ip_fw2.c revision 164033
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 164033 2006-11-06 13:42:10Z rwatson $ 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#include "opt_mac.h" 46 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/condvar.h> 50#include <sys/malloc.h> 51#include <sys/mbuf.h> 52#include <sys/kernel.h> 53#include <sys/lock.h> 54#include <sys/jail.h> 55#include <sys/module.h> 56#include <sys/priv.h> 57#include <sys/proc.h> 58#include <sys/rwlock.h> 59#include <sys/socket.h> 60#include <sys/socketvar.h> 61#include <sys/sysctl.h> 62#include <sys/syslog.h> 63#include <sys/ucred.h> 64#include <net/if.h> 65#include <net/radix.h> 66#include <net/route.h> 67#include <netinet/in.h> 68#include <netinet/in_systm.h> 69#include <netinet/in_var.h> 70#include <netinet/in_pcb.h> 71#include <netinet/ip.h> 72#include <netinet/ip_var.h> 73#include <netinet/ip_icmp.h> 74#include <netinet/ip_fw.h> 75#include <netinet/ip_divert.h> 76#include <netinet/ip_dummynet.h> 77#include <netinet/ip_carp.h> 78#include <netinet/pim.h> 79#include <netinet/tcp.h> 80#include <netinet/tcp_timer.h> 81#include <netinet/tcp_var.h> 82#include <netinet/tcpip.h> 83#include <netinet/udp.h> 84#include <netinet/udp_var.h> 85 86#include <netgraph/ng_ipfw.h> 87 88#include <altq/if_altq.h> 89 90#ifdef IPSEC 91#include <netinet6/ipsec.h> 92#endif 93 94#include <netinet/ip6.h> 95#include <netinet/icmp6.h> 96#ifdef INET6 97#include <netinet6/scope6_var.h> 98#endif 99 100#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 101 102#include <machine/in_cksum.h> /* XXX for in_cksum */ 103 104#include <security/mac/mac_framework.h> 105 106/* 107 * set_disable contains one bit per set value (0..31). 108 * If the bit is set, all rules with the corresponding set 109 * are disabled. Set RESVD_SET(31) is reserved for the default rule 110 * and rules that are not deleted by the flush command, 111 * and CANNOT be disabled. 112 * Rules in set RESVD_SET can only be deleted explicitly. 113 */ 114static u_int32_t set_disable; 115 116static int fw_verbose; 117static int verbose_limit; 118 119static struct callout ipfw_timeout; 120static uma_zone_t ipfw_dyn_rule_zone; 121#define IPFW_DEFAULT_RULE 65535 122 123/* 124 * Data structure to cache our ucred related 125 * information. This structure only gets used if 126 * the user specified UID/GID based constraints in 127 * a firewall rule. 128 */ 129struct ip_fw_ugid { 130 gid_t fw_groups[NGROUPS]; 131 int fw_ngroups; 132 uid_t fw_uid; 133 int fw_prid; 134}; 135 136#define IPFW_TABLES_MAX 128 137struct ip_fw_chain { 138 struct ip_fw *rules; /* list of rules */ 139 struct ip_fw *reap; /* list of rules to reap */ 140 struct radix_node_head *tables[IPFW_TABLES_MAX]; 141 struct rwlock rwmtx; 142}; 143#define IPFW_LOCK_INIT(_chain) \ 144 rw_init(&(_chain)->rwmtx, "IPFW static rules") 145#define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx) 146#define IPFW_WLOCK_ASSERT(_chain) do { \ 147 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \ 148 NET_ASSERT_GIANT(); \ 149} while (0) 150 151#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 152#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 153#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 154#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 155 156/* 157 * list of rules for layer 3 158 */ 159static struct ip_fw_chain layer3_chain; 160 161MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 162MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 163 164struct table_entry { 165 struct radix_node rn[2]; 166 struct sockaddr_in addr, mask; 167 u_int32_t value; 168}; 169 170static int fw_debug = 1; 171static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 172 173extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 174 175#ifdef SYSCTL_NODE 176SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 177SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 178 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 179 ipfw_chg_hook, "I", "Enable ipfw"); 180SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 181 &autoinc_step, 0, "Rule number autincrement step"); 182SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 183 CTLFLAG_RW | CTLFLAG_SECURE3, 184 &fw_one_pass, 0, 185 "Only do a single pass through ipfw when using dummynet(4)"); 186SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 187 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 188SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 189 CTLFLAG_RW | CTLFLAG_SECURE3, 190 &fw_verbose, 0, "Log matches to ipfw rules"); 191SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 192 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 193 194/* 195 * Description of dynamic rules. 196 * 197 * Dynamic rules are stored in lists accessed through a hash table 198 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 199 * be modified through the sysctl variable dyn_buckets which is 200 * updated when the table becomes empty. 201 * 202 * XXX currently there is only one list, ipfw_dyn. 203 * 204 * When a packet is received, its address fields are first masked 205 * with the mask defined for the rule, then hashed, then matched 206 * against the entries in the corresponding list. 207 * Dynamic rules can be used for different purposes: 208 * + stateful rules; 209 * + enforcing limits on the number of sessions; 210 * + in-kernel NAT (not implemented yet) 211 * 212 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 213 * measured in seconds and depending on the flags. 214 * 215 * The total number of dynamic rules is stored in dyn_count. 216 * The max number of dynamic rules is dyn_max. When we reach 217 * the maximum number of rules we do not create anymore. This is 218 * done to avoid consuming too much memory, but also too much 219 * time when searching on each packet (ideally, we should try instead 220 * to put a limit on the length of the list on each bucket...). 221 * 222 * Each dynamic rule holds a pointer to the parent ipfw rule so 223 * we know what action to perform. Dynamic rules are removed when 224 * the parent rule is deleted. XXX we should make them survive. 225 * 226 * There are some limitations with dynamic rules -- we do not 227 * obey the 'randomized match', and we do not do multiple 228 * passes through the firewall. XXX check the latter!!! 229 */ 230static ipfw_dyn_rule **ipfw_dyn_v = NULL; 231static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 232static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 233 234static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 235#define IPFW_DYN_LOCK_INIT() \ 236 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 237#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 238#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 239#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 240#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 241 242/* 243 * Timeouts for various events in handing dynamic rules. 244 */ 245static u_int32_t dyn_ack_lifetime = 300; 246static u_int32_t dyn_syn_lifetime = 20; 247static u_int32_t dyn_fin_lifetime = 1; 248static u_int32_t dyn_rst_lifetime = 1; 249static u_int32_t dyn_udp_lifetime = 10; 250static u_int32_t dyn_short_lifetime = 5; 251 252/* 253 * Keepalives are sent if dyn_keepalive is set. They are sent every 254 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 255 * seconds of lifetime of a rule. 256 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 257 * than dyn_keepalive_period. 258 */ 259 260static u_int32_t dyn_keepalive_interval = 20; 261static u_int32_t dyn_keepalive_period = 5; 262static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 263 264static u_int32_t static_count; /* # of static rules */ 265static u_int32_t static_len; /* size in bytes of static rules */ 266static u_int32_t dyn_count; /* # of dynamic rules */ 267static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 268 269SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 270 &dyn_buckets, 0, "Number of dyn. buckets"); 271SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 272 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 273SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 274 &dyn_count, 0, "Number of dyn. rules"); 275SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 276 &dyn_max, 0, "Max number of dyn. rules"); 277SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 278 &static_count, 0, "Number of static rules"); 279SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 280 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 281SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 282 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 283SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 284 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 285SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 286 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 287SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 288 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 289SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 290 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 291SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 292 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 293 294#ifdef INET6 295/* 296 * IPv6 specific variables 297 */ 298SYSCTL_DECL(_net_inet6_ip6); 299 300static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 301static struct sysctl_oid *ip6_fw_sysctl_tree; 302#endif /* INET6 */ 303#endif /* SYSCTL_NODE */ 304 305static int fw_deny_unknown_exthdrs = 1; 306 307 308/* 309 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 310 * Other macros just cast void * into the appropriate type 311 */ 312#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 313#define TCP(p) ((struct tcphdr *)(p)) 314#define UDP(p) ((struct udphdr *)(p)) 315#define ICMP(p) ((struct icmphdr *)(p)) 316#define ICMP6(p) ((struct icmp6_hdr *)(p)) 317 318static __inline int 319icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 320{ 321 int type = icmp->icmp_type; 322 323 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 324} 325 326#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 327 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 328 329static int 330is_icmp_query(struct icmphdr *icmp) 331{ 332 int type = icmp->icmp_type; 333 334 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 335} 336#undef TT 337 338/* 339 * The following checks use two arrays of 8 or 16 bits to store the 340 * bits that we want set or clear, respectively. They are in the 341 * low and high half of cmd->arg1 or cmd->d[0]. 342 * 343 * We scan options and store the bits we find set. We succeed if 344 * 345 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 346 * 347 * The code is sometimes optimized not to store additional variables. 348 */ 349 350static int 351flags_match(ipfw_insn *cmd, u_int8_t bits) 352{ 353 u_char want_clear; 354 bits = ~bits; 355 356 if ( ((cmd->arg1 & 0xff) & bits) != 0) 357 return 0; /* some bits we want set were clear */ 358 want_clear = (cmd->arg1 >> 8) & 0xff; 359 if ( (want_clear & bits) != want_clear) 360 return 0; /* some bits we want clear were set */ 361 return 1; 362} 363 364static int 365ipopts_match(struct ip *ip, ipfw_insn *cmd) 366{ 367 int optlen, bits = 0; 368 u_char *cp = (u_char *)(ip + 1); 369 int x = (ip->ip_hl << 2) - sizeof (struct ip); 370 371 for (; x > 0; x -= optlen, cp += optlen) { 372 int opt = cp[IPOPT_OPTVAL]; 373 374 if (opt == IPOPT_EOL) 375 break; 376 if (opt == IPOPT_NOP) 377 optlen = 1; 378 else { 379 optlen = cp[IPOPT_OLEN]; 380 if (optlen <= 0 || optlen > x) 381 return 0; /* invalid or truncated */ 382 } 383 switch (opt) { 384 385 default: 386 break; 387 388 case IPOPT_LSRR: 389 bits |= IP_FW_IPOPT_LSRR; 390 break; 391 392 case IPOPT_SSRR: 393 bits |= IP_FW_IPOPT_SSRR; 394 break; 395 396 case IPOPT_RR: 397 bits |= IP_FW_IPOPT_RR; 398 break; 399 400 case IPOPT_TS: 401 bits |= IP_FW_IPOPT_TS; 402 break; 403 } 404 } 405 return (flags_match(cmd, bits)); 406} 407 408static int 409tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 410{ 411 int optlen, bits = 0; 412 u_char *cp = (u_char *)(tcp + 1); 413 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 414 415 for (; x > 0; x -= optlen, cp += optlen) { 416 int opt = cp[0]; 417 if (opt == TCPOPT_EOL) 418 break; 419 if (opt == TCPOPT_NOP) 420 optlen = 1; 421 else { 422 optlen = cp[1]; 423 if (optlen <= 0) 424 break; 425 } 426 427 switch (opt) { 428 429 default: 430 break; 431 432 case TCPOPT_MAXSEG: 433 bits |= IP_FW_TCPOPT_MSS; 434 break; 435 436 case TCPOPT_WINDOW: 437 bits |= IP_FW_TCPOPT_WINDOW; 438 break; 439 440 case TCPOPT_SACK_PERMITTED: 441 case TCPOPT_SACK: 442 bits |= IP_FW_TCPOPT_SACK; 443 break; 444 445 case TCPOPT_TIMESTAMP: 446 bits |= IP_FW_TCPOPT_TS; 447 break; 448 449 } 450 } 451 return (flags_match(cmd, bits)); 452} 453 454static int 455iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 456{ 457 if (ifp == NULL) /* no iface with this packet, match fails */ 458 return 0; 459 /* Check by name or by IP address */ 460 if (cmd->name[0] != '\0') { /* match by name */ 461 /* Check name */ 462 if (cmd->p.glob) { 463 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 464 return(1); 465 } else { 466 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 467 return(1); 468 } 469 } else { 470 struct ifaddr *ia; 471 472 /* XXX lock? */ 473 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 474 if (ia->ifa_addr->sa_family != AF_INET) 475 continue; 476 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 477 (ia->ifa_addr))->sin_addr.s_addr) 478 return(1); /* match */ 479 } 480 } 481 return(0); /* no match, fail ... */ 482} 483 484/* 485 * The verify_path function checks if a route to the src exists and 486 * if it is reachable via ifp (when provided). 487 * 488 * The 'verrevpath' option checks that the interface that an IP packet 489 * arrives on is the same interface that traffic destined for the 490 * packet's source address would be routed out of. The 'versrcreach' 491 * option just checks that the source address is reachable via any route 492 * (except default) in the routing table. These two are a measure to block 493 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 494 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 495 * is purposely reminiscent of the Cisco IOS command, 496 * 497 * ip verify unicast reverse-path 498 * ip verify unicast source reachable-via any 499 * 500 * which implements the same functionality. But note that syntax is 501 * misleading. The check may be performed on all IP packets whether unicast, 502 * multicast, or broadcast. 503 */ 504static int 505verify_path(struct in_addr src, struct ifnet *ifp) 506{ 507 struct route ro; 508 struct sockaddr_in *dst; 509 510 bzero(&ro, sizeof(ro)); 511 512 dst = (struct sockaddr_in *)&(ro.ro_dst); 513 dst->sin_family = AF_INET; 514 dst->sin_len = sizeof(*dst); 515 dst->sin_addr = src; 516 rtalloc_ign(&ro, RTF_CLONING); 517 518 if (ro.ro_rt == NULL) 519 return 0; 520 521 /* 522 * If ifp is provided, check for equality with rtentry. 523 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 524 * in order to pass packets injected back by if_simloop(): 525 * if useloopback == 1 routing entry (via lo0) for our own address 526 * may exist, so we need to handle routing assymetry. 527 */ 528 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 529 RTFREE(ro.ro_rt); 530 return 0; 531 } 532 533 /* if no ifp provided, check if rtentry is not default route */ 534 if (ifp == NULL && 535 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 536 RTFREE(ro.ro_rt); 537 return 0; 538 } 539 540 /* or if this is a blackhole/reject route */ 541 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 542 RTFREE(ro.ro_rt); 543 return 0; 544 } 545 546 /* found valid route */ 547 RTFREE(ro.ro_rt); 548 return 1; 549} 550 551#ifdef INET6 552/* 553 * ipv6 specific rules here... 554 */ 555static __inline int 556icmp6type_match (int type, ipfw_insn_u32 *cmd) 557{ 558 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 559} 560 561static int 562flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 563{ 564 int i; 565 for (i=0; i <= cmd->o.arg1; ++i ) 566 if (curr_flow == cmd->d[i] ) 567 return 1; 568 return 0; 569} 570 571/* support for IP6_*_ME opcodes */ 572static int 573search_ip6_addr_net (struct in6_addr * ip6_addr) 574{ 575 struct ifnet *mdc; 576 struct ifaddr *mdc2; 577 struct in6_ifaddr *fdm; 578 struct in6_addr copia; 579 580 TAILQ_FOREACH(mdc, &ifnet, if_link) 581 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 582 if (mdc2->ifa_addr->sa_family == AF_INET6) { 583 fdm = (struct in6_ifaddr *)mdc2; 584 copia = fdm->ia_addr.sin6_addr; 585 /* need for leaving scope_id in the sock_addr */ 586 in6_clearscope(&copia); 587 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 588 return 1; 589 } 590 } 591 return 0; 592} 593 594static int 595verify_path6(struct in6_addr *src, struct ifnet *ifp) 596{ 597 struct route_in6 ro; 598 struct sockaddr_in6 *dst; 599 600 bzero(&ro, sizeof(ro)); 601 602 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 603 dst->sin6_family = AF_INET6; 604 dst->sin6_len = sizeof(*dst); 605 dst->sin6_addr = *src; 606 rtalloc_ign((struct route *)&ro, RTF_CLONING); 607 608 if (ro.ro_rt == NULL) 609 return 0; 610 611 /* 612 * if ifp is provided, check for equality with rtentry 613 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 614 * to support the case of sending packets to an address of our own. 615 * (where the former interface is the first argument of if_simloop() 616 * (=ifp), the latter is lo0) 617 */ 618 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 619 RTFREE(ro.ro_rt); 620 return 0; 621 } 622 623 /* if no ifp provided, check if rtentry is not default route */ 624 if (ifp == NULL && 625 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 626 RTFREE(ro.ro_rt); 627 return 0; 628 } 629 630 /* or if this is a blackhole/reject route */ 631 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 632 RTFREE(ro.ro_rt); 633 return 0; 634 } 635 636 /* found valid route */ 637 RTFREE(ro.ro_rt); 638 return 1; 639 640} 641static __inline int 642hash_packet6(struct ipfw_flow_id *id) 643{ 644 u_int32_t i; 645 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 646 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 647 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 648 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 649 (id->dst_port) ^ (id->src_port); 650 return i; 651} 652 653static int 654is_icmp6_query(int icmp6_type) 655{ 656 if ((icmp6_type <= ICMP6_MAXTYPE) && 657 (icmp6_type == ICMP6_ECHO_REQUEST || 658 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 659 icmp6_type == ICMP6_WRUREQUEST || 660 icmp6_type == ICMP6_FQDN_QUERY || 661 icmp6_type == ICMP6_NI_QUERY)) 662 return (1); 663 664 return (0); 665} 666 667static void 668send_reject6(struct ip_fw_args *args, int code, u_int hlen) 669{ 670 if (code == ICMP6_UNREACH_RST && 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, uint32_t tablearg) 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 if (cmd->opcode == O_TAG) 785 cmd += F_LEN(cmd); 786 787 action = action2; 788 switch (cmd->opcode) { 789 case O_DENY: 790 action = "Deny"; 791 break; 792 793 case O_REJECT: 794 if (cmd->arg1==ICMP_REJECT_RST) 795 action = "Reset"; 796 else if (cmd->arg1==ICMP_UNREACH_HOST) 797 action = "Reject"; 798 else 799 snprintf(SNPARGS(action2, 0), "Unreach %d", 800 cmd->arg1); 801 break; 802 803 case O_UNREACH6: 804 if (cmd->arg1==ICMP6_UNREACH_RST) 805 action = "Reset"; 806 else 807 snprintf(SNPARGS(action2, 0), "Unreach %d", 808 cmd->arg1); 809 break; 810 811 case O_ACCEPT: 812 action = "Accept"; 813 break; 814 case O_COUNT: 815 action = "Count"; 816 break; 817 case O_DIVERT: 818 snprintf(SNPARGS(action2, 0), "Divert %d", 819 cmd->arg1); 820 break; 821 case O_TEE: 822 snprintf(SNPARGS(action2, 0), "Tee %d", 823 cmd->arg1); 824 break; 825 case O_SKIPTO: 826 snprintf(SNPARGS(action2, 0), "SkipTo %d", 827 cmd->arg1); 828 break; 829 case O_PIPE: 830 snprintf(SNPARGS(action2, 0), "Pipe %d", 831 cmd->arg1); 832 break; 833 case O_QUEUE: 834 snprintf(SNPARGS(action2, 0), "Queue %d", 835 cmd->arg1); 836 break; 837 case O_FORWARD_IP: { 838 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 839 int len; 840 struct in_addr dummyaddr; 841 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 842 dummyaddr.s_addr = htonl(tablearg); 843 else 844 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 845 846 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 847 inet_ntoa(dummyaddr)); 848 849 if (sa->sa.sin_port) 850 snprintf(SNPARGS(action2, len), ":%d", 851 sa->sa.sin_port); 852 } 853 break; 854 case O_NETGRAPH: 855 snprintf(SNPARGS(action2, 0), "Netgraph %d", 856 cmd->arg1); 857 break; 858 case O_NGTEE: 859 snprintf(SNPARGS(action2, 0), "Ngtee %d", 860 cmd->arg1); 861 break; 862 default: 863 action = "UNKNOWN"; 864 break; 865 } 866 } 867 868 if (hlen == 0) { /* non-ip */ 869 snprintf(SNPARGS(proto, 0), "MAC"); 870 871 } else { 872 int len; 873 char src[48], dst[48]; 874 struct icmphdr *icmp; 875 struct tcphdr *tcp; 876 struct udphdr *udp; 877 /* Initialize to make compiler happy. */ 878 struct ip *ip = NULL; 879#ifdef INET6 880 struct ip6_hdr *ip6 = NULL; 881 struct icmp6_hdr *icmp6; 882#endif 883 src[0] = '\0'; 884 dst[0] = '\0'; 885#ifdef INET6 886 if (IS_IP6_FLOW_ID(&(args->f_id))) { 887 snprintf(src, sizeof(src), "[%s]", 888 ip6_sprintf(&args->f_id.src_ip6)); 889 snprintf(dst, sizeof(dst), "[%s]", 890 ip6_sprintf(&args->f_id.dst_ip6)); 891 892 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 893 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 894 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 895 } else 896#endif 897 { 898 ip = mtod(m, struct ip *); 899 tcp = L3HDR(struct tcphdr, ip); 900 udp = L3HDR(struct udphdr, ip); 901 902 inet_ntoa_r(ip->ip_src, src); 903 inet_ntoa_r(ip->ip_dst, dst); 904 } 905 906 switch (args->f_id.proto) { 907 case IPPROTO_TCP: 908 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 909 if (offset == 0) 910 snprintf(SNPARGS(proto, len), ":%d %s:%d", 911 ntohs(tcp->th_sport), 912 dst, 913 ntohs(tcp->th_dport)); 914 else 915 snprintf(SNPARGS(proto, len), " %s", dst); 916 break; 917 918 case IPPROTO_UDP: 919 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 920 if (offset == 0) 921 snprintf(SNPARGS(proto, len), ":%d %s:%d", 922 ntohs(udp->uh_sport), 923 dst, 924 ntohs(udp->uh_dport)); 925 else 926 snprintf(SNPARGS(proto, len), " %s", dst); 927 break; 928 929 case IPPROTO_ICMP: 930 icmp = L3HDR(struct icmphdr, ip); 931 if (offset == 0) 932 len = snprintf(SNPARGS(proto, 0), 933 "ICMP:%u.%u ", 934 icmp->icmp_type, icmp->icmp_code); 935 else 936 len = snprintf(SNPARGS(proto, 0), "ICMP "); 937 len += snprintf(SNPARGS(proto, len), "%s", src); 938 snprintf(SNPARGS(proto, len), " %s", dst); 939 break; 940#ifdef INET6 941 case IPPROTO_ICMPV6: 942 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 943 if (offset == 0) 944 len = snprintf(SNPARGS(proto, 0), 945 "ICMPv6:%u.%u ", 946 icmp6->icmp6_type, icmp6->icmp6_code); 947 else 948 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 949 len += snprintf(SNPARGS(proto, len), "%s", src); 950 snprintf(SNPARGS(proto, len), " %s", dst); 951 break; 952#endif 953 default: 954 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 955 args->f_id.proto, src); 956 snprintf(SNPARGS(proto, len), " %s", dst); 957 break; 958 } 959 960#ifdef INET6 961 if (IS_IP6_FLOW_ID(&(args->f_id))) { 962 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 963 snprintf(SNPARGS(fragment, 0), 964 " (frag %08x:%d@%d%s)", 965 args->f_id.frag_id6, 966 ntohs(ip6->ip6_plen) - hlen, 967 ntohs(offset & IP6F_OFF_MASK) << 3, 968 (offset & IP6F_MORE_FRAG) ? "+" : ""); 969 } else 970#endif 971 { 972 int ip_off, ip_len; 973 if (eh != NULL) { /* layer 2 packets are as on the wire */ 974 ip_off = ntohs(ip->ip_off); 975 ip_len = ntohs(ip->ip_len); 976 } else { 977 ip_off = ip->ip_off; 978 ip_len = ip->ip_len; 979 } 980 if (ip_off & (IP_MF | IP_OFFMASK)) 981 snprintf(SNPARGS(fragment, 0), 982 " (frag %d:%d@%d%s)", 983 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 984 offset << 3, 985 (ip_off & IP_MF) ? "+" : ""); 986 } 987 } 988 if (oif || m->m_pkthdr.rcvif) 989 log(LOG_SECURITY | LOG_INFO, 990 "ipfw: %d %s %s %s via %s%s\n", 991 f ? f->rulenum : -1, 992 action, proto, oif ? "out" : "in", 993 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 994 fragment); 995 else 996 log(LOG_SECURITY | LOG_INFO, 997 "ipfw: %d %s %s [no if info]%s\n", 998 f ? f->rulenum : -1, 999 action, proto, fragment); 1000 if (limit_reached) 1001 log(LOG_SECURITY | LOG_NOTICE, 1002 "ipfw: limit %d reached on entry %d\n", 1003 limit_reached, f ? f->rulenum : -1); 1004} 1005 1006/* 1007 * IMPORTANT: the hash function for dynamic rules must be commutative 1008 * in source and destination (ip,port), because rules are bidirectional 1009 * and we want to find both in the same bucket. 1010 */ 1011static __inline int 1012hash_packet(struct ipfw_flow_id *id) 1013{ 1014 u_int32_t i; 1015 1016#ifdef INET6 1017 if (IS_IP6_FLOW_ID(id)) 1018 i = hash_packet6(id); 1019 else 1020#endif /* INET6 */ 1021 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1022 i &= (curr_dyn_buckets - 1); 1023 return i; 1024} 1025 1026/** 1027 * unlink a dynamic rule from a chain. prev is a pointer to 1028 * the previous one, q is a pointer to the rule to delete, 1029 * head is a pointer to the head of the queue. 1030 * Modifies q and potentially also head. 1031 */ 1032#define UNLINK_DYN_RULE(prev, head, q) { \ 1033 ipfw_dyn_rule *old_q = q; \ 1034 \ 1035 /* remove a refcount to the parent */ \ 1036 if (q->dyn_type == O_LIMIT) \ 1037 q->parent->count--; \ 1038 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1039 (q->id.src_ip), (q->id.src_port), \ 1040 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1041 if (prev != NULL) \ 1042 prev->next = q = q->next; \ 1043 else \ 1044 head = q = q->next; \ 1045 dyn_count--; \ 1046 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1047 1048#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1049 1050/** 1051 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1052 * 1053 * If keep_me == NULL, rules are deleted even if not expired, 1054 * otherwise only expired rules are removed. 1055 * 1056 * The value of the second parameter is also used to point to identify 1057 * a rule we absolutely do not want to remove (e.g. because we are 1058 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1059 * rules). The pointer is only used for comparison, so any non-null 1060 * value will do. 1061 */ 1062static void 1063remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1064{ 1065 static u_int32_t last_remove = 0; 1066 1067#define FORCE (keep_me == NULL) 1068 1069 ipfw_dyn_rule *prev, *q; 1070 int i, pass = 0, max_pass = 0; 1071 1072 IPFW_DYN_LOCK_ASSERT(); 1073 1074 if (ipfw_dyn_v == NULL || dyn_count == 0) 1075 return; 1076 /* do not expire more than once per second, it is useless */ 1077 if (!FORCE && last_remove == time_uptime) 1078 return; 1079 last_remove = time_uptime; 1080 1081 /* 1082 * because O_LIMIT refer to parent rules, during the first pass only 1083 * remove child and mark any pending LIMIT_PARENT, and remove 1084 * them in a second pass. 1085 */ 1086next_pass: 1087 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1088 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1089 /* 1090 * Logic can become complex here, so we split tests. 1091 */ 1092 if (q == keep_me) 1093 goto next; 1094 if (rule != NULL && rule != q->rule) 1095 goto next; /* not the one we are looking for */ 1096 if (q->dyn_type == O_LIMIT_PARENT) { 1097 /* 1098 * handle parent in the second pass, 1099 * record we need one. 1100 */ 1101 max_pass = 1; 1102 if (pass == 0) 1103 goto next; 1104 if (FORCE && q->count != 0 ) { 1105 /* XXX should not happen! */ 1106 printf("ipfw: OUCH! cannot remove rule," 1107 " count %d\n", q->count); 1108 } 1109 } else { 1110 if (!FORCE && 1111 !TIME_LEQ( q->expire, time_uptime )) 1112 goto next; 1113 } 1114 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1115 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1116 continue; 1117 } 1118next: 1119 prev=q; 1120 q=q->next; 1121 } 1122 } 1123 if (pass++ < max_pass) 1124 goto next_pass; 1125} 1126 1127 1128/** 1129 * lookup a dynamic rule. 1130 */ 1131static ipfw_dyn_rule * 1132lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1133 struct tcphdr *tcp) 1134{ 1135 /* 1136 * stateful ipfw extensions. 1137 * Lookup into dynamic session queue 1138 */ 1139#define MATCH_REVERSE 0 1140#define MATCH_FORWARD 1 1141#define MATCH_NONE 2 1142#define MATCH_UNKNOWN 3 1143 int i, dir = MATCH_NONE; 1144 ipfw_dyn_rule *prev, *q=NULL; 1145 1146 IPFW_DYN_LOCK_ASSERT(); 1147 1148 if (ipfw_dyn_v == NULL) 1149 goto done; /* not found */ 1150 i = hash_packet( pkt ); 1151 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1152 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1153 goto next; 1154 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1155 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1156 continue; 1157 } 1158 if (pkt->proto == q->id.proto && 1159 q->dyn_type != O_LIMIT_PARENT) { 1160 if (IS_IP6_FLOW_ID(pkt)) { 1161 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1162 &(q->id.src_ip6)) && 1163 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1164 &(q->id.dst_ip6)) && 1165 pkt->src_port == q->id.src_port && 1166 pkt->dst_port == q->id.dst_port ) { 1167 dir = MATCH_FORWARD; 1168 break; 1169 } 1170 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1171 &(q->id.dst_ip6)) && 1172 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1173 &(q->id.src_ip6)) && 1174 pkt->src_port == q->id.dst_port && 1175 pkt->dst_port == q->id.src_port ) { 1176 dir = MATCH_REVERSE; 1177 break; 1178 } 1179 } else { 1180 if (pkt->src_ip == q->id.src_ip && 1181 pkt->dst_ip == q->id.dst_ip && 1182 pkt->src_port == q->id.src_port && 1183 pkt->dst_port == q->id.dst_port ) { 1184 dir = MATCH_FORWARD; 1185 break; 1186 } 1187 if (pkt->src_ip == q->id.dst_ip && 1188 pkt->dst_ip == q->id.src_ip && 1189 pkt->src_port == q->id.dst_port && 1190 pkt->dst_port == q->id.src_port ) { 1191 dir = MATCH_REVERSE; 1192 break; 1193 } 1194 } 1195 } 1196next: 1197 prev = q; 1198 q = q->next; 1199 } 1200 if (q == NULL) 1201 goto done; /* q = NULL, not found */ 1202 1203 if ( prev != NULL) { /* found and not in front */ 1204 prev->next = q->next; 1205 q->next = ipfw_dyn_v[i]; 1206 ipfw_dyn_v[i] = q; 1207 } 1208 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1209 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1210 1211#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1212#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1213 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1214 switch (q->state) { 1215 case TH_SYN: /* opening */ 1216 q->expire = time_uptime + dyn_syn_lifetime; 1217 break; 1218 1219 case BOTH_SYN: /* move to established */ 1220 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1221 case BOTH_SYN | (TH_FIN << 8) : 1222 if (tcp) { 1223#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1224 u_int32_t ack = ntohl(tcp->th_ack); 1225 if (dir == MATCH_FORWARD) { 1226 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1227 q->ack_fwd = ack; 1228 else { /* ignore out-of-sequence */ 1229 break; 1230 } 1231 } else { 1232 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1233 q->ack_rev = ack; 1234 else { /* ignore out-of-sequence */ 1235 break; 1236 } 1237 } 1238 } 1239 q->expire = time_uptime + dyn_ack_lifetime; 1240 break; 1241 1242 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1243 if (dyn_fin_lifetime >= dyn_keepalive_period) 1244 dyn_fin_lifetime = dyn_keepalive_period - 1; 1245 q->expire = time_uptime + dyn_fin_lifetime; 1246 break; 1247 1248 default: 1249#if 0 1250 /* 1251 * reset or some invalid combination, but can also 1252 * occur if we use keep-state the wrong way. 1253 */ 1254 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1255 printf("invalid state: 0x%x\n", q->state); 1256#endif 1257 if (dyn_rst_lifetime >= dyn_keepalive_period) 1258 dyn_rst_lifetime = dyn_keepalive_period - 1; 1259 q->expire = time_uptime + dyn_rst_lifetime; 1260 break; 1261 } 1262 } else if (pkt->proto == IPPROTO_UDP) { 1263 q->expire = time_uptime + dyn_udp_lifetime; 1264 } else { 1265 /* other protocols */ 1266 q->expire = time_uptime + dyn_short_lifetime; 1267 } 1268done: 1269 if (match_direction) 1270 *match_direction = dir; 1271 return q; 1272} 1273 1274static ipfw_dyn_rule * 1275lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1276 struct tcphdr *tcp) 1277{ 1278 ipfw_dyn_rule *q; 1279 1280 IPFW_DYN_LOCK(); 1281 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1282 if (q == NULL) 1283 IPFW_DYN_UNLOCK(); 1284 /* NB: return table locked when q is not NULL */ 1285 return q; 1286} 1287 1288static void 1289realloc_dynamic_table(void) 1290{ 1291 IPFW_DYN_LOCK_ASSERT(); 1292 1293 /* 1294 * Try reallocation, make sure we have a power of 2 and do 1295 * not allow more than 64k entries. In case of overflow, 1296 * default to 1024. 1297 */ 1298 1299 if (dyn_buckets > 65536) 1300 dyn_buckets = 1024; 1301 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1302 dyn_buckets = curr_dyn_buckets; /* reset */ 1303 return; 1304 } 1305 curr_dyn_buckets = dyn_buckets; 1306 if (ipfw_dyn_v != NULL) 1307 free(ipfw_dyn_v, M_IPFW); 1308 for (;;) { 1309 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1310 M_IPFW, M_NOWAIT | M_ZERO); 1311 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1312 break; 1313 curr_dyn_buckets /= 2; 1314 } 1315} 1316 1317/** 1318 * Install state of type 'type' for a dynamic session. 1319 * The hash table contains two type of rules: 1320 * - regular rules (O_KEEP_STATE) 1321 * - rules for sessions with limited number of sess per user 1322 * (O_LIMIT). When they are created, the parent is 1323 * increased by 1, and decreased on delete. In this case, 1324 * the third parameter is the parent rule and not the chain. 1325 * - "parent" rules for the above (O_LIMIT_PARENT). 1326 */ 1327static ipfw_dyn_rule * 1328add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1329{ 1330 ipfw_dyn_rule *r; 1331 int i; 1332 1333 IPFW_DYN_LOCK_ASSERT(); 1334 1335 if (ipfw_dyn_v == NULL || 1336 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1337 realloc_dynamic_table(); 1338 if (ipfw_dyn_v == NULL) 1339 return NULL; /* failed ! */ 1340 } 1341 i = hash_packet(id); 1342 1343 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1344 if (r == NULL) { 1345 printf ("ipfw: sorry cannot allocate state\n"); 1346 return NULL; 1347 } 1348 1349 /* increase refcount on parent, and set pointer */ 1350 if (dyn_type == O_LIMIT) { 1351 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1352 if ( parent->dyn_type != O_LIMIT_PARENT) 1353 panic("invalid parent"); 1354 parent->count++; 1355 r->parent = parent; 1356 rule = parent->rule; 1357 } 1358 1359 r->id = *id; 1360 r->expire = time_uptime + dyn_syn_lifetime; 1361 r->rule = rule; 1362 r->dyn_type = dyn_type; 1363 r->pcnt = r->bcnt = 0; 1364 r->count = 0; 1365 1366 r->bucket = i; 1367 r->next = ipfw_dyn_v[i]; 1368 ipfw_dyn_v[i] = r; 1369 dyn_count++; 1370 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1371 dyn_type, 1372 (r->id.src_ip), (r->id.src_port), 1373 (r->id.dst_ip), (r->id.dst_port), 1374 dyn_count ); ) 1375 return r; 1376} 1377 1378/** 1379 * lookup dynamic parent rule using pkt and rule as search keys. 1380 * If the lookup fails, then install one. 1381 */ 1382static ipfw_dyn_rule * 1383lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1384{ 1385 ipfw_dyn_rule *q; 1386 int i; 1387 1388 IPFW_DYN_LOCK_ASSERT(); 1389 1390 if (ipfw_dyn_v) { 1391 int is_v6 = IS_IP6_FLOW_ID(pkt); 1392 i = hash_packet( pkt ); 1393 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1394 if (q->dyn_type == O_LIMIT_PARENT && 1395 rule== q->rule && 1396 pkt->proto == q->id.proto && 1397 pkt->src_port == q->id.src_port && 1398 pkt->dst_port == q->id.dst_port && 1399 ( 1400 (is_v6 && 1401 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1402 &(q->id.src_ip6)) && 1403 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1404 &(q->id.dst_ip6))) || 1405 (!is_v6 && 1406 pkt->src_ip == q->id.src_ip && 1407 pkt->dst_ip == q->id.dst_ip) 1408 ) 1409 ) { 1410 q->expire = time_uptime + dyn_short_lifetime; 1411 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1412 return q; 1413 } 1414 } 1415 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1416} 1417 1418/** 1419 * Install dynamic state for rule type cmd->o.opcode 1420 * 1421 * Returns 1 (failure) if state is not installed because of errors or because 1422 * session limitations are enforced. 1423 */ 1424static int 1425install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1426 struct ip_fw_args *args, uint32_t tablearg) 1427{ 1428 static int last_log; 1429 ipfw_dyn_rule *q; 1430 struct in_addr da; 1431 char src[48], dst[48]; 1432 1433 src[0] = '\0'; 1434 dst[0] = '\0'; 1435 1436 DEB( 1437 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1438 __func__, cmd->o.opcode, 1439 (args->f_id.src_ip), (args->f_id.src_port), 1440 (args->f_id.dst_ip), (args->f_id.dst_port)); 1441 ) 1442 1443 IPFW_DYN_LOCK(); 1444 1445 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1446 1447 if (q != NULL) { /* should never occur */ 1448 if (last_log != time_uptime) { 1449 last_log = time_uptime; 1450 printf("ipfw: %s: entry already present, done\n", 1451 __func__); 1452 } 1453 IPFW_DYN_UNLOCK(); 1454 return (0); 1455 } 1456 1457 if (dyn_count >= dyn_max) 1458 /* Run out of slots, try to remove any expired rule. */ 1459 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1460 1461 if (dyn_count >= dyn_max) { 1462 if (last_log != time_uptime) { 1463 last_log = time_uptime; 1464 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1465 } 1466 IPFW_DYN_UNLOCK(); 1467 return (1); /* cannot install, notify caller */ 1468 } 1469 1470 switch (cmd->o.opcode) { 1471 case O_KEEP_STATE: /* bidir rule */ 1472 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1473 break; 1474 1475 case O_LIMIT: { /* limit number of sessions */ 1476 struct ipfw_flow_id id; 1477 ipfw_dyn_rule *parent; 1478 uint32_t conn_limit; 1479 uint16_t limit_mask = cmd->limit_mask; 1480 1481 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1482 tablearg : cmd->conn_limit; 1483 1484 DEB( 1485 if (cmd->conn_limit == IP_FW_TABLEARG) 1486 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1487 "(tablearg)\n", __func__, conn_limit); 1488 else 1489 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1490 __func__, conn_limit); 1491 ) 1492 1493 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1494 id.proto = args->f_id.proto; 1495 id.addr_type = args->f_id.addr_type; 1496 1497 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1498 if (limit_mask & DYN_SRC_ADDR) 1499 id.src_ip6 = args->f_id.src_ip6; 1500 if (limit_mask & DYN_DST_ADDR) 1501 id.dst_ip6 = args->f_id.dst_ip6; 1502 } else { 1503 if (limit_mask & DYN_SRC_ADDR) 1504 id.src_ip = args->f_id.src_ip; 1505 if (limit_mask & DYN_DST_ADDR) 1506 id.dst_ip = args->f_id.dst_ip; 1507 } 1508 if (limit_mask & DYN_SRC_PORT) 1509 id.src_port = args->f_id.src_port; 1510 if (limit_mask & DYN_DST_PORT) 1511 id.dst_port = args->f_id.dst_port; 1512 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1513 printf("ipfw: %s: add parent failed\n", __func__); 1514 IPFW_DYN_UNLOCK(); 1515 return (1); 1516 } 1517 1518 if (parent->count >= conn_limit) { 1519 /* See if we can remove some expired rule. */ 1520 remove_dyn_rule(rule, parent); 1521 if (parent->count >= conn_limit) { 1522 if (fw_verbose && last_log != time_uptime) { 1523 last_log = time_uptime; 1524#ifdef INET6 1525 /* 1526 * XXX IPv6 flows are not 1527 * supported yet. 1528 */ 1529 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1530 snprintf(src, sizeof(src), 1531 "[%s]", ip6_sprintf( 1532 &args->f_id.src_ip6)); 1533 snprintf(dst, sizeof(dst), 1534 "[%s]", ip6_sprintf( 1535 &args->f_id.dst_ip6)); 1536 } else 1537#endif 1538 { 1539 da.s_addr = 1540 htonl(args->f_id.src_ip); 1541 inet_ntoa_r(da, src); 1542 da.s_addr = 1543 htonl(args->f_id.dst_ip); 1544 inet_ntoa_r(da, dst); 1545 } 1546 log(LOG_SECURITY | LOG_DEBUG, 1547 "%s %s:%u -> %s:%u, %s\n", 1548 "drop session", 1549 src, (args->f_id.src_port), 1550 dst, (args->f_id.dst_port), 1551 "too many entries"); 1552 } 1553 IPFW_DYN_UNLOCK(); 1554 return (1); 1555 } 1556 } 1557 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1558 break; 1559 } 1560 default: 1561 printf("ipfw: %s: unknown dynamic rule type %u\n", 1562 __func__, cmd->o.opcode); 1563 IPFW_DYN_UNLOCK(); 1564 return (1); 1565 } 1566 1567 /* XXX just set lifetime */ 1568 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1569 1570 IPFW_DYN_UNLOCK(); 1571 return (0); 1572} 1573 1574/* 1575 * Generate a TCP packet, containing either a RST or a keepalive. 1576 * When flags & TH_RST, we are sending a RST packet, because of a 1577 * "reset" action matched the packet. 1578 * Otherwise we are sending a keepalive, and flags & TH_ 1579 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1580 * so that MAC can label the reply appropriately. 1581 */ 1582static struct mbuf * 1583send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1584 u_int32_t ack, int flags) 1585{ 1586 struct mbuf *m; 1587 struct ip *ip; 1588 struct tcphdr *tcp; 1589 1590 MGETHDR(m, M_DONTWAIT, MT_DATA); 1591 if (m == 0) 1592 return (NULL); 1593 m->m_pkthdr.rcvif = (struct ifnet *)0; 1594 1595#ifdef MAC 1596 if (replyto != NULL) 1597 mac_create_mbuf_netlayer(replyto, m); 1598 else 1599 mac_create_mbuf_from_firewall(m); 1600#else 1601 (void)replyto; /* don't warn about unused arg */ 1602#endif 1603 1604 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1605 m->m_data += max_linkhdr; 1606 1607 ip = mtod(m, struct ip *); 1608 bzero(ip, m->m_len); 1609 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1610 ip->ip_p = IPPROTO_TCP; 1611 tcp->th_off = 5; 1612 /* 1613 * Assume we are sending a RST (or a keepalive in the reverse 1614 * direction), swap src and destination addresses and ports. 1615 */ 1616 ip->ip_src.s_addr = htonl(id->dst_ip); 1617 ip->ip_dst.s_addr = htonl(id->src_ip); 1618 tcp->th_sport = htons(id->dst_port); 1619 tcp->th_dport = htons(id->src_port); 1620 if (flags & TH_RST) { /* we are sending a RST */ 1621 if (flags & TH_ACK) { 1622 tcp->th_seq = htonl(ack); 1623 tcp->th_ack = htonl(0); 1624 tcp->th_flags = TH_RST; 1625 } else { 1626 if (flags & TH_SYN) 1627 seq++; 1628 tcp->th_seq = htonl(0); 1629 tcp->th_ack = htonl(seq); 1630 tcp->th_flags = TH_RST | TH_ACK; 1631 } 1632 } else { 1633 /* 1634 * We are sending a keepalive. flags & TH_SYN determines 1635 * the direction, forward if set, reverse if clear. 1636 * NOTE: seq and ack are always assumed to be correct 1637 * as set by the caller. This may be confusing... 1638 */ 1639 if (flags & TH_SYN) { 1640 /* 1641 * we have to rewrite the correct addresses! 1642 */ 1643 ip->ip_dst.s_addr = htonl(id->dst_ip); 1644 ip->ip_src.s_addr = htonl(id->src_ip); 1645 tcp->th_dport = htons(id->dst_port); 1646 tcp->th_sport = htons(id->src_port); 1647 } 1648 tcp->th_seq = htonl(seq); 1649 tcp->th_ack = htonl(ack); 1650 tcp->th_flags = TH_ACK; 1651 } 1652 /* 1653 * set ip_len to the payload size so we can compute 1654 * the tcp checksum on the pseudoheader 1655 * XXX check this, could save a couple of words ? 1656 */ 1657 ip->ip_len = htons(sizeof(struct tcphdr)); 1658 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1659 /* 1660 * now fill fields left out earlier 1661 */ 1662 ip->ip_ttl = ip_defttl; 1663 ip->ip_len = m->m_pkthdr.len; 1664 m->m_flags |= M_SKIP_FIREWALL; 1665 return (m); 1666} 1667 1668/* 1669 * sends a reject message, consuming the mbuf passed as an argument. 1670 */ 1671static void 1672send_reject(struct ip_fw_args *args, int code, int ip_len) 1673{ 1674 1675 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1676 /* We need the IP header in host order for icmp_error(). */ 1677 if (args->eh != NULL) { 1678 struct ip *ip = mtod(args->m, struct ip *); 1679 ip->ip_len = ntohs(ip->ip_len); 1680 ip->ip_off = ntohs(ip->ip_off); 1681 } 1682 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1683 } else if (args->f_id.proto == IPPROTO_TCP) { 1684 struct tcphdr *const tcp = 1685 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1686 if ( (tcp->th_flags & TH_RST) == 0) { 1687 struct mbuf *m; 1688 m = send_pkt(args->m, &(args->f_id), 1689 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1690 tcp->th_flags | TH_RST); 1691 if (m != NULL) 1692 ip_output(m, NULL, NULL, 0, NULL, NULL); 1693 } 1694 m_freem(args->m); 1695 } else 1696 m_freem(args->m); 1697 args->m = NULL; 1698} 1699 1700/** 1701 * 1702 * Given an ip_fw *, lookup_next_rule will return a pointer 1703 * to the next rule, which can be either the jump 1704 * target (for skipto instructions) or the next one in the list (in 1705 * all other cases including a missing jump target). 1706 * The result is also written in the "next_rule" field of the rule. 1707 * Backward jumps are not allowed, so start looking from the next 1708 * rule... 1709 * 1710 * This never returns NULL -- in case we do not have an exact match, 1711 * the next rule is returned. When the ruleset is changed, 1712 * pointers are flushed so we are always correct. 1713 */ 1714 1715static struct ip_fw * 1716lookup_next_rule(struct ip_fw *me) 1717{ 1718 struct ip_fw *rule = NULL; 1719 ipfw_insn *cmd; 1720 1721 /* look for action, in case it is a skipto */ 1722 cmd = ACTION_PTR(me); 1723 if (cmd->opcode == O_LOG) 1724 cmd += F_LEN(cmd); 1725 if (cmd->opcode == O_ALTQ) 1726 cmd += F_LEN(cmd); 1727 if (cmd->opcode == O_TAG) 1728 cmd += F_LEN(cmd); 1729 if ( cmd->opcode == O_SKIPTO ) 1730 for (rule = me->next; rule ; rule = rule->next) 1731 if (rule->rulenum >= cmd->arg1) 1732 break; 1733 if (rule == NULL) /* failure or not a skipto */ 1734 rule = me->next; 1735 me->next_rule = rule; 1736 return rule; 1737} 1738 1739static int 1740add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1741 uint8_t mlen, uint32_t value) 1742{ 1743 struct radix_node_head *rnh; 1744 struct table_entry *ent; 1745 1746 if (tbl >= IPFW_TABLES_MAX) 1747 return (EINVAL); 1748 rnh = ch->tables[tbl]; 1749 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1750 if (ent == NULL) 1751 return (ENOMEM); 1752 ent->value = value; 1753 ent->addr.sin_len = ent->mask.sin_len = 8; 1754 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1755 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1756 IPFW_WLOCK(&layer3_chain); 1757 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1758 NULL) { 1759 IPFW_WUNLOCK(&layer3_chain); 1760 free(ent, M_IPFW_TBL); 1761 return (EEXIST); 1762 } 1763 IPFW_WUNLOCK(&layer3_chain); 1764 return (0); 1765} 1766 1767static int 1768del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1769 uint8_t mlen) 1770{ 1771 struct radix_node_head *rnh; 1772 struct table_entry *ent; 1773 struct sockaddr_in sa, mask; 1774 1775 if (tbl >= IPFW_TABLES_MAX) 1776 return (EINVAL); 1777 rnh = ch->tables[tbl]; 1778 sa.sin_len = mask.sin_len = 8; 1779 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1780 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1781 IPFW_WLOCK(ch); 1782 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1783 if (ent == NULL) { 1784 IPFW_WUNLOCK(ch); 1785 return (ESRCH); 1786 } 1787 IPFW_WUNLOCK(ch); 1788 free(ent, M_IPFW_TBL); 1789 return (0); 1790} 1791 1792static int 1793flush_table_entry(struct radix_node *rn, void *arg) 1794{ 1795 struct radix_node_head * const rnh = arg; 1796 struct table_entry *ent; 1797 1798 ent = (struct table_entry *) 1799 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1800 if (ent != NULL) 1801 free(ent, M_IPFW_TBL); 1802 return (0); 1803} 1804 1805static int 1806flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1807{ 1808 struct radix_node_head *rnh; 1809 1810 IPFW_WLOCK_ASSERT(ch); 1811 1812 if (tbl >= IPFW_TABLES_MAX) 1813 return (EINVAL); 1814 rnh = ch->tables[tbl]; 1815 KASSERT(rnh != NULL, ("NULL IPFW table")); 1816 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1817 return (0); 1818} 1819 1820static void 1821flush_tables(struct ip_fw_chain *ch) 1822{ 1823 uint16_t tbl; 1824 1825 IPFW_WLOCK_ASSERT(ch); 1826 1827 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1828 flush_table(ch, tbl); 1829} 1830 1831static int 1832init_tables(struct ip_fw_chain *ch) 1833{ 1834 int i; 1835 uint16_t j; 1836 1837 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1838 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1839 for (j = 0; j < i; j++) { 1840 (void) flush_table(ch, j); 1841 } 1842 return (ENOMEM); 1843 } 1844 } 1845 return (0); 1846} 1847 1848static int 1849lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1850 uint32_t *val) 1851{ 1852 struct radix_node_head *rnh; 1853 struct table_entry *ent; 1854 struct sockaddr_in sa; 1855 1856 if (tbl >= IPFW_TABLES_MAX) 1857 return (0); 1858 rnh = ch->tables[tbl]; 1859 sa.sin_len = 8; 1860 sa.sin_addr.s_addr = addr; 1861 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1862 if (ent != NULL) { 1863 *val = ent->value; 1864 return (1); 1865 } 1866 return (0); 1867} 1868 1869static int 1870count_table_entry(struct radix_node *rn, void *arg) 1871{ 1872 u_int32_t * const cnt = arg; 1873 1874 (*cnt)++; 1875 return (0); 1876} 1877 1878static int 1879count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1880{ 1881 struct radix_node_head *rnh; 1882 1883 if (tbl >= IPFW_TABLES_MAX) 1884 return (EINVAL); 1885 rnh = ch->tables[tbl]; 1886 *cnt = 0; 1887 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1888 return (0); 1889} 1890 1891static int 1892dump_table_entry(struct radix_node *rn, void *arg) 1893{ 1894 struct table_entry * const n = (struct table_entry *)rn; 1895 ipfw_table * const tbl = arg; 1896 ipfw_table_entry *ent; 1897 1898 if (tbl->cnt == tbl->size) 1899 return (1); 1900 ent = &tbl->ent[tbl->cnt]; 1901 ent->tbl = tbl->tbl; 1902 if (in_nullhost(n->mask.sin_addr)) 1903 ent->masklen = 0; 1904 else 1905 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1906 ent->addr = n->addr.sin_addr.s_addr; 1907 ent->value = n->value; 1908 tbl->cnt++; 1909 return (0); 1910} 1911 1912static int 1913dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1914{ 1915 struct radix_node_head *rnh; 1916 1917 if (tbl->tbl >= IPFW_TABLES_MAX) 1918 return (EINVAL); 1919 rnh = ch->tables[tbl->tbl]; 1920 tbl->cnt = 0; 1921 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1922 return (0); 1923} 1924 1925static void 1926fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1927{ 1928 struct ucred *cr; 1929 1930 if (inp->inp_socket != NULL) { 1931 cr = inp->inp_socket->so_cred; 1932 ugp->fw_prid = jailed(cr) ? 1933 cr->cr_prison->pr_id : -1; 1934 ugp->fw_uid = cr->cr_uid; 1935 ugp->fw_ngroups = cr->cr_ngroups; 1936 bcopy(cr->cr_groups, ugp->fw_groups, 1937 sizeof(ugp->fw_groups)); 1938 } 1939} 1940 1941static int 1942check_uidgid(ipfw_insn_u32 *insn, 1943 int proto, struct ifnet *oif, 1944 struct in_addr dst_ip, u_int16_t dst_port, 1945 struct in_addr src_ip, u_int16_t src_port, 1946 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1947{ 1948 struct inpcbinfo *pi; 1949 int wildcard; 1950 struct inpcb *pcb; 1951 int match; 1952 gid_t *gp; 1953 1954 /* 1955 * Check to see if the UDP or TCP stack supplied us with 1956 * the PCB. If so, rather then holding a lock and looking 1957 * up the PCB, we can use the one that was supplied. 1958 */ 1959 if (inp && *lookup == 0) { 1960 INP_LOCK_ASSERT(inp); 1961 if (inp->inp_socket != NULL) { 1962 fill_ugid_cache(inp, ugp); 1963 *lookup = 1; 1964 } 1965 } 1966 /* 1967 * If we have already been here and the packet has no 1968 * PCB entry associated with it, then we can safely 1969 * assume that this is a no match. 1970 */ 1971 if (*lookup == -1) 1972 return (0); 1973 if (proto == IPPROTO_TCP) { 1974 wildcard = 0; 1975 pi = &tcbinfo; 1976 } else if (proto == IPPROTO_UDP) { 1977 wildcard = INPLOOKUP_WILDCARD; 1978 pi = &udbinfo; 1979 } else 1980 return 0; 1981 match = 0; 1982 if (*lookup == 0) { 1983 INP_INFO_RLOCK(pi); 1984 pcb = (oif) ? 1985 in_pcblookup_hash(pi, 1986 dst_ip, htons(dst_port), 1987 src_ip, htons(src_port), 1988 wildcard, oif) : 1989 in_pcblookup_hash(pi, 1990 src_ip, htons(src_port), 1991 dst_ip, htons(dst_port), 1992 wildcard, NULL); 1993 if (pcb != NULL) { 1994 INP_LOCK(pcb); 1995 if (pcb->inp_socket != NULL) { 1996 fill_ugid_cache(pcb, ugp); 1997 *lookup = 1; 1998 } 1999 INP_UNLOCK(pcb); 2000 } 2001 INP_INFO_RUNLOCK(pi); 2002 if (*lookup == 0) { 2003 /* 2004 * If the lookup did not yield any results, there 2005 * is no sense in coming back and trying again. So 2006 * we can set lookup to -1 and ensure that we wont 2007 * bother the pcb system again. 2008 */ 2009 *lookup = -1; 2010 return (0); 2011 } 2012 } 2013 if (insn->o.opcode == O_UID) 2014 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2015 else if (insn->o.opcode == O_GID) { 2016 for (gp = ugp->fw_groups; 2017 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2018 if (*gp == (gid_t)insn->d[0]) { 2019 match = 1; 2020 break; 2021 } 2022 } else if (insn->o.opcode == O_JAIL) 2023 match = (ugp->fw_prid == (int)insn->d[0]); 2024 return match; 2025} 2026 2027/* 2028 * The main check routine for the firewall. 2029 * 2030 * All arguments are in args so we can modify them and return them 2031 * back to the caller. 2032 * 2033 * Parameters: 2034 * 2035 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2036 * Starts with the IP header. 2037 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2038 * args->oif Outgoing interface, or NULL if packet is incoming. 2039 * The incoming interface is in the mbuf. (in) 2040 * args->divert_rule (in/out) 2041 * Skip up to the first rule past this rule number; 2042 * upon return, non-zero port number for divert or tee. 2043 * 2044 * args->rule Pointer to the last matching rule (in/out) 2045 * args->next_hop Socket we are forwarding to (out). 2046 * args->f_id Addresses grabbed from the packet (out) 2047 * args->cookie a cookie depending on rule action 2048 * 2049 * Return value: 2050 * 2051 * IP_FW_PASS the packet must be accepted 2052 * IP_FW_DENY the packet must be dropped 2053 * IP_FW_DIVERT divert packet, port in m_tag 2054 * IP_FW_TEE tee packet, port in m_tag 2055 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2056 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2057 * 2058 */ 2059 2060int 2061ipfw_chk(struct ip_fw_args *args) 2062{ 2063 /* 2064 * Local variables hold state during the processing of a packet. 2065 * 2066 * IMPORTANT NOTE: to speed up the processing of rules, there 2067 * are some assumption on the values of the variables, which 2068 * are documented here. Should you change them, please check 2069 * the implementation of the various instructions to make sure 2070 * that they still work. 2071 * 2072 * args->eh The MAC header. It is non-null for a layer2 2073 * packet, it is NULL for a layer-3 packet. 2074 * 2075 * m | args->m Pointer to the mbuf, as received from the caller. 2076 * It may change if ipfw_chk() does an m_pullup, or if it 2077 * consumes the packet because it calls send_reject(). 2078 * XXX This has to change, so that ipfw_chk() never modifies 2079 * or consumes the buffer. 2080 * ip is simply an alias of the value of m, and it is kept 2081 * in sync with it (the packet is supposed to start with 2082 * the ip header). 2083 */ 2084 struct mbuf *m = args->m; 2085 struct ip *ip = mtod(m, struct ip *); 2086 2087 /* 2088 * For rules which contain uid/gid or jail constraints, cache 2089 * a copy of the users credentials after the pcb lookup has been 2090 * executed. This will speed up the processing of rules with 2091 * these types of constraints, as well as decrease contention 2092 * on pcb related locks. 2093 */ 2094 struct ip_fw_ugid fw_ugid_cache; 2095 int ugid_lookup = 0; 2096 2097 /* 2098 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2099 * associated with a packet input on a divert socket. This 2100 * will allow to distinguish traffic and its direction when 2101 * it originates from a divert socket. 2102 */ 2103 u_int divinput_flags = 0; 2104 2105 /* 2106 * oif | args->oif If NULL, ipfw_chk has been called on the 2107 * inbound path (ether_input, ip_input). 2108 * If non-NULL, ipfw_chk has been called on the outbound path 2109 * (ether_output, ip_output). 2110 */ 2111 struct ifnet *oif = args->oif; 2112 2113 struct ip_fw *f = NULL; /* matching rule */ 2114 int retval = 0; 2115 2116 /* 2117 * hlen The length of the IP header. 2118 */ 2119 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2120 2121 /* 2122 * offset The offset of a fragment. offset != 0 means that 2123 * we have a fragment at this offset of an IPv4 packet. 2124 * offset == 0 means that (if this is an IPv4 packet) 2125 * this is the first or only fragment. 2126 * For IPv6 offset == 0 means there is no Fragment Header. 2127 * If offset != 0 for IPv6 always use correct mask to 2128 * get the correct offset because we add IP6F_MORE_FRAG 2129 * to be able to dectect the first fragment which would 2130 * otherwise have offset = 0. 2131 */ 2132 u_short offset = 0; 2133 2134 /* 2135 * Local copies of addresses. They are only valid if we have 2136 * an IP packet. 2137 * 2138 * proto The protocol. Set to 0 for non-ip packets, 2139 * or to the protocol read from the packet otherwise. 2140 * proto != 0 means that we have an IPv4 packet. 2141 * 2142 * src_port, dst_port port numbers, in HOST format. Only 2143 * valid for TCP and UDP packets. 2144 * 2145 * src_ip, dst_ip ip addresses, in NETWORK format. 2146 * Only valid for IPv4 packets. 2147 */ 2148 u_int8_t proto; 2149 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2150 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2151 u_int16_t ip_len=0; 2152 int pktlen; 2153 2154 /* 2155 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2156 * MATCH_NONE when checked and not matched (q = NULL), 2157 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2158 */ 2159 int dyn_dir = MATCH_UNKNOWN; 2160 ipfw_dyn_rule *q = NULL; 2161 struct ip_fw_chain *chain = &layer3_chain; 2162 struct m_tag *mtag; 2163 2164 /* 2165 * We store in ulp a pointer to the upper layer protocol header. 2166 * In the ipv4 case this is easy to determine from the header, 2167 * but for ipv6 we might have some additional headers in the middle. 2168 * ulp is NULL if not found. 2169 */ 2170 void *ulp = NULL; /* upper layer protocol pointer. */ 2171 /* XXX ipv6 variables */ 2172 int is_ipv6 = 0; 2173 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2174 /* end of ipv6 variables */ 2175 int is_ipv4 = 0; 2176 2177 if (m->m_flags & M_SKIP_FIREWALL) 2178 return (IP_FW_PASS); /* accept */ 2179 2180 pktlen = m->m_pkthdr.len; 2181 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2182 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2183 2184/* 2185 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2186 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2187 * pointer might become stale after other pullups (but we never use it 2188 * this way). 2189 */ 2190#define PULLUP_TO(len, p, T) \ 2191do { \ 2192 int x = (len) + sizeof(T); \ 2193 if ((m)->m_len < x) { \ 2194 args->m = m = m_pullup(m, x); \ 2195 if (m == NULL) \ 2196 goto pullup_failed; \ 2197 } \ 2198 p = (mtod(m, char *) + (len)); \ 2199} while (0) 2200 2201 /* Identify IP packets and fill up variables. */ 2202 if (pktlen >= sizeof(struct ip6_hdr) && 2203 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2204 mtod(m, struct ip *)->ip_v == 6) { 2205 is_ipv6 = 1; 2206 args->f_id.addr_type = 6; 2207 hlen = sizeof(struct ip6_hdr); 2208 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2209 2210 /* Search extension headers to find upper layer protocols */ 2211 while (ulp == NULL) { 2212 switch (proto) { 2213 case IPPROTO_ICMPV6: 2214 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2215 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2216 break; 2217 2218 case IPPROTO_TCP: 2219 PULLUP_TO(hlen, ulp, struct tcphdr); 2220 dst_port = TCP(ulp)->th_dport; 2221 src_port = TCP(ulp)->th_sport; 2222 args->f_id.flags = TCP(ulp)->th_flags; 2223 break; 2224 2225 case IPPROTO_UDP: 2226 PULLUP_TO(hlen, ulp, struct udphdr); 2227 dst_port = UDP(ulp)->uh_dport; 2228 src_port = UDP(ulp)->uh_sport; 2229 break; 2230 2231 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2232 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2233 ext_hd |= EXT_HOPOPTS; 2234 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2235 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2236 ulp = NULL; 2237 break; 2238 2239 case IPPROTO_ROUTING: /* RFC 2460 */ 2240 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2241 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2242 case 0: 2243 break; 2244 default: 2245 printf("IPFW2: IPV6 - Unknown Routing " 2246 "Header type(%d)\n", 2247 ((struct ip6_rthdr *)ulp)->ip6r_type); 2248 if (fw_deny_unknown_exthdrs) 2249 return (IP_FW_DENY); 2250 break; 2251 } 2252 ext_hd |= EXT_ROUTING; 2253 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2254 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2255 ulp = NULL; 2256 break; 2257 2258 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2259 PULLUP_TO(hlen, ulp, struct ip6_frag); 2260 ext_hd |= EXT_FRAGMENT; 2261 hlen += sizeof (struct ip6_frag); 2262 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2263 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2264 IP6F_OFF_MASK; 2265 /* Add IP6F_MORE_FRAG for offset of first 2266 * fragment to be != 0. */ 2267 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2268 IP6F_MORE_FRAG; 2269 if (offset == 0) { 2270 printf("IPFW2: IPV6 - Invalid Fragment " 2271 "Header\n"); 2272 if (fw_deny_unknown_exthdrs) 2273 return (IP_FW_DENY); 2274 break; 2275 } 2276 args->f_id.frag_id6 = 2277 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2278 ulp = NULL; 2279 break; 2280 2281 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2282 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2283 ext_hd |= EXT_DSTOPTS; 2284 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2285 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2286 ulp = NULL; 2287 break; 2288 2289 case IPPROTO_AH: /* RFC 2402 */ 2290 PULLUP_TO(hlen, ulp, struct ip6_ext); 2291 ext_hd |= EXT_AH; 2292 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2293 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2294 ulp = NULL; 2295 break; 2296 2297 case IPPROTO_ESP: /* RFC 2406 */ 2298 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2299 /* Anything past Seq# is variable length and 2300 * data past this ext. header is encrypted. */ 2301 ext_hd |= EXT_ESP; 2302 break; 2303 2304 case IPPROTO_NONE: /* RFC 2460 */ 2305 PULLUP_TO(hlen, ulp, struct ip6_ext); 2306 /* Packet ends here. if ip6e_len!=0 octets 2307 * must be ignored. */ 2308 break; 2309 2310 case IPPROTO_OSPFIGP: 2311 /* XXX OSPF header check? */ 2312 PULLUP_TO(hlen, ulp, struct ip6_ext); 2313 break; 2314 2315 case IPPROTO_PIM: 2316 /* XXX PIM header check? */ 2317 PULLUP_TO(hlen, ulp, struct pim); 2318 break; 2319 2320 case IPPROTO_CARP: 2321 PULLUP_TO(hlen, ulp, struct carp_header); 2322 if (((struct carp_header *)ulp)->carp_version != 2323 CARP_VERSION) 2324 return (IP_FW_DENY); 2325 if (((struct carp_header *)ulp)->carp_type != 2326 CARP_ADVERTISEMENT) 2327 return (IP_FW_DENY); 2328 break; 2329 2330 case IPPROTO_IPV6: /* RFC 2893 */ 2331 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2332 break; 2333 2334 case IPPROTO_IPV4: /* RFC 2893 */ 2335 PULLUP_TO(hlen, ulp, struct ip); 2336 break; 2337 2338 default: 2339 printf("IPFW2: IPV6 - Unknown Extension " 2340 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2341 if (fw_deny_unknown_exthdrs) 2342 return (IP_FW_DENY); 2343 PULLUP_TO(hlen, ulp, struct ip6_ext); 2344 break; 2345 } /*switch */ 2346 } 2347 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2348 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2349 args->f_id.src_ip = 0; 2350 args->f_id.dst_ip = 0; 2351 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2352 } else if (pktlen >= sizeof(struct ip) && 2353 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2354 mtod(m, struct ip *)->ip_v == 4) { 2355 is_ipv4 = 1; 2356 ip = mtod(m, struct ip *); 2357 hlen = ip->ip_hl << 2; 2358 args->f_id.addr_type = 4; 2359 2360 /* 2361 * Collect parameters into local variables for faster matching. 2362 */ 2363 proto = ip->ip_p; 2364 src_ip = ip->ip_src; 2365 dst_ip = ip->ip_dst; 2366 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2367 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2368 ip_len = ntohs(ip->ip_len); 2369 } else { 2370 offset = ip->ip_off & IP_OFFMASK; 2371 ip_len = ip->ip_len; 2372 } 2373 pktlen = ip_len < pktlen ? ip_len : pktlen; 2374 2375 if (offset == 0) { 2376 switch (proto) { 2377 case IPPROTO_TCP: 2378 PULLUP_TO(hlen, ulp, struct tcphdr); 2379 dst_port = TCP(ulp)->th_dport; 2380 src_port = TCP(ulp)->th_sport; 2381 args->f_id.flags = TCP(ulp)->th_flags; 2382 break; 2383 2384 case IPPROTO_UDP: 2385 PULLUP_TO(hlen, ulp, struct udphdr); 2386 dst_port = UDP(ulp)->uh_dport; 2387 src_port = UDP(ulp)->uh_sport; 2388 break; 2389 2390 case IPPROTO_ICMP: 2391 PULLUP_TO(hlen, ulp, struct icmphdr); 2392 args->f_id.flags = ICMP(ulp)->icmp_type; 2393 break; 2394 2395 default: 2396 break; 2397 } 2398 } 2399 2400 args->f_id.src_ip = ntohl(src_ip.s_addr); 2401 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2402 } 2403#undef PULLUP_TO 2404 if (proto) { /* we may have port numbers, store them */ 2405 args->f_id.proto = proto; 2406 args->f_id.src_port = src_port = ntohs(src_port); 2407 args->f_id.dst_port = dst_port = ntohs(dst_port); 2408 } 2409 2410 IPFW_RLOCK(chain); 2411 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2412 if (args->rule) { 2413 /* 2414 * Packet has already been tagged. Look for the next rule 2415 * to restart processing. 2416 * 2417 * If fw_one_pass != 0 then just accept it. 2418 * XXX should not happen here, but optimized out in 2419 * the caller. 2420 */ 2421 if (fw_one_pass) { 2422 IPFW_RUNLOCK(chain); 2423 return (IP_FW_PASS); 2424 } 2425 2426 f = args->rule->next_rule; 2427 if (f == NULL) 2428 f = lookup_next_rule(args->rule); 2429 } else { 2430 /* 2431 * Find the starting rule. It can be either the first 2432 * one, or the one after divert_rule if asked so. 2433 */ 2434 int skipto = mtag ? divert_cookie(mtag) : 0; 2435 2436 f = chain->rules; 2437 if (args->eh == NULL && skipto != 0) { 2438 if (skipto >= IPFW_DEFAULT_RULE) { 2439 IPFW_RUNLOCK(chain); 2440 return (IP_FW_DENY); /* invalid */ 2441 } 2442 while (f && f->rulenum <= skipto) 2443 f = f->next; 2444 if (f == NULL) { /* drop packet */ 2445 IPFW_RUNLOCK(chain); 2446 return (IP_FW_DENY); 2447 } 2448 } 2449 } 2450 /* reset divert rule to avoid confusion later */ 2451 if (mtag) { 2452 divinput_flags = divert_info(mtag) & 2453 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2454 m_tag_delete(m, mtag); 2455 } 2456 2457 /* 2458 * Now scan the rules, and parse microinstructions for each rule. 2459 */ 2460 for (; f; f = f->next) { 2461 ipfw_insn *cmd; 2462 uint32_t tablearg = 0; 2463 int l, cmdlen, skip_or; /* skip rest of OR block */ 2464 2465again: 2466 if (set_disable & (1 << f->set) ) 2467 continue; 2468 2469 skip_or = 0; 2470 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2471 l -= cmdlen, cmd += cmdlen) { 2472 int match; 2473 2474 /* 2475 * check_body is a jump target used when we find a 2476 * CHECK_STATE, and need to jump to the body of 2477 * the target rule. 2478 */ 2479 2480check_body: 2481 cmdlen = F_LEN(cmd); 2482 /* 2483 * An OR block (insn_1 || .. || insn_n) has the 2484 * F_OR bit set in all but the last instruction. 2485 * The first match will set "skip_or", and cause 2486 * the following instructions to be skipped until 2487 * past the one with the F_OR bit clear. 2488 */ 2489 if (skip_or) { /* skip this instruction */ 2490 if ((cmd->len & F_OR) == 0) 2491 skip_or = 0; /* next one is good */ 2492 continue; 2493 } 2494 match = 0; /* set to 1 if we succeed */ 2495 2496 switch (cmd->opcode) { 2497 /* 2498 * The first set of opcodes compares the packet's 2499 * fields with some pattern, setting 'match' if a 2500 * match is found. At the end of the loop there is 2501 * logic to deal with F_NOT and F_OR flags associated 2502 * with the opcode. 2503 */ 2504 case O_NOP: 2505 match = 1; 2506 break; 2507 2508 case O_FORWARD_MAC: 2509 printf("ipfw: opcode %d unimplemented\n", 2510 cmd->opcode); 2511 break; 2512 2513 case O_GID: 2514 case O_UID: 2515 case O_JAIL: 2516 /* 2517 * We only check offset == 0 && proto != 0, 2518 * as this ensures that we have a 2519 * packet with the ports info. 2520 */ 2521 if (offset!=0) 2522 break; 2523 if (is_ipv6) /* XXX to be fixed later */ 2524 break; 2525 if (proto == IPPROTO_TCP || 2526 proto == IPPROTO_UDP) 2527 match = check_uidgid( 2528 (ipfw_insn_u32 *)cmd, 2529 proto, oif, 2530 dst_ip, dst_port, 2531 src_ip, src_port, &fw_ugid_cache, 2532 &ugid_lookup, args->inp); 2533 break; 2534 2535 case O_RECV: 2536 match = iface_match(m->m_pkthdr.rcvif, 2537 (ipfw_insn_if *)cmd); 2538 break; 2539 2540 case O_XMIT: 2541 match = iface_match(oif, (ipfw_insn_if *)cmd); 2542 break; 2543 2544 case O_VIA: 2545 match = iface_match(oif ? oif : 2546 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2547 break; 2548 2549 case O_MACADDR2: 2550 if (args->eh != NULL) { /* have MAC header */ 2551 u_int32_t *want = (u_int32_t *) 2552 ((ipfw_insn_mac *)cmd)->addr; 2553 u_int32_t *mask = (u_int32_t *) 2554 ((ipfw_insn_mac *)cmd)->mask; 2555 u_int32_t *hdr = (u_int32_t *)args->eh; 2556 2557 match = 2558 ( want[0] == (hdr[0] & mask[0]) && 2559 want[1] == (hdr[1] & mask[1]) && 2560 want[2] == (hdr[2] & mask[2]) ); 2561 } 2562 break; 2563 2564 case O_MAC_TYPE: 2565 if (args->eh != NULL) { 2566 u_int16_t t = 2567 ntohs(args->eh->ether_type); 2568 u_int16_t *p = 2569 ((ipfw_insn_u16 *)cmd)->ports; 2570 int i; 2571 2572 for (i = cmdlen - 1; !match && i>0; 2573 i--, p += 2) 2574 match = (t>=p[0] && t<=p[1]); 2575 } 2576 break; 2577 2578 case O_FRAG: 2579 match = (offset != 0); 2580 break; 2581 2582 case O_IN: /* "out" is "not in" */ 2583 match = (oif == NULL); 2584 break; 2585 2586 case O_LAYER2: 2587 match = (args->eh != NULL); 2588 break; 2589 2590 case O_DIVERTED: 2591 match = (cmd->arg1 & 1 && divinput_flags & 2592 IP_FW_DIVERT_LOOPBACK_FLAG) || 2593 (cmd->arg1 & 2 && divinput_flags & 2594 IP_FW_DIVERT_OUTPUT_FLAG); 2595 break; 2596 2597 case O_PROTO: 2598 /* 2599 * We do not allow an arg of 0 so the 2600 * check of "proto" only suffices. 2601 */ 2602 match = (proto == cmd->arg1); 2603 break; 2604 2605 case O_IP_SRC: 2606 match = is_ipv4 && 2607 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2608 src_ip.s_addr); 2609 break; 2610 2611 case O_IP_SRC_LOOKUP: 2612 case O_IP_DST_LOOKUP: 2613 if (is_ipv4) { 2614 uint32_t a = 2615 (cmd->opcode == O_IP_DST_LOOKUP) ? 2616 dst_ip.s_addr : src_ip.s_addr; 2617 uint32_t v; 2618 2619 match = lookup_table(chain, cmd->arg1, a, 2620 &v); 2621 if (!match) 2622 break; 2623 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2624 match = 2625 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2626 else 2627 tablearg = v; 2628 } 2629 break; 2630 2631 case O_IP_SRC_MASK: 2632 case O_IP_DST_MASK: 2633 if (is_ipv4) { 2634 uint32_t a = 2635 (cmd->opcode == O_IP_DST_MASK) ? 2636 dst_ip.s_addr : src_ip.s_addr; 2637 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2638 int i = cmdlen-1; 2639 2640 for (; !match && i>0; i-= 2, p+= 2) 2641 match = (p[0] == (a & p[1])); 2642 } 2643 break; 2644 2645 case O_IP_SRC_ME: 2646 if (is_ipv4) { 2647 struct ifnet *tif; 2648 2649 INADDR_TO_IFP(src_ip, tif); 2650 match = (tif != NULL); 2651 } 2652 break; 2653 2654 case O_IP_DST_SET: 2655 case O_IP_SRC_SET: 2656 if (is_ipv4) { 2657 u_int32_t *d = (u_int32_t *)(cmd+1); 2658 u_int32_t addr = 2659 cmd->opcode == O_IP_DST_SET ? 2660 args->f_id.dst_ip : 2661 args->f_id.src_ip; 2662 2663 if (addr < d[0]) 2664 break; 2665 addr -= d[0]; /* subtract base */ 2666 match = (addr < cmd->arg1) && 2667 ( d[ 1 + (addr>>5)] & 2668 (1<<(addr & 0x1f)) ); 2669 } 2670 break; 2671 2672 case O_IP_DST: 2673 match = is_ipv4 && 2674 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2675 dst_ip.s_addr); 2676 break; 2677 2678 case O_IP_DST_ME: 2679 if (is_ipv4) { 2680 struct ifnet *tif; 2681 2682 INADDR_TO_IFP(dst_ip, tif); 2683 match = (tif != NULL); 2684 } 2685 break; 2686 2687 case O_IP_SRCPORT: 2688 case O_IP_DSTPORT: 2689 /* 2690 * offset == 0 && proto != 0 is enough 2691 * to guarantee that we have a 2692 * packet with port info. 2693 */ 2694 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2695 && offset == 0) { 2696 u_int16_t x = 2697 (cmd->opcode == O_IP_SRCPORT) ? 2698 src_port : dst_port ; 2699 u_int16_t *p = 2700 ((ipfw_insn_u16 *)cmd)->ports; 2701 int i; 2702 2703 for (i = cmdlen - 1; !match && i>0; 2704 i--, p += 2) 2705 match = (x>=p[0] && x<=p[1]); 2706 } 2707 break; 2708 2709 case O_ICMPTYPE: 2710 match = (offset == 0 && proto==IPPROTO_ICMP && 2711 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2712 break; 2713 2714#ifdef INET6 2715 case O_ICMP6TYPE: 2716 match = is_ipv6 && offset == 0 && 2717 proto==IPPROTO_ICMPV6 && 2718 icmp6type_match( 2719 ICMP6(ulp)->icmp6_type, 2720 (ipfw_insn_u32 *)cmd); 2721 break; 2722#endif /* INET6 */ 2723 2724 case O_IPOPT: 2725 match = (is_ipv4 && 2726 ipopts_match(mtod(m, struct ip *), cmd) ); 2727 break; 2728 2729 case O_IPVER: 2730 match = (is_ipv4 && 2731 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2732 break; 2733 2734 case O_IPID: 2735 case O_IPLEN: 2736 case O_IPTTL: 2737 if (is_ipv4) { /* only for IP packets */ 2738 uint16_t x; 2739 uint16_t *p; 2740 int i; 2741 2742 if (cmd->opcode == O_IPLEN) 2743 x = ip_len; 2744 else if (cmd->opcode == O_IPTTL) 2745 x = mtod(m, struct ip *)->ip_ttl; 2746 else /* must be IPID */ 2747 x = ntohs(mtod(m, struct ip *)->ip_id); 2748 if (cmdlen == 1) { 2749 match = (cmd->arg1 == x); 2750 break; 2751 } 2752 /* otherwise we have ranges */ 2753 p = ((ipfw_insn_u16 *)cmd)->ports; 2754 i = cmdlen - 1; 2755 for (; !match && i>0; i--, p += 2) 2756 match = (x >= p[0] && x <= p[1]); 2757 } 2758 break; 2759 2760 case O_IPPRECEDENCE: 2761 match = (is_ipv4 && 2762 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2763 break; 2764 2765 case O_IPTOS: 2766 match = (is_ipv4 && 2767 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2768 break; 2769 2770 case O_TCPDATALEN: 2771 if (proto == IPPROTO_TCP && offset == 0) { 2772 struct tcphdr *tcp; 2773 uint16_t x; 2774 uint16_t *p; 2775 int i; 2776 2777 tcp = TCP(ulp); 2778 x = ip_len - 2779 ((ip->ip_hl + tcp->th_off) << 2); 2780 if (cmdlen == 1) { 2781 match = (cmd->arg1 == x); 2782 break; 2783 } 2784 /* otherwise we have ranges */ 2785 p = ((ipfw_insn_u16 *)cmd)->ports; 2786 i = cmdlen - 1; 2787 for (; !match && i>0; i--, p += 2) 2788 match = (x >= p[0] && x <= p[1]); 2789 } 2790 break; 2791 2792 case O_TCPFLAGS: 2793 match = (proto == IPPROTO_TCP && offset == 0 && 2794 flags_match(cmd, TCP(ulp)->th_flags)); 2795 break; 2796 2797 case O_TCPOPTS: 2798 match = (proto == IPPROTO_TCP && offset == 0 && 2799 tcpopts_match(TCP(ulp), cmd)); 2800 break; 2801 2802 case O_TCPSEQ: 2803 match = (proto == IPPROTO_TCP && offset == 0 && 2804 ((ipfw_insn_u32 *)cmd)->d[0] == 2805 TCP(ulp)->th_seq); 2806 break; 2807 2808 case O_TCPACK: 2809 match = (proto == IPPROTO_TCP && offset == 0 && 2810 ((ipfw_insn_u32 *)cmd)->d[0] == 2811 TCP(ulp)->th_ack); 2812 break; 2813 2814 case O_TCPWIN: 2815 match = (proto == IPPROTO_TCP && offset == 0 && 2816 cmd->arg1 == TCP(ulp)->th_win); 2817 break; 2818 2819 case O_ESTAB: 2820 /* reject packets which have SYN only */ 2821 /* XXX should i also check for TH_ACK ? */ 2822 match = (proto == IPPROTO_TCP && offset == 0 && 2823 (TCP(ulp)->th_flags & 2824 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2825 break; 2826 2827 case O_ALTQ: { 2828 struct altq_tag *at; 2829 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2830 2831 match = 1; 2832 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2833 if (mtag != NULL) 2834 break; 2835 mtag = m_tag_get(PACKET_TAG_PF_QID, 2836 sizeof(struct altq_tag), 2837 M_NOWAIT); 2838 if (mtag == NULL) { 2839 /* 2840 * Let the packet fall back to the 2841 * default ALTQ. 2842 */ 2843 break; 2844 } 2845 at = (struct altq_tag *)(mtag+1); 2846 at->qid = altq->qid; 2847 if (is_ipv4) 2848 at->af = AF_INET; 2849 else 2850 at->af = AF_LINK; 2851 at->hdr = ip; 2852 m_tag_prepend(m, mtag); 2853 break; 2854 } 2855 2856 case O_LOG: 2857 if (fw_verbose) 2858 ipfw_log(f, hlen, args, m, 2859 oif, offset, tablearg); 2860 match = 1; 2861 break; 2862 2863 case O_PROB: 2864 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2865 break; 2866 2867 case O_VERREVPATH: 2868 /* Outgoing packets automatically pass/match */ 2869 match = ((oif != NULL) || 2870 (m->m_pkthdr.rcvif == NULL) || 2871 ( 2872#ifdef INET6 2873 is_ipv6 ? 2874 verify_path6(&(args->f_id.src_ip6), 2875 m->m_pkthdr.rcvif) : 2876#endif 2877 verify_path(src_ip, m->m_pkthdr.rcvif))); 2878 break; 2879 2880 case O_VERSRCREACH: 2881 /* Outgoing packets automatically pass/match */ 2882 match = (hlen > 0 && ((oif != NULL) || 2883#ifdef INET6 2884 is_ipv6 ? 2885 verify_path6(&(args->f_id.src_ip6), 2886 NULL) : 2887#endif 2888 verify_path(src_ip, NULL))); 2889 break; 2890 2891 case O_ANTISPOOF: 2892 /* Outgoing packets automatically pass/match */ 2893 if (oif == NULL && hlen > 0 && 2894 ( (is_ipv4 && in_localaddr(src_ip)) 2895#ifdef INET6 2896 || (is_ipv6 && 2897 in6_localaddr(&(args->f_id.src_ip6))) 2898#endif 2899 )) 2900 match = 2901#ifdef INET6 2902 is_ipv6 ? verify_path6( 2903 &(args->f_id.src_ip6), 2904 m->m_pkthdr.rcvif) : 2905#endif 2906 verify_path(src_ip, 2907 m->m_pkthdr.rcvif); 2908 else 2909 match = 1; 2910 break; 2911 2912 case O_IPSEC: 2913#ifdef FAST_IPSEC 2914 match = (m_tag_find(m, 2915 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2916#endif 2917#ifdef IPSEC 2918 match = (ipsec_getnhist(m) != 0); 2919#endif 2920 /* otherwise no match */ 2921 break; 2922 2923#ifdef INET6 2924 case O_IP6_SRC: 2925 match = is_ipv6 && 2926 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2927 &((ipfw_insn_ip6 *)cmd)->addr6); 2928 break; 2929 2930 case O_IP6_DST: 2931 match = is_ipv6 && 2932 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2933 &((ipfw_insn_ip6 *)cmd)->addr6); 2934 break; 2935 case O_IP6_SRC_MASK: 2936 case O_IP6_DST_MASK: 2937 if (is_ipv6) { 2938 int i = cmdlen - 1; 2939 struct in6_addr p; 2940 struct in6_addr *d = 2941 &((ipfw_insn_ip6 *)cmd)->addr6; 2942 2943 for (; !match && i > 0; d += 2, 2944 i -= F_INSN_SIZE(struct in6_addr) 2945 * 2) { 2946 p = (cmd->opcode == 2947 O_IP6_SRC_MASK) ? 2948 args->f_id.src_ip6: 2949 args->f_id.dst_ip6; 2950 APPLY_MASK(&p, &d[1]); 2951 match = 2952 IN6_ARE_ADDR_EQUAL(&d[0], 2953 &p); 2954 } 2955 } 2956 break; 2957 2958 case O_IP6_SRC_ME: 2959 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2960 break; 2961 2962 case O_IP6_DST_ME: 2963 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2964 break; 2965 2966 case O_FLOW6ID: 2967 match = is_ipv6 && 2968 flow6id_match(args->f_id.flow_id6, 2969 (ipfw_insn_u32 *) cmd); 2970 break; 2971 2972 case O_EXT_HDR: 2973 match = is_ipv6 && 2974 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2975 break; 2976 2977 case O_IP6: 2978 match = is_ipv6; 2979 break; 2980#endif 2981 2982 case O_IP4: 2983 match = is_ipv4; 2984 break; 2985 2986 case O_TAG: { 2987 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2988 tablearg : cmd->arg1; 2989 2990 /* Packet is already tagged with this tag? */ 2991 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2992 2993 /* We have `untag' action when F_NOT flag is 2994 * present. And we must remove this mtag from 2995 * mbuf and reset `match' to zero (`match' will 2996 * be inversed later). 2997 * Otherwise we should allocate new mtag and 2998 * push it into mbuf. 2999 */ 3000 if (cmd->len & F_NOT) { /* `untag' action */ 3001 if (mtag != NULL) 3002 m_tag_delete(m, mtag); 3003 } else if (mtag == NULL) { 3004 if ((mtag = m_tag_alloc(MTAG_IPFW, 3005 tag, 0, M_NOWAIT)) != NULL) 3006 m_tag_prepend(m, mtag); 3007 } 3008 match = (cmd->len & F_NOT) ? 0: 1; 3009 break; 3010 } 3011 3012 case O_TAGGED: { 3013 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3014 tablearg : cmd->arg1; 3015 3016 if (cmdlen == 1) { 3017 match = m_tag_locate(m, MTAG_IPFW, 3018 tag, NULL) != NULL; 3019 break; 3020 } 3021 3022 /* we have ranges */ 3023 for (mtag = m_tag_first(m); 3024 mtag != NULL && !match; 3025 mtag = m_tag_next(m, mtag)) { 3026 uint16_t *p; 3027 int i; 3028 3029 if (mtag->m_tag_cookie != MTAG_IPFW) 3030 continue; 3031 3032 p = ((ipfw_insn_u16 *)cmd)->ports; 3033 i = cmdlen - 1; 3034 for(; !match && i > 0; i--, p += 2) 3035 match = 3036 mtag->m_tag_id >= p[0] && 3037 mtag->m_tag_id <= p[1]; 3038 } 3039 break; 3040 } 3041 3042 /* 3043 * The second set of opcodes represents 'actions', 3044 * i.e. the terminal part of a rule once the packet 3045 * matches all previous patterns. 3046 * Typically there is only one action for each rule, 3047 * and the opcode is stored at the end of the rule 3048 * (but there are exceptions -- see below). 3049 * 3050 * In general, here we set retval and terminate the 3051 * outer loop (would be a 'break 3' in some language, 3052 * but we need to do a 'goto done'). 3053 * 3054 * Exceptions: 3055 * O_COUNT and O_SKIPTO actions: 3056 * instead of terminating, we jump to the next rule 3057 * ('goto next_rule', equivalent to a 'break 2'), 3058 * or to the SKIPTO target ('goto again' after 3059 * having set f, cmd and l), respectively. 3060 * 3061 * O_TAG, O_LOG and O_ALTQ action parameters: 3062 * perform some action and set match = 1; 3063 * 3064 * O_LIMIT and O_KEEP_STATE: these opcodes are 3065 * not real 'actions', and are stored right 3066 * before the 'action' part of the rule. 3067 * These opcodes try to install an entry in the 3068 * state tables; if successful, we continue with 3069 * the next opcode (match=1; break;), otherwise 3070 * the packet * must be dropped 3071 * ('goto done' after setting retval); 3072 * 3073 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3074 * cause a lookup of the state table, and a jump 3075 * to the 'action' part of the parent rule 3076 * ('goto check_body') if an entry is found, or 3077 * (CHECK_STATE only) a jump to the next rule if 3078 * the entry is not found ('goto next_rule'). 3079 * The result of the lookup is cached to make 3080 * further instances of these opcodes are 3081 * effectively NOPs. 3082 */ 3083 case O_LIMIT: 3084 case O_KEEP_STATE: 3085 if (install_state(f, 3086 (ipfw_insn_limit *)cmd, args, tablearg)) { 3087 retval = IP_FW_DENY; 3088 goto done; /* error/limit violation */ 3089 } 3090 match = 1; 3091 break; 3092 3093 case O_PROBE_STATE: 3094 case O_CHECK_STATE: 3095 /* 3096 * dynamic rules are checked at the first 3097 * keep-state or check-state occurrence, 3098 * with the result being stored in dyn_dir. 3099 * The compiler introduces a PROBE_STATE 3100 * instruction for us when we have a 3101 * KEEP_STATE (because PROBE_STATE needs 3102 * to be run first). 3103 */ 3104 if (dyn_dir == MATCH_UNKNOWN && 3105 (q = lookup_dyn_rule(&args->f_id, 3106 &dyn_dir, proto == IPPROTO_TCP ? 3107 TCP(ulp) : NULL)) 3108 != NULL) { 3109 /* 3110 * Found dynamic entry, update stats 3111 * and jump to the 'action' part of 3112 * the parent rule. 3113 */ 3114 q->pcnt++; 3115 q->bcnt += pktlen; 3116 f = q->rule; 3117 cmd = ACTION_PTR(f); 3118 l = f->cmd_len - f->act_ofs; 3119 IPFW_DYN_UNLOCK(); 3120 goto check_body; 3121 } 3122 /* 3123 * Dynamic entry not found. If CHECK_STATE, 3124 * skip to next rule, if PROBE_STATE just 3125 * ignore and continue with next opcode. 3126 */ 3127 if (cmd->opcode == O_CHECK_STATE) 3128 goto next_rule; 3129 match = 1; 3130 break; 3131 3132 case O_ACCEPT: 3133 retval = 0; /* accept */ 3134 goto done; 3135 3136 case O_PIPE: 3137 case O_QUEUE: 3138 args->rule = f; /* report matching rule */ 3139 if (cmd->arg1 == IP_FW_TABLEARG) 3140 args->cookie = tablearg; 3141 else 3142 args->cookie = cmd->arg1; 3143 retval = IP_FW_DUMMYNET; 3144 goto done; 3145 3146 case O_DIVERT: 3147 case O_TEE: { 3148 struct divert_tag *dt; 3149 3150 if (args->eh) /* not on layer 2 */ 3151 break; 3152 mtag = m_tag_get(PACKET_TAG_DIVERT, 3153 sizeof(struct divert_tag), 3154 M_NOWAIT); 3155 if (mtag == NULL) { 3156 /* XXX statistic */ 3157 /* drop packet */ 3158 IPFW_RUNLOCK(chain); 3159 return (IP_FW_DENY); 3160 } 3161 dt = (struct divert_tag *)(mtag+1); 3162 dt->cookie = f->rulenum; 3163 if (cmd->arg1 == IP_FW_TABLEARG) 3164 dt->info = tablearg; 3165 else 3166 dt->info = cmd->arg1; 3167 m_tag_prepend(m, mtag); 3168 retval = (cmd->opcode == O_DIVERT) ? 3169 IP_FW_DIVERT : IP_FW_TEE; 3170 goto done; 3171 } 3172 3173 case O_COUNT: 3174 case O_SKIPTO: 3175 f->pcnt++; /* update stats */ 3176 f->bcnt += pktlen; 3177 f->timestamp = time_uptime; 3178 if (cmd->opcode == O_COUNT) 3179 goto next_rule; 3180 /* handle skipto */ 3181 if (f->next_rule == NULL) 3182 lookup_next_rule(f); 3183 f = f->next_rule; 3184 goto again; 3185 3186 case O_REJECT: 3187 /* 3188 * Drop the packet and send a reject notice 3189 * if the packet is not ICMP (or is an ICMP 3190 * query), and it is not multicast/broadcast. 3191 */ 3192 if (hlen > 0 && is_ipv4 && offset == 0 && 3193 (proto != IPPROTO_ICMP || 3194 is_icmp_query(ICMP(ulp))) && 3195 !(m->m_flags & (M_BCAST|M_MCAST)) && 3196 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3197 send_reject(args, cmd->arg1, ip_len); 3198 m = args->m; 3199 } 3200 /* FALLTHROUGH */ 3201#ifdef INET6 3202 case O_UNREACH6: 3203 if (hlen > 0 && is_ipv6 && 3204 ((offset & IP6F_OFF_MASK) == 0) && 3205 (proto != IPPROTO_ICMPV6 || 3206 (is_icmp6_query(args->f_id.flags) == 1)) && 3207 !(m->m_flags & (M_BCAST|M_MCAST)) && 3208 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3209 send_reject6(args, cmd->arg1, hlen); 3210 m = args->m; 3211 } 3212 /* FALLTHROUGH */ 3213#endif 3214 case O_DENY: 3215 retval = IP_FW_DENY; 3216 goto done; 3217 3218 case O_FORWARD_IP: { 3219 struct sockaddr_in *sa; 3220 sa = &(((ipfw_insn_sa *)cmd)->sa); 3221 if (args->eh) /* not valid on layer2 pkts */ 3222 break; 3223 if (!q || dyn_dir == MATCH_FORWARD) { 3224 if (sa->sin_addr.s_addr == INADDR_ANY) { 3225 bcopy(sa, &args->hopstore, 3226 sizeof(*sa)); 3227 args->hopstore.sin_addr.s_addr = 3228 htonl(tablearg); 3229 args->next_hop = 3230 &args->hopstore; 3231 } else { 3232 args->next_hop = sa; 3233 } 3234 } 3235 retval = IP_FW_PASS; 3236 } 3237 goto done; 3238 3239 case O_NETGRAPH: 3240 case O_NGTEE: 3241 args->rule = f; /* report matching rule */ 3242 if (cmd->arg1 == IP_FW_TABLEARG) 3243 args->cookie = tablearg; 3244 else 3245 args->cookie = cmd->arg1; 3246 retval = (cmd->opcode == O_NETGRAPH) ? 3247 IP_FW_NETGRAPH : IP_FW_NGTEE; 3248 goto done; 3249 3250 default: 3251 panic("-- unknown opcode %d\n", cmd->opcode); 3252 } /* end of switch() on opcodes */ 3253 3254 if (cmd->len & F_NOT) 3255 match = !match; 3256 3257 if (match) { 3258 if (cmd->len & F_OR) 3259 skip_or = 1; 3260 } else { 3261 if (!(cmd->len & F_OR)) /* not an OR block, */ 3262 break; /* try next rule */ 3263 } 3264 3265 } /* end of inner for, scan opcodes */ 3266 3267next_rule:; /* try next rule */ 3268 3269 } /* end of outer for, scan rules */ 3270 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3271 IPFW_RUNLOCK(chain); 3272 return (IP_FW_DENY); 3273 3274done: 3275 /* Update statistics */ 3276 f->pcnt++; 3277 f->bcnt += pktlen; 3278 f->timestamp = time_uptime; 3279 IPFW_RUNLOCK(chain); 3280 return (retval); 3281 3282pullup_failed: 3283 if (fw_verbose) 3284 printf("ipfw: pullup failed\n"); 3285 return (IP_FW_DENY); 3286} 3287 3288/* 3289 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3290 * These will be reconstructed on the fly as packets are matched. 3291 */ 3292static void 3293flush_rule_ptrs(struct ip_fw_chain *chain) 3294{ 3295 struct ip_fw *rule; 3296 3297 IPFW_WLOCK_ASSERT(chain); 3298 3299 for (rule = chain->rules; rule; rule = rule->next) 3300 rule->next_rule = NULL; 3301} 3302 3303/* 3304 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3305 * possibly create a rule number and add the rule to the list. 3306 * Update the rule_number in the input struct so the caller knows it as well. 3307 */ 3308static int 3309add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3310{ 3311 struct ip_fw *rule, *f, *prev; 3312 int l = RULESIZE(input_rule); 3313 3314 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3315 return (EINVAL); 3316 3317 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3318 if (rule == NULL) 3319 return (ENOSPC); 3320 3321 bcopy(input_rule, rule, l); 3322 3323 rule->next = NULL; 3324 rule->next_rule = NULL; 3325 3326 rule->pcnt = 0; 3327 rule->bcnt = 0; 3328 rule->timestamp = 0; 3329 3330 IPFW_WLOCK(chain); 3331 3332 if (chain->rules == NULL) { /* default rule */ 3333 chain->rules = rule; 3334 goto done; 3335 } 3336 3337 /* 3338 * If rulenum is 0, find highest numbered rule before the 3339 * default rule, and add autoinc_step 3340 */ 3341 if (autoinc_step < 1) 3342 autoinc_step = 1; 3343 else if (autoinc_step > 1000) 3344 autoinc_step = 1000; 3345 if (rule->rulenum == 0) { 3346 /* 3347 * locate the highest numbered rule before default 3348 */ 3349 for (f = chain->rules; f; f = f->next) { 3350 if (f->rulenum == IPFW_DEFAULT_RULE) 3351 break; 3352 rule->rulenum = f->rulenum; 3353 } 3354 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3355 rule->rulenum += autoinc_step; 3356 input_rule->rulenum = rule->rulenum; 3357 } 3358 3359 /* 3360 * Now insert the new rule in the right place in the sorted list. 3361 */ 3362 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3363 if (f->rulenum > rule->rulenum) { /* found the location */ 3364 if (prev) { 3365 rule->next = f; 3366 prev->next = rule; 3367 } else { /* head insert */ 3368 rule->next = chain->rules; 3369 chain->rules = rule; 3370 } 3371 break; 3372 } 3373 } 3374 flush_rule_ptrs(chain); 3375done: 3376 static_count++; 3377 static_len += l; 3378 IPFW_WUNLOCK(chain); 3379 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3380 rule->rulenum, static_count);) 3381 return (0); 3382} 3383 3384/** 3385 * Remove a static rule (including derived * dynamic rules) 3386 * and place it on the ``reap list'' for later reclamation. 3387 * The caller is in charge of clearing rule pointers to avoid 3388 * dangling pointers. 3389 * @return a pointer to the next entry. 3390 * Arguments are not checked, so they better be correct. 3391 */ 3392static struct ip_fw * 3393remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3394{ 3395 struct ip_fw *n; 3396 int l = RULESIZE(rule); 3397 3398 IPFW_WLOCK_ASSERT(chain); 3399 3400 n = rule->next; 3401 IPFW_DYN_LOCK(); 3402 remove_dyn_rule(rule, NULL /* force removal */); 3403 IPFW_DYN_UNLOCK(); 3404 if (prev == NULL) 3405 chain->rules = n; 3406 else 3407 prev->next = n; 3408 static_count--; 3409 static_len -= l; 3410 3411 rule->next = chain->reap; 3412 chain->reap = rule; 3413 3414 return n; 3415} 3416 3417/** 3418 * Reclaim storage associated with a list of rules. This is 3419 * typically the list created using remove_rule. 3420 */ 3421static void 3422reap_rules(struct ip_fw *head) 3423{ 3424 struct ip_fw *rule; 3425 3426 while ((rule = head) != NULL) { 3427 head = head->next; 3428 if (DUMMYNET_LOADED) 3429 ip_dn_ruledel_ptr(rule); 3430 free(rule, M_IPFW); 3431 } 3432} 3433 3434/* 3435 * Remove all rules from a chain (except rules in set RESVD_SET 3436 * unless kill_default = 1). The caller is responsible for 3437 * reclaiming storage for the rules left in chain->reap. 3438 */ 3439static void 3440free_chain(struct ip_fw_chain *chain, int kill_default) 3441{ 3442 struct ip_fw *prev, *rule; 3443 3444 IPFW_WLOCK_ASSERT(chain); 3445 3446 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3447 for (prev = NULL, rule = chain->rules; rule ; ) 3448 if (kill_default || rule->set != RESVD_SET) 3449 rule = remove_rule(chain, rule, prev); 3450 else { 3451 prev = rule; 3452 rule = rule->next; 3453 } 3454} 3455 3456/** 3457 * Remove all rules with given number, and also do set manipulation. 3458 * Assumes chain != NULL && *chain != NULL. 3459 * 3460 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3461 * the next 8 bits are the new set, the top 8 bits are the command: 3462 * 3463 * 0 delete rules with given number 3464 * 1 delete rules with given set number 3465 * 2 move rules with given number to new set 3466 * 3 move rules with given set number to new set 3467 * 4 swap sets with given numbers 3468 */ 3469static int 3470del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3471{ 3472 struct ip_fw *prev = NULL, *rule; 3473 u_int16_t rulenum; /* rule or old_set */ 3474 u_int8_t cmd, new_set; 3475 3476 rulenum = arg & 0xffff; 3477 cmd = (arg >> 24) & 0xff; 3478 new_set = (arg >> 16) & 0xff; 3479 3480 if (cmd > 4) 3481 return EINVAL; 3482 if (new_set > RESVD_SET) 3483 return EINVAL; 3484 if (cmd == 0 || cmd == 2) { 3485 if (rulenum >= IPFW_DEFAULT_RULE) 3486 return EINVAL; 3487 } else { 3488 if (rulenum > RESVD_SET) /* old_set */ 3489 return EINVAL; 3490 } 3491 3492 IPFW_WLOCK(chain); 3493 rule = chain->rules; 3494 chain->reap = NULL; 3495 switch (cmd) { 3496 case 0: /* delete rules with given number */ 3497 /* 3498 * locate first rule to delete 3499 */ 3500 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3501 ; 3502 if (rule->rulenum != rulenum) { 3503 IPFW_WUNLOCK(chain); 3504 return EINVAL; 3505 } 3506 3507 /* 3508 * flush pointers outside the loop, then delete all matching 3509 * rules. prev remains the same throughout the cycle. 3510 */ 3511 flush_rule_ptrs(chain); 3512 while (rule->rulenum == rulenum) 3513 rule = remove_rule(chain, rule, prev); 3514 break; 3515 3516 case 1: /* delete all rules with given set number */ 3517 flush_rule_ptrs(chain); 3518 rule = chain->rules; 3519 while (rule->rulenum < IPFW_DEFAULT_RULE) 3520 if (rule->set == rulenum) 3521 rule = remove_rule(chain, rule, prev); 3522 else { 3523 prev = rule; 3524 rule = rule->next; 3525 } 3526 break; 3527 3528 case 2: /* move rules with given number to new set */ 3529 rule = chain->rules; 3530 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3531 if (rule->rulenum == rulenum) 3532 rule->set = new_set; 3533 break; 3534 3535 case 3: /* move rules with given set number to new set */ 3536 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3537 if (rule->set == rulenum) 3538 rule->set = new_set; 3539 break; 3540 3541 case 4: /* swap two sets */ 3542 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3543 if (rule->set == rulenum) 3544 rule->set = new_set; 3545 else if (rule->set == new_set) 3546 rule->set = rulenum; 3547 break; 3548 } 3549 /* 3550 * Look for rules to reclaim. We grab the list before 3551 * releasing the lock then reclaim them w/o the lock to 3552 * avoid a LOR with dummynet. 3553 */ 3554 rule = chain->reap; 3555 chain->reap = NULL; 3556 IPFW_WUNLOCK(chain); 3557 if (rule) 3558 reap_rules(rule); 3559 return 0; 3560} 3561 3562/* 3563 * Clear counters for a specific rule. 3564 * The enclosing "table" is assumed locked. 3565 */ 3566static void 3567clear_counters(struct ip_fw *rule, int log_only) 3568{ 3569 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3570 3571 if (log_only == 0) { 3572 rule->bcnt = rule->pcnt = 0; 3573 rule->timestamp = 0; 3574 } 3575 if (l->o.opcode == O_LOG) 3576 l->log_left = l->max_log; 3577} 3578 3579/** 3580 * Reset some or all counters on firewall rules. 3581 * @arg frwl is null to clear all entries, or contains a specific 3582 * rule number. 3583 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3584 */ 3585static int 3586zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3587{ 3588 struct ip_fw *rule; 3589 char *msg; 3590 3591 IPFW_WLOCK(chain); 3592 if (rulenum == 0) { 3593 norule_counter = 0; 3594 for (rule = chain->rules; rule; rule = rule->next) 3595 clear_counters(rule, log_only); 3596 msg = log_only ? "ipfw: All logging counts reset.\n" : 3597 "ipfw: Accounting cleared.\n"; 3598 } else { 3599 int cleared = 0; 3600 /* 3601 * We can have multiple rules with the same number, so we 3602 * need to clear them all. 3603 */ 3604 for (rule = chain->rules; rule; rule = rule->next) 3605 if (rule->rulenum == rulenum) { 3606 while (rule && rule->rulenum == rulenum) { 3607 clear_counters(rule, log_only); 3608 rule = rule->next; 3609 } 3610 cleared = 1; 3611 break; 3612 } 3613 if (!cleared) { /* we did not find any matching rules */ 3614 IPFW_WUNLOCK(chain); 3615 return (EINVAL); 3616 } 3617 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3618 "ipfw: Entry %d cleared.\n"; 3619 } 3620 IPFW_WUNLOCK(chain); 3621 3622 if (fw_verbose) 3623 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3624 return (0); 3625} 3626 3627/* 3628 * Check validity of the structure before insert. 3629 * Fortunately rules are simple, so this mostly need to check rule sizes. 3630 */ 3631static int 3632check_ipfw_struct(struct ip_fw *rule, int size) 3633{ 3634 int l, cmdlen = 0; 3635 int have_action=0; 3636 ipfw_insn *cmd; 3637 3638 if (size < sizeof(*rule)) { 3639 printf("ipfw: rule too short\n"); 3640 return (EINVAL); 3641 } 3642 /* first, check for valid size */ 3643 l = RULESIZE(rule); 3644 if (l != size) { 3645 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3646 return (EINVAL); 3647 } 3648 if (rule->act_ofs >= rule->cmd_len) { 3649 printf("ipfw: bogus action offset (%u > %u)\n", 3650 rule->act_ofs, rule->cmd_len - 1); 3651 return (EINVAL); 3652 } 3653 /* 3654 * Now go for the individual checks. Very simple ones, basically only 3655 * instruction sizes. 3656 */ 3657 for (l = rule->cmd_len, cmd = rule->cmd ; 3658 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3659 cmdlen = F_LEN(cmd); 3660 if (cmdlen > l) { 3661 printf("ipfw: opcode %d size truncated\n", 3662 cmd->opcode); 3663 return EINVAL; 3664 } 3665 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3666 switch (cmd->opcode) { 3667 case O_PROBE_STATE: 3668 case O_KEEP_STATE: 3669 case O_PROTO: 3670 case O_IP_SRC_ME: 3671 case O_IP_DST_ME: 3672 case O_LAYER2: 3673 case O_IN: 3674 case O_FRAG: 3675 case O_DIVERTED: 3676 case O_IPOPT: 3677 case O_IPTOS: 3678 case O_IPPRECEDENCE: 3679 case O_IPVER: 3680 case O_TCPWIN: 3681 case O_TCPFLAGS: 3682 case O_TCPOPTS: 3683 case O_ESTAB: 3684 case O_VERREVPATH: 3685 case O_VERSRCREACH: 3686 case O_ANTISPOOF: 3687 case O_IPSEC: 3688#ifdef INET6 3689 case O_IP6_SRC_ME: 3690 case O_IP6_DST_ME: 3691 case O_EXT_HDR: 3692 case O_IP6: 3693#endif 3694 case O_IP4: 3695 case O_TAG: 3696 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3697 goto bad_size; 3698 break; 3699 3700 case O_UID: 3701 case O_GID: 3702 case O_JAIL: 3703 case O_IP_SRC: 3704 case O_IP_DST: 3705 case O_TCPSEQ: 3706 case O_TCPACK: 3707 case O_PROB: 3708 case O_ICMPTYPE: 3709 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3710 goto bad_size; 3711 break; 3712 3713 case O_LIMIT: 3714 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3715 goto bad_size; 3716 break; 3717 3718 case O_LOG: 3719 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3720 goto bad_size; 3721 3722 ((ipfw_insn_log *)cmd)->log_left = 3723 ((ipfw_insn_log *)cmd)->max_log; 3724 3725 break; 3726 3727 case O_IP_SRC_MASK: 3728 case O_IP_DST_MASK: 3729 /* only odd command lengths */ 3730 if ( !(cmdlen & 1) || cmdlen > 31) 3731 goto bad_size; 3732 break; 3733 3734 case O_IP_SRC_SET: 3735 case O_IP_DST_SET: 3736 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3737 printf("ipfw: invalid set size %d\n", 3738 cmd->arg1); 3739 return EINVAL; 3740 } 3741 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3742 (cmd->arg1+31)/32 ) 3743 goto bad_size; 3744 break; 3745 3746 case O_IP_SRC_LOOKUP: 3747 case O_IP_DST_LOOKUP: 3748 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3749 printf("ipfw: invalid table number %d\n", 3750 cmd->arg1); 3751 return (EINVAL); 3752 } 3753 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3754 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3755 goto bad_size; 3756 break; 3757 3758 case O_MACADDR2: 3759 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3760 goto bad_size; 3761 break; 3762 3763 case O_NOP: 3764 case O_IPID: 3765 case O_IPTTL: 3766 case O_IPLEN: 3767 case O_TCPDATALEN: 3768 case O_TAGGED: 3769 if (cmdlen < 1 || cmdlen > 31) 3770 goto bad_size; 3771 break; 3772 3773 case O_MAC_TYPE: 3774 case O_IP_SRCPORT: 3775 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3776 if (cmdlen < 2 || cmdlen > 31) 3777 goto bad_size; 3778 break; 3779 3780 case O_RECV: 3781 case O_XMIT: 3782 case O_VIA: 3783 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3784 goto bad_size; 3785 break; 3786 3787 case O_ALTQ: 3788 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3789 goto bad_size; 3790 break; 3791 3792 case O_PIPE: 3793 case O_QUEUE: 3794 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3795 goto bad_size; 3796 goto check_action; 3797 3798 case O_FORWARD_IP: 3799#ifdef IPFIREWALL_FORWARD 3800 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3801 goto bad_size; 3802 goto check_action; 3803#else 3804 return EINVAL; 3805#endif 3806 3807 case O_DIVERT: 3808 case O_TEE: 3809 if (ip_divert_ptr == NULL) 3810 return EINVAL; 3811 else 3812 goto check_size; 3813 case O_NETGRAPH: 3814 case O_NGTEE: 3815 if (!NG_IPFW_LOADED) 3816 return EINVAL; 3817 else 3818 goto check_size; 3819 case O_FORWARD_MAC: /* XXX not implemented yet */ 3820 case O_CHECK_STATE: 3821 case O_COUNT: 3822 case O_ACCEPT: 3823 case O_DENY: 3824 case O_REJECT: 3825#ifdef INET6 3826 case O_UNREACH6: 3827#endif 3828 case O_SKIPTO: 3829check_size: 3830 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3831 goto bad_size; 3832check_action: 3833 if (have_action) { 3834 printf("ipfw: opcode %d, multiple actions" 3835 " not allowed\n", 3836 cmd->opcode); 3837 return EINVAL; 3838 } 3839 have_action = 1; 3840 if (l != cmdlen) { 3841 printf("ipfw: opcode %d, action must be" 3842 " last opcode\n", 3843 cmd->opcode); 3844 return EINVAL; 3845 } 3846 break; 3847#ifdef INET6 3848 case O_IP6_SRC: 3849 case O_IP6_DST: 3850 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3851 F_INSN_SIZE(ipfw_insn)) 3852 goto bad_size; 3853 break; 3854 3855 case O_FLOW6ID: 3856 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3857 ((ipfw_insn_u32 *)cmd)->o.arg1) 3858 goto bad_size; 3859 break; 3860 3861 case O_IP6_SRC_MASK: 3862 case O_IP6_DST_MASK: 3863 if ( !(cmdlen & 1) || cmdlen > 127) 3864 goto bad_size; 3865 break; 3866 case O_ICMP6TYPE: 3867 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3868 goto bad_size; 3869 break; 3870#endif 3871 3872 default: 3873 switch (cmd->opcode) { 3874#ifndef INET6 3875 case O_IP6_SRC_ME: 3876 case O_IP6_DST_ME: 3877 case O_EXT_HDR: 3878 case O_IP6: 3879 case O_UNREACH6: 3880 case O_IP6_SRC: 3881 case O_IP6_DST: 3882 case O_FLOW6ID: 3883 case O_IP6_SRC_MASK: 3884 case O_IP6_DST_MASK: 3885 case O_ICMP6TYPE: 3886 printf("ipfw: no IPv6 support in kernel\n"); 3887 return EPROTONOSUPPORT; 3888#endif 3889 default: 3890 printf("ipfw: opcode %d, unknown opcode\n", 3891 cmd->opcode); 3892 return EINVAL; 3893 } 3894 } 3895 } 3896 if (have_action == 0) { 3897 printf("ipfw: missing action\n"); 3898 return EINVAL; 3899 } 3900 return 0; 3901 3902bad_size: 3903 printf("ipfw: opcode %d size %d wrong\n", 3904 cmd->opcode, cmdlen); 3905 return EINVAL; 3906} 3907 3908/* 3909 * Copy the static and dynamic rules to the supplied buffer 3910 * and return the amount of space actually used. 3911 */ 3912static size_t 3913ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3914{ 3915 char *bp = buf; 3916 char *ep = bp + space; 3917 struct ip_fw *rule; 3918 int i; 3919 3920 /* XXX this can take a long time and locking will block packet flow */ 3921 IPFW_RLOCK(chain); 3922 for (rule = chain->rules; rule ; rule = rule->next) { 3923 /* 3924 * Verify the entry fits in the buffer in case the 3925 * rules changed between calculating buffer space and 3926 * now. This would be better done using a generation 3927 * number but should suffice for now. 3928 */ 3929 i = RULESIZE(rule); 3930 if (bp + i <= ep) { 3931 bcopy(rule, bp, i); 3932 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3933 sizeof(set_disable)); 3934 bp += i; 3935 } 3936 } 3937 IPFW_RUNLOCK(chain); 3938 if (ipfw_dyn_v) { 3939 ipfw_dyn_rule *p, *last = NULL; 3940 3941 IPFW_DYN_LOCK(); 3942 for (i = 0 ; i < curr_dyn_buckets; i++) 3943 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3944 if (bp + sizeof *p <= ep) { 3945 ipfw_dyn_rule *dst = 3946 (ipfw_dyn_rule *)bp; 3947 bcopy(p, dst, sizeof *p); 3948 bcopy(&(p->rule->rulenum), &(dst->rule), 3949 sizeof(p->rule->rulenum)); 3950 /* 3951 * store a non-null value in "next". 3952 * The userland code will interpret a 3953 * NULL here as a marker 3954 * for the last dynamic rule. 3955 */ 3956 bcopy(&dst, &dst->next, sizeof(dst)); 3957 last = dst; 3958 dst->expire = 3959 TIME_LEQ(dst->expire, time_uptime) ? 3960 0 : dst->expire - time_uptime ; 3961 bp += sizeof(ipfw_dyn_rule); 3962 } 3963 } 3964 IPFW_DYN_UNLOCK(); 3965 if (last != NULL) /* mark last dynamic rule */ 3966 bzero(&last->next, sizeof(last)); 3967 } 3968 return (bp - (char *)buf); 3969} 3970 3971 3972/** 3973 * {set|get}sockopt parser. 3974 */ 3975static int 3976ipfw_ctl(struct sockopt *sopt) 3977{ 3978#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3979 int error, rule_num; 3980 size_t size; 3981 struct ip_fw *buf, *rule; 3982 u_int32_t rulenum[2]; 3983 3984 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 3985 if (error) 3986 return (error); 3987 3988 /* 3989 * Disallow modifications in really-really secure mode, but still allow 3990 * the logging counters to be reset. 3991 */ 3992 if (sopt->sopt_name == IP_FW_ADD || 3993 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3994 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3995 if (error) 3996 return (error); 3997 } 3998 3999 error = 0; 4000 4001 switch (sopt->sopt_name) { 4002 case IP_FW_GET: 4003 /* 4004 * pass up a copy of the current rules. Static rules 4005 * come first (the last of which has number IPFW_DEFAULT_RULE), 4006 * followed by a possibly empty list of dynamic rule. 4007 * The last dynamic rule has NULL in the "next" field. 4008 * 4009 * Note that the calculated size is used to bound the 4010 * amount of data returned to the user. The rule set may 4011 * change between calculating the size and returning the 4012 * data in which case we'll just return what fits. 4013 */ 4014 size = static_len; /* size of static rules */ 4015 if (ipfw_dyn_v) /* add size of dyn.rules */ 4016 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4017 4018 /* 4019 * XXX todo: if the user passes a short length just to know 4020 * how much room is needed, do not bother filling up the 4021 * buffer, just jump to the sooptcopyout. 4022 */ 4023 buf = malloc(size, M_TEMP, M_WAITOK); 4024 error = sooptcopyout(sopt, buf, 4025 ipfw_getrules(&layer3_chain, buf, size)); 4026 free(buf, M_TEMP); 4027 break; 4028 4029 case IP_FW_FLUSH: 4030 /* 4031 * Normally we cannot release the lock on each iteration. 4032 * We could do it here only because we start from the head all 4033 * the times so there is no risk of missing some entries. 4034 * On the other hand, the risk is that we end up with 4035 * a very inconsistent ruleset, so better keep the lock 4036 * around the whole cycle. 4037 * 4038 * XXX this code can be improved by resetting the head of 4039 * the list to point to the default rule, and then freeing 4040 * the old list without the need for a lock. 4041 */ 4042 4043 IPFW_WLOCK(&layer3_chain); 4044 layer3_chain.reap = NULL; 4045 free_chain(&layer3_chain, 0 /* keep default rule */); 4046 rule = layer3_chain.reap; 4047 layer3_chain.reap = NULL; 4048 IPFW_WUNLOCK(&layer3_chain); 4049 if (rule != NULL) 4050 reap_rules(rule); 4051 break; 4052 4053 case IP_FW_ADD: 4054 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4055 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4056 sizeof(struct ip_fw) ); 4057 if (error == 0) 4058 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4059 if (error == 0) { 4060 error = add_rule(&layer3_chain, rule); 4061 size = RULESIZE(rule); 4062 if (!error && sopt->sopt_dir == SOPT_GET) 4063 error = sooptcopyout(sopt, rule, size); 4064 } 4065 free(rule, M_TEMP); 4066 break; 4067 4068 case IP_FW_DEL: 4069 /* 4070 * IP_FW_DEL is used for deleting single rules or sets, 4071 * and (ab)used to atomically manipulate sets. Argument size 4072 * is used to distinguish between the two: 4073 * sizeof(u_int32_t) 4074 * delete single rule or set of rules, 4075 * or reassign rules (or sets) to a different set. 4076 * 2*sizeof(u_int32_t) 4077 * atomic disable/enable sets. 4078 * first u_int32_t contains sets to be disabled, 4079 * second u_int32_t contains sets to be enabled. 4080 */ 4081 error = sooptcopyin(sopt, rulenum, 4082 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4083 if (error) 4084 break; 4085 size = sopt->sopt_valsize; 4086 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4087 error = del_entry(&layer3_chain, rulenum[0]); 4088 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4089 set_disable = 4090 (set_disable | rulenum[0]) & ~rulenum[1] & 4091 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4092 else 4093 error = EINVAL; 4094 break; 4095 4096 case IP_FW_ZERO: 4097 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 4098 rule_num = 0; 4099 if (sopt->sopt_val != 0) { 4100 error = sooptcopyin(sopt, &rule_num, 4101 sizeof(int), sizeof(int)); 4102 if (error) 4103 break; 4104 } 4105 error = zero_entry(&layer3_chain, rule_num, 4106 sopt->sopt_name == IP_FW_RESETLOG); 4107 break; 4108 4109 case IP_FW_TABLE_ADD: 4110 { 4111 ipfw_table_entry ent; 4112 4113 error = sooptcopyin(sopt, &ent, 4114 sizeof(ent), sizeof(ent)); 4115 if (error) 4116 break; 4117 error = add_table_entry(&layer3_chain, ent.tbl, 4118 ent.addr, ent.masklen, ent.value); 4119 } 4120 break; 4121 4122 case IP_FW_TABLE_DEL: 4123 { 4124 ipfw_table_entry ent; 4125 4126 error = sooptcopyin(sopt, &ent, 4127 sizeof(ent), sizeof(ent)); 4128 if (error) 4129 break; 4130 error = del_table_entry(&layer3_chain, ent.tbl, 4131 ent.addr, ent.masklen); 4132 } 4133 break; 4134 4135 case IP_FW_TABLE_FLUSH: 4136 { 4137 u_int16_t tbl; 4138 4139 error = sooptcopyin(sopt, &tbl, 4140 sizeof(tbl), sizeof(tbl)); 4141 if (error) 4142 break; 4143 IPFW_WLOCK(&layer3_chain); 4144 error = flush_table(&layer3_chain, tbl); 4145 IPFW_WUNLOCK(&layer3_chain); 4146 } 4147 break; 4148 4149 case IP_FW_TABLE_GETSIZE: 4150 { 4151 u_int32_t tbl, cnt; 4152 4153 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4154 sizeof(tbl)))) 4155 break; 4156 IPFW_RLOCK(&layer3_chain); 4157 error = count_table(&layer3_chain, tbl, &cnt); 4158 IPFW_RUNLOCK(&layer3_chain); 4159 if (error) 4160 break; 4161 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4162 } 4163 break; 4164 4165 case IP_FW_TABLE_LIST: 4166 { 4167 ipfw_table *tbl; 4168 4169 if (sopt->sopt_valsize < sizeof(*tbl)) { 4170 error = EINVAL; 4171 break; 4172 } 4173 size = sopt->sopt_valsize; 4174 tbl = malloc(size, M_TEMP, M_WAITOK); 4175 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4176 if (error) { 4177 free(tbl, M_TEMP); 4178 break; 4179 } 4180 tbl->size = (size - sizeof(*tbl)) / 4181 sizeof(ipfw_table_entry); 4182 IPFW_RLOCK(&layer3_chain); 4183 error = dump_table(&layer3_chain, tbl); 4184 IPFW_RUNLOCK(&layer3_chain); 4185 if (error) { 4186 free(tbl, M_TEMP); 4187 break; 4188 } 4189 error = sooptcopyout(sopt, tbl, size); 4190 free(tbl, M_TEMP); 4191 } 4192 break; 4193 4194 default: 4195 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4196 error = EINVAL; 4197 } 4198 4199 return (error); 4200#undef RULE_MAXSIZE 4201} 4202 4203/** 4204 * dummynet needs a reference to the default rule, because rules can be 4205 * deleted while packets hold a reference to them. When this happens, 4206 * dummynet changes the reference to the default rule (it could well be a 4207 * NULL pointer, but this way we do not need to check for the special 4208 * case, plus here he have info on the default behaviour). 4209 */ 4210struct ip_fw *ip_fw_default_rule; 4211 4212/* 4213 * This procedure is only used to handle keepalives. It is invoked 4214 * every dyn_keepalive_period 4215 */ 4216static void 4217ipfw_tick(void * __unused unused) 4218{ 4219 struct mbuf *m0, *m, *mnext, **mtailp; 4220 int i; 4221 ipfw_dyn_rule *q; 4222 4223 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4224 goto done; 4225 4226 /* 4227 * We make a chain of packets to go out here -- not deferring 4228 * until after we drop the IPFW dynamic rule lock would result 4229 * in a lock order reversal with the normal packet input -> ipfw 4230 * call stack. 4231 */ 4232 m0 = NULL; 4233 mtailp = &m0; 4234 IPFW_DYN_LOCK(); 4235 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4236 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4237 if (q->dyn_type == O_LIMIT_PARENT) 4238 continue; 4239 if (q->id.proto != IPPROTO_TCP) 4240 continue; 4241 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4242 continue; 4243 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4244 q->expire)) 4245 continue; /* too early */ 4246 if (TIME_LEQ(q->expire, time_uptime)) 4247 continue; /* too late, rule expired */ 4248 4249 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4250 q->ack_fwd, TH_SYN); 4251 if (*mtailp != NULL) 4252 mtailp = &(*mtailp)->m_nextpkt; 4253 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4254 q->ack_rev, 0); 4255 if (*mtailp != NULL) 4256 mtailp = &(*mtailp)->m_nextpkt; 4257 } 4258 } 4259 IPFW_DYN_UNLOCK(); 4260 for (m = mnext = m0; m != NULL; m = mnext) { 4261 mnext = m->m_nextpkt; 4262 m->m_nextpkt = NULL; 4263 ip_output(m, NULL, NULL, 0, NULL, NULL); 4264 } 4265done: 4266 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4267} 4268 4269int 4270ipfw_init(void) 4271{ 4272 struct ip_fw default_rule; 4273 int error; 4274 4275#ifdef INET6 4276 /* Setup IPv6 fw sysctl tree. */ 4277 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4278 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4279 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4280 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4281 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4282 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4283 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4284 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4285 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4286 &fw_deny_unknown_exthdrs, 0, 4287 "Deny packets with unknown IPv6 Extension Headers"); 4288#endif 4289 4290 layer3_chain.rules = NULL; 4291 IPFW_LOCK_INIT(&layer3_chain); 4292 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4293 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4294 UMA_ALIGN_PTR, 0); 4295 IPFW_DYN_LOCK_INIT(); 4296 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4297 4298 bzero(&default_rule, sizeof default_rule); 4299 4300 default_rule.act_ofs = 0; 4301 default_rule.rulenum = IPFW_DEFAULT_RULE; 4302 default_rule.cmd_len = 1; 4303 default_rule.set = RESVD_SET; 4304 4305 default_rule.cmd[0].len = 1; 4306 default_rule.cmd[0].opcode = 4307#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4308 1 ? O_ACCEPT : 4309#endif 4310 O_DENY; 4311 4312 error = add_rule(&layer3_chain, &default_rule); 4313 if (error != 0) { 4314 printf("ipfw2: error %u initializing default rule " 4315 "(support disabled)\n", error); 4316 IPFW_DYN_LOCK_DESTROY(); 4317 IPFW_LOCK_DESTROY(&layer3_chain); 4318 uma_zdestroy(ipfw_dyn_rule_zone); 4319 return (error); 4320 } 4321 4322 ip_fw_default_rule = layer3_chain.rules; 4323 printf("ipfw2 " 4324#ifdef INET6 4325 "(+ipv6) " 4326#endif 4327 "initialized, divert %s, " 4328 "rule-based forwarding " 4329#ifdef IPFIREWALL_FORWARD 4330 "enabled, " 4331#else 4332 "disabled, " 4333#endif 4334 "default to %s, logging ", 4335#ifdef IPDIVERT 4336 "enabled", 4337#else 4338 "loadable", 4339#endif 4340 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4341 4342#ifdef IPFIREWALL_VERBOSE 4343 fw_verbose = 1; 4344#endif 4345#ifdef IPFIREWALL_VERBOSE_LIMIT 4346 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4347#endif 4348 if (fw_verbose == 0) 4349 printf("disabled\n"); 4350 else if (verbose_limit == 0) 4351 printf("unlimited\n"); 4352 else 4353 printf("limited to %d packets/entry by default\n", 4354 verbose_limit); 4355 4356 error = init_tables(&layer3_chain); 4357 if (error) { 4358 IPFW_DYN_LOCK_DESTROY(); 4359 IPFW_LOCK_DESTROY(&layer3_chain); 4360 uma_zdestroy(ipfw_dyn_rule_zone); 4361 return (error); 4362 } 4363 ip_fw_ctl_ptr = ipfw_ctl; 4364 ip_fw_chk_ptr = ipfw_chk; 4365 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4366 4367 return (0); 4368} 4369 4370void 4371ipfw_destroy(void) 4372{ 4373 struct ip_fw *reap; 4374 4375 ip_fw_chk_ptr = NULL; 4376 ip_fw_ctl_ptr = NULL; 4377 callout_drain(&ipfw_timeout); 4378 IPFW_WLOCK(&layer3_chain); 4379 flush_tables(&layer3_chain); 4380 layer3_chain.reap = NULL; 4381 free_chain(&layer3_chain, 1 /* kill default rule */); 4382 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4383 IPFW_WUNLOCK(&layer3_chain); 4384 if (reap != NULL) 4385 reap_rules(reap); 4386 IPFW_DYN_LOCK_DESTROY(); 4387 uma_zdestroy(ipfw_dyn_rule_zone); 4388 IPFW_LOCK_DESTROY(&layer3_chain); 4389 4390#ifdef INET6 4391 /* Free IPv6 fw sysctl tree. */ 4392 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4393#endif 4394 4395 printf("IP firewall unloaded\n"); 4396} 4397