ip_fw2.c revision 158580
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 158580 2006-05-14 23:42:24Z mlaier $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdn.h" 38#include "opt_inet.h" 39#ifndef INET 40#error IPFIREWALL requires INET. 41#endif /* INET */ 42#endif 43#include "opt_inet6.h" 44#include "opt_ipsec.h" 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/condvar.h> 49#include <sys/malloc.h> 50#include <sys/mbuf.h> 51#include <sys/kernel.h> 52#include <sys/lock.h> 53#include <sys/jail.h> 54#include <sys/module.h> 55#include <sys/proc.h> 56#include <sys/rwlock.h> 57#include <sys/socket.h> 58#include <sys/socketvar.h> 59#include <sys/sysctl.h> 60#include <sys/syslog.h> 61#include <sys/ucred.h> 62#include <net/if.h> 63#include <net/radix.h> 64#include <net/route.h> 65#include <netinet/in.h> 66#include <netinet/in_systm.h> 67#include <netinet/in_var.h> 68#include <netinet/in_pcb.h> 69#include <netinet/ip.h> 70#include <netinet/ip_var.h> 71#include <netinet/ip_icmp.h> 72#include <netinet/ip_fw.h> 73#include <netinet/ip_divert.h> 74#include <netinet/ip_dummynet.h> 75#include <netinet/tcp.h> 76#include <netinet/tcp_timer.h> 77#include <netinet/tcp_var.h> 78#include <netinet/tcpip.h> 79#include <netinet/udp.h> 80#include <netinet/udp_var.h> 81 82#include <netgraph/ng_ipfw.h> 83 84#include <altq/if_altq.h> 85 86#ifdef IPSEC 87#include <netinet6/ipsec.h> 88#endif 89 90#include <netinet/ip6.h> 91#include <netinet/icmp6.h> 92#ifdef INET6 93#include <netinet6/scope6_var.h> 94#endif 95 96#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 97 98#include <machine/in_cksum.h> /* XXX for in_cksum */ 99 100/* 101 * set_disable contains one bit per set value (0..31). 102 * If the bit is set, all rules with the corresponding set 103 * are disabled. Set RESVD_SET(31) is reserved for the default rule 104 * and rules that are not deleted by the flush command, 105 * and CANNOT be disabled. 106 * Rules in set RESVD_SET can only be deleted explicitly. 107 */ 108static u_int32_t set_disable; 109 110static int fw_verbose; 111static int verbose_limit; 112 113static struct callout ipfw_timeout; 114static uma_zone_t ipfw_dyn_rule_zone; 115#define IPFW_DEFAULT_RULE 65535 116 117/* 118 * Data structure to cache our ucred related 119 * information. This structure only gets used if 120 * the user specified UID/GID based constraints in 121 * a firewall rule. 122 */ 123struct ip_fw_ugid { 124 gid_t fw_groups[NGROUPS]; 125 int fw_ngroups; 126 uid_t fw_uid; 127 int fw_prid; 128}; 129 130#define IPFW_TABLES_MAX 128 131struct ip_fw_chain { 132 struct ip_fw *rules; /* list of rules */ 133 struct ip_fw *reap; /* list of rules to reap */ 134 struct radix_node_head *tables[IPFW_TABLES_MAX]; 135 struct rwlock rwmtx; 136}; 137#define IPFW_LOCK_INIT(_chain) \ 138 rw_init(&(_chain)->rwmtx, "IPFW static rules") 139#define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx) 140#define IPFW_WLOCK_ASSERT(_chain) do { \ 141 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \ 142 NET_ASSERT_GIANT(); \ 143} while (0) 144 145#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 146#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 147#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 148#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 149 150/* 151 * list of rules for layer 3 152 */ 153static struct ip_fw_chain layer3_chain; 154 155MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 156MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 157 158struct table_entry { 159 struct radix_node rn[2]; 160 struct sockaddr_in addr, mask; 161 u_int32_t value; 162}; 163 164static int fw_debug = 1; 165static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 166 167extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 168 169#ifdef SYSCTL_NODE 170SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 171SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 172 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 173 ipfw_chg_hook, "I", "Enable ipfw"); 174SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 175 &autoinc_step, 0, "Rule number autincrement step"); 176SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 177 CTLFLAG_RW | CTLFLAG_SECURE3, 178 &fw_one_pass, 0, 179 "Only do a single pass through ipfw when using dummynet(4)"); 180SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 181 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 182SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 183 CTLFLAG_RW | CTLFLAG_SECURE3, 184 &fw_verbose, 0, "Log matches to ipfw rules"); 185SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 186 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 187 188/* 189 * Description of dynamic rules. 190 * 191 * Dynamic rules are stored in lists accessed through a hash table 192 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 193 * be modified through the sysctl variable dyn_buckets which is 194 * updated when the table becomes empty. 195 * 196 * XXX currently there is only one list, ipfw_dyn. 197 * 198 * When a packet is received, its address fields are first masked 199 * with the mask defined for the rule, then hashed, then matched 200 * against the entries in the corresponding list. 201 * Dynamic rules can be used for different purposes: 202 * + stateful rules; 203 * + enforcing limits on the number of sessions; 204 * + in-kernel NAT (not implemented yet) 205 * 206 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 207 * measured in seconds and depending on the flags. 208 * 209 * The total number of dynamic rules is stored in dyn_count. 210 * The max number of dynamic rules is dyn_max. When we reach 211 * the maximum number of rules we do not create anymore. This is 212 * done to avoid consuming too much memory, but also too much 213 * time when searching on each packet (ideally, we should try instead 214 * to put a limit on the length of the list on each bucket...). 215 * 216 * Each dynamic rule holds a pointer to the parent ipfw rule so 217 * we know what action to perform. Dynamic rules are removed when 218 * the parent rule is deleted. XXX we should make them survive. 219 * 220 * There are some limitations with dynamic rules -- we do not 221 * obey the 'randomized match', and we do not do multiple 222 * passes through the firewall. XXX check the latter!!! 223 */ 224static ipfw_dyn_rule **ipfw_dyn_v = NULL; 225static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 226static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 227 228static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 229#define IPFW_DYN_LOCK_INIT() \ 230 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 231#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 232#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 233#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 234#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 235 236/* 237 * Timeouts for various events in handing dynamic rules. 238 */ 239static u_int32_t dyn_ack_lifetime = 300; 240static u_int32_t dyn_syn_lifetime = 20; 241static u_int32_t dyn_fin_lifetime = 1; 242static u_int32_t dyn_rst_lifetime = 1; 243static u_int32_t dyn_udp_lifetime = 10; 244static u_int32_t dyn_short_lifetime = 5; 245 246/* 247 * Keepalives are sent if dyn_keepalive is set. They are sent every 248 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 249 * seconds of lifetime of a rule. 250 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 251 * than dyn_keepalive_period. 252 */ 253 254static u_int32_t dyn_keepalive_interval = 20; 255static u_int32_t dyn_keepalive_period = 5; 256static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 257 258static u_int32_t static_count; /* # of static rules */ 259static u_int32_t static_len; /* size in bytes of static rules */ 260static u_int32_t dyn_count; /* # of dynamic rules */ 261static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 262 263SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 264 &dyn_buckets, 0, "Number of dyn. buckets"); 265SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 266 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 267SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 268 &dyn_count, 0, "Number of dyn. rules"); 269SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 270 &dyn_max, 0, "Max number of dyn. rules"); 271SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 272 &static_count, 0, "Number of static rules"); 273SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 274 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 275SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 276 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 277SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 278 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 279SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 280 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 281SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 282 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 283SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 284 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 285SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 286 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 287 288#ifdef INET6 289/* 290 * IPv6 specific variables 291 */ 292SYSCTL_DECL(_net_inet6_ip6); 293 294static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 295static struct sysctl_oid *ip6_fw_sysctl_tree; 296#endif /* INET6 */ 297#endif /* SYSCTL_NODE */ 298 299static int fw_deny_unknown_exthdrs = 1; 300 301 302/* 303 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 304 * Other macros just cast void * into the appropriate type 305 */ 306#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 307#define TCP(p) ((struct tcphdr *)(p)) 308#define UDP(p) ((struct udphdr *)(p)) 309#define ICMP(p) ((struct icmphdr *)(p)) 310#define ICMP6(p) ((struct icmp6_hdr *)(p)) 311 312static __inline int 313icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 314{ 315 int type = icmp->icmp_type; 316 317 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 318} 319 320#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 321 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 322 323static int 324is_icmp_query(struct icmphdr *icmp) 325{ 326 int type = icmp->icmp_type; 327 328 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 329} 330#undef TT 331 332/* 333 * The following checks use two arrays of 8 or 16 bits to store the 334 * bits that we want set or clear, respectively. They are in the 335 * low and high half of cmd->arg1 or cmd->d[0]. 336 * 337 * We scan options and store the bits we find set. We succeed if 338 * 339 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 340 * 341 * The code is sometimes optimized not to store additional variables. 342 */ 343 344static int 345flags_match(ipfw_insn *cmd, u_int8_t bits) 346{ 347 u_char want_clear; 348 bits = ~bits; 349 350 if ( ((cmd->arg1 & 0xff) & bits) != 0) 351 return 0; /* some bits we want set were clear */ 352 want_clear = (cmd->arg1 >> 8) & 0xff; 353 if ( (want_clear & bits) != want_clear) 354 return 0; /* some bits we want clear were set */ 355 return 1; 356} 357 358static int 359ipopts_match(struct ip *ip, ipfw_insn *cmd) 360{ 361 int optlen, bits = 0; 362 u_char *cp = (u_char *)(ip + 1); 363 int x = (ip->ip_hl << 2) - sizeof (struct ip); 364 365 for (; x > 0; x -= optlen, cp += optlen) { 366 int opt = cp[IPOPT_OPTVAL]; 367 368 if (opt == IPOPT_EOL) 369 break; 370 if (opt == IPOPT_NOP) 371 optlen = 1; 372 else { 373 optlen = cp[IPOPT_OLEN]; 374 if (optlen <= 0 || optlen > x) 375 return 0; /* invalid or truncated */ 376 } 377 switch (opt) { 378 379 default: 380 break; 381 382 case IPOPT_LSRR: 383 bits |= IP_FW_IPOPT_LSRR; 384 break; 385 386 case IPOPT_SSRR: 387 bits |= IP_FW_IPOPT_SSRR; 388 break; 389 390 case IPOPT_RR: 391 bits |= IP_FW_IPOPT_RR; 392 break; 393 394 case IPOPT_TS: 395 bits |= IP_FW_IPOPT_TS; 396 break; 397 } 398 } 399 return (flags_match(cmd, bits)); 400} 401 402static int 403tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 404{ 405 int optlen, bits = 0; 406 u_char *cp = (u_char *)(tcp + 1); 407 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 408 409 for (; x > 0; x -= optlen, cp += optlen) { 410 int opt = cp[0]; 411 if (opt == TCPOPT_EOL) 412 break; 413 if (opt == TCPOPT_NOP) 414 optlen = 1; 415 else { 416 optlen = cp[1]; 417 if (optlen <= 0) 418 break; 419 } 420 421 switch (opt) { 422 423 default: 424 break; 425 426 case TCPOPT_MAXSEG: 427 bits |= IP_FW_TCPOPT_MSS; 428 break; 429 430 case TCPOPT_WINDOW: 431 bits |= IP_FW_TCPOPT_WINDOW; 432 break; 433 434 case TCPOPT_SACK_PERMITTED: 435 case TCPOPT_SACK: 436 bits |= IP_FW_TCPOPT_SACK; 437 break; 438 439 case TCPOPT_TIMESTAMP: 440 bits |= IP_FW_TCPOPT_TS; 441 break; 442 443 } 444 } 445 return (flags_match(cmd, bits)); 446} 447 448static int 449iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 450{ 451 if (ifp == NULL) /* no iface with this packet, match fails */ 452 return 0; 453 /* Check by name or by IP address */ 454 if (cmd->name[0] != '\0') { /* match by name */ 455 /* Check name */ 456 if (cmd->p.glob) { 457 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 458 return(1); 459 } else { 460 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 461 return(1); 462 } 463 } else { 464 struct ifaddr *ia; 465 466 /* XXX lock? */ 467 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 468 if (ia->ifa_addr == NULL) 469 continue; 470 if (ia->ifa_addr->sa_family != AF_INET) 471 continue; 472 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 473 (ia->ifa_addr))->sin_addr.s_addr) 474 return(1); /* match */ 475 } 476 } 477 return(0); /* no match, fail ... */ 478} 479 480/* 481 * The verify_path function checks if a route to the src exists and 482 * if it is reachable via ifp (when provided). 483 * 484 * The 'verrevpath' option checks that the interface that an IP packet 485 * arrives on is the same interface that traffic destined for the 486 * packet's source address would be routed out of. The 'versrcreach' 487 * option just checks that the source address is reachable via any route 488 * (except default) in the routing table. These two are a measure to block 489 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 490 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 491 * is purposely reminiscent of the Cisco IOS command, 492 * 493 * ip verify unicast reverse-path 494 * ip verify unicast source reachable-via any 495 * 496 * which implements the same functionality. But note that syntax is 497 * misleading. The check may be performed on all IP packets whether unicast, 498 * multicast, or broadcast. 499 */ 500static int 501verify_path(struct in_addr src, struct ifnet *ifp) 502{ 503 struct route ro; 504 struct sockaddr_in *dst; 505 506 bzero(&ro, sizeof(ro)); 507 508 dst = (struct sockaddr_in *)&(ro.ro_dst); 509 dst->sin_family = AF_INET; 510 dst->sin_len = sizeof(*dst); 511 dst->sin_addr = src; 512 rtalloc_ign(&ro, RTF_CLONING); 513 514 if (ro.ro_rt == NULL) 515 return 0; 516 517 /* 518 * If ifp is provided, check for equality with rtentry. 519 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 520 * in order to pass packets injected back by if_simloop(): 521 * if useloopback == 1 routing entry (via lo0) for our own address 522 * may exist, so we need to handle routing assymetry. 523 */ 524 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 525 RTFREE(ro.ro_rt); 526 return 0; 527 } 528 529 /* if no ifp provided, check if rtentry is not default route */ 530 if (ifp == NULL && 531 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 532 RTFREE(ro.ro_rt); 533 return 0; 534 } 535 536 /* or if this is a blackhole/reject route */ 537 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 538 RTFREE(ro.ro_rt); 539 return 0; 540 } 541 542 /* found valid route */ 543 RTFREE(ro.ro_rt); 544 return 1; 545} 546 547#ifdef INET6 548/* 549 * ipv6 specific rules here... 550 */ 551static __inline int 552icmp6type_match (int type, ipfw_insn_u32 *cmd) 553{ 554 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 555} 556 557static int 558flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 559{ 560 int i; 561 for (i=0; i <= cmd->o.arg1; ++i ) 562 if (curr_flow == cmd->d[i] ) 563 return 1; 564 return 0; 565} 566 567/* support for IP6_*_ME opcodes */ 568static int 569search_ip6_addr_net (struct in6_addr * ip6_addr) 570{ 571 struct ifnet *mdc; 572 struct ifaddr *mdc2; 573 struct in6_ifaddr *fdm; 574 struct in6_addr copia; 575 576 TAILQ_FOREACH(mdc, &ifnet, if_link) 577 for (mdc2 = mdc->if_addrlist.tqh_first; mdc2; 578 mdc2 = mdc2->ifa_list.tqe_next) { 579 if (!mdc2->ifa_addr) 580 continue; 581 if (mdc2->ifa_addr->sa_family == AF_INET6) { 582 fdm = (struct in6_ifaddr *)mdc2; 583 copia = fdm->ia_addr.sin6_addr; 584 /* need for leaving scope_id in the sock_addr */ 585 in6_clearscope(&copia); 586 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 587 return 1; 588 } 589 } 590 return 0; 591} 592 593static int 594verify_path6(struct in6_addr *src, struct ifnet *ifp) 595{ 596 struct route_in6 ro; 597 struct sockaddr_in6 *dst; 598 599 bzero(&ro, sizeof(ro)); 600 601 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 602 dst->sin6_family = AF_INET6; 603 dst->sin6_len = sizeof(*dst); 604 dst->sin6_addr = *src; 605 rtalloc_ign((struct route *)&ro, RTF_CLONING); 606 607 if (ro.ro_rt == NULL) 608 return 0; 609 610 /* 611 * if ifp is provided, check for equality with rtentry 612 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 613 * to support the case of sending packets to an address of our own. 614 * (where the former interface is the first argument of if_simloop() 615 * (=ifp), the latter is lo0) 616 */ 617 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 618 RTFREE(ro.ro_rt); 619 return 0; 620 } 621 622 /* if no ifp provided, check if rtentry is not default route */ 623 if (ifp == NULL && 624 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 625 RTFREE(ro.ro_rt); 626 return 0; 627 } 628 629 /* or if this is a blackhole/reject route */ 630 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 631 RTFREE(ro.ro_rt); 632 return 0; 633 } 634 635 /* found valid route */ 636 RTFREE(ro.ro_rt); 637 return 1; 638 639} 640static __inline int 641hash_packet6(struct ipfw_flow_id *id) 642{ 643 u_int32_t i; 644 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 645 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 646 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 647 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 648 (id->dst_port) ^ (id->src_port); 649 return i; 650} 651 652static int 653is_icmp6_query(int icmp6_type) 654{ 655 if ((icmp6_type <= ICMP6_MAXTYPE) && 656 (icmp6_type == ICMP6_ECHO_REQUEST || 657 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 658 icmp6_type == ICMP6_WRUREQUEST || 659 icmp6_type == ICMP6_FQDN_QUERY || 660 icmp6_type == ICMP6_NI_QUERY)) 661 return (1); 662 663 return (0); 664} 665 666static void 667send_reject6(struct ip_fw_args *args, int code, u_short offset, u_int hlen) 668{ 669 if (code == ICMP6_UNREACH_RST && offset == 0 && 670 args->f_id.proto == IPPROTO_TCP) { 671 struct ip6_hdr *ip6; 672 struct tcphdr *tcp; 673 tcp_seq ack, seq; 674 int flags; 675 struct { 676 struct ip6_hdr ip6; 677 struct tcphdr th; 678 } ti; 679 680 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) { 681 args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr)); 682 if (args->m == NULL) 683 return; 684 } 685 686 ip6 = mtod(args->m, struct ip6_hdr *); 687 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 688 689 if ((tcp->th_flags & TH_RST) != 0) { 690 m_freem(args->m); 691 return; 692 } 693 694 ti.ip6 = *ip6; 695 ti.th = *tcp; 696 ti.th.th_seq = ntohl(ti.th.th_seq); 697 ti.th.th_ack = ntohl(ti.th.th_ack); 698 ti.ip6.ip6_nxt = IPPROTO_TCP; 699 700 if (ti.th.th_flags & TH_ACK) { 701 ack = 0; 702 seq = ti.th.th_ack; 703 flags = TH_RST; 704 } else { 705 ack = ti.th.th_seq; 706 if (((args->m)->m_flags & M_PKTHDR) != 0) { 707 ack += (args->m)->m_pkthdr.len - hlen 708 - (ti.th.th_off << 2); 709 } else if (ip6->ip6_plen) { 710 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) 711 - hlen - (ti.th.th_off << 2); 712 } else { 713 m_freem(args->m); 714 return; 715 } 716 if (tcp->th_flags & TH_SYN) 717 ack++; 718 seq = 0; 719 flags = TH_RST|TH_ACK; 720 } 721 bcopy(&ti, ip6, sizeof(ti)); 722 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1), 723 args->m, ack, seq, flags); 724 725 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 726 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0); 727 728 } else 729 m_freem(args->m); 730 731 args->m = NULL; 732} 733 734#endif /* INET6 */ 735 736static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 737 738#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 739#define SNP(buf) buf, sizeof(buf) 740 741/* 742 * We enter here when we have a rule with O_LOG. 743 * XXX this function alone takes about 2Kbytes of code! 744 */ 745static void 746ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 747 struct mbuf *m, struct ifnet *oif, u_short offset) 748{ 749 struct ether_header *eh = args->eh; 750 char *action; 751 int limit_reached = 0; 752 char action2[40], proto[128], fragment[32]; 753 754 fragment[0] = '\0'; 755 proto[0] = '\0'; 756 757 if (f == NULL) { /* bogus pkt */ 758 if (verbose_limit != 0 && norule_counter >= verbose_limit) 759 return; 760 norule_counter++; 761 if (norule_counter == verbose_limit) 762 limit_reached = verbose_limit; 763 action = "Refuse"; 764 } else { /* O_LOG is the first action, find the real one */ 765 ipfw_insn *cmd = ACTION_PTR(f); 766 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 767 768 if (l->max_log != 0 && l->log_left == 0) 769 return; 770 l->log_left--; 771 if (l->log_left == 0) 772 limit_reached = l->max_log; 773 cmd += F_LEN(cmd); /* point to first action */ 774 if (cmd->opcode == O_ALTQ) { 775 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 776 777 snprintf(SNPARGS(action2, 0), "Altq %d", 778 altq->qid); 779 cmd += F_LEN(cmd); 780 } 781 if (cmd->opcode == O_PROB) 782 cmd += F_LEN(cmd); 783 784 action = action2; 785 switch (cmd->opcode) { 786 case O_DENY: 787 action = "Deny"; 788 break; 789 790 case O_REJECT: 791 if (cmd->arg1==ICMP_REJECT_RST) 792 action = "Reset"; 793 else if (cmd->arg1==ICMP_UNREACH_HOST) 794 action = "Reject"; 795 else 796 snprintf(SNPARGS(action2, 0), "Unreach %d", 797 cmd->arg1); 798 break; 799 800 case O_UNREACH6: 801 if (cmd->arg1==ICMP6_UNREACH_RST) 802 action = "Reset"; 803 else 804 snprintf(SNPARGS(action2, 0), "Unreach %d", 805 cmd->arg1); 806 break; 807 808 case O_ACCEPT: 809 action = "Accept"; 810 break; 811 case O_COUNT: 812 action = "Count"; 813 break; 814 case O_DIVERT: 815 snprintf(SNPARGS(action2, 0), "Divert %d", 816 cmd->arg1); 817 break; 818 case O_TEE: 819 snprintf(SNPARGS(action2, 0), "Tee %d", 820 cmd->arg1); 821 break; 822 case O_SKIPTO: 823 snprintf(SNPARGS(action2, 0), "SkipTo %d", 824 cmd->arg1); 825 break; 826 case O_PIPE: 827 snprintf(SNPARGS(action2, 0), "Pipe %d", 828 cmd->arg1); 829 break; 830 case O_QUEUE: 831 snprintf(SNPARGS(action2, 0), "Queue %d", 832 cmd->arg1); 833 break; 834 case O_FORWARD_IP: { 835 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 836 int len; 837 838 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 839 inet_ntoa(sa->sa.sin_addr)); 840 if (sa->sa.sin_port) 841 snprintf(SNPARGS(action2, len), ":%d", 842 sa->sa.sin_port); 843 } 844 break; 845 case O_NETGRAPH: 846 snprintf(SNPARGS(action2, 0), "Netgraph %d", 847 cmd->arg1); 848 break; 849 case O_NGTEE: 850 snprintf(SNPARGS(action2, 0), "Ngtee %d", 851 cmd->arg1); 852 break; 853 default: 854 action = "UNKNOWN"; 855 break; 856 } 857 } 858 859 if (hlen == 0) { /* non-ip */ 860 snprintf(SNPARGS(proto, 0), "MAC"); 861 862 } else { 863 int len; 864 char src[48], dst[48]; 865 struct icmphdr *icmp; 866 struct tcphdr *tcp; 867 struct udphdr *udp; 868 /* Initialize to make compiler happy. */ 869 struct ip *ip = NULL; 870#ifdef INET6 871 struct ip6_hdr *ip6 = NULL; 872 struct icmp6_hdr *icmp6; 873#endif 874 src[0] = '\0'; 875 dst[0] = '\0'; 876#ifdef INET6 877 if (args->f_id.addr_type == 6) { 878 snprintf(src, sizeof(src), "[%s]", 879 ip6_sprintf(&args->f_id.src_ip6)); 880 snprintf(dst, sizeof(dst), "[%s]", 881 ip6_sprintf(&args->f_id.dst_ip6)); 882 883 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 884 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 885 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 886 } else 887#endif 888 { 889 ip = mtod(m, struct ip *); 890 tcp = L3HDR(struct tcphdr, ip); 891 udp = L3HDR(struct udphdr, ip); 892 893 inet_ntoa_r(ip->ip_src, src); 894 inet_ntoa_r(ip->ip_dst, dst); 895 } 896 897 switch (args->f_id.proto) { 898 case IPPROTO_TCP: 899 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 900 if (offset == 0) 901 snprintf(SNPARGS(proto, len), ":%d %s:%d", 902 ntohs(tcp->th_sport), 903 dst, 904 ntohs(tcp->th_dport)); 905 else 906 snprintf(SNPARGS(proto, len), " %s", dst); 907 break; 908 909 case IPPROTO_UDP: 910 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 911 if (offset == 0) 912 snprintf(SNPARGS(proto, len), ":%d %s:%d", 913 ntohs(udp->uh_sport), 914 dst, 915 ntohs(udp->uh_dport)); 916 else 917 snprintf(SNPARGS(proto, len), " %s", dst); 918 break; 919 920 case IPPROTO_ICMP: 921 icmp = L3HDR(struct icmphdr, ip); 922 if (offset == 0) 923 len = snprintf(SNPARGS(proto, 0), 924 "ICMP:%u.%u ", 925 icmp->icmp_type, icmp->icmp_code); 926 else 927 len = snprintf(SNPARGS(proto, 0), "ICMP "); 928 len += snprintf(SNPARGS(proto, len), "%s", src); 929 snprintf(SNPARGS(proto, len), " %s", dst); 930 break; 931#ifdef INET6 932 case IPPROTO_ICMPV6: 933 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 934 if (offset == 0) 935 len = snprintf(SNPARGS(proto, 0), 936 "ICMPv6:%u.%u ", 937 icmp6->icmp6_type, icmp6->icmp6_code); 938 else 939 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 940 len += snprintf(SNPARGS(proto, len), "%s", src); 941 snprintf(SNPARGS(proto, len), " %s", dst); 942 break; 943#endif 944 default: 945 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 946 args->f_id.proto, src); 947 snprintf(SNPARGS(proto, len), " %s", dst); 948 break; 949 } 950 951#ifdef INET6 952 if (args->f_id.addr_type == 6) { 953 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 954 snprintf(SNPARGS(fragment, 0), 955 " (frag %08x:%d@%d%s)", 956 args->f_id.frag_id6, 957 ntohs(ip6->ip6_plen) - hlen, 958 ntohs(offset & IP6F_OFF_MASK) << 3, 959 (offset & IP6F_MORE_FRAG) ? "+" : ""); 960 } else 961#endif 962 { 963 int ip_off, ip_len; 964 if (eh != NULL) { /* layer 2 packets are as on the wire */ 965 ip_off = ntohs(ip->ip_off); 966 ip_len = ntohs(ip->ip_len); 967 } else { 968 ip_off = ip->ip_off; 969 ip_len = ip->ip_len; 970 } 971 if (ip_off & (IP_MF | IP_OFFMASK)) 972 snprintf(SNPARGS(fragment, 0), 973 " (frag %d:%d@%d%s)", 974 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 975 offset << 3, 976 (ip_off & IP_MF) ? "+" : ""); 977 } 978 } 979 if (oif || m->m_pkthdr.rcvif) 980 log(LOG_SECURITY | LOG_INFO, 981 "ipfw: %d %s %s %s via %s%s\n", 982 f ? f->rulenum : -1, 983 action, proto, oif ? "out" : "in", 984 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 985 fragment); 986 else 987 log(LOG_SECURITY | LOG_INFO, 988 "ipfw: %d %s %s [no if info]%s\n", 989 f ? f->rulenum : -1, 990 action, proto, fragment); 991 if (limit_reached) 992 log(LOG_SECURITY | LOG_NOTICE, 993 "ipfw: limit %d reached on entry %d\n", 994 limit_reached, f ? f->rulenum : -1); 995} 996 997/* 998 * IMPORTANT: the hash function for dynamic rules must be commutative 999 * in source and destination (ip,port), because rules are bidirectional 1000 * and we want to find both in the same bucket. 1001 */ 1002static __inline int 1003hash_packet(struct ipfw_flow_id *id) 1004{ 1005 u_int32_t i; 1006 1007#ifdef INET6 1008 if (IS_IP6_FLOW_ID(id)) 1009 i = hash_packet6(id); 1010 else 1011#endif /* INET6 */ 1012 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1013 i &= (curr_dyn_buckets - 1); 1014 return i; 1015} 1016 1017/** 1018 * unlink a dynamic rule from a chain. prev is a pointer to 1019 * the previous one, q is a pointer to the rule to delete, 1020 * head is a pointer to the head of the queue. 1021 * Modifies q and potentially also head. 1022 */ 1023#define UNLINK_DYN_RULE(prev, head, q) { \ 1024 ipfw_dyn_rule *old_q = q; \ 1025 \ 1026 /* remove a refcount to the parent */ \ 1027 if (q->dyn_type == O_LIMIT) \ 1028 q->parent->count--; \ 1029 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1030 (q->id.src_ip), (q->id.src_port), \ 1031 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1032 if (prev != NULL) \ 1033 prev->next = q = q->next; \ 1034 else \ 1035 head = q = q->next; \ 1036 dyn_count--; \ 1037 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1038 1039#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1040 1041/** 1042 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1043 * 1044 * If keep_me == NULL, rules are deleted even if not expired, 1045 * otherwise only expired rules are removed. 1046 * 1047 * The value of the second parameter is also used to point to identify 1048 * a rule we absolutely do not want to remove (e.g. because we are 1049 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1050 * rules). The pointer is only used for comparison, so any non-null 1051 * value will do. 1052 */ 1053static void 1054remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1055{ 1056 static u_int32_t last_remove = 0; 1057 1058#define FORCE (keep_me == NULL) 1059 1060 ipfw_dyn_rule *prev, *q; 1061 int i, pass = 0, max_pass = 0; 1062 1063 IPFW_DYN_LOCK_ASSERT(); 1064 1065 if (ipfw_dyn_v == NULL || dyn_count == 0) 1066 return; 1067 /* do not expire more than once per second, it is useless */ 1068 if (!FORCE && last_remove == time_uptime) 1069 return; 1070 last_remove = time_uptime; 1071 1072 /* 1073 * because O_LIMIT refer to parent rules, during the first pass only 1074 * remove child and mark any pending LIMIT_PARENT, and remove 1075 * them in a second pass. 1076 */ 1077next_pass: 1078 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1079 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1080 /* 1081 * Logic can become complex here, so we split tests. 1082 */ 1083 if (q == keep_me) 1084 goto next; 1085 if (rule != NULL && rule != q->rule) 1086 goto next; /* not the one we are looking for */ 1087 if (q->dyn_type == O_LIMIT_PARENT) { 1088 /* 1089 * handle parent in the second pass, 1090 * record we need one. 1091 */ 1092 max_pass = 1; 1093 if (pass == 0) 1094 goto next; 1095 if (FORCE && q->count != 0 ) { 1096 /* XXX should not happen! */ 1097 printf("ipfw: OUCH! cannot remove rule," 1098 " count %d\n", q->count); 1099 } 1100 } else { 1101 if (!FORCE && 1102 !TIME_LEQ( q->expire, time_uptime )) 1103 goto next; 1104 } 1105 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1106 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1107 continue; 1108 } 1109next: 1110 prev=q; 1111 q=q->next; 1112 } 1113 } 1114 if (pass++ < max_pass) 1115 goto next_pass; 1116} 1117 1118 1119/** 1120 * lookup a dynamic rule. 1121 */ 1122static ipfw_dyn_rule * 1123lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1124 struct tcphdr *tcp) 1125{ 1126 /* 1127 * stateful ipfw extensions. 1128 * Lookup into dynamic session queue 1129 */ 1130#define MATCH_REVERSE 0 1131#define MATCH_FORWARD 1 1132#define MATCH_NONE 2 1133#define MATCH_UNKNOWN 3 1134 int i, dir = MATCH_NONE; 1135 ipfw_dyn_rule *prev, *q=NULL; 1136 1137 IPFW_DYN_LOCK_ASSERT(); 1138 1139 if (ipfw_dyn_v == NULL) 1140 goto done; /* not found */ 1141 i = hash_packet( pkt ); 1142 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1143 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1144 goto next; 1145 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1146 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1147 continue; 1148 } 1149 if (pkt->proto == q->id.proto && 1150 q->dyn_type != O_LIMIT_PARENT) { 1151 if (IS_IP6_FLOW_ID(pkt)) { 1152 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1153 &(q->id.src_ip6)) && 1154 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1155 &(q->id.dst_ip6)) && 1156 pkt->src_port == q->id.src_port && 1157 pkt->dst_port == q->id.dst_port ) { 1158 dir = MATCH_FORWARD; 1159 break; 1160 } 1161 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1162 &(q->id.dst_ip6)) && 1163 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1164 &(q->id.src_ip6)) && 1165 pkt->src_port == q->id.dst_port && 1166 pkt->dst_port == q->id.src_port ) { 1167 dir = MATCH_REVERSE; 1168 break; 1169 } 1170 } else { 1171 if (pkt->src_ip == q->id.src_ip && 1172 pkt->dst_ip == q->id.dst_ip && 1173 pkt->src_port == q->id.src_port && 1174 pkt->dst_port == q->id.dst_port ) { 1175 dir = MATCH_FORWARD; 1176 break; 1177 } 1178 if (pkt->src_ip == q->id.dst_ip && 1179 pkt->dst_ip == q->id.src_ip && 1180 pkt->src_port == q->id.dst_port && 1181 pkt->dst_port == q->id.src_port ) { 1182 dir = MATCH_REVERSE; 1183 break; 1184 } 1185 } 1186 } 1187next: 1188 prev = q; 1189 q = q->next; 1190 } 1191 if (q == NULL) 1192 goto done; /* q = NULL, not found */ 1193 1194 if ( prev != NULL) { /* found and not in front */ 1195 prev->next = q->next; 1196 q->next = ipfw_dyn_v[i]; 1197 ipfw_dyn_v[i] = q; 1198 } 1199 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1200 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1201 1202#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1203#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1204 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1205 switch (q->state) { 1206 case TH_SYN: /* opening */ 1207 q->expire = time_uptime + dyn_syn_lifetime; 1208 break; 1209 1210 case BOTH_SYN: /* move to established */ 1211 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1212 case BOTH_SYN | (TH_FIN << 8) : 1213 if (tcp) { 1214#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1215 u_int32_t ack = ntohl(tcp->th_ack); 1216 if (dir == MATCH_FORWARD) { 1217 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1218 q->ack_fwd = ack; 1219 else { /* ignore out-of-sequence */ 1220 break; 1221 } 1222 } else { 1223 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1224 q->ack_rev = ack; 1225 else { /* ignore out-of-sequence */ 1226 break; 1227 } 1228 } 1229 } 1230 q->expire = time_uptime + dyn_ack_lifetime; 1231 break; 1232 1233 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1234 if (dyn_fin_lifetime >= dyn_keepalive_period) 1235 dyn_fin_lifetime = dyn_keepalive_period - 1; 1236 q->expire = time_uptime + dyn_fin_lifetime; 1237 break; 1238 1239 default: 1240#if 0 1241 /* 1242 * reset or some invalid combination, but can also 1243 * occur if we use keep-state the wrong way. 1244 */ 1245 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1246 printf("invalid state: 0x%x\n", q->state); 1247#endif 1248 if (dyn_rst_lifetime >= dyn_keepalive_period) 1249 dyn_rst_lifetime = dyn_keepalive_period - 1; 1250 q->expire = time_uptime + dyn_rst_lifetime; 1251 break; 1252 } 1253 } else if (pkt->proto == IPPROTO_UDP) { 1254 q->expire = time_uptime + dyn_udp_lifetime; 1255 } else { 1256 /* other protocols */ 1257 q->expire = time_uptime + dyn_short_lifetime; 1258 } 1259done: 1260 if (match_direction) 1261 *match_direction = dir; 1262 return q; 1263} 1264 1265static ipfw_dyn_rule * 1266lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1267 struct tcphdr *tcp) 1268{ 1269 ipfw_dyn_rule *q; 1270 1271 IPFW_DYN_LOCK(); 1272 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1273 if (q == NULL) 1274 IPFW_DYN_UNLOCK(); 1275 /* NB: return table locked when q is not NULL */ 1276 return q; 1277} 1278 1279static void 1280realloc_dynamic_table(void) 1281{ 1282 IPFW_DYN_LOCK_ASSERT(); 1283 1284 /* 1285 * Try reallocation, make sure we have a power of 2 and do 1286 * not allow more than 64k entries. In case of overflow, 1287 * default to 1024. 1288 */ 1289 1290 if (dyn_buckets > 65536) 1291 dyn_buckets = 1024; 1292 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1293 dyn_buckets = curr_dyn_buckets; /* reset */ 1294 return; 1295 } 1296 curr_dyn_buckets = dyn_buckets; 1297 if (ipfw_dyn_v != NULL) 1298 free(ipfw_dyn_v, M_IPFW); 1299 for (;;) { 1300 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1301 M_IPFW, M_NOWAIT | M_ZERO); 1302 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1303 break; 1304 curr_dyn_buckets /= 2; 1305 } 1306} 1307 1308/** 1309 * Install state of type 'type' for a dynamic session. 1310 * The hash table contains two type of rules: 1311 * - regular rules (O_KEEP_STATE) 1312 * - rules for sessions with limited number of sess per user 1313 * (O_LIMIT). When they are created, the parent is 1314 * increased by 1, and decreased on delete. In this case, 1315 * the third parameter is the parent rule and not the chain. 1316 * - "parent" rules for the above (O_LIMIT_PARENT). 1317 */ 1318static ipfw_dyn_rule * 1319add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1320{ 1321 ipfw_dyn_rule *r; 1322 int i; 1323 1324 IPFW_DYN_LOCK_ASSERT(); 1325 1326 if (ipfw_dyn_v == NULL || 1327 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1328 realloc_dynamic_table(); 1329 if (ipfw_dyn_v == NULL) 1330 return NULL; /* failed ! */ 1331 } 1332 i = hash_packet(id); 1333 1334 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1335 if (r == NULL) { 1336 printf ("ipfw: sorry cannot allocate state\n"); 1337 return NULL; 1338 } 1339 1340 /* increase refcount on parent, and set pointer */ 1341 if (dyn_type == O_LIMIT) { 1342 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1343 if ( parent->dyn_type != O_LIMIT_PARENT) 1344 panic("invalid parent"); 1345 parent->count++; 1346 r->parent = parent; 1347 rule = parent->rule; 1348 } 1349 1350 r->id = *id; 1351 r->expire = time_uptime + dyn_syn_lifetime; 1352 r->rule = rule; 1353 r->dyn_type = dyn_type; 1354 r->pcnt = r->bcnt = 0; 1355 r->count = 0; 1356 1357 r->bucket = i; 1358 r->next = ipfw_dyn_v[i]; 1359 ipfw_dyn_v[i] = r; 1360 dyn_count++; 1361 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1362 dyn_type, 1363 (r->id.src_ip), (r->id.src_port), 1364 (r->id.dst_ip), (r->id.dst_port), 1365 dyn_count ); ) 1366 return r; 1367} 1368 1369/** 1370 * lookup dynamic parent rule using pkt and rule as search keys. 1371 * If the lookup fails, then install one. 1372 */ 1373static ipfw_dyn_rule * 1374lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1375{ 1376 ipfw_dyn_rule *q; 1377 int i; 1378 1379 IPFW_DYN_LOCK_ASSERT(); 1380 1381 if (ipfw_dyn_v) { 1382 int is_v6 = IS_IP6_FLOW_ID(pkt); 1383 i = hash_packet( pkt ); 1384 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1385 if (q->dyn_type == O_LIMIT_PARENT && 1386 rule== q->rule && 1387 pkt->proto == q->id.proto && 1388 pkt->src_port == q->id.src_port && 1389 pkt->dst_port == q->id.dst_port && 1390 ( 1391 (is_v6 && 1392 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1393 &(q->id.src_ip6)) && 1394 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1395 &(q->id.dst_ip6))) || 1396 (!is_v6 && 1397 pkt->src_ip == q->id.src_ip && 1398 pkt->dst_ip == q->id.dst_ip) 1399 ) 1400 ) { 1401 q->expire = time_uptime + dyn_short_lifetime; 1402 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1403 return q; 1404 } 1405 } 1406 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1407} 1408 1409/** 1410 * Install dynamic state for rule type cmd->o.opcode 1411 * 1412 * Returns 1 (failure) if state is not installed because of errors or because 1413 * session limitations are enforced. 1414 */ 1415static int 1416install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1417 struct ip_fw_args *args) 1418{ 1419 static int last_log; 1420 1421 ipfw_dyn_rule *q; 1422 1423 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 1424 cmd->o.opcode, 1425 (args->f_id.src_ip), (args->f_id.src_port), 1426 (args->f_id.dst_ip), (args->f_id.dst_port) );) 1427 1428 IPFW_DYN_LOCK(); 1429 1430 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1431 1432 if (q != NULL) { /* should never occur */ 1433 if (last_log != time_uptime) { 1434 last_log = time_uptime; 1435 printf("ipfw: install_state: entry already present, done\n"); 1436 } 1437 IPFW_DYN_UNLOCK(); 1438 return 0; 1439 } 1440 1441 if (dyn_count >= dyn_max) 1442 /* 1443 * Run out of slots, try to remove any expired rule. 1444 */ 1445 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1446 1447 if (dyn_count >= dyn_max) { 1448 if (last_log != time_uptime) { 1449 last_log = time_uptime; 1450 printf("ipfw: install_state: Too many dynamic rules\n"); 1451 } 1452 IPFW_DYN_UNLOCK(); 1453 return 1; /* cannot install, notify caller */ 1454 } 1455 1456 switch (cmd->o.opcode) { 1457 case O_KEEP_STATE: /* bidir rule */ 1458 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1459 break; 1460 1461 case O_LIMIT: /* limit number of sessions */ 1462 { 1463 u_int16_t limit_mask = cmd->limit_mask; 1464 struct ipfw_flow_id id; 1465 ipfw_dyn_rule *parent; 1466 1467 DEB(printf("ipfw: installing dyn-limit rule %d\n", 1468 cmd->conn_limit);) 1469 1470 id.dst_ip = id.src_ip = 0; 1471 id.dst_port = id.src_port = 0; 1472 id.proto = args->f_id.proto; 1473 1474 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1475 if (limit_mask & DYN_SRC_ADDR) 1476 id.src_ip6 = args->f_id.src_ip6; 1477 if (limit_mask & DYN_DST_ADDR) 1478 id.dst_ip6 = args->f_id.dst_ip6; 1479 } else { 1480 if (limit_mask & DYN_SRC_ADDR) 1481 id.src_ip = args->f_id.src_ip; 1482 if (limit_mask & DYN_DST_ADDR) 1483 id.dst_ip = args->f_id.dst_ip; 1484 } 1485 if (limit_mask & DYN_SRC_PORT) 1486 id.src_port = args->f_id.src_port; 1487 if (limit_mask & DYN_DST_PORT) 1488 id.dst_port = args->f_id.dst_port; 1489 parent = lookup_dyn_parent(&id, rule); 1490 if (parent == NULL) { 1491 printf("ipfw: add parent failed\n"); 1492 IPFW_DYN_UNLOCK(); 1493 return 1; 1494 } 1495 if (parent->count >= cmd->conn_limit) { 1496 /* 1497 * See if we can remove some expired rule. 1498 */ 1499 remove_dyn_rule(rule, parent); 1500 if (parent->count >= cmd->conn_limit) { 1501 if (fw_verbose && last_log != time_uptime) { 1502 last_log = time_uptime; 1503 log(LOG_SECURITY | LOG_DEBUG, 1504 "drop session, too many entries\n"); 1505 } 1506 IPFW_DYN_UNLOCK(); 1507 return 1; 1508 } 1509 } 1510 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1511 } 1512 break; 1513 default: 1514 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 1515 IPFW_DYN_UNLOCK(); 1516 return 1; 1517 } 1518 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1519 IPFW_DYN_UNLOCK(); 1520 return 0; 1521} 1522 1523/* 1524 * Generate a TCP packet, containing either a RST or a keepalive. 1525 * When flags & TH_RST, we are sending a RST packet, because of a 1526 * "reset" action matched the packet. 1527 * Otherwise we are sending a keepalive, and flags & TH_ 1528 */ 1529static struct mbuf * 1530send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1531{ 1532 struct mbuf *m; 1533 struct ip *ip; 1534 struct tcphdr *tcp; 1535 1536 MGETHDR(m, M_DONTWAIT, MT_DATA); 1537 if (m == 0) 1538 return (NULL); 1539 m->m_pkthdr.rcvif = (struct ifnet *)0; 1540 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1541 m->m_data += max_linkhdr; 1542 1543 ip = mtod(m, struct ip *); 1544 bzero(ip, m->m_len); 1545 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1546 ip->ip_p = IPPROTO_TCP; 1547 tcp->th_off = 5; 1548 /* 1549 * Assume we are sending a RST (or a keepalive in the reverse 1550 * direction), swap src and destination addresses and ports. 1551 */ 1552 ip->ip_src.s_addr = htonl(id->dst_ip); 1553 ip->ip_dst.s_addr = htonl(id->src_ip); 1554 tcp->th_sport = htons(id->dst_port); 1555 tcp->th_dport = htons(id->src_port); 1556 if (flags & TH_RST) { /* we are sending a RST */ 1557 if (flags & TH_ACK) { 1558 tcp->th_seq = htonl(ack); 1559 tcp->th_ack = htonl(0); 1560 tcp->th_flags = TH_RST; 1561 } else { 1562 if (flags & TH_SYN) 1563 seq++; 1564 tcp->th_seq = htonl(0); 1565 tcp->th_ack = htonl(seq); 1566 tcp->th_flags = TH_RST | TH_ACK; 1567 } 1568 } else { 1569 /* 1570 * We are sending a keepalive. flags & TH_SYN determines 1571 * the direction, forward if set, reverse if clear. 1572 * NOTE: seq and ack are always assumed to be correct 1573 * as set by the caller. This may be confusing... 1574 */ 1575 if (flags & TH_SYN) { 1576 /* 1577 * we have to rewrite the correct addresses! 1578 */ 1579 ip->ip_dst.s_addr = htonl(id->dst_ip); 1580 ip->ip_src.s_addr = htonl(id->src_ip); 1581 tcp->th_dport = htons(id->dst_port); 1582 tcp->th_sport = htons(id->src_port); 1583 } 1584 tcp->th_seq = htonl(seq); 1585 tcp->th_ack = htonl(ack); 1586 tcp->th_flags = TH_ACK; 1587 } 1588 /* 1589 * set ip_len to the payload size so we can compute 1590 * the tcp checksum on the pseudoheader 1591 * XXX check this, could save a couple of words ? 1592 */ 1593 ip->ip_len = htons(sizeof(struct tcphdr)); 1594 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1595 /* 1596 * now fill fields left out earlier 1597 */ 1598 ip->ip_ttl = ip_defttl; 1599 ip->ip_len = m->m_pkthdr.len; 1600 m->m_flags |= M_SKIP_FIREWALL; 1601 return (m); 1602} 1603 1604/* 1605 * sends a reject message, consuming the mbuf passed as an argument. 1606 */ 1607static void 1608send_reject(struct ip_fw_args *args, int code, u_short offset, int ip_len) 1609{ 1610 1611 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1612 /* We need the IP header in host order for icmp_error(). */ 1613 if (args->eh != NULL) { 1614 struct ip *ip = mtod(args->m, struct ip *); 1615 ip->ip_len = ntohs(ip->ip_len); 1616 ip->ip_off = ntohs(ip->ip_off); 1617 } 1618 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1619 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1620 struct tcphdr *const tcp = 1621 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1622 if ( (tcp->th_flags & TH_RST) == 0) { 1623 struct mbuf *m; 1624 m = send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1625 ntohl(tcp->th_ack), 1626 tcp->th_flags | TH_RST); 1627 if (m != NULL) 1628 ip_output(m, NULL, NULL, 0, NULL, NULL); 1629 } 1630 m_freem(args->m); 1631 } else 1632 m_freem(args->m); 1633 args->m = NULL; 1634} 1635 1636/** 1637 * 1638 * Given an ip_fw *, lookup_next_rule will return a pointer 1639 * to the next rule, which can be either the jump 1640 * target (for skipto instructions) or the next one in the list (in 1641 * all other cases including a missing jump target). 1642 * The result is also written in the "next_rule" field of the rule. 1643 * Backward jumps are not allowed, so start looking from the next 1644 * rule... 1645 * 1646 * This never returns NULL -- in case we do not have an exact match, 1647 * the next rule is returned. When the ruleset is changed, 1648 * pointers are flushed so we are always correct. 1649 */ 1650 1651static struct ip_fw * 1652lookup_next_rule(struct ip_fw *me) 1653{ 1654 struct ip_fw *rule = NULL; 1655 ipfw_insn *cmd; 1656 1657 /* look for action, in case it is a skipto */ 1658 cmd = ACTION_PTR(me); 1659 if (cmd->opcode == O_LOG) 1660 cmd += F_LEN(cmd); 1661 if (cmd->opcode == O_ALTQ) 1662 cmd += F_LEN(cmd); 1663 if ( cmd->opcode == O_SKIPTO ) 1664 for (rule = me->next; rule ; rule = rule->next) 1665 if (rule->rulenum >= cmd->arg1) 1666 break; 1667 if (rule == NULL) /* failure or not a skipto */ 1668 rule = me->next; 1669 me->next_rule = rule; 1670 return rule; 1671} 1672 1673static int 1674add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1675 uint8_t mlen, uint32_t value) 1676{ 1677 struct radix_node_head *rnh; 1678 struct table_entry *ent; 1679 1680 if (tbl >= IPFW_TABLES_MAX) 1681 return (EINVAL); 1682 rnh = ch->tables[tbl]; 1683 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1684 if (ent == NULL) 1685 return (ENOMEM); 1686 ent->value = value; 1687 ent->addr.sin_len = ent->mask.sin_len = 8; 1688 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1689 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1690 IPFW_WLOCK(&layer3_chain); 1691 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1692 NULL) { 1693 IPFW_WUNLOCK(&layer3_chain); 1694 free(ent, M_IPFW_TBL); 1695 return (EEXIST); 1696 } 1697 IPFW_WUNLOCK(&layer3_chain); 1698 return (0); 1699} 1700 1701static int 1702del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1703 uint8_t mlen) 1704{ 1705 struct radix_node_head *rnh; 1706 struct table_entry *ent; 1707 struct sockaddr_in sa, mask; 1708 1709 if (tbl >= IPFW_TABLES_MAX) 1710 return (EINVAL); 1711 rnh = ch->tables[tbl]; 1712 sa.sin_len = mask.sin_len = 8; 1713 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1714 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1715 IPFW_WLOCK(ch); 1716 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1717 if (ent == NULL) { 1718 IPFW_WUNLOCK(ch); 1719 return (ESRCH); 1720 } 1721 IPFW_WUNLOCK(ch); 1722 free(ent, M_IPFW_TBL); 1723 return (0); 1724} 1725 1726static int 1727flush_table_entry(struct radix_node *rn, void *arg) 1728{ 1729 struct radix_node_head * const rnh = arg; 1730 struct table_entry *ent; 1731 1732 ent = (struct table_entry *) 1733 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1734 if (ent != NULL) 1735 free(ent, M_IPFW_TBL); 1736 return (0); 1737} 1738 1739static int 1740flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1741{ 1742 struct radix_node_head *rnh; 1743 1744 IPFW_WLOCK_ASSERT(ch); 1745 1746 if (tbl >= IPFW_TABLES_MAX) 1747 return (EINVAL); 1748 rnh = ch->tables[tbl]; 1749 KASSERT(rnh != NULL, ("NULL IPFW table")); 1750 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1751 return (0); 1752} 1753 1754static void 1755flush_tables(struct ip_fw_chain *ch) 1756{ 1757 uint16_t tbl; 1758 1759 IPFW_WLOCK_ASSERT(ch); 1760 1761 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1762 flush_table(ch, tbl); 1763} 1764 1765static int 1766init_tables(struct ip_fw_chain *ch) 1767{ 1768 int i; 1769 uint16_t j; 1770 1771 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1772 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1773 for (j = 0; j < i; j++) { 1774 (void) flush_table(ch, j); 1775 } 1776 return (ENOMEM); 1777 } 1778 } 1779 return (0); 1780} 1781 1782static int 1783lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1784 uint32_t *val) 1785{ 1786 struct radix_node_head *rnh; 1787 struct table_entry *ent; 1788 struct sockaddr_in sa; 1789 1790 if (tbl >= IPFW_TABLES_MAX) 1791 return (0); 1792 rnh = ch->tables[tbl]; 1793 sa.sin_len = 8; 1794 sa.sin_addr.s_addr = addr; 1795 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1796 if (ent != NULL) { 1797 *val = ent->value; 1798 return (1); 1799 } 1800 return (0); 1801} 1802 1803static int 1804count_table_entry(struct radix_node *rn, void *arg) 1805{ 1806 u_int32_t * const cnt = arg; 1807 1808 (*cnt)++; 1809 return (0); 1810} 1811 1812static int 1813count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1814{ 1815 struct radix_node_head *rnh; 1816 1817 if (tbl >= IPFW_TABLES_MAX) 1818 return (EINVAL); 1819 rnh = ch->tables[tbl]; 1820 *cnt = 0; 1821 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1822 return (0); 1823} 1824 1825static int 1826dump_table_entry(struct radix_node *rn, void *arg) 1827{ 1828 struct table_entry * const n = (struct table_entry *)rn; 1829 ipfw_table * const tbl = arg; 1830 ipfw_table_entry *ent; 1831 1832 if (tbl->cnt == tbl->size) 1833 return (1); 1834 ent = &tbl->ent[tbl->cnt]; 1835 ent->tbl = tbl->tbl; 1836 if (in_nullhost(n->mask.sin_addr)) 1837 ent->masklen = 0; 1838 else 1839 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1840 ent->addr = n->addr.sin_addr.s_addr; 1841 ent->value = n->value; 1842 tbl->cnt++; 1843 return (0); 1844} 1845 1846static int 1847dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1848{ 1849 struct radix_node_head *rnh; 1850 1851 if (tbl->tbl >= IPFW_TABLES_MAX) 1852 return (EINVAL); 1853 rnh = ch->tables[tbl->tbl]; 1854 tbl->cnt = 0; 1855 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1856 return (0); 1857} 1858 1859static void 1860fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1861{ 1862 struct ucred *cr; 1863 1864 if (inp->inp_socket != NULL) { 1865 cr = inp->inp_socket->so_cred; 1866 ugp->fw_prid = jailed(cr) ? 1867 cr->cr_prison->pr_id : -1; 1868 ugp->fw_uid = cr->cr_uid; 1869 ugp->fw_ngroups = cr->cr_ngroups; 1870 bcopy(cr->cr_groups, ugp->fw_groups, 1871 sizeof(ugp->fw_groups)); 1872 } 1873} 1874 1875static int 1876check_uidgid(ipfw_insn_u32 *insn, 1877 int proto, struct ifnet *oif, 1878 struct in_addr dst_ip, u_int16_t dst_port, 1879 struct in_addr src_ip, u_int16_t src_port, 1880 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1881{ 1882 struct inpcbinfo *pi; 1883 int wildcard; 1884 struct inpcb *pcb; 1885 int match; 1886 gid_t *gp; 1887 1888 /* 1889 * Check to see if the UDP or TCP stack supplied us with 1890 * the PCB. If so, rather then holding a lock and looking 1891 * up the PCB, we can use the one that was supplied. 1892 */ 1893 if (inp && *lookup == 0) { 1894 INP_LOCK_ASSERT(inp); 1895 if (inp->inp_socket != NULL) { 1896 fill_ugid_cache(inp, ugp); 1897 *lookup = 1; 1898 } 1899 } 1900 /* 1901 * If we have already been here and the packet has no 1902 * PCB entry associated with it, then we can safely 1903 * assume that this is a no match. 1904 */ 1905 if (*lookup == -1) 1906 return (0); 1907 if (proto == IPPROTO_TCP) { 1908 wildcard = 0; 1909 pi = &tcbinfo; 1910 } else if (proto == IPPROTO_UDP) { 1911 wildcard = 1; 1912 pi = &udbinfo; 1913 } else 1914 return 0; 1915 match = 0; 1916 if (*lookup == 0) { 1917 INP_INFO_RLOCK(pi); 1918 pcb = (oif) ? 1919 in_pcblookup_hash(pi, 1920 dst_ip, htons(dst_port), 1921 src_ip, htons(src_port), 1922 wildcard, oif) : 1923 in_pcblookup_hash(pi, 1924 src_ip, htons(src_port), 1925 dst_ip, htons(dst_port), 1926 wildcard, NULL); 1927 if (pcb != NULL) { 1928 INP_LOCK(pcb); 1929 if (pcb->inp_socket != NULL) { 1930 fill_ugid_cache(pcb, ugp); 1931 *lookup = 1; 1932 } 1933 INP_UNLOCK(pcb); 1934 } 1935 INP_INFO_RUNLOCK(pi); 1936 if (*lookup == 0) { 1937 /* 1938 * If the lookup did not yield any results, there 1939 * is no sense in coming back and trying again. So 1940 * we can set lookup to -1 and ensure that we wont 1941 * bother the pcb system again. 1942 */ 1943 *lookup = -1; 1944 return (0); 1945 } 1946 } 1947 if (insn->o.opcode == O_UID) 1948 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1949 else if (insn->o.opcode == O_GID) { 1950 for (gp = ugp->fw_groups; 1951 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1952 if (*gp == (gid_t)insn->d[0]) { 1953 match = 1; 1954 break; 1955 } 1956 } else if (insn->o.opcode == O_JAIL) 1957 match = (ugp->fw_prid == (int)insn->d[0]); 1958 return match; 1959} 1960 1961/* 1962 * The main check routine for the firewall. 1963 * 1964 * All arguments are in args so we can modify them and return them 1965 * back to the caller. 1966 * 1967 * Parameters: 1968 * 1969 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1970 * Starts with the IP header. 1971 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1972 * args->oif Outgoing interface, or NULL if packet is incoming. 1973 * The incoming interface is in the mbuf. (in) 1974 * args->divert_rule (in/out) 1975 * Skip up to the first rule past this rule number; 1976 * upon return, non-zero port number for divert or tee. 1977 * 1978 * args->rule Pointer to the last matching rule (in/out) 1979 * args->next_hop Socket we are forwarding to (out). 1980 * args->f_id Addresses grabbed from the packet (out) 1981 * args->cookie a cookie depending on rule action 1982 * 1983 * Return value: 1984 * 1985 * IP_FW_PASS the packet must be accepted 1986 * IP_FW_DENY the packet must be dropped 1987 * IP_FW_DIVERT divert packet, port in m_tag 1988 * IP_FW_TEE tee packet, port in m_tag 1989 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 1990 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 1991 * 1992 */ 1993 1994int 1995ipfw_chk(struct ip_fw_args *args) 1996{ 1997 /* 1998 * Local variables hold state during the processing of a packet. 1999 * 2000 * IMPORTANT NOTE: to speed up the processing of rules, there 2001 * are some assumption on the values of the variables, which 2002 * are documented here. Should you change them, please check 2003 * the implementation of the various instructions to make sure 2004 * that they still work. 2005 * 2006 * args->eh The MAC header. It is non-null for a layer2 2007 * packet, it is NULL for a layer-3 packet. 2008 * 2009 * m | args->m Pointer to the mbuf, as received from the caller. 2010 * It may change if ipfw_chk() does an m_pullup, or if it 2011 * consumes the packet because it calls send_reject(). 2012 * XXX This has to change, so that ipfw_chk() never modifies 2013 * or consumes the buffer. 2014 * ip is simply an alias of the value of m, and it is kept 2015 * in sync with it (the packet is supposed to start with 2016 * the ip header). 2017 */ 2018 struct mbuf *m = args->m; 2019 struct ip *ip = mtod(m, struct ip *); 2020 2021 /* 2022 * For rules which contain uid/gid or jail constraints, cache 2023 * a copy of the users credentials after the pcb lookup has been 2024 * executed. This will speed up the processing of rules with 2025 * these types of constraints, as well as decrease contention 2026 * on pcb related locks. 2027 */ 2028 struct ip_fw_ugid fw_ugid_cache; 2029 int ugid_lookup = 0; 2030 2031 /* 2032 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2033 * associated with a packet input on a divert socket. This 2034 * will allow to distinguish traffic and its direction when 2035 * it originates from a divert socket. 2036 */ 2037 u_int divinput_flags = 0; 2038 2039 /* 2040 * oif | args->oif If NULL, ipfw_chk has been called on the 2041 * inbound path (ether_input, ip_input). 2042 * If non-NULL, ipfw_chk has been called on the outbound path 2043 * (ether_output, ip_output). 2044 */ 2045 struct ifnet *oif = args->oif; 2046 2047 struct ip_fw *f = NULL; /* matching rule */ 2048 int retval = 0; 2049 2050 /* 2051 * hlen The length of the IP header. 2052 */ 2053 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2054 2055 /* 2056 * offset The offset of a fragment. offset != 0 means that 2057 * we have a fragment at this offset of an IPv4 packet. 2058 * offset == 0 means that (if this is an IPv4 packet) 2059 * this is the first or only fragment. 2060 * For IPv6 offset == 0 means there is no Fragment Header. 2061 * If offset != 0 for IPv6 always use correct mask to 2062 * get the correct offset because we add IP6F_MORE_FRAG 2063 * to be able to dectect the first fragment which would 2064 * otherwise have offset = 0. 2065 */ 2066 u_short offset = 0; 2067 2068 /* 2069 * Local copies of addresses. They are only valid if we have 2070 * an IP packet. 2071 * 2072 * proto The protocol. Set to 0 for non-ip packets, 2073 * or to the protocol read from the packet otherwise. 2074 * proto != 0 means that we have an IPv4 packet. 2075 * 2076 * src_port, dst_port port numbers, in HOST format. Only 2077 * valid for TCP and UDP packets. 2078 * 2079 * src_ip, dst_ip ip addresses, in NETWORK format. 2080 * Only valid for IPv4 packets. 2081 */ 2082 u_int8_t proto; 2083 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2084 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2085 u_int16_t ip_len=0; 2086 int pktlen; 2087 2088 /* 2089 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2090 * MATCH_NONE when checked and not matched (q = NULL), 2091 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2092 */ 2093 int dyn_dir = MATCH_UNKNOWN; 2094 ipfw_dyn_rule *q = NULL; 2095 struct ip_fw_chain *chain = &layer3_chain; 2096 struct m_tag *mtag; 2097 2098 /* 2099 * We store in ulp a pointer to the upper layer protocol header. 2100 * In the ipv4 case this is easy to determine from the header, 2101 * but for ipv6 we might have some additional headers in the middle. 2102 * ulp is NULL if not found. 2103 */ 2104 void *ulp = NULL; /* upper layer protocol pointer. */ 2105 /* XXX ipv6 variables */ 2106 int is_ipv6 = 0; 2107 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2108 /* end of ipv6 variables */ 2109 int is_ipv4 = 0; 2110 2111 if (m->m_flags & M_SKIP_FIREWALL) 2112 return (IP_FW_PASS); /* accept */ 2113 2114 pktlen = m->m_pkthdr.len; 2115 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2116 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2117 2118/* 2119 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2120 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2121 * pointer might become stale after other pullups (but we never use it 2122 * this way). 2123 */ 2124#define PULLUP_TO(len, p, T) \ 2125do { \ 2126 int x = (len) + sizeof(T); \ 2127 if ((m)->m_len < x) { \ 2128 args->m = m = m_pullup(m, x); \ 2129 if (m == NULL) \ 2130 goto pullup_failed; \ 2131 } \ 2132 p = (mtod(m, char *) + (len)); \ 2133} while (0) 2134 2135 /* Identify IP packets and fill up variables. */ 2136 if (pktlen >= sizeof(struct ip6_hdr) && 2137 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2138 mtod(m, struct ip *)->ip_v == 6) { 2139 is_ipv6 = 1; 2140 args->f_id.addr_type = 6; 2141 hlen = sizeof(struct ip6_hdr); 2142 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2143 2144 /* Search extension headers to find upper layer protocols */ 2145 while (ulp == NULL) { 2146 switch (proto) { 2147 case IPPROTO_ICMPV6: 2148 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2149 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2150 break; 2151 2152 case IPPROTO_TCP: 2153 PULLUP_TO(hlen, ulp, struct tcphdr); 2154 dst_port = TCP(ulp)->th_dport; 2155 src_port = TCP(ulp)->th_sport; 2156 args->f_id.flags = TCP(ulp)->th_flags; 2157 break; 2158 2159 case IPPROTO_UDP: 2160 PULLUP_TO(hlen, ulp, struct udphdr); 2161 dst_port = UDP(ulp)->uh_dport; 2162 src_port = UDP(ulp)->uh_sport; 2163 break; 2164 2165 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2166 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2167 ext_hd |= EXT_HOPOPTS; 2168 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2169 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2170 ulp = NULL; 2171 break; 2172 2173 case IPPROTO_ROUTING: /* RFC 2460 */ 2174 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2175 if (((struct ip6_rthdr *)ulp)->ip6r_type != 0) { 2176 printf("IPFW2: IPV6 - Unknown Routing " 2177 "Header type(%d)\n", 2178 ((struct ip6_rthdr *)ulp)->ip6r_type); 2179 if (fw_deny_unknown_exthdrs) 2180 return (IP_FW_DENY); 2181 break; 2182 } 2183 ext_hd |= EXT_ROUTING; 2184 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2185 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2186 ulp = NULL; 2187 break; 2188 2189 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2190 PULLUP_TO(hlen, ulp, struct ip6_frag); 2191 ext_hd |= EXT_FRAGMENT; 2192 hlen += sizeof (struct ip6_frag); 2193 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2194 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2195 IP6F_OFF_MASK; 2196 /* Add IP6F_MORE_FRAG for offset of first 2197 * fragment to be != 0. */ 2198 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2199 IP6F_MORE_FRAG; 2200 if (offset == 0) { 2201 printf("IPFW2: IPV6 - Invalid Fragment " 2202 "Header\n"); 2203 if (fw_deny_unknown_exthdrs) 2204 return (IP_FW_DENY); 2205 break; 2206 } 2207 args->f_id.frag_id6 = 2208 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2209 ulp = NULL; 2210 break; 2211 2212 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2213 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2214 ext_hd |= EXT_DSTOPTS; 2215 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2216 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2217 ulp = NULL; 2218 break; 2219 2220 case IPPROTO_AH: /* RFC 2402 */ 2221 PULLUP_TO(hlen, ulp, struct ip6_ext); 2222 ext_hd |= EXT_AH; 2223 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2224 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2225 ulp = NULL; 2226 break; 2227 2228 case IPPROTO_ESP: /* RFC 2406 */ 2229 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2230 /* Anything past Seq# is variable length and 2231 * data past this ext. header is encrypted. */ 2232 ext_hd |= EXT_ESP; 2233 break; 2234 2235 case IPPROTO_NONE: /* RFC 2460 */ 2236 PULLUP_TO(hlen, ulp, struct ip6_ext); 2237 /* Packet ends here. if ip6e_len!=0 octets 2238 * must be ignored. */ 2239 break; 2240 2241 case IPPROTO_OSPFIGP: 2242 /* XXX OSPF header check? */ 2243 PULLUP_TO(hlen, ulp, struct ip6_ext); 2244 break; 2245 2246 default: 2247 printf("IPFW2: IPV6 - Unknown Extension " 2248 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2249 if (fw_deny_unknown_exthdrs) 2250 return (IP_FW_DENY); 2251 break; 2252 } /*switch */ 2253 } 2254 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2255 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2256 args->f_id.src_ip = 0; 2257 args->f_id.dst_ip = 0; 2258 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2259 } else if (pktlen >= sizeof(struct ip) && 2260 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2261 mtod(m, struct ip *)->ip_v == 4) { 2262 is_ipv4 = 1; 2263 ip = mtod(m, struct ip *); 2264 hlen = ip->ip_hl << 2; 2265 args->f_id.addr_type = 4; 2266 2267 /* 2268 * Collect parameters into local variables for faster matching. 2269 */ 2270 proto = ip->ip_p; 2271 src_ip = ip->ip_src; 2272 dst_ip = ip->ip_dst; 2273 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2274 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2275 ip_len = ntohs(ip->ip_len); 2276 } else { 2277 offset = ip->ip_off & IP_OFFMASK; 2278 ip_len = ip->ip_len; 2279 } 2280 pktlen = ip_len < pktlen ? ip_len : pktlen; 2281 2282 if (offset == 0) { 2283 switch (proto) { 2284 case IPPROTO_TCP: 2285 PULLUP_TO(hlen, ulp, struct tcphdr); 2286 dst_port = TCP(ulp)->th_dport; 2287 src_port = TCP(ulp)->th_sport; 2288 args->f_id.flags = TCP(ulp)->th_flags; 2289 break; 2290 2291 case IPPROTO_UDP: 2292 PULLUP_TO(hlen, ulp, struct udphdr); 2293 dst_port = UDP(ulp)->uh_dport; 2294 src_port = UDP(ulp)->uh_sport; 2295 break; 2296 2297 case IPPROTO_ICMP: 2298 PULLUP_TO(hlen, ulp, struct icmphdr); 2299 args->f_id.flags = ICMP(ulp)->icmp_type; 2300 break; 2301 2302 default: 2303 break; 2304 } 2305 } 2306 2307 args->f_id.src_ip = ntohl(src_ip.s_addr); 2308 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2309 } 2310#undef PULLUP_TO 2311 if (proto) { /* we may have port numbers, store them */ 2312 args->f_id.proto = proto; 2313 args->f_id.src_port = src_port = ntohs(src_port); 2314 args->f_id.dst_port = dst_port = ntohs(dst_port); 2315 } 2316 2317 IPFW_RLOCK(chain); 2318 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2319 if (args->rule) { 2320 /* 2321 * Packet has already been tagged. Look for the next rule 2322 * to restart processing. 2323 * 2324 * If fw_one_pass != 0 then just accept it. 2325 * XXX should not happen here, but optimized out in 2326 * the caller. 2327 */ 2328 if (fw_one_pass) { 2329 IPFW_RUNLOCK(chain); 2330 return (IP_FW_PASS); 2331 } 2332 2333 f = args->rule->next_rule; 2334 if (f == NULL) 2335 f = lookup_next_rule(args->rule); 2336 } else { 2337 /* 2338 * Find the starting rule. It can be either the first 2339 * one, or the one after divert_rule if asked so. 2340 */ 2341 int skipto = mtag ? divert_cookie(mtag) : 0; 2342 2343 f = chain->rules; 2344 if (args->eh == NULL && skipto != 0) { 2345 if (skipto >= IPFW_DEFAULT_RULE) { 2346 IPFW_RUNLOCK(chain); 2347 return (IP_FW_DENY); /* invalid */ 2348 } 2349 while (f && f->rulenum <= skipto) 2350 f = f->next; 2351 if (f == NULL) { /* drop packet */ 2352 IPFW_RUNLOCK(chain); 2353 return (IP_FW_DENY); 2354 } 2355 } 2356 } 2357 /* reset divert rule to avoid confusion later */ 2358 if (mtag) { 2359 divinput_flags = divert_info(mtag) & 2360 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2361 m_tag_delete(m, mtag); 2362 } 2363 2364 /* 2365 * Now scan the rules, and parse microinstructions for each rule. 2366 */ 2367 for (; f; f = f->next) { 2368 ipfw_insn *cmd; 2369 uint32_t tablearg = 0; 2370 int l, cmdlen, skip_or; /* skip rest of OR block */ 2371 2372again: 2373 if (set_disable & (1 << f->set) ) 2374 continue; 2375 2376 skip_or = 0; 2377 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2378 l -= cmdlen, cmd += cmdlen) { 2379 int match; 2380 2381 /* 2382 * check_body is a jump target used when we find a 2383 * CHECK_STATE, and need to jump to the body of 2384 * the target rule. 2385 */ 2386 2387check_body: 2388 cmdlen = F_LEN(cmd); 2389 /* 2390 * An OR block (insn_1 || .. || insn_n) has the 2391 * F_OR bit set in all but the last instruction. 2392 * The first match will set "skip_or", and cause 2393 * the following instructions to be skipped until 2394 * past the one with the F_OR bit clear. 2395 */ 2396 if (skip_or) { /* skip this instruction */ 2397 if ((cmd->len & F_OR) == 0) 2398 skip_or = 0; /* next one is good */ 2399 continue; 2400 } 2401 match = 0; /* set to 1 if we succeed */ 2402 2403 switch (cmd->opcode) { 2404 /* 2405 * The first set of opcodes compares the packet's 2406 * fields with some pattern, setting 'match' if a 2407 * match is found. At the end of the loop there is 2408 * logic to deal with F_NOT and F_OR flags associated 2409 * with the opcode. 2410 */ 2411 case O_NOP: 2412 match = 1; 2413 break; 2414 2415 case O_FORWARD_MAC: 2416 printf("ipfw: opcode %d unimplemented\n", 2417 cmd->opcode); 2418 break; 2419 2420 case O_GID: 2421 case O_UID: 2422 case O_JAIL: 2423 /* 2424 * We only check offset == 0 && proto != 0, 2425 * as this ensures that we have a 2426 * packet with the ports info. 2427 */ 2428 if (offset!=0) 2429 break; 2430 if (is_ipv6) /* XXX to be fixed later */ 2431 break; 2432 if (proto == IPPROTO_TCP || 2433 proto == IPPROTO_UDP) 2434 match = check_uidgid( 2435 (ipfw_insn_u32 *)cmd, 2436 proto, oif, 2437 dst_ip, dst_port, 2438 src_ip, src_port, &fw_ugid_cache, 2439 &ugid_lookup, args->inp); 2440 break; 2441 2442 case O_RECV: 2443 match = iface_match(m->m_pkthdr.rcvif, 2444 (ipfw_insn_if *)cmd); 2445 break; 2446 2447 case O_XMIT: 2448 match = iface_match(oif, (ipfw_insn_if *)cmd); 2449 break; 2450 2451 case O_VIA: 2452 match = iface_match(oif ? oif : 2453 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2454 break; 2455 2456 case O_MACADDR2: 2457 if (args->eh != NULL) { /* have MAC header */ 2458 u_int32_t *want = (u_int32_t *) 2459 ((ipfw_insn_mac *)cmd)->addr; 2460 u_int32_t *mask = (u_int32_t *) 2461 ((ipfw_insn_mac *)cmd)->mask; 2462 u_int32_t *hdr = (u_int32_t *)args->eh; 2463 2464 match = 2465 ( want[0] == (hdr[0] & mask[0]) && 2466 want[1] == (hdr[1] & mask[1]) && 2467 want[2] == (hdr[2] & mask[2]) ); 2468 } 2469 break; 2470 2471 case O_MAC_TYPE: 2472 if (args->eh != NULL) { 2473 u_int16_t t = 2474 ntohs(args->eh->ether_type); 2475 u_int16_t *p = 2476 ((ipfw_insn_u16 *)cmd)->ports; 2477 int i; 2478 2479 for (i = cmdlen - 1; !match && i>0; 2480 i--, p += 2) 2481 match = (t>=p[0] && t<=p[1]); 2482 } 2483 break; 2484 2485 case O_FRAG: 2486 match = (offset != 0); 2487 break; 2488 2489 case O_IN: /* "out" is "not in" */ 2490 match = (oif == NULL); 2491 break; 2492 2493 case O_LAYER2: 2494 match = (args->eh != NULL); 2495 break; 2496 2497 case O_DIVERTED: 2498 match = (cmd->arg1 & 1 && divinput_flags & 2499 IP_FW_DIVERT_LOOPBACK_FLAG) || 2500 (cmd->arg1 & 2 && divinput_flags & 2501 IP_FW_DIVERT_OUTPUT_FLAG); 2502 break; 2503 2504 case O_PROTO: 2505 /* 2506 * We do not allow an arg of 0 so the 2507 * check of "proto" only suffices. 2508 */ 2509 match = (proto == cmd->arg1); 2510 break; 2511 2512 case O_IP_SRC: 2513 match = is_ipv4 && 2514 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2515 src_ip.s_addr); 2516 break; 2517 2518 case O_IP_SRC_LOOKUP: 2519 case O_IP_DST_LOOKUP: 2520 if (is_ipv4) { 2521 uint32_t a = 2522 (cmd->opcode == O_IP_DST_LOOKUP) ? 2523 dst_ip.s_addr : src_ip.s_addr; 2524 uint32_t v; 2525 2526 match = lookup_table(chain, cmd->arg1, a, 2527 &v); 2528 if (!match) 2529 break; 2530 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2531 match = 2532 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2533 else 2534 tablearg = v; 2535 } 2536 break; 2537 2538 case O_IP_SRC_MASK: 2539 case O_IP_DST_MASK: 2540 if (is_ipv4) { 2541 uint32_t a = 2542 (cmd->opcode == O_IP_DST_MASK) ? 2543 dst_ip.s_addr : src_ip.s_addr; 2544 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2545 int i = cmdlen-1; 2546 2547 for (; !match && i>0; i-= 2, p+= 2) 2548 match = (p[0] == (a & p[1])); 2549 } 2550 break; 2551 2552 case O_IP_SRC_ME: 2553 if (is_ipv4) { 2554 struct ifnet *tif; 2555 2556 INADDR_TO_IFP(src_ip, tif); 2557 match = (tif != NULL); 2558 } 2559 break; 2560 2561 case O_IP_DST_SET: 2562 case O_IP_SRC_SET: 2563 if (is_ipv4) { 2564 u_int32_t *d = (u_int32_t *)(cmd+1); 2565 u_int32_t addr = 2566 cmd->opcode == O_IP_DST_SET ? 2567 args->f_id.dst_ip : 2568 args->f_id.src_ip; 2569 2570 if (addr < d[0]) 2571 break; 2572 addr -= d[0]; /* subtract base */ 2573 match = (addr < cmd->arg1) && 2574 ( d[ 1 + (addr>>5)] & 2575 (1<<(addr & 0x1f)) ); 2576 } 2577 break; 2578 2579 case O_IP_DST: 2580 match = is_ipv4 && 2581 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2582 dst_ip.s_addr); 2583 break; 2584 2585 case O_IP_DST_ME: 2586 if (is_ipv4) { 2587 struct ifnet *tif; 2588 2589 INADDR_TO_IFP(dst_ip, tif); 2590 match = (tif != NULL); 2591 } 2592 break; 2593 2594 case O_IP_SRCPORT: 2595 case O_IP_DSTPORT: 2596 /* 2597 * offset == 0 && proto != 0 is enough 2598 * to guarantee that we have a 2599 * packet with port info. 2600 */ 2601 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2602 && offset == 0) { 2603 u_int16_t x = 2604 (cmd->opcode == O_IP_SRCPORT) ? 2605 src_port : dst_port ; 2606 u_int16_t *p = 2607 ((ipfw_insn_u16 *)cmd)->ports; 2608 int i; 2609 2610 for (i = cmdlen - 1; !match && i>0; 2611 i--, p += 2) 2612 match = (x>=p[0] && x<=p[1]); 2613 } 2614 break; 2615 2616 case O_ICMPTYPE: 2617 match = (offset == 0 && proto==IPPROTO_ICMP && 2618 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2619 break; 2620 2621#ifdef INET6 2622 case O_ICMP6TYPE: 2623 match = is_ipv6 && offset == 0 && 2624 proto==IPPROTO_ICMPV6 && 2625 icmp6type_match( 2626 ICMP6(ulp)->icmp6_type, 2627 (ipfw_insn_u32 *)cmd); 2628 break; 2629#endif /* INET6 */ 2630 2631 case O_IPOPT: 2632 match = (is_ipv4 && 2633 ipopts_match(mtod(m, struct ip *), cmd) ); 2634 break; 2635 2636 case O_IPVER: 2637 match = (is_ipv4 && 2638 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2639 break; 2640 2641 case O_IPID: 2642 case O_IPLEN: 2643 case O_IPTTL: 2644 if (is_ipv4) { /* only for IP packets */ 2645 uint16_t x; 2646 uint16_t *p; 2647 int i; 2648 2649 if (cmd->opcode == O_IPLEN) 2650 x = ip_len; 2651 else if (cmd->opcode == O_IPTTL) 2652 x = mtod(m, struct ip *)->ip_ttl; 2653 else /* must be IPID */ 2654 x = ntohs(mtod(m, struct ip *)->ip_id); 2655 if (cmdlen == 1) { 2656 match = (cmd->arg1 == x); 2657 break; 2658 } 2659 /* otherwise we have ranges */ 2660 p = ((ipfw_insn_u16 *)cmd)->ports; 2661 i = cmdlen - 1; 2662 for (; !match && i>0; i--, p += 2) 2663 match = (x >= p[0] && x <= p[1]); 2664 } 2665 break; 2666 2667 case O_IPPRECEDENCE: 2668 match = (is_ipv4 && 2669 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2670 break; 2671 2672 case O_IPTOS: 2673 match = (is_ipv4 && 2674 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2675 break; 2676 2677 case O_TCPDATALEN: 2678 if (proto == IPPROTO_TCP && offset == 0) { 2679 struct tcphdr *tcp; 2680 uint16_t x; 2681 uint16_t *p; 2682 int i; 2683 2684 tcp = TCP(ulp); 2685 x = ip_len - 2686 ((ip->ip_hl + tcp->th_off) << 2); 2687 if (cmdlen == 1) { 2688 match = (cmd->arg1 == x); 2689 break; 2690 } 2691 /* otherwise we have ranges */ 2692 p = ((ipfw_insn_u16 *)cmd)->ports; 2693 i = cmdlen - 1; 2694 for (; !match && i>0; i--, p += 2) 2695 match = (x >= p[0] && x <= p[1]); 2696 } 2697 break; 2698 2699 case O_TCPFLAGS: 2700 match = (proto == IPPROTO_TCP && offset == 0 && 2701 flags_match(cmd, TCP(ulp)->th_flags)); 2702 break; 2703 2704 case O_TCPOPTS: 2705 match = (proto == IPPROTO_TCP && offset == 0 && 2706 tcpopts_match(TCP(ulp), cmd)); 2707 break; 2708 2709 case O_TCPSEQ: 2710 match = (proto == IPPROTO_TCP && offset == 0 && 2711 ((ipfw_insn_u32 *)cmd)->d[0] == 2712 TCP(ulp)->th_seq); 2713 break; 2714 2715 case O_TCPACK: 2716 match = (proto == IPPROTO_TCP && offset == 0 && 2717 ((ipfw_insn_u32 *)cmd)->d[0] == 2718 TCP(ulp)->th_ack); 2719 break; 2720 2721 case O_TCPWIN: 2722 match = (proto == IPPROTO_TCP && offset == 0 && 2723 cmd->arg1 == TCP(ulp)->th_win); 2724 break; 2725 2726 case O_ESTAB: 2727 /* reject packets which have SYN only */ 2728 /* XXX should i also check for TH_ACK ? */ 2729 match = (proto == IPPROTO_TCP && offset == 0 && 2730 (TCP(ulp)->th_flags & 2731 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2732 break; 2733 2734 case O_ALTQ: { 2735 struct altq_tag *at; 2736 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2737 2738 match = 1; 2739 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2740 if (mtag != NULL) 2741 break; 2742 mtag = m_tag_get(PACKET_TAG_PF_QID, 2743 sizeof(struct altq_tag), 2744 M_NOWAIT); 2745 if (mtag == NULL) { 2746 /* 2747 * Let the packet fall back to the 2748 * default ALTQ. 2749 */ 2750 break; 2751 } 2752 at = (struct altq_tag *)(mtag+1); 2753 at->qid = altq->qid; 2754 if (is_ipv4) 2755 at->af = AF_INET; 2756 else 2757 at->af = AF_LINK; 2758 at->hdr = ip; 2759 m_tag_prepend(m, mtag); 2760 break; 2761 } 2762 2763 case O_LOG: 2764 if (fw_verbose) 2765 ipfw_log(f, hlen, args, m, oif, offset); 2766 match = 1; 2767 break; 2768 2769 case O_PROB: 2770 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2771 break; 2772 2773 case O_VERREVPATH: 2774 /* Outgoing packets automatically pass/match */ 2775 match = ((oif != NULL) || 2776 (m->m_pkthdr.rcvif == NULL) || 2777 ( 2778#ifdef INET6 2779 is_ipv6 ? 2780 verify_path6(&(args->f_id.src_ip6), 2781 m->m_pkthdr.rcvif) : 2782#endif 2783 verify_path(src_ip, m->m_pkthdr.rcvif))); 2784 break; 2785 2786 case O_VERSRCREACH: 2787 /* Outgoing packets automatically pass/match */ 2788 match = (hlen > 0 && ((oif != NULL) || 2789#ifdef INET6 2790 is_ipv6 ? 2791 verify_path6(&(args->f_id.src_ip6), 2792 NULL) : 2793#endif 2794 verify_path(src_ip, NULL))); 2795 break; 2796 2797 case O_ANTISPOOF: 2798 /* Outgoing packets automatically pass/match */ 2799 if (oif == NULL && hlen > 0 && 2800 ( (is_ipv4 && in_localaddr(src_ip)) 2801#ifdef INET6 2802 || (is_ipv6 && 2803 in6_localaddr(&(args->f_id.src_ip6))) 2804#endif 2805 )) 2806 match = 2807#ifdef INET6 2808 is_ipv6 ? verify_path6( 2809 &(args->f_id.src_ip6), 2810 m->m_pkthdr.rcvif) : 2811#endif 2812 verify_path(src_ip, 2813 m->m_pkthdr.rcvif); 2814 else 2815 match = 1; 2816 break; 2817 2818 case O_IPSEC: 2819#ifdef FAST_IPSEC 2820 match = (m_tag_find(m, 2821 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2822#endif 2823#ifdef IPSEC 2824 match = (ipsec_getnhist(m) != 0); 2825#endif 2826 /* otherwise no match */ 2827 break; 2828 2829#ifdef INET6 2830 case O_IP6_SRC: 2831 match = is_ipv6 && 2832 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2833 &((ipfw_insn_ip6 *)cmd)->addr6); 2834 break; 2835 2836 case O_IP6_DST: 2837 match = is_ipv6 && 2838 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2839 &((ipfw_insn_ip6 *)cmd)->addr6); 2840 break; 2841 case O_IP6_SRC_MASK: 2842 if (is_ipv6) { 2843 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2844 struct in6_addr p = args->f_id.src_ip6; 2845 2846 APPLY_MASK(&p, &te->mask6); 2847 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2848 } 2849 break; 2850 2851 case O_IP6_DST_MASK: 2852 if (is_ipv6) { 2853 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2854 struct in6_addr p = args->f_id.dst_ip6; 2855 2856 APPLY_MASK(&p, &te->mask6); 2857 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2858 } 2859 break; 2860 2861 case O_IP6_SRC_ME: 2862 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2863 break; 2864 2865 case O_IP6_DST_ME: 2866 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2867 break; 2868 2869 case O_FLOW6ID: 2870 match = is_ipv6 && 2871 flow6id_match(args->f_id.flow_id6, 2872 (ipfw_insn_u32 *) cmd); 2873 break; 2874 2875 case O_EXT_HDR: 2876 match = is_ipv6 && 2877 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2878 break; 2879 2880 case O_IP6: 2881 match = is_ipv6; 2882 break; 2883#endif 2884 2885 case O_IP4: 2886 match = is_ipv4; 2887 break; 2888 2889 /* 2890 * The second set of opcodes represents 'actions', 2891 * i.e. the terminal part of a rule once the packet 2892 * matches all previous patterns. 2893 * Typically there is only one action for each rule, 2894 * and the opcode is stored at the end of the rule 2895 * (but there are exceptions -- see below). 2896 * 2897 * In general, here we set retval and terminate the 2898 * outer loop (would be a 'break 3' in some language, 2899 * but we need to do a 'goto done'). 2900 * 2901 * Exceptions: 2902 * O_COUNT and O_SKIPTO actions: 2903 * instead of terminating, we jump to the next rule 2904 * ('goto next_rule', equivalent to a 'break 2'), 2905 * or to the SKIPTO target ('goto again' after 2906 * having set f, cmd and l), respectively. 2907 * 2908 * O_LOG and O_ALTQ action parameters: 2909 * perform some action and set match = 1; 2910 * 2911 * O_LIMIT and O_KEEP_STATE: these opcodes are 2912 * not real 'actions', and are stored right 2913 * before the 'action' part of the rule. 2914 * These opcodes try to install an entry in the 2915 * state tables; if successful, we continue with 2916 * the next opcode (match=1; break;), otherwise 2917 * the packet * must be dropped 2918 * ('goto done' after setting retval); 2919 * 2920 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2921 * cause a lookup of the state table, and a jump 2922 * to the 'action' part of the parent rule 2923 * ('goto check_body') if an entry is found, or 2924 * (CHECK_STATE only) a jump to the next rule if 2925 * the entry is not found ('goto next_rule'). 2926 * The result of the lookup is cached to make 2927 * further instances of these opcodes are 2928 * effectively NOPs. 2929 */ 2930 case O_LIMIT: 2931 case O_KEEP_STATE: 2932 if (install_state(f, 2933 (ipfw_insn_limit *)cmd, args)) { 2934 retval = IP_FW_DENY; 2935 goto done; /* error/limit violation */ 2936 } 2937 match = 1; 2938 break; 2939 2940 case O_PROBE_STATE: 2941 case O_CHECK_STATE: 2942 /* 2943 * dynamic rules are checked at the first 2944 * keep-state or check-state occurrence, 2945 * with the result being stored in dyn_dir. 2946 * The compiler introduces a PROBE_STATE 2947 * instruction for us when we have a 2948 * KEEP_STATE (because PROBE_STATE needs 2949 * to be run first). 2950 */ 2951 if (dyn_dir == MATCH_UNKNOWN && 2952 (q = lookup_dyn_rule(&args->f_id, 2953 &dyn_dir, proto == IPPROTO_TCP ? 2954 TCP(ulp) : NULL)) 2955 != NULL) { 2956 /* 2957 * Found dynamic entry, update stats 2958 * and jump to the 'action' part of 2959 * the parent rule. 2960 */ 2961 q->pcnt++; 2962 q->bcnt += pktlen; 2963 f = q->rule; 2964 cmd = ACTION_PTR(f); 2965 l = f->cmd_len - f->act_ofs; 2966 IPFW_DYN_UNLOCK(); 2967 goto check_body; 2968 } 2969 /* 2970 * Dynamic entry not found. If CHECK_STATE, 2971 * skip to next rule, if PROBE_STATE just 2972 * ignore and continue with next opcode. 2973 */ 2974 if (cmd->opcode == O_CHECK_STATE) 2975 goto next_rule; 2976 match = 1; 2977 break; 2978 2979 case O_ACCEPT: 2980 retval = 0; /* accept */ 2981 goto done; 2982 2983 case O_PIPE: 2984 case O_QUEUE: 2985 args->rule = f; /* report matching rule */ 2986 if (cmd->arg1 == IP_FW_TABLEARG) 2987 args->cookie = tablearg; 2988 else 2989 args->cookie = cmd->arg1; 2990 retval = IP_FW_DUMMYNET; 2991 goto done; 2992 2993 case O_DIVERT: 2994 case O_TEE: { 2995 struct divert_tag *dt; 2996 2997 if (args->eh) /* not on layer 2 */ 2998 break; 2999 mtag = m_tag_get(PACKET_TAG_DIVERT, 3000 sizeof(struct divert_tag), 3001 M_NOWAIT); 3002 if (mtag == NULL) { 3003 /* XXX statistic */ 3004 /* drop packet */ 3005 IPFW_RUNLOCK(chain); 3006 return (IP_FW_DENY); 3007 } 3008 dt = (struct divert_tag *)(mtag+1); 3009 dt->cookie = f->rulenum; 3010 if (cmd->arg1 == IP_FW_TABLEARG) 3011 dt->info = tablearg; 3012 else 3013 dt->info = cmd->arg1; 3014 m_tag_prepend(m, mtag); 3015 retval = (cmd->opcode == O_DIVERT) ? 3016 IP_FW_DIVERT : IP_FW_TEE; 3017 goto done; 3018 } 3019 3020 case O_COUNT: 3021 case O_SKIPTO: 3022 f->pcnt++; /* update stats */ 3023 f->bcnt += pktlen; 3024 f->timestamp = time_uptime; 3025 if (cmd->opcode == O_COUNT) 3026 goto next_rule; 3027 /* handle skipto */ 3028 if (f->next_rule == NULL) 3029 lookup_next_rule(f); 3030 f = f->next_rule; 3031 goto again; 3032 3033 case O_REJECT: 3034 /* 3035 * Drop the packet and send a reject notice 3036 * if the packet is not ICMP (or is an ICMP 3037 * query), and it is not multicast/broadcast. 3038 */ 3039 if (hlen > 0 && is_ipv4 && offset == 0 && 3040 (proto != IPPROTO_ICMP || 3041 is_icmp_query(ICMP(ulp))) && 3042 !(m->m_flags & (M_BCAST|M_MCAST)) && 3043 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3044 send_reject(args, cmd->arg1, 3045 offset,ip_len); 3046 m = args->m; 3047 } 3048 /* FALLTHROUGH */ 3049#ifdef INET6 3050 case O_UNREACH6: 3051 if (hlen > 0 && is_ipv6 && 3052 (proto != IPPROTO_ICMPV6 || 3053 (is_icmp6_query(args->f_id.flags) == 1)) && 3054 !(m->m_flags & (M_BCAST|M_MCAST)) && 3055 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3056 send_reject6(args, cmd->arg1, 3057 offset, hlen); 3058 m = args->m; 3059 } 3060 /* FALLTHROUGH */ 3061#endif 3062 case O_DENY: 3063 retval = IP_FW_DENY; 3064 goto done; 3065 3066 case O_FORWARD_IP: 3067 if (args->eh) /* not valid on layer2 pkts */ 3068 break; 3069 if (!q || dyn_dir == MATCH_FORWARD) 3070 args->next_hop = 3071 &((ipfw_insn_sa *)cmd)->sa; 3072 retval = IP_FW_PASS; 3073 goto done; 3074 3075 case O_NETGRAPH: 3076 case O_NGTEE: 3077 args->rule = f; /* report matching rule */ 3078 if (cmd->arg1 == IP_FW_TABLEARG) 3079 args->cookie = tablearg; 3080 else 3081 args->cookie = cmd->arg1; 3082 retval = (cmd->opcode == O_NETGRAPH) ? 3083 IP_FW_NETGRAPH : IP_FW_NGTEE; 3084 goto done; 3085 3086 default: 3087 panic("-- unknown opcode %d\n", cmd->opcode); 3088 } /* end of switch() on opcodes */ 3089 3090 if (cmd->len & F_NOT) 3091 match = !match; 3092 3093 if (match) { 3094 if (cmd->len & F_OR) 3095 skip_or = 1; 3096 } else { 3097 if (!(cmd->len & F_OR)) /* not an OR block, */ 3098 break; /* try next rule */ 3099 } 3100 3101 } /* end of inner for, scan opcodes */ 3102 3103next_rule:; /* try next rule */ 3104 3105 } /* end of outer for, scan rules */ 3106 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3107 IPFW_RUNLOCK(chain); 3108 return (IP_FW_DENY); 3109 3110done: 3111 /* Update statistics */ 3112 f->pcnt++; 3113 f->bcnt += pktlen; 3114 f->timestamp = time_uptime; 3115 IPFW_RUNLOCK(chain); 3116 return (retval); 3117 3118pullup_failed: 3119 if (fw_verbose) 3120 printf("ipfw: pullup failed\n"); 3121 return (IP_FW_DENY); 3122} 3123 3124/* 3125 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3126 * These will be reconstructed on the fly as packets are matched. 3127 */ 3128static void 3129flush_rule_ptrs(struct ip_fw_chain *chain) 3130{ 3131 struct ip_fw *rule; 3132 3133 IPFW_WLOCK_ASSERT(chain); 3134 3135 for (rule = chain->rules; rule; rule = rule->next) 3136 rule->next_rule = NULL; 3137} 3138 3139/* 3140 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3141 * possibly create a rule number and add the rule to the list. 3142 * Update the rule_number in the input struct so the caller knows it as well. 3143 */ 3144static int 3145add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3146{ 3147 struct ip_fw *rule, *f, *prev; 3148 int l = RULESIZE(input_rule); 3149 3150 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3151 return (EINVAL); 3152 3153 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3154 if (rule == NULL) 3155 return (ENOSPC); 3156 3157 bcopy(input_rule, rule, l); 3158 3159 rule->next = NULL; 3160 rule->next_rule = NULL; 3161 3162 rule->pcnt = 0; 3163 rule->bcnt = 0; 3164 rule->timestamp = 0; 3165 3166 IPFW_WLOCK(chain); 3167 3168 if (chain->rules == NULL) { /* default rule */ 3169 chain->rules = rule; 3170 goto done; 3171 } 3172 3173 /* 3174 * If rulenum is 0, find highest numbered rule before the 3175 * default rule, and add autoinc_step 3176 */ 3177 if (autoinc_step < 1) 3178 autoinc_step = 1; 3179 else if (autoinc_step > 1000) 3180 autoinc_step = 1000; 3181 if (rule->rulenum == 0) { 3182 /* 3183 * locate the highest numbered rule before default 3184 */ 3185 for (f = chain->rules; f; f = f->next) { 3186 if (f->rulenum == IPFW_DEFAULT_RULE) 3187 break; 3188 rule->rulenum = f->rulenum; 3189 } 3190 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3191 rule->rulenum += autoinc_step; 3192 input_rule->rulenum = rule->rulenum; 3193 } 3194 3195 /* 3196 * Now insert the new rule in the right place in the sorted list. 3197 */ 3198 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3199 if (f->rulenum > rule->rulenum) { /* found the location */ 3200 if (prev) { 3201 rule->next = f; 3202 prev->next = rule; 3203 } else { /* head insert */ 3204 rule->next = chain->rules; 3205 chain->rules = rule; 3206 } 3207 break; 3208 } 3209 } 3210 flush_rule_ptrs(chain); 3211done: 3212 static_count++; 3213 static_len += l; 3214 IPFW_WUNLOCK(chain); 3215 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3216 rule->rulenum, static_count);) 3217 return (0); 3218} 3219 3220/** 3221 * Remove a static rule (including derived * dynamic rules) 3222 * and place it on the ``reap list'' for later reclamation. 3223 * The caller is in charge of clearing rule pointers to avoid 3224 * dangling pointers. 3225 * @return a pointer to the next entry. 3226 * Arguments are not checked, so they better be correct. 3227 */ 3228static struct ip_fw * 3229remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3230{ 3231 struct ip_fw *n; 3232 int l = RULESIZE(rule); 3233 3234 IPFW_WLOCK_ASSERT(chain); 3235 3236 n = rule->next; 3237 IPFW_DYN_LOCK(); 3238 remove_dyn_rule(rule, NULL /* force removal */); 3239 IPFW_DYN_UNLOCK(); 3240 if (prev == NULL) 3241 chain->rules = n; 3242 else 3243 prev->next = n; 3244 static_count--; 3245 static_len -= l; 3246 3247 rule->next = chain->reap; 3248 chain->reap = rule; 3249 3250 return n; 3251} 3252 3253/** 3254 * Reclaim storage associated with a list of rules. This is 3255 * typically the list created using remove_rule. 3256 */ 3257static void 3258reap_rules(struct ip_fw *head) 3259{ 3260 struct ip_fw *rule; 3261 3262 while ((rule = head) != NULL) { 3263 head = head->next; 3264 if (DUMMYNET_LOADED) 3265 ip_dn_ruledel_ptr(rule); 3266 free(rule, M_IPFW); 3267 } 3268} 3269 3270/* 3271 * Remove all rules from a chain (except rules in set RESVD_SET 3272 * unless kill_default = 1). The caller is responsible for 3273 * reclaiming storage for the rules left in chain->reap. 3274 */ 3275static void 3276free_chain(struct ip_fw_chain *chain, int kill_default) 3277{ 3278 struct ip_fw *prev, *rule; 3279 3280 IPFW_WLOCK_ASSERT(chain); 3281 3282 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3283 for (prev = NULL, rule = chain->rules; rule ; ) 3284 if (kill_default || rule->set != RESVD_SET) 3285 rule = remove_rule(chain, rule, prev); 3286 else { 3287 prev = rule; 3288 rule = rule->next; 3289 } 3290} 3291 3292/** 3293 * Remove all rules with given number, and also do set manipulation. 3294 * Assumes chain != NULL && *chain != NULL. 3295 * 3296 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3297 * the next 8 bits are the new set, the top 8 bits are the command: 3298 * 3299 * 0 delete rules with given number 3300 * 1 delete rules with given set number 3301 * 2 move rules with given number to new set 3302 * 3 move rules with given set number to new set 3303 * 4 swap sets with given numbers 3304 */ 3305static int 3306del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3307{ 3308 struct ip_fw *prev = NULL, *rule; 3309 u_int16_t rulenum; /* rule or old_set */ 3310 u_int8_t cmd, new_set; 3311 3312 rulenum = arg & 0xffff; 3313 cmd = (arg >> 24) & 0xff; 3314 new_set = (arg >> 16) & 0xff; 3315 3316 if (cmd > 4) 3317 return EINVAL; 3318 if (new_set > RESVD_SET) 3319 return EINVAL; 3320 if (cmd == 0 || cmd == 2) { 3321 if (rulenum >= IPFW_DEFAULT_RULE) 3322 return EINVAL; 3323 } else { 3324 if (rulenum > RESVD_SET) /* old_set */ 3325 return EINVAL; 3326 } 3327 3328 IPFW_WLOCK(chain); 3329 rule = chain->rules; 3330 chain->reap = NULL; 3331 switch (cmd) { 3332 case 0: /* delete rules with given number */ 3333 /* 3334 * locate first rule to delete 3335 */ 3336 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3337 ; 3338 if (rule->rulenum != rulenum) { 3339 IPFW_WUNLOCK(chain); 3340 return EINVAL; 3341 } 3342 3343 /* 3344 * flush pointers outside the loop, then delete all matching 3345 * rules. prev remains the same throughout the cycle. 3346 */ 3347 flush_rule_ptrs(chain); 3348 while (rule->rulenum == rulenum) 3349 rule = remove_rule(chain, rule, prev); 3350 break; 3351 3352 case 1: /* delete all rules with given set number */ 3353 flush_rule_ptrs(chain); 3354 rule = chain->rules; 3355 while (rule->rulenum < IPFW_DEFAULT_RULE) 3356 if (rule->set == rulenum) 3357 rule = remove_rule(chain, rule, prev); 3358 else { 3359 prev = rule; 3360 rule = rule->next; 3361 } 3362 break; 3363 3364 case 2: /* move rules with given number to new set */ 3365 rule = chain->rules; 3366 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3367 if (rule->rulenum == rulenum) 3368 rule->set = new_set; 3369 break; 3370 3371 case 3: /* move rules with given set number to new set */ 3372 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3373 if (rule->set == rulenum) 3374 rule->set = new_set; 3375 break; 3376 3377 case 4: /* swap two sets */ 3378 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3379 if (rule->set == rulenum) 3380 rule->set = new_set; 3381 else if (rule->set == new_set) 3382 rule->set = rulenum; 3383 break; 3384 } 3385 /* 3386 * Look for rules to reclaim. We grab the list before 3387 * releasing the lock then reclaim them w/o the lock to 3388 * avoid a LOR with dummynet. 3389 */ 3390 rule = chain->reap; 3391 chain->reap = NULL; 3392 IPFW_WUNLOCK(chain); 3393 if (rule) 3394 reap_rules(rule); 3395 return 0; 3396} 3397 3398/* 3399 * Clear counters for a specific rule. 3400 * The enclosing "table" is assumed locked. 3401 */ 3402static void 3403clear_counters(struct ip_fw *rule, int log_only) 3404{ 3405 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3406 3407 if (log_only == 0) { 3408 rule->bcnt = rule->pcnt = 0; 3409 rule->timestamp = 0; 3410 } 3411 if (l->o.opcode == O_LOG) 3412 l->log_left = l->max_log; 3413} 3414 3415/** 3416 * Reset some or all counters on firewall rules. 3417 * @arg frwl is null to clear all entries, or contains a specific 3418 * rule number. 3419 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3420 */ 3421static int 3422zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3423{ 3424 struct ip_fw *rule; 3425 char *msg; 3426 3427 IPFW_WLOCK(chain); 3428 if (rulenum == 0) { 3429 norule_counter = 0; 3430 for (rule = chain->rules; rule; rule = rule->next) 3431 clear_counters(rule, log_only); 3432 msg = log_only ? "ipfw: All logging counts reset.\n" : 3433 "ipfw: Accounting cleared.\n"; 3434 } else { 3435 int cleared = 0; 3436 /* 3437 * We can have multiple rules with the same number, so we 3438 * need to clear them all. 3439 */ 3440 for (rule = chain->rules; rule; rule = rule->next) 3441 if (rule->rulenum == rulenum) { 3442 while (rule && rule->rulenum == rulenum) { 3443 clear_counters(rule, log_only); 3444 rule = rule->next; 3445 } 3446 cleared = 1; 3447 break; 3448 } 3449 if (!cleared) { /* we did not find any matching rules */ 3450 IPFW_WUNLOCK(chain); 3451 return (EINVAL); 3452 } 3453 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3454 "ipfw: Entry %d cleared.\n"; 3455 } 3456 IPFW_WUNLOCK(chain); 3457 3458 if (fw_verbose) 3459 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3460 return (0); 3461} 3462 3463/* 3464 * Check validity of the structure before insert. 3465 * Fortunately rules are simple, so this mostly need to check rule sizes. 3466 */ 3467static int 3468check_ipfw_struct(struct ip_fw *rule, int size) 3469{ 3470 int l, cmdlen = 0; 3471 int have_action=0; 3472 ipfw_insn *cmd; 3473 3474 if (size < sizeof(*rule)) { 3475 printf("ipfw: rule too short\n"); 3476 return (EINVAL); 3477 } 3478 /* first, check for valid size */ 3479 l = RULESIZE(rule); 3480 if (l != size) { 3481 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3482 return (EINVAL); 3483 } 3484 if (rule->act_ofs >= rule->cmd_len) { 3485 printf("ipfw: bogus action offset (%u > %u)\n", 3486 rule->act_ofs, rule->cmd_len - 1); 3487 return (EINVAL); 3488 } 3489 /* 3490 * Now go for the individual checks. Very simple ones, basically only 3491 * instruction sizes. 3492 */ 3493 for (l = rule->cmd_len, cmd = rule->cmd ; 3494 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3495 cmdlen = F_LEN(cmd); 3496 if (cmdlen > l) { 3497 printf("ipfw: opcode %d size truncated\n", 3498 cmd->opcode); 3499 return EINVAL; 3500 } 3501 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3502 switch (cmd->opcode) { 3503 case O_PROBE_STATE: 3504 case O_KEEP_STATE: 3505 case O_PROTO: 3506 case O_IP_SRC_ME: 3507 case O_IP_DST_ME: 3508 case O_LAYER2: 3509 case O_IN: 3510 case O_FRAG: 3511 case O_DIVERTED: 3512 case O_IPOPT: 3513 case O_IPTOS: 3514 case O_IPPRECEDENCE: 3515 case O_IPVER: 3516 case O_TCPWIN: 3517 case O_TCPFLAGS: 3518 case O_TCPOPTS: 3519 case O_ESTAB: 3520 case O_VERREVPATH: 3521 case O_VERSRCREACH: 3522 case O_ANTISPOOF: 3523 case O_IPSEC: 3524#ifdef INET6 3525 case O_IP6_SRC_ME: 3526 case O_IP6_DST_ME: 3527 case O_EXT_HDR: 3528 case O_IP6: 3529#endif 3530 case O_IP4: 3531 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3532 goto bad_size; 3533 break; 3534 3535 case O_UID: 3536 case O_GID: 3537 case O_JAIL: 3538 case O_IP_SRC: 3539 case O_IP_DST: 3540 case O_TCPSEQ: 3541 case O_TCPACK: 3542 case O_PROB: 3543 case O_ICMPTYPE: 3544 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3545 goto bad_size; 3546 break; 3547 3548 case O_LIMIT: 3549 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3550 goto bad_size; 3551 break; 3552 3553 case O_LOG: 3554 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3555 goto bad_size; 3556 3557 ((ipfw_insn_log *)cmd)->log_left = 3558 ((ipfw_insn_log *)cmd)->max_log; 3559 3560 break; 3561 3562 case O_IP_SRC_MASK: 3563 case O_IP_DST_MASK: 3564 /* only odd command lengths */ 3565 if ( !(cmdlen & 1) || cmdlen > 31) 3566 goto bad_size; 3567 break; 3568 3569 case O_IP_SRC_SET: 3570 case O_IP_DST_SET: 3571 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3572 printf("ipfw: invalid set size %d\n", 3573 cmd->arg1); 3574 return EINVAL; 3575 } 3576 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3577 (cmd->arg1+31)/32 ) 3578 goto bad_size; 3579 break; 3580 3581 case O_IP_SRC_LOOKUP: 3582 case O_IP_DST_LOOKUP: 3583 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3584 printf("ipfw: invalid table number %d\n", 3585 cmd->arg1); 3586 return (EINVAL); 3587 } 3588 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3589 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3590 goto bad_size; 3591 break; 3592 3593 case O_MACADDR2: 3594 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3595 goto bad_size; 3596 break; 3597 3598 case O_NOP: 3599 case O_IPID: 3600 case O_IPTTL: 3601 case O_IPLEN: 3602 case O_TCPDATALEN: 3603 if (cmdlen < 1 || cmdlen > 31) 3604 goto bad_size; 3605 break; 3606 3607 case O_MAC_TYPE: 3608 case O_IP_SRCPORT: 3609 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3610 if (cmdlen < 2 || cmdlen > 31) 3611 goto bad_size; 3612 break; 3613 3614 case O_RECV: 3615 case O_XMIT: 3616 case O_VIA: 3617 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3618 goto bad_size; 3619 break; 3620 3621 case O_ALTQ: 3622 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3623 goto bad_size; 3624 break; 3625 3626 case O_PIPE: 3627 case O_QUEUE: 3628 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3629 goto bad_size; 3630 goto check_action; 3631 3632 case O_FORWARD_IP: 3633#ifdef IPFIREWALL_FORWARD 3634 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3635 goto bad_size; 3636 goto check_action; 3637#else 3638 return EINVAL; 3639#endif 3640 3641 case O_DIVERT: 3642 case O_TEE: 3643 if (ip_divert_ptr == NULL) 3644 return EINVAL; 3645 else 3646 goto check_size; 3647 case O_NETGRAPH: 3648 case O_NGTEE: 3649 if (!NG_IPFW_LOADED) 3650 return EINVAL; 3651 else 3652 goto check_size; 3653 case O_FORWARD_MAC: /* XXX not implemented yet */ 3654 case O_CHECK_STATE: 3655 case O_COUNT: 3656 case O_ACCEPT: 3657 case O_DENY: 3658 case O_REJECT: 3659#ifdef INET6 3660 case O_UNREACH6: 3661#endif 3662 case O_SKIPTO: 3663check_size: 3664 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3665 goto bad_size; 3666check_action: 3667 if (have_action) { 3668 printf("ipfw: opcode %d, multiple actions" 3669 " not allowed\n", 3670 cmd->opcode); 3671 return EINVAL; 3672 } 3673 have_action = 1; 3674 if (l != cmdlen) { 3675 printf("ipfw: opcode %d, action must be" 3676 " last opcode\n", 3677 cmd->opcode); 3678 return EINVAL; 3679 } 3680 break; 3681#ifdef INET6 3682 case O_IP6_SRC: 3683 case O_IP6_DST: 3684 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3685 F_INSN_SIZE(ipfw_insn)) 3686 goto bad_size; 3687 break; 3688 3689 case O_FLOW6ID: 3690 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3691 ((ipfw_insn_u32 *)cmd)->o.arg1) 3692 goto bad_size; 3693 break; 3694 3695 case O_IP6_SRC_MASK: 3696 case O_IP6_DST_MASK: 3697 if ( !(cmdlen & 1) || cmdlen > 127) 3698 goto bad_size; 3699 break; 3700 case O_ICMP6TYPE: 3701 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3702 goto bad_size; 3703 break; 3704#endif 3705 3706 default: 3707 switch (cmd->opcode) { 3708#ifndef INET6 3709 case O_IP6_SRC_ME: 3710 case O_IP6_DST_ME: 3711 case O_EXT_HDR: 3712 case O_IP6: 3713 case O_UNREACH6: 3714 case O_IP6_SRC: 3715 case O_IP6_DST: 3716 case O_FLOW6ID: 3717 case O_IP6_SRC_MASK: 3718 case O_IP6_DST_MASK: 3719 case O_ICMP6TYPE: 3720 printf("ipfw: no IPv6 support in kernel\n"); 3721 return EPROTONOSUPPORT; 3722#endif 3723 default: 3724 printf("ipfw: opcode %d, unknown opcode\n", 3725 cmd->opcode); 3726 return EINVAL; 3727 } 3728 } 3729 } 3730 if (have_action == 0) { 3731 printf("ipfw: missing action\n"); 3732 return EINVAL; 3733 } 3734 return 0; 3735 3736bad_size: 3737 printf("ipfw: opcode %d size %d wrong\n", 3738 cmd->opcode, cmdlen); 3739 return EINVAL; 3740} 3741 3742/* 3743 * Copy the static and dynamic rules to the supplied buffer 3744 * and return the amount of space actually used. 3745 */ 3746static size_t 3747ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3748{ 3749 char *bp = buf; 3750 char *ep = bp + space; 3751 struct ip_fw *rule; 3752 int i; 3753 3754 /* XXX this can take a long time and locking will block packet flow */ 3755 IPFW_RLOCK(chain); 3756 for (rule = chain->rules; rule ; rule = rule->next) { 3757 /* 3758 * Verify the entry fits in the buffer in case the 3759 * rules changed between calculating buffer space and 3760 * now. This would be better done using a generation 3761 * number but should suffice for now. 3762 */ 3763 i = RULESIZE(rule); 3764 if (bp + i <= ep) { 3765 bcopy(rule, bp, i); 3766 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3767 sizeof(set_disable)); 3768 bp += i; 3769 } 3770 } 3771 IPFW_RUNLOCK(chain); 3772 if (ipfw_dyn_v) { 3773 ipfw_dyn_rule *p, *last = NULL; 3774 3775 IPFW_DYN_LOCK(); 3776 for (i = 0 ; i < curr_dyn_buckets; i++) 3777 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3778 if (bp + sizeof *p <= ep) { 3779 ipfw_dyn_rule *dst = 3780 (ipfw_dyn_rule *)bp; 3781 bcopy(p, dst, sizeof *p); 3782 bcopy(&(p->rule->rulenum), &(dst->rule), 3783 sizeof(p->rule->rulenum)); 3784 /* 3785 * store a non-null value in "next". 3786 * The userland code will interpret a 3787 * NULL here as a marker 3788 * for the last dynamic rule. 3789 */ 3790 bcopy(&dst, &dst->next, sizeof(dst)); 3791 last = dst; 3792 dst->expire = 3793 TIME_LEQ(dst->expire, time_uptime) ? 3794 0 : dst->expire - time_uptime ; 3795 bp += sizeof(ipfw_dyn_rule); 3796 } 3797 } 3798 IPFW_DYN_UNLOCK(); 3799 if (last != NULL) /* mark last dynamic rule */ 3800 bzero(&last->next, sizeof(last)); 3801 } 3802 return (bp - (char *)buf); 3803} 3804 3805 3806/** 3807 * {set|get}sockopt parser. 3808 */ 3809static int 3810ipfw_ctl(struct sockopt *sopt) 3811{ 3812#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3813 int error, rule_num; 3814 size_t size; 3815 struct ip_fw *buf, *rule; 3816 u_int32_t rulenum[2]; 3817 3818 error = suser(sopt->sopt_td); 3819 if (error) 3820 return (error); 3821 3822 /* 3823 * Disallow modifications in really-really secure mode, but still allow 3824 * the logging counters to be reset. 3825 */ 3826 if (sopt->sopt_name == IP_FW_ADD || 3827 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3828 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3829 if (error) 3830 return (error); 3831 } 3832 3833 error = 0; 3834 3835 switch (sopt->sopt_name) { 3836 case IP_FW_GET: 3837 /* 3838 * pass up a copy of the current rules. Static rules 3839 * come first (the last of which has number IPFW_DEFAULT_RULE), 3840 * followed by a possibly empty list of dynamic rule. 3841 * The last dynamic rule has NULL in the "next" field. 3842 * 3843 * Note that the calculated size is used to bound the 3844 * amount of data returned to the user. The rule set may 3845 * change between calculating the size and returning the 3846 * data in which case we'll just return what fits. 3847 */ 3848 size = static_len; /* size of static rules */ 3849 if (ipfw_dyn_v) /* add size of dyn.rules */ 3850 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3851 3852 /* 3853 * XXX todo: if the user passes a short length just to know 3854 * how much room is needed, do not bother filling up the 3855 * buffer, just jump to the sooptcopyout. 3856 */ 3857 buf = malloc(size, M_TEMP, M_WAITOK); 3858 error = sooptcopyout(sopt, buf, 3859 ipfw_getrules(&layer3_chain, buf, size)); 3860 free(buf, M_TEMP); 3861 break; 3862 3863 case IP_FW_FLUSH: 3864 /* 3865 * Normally we cannot release the lock on each iteration. 3866 * We could do it here only because we start from the head all 3867 * the times so there is no risk of missing some entries. 3868 * On the other hand, the risk is that we end up with 3869 * a very inconsistent ruleset, so better keep the lock 3870 * around the whole cycle. 3871 * 3872 * XXX this code can be improved by resetting the head of 3873 * the list to point to the default rule, and then freeing 3874 * the old list without the need for a lock. 3875 */ 3876 3877 IPFW_WLOCK(&layer3_chain); 3878 layer3_chain.reap = NULL; 3879 free_chain(&layer3_chain, 0 /* keep default rule */); 3880 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3881 IPFW_WUNLOCK(&layer3_chain); 3882 if (layer3_chain.reap != NULL) 3883 reap_rules(rule); 3884 break; 3885 3886 case IP_FW_ADD: 3887 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3888 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3889 sizeof(struct ip_fw) ); 3890 if (error == 0) 3891 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3892 if (error == 0) { 3893 error = add_rule(&layer3_chain, rule); 3894 size = RULESIZE(rule); 3895 if (!error && sopt->sopt_dir == SOPT_GET) 3896 error = sooptcopyout(sopt, rule, size); 3897 } 3898 free(rule, M_TEMP); 3899 break; 3900 3901 case IP_FW_DEL: 3902 /* 3903 * IP_FW_DEL is used for deleting single rules or sets, 3904 * and (ab)used to atomically manipulate sets. Argument size 3905 * is used to distinguish between the two: 3906 * sizeof(u_int32_t) 3907 * delete single rule or set of rules, 3908 * or reassign rules (or sets) to a different set. 3909 * 2*sizeof(u_int32_t) 3910 * atomic disable/enable sets. 3911 * first u_int32_t contains sets to be disabled, 3912 * second u_int32_t contains sets to be enabled. 3913 */ 3914 error = sooptcopyin(sopt, rulenum, 3915 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3916 if (error) 3917 break; 3918 size = sopt->sopt_valsize; 3919 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3920 error = del_entry(&layer3_chain, rulenum[0]); 3921 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3922 set_disable = 3923 (set_disable | rulenum[0]) & ~rulenum[1] & 3924 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3925 else 3926 error = EINVAL; 3927 break; 3928 3929 case IP_FW_ZERO: 3930 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3931 rule_num = 0; 3932 if (sopt->sopt_val != 0) { 3933 error = sooptcopyin(sopt, &rule_num, 3934 sizeof(int), sizeof(int)); 3935 if (error) 3936 break; 3937 } 3938 error = zero_entry(&layer3_chain, rule_num, 3939 sopt->sopt_name == IP_FW_RESETLOG); 3940 break; 3941 3942 case IP_FW_TABLE_ADD: 3943 { 3944 ipfw_table_entry ent; 3945 3946 error = sooptcopyin(sopt, &ent, 3947 sizeof(ent), sizeof(ent)); 3948 if (error) 3949 break; 3950 error = add_table_entry(&layer3_chain, ent.tbl, 3951 ent.addr, ent.masklen, ent.value); 3952 } 3953 break; 3954 3955 case IP_FW_TABLE_DEL: 3956 { 3957 ipfw_table_entry ent; 3958 3959 error = sooptcopyin(sopt, &ent, 3960 sizeof(ent), sizeof(ent)); 3961 if (error) 3962 break; 3963 error = del_table_entry(&layer3_chain, ent.tbl, 3964 ent.addr, ent.masklen); 3965 } 3966 break; 3967 3968 case IP_FW_TABLE_FLUSH: 3969 { 3970 u_int16_t tbl; 3971 3972 error = sooptcopyin(sopt, &tbl, 3973 sizeof(tbl), sizeof(tbl)); 3974 if (error) 3975 break; 3976 IPFW_WLOCK(&layer3_chain); 3977 error = flush_table(&layer3_chain, tbl); 3978 IPFW_WUNLOCK(&layer3_chain); 3979 } 3980 break; 3981 3982 case IP_FW_TABLE_GETSIZE: 3983 { 3984 u_int32_t tbl, cnt; 3985 3986 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 3987 sizeof(tbl)))) 3988 break; 3989 IPFW_RLOCK(&layer3_chain); 3990 error = count_table(&layer3_chain, tbl, &cnt); 3991 IPFW_RUNLOCK(&layer3_chain); 3992 if (error) 3993 break; 3994 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 3995 } 3996 break; 3997 3998 case IP_FW_TABLE_LIST: 3999 { 4000 ipfw_table *tbl; 4001 4002 if (sopt->sopt_valsize < sizeof(*tbl)) { 4003 error = EINVAL; 4004 break; 4005 } 4006 size = sopt->sopt_valsize; 4007 tbl = malloc(size, M_TEMP, M_WAITOK); 4008 if (tbl == NULL) { 4009 error = ENOMEM; 4010 break; 4011 } 4012 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4013 if (error) { 4014 free(tbl, M_TEMP); 4015 break; 4016 } 4017 tbl->size = (size - sizeof(*tbl)) / 4018 sizeof(ipfw_table_entry); 4019 IPFW_RLOCK(&layer3_chain); 4020 error = dump_table(&layer3_chain, tbl); 4021 IPFW_RUNLOCK(&layer3_chain); 4022 if (error) { 4023 free(tbl, M_TEMP); 4024 break; 4025 } 4026 error = sooptcopyout(sopt, tbl, size); 4027 free(tbl, M_TEMP); 4028 } 4029 break; 4030 4031 default: 4032 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4033 error = EINVAL; 4034 } 4035 4036 return (error); 4037#undef RULE_MAXSIZE 4038} 4039 4040/** 4041 * dummynet needs a reference to the default rule, because rules can be 4042 * deleted while packets hold a reference to them. When this happens, 4043 * dummynet changes the reference to the default rule (it could well be a 4044 * NULL pointer, but this way we do not need to check for the special 4045 * case, plus here he have info on the default behaviour). 4046 */ 4047struct ip_fw *ip_fw_default_rule; 4048 4049/* 4050 * This procedure is only used to handle keepalives. It is invoked 4051 * every dyn_keepalive_period 4052 */ 4053static void 4054ipfw_tick(void * __unused unused) 4055{ 4056 struct mbuf *m0, *m, *mnext, **mtailp; 4057 int i; 4058 ipfw_dyn_rule *q; 4059 4060 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4061 goto done; 4062 4063 /* 4064 * We make a chain of packets to go out here -- not deferring 4065 * until after we drop the IPFW dynamic rule lock would result 4066 * in a lock order reversal with the normal packet input -> ipfw 4067 * call stack. 4068 */ 4069 m0 = NULL; 4070 mtailp = &m0; 4071 IPFW_DYN_LOCK(); 4072 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4073 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4074 if (q->dyn_type == O_LIMIT_PARENT) 4075 continue; 4076 if (q->id.proto != IPPROTO_TCP) 4077 continue; 4078 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4079 continue; 4080 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4081 q->expire)) 4082 continue; /* too early */ 4083 if (TIME_LEQ(q->expire, time_uptime)) 4084 continue; /* too late, rule expired */ 4085 4086 *mtailp = send_pkt(&(q->id), q->ack_rev - 1, 4087 q->ack_fwd, TH_SYN); 4088 if (*mtailp != NULL) 4089 mtailp = &(*mtailp)->m_nextpkt; 4090 *mtailp = send_pkt(&(q->id), q->ack_fwd - 1, 4091 q->ack_rev, 0); 4092 if (*mtailp != NULL) 4093 mtailp = &(*mtailp)->m_nextpkt; 4094 } 4095 } 4096 IPFW_DYN_UNLOCK(); 4097 for (m = mnext = m0; m != NULL; m = mnext) { 4098 mnext = m->m_nextpkt; 4099 m->m_nextpkt = NULL; 4100 ip_output(m, NULL, NULL, 0, NULL, NULL); 4101 } 4102done: 4103 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4104} 4105 4106int 4107ipfw_init(void) 4108{ 4109 struct ip_fw default_rule; 4110 int error; 4111 4112#ifdef INET6 4113 /* Setup IPv6 fw sysctl tree. */ 4114 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4115 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4116 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4117 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4118 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4119 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4120 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4121 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4122 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4123 &fw_deny_unknown_exthdrs, 0, 4124 "Deny packets with unknown IPv6 Extension Headers"); 4125#endif 4126 4127 layer3_chain.rules = NULL; 4128 IPFW_LOCK_INIT(&layer3_chain); 4129 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4130 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4131 UMA_ALIGN_PTR, 0); 4132 IPFW_DYN_LOCK_INIT(); 4133 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4134 4135 bzero(&default_rule, sizeof default_rule); 4136 4137 default_rule.act_ofs = 0; 4138 default_rule.rulenum = IPFW_DEFAULT_RULE; 4139 default_rule.cmd_len = 1; 4140 default_rule.set = RESVD_SET; 4141 4142 default_rule.cmd[0].len = 1; 4143 default_rule.cmd[0].opcode = 4144#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4145 1 ? O_ACCEPT : 4146#endif 4147 O_DENY; 4148 4149 error = add_rule(&layer3_chain, &default_rule); 4150 if (error != 0) { 4151 printf("ipfw2: error %u initializing default rule " 4152 "(support disabled)\n", error); 4153 IPFW_DYN_LOCK_DESTROY(); 4154 IPFW_LOCK_DESTROY(&layer3_chain); 4155 uma_zdestroy(ipfw_dyn_rule_zone); 4156 return (error); 4157 } 4158 4159 ip_fw_default_rule = layer3_chain.rules; 4160 printf("ipfw2 " 4161#ifdef INET6 4162 "(+ipv6) " 4163#endif 4164 "initialized, divert %s, " 4165 "rule-based forwarding " 4166#ifdef IPFIREWALL_FORWARD 4167 "enabled, " 4168#else 4169 "disabled, " 4170#endif 4171 "default to %s, logging ", 4172#ifdef IPDIVERT 4173 "enabled", 4174#else 4175 "loadable", 4176#endif 4177 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4178 4179#ifdef IPFIREWALL_VERBOSE 4180 fw_verbose = 1; 4181#endif 4182#ifdef IPFIREWALL_VERBOSE_LIMIT 4183 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4184#endif 4185 if (fw_verbose == 0) 4186 printf("disabled\n"); 4187 else if (verbose_limit == 0) 4188 printf("unlimited\n"); 4189 else 4190 printf("limited to %d packets/entry by default\n", 4191 verbose_limit); 4192 4193 error = init_tables(&layer3_chain); 4194 if (error) { 4195 IPFW_DYN_LOCK_DESTROY(); 4196 IPFW_LOCK_DESTROY(&layer3_chain); 4197 uma_zdestroy(ipfw_dyn_rule_zone); 4198 return (error); 4199 } 4200 ip_fw_ctl_ptr = ipfw_ctl; 4201 ip_fw_chk_ptr = ipfw_chk; 4202 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4203 4204 return (0); 4205} 4206 4207void 4208ipfw_destroy(void) 4209{ 4210 struct ip_fw *reap; 4211 4212 ip_fw_chk_ptr = NULL; 4213 ip_fw_ctl_ptr = NULL; 4214 callout_drain(&ipfw_timeout); 4215 IPFW_WLOCK(&layer3_chain); 4216 flush_tables(&layer3_chain); 4217 layer3_chain.reap = NULL; 4218 free_chain(&layer3_chain, 1 /* kill default rule */); 4219 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4220 IPFW_WUNLOCK(&layer3_chain); 4221 if (reap != NULL) 4222 reap_rules(reap); 4223 IPFW_DYN_LOCK_DESTROY(); 4224 uma_zdestroy(ipfw_dyn_rule_zone); 4225 IPFW_LOCK_DESTROY(&layer3_chain); 4226 4227#ifdef INET6 4228 /* Free IPv6 fw sysctl tree. */ 4229 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4230#endif 4231 4232 printf("IP firewall unloaded\n"); 4233} 4234