ip_fw2.c revision 165738
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 165738 2007-01-02 19:57:31Z julian $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdn.h" 38#include "opt_inet.h" 39#ifndef INET 40#error IPFIREWALL requires INET. 41#endif /* INET */ 42#endif 43#include "opt_inet6.h" 44#include "opt_ipsec.h" 45#include "opt_mac.h" 46 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/condvar.h> 50#include <sys/eventhandler.h> 51#include <sys/malloc.h> 52#include <sys/mbuf.h> 53#include <sys/kernel.h> 54#include <sys/lock.h> 55#include <sys/jail.h> 56#include <sys/module.h> 57#include <sys/priv.h> 58#include <sys/proc.h> 59#include <sys/rwlock.h> 60#include <sys/socket.h> 61#include <sys/socketvar.h> 62#include <sys/sysctl.h> 63#include <sys/syslog.h> 64#include <sys/ucred.h> 65#include <net/if.h> 66#include <net/radix.h> 67#include <net/route.h> 68#include <netinet/in.h> 69#include <netinet/in_systm.h> 70#include <netinet/in_var.h> 71#include <netinet/in_pcb.h> 72#include <netinet/ip.h> 73#include <netinet/ip_var.h> 74#include <netinet/ip_icmp.h> 75#include <netinet/ip_fw.h> 76#include <netinet/ip_divert.h> 77#include <netinet/ip_dummynet.h> 78#include <netinet/ip_carp.h> 79#include <netinet/pim.h> 80#include <netinet/tcp.h> 81#include <netinet/tcp_timer.h> 82#include <netinet/tcp_var.h> 83#include <netinet/tcpip.h> 84#include <netinet/udp.h> 85#include <netinet/udp_var.h> 86#include <netinet/sctp.h> 87 88#include <netinet/libalias/alias.h> 89#include <netinet/libalias/alias_local.h> 90#include <netgraph/ng_ipfw.h> 91 92#include <altq/if_altq.h> 93 94#ifdef IPSEC 95#include <netinet6/ipsec.h> 96#endif 97 98#include <netinet/ip6.h> 99#include <netinet/icmp6.h> 100#ifdef INET6 101#include <netinet6/scope6_var.h> 102#endif 103 104#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 105 106#include <machine/in_cksum.h> /* XXX for in_cksum */ 107 108#include <security/mac/mac_framework.h> 109 110/* 111 * set_disable contains one bit per set value (0..31). 112 * If the bit is set, all rules with the corresponding set 113 * are disabled. Set RESVD_SET(31) is reserved for the default rule 114 * and rules that are not deleted by the flush command, 115 * and CANNOT be disabled. 116 * Rules in set RESVD_SET can only be deleted explicitly. 117 */ 118static u_int32_t set_disable; 119 120static int fw_verbose; 121static int verbose_limit; 122 123static struct callout ipfw_timeout; 124static uma_zone_t ipfw_dyn_rule_zone; 125#define IPFW_DEFAULT_RULE 65535 126 127/* 128 * Data structure to cache our ucred related 129 * information. This structure only gets used if 130 * the user specified UID/GID based constraints in 131 * a firewall rule. 132 */ 133struct ip_fw_ugid { 134 gid_t fw_groups[NGROUPS]; 135 int fw_ngroups; 136 uid_t fw_uid; 137 int fw_prid; 138}; 139 140#define IPFW_TABLES_MAX 128 141struct ip_fw_chain { 142 struct ip_fw *rules; /* list of rules */ 143 struct ip_fw *reap; /* list of rules to reap */ 144 LIST_HEAD(, cfg_nat) nat; /* list of nat entries */ 145 struct radix_node_head *tables[IPFW_TABLES_MAX]; 146 struct rwlock rwmtx; 147}; 148#define IPFW_LOCK_INIT(_chain) \ 149 rw_init(&(_chain)->rwmtx, "IPFW static rules") 150#define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx) 151#define IPFW_WLOCK_ASSERT(_chain) do { \ 152 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \ 153 NET_ASSERT_GIANT(); \ 154} while (0) 155 156#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 157#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 158#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 159#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 160 161/* 162 * list of rules for layer 3 163 */ 164static struct ip_fw_chain layer3_chain; 165 166MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 167MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 168 169struct table_entry { 170 struct radix_node rn[2]; 171 struct sockaddr_in addr, mask; 172 u_int32_t value; 173}; 174 175static int fw_debug = 1; 176static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 177 178extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 179 180#ifdef SYSCTL_NODE 181SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 182SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 183 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 184 ipfw_chg_hook, "I", "Enable ipfw"); 185SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 186 &autoinc_step, 0, "Rule number autincrement step"); 187SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 188 CTLFLAG_RW | CTLFLAG_SECURE3, 189 &fw_one_pass, 0, 190 "Only do a single pass through ipfw when using dummynet(4)"); 191SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 192 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 193SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 194 CTLFLAG_RW | CTLFLAG_SECURE3, 195 &fw_verbose, 0, "Log matches to ipfw rules"); 196SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 197 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 198 199/* 200 * Description of dynamic rules. 201 * 202 * Dynamic rules are stored in lists accessed through a hash table 203 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 204 * be modified through the sysctl variable dyn_buckets which is 205 * updated when the table becomes empty. 206 * 207 * XXX currently there is only one list, ipfw_dyn. 208 * 209 * When a packet is received, its address fields are first masked 210 * with the mask defined for the rule, then hashed, then matched 211 * against the entries in the corresponding list. 212 * Dynamic rules can be used for different purposes: 213 * + stateful rules; 214 * + enforcing limits on the number of sessions; 215 * + in-kernel NAT (not implemented yet) 216 * 217 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 218 * measured in seconds and depending on the flags. 219 * 220 * The total number of dynamic rules is stored in dyn_count. 221 * The max number of dynamic rules is dyn_max. When we reach 222 * the maximum number of rules we do not create anymore. This is 223 * done to avoid consuming too much memory, but also too much 224 * time when searching on each packet (ideally, we should try instead 225 * to put a limit on the length of the list on each bucket...). 226 * 227 * Each dynamic rule holds a pointer to the parent ipfw rule so 228 * we know what action to perform. Dynamic rules are removed when 229 * the parent rule is deleted. XXX we should make them survive. 230 * 231 * There are some limitations with dynamic rules -- we do not 232 * obey the 'randomized match', and we do not do multiple 233 * passes through the firewall. XXX check the latter!!! 234 */ 235static ipfw_dyn_rule **ipfw_dyn_v = NULL; 236static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 237static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 238 239static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 240#define IPFW_DYN_LOCK_INIT() \ 241 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 242#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 243#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 244#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 245#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 246 247/* 248 * Timeouts for various events in handing dynamic rules. 249 */ 250static u_int32_t dyn_ack_lifetime = 300; 251static u_int32_t dyn_syn_lifetime = 20; 252static u_int32_t dyn_fin_lifetime = 1; 253static u_int32_t dyn_rst_lifetime = 1; 254static u_int32_t dyn_udp_lifetime = 10; 255static u_int32_t dyn_short_lifetime = 5; 256 257/* 258 * Keepalives are sent if dyn_keepalive is set. They are sent every 259 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 260 * seconds of lifetime of a rule. 261 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 262 * than dyn_keepalive_period. 263 */ 264 265static u_int32_t dyn_keepalive_interval = 20; 266static u_int32_t dyn_keepalive_period = 5; 267static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 268 269static u_int32_t static_count; /* # of static rules */ 270static u_int32_t static_len; /* size in bytes of static rules */ 271static u_int32_t dyn_count; /* # of dynamic rules */ 272static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 273 274SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 275 &dyn_buckets, 0, "Number of dyn. buckets"); 276SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 277 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 278SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 279 &dyn_count, 0, "Number of dyn. rules"); 280SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 281 &dyn_max, 0, "Max number of dyn. rules"); 282SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 283 &static_count, 0, "Number of static rules"); 284SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 285 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 286SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 287 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 288SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 289 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 290SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 291 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 292SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 293 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 294SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 295 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 296SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 297 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 298 299#ifdef INET6 300/* 301 * IPv6 specific variables 302 */ 303SYSCTL_DECL(_net_inet6_ip6); 304 305static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 306static struct sysctl_oid *ip6_fw_sysctl_tree; 307#endif /* INET6 */ 308#endif /* SYSCTL_NODE */ 309 310MODULE_DEPEND(ipfw, libalias, 1, 1, 1); 311static int fw_deny_unknown_exthdrs = 1; 312 313 314/* 315 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 316 * Other macros just cast void * into the appropriate type 317 */ 318#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 319#define TCP(p) ((struct tcphdr *)(p)) 320#define SCTP(p) ((struct sctphdr *)(p)) 321#define UDP(p) ((struct udphdr *)(p)) 322#define ICMP(p) ((struct icmphdr *)(p)) 323#define ICMP6(p) ((struct icmp6_hdr *)(p)) 324 325static __inline int 326icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 327{ 328 int type = icmp->icmp_type; 329 330 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 331} 332 333#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 334 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 335 336static int 337is_icmp_query(struct icmphdr *icmp) 338{ 339 int type = icmp->icmp_type; 340 341 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 342} 343#undef TT 344 345/* 346 * The following checks use two arrays of 8 or 16 bits to store the 347 * bits that we want set or clear, respectively. They are in the 348 * low and high half of cmd->arg1 or cmd->d[0]. 349 * 350 * We scan options and store the bits we find set. We succeed if 351 * 352 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 353 * 354 * The code is sometimes optimized not to store additional variables. 355 */ 356 357static int 358flags_match(ipfw_insn *cmd, u_int8_t bits) 359{ 360 u_char want_clear; 361 bits = ~bits; 362 363 if ( ((cmd->arg1 & 0xff) & bits) != 0) 364 return 0; /* some bits we want set were clear */ 365 want_clear = (cmd->arg1 >> 8) & 0xff; 366 if ( (want_clear & bits) != want_clear) 367 return 0; /* some bits we want clear were set */ 368 return 1; 369} 370 371static int 372ipopts_match(struct ip *ip, ipfw_insn *cmd) 373{ 374 int optlen, bits = 0; 375 u_char *cp = (u_char *)(ip + 1); 376 int x = (ip->ip_hl << 2) - sizeof (struct ip); 377 378 for (; x > 0; x -= optlen, cp += optlen) { 379 int opt = cp[IPOPT_OPTVAL]; 380 381 if (opt == IPOPT_EOL) 382 break; 383 if (opt == IPOPT_NOP) 384 optlen = 1; 385 else { 386 optlen = cp[IPOPT_OLEN]; 387 if (optlen <= 0 || optlen > x) 388 return 0; /* invalid or truncated */ 389 } 390 switch (opt) { 391 392 default: 393 break; 394 395 case IPOPT_LSRR: 396 bits |= IP_FW_IPOPT_LSRR; 397 break; 398 399 case IPOPT_SSRR: 400 bits |= IP_FW_IPOPT_SSRR; 401 break; 402 403 case IPOPT_RR: 404 bits |= IP_FW_IPOPT_RR; 405 break; 406 407 case IPOPT_TS: 408 bits |= IP_FW_IPOPT_TS; 409 break; 410 } 411 } 412 return (flags_match(cmd, bits)); 413} 414 415static int 416tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 417{ 418 int optlen, bits = 0; 419 u_char *cp = (u_char *)(tcp + 1); 420 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 421 422 for (; x > 0; x -= optlen, cp += optlen) { 423 int opt = cp[0]; 424 if (opt == TCPOPT_EOL) 425 break; 426 if (opt == TCPOPT_NOP) 427 optlen = 1; 428 else { 429 optlen = cp[1]; 430 if (optlen <= 0) 431 break; 432 } 433 434 switch (opt) { 435 436 default: 437 break; 438 439 case TCPOPT_MAXSEG: 440 bits |= IP_FW_TCPOPT_MSS; 441 break; 442 443 case TCPOPT_WINDOW: 444 bits |= IP_FW_TCPOPT_WINDOW; 445 break; 446 447 case TCPOPT_SACK_PERMITTED: 448 case TCPOPT_SACK: 449 bits |= IP_FW_TCPOPT_SACK; 450 break; 451 452 case TCPOPT_TIMESTAMP: 453 bits |= IP_FW_TCPOPT_TS; 454 break; 455 456 } 457 } 458 return (flags_match(cmd, bits)); 459} 460 461static int 462iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 463{ 464 if (ifp == NULL) /* no iface with this packet, match fails */ 465 return 0; 466 /* Check by name or by IP address */ 467 if (cmd->name[0] != '\0') { /* match by name */ 468 /* Check name */ 469 if (cmd->p.glob) { 470 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 471 return(1); 472 } else { 473 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 474 return(1); 475 } 476 } else { 477 struct ifaddr *ia; 478 479 /* XXX lock? */ 480 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 481 if (ia->ifa_addr->sa_family != AF_INET) 482 continue; 483 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 484 (ia->ifa_addr))->sin_addr.s_addr) 485 return(1); /* match */ 486 } 487 } 488 return(0); /* no match, fail ... */ 489} 490 491/* 492 * The verify_path function checks if a route to the src exists and 493 * if it is reachable via ifp (when provided). 494 * 495 * The 'verrevpath' option checks that the interface that an IP packet 496 * arrives on is the same interface that traffic destined for the 497 * packet's source address would be routed out of. The 'versrcreach' 498 * option just checks that the source address is reachable via any route 499 * (except default) in the routing table. These two are a measure to block 500 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 501 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 502 * is purposely reminiscent of the Cisco IOS command, 503 * 504 * ip verify unicast reverse-path 505 * ip verify unicast source reachable-via any 506 * 507 * which implements the same functionality. But note that syntax is 508 * misleading. The check may be performed on all IP packets whether unicast, 509 * multicast, or broadcast. 510 */ 511static int 512verify_path(struct in_addr src, struct ifnet *ifp) 513{ 514 struct route ro; 515 struct sockaddr_in *dst; 516 517 bzero(&ro, sizeof(ro)); 518 519 dst = (struct sockaddr_in *)&(ro.ro_dst); 520 dst->sin_family = AF_INET; 521 dst->sin_len = sizeof(*dst); 522 dst->sin_addr = src; 523 rtalloc_ign(&ro, RTF_CLONING); 524 525 if (ro.ro_rt == NULL) 526 return 0; 527 528 /* 529 * If ifp is provided, check for equality with rtentry. 530 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 531 * in order to pass packets injected back by if_simloop(): 532 * if useloopback == 1 routing entry (via lo0) for our own address 533 * may exist, so we need to handle routing assymetry. 534 */ 535 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 536 RTFREE(ro.ro_rt); 537 return 0; 538 } 539 540 /* if no ifp provided, check if rtentry is not default route */ 541 if (ifp == NULL && 542 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 543 RTFREE(ro.ro_rt); 544 return 0; 545 } 546 547 /* or if this is a blackhole/reject route */ 548 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 549 RTFREE(ro.ro_rt); 550 return 0; 551 } 552 553 /* found valid route */ 554 RTFREE(ro.ro_rt); 555 return 1; 556} 557 558#ifdef INET6 559/* 560 * ipv6 specific rules here... 561 */ 562static __inline int 563icmp6type_match (int type, ipfw_insn_u32 *cmd) 564{ 565 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 566} 567 568static int 569flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 570{ 571 int i; 572 for (i=0; i <= cmd->o.arg1; ++i ) 573 if (curr_flow == cmd->d[i] ) 574 return 1; 575 return 0; 576} 577 578/* support for IP6_*_ME opcodes */ 579static int 580search_ip6_addr_net (struct in6_addr * ip6_addr) 581{ 582 struct ifnet *mdc; 583 struct ifaddr *mdc2; 584 struct in6_ifaddr *fdm; 585 struct in6_addr copia; 586 587 TAILQ_FOREACH(mdc, &ifnet, if_link) 588 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 589 if (mdc2->ifa_addr->sa_family == AF_INET6) { 590 fdm = (struct in6_ifaddr *)mdc2; 591 copia = fdm->ia_addr.sin6_addr; 592 /* need for leaving scope_id in the sock_addr */ 593 in6_clearscope(&copia); 594 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 595 return 1; 596 } 597 } 598 return 0; 599} 600 601static int 602verify_path6(struct in6_addr *src, struct ifnet *ifp) 603{ 604 struct route_in6 ro; 605 struct sockaddr_in6 *dst; 606 607 bzero(&ro, sizeof(ro)); 608 609 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 610 dst->sin6_family = AF_INET6; 611 dst->sin6_len = sizeof(*dst); 612 dst->sin6_addr = *src; 613 rtalloc_ign((struct route *)&ro, RTF_CLONING); 614 615 if (ro.ro_rt == NULL) 616 return 0; 617 618 /* 619 * if ifp is provided, check for equality with rtentry 620 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 621 * to support the case of sending packets to an address of our own. 622 * (where the former interface is the first argument of if_simloop() 623 * (=ifp), the latter is lo0) 624 */ 625 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 626 RTFREE(ro.ro_rt); 627 return 0; 628 } 629 630 /* if no ifp provided, check if rtentry is not default route */ 631 if (ifp == NULL && 632 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 633 RTFREE(ro.ro_rt); 634 return 0; 635 } 636 637 /* or if this is a blackhole/reject route */ 638 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 639 RTFREE(ro.ro_rt); 640 return 0; 641 } 642 643 /* found valid route */ 644 RTFREE(ro.ro_rt); 645 return 1; 646 647} 648static __inline int 649hash_packet6(struct ipfw_flow_id *id) 650{ 651 u_int32_t i; 652 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 653 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 654 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 655 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 656 (id->dst_port) ^ (id->src_port); 657 return i; 658} 659 660static int 661is_icmp6_query(int icmp6_type) 662{ 663 if ((icmp6_type <= ICMP6_MAXTYPE) && 664 (icmp6_type == ICMP6_ECHO_REQUEST || 665 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 666 icmp6_type == ICMP6_WRUREQUEST || 667 icmp6_type == ICMP6_FQDN_QUERY || 668 icmp6_type == ICMP6_NI_QUERY)) 669 return (1); 670 671 return (0); 672} 673 674static void 675send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 676{ 677 struct mbuf *m; 678 679 m = args->m; 680 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 681 struct tcphdr *tcp; 682 tcp_seq ack, seq; 683 int flags; 684 struct { 685 struct ip6_hdr ip6; 686 struct tcphdr th; 687 } ti; 688 tcp = (struct tcphdr *)((char *)ip6 + hlen); 689 690 if ((tcp->th_flags & TH_RST) != 0) { 691 m_freem(m); 692 args->m = NULL; 693 return; 694 } 695 696 ti.ip6 = *ip6; 697 ti.th = *tcp; 698 ti.th.th_seq = ntohl(ti.th.th_seq); 699 ti.th.th_ack = ntohl(ti.th.th_ack); 700 ti.ip6.ip6_nxt = IPPROTO_TCP; 701 702 if (ti.th.th_flags & TH_ACK) { 703 ack = 0; 704 seq = ti.th.th_ack; 705 flags = TH_RST; 706 } else { 707 ack = ti.th.th_seq; 708 if ((m->m_flags & M_PKTHDR) != 0) { 709 /* 710 * total new data to ACK is: 711 * total packet length, 712 * minus the header length, 713 * minus the tcp header length. 714 */ 715 ack += m->m_pkthdr.len - hlen 716 - (ti.th.th_off << 2); 717 } else if (ip6->ip6_plen) { 718 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 719 hlen - (ti.th.th_off << 2); 720 } else { 721 m_freem(m); 722 return; 723 } 724 if (tcp->th_flags & TH_SYN) 725 ack++; 726 seq = 0; 727 flags = TH_RST|TH_ACK; 728 } 729 bcopy(&ti, ip6, sizeof(ti)); 730 /* 731 * m is only used to recycle the mbuf 732 * The data in it is never read so we don't need 733 * to correct the offsets or anything 734 */ 735 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 736 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 737#if 0 738 /* 739 * Unlike above, the mbufs need to line up with the ip6 hdr, 740 * as the contents are read. We need to m_adj() the 741 * needed amount. 742 * The mbuf will however be thrown away so we can adjust it. 743 * Remember we did an m_pullup on it already so we 744 * can make some assumptions about contiguousness. 745 */ 746 if (args->L3offset) 747 m_adj(m, args->L3offset); 748#endif 749 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 750 } else 751 m_freem(m); 752 753 args->m = NULL; 754} 755 756#endif /* INET6 */ 757 758static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 759 760#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 761#define SNP(buf) buf, sizeof(buf) 762 763/* 764 * We enter here when we have a rule with O_LOG. 765 * XXX this function alone takes about 2Kbytes of code! 766 */ 767static void 768ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 769 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 770 struct ip *ip) 771{ 772 struct ether_header *eh = args->eh; 773 char *action; 774 int limit_reached = 0; 775 char action2[40], proto[128], fragment[32]; 776 777 fragment[0] = '\0'; 778 proto[0] = '\0'; 779 780 if (f == NULL) { /* bogus pkt */ 781 if (verbose_limit != 0 && norule_counter >= verbose_limit) 782 return; 783 norule_counter++; 784 if (norule_counter == verbose_limit) 785 limit_reached = verbose_limit; 786 action = "Refuse"; 787 } else { /* O_LOG is the first action, find the real one */ 788 ipfw_insn *cmd = ACTION_PTR(f); 789 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 790 791 if (l->max_log != 0 && l->log_left == 0) 792 return; 793 l->log_left--; 794 if (l->log_left == 0) 795 limit_reached = l->max_log; 796 cmd += F_LEN(cmd); /* point to first action */ 797 if (cmd->opcode == O_ALTQ) { 798 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 799 800 snprintf(SNPARGS(action2, 0), "Altq %d", 801 altq->qid); 802 cmd += F_LEN(cmd); 803 } 804 if (cmd->opcode == O_PROB) 805 cmd += F_LEN(cmd); 806 807 if (cmd->opcode == O_TAG) 808 cmd += F_LEN(cmd); 809 810 action = action2; 811 switch (cmd->opcode) { 812 case O_DENY: 813 action = "Deny"; 814 break; 815 816 case O_REJECT: 817 if (cmd->arg1==ICMP_REJECT_RST) 818 action = "Reset"; 819 else if (cmd->arg1==ICMP_UNREACH_HOST) 820 action = "Reject"; 821 else 822 snprintf(SNPARGS(action2, 0), "Unreach %d", 823 cmd->arg1); 824 break; 825 826 case O_UNREACH6: 827 if (cmd->arg1==ICMP6_UNREACH_RST) 828 action = "Reset"; 829 else 830 snprintf(SNPARGS(action2, 0), "Unreach %d", 831 cmd->arg1); 832 break; 833 834 case O_ACCEPT: 835 action = "Accept"; 836 break; 837 case O_COUNT: 838 action = "Count"; 839 break; 840 case O_DIVERT: 841 snprintf(SNPARGS(action2, 0), "Divert %d", 842 cmd->arg1); 843 break; 844 case O_TEE: 845 snprintf(SNPARGS(action2, 0), "Tee %d", 846 cmd->arg1); 847 break; 848 case O_SKIPTO: 849 snprintf(SNPARGS(action2, 0), "SkipTo %d", 850 cmd->arg1); 851 break; 852 case O_PIPE: 853 snprintf(SNPARGS(action2, 0), "Pipe %d", 854 cmd->arg1); 855 break; 856 case O_QUEUE: 857 snprintf(SNPARGS(action2, 0), "Queue %d", 858 cmd->arg1); 859 break; 860 case O_FORWARD_IP: { 861 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 862 int len; 863 struct in_addr dummyaddr; 864 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 865 dummyaddr.s_addr = htonl(tablearg); 866 else 867 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 868 869 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 870 inet_ntoa(dummyaddr)); 871 872 if (sa->sa.sin_port) 873 snprintf(SNPARGS(action2, len), ":%d", 874 sa->sa.sin_port); 875 } 876 break; 877 case O_NETGRAPH: 878 snprintf(SNPARGS(action2, 0), "Netgraph %d", 879 cmd->arg1); 880 break; 881 case O_NGTEE: 882 snprintf(SNPARGS(action2, 0), "Ngtee %d", 883 cmd->arg1); 884 break; 885 case O_NAT: 886 action = "Nat"; 887 break; 888 default: 889 action = "UNKNOWN"; 890 break; 891 } 892 } 893 894 if (hlen == 0) { /* non-ip */ 895 snprintf(SNPARGS(proto, 0), "MAC"); 896 897 } else { 898 int len; 899 char src[48], dst[48]; 900 struct icmphdr *icmp; 901 struct tcphdr *tcp; 902 struct udphdr *udp; 903 /* Initialize to make compiler happy. */ 904 struct ip *ip = NULL; 905#ifdef INET6 906 struct ip6_hdr *ip6 = NULL; 907 struct icmp6_hdr *icmp6; 908#endif 909 src[0] = '\0'; 910 dst[0] = '\0'; 911#ifdef INET6 912 if (IS_IP6_FLOW_ID(&(args->f_id))) { 913 char ip6buf[INET6_ADDRSTRLEN]; 914 snprintf(src, sizeof(src), "[%s]", 915 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 916 snprintf(dst, sizeof(dst), "[%s]", 917 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 918 919 ip6 = (struct ip6_hdr *)ip; 920 tcp = (struct tcphdr *)(((char *)ip) + hlen); 921 udp = (struct udphdr *)(((char *)ip) + hlen); 922 } else 923#endif 924 { 925 tcp = L3HDR(struct tcphdr, ip); 926 udp = L3HDR(struct udphdr, ip); 927 928 inet_ntoa_r(ip->ip_src, src); 929 inet_ntoa_r(ip->ip_dst, dst); 930 } 931 932 switch (args->f_id.proto) { 933 case IPPROTO_TCP: 934 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 935 if (offset == 0) 936 snprintf(SNPARGS(proto, len), ":%d %s:%d", 937 ntohs(tcp->th_sport), 938 dst, 939 ntohs(tcp->th_dport)); 940 else 941 snprintf(SNPARGS(proto, len), " %s", dst); 942 break; 943 944 case IPPROTO_UDP: 945 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 946 if (offset == 0) 947 snprintf(SNPARGS(proto, len), ":%d %s:%d", 948 ntohs(udp->uh_sport), 949 dst, 950 ntohs(udp->uh_dport)); 951 else 952 snprintf(SNPARGS(proto, len), " %s", dst); 953 break; 954 955 case IPPROTO_ICMP: 956 icmp = L3HDR(struct icmphdr, ip); 957 if (offset == 0) 958 len = snprintf(SNPARGS(proto, 0), 959 "ICMP:%u.%u ", 960 icmp->icmp_type, icmp->icmp_code); 961 else 962 len = snprintf(SNPARGS(proto, 0), "ICMP "); 963 len += snprintf(SNPARGS(proto, len), "%s", src); 964 snprintf(SNPARGS(proto, len), " %s", dst); 965 break; 966#ifdef INET6 967 case IPPROTO_ICMPV6: 968 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 969 if (offset == 0) 970 len = snprintf(SNPARGS(proto, 0), 971 "ICMPv6:%u.%u ", 972 icmp6->icmp6_type, icmp6->icmp6_code); 973 else 974 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 975 len += snprintf(SNPARGS(proto, len), "%s", src); 976 snprintf(SNPARGS(proto, len), " %s", dst); 977 break; 978#endif 979 default: 980 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 981 args->f_id.proto, src); 982 snprintf(SNPARGS(proto, len), " %s", dst); 983 break; 984 } 985 986#ifdef INET6 987 if (IS_IP6_FLOW_ID(&(args->f_id))) { 988 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 989 snprintf(SNPARGS(fragment, 0), 990 " (frag %08x:%d@%d%s)", 991 args->f_id.frag_id6, 992 ntohs(ip6->ip6_plen) - hlen, 993 ntohs(offset & IP6F_OFF_MASK) << 3, 994 (offset & IP6F_MORE_FRAG) ? "+" : ""); 995 } else 996#endif 997 { 998 int ip_off, ip_len; 999 if (eh != NULL) { /* layer 2 packets are as on the wire */ 1000 ip_off = ntohs(ip->ip_off); 1001 ip_len = ntohs(ip->ip_len); 1002 } else { 1003 ip_off = ip->ip_off; 1004 ip_len = ip->ip_len; 1005 } 1006 if (ip_off & (IP_MF | IP_OFFMASK)) 1007 snprintf(SNPARGS(fragment, 0), 1008 " (frag %d:%d@%d%s)", 1009 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1010 offset << 3, 1011 (ip_off & IP_MF) ? "+" : ""); 1012 } 1013 } 1014 if (oif || m->m_pkthdr.rcvif) 1015 log(LOG_SECURITY | LOG_INFO, 1016 "ipfw: %d %s %s %s via %s%s\n", 1017 f ? f->rulenum : -1, 1018 action, proto, oif ? "out" : "in", 1019 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1020 fragment); 1021 else 1022 log(LOG_SECURITY | LOG_INFO, 1023 "ipfw: %d %s %s [no if info]%s\n", 1024 f ? f->rulenum : -1, 1025 action, proto, fragment); 1026 if (limit_reached) 1027 log(LOG_SECURITY | LOG_NOTICE, 1028 "ipfw: limit %d reached on entry %d\n", 1029 limit_reached, f ? f->rulenum : -1); 1030} 1031 1032/* 1033 * IMPORTANT: the hash function for dynamic rules must be commutative 1034 * in source and destination (ip,port), because rules are bidirectional 1035 * and we want to find both in the same bucket. 1036 */ 1037static __inline int 1038hash_packet(struct ipfw_flow_id *id) 1039{ 1040 u_int32_t i; 1041 1042#ifdef INET6 1043 if (IS_IP6_FLOW_ID(id)) 1044 i = hash_packet6(id); 1045 else 1046#endif /* INET6 */ 1047 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1048 i &= (curr_dyn_buckets - 1); 1049 return i; 1050} 1051 1052/** 1053 * unlink a dynamic rule from a chain. prev is a pointer to 1054 * the previous one, q is a pointer to the rule to delete, 1055 * head is a pointer to the head of the queue. 1056 * Modifies q and potentially also head. 1057 */ 1058#define UNLINK_DYN_RULE(prev, head, q) { \ 1059 ipfw_dyn_rule *old_q = q; \ 1060 \ 1061 /* remove a refcount to the parent */ \ 1062 if (q->dyn_type == O_LIMIT) \ 1063 q->parent->count--; \ 1064 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1065 (q->id.src_ip), (q->id.src_port), \ 1066 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1067 if (prev != NULL) \ 1068 prev->next = q = q->next; \ 1069 else \ 1070 head = q = q->next; \ 1071 dyn_count--; \ 1072 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1073 1074#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1075 1076/** 1077 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1078 * 1079 * If keep_me == NULL, rules are deleted even if not expired, 1080 * otherwise only expired rules are removed. 1081 * 1082 * The value of the second parameter is also used to point to identify 1083 * a rule we absolutely do not want to remove (e.g. because we are 1084 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1085 * rules). The pointer is only used for comparison, so any non-null 1086 * value will do. 1087 */ 1088static void 1089remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1090{ 1091 static u_int32_t last_remove = 0; 1092 1093#define FORCE (keep_me == NULL) 1094 1095 ipfw_dyn_rule *prev, *q; 1096 int i, pass = 0, max_pass = 0; 1097 1098 IPFW_DYN_LOCK_ASSERT(); 1099 1100 if (ipfw_dyn_v == NULL || dyn_count == 0) 1101 return; 1102 /* do not expire more than once per second, it is useless */ 1103 if (!FORCE && last_remove == time_uptime) 1104 return; 1105 last_remove = time_uptime; 1106 1107 /* 1108 * because O_LIMIT refer to parent rules, during the first pass only 1109 * remove child and mark any pending LIMIT_PARENT, and remove 1110 * them in a second pass. 1111 */ 1112next_pass: 1113 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1114 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1115 /* 1116 * Logic can become complex here, so we split tests. 1117 */ 1118 if (q == keep_me) 1119 goto next; 1120 if (rule != NULL && rule != q->rule) 1121 goto next; /* not the one we are looking for */ 1122 if (q->dyn_type == O_LIMIT_PARENT) { 1123 /* 1124 * handle parent in the second pass, 1125 * record we need one. 1126 */ 1127 max_pass = 1; 1128 if (pass == 0) 1129 goto next; 1130 if (FORCE && q->count != 0 ) { 1131 /* XXX should not happen! */ 1132 printf("ipfw: OUCH! cannot remove rule," 1133 " count %d\n", q->count); 1134 } 1135 } else { 1136 if (!FORCE && 1137 !TIME_LEQ( q->expire, time_uptime )) 1138 goto next; 1139 } 1140 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1141 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1142 continue; 1143 } 1144next: 1145 prev=q; 1146 q=q->next; 1147 } 1148 } 1149 if (pass++ < max_pass) 1150 goto next_pass; 1151} 1152 1153 1154/** 1155 * lookup a dynamic rule. 1156 */ 1157static ipfw_dyn_rule * 1158lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1159 struct tcphdr *tcp) 1160{ 1161 /* 1162 * stateful ipfw extensions. 1163 * Lookup into dynamic session queue 1164 */ 1165#define MATCH_REVERSE 0 1166#define MATCH_FORWARD 1 1167#define MATCH_NONE 2 1168#define MATCH_UNKNOWN 3 1169 int i, dir = MATCH_NONE; 1170 ipfw_dyn_rule *prev, *q=NULL; 1171 1172 IPFW_DYN_LOCK_ASSERT(); 1173 1174 if (ipfw_dyn_v == NULL) 1175 goto done; /* not found */ 1176 i = hash_packet( pkt ); 1177 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1178 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1179 goto next; 1180 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1181 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1182 continue; 1183 } 1184 if (pkt->proto == q->id.proto && 1185 q->dyn_type != O_LIMIT_PARENT) { 1186 if (IS_IP6_FLOW_ID(pkt)) { 1187 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1188 &(q->id.src_ip6)) && 1189 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1190 &(q->id.dst_ip6)) && 1191 pkt->src_port == q->id.src_port && 1192 pkt->dst_port == q->id.dst_port ) { 1193 dir = MATCH_FORWARD; 1194 break; 1195 } 1196 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1197 &(q->id.dst_ip6)) && 1198 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1199 &(q->id.src_ip6)) && 1200 pkt->src_port == q->id.dst_port && 1201 pkt->dst_port == q->id.src_port ) { 1202 dir = MATCH_REVERSE; 1203 break; 1204 } 1205 } else { 1206 if (pkt->src_ip == q->id.src_ip && 1207 pkt->dst_ip == q->id.dst_ip && 1208 pkt->src_port == q->id.src_port && 1209 pkt->dst_port == q->id.dst_port ) { 1210 dir = MATCH_FORWARD; 1211 break; 1212 } 1213 if (pkt->src_ip == q->id.dst_ip && 1214 pkt->dst_ip == q->id.src_ip && 1215 pkt->src_port == q->id.dst_port && 1216 pkt->dst_port == q->id.src_port ) { 1217 dir = MATCH_REVERSE; 1218 break; 1219 } 1220 } 1221 } 1222next: 1223 prev = q; 1224 q = q->next; 1225 } 1226 if (q == NULL) 1227 goto done; /* q = NULL, not found */ 1228 1229 if ( prev != NULL) { /* found and not in front */ 1230 prev->next = q->next; 1231 q->next = ipfw_dyn_v[i]; 1232 ipfw_dyn_v[i] = q; 1233 } 1234 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1235 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1236 1237#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1238#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1239 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1240 switch (q->state) { 1241 case TH_SYN: /* opening */ 1242 q->expire = time_uptime + dyn_syn_lifetime; 1243 break; 1244 1245 case BOTH_SYN: /* move to established */ 1246 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1247 case BOTH_SYN | (TH_FIN << 8) : 1248 if (tcp) { 1249#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1250 u_int32_t ack = ntohl(tcp->th_ack); 1251 if (dir == MATCH_FORWARD) { 1252 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1253 q->ack_fwd = ack; 1254 else { /* ignore out-of-sequence */ 1255 break; 1256 } 1257 } else { 1258 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1259 q->ack_rev = ack; 1260 else { /* ignore out-of-sequence */ 1261 break; 1262 } 1263 } 1264 } 1265 q->expire = time_uptime + dyn_ack_lifetime; 1266 break; 1267 1268 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1269 if (dyn_fin_lifetime >= dyn_keepalive_period) 1270 dyn_fin_lifetime = dyn_keepalive_period - 1; 1271 q->expire = time_uptime + dyn_fin_lifetime; 1272 break; 1273 1274 default: 1275#if 0 1276 /* 1277 * reset or some invalid combination, but can also 1278 * occur if we use keep-state the wrong way. 1279 */ 1280 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1281 printf("invalid state: 0x%x\n", q->state); 1282#endif 1283 if (dyn_rst_lifetime >= dyn_keepalive_period) 1284 dyn_rst_lifetime = dyn_keepalive_period - 1; 1285 q->expire = time_uptime + dyn_rst_lifetime; 1286 break; 1287 } 1288 } else if (pkt->proto == IPPROTO_UDP) { 1289 q->expire = time_uptime + dyn_udp_lifetime; 1290 } else { 1291 /* other protocols */ 1292 q->expire = time_uptime + dyn_short_lifetime; 1293 } 1294done: 1295 if (match_direction) 1296 *match_direction = dir; 1297 return q; 1298} 1299 1300static ipfw_dyn_rule * 1301lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1302 struct tcphdr *tcp) 1303{ 1304 ipfw_dyn_rule *q; 1305 1306 IPFW_DYN_LOCK(); 1307 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1308 if (q == NULL) 1309 IPFW_DYN_UNLOCK(); 1310 /* NB: return table locked when q is not NULL */ 1311 return q; 1312} 1313 1314static void 1315realloc_dynamic_table(void) 1316{ 1317 IPFW_DYN_LOCK_ASSERT(); 1318 1319 /* 1320 * Try reallocation, make sure we have a power of 2 and do 1321 * not allow more than 64k entries. In case of overflow, 1322 * default to 1024. 1323 */ 1324 1325 if (dyn_buckets > 65536) 1326 dyn_buckets = 1024; 1327 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1328 dyn_buckets = curr_dyn_buckets; /* reset */ 1329 return; 1330 } 1331 curr_dyn_buckets = dyn_buckets; 1332 if (ipfw_dyn_v != NULL) 1333 free(ipfw_dyn_v, M_IPFW); 1334 for (;;) { 1335 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1336 M_IPFW, M_NOWAIT | M_ZERO); 1337 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1338 break; 1339 curr_dyn_buckets /= 2; 1340 } 1341} 1342 1343/** 1344 * Install state of type 'type' for a dynamic session. 1345 * The hash table contains two type of rules: 1346 * - regular rules (O_KEEP_STATE) 1347 * - rules for sessions with limited number of sess per user 1348 * (O_LIMIT). When they are created, the parent is 1349 * increased by 1, and decreased on delete. In this case, 1350 * the third parameter is the parent rule and not the chain. 1351 * - "parent" rules for the above (O_LIMIT_PARENT). 1352 */ 1353static ipfw_dyn_rule * 1354add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1355{ 1356 ipfw_dyn_rule *r; 1357 int i; 1358 1359 IPFW_DYN_LOCK_ASSERT(); 1360 1361 if (ipfw_dyn_v == NULL || 1362 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1363 realloc_dynamic_table(); 1364 if (ipfw_dyn_v == NULL) 1365 return NULL; /* failed ! */ 1366 } 1367 i = hash_packet(id); 1368 1369 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1370 if (r == NULL) { 1371 printf ("ipfw: sorry cannot allocate state\n"); 1372 return NULL; 1373 } 1374 1375 /* increase refcount on parent, and set pointer */ 1376 if (dyn_type == O_LIMIT) { 1377 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1378 if ( parent->dyn_type != O_LIMIT_PARENT) 1379 panic("invalid parent"); 1380 parent->count++; 1381 r->parent = parent; 1382 rule = parent->rule; 1383 } 1384 1385 r->id = *id; 1386 r->expire = time_uptime + dyn_syn_lifetime; 1387 r->rule = rule; 1388 r->dyn_type = dyn_type; 1389 r->pcnt = r->bcnt = 0; 1390 r->count = 0; 1391 1392 r->bucket = i; 1393 r->next = ipfw_dyn_v[i]; 1394 ipfw_dyn_v[i] = r; 1395 dyn_count++; 1396 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1397 dyn_type, 1398 (r->id.src_ip), (r->id.src_port), 1399 (r->id.dst_ip), (r->id.dst_port), 1400 dyn_count ); ) 1401 return r; 1402} 1403 1404/** 1405 * lookup dynamic parent rule using pkt and rule as search keys. 1406 * If the lookup fails, then install one. 1407 */ 1408static ipfw_dyn_rule * 1409lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1410{ 1411 ipfw_dyn_rule *q; 1412 int i; 1413 1414 IPFW_DYN_LOCK_ASSERT(); 1415 1416 if (ipfw_dyn_v) { 1417 int is_v6 = IS_IP6_FLOW_ID(pkt); 1418 i = hash_packet( pkt ); 1419 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1420 if (q->dyn_type == O_LIMIT_PARENT && 1421 rule== q->rule && 1422 pkt->proto == q->id.proto && 1423 pkt->src_port == q->id.src_port && 1424 pkt->dst_port == q->id.dst_port && 1425 ( 1426 (is_v6 && 1427 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1428 &(q->id.src_ip6)) && 1429 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1430 &(q->id.dst_ip6))) || 1431 (!is_v6 && 1432 pkt->src_ip == q->id.src_ip && 1433 pkt->dst_ip == q->id.dst_ip) 1434 ) 1435 ) { 1436 q->expire = time_uptime + dyn_short_lifetime; 1437 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1438 return q; 1439 } 1440 } 1441 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1442} 1443 1444/** 1445 * Install dynamic state for rule type cmd->o.opcode 1446 * 1447 * Returns 1 (failure) if state is not installed because of errors or because 1448 * session limitations are enforced. 1449 */ 1450static int 1451install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1452 struct ip_fw_args *args, uint32_t tablearg) 1453{ 1454 static int last_log; 1455 ipfw_dyn_rule *q; 1456 struct in_addr da; 1457 char src[48], dst[48]; 1458 1459 src[0] = '\0'; 1460 dst[0] = '\0'; 1461 1462 DEB( 1463 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1464 __func__, cmd->o.opcode, 1465 (args->f_id.src_ip), (args->f_id.src_port), 1466 (args->f_id.dst_ip), (args->f_id.dst_port)); 1467 ) 1468 1469 IPFW_DYN_LOCK(); 1470 1471 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1472 1473 if (q != NULL) { /* should never occur */ 1474 if (last_log != time_uptime) { 1475 last_log = time_uptime; 1476 printf("ipfw: %s: entry already present, done\n", 1477 __func__); 1478 } 1479 IPFW_DYN_UNLOCK(); 1480 return (0); 1481 } 1482 1483 if (dyn_count >= dyn_max) 1484 /* Run out of slots, try to remove any expired rule. */ 1485 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1486 1487 if (dyn_count >= dyn_max) { 1488 if (last_log != time_uptime) { 1489 last_log = time_uptime; 1490 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1491 } 1492 IPFW_DYN_UNLOCK(); 1493 return (1); /* cannot install, notify caller */ 1494 } 1495 1496 switch (cmd->o.opcode) { 1497 case O_KEEP_STATE: /* bidir rule */ 1498 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1499 break; 1500 1501 case O_LIMIT: { /* limit number of sessions */ 1502 struct ipfw_flow_id id; 1503 ipfw_dyn_rule *parent; 1504 uint32_t conn_limit; 1505 uint16_t limit_mask = cmd->limit_mask; 1506 1507 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1508 tablearg : cmd->conn_limit; 1509 1510 DEB( 1511 if (cmd->conn_limit == IP_FW_TABLEARG) 1512 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1513 "(tablearg)\n", __func__, conn_limit); 1514 else 1515 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1516 __func__, conn_limit); 1517 ) 1518 1519 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1520 id.proto = args->f_id.proto; 1521 id.addr_type = args->f_id.addr_type; 1522 1523 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1524 if (limit_mask & DYN_SRC_ADDR) 1525 id.src_ip6 = args->f_id.src_ip6; 1526 if (limit_mask & DYN_DST_ADDR) 1527 id.dst_ip6 = args->f_id.dst_ip6; 1528 } else { 1529 if (limit_mask & DYN_SRC_ADDR) 1530 id.src_ip = args->f_id.src_ip; 1531 if (limit_mask & DYN_DST_ADDR) 1532 id.dst_ip = args->f_id.dst_ip; 1533 } 1534 if (limit_mask & DYN_SRC_PORT) 1535 id.src_port = args->f_id.src_port; 1536 if (limit_mask & DYN_DST_PORT) 1537 id.dst_port = args->f_id.dst_port; 1538 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1539 printf("ipfw: %s: add parent failed\n", __func__); 1540 IPFW_DYN_UNLOCK(); 1541 return (1); 1542 } 1543 1544 if (parent->count >= conn_limit) { 1545 /* See if we can remove some expired rule. */ 1546 remove_dyn_rule(rule, parent); 1547 if (parent->count >= conn_limit) { 1548 if (fw_verbose && last_log != time_uptime) { 1549 last_log = time_uptime; 1550#ifdef INET6 1551 /* 1552 * XXX IPv6 flows are not 1553 * supported yet. 1554 */ 1555 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1556 char ip6buf[INET6_ADDRSTRLEN]; 1557 snprintf(src, sizeof(src), 1558 "[%s]", ip6_sprintf(ip6buf, 1559 &args->f_id.src_ip6)); 1560 snprintf(dst, sizeof(dst), 1561 "[%s]", ip6_sprintf(ip6buf, 1562 &args->f_id.dst_ip6)); 1563 } else 1564#endif 1565 { 1566 da.s_addr = 1567 htonl(args->f_id.src_ip); 1568 inet_ntoa_r(da, src); 1569 da.s_addr = 1570 htonl(args->f_id.dst_ip); 1571 inet_ntoa_r(da, dst); 1572 } 1573 log(LOG_SECURITY | LOG_DEBUG, 1574 "%s %s:%u -> %s:%u, %s\n", 1575 "drop session", 1576 src, (args->f_id.src_port), 1577 dst, (args->f_id.dst_port), 1578 "too many entries"); 1579 } 1580 IPFW_DYN_UNLOCK(); 1581 return (1); 1582 } 1583 } 1584 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1585 break; 1586 } 1587 default: 1588 printf("ipfw: %s: unknown dynamic rule type %u\n", 1589 __func__, cmd->o.opcode); 1590 IPFW_DYN_UNLOCK(); 1591 return (1); 1592 } 1593 1594 /* XXX just set lifetime */ 1595 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1596 1597 IPFW_DYN_UNLOCK(); 1598 return (0); 1599} 1600 1601/* 1602 * Generate a TCP packet, containing either a RST or a keepalive. 1603 * When flags & TH_RST, we are sending a RST packet, because of a 1604 * "reset" action matched the packet. 1605 * Otherwise we are sending a keepalive, and flags & TH_ 1606 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1607 * so that MAC can label the reply appropriately. 1608 */ 1609static struct mbuf * 1610send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1611 u_int32_t ack, int flags) 1612{ 1613 struct mbuf *m; 1614 struct ip *ip; 1615 struct tcphdr *tcp; 1616 1617 MGETHDR(m, M_DONTWAIT, MT_DATA); 1618 if (m == 0) 1619 return (NULL); 1620 m->m_pkthdr.rcvif = (struct ifnet *)0; 1621 1622#ifdef MAC 1623 if (replyto != NULL) 1624 mac_create_mbuf_netlayer(replyto, m); 1625 else 1626 mac_create_mbuf_from_firewall(m); 1627#else 1628 (void)replyto; /* don't warn about unused arg */ 1629#endif 1630 1631 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1632 m->m_data += max_linkhdr; 1633 1634 ip = mtod(m, struct ip *); 1635 bzero(ip, m->m_len); 1636 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1637 ip->ip_p = IPPROTO_TCP; 1638 tcp->th_off = 5; 1639 /* 1640 * Assume we are sending a RST (or a keepalive in the reverse 1641 * direction), swap src and destination addresses and ports. 1642 */ 1643 ip->ip_src.s_addr = htonl(id->dst_ip); 1644 ip->ip_dst.s_addr = htonl(id->src_ip); 1645 tcp->th_sport = htons(id->dst_port); 1646 tcp->th_dport = htons(id->src_port); 1647 if (flags & TH_RST) { /* we are sending a RST */ 1648 if (flags & TH_ACK) { 1649 tcp->th_seq = htonl(ack); 1650 tcp->th_ack = htonl(0); 1651 tcp->th_flags = TH_RST; 1652 } else { 1653 if (flags & TH_SYN) 1654 seq++; 1655 tcp->th_seq = htonl(0); 1656 tcp->th_ack = htonl(seq); 1657 tcp->th_flags = TH_RST | TH_ACK; 1658 } 1659 } else { 1660 /* 1661 * We are sending a keepalive. flags & TH_SYN determines 1662 * the direction, forward if set, reverse if clear. 1663 * NOTE: seq and ack are always assumed to be correct 1664 * as set by the caller. This may be confusing... 1665 */ 1666 if (flags & TH_SYN) { 1667 /* 1668 * we have to rewrite the correct addresses! 1669 */ 1670 ip->ip_dst.s_addr = htonl(id->dst_ip); 1671 ip->ip_src.s_addr = htonl(id->src_ip); 1672 tcp->th_dport = htons(id->dst_port); 1673 tcp->th_sport = htons(id->src_port); 1674 } 1675 tcp->th_seq = htonl(seq); 1676 tcp->th_ack = htonl(ack); 1677 tcp->th_flags = TH_ACK; 1678 } 1679 /* 1680 * set ip_len to the payload size so we can compute 1681 * the tcp checksum on the pseudoheader 1682 * XXX check this, could save a couple of words ? 1683 */ 1684 ip->ip_len = htons(sizeof(struct tcphdr)); 1685 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1686 /* 1687 * now fill fields left out earlier 1688 */ 1689 ip->ip_ttl = ip_defttl; 1690 ip->ip_len = m->m_pkthdr.len; 1691 m->m_flags |= M_SKIP_FIREWALL; 1692 return (m); 1693} 1694 1695/* 1696 * sends a reject message, consuming the mbuf passed as an argument. 1697 */ 1698static void 1699send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1700{ 1701 1702#if 0 1703 /* XXX When ip is not guaranteed to be at mtod() we will 1704 * need to account for this */ 1705 * The mbuf will however be thrown away so we can adjust it. 1706 * Remember we did an m_pullup on it already so we 1707 * can make some assumptions about contiguousness. 1708 */ 1709 if (args->L3offset) 1710 m_adj(m, args->L3offset); 1711#endif 1712 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1713 /* We need the IP header in host order for icmp_error(). */ 1714 if (args->eh != NULL) { 1715 ip->ip_len = ntohs(ip->ip_len); 1716 ip->ip_off = ntohs(ip->ip_off); 1717 } 1718 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1719 } else if (args->f_id.proto == IPPROTO_TCP) { 1720 struct tcphdr *const tcp = 1721 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1722 if ( (tcp->th_flags & TH_RST) == 0) { 1723 struct mbuf *m; 1724 m = send_pkt(args->m, &(args->f_id), 1725 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1726 tcp->th_flags | TH_RST); 1727 if (m != NULL) 1728 ip_output(m, NULL, NULL, 0, NULL, NULL); 1729 } 1730 m_freem(args->m); 1731 } else 1732 m_freem(args->m); 1733 args->m = NULL; 1734} 1735 1736/** 1737 * 1738 * Given an ip_fw *, lookup_next_rule will return a pointer 1739 * to the next rule, which can be either the jump 1740 * target (for skipto instructions) or the next one in the list (in 1741 * all other cases including a missing jump target). 1742 * The result is also written in the "next_rule" field of the rule. 1743 * Backward jumps are not allowed, so start looking from the next 1744 * rule... 1745 * 1746 * This never returns NULL -- in case we do not have an exact match, 1747 * the next rule is returned. When the ruleset is changed, 1748 * pointers are flushed so we are always correct. 1749 */ 1750 1751static struct ip_fw * 1752lookup_next_rule(struct ip_fw *me) 1753{ 1754 struct ip_fw *rule = NULL; 1755 ipfw_insn *cmd; 1756 1757 /* look for action, in case it is a skipto */ 1758 cmd = ACTION_PTR(me); 1759 if (cmd->opcode == O_LOG) 1760 cmd += F_LEN(cmd); 1761 if (cmd->opcode == O_ALTQ) 1762 cmd += F_LEN(cmd); 1763 if (cmd->opcode == O_TAG) 1764 cmd += F_LEN(cmd); 1765 if ( cmd->opcode == O_SKIPTO ) 1766 for (rule = me->next; rule ; rule = rule->next) 1767 if (rule->rulenum >= cmd->arg1) 1768 break; 1769 if (rule == NULL) /* failure or not a skipto */ 1770 rule = me->next; 1771 me->next_rule = rule; 1772 return rule; 1773} 1774 1775static int 1776add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1777 uint8_t mlen, uint32_t value) 1778{ 1779 struct radix_node_head *rnh; 1780 struct table_entry *ent; 1781 1782 if (tbl >= IPFW_TABLES_MAX) 1783 return (EINVAL); 1784 rnh = ch->tables[tbl]; 1785 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1786 if (ent == NULL) 1787 return (ENOMEM); 1788 ent->value = value; 1789 ent->addr.sin_len = ent->mask.sin_len = 8; 1790 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1791 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1792 IPFW_WLOCK(&layer3_chain); 1793 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1794 NULL) { 1795 IPFW_WUNLOCK(&layer3_chain); 1796 free(ent, M_IPFW_TBL); 1797 return (EEXIST); 1798 } 1799 IPFW_WUNLOCK(&layer3_chain); 1800 return (0); 1801} 1802 1803static int 1804del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1805 uint8_t mlen) 1806{ 1807 struct radix_node_head *rnh; 1808 struct table_entry *ent; 1809 struct sockaddr_in sa, mask; 1810 1811 if (tbl >= IPFW_TABLES_MAX) 1812 return (EINVAL); 1813 rnh = ch->tables[tbl]; 1814 sa.sin_len = mask.sin_len = 8; 1815 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1816 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1817 IPFW_WLOCK(ch); 1818 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1819 if (ent == NULL) { 1820 IPFW_WUNLOCK(ch); 1821 return (ESRCH); 1822 } 1823 IPFW_WUNLOCK(ch); 1824 free(ent, M_IPFW_TBL); 1825 return (0); 1826} 1827 1828static int 1829flush_table_entry(struct radix_node *rn, void *arg) 1830{ 1831 struct radix_node_head * const rnh = arg; 1832 struct table_entry *ent; 1833 1834 ent = (struct table_entry *) 1835 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1836 if (ent != NULL) 1837 free(ent, M_IPFW_TBL); 1838 return (0); 1839} 1840 1841static int 1842flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1843{ 1844 struct radix_node_head *rnh; 1845 1846 IPFW_WLOCK_ASSERT(ch); 1847 1848 if (tbl >= IPFW_TABLES_MAX) 1849 return (EINVAL); 1850 rnh = ch->tables[tbl]; 1851 KASSERT(rnh != NULL, ("NULL IPFW table")); 1852 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1853 return (0); 1854} 1855 1856static void 1857flush_tables(struct ip_fw_chain *ch) 1858{ 1859 uint16_t tbl; 1860 1861 IPFW_WLOCK_ASSERT(ch); 1862 1863 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1864 flush_table(ch, tbl); 1865} 1866 1867static int 1868init_tables(struct ip_fw_chain *ch) 1869{ 1870 int i; 1871 uint16_t j; 1872 1873 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1874 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1875 for (j = 0; j < i; j++) { 1876 (void) flush_table(ch, j); 1877 } 1878 return (ENOMEM); 1879 } 1880 } 1881 return (0); 1882} 1883 1884static int 1885lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1886 uint32_t *val) 1887{ 1888 struct radix_node_head *rnh; 1889 struct table_entry *ent; 1890 struct sockaddr_in sa; 1891 1892 if (tbl >= IPFW_TABLES_MAX) 1893 return (0); 1894 rnh = ch->tables[tbl]; 1895 sa.sin_len = 8; 1896 sa.sin_addr.s_addr = addr; 1897 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1898 if (ent != NULL) { 1899 *val = ent->value; 1900 return (1); 1901 } 1902 return (0); 1903} 1904 1905static int 1906count_table_entry(struct radix_node *rn, void *arg) 1907{ 1908 u_int32_t * const cnt = arg; 1909 1910 (*cnt)++; 1911 return (0); 1912} 1913 1914static int 1915count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1916{ 1917 struct radix_node_head *rnh; 1918 1919 if (tbl >= IPFW_TABLES_MAX) 1920 return (EINVAL); 1921 rnh = ch->tables[tbl]; 1922 *cnt = 0; 1923 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1924 return (0); 1925} 1926 1927static int 1928dump_table_entry(struct radix_node *rn, void *arg) 1929{ 1930 struct table_entry * const n = (struct table_entry *)rn; 1931 ipfw_table * const tbl = arg; 1932 ipfw_table_entry *ent; 1933 1934 if (tbl->cnt == tbl->size) 1935 return (1); 1936 ent = &tbl->ent[tbl->cnt]; 1937 ent->tbl = tbl->tbl; 1938 if (in_nullhost(n->mask.sin_addr)) 1939 ent->masklen = 0; 1940 else 1941 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1942 ent->addr = n->addr.sin_addr.s_addr; 1943 ent->value = n->value; 1944 tbl->cnt++; 1945 return (0); 1946} 1947 1948static int 1949dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1950{ 1951 struct radix_node_head *rnh; 1952 1953 if (tbl->tbl >= IPFW_TABLES_MAX) 1954 return (EINVAL); 1955 rnh = ch->tables[tbl->tbl]; 1956 tbl->cnt = 0; 1957 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1958 return (0); 1959} 1960 1961static void 1962fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1963{ 1964 struct ucred *cr; 1965 1966 if (inp->inp_socket != NULL) { 1967 cr = inp->inp_socket->so_cred; 1968 ugp->fw_prid = jailed(cr) ? 1969 cr->cr_prison->pr_id : -1; 1970 ugp->fw_uid = cr->cr_uid; 1971 ugp->fw_ngroups = cr->cr_ngroups; 1972 bcopy(cr->cr_groups, ugp->fw_groups, 1973 sizeof(ugp->fw_groups)); 1974 } 1975} 1976 1977static int 1978check_uidgid(ipfw_insn_u32 *insn, 1979 int proto, struct ifnet *oif, 1980 struct in_addr dst_ip, u_int16_t dst_port, 1981 struct in_addr src_ip, u_int16_t src_port, 1982 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1983{ 1984 struct inpcbinfo *pi; 1985 int wildcard; 1986 struct inpcb *pcb; 1987 int match; 1988 gid_t *gp; 1989 1990 /* 1991 * Check to see if the UDP or TCP stack supplied us with 1992 * the PCB. If so, rather then holding a lock and looking 1993 * up the PCB, we can use the one that was supplied. 1994 */ 1995 if (inp && *lookup == 0) { 1996 INP_LOCK_ASSERT(inp); 1997 if (inp->inp_socket != NULL) { 1998 fill_ugid_cache(inp, ugp); 1999 *lookup = 1; 2000 } 2001 } 2002 /* 2003 * If we have already been here and the packet has no 2004 * PCB entry associated with it, then we can safely 2005 * assume that this is a no match. 2006 */ 2007 if (*lookup == -1) 2008 return (0); 2009 if (proto == IPPROTO_TCP) { 2010 wildcard = 0; 2011 pi = &tcbinfo; 2012 } else if (proto == IPPROTO_UDP) { 2013 wildcard = INPLOOKUP_WILDCARD; 2014 pi = &udbinfo; 2015 } else 2016 return 0; 2017 match = 0; 2018 if (*lookup == 0) { 2019 INP_INFO_RLOCK(pi); 2020 pcb = (oif) ? 2021 in_pcblookup_hash(pi, 2022 dst_ip, htons(dst_port), 2023 src_ip, htons(src_port), 2024 wildcard, oif) : 2025 in_pcblookup_hash(pi, 2026 src_ip, htons(src_port), 2027 dst_ip, htons(dst_port), 2028 wildcard, NULL); 2029 if (pcb != NULL) { 2030 INP_LOCK(pcb); 2031 if (pcb->inp_socket != NULL) { 2032 fill_ugid_cache(pcb, ugp); 2033 *lookup = 1; 2034 } 2035 INP_UNLOCK(pcb); 2036 } 2037 INP_INFO_RUNLOCK(pi); 2038 if (*lookup == 0) { 2039 /* 2040 * If the lookup did not yield any results, there 2041 * is no sense in coming back and trying again. So 2042 * we can set lookup to -1 and ensure that we wont 2043 * bother the pcb system again. 2044 */ 2045 *lookup = -1; 2046 return (0); 2047 } 2048 } 2049 if (insn->o.opcode == O_UID) 2050 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2051 else if (insn->o.opcode == O_GID) { 2052 for (gp = ugp->fw_groups; 2053 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2054 if (*gp == (gid_t)insn->d[0]) { 2055 match = 1; 2056 break; 2057 } 2058 } else if (insn->o.opcode == O_JAIL) 2059 match = (ugp->fw_prid == (int)insn->d[0]); 2060 return match; 2061} 2062 2063static eventhandler_tag ifaddr_event_tag; 2064 2065static void 2066ifaddr_change(void *arg __unused, struct ifnet *ifp) { 2067 struct cfg_nat *ptr; 2068 struct ifaddr *ifa; 2069 2070 IPFW_WLOCK(&layer3_chain); 2071 /* Check every nat entry... */ 2072 LIST_FOREACH(ptr, &layer3_chain.nat, _next) { 2073 /* ...using nic 'ifp->if_xname' as dynamic alias address. */ 2074 if (strncmp(ptr->if_name, ifp->if_xname, IF_NAMESIZE) == 0) { 2075 mtx_lock(&ifp->if_addr_mtx); 2076 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) { 2077 if (ifa->ifa_addr == NULL) 2078 continue; 2079 if (ifa->ifa_addr->sa_family != AF_INET) 2080 continue; 2081 ptr->ip = ((struct sockaddr_in *) 2082 (ifa->ifa_addr))->sin_addr; 2083 LibAliasSetAddress(ptr->lib, ptr->ip); 2084 } 2085 mtx_unlock(&ifp->if_addr_mtx); 2086 } 2087 } 2088 IPFW_WUNLOCK(&layer3_chain); 2089} 2090 2091static void 2092flush_nat_ptrs(const int i) { 2093 struct ip_fw *rule; 2094 2095 IPFW_WLOCK_ASSERT(&layer3_chain); 2096 for (rule = layer3_chain.rules; rule; rule = rule->next) { 2097 ipfw_insn_nat *cmd = (ipfw_insn_nat *)ACTION_PTR(rule); 2098 if (cmd->o.opcode != O_NAT) 2099 continue; 2100 if (cmd->nat != NULL && cmd->nat->id == i) 2101 cmd->nat = NULL; 2102 } 2103} 2104 2105static struct cfg_nat * 2106lookup_nat(const int i) { 2107 struct cfg_nat *ptr; 2108 2109 LIST_FOREACH(ptr, &layer3_chain.nat, _next) 2110 if (ptr->id == i) 2111 return(ptr); 2112 return (NULL); 2113} 2114 2115#define HOOK_NAT(b, p) do { \ 2116 IPFW_WLOCK_ASSERT(&layer3_chain); \ 2117 LIST_INSERT_HEAD(b, p, _next); \ 2118} while (0) 2119 2120#define UNHOOK_NAT(p) do { \ 2121 IPFW_WLOCK_ASSERT(&layer3_chain); \ 2122 LIST_REMOVE(p, _next); \ 2123} while (0) 2124 2125#define HOOK_REDIR(b, p) do { \ 2126 LIST_INSERT_HEAD(b, p, _next); \ 2127} while (0) 2128 2129#define HOOK_SPOOL(b, p) do { \ 2130 LIST_INSERT_HEAD(b, p, _next); \ 2131} while (0) 2132 2133static void 2134del_redir_spool_cfg(struct cfg_nat *n, struct redir_chain *head) { 2135 struct cfg_redir *r, *tmp_r; 2136 struct cfg_spool *s, *tmp_s; 2137 int i, num; 2138 2139 LIST_FOREACH_SAFE(r, head, _next, tmp_r) { 2140 num = 1; /* Number of alias_link to delete. */ 2141 switch (r->mode) { 2142 case REDIR_PORT: 2143 num = r->pport_cnt; 2144 /* FALLTHROUGH */ 2145 case REDIR_ADDR: 2146 case REDIR_PROTO: 2147 /* Delete all libalias redirect entry. */ 2148 for (i = 0; i < num; i++) 2149 LibAliasRedirectDelete(n->lib, r->alink[i]); 2150 /* Del spool cfg if any. */ 2151 LIST_FOREACH_SAFE(s, &r->spool_chain, _next, tmp_s) { 2152 LIST_REMOVE(s, _next); 2153 free(s, M_IPFW); 2154 } 2155 free(r->alink, M_IPFW); 2156 LIST_REMOVE(r, _next); 2157 free(r, M_IPFW); 2158 break; 2159 default: 2160 printf("unknown redirect mode: %u\n", r->mode); 2161 /* XXX - panic?!?!? */ 2162 break; 2163 } 2164 } 2165} 2166 2167static int 2168add_redir_spool_cfg(char *buf, struct cfg_nat *ptr) { 2169 struct cfg_redir *r, *ser_r; 2170 struct cfg_spool *s, *ser_s; 2171 int cnt, off, i; 2172 char *panic_err; 2173 2174 for (cnt = 0, off = 0; cnt < ptr->redir_cnt; cnt++) { 2175 ser_r = (struct cfg_redir *)&buf[off]; 2176 r = malloc(SOF_REDIR, M_IPFW, M_WAITOK | M_ZERO); 2177 memcpy(r, ser_r, SOF_REDIR); 2178 LIST_INIT(&r->spool_chain); 2179 off += SOF_REDIR; 2180 r->alink = malloc(sizeof(struct alias_link *) * r->pport_cnt, 2181 M_IPFW, M_WAITOK | M_ZERO); 2182 switch (r->mode) { 2183 case REDIR_ADDR: 2184 r->alink[0] = LibAliasRedirectAddr(ptr->lib, r->laddr, 2185 r->paddr); 2186 break; 2187 case REDIR_PORT: 2188 for (i = 0 ; i < r->pport_cnt; i++) { 2189 /* If remotePort is all ports, set it to 0. */ 2190 u_short remotePortCopy = r->rport + i; 2191 if (r->rport_cnt == 1 && r->rport == 0) 2192 remotePortCopy = 0; 2193 r->alink[i] = LibAliasRedirectPort(ptr->lib, 2194 r->laddr, htons(r->lport + i), r->raddr, 2195 htons(remotePortCopy), r->paddr, 2196 htons(r->pport + i), r->proto); 2197 if (r->alink[i] == NULL) { 2198 r->alink[0] = NULL; 2199 break; 2200 } 2201 } 2202 break; 2203 case REDIR_PROTO: 2204 r->alink[0] = LibAliasRedirectProto(ptr->lib ,r->laddr, 2205 r->raddr, r->paddr, r->proto); 2206 break; 2207 default: 2208 printf("unknown redirect mode: %u\n", r->mode); 2209 break; 2210 } 2211 if (r->alink[0] == NULL) { 2212 panic_err = "LibAliasRedirect* returned NULL"; 2213 goto bad; 2214 } else /* LSNAT handling. */ 2215 for (i = 0; i < r->spool_cnt; i++) { 2216 ser_s = (struct cfg_spool *)&buf[off]; 2217 s = malloc(SOF_REDIR, M_IPFW, 2218 M_WAITOK | M_ZERO); 2219 memcpy(s, ser_s, SOF_SPOOL); 2220 LibAliasAddServer(ptr->lib, r->alink[0], 2221 s->addr, htons(s->port)); 2222 off += SOF_SPOOL; 2223 /* Hook spool entry. */ 2224 HOOK_SPOOL(&r->spool_chain, s); 2225 } 2226 /* And finally hook this redir entry. */ 2227 HOOK_REDIR(&ptr->redir_chain, r); 2228 } 2229 return (1); 2230bad: 2231 /* something really bad happened: panic! */ 2232 panic("%s\n", panic_err); 2233} 2234 2235/* 2236 * The main check routine for the firewall. 2237 * 2238 * All arguments are in args so we can modify them and return them 2239 * back to the caller. 2240 * 2241 * Parameters: 2242 * 2243 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2244 * Starts with the IP header. 2245 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2246 * args->L3offset Number of bytes bypassed if we came from L2. 2247 * e.g. often sizeof(eh) ** NOTYET ** 2248 * args->oif Outgoing interface, or NULL if packet is incoming. 2249 * The incoming interface is in the mbuf. (in) 2250 * args->divert_rule (in/out) 2251 * Skip up to the first rule past this rule number; 2252 * upon return, non-zero port number for divert or tee. 2253 * 2254 * args->rule Pointer to the last matching rule (in/out) 2255 * args->next_hop Socket we are forwarding to (out). 2256 * args->f_id Addresses grabbed from the packet (out) 2257 * args->cookie a cookie depending on rule action 2258 * 2259 * Return value: 2260 * 2261 * IP_FW_PASS the packet must be accepted 2262 * IP_FW_DENY the packet must be dropped 2263 * IP_FW_DIVERT divert packet, port in m_tag 2264 * IP_FW_TEE tee packet, port in m_tag 2265 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2266 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2267 * 2268 */ 2269int 2270ipfw_chk(struct ip_fw_args *args) 2271{ 2272 /* 2273 * Local variables holding state during the processing of a packet: 2274 * 2275 * IMPORTANT NOTE: to speed up the processing of rules, there 2276 * are some assumption on the values of the variables, which 2277 * are documented here. Should you change them, please check 2278 * the implementation of the various instructions to make sure 2279 * that they still work. 2280 * 2281 * args->eh The MAC header. It is non-null for a layer2 2282 * packet, it is NULL for a layer-3 packet. 2283 * **notyet** 2284 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2285 * 2286 * m | args->m Pointer to the mbuf, as received from the caller. 2287 * It may change if ipfw_chk() does an m_pullup, or if it 2288 * consumes the packet because it calls send_reject(). 2289 * XXX This has to change, so that ipfw_chk() never modifies 2290 * or consumes the buffer. 2291 * ip is the beginning of the ip(4 or 6) header. 2292 * Calculated by adding the L3offset to the start of data. 2293 * (Until we start using L3offset, the packet is 2294 * supposed to start with the ip header). 2295 */ 2296 struct mbuf *m = args->m; 2297 struct ip *ip = mtod(m, struct ip *); 2298 2299 /* 2300 * For rules which contain uid/gid or jail constraints, cache 2301 * a copy of the users credentials after the pcb lookup has been 2302 * executed. This will speed up the processing of rules with 2303 * these types of constraints, as well as decrease contention 2304 * on pcb related locks. 2305 */ 2306 struct ip_fw_ugid fw_ugid_cache; 2307 int ugid_lookup = 0; 2308 2309 /* 2310 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2311 * associated with a packet input on a divert socket. This 2312 * will allow to distinguish traffic and its direction when 2313 * it originates from a divert socket. 2314 */ 2315 u_int divinput_flags = 0; 2316 2317 /* 2318 * oif | args->oif If NULL, ipfw_chk has been called on the 2319 * inbound path (ether_input, ip_input). 2320 * If non-NULL, ipfw_chk has been called on the outbound path 2321 * (ether_output, ip_output). 2322 */ 2323 struct ifnet *oif = args->oif; 2324 2325 struct ip_fw *f = NULL; /* matching rule */ 2326 int retval = 0; 2327 2328 /* 2329 * hlen The length of the IP header. 2330 */ 2331 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2332 2333 /* 2334 * offset The offset of a fragment. offset != 0 means that 2335 * we have a fragment at this offset of an IPv4 packet. 2336 * offset == 0 means that (if this is an IPv4 packet) 2337 * this is the first or only fragment. 2338 * For IPv6 offset == 0 means there is no Fragment Header. 2339 * If offset != 0 for IPv6 always use correct mask to 2340 * get the correct offset because we add IP6F_MORE_FRAG 2341 * to be able to dectect the first fragment which would 2342 * otherwise have offset = 0. 2343 */ 2344 u_short offset = 0; 2345 2346 /* 2347 * Local copies of addresses. They are only valid if we have 2348 * an IP packet. 2349 * 2350 * proto The protocol. Set to 0 for non-ip packets, 2351 * or to the protocol read from the packet otherwise. 2352 * proto != 0 means that we have an IPv4 packet. 2353 * 2354 * src_port, dst_port port numbers, in HOST format. Only 2355 * valid for TCP and UDP packets. 2356 * 2357 * src_ip, dst_ip ip addresses, in NETWORK format. 2358 * Only valid for IPv4 packets. 2359 */ 2360 u_int8_t proto; 2361 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2362 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2363 u_int16_t ip_len=0; 2364 int pktlen; 2365 u_int16_t etype = 0; /* Host order stored ether type */ 2366 2367 /* 2368 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2369 * MATCH_NONE when checked and not matched (q = NULL), 2370 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2371 */ 2372 int dyn_dir = MATCH_UNKNOWN; 2373 ipfw_dyn_rule *q = NULL; 2374 struct ip_fw_chain *chain = &layer3_chain; 2375 struct m_tag *mtag; 2376 2377 /* 2378 * We store in ulp a pointer to the upper layer protocol header. 2379 * In the ipv4 case this is easy to determine from the header, 2380 * but for ipv6 we might have some additional headers in the middle. 2381 * ulp is NULL if not found. 2382 */ 2383 void *ulp = NULL; /* upper layer protocol pointer. */ 2384 /* XXX ipv6 variables */ 2385 int is_ipv6 = 0; 2386 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2387 /* end of ipv6 variables */ 2388 int is_ipv4 = 0; 2389 2390 if (m->m_flags & M_SKIP_FIREWALL) 2391 return (IP_FW_PASS); /* accept */ 2392 2393 pktlen = m->m_pkthdr.len; 2394 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2395 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2396 2397/* 2398 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2399 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2400 * pointer might become stale after other pullups (but we never use it 2401 * this way). 2402 */ 2403#define PULLUP_TO(len, p, T) \ 2404do { \ 2405 int x = (len) + sizeof(T); \ 2406 if ((m)->m_len < x) { \ 2407 args->m = m = m_pullup(m, x); \ 2408 if (m == NULL) \ 2409 goto pullup_failed; \ 2410 } \ 2411 p = (mtod(m, char *) + (len)); \ 2412} while (0) 2413 2414 /* 2415 * if we have an ether header, 2416 */ 2417 if (args->eh) 2418 etype = ntohs(args->eh->ether_type); 2419 2420 /* Identify IP packets and fill up variables. */ 2421 if (pktlen >= sizeof(struct ip6_hdr) && 2422 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2423 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2424 is_ipv6 = 1; 2425 args->f_id.addr_type = 6; 2426 hlen = sizeof(struct ip6_hdr); 2427 proto = ip6->ip6_nxt; 2428 2429 /* Search extension headers to find upper layer protocols */ 2430 while (ulp == NULL) { 2431 switch (proto) { 2432 case IPPROTO_ICMPV6: 2433 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2434 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2435 break; 2436 2437 case IPPROTO_TCP: 2438 PULLUP_TO(hlen, ulp, struct tcphdr); 2439 dst_port = TCP(ulp)->th_dport; 2440 src_port = TCP(ulp)->th_sport; 2441 args->f_id.flags = TCP(ulp)->th_flags; 2442 break; 2443 2444 case IPPROTO_SCTP: 2445 PULLUP_TO(hlen, ulp, struct sctphdr); 2446 src_port = SCTP(ulp)->src_port; 2447 dst_port = SCTP(ulp)->dest_port; 2448 break; 2449 2450 case IPPROTO_UDP: 2451 PULLUP_TO(hlen, ulp, struct udphdr); 2452 dst_port = UDP(ulp)->uh_dport; 2453 src_port = UDP(ulp)->uh_sport; 2454 break; 2455 2456 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2457 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2458 ext_hd |= EXT_HOPOPTS; 2459 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2460 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2461 ulp = NULL; 2462 break; 2463 2464 case IPPROTO_ROUTING: /* RFC 2460 */ 2465 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2466 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2467 case 0: 2468 break; 2469 default: 2470 printf("IPFW2: IPV6 - Unknown Routing " 2471 "Header type(%d)\n", 2472 ((struct ip6_rthdr *)ulp)->ip6r_type); 2473 if (fw_deny_unknown_exthdrs) 2474 return (IP_FW_DENY); 2475 break; 2476 } 2477 ext_hd |= EXT_ROUTING; 2478 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2479 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2480 ulp = NULL; 2481 break; 2482 2483 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2484 PULLUP_TO(hlen, ulp, struct ip6_frag); 2485 ext_hd |= EXT_FRAGMENT; 2486 hlen += sizeof (struct ip6_frag); 2487 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2488 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2489 IP6F_OFF_MASK; 2490 /* Add IP6F_MORE_FRAG for offset of first 2491 * fragment to be != 0. */ 2492 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2493 IP6F_MORE_FRAG; 2494 if (offset == 0) { 2495 printf("IPFW2: IPV6 - Invalid Fragment " 2496 "Header\n"); 2497 if (fw_deny_unknown_exthdrs) 2498 return (IP_FW_DENY); 2499 break; 2500 } 2501 args->f_id.frag_id6 = 2502 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2503 ulp = NULL; 2504 break; 2505 2506 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2507 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2508 ext_hd |= EXT_DSTOPTS; 2509 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2510 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2511 ulp = NULL; 2512 break; 2513 2514 case IPPROTO_AH: /* RFC 2402 */ 2515 PULLUP_TO(hlen, ulp, struct ip6_ext); 2516 ext_hd |= EXT_AH; 2517 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2518 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2519 ulp = NULL; 2520 break; 2521 2522 case IPPROTO_ESP: /* RFC 2406 */ 2523 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2524 /* Anything past Seq# is variable length and 2525 * data past this ext. header is encrypted. */ 2526 ext_hd |= EXT_ESP; 2527 break; 2528 2529 case IPPROTO_NONE: /* RFC 2460 */ 2530 PULLUP_TO(hlen, ulp, struct ip6_ext); 2531 /* Packet ends here. if ip6e_len!=0 octets 2532 * must be ignored. */ 2533 break; 2534 2535 case IPPROTO_OSPFIGP: 2536 /* XXX OSPF header check? */ 2537 PULLUP_TO(hlen, ulp, struct ip6_ext); 2538 break; 2539 2540 case IPPROTO_PIM: 2541 /* XXX PIM header check? */ 2542 PULLUP_TO(hlen, ulp, struct pim); 2543 break; 2544 2545 case IPPROTO_CARP: 2546 PULLUP_TO(hlen, ulp, struct carp_header); 2547 if (((struct carp_header *)ulp)->carp_version != 2548 CARP_VERSION) 2549 return (IP_FW_DENY); 2550 if (((struct carp_header *)ulp)->carp_type != 2551 CARP_ADVERTISEMENT) 2552 return (IP_FW_DENY); 2553 break; 2554 2555 case IPPROTO_IPV6: /* RFC 2893 */ 2556 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2557 break; 2558 2559 case IPPROTO_IPV4: /* RFC 2893 */ 2560 PULLUP_TO(hlen, ulp, struct ip); 2561 break; 2562 2563 default: 2564 printf("IPFW2: IPV6 - Unknown Extension " 2565 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2566 if (fw_deny_unknown_exthdrs) 2567 return (IP_FW_DENY); 2568 PULLUP_TO(hlen, ulp, struct ip6_ext); 2569 break; 2570 } /*switch */ 2571 } 2572 ip = mtod(m, struct ip *); 2573 ip6 = (struct ip6_hdr *)ip; 2574 args->f_id.src_ip6 = ip6->ip6_src; 2575 args->f_id.dst_ip6 = ip6->ip6_dst; 2576 args->f_id.src_ip = 0; 2577 args->f_id.dst_ip = 0; 2578 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2579 } else if (pktlen >= sizeof(struct ip) && 2580 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2581 is_ipv4 = 1; 2582 hlen = ip->ip_hl << 2; 2583 args->f_id.addr_type = 4; 2584 2585 /* 2586 * Collect parameters into local variables for faster matching. 2587 */ 2588 proto = ip->ip_p; 2589 src_ip = ip->ip_src; 2590 dst_ip = ip->ip_dst; 2591 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2592 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2593 ip_len = ntohs(ip->ip_len); 2594 } else { 2595 offset = ip->ip_off & IP_OFFMASK; 2596 ip_len = ip->ip_len; 2597 } 2598 pktlen = ip_len < pktlen ? ip_len : pktlen; 2599 2600 if (offset == 0) { 2601 switch (proto) { 2602 case IPPROTO_TCP: 2603 PULLUP_TO(hlen, ulp, struct tcphdr); 2604 dst_port = TCP(ulp)->th_dport; 2605 src_port = TCP(ulp)->th_sport; 2606 args->f_id.flags = TCP(ulp)->th_flags; 2607 break; 2608 2609 case IPPROTO_UDP: 2610 PULLUP_TO(hlen, ulp, struct udphdr); 2611 dst_port = UDP(ulp)->uh_dport; 2612 src_port = UDP(ulp)->uh_sport; 2613 break; 2614 2615 case IPPROTO_ICMP: 2616 PULLUP_TO(hlen, ulp, struct icmphdr); 2617 args->f_id.flags = ICMP(ulp)->icmp_type; 2618 break; 2619 2620 default: 2621 break; 2622 } 2623 } 2624 2625 ip = mtod(m, struct ip *); 2626 args->f_id.src_ip = ntohl(src_ip.s_addr); 2627 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2628 } 2629#undef PULLUP_TO 2630 if (proto) { /* we may have port numbers, store them */ 2631 args->f_id.proto = proto; 2632 args->f_id.src_port = src_port = ntohs(src_port); 2633 args->f_id.dst_port = dst_port = ntohs(dst_port); 2634 } 2635 2636 IPFW_RLOCK(chain); 2637 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2638 if (args->rule) { 2639 /* 2640 * Packet has already been tagged. Look for the next rule 2641 * to restart processing. 2642 * 2643 * If fw_one_pass != 0 then just accept it. 2644 * XXX should not happen here, but optimized out in 2645 * the caller. 2646 */ 2647 if (fw_one_pass) { 2648 IPFW_RUNLOCK(chain); 2649 return (IP_FW_PASS); 2650 } 2651 2652 f = args->rule->next_rule; 2653 if (f == NULL) 2654 f = lookup_next_rule(args->rule); 2655 } else { 2656 /* 2657 * Find the starting rule. It can be either the first 2658 * one, or the one after divert_rule if asked so. 2659 */ 2660 int skipto = mtag ? divert_cookie(mtag) : 0; 2661 2662 f = chain->rules; 2663 if (args->eh == NULL && skipto != 0) { 2664 if (skipto >= IPFW_DEFAULT_RULE) { 2665 IPFW_RUNLOCK(chain); 2666 return (IP_FW_DENY); /* invalid */ 2667 } 2668 while (f && f->rulenum <= skipto) 2669 f = f->next; 2670 if (f == NULL) { /* drop packet */ 2671 IPFW_RUNLOCK(chain); 2672 return (IP_FW_DENY); 2673 } 2674 } 2675 } 2676 /* reset divert rule to avoid confusion later */ 2677 if (mtag) { 2678 divinput_flags = divert_info(mtag) & 2679 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2680 m_tag_delete(m, mtag); 2681 } 2682 2683 /* 2684 * Now scan the rules, and parse microinstructions for each rule. 2685 */ 2686 for (; f; f = f->next) { 2687 ipfw_insn *cmd; 2688 uint32_t tablearg = 0; 2689 int l, cmdlen, skip_or; /* skip rest of OR block */ 2690 2691again: 2692 if (set_disable & (1 << f->set) ) 2693 continue; 2694 2695 skip_or = 0; 2696 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2697 l -= cmdlen, cmd += cmdlen) { 2698 int match; 2699 2700 /* 2701 * check_body is a jump target used when we find a 2702 * CHECK_STATE, and need to jump to the body of 2703 * the target rule. 2704 */ 2705 2706check_body: 2707 cmdlen = F_LEN(cmd); 2708 /* 2709 * An OR block (insn_1 || .. || insn_n) has the 2710 * F_OR bit set in all but the last instruction. 2711 * The first match will set "skip_or", and cause 2712 * the following instructions to be skipped until 2713 * past the one with the F_OR bit clear. 2714 */ 2715 if (skip_or) { /* skip this instruction */ 2716 if ((cmd->len & F_OR) == 0) 2717 skip_or = 0; /* next one is good */ 2718 continue; 2719 } 2720 match = 0; /* set to 1 if we succeed */ 2721 2722 switch (cmd->opcode) { 2723 /* 2724 * The first set of opcodes compares the packet's 2725 * fields with some pattern, setting 'match' if a 2726 * match is found. At the end of the loop there is 2727 * logic to deal with F_NOT and F_OR flags associated 2728 * with the opcode. 2729 */ 2730 case O_NOP: 2731 match = 1; 2732 break; 2733 2734 case O_FORWARD_MAC: 2735 printf("ipfw: opcode %d unimplemented\n", 2736 cmd->opcode); 2737 break; 2738 2739 case O_GID: 2740 case O_UID: 2741 case O_JAIL: 2742 /* 2743 * We only check offset == 0 && proto != 0, 2744 * as this ensures that we have a 2745 * packet with the ports info. 2746 */ 2747 if (offset!=0) 2748 break; 2749 if (is_ipv6) /* XXX to be fixed later */ 2750 break; 2751 if (proto == IPPROTO_TCP || 2752 proto == IPPROTO_UDP) 2753 match = check_uidgid( 2754 (ipfw_insn_u32 *)cmd, 2755 proto, oif, 2756 dst_ip, dst_port, 2757 src_ip, src_port, &fw_ugid_cache, 2758 &ugid_lookup, args->inp); 2759 break; 2760 2761 case O_RECV: 2762 match = iface_match(m->m_pkthdr.rcvif, 2763 (ipfw_insn_if *)cmd); 2764 break; 2765 2766 case O_XMIT: 2767 match = iface_match(oif, (ipfw_insn_if *)cmd); 2768 break; 2769 2770 case O_VIA: 2771 match = iface_match(oif ? oif : 2772 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2773 break; 2774 2775 case O_MACADDR2: 2776 if (args->eh != NULL) { /* have MAC header */ 2777 u_int32_t *want = (u_int32_t *) 2778 ((ipfw_insn_mac *)cmd)->addr; 2779 u_int32_t *mask = (u_int32_t *) 2780 ((ipfw_insn_mac *)cmd)->mask; 2781 u_int32_t *hdr = (u_int32_t *)args->eh; 2782 2783 match = 2784 ( want[0] == (hdr[0] & mask[0]) && 2785 want[1] == (hdr[1] & mask[1]) && 2786 want[2] == (hdr[2] & mask[2]) ); 2787 } 2788 break; 2789 2790 case O_MAC_TYPE: 2791 if (args->eh != NULL) { 2792 u_int16_t *p = 2793 ((ipfw_insn_u16 *)cmd)->ports; 2794 int i; 2795 2796 for (i = cmdlen - 1; !match && i>0; 2797 i--, p += 2) 2798 match = (etype >= p[0] && 2799 etype <= p[1]); 2800 } 2801 break; 2802 2803 case O_FRAG: 2804 match = (offset != 0); 2805 break; 2806 2807 case O_IN: /* "out" is "not in" */ 2808 match = (oif == NULL); 2809 break; 2810 2811 case O_LAYER2: 2812 match = (args->eh != NULL); 2813 break; 2814 2815 case O_DIVERTED: 2816 match = (cmd->arg1 & 1 && divinput_flags & 2817 IP_FW_DIVERT_LOOPBACK_FLAG) || 2818 (cmd->arg1 & 2 && divinput_flags & 2819 IP_FW_DIVERT_OUTPUT_FLAG); 2820 break; 2821 2822 case O_PROTO: 2823 /* 2824 * We do not allow an arg of 0 so the 2825 * check of "proto" only suffices. 2826 */ 2827 match = (proto == cmd->arg1); 2828 break; 2829 2830 case O_IP_SRC: 2831 match = is_ipv4 && 2832 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2833 src_ip.s_addr); 2834 break; 2835 2836 case O_IP_SRC_LOOKUP: 2837 case O_IP_DST_LOOKUP: 2838 if (is_ipv4) { 2839 uint32_t a = 2840 (cmd->opcode == O_IP_DST_LOOKUP) ? 2841 dst_ip.s_addr : src_ip.s_addr; 2842 uint32_t v; 2843 2844 match = lookup_table(chain, cmd->arg1, a, 2845 &v); 2846 if (!match) 2847 break; 2848 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2849 match = 2850 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2851 else 2852 tablearg = v; 2853 } 2854 break; 2855 2856 case O_IP_SRC_MASK: 2857 case O_IP_DST_MASK: 2858 if (is_ipv4) { 2859 uint32_t a = 2860 (cmd->opcode == O_IP_DST_MASK) ? 2861 dst_ip.s_addr : src_ip.s_addr; 2862 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2863 int i = cmdlen-1; 2864 2865 for (; !match && i>0; i-= 2, p+= 2) 2866 match = (p[0] == (a & p[1])); 2867 } 2868 break; 2869 2870 case O_IP_SRC_ME: 2871 if (is_ipv4) { 2872 struct ifnet *tif; 2873 2874 INADDR_TO_IFP(src_ip, tif); 2875 match = (tif != NULL); 2876 } 2877 break; 2878 2879 case O_IP_DST_SET: 2880 case O_IP_SRC_SET: 2881 if (is_ipv4) { 2882 u_int32_t *d = (u_int32_t *)(cmd+1); 2883 u_int32_t addr = 2884 cmd->opcode == O_IP_DST_SET ? 2885 args->f_id.dst_ip : 2886 args->f_id.src_ip; 2887 2888 if (addr < d[0]) 2889 break; 2890 addr -= d[0]; /* subtract base */ 2891 match = (addr < cmd->arg1) && 2892 ( d[ 1 + (addr>>5)] & 2893 (1<<(addr & 0x1f)) ); 2894 } 2895 break; 2896 2897 case O_IP_DST: 2898 match = is_ipv4 && 2899 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2900 dst_ip.s_addr); 2901 break; 2902 2903 case O_IP_DST_ME: 2904 if (is_ipv4) { 2905 struct ifnet *tif; 2906 2907 INADDR_TO_IFP(dst_ip, tif); 2908 match = (tif != NULL); 2909 } 2910 break; 2911 2912 case O_IP_SRCPORT: 2913 case O_IP_DSTPORT: 2914 /* 2915 * offset == 0 && proto != 0 is enough 2916 * to guarantee that we have a 2917 * packet with port info. 2918 */ 2919 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2920 && offset == 0) { 2921 u_int16_t x = 2922 (cmd->opcode == O_IP_SRCPORT) ? 2923 src_port : dst_port ; 2924 u_int16_t *p = 2925 ((ipfw_insn_u16 *)cmd)->ports; 2926 int i; 2927 2928 for (i = cmdlen - 1; !match && i>0; 2929 i--, p += 2) 2930 match = (x>=p[0] && x<=p[1]); 2931 } 2932 break; 2933 2934 case O_ICMPTYPE: 2935 match = (offset == 0 && proto==IPPROTO_ICMP && 2936 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2937 break; 2938 2939#ifdef INET6 2940 case O_ICMP6TYPE: 2941 match = is_ipv6 && offset == 0 && 2942 proto==IPPROTO_ICMPV6 && 2943 icmp6type_match( 2944 ICMP6(ulp)->icmp6_type, 2945 (ipfw_insn_u32 *)cmd); 2946 break; 2947#endif /* INET6 */ 2948 2949 case O_IPOPT: 2950 match = (is_ipv4 && 2951 ipopts_match(ip, cmd) ); 2952 break; 2953 2954 case O_IPVER: 2955 match = (is_ipv4 && 2956 cmd->arg1 == ip->ip_v); 2957 break; 2958 2959 case O_IPID: 2960 case O_IPLEN: 2961 case O_IPTTL: 2962 if (is_ipv4) { /* only for IP packets */ 2963 uint16_t x; 2964 uint16_t *p; 2965 int i; 2966 2967 if (cmd->opcode == O_IPLEN) 2968 x = ip_len; 2969 else if (cmd->opcode == O_IPTTL) 2970 x = ip->ip_ttl; 2971 else /* must be IPID */ 2972 x = ntohs(ip->ip_id); 2973 if (cmdlen == 1) { 2974 match = (cmd->arg1 == x); 2975 break; 2976 } 2977 /* otherwise we have ranges */ 2978 p = ((ipfw_insn_u16 *)cmd)->ports; 2979 i = cmdlen - 1; 2980 for (; !match && i>0; i--, p += 2) 2981 match = (x >= p[0] && x <= p[1]); 2982 } 2983 break; 2984 2985 case O_IPPRECEDENCE: 2986 match = (is_ipv4 && 2987 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2988 break; 2989 2990 case O_IPTOS: 2991 match = (is_ipv4 && 2992 flags_match(cmd, ip->ip_tos)); 2993 break; 2994 2995 case O_TCPDATALEN: 2996 if (proto == IPPROTO_TCP && offset == 0) { 2997 struct tcphdr *tcp; 2998 uint16_t x; 2999 uint16_t *p; 3000 int i; 3001 3002 tcp = TCP(ulp); 3003 x = ip_len - 3004 ((ip->ip_hl + tcp->th_off) << 2); 3005 if (cmdlen == 1) { 3006 match = (cmd->arg1 == x); 3007 break; 3008 } 3009 /* otherwise we have ranges */ 3010 p = ((ipfw_insn_u16 *)cmd)->ports; 3011 i = cmdlen - 1; 3012 for (; !match && i>0; i--, p += 2) 3013 match = (x >= p[0] && x <= p[1]); 3014 } 3015 break; 3016 3017 case O_TCPFLAGS: 3018 match = (proto == IPPROTO_TCP && offset == 0 && 3019 flags_match(cmd, TCP(ulp)->th_flags)); 3020 break; 3021 3022 case O_TCPOPTS: 3023 match = (proto == IPPROTO_TCP && offset == 0 && 3024 tcpopts_match(TCP(ulp), cmd)); 3025 break; 3026 3027 case O_TCPSEQ: 3028 match = (proto == IPPROTO_TCP && offset == 0 && 3029 ((ipfw_insn_u32 *)cmd)->d[0] == 3030 TCP(ulp)->th_seq); 3031 break; 3032 3033 case O_TCPACK: 3034 match = (proto == IPPROTO_TCP && offset == 0 && 3035 ((ipfw_insn_u32 *)cmd)->d[0] == 3036 TCP(ulp)->th_ack); 3037 break; 3038 3039 case O_TCPWIN: 3040 match = (proto == IPPROTO_TCP && offset == 0 && 3041 cmd->arg1 == TCP(ulp)->th_win); 3042 break; 3043 3044 case O_ESTAB: 3045 /* reject packets which have SYN only */ 3046 /* XXX should i also check for TH_ACK ? */ 3047 match = (proto == IPPROTO_TCP && offset == 0 && 3048 (TCP(ulp)->th_flags & 3049 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 3050 break; 3051 3052 case O_ALTQ: { 3053 struct altq_tag *at; 3054 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 3055 3056 match = 1; 3057 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 3058 if (mtag != NULL) 3059 break; 3060 mtag = m_tag_get(PACKET_TAG_PF_QID, 3061 sizeof(struct altq_tag), 3062 M_NOWAIT); 3063 if (mtag == NULL) { 3064 /* 3065 * Let the packet fall back to the 3066 * default ALTQ. 3067 */ 3068 break; 3069 } 3070 at = (struct altq_tag *)(mtag+1); 3071 at->qid = altq->qid; 3072 if (is_ipv4) 3073 at->af = AF_INET; 3074 else 3075 at->af = AF_LINK; 3076 at->hdr = ip; 3077 m_tag_prepend(m, mtag); 3078 break; 3079 } 3080 3081 case O_LOG: 3082 if (fw_verbose) 3083 ipfw_log(f, hlen, args, m, 3084 oif, offset, tablearg, ip); 3085 match = 1; 3086 break; 3087 3088 case O_PROB: 3089 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 3090 break; 3091 3092 case O_VERREVPATH: 3093 /* Outgoing packets automatically pass/match */ 3094 match = ((oif != NULL) || 3095 (m->m_pkthdr.rcvif == NULL) || 3096 ( 3097#ifdef INET6 3098 is_ipv6 ? 3099 verify_path6(&(args->f_id.src_ip6), 3100 m->m_pkthdr.rcvif) : 3101#endif 3102 verify_path(src_ip, m->m_pkthdr.rcvif))); 3103 break; 3104 3105 case O_VERSRCREACH: 3106 /* Outgoing packets automatically pass/match */ 3107 match = (hlen > 0 && ((oif != NULL) || 3108#ifdef INET6 3109 is_ipv6 ? 3110 verify_path6(&(args->f_id.src_ip6), 3111 NULL) : 3112#endif 3113 verify_path(src_ip, NULL))); 3114 break; 3115 3116 case O_ANTISPOOF: 3117 /* Outgoing packets automatically pass/match */ 3118 if (oif == NULL && hlen > 0 && 3119 ( (is_ipv4 && in_localaddr(src_ip)) 3120#ifdef INET6 3121 || (is_ipv6 && 3122 in6_localaddr(&(args->f_id.src_ip6))) 3123#endif 3124 )) 3125 match = 3126#ifdef INET6 3127 is_ipv6 ? verify_path6( 3128 &(args->f_id.src_ip6), 3129 m->m_pkthdr.rcvif) : 3130#endif 3131 verify_path(src_ip, 3132 m->m_pkthdr.rcvif); 3133 else 3134 match = 1; 3135 break; 3136 3137 case O_IPSEC: 3138#ifdef FAST_IPSEC 3139 match = (m_tag_find(m, 3140 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 3141#endif 3142#ifdef IPSEC 3143 match = (ipsec_getnhist(m) != 0); 3144#endif 3145 /* otherwise no match */ 3146 break; 3147 3148#ifdef INET6 3149 case O_IP6_SRC: 3150 match = is_ipv6 && 3151 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 3152 &((ipfw_insn_ip6 *)cmd)->addr6); 3153 break; 3154 3155 case O_IP6_DST: 3156 match = is_ipv6 && 3157 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 3158 &((ipfw_insn_ip6 *)cmd)->addr6); 3159 break; 3160 case O_IP6_SRC_MASK: 3161 case O_IP6_DST_MASK: 3162 if (is_ipv6) { 3163 int i = cmdlen - 1; 3164 struct in6_addr p; 3165 struct in6_addr *d = 3166 &((ipfw_insn_ip6 *)cmd)->addr6; 3167 3168 for (; !match && i > 0; d += 2, 3169 i -= F_INSN_SIZE(struct in6_addr) 3170 * 2) { 3171 p = (cmd->opcode == 3172 O_IP6_SRC_MASK) ? 3173 args->f_id.src_ip6: 3174 args->f_id.dst_ip6; 3175 APPLY_MASK(&p, &d[1]); 3176 match = 3177 IN6_ARE_ADDR_EQUAL(&d[0], 3178 &p); 3179 } 3180 } 3181 break; 3182 3183 case O_IP6_SRC_ME: 3184 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3185 break; 3186 3187 case O_IP6_DST_ME: 3188 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3189 break; 3190 3191 case O_FLOW6ID: 3192 match = is_ipv6 && 3193 flow6id_match(args->f_id.flow_id6, 3194 (ipfw_insn_u32 *) cmd); 3195 break; 3196 3197 case O_EXT_HDR: 3198 match = is_ipv6 && 3199 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3200 break; 3201 3202 case O_IP6: 3203 match = is_ipv6; 3204 break; 3205#endif 3206 3207 case O_IP4: 3208 match = is_ipv4; 3209 break; 3210 3211 case O_TAG: { 3212 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3213 tablearg : cmd->arg1; 3214 3215 /* Packet is already tagged with this tag? */ 3216 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3217 3218 /* We have `untag' action when F_NOT flag is 3219 * present. And we must remove this mtag from 3220 * mbuf and reset `match' to zero (`match' will 3221 * be inversed later). 3222 * Otherwise we should allocate new mtag and 3223 * push it into mbuf. 3224 */ 3225 if (cmd->len & F_NOT) { /* `untag' action */ 3226 if (mtag != NULL) 3227 m_tag_delete(m, mtag); 3228 } else if (mtag == NULL) { 3229 if ((mtag = m_tag_alloc(MTAG_IPFW, 3230 tag, 0, M_NOWAIT)) != NULL) 3231 m_tag_prepend(m, mtag); 3232 } 3233 match = (cmd->len & F_NOT) ? 0: 1; 3234 break; 3235 } 3236 3237 case O_TAGGED: { 3238 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3239 tablearg : cmd->arg1; 3240 3241 if (cmdlen == 1) { 3242 match = m_tag_locate(m, MTAG_IPFW, 3243 tag, NULL) != NULL; 3244 break; 3245 } 3246 3247 /* we have ranges */ 3248 for (mtag = m_tag_first(m); 3249 mtag != NULL && !match; 3250 mtag = m_tag_next(m, mtag)) { 3251 uint16_t *p; 3252 int i; 3253 3254 if (mtag->m_tag_cookie != MTAG_IPFW) 3255 continue; 3256 3257 p = ((ipfw_insn_u16 *)cmd)->ports; 3258 i = cmdlen - 1; 3259 for(; !match && i > 0; i--, p += 2) 3260 match = 3261 mtag->m_tag_id >= p[0] && 3262 mtag->m_tag_id <= p[1]; 3263 } 3264 break; 3265 } 3266 3267 /* 3268 * The second set of opcodes represents 'actions', 3269 * i.e. the terminal part of a rule once the packet 3270 * matches all previous patterns. 3271 * Typically there is only one action for each rule, 3272 * and the opcode is stored at the end of the rule 3273 * (but there are exceptions -- see below). 3274 * 3275 * In general, here we set retval and terminate the 3276 * outer loop (would be a 'break 3' in some language, 3277 * but we need to do a 'goto done'). 3278 * 3279 * Exceptions: 3280 * O_COUNT and O_SKIPTO actions: 3281 * instead of terminating, we jump to the next rule 3282 * ('goto next_rule', equivalent to a 'break 2'), 3283 * or to the SKIPTO target ('goto again' after 3284 * having set f, cmd and l), respectively. 3285 * 3286 * O_TAG, O_LOG and O_ALTQ action parameters: 3287 * perform some action and set match = 1; 3288 * 3289 * O_LIMIT and O_KEEP_STATE: these opcodes are 3290 * not real 'actions', and are stored right 3291 * before the 'action' part of the rule. 3292 * These opcodes try to install an entry in the 3293 * state tables; if successful, we continue with 3294 * the next opcode (match=1; break;), otherwise 3295 * the packet * must be dropped 3296 * ('goto done' after setting retval); 3297 * 3298 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3299 * cause a lookup of the state table, and a jump 3300 * to the 'action' part of the parent rule 3301 * ('goto check_body') if an entry is found, or 3302 * (CHECK_STATE only) a jump to the next rule if 3303 * the entry is not found ('goto next_rule'). 3304 * The result of the lookup is cached to make 3305 * further instances of these opcodes are 3306 * effectively NOPs. 3307 */ 3308 case O_LIMIT: 3309 case O_KEEP_STATE: 3310 if (install_state(f, 3311 (ipfw_insn_limit *)cmd, args, tablearg)) { 3312 retval = IP_FW_DENY; 3313 goto done; /* error/limit violation */ 3314 } 3315 match = 1; 3316 break; 3317 3318 case O_PROBE_STATE: 3319 case O_CHECK_STATE: 3320 /* 3321 * dynamic rules are checked at the first 3322 * keep-state or check-state occurrence, 3323 * with the result being stored in dyn_dir. 3324 * The compiler introduces a PROBE_STATE 3325 * instruction for us when we have a 3326 * KEEP_STATE (because PROBE_STATE needs 3327 * to be run first). 3328 */ 3329 if (dyn_dir == MATCH_UNKNOWN && 3330 (q = lookup_dyn_rule(&args->f_id, 3331 &dyn_dir, proto == IPPROTO_TCP ? 3332 TCP(ulp) : NULL)) 3333 != NULL) { 3334 /* 3335 * Found dynamic entry, update stats 3336 * and jump to the 'action' part of 3337 * the parent rule. 3338 */ 3339 q->pcnt++; 3340 q->bcnt += pktlen; 3341 f = q->rule; 3342 cmd = ACTION_PTR(f); 3343 l = f->cmd_len - f->act_ofs; 3344 IPFW_DYN_UNLOCK(); 3345 goto check_body; 3346 } 3347 /* 3348 * Dynamic entry not found. If CHECK_STATE, 3349 * skip to next rule, if PROBE_STATE just 3350 * ignore and continue with next opcode. 3351 */ 3352 if (cmd->opcode == O_CHECK_STATE) 3353 goto next_rule; 3354 match = 1; 3355 break; 3356 3357 case O_ACCEPT: 3358 retval = 0; /* accept */ 3359 goto done; 3360 3361 case O_PIPE: 3362 case O_QUEUE: 3363 args->rule = f; /* report matching rule */ 3364 if (cmd->arg1 == IP_FW_TABLEARG) 3365 args->cookie = tablearg; 3366 else 3367 args->cookie = cmd->arg1; 3368 retval = IP_FW_DUMMYNET; 3369 goto done; 3370 3371 case O_DIVERT: 3372 case O_TEE: { 3373 struct divert_tag *dt; 3374 3375 if (args->eh) /* not on layer 2 */ 3376 break; 3377 mtag = m_tag_get(PACKET_TAG_DIVERT, 3378 sizeof(struct divert_tag), 3379 M_NOWAIT); 3380 if (mtag == NULL) { 3381 /* XXX statistic */ 3382 /* drop packet */ 3383 IPFW_RUNLOCK(chain); 3384 return (IP_FW_DENY); 3385 } 3386 dt = (struct divert_tag *)(mtag+1); 3387 dt->cookie = f->rulenum; 3388 if (cmd->arg1 == IP_FW_TABLEARG) 3389 dt->info = tablearg; 3390 else 3391 dt->info = cmd->arg1; 3392 m_tag_prepend(m, mtag); 3393 retval = (cmd->opcode == O_DIVERT) ? 3394 IP_FW_DIVERT : IP_FW_TEE; 3395 goto done; 3396 } 3397 3398 case O_COUNT: 3399 case O_SKIPTO: 3400 f->pcnt++; /* update stats */ 3401 f->bcnt += pktlen; 3402 f->timestamp = time_uptime; 3403 if (cmd->opcode == O_COUNT) 3404 goto next_rule; 3405 /* handle skipto */ 3406 if (f->next_rule == NULL) 3407 lookup_next_rule(f); 3408 f = f->next_rule; 3409 goto again; 3410 3411 case O_REJECT: 3412 /* 3413 * Drop the packet and send a reject notice 3414 * if the packet is not ICMP (or is an ICMP 3415 * query), and it is not multicast/broadcast. 3416 */ 3417 if (hlen > 0 && is_ipv4 && offset == 0 && 3418 (proto != IPPROTO_ICMP || 3419 is_icmp_query(ICMP(ulp))) && 3420 !(m->m_flags & (M_BCAST|M_MCAST)) && 3421 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3422 send_reject(args, cmd->arg1, ip_len, ip); 3423 m = args->m; 3424 } 3425 /* FALLTHROUGH */ 3426#ifdef INET6 3427 case O_UNREACH6: 3428 if (hlen > 0 && is_ipv6 && 3429 ((offset & IP6F_OFF_MASK) == 0) && 3430 (proto != IPPROTO_ICMPV6 || 3431 (is_icmp6_query(args->f_id.flags) == 1)) && 3432 !(m->m_flags & (M_BCAST|M_MCAST)) && 3433 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3434 send_reject6( 3435 args, cmd->arg1, hlen, 3436 (struct ip6_hdr *)ip); 3437 m = args->m; 3438 } 3439 /* FALLTHROUGH */ 3440#endif 3441 case O_DENY: 3442 retval = IP_FW_DENY; 3443 goto done; 3444 3445 case O_FORWARD_IP: { 3446 struct sockaddr_in *sa; 3447 sa = &(((ipfw_insn_sa *)cmd)->sa); 3448 if (args->eh) /* not valid on layer2 pkts */ 3449 break; 3450 if (!q || dyn_dir == MATCH_FORWARD) { 3451 if (sa->sin_addr.s_addr == INADDR_ANY) { 3452 bcopy(sa, &args->hopstore, 3453 sizeof(*sa)); 3454 args->hopstore.sin_addr.s_addr = 3455 htonl(tablearg); 3456 args->next_hop = 3457 &args->hopstore; 3458 } else { 3459 args->next_hop = sa; 3460 } 3461 } 3462 retval = IP_FW_PASS; 3463 } 3464 goto done; 3465 3466 case O_NETGRAPH: 3467 case O_NGTEE: 3468 args->rule = f; /* report matching rule */ 3469 if (cmd->arg1 == IP_FW_TABLEARG) 3470 args->cookie = tablearg; 3471 else 3472 args->cookie = cmd->arg1; 3473 retval = (cmd->opcode == O_NETGRAPH) ? 3474 IP_FW_NETGRAPH : IP_FW_NGTEE; 3475 goto done; 3476 3477 case O_NAT: { 3478 struct cfg_nat *t; 3479 struct mbuf *mcl; 3480 /* XXX - libalias duct tape */ 3481 int ldt; 3482 char *c; 3483 3484 ldt = 0; 3485 args->rule = f; /* Report matching rule. */ 3486 retval = 0; 3487 t = ((ipfw_insn_nat *)cmd)->nat; 3488 if (t == NULL) { 3489 t = lookup_nat(cmd->arg1); 3490 if (t == NULL) { 3491 retval = IP_FW_DENY; 3492 goto done; 3493 } else 3494 ((ipfw_insn_nat *)cmd)->nat = 3495 t; 3496 } 3497 if ((mcl = m_megapullup(m, m->m_pkthdr.len)) == 3498 NULL) 3499 goto badnat; 3500 ip = mtod(mcl, struct ip *); 3501 if (args->eh == NULL) { 3502 ip->ip_len = htons(ip->ip_len); 3503 ip->ip_off = htons(ip->ip_off); 3504 } 3505 3506 /* 3507 * XXX - Libalias checksum offload 'duct tape': 3508 * 3509 * locally generated packets have only 3510 * pseudo-header checksum calculated 3511 * and libalias will screw it[1], so 3512 * mark them for later fix. Moreover 3513 * there are cases when libalias 3514 * modify tcp packet data[2], mark it 3515 * for later fix too. 3516 * 3517 * [1] libalias was never meant to run 3518 * in kernel, so it doesn't have any 3519 * knowledge about checksum 3520 * offloading, and it expects a packet 3521 * with a full internet 3522 * checksum. Unfortunately, packets 3523 * generated locally will have just the 3524 * pseudo header calculated, and when 3525 * libalias tries to adjust the 3526 * checksum it will actually screw it. 3527 * 3528 * [2] when libalias modify tcp's data 3529 * content, full TCP checksum has to 3530 * be recomputed: the problem is that 3531 * libalias doesn't have any idea 3532 * about checksum offloading To 3533 * workaround this, we do not do 3534 * checksumming in LibAlias, but only 3535 * mark the packets in th_x2 field. If 3536 * we receive a marked packet, we 3537 * calculate correct checksum for it 3538 * aware of offloading. Why such a 3539 * terrible hack instead of 3540 * recalculating checksum for each 3541 * packet? Because the previous 3542 * checksum was not checked! 3543 * Recalculating checksums for EVERY 3544 * packet will hide ALL transmission 3545 * errors. Yes, marked packets still 3546 * suffer from this problem. But, 3547 * sigh, natd(8) has this problem, 3548 * too. 3549 * 3550 * TODO: -make libalias mbuf aware (so 3551 * it can handle delayed checksum and tso) 3552 */ 3553 3554 if (mcl->m_pkthdr.rcvif == NULL && 3555 mcl->m_pkthdr.csum_flags & 3556 CSUM_DELAY_DATA) 3557 ldt = 1; 3558 3559 c = mtod(mcl, char *); 3560 if (oif == NULL) 3561 retval = LibAliasIn(t->lib, c, 3562 MCLBYTES); 3563 else 3564 retval = LibAliasOut(t->lib, c, 3565 MCLBYTES); 3566 if (retval != PKT_ALIAS_OK) { 3567 /* XXX - should i add some logging? */ 3568 m_free(mcl); 3569 badnat: 3570 args->m = NULL; 3571 retval = IP_FW_DENY; 3572 goto done; 3573 } 3574 mcl->m_pkthdr.len = mcl->m_len = 3575 ntohs(ip->ip_len); 3576 3577 /* 3578 * XXX - libalias checksum offload 3579 * 'duct tape' (see above) 3580 */ 3581 3582 if ((ip->ip_off & htons(IP_OFFMASK)) == 0 && 3583 ip->ip_p == IPPROTO_TCP) { 3584 struct tcphdr *th; 3585 3586 th = (struct tcphdr *)(ip + 1); 3587 if (th->th_x2) 3588 ldt = 1; 3589 } 3590 3591 if (ldt) { 3592 struct tcphdr *th; 3593 struct udphdr *uh; 3594 u_short cksum; 3595 3596 ip->ip_len = ntohs(ip->ip_len); 3597 cksum = in_pseudo( 3598 ip->ip_src.s_addr, 3599 ip->ip_dst.s_addr, 3600 htons(ip->ip_p + ip->ip_len - 3601 (ip->ip_hl << 2)) 3602 ); 3603 3604 switch (ip->ip_p) { 3605 case IPPROTO_TCP: 3606 th = (struct tcphdr *)(ip + 1); 3607 /* 3608 * Maybe it was set in 3609 * libalias... 3610 */ 3611 th->th_x2 = 0; 3612 th->th_sum = cksum; 3613 mcl->m_pkthdr.csum_data = 3614 offsetof(struct tcphdr, 3615 th_sum); 3616 break; 3617 case IPPROTO_UDP: 3618 uh = (struct udphdr *)(ip + 1); 3619 uh->uh_sum = cksum; 3620 mcl->m_pkthdr.csum_data = 3621 offsetof(struct udphdr, 3622 uh_sum); 3623 break; 3624 } 3625 /* 3626 * No hw checksum offloading: do it 3627 * by ourself. 3628 */ 3629 if ((mcl->m_pkthdr.csum_flags & 3630 CSUM_DELAY_DATA) == 0) { 3631 in_delayed_cksum(mcl); 3632 mcl->m_pkthdr.csum_flags &= 3633 ~CSUM_DELAY_DATA; 3634 } 3635 ip->ip_len = htons(ip->ip_len); 3636 } 3637 3638 if (args->eh == NULL) { 3639 ip->ip_len = ntohs(ip->ip_len); 3640 ip->ip_off = ntohs(ip->ip_off); 3641 } 3642 3643 args->m = mcl; 3644 retval = IP_FW_NAT; 3645 goto done; 3646 } 3647 3648 default: 3649 panic("-- unknown opcode %d\n", cmd->opcode); 3650 } /* end of switch() on opcodes */ 3651 3652 if (cmd->len & F_NOT) 3653 match = !match; 3654 3655 if (match) { 3656 if (cmd->len & F_OR) 3657 skip_or = 1; 3658 } else { 3659 if (!(cmd->len & F_OR)) /* not an OR block, */ 3660 break; /* try next rule */ 3661 } 3662 3663 } /* end of inner for, scan opcodes */ 3664 3665next_rule:; /* try next rule */ 3666 3667 } /* end of outer for, scan rules */ 3668 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3669 IPFW_RUNLOCK(chain); 3670 return (IP_FW_DENY); 3671 3672done: 3673 /* Update statistics */ 3674 f->pcnt++; 3675 f->bcnt += pktlen; 3676 f->timestamp = time_uptime; 3677 IPFW_RUNLOCK(chain); 3678 return (retval); 3679 3680pullup_failed: 3681 if (fw_verbose) 3682 printf("ipfw: pullup failed\n"); 3683 return (IP_FW_DENY); 3684} 3685 3686/* 3687 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3688 * These will be reconstructed on the fly as packets are matched. 3689 */ 3690static void 3691flush_rule_ptrs(struct ip_fw_chain *chain) 3692{ 3693 struct ip_fw *rule; 3694 3695 IPFW_WLOCK_ASSERT(chain); 3696 3697 for (rule = chain->rules; rule; rule = rule->next) 3698 rule->next_rule = NULL; 3699} 3700 3701/* 3702 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3703 * possibly create a rule number and add the rule to the list. 3704 * Update the rule_number in the input struct so the caller knows it as well. 3705 */ 3706static int 3707add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3708{ 3709 struct ip_fw *rule, *f, *prev; 3710 int l = RULESIZE(input_rule); 3711 3712 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3713 return (EINVAL); 3714 3715 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3716 if (rule == NULL) 3717 return (ENOSPC); 3718 3719 bcopy(input_rule, rule, l); 3720 3721 rule->next = NULL; 3722 rule->next_rule = NULL; 3723 3724 rule->pcnt = 0; 3725 rule->bcnt = 0; 3726 rule->timestamp = 0; 3727 3728 IPFW_WLOCK(chain); 3729 3730 if (chain->rules == NULL) { /* default rule */ 3731 chain->rules = rule; 3732 goto done; 3733 } 3734 3735 /* 3736 * If rulenum is 0, find highest numbered rule before the 3737 * default rule, and add autoinc_step 3738 */ 3739 if (autoinc_step < 1) 3740 autoinc_step = 1; 3741 else if (autoinc_step > 1000) 3742 autoinc_step = 1000; 3743 if (rule->rulenum == 0) { 3744 /* 3745 * locate the highest numbered rule before default 3746 */ 3747 for (f = chain->rules; f; f = f->next) { 3748 if (f->rulenum == IPFW_DEFAULT_RULE) 3749 break; 3750 rule->rulenum = f->rulenum; 3751 } 3752 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3753 rule->rulenum += autoinc_step; 3754 input_rule->rulenum = rule->rulenum; 3755 } 3756 3757 /* 3758 * Now insert the new rule in the right place in the sorted list. 3759 */ 3760 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3761 if (f->rulenum > rule->rulenum) { /* found the location */ 3762 if (prev) { 3763 rule->next = f; 3764 prev->next = rule; 3765 } else { /* head insert */ 3766 rule->next = chain->rules; 3767 chain->rules = rule; 3768 } 3769 break; 3770 } 3771 } 3772 flush_rule_ptrs(chain); 3773done: 3774 static_count++; 3775 static_len += l; 3776 IPFW_WUNLOCK(chain); 3777 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3778 rule->rulenum, static_count);) 3779 return (0); 3780} 3781 3782/** 3783 * Remove a static rule (including derived * dynamic rules) 3784 * and place it on the ``reap list'' for later reclamation. 3785 * The caller is in charge of clearing rule pointers to avoid 3786 * dangling pointers. 3787 * @return a pointer to the next entry. 3788 * Arguments are not checked, so they better be correct. 3789 */ 3790static struct ip_fw * 3791remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3792{ 3793 struct ip_fw *n; 3794 int l = RULESIZE(rule); 3795 3796 IPFW_WLOCK_ASSERT(chain); 3797 3798 n = rule->next; 3799 IPFW_DYN_LOCK(); 3800 remove_dyn_rule(rule, NULL /* force removal */); 3801 IPFW_DYN_UNLOCK(); 3802 if (prev == NULL) 3803 chain->rules = n; 3804 else 3805 prev->next = n; 3806 static_count--; 3807 static_len -= l; 3808 3809 rule->next = chain->reap; 3810 chain->reap = rule; 3811 3812 return n; 3813} 3814 3815/** 3816 * Reclaim storage associated with a list of rules. This is 3817 * typically the list created using remove_rule. 3818 */ 3819static void 3820reap_rules(struct ip_fw *head) 3821{ 3822 struct ip_fw *rule; 3823 3824 while ((rule = head) != NULL) { 3825 head = head->next; 3826 if (DUMMYNET_LOADED) 3827 ip_dn_ruledel_ptr(rule); 3828 free(rule, M_IPFW); 3829 } 3830} 3831 3832/* 3833 * Remove all rules from a chain (except rules in set RESVD_SET 3834 * unless kill_default = 1). The caller is responsible for 3835 * reclaiming storage for the rules left in chain->reap. 3836 */ 3837static void 3838free_chain(struct ip_fw_chain *chain, int kill_default) 3839{ 3840 struct ip_fw *prev, *rule; 3841 3842 IPFW_WLOCK_ASSERT(chain); 3843 3844 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3845 for (prev = NULL, rule = chain->rules; rule ; ) 3846 if (kill_default || rule->set != RESVD_SET) 3847 rule = remove_rule(chain, rule, prev); 3848 else { 3849 prev = rule; 3850 rule = rule->next; 3851 } 3852} 3853 3854/** 3855 * Remove all rules with given number, and also do set manipulation. 3856 * Assumes chain != NULL && *chain != NULL. 3857 * 3858 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3859 * the next 8 bits are the new set, the top 8 bits are the command: 3860 * 3861 * 0 delete rules with given number 3862 * 1 delete rules with given set number 3863 * 2 move rules with given number to new set 3864 * 3 move rules with given set number to new set 3865 * 4 swap sets with given numbers 3866 */ 3867static int 3868del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3869{ 3870 struct ip_fw *prev = NULL, *rule; 3871 u_int16_t rulenum; /* rule or old_set */ 3872 u_int8_t cmd, new_set; 3873 3874 rulenum = arg & 0xffff; 3875 cmd = (arg >> 24) & 0xff; 3876 new_set = (arg >> 16) & 0xff; 3877 3878 if (cmd > 4) 3879 return EINVAL; 3880 if (new_set > RESVD_SET) 3881 return EINVAL; 3882 if (cmd == 0 || cmd == 2) { 3883 if (rulenum >= IPFW_DEFAULT_RULE) 3884 return EINVAL; 3885 } else { 3886 if (rulenum > RESVD_SET) /* old_set */ 3887 return EINVAL; 3888 } 3889 3890 IPFW_WLOCK(chain); 3891 rule = chain->rules; 3892 chain->reap = NULL; 3893 switch (cmd) { 3894 case 0: /* delete rules with given number */ 3895 /* 3896 * locate first rule to delete 3897 */ 3898 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3899 ; 3900 if (rule->rulenum != rulenum) { 3901 IPFW_WUNLOCK(chain); 3902 return EINVAL; 3903 } 3904 3905 /* 3906 * flush pointers outside the loop, then delete all matching 3907 * rules. prev remains the same throughout the cycle. 3908 */ 3909 flush_rule_ptrs(chain); 3910 while (rule->rulenum == rulenum) 3911 rule = remove_rule(chain, rule, prev); 3912 break; 3913 3914 case 1: /* delete all rules with given set number */ 3915 flush_rule_ptrs(chain); 3916 rule = chain->rules; 3917 while (rule->rulenum < IPFW_DEFAULT_RULE) 3918 if (rule->set == rulenum) 3919 rule = remove_rule(chain, rule, prev); 3920 else { 3921 prev = rule; 3922 rule = rule->next; 3923 } 3924 break; 3925 3926 case 2: /* move rules with given number to new set */ 3927 rule = chain->rules; 3928 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3929 if (rule->rulenum == rulenum) 3930 rule->set = new_set; 3931 break; 3932 3933 case 3: /* move rules with given set number to new set */ 3934 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3935 if (rule->set == rulenum) 3936 rule->set = new_set; 3937 break; 3938 3939 case 4: /* swap two sets */ 3940 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3941 if (rule->set == rulenum) 3942 rule->set = new_set; 3943 else if (rule->set == new_set) 3944 rule->set = rulenum; 3945 break; 3946 } 3947 /* 3948 * Look for rules to reclaim. We grab the list before 3949 * releasing the lock then reclaim them w/o the lock to 3950 * avoid a LOR with dummynet. 3951 */ 3952 rule = chain->reap; 3953 chain->reap = NULL; 3954 IPFW_WUNLOCK(chain); 3955 if (rule) 3956 reap_rules(rule); 3957 return 0; 3958} 3959 3960/* 3961 * Clear counters for a specific rule. 3962 * The enclosing "table" is assumed locked. 3963 */ 3964static void 3965clear_counters(struct ip_fw *rule, int log_only) 3966{ 3967 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3968 3969 if (log_only == 0) { 3970 rule->bcnt = rule->pcnt = 0; 3971 rule->timestamp = 0; 3972 } 3973 if (l->o.opcode == O_LOG) 3974 l->log_left = l->max_log; 3975} 3976 3977/** 3978 * Reset some or all counters on firewall rules. 3979 * @arg frwl is null to clear all entries, or contains a specific 3980 * rule number. 3981 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3982 */ 3983static int 3984zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3985{ 3986 struct ip_fw *rule; 3987 char *msg; 3988 3989 IPFW_WLOCK(chain); 3990 if (rulenum == 0) { 3991 norule_counter = 0; 3992 for (rule = chain->rules; rule; rule = rule->next) 3993 clear_counters(rule, log_only); 3994 msg = log_only ? "ipfw: All logging counts reset.\n" : 3995 "ipfw: Accounting cleared.\n"; 3996 } else { 3997 int cleared = 0; 3998 /* 3999 * We can have multiple rules with the same number, so we 4000 * need to clear them all. 4001 */ 4002 for (rule = chain->rules; rule; rule = rule->next) 4003 if (rule->rulenum == rulenum) { 4004 while (rule && rule->rulenum == rulenum) { 4005 clear_counters(rule, log_only); 4006 rule = rule->next; 4007 } 4008 cleared = 1; 4009 break; 4010 } 4011 if (!cleared) { /* we did not find any matching rules */ 4012 IPFW_WUNLOCK(chain); 4013 return (EINVAL); 4014 } 4015 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 4016 "ipfw: Entry %d cleared.\n"; 4017 } 4018 IPFW_WUNLOCK(chain); 4019 4020 if (fw_verbose) 4021 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 4022 return (0); 4023} 4024 4025/* 4026 * Check validity of the structure before insert. 4027 * Fortunately rules are simple, so this mostly need to check rule sizes. 4028 */ 4029static int 4030check_ipfw_struct(struct ip_fw *rule, int size) 4031{ 4032 int l, cmdlen = 0; 4033 int have_action=0; 4034 ipfw_insn *cmd; 4035 4036 if (size < sizeof(*rule)) { 4037 printf("ipfw: rule too short\n"); 4038 return (EINVAL); 4039 } 4040 /* first, check for valid size */ 4041 l = RULESIZE(rule); 4042 if (l != size) { 4043 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 4044 return (EINVAL); 4045 } 4046 if (rule->act_ofs >= rule->cmd_len) { 4047 printf("ipfw: bogus action offset (%u > %u)\n", 4048 rule->act_ofs, rule->cmd_len - 1); 4049 return (EINVAL); 4050 } 4051 /* 4052 * Now go for the individual checks. Very simple ones, basically only 4053 * instruction sizes. 4054 */ 4055 for (l = rule->cmd_len, cmd = rule->cmd ; 4056 l > 0 ; l -= cmdlen, cmd += cmdlen) { 4057 cmdlen = F_LEN(cmd); 4058 if (cmdlen > l) { 4059 printf("ipfw: opcode %d size truncated\n", 4060 cmd->opcode); 4061 return EINVAL; 4062 } 4063 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 4064 switch (cmd->opcode) { 4065 case O_PROBE_STATE: 4066 case O_KEEP_STATE: 4067 case O_PROTO: 4068 case O_IP_SRC_ME: 4069 case O_IP_DST_ME: 4070 case O_LAYER2: 4071 case O_IN: 4072 case O_FRAG: 4073 case O_DIVERTED: 4074 case O_IPOPT: 4075 case O_IPTOS: 4076 case O_IPPRECEDENCE: 4077 case O_IPVER: 4078 case O_TCPWIN: 4079 case O_TCPFLAGS: 4080 case O_TCPOPTS: 4081 case O_ESTAB: 4082 case O_VERREVPATH: 4083 case O_VERSRCREACH: 4084 case O_ANTISPOOF: 4085 case O_IPSEC: 4086#ifdef INET6 4087 case O_IP6_SRC_ME: 4088 case O_IP6_DST_ME: 4089 case O_EXT_HDR: 4090 case O_IP6: 4091#endif 4092 case O_IP4: 4093 case O_TAG: 4094 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4095 goto bad_size; 4096 break; 4097 4098 case O_UID: 4099 case O_GID: 4100 case O_JAIL: 4101 case O_IP_SRC: 4102 case O_IP_DST: 4103 case O_TCPSEQ: 4104 case O_TCPACK: 4105 case O_PROB: 4106 case O_ICMPTYPE: 4107 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 4108 goto bad_size; 4109 break; 4110 4111 case O_LIMIT: 4112 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 4113 goto bad_size; 4114 break; 4115 4116 case O_LOG: 4117 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 4118 goto bad_size; 4119 4120 ((ipfw_insn_log *)cmd)->log_left = 4121 ((ipfw_insn_log *)cmd)->max_log; 4122 4123 break; 4124 4125 case O_IP_SRC_MASK: 4126 case O_IP_DST_MASK: 4127 /* only odd command lengths */ 4128 if ( !(cmdlen & 1) || cmdlen > 31) 4129 goto bad_size; 4130 break; 4131 4132 case O_IP_SRC_SET: 4133 case O_IP_DST_SET: 4134 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 4135 printf("ipfw: invalid set size %d\n", 4136 cmd->arg1); 4137 return EINVAL; 4138 } 4139 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4140 (cmd->arg1+31)/32 ) 4141 goto bad_size; 4142 break; 4143 4144 case O_IP_SRC_LOOKUP: 4145 case O_IP_DST_LOOKUP: 4146 if (cmd->arg1 >= IPFW_TABLES_MAX) { 4147 printf("ipfw: invalid table number %d\n", 4148 cmd->arg1); 4149 return (EINVAL); 4150 } 4151 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 4152 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 4153 goto bad_size; 4154 break; 4155 4156 case O_MACADDR2: 4157 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 4158 goto bad_size; 4159 break; 4160 4161 case O_NOP: 4162 case O_IPID: 4163 case O_IPTTL: 4164 case O_IPLEN: 4165 case O_TCPDATALEN: 4166 case O_TAGGED: 4167 if (cmdlen < 1 || cmdlen > 31) 4168 goto bad_size; 4169 break; 4170 4171 case O_MAC_TYPE: 4172 case O_IP_SRCPORT: 4173 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 4174 if (cmdlen < 2 || cmdlen > 31) 4175 goto bad_size; 4176 break; 4177 4178 case O_RECV: 4179 case O_XMIT: 4180 case O_VIA: 4181 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 4182 goto bad_size; 4183 break; 4184 4185 case O_ALTQ: 4186 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 4187 goto bad_size; 4188 break; 4189 4190 case O_PIPE: 4191 case O_QUEUE: 4192 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4193 goto bad_size; 4194 goto check_action; 4195 4196 case O_FORWARD_IP: 4197#ifdef IPFIREWALL_FORWARD 4198 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 4199 goto bad_size; 4200 goto check_action; 4201#else 4202 return EINVAL; 4203#endif 4204 4205 case O_DIVERT: 4206 case O_TEE: 4207 if (ip_divert_ptr == NULL) 4208 return EINVAL; 4209 else 4210 goto check_size; 4211 case O_NETGRAPH: 4212 case O_NGTEE: 4213 if (!NG_IPFW_LOADED) 4214 return EINVAL; 4215 else 4216 goto check_size; 4217 case O_NAT: 4218 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 4219 goto bad_size; 4220 goto check_action; 4221 case O_FORWARD_MAC: /* XXX not implemented yet */ 4222 case O_CHECK_STATE: 4223 case O_COUNT: 4224 case O_ACCEPT: 4225 case O_DENY: 4226 case O_REJECT: 4227#ifdef INET6 4228 case O_UNREACH6: 4229#endif 4230 case O_SKIPTO: 4231check_size: 4232 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4233 goto bad_size; 4234check_action: 4235 if (have_action) { 4236 printf("ipfw: opcode %d, multiple actions" 4237 " not allowed\n", 4238 cmd->opcode); 4239 return EINVAL; 4240 } 4241 have_action = 1; 4242 if (l != cmdlen) { 4243 printf("ipfw: opcode %d, action must be" 4244 " last opcode\n", 4245 cmd->opcode); 4246 return EINVAL; 4247 } 4248 break; 4249#ifdef INET6 4250 case O_IP6_SRC: 4251 case O_IP6_DST: 4252 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 4253 F_INSN_SIZE(ipfw_insn)) 4254 goto bad_size; 4255 break; 4256 4257 case O_FLOW6ID: 4258 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4259 ((ipfw_insn_u32 *)cmd)->o.arg1) 4260 goto bad_size; 4261 break; 4262 4263 case O_IP6_SRC_MASK: 4264 case O_IP6_DST_MASK: 4265 if ( !(cmdlen & 1) || cmdlen > 127) 4266 goto bad_size; 4267 break; 4268 case O_ICMP6TYPE: 4269 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4270 goto bad_size; 4271 break; 4272#endif 4273 4274 default: 4275 switch (cmd->opcode) { 4276#ifndef INET6 4277 case O_IP6_SRC_ME: 4278 case O_IP6_DST_ME: 4279 case O_EXT_HDR: 4280 case O_IP6: 4281 case O_UNREACH6: 4282 case O_IP6_SRC: 4283 case O_IP6_DST: 4284 case O_FLOW6ID: 4285 case O_IP6_SRC_MASK: 4286 case O_IP6_DST_MASK: 4287 case O_ICMP6TYPE: 4288 printf("ipfw: no IPv6 support in kernel\n"); 4289 return EPROTONOSUPPORT; 4290#endif 4291 default: 4292 printf("ipfw: opcode %d, unknown opcode\n", 4293 cmd->opcode); 4294 return EINVAL; 4295 } 4296 } 4297 } 4298 if (have_action == 0) { 4299 printf("ipfw: missing action\n"); 4300 return EINVAL; 4301 } 4302 return 0; 4303 4304bad_size: 4305 printf("ipfw: opcode %d size %d wrong\n", 4306 cmd->opcode, cmdlen); 4307 return EINVAL; 4308} 4309 4310/* 4311 * Copy the static and dynamic rules to the supplied buffer 4312 * and return the amount of space actually used. 4313 */ 4314static size_t 4315ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4316{ 4317 char *bp = buf; 4318 char *ep = bp + space; 4319 struct ip_fw *rule; 4320 int i; 4321 4322 /* XXX this can take a long time and locking will block packet flow */ 4323 IPFW_RLOCK(chain); 4324 for (rule = chain->rules; rule ; rule = rule->next) { 4325 /* 4326 * Verify the entry fits in the buffer in case the 4327 * rules changed between calculating buffer space and 4328 * now. This would be better done using a generation 4329 * number but should suffice for now. 4330 */ 4331 i = RULESIZE(rule); 4332 if (bp + i <= ep) { 4333 bcopy(rule, bp, i); 4334 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 4335 sizeof(set_disable)); 4336 bp += i; 4337 } 4338 } 4339 IPFW_RUNLOCK(chain); 4340 if (ipfw_dyn_v) { 4341 ipfw_dyn_rule *p, *last = NULL; 4342 4343 IPFW_DYN_LOCK(); 4344 for (i = 0 ; i < curr_dyn_buckets; i++) 4345 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4346 if (bp + sizeof *p <= ep) { 4347 ipfw_dyn_rule *dst = 4348 (ipfw_dyn_rule *)bp; 4349 bcopy(p, dst, sizeof *p); 4350 bcopy(&(p->rule->rulenum), &(dst->rule), 4351 sizeof(p->rule->rulenum)); 4352 /* 4353 * store a non-null value in "next". 4354 * The userland code will interpret a 4355 * NULL here as a marker 4356 * for the last dynamic rule. 4357 */ 4358 bcopy(&dst, &dst->next, sizeof(dst)); 4359 last = dst; 4360 dst->expire = 4361 TIME_LEQ(dst->expire, time_uptime) ? 4362 0 : dst->expire - time_uptime ; 4363 bp += sizeof(ipfw_dyn_rule); 4364 } 4365 } 4366 IPFW_DYN_UNLOCK(); 4367 if (last != NULL) /* mark last dynamic rule */ 4368 bzero(&last->next, sizeof(last)); 4369 } 4370 return (bp - (char *)buf); 4371} 4372 4373 4374/** 4375 * {set|get}sockopt parser. 4376 */ 4377static int 4378ipfw_ctl(struct sockopt *sopt) 4379{ 4380#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4381 int error, rule_num; 4382 size_t size; 4383 struct ip_fw *buf, *rule; 4384 u_int32_t rulenum[2]; 4385 4386 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4387 if (error) 4388 return (error); 4389 4390 /* 4391 * Disallow modifications in really-really secure mode, but still allow 4392 * the logging counters to be reset. 4393 */ 4394 if (sopt->sopt_name == IP_FW_ADD || 4395 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4396 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4397 if (error) 4398 return (error); 4399 } 4400 4401 error = 0; 4402 4403 switch (sopt->sopt_name) { 4404 case IP_FW_GET: 4405 /* 4406 * pass up a copy of the current rules. Static rules 4407 * come first (the last of which has number IPFW_DEFAULT_RULE), 4408 * followed by a possibly empty list of dynamic rule. 4409 * The last dynamic rule has NULL in the "next" field. 4410 * 4411 * Note that the calculated size is used to bound the 4412 * amount of data returned to the user. The rule set may 4413 * change between calculating the size and returning the 4414 * data in which case we'll just return what fits. 4415 */ 4416 size = static_len; /* size of static rules */ 4417 if (ipfw_dyn_v) /* add size of dyn.rules */ 4418 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4419 4420 /* 4421 * XXX todo: if the user passes a short length just to know 4422 * how much room is needed, do not bother filling up the 4423 * buffer, just jump to the sooptcopyout. 4424 */ 4425 buf = malloc(size, M_TEMP, M_WAITOK); 4426 error = sooptcopyout(sopt, buf, 4427 ipfw_getrules(&layer3_chain, buf, size)); 4428 free(buf, M_TEMP); 4429 break; 4430 4431 case IP_FW_FLUSH: 4432 /* 4433 * Normally we cannot release the lock on each iteration. 4434 * We could do it here only because we start from the head all 4435 * the times so there is no risk of missing some entries. 4436 * On the other hand, the risk is that we end up with 4437 * a very inconsistent ruleset, so better keep the lock 4438 * around the whole cycle. 4439 * 4440 * XXX this code can be improved by resetting the head of 4441 * the list to point to the default rule, and then freeing 4442 * the old list without the need for a lock. 4443 */ 4444 4445 IPFW_WLOCK(&layer3_chain); 4446 layer3_chain.reap = NULL; 4447 free_chain(&layer3_chain, 0 /* keep default rule */); 4448 rule = layer3_chain.reap; 4449 layer3_chain.reap = NULL; 4450 IPFW_WUNLOCK(&layer3_chain); 4451 if (rule != NULL) 4452 reap_rules(rule); 4453 break; 4454 4455 case IP_FW_ADD: 4456 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4457 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4458 sizeof(struct ip_fw) ); 4459 if (error == 0) 4460 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4461 if (error == 0) { 4462 error = add_rule(&layer3_chain, rule); 4463 size = RULESIZE(rule); 4464 if (!error && sopt->sopt_dir == SOPT_GET) 4465 error = sooptcopyout(sopt, rule, size); 4466 } 4467 free(rule, M_TEMP); 4468 break; 4469 4470 case IP_FW_DEL: 4471 /* 4472 * IP_FW_DEL is used for deleting single rules or sets, 4473 * and (ab)used to atomically manipulate sets. Argument size 4474 * is used to distinguish between the two: 4475 * sizeof(u_int32_t) 4476 * delete single rule or set of rules, 4477 * or reassign rules (or sets) to a different set. 4478 * 2*sizeof(u_int32_t) 4479 * atomic disable/enable sets. 4480 * first u_int32_t contains sets to be disabled, 4481 * second u_int32_t contains sets to be enabled. 4482 */ 4483 error = sooptcopyin(sopt, rulenum, 4484 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4485 if (error) 4486 break; 4487 size = sopt->sopt_valsize; 4488 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4489 error = del_entry(&layer3_chain, rulenum[0]); 4490 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4491 set_disable = 4492 (set_disable | rulenum[0]) & ~rulenum[1] & 4493 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4494 else 4495 error = EINVAL; 4496 break; 4497 4498 case IP_FW_ZERO: 4499 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 4500 rule_num = 0; 4501 if (sopt->sopt_val != 0) { 4502 error = sooptcopyin(sopt, &rule_num, 4503 sizeof(int), sizeof(int)); 4504 if (error) 4505 break; 4506 } 4507 error = zero_entry(&layer3_chain, rule_num, 4508 sopt->sopt_name == IP_FW_RESETLOG); 4509 break; 4510 4511 case IP_FW_TABLE_ADD: 4512 { 4513 ipfw_table_entry ent; 4514 4515 error = sooptcopyin(sopt, &ent, 4516 sizeof(ent), sizeof(ent)); 4517 if (error) 4518 break; 4519 error = add_table_entry(&layer3_chain, ent.tbl, 4520 ent.addr, ent.masklen, ent.value); 4521 } 4522 break; 4523 4524 case IP_FW_TABLE_DEL: 4525 { 4526 ipfw_table_entry ent; 4527 4528 error = sooptcopyin(sopt, &ent, 4529 sizeof(ent), sizeof(ent)); 4530 if (error) 4531 break; 4532 error = del_table_entry(&layer3_chain, ent.tbl, 4533 ent.addr, ent.masklen); 4534 } 4535 break; 4536 4537 case IP_FW_TABLE_FLUSH: 4538 { 4539 u_int16_t tbl; 4540 4541 error = sooptcopyin(sopt, &tbl, 4542 sizeof(tbl), sizeof(tbl)); 4543 if (error) 4544 break; 4545 IPFW_WLOCK(&layer3_chain); 4546 error = flush_table(&layer3_chain, tbl); 4547 IPFW_WUNLOCK(&layer3_chain); 4548 } 4549 break; 4550 4551 case IP_FW_TABLE_GETSIZE: 4552 { 4553 u_int32_t tbl, cnt; 4554 4555 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4556 sizeof(tbl)))) 4557 break; 4558 IPFW_RLOCK(&layer3_chain); 4559 error = count_table(&layer3_chain, tbl, &cnt); 4560 IPFW_RUNLOCK(&layer3_chain); 4561 if (error) 4562 break; 4563 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4564 } 4565 break; 4566 4567 case IP_FW_TABLE_LIST: 4568 { 4569 ipfw_table *tbl; 4570 4571 if (sopt->sopt_valsize < sizeof(*tbl)) { 4572 error = EINVAL; 4573 break; 4574 } 4575 size = sopt->sopt_valsize; 4576 tbl = malloc(size, M_TEMP, M_WAITOK); 4577 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4578 if (error) { 4579 free(tbl, M_TEMP); 4580 break; 4581 } 4582 tbl->size = (size - sizeof(*tbl)) / 4583 sizeof(ipfw_table_entry); 4584 IPFW_RLOCK(&layer3_chain); 4585 error = dump_table(&layer3_chain, tbl); 4586 IPFW_RUNLOCK(&layer3_chain); 4587 if (error) { 4588 free(tbl, M_TEMP); 4589 break; 4590 } 4591 error = sooptcopyout(sopt, tbl, size); 4592 free(tbl, M_TEMP); 4593 } 4594 break; 4595 4596 case IP_FW_NAT_CFG: 4597 { 4598 struct cfg_nat *ptr, *ser_n; 4599 char *buf; 4600 4601 buf = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO); 4602 error = sooptcopyin(sopt, buf, NAT_BUF_LEN, 4603 sizeof(struct cfg_nat)); 4604 ser_n = (struct cfg_nat *)buf; 4605 4606 /* 4607 * Find/create nat rule. 4608 */ 4609 IPFW_WLOCK(&layer3_chain); 4610 ptr = lookup_nat(ser_n->id); 4611 if (ptr == NULL) { 4612 /* New rule: allocate and init new instance. */ 4613 ptr = malloc(sizeof(struct cfg_nat), 4614 M_IPFW, M_NOWAIT | M_ZERO); 4615 if (ptr == NULL) { 4616 IPFW_WUNLOCK(&layer3_chain); 4617 free(buf, M_IPFW); 4618 return (ENOSPC); 4619 } 4620 ptr->lib = LibAliasInit(NULL); 4621 if (ptr->lib == NULL) { 4622 IPFW_WUNLOCK(&layer3_chain); 4623 free(ptr, M_IPFW); 4624 free(buf, M_IPFW); 4625 return (EINVAL); 4626 } 4627 LIST_INIT(&ptr->redir_chain); 4628 } else { 4629 /* Entry already present: temporarly unhook it. */ 4630 UNHOOK_NAT(ptr); 4631 flush_nat_ptrs(ser_n->id); 4632 } 4633 IPFW_WUNLOCK(&layer3_chain); 4634 4635 /* 4636 * Basic nat configuration. 4637 */ 4638 ptr->id = ser_n->id; 4639 /* 4640 * XXX - what if this rule doesn't nat any ip and just 4641 * redirect? 4642 * do we set aliasaddress to 0.0.0.0? 4643 */ 4644 ptr->ip = ser_n->ip; 4645 ptr->redir_cnt = ser_n->redir_cnt; 4646 ptr->mode = ser_n->mode; 4647 LibAliasSetMode(ptr->lib, ser_n->mode, ser_n->mode); 4648 LibAliasSetAddress(ptr->lib, ptr->ip); 4649 memcpy(ptr->if_name, ser_n->if_name, IF_NAMESIZE); 4650 4651 /* 4652 * Redir and LSNAT configuration. 4653 */ 4654 /* Delete old cfgs. */ 4655 del_redir_spool_cfg(ptr, &ptr->redir_chain); 4656 /* Add new entries. */ 4657 add_redir_spool_cfg(&buf[(sizeof(struct cfg_nat))], ptr); 4658 free(buf, M_IPFW); 4659 IPFW_WLOCK(&layer3_chain); 4660 HOOK_NAT(&layer3_chain.nat, ptr); 4661 IPFW_WUNLOCK(&layer3_chain); 4662 } 4663 break; 4664 4665 case IP_FW_NAT_DEL: 4666 { 4667 struct cfg_nat *ptr; 4668 int i; 4669 4670 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 4671 IPFW_WLOCK(&layer3_chain); 4672 ptr = lookup_nat(i); 4673 if (ptr == NULL) { 4674 error = EINVAL; 4675 IPFW_WUNLOCK(&layer3_chain); 4676 break; 4677 } 4678 UNHOOK_NAT(ptr); 4679 flush_nat_ptrs(i); 4680 IPFW_WUNLOCK(&layer3_chain); 4681 del_redir_spool_cfg(ptr, &ptr->redir_chain); 4682 LibAliasUninit(ptr->lib); 4683 free(ptr, M_IPFW); 4684 } 4685 break; 4686 4687 case IP_FW_NAT_GET_CONFIG: 4688 { 4689 uint8_t *data; 4690 struct cfg_nat *n; 4691 struct cfg_redir *r; 4692 struct cfg_spool *s; 4693 int nat_cnt, off; 4694 4695 nat_cnt = 0; 4696 off = sizeof(nat_cnt); 4697 4698 data = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO); 4699 IPFW_RLOCK(&layer3_chain); 4700 /* Serialize all the data. */ 4701 LIST_FOREACH(n, &layer3_chain.nat, _next) { 4702 nat_cnt++; 4703 if (off + SOF_NAT < NAT_BUF_LEN) { 4704 bcopy(n, &data[off], SOF_NAT); 4705 off += SOF_NAT; 4706 LIST_FOREACH(r, &n->redir_chain, _next) { 4707 if (off + SOF_REDIR < NAT_BUF_LEN) { 4708 bcopy(r, &data[off], 4709 SOF_REDIR); 4710 off += SOF_REDIR; 4711 LIST_FOREACH(s, &r->spool_chain, 4712 _next) { 4713 if (off + SOF_SPOOL < 4714 NAT_BUF_LEN) { 4715 bcopy(s, 4716 &data[off], 4717 SOF_SPOOL); 4718 off += 4719 SOF_SPOOL; 4720 } else 4721 goto nospace; 4722 } 4723 } else 4724 goto nospace; 4725 } 4726 } else 4727 goto nospace; 4728 } 4729 bcopy(&nat_cnt, data, sizeof(nat_cnt)); 4730 IPFW_RUNLOCK(&layer3_chain); 4731 error = sooptcopyout(sopt, data, NAT_BUF_LEN); 4732 free(data, M_IPFW); 4733 break; 4734 nospace: 4735 IPFW_RUNLOCK(&layer3_chain); 4736 printf("serialized data buffer not big enough:" 4737 "please increase NAT_BUF_LEN\n"); 4738 free(data, M_IPFW); 4739 } 4740 break; 4741 4742 case IP_FW_NAT_GET_LOG: 4743 { 4744 uint8_t *data; 4745 struct cfg_nat *ptr; 4746 int i, size, cnt, sof; 4747 4748 data = NULL; 4749 sof = LIBALIAS_BUF_SIZE; 4750 cnt = 0; 4751 4752 IPFW_RLOCK(&layer3_chain); 4753 size = i = 0; 4754 LIST_FOREACH(ptr, &layer3_chain.nat, _next) { 4755 if (ptr->lib->logDesc == NULL) 4756 continue; 4757 cnt++; 4758 size = cnt * (sof + sizeof(int)); 4759 data = realloc(data, size, M_IPFW, M_NOWAIT | M_ZERO); 4760 if (data == NULL) { 4761 IPFW_RUNLOCK(&layer3_chain); 4762 return (ENOSPC); 4763 } 4764 bcopy(&ptr->id, &data[i], sizeof(int)); 4765 i += sizeof(int); 4766 bcopy(ptr->lib->logDesc, &data[i], sof); 4767 i += sof; 4768 } 4769 IPFW_RUNLOCK(&layer3_chain); 4770 error = sooptcopyout(sopt, data, size); 4771 free(data, M_IPFW); 4772 } 4773 break; 4774 4775 default: 4776 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4777 error = EINVAL; 4778 } 4779 4780 return (error); 4781#undef RULE_MAXSIZE 4782} 4783 4784/** 4785 * dummynet needs a reference to the default rule, because rules can be 4786 * deleted while packets hold a reference to them. When this happens, 4787 * dummynet changes the reference to the default rule (it could well be a 4788 * NULL pointer, but this way we do not need to check for the special 4789 * case, plus here he have info on the default behaviour). 4790 */ 4791struct ip_fw *ip_fw_default_rule; 4792 4793/* 4794 * This procedure is only used to handle keepalives. It is invoked 4795 * every dyn_keepalive_period 4796 */ 4797static void 4798ipfw_tick(void * __unused unused) 4799{ 4800 struct mbuf *m0, *m, *mnext, **mtailp; 4801 int i; 4802 ipfw_dyn_rule *q; 4803 4804 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4805 goto done; 4806 4807 /* 4808 * We make a chain of packets to go out here -- not deferring 4809 * until after we drop the IPFW dynamic rule lock would result 4810 * in a lock order reversal with the normal packet input -> ipfw 4811 * call stack. 4812 */ 4813 m0 = NULL; 4814 mtailp = &m0; 4815 IPFW_DYN_LOCK(); 4816 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4817 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4818 if (q->dyn_type == O_LIMIT_PARENT) 4819 continue; 4820 if (q->id.proto != IPPROTO_TCP) 4821 continue; 4822 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4823 continue; 4824 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4825 q->expire)) 4826 continue; /* too early */ 4827 if (TIME_LEQ(q->expire, time_uptime)) 4828 continue; /* too late, rule expired */ 4829 4830 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4831 q->ack_fwd, TH_SYN); 4832 if (*mtailp != NULL) 4833 mtailp = &(*mtailp)->m_nextpkt; 4834 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4835 q->ack_rev, 0); 4836 if (*mtailp != NULL) 4837 mtailp = &(*mtailp)->m_nextpkt; 4838 } 4839 } 4840 IPFW_DYN_UNLOCK(); 4841 for (m = mnext = m0; m != NULL; m = mnext) { 4842 mnext = m->m_nextpkt; 4843 m->m_nextpkt = NULL; 4844 ip_output(m, NULL, NULL, 0, NULL, NULL); 4845 } 4846done: 4847 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4848} 4849 4850int 4851ipfw_init(void) 4852{ 4853 struct ip_fw default_rule; 4854 int error; 4855 4856#ifdef INET6 4857 /* Setup IPv6 fw sysctl tree. */ 4858 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4859 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4860 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4861 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4862 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4863 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4864 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4865 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4866 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4867 &fw_deny_unknown_exthdrs, 0, 4868 "Deny packets with unknown IPv6 Extension Headers"); 4869#endif 4870 4871 layer3_chain.rules = NULL; 4872 IPFW_LOCK_INIT(&layer3_chain); 4873 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4874 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4875 UMA_ALIGN_PTR, 0); 4876 IPFW_DYN_LOCK_INIT(); 4877 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4878 4879 bzero(&default_rule, sizeof default_rule); 4880 4881 default_rule.act_ofs = 0; 4882 default_rule.rulenum = IPFW_DEFAULT_RULE; 4883 default_rule.cmd_len = 1; 4884 default_rule.set = RESVD_SET; 4885 4886 default_rule.cmd[0].len = 1; 4887 default_rule.cmd[0].opcode = 4888#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4889 1 ? O_ACCEPT : 4890#endif 4891 O_DENY; 4892 4893 error = add_rule(&layer3_chain, &default_rule); 4894 if (error != 0) { 4895 printf("ipfw2: error %u initializing default rule " 4896 "(support disabled)\n", error); 4897 IPFW_DYN_LOCK_DESTROY(); 4898 IPFW_LOCK_DESTROY(&layer3_chain); 4899 uma_zdestroy(ipfw_dyn_rule_zone); 4900 return (error); 4901 } 4902 4903 ip_fw_default_rule = layer3_chain.rules; 4904 printf("ipfw2 " 4905#ifdef INET6 4906 "(+ipv6) " 4907#endif 4908 "initialized, divert %s, " 4909 "rule-based forwarding " 4910#ifdef IPFIREWALL_FORWARD 4911 "enabled, " 4912#else 4913 "disabled, " 4914#endif 4915 "default to %s, logging ", 4916#ifdef IPDIVERT 4917 "enabled", 4918#else 4919 "loadable", 4920#endif 4921 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4922 4923#ifdef IPFIREWALL_VERBOSE 4924 fw_verbose = 1; 4925#endif 4926#ifdef IPFIREWALL_VERBOSE_LIMIT 4927 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4928#endif 4929 if (fw_verbose == 0) 4930 printf("disabled\n"); 4931 else if (verbose_limit == 0) 4932 printf("unlimited\n"); 4933 else 4934 printf("limited to %d packets/entry by default\n", 4935 verbose_limit); 4936 4937 error = init_tables(&layer3_chain); 4938 if (error) { 4939 IPFW_DYN_LOCK_DESTROY(); 4940 IPFW_LOCK_DESTROY(&layer3_chain); 4941 uma_zdestroy(ipfw_dyn_rule_zone); 4942 return (error); 4943 } 4944 ip_fw_ctl_ptr = ipfw_ctl; 4945 ip_fw_chk_ptr = ipfw_chk; 4946 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4947 LIST_INIT(&layer3_chain.nat); 4948 ifaddr_event_tag = EVENTHANDLER_REGISTER(ifaddr_event, ifaddr_change, 4949 NULL, EVENTHANDLER_PRI_ANY); 4950 return (0); 4951} 4952 4953void 4954ipfw_destroy(void) 4955{ 4956 struct ip_fw *reap; 4957 struct cfg_nat *ptr, *ptr_temp; 4958 4959 ip_fw_chk_ptr = NULL; 4960 ip_fw_ctl_ptr = NULL; 4961 callout_drain(&ipfw_timeout); 4962 IPFW_WLOCK(&layer3_chain); 4963 flush_tables(&layer3_chain); 4964 LIST_FOREACH_SAFE(ptr, &layer3_chain.nat, _next, ptr_temp) { 4965 LIST_REMOVE(ptr, _next); 4966 del_redir_spool_cfg(ptr, &ptr->redir_chain); 4967 LibAliasUninit(ptr->lib); 4968 free(ptr, M_IPFW); 4969 } 4970 EVENTHANDLER_DEREGISTER(ifaddr_event, ifaddr_event_tag); 4971 layer3_chain.reap = NULL; 4972 free_chain(&layer3_chain, 1 /* kill default rule */); 4973 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4974 IPFW_WUNLOCK(&layer3_chain); 4975 if (reap != NULL) 4976 reap_rules(reap); 4977 IPFW_DYN_LOCK_DESTROY(); 4978 uma_zdestroy(ipfw_dyn_rule_zone); 4979 IPFW_LOCK_DESTROY(&layer3_chain); 4980 4981#ifdef INET6 4982 /* Free IPv6 fw sysctl tree. */ 4983 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4984#endif 4985 4986 printf("IP firewall unloaded\n"); 4987} 4988