ip_fw2.c revision 163606
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 163606 2006-10-22 11:52:19Z 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/proc.h> 57#include <sys/rwlock.h> 58#include <sys/socket.h> 59#include <sys/socketvar.h> 60#include <sys/sysctl.h> 61#include <sys/syslog.h> 62#include <sys/ucred.h> 63#include <net/if.h> 64#include <net/radix.h> 65#include <net/route.h> 66#include <netinet/in.h> 67#include <netinet/in_systm.h> 68#include <netinet/in_var.h> 69#include <netinet/in_pcb.h> 70#include <netinet/ip.h> 71#include <netinet/ip_var.h> 72#include <netinet/ip_icmp.h> 73#include <netinet/ip_fw.h> 74#include <netinet/ip_divert.h> 75#include <netinet/ip_dummynet.h> 76#include <netinet/ip_carp.h> 77#include <netinet/pim.h> 78#include <netinet/tcp.h> 79#include <netinet/tcp_timer.h> 80#include <netinet/tcp_var.h> 81#include <netinet/tcpip.h> 82#include <netinet/udp.h> 83#include <netinet/udp_var.h> 84 85#include <netgraph/ng_ipfw.h> 86 87#include <altq/if_altq.h> 88 89#ifdef IPSEC 90#include <netinet6/ipsec.h> 91#endif 92 93#include <netinet/ip6.h> 94#include <netinet/icmp6.h> 95#ifdef INET6 96#include <netinet6/scope6_var.h> 97#endif 98 99#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 100 101#include <machine/in_cksum.h> /* XXX for in_cksum */ 102 103#include <security/mac/mac_framework.h> 104 105/* 106 * set_disable contains one bit per set value (0..31). 107 * If the bit is set, all rules with the corresponding set 108 * are disabled. Set RESVD_SET(31) is reserved for the default rule 109 * and rules that are not deleted by the flush command, 110 * and CANNOT be disabled. 111 * Rules in set RESVD_SET can only be deleted explicitly. 112 */ 113static u_int32_t set_disable; 114 115static int fw_verbose; 116static int verbose_limit; 117 118static struct callout ipfw_timeout; 119static uma_zone_t ipfw_dyn_rule_zone; 120#define IPFW_DEFAULT_RULE 65535 121 122/* 123 * Data structure to cache our ucred related 124 * information. This structure only gets used if 125 * the user specified UID/GID based constraints in 126 * a firewall rule. 127 */ 128struct ip_fw_ugid { 129 gid_t fw_groups[NGROUPS]; 130 int fw_ngroups; 131 uid_t fw_uid; 132 int fw_prid; 133}; 134 135#define IPFW_TABLES_MAX 128 136struct ip_fw_chain { 137 struct ip_fw *rules; /* list of rules */ 138 struct ip_fw *reap; /* list of rules to reap */ 139 struct radix_node_head *tables[IPFW_TABLES_MAX]; 140 struct rwlock rwmtx; 141}; 142#define IPFW_LOCK_INIT(_chain) \ 143 rw_init(&(_chain)->rwmtx, "IPFW static rules") 144#define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx) 145#define IPFW_WLOCK_ASSERT(_chain) do { \ 146 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \ 147 NET_ASSERT_GIANT(); \ 148} while (0) 149 150#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 151#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 152#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 153#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 154 155/* 156 * list of rules for layer 3 157 */ 158static struct ip_fw_chain layer3_chain; 159 160MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 161MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 162 163struct table_entry { 164 struct radix_node rn[2]; 165 struct sockaddr_in addr, mask; 166 u_int32_t value; 167}; 168 169static int fw_debug = 1; 170static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 171 172extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 173 174#ifdef SYSCTL_NODE 175SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 176SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 177 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 178 ipfw_chg_hook, "I", "Enable ipfw"); 179SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 180 &autoinc_step, 0, "Rule number autincrement step"); 181SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 182 CTLFLAG_RW | CTLFLAG_SECURE3, 183 &fw_one_pass, 0, 184 "Only do a single pass through ipfw when using dummynet(4)"); 185SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 186 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 187SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 188 CTLFLAG_RW | CTLFLAG_SECURE3, 189 &fw_verbose, 0, "Log matches to ipfw rules"); 190SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 191 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 192 193/* 194 * Description of dynamic rules. 195 * 196 * Dynamic rules are stored in lists accessed through a hash table 197 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 198 * be modified through the sysctl variable dyn_buckets which is 199 * updated when the table becomes empty. 200 * 201 * XXX currently there is only one list, ipfw_dyn. 202 * 203 * When a packet is received, its address fields are first masked 204 * with the mask defined for the rule, then hashed, then matched 205 * against the entries in the corresponding list. 206 * Dynamic rules can be used for different purposes: 207 * + stateful rules; 208 * + enforcing limits on the number of sessions; 209 * + in-kernel NAT (not implemented yet) 210 * 211 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 212 * measured in seconds and depending on the flags. 213 * 214 * The total number of dynamic rules is stored in dyn_count. 215 * The max number of dynamic rules is dyn_max. When we reach 216 * the maximum number of rules we do not create anymore. This is 217 * done to avoid consuming too much memory, but also too much 218 * time when searching on each packet (ideally, we should try instead 219 * to put a limit on the length of the list on each bucket...). 220 * 221 * Each dynamic rule holds a pointer to the parent ipfw rule so 222 * we know what action to perform. Dynamic rules are removed when 223 * the parent rule is deleted. XXX we should make them survive. 224 * 225 * There are some limitations with dynamic rules -- we do not 226 * obey the 'randomized match', and we do not do multiple 227 * passes through the firewall. XXX check the latter!!! 228 */ 229static ipfw_dyn_rule **ipfw_dyn_v = NULL; 230static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 231static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 232 233static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 234#define IPFW_DYN_LOCK_INIT() \ 235 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 236#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 237#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 238#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 239#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 240 241/* 242 * Timeouts for various events in handing dynamic rules. 243 */ 244static u_int32_t dyn_ack_lifetime = 300; 245static u_int32_t dyn_syn_lifetime = 20; 246static u_int32_t dyn_fin_lifetime = 1; 247static u_int32_t dyn_rst_lifetime = 1; 248static u_int32_t dyn_udp_lifetime = 10; 249static u_int32_t dyn_short_lifetime = 5; 250 251/* 252 * Keepalives are sent if dyn_keepalive is set. They are sent every 253 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 254 * seconds of lifetime of a rule. 255 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 256 * than dyn_keepalive_period. 257 */ 258 259static u_int32_t dyn_keepalive_interval = 20; 260static u_int32_t dyn_keepalive_period = 5; 261static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 262 263static u_int32_t static_count; /* # of static rules */ 264static u_int32_t static_len; /* size in bytes of static rules */ 265static u_int32_t dyn_count; /* # of dynamic rules */ 266static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 267 268SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 269 &dyn_buckets, 0, "Number of dyn. buckets"); 270SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 271 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 272SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 273 &dyn_count, 0, "Number of dyn. rules"); 274SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 275 &dyn_max, 0, "Max number of dyn. rules"); 276SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 277 &static_count, 0, "Number of static rules"); 278SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 279 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 280SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 281 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 282SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 283 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 284SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 285 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 286SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 287 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 288SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 289 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 290SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 291 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 292 293#ifdef INET6 294/* 295 * IPv6 specific variables 296 */ 297SYSCTL_DECL(_net_inet6_ip6); 298 299static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 300static struct sysctl_oid *ip6_fw_sysctl_tree; 301#endif /* INET6 */ 302#endif /* SYSCTL_NODE */ 303 304static int fw_deny_unknown_exthdrs = 1; 305 306 307/* 308 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 309 * Other macros just cast void * into the appropriate type 310 */ 311#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 312#define TCP(p) ((struct tcphdr *)(p)) 313#define UDP(p) ((struct udphdr *)(p)) 314#define ICMP(p) ((struct icmphdr *)(p)) 315#define ICMP6(p) ((struct icmp6_hdr *)(p)) 316 317static __inline int 318icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 319{ 320 int type = icmp->icmp_type; 321 322 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 323} 324 325#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 326 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 327 328static int 329is_icmp_query(struct icmphdr *icmp) 330{ 331 int type = icmp->icmp_type; 332 333 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 334} 335#undef TT 336 337/* 338 * The following checks use two arrays of 8 or 16 bits to store the 339 * bits that we want set or clear, respectively. They are in the 340 * low and high half of cmd->arg1 or cmd->d[0]. 341 * 342 * We scan options and store the bits we find set. We succeed if 343 * 344 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 345 * 346 * The code is sometimes optimized not to store additional variables. 347 */ 348 349static int 350flags_match(ipfw_insn *cmd, u_int8_t bits) 351{ 352 u_char want_clear; 353 bits = ~bits; 354 355 if ( ((cmd->arg1 & 0xff) & bits) != 0) 356 return 0; /* some bits we want set were clear */ 357 want_clear = (cmd->arg1 >> 8) & 0xff; 358 if ( (want_clear & bits) != want_clear) 359 return 0; /* some bits we want clear were set */ 360 return 1; 361} 362 363static int 364ipopts_match(struct ip *ip, ipfw_insn *cmd) 365{ 366 int optlen, bits = 0; 367 u_char *cp = (u_char *)(ip + 1); 368 int x = (ip->ip_hl << 2) - sizeof (struct ip); 369 370 for (; x > 0; x -= optlen, cp += optlen) { 371 int opt = cp[IPOPT_OPTVAL]; 372 373 if (opt == IPOPT_EOL) 374 break; 375 if (opt == IPOPT_NOP) 376 optlen = 1; 377 else { 378 optlen = cp[IPOPT_OLEN]; 379 if (optlen <= 0 || optlen > x) 380 return 0; /* invalid or truncated */ 381 } 382 switch (opt) { 383 384 default: 385 break; 386 387 case IPOPT_LSRR: 388 bits |= IP_FW_IPOPT_LSRR; 389 break; 390 391 case IPOPT_SSRR: 392 bits |= IP_FW_IPOPT_SSRR; 393 break; 394 395 case IPOPT_RR: 396 bits |= IP_FW_IPOPT_RR; 397 break; 398 399 case IPOPT_TS: 400 bits |= IP_FW_IPOPT_TS; 401 break; 402 } 403 } 404 return (flags_match(cmd, bits)); 405} 406 407static int 408tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 409{ 410 int optlen, bits = 0; 411 u_char *cp = (u_char *)(tcp + 1); 412 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 413 414 for (; x > 0; x -= optlen, cp += optlen) { 415 int opt = cp[0]; 416 if (opt == TCPOPT_EOL) 417 break; 418 if (opt == TCPOPT_NOP) 419 optlen = 1; 420 else { 421 optlen = cp[1]; 422 if (optlen <= 0) 423 break; 424 } 425 426 switch (opt) { 427 428 default: 429 break; 430 431 case TCPOPT_MAXSEG: 432 bits |= IP_FW_TCPOPT_MSS; 433 break; 434 435 case TCPOPT_WINDOW: 436 bits |= IP_FW_TCPOPT_WINDOW; 437 break; 438 439 case TCPOPT_SACK_PERMITTED: 440 case TCPOPT_SACK: 441 bits |= IP_FW_TCPOPT_SACK; 442 break; 443 444 case TCPOPT_TIMESTAMP: 445 bits |= IP_FW_TCPOPT_TS; 446 break; 447 448 } 449 } 450 return (flags_match(cmd, bits)); 451} 452 453static int 454iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 455{ 456 if (ifp == NULL) /* no iface with this packet, match fails */ 457 return 0; 458 /* Check by name or by IP address */ 459 if (cmd->name[0] != '\0') { /* match by name */ 460 /* Check name */ 461 if (cmd->p.glob) { 462 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 463 return(1); 464 } else { 465 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 466 return(1); 467 } 468 } else { 469 struct ifaddr *ia; 470 471 /* XXX lock? */ 472 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 473 if (ia->ifa_addr->sa_family != AF_INET) 474 continue; 475 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 476 (ia->ifa_addr))->sin_addr.s_addr) 477 return(1); /* match */ 478 } 479 } 480 return(0); /* no match, fail ... */ 481} 482 483/* 484 * The verify_path function checks if a route to the src exists and 485 * if it is reachable via ifp (when provided). 486 * 487 * The 'verrevpath' option checks that the interface that an IP packet 488 * arrives on is the same interface that traffic destined for the 489 * packet's source address would be routed out of. The 'versrcreach' 490 * option just checks that the source address is reachable via any route 491 * (except default) in the routing table. These two are a measure to block 492 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 493 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 494 * is purposely reminiscent of the Cisco IOS command, 495 * 496 * ip verify unicast reverse-path 497 * ip verify unicast source reachable-via any 498 * 499 * which implements the same functionality. But note that syntax is 500 * misleading. The check may be performed on all IP packets whether unicast, 501 * multicast, or broadcast. 502 */ 503static int 504verify_path(struct in_addr src, struct ifnet *ifp) 505{ 506 struct route ro; 507 struct sockaddr_in *dst; 508 509 bzero(&ro, sizeof(ro)); 510 511 dst = (struct sockaddr_in *)&(ro.ro_dst); 512 dst->sin_family = AF_INET; 513 dst->sin_len = sizeof(*dst); 514 dst->sin_addr = src; 515 rtalloc_ign(&ro, RTF_CLONING); 516 517 if (ro.ro_rt == NULL) 518 return 0; 519 520 /* 521 * If ifp is provided, check for equality with rtentry. 522 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 523 * in order to pass packets injected back by if_simloop(): 524 * if useloopback == 1 routing entry (via lo0) for our own address 525 * may exist, so we need to handle routing assymetry. 526 */ 527 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 528 RTFREE(ro.ro_rt); 529 return 0; 530 } 531 532 /* if no ifp provided, check if rtentry is not default route */ 533 if (ifp == NULL && 534 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 535 RTFREE(ro.ro_rt); 536 return 0; 537 } 538 539 /* or if this is a blackhole/reject route */ 540 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 541 RTFREE(ro.ro_rt); 542 return 0; 543 } 544 545 /* found valid route */ 546 RTFREE(ro.ro_rt); 547 return 1; 548} 549 550#ifdef INET6 551/* 552 * ipv6 specific rules here... 553 */ 554static __inline int 555icmp6type_match (int type, ipfw_insn_u32 *cmd) 556{ 557 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 558} 559 560static int 561flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 562{ 563 int i; 564 for (i=0; i <= cmd->o.arg1; ++i ) 565 if (curr_flow == cmd->d[i] ) 566 return 1; 567 return 0; 568} 569 570/* support for IP6_*_ME opcodes */ 571static int 572search_ip6_addr_net (struct in6_addr * ip6_addr) 573{ 574 struct ifnet *mdc; 575 struct ifaddr *mdc2; 576 struct in6_ifaddr *fdm; 577 struct in6_addr copia; 578 579 TAILQ_FOREACH(mdc, &ifnet, if_link) 580 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 581 if (mdc2->ifa_addr->sa_family == AF_INET6) { 582 fdm = (struct in6_ifaddr *)mdc2; 583 copia = fdm->ia_addr.sin6_addr; 584 /* need for leaving scope_id in the sock_addr */ 585 in6_clearscope(&copia); 586 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 587 return 1; 588 } 589 } 590 return 0; 591} 592 593static int 594verify_path6(struct in6_addr *src, struct ifnet *ifp) 595{ 596 struct route_in6 ro; 597 struct sockaddr_in6 *dst; 598 599 bzero(&ro, sizeof(ro)); 600 601 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 602 dst->sin6_family = AF_INET6; 603 dst->sin6_len = sizeof(*dst); 604 dst->sin6_addr = *src; 605 rtalloc_ign((struct route *)&ro, RTF_CLONING); 606 607 if (ro.ro_rt == NULL) 608 return 0; 609 610 /* 611 * if ifp is provided, check for equality with rtentry 612 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 613 * to support the case of sending packets to an address of our own. 614 * (where the former interface is the first argument of if_simloop() 615 * (=ifp), the latter is lo0) 616 */ 617 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 618 RTFREE(ro.ro_rt); 619 return 0; 620 } 621 622 /* if no ifp provided, check if rtentry is not default route */ 623 if (ifp == NULL && 624 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 625 RTFREE(ro.ro_rt); 626 return 0; 627 } 628 629 /* or if this is a blackhole/reject route */ 630 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 631 RTFREE(ro.ro_rt); 632 return 0; 633 } 634 635 /* found valid route */ 636 RTFREE(ro.ro_rt); 637 return 1; 638 639} 640static __inline int 641hash_packet6(struct ipfw_flow_id *id) 642{ 643 u_int32_t i; 644 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 645 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 646 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 647 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 648 (id->dst_port) ^ (id->src_port); 649 return i; 650} 651 652static int 653is_icmp6_query(int icmp6_type) 654{ 655 if ((icmp6_type <= ICMP6_MAXTYPE) && 656 (icmp6_type == ICMP6_ECHO_REQUEST || 657 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 658 icmp6_type == ICMP6_WRUREQUEST || 659 icmp6_type == ICMP6_FQDN_QUERY || 660 icmp6_type == ICMP6_NI_QUERY)) 661 return (1); 662 663 return (0); 664} 665 666static void 667send_reject6(struct ip_fw_args *args, int code, u_int hlen) 668{ 669 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 670 struct ip6_hdr *ip6; 671 struct tcphdr *tcp; 672 tcp_seq ack, seq; 673 int flags; 674 struct { 675 struct ip6_hdr ip6; 676 struct tcphdr th; 677 } ti; 678 679 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) { 680 args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr)); 681 if (args->m == NULL) 682 return; 683 } 684 685 ip6 = mtod(args->m, struct ip6_hdr *); 686 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 687 688 if ((tcp->th_flags & TH_RST) != 0) { 689 m_freem(args->m); 690 return; 691 } 692 693 ti.ip6 = *ip6; 694 ti.th = *tcp; 695 ti.th.th_seq = ntohl(ti.th.th_seq); 696 ti.th.th_ack = ntohl(ti.th.th_ack); 697 ti.ip6.ip6_nxt = IPPROTO_TCP; 698 699 if (ti.th.th_flags & TH_ACK) { 700 ack = 0; 701 seq = ti.th.th_ack; 702 flags = TH_RST; 703 } else { 704 ack = ti.th.th_seq; 705 if (((args->m)->m_flags & M_PKTHDR) != 0) { 706 ack += (args->m)->m_pkthdr.len - hlen 707 - (ti.th.th_off << 2); 708 } else if (ip6->ip6_plen) { 709 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) 710 - hlen - (ti.th.th_off << 2); 711 } else { 712 m_freem(args->m); 713 return; 714 } 715 if (tcp->th_flags & TH_SYN) 716 ack++; 717 seq = 0; 718 flags = TH_RST|TH_ACK; 719 } 720 bcopy(&ti, ip6, sizeof(ti)); 721 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1), 722 args->m, ack, seq, flags); 723 724 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 725 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0); 726 727 } else 728 m_freem(args->m); 729 730 args->m = NULL; 731} 732 733#endif /* INET6 */ 734 735static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 736 737#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 738#define SNP(buf) buf, sizeof(buf) 739 740/* 741 * We enter here when we have a rule with O_LOG. 742 * XXX this function alone takes about 2Kbytes of code! 743 */ 744static void 745ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 746 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg) 747{ 748 struct ether_header *eh = args->eh; 749 char *action; 750 int limit_reached = 0; 751 char action2[40], proto[128], fragment[32]; 752 753 fragment[0] = '\0'; 754 proto[0] = '\0'; 755 756 if (f == NULL) { /* bogus pkt */ 757 if (verbose_limit != 0 && norule_counter >= verbose_limit) 758 return; 759 norule_counter++; 760 if (norule_counter == verbose_limit) 761 limit_reached = verbose_limit; 762 action = "Refuse"; 763 } else { /* O_LOG is the first action, find the real one */ 764 ipfw_insn *cmd = ACTION_PTR(f); 765 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 766 767 if (l->max_log != 0 && l->log_left == 0) 768 return; 769 l->log_left--; 770 if (l->log_left == 0) 771 limit_reached = l->max_log; 772 cmd += F_LEN(cmd); /* point to first action */ 773 if (cmd->opcode == O_ALTQ) { 774 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 775 776 snprintf(SNPARGS(action2, 0), "Altq %d", 777 altq->qid); 778 cmd += F_LEN(cmd); 779 } 780 if (cmd->opcode == O_PROB) 781 cmd += F_LEN(cmd); 782 783 if (cmd->opcode == O_TAG) 784 cmd += F_LEN(cmd); 785 786 action = action2; 787 switch (cmd->opcode) { 788 case O_DENY: 789 action = "Deny"; 790 break; 791 792 case O_REJECT: 793 if (cmd->arg1==ICMP_REJECT_RST) 794 action = "Reset"; 795 else if (cmd->arg1==ICMP_UNREACH_HOST) 796 action = "Reject"; 797 else 798 snprintf(SNPARGS(action2, 0), "Unreach %d", 799 cmd->arg1); 800 break; 801 802 case O_UNREACH6: 803 if (cmd->arg1==ICMP6_UNREACH_RST) 804 action = "Reset"; 805 else 806 snprintf(SNPARGS(action2, 0), "Unreach %d", 807 cmd->arg1); 808 break; 809 810 case O_ACCEPT: 811 action = "Accept"; 812 break; 813 case O_COUNT: 814 action = "Count"; 815 break; 816 case O_DIVERT: 817 snprintf(SNPARGS(action2, 0), "Divert %d", 818 cmd->arg1); 819 break; 820 case O_TEE: 821 snprintf(SNPARGS(action2, 0), "Tee %d", 822 cmd->arg1); 823 break; 824 case O_SKIPTO: 825 snprintf(SNPARGS(action2, 0), "SkipTo %d", 826 cmd->arg1); 827 break; 828 case O_PIPE: 829 snprintf(SNPARGS(action2, 0), "Pipe %d", 830 cmd->arg1); 831 break; 832 case O_QUEUE: 833 snprintf(SNPARGS(action2, 0), "Queue %d", 834 cmd->arg1); 835 break; 836 case O_FORWARD_IP: { 837 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 838 int len; 839 struct in_addr dummyaddr; 840 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 841 dummyaddr.s_addr = htonl(tablearg); 842 else 843 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 844 845 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 846 inet_ntoa(dummyaddr)); 847 848 if (sa->sa.sin_port) 849 snprintf(SNPARGS(action2, len), ":%d", 850 sa->sa.sin_port); 851 } 852 break; 853 case O_NETGRAPH: 854 snprintf(SNPARGS(action2, 0), "Netgraph %d", 855 cmd->arg1); 856 break; 857 case O_NGTEE: 858 snprintf(SNPARGS(action2, 0), "Ngtee %d", 859 cmd->arg1); 860 break; 861 default: 862 action = "UNKNOWN"; 863 break; 864 } 865 } 866 867 if (hlen == 0) { /* non-ip */ 868 snprintf(SNPARGS(proto, 0), "MAC"); 869 870 } else { 871 int len; 872 char src[48], dst[48]; 873 struct icmphdr *icmp; 874 struct tcphdr *tcp; 875 struct udphdr *udp; 876 /* Initialize to make compiler happy. */ 877 struct ip *ip = NULL; 878#ifdef INET6 879 struct ip6_hdr *ip6 = NULL; 880 struct icmp6_hdr *icmp6; 881#endif 882 src[0] = '\0'; 883 dst[0] = '\0'; 884#ifdef INET6 885 if (IS_IP6_FLOW_ID(&(args->f_id))) { 886 snprintf(src, sizeof(src), "[%s]", 887 ip6_sprintf(&args->f_id.src_ip6)); 888 snprintf(dst, sizeof(dst), "[%s]", 889 ip6_sprintf(&args->f_id.dst_ip6)); 890 891 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 892 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 893 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 894 } else 895#endif 896 { 897 ip = mtod(m, struct ip *); 898 tcp = L3HDR(struct tcphdr, ip); 899 udp = L3HDR(struct udphdr, ip); 900 901 inet_ntoa_r(ip->ip_src, src); 902 inet_ntoa_r(ip->ip_dst, dst); 903 } 904 905 switch (args->f_id.proto) { 906 case IPPROTO_TCP: 907 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 908 if (offset == 0) 909 snprintf(SNPARGS(proto, len), ":%d %s:%d", 910 ntohs(tcp->th_sport), 911 dst, 912 ntohs(tcp->th_dport)); 913 else 914 snprintf(SNPARGS(proto, len), " %s", dst); 915 break; 916 917 case IPPROTO_UDP: 918 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 919 if (offset == 0) 920 snprintf(SNPARGS(proto, len), ":%d %s:%d", 921 ntohs(udp->uh_sport), 922 dst, 923 ntohs(udp->uh_dport)); 924 else 925 snprintf(SNPARGS(proto, len), " %s", dst); 926 break; 927 928 case IPPROTO_ICMP: 929 icmp = L3HDR(struct icmphdr, ip); 930 if (offset == 0) 931 len = snprintf(SNPARGS(proto, 0), 932 "ICMP:%u.%u ", 933 icmp->icmp_type, icmp->icmp_code); 934 else 935 len = snprintf(SNPARGS(proto, 0), "ICMP "); 936 len += snprintf(SNPARGS(proto, len), "%s", src); 937 snprintf(SNPARGS(proto, len), " %s", dst); 938 break; 939#ifdef INET6 940 case IPPROTO_ICMPV6: 941 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 942 if (offset == 0) 943 len = snprintf(SNPARGS(proto, 0), 944 "ICMPv6:%u.%u ", 945 icmp6->icmp6_type, icmp6->icmp6_code); 946 else 947 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 948 len += snprintf(SNPARGS(proto, len), "%s", src); 949 snprintf(SNPARGS(proto, len), " %s", dst); 950 break; 951#endif 952 default: 953 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 954 args->f_id.proto, src); 955 snprintf(SNPARGS(proto, len), " %s", dst); 956 break; 957 } 958 959#ifdef INET6 960 if (IS_IP6_FLOW_ID(&(args->f_id))) { 961 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 962 snprintf(SNPARGS(fragment, 0), 963 " (frag %08x:%d@%d%s)", 964 args->f_id.frag_id6, 965 ntohs(ip6->ip6_plen) - hlen, 966 ntohs(offset & IP6F_OFF_MASK) << 3, 967 (offset & IP6F_MORE_FRAG) ? "+" : ""); 968 } else 969#endif 970 { 971 int ip_off, ip_len; 972 if (eh != NULL) { /* layer 2 packets are as on the wire */ 973 ip_off = ntohs(ip->ip_off); 974 ip_len = ntohs(ip->ip_len); 975 } else { 976 ip_off = ip->ip_off; 977 ip_len = ip->ip_len; 978 } 979 if (ip_off & (IP_MF | IP_OFFMASK)) 980 snprintf(SNPARGS(fragment, 0), 981 " (frag %d:%d@%d%s)", 982 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 983 offset << 3, 984 (ip_off & IP_MF) ? "+" : ""); 985 } 986 } 987 if (oif || m->m_pkthdr.rcvif) 988 log(LOG_SECURITY | LOG_INFO, 989 "ipfw: %d %s %s %s via %s%s\n", 990 f ? f->rulenum : -1, 991 action, proto, oif ? "out" : "in", 992 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 993 fragment); 994 else 995 log(LOG_SECURITY | LOG_INFO, 996 "ipfw: %d %s %s [no if info]%s\n", 997 f ? f->rulenum : -1, 998 action, proto, fragment); 999 if (limit_reached) 1000 log(LOG_SECURITY | LOG_NOTICE, 1001 "ipfw: limit %d reached on entry %d\n", 1002 limit_reached, f ? f->rulenum : -1); 1003} 1004 1005/* 1006 * IMPORTANT: the hash function for dynamic rules must be commutative 1007 * in source and destination (ip,port), because rules are bidirectional 1008 * and we want to find both in the same bucket. 1009 */ 1010static __inline int 1011hash_packet(struct ipfw_flow_id *id) 1012{ 1013 u_int32_t i; 1014 1015#ifdef INET6 1016 if (IS_IP6_FLOW_ID(id)) 1017 i = hash_packet6(id); 1018 else 1019#endif /* INET6 */ 1020 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1021 i &= (curr_dyn_buckets - 1); 1022 return i; 1023} 1024 1025/** 1026 * unlink a dynamic rule from a chain. prev is a pointer to 1027 * the previous one, q is a pointer to the rule to delete, 1028 * head is a pointer to the head of the queue. 1029 * Modifies q and potentially also head. 1030 */ 1031#define UNLINK_DYN_RULE(prev, head, q) { \ 1032 ipfw_dyn_rule *old_q = q; \ 1033 \ 1034 /* remove a refcount to the parent */ \ 1035 if (q->dyn_type == O_LIMIT) \ 1036 q->parent->count--; \ 1037 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1038 (q->id.src_ip), (q->id.src_port), \ 1039 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1040 if (prev != NULL) \ 1041 prev->next = q = q->next; \ 1042 else \ 1043 head = q = q->next; \ 1044 dyn_count--; \ 1045 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1046 1047#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1048 1049/** 1050 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1051 * 1052 * If keep_me == NULL, rules are deleted even if not expired, 1053 * otherwise only expired rules are removed. 1054 * 1055 * The value of the second parameter is also used to point to identify 1056 * a rule we absolutely do not want to remove (e.g. because we are 1057 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1058 * rules). The pointer is only used for comparison, so any non-null 1059 * value will do. 1060 */ 1061static void 1062remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1063{ 1064 static u_int32_t last_remove = 0; 1065 1066#define FORCE (keep_me == NULL) 1067 1068 ipfw_dyn_rule *prev, *q; 1069 int i, pass = 0, max_pass = 0; 1070 1071 IPFW_DYN_LOCK_ASSERT(); 1072 1073 if (ipfw_dyn_v == NULL || dyn_count == 0) 1074 return; 1075 /* do not expire more than once per second, it is useless */ 1076 if (!FORCE && last_remove == time_uptime) 1077 return; 1078 last_remove = time_uptime; 1079 1080 /* 1081 * because O_LIMIT refer to parent rules, during the first pass only 1082 * remove child and mark any pending LIMIT_PARENT, and remove 1083 * them in a second pass. 1084 */ 1085next_pass: 1086 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1087 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1088 /* 1089 * Logic can become complex here, so we split tests. 1090 */ 1091 if (q == keep_me) 1092 goto next; 1093 if (rule != NULL && rule != q->rule) 1094 goto next; /* not the one we are looking for */ 1095 if (q->dyn_type == O_LIMIT_PARENT) { 1096 /* 1097 * handle parent in the second pass, 1098 * record we need one. 1099 */ 1100 max_pass = 1; 1101 if (pass == 0) 1102 goto next; 1103 if (FORCE && q->count != 0 ) { 1104 /* XXX should not happen! */ 1105 printf("ipfw: OUCH! cannot remove rule," 1106 " count %d\n", q->count); 1107 } 1108 } else { 1109 if (!FORCE && 1110 !TIME_LEQ( q->expire, time_uptime )) 1111 goto next; 1112 } 1113 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1114 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1115 continue; 1116 } 1117next: 1118 prev=q; 1119 q=q->next; 1120 } 1121 } 1122 if (pass++ < max_pass) 1123 goto next_pass; 1124} 1125 1126 1127/** 1128 * lookup a dynamic rule. 1129 */ 1130static ipfw_dyn_rule * 1131lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1132 struct tcphdr *tcp) 1133{ 1134 /* 1135 * stateful ipfw extensions. 1136 * Lookup into dynamic session queue 1137 */ 1138#define MATCH_REVERSE 0 1139#define MATCH_FORWARD 1 1140#define MATCH_NONE 2 1141#define MATCH_UNKNOWN 3 1142 int i, dir = MATCH_NONE; 1143 ipfw_dyn_rule *prev, *q=NULL; 1144 1145 IPFW_DYN_LOCK_ASSERT(); 1146 1147 if (ipfw_dyn_v == NULL) 1148 goto done; /* not found */ 1149 i = hash_packet( pkt ); 1150 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1151 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1152 goto next; 1153 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1154 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1155 continue; 1156 } 1157 if (pkt->proto == q->id.proto && 1158 q->dyn_type != O_LIMIT_PARENT) { 1159 if (IS_IP6_FLOW_ID(pkt)) { 1160 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1161 &(q->id.src_ip6)) && 1162 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1163 &(q->id.dst_ip6)) && 1164 pkt->src_port == q->id.src_port && 1165 pkt->dst_port == q->id.dst_port ) { 1166 dir = MATCH_FORWARD; 1167 break; 1168 } 1169 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1170 &(q->id.dst_ip6)) && 1171 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1172 &(q->id.src_ip6)) && 1173 pkt->src_port == q->id.dst_port && 1174 pkt->dst_port == q->id.src_port ) { 1175 dir = MATCH_REVERSE; 1176 break; 1177 } 1178 } else { 1179 if (pkt->src_ip == q->id.src_ip && 1180 pkt->dst_ip == q->id.dst_ip && 1181 pkt->src_port == q->id.src_port && 1182 pkt->dst_port == q->id.dst_port ) { 1183 dir = MATCH_FORWARD; 1184 break; 1185 } 1186 if (pkt->src_ip == q->id.dst_ip && 1187 pkt->dst_ip == q->id.src_ip && 1188 pkt->src_port == q->id.dst_port && 1189 pkt->dst_port == q->id.src_port ) { 1190 dir = MATCH_REVERSE; 1191 break; 1192 } 1193 } 1194 } 1195next: 1196 prev = q; 1197 q = q->next; 1198 } 1199 if (q == NULL) 1200 goto done; /* q = NULL, not found */ 1201 1202 if ( prev != NULL) { /* found and not in front */ 1203 prev->next = q->next; 1204 q->next = ipfw_dyn_v[i]; 1205 ipfw_dyn_v[i] = q; 1206 } 1207 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1208 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1209 1210#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1211#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1212 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1213 switch (q->state) { 1214 case TH_SYN: /* opening */ 1215 q->expire = time_uptime + dyn_syn_lifetime; 1216 break; 1217 1218 case BOTH_SYN: /* move to established */ 1219 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1220 case BOTH_SYN | (TH_FIN << 8) : 1221 if (tcp) { 1222#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1223 u_int32_t ack = ntohl(tcp->th_ack); 1224 if (dir == MATCH_FORWARD) { 1225 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1226 q->ack_fwd = ack; 1227 else { /* ignore out-of-sequence */ 1228 break; 1229 } 1230 } else { 1231 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1232 q->ack_rev = ack; 1233 else { /* ignore out-of-sequence */ 1234 break; 1235 } 1236 } 1237 } 1238 q->expire = time_uptime + dyn_ack_lifetime; 1239 break; 1240 1241 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1242 if (dyn_fin_lifetime >= dyn_keepalive_period) 1243 dyn_fin_lifetime = dyn_keepalive_period - 1; 1244 q->expire = time_uptime + dyn_fin_lifetime; 1245 break; 1246 1247 default: 1248#if 0 1249 /* 1250 * reset or some invalid combination, but can also 1251 * occur if we use keep-state the wrong way. 1252 */ 1253 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1254 printf("invalid state: 0x%x\n", q->state); 1255#endif 1256 if (dyn_rst_lifetime >= dyn_keepalive_period) 1257 dyn_rst_lifetime = dyn_keepalive_period - 1; 1258 q->expire = time_uptime + dyn_rst_lifetime; 1259 break; 1260 } 1261 } else if (pkt->proto == IPPROTO_UDP) { 1262 q->expire = time_uptime + dyn_udp_lifetime; 1263 } else { 1264 /* other protocols */ 1265 q->expire = time_uptime + dyn_short_lifetime; 1266 } 1267done: 1268 if (match_direction) 1269 *match_direction = dir; 1270 return q; 1271} 1272 1273static ipfw_dyn_rule * 1274lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1275 struct tcphdr *tcp) 1276{ 1277 ipfw_dyn_rule *q; 1278 1279 IPFW_DYN_LOCK(); 1280 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1281 if (q == NULL) 1282 IPFW_DYN_UNLOCK(); 1283 /* NB: return table locked when q is not NULL */ 1284 return q; 1285} 1286 1287static void 1288realloc_dynamic_table(void) 1289{ 1290 IPFW_DYN_LOCK_ASSERT(); 1291 1292 /* 1293 * Try reallocation, make sure we have a power of 2 and do 1294 * not allow more than 64k entries. In case of overflow, 1295 * default to 1024. 1296 */ 1297 1298 if (dyn_buckets > 65536) 1299 dyn_buckets = 1024; 1300 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1301 dyn_buckets = curr_dyn_buckets; /* reset */ 1302 return; 1303 } 1304 curr_dyn_buckets = dyn_buckets; 1305 if (ipfw_dyn_v != NULL) 1306 free(ipfw_dyn_v, M_IPFW); 1307 for (;;) { 1308 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1309 M_IPFW, M_NOWAIT | M_ZERO); 1310 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1311 break; 1312 curr_dyn_buckets /= 2; 1313 } 1314} 1315 1316/** 1317 * Install state of type 'type' for a dynamic session. 1318 * The hash table contains two type of rules: 1319 * - regular rules (O_KEEP_STATE) 1320 * - rules for sessions with limited number of sess per user 1321 * (O_LIMIT). When they are created, the parent is 1322 * increased by 1, and decreased on delete. In this case, 1323 * the third parameter is the parent rule and not the chain. 1324 * - "parent" rules for the above (O_LIMIT_PARENT). 1325 */ 1326static ipfw_dyn_rule * 1327add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1328{ 1329 ipfw_dyn_rule *r; 1330 int i; 1331 1332 IPFW_DYN_LOCK_ASSERT(); 1333 1334 if (ipfw_dyn_v == NULL || 1335 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1336 realloc_dynamic_table(); 1337 if (ipfw_dyn_v == NULL) 1338 return NULL; /* failed ! */ 1339 } 1340 i = hash_packet(id); 1341 1342 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1343 if (r == NULL) { 1344 printf ("ipfw: sorry cannot allocate state\n"); 1345 return NULL; 1346 } 1347 1348 /* increase refcount on parent, and set pointer */ 1349 if (dyn_type == O_LIMIT) { 1350 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1351 if ( parent->dyn_type != O_LIMIT_PARENT) 1352 panic("invalid parent"); 1353 parent->count++; 1354 r->parent = parent; 1355 rule = parent->rule; 1356 } 1357 1358 r->id = *id; 1359 r->expire = time_uptime + dyn_syn_lifetime; 1360 r->rule = rule; 1361 r->dyn_type = dyn_type; 1362 r->pcnt = r->bcnt = 0; 1363 r->count = 0; 1364 1365 r->bucket = i; 1366 r->next = ipfw_dyn_v[i]; 1367 ipfw_dyn_v[i] = r; 1368 dyn_count++; 1369 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1370 dyn_type, 1371 (r->id.src_ip), (r->id.src_port), 1372 (r->id.dst_ip), (r->id.dst_port), 1373 dyn_count ); ) 1374 return r; 1375} 1376 1377/** 1378 * lookup dynamic parent rule using pkt and rule as search keys. 1379 * If the lookup fails, then install one. 1380 */ 1381static ipfw_dyn_rule * 1382lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1383{ 1384 ipfw_dyn_rule *q; 1385 int i; 1386 1387 IPFW_DYN_LOCK_ASSERT(); 1388 1389 if (ipfw_dyn_v) { 1390 int is_v6 = IS_IP6_FLOW_ID(pkt); 1391 i = hash_packet( pkt ); 1392 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1393 if (q->dyn_type == O_LIMIT_PARENT && 1394 rule== q->rule && 1395 pkt->proto == q->id.proto && 1396 pkt->src_port == q->id.src_port && 1397 pkt->dst_port == q->id.dst_port && 1398 ( 1399 (is_v6 && 1400 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1401 &(q->id.src_ip6)) && 1402 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1403 &(q->id.dst_ip6))) || 1404 (!is_v6 && 1405 pkt->src_ip == q->id.src_ip && 1406 pkt->dst_ip == q->id.dst_ip) 1407 ) 1408 ) { 1409 q->expire = time_uptime + dyn_short_lifetime; 1410 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1411 return q; 1412 } 1413 } 1414 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1415} 1416 1417/** 1418 * Install dynamic state for rule type cmd->o.opcode 1419 * 1420 * Returns 1 (failure) if state is not installed because of errors or because 1421 * session limitations are enforced. 1422 */ 1423static int 1424install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1425 struct ip_fw_args *args, uint32_t tablearg) 1426{ 1427 static int last_log; 1428 ipfw_dyn_rule *q; 1429 struct in_addr da; 1430 char src[48], dst[48]; 1431 1432 src[0] = '\0'; 1433 dst[0] = '\0'; 1434 1435 DEB( 1436 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1437 __func__, cmd->o.opcode, 1438 (args->f_id.src_ip), (args->f_id.src_port), 1439 (args->f_id.dst_ip), (args->f_id.dst_port)); 1440 ) 1441 1442 IPFW_DYN_LOCK(); 1443 1444 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1445 1446 if (q != NULL) { /* should never occur */ 1447 if (last_log != time_uptime) { 1448 last_log = time_uptime; 1449 printf("ipfw: %s: entry already present, done\n", 1450 __func__); 1451 } 1452 IPFW_DYN_UNLOCK(); 1453 return (0); 1454 } 1455 1456 if (dyn_count >= dyn_max) 1457 /* Run out of slots, try to remove any expired rule. */ 1458 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1459 1460 if (dyn_count >= dyn_max) { 1461 if (last_log != time_uptime) { 1462 last_log = time_uptime; 1463 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1464 } 1465 IPFW_DYN_UNLOCK(); 1466 return (1); /* cannot install, notify caller */ 1467 } 1468 1469 switch (cmd->o.opcode) { 1470 case O_KEEP_STATE: /* bidir rule */ 1471 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1472 break; 1473 1474 case O_LIMIT: { /* limit number of sessions */ 1475 struct ipfw_flow_id id; 1476 ipfw_dyn_rule *parent; 1477 uint32_t conn_limit; 1478 uint16_t limit_mask = cmd->limit_mask; 1479 1480 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1481 tablearg : cmd->conn_limit; 1482 1483 DEB( 1484 if (cmd->conn_limit == IP_FW_TABLEARG) 1485 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1486 "(tablearg)\n", __func__, conn_limit); 1487 else 1488 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1489 __func__, conn_limit); 1490 ) 1491 1492 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1493 id.proto = args->f_id.proto; 1494 id.addr_type = args->f_id.addr_type; 1495 1496 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1497 if (limit_mask & DYN_SRC_ADDR) 1498 id.src_ip6 = args->f_id.src_ip6; 1499 if (limit_mask & DYN_DST_ADDR) 1500 id.dst_ip6 = args->f_id.dst_ip6; 1501 } else { 1502 if (limit_mask & DYN_SRC_ADDR) 1503 id.src_ip = args->f_id.src_ip; 1504 if (limit_mask & DYN_DST_ADDR) 1505 id.dst_ip = args->f_id.dst_ip; 1506 } 1507 if (limit_mask & DYN_SRC_PORT) 1508 id.src_port = args->f_id.src_port; 1509 if (limit_mask & DYN_DST_PORT) 1510 id.dst_port = args->f_id.dst_port; 1511 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1512 printf("ipfw: %s: add parent failed\n", __func__); 1513 IPFW_DYN_UNLOCK(); 1514 return (1); 1515 } 1516 1517 if (parent->count >= conn_limit) { 1518 /* See if we can remove some expired rule. */ 1519 remove_dyn_rule(rule, parent); 1520 if (parent->count >= conn_limit) { 1521 if (fw_verbose && last_log != time_uptime) { 1522 last_log = time_uptime; 1523#ifdef INET6 1524 /* 1525 * XXX IPv6 flows are not 1526 * supported yet. 1527 */ 1528 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1529 snprintf(src, sizeof(src), 1530 "[%s]", ip6_sprintf( 1531 &args->f_id.src_ip6)); 1532 snprintf(dst, sizeof(dst), 1533 "[%s]", ip6_sprintf( 1534 &args->f_id.dst_ip6)); 1535 } else 1536#endif 1537 { 1538 da.s_addr = 1539 htonl(args->f_id.src_ip); 1540 inet_ntoa_r(da, src); 1541 da.s_addr = 1542 htonl(args->f_id.dst_ip); 1543 inet_ntoa_r(da, dst); 1544 } 1545 log(LOG_SECURITY | LOG_DEBUG, 1546 "%s %s:%u -> %s:%u, %s\n", 1547 "drop session", 1548 src, (args->f_id.src_port), 1549 dst, (args->f_id.dst_port), 1550 "too many entries"); 1551 } 1552 IPFW_DYN_UNLOCK(); 1553 return (1); 1554 } 1555 } 1556 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1557 break; 1558 } 1559 default: 1560 printf("ipfw: %s: unknown dynamic rule type %u\n", 1561 __func__, cmd->o.opcode); 1562 IPFW_DYN_UNLOCK(); 1563 return (1); 1564 } 1565 1566 /* XXX just set lifetime */ 1567 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1568 1569 IPFW_DYN_UNLOCK(); 1570 return (0); 1571} 1572 1573/* 1574 * Generate a TCP packet, containing either a RST or a keepalive. 1575 * When flags & TH_RST, we are sending a RST packet, because of a 1576 * "reset" action matched the packet. 1577 * Otherwise we are sending a keepalive, and flags & TH_ 1578 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1579 * so that MAC can label the reply appropriately. 1580 */ 1581static struct mbuf * 1582send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1583 u_int32_t ack, int flags) 1584{ 1585 struct mbuf *m; 1586 struct ip *ip; 1587 struct tcphdr *tcp; 1588 1589 MGETHDR(m, M_DONTWAIT, MT_DATA); 1590 if (m == 0) 1591 return (NULL); 1592 m->m_pkthdr.rcvif = (struct ifnet *)0; 1593 1594#ifdef MAC 1595 if (replyto != NULL) 1596 mac_create_mbuf_netlayer(replyto, m); 1597 else 1598 mac_create_mbuf_from_firewall(m); 1599#else 1600 (void)replyto; /* don't warn about unused arg */ 1601#endif 1602 1603 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1604 m->m_data += max_linkhdr; 1605 1606 ip = mtod(m, struct ip *); 1607 bzero(ip, m->m_len); 1608 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1609 ip->ip_p = IPPROTO_TCP; 1610 tcp->th_off = 5; 1611 /* 1612 * Assume we are sending a RST (or a keepalive in the reverse 1613 * direction), swap src and destination addresses and ports. 1614 */ 1615 ip->ip_src.s_addr = htonl(id->dst_ip); 1616 ip->ip_dst.s_addr = htonl(id->src_ip); 1617 tcp->th_sport = htons(id->dst_port); 1618 tcp->th_dport = htons(id->src_port); 1619 if (flags & TH_RST) { /* we are sending a RST */ 1620 if (flags & TH_ACK) { 1621 tcp->th_seq = htonl(ack); 1622 tcp->th_ack = htonl(0); 1623 tcp->th_flags = TH_RST; 1624 } else { 1625 if (flags & TH_SYN) 1626 seq++; 1627 tcp->th_seq = htonl(0); 1628 tcp->th_ack = htonl(seq); 1629 tcp->th_flags = TH_RST | TH_ACK; 1630 } 1631 } else { 1632 /* 1633 * We are sending a keepalive. flags & TH_SYN determines 1634 * the direction, forward if set, reverse if clear. 1635 * NOTE: seq and ack are always assumed to be correct 1636 * as set by the caller. This may be confusing... 1637 */ 1638 if (flags & TH_SYN) { 1639 /* 1640 * we have to rewrite the correct addresses! 1641 */ 1642 ip->ip_dst.s_addr = htonl(id->dst_ip); 1643 ip->ip_src.s_addr = htonl(id->src_ip); 1644 tcp->th_dport = htons(id->dst_port); 1645 tcp->th_sport = htons(id->src_port); 1646 } 1647 tcp->th_seq = htonl(seq); 1648 tcp->th_ack = htonl(ack); 1649 tcp->th_flags = TH_ACK; 1650 } 1651 /* 1652 * set ip_len to the payload size so we can compute 1653 * the tcp checksum on the pseudoheader 1654 * XXX check this, could save a couple of words ? 1655 */ 1656 ip->ip_len = htons(sizeof(struct tcphdr)); 1657 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1658 /* 1659 * now fill fields left out earlier 1660 */ 1661 ip->ip_ttl = ip_defttl; 1662 ip->ip_len = m->m_pkthdr.len; 1663 m->m_flags |= M_SKIP_FIREWALL; 1664 return (m); 1665} 1666 1667/* 1668 * sends a reject message, consuming the mbuf passed as an argument. 1669 */ 1670static void 1671send_reject(struct ip_fw_args *args, int code, int ip_len) 1672{ 1673 1674 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1675 /* We need the IP header in host order for icmp_error(). */ 1676 if (args->eh != NULL) { 1677 struct ip *ip = mtod(args->m, struct ip *); 1678 ip->ip_len = ntohs(ip->ip_len); 1679 ip->ip_off = ntohs(ip->ip_off); 1680 } 1681 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1682 } else if (args->f_id.proto == IPPROTO_TCP) { 1683 struct tcphdr *const tcp = 1684 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1685 if ( (tcp->th_flags & TH_RST) == 0) { 1686 struct mbuf *m; 1687 m = send_pkt(args->m, &(args->f_id), 1688 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1689 tcp->th_flags | TH_RST); 1690 if (m != NULL) 1691 ip_output(m, NULL, NULL, 0, NULL, NULL); 1692 } 1693 m_freem(args->m); 1694 } else 1695 m_freem(args->m); 1696 args->m = NULL; 1697} 1698 1699/** 1700 * 1701 * Given an ip_fw *, lookup_next_rule will return a pointer 1702 * to the next rule, which can be either the jump 1703 * target (for skipto instructions) or the next one in the list (in 1704 * all other cases including a missing jump target). 1705 * The result is also written in the "next_rule" field of the rule. 1706 * Backward jumps are not allowed, so start looking from the next 1707 * rule... 1708 * 1709 * This never returns NULL -- in case we do not have an exact match, 1710 * the next rule is returned. When the ruleset is changed, 1711 * pointers are flushed so we are always correct. 1712 */ 1713 1714static struct ip_fw * 1715lookup_next_rule(struct ip_fw *me) 1716{ 1717 struct ip_fw *rule = NULL; 1718 ipfw_insn *cmd; 1719 1720 /* look for action, in case it is a skipto */ 1721 cmd = ACTION_PTR(me); 1722 if (cmd->opcode == O_LOG) 1723 cmd += F_LEN(cmd); 1724 if (cmd->opcode == O_ALTQ) 1725 cmd += F_LEN(cmd); 1726 if (cmd->opcode == O_TAG) 1727 cmd += F_LEN(cmd); 1728 if ( cmd->opcode == O_SKIPTO ) 1729 for (rule = me->next; rule ; rule = rule->next) 1730 if (rule->rulenum >= cmd->arg1) 1731 break; 1732 if (rule == NULL) /* failure or not a skipto */ 1733 rule = me->next; 1734 me->next_rule = rule; 1735 return rule; 1736} 1737 1738static int 1739add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1740 uint8_t mlen, uint32_t value) 1741{ 1742 struct radix_node_head *rnh; 1743 struct table_entry *ent; 1744 1745 if (tbl >= IPFW_TABLES_MAX) 1746 return (EINVAL); 1747 rnh = ch->tables[tbl]; 1748 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1749 if (ent == NULL) 1750 return (ENOMEM); 1751 ent->value = value; 1752 ent->addr.sin_len = ent->mask.sin_len = 8; 1753 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1754 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1755 IPFW_WLOCK(&layer3_chain); 1756 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1757 NULL) { 1758 IPFW_WUNLOCK(&layer3_chain); 1759 free(ent, M_IPFW_TBL); 1760 return (EEXIST); 1761 } 1762 IPFW_WUNLOCK(&layer3_chain); 1763 return (0); 1764} 1765 1766static int 1767del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1768 uint8_t mlen) 1769{ 1770 struct radix_node_head *rnh; 1771 struct table_entry *ent; 1772 struct sockaddr_in sa, mask; 1773 1774 if (tbl >= IPFW_TABLES_MAX) 1775 return (EINVAL); 1776 rnh = ch->tables[tbl]; 1777 sa.sin_len = mask.sin_len = 8; 1778 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1779 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1780 IPFW_WLOCK(ch); 1781 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1782 if (ent == NULL) { 1783 IPFW_WUNLOCK(ch); 1784 return (ESRCH); 1785 } 1786 IPFW_WUNLOCK(ch); 1787 free(ent, M_IPFW_TBL); 1788 return (0); 1789} 1790 1791static int 1792flush_table_entry(struct radix_node *rn, void *arg) 1793{ 1794 struct radix_node_head * const rnh = arg; 1795 struct table_entry *ent; 1796 1797 ent = (struct table_entry *) 1798 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1799 if (ent != NULL) 1800 free(ent, M_IPFW_TBL); 1801 return (0); 1802} 1803 1804static int 1805flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1806{ 1807 struct radix_node_head *rnh; 1808 1809 IPFW_WLOCK_ASSERT(ch); 1810 1811 if (tbl >= IPFW_TABLES_MAX) 1812 return (EINVAL); 1813 rnh = ch->tables[tbl]; 1814 KASSERT(rnh != NULL, ("NULL IPFW table")); 1815 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1816 return (0); 1817} 1818 1819static void 1820flush_tables(struct ip_fw_chain *ch) 1821{ 1822 uint16_t tbl; 1823 1824 IPFW_WLOCK_ASSERT(ch); 1825 1826 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1827 flush_table(ch, tbl); 1828} 1829 1830static int 1831init_tables(struct ip_fw_chain *ch) 1832{ 1833 int i; 1834 uint16_t j; 1835 1836 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1837 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1838 for (j = 0; j < i; j++) { 1839 (void) flush_table(ch, j); 1840 } 1841 return (ENOMEM); 1842 } 1843 } 1844 return (0); 1845} 1846 1847static int 1848lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1849 uint32_t *val) 1850{ 1851 struct radix_node_head *rnh; 1852 struct table_entry *ent; 1853 struct sockaddr_in sa; 1854 1855 if (tbl >= IPFW_TABLES_MAX) 1856 return (0); 1857 rnh = ch->tables[tbl]; 1858 sa.sin_len = 8; 1859 sa.sin_addr.s_addr = addr; 1860 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1861 if (ent != NULL) { 1862 *val = ent->value; 1863 return (1); 1864 } 1865 return (0); 1866} 1867 1868static int 1869count_table_entry(struct radix_node *rn, void *arg) 1870{ 1871 u_int32_t * const cnt = arg; 1872 1873 (*cnt)++; 1874 return (0); 1875} 1876 1877static int 1878count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1879{ 1880 struct radix_node_head *rnh; 1881 1882 if (tbl >= IPFW_TABLES_MAX) 1883 return (EINVAL); 1884 rnh = ch->tables[tbl]; 1885 *cnt = 0; 1886 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1887 return (0); 1888} 1889 1890static int 1891dump_table_entry(struct radix_node *rn, void *arg) 1892{ 1893 struct table_entry * const n = (struct table_entry *)rn; 1894 ipfw_table * const tbl = arg; 1895 ipfw_table_entry *ent; 1896 1897 if (tbl->cnt == tbl->size) 1898 return (1); 1899 ent = &tbl->ent[tbl->cnt]; 1900 ent->tbl = tbl->tbl; 1901 if (in_nullhost(n->mask.sin_addr)) 1902 ent->masklen = 0; 1903 else 1904 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1905 ent->addr = n->addr.sin_addr.s_addr; 1906 ent->value = n->value; 1907 tbl->cnt++; 1908 return (0); 1909} 1910 1911static int 1912dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1913{ 1914 struct radix_node_head *rnh; 1915 1916 if (tbl->tbl >= IPFW_TABLES_MAX) 1917 return (EINVAL); 1918 rnh = ch->tables[tbl->tbl]; 1919 tbl->cnt = 0; 1920 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1921 return (0); 1922} 1923 1924static void 1925fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1926{ 1927 struct ucred *cr; 1928 1929 if (inp->inp_socket != NULL) { 1930 cr = inp->inp_socket->so_cred; 1931 ugp->fw_prid = jailed(cr) ? 1932 cr->cr_prison->pr_id : -1; 1933 ugp->fw_uid = cr->cr_uid; 1934 ugp->fw_ngroups = cr->cr_ngroups; 1935 bcopy(cr->cr_groups, ugp->fw_groups, 1936 sizeof(ugp->fw_groups)); 1937 } 1938} 1939 1940static int 1941check_uidgid(ipfw_insn_u32 *insn, 1942 int proto, struct ifnet *oif, 1943 struct in_addr dst_ip, u_int16_t dst_port, 1944 struct in_addr src_ip, u_int16_t src_port, 1945 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1946{ 1947 struct inpcbinfo *pi; 1948 int wildcard; 1949 struct inpcb *pcb; 1950 int match; 1951 gid_t *gp; 1952 1953 /* 1954 * Check to see if the UDP or TCP stack supplied us with 1955 * the PCB. If so, rather then holding a lock and looking 1956 * up the PCB, we can use the one that was supplied. 1957 */ 1958 if (inp && *lookup == 0) { 1959 INP_LOCK_ASSERT(inp); 1960 if (inp->inp_socket != NULL) { 1961 fill_ugid_cache(inp, ugp); 1962 *lookup = 1; 1963 } 1964 } 1965 /* 1966 * If we have already been here and the packet has no 1967 * PCB entry associated with it, then we can safely 1968 * assume that this is a no match. 1969 */ 1970 if (*lookup == -1) 1971 return (0); 1972 if (proto == IPPROTO_TCP) { 1973 wildcard = 0; 1974 pi = &tcbinfo; 1975 } else if (proto == IPPROTO_UDP) { 1976 wildcard = INPLOOKUP_WILDCARD; 1977 pi = &udbinfo; 1978 } else 1979 return 0; 1980 match = 0; 1981 if (*lookup == 0) { 1982 INP_INFO_RLOCK(pi); 1983 pcb = (oif) ? 1984 in_pcblookup_hash(pi, 1985 dst_ip, htons(dst_port), 1986 src_ip, htons(src_port), 1987 wildcard, oif) : 1988 in_pcblookup_hash(pi, 1989 src_ip, htons(src_port), 1990 dst_ip, htons(dst_port), 1991 wildcard, NULL); 1992 if (pcb != NULL) { 1993 INP_LOCK(pcb); 1994 if (pcb->inp_socket != NULL) { 1995 fill_ugid_cache(pcb, ugp); 1996 *lookup = 1; 1997 } 1998 INP_UNLOCK(pcb); 1999 } 2000 INP_INFO_RUNLOCK(pi); 2001 if (*lookup == 0) { 2002 /* 2003 * If the lookup did not yield any results, there 2004 * is no sense in coming back and trying again. So 2005 * we can set lookup to -1 and ensure that we wont 2006 * bother the pcb system again. 2007 */ 2008 *lookup = -1; 2009 return (0); 2010 } 2011 } 2012 if (insn->o.opcode == O_UID) 2013 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2014 else if (insn->o.opcode == O_GID) { 2015 for (gp = ugp->fw_groups; 2016 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2017 if (*gp == (gid_t)insn->d[0]) { 2018 match = 1; 2019 break; 2020 } 2021 } else if (insn->o.opcode == O_JAIL) 2022 match = (ugp->fw_prid == (int)insn->d[0]); 2023 return match; 2024} 2025 2026/* 2027 * The main check routine for the firewall. 2028 * 2029 * All arguments are in args so we can modify them and return them 2030 * back to the caller. 2031 * 2032 * Parameters: 2033 * 2034 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2035 * Starts with the IP header. 2036 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2037 * args->oif Outgoing interface, or NULL if packet is incoming. 2038 * The incoming interface is in the mbuf. (in) 2039 * args->divert_rule (in/out) 2040 * Skip up to the first rule past this rule number; 2041 * upon return, non-zero port number for divert or tee. 2042 * 2043 * args->rule Pointer to the last matching rule (in/out) 2044 * args->next_hop Socket we are forwarding to (out). 2045 * args->f_id Addresses grabbed from the packet (out) 2046 * args->cookie a cookie depending on rule action 2047 * 2048 * Return value: 2049 * 2050 * IP_FW_PASS the packet must be accepted 2051 * IP_FW_DENY the packet must be dropped 2052 * IP_FW_DIVERT divert packet, port in m_tag 2053 * IP_FW_TEE tee packet, port in m_tag 2054 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2055 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2056 * 2057 */ 2058 2059int 2060ipfw_chk(struct ip_fw_args *args) 2061{ 2062 /* 2063 * Local variables hold state during the processing of a packet. 2064 * 2065 * IMPORTANT NOTE: to speed up the processing of rules, there 2066 * are some assumption on the values of the variables, which 2067 * are documented here. Should you change them, please check 2068 * the implementation of the various instructions to make sure 2069 * that they still work. 2070 * 2071 * args->eh The MAC header. It is non-null for a layer2 2072 * packet, it is NULL for a layer-3 packet. 2073 * 2074 * m | args->m Pointer to the mbuf, as received from the caller. 2075 * It may change if ipfw_chk() does an m_pullup, or if it 2076 * consumes the packet because it calls send_reject(). 2077 * XXX This has to change, so that ipfw_chk() never modifies 2078 * or consumes the buffer. 2079 * ip is simply an alias of the value of m, and it is kept 2080 * in sync with it (the packet is supposed to start with 2081 * the ip header). 2082 */ 2083 struct mbuf *m = args->m; 2084 struct ip *ip = mtod(m, struct ip *); 2085 2086 /* 2087 * For rules which contain uid/gid or jail constraints, cache 2088 * a copy of the users credentials after the pcb lookup has been 2089 * executed. This will speed up the processing of rules with 2090 * these types of constraints, as well as decrease contention 2091 * on pcb related locks. 2092 */ 2093 struct ip_fw_ugid fw_ugid_cache; 2094 int ugid_lookup = 0; 2095 2096 /* 2097 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2098 * associated with a packet input on a divert socket. This 2099 * will allow to distinguish traffic and its direction when 2100 * it originates from a divert socket. 2101 */ 2102 u_int divinput_flags = 0; 2103 2104 /* 2105 * oif | args->oif If NULL, ipfw_chk has been called on the 2106 * inbound path (ether_input, ip_input). 2107 * If non-NULL, ipfw_chk has been called on the outbound path 2108 * (ether_output, ip_output). 2109 */ 2110 struct ifnet *oif = args->oif; 2111 2112 struct ip_fw *f = NULL; /* matching rule */ 2113 int retval = 0; 2114 2115 /* 2116 * hlen The length of the IP header. 2117 */ 2118 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2119 2120 /* 2121 * offset The offset of a fragment. offset != 0 means that 2122 * we have a fragment at this offset of an IPv4 packet. 2123 * offset == 0 means that (if this is an IPv4 packet) 2124 * this is the first or only fragment. 2125 * For IPv6 offset == 0 means there is no Fragment Header. 2126 * If offset != 0 for IPv6 always use correct mask to 2127 * get the correct offset because we add IP6F_MORE_FRAG 2128 * to be able to dectect the first fragment which would 2129 * otherwise have offset = 0. 2130 */ 2131 u_short offset = 0; 2132 2133 /* 2134 * Local copies of addresses. They are only valid if we have 2135 * an IP packet. 2136 * 2137 * proto The protocol. Set to 0 for non-ip packets, 2138 * or to the protocol read from the packet otherwise. 2139 * proto != 0 means that we have an IPv4 packet. 2140 * 2141 * src_port, dst_port port numbers, in HOST format. Only 2142 * valid for TCP and UDP packets. 2143 * 2144 * src_ip, dst_ip ip addresses, in NETWORK format. 2145 * Only valid for IPv4 packets. 2146 */ 2147 u_int8_t proto; 2148 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2149 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2150 u_int16_t ip_len=0; 2151 int pktlen; 2152 2153 /* 2154 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2155 * MATCH_NONE when checked and not matched (q = NULL), 2156 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2157 */ 2158 int dyn_dir = MATCH_UNKNOWN; 2159 ipfw_dyn_rule *q = NULL; 2160 struct ip_fw_chain *chain = &layer3_chain; 2161 struct m_tag *mtag; 2162 2163 /* 2164 * We store in ulp a pointer to the upper layer protocol header. 2165 * In the ipv4 case this is easy to determine from the header, 2166 * but for ipv6 we might have some additional headers in the middle. 2167 * ulp is NULL if not found. 2168 */ 2169 void *ulp = NULL; /* upper layer protocol pointer. */ 2170 /* XXX ipv6 variables */ 2171 int is_ipv6 = 0; 2172 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2173 /* end of ipv6 variables */ 2174 int is_ipv4 = 0; 2175 2176 if (m->m_flags & M_SKIP_FIREWALL) 2177 return (IP_FW_PASS); /* accept */ 2178 2179 pktlen = m->m_pkthdr.len; 2180 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2181 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2182 2183/* 2184 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2185 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2186 * pointer might become stale after other pullups (but we never use it 2187 * this way). 2188 */ 2189#define PULLUP_TO(len, p, T) \ 2190do { \ 2191 int x = (len) + sizeof(T); \ 2192 if ((m)->m_len < x) { \ 2193 args->m = m = m_pullup(m, x); \ 2194 if (m == NULL) \ 2195 goto pullup_failed; \ 2196 } \ 2197 p = (mtod(m, char *) + (len)); \ 2198} while (0) 2199 2200 /* Identify IP packets and fill up variables. */ 2201 if (pktlen >= sizeof(struct ip6_hdr) && 2202 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2203 mtod(m, struct ip *)->ip_v == 6) { 2204 is_ipv6 = 1; 2205 args->f_id.addr_type = 6; 2206 hlen = sizeof(struct ip6_hdr); 2207 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2208 2209 /* Search extension headers to find upper layer protocols */ 2210 while (ulp == NULL) { 2211 switch (proto) { 2212 case IPPROTO_ICMPV6: 2213 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2214 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2215 break; 2216 2217 case IPPROTO_TCP: 2218 PULLUP_TO(hlen, ulp, struct tcphdr); 2219 dst_port = TCP(ulp)->th_dport; 2220 src_port = TCP(ulp)->th_sport; 2221 args->f_id.flags = TCP(ulp)->th_flags; 2222 break; 2223 2224 case IPPROTO_UDP: 2225 PULLUP_TO(hlen, ulp, struct udphdr); 2226 dst_port = UDP(ulp)->uh_dport; 2227 src_port = UDP(ulp)->uh_sport; 2228 break; 2229 2230 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2231 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2232 ext_hd |= EXT_HOPOPTS; 2233 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2234 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2235 ulp = NULL; 2236 break; 2237 2238 case IPPROTO_ROUTING: /* RFC 2460 */ 2239 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2240 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2241 case 0: 2242 break; 2243 default: 2244 printf("IPFW2: IPV6 - Unknown Routing " 2245 "Header type(%d)\n", 2246 ((struct ip6_rthdr *)ulp)->ip6r_type); 2247 if (fw_deny_unknown_exthdrs) 2248 return (IP_FW_DENY); 2249 break; 2250 } 2251 ext_hd |= EXT_ROUTING; 2252 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2253 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2254 ulp = NULL; 2255 break; 2256 2257 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2258 PULLUP_TO(hlen, ulp, struct ip6_frag); 2259 ext_hd |= EXT_FRAGMENT; 2260 hlen += sizeof (struct ip6_frag); 2261 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2262 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2263 IP6F_OFF_MASK; 2264 /* Add IP6F_MORE_FRAG for offset of first 2265 * fragment to be != 0. */ 2266 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2267 IP6F_MORE_FRAG; 2268 if (offset == 0) { 2269 printf("IPFW2: IPV6 - Invalid Fragment " 2270 "Header\n"); 2271 if (fw_deny_unknown_exthdrs) 2272 return (IP_FW_DENY); 2273 break; 2274 } 2275 args->f_id.frag_id6 = 2276 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2277 ulp = NULL; 2278 break; 2279 2280 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2281 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2282 ext_hd |= EXT_DSTOPTS; 2283 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2284 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2285 ulp = NULL; 2286 break; 2287 2288 case IPPROTO_AH: /* RFC 2402 */ 2289 PULLUP_TO(hlen, ulp, struct ip6_ext); 2290 ext_hd |= EXT_AH; 2291 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2292 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2293 ulp = NULL; 2294 break; 2295 2296 case IPPROTO_ESP: /* RFC 2406 */ 2297 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2298 /* Anything past Seq# is variable length and 2299 * data past this ext. header is encrypted. */ 2300 ext_hd |= EXT_ESP; 2301 break; 2302 2303 case IPPROTO_NONE: /* RFC 2460 */ 2304 PULLUP_TO(hlen, ulp, struct ip6_ext); 2305 /* Packet ends here. if ip6e_len!=0 octets 2306 * must be ignored. */ 2307 break; 2308 2309 case IPPROTO_OSPFIGP: 2310 /* XXX OSPF header check? */ 2311 PULLUP_TO(hlen, ulp, struct ip6_ext); 2312 break; 2313 2314 case IPPROTO_PIM: 2315 /* XXX PIM header check? */ 2316 PULLUP_TO(hlen, ulp, struct pim); 2317 break; 2318 2319 case IPPROTO_CARP: 2320 PULLUP_TO(hlen, ulp, struct carp_header); 2321 if (((struct carp_header *)ulp)->carp_version != 2322 CARP_VERSION) 2323 return (IP_FW_DENY); 2324 if (((struct carp_header *)ulp)->carp_type != 2325 CARP_ADVERTISEMENT) 2326 return (IP_FW_DENY); 2327 break; 2328 2329 case IPPROTO_IPV6: /* RFC 2893 */ 2330 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2331 break; 2332 2333 case IPPROTO_IPV4: /* RFC 2893 */ 2334 PULLUP_TO(hlen, ulp, struct ip); 2335 break; 2336 2337 default: 2338 printf("IPFW2: IPV6 - Unknown Extension " 2339 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2340 if (fw_deny_unknown_exthdrs) 2341 return (IP_FW_DENY); 2342 PULLUP_TO(hlen, ulp, struct ip6_ext); 2343 break; 2344 } /*switch */ 2345 } 2346 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2347 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2348 args->f_id.src_ip = 0; 2349 args->f_id.dst_ip = 0; 2350 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2351 } else if (pktlen >= sizeof(struct ip) && 2352 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2353 mtod(m, struct ip *)->ip_v == 4) { 2354 is_ipv4 = 1; 2355 ip = mtod(m, struct ip *); 2356 hlen = ip->ip_hl << 2; 2357 args->f_id.addr_type = 4; 2358 2359 /* 2360 * Collect parameters into local variables for faster matching. 2361 */ 2362 proto = ip->ip_p; 2363 src_ip = ip->ip_src; 2364 dst_ip = ip->ip_dst; 2365 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2366 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2367 ip_len = ntohs(ip->ip_len); 2368 } else { 2369 offset = ip->ip_off & IP_OFFMASK; 2370 ip_len = ip->ip_len; 2371 } 2372 pktlen = ip_len < pktlen ? ip_len : pktlen; 2373 2374 if (offset == 0) { 2375 switch (proto) { 2376 case IPPROTO_TCP: 2377 PULLUP_TO(hlen, ulp, struct tcphdr); 2378 dst_port = TCP(ulp)->th_dport; 2379 src_port = TCP(ulp)->th_sport; 2380 args->f_id.flags = TCP(ulp)->th_flags; 2381 break; 2382 2383 case IPPROTO_UDP: 2384 PULLUP_TO(hlen, ulp, struct udphdr); 2385 dst_port = UDP(ulp)->uh_dport; 2386 src_port = UDP(ulp)->uh_sport; 2387 break; 2388 2389 case IPPROTO_ICMP: 2390 PULLUP_TO(hlen, ulp, struct icmphdr); 2391 args->f_id.flags = ICMP(ulp)->icmp_type; 2392 break; 2393 2394 default: 2395 break; 2396 } 2397 } 2398 2399 args->f_id.src_ip = ntohl(src_ip.s_addr); 2400 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2401 } 2402#undef PULLUP_TO 2403 if (proto) { /* we may have port numbers, store them */ 2404 args->f_id.proto = proto; 2405 args->f_id.src_port = src_port = ntohs(src_port); 2406 args->f_id.dst_port = dst_port = ntohs(dst_port); 2407 } 2408 2409 IPFW_RLOCK(chain); 2410 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2411 if (args->rule) { 2412 /* 2413 * Packet has already been tagged. Look for the next rule 2414 * to restart processing. 2415 * 2416 * If fw_one_pass != 0 then just accept it. 2417 * XXX should not happen here, but optimized out in 2418 * the caller. 2419 */ 2420 if (fw_one_pass) { 2421 IPFW_RUNLOCK(chain); 2422 return (IP_FW_PASS); 2423 } 2424 2425 f = args->rule->next_rule; 2426 if (f == NULL) 2427 f = lookup_next_rule(args->rule); 2428 } else { 2429 /* 2430 * Find the starting rule. It can be either the first 2431 * one, or the one after divert_rule if asked so. 2432 */ 2433 int skipto = mtag ? divert_cookie(mtag) : 0; 2434 2435 f = chain->rules; 2436 if (args->eh == NULL && skipto != 0) { 2437 if (skipto >= IPFW_DEFAULT_RULE) { 2438 IPFW_RUNLOCK(chain); 2439 return (IP_FW_DENY); /* invalid */ 2440 } 2441 while (f && f->rulenum <= skipto) 2442 f = f->next; 2443 if (f == NULL) { /* drop packet */ 2444 IPFW_RUNLOCK(chain); 2445 return (IP_FW_DENY); 2446 } 2447 } 2448 } 2449 /* reset divert rule to avoid confusion later */ 2450 if (mtag) { 2451 divinput_flags = divert_info(mtag) & 2452 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2453 m_tag_delete(m, mtag); 2454 } 2455 2456 /* 2457 * Now scan the rules, and parse microinstructions for each rule. 2458 */ 2459 for (; f; f = f->next) { 2460 ipfw_insn *cmd; 2461 uint32_t tablearg = 0; 2462 int l, cmdlen, skip_or; /* skip rest of OR block */ 2463 2464again: 2465 if (set_disable & (1 << f->set) ) 2466 continue; 2467 2468 skip_or = 0; 2469 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2470 l -= cmdlen, cmd += cmdlen) { 2471 int match; 2472 2473 /* 2474 * check_body is a jump target used when we find a 2475 * CHECK_STATE, and need to jump to the body of 2476 * the target rule. 2477 */ 2478 2479check_body: 2480 cmdlen = F_LEN(cmd); 2481 /* 2482 * An OR block (insn_1 || .. || insn_n) has the 2483 * F_OR bit set in all but the last instruction. 2484 * The first match will set "skip_or", and cause 2485 * the following instructions to be skipped until 2486 * past the one with the F_OR bit clear. 2487 */ 2488 if (skip_or) { /* skip this instruction */ 2489 if ((cmd->len & F_OR) == 0) 2490 skip_or = 0; /* next one is good */ 2491 continue; 2492 } 2493 match = 0; /* set to 1 if we succeed */ 2494 2495 switch (cmd->opcode) { 2496 /* 2497 * The first set of opcodes compares the packet's 2498 * fields with some pattern, setting 'match' if a 2499 * match is found. At the end of the loop there is 2500 * logic to deal with F_NOT and F_OR flags associated 2501 * with the opcode. 2502 */ 2503 case O_NOP: 2504 match = 1; 2505 break; 2506 2507 case O_FORWARD_MAC: 2508 printf("ipfw: opcode %d unimplemented\n", 2509 cmd->opcode); 2510 break; 2511 2512 case O_GID: 2513 case O_UID: 2514 case O_JAIL: 2515 /* 2516 * We only check offset == 0 && proto != 0, 2517 * as this ensures that we have a 2518 * packet with the ports info. 2519 */ 2520 if (offset!=0) 2521 break; 2522 if (is_ipv6) /* XXX to be fixed later */ 2523 break; 2524 if (proto == IPPROTO_TCP || 2525 proto == IPPROTO_UDP) 2526 match = check_uidgid( 2527 (ipfw_insn_u32 *)cmd, 2528 proto, oif, 2529 dst_ip, dst_port, 2530 src_ip, src_port, &fw_ugid_cache, 2531 &ugid_lookup, args->inp); 2532 break; 2533 2534 case O_RECV: 2535 match = iface_match(m->m_pkthdr.rcvif, 2536 (ipfw_insn_if *)cmd); 2537 break; 2538 2539 case O_XMIT: 2540 match = iface_match(oif, (ipfw_insn_if *)cmd); 2541 break; 2542 2543 case O_VIA: 2544 match = iface_match(oif ? oif : 2545 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2546 break; 2547 2548 case O_MACADDR2: 2549 if (args->eh != NULL) { /* have MAC header */ 2550 u_int32_t *want = (u_int32_t *) 2551 ((ipfw_insn_mac *)cmd)->addr; 2552 u_int32_t *mask = (u_int32_t *) 2553 ((ipfw_insn_mac *)cmd)->mask; 2554 u_int32_t *hdr = (u_int32_t *)args->eh; 2555 2556 match = 2557 ( want[0] == (hdr[0] & mask[0]) && 2558 want[1] == (hdr[1] & mask[1]) && 2559 want[2] == (hdr[2] & mask[2]) ); 2560 } 2561 break; 2562 2563 case O_MAC_TYPE: 2564 if (args->eh != NULL) { 2565 u_int16_t t = 2566 ntohs(args->eh->ether_type); 2567 u_int16_t *p = 2568 ((ipfw_insn_u16 *)cmd)->ports; 2569 int i; 2570 2571 for (i = cmdlen - 1; !match && i>0; 2572 i--, p += 2) 2573 match = (t>=p[0] && t<=p[1]); 2574 } 2575 break; 2576 2577 case O_FRAG: 2578 match = (offset != 0); 2579 break; 2580 2581 case O_IN: /* "out" is "not in" */ 2582 match = (oif == NULL); 2583 break; 2584 2585 case O_LAYER2: 2586 match = (args->eh != NULL); 2587 break; 2588 2589 case O_DIVERTED: 2590 match = (cmd->arg1 & 1 && divinput_flags & 2591 IP_FW_DIVERT_LOOPBACK_FLAG) || 2592 (cmd->arg1 & 2 && divinput_flags & 2593 IP_FW_DIVERT_OUTPUT_FLAG); 2594 break; 2595 2596 case O_PROTO: 2597 /* 2598 * We do not allow an arg of 0 so the 2599 * check of "proto" only suffices. 2600 */ 2601 match = (proto == cmd->arg1); 2602 break; 2603 2604 case O_IP_SRC: 2605 match = is_ipv4 && 2606 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2607 src_ip.s_addr); 2608 break; 2609 2610 case O_IP_SRC_LOOKUP: 2611 case O_IP_DST_LOOKUP: 2612 if (is_ipv4) { 2613 uint32_t a = 2614 (cmd->opcode == O_IP_DST_LOOKUP) ? 2615 dst_ip.s_addr : src_ip.s_addr; 2616 uint32_t v; 2617 2618 match = lookup_table(chain, cmd->arg1, a, 2619 &v); 2620 if (!match) 2621 break; 2622 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2623 match = 2624 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2625 else 2626 tablearg = v; 2627 } 2628 break; 2629 2630 case O_IP_SRC_MASK: 2631 case O_IP_DST_MASK: 2632 if (is_ipv4) { 2633 uint32_t a = 2634 (cmd->opcode == O_IP_DST_MASK) ? 2635 dst_ip.s_addr : src_ip.s_addr; 2636 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2637 int i = cmdlen-1; 2638 2639 for (; !match && i>0; i-= 2, p+= 2) 2640 match = (p[0] == (a & p[1])); 2641 } 2642 break; 2643 2644 case O_IP_SRC_ME: 2645 if (is_ipv4) { 2646 struct ifnet *tif; 2647 2648 INADDR_TO_IFP(src_ip, tif); 2649 match = (tif != NULL); 2650 } 2651 break; 2652 2653 case O_IP_DST_SET: 2654 case O_IP_SRC_SET: 2655 if (is_ipv4) { 2656 u_int32_t *d = (u_int32_t *)(cmd+1); 2657 u_int32_t addr = 2658 cmd->opcode == O_IP_DST_SET ? 2659 args->f_id.dst_ip : 2660 args->f_id.src_ip; 2661 2662 if (addr < d[0]) 2663 break; 2664 addr -= d[0]; /* subtract base */ 2665 match = (addr < cmd->arg1) && 2666 ( d[ 1 + (addr>>5)] & 2667 (1<<(addr & 0x1f)) ); 2668 } 2669 break; 2670 2671 case O_IP_DST: 2672 match = is_ipv4 && 2673 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2674 dst_ip.s_addr); 2675 break; 2676 2677 case O_IP_DST_ME: 2678 if (is_ipv4) { 2679 struct ifnet *tif; 2680 2681 INADDR_TO_IFP(dst_ip, tif); 2682 match = (tif != NULL); 2683 } 2684 break; 2685 2686 case O_IP_SRCPORT: 2687 case O_IP_DSTPORT: 2688 /* 2689 * offset == 0 && proto != 0 is enough 2690 * to guarantee that we have a 2691 * packet with port info. 2692 */ 2693 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2694 && offset == 0) { 2695 u_int16_t x = 2696 (cmd->opcode == O_IP_SRCPORT) ? 2697 src_port : dst_port ; 2698 u_int16_t *p = 2699 ((ipfw_insn_u16 *)cmd)->ports; 2700 int i; 2701 2702 for (i = cmdlen - 1; !match && i>0; 2703 i--, p += 2) 2704 match = (x>=p[0] && x<=p[1]); 2705 } 2706 break; 2707 2708 case O_ICMPTYPE: 2709 match = (offset == 0 && proto==IPPROTO_ICMP && 2710 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2711 break; 2712 2713#ifdef INET6 2714 case O_ICMP6TYPE: 2715 match = is_ipv6 && offset == 0 && 2716 proto==IPPROTO_ICMPV6 && 2717 icmp6type_match( 2718 ICMP6(ulp)->icmp6_type, 2719 (ipfw_insn_u32 *)cmd); 2720 break; 2721#endif /* INET6 */ 2722 2723 case O_IPOPT: 2724 match = (is_ipv4 && 2725 ipopts_match(mtod(m, struct ip *), cmd) ); 2726 break; 2727 2728 case O_IPVER: 2729 match = (is_ipv4 && 2730 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2731 break; 2732 2733 case O_IPID: 2734 case O_IPLEN: 2735 case O_IPTTL: 2736 if (is_ipv4) { /* only for IP packets */ 2737 uint16_t x; 2738 uint16_t *p; 2739 int i; 2740 2741 if (cmd->opcode == O_IPLEN) 2742 x = ip_len; 2743 else if (cmd->opcode == O_IPTTL) 2744 x = mtod(m, struct ip *)->ip_ttl; 2745 else /* must be IPID */ 2746 x = ntohs(mtod(m, struct ip *)->ip_id); 2747 if (cmdlen == 1) { 2748 match = (cmd->arg1 == x); 2749 break; 2750 } 2751 /* otherwise we have ranges */ 2752 p = ((ipfw_insn_u16 *)cmd)->ports; 2753 i = cmdlen - 1; 2754 for (; !match && i>0; i--, p += 2) 2755 match = (x >= p[0] && x <= p[1]); 2756 } 2757 break; 2758 2759 case O_IPPRECEDENCE: 2760 match = (is_ipv4 && 2761 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2762 break; 2763 2764 case O_IPTOS: 2765 match = (is_ipv4 && 2766 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2767 break; 2768 2769 case O_TCPDATALEN: 2770 if (proto == IPPROTO_TCP && offset == 0) { 2771 struct tcphdr *tcp; 2772 uint16_t x; 2773 uint16_t *p; 2774 int i; 2775 2776 tcp = TCP(ulp); 2777 x = ip_len - 2778 ((ip->ip_hl + tcp->th_off) << 2); 2779 if (cmdlen == 1) { 2780 match = (cmd->arg1 == x); 2781 break; 2782 } 2783 /* otherwise we have ranges */ 2784 p = ((ipfw_insn_u16 *)cmd)->ports; 2785 i = cmdlen - 1; 2786 for (; !match && i>0; i--, p += 2) 2787 match = (x >= p[0] && x <= p[1]); 2788 } 2789 break; 2790 2791 case O_TCPFLAGS: 2792 match = (proto == IPPROTO_TCP && offset == 0 && 2793 flags_match(cmd, TCP(ulp)->th_flags)); 2794 break; 2795 2796 case O_TCPOPTS: 2797 match = (proto == IPPROTO_TCP && offset == 0 && 2798 tcpopts_match(TCP(ulp), cmd)); 2799 break; 2800 2801 case O_TCPSEQ: 2802 match = (proto == IPPROTO_TCP && offset == 0 && 2803 ((ipfw_insn_u32 *)cmd)->d[0] == 2804 TCP(ulp)->th_seq); 2805 break; 2806 2807 case O_TCPACK: 2808 match = (proto == IPPROTO_TCP && offset == 0 && 2809 ((ipfw_insn_u32 *)cmd)->d[0] == 2810 TCP(ulp)->th_ack); 2811 break; 2812 2813 case O_TCPWIN: 2814 match = (proto == IPPROTO_TCP && offset == 0 && 2815 cmd->arg1 == TCP(ulp)->th_win); 2816 break; 2817 2818 case O_ESTAB: 2819 /* reject packets which have SYN only */ 2820 /* XXX should i also check for TH_ACK ? */ 2821 match = (proto == IPPROTO_TCP && offset == 0 && 2822 (TCP(ulp)->th_flags & 2823 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2824 break; 2825 2826 case O_ALTQ: { 2827 struct altq_tag *at; 2828 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2829 2830 match = 1; 2831 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2832 if (mtag != NULL) 2833 break; 2834 mtag = m_tag_get(PACKET_TAG_PF_QID, 2835 sizeof(struct altq_tag), 2836 M_NOWAIT); 2837 if (mtag == NULL) { 2838 /* 2839 * Let the packet fall back to the 2840 * default ALTQ. 2841 */ 2842 break; 2843 } 2844 at = (struct altq_tag *)(mtag+1); 2845 at->qid = altq->qid; 2846 if (is_ipv4) 2847 at->af = AF_INET; 2848 else 2849 at->af = AF_LINK; 2850 at->hdr = ip; 2851 m_tag_prepend(m, mtag); 2852 break; 2853 } 2854 2855 case O_LOG: 2856 if (fw_verbose) 2857 ipfw_log(f, hlen, args, m, 2858 oif, offset, tablearg); 2859 match = 1; 2860 break; 2861 2862 case O_PROB: 2863 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2864 break; 2865 2866 case O_VERREVPATH: 2867 /* Outgoing packets automatically pass/match */ 2868 match = ((oif != NULL) || 2869 (m->m_pkthdr.rcvif == NULL) || 2870 ( 2871#ifdef INET6 2872 is_ipv6 ? 2873 verify_path6(&(args->f_id.src_ip6), 2874 m->m_pkthdr.rcvif) : 2875#endif 2876 verify_path(src_ip, m->m_pkthdr.rcvif))); 2877 break; 2878 2879 case O_VERSRCREACH: 2880 /* Outgoing packets automatically pass/match */ 2881 match = (hlen > 0 && ((oif != NULL) || 2882#ifdef INET6 2883 is_ipv6 ? 2884 verify_path6(&(args->f_id.src_ip6), 2885 NULL) : 2886#endif 2887 verify_path(src_ip, NULL))); 2888 break; 2889 2890 case O_ANTISPOOF: 2891 /* Outgoing packets automatically pass/match */ 2892 if (oif == NULL && hlen > 0 && 2893 ( (is_ipv4 && in_localaddr(src_ip)) 2894#ifdef INET6 2895 || (is_ipv6 && 2896 in6_localaddr(&(args->f_id.src_ip6))) 2897#endif 2898 )) 2899 match = 2900#ifdef INET6 2901 is_ipv6 ? verify_path6( 2902 &(args->f_id.src_ip6), 2903 m->m_pkthdr.rcvif) : 2904#endif 2905 verify_path(src_ip, 2906 m->m_pkthdr.rcvif); 2907 else 2908 match = 1; 2909 break; 2910 2911 case O_IPSEC: 2912#ifdef FAST_IPSEC 2913 match = (m_tag_find(m, 2914 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2915#endif 2916#ifdef IPSEC 2917 match = (ipsec_getnhist(m) != 0); 2918#endif 2919 /* otherwise no match */ 2920 break; 2921 2922#ifdef INET6 2923 case O_IP6_SRC: 2924 match = is_ipv6 && 2925 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2926 &((ipfw_insn_ip6 *)cmd)->addr6); 2927 break; 2928 2929 case O_IP6_DST: 2930 match = is_ipv6 && 2931 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2932 &((ipfw_insn_ip6 *)cmd)->addr6); 2933 break; 2934 case O_IP6_SRC_MASK: 2935 case O_IP6_DST_MASK: 2936 if (is_ipv6) { 2937 int i = cmdlen - 1; 2938 struct in6_addr p; 2939 struct in6_addr *d = 2940 &((ipfw_insn_ip6 *)cmd)->addr6; 2941 2942 for (; !match && i > 0; d += 2, 2943 i -= F_INSN_SIZE(struct in6_addr) 2944 * 2) { 2945 p = (cmd->opcode == 2946 O_IP6_SRC_MASK) ? 2947 args->f_id.src_ip6: 2948 args->f_id.dst_ip6; 2949 APPLY_MASK(&p, &d[1]); 2950 match = 2951 IN6_ARE_ADDR_EQUAL(&d[0], 2952 &p); 2953 } 2954 } 2955 break; 2956 2957 case O_IP6_SRC_ME: 2958 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2959 break; 2960 2961 case O_IP6_DST_ME: 2962 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2963 break; 2964 2965 case O_FLOW6ID: 2966 match = is_ipv6 && 2967 flow6id_match(args->f_id.flow_id6, 2968 (ipfw_insn_u32 *) cmd); 2969 break; 2970 2971 case O_EXT_HDR: 2972 match = is_ipv6 && 2973 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2974 break; 2975 2976 case O_IP6: 2977 match = is_ipv6; 2978 break; 2979#endif 2980 2981 case O_IP4: 2982 match = is_ipv4; 2983 break; 2984 2985 case O_TAG: { 2986 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2987 tablearg : cmd->arg1; 2988 2989 /* Packet is already tagged with this tag? */ 2990 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2991 2992 /* We have `untag' action when F_NOT flag is 2993 * present. And we must remove this mtag from 2994 * mbuf and reset `match' to zero (`match' will 2995 * be inversed later). 2996 * Otherwise we should allocate new mtag and 2997 * push it into mbuf. 2998 */ 2999 if (cmd->len & F_NOT) { /* `untag' action */ 3000 if (mtag != NULL) 3001 m_tag_delete(m, mtag); 3002 } else if (mtag == NULL) { 3003 if ((mtag = m_tag_alloc(MTAG_IPFW, 3004 tag, 0, M_NOWAIT)) != NULL) 3005 m_tag_prepend(m, mtag); 3006 } 3007 match = (cmd->len & F_NOT) ? 0: 1; 3008 break; 3009 } 3010 3011 case O_TAGGED: { 3012 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3013 tablearg : cmd->arg1; 3014 3015 if (cmdlen == 1) { 3016 match = m_tag_locate(m, MTAG_IPFW, 3017 tag, NULL) != NULL; 3018 break; 3019 } 3020 3021 /* we have ranges */ 3022 for (mtag = m_tag_first(m); 3023 mtag != NULL && !match; 3024 mtag = m_tag_next(m, mtag)) { 3025 uint16_t *p; 3026 int i; 3027 3028 if (mtag->m_tag_cookie != MTAG_IPFW) 3029 continue; 3030 3031 p = ((ipfw_insn_u16 *)cmd)->ports; 3032 i = cmdlen - 1; 3033 for(; !match && i > 0; i--, p += 2) 3034 match = 3035 mtag->m_tag_id >= p[0] && 3036 mtag->m_tag_id <= p[1]; 3037 } 3038 break; 3039 } 3040 3041 /* 3042 * The second set of opcodes represents 'actions', 3043 * i.e. the terminal part of a rule once the packet 3044 * matches all previous patterns. 3045 * Typically there is only one action for each rule, 3046 * and the opcode is stored at the end of the rule 3047 * (but there are exceptions -- see below). 3048 * 3049 * In general, here we set retval and terminate the 3050 * outer loop (would be a 'break 3' in some language, 3051 * but we need to do a 'goto done'). 3052 * 3053 * Exceptions: 3054 * O_COUNT and O_SKIPTO actions: 3055 * instead of terminating, we jump to the next rule 3056 * ('goto next_rule', equivalent to a 'break 2'), 3057 * or to the SKIPTO target ('goto again' after 3058 * having set f, cmd and l), respectively. 3059 * 3060 * O_TAG, O_LOG and O_ALTQ action parameters: 3061 * perform some action and set match = 1; 3062 * 3063 * O_LIMIT and O_KEEP_STATE: these opcodes are 3064 * not real 'actions', and are stored right 3065 * before the 'action' part of the rule. 3066 * These opcodes try to install an entry in the 3067 * state tables; if successful, we continue with 3068 * the next opcode (match=1; break;), otherwise 3069 * the packet * must be dropped 3070 * ('goto done' after setting retval); 3071 * 3072 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3073 * cause a lookup of the state table, and a jump 3074 * to the 'action' part of the parent rule 3075 * ('goto check_body') if an entry is found, or 3076 * (CHECK_STATE only) a jump to the next rule if 3077 * the entry is not found ('goto next_rule'). 3078 * The result of the lookup is cached to make 3079 * further instances of these opcodes are 3080 * effectively NOPs. 3081 */ 3082 case O_LIMIT: 3083 case O_KEEP_STATE: 3084 if (install_state(f, 3085 (ipfw_insn_limit *)cmd, args, tablearg)) { 3086 retval = IP_FW_DENY; 3087 goto done; /* error/limit violation */ 3088 } 3089 match = 1; 3090 break; 3091 3092 case O_PROBE_STATE: 3093 case O_CHECK_STATE: 3094 /* 3095 * dynamic rules are checked at the first 3096 * keep-state or check-state occurrence, 3097 * with the result being stored in dyn_dir. 3098 * The compiler introduces a PROBE_STATE 3099 * instruction for us when we have a 3100 * KEEP_STATE (because PROBE_STATE needs 3101 * to be run first). 3102 */ 3103 if (dyn_dir == MATCH_UNKNOWN && 3104 (q = lookup_dyn_rule(&args->f_id, 3105 &dyn_dir, proto == IPPROTO_TCP ? 3106 TCP(ulp) : NULL)) 3107 != NULL) { 3108 /* 3109 * Found dynamic entry, update stats 3110 * and jump to the 'action' part of 3111 * the parent rule. 3112 */ 3113 q->pcnt++; 3114 q->bcnt += pktlen; 3115 f = q->rule; 3116 cmd = ACTION_PTR(f); 3117 l = f->cmd_len - f->act_ofs; 3118 IPFW_DYN_UNLOCK(); 3119 goto check_body; 3120 } 3121 /* 3122 * Dynamic entry not found. If CHECK_STATE, 3123 * skip to next rule, if PROBE_STATE just 3124 * ignore and continue with next opcode. 3125 */ 3126 if (cmd->opcode == O_CHECK_STATE) 3127 goto next_rule; 3128 match = 1; 3129 break; 3130 3131 case O_ACCEPT: 3132 retval = 0; /* accept */ 3133 goto done; 3134 3135 case O_PIPE: 3136 case O_QUEUE: 3137 args->rule = f; /* report matching rule */ 3138 if (cmd->arg1 == IP_FW_TABLEARG) 3139 args->cookie = tablearg; 3140 else 3141 args->cookie = cmd->arg1; 3142 retval = IP_FW_DUMMYNET; 3143 goto done; 3144 3145 case O_DIVERT: 3146 case O_TEE: { 3147 struct divert_tag *dt; 3148 3149 if (args->eh) /* not on layer 2 */ 3150 break; 3151 mtag = m_tag_get(PACKET_TAG_DIVERT, 3152 sizeof(struct divert_tag), 3153 M_NOWAIT); 3154 if (mtag == NULL) { 3155 /* XXX statistic */ 3156 /* drop packet */ 3157 IPFW_RUNLOCK(chain); 3158 return (IP_FW_DENY); 3159 } 3160 dt = (struct divert_tag *)(mtag+1); 3161 dt->cookie = f->rulenum; 3162 if (cmd->arg1 == IP_FW_TABLEARG) 3163 dt->info = tablearg; 3164 else 3165 dt->info = cmd->arg1; 3166 m_tag_prepend(m, mtag); 3167 retval = (cmd->opcode == O_DIVERT) ? 3168 IP_FW_DIVERT : IP_FW_TEE; 3169 goto done; 3170 } 3171 3172 case O_COUNT: 3173 case O_SKIPTO: 3174 f->pcnt++; /* update stats */ 3175 f->bcnt += pktlen; 3176 f->timestamp = time_uptime; 3177 if (cmd->opcode == O_COUNT) 3178 goto next_rule; 3179 /* handle skipto */ 3180 if (f->next_rule == NULL) 3181 lookup_next_rule(f); 3182 f = f->next_rule; 3183 goto again; 3184 3185 case O_REJECT: 3186 /* 3187 * Drop the packet and send a reject notice 3188 * if the packet is not ICMP (or is an ICMP 3189 * query), and it is not multicast/broadcast. 3190 */ 3191 if (hlen > 0 && is_ipv4 && offset == 0 && 3192 (proto != IPPROTO_ICMP || 3193 is_icmp_query(ICMP(ulp))) && 3194 !(m->m_flags & (M_BCAST|M_MCAST)) && 3195 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3196 send_reject(args, cmd->arg1, ip_len); 3197 m = args->m; 3198 } 3199 /* FALLTHROUGH */ 3200#ifdef INET6 3201 case O_UNREACH6: 3202 if (hlen > 0 && is_ipv6 && 3203 ((offset & IP6F_OFF_MASK) == 0) && 3204 (proto != IPPROTO_ICMPV6 || 3205 (is_icmp6_query(args->f_id.flags) == 1)) && 3206 !(m->m_flags & (M_BCAST|M_MCAST)) && 3207 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3208 send_reject6(args, cmd->arg1, hlen); 3209 m = args->m; 3210 } 3211 /* FALLTHROUGH */ 3212#endif 3213 case O_DENY: 3214 retval = IP_FW_DENY; 3215 goto done; 3216 3217 case O_FORWARD_IP: { 3218 struct sockaddr_in *sa; 3219 sa = &(((ipfw_insn_sa *)cmd)->sa); 3220 if (args->eh) /* not valid on layer2 pkts */ 3221 break; 3222 if (!q || dyn_dir == MATCH_FORWARD) { 3223 if (sa->sin_addr.s_addr == INADDR_ANY) { 3224 bcopy(sa, &args->hopstore, 3225 sizeof(*sa)); 3226 args->hopstore.sin_addr.s_addr = 3227 htonl(tablearg); 3228 args->next_hop = 3229 &args->hopstore; 3230 } else { 3231 args->next_hop = sa; 3232 } 3233 } 3234 retval = IP_FW_PASS; 3235 } 3236 goto done; 3237 3238 case O_NETGRAPH: 3239 case O_NGTEE: 3240 args->rule = f; /* report matching rule */ 3241 if (cmd->arg1 == IP_FW_TABLEARG) 3242 args->cookie = tablearg; 3243 else 3244 args->cookie = cmd->arg1; 3245 retval = (cmd->opcode == O_NETGRAPH) ? 3246 IP_FW_NETGRAPH : IP_FW_NGTEE; 3247 goto done; 3248 3249 default: 3250 panic("-- unknown opcode %d\n", cmd->opcode); 3251 } /* end of switch() on opcodes */ 3252 3253 if (cmd->len & F_NOT) 3254 match = !match; 3255 3256 if (match) { 3257 if (cmd->len & F_OR) 3258 skip_or = 1; 3259 } else { 3260 if (!(cmd->len & F_OR)) /* not an OR block, */ 3261 break; /* try next rule */ 3262 } 3263 3264 } /* end of inner for, scan opcodes */ 3265 3266next_rule:; /* try next rule */ 3267 3268 } /* end of outer for, scan rules */ 3269 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3270 IPFW_RUNLOCK(chain); 3271 return (IP_FW_DENY); 3272 3273done: 3274 /* Update statistics */ 3275 f->pcnt++; 3276 f->bcnt += pktlen; 3277 f->timestamp = time_uptime; 3278 IPFW_RUNLOCK(chain); 3279 return (retval); 3280 3281pullup_failed: 3282 if (fw_verbose) 3283 printf("ipfw: pullup failed\n"); 3284 return (IP_FW_DENY); 3285} 3286 3287/* 3288 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3289 * These will be reconstructed on the fly as packets are matched. 3290 */ 3291static void 3292flush_rule_ptrs(struct ip_fw_chain *chain) 3293{ 3294 struct ip_fw *rule; 3295 3296 IPFW_WLOCK_ASSERT(chain); 3297 3298 for (rule = chain->rules; rule; rule = rule->next) 3299 rule->next_rule = NULL; 3300} 3301 3302/* 3303 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3304 * possibly create a rule number and add the rule to the list. 3305 * Update the rule_number in the input struct so the caller knows it as well. 3306 */ 3307static int 3308add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3309{ 3310 struct ip_fw *rule, *f, *prev; 3311 int l = RULESIZE(input_rule); 3312 3313 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3314 return (EINVAL); 3315 3316 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3317 if (rule == NULL) 3318 return (ENOSPC); 3319 3320 bcopy(input_rule, rule, l); 3321 3322 rule->next = NULL; 3323 rule->next_rule = NULL; 3324 3325 rule->pcnt = 0; 3326 rule->bcnt = 0; 3327 rule->timestamp = 0; 3328 3329 IPFW_WLOCK(chain); 3330 3331 if (chain->rules == NULL) { /* default rule */ 3332 chain->rules = rule; 3333 goto done; 3334 } 3335 3336 /* 3337 * If rulenum is 0, find highest numbered rule before the 3338 * default rule, and add autoinc_step 3339 */ 3340 if (autoinc_step < 1) 3341 autoinc_step = 1; 3342 else if (autoinc_step > 1000) 3343 autoinc_step = 1000; 3344 if (rule->rulenum == 0) { 3345 /* 3346 * locate the highest numbered rule before default 3347 */ 3348 for (f = chain->rules; f; f = f->next) { 3349 if (f->rulenum == IPFW_DEFAULT_RULE) 3350 break; 3351 rule->rulenum = f->rulenum; 3352 } 3353 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3354 rule->rulenum += autoinc_step; 3355 input_rule->rulenum = rule->rulenum; 3356 } 3357 3358 /* 3359 * Now insert the new rule in the right place in the sorted list. 3360 */ 3361 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3362 if (f->rulenum > rule->rulenum) { /* found the location */ 3363 if (prev) { 3364 rule->next = f; 3365 prev->next = rule; 3366 } else { /* head insert */ 3367 rule->next = chain->rules; 3368 chain->rules = rule; 3369 } 3370 break; 3371 } 3372 } 3373 flush_rule_ptrs(chain); 3374done: 3375 static_count++; 3376 static_len += l; 3377 IPFW_WUNLOCK(chain); 3378 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3379 rule->rulenum, static_count);) 3380 return (0); 3381} 3382 3383/** 3384 * Remove a static rule (including derived * dynamic rules) 3385 * and place it on the ``reap list'' for later reclamation. 3386 * The caller is in charge of clearing rule pointers to avoid 3387 * dangling pointers. 3388 * @return a pointer to the next entry. 3389 * Arguments are not checked, so they better be correct. 3390 */ 3391static struct ip_fw * 3392remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3393{ 3394 struct ip_fw *n; 3395 int l = RULESIZE(rule); 3396 3397 IPFW_WLOCK_ASSERT(chain); 3398 3399 n = rule->next; 3400 IPFW_DYN_LOCK(); 3401 remove_dyn_rule(rule, NULL /* force removal */); 3402 IPFW_DYN_UNLOCK(); 3403 if (prev == NULL) 3404 chain->rules = n; 3405 else 3406 prev->next = n; 3407 static_count--; 3408 static_len -= l; 3409 3410 rule->next = chain->reap; 3411 chain->reap = rule; 3412 3413 return n; 3414} 3415 3416/** 3417 * Reclaim storage associated with a list of rules. This is 3418 * typically the list created using remove_rule. 3419 */ 3420static void 3421reap_rules(struct ip_fw *head) 3422{ 3423 struct ip_fw *rule; 3424 3425 while ((rule = head) != NULL) { 3426 head = head->next; 3427 if (DUMMYNET_LOADED) 3428 ip_dn_ruledel_ptr(rule); 3429 free(rule, M_IPFW); 3430 } 3431} 3432 3433/* 3434 * Remove all rules from a chain (except rules in set RESVD_SET 3435 * unless kill_default = 1). The caller is responsible for 3436 * reclaiming storage for the rules left in chain->reap. 3437 */ 3438static void 3439free_chain(struct ip_fw_chain *chain, int kill_default) 3440{ 3441 struct ip_fw *prev, *rule; 3442 3443 IPFW_WLOCK_ASSERT(chain); 3444 3445 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3446 for (prev = NULL, rule = chain->rules; rule ; ) 3447 if (kill_default || rule->set != RESVD_SET) 3448 rule = remove_rule(chain, rule, prev); 3449 else { 3450 prev = rule; 3451 rule = rule->next; 3452 } 3453} 3454 3455/** 3456 * Remove all rules with given number, and also do set manipulation. 3457 * Assumes chain != NULL && *chain != NULL. 3458 * 3459 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3460 * the next 8 bits are the new set, the top 8 bits are the command: 3461 * 3462 * 0 delete rules with given number 3463 * 1 delete rules with given set number 3464 * 2 move rules with given number to new set 3465 * 3 move rules with given set number to new set 3466 * 4 swap sets with given numbers 3467 */ 3468static int 3469del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3470{ 3471 struct ip_fw *prev = NULL, *rule; 3472 u_int16_t rulenum; /* rule or old_set */ 3473 u_int8_t cmd, new_set; 3474 3475 rulenum = arg & 0xffff; 3476 cmd = (arg >> 24) & 0xff; 3477 new_set = (arg >> 16) & 0xff; 3478 3479 if (cmd > 4) 3480 return EINVAL; 3481 if (new_set > RESVD_SET) 3482 return EINVAL; 3483 if (cmd == 0 || cmd == 2) { 3484 if (rulenum >= IPFW_DEFAULT_RULE) 3485 return EINVAL; 3486 } else { 3487 if (rulenum > RESVD_SET) /* old_set */ 3488 return EINVAL; 3489 } 3490 3491 IPFW_WLOCK(chain); 3492 rule = chain->rules; 3493 chain->reap = NULL; 3494 switch (cmd) { 3495 case 0: /* delete rules with given number */ 3496 /* 3497 * locate first rule to delete 3498 */ 3499 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3500 ; 3501 if (rule->rulenum != rulenum) { 3502 IPFW_WUNLOCK(chain); 3503 return EINVAL; 3504 } 3505 3506 /* 3507 * flush pointers outside the loop, then delete all matching 3508 * rules. prev remains the same throughout the cycle. 3509 */ 3510 flush_rule_ptrs(chain); 3511 while (rule->rulenum == rulenum) 3512 rule = remove_rule(chain, rule, prev); 3513 break; 3514 3515 case 1: /* delete all rules with given set number */ 3516 flush_rule_ptrs(chain); 3517 rule = chain->rules; 3518 while (rule->rulenum < IPFW_DEFAULT_RULE) 3519 if (rule->set == rulenum) 3520 rule = remove_rule(chain, rule, prev); 3521 else { 3522 prev = rule; 3523 rule = rule->next; 3524 } 3525 break; 3526 3527 case 2: /* move rules with given number to new set */ 3528 rule = chain->rules; 3529 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3530 if (rule->rulenum == rulenum) 3531 rule->set = new_set; 3532 break; 3533 3534 case 3: /* move rules with given set number to new set */ 3535 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3536 if (rule->set == rulenum) 3537 rule->set = new_set; 3538 break; 3539 3540 case 4: /* swap two sets */ 3541 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3542 if (rule->set == rulenum) 3543 rule->set = new_set; 3544 else if (rule->set == new_set) 3545 rule->set = rulenum; 3546 break; 3547 } 3548 /* 3549 * Look for rules to reclaim. We grab the list before 3550 * releasing the lock then reclaim them w/o the lock to 3551 * avoid a LOR with dummynet. 3552 */ 3553 rule = chain->reap; 3554 chain->reap = NULL; 3555 IPFW_WUNLOCK(chain); 3556 if (rule) 3557 reap_rules(rule); 3558 return 0; 3559} 3560 3561/* 3562 * Clear counters for a specific rule. 3563 * The enclosing "table" is assumed locked. 3564 */ 3565static void 3566clear_counters(struct ip_fw *rule, int log_only) 3567{ 3568 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3569 3570 if (log_only == 0) { 3571 rule->bcnt = rule->pcnt = 0; 3572 rule->timestamp = 0; 3573 } 3574 if (l->o.opcode == O_LOG) 3575 l->log_left = l->max_log; 3576} 3577 3578/** 3579 * Reset some or all counters on firewall rules. 3580 * @arg frwl is null to clear all entries, or contains a specific 3581 * rule number. 3582 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3583 */ 3584static int 3585zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3586{ 3587 struct ip_fw *rule; 3588 char *msg; 3589 3590 IPFW_WLOCK(chain); 3591 if (rulenum == 0) { 3592 norule_counter = 0; 3593 for (rule = chain->rules; rule; rule = rule->next) 3594 clear_counters(rule, log_only); 3595 msg = log_only ? "ipfw: All logging counts reset.\n" : 3596 "ipfw: Accounting cleared.\n"; 3597 } else { 3598 int cleared = 0; 3599 /* 3600 * We can have multiple rules with the same number, so we 3601 * need to clear them all. 3602 */ 3603 for (rule = chain->rules; rule; rule = rule->next) 3604 if (rule->rulenum == rulenum) { 3605 while (rule && rule->rulenum == rulenum) { 3606 clear_counters(rule, log_only); 3607 rule = rule->next; 3608 } 3609 cleared = 1; 3610 break; 3611 } 3612 if (!cleared) { /* we did not find any matching rules */ 3613 IPFW_WUNLOCK(chain); 3614 return (EINVAL); 3615 } 3616 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3617 "ipfw: Entry %d cleared.\n"; 3618 } 3619 IPFW_WUNLOCK(chain); 3620 3621 if (fw_verbose) 3622 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3623 return (0); 3624} 3625 3626/* 3627 * Check validity of the structure before insert. 3628 * Fortunately rules are simple, so this mostly need to check rule sizes. 3629 */ 3630static int 3631check_ipfw_struct(struct ip_fw *rule, int size) 3632{ 3633 int l, cmdlen = 0; 3634 int have_action=0; 3635 ipfw_insn *cmd; 3636 3637 if (size < sizeof(*rule)) { 3638 printf("ipfw: rule too short\n"); 3639 return (EINVAL); 3640 } 3641 /* first, check for valid size */ 3642 l = RULESIZE(rule); 3643 if (l != size) { 3644 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3645 return (EINVAL); 3646 } 3647 if (rule->act_ofs >= rule->cmd_len) { 3648 printf("ipfw: bogus action offset (%u > %u)\n", 3649 rule->act_ofs, rule->cmd_len - 1); 3650 return (EINVAL); 3651 } 3652 /* 3653 * Now go for the individual checks. Very simple ones, basically only 3654 * instruction sizes. 3655 */ 3656 for (l = rule->cmd_len, cmd = rule->cmd ; 3657 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3658 cmdlen = F_LEN(cmd); 3659 if (cmdlen > l) { 3660 printf("ipfw: opcode %d size truncated\n", 3661 cmd->opcode); 3662 return EINVAL; 3663 } 3664 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3665 switch (cmd->opcode) { 3666 case O_PROBE_STATE: 3667 case O_KEEP_STATE: 3668 case O_PROTO: 3669 case O_IP_SRC_ME: 3670 case O_IP_DST_ME: 3671 case O_LAYER2: 3672 case O_IN: 3673 case O_FRAG: 3674 case O_DIVERTED: 3675 case O_IPOPT: 3676 case O_IPTOS: 3677 case O_IPPRECEDENCE: 3678 case O_IPVER: 3679 case O_TCPWIN: 3680 case O_TCPFLAGS: 3681 case O_TCPOPTS: 3682 case O_ESTAB: 3683 case O_VERREVPATH: 3684 case O_VERSRCREACH: 3685 case O_ANTISPOOF: 3686 case O_IPSEC: 3687#ifdef INET6 3688 case O_IP6_SRC_ME: 3689 case O_IP6_DST_ME: 3690 case O_EXT_HDR: 3691 case O_IP6: 3692#endif 3693 case O_IP4: 3694 case O_TAG: 3695 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3696 goto bad_size; 3697 break; 3698 3699 case O_UID: 3700 case O_GID: 3701 case O_JAIL: 3702 case O_IP_SRC: 3703 case O_IP_DST: 3704 case O_TCPSEQ: 3705 case O_TCPACK: 3706 case O_PROB: 3707 case O_ICMPTYPE: 3708 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3709 goto bad_size; 3710 break; 3711 3712 case O_LIMIT: 3713 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3714 goto bad_size; 3715 break; 3716 3717 case O_LOG: 3718 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3719 goto bad_size; 3720 3721 ((ipfw_insn_log *)cmd)->log_left = 3722 ((ipfw_insn_log *)cmd)->max_log; 3723 3724 break; 3725 3726 case O_IP_SRC_MASK: 3727 case O_IP_DST_MASK: 3728 /* only odd command lengths */ 3729 if ( !(cmdlen & 1) || cmdlen > 31) 3730 goto bad_size; 3731 break; 3732 3733 case O_IP_SRC_SET: 3734 case O_IP_DST_SET: 3735 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3736 printf("ipfw: invalid set size %d\n", 3737 cmd->arg1); 3738 return EINVAL; 3739 } 3740 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3741 (cmd->arg1+31)/32 ) 3742 goto bad_size; 3743 break; 3744 3745 case O_IP_SRC_LOOKUP: 3746 case O_IP_DST_LOOKUP: 3747 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3748 printf("ipfw: invalid table number %d\n", 3749 cmd->arg1); 3750 return (EINVAL); 3751 } 3752 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3753 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3754 goto bad_size; 3755 break; 3756 3757 case O_MACADDR2: 3758 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3759 goto bad_size; 3760 break; 3761 3762 case O_NOP: 3763 case O_IPID: 3764 case O_IPTTL: 3765 case O_IPLEN: 3766 case O_TCPDATALEN: 3767 case O_TAGGED: 3768 if (cmdlen < 1 || cmdlen > 31) 3769 goto bad_size; 3770 break; 3771 3772 case O_MAC_TYPE: 3773 case O_IP_SRCPORT: 3774 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3775 if (cmdlen < 2 || cmdlen > 31) 3776 goto bad_size; 3777 break; 3778 3779 case O_RECV: 3780 case O_XMIT: 3781 case O_VIA: 3782 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3783 goto bad_size; 3784 break; 3785 3786 case O_ALTQ: 3787 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3788 goto bad_size; 3789 break; 3790 3791 case O_PIPE: 3792 case O_QUEUE: 3793 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3794 goto bad_size; 3795 goto check_action; 3796 3797 case O_FORWARD_IP: 3798#ifdef IPFIREWALL_FORWARD 3799 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3800 goto bad_size; 3801 goto check_action; 3802#else 3803 return EINVAL; 3804#endif 3805 3806 case O_DIVERT: 3807 case O_TEE: 3808 if (ip_divert_ptr == NULL) 3809 return EINVAL; 3810 else 3811 goto check_size; 3812 case O_NETGRAPH: 3813 case O_NGTEE: 3814 if (!NG_IPFW_LOADED) 3815 return EINVAL; 3816 else 3817 goto check_size; 3818 case O_FORWARD_MAC: /* XXX not implemented yet */ 3819 case O_CHECK_STATE: 3820 case O_COUNT: 3821 case O_ACCEPT: 3822 case O_DENY: 3823 case O_REJECT: 3824#ifdef INET6 3825 case O_UNREACH6: 3826#endif 3827 case O_SKIPTO: 3828check_size: 3829 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3830 goto bad_size; 3831check_action: 3832 if (have_action) { 3833 printf("ipfw: opcode %d, multiple actions" 3834 " not allowed\n", 3835 cmd->opcode); 3836 return EINVAL; 3837 } 3838 have_action = 1; 3839 if (l != cmdlen) { 3840 printf("ipfw: opcode %d, action must be" 3841 " last opcode\n", 3842 cmd->opcode); 3843 return EINVAL; 3844 } 3845 break; 3846#ifdef INET6 3847 case O_IP6_SRC: 3848 case O_IP6_DST: 3849 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3850 F_INSN_SIZE(ipfw_insn)) 3851 goto bad_size; 3852 break; 3853 3854 case O_FLOW6ID: 3855 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3856 ((ipfw_insn_u32 *)cmd)->o.arg1) 3857 goto bad_size; 3858 break; 3859 3860 case O_IP6_SRC_MASK: 3861 case O_IP6_DST_MASK: 3862 if ( !(cmdlen & 1) || cmdlen > 127) 3863 goto bad_size; 3864 break; 3865 case O_ICMP6TYPE: 3866 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3867 goto bad_size; 3868 break; 3869#endif 3870 3871 default: 3872 switch (cmd->opcode) { 3873#ifndef INET6 3874 case O_IP6_SRC_ME: 3875 case O_IP6_DST_ME: 3876 case O_EXT_HDR: 3877 case O_IP6: 3878 case O_UNREACH6: 3879 case O_IP6_SRC: 3880 case O_IP6_DST: 3881 case O_FLOW6ID: 3882 case O_IP6_SRC_MASK: 3883 case O_IP6_DST_MASK: 3884 case O_ICMP6TYPE: 3885 printf("ipfw: no IPv6 support in kernel\n"); 3886 return EPROTONOSUPPORT; 3887#endif 3888 default: 3889 printf("ipfw: opcode %d, unknown opcode\n", 3890 cmd->opcode); 3891 return EINVAL; 3892 } 3893 } 3894 } 3895 if (have_action == 0) { 3896 printf("ipfw: missing action\n"); 3897 return EINVAL; 3898 } 3899 return 0; 3900 3901bad_size: 3902 printf("ipfw: opcode %d size %d wrong\n", 3903 cmd->opcode, cmdlen); 3904 return EINVAL; 3905} 3906 3907/* 3908 * Copy the static and dynamic rules to the supplied buffer 3909 * and return the amount of space actually used. 3910 */ 3911static size_t 3912ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3913{ 3914 char *bp = buf; 3915 char *ep = bp + space; 3916 struct ip_fw *rule; 3917 int i; 3918 3919 /* XXX this can take a long time and locking will block packet flow */ 3920 IPFW_RLOCK(chain); 3921 for (rule = chain->rules; rule ; rule = rule->next) { 3922 /* 3923 * Verify the entry fits in the buffer in case the 3924 * rules changed between calculating buffer space and 3925 * now. This would be better done using a generation 3926 * number but should suffice for now. 3927 */ 3928 i = RULESIZE(rule); 3929 if (bp + i <= ep) { 3930 bcopy(rule, bp, i); 3931 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3932 sizeof(set_disable)); 3933 bp += i; 3934 } 3935 } 3936 IPFW_RUNLOCK(chain); 3937 if (ipfw_dyn_v) { 3938 ipfw_dyn_rule *p, *last = NULL; 3939 3940 IPFW_DYN_LOCK(); 3941 for (i = 0 ; i < curr_dyn_buckets; i++) 3942 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3943 if (bp + sizeof *p <= ep) { 3944 ipfw_dyn_rule *dst = 3945 (ipfw_dyn_rule *)bp; 3946 bcopy(p, dst, sizeof *p); 3947 bcopy(&(p->rule->rulenum), &(dst->rule), 3948 sizeof(p->rule->rulenum)); 3949 /* 3950 * store a non-null value in "next". 3951 * The userland code will interpret a 3952 * NULL here as a marker 3953 * for the last dynamic rule. 3954 */ 3955 bcopy(&dst, &dst->next, sizeof(dst)); 3956 last = dst; 3957 dst->expire = 3958 TIME_LEQ(dst->expire, time_uptime) ? 3959 0 : dst->expire - time_uptime ; 3960 bp += sizeof(ipfw_dyn_rule); 3961 } 3962 } 3963 IPFW_DYN_UNLOCK(); 3964 if (last != NULL) /* mark last dynamic rule */ 3965 bzero(&last->next, sizeof(last)); 3966 } 3967 return (bp - (char *)buf); 3968} 3969 3970 3971/** 3972 * {set|get}sockopt parser. 3973 */ 3974static int 3975ipfw_ctl(struct sockopt *sopt) 3976{ 3977#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3978 int error, rule_num; 3979 size_t size; 3980 struct ip_fw *buf, *rule; 3981 u_int32_t rulenum[2]; 3982 3983 error = suser(sopt->sopt_td); 3984 if (error) 3985 return (error); 3986 3987 /* 3988 * Disallow modifications in really-really secure mode, but still allow 3989 * the logging counters to be reset. 3990 */ 3991 if (sopt->sopt_name == IP_FW_ADD || 3992 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3993 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3994 if (error) 3995 return (error); 3996 } 3997 3998 error = 0; 3999 4000 switch (sopt->sopt_name) { 4001 case IP_FW_GET: 4002 /* 4003 * pass up a copy of the current rules. Static rules 4004 * come first (the last of which has number IPFW_DEFAULT_RULE), 4005 * followed by a possibly empty list of dynamic rule. 4006 * The last dynamic rule has NULL in the "next" field. 4007 * 4008 * Note that the calculated size is used to bound the 4009 * amount of data returned to the user. The rule set may 4010 * change between calculating the size and returning the 4011 * data in which case we'll just return what fits. 4012 */ 4013 size = static_len; /* size of static rules */ 4014 if (ipfw_dyn_v) /* add size of dyn.rules */ 4015 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4016 4017 /* 4018 * XXX todo: if the user passes a short length just to know 4019 * how much room is needed, do not bother filling up the 4020 * buffer, just jump to the sooptcopyout. 4021 */ 4022 buf = malloc(size, M_TEMP, M_WAITOK); 4023 error = sooptcopyout(sopt, buf, 4024 ipfw_getrules(&layer3_chain, buf, size)); 4025 free(buf, M_TEMP); 4026 break; 4027 4028 case IP_FW_FLUSH: 4029 /* 4030 * Normally we cannot release the lock on each iteration. 4031 * We could do it here only because we start from the head all 4032 * the times so there is no risk of missing some entries. 4033 * On the other hand, the risk is that we end up with 4034 * a very inconsistent ruleset, so better keep the lock 4035 * around the whole cycle. 4036 * 4037 * XXX this code can be improved by resetting the head of 4038 * the list to point to the default rule, and then freeing 4039 * the old list without the need for a lock. 4040 */ 4041 4042 IPFW_WLOCK(&layer3_chain); 4043 layer3_chain.reap = NULL; 4044 free_chain(&layer3_chain, 0 /* keep default rule */); 4045 rule = layer3_chain.reap; 4046 layer3_chain.reap = NULL; 4047 IPFW_WUNLOCK(&layer3_chain); 4048 if (rule != NULL) 4049 reap_rules(rule); 4050 break; 4051 4052 case IP_FW_ADD: 4053 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4054 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4055 sizeof(struct ip_fw) ); 4056 if (error == 0) 4057 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4058 if (error == 0) { 4059 error = add_rule(&layer3_chain, rule); 4060 size = RULESIZE(rule); 4061 if (!error && sopt->sopt_dir == SOPT_GET) 4062 error = sooptcopyout(sopt, rule, size); 4063 } 4064 free(rule, M_TEMP); 4065 break; 4066 4067 case IP_FW_DEL: 4068 /* 4069 * IP_FW_DEL is used for deleting single rules or sets, 4070 * and (ab)used to atomically manipulate sets. Argument size 4071 * is used to distinguish between the two: 4072 * sizeof(u_int32_t) 4073 * delete single rule or set of rules, 4074 * or reassign rules (or sets) to a different set. 4075 * 2*sizeof(u_int32_t) 4076 * atomic disable/enable sets. 4077 * first u_int32_t contains sets to be disabled, 4078 * second u_int32_t contains sets to be enabled. 4079 */ 4080 error = sooptcopyin(sopt, rulenum, 4081 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4082 if (error) 4083 break; 4084 size = sopt->sopt_valsize; 4085 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4086 error = del_entry(&layer3_chain, rulenum[0]); 4087 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4088 set_disable = 4089 (set_disable | rulenum[0]) & ~rulenum[1] & 4090 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4091 else 4092 error = EINVAL; 4093 break; 4094 4095 case IP_FW_ZERO: 4096 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 4097 rule_num = 0; 4098 if (sopt->sopt_val != 0) { 4099 error = sooptcopyin(sopt, &rule_num, 4100 sizeof(int), sizeof(int)); 4101 if (error) 4102 break; 4103 } 4104 error = zero_entry(&layer3_chain, rule_num, 4105 sopt->sopt_name == IP_FW_RESETLOG); 4106 break; 4107 4108 case IP_FW_TABLE_ADD: 4109 { 4110 ipfw_table_entry ent; 4111 4112 error = sooptcopyin(sopt, &ent, 4113 sizeof(ent), sizeof(ent)); 4114 if (error) 4115 break; 4116 error = add_table_entry(&layer3_chain, ent.tbl, 4117 ent.addr, ent.masklen, ent.value); 4118 } 4119 break; 4120 4121 case IP_FW_TABLE_DEL: 4122 { 4123 ipfw_table_entry ent; 4124 4125 error = sooptcopyin(sopt, &ent, 4126 sizeof(ent), sizeof(ent)); 4127 if (error) 4128 break; 4129 error = del_table_entry(&layer3_chain, ent.tbl, 4130 ent.addr, ent.masklen); 4131 } 4132 break; 4133 4134 case IP_FW_TABLE_FLUSH: 4135 { 4136 u_int16_t tbl; 4137 4138 error = sooptcopyin(sopt, &tbl, 4139 sizeof(tbl), sizeof(tbl)); 4140 if (error) 4141 break; 4142 IPFW_WLOCK(&layer3_chain); 4143 error = flush_table(&layer3_chain, tbl); 4144 IPFW_WUNLOCK(&layer3_chain); 4145 } 4146 break; 4147 4148 case IP_FW_TABLE_GETSIZE: 4149 { 4150 u_int32_t tbl, cnt; 4151 4152 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4153 sizeof(tbl)))) 4154 break; 4155 IPFW_RLOCK(&layer3_chain); 4156 error = count_table(&layer3_chain, tbl, &cnt); 4157 IPFW_RUNLOCK(&layer3_chain); 4158 if (error) 4159 break; 4160 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4161 } 4162 break; 4163 4164 case IP_FW_TABLE_LIST: 4165 { 4166 ipfw_table *tbl; 4167 4168 if (sopt->sopt_valsize < sizeof(*tbl)) { 4169 error = EINVAL; 4170 break; 4171 } 4172 size = sopt->sopt_valsize; 4173 tbl = malloc(size, M_TEMP, M_WAITOK); 4174 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4175 if (error) { 4176 free(tbl, M_TEMP); 4177 break; 4178 } 4179 tbl->size = (size - sizeof(*tbl)) / 4180 sizeof(ipfw_table_entry); 4181 IPFW_RLOCK(&layer3_chain); 4182 error = dump_table(&layer3_chain, tbl); 4183 IPFW_RUNLOCK(&layer3_chain); 4184 if (error) { 4185 free(tbl, M_TEMP); 4186 break; 4187 } 4188 error = sooptcopyout(sopt, tbl, size); 4189 free(tbl, M_TEMP); 4190 } 4191 break; 4192 4193 default: 4194 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4195 error = EINVAL; 4196 } 4197 4198 return (error); 4199#undef RULE_MAXSIZE 4200} 4201 4202/** 4203 * dummynet needs a reference to the default rule, because rules can be 4204 * deleted while packets hold a reference to them. When this happens, 4205 * dummynet changes the reference to the default rule (it could well be a 4206 * NULL pointer, but this way we do not need to check for the special 4207 * case, plus here he have info on the default behaviour). 4208 */ 4209struct ip_fw *ip_fw_default_rule; 4210 4211/* 4212 * This procedure is only used to handle keepalives. It is invoked 4213 * every dyn_keepalive_period 4214 */ 4215static void 4216ipfw_tick(void * __unused unused) 4217{ 4218 struct mbuf *m0, *m, *mnext, **mtailp; 4219 int i; 4220 ipfw_dyn_rule *q; 4221 4222 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4223 goto done; 4224 4225 /* 4226 * We make a chain of packets to go out here -- not deferring 4227 * until after we drop the IPFW dynamic rule lock would result 4228 * in a lock order reversal with the normal packet input -> ipfw 4229 * call stack. 4230 */ 4231 m0 = NULL; 4232 mtailp = &m0; 4233 IPFW_DYN_LOCK(); 4234 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4235 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4236 if (q->dyn_type == O_LIMIT_PARENT) 4237 continue; 4238 if (q->id.proto != IPPROTO_TCP) 4239 continue; 4240 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4241 continue; 4242 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4243 q->expire)) 4244 continue; /* too early */ 4245 if (TIME_LEQ(q->expire, time_uptime)) 4246 continue; /* too late, rule expired */ 4247 4248 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4249 q->ack_fwd, TH_SYN); 4250 if (*mtailp != NULL) 4251 mtailp = &(*mtailp)->m_nextpkt; 4252 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4253 q->ack_rev, 0); 4254 if (*mtailp != NULL) 4255 mtailp = &(*mtailp)->m_nextpkt; 4256 } 4257 } 4258 IPFW_DYN_UNLOCK(); 4259 for (m = mnext = m0; m != NULL; m = mnext) { 4260 mnext = m->m_nextpkt; 4261 m->m_nextpkt = NULL; 4262 ip_output(m, NULL, NULL, 0, NULL, NULL); 4263 } 4264done: 4265 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4266} 4267 4268int 4269ipfw_init(void) 4270{ 4271 struct ip_fw default_rule; 4272 int error; 4273 4274#ifdef INET6 4275 /* Setup IPv6 fw sysctl tree. */ 4276 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4277 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4278 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4279 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4280 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4281 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4282 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4283 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4284 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4285 &fw_deny_unknown_exthdrs, 0, 4286 "Deny packets with unknown IPv6 Extension Headers"); 4287#endif 4288 4289 layer3_chain.rules = NULL; 4290 IPFW_LOCK_INIT(&layer3_chain); 4291 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4292 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4293 UMA_ALIGN_PTR, 0); 4294 IPFW_DYN_LOCK_INIT(); 4295 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4296 4297 bzero(&default_rule, sizeof default_rule); 4298 4299 default_rule.act_ofs = 0; 4300 default_rule.rulenum = IPFW_DEFAULT_RULE; 4301 default_rule.cmd_len = 1; 4302 default_rule.set = RESVD_SET; 4303 4304 default_rule.cmd[0].len = 1; 4305 default_rule.cmd[0].opcode = 4306#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4307 1 ? O_ACCEPT : 4308#endif 4309 O_DENY; 4310 4311 error = add_rule(&layer3_chain, &default_rule); 4312 if (error != 0) { 4313 printf("ipfw2: error %u initializing default rule " 4314 "(support disabled)\n", error); 4315 IPFW_DYN_LOCK_DESTROY(); 4316 IPFW_LOCK_DESTROY(&layer3_chain); 4317 uma_zdestroy(ipfw_dyn_rule_zone); 4318 return (error); 4319 } 4320 4321 ip_fw_default_rule = layer3_chain.rules; 4322 printf("ipfw2 " 4323#ifdef INET6 4324 "(+ipv6) " 4325#endif 4326 "initialized, divert %s, " 4327 "rule-based forwarding " 4328#ifdef IPFIREWALL_FORWARD 4329 "enabled, " 4330#else 4331 "disabled, " 4332#endif 4333 "default to %s, logging ", 4334#ifdef IPDIVERT 4335 "enabled", 4336#else 4337 "loadable", 4338#endif 4339 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4340 4341#ifdef IPFIREWALL_VERBOSE 4342 fw_verbose = 1; 4343#endif 4344#ifdef IPFIREWALL_VERBOSE_LIMIT 4345 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4346#endif 4347 if (fw_verbose == 0) 4348 printf("disabled\n"); 4349 else if (verbose_limit == 0) 4350 printf("unlimited\n"); 4351 else 4352 printf("limited to %d packets/entry by default\n", 4353 verbose_limit); 4354 4355 error = init_tables(&layer3_chain); 4356 if (error) { 4357 IPFW_DYN_LOCK_DESTROY(); 4358 IPFW_LOCK_DESTROY(&layer3_chain); 4359 uma_zdestroy(ipfw_dyn_rule_zone); 4360 return (error); 4361 } 4362 ip_fw_ctl_ptr = ipfw_ctl; 4363 ip_fw_chk_ptr = ipfw_chk; 4364 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4365 4366 return (0); 4367} 4368 4369void 4370ipfw_destroy(void) 4371{ 4372 struct ip_fw *reap; 4373 4374 ip_fw_chk_ptr = NULL; 4375 ip_fw_ctl_ptr = NULL; 4376 callout_drain(&ipfw_timeout); 4377 IPFW_WLOCK(&layer3_chain); 4378 flush_tables(&layer3_chain); 4379 layer3_chain.reap = NULL; 4380 free_chain(&layer3_chain, 1 /* kill default rule */); 4381 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4382 IPFW_WUNLOCK(&layer3_chain); 4383 if (reap != NULL) 4384 reap_rules(reap); 4385 IPFW_DYN_LOCK_DESTROY(); 4386 uma_zdestroy(ipfw_dyn_rule_zone); 4387 IPFW_LOCK_DESTROY(&layer3_chain); 4388 4389#ifdef INET6 4390 /* Free IPv6 fw sysctl tree. */ 4391 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4392#endif 4393 4394 printf("IP firewall unloaded\n"); 4395} 4396