtailp = &(*mtailp)->m_nextpkt; 4174 *mtailp = send_pkt(&(q->id), q->ack_fwd - 1, 4175 q->ack_rev, 0); 4176 if (*mtailp != NULL) 4177 mtailp = &(*mtailp)->m_nextpkt; 4178 } 4179 } 4180 IPFW_DYN_UNLOCK(); 4181 for (m = mnext = m0; m != NULL; m = mnext) { 4182 mnext = m->m_nextpkt; 4183 m->m_nextpkt = NULL; 4184 ip_output(m, NULL, NULL, 0, NULL, NULL); 4185 } 4186done: 4187 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4188} 4189 4190int 4191ipfw_init(void) 4192{ 4193 struct ip_fw default_rule; 4194 int error; 4195 4196#ifdef INET6 4197 /* Setup IPv6 fw sysctl tree. */ 4198 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4199 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4200 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4201 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4202 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4203 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4204 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4205 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4206 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4207 &fw_deny_unknown_exthdrs, 0, 4208 "Deny packets with unknown IPv6 Extension Headers"); 4209#endif 4210 4211 layer3_chain.rules = NULL; 4212 IPFW_LOCK_INIT(&layer3_chain); 4213 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4214 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4215 UMA_ALIGN_PTR, 0); 4216 IPFW_DYN_LOCK_INIT(); 4217 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4218 4219 bzero(&default_rule, sizeof default_rule); 4220 4221 default_rule.act_ofs = 0; 4222 default_rule.rulenum = IPFW_DEFAULT_RULE; 4223 default_rule.cmd_len = 1; 4224 default_rule.set = RESVD_SET; 4225 4226 default_rule.cmd[0].len = 1; 4227 default_rule.cmd[0].opcode = 4228#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4229 1 ? O_ACCEPT : 4230#endif 4231 O_DENY; 4232 4233 error = add_rule(&layer3_chain, &default_rule); 4234 if (error != 0) { 4235 printf("ipfw2: error %u initializing default rule " 4236 "(support disabled)\n", error); 4237 IPFW_DYN_LOCK_DESTROY(); 4238 IPFW_LOCK_DESTROY(&layer3_chain); 4239 uma_zdestroy(ipfw_dyn_rule_zone); 4240 return (error); 4241 } 4242 4243 ip_fw_default_rule = layer3_chain.rules; 4244 printf("ipfw2 " 4245#ifdef INET6 4246 "(+ipv6) " 4247#endif 4248 "initialized, divert %s, " 4249 "rule-based forwarding " 4250#ifdef IPFIREWALL_FORWARD 4251 "enabled, " 4252#else 4253 "disabled, " 4254#endif 4255 "default to %s, logging ", 4256#ifdef IPDIVERT 4257 "enabled", 4258#else 4259 "loadable", 4260#endif 4261 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4262 4263#ifdef IPFIREWALL_VERBOSE 4264 fw_verbose = 1; 4265#endif 4266#ifdef IPFIREWALL_VERBOSE_LIMIT 4267 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4268#endif 4269 if (fw_verbose == 0) 4270 printf("disabled\n"); 4271 else if (verbose_limit == 0) 4272 printf("unlimited\n"); 4273 else 4274 printf("limited to %d packets/entry by default\n", 4275 verbose_limit); 4276 4277 error = init_tables(&layer3_chain); 4278 if (error) { 4279 IPFW_DYN_LOCK_DESTROY(); 4280 IPFW_LOCK_DESTROY(&layer3_chain); 4281 uma_zdestroy(ipfw_dyn_rule_zone); 4282 return (error); 4283 } 4284 ip_fw_ctl_ptr = ipfw_ctl; 4285 ip_fw_chk_ptr = ipfw_chk; 4286 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4287 4288 return (0); 4289} 4290 4291void 4292ipfw_destroy(void) 4293{ 4294 struct ip_fw *reap; 4295 4296 ip_fw_chk_ptr = NULL; 4297 ip_fw_ctl_ptr = NULL; 4298 callout_drain(&ipfw_timeout); 4299 IPFW_WLOCK(&layer3_chain); 4300 flush_tables(&layer3_chain); 4301 layer3_chain.reap = NULL; 4302 free_chain(&layer3_chain, 1 /* kill default rule */); 4303 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4304 IPFW_WUNLOCK(&layer3_chain); 4305 if (reap != NULL) 4306 reap_rules(reap); 4307 IPFW_DYN_LOCK_DESTROY(); 4308 uma_zdestroy(ipfw_dyn_rule_zone); 4309 IPFW_LOCK_DESTROY(&layer3_chain); 4310 4311#ifdef INET6 4312 /* Free IPv6 fw sysctl tree. */ 4313 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4314#endif 4315 4316 printf("IP firewall unloaded\n"); 4317} 4318