ip_fw2.c revision 191288
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 26#include <sys/cdefs.h> 27__FBSDID("$FreeBSD: head/sys/netinet/ip_fw2.c 191288 2009-04-19 22:34:35Z rwatson $"); 28 29#define DEB(x) 30#define DDB(x) x 31 32/* 33 * Implement IP packet firewall (new version) 34 */ 35 36#if !defined(KLD_MODULE) 37#include "opt_ipfw.h" 38#include "opt_ipdivert.h" 39#include "opt_ipdn.h" 40#include "opt_inet.h" 41#ifndef INET 42#error IPFIREWALL requires INET. 43#endif /* INET */ 44#endif 45#include "opt_inet6.h" 46#include "opt_ipsec.h" 47#include "opt_mac.h" 48#include "opt_route.h" 49 50#include <sys/param.h> 51#include <sys/systm.h> 52#include <sys/condvar.h> 53#include <sys/eventhandler.h> 54#include <sys/malloc.h> 55#include <sys/mbuf.h> 56#include <sys/kernel.h> 57#include <sys/lock.h> 58#include <sys/jail.h> 59#include <sys/module.h> 60#include <sys/priv.h> 61#include <sys/proc.h> 62#include <sys/rwlock.h> 63#include <sys/socket.h> 64#include <sys/socketvar.h> 65#include <sys/sysctl.h> 66#include <sys/syslog.h> 67#include <sys/ucred.h> 68#include <sys/vimage.h> 69#include <net/ethernet.h> /* for ETHERTYPE_IP */ 70#include <net/if.h> 71#include <net/radix.h> 72#include <net/route.h> 73#include <net/pf_mtag.h> 74#include <net/vnet.h> 75 76#define IPFW_INTERNAL /* Access to protected data structures in ip_fw.h. */ 77 78#include <netinet/in.h> 79#include <netinet/in_var.h> 80#include <netinet/in_pcb.h> 81#include <netinet/ip.h> 82#include <netinet/ip_var.h> 83#include <netinet/ip_icmp.h> 84#include <netinet/ip_fw.h> 85#include <netinet/ip_divert.h> 86#include <netinet/ip_dummynet.h> 87#include <netinet/ip_carp.h> 88#include <netinet/pim.h> 89#include <netinet/tcp_var.h> 90#include <netinet/udp.h> 91#include <netinet/udp_var.h> 92#include <netinet/sctp.h> 93#include <netinet/vinet.h> 94 95#include <netgraph/ng_ipfw.h> 96 97#include <netinet/ip6.h> 98#include <netinet/icmp6.h> 99#ifdef INET6 100#include <netinet6/scope6_var.h> 101#endif 102 103#include <machine/in_cksum.h> /* XXX for in_cksum */ 104 105#ifdef MAC 106#include <security/mac/mac_framework.h> 107#endif 108 109#ifndef VIMAGE 110#ifndef VIMAGE_GLOBALS 111struct vnet_ipfw vnet_ipfw_0; 112#endif 113#endif 114 115/* 116 * set_disable contains one bit per set value (0..31). 117 * If the bit is set, all rules with the corresponding set 118 * are disabled. Set RESVD_SET(31) is reserved for the default rule 119 * and rules that are not deleted by the flush command, 120 * and CANNOT be disabled. 121 * Rules in set RESVD_SET can only be deleted explicitly. 122 */ 123#ifdef VIMAGE_GLOBALS 124static u_int32_t set_disable; 125static int fw_verbose; 126static struct callout ipfw_timeout; 127static int verbose_limit; 128#endif 129 130static uma_zone_t ipfw_dyn_rule_zone; 131 132/* 133 * Data structure to cache our ucred related 134 * information. This structure only gets used if 135 * the user specified UID/GID based constraints in 136 * a firewall rule. 137 */ 138struct ip_fw_ugid { 139 gid_t fw_groups[NGROUPS]; 140 int fw_ngroups; 141 uid_t fw_uid; 142 int fw_prid; 143}; 144 145/* 146 * list of rules for layer 3 147 */ 148#ifdef VIMAGE_GLOBALS 149struct ip_fw_chain layer3_chain; 150#endif 151 152MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 153MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 154#define IPFW_NAT_LOADED (ipfw_nat_ptr != NULL) 155ipfw_nat_t *ipfw_nat_ptr = NULL; 156ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 157ipfw_nat_cfg_t *ipfw_nat_del_ptr; 158ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 159ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 160 161struct table_entry { 162 struct radix_node rn[2]; 163 struct sockaddr_in addr, mask; 164 u_int32_t value; 165}; 166 167#ifdef VIMAGE_GLOBALS 168static int autoinc_step; 169#endif 170 171extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 172 173#ifdef SYSCTL_NODE 174SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 175SYSCTL_V_PROC(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, enable, 176 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, fw_enable, 0, 177 ipfw_chg_hook, "I", "Enable ipfw"); 178SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, autoinc_step, 179 CTLFLAG_RW, autoinc_step, 0, "Rule number autincrement step"); 180SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_fw, OID_AUTO, one_pass, 181 CTLFLAG_RW | CTLFLAG_SECURE3, fw_one_pass, 0, 182 "Only do a single pass through ipfw when using dummynet(4)"); 183SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, verbose, 184 CTLFLAG_RW | CTLFLAG_SECURE3, 185 fw_verbose, 0, "Log matches to ipfw rules"); 186SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, verbose_limit, 187 CTLFLAG_RW, verbose_limit, 0, 188 "Set upper limit of matches of ipfw rules logged"); 189SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 190 NULL, IPFW_DEFAULT_RULE, "The default/max possible rule number."); 191SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 192 NULL, IPFW_TABLES_MAX, "The maximum number of tables."); 193#endif /* SYSCTL_NODE */ 194 195/* 196 * Description of dynamic rules. 197 * 198 * Dynamic rules are stored in lists accessed through a hash table 199 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 200 * be modified through the sysctl variable dyn_buckets which is 201 * updated when the table becomes empty. 202 * 203 * XXX currently there is only one list, ipfw_dyn. 204 * 205 * When a packet is received, its address fields are first masked 206 * with the mask defined for the rule, then hashed, then matched 207 * against the entries in the corresponding list. 208 * Dynamic rules can be used for different purposes: 209 * + stateful rules; 210 * + enforcing limits on the number of sessions; 211 * + in-kernel NAT (not implemented yet) 212 * 213 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 214 * measured in seconds and depending on the flags. 215 * 216 * The total number of dynamic rules is stored in dyn_count. 217 * The max number of dynamic rules is dyn_max. When we reach 218 * the maximum number of rules we do not create anymore. This is 219 * done to avoid consuming too much memory, but also too much 220 * time when searching on each packet (ideally, we should try instead 221 * to put a limit on the length of the list on each bucket...). 222 * 223 * Each dynamic rule holds a pointer to the parent ipfw rule so 224 * we know what action to perform. Dynamic rules are removed when 225 * the parent rule is deleted. XXX we should make them survive. 226 * 227 * There are some limitations with dynamic rules -- we do not 228 * obey the 'randomized match', and we do not do multiple 229 * passes through the firewall. XXX check the latter!!! 230 */ 231#ifdef VIMAGE_GLOBALS 232static ipfw_dyn_rule **ipfw_dyn_v; 233static u_int32_t dyn_buckets; 234static u_int32_t curr_dyn_buckets; 235#endif 236 237static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 238#define IPFW_DYN_LOCK_INIT() \ 239 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 240#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 241#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 242#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 243#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 244 245/* 246 * Timeouts for various events in handing dynamic rules. 247 */ 248#ifdef VIMAGE_GLOBALS 249static u_int32_t dyn_ack_lifetime; 250static u_int32_t dyn_syn_lifetime; 251static u_int32_t dyn_fin_lifetime; 252static u_int32_t dyn_rst_lifetime; 253static u_int32_t dyn_udp_lifetime; 254static u_int32_t dyn_short_lifetime; 255 256/* 257 * Keepalives are sent if dyn_keepalive is set. They are sent every 258 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 259 * seconds of lifetime of a rule. 260 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 261 * than dyn_keepalive_period. 262 */ 263 264static u_int32_t dyn_keepalive_interval; 265static u_int32_t dyn_keepalive_period; 266static u_int32_t dyn_keepalive; 267 268static u_int32_t static_count; /* # of static rules */ 269static u_int32_t static_len; /* size in bytes of static rules */ 270static u_int32_t dyn_count; /* # of dynamic rules */ 271static u_int32_t dyn_max; /* max # of dynamic rules */ 272#endif /* VIMAGE_GLOBALS */ 273 274#ifdef SYSCTL_NODE 275SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_buckets, 276 CTLFLAG_RW, dyn_buckets, 0, "Number of dyn. buckets"); 277SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, 278 CTLFLAG_RD, curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 279SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_count, 280 CTLFLAG_RD, dyn_count, 0, "Number of dyn. rules"); 281SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_max, 282 CTLFLAG_RW, dyn_max, 0, "Max number of dyn. rules"); 283SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, static_count, 284 CTLFLAG_RD, static_count, 0, "Number of static rules"); 285SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, 286 CTLFLAG_RW, dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 287SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, 288 CTLFLAG_RW, dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 289SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, 290 CTLFLAG_RW, dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 291SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, 292 CTLFLAG_RW, dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 293SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, 294 CTLFLAG_RW, dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 295SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, 296 CTLFLAG_RW, dyn_short_lifetime, 0, 297 "Lifetime of dyn. rules for other situations"); 298SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_keepalive, 299 CTLFLAG_RW, dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 300#endif /* SYSCTL_NODE */ 301 302#ifdef INET6 303/* 304 * IPv6 specific variables 305 */ 306#ifdef SYSCTL_NODE 307SYSCTL_DECL(_net_inet6_ip6); 308#endif /* SYSCTL_NODE */ 309 310static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 311static struct sysctl_oid *ip6_fw_sysctl_tree; 312#endif /* INET6 */ 313 314#ifdef VIMAGE_GLOBALS 315static int fw_deny_unknown_exthdrs; 316#endif 317 318/* 319 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 320 * Other macros just cast void * into the appropriate type 321 */ 322#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 323#define TCP(p) ((struct tcphdr *)(p)) 324#define SCTP(p) ((struct sctphdr *)(p)) 325#define UDP(p) ((struct udphdr *)(p)) 326#define ICMP(p) ((struct icmphdr *)(p)) 327#define ICMP6(p) ((struct icmp6_hdr *)(p)) 328 329static __inline int 330icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 331{ 332 int type = icmp->icmp_type; 333 334 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 335} 336 337#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 338 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 339 340static int 341is_icmp_query(struct icmphdr *icmp) 342{ 343 int type = icmp->icmp_type; 344 345 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 346} 347#undef TT 348 349/* 350 * The following checks use two arrays of 8 or 16 bits to store the 351 * bits that we want set or clear, respectively. They are in the 352 * low and high half of cmd->arg1 or cmd->d[0]. 353 * 354 * We scan options and store the bits we find set. We succeed if 355 * 356 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 357 * 358 * The code is sometimes optimized not to store additional variables. 359 */ 360 361static int 362flags_match(ipfw_insn *cmd, u_int8_t bits) 363{ 364 u_char want_clear; 365 bits = ~bits; 366 367 if ( ((cmd->arg1 & 0xff) & bits) != 0) 368 return 0; /* some bits we want set were clear */ 369 want_clear = (cmd->arg1 >> 8) & 0xff; 370 if ( (want_clear & bits) != want_clear) 371 return 0; /* some bits we want clear were set */ 372 return 1; 373} 374 375static int 376ipopts_match(struct ip *ip, ipfw_insn *cmd) 377{ 378 int optlen, bits = 0; 379 u_char *cp = (u_char *)(ip + 1); 380 int x = (ip->ip_hl << 2) - sizeof (struct ip); 381 382 for (; x > 0; x -= optlen, cp += optlen) { 383 int opt = cp[IPOPT_OPTVAL]; 384 385 if (opt == IPOPT_EOL) 386 break; 387 if (opt == IPOPT_NOP) 388 optlen = 1; 389 else { 390 optlen = cp[IPOPT_OLEN]; 391 if (optlen <= 0 || optlen > x) 392 return 0; /* invalid or truncated */ 393 } 394 switch (opt) { 395 396 default: 397 break; 398 399 case IPOPT_LSRR: 400 bits |= IP_FW_IPOPT_LSRR; 401 break; 402 403 case IPOPT_SSRR: 404 bits |= IP_FW_IPOPT_SSRR; 405 break; 406 407 case IPOPT_RR: 408 bits |= IP_FW_IPOPT_RR; 409 break; 410 411 case IPOPT_TS: 412 bits |= IP_FW_IPOPT_TS; 413 break; 414 } 415 } 416 return (flags_match(cmd, bits)); 417} 418 419static int 420tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 421{ 422 int optlen, bits = 0; 423 u_char *cp = (u_char *)(tcp + 1); 424 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 425 426 for (; x > 0; x -= optlen, cp += optlen) { 427 int opt = cp[0]; 428 if (opt == TCPOPT_EOL) 429 break; 430 if (opt == TCPOPT_NOP) 431 optlen = 1; 432 else { 433 optlen = cp[1]; 434 if (optlen <= 0) 435 break; 436 } 437 438 switch (opt) { 439 440 default: 441 break; 442 443 case TCPOPT_MAXSEG: 444 bits |= IP_FW_TCPOPT_MSS; 445 break; 446 447 case TCPOPT_WINDOW: 448 bits |= IP_FW_TCPOPT_WINDOW; 449 break; 450 451 case TCPOPT_SACK_PERMITTED: 452 case TCPOPT_SACK: 453 bits |= IP_FW_TCPOPT_SACK; 454 break; 455 456 case TCPOPT_TIMESTAMP: 457 bits |= IP_FW_TCPOPT_TS; 458 break; 459 460 } 461 } 462 return (flags_match(cmd, bits)); 463} 464 465static int 466iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 467{ 468 if (ifp == NULL) /* no iface with this packet, match fails */ 469 return 0; 470 /* Check by name or by IP address */ 471 if (cmd->name[0] != '\0') { /* match by name */ 472 /* Check name */ 473 if (cmd->p.glob) { 474 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 475 return(1); 476 } else { 477 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 478 return(1); 479 } 480 } else { 481 struct ifaddr *ia; 482 483 IF_ADDR_LOCK(ifp); 484 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 485 if (ia->ifa_addr->sa_family != AF_INET) 486 continue; 487 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 488 (ia->ifa_addr))->sin_addr.s_addr) { 489 IF_ADDR_UNLOCK(ifp); 490 return(1); /* match */ 491 } 492 } 493 IF_ADDR_UNLOCK(ifp); 494 } 495 return(0); /* no match, fail ... */ 496} 497 498/* 499 * The verify_path function checks if a route to the src exists and 500 * if it is reachable via ifp (when provided). 501 * 502 * The 'verrevpath' option checks that the interface that an IP packet 503 * arrives on is the same interface that traffic destined for the 504 * packet's source address would be routed out of. The 'versrcreach' 505 * option just checks that the source address is reachable via any route 506 * (except default) in the routing table. These two are a measure to block 507 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 508 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 509 * is purposely reminiscent of the Cisco IOS command, 510 * 511 * ip verify unicast reverse-path 512 * ip verify unicast source reachable-via any 513 * 514 * which implements the same functionality. But note that syntax is 515 * misleading. The check may be performed on all IP packets whether unicast, 516 * multicast, or broadcast. 517 */ 518static int 519verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 520{ 521 struct route ro; 522 struct sockaddr_in *dst; 523 524 bzero(&ro, sizeof(ro)); 525 526 dst = (struct sockaddr_in *)&(ro.ro_dst); 527 dst->sin_family = AF_INET; 528 dst->sin_len = sizeof(*dst); 529 dst->sin_addr = src; 530 in_rtalloc_ign(&ro, 0, fib); 531 532 if (ro.ro_rt == NULL) 533 return 0; 534 535 /* 536 * If ifp is provided, check for equality with rtentry. 537 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 538 * in order to pass packets injected back by if_simloop(): 539 * if useloopback == 1 routing entry (via lo0) for our own address 540 * may exist, so we need to handle routing assymetry. 541 */ 542 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 543 RTFREE(ro.ro_rt); 544 return 0; 545 } 546 547 /* if no ifp provided, check if rtentry is not default route */ 548 if (ifp == NULL && 549 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 550 RTFREE(ro.ro_rt); 551 return 0; 552 } 553 554 /* or if this is a blackhole/reject route */ 555 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 556 RTFREE(ro.ro_rt); 557 return 0; 558 } 559 560 /* found valid route */ 561 RTFREE(ro.ro_rt); 562 return 1; 563} 564 565#ifdef INET6 566/* 567 * ipv6 specific rules here... 568 */ 569static __inline int 570icmp6type_match (int type, ipfw_insn_u32 *cmd) 571{ 572 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 573} 574 575static int 576flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 577{ 578 int i; 579 for (i=0; i <= cmd->o.arg1; ++i ) 580 if (curr_flow == cmd->d[i] ) 581 return 1; 582 return 0; 583} 584 585/* support for IP6_*_ME opcodes */ 586static int 587search_ip6_addr_net (struct in6_addr * ip6_addr) 588{ 589 INIT_VNET_NET(curvnet); 590 struct ifnet *mdc; 591 struct ifaddr *mdc2; 592 struct in6_ifaddr *fdm; 593 struct in6_addr copia; 594 595 TAILQ_FOREACH(mdc, &V_ifnet, if_link) { 596 IF_ADDR_LOCK(mdc); 597 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_list) { 598 if (mdc2->ifa_addr->sa_family == AF_INET6) { 599 fdm = (struct in6_ifaddr *)mdc2; 600 copia = fdm->ia_addr.sin6_addr; 601 /* need for leaving scope_id in the sock_addr */ 602 in6_clearscope(&copia); 603 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { 604 IF_ADDR_UNLOCK(mdc); 605 return 1; 606 } 607 } 608 } 609 IF_ADDR_UNLOCK(mdc); 610 } 611 return 0; 612} 613 614static int 615verify_path6(struct in6_addr *src, struct ifnet *ifp) 616{ 617 struct route_in6 ro; 618 struct sockaddr_in6 *dst; 619 620 bzero(&ro, sizeof(ro)); 621 622 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 623 dst->sin6_family = AF_INET6; 624 dst->sin6_len = sizeof(*dst); 625 dst->sin6_addr = *src; 626 /* XXX MRT 0 for ipv6 at this time */ 627 rtalloc_ign((struct route *)&ro, 0); 628 629 if (ro.ro_rt == NULL) 630 return 0; 631 632 /* 633 * if ifp is provided, check for equality with rtentry 634 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 635 * to support the case of sending packets to an address of our own. 636 * (where the former interface is the first argument of if_simloop() 637 * (=ifp), the latter is lo0) 638 */ 639 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 640 RTFREE(ro.ro_rt); 641 return 0; 642 } 643 644 /* if no ifp provided, check if rtentry is not default route */ 645 if (ifp == NULL && 646 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 647 RTFREE(ro.ro_rt); 648 return 0; 649 } 650 651 /* or if this is a blackhole/reject route */ 652 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 653 RTFREE(ro.ro_rt); 654 return 0; 655 } 656 657 /* found valid route */ 658 RTFREE(ro.ro_rt); 659 return 1; 660 661} 662static __inline int 663hash_packet6(struct ipfw_flow_id *id) 664{ 665 u_int32_t i; 666 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 667 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 668 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 669 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 670 (id->dst_port) ^ (id->src_port); 671 return i; 672} 673 674static int 675is_icmp6_query(int icmp6_type) 676{ 677 if ((icmp6_type <= ICMP6_MAXTYPE) && 678 (icmp6_type == ICMP6_ECHO_REQUEST || 679 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 680 icmp6_type == ICMP6_WRUREQUEST || 681 icmp6_type == ICMP6_FQDN_QUERY || 682 icmp6_type == ICMP6_NI_QUERY)) 683 return (1); 684 685 return (0); 686} 687 688static void 689send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 690{ 691 struct mbuf *m; 692 693 m = args->m; 694 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 695 struct tcphdr *tcp; 696 tcp_seq ack, seq; 697 int flags; 698 struct { 699 struct ip6_hdr ip6; 700 struct tcphdr th; 701 } ti; 702 tcp = (struct tcphdr *)((char *)ip6 + hlen); 703 704 if ((tcp->th_flags & TH_RST) != 0) { 705 m_freem(m); 706 args->m = NULL; 707 return; 708 } 709 710 ti.ip6 = *ip6; 711 ti.th = *tcp; 712 ti.th.th_seq = ntohl(ti.th.th_seq); 713 ti.th.th_ack = ntohl(ti.th.th_ack); 714 ti.ip6.ip6_nxt = IPPROTO_TCP; 715 716 if (ti.th.th_flags & TH_ACK) { 717 ack = 0; 718 seq = ti.th.th_ack; 719 flags = TH_RST; 720 } else { 721 ack = ti.th.th_seq; 722 if ((m->m_flags & M_PKTHDR) != 0) { 723 /* 724 * total new data to ACK is: 725 * total packet length, 726 * minus the header length, 727 * minus the tcp header length. 728 */ 729 ack += m->m_pkthdr.len - hlen 730 - (ti.th.th_off << 2); 731 } else if (ip6->ip6_plen) { 732 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 733 hlen - (ti.th.th_off << 2); 734 } else { 735 m_freem(m); 736 return; 737 } 738 if (tcp->th_flags & TH_SYN) 739 ack++; 740 seq = 0; 741 flags = TH_RST|TH_ACK; 742 } 743 bcopy(&ti, ip6, sizeof(ti)); 744 /* 745 * m is only used to recycle the mbuf 746 * The data in it is never read so we don't need 747 * to correct the offsets or anything 748 */ 749 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 750 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 751#if 0 752 /* 753 * Unlike above, the mbufs need to line up with the ip6 hdr, 754 * as the contents are read. We need to m_adj() the 755 * needed amount. 756 * The mbuf will however be thrown away so we can adjust it. 757 * Remember we did an m_pullup on it already so we 758 * can make some assumptions about contiguousness. 759 */ 760 if (args->L3offset) 761 m_adj(m, args->L3offset); 762#endif 763 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 764 } else 765 m_freem(m); 766 767 args->m = NULL; 768} 769 770#endif /* INET6 */ 771 772#ifdef VIMAGE_GLOBALS 773static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 774#endif 775 776#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 777#define SNP(buf) buf, sizeof(buf) 778 779/* 780 * We enter here when we have a rule with O_LOG. 781 * XXX this function alone takes about 2Kbytes of code! 782 */ 783static void 784ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 785 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 786 struct ip *ip) 787{ 788 INIT_VNET_IPFW(curvnet); 789 struct ether_header *eh = args->eh; 790 char *action; 791 int limit_reached = 0; 792 char action2[40], proto[128], fragment[32]; 793 794 fragment[0] = '\0'; 795 proto[0] = '\0'; 796 797 if (f == NULL) { /* bogus pkt */ 798 if (V_verbose_limit != 0 && V_norule_counter >= V_verbose_limit) 799 return; 800 V_norule_counter++; 801 if (V_norule_counter == V_verbose_limit) 802 limit_reached = V_verbose_limit; 803 action = "Refuse"; 804 } else { /* O_LOG is the first action, find the real one */ 805 ipfw_insn *cmd = ACTION_PTR(f); 806 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 807 808 if (l->max_log != 0 && l->log_left == 0) 809 return; 810 l->log_left--; 811 if (l->log_left == 0) 812 limit_reached = l->max_log; 813 cmd += F_LEN(cmd); /* point to first action */ 814 if (cmd->opcode == O_ALTQ) { 815 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 816 817 snprintf(SNPARGS(action2, 0), "Altq %d", 818 altq->qid); 819 cmd += F_LEN(cmd); 820 } 821 if (cmd->opcode == O_PROB) 822 cmd += F_LEN(cmd); 823 824 if (cmd->opcode == O_TAG) 825 cmd += F_LEN(cmd); 826 827 action = action2; 828 switch (cmd->opcode) { 829 case O_DENY: 830 action = "Deny"; 831 break; 832 833 case O_REJECT: 834 if (cmd->arg1==ICMP_REJECT_RST) 835 action = "Reset"; 836 else if (cmd->arg1==ICMP_UNREACH_HOST) 837 action = "Reject"; 838 else 839 snprintf(SNPARGS(action2, 0), "Unreach %d", 840 cmd->arg1); 841 break; 842 843 case O_UNREACH6: 844 if (cmd->arg1==ICMP6_UNREACH_RST) 845 action = "Reset"; 846 else 847 snprintf(SNPARGS(action2, 0), "Unreach %d", 848 cmd->arg1); 849 break; 850 851 case O_ACCEPT: 852 action = "Accept"; 853 break; 854 case O_COUNT: 855 action = "Count"; 856 break; 857 case O_DIVERT: 858 snprintf(SNPARGS(action2, 0), "Divert %d", 859 cmd->arg1); 860 break; 861 case O_TEE: 862 snprintf(SNPARGS(action2, 0), "Tee %d", 863 cmd->arg1); 864 break; 865 case O_SETFIB: 866 snprintf(SNPARGS(action2, 0), "SetFib %d", 867 cmd->arg1); 868 break; 869 case O_SKIPTO: 870 snprintf(SNPARGS(action2, 0), "SkipTo %d", 871 cmd->arg1); 872 break; 873 case O_PIPE: 874 snprintf(SNPARGS(action2, 0), "Pipe %d", 875 cmd->arg1); 876 break; 877 case O_QUEUE: 878 snprintf(SNPARGS(action2, 0), "Queue %d", 879 cmd->arg1); 880 break; 881 case O_FORWARD_IP: { 882 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 883 int len; 884 struct in_addr dummyaddr; 885 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 886 dummyaddr.s_addr = htonl(tablearg); 887 else 888 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 889 890 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 891 inet_ntoa(dummyaddr)); 892 893 if (sa->sa.sin_port) 894 snprintf(SNPARGS(action2, len), ":%d", 895 sa->sa.sin_port); 896 } 897 break; 898 case O_NETGRAPH: 899 snprintf(SNPARGS(action2, 0), "Netgraph %d", 900 cmd->arg1); 901 break; 902 case O_NGTEE: 903 snprintf(SNPARGS(action2, 0), "Ngtee %d", 904 cmd->arg1); 905 break; 906 case O_NAT: 907 action = "Nat"; 908 break; 909 case O_REASS: 910 action = "Reass"; 911 break; 912 default: 913 action = "UNKNOWN"; 914 break; 915 } 916 } 917 918 if (hlen == 0) { /* non-ip */ 919 snprintf(SNPARGS(proto, 0), "MAC"); 920 921 } else { 922 int len; 923 char src[48], dst[48]; 924 struct icmphdr *icmp; 925 struct tcphdr *tcp; 926 struct udphdr *udp; 927#ifdef INET6 928 struct ip6_hdr *ip6 = NULL; 929 struct icmp6_hdr *icmp6; 930#endif 931 src[0] = '\0'; 932 dst[0] = '\0'; 933#ifdef INET6 934 if (IS_IP6_FLOW_ID(&(args->f_id))) { 935 char ip6buf[INET6_ADDRSTRLEN]; 936 snprintf(src, sizeof(src), "[%s]", 937 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 938 snprintf(dst, sizeof(dst), "[%s]", 939 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 940 941 ip6 = (struct ip6_hdr *)ip; 942 tcp = (struct tcphdr *)(((char *)ip) + hlen); 943 udp = (struct udphdr *)(((char *)ip) + hlen); 944 } else 945#endif 946 { 947 tcp = L3HDR(struct tcphdr, ip); 948 udp = L3HDR(struct udphdr, ip); 949 950 inet_ntoa_r(ip->ip_src, src); 951 inet_ntoa_r(ip->ip_dst, dst); 952 } 953 954 switch (args->f_id.proto) { 955 case IPPROTO_TCP: 956 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 957 if (offset == 0) 958 snprintf(SNPARGS(proto, len), ":%d %s:%d", 959 ntohs(tcp->th_sport), 960 dst, 961 ntohs(tcp->th_dport)); 962 else 963 snprintf(SNPARGS(proto, len), " %s", dst); 964 break; 965 966 case IPPROTO_UDP: 967 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 968 if (offset == 0) 969 snprintf(SNPARGS(proto, len), ":%d %s:%d", 970 ntohs(udp->uh_sport), 971 dst, 972 ntohs(udp->uh_dport)); 973 else 974 snprintf(SNPARGS(proto, len), " %s", dst); 975 break; 976 977 case IPPROTO_ICMP: 978 icmp = L3HDR(struct icmphdr, ip); 979 if (offset == 0) 980 len = snprintf(SNPARGS(proto, 0), 981 "ICMP:%u.%u ", 982 icmp->icmp_type, icmp->icmp_code); 983 else 984 len = snprintf(SNPARGS(proto, 0), "ICMP "); 985 len += snprintf(SNPARGS(proto, len), "%s", src); 986 snprintf(SNPARGS(proto, len), " %s", dst); 987 break; 988#ifdef INET6 989 case IPPROTO_ICMPV6: 990 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 991 if (offset == 0) 992 len = snprintf(SNPARGS(proto, 0), 993 "ICMPv6:%u.%u ", 994 icmp6->icmp6_type, icmp6->icmp6_code); 995 else 996 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 997 len += snprintf(SNPARGS(proto, len), "%s", src); 998 snprintf(SNPARGS(proto, len), " %s", dst); 999 break; 1000#endif 1001 default: 1002 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 1003 args->f_id.proto, src); 1004 snprintf(SNPARGS(proto, len), " %s", dst); 1005 break; 1006 } 1007 1008#ifdef INET6 1009 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1010 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 1011 snprintf(SNPARGS(fragment, 0), 1012 " (frag %08x:%d@%d%s)", 1013 args->f_id.frag_id6, 1014 ntohs(ip6->ip6_plen) - hlen, 1015 ntohs(offset & IP6F_OFF_MASK) << 3, 1016 (offset & IP6F_MORE_FRAG) ? "+" : ""); 1017 } else 1018#endif 1019 { 1020 int ip_off, ip_len; 1021 if (eh != NULL) { /* layer 2 packets are as on the wire */ 1022 ip_off = ntohs(ip->ip_off); 1023 ip_len = ntohs(ip->ip_len); 1024 } else { 1025 ip_off = ip->ip_off; 1026 ip_len = ip->ip_len; 1027 } 1028 if (ip_off & (IP_MF | IP_OFFMASK)) 1029 snprintf(SNPARGS(fragment, 0), 1030 " (frag %d:%d@%d%s)", 1031 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1032 offset << 3, 1033 (ip_off & IP_MF) ? "+" : ""); 1034 } 1035 } 1036 if (oif || m->m_pkthdr.rcvif) 1037 log(LOG_SECURITY | LOG_INFO, 1038 "ipfw: %d %s %s %s via %s%s\n", 1039 f ? f->rulenum : -1, 1040 action, proto, oif ? "out" : "in", 1041 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1042 fragment); 1043 else 1044 log(LOG_SECURITY | LOG_INFO, 1045 "ipfw: %d %s %s [no if info]%s\n", 1046 f ? f->rulenum : -1, 1047 action, proto, fragment); 1048 if (limit_reached) 1049 log(LOG_SECURITY | LOG_NOTICE, 1050 "ipfw: limit %d reached on entry %d\n", 1051 limit_reached, f ? f->rulenum : -1); 1052} 1053 1054/* 1055 * IMPORTANT: the hash function for dynamic rules must be commutative 1056 * in source and destination (ip,port), because rules are bidirectional 1057 * and we want to find both in the same bucket. 1058 */ 1059static __inline int 1060hash_packet(struct ipfw_flow_id *id) 1061{ 1062 INIT_VNET_IPFW(curvnet); 1063 u_int32_t i; 1064 1065#ifdef INET6 1066 if (IS_IP6_FLOW_ID(id)) 1067 i = hash_packet6(id); 1068 else 1069#endif /* INET6 */ 1070 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1071 i &= (V_curr_dyn_buckets - 1); 1072 return i; 1073} 1074 1075/** 1076 * unlink a dynamic rule from a chain. prev is a pointer to 1077 * the previous one, q is a pointer to the rule to delete, 1078 * head is a pointer to the head of the queue. 1079 * Modifies q and potentially also head. 1080 */ 1081#define UNLINK_DYN_RULE(prev, head, q) { \ 1082 ipfw_dyn_rule *old_q = q; \ 1083 \ 1084 /* remove a refcount to the parent */ \ 1085 if (q->dyn_type == O_LIMIT) \ 1086 q->parent->count--; \ 1087 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1088 (q->id.src_ip), (q->id.src_port), \ 1089 (q->id.dst_ip), (q->id.dst_port), V_dyn_count-1 ); ) \ 1090 if (prev != NULL) \ 1091 prev->next = q = q->next; \ 1092 else \ 1093 head = q = q->next; \ 1094 V_dyn_count--; \ 1095 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1096 1097#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1098 1099/** 1100 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1101 * 1102 * If keep_me == NULL, rules are deleted even if not expired, 1103 * otherwise only expired rules are removed. 1104 * 1105 * The value of the second parameter is also used to point to identify 1106 * a rule we absolutely do not want to remove (e.g. because we are 1107 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1108 * rules). The pointer is only used for comparison, so any non-null 1109 * value will do. 1110 */ 1111static void 1112remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1113{ 1114 INIT_VNET_IPFW(curvnet); 1115 static u_int32_t last_remove = 0; 1116 1117#define FORCE (keep_me == NULL) 1118 1119 ipfw_dyn_rule *prev, *q; 1120 int i, pass = 0, max_pass = 0; 1121 1122 IPFW_DYN_LOCK_ASSERT(); 1123 1124 if (V_ipfw_dyn_v == NULL || V_dyn_count == 0) 1125 return; 1126 /* do not expire more than once per second, it is useless */ 1127 if (!FORCE && last_remove == time_uptime) 1128 return; 1129 last_remove = time_uptime; 1130 1131 /* 1132 * because O_LIMIT refer to parent rules, during the first pass only 1133 * remove child and mark any pending LIMIT_PARENT, and remove 1134 * them in a second pass. 1135 */ 1136next_pass: 1137 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 1138 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q ; ) { 1139 /* 1140 * Logic can become complex here, so we split tests. 1141 */ 1142 if (q == keep_me) 1143 goto next; 1144 if (rule != NULL && rule != q->rule) 1145 goto next; /* not the one we are looking for */ 1146 if (q->dyn_type == O_LIMIT_PARENT) { 1147 /* 1148 * handle parent in the second pass, 1149 * record we need one. 1150 */ 1151 max_pass = 1; 1152 if (pass == 0) 1153 goto next; 1154 if (FORCE && q->count != 0 ) { 1155 /* XXX should not happen! */ 1156 printf("ipfw: OUCH! cannot remove rule," 1157 " count %d\n", q->count); 1158 } 1159 } else { 1160 if (!FORCE && 1161 !TIME_LEQ( q->expire, time_uptime )) 1162 goto next; 1163 } 1164 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1165 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1166 continue; 1167 } 1168next: 1169 prev=q; 1170 q=q->next; 1171 } 1172 } 1173 if (pass++ < max_pass) 1174 goto next_pass; 1175} 1176 1177 1178/** 1179 * lookup a dynamic rule. 1180 */ 1181static ipfw_dyn_rule * 1182lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1183 struct tcphdr *tcp) 1184{ 1185 INIT_VNET_IPFW(curvnet); 1186 /* 1187 * stateful ipfw extensions. 1188 * Lookup into dynamic session queue 1189 */ 1190#define MATCH_REVERSE 0 1191#define MATCH_FORWARD 1 1192#define MATCH_NONE 2 1193#define MATCH_UNKNOWN 3 1194 int i, dir = MATCH_NONE; 1195 ipfw_dyn_rule *prev, *q=NULL; 1196 1197 IPFW_DYN_LOCK_ASSERT(); 1198 1199 if (V_ipfw_dyn_v == NULL) 1200 goto done; /* not found */ 1201 i = hash_packet( pkt ); 1202 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q != NULL ; ) { 1203 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1204 goto next; 1205 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1206 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1207 continue; 1208 } 1209 if (pkt->proto == q->id.proto && 1210 q->dyn_type != O_LIMIT_PARENT) { 1211 if (IS_IP6_FLOW_ID(pkt)) { 1212 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1213 &(q->id.src_ip6)) && 1214 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1215 &(q->id.dst_ip6)) && 1216 pkt->src_port == q->id.src_port && 1217 pkt->dst_port == q->id.dst_port ) { 1218 dir = MATCH_FORWARD; 1219 break; 1220 } 1221 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1222 &(q->id.dst_ip6)) && 1223 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1224 &(q->id.src_ip6)) && 1225 pkt->src_port == q->id.dst_port && 1226 pkt->dst_port == q->id.src_port ) { 1227 dir = MATCH_REVERSE; 1228 break; 1229 } 1230 } else { 1231 if (pkt->src_ip == q->id.src_ip && 1232 pkt->dst_ip == q->id.dst_ip && 1233 pkt->src_port == q->id.src_port && 1234 pkt->dst_port == q->id.dst_port ) { 1235 dir = MATCH_FORWARD; 1236 break; 1237 } 1238 if (pkt->src_ip == q->id.dst_ip && 1239 pkt->dst_ip == q->id.src_ip && 1240 pkt->src_port == q->id.dst_port && 1241 pkt->dst_port == q->id.src_port ) { 1242 dir = MATCH_REVERSE; 1243 break; 1244 } 1245 } 1246 } 1247next: 1248 prev = q; 1249 q = q->next; 1250 } 1251 if (q == NULL) 1252 goto done; /* q = NULL, not found */ 1253 1254 if ( prev != NULL) { /* found and not in front */ 1255 prev->next = q->next; 1256 q->next = V_ipfw_dyn_v[i]; 1257 V_ipfw_dyn_v[i] = q; 1258 } 1259 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1260 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1261 1262#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1263#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1264 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1265 switch (q->state) { 1266 case TH_SYN: /* opening */ 1267 q->expire = time_uptime + V_dyn_syn_lifetime; 1268 break; 1269 1270 case BOTH_SYN: /* move to established */ 1271 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1272 case BOTH_SYN | (TH_FIN << 8) : 1273 if (tcp) { 1274#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1275 u_int32_t ack = ntohl(tcp->th_ack); 1276 if (dir == MATCH_FORWARD) { 1277 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1278 q->ack_fwd = ack; 1279 else { /* ignore out-of-sequence */ 1280 break; 1281 } 1282 } else { 1283 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1284 q->ack_rev = ack; 1285 else { /* ignore out-of-sequence */ 1286 break; 1287 } 1288 } 1289 } 1290 q->expire = time_uptime + V_dyn_ack_lifetime; 1291 break; 1292 1293 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1294 if (V_dyn_fin_lifetime >= V_dyn_keepalive_period) 1295 V_dyn_fin_lifetime = V_dyn_keepalive_period - 1; 1296 q->expire = time_uptime + V_dyn_fin_lifetime; 1297 break; 1298 1299 default: 1300#if 0 1301 /* 1302 * reset or some invalid combination, but can also 1303 * occur if we use keep-state the wrong way. 1304 */ 1305 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1306 printf("invalid state: 0x%x\n", q->state); 1307#endif 1308 if (V_dyn_rst_lifetime >= V_dyn_keepalive_period) 1309 V_dyn_rst_lifetime = V_dyn_keepalive_period - 1; 1310 q->expire = time_uptime + V_dyn_rst_lifetime; 1311 break; 1312 } 1313 } else if (pkt->proto == IPPROTO_UDP) { 1314 q->expire = time_uptime + V_dyn_udp_lifetime; 1315 } else { 1316 /* other protocols */ 1317 q->expire = time_uptime + V_dyn_short_lifetime; 1318 } 1319done: 1320 if (match_direction) 1321 *match_direction = dir; 1322 return q; 1323} 1324 1325static ipfw_dyn_rule * 1326lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1327 struct tcphdr *tcp) 1328{ 1329 ipfw_dyn_rule *q; 1330 1331 IPFW_DYN_LOCK(); 1332 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1333 if (q == NULL) 1334 IPFW_DYN_UNLOCK(); 1335 /* NB: return table locked when q is not NULL */ 1336 return q; 1337} 1338 1339static void 1340realloc_dynamic_table(void) 1341{ 1342 INIT_VNET_IPFW(curvnet); 1343 IPFW_DYN_LOCK_ASSERT(); 1344 1345 /* 1346 * Try reallocation, make sure we have a power of 2 and do 1347 * not allow more than 64k entries. In case of overflow, 1348 * default to 1024. 1349 */ 1350 1351 if (V_dyn_buckets > 65536) 1352 V_dyn_buckets = 1024; 1353 if ((V_dyn_buckets & (V_dyn_buckets-1)) != 0) { /* not a power of 2 */ 1354 V_dyn_buckets = V_curr_dyn_buckets; /* reset */ 1355 return; 1356 } 1357 V_curr_dyn_buckets = V_dyn_buckets; 1358 if (V_ipfw_dyn_v != NULL) 1359 free(V_ipfw_dyn_v, M_IPFW); 1360 for (;;) { 1361 V_ipfw_dyn_v = malloc(V_curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1362 M_IPFW, M_NOWAIT | M_ZERO); 1363 if (V_ipfw_dyn_v != NULL || V_curr_dyn_buckets <= 2) 1364 break; 1365 V_curr_dyn_buckets /= 2; 1366 } 1367} 1368 1369/** 1370 * Install state of type 'type' for a dynamic session. 1371 * The hash table contains two type of rules: 1372 * - regular rules (O_KEEP_STATE) 1373 * - rules for sessions with limited number of sess per user 1374 * (O_LIMIT). When they are created, the parent is 1375 * increased by 1, and decreased on delete. In this case, 1376 * the third parameter is the parent rule and not the chain. 1377 * - "parent" rules for the above (O_LIMIT_PARENT). 1378 */ 1379static ipfw_dyn_rule * 1380add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1381{ 1382 INIT_VNET_IPFW(curvnet); 1383 ipfw_dyn_rule *r; 1384 int i; 1385 1386 IPFW_DYN_LOCK_ASSERT(); 1387 1388 if (V_ipfw_dyn_v == NULL || 1389 (V_dyn_count == 0 && V_dyn_buckets != V_curr_dyn_buckets)) { 1390 realloc_dynamic_table(); 1391 if (V_ipfw_dyn_v == NULL) 1392 return NULL; /* failed ! */ 1393 } 1394 i = hash_packet(id); 1395 1396 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1397 if (r == NULL) { 1398 printf ("ipfw: sorry cannot allocate state\n"); 1399 return NULL; 1400 } 1401 1402 /* increase refcount on parent, and set pointer */ 1403 if (dyn_type == O_LIMIT) { 1404 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1405 if ( parent->dyn_type != O_LIMIT_PARENT) 1406 panic("invalid parent"); 1407 parent->count++; 1408 r->parent = parent; 1409 rule = parent->rule; 1410 } 1411 1412 r->id = *id; 1413 r->expire = time_uptime + V_dyn_syn_lifetime; 1414 r->rule = rule; 1415 r->dyn_type = dyn_type; 1416 r->pcnt = r->bcnt = 0; 1417 r->count = 0; 1418 1419 r->bucket = i; 1420 r->next = V_ipfw_dyn_v[i]; 1421 V_ipfw_dyn_v[i] = r; 1422 V_dyn_count++; 1423 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1424 dyn_type, 1425 (r->id.src_ip), (r->id.src_port), 1426 (r->id.dst_ip), (r->id.dst_port), 1427 V_dyn_count ); ) 1428 return r; 1429} 1430 1431/** 1432 * lookup dynamic parent rule using pkt and rule as search keys. 1433 * If the lookup fails, then install one. 1434 */ 1435static ipfw_dyn_rule * 1436lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1437{ 1438 INIT_VNET_IPFW(curvnet); 1439 ipfw_dyn_rule *q; 1440 int i; 1441 1442 IPFW_DYN_LOCK_ASSERT(); 1443 1444 if (V_ipfw_dyn_v) { 1445 int is_v6 = IS_IP6_FLOW_ID(pkt); 1446 i = hash_packet( pkt ); 1447 for (q = V_ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1448 if (q->dyn_type == O_LIMIT_PARENT && 1449 rule== q->rule && 1450 pkt->proto == q->id.proto && 1451 pkt->src_port == q->id.src_port && 1452 pkt->dst_port == q->id.dst_port && 1453 ( 1454 (is_v6 && 1455 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1456 &(q->id.src_ip6)) && 1457 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1458 &(q->id.dst_ip6))) || 1459 (!is_v6 && 1460 pkt->src_ip == q->id.src_ip && 1461 pkt->dst_ip == q->id.dst_ip) 1462 ) 1463 ) { 1464 q->expire = time_uptime + V_dyn_short_lifetime; 1465 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1466 return q; 1467 } 1468 } 1469 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1470} 1471 1472/** 1473 * Install dynamic state for rule type cmd->o.opcode 1474 * 1475 * Returns 1 (failure) if state is not installed because of errors or because 1476 * session limitations are enforced. 1477 */ 1478static int 1479install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1480 struct ip_fw_args *args, uint32_t tablearg) 1481{ 1482 INIT_VNET_IPFW(curvnet); 1483 static int last_log; 1484 ipfw_dyn_rule *q; 1485 struct in_addr da; 1486 char src[48], dst[48]; 1487 1488 src[0] = '\0'; 1489 dst[0] = '\0'; 1490 1491 DEB( 1492 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1493 __func__, cmd->o.opcode, 1494 (args->f_id.src_ip), (args->f_id.src_port), 1495 (args->f_id.dst_ip), (args->f_id.dst_port)); 1496 ) 1497 1498 IPFW_DYN_LOCK(); 1499 1500 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1501 1502 if (q != NULL) { /* should never occur */ 1503 if (last_log != time_uptime) { 1504 last_log = time_uptime; 1505 printf("ipfw: %s: entry already present, done\n", 1506 __func__); 1507 } 1508 IPFW_DYN_UNLOCK(); 1509 return (0); 1510 } 1511 1512 if (V_dyn_count >= V_dyn_max) 1513 /* Run out of slots, try to remove any expired rule. */ 1514 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1515 1516 if (V_dyn_count >= V_dyn_max) { 1517 if (last_log != time_uptime) { 1518 last_log = time_uptime; 1519 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1520 } 1521 IPFW_DYN_UNLOCK(); 1522 return (1); /* cannot install, notify caller */ 1523 } 1524 1525 switch (cmd->o.opcode) { 1526 case O_KEEP_STATE: /* bidir rule */ 1527 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1528 break; 1529 1530 case O_LIMIT: { /* limit number of sessions */ 1531 struct ipfw_flow_id id; 1532 ipfw_dyn_rule *parent; 1533 uint32_t conn_limit; 1534 uint16_t limit_mask = cmd->limit_mask; 1535 1536 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1537 tablearg : cmd->conn_limit; 1538 1539 DEB( 1540 if (cmd->conn_limit == IP_FW_TABLEARG) 1541 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1542 "(tablearg)\n", __func__, conn_limit); 1543 else 1544 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1545 __func__, conn_limit); 1546 ) 1547 1548 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1549 id.proto = args->f_id.proto; 1550 id.addr_type = args->f_id.addr_type; 1551 id.fib = M_GETFIB(args->m); 1552 1553 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1554 if (limit_mask & DYN_SRC_ADDR) 1555 id.src_ip6 = args->f_id.src_ip6; 1556 if (limit_mask & DYN_DST_ADDR) 1557 id.dst_ip6 = args->f_id.dst_ip6; 1558 } else { 1559 if (limit_mask & DYN_SRC_ADDR) 1560 id.src_ip = args->f_id.src_ip; 1561 if (limit_mask & DYN_DST_ADDR) 1562 id.dst_ip = args->f_id.dst_ip; 1563 } 1564 if (limit_mask & DYN_SRC_PORT) 1565 id.src_port = args->f_id.src_port; 1566 if (limit_mask & DYN_DST_PORT) 1567 id.dst_port = args->f_id.dst_port; 1568 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1569 printf("ipfw: %s: add parent failed\n", __func__); 1570 IPFW_DYN_UNLOCK(); 1571 return (1); 1572 } 1573 1574 if (parent->count >= conn_limit) { 1575 /* See if we can remove some expired rule. */ 1576 remove_dyn_rule(rule, parent); 1577 if (parent->count >= conn_limit) { 1578 if (V_fw_verbose && last_log != time_uptime) { 1579 last_log = time_uptime; 1580#ifdef INET6 1581 /* 1582 * XXX IPv6 flows are not 1583 * supported yet. 1584 */ 1585 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1586 char ip6buf[INET6_ADDRSTRLEN]; 1587 snprintf(src, sizeof(src), 1588 "[%s]", ip6_sprintf(ip6buf, 1589 &args->f_id.src_ip6)); 1590 snprintf(dst, sizeof(dst), 1591 "[%s]", ip6_sprintf(ip6buf, 1592 &args->f_id.dst_ip6)); 1593 } else 1594#endif 1595 { 1596 da.s_addr = 1597 htonl(args->f_id.src_ip); 1598 inet_ntoa_r(da, src); 1599 da.s_addr = 1600 htonl(args->f_id.dst_ip); 1601 inet_ntoa_r(da, dst); 1602 } 1603 log(LOG_SECURITY | LOG_DEBUG, 1604 "ipfw: %d %s %s:%u -> %s:%u, %s\n", 1605 parent->rule->rulenum, 1606 "drop session", 1607 src, (args->f_id.src_port), 1608 dst, (args->f_id.dst_port), 1609 "too many entries"); 1610 } 1611 IPFW_DYN_UNLOCK(); 1612 return (1); 1613 } 1614 } 1615 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1616 break; 1617 } 1618 default: 1619 printf("ipfw: %s: unknown dynamic rule type %u\n", 1620 __func__, cmd->o.opcode); 1621 IPFW_DYN_UNLOCK(); 1622 return (1); 1623 } 1624 1625 /* XXX just set lifetime */ 1626 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1627 1628 IPFW_DYN_UNLOCK(); 1629 return (0); 1630} 1631 1632/* 1633 * Generate a TCP packet, containing either a RST or a keepalive. 1634 * When flags & TH_RST, we are sending a RST packet, because of a 1635 * "reset" action matched the packet. 1636 * Otherwise we are sending a keepalive, and flags & TH_ 1637 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1638 * so that MAC can label the reply appropriately. 1639 */ 1640static struct mbuf * 1641send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1642 u_int32_t ack, int flags) 1643{ 1644 INIT_VNET_INET(curvnet); 1645 struct mbuf *m; 1646 struct ip *ip; 1647 struct tcphdr *tcp; 1648 1649 MGETHDR(m, M_DONTWAIT, MT_DATA); 1650 if (m == 0) 1651 return (NULL); 1652 m->m_pkthdr.rcvif = (struct ifnet *)0; 1653 1654 M_SETFIB(m, id->fib); 1655#ifdef MAC 1656 if (replyto != NULL) 1657 mac_netinet_firewall_reply(replyto, m); 1658 else 1659 mac_netinet_firewall_send(m); 1660#else 1661 (void)replyto; /* don't warn about unused arg */ 1662#endif 1663 1664 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1665 m->m_data += max_linkhdr; 1666 1667 ip = mtod(m, struct ip *); 1668 bzero(ip, m->m_len); 1669 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1670 ip->ip_p = IPPROTO_TCP; 1671 tcp->th_off = 5; 1672 /* 1673 * Assume we are sending a RST (or a keepalive in the reverse 1674 * direction), swap src and destination addresses and ports. 1675 */ 1676 ip->ip_src.s_addr = htonl(id->dst_ip); 1677 ip->ip_dst.s_addr = htonl(id->src_ip); 1678 tcp->th_sport = htons(id->dst_port); 1679 tcp->th_dport = htons(id->src_port); 1680 if (flags & TH_RST) { /* we are sending a RST */ 1681 if (flags & TH_ACK) { 1682 tcp->th_seq = htonl(ack); 1683 tcp->th_ack = htonl(0); 1684 tcp->th_flags = TH_RST; 1685 } else { 1686 if (flags & TH_SYN) 1687 seq++; 1688 tcp->th_seq = htonl(0); 1689 tcp->th_ack = htonl(seq); 1690 tcp->th_flags = TH_RST | TH_ACK; 1691 } 1692 } else { 1693 /* 1694 * We are sending a keepalive. flags & TH_SYN determines 1695 * the direction, forward if set, reverse if clear. 1696 * NOTE: seq and ack are always assumed to be correct 1697 * as set by the caller. This may be confusing... 1698 */ 1699 if (flags & TH_SYN) { 1700 /* 1701 * we have to rewrite the correct addresses! 1702 */ 1703 ip->ip_dst.s_addr = htonl(id->dst_ip); 1704 ip->ip_src.s_addr = htonl(id->src_ip); 1705 tcp->th_dport = htons(id->dst_port); 1706 tcp->th_sport = htons(id->src_port); 1707 } 1708 tcp->th_seq = htonl(seq); 1709 tcp->th_ack = htonl(ack); 1710 tcp->th_flags = TH_ACK; 1711 } 1712 /* 1713 * set ip_len to the payload size so we can compute 1714 * the tcp checksum on the pseudoheader 1715 * XXX check this, could save a couple of words ? 1716 */ 1717 ip->ip_len = htons(sizeof(struct tcphdr)); 1718 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1719 /* 1720 * now fill fields left out earlier 1721 */ 1722 ip->ip_ttl = V_ip_defttl; 1723 ip->ip_len = m->m_pkthdr.len; 1724 m->m_flags |= M_SKIP_FIREWALL; 1725 return (m); 1726} 1727 1728/* 1729 * sends a reject message, consuming the mbuf passed as an argument. 1730 */ 1731static void 1732send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1733{ 1734 1735#if 0 1736 /* XXX When ip is not guaranteed to be at mtod() we will 1737 * need to account for this */ 1738 * The mbuf will however be thrown away so we can adjust it. 1739 * Remember we did an m_pullup on it already so we 1740 * can make some assumptions about contiguousness. 1741 */ 1742 if (args->L3offset) 1743 m_adj(m, args->L3offset); 1744#endif 1745 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1746 /* We need the IP header in host order for icmp_error(). */ 1747 if (args->eh != NULL) { 1748 ip->ip_len = ntohs(ip->ip_len); 1749 ip->ip_off = ntohs(ip->ip_off); 1750 } 1751 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1752 } else if (args->f_id.proto == IPPROTO_TCP) { 1753 struct tcphdr *const tcp = 1754 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1755 if ( (tcp->th_flags & TH_RST) == 0) { 1756 struct mbuf *m; 1757 m = send_pkt(args->m, &(args->f_id), 1758 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1759 tcp->th_flags | TH_RST); 1760 if (m != NULL) 1761 ip_output(m, NULL, NULL, 0, NULL, NULL); 1762 } 1763 m_freem(args->m); 1764 } else 1765 m_freem(args->m); 1766 args->m = NULL; 1767} 1768 1769/** 1770 * 1771 * Given an ip_fw *, lookup_next_rule will return a pointer 1772 * to the next rule, which can be either the jump 1773 * target (for skipto instructions) or the next one in the list (in 1774 * all other cases including a missing jump target). 1775 * The result is also written in the "next_rule" field of the rule. 1776 * Backward jumps are not allowed, so start looking from the next 1777 * rule... 1778 * 1779 * This never returns NULL -- in case we do not have an exact match, 1780 * the next rule is returned. When the ruleset is changed, 1781 * pointers are flushed so we are always correct. 1782 */ 1783 1784static struct ip_fw * 1785lookup_next_rule(struct ip_fw *me, u_int32_t tablearg) 1786{ 1787 struct ip_fw *rule = NULL; 1788 ipfw_insn *cmd; 1789 u_int16_t rulenum; 1790 1791 /* look for action, in case it is a skipto */ 1792 cmd = ACTION_PTR(me); 1793 if (cmd->opcode == O_LOG) 1794 cmd += F_LEN(cmd); 1795 if (cmd->opcode == O_ALTQ) 1796 cmd += F_LEN(cmd); 1797 if (cmd->opcode == O_TAG) 1798 cmd += F_LEN(cmd); 1799 if (cmd->opcode == O_SKIPTO ) { 1800 if (tablearg != 0) { 1801 rulenum = (u_int16_t)tablearg; 1802 } else { 1803 rulenum = cmd->arg1; 1804 } 1805 for (rule = me->next; rule ; rule = rule->next) { 1806 if (rule->rulenum >= rulenum) { 1807 break; 1808 } 1809 } 1810 } 1811 if (rule == NULL) /* failure or not a skipto */ 1812 rule = me->next; 1813 me->next_rule = rule; 1814 return rule; 1815} 1816 1817static int 1818add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1819 uint8_t mlen, uint32_t value) 1820{ 1821 INIT_VNET_IPFW(curvnet); 1822 struct radix_node_head *rnh; 1823 struct table_entry *ent; 1824 struct radix_node *rn; 1825 1826 if (tbl >= IPFW_TABLES_MAX) 1827 return (EINVAL); 1828 rnh = ch->tables[tbl]; 1829 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1830 if (ent == NULL) 1831 return (ENOMEM); 1832 ent->value = value; 1833 ent->addr.sin_len = ent->mask.sin_len = 8; 1834 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1835 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1836 IPFW_WLOCK(ch); 1837 rn = rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent); 1838 if (rn == NULL) { 1839 IPFW_WUNLOCK(ch); 1840 free(ent, M_IPFW_TBL); 1841 return (EEXIST); 1842 } 1843 IPFW_WUNLOCK(ch); 1844 return (0); 1845} 1846 1847static int 1848del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1849 uint8_t mlen) 1850{ 1851 struct radix_node_head *rnh; 1852 struct table_entry *ent; 1853 struct sockaddr_in sa, mask; 1854 1855 if (tbl >= IPFW_TABLES_MAX) 1856 return (EINVAL); 1857 rnh = ch->tables[tbl]; 1858 sa.sin_len = mask.sin_len = 8; 1859 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1860 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1861 IPFW_WLOCK(ch); 1862 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1863 if (ent == NULL) { 1864 IPFW_WUNLOCK(ch); 1865 return (ESRCH); 1866 } 1867 IPFW_WUNLOCK(ch); 1868 free(ent, M_IPFW_TBL); 1869 return (0); 1870} 1871 1872static int 1873flush_table_entry(struct radix_node *rn, void *arg) 1874{ 1875 struct radix_node_head * const rnh = arg; 1876 struct table_entry *ent; 1877 1878 ent = (struct table_entry *) 1879 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1880 if (ent != NULL) 1881 free(ent, M_IPFW_TBL); 1882 return (0); 1883} 1884 1885static int 1886flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1887{ 1888 struct radix_node_head *rnh; 1889 1890 IPFW_WLOCK_ASSERT(ch); 1891 1892 if (tbl >= IPFW_TABLES_MAX) 1893 return (EINVAL); 1894 rnh = ch->tables[tbl]; 1895 KASSERT(rnh != NULL, ("NULL IPFW table")); 1896 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1897 return (0); 1898} 1899 1900static void 1901flush_tables(struct ip_fw_chain *ch) 1902{ 1903 uint16_t tbl; 1904 1905 IPFW_WLOCK_ASSERT(ch); 1906 1907 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1908 flush_table(ch, tbl); 1909} 1910 1911static int 1912init_tables(struct ip_fw_chain *ch) 1913{ 1914 int i; 1915 uint16_t j; 1916 1917 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1918 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1919 for (j = 0; j < i; j++) { 1920 (void) flush_table(ch, j); 1921 } 1922 return (ENOMEM); 1923 } 1924 } 1925 return (0); 1926} 1927 1928static int 1929lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1930 uint32_t *val) 1931{ 1932 struct radix_node_head *rnh; 1933 struct table_entry *ent; 1934 struct sockaddr_in sa; 1935 1936 if (tbl >= IPFW_TABLES_MAX) 1937 return (0); 1938 rnh = ch->tables[tbl]; 1939 sa.sin_len = 8; 1940 sa.sin_addr.s_addr = addr; 1941 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1942 if (ent != NULL) { 1943 *val = ent->value; 1944 return (1); 1945 } 1946 return (0); 1947} 1948 1949static int 1950count_table_entry(struct radix_node *rn, void *arg) 1951{ 1952 u_int32_t * const cnt = arg; 1953 1954 (*cnt)++; 1955 return (0); 1956} 1957 1958static int 1959count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1960{ 1961 struct radix_node_head *rnh; 1962 1963 if (tbl >= IPFW_TABLES_MAX) 1964 return (EINVAL); 1965 rnh = ch->tables[tbl]; 1966 *cnt = 0; 1967 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1968 return (0); 1969} 1970 1971static int 1972dump_table_entry(struct radix_node *rn, void *arg) 1973{ 1974 struct table_entry * const n = (struct table_entry *)rn; 1975 ipfw_table * const tbl = arg; 1976 ipfw_table_entry *ent; 1977 1978 if (tbl->cnt == tbl->size) 1979 return (1); 1980 ent = &tbl->ent[tbl->cnt]; 1981 ent->tbl = tbl->tbl; 1982 if (in_nullhost(n->mask.sin_addr)) 1983 ent->masklen = 0; 1984 else 1985 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1986 ent->addr = n->addr.sin_addr.s_addr; 1987 ent->value = n->value; 1988 tbl->cnt++; 1989 return (0); 1990} 1991 1992static int 1993dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1994{ 1995 struct radix_node_head *rnh; 1996 1997 if (tbl->tbl >= IPFW_TABLES_MAX) 1998 return (EINVAL); 1999 rnh = ch->tables[tbl->tbl]; 2000 tbl->cnt = 0; 2001 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 2002 return (0); 2003} 2004 2005static void 2006fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 2007{ 2008 struct ucred *cr; 2009 2010 cr = inp->inp_cred; 2011 ugp->fw_prid = jailed(cr) ? cr->cr_prison->pr_id : -1; 2012 ugp->fw_uid = cr->cr_uid; 2013 ugp->fw_ngroups = cr->cr_ngroups; 2014 bcopy(cr->cr_groups, ugp->fw_groups, sizeof(ugp->fw_groups)); 2015} 2016 2017static int 2018check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 2019 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 2020 u_int16_t src_port, struct ip_fw_ugid *ugp, int *ugid_lookupp, 2021 struct inpcb *inp) 2022{ 2023 INIT_VNET_INET(curvnet); 2024 struct inpcbinfo *pi; 2025 int wildcard; 2026 struct inpcb *pcb; 2027 int match; 2028 gid_t *gp; 2029 2030 /* 2031 * Check to see if the UDP or TCP stack supplied us with 2032 * the PCB. If so, rather then holding a lock and looking 2033 * up the PCB, we can use the one that was supplied. 2034 */ 2035 if (inp && *ugid_lookupp == 0) { 2036 INP_LOCK_ASSERT(inp); 2037 if (inp->inp_socket != NULL) { 2038 fill_ugid_cache(inp, ugp); 2039 *ugid_lookupp = 1; 2040 } else 2041 *ugid_lookupp = -1; 2042 } 2043 /* 2044 * If we have already been here and the packet has no 2045 * PCB entry associated with it, then we can safely 2046 * assume that this is a no match. 2047 */ 2048 if (*ugid_lookupp == -1) 2049 return (0); 2050 if (proto == IPPROTO_TCP) { 2051 wildcard = 0; 2052 pi = &V_tcbinfo; 2053 } else if (proto == IPPROTO_UDP) { 2054 wildcard = INPLOOKUP_WILDCARD; 2055 pi = &V_udbinfo; 2056 } else 2057 return 0; 2058 match = 0; 2059 if (*ugid_lookupp == 0) { 2060 INP_INFO_RLOCK(pi); 2061 pcb = (oif) ? 2062 in_pcblookup_hash(pi, 2063 dst_ip, htons(dst_port), 2064 src_ip, htons(src_port), 2065 wildcard, oif) : 2066 in_pcblookup_hash(pi, 2067 src_ip, htons(src_port), 2068 dst_ip, htons(dst_port), 2069 wildcard, NULL); 2070 if (pcb != NULL) { 2071 fill_ugid_cache(pcb, ugp); 2072 *ugid_lookupp = 1; 2073 } 2074 INP_INFO_RUNLOCK(pi); 2075 if (*ugid_lookupp == 0) { 2076 /* 2077 * If the lookup did not yield any results, there 2078 * is no sense in coming back and trying again. So 2079 * we can set lookup to -1 and ensure that we wont 2080 * bother the pcb system again. 2081 */ 2082 *ugid_lookupp = -1; 2083 return (0); 2084 } 2085 } 2086 if (insn->o.opcode == O_UID) 2087 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2088 else if (insn->o.opcode == O_GID) { 2089 for (gp = ugp->fw_groups; 2090 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2091 if (*gp == (gid_t)insn->d[0]) { 2092 match = 1; 2093 break; 2094 } 2095 } else if (insn->o.opcode == O_JAIL) 2096 match = (ugp->fw_prid == (int)insn->d[0]); 2097 return match; 2098} 2099 2100/* 2101 * The main check routine for the firewall. 2102 * 2103 * All arguments are in args so we can modify them and return them 2104 * back to the caller. 2105 * 2106 * Parameters: 2107 * 2108 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2109 * Starts with the IP header. 2110 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2111 * args->L3offset Number of bytes bypassed if we came from L2. 2112 * e.g. often sizeof(eh) ** NOTYET ** 2113 * args->oif Outgoing interface, or NULL if packet is incoming. 2114 * The incoming interface is in the mbuf. (in) 2115 * args->divert_rule (in/out) 2116 * Skip up to the first rule past this rule number; 2117 * upon return, non-zero port number for divert or tee. 2118 * 2119 * args->rule Pointer to the last matching rule (in/out) 2120 * args->next_hop Socket we are forwarding to (out). 2121 * args->f_id Addresses grabbed from the packet (out) 2122 * args->cookie a cookie depending on rule action 2123 * 2124 * Return value: 2125 * 2126 * IP_FW_PASS the packet must be accepted 2127 * IP_FW_DENY the packet must be dropped 2128 * IP_FW_DIVERT divert packet, port in m_tag 2129 * IP_FW_TEE tee packet, port in m_tag 2130 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2131 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2132 * 2133 */ 2134int 2135ipfw_chk(struct ip_fw_args *args) 2136{ 2137 INIT_VNET_INET(curvnet); 2138 INIT_VNET_IPFW(curvnet); 2139 2140 /* 2141 * Local variables holding state during the processing of a packet: 2142 * 2143 * IMPORTANT NOTE: to speed up the processing of rules, there 2144 * are some assumption on the values of the variables, which 2145 * are documented here. Should you change them, please check 2146 * the implementation of the various instructions to make sure 2147 * that they still work. 2148 * 2149 * args->eh The MAC header. It is non-null for a layer2 2150 * packet, it is NULL for a layer-3 packet. 2151 * **notyet** 2152 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2153 * 2154 * m | args->m Pointer to the mbuf, as received from the caller. 2155 * It may change if ipfw_chk() does an m_pullup, or if it 2156 * consumes the packet because it calls send_reject(). 2157 * XXX This has to change, so that ipfw_chk() never modifies 2158 * or consumes the buffer. 2159 * ip is the beginning of the ip(4 or 6) header. 2160 * Calculated by adding the L3offset to the start of data. 2161 * (Until we start using L3offset, the packet is 2162 * supposed to start with the ip header). 2163 */ 2164 struct mbuf *m = args->m; 2165 struct ip *ip = mtod(m, struct ip *); 2166 2167 /* 2168 * For rules which contain uid/gid or jail constraints, cache 2169 * a copy of the users credentials after the pcb lookup has been 2170 * executed. This will speed up the processing of rules with 2171 * these types of constraints, as well as decrease contention 2172 * on pcb related locks. 2173 */ 2174 struct ip_fw_ugid fw_ugid_cache; 2175 int ugid_lookup = 0; 2176 2177 /* 2178 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2179 * associated with a packet input on a divert socket. This 2180 * will allow to distinguish traffic and its direction when 2181 * it originates from a divert socket. 2182 */ 2183 u_int divinput_flags = 0; 2184 2185 /* 2186 * oif | args->oif If NULL, ipfw_chk has been called on the 2187 * inbound path (ether_input, ip_input). 2188 * If non-NULL, ipfw_chk has been called on the outbound path 2189 * (ether_output, ip_output). 2190 */ 2191 struct ifnet *oif = args->oif; 2192 2193 struct ip_fw *f = NULL; /* matching rule */ 2194 int retval = 0; 2195 2196 /* 2197 * hlen The length of the IP header. 2198 */ 2199 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2200 2201 /* 2202 * offset The offset of a fragment. offset != 0 means that 2203 * we have a fragment at this offset of an IPv4 packet. 2204 * offset == 0 means that (if this is an IPv4 packet) 2205 * this is the first or only fragment. 2206 * For IPv6 offset == 0 means there is no Fragment Header. 2207 * If offset != 0 for IPv6 always use correct mask to 2208 * get the correct offset because we add IP6F_MORE_FRAG 2209 * to be able to dectect the first fragment which would 2210 * otherwise have offset = 0. 2211 */ 2212 u_short offset = 0; 2213 2214 /* 2215 * Local copies of addresses. They are only valid if we have 2216 * an IP packet. 2217 * 2218 * proto The protocol. Set to 0 for non-ip packets, 2219 * or to the protocol read from the packet otherwise. 2220 * proto != 0 means that we have an IPv4 packet. 2221 * 2222 * src_port, dst_port port numbers, in HOST format. Only 2223 * valid for TCP and UDP packets. 2224 * 2225 * src_ip, dst_ip ip addresses, in NETWORK format. 2226 * Only valid for IPv4 packets. 2227 */ 2228 u_int8_t proto; 2229 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2230 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2231 u_int16_t ip_len=0; 2232 int pktlen; 2233 u_int16_t etype = 0; /* Host order stored ether type */ 2234 2235 /* 2236 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2237 * MATCH_NONE when checked and not matched (q = NULL), 2238 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2239 */ 2240 int dyn_dir = MATCH_UNKNOWN; 2241 ipfw_dyn_rule *q = NULL; 2242 struct ip_fw_chain *chain = &V_layer3_chain; 2243 struct m_tag *mtag; 2244 2245 /* 2246 * We store in ulp a pointer to the upper layer protocol header. 2247 * In the ipv4 case this is easy to determine from the header, 2248 * but for ipv6 we might have some additional headers in the middle. 2249 * ulp is NULL if not found. 2250 */ 2251 void *ulp = NULL; /* upper layer protocol pointer. */ 2252 /* XXX ipv6 variables */ 2253 int is_ipv6 = 0; 2254 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2255 /* end of ipv6 variables */ 2256 int is_ipv4 = 0; 2257 2258 if (m->m_flags & M_SKIP_FIREWALL) 2259 return (IP_FW_PASS); /* accept */ 2260 2261 dst_ip.s_addr = 0; /* make sure it is initialized */ 2262 pktlen = m->m_pkthdr.len; 2263 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 2264 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2265 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2266 2267/* 2268 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2269 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2270 * pointer might become stale after other pullups (but we never use it 2271 * this way). 2272 */ 2273#define PULLUP_TO(len, p, T) \ 2274do { \ 2275 int x = (len) + sizeof(T); \ 2276 if ((m)->m_len < x) { \ 2277 args->m = m = m_pullup(m, x); \ 2278 if (m == NULL) \ 2279 goto pullup_failed; \ 2280 } \ 2281 p = (mtod(m, char *) + (len)); \ 2282} while (0) 2283 2284 /* 2285 * if we have an ether header, 2286 */ 2287 if (args->eh) 2288 etype = ntohs(args->eh->ether_type); 2289 2290 /* Identify IP packets and fill up variables. */ 2291 if (pktlen >= sizeof(struct ip6_hdr) && 2292 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2293 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2294 is_ipv6 = 1; 2295 args->f_id.addr_type = 6; 2296 hlen = sizeof(struct ip6_hdr); 2297 proto = ip6->ip6_nxt; 2298 2299 /* Search extension headers to find upper layer protocols */ 2300 while (ulp == NULL) { 2301 switch (proto) { 2302 case IPPROTO_ICMPV6: 2303 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2304 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2305 break; 2306 2307 case IPPROTO_TCP: 2308 PULLUP_TO(hlen, ulp, struct tcphdr); 2309 dst_port = TCP(ulp)->th_dport; 2310 src_port = TCP(ulp)->th_sport; 2311 args->f_id.flags = TCP(ulp)->th_flags; 2312 break; 2313 2314 case IPPROTO_SCTP: 2315 PULLUP_TO(hlen, ulp, struct sctphdr); 2316 src_port = SCTP(ulp)->src_port; 2317 dst_port = SCTP(ulp)->dest_port; 2318 break; 2319 2320 case IPPROTO_UDP: 2321 PULLUP_TO(hlen, ulp, struct udphdr); 2322 dst_port = UDP(ulp)->uh_dport; 2323 src_port = UDP(ulp)->uh_sport; 2324 break; 2325 2326 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2327 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2328 ext_hd |= EXT_HOPOPTS; 2329 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2330 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2331 ulp = NULL; 2332 break; 2333 2334 case IPPROTO_ROUTING: /* RFC 2460 */ 2335 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2336 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2337 case 0: 2338 ext_hd |= EXT_RTHDR0; 2339 break; 2340 case 2: 2341 ext_hd |= EXT_RTHDR2; 2342 break; 2343 default: 2344 printf("IPFW2: IPV6 - Unknown Routing " 2345 "Header type(%d)\n", 2346 ((struct ip6_rthdr *)ulp)->ip6r_type); 2347 if (V_fw_deny_unknown_exthdrs) 2348 return (IP_FW_DENY); 2349 break; 2350 } 2351 ext_hd |= EXT_ROUTING; 2352 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2353 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2354 ulp = NULL; 2355 break; 2356 2357 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2358 PULLUP_TO(hlen, ulp, struct ip6_frag); 2359 ext_hd |= EXT_FRAGMENT; 2360 hlen += sizeof (struct ip6_frag); 2361 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2362 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2363 IP6F_OFF_MASK; 2364 /* Add IP6F_MORE_FRAG for offset of first 2365 * fragment to be != 0. */ 2366 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2367 IP6F_MORE_FRAG; 2368 if (offset == 0) { 2369 printf("IPFW2: IPV6 - Invalid Fragment " 2370 "Header\n"); 2371 if (V_fw_deny_unknown_exthdrs) 2372 return (IP_FW_DENY); 2373 break; 2374 } 2375 args->f_id.frag_id6 = 2376 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2377 ulp = NULL; 2378 break; 2379 2380 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2381 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2382 ext_hd |= EXT_DSTOPTS; 2383 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2384 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2385 ulp = NULL; 2386 break; 2387 2388 case IPPROTO_AH: /* RFC 2402 */ 2389 PULLUP_TO(hlen, ulp, struct ip6_ext); 2390 ext_hd |= EXT_AH; 2391 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2392 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2393 ulp = NULL; 2394 break; 2395 2396 case IPPROTO_ESP: /* RFC 2406 */ 2397 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2398 /* Anything past Seq# is variable length and 2399 * data past this ext. header is encrypted. */ 2400 ext_hd |= EXT_ESP; 2401 break; 2402 2403 case IPPROTO_NONE: /* RFC 2460 */ 2404 /* 2405 * Packet ends here, and IPv6 header has 2406 * already been pulled up. If ip6e_len!=0 2407 * then octets must be ignored. 2408 */ 2409 ulp = ip; /* non-NULL to get out of loop. */ 2410 break; 2411 2412 case IPPROTO_OSPFIGP: 2413 /* XXX OSPF header check? */ 2414 PULLUP_TO(hlen, ulp, struct ip6_ext); 2415 break; 2416 2417 case IPPROTO_PIM: 2418 /* XXX PIM header check? */ 2419 PULLUP_TO(hlen, ulp, struct pim); 2420 break; 2421 2422 case IPPROTO_CARP: 2423 PULLUP_TO(hlen, ulp, struct carp_header); 2424 if (((struct carp_header *)ulp)->carp_version != 2425 CARP_VERSION) 2426 return (IP_FW_DENY); 2427 if (((struct carp_header *)ulp)->carp_type != 2428 CARP_ADVERTISEMENT) 2429 return (IP_FW_DENY); 2430 break; 2431 2432 case IPPROTO_IPV6: /* RFC 2893 */ 2433 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2434 break; 2435 2436 case IPPROTO_IPV4: /* RFC 2893 */ 2437 PULLUP_TO(hlen, ulp, struct ip); 2438 break; 2439 2440 default: 2441 printf("IPFW2: IPV6 - Unknown Extension " 2442 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2443 if (V_fw_deny_unknown_exthdrs) 2444 return (IP_FW_DENY); 2445 PULLUP_TO(hlen, ulp, struct ip6_ext); 2446 break; 2447 } /*switch */ 2448 } 2449 ip = mtod(m, struct ip *); 2450 ip6 = (struct ip6_hdr *)ip; 2451 args->f_id.src_ip6 = ip6->ip6_src; 2452 args->f_id.dst_ip6 = ip6->ip6_dst; 2453 args->f_id.src_ip = 0; 2454 args->f_id.dst_ip = 0; 2455 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2456 } else if (pktlen >= sizeof(struct ip) && 2457 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2458 is_ipv4 = 1; 2459 hlen = ip->ip_hl << 2; 2460 args->f_id.addr_type = 4; 2461 2462 /* 2463 * Collect parameters into local variables for faster matching. 2464 */ 2465 proto = ip->ip_p; 2466 src_ip = ip->ip_src; 2467 dst_ip = ip->ip_dst; 2468 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2469 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2470 ip_len = ntohs(ip->ip_len); 2471 } else { 2472 offset = ip->ip_off & IP_OFFMASK; 2473 ip_len = ip->ip_len; 2474 } 2475 pktlen = ip_len < pktlen ? ip_len : pktlen; 2476 2477 if (offset == 0) { 2478 switch (proto) { 2479 case IPPROTO_TCP: 2480 PULLUP_TO(hlen, ulp, struct tcphdr); 2481 dst_port = TCP(ulp)->th_dport; 2482 src_port = TCP(ulp)->th_sport; 2483 args->f_id.flags = TCP(ulp)->th_flags; 2484 break; 2485 2486 case IPPROTO_UDP: 2487 PULLUP_TO(hlen, ulp, struct udphdr); 2488 dst_port = UDP(ulp)->uh_dport; 2489 src_port = UDP(ulp)->uh_sport; 2490 break; 2491 2492 case IPPROTO_ICMP: 2493 PULLUP_TO(hlen, ulp, struct icmphdr); 2494 args->f_id.flags = ICMP(ulp)->icmp_type; 2495 break; 2496 2497 default: 2498 break; 2499 } 2500 } 2501 2502 ip = mtod(m, struct ip *); 2503 args->f_id.src_ip = ntohl(src_ip.s_addr); 2504 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2505 } 2506#undef PULLUP_TO 2507 if (proto) { /* we may have port numbers, store them */ 2508 args->f_id.proto = proto; 2509 args->f_id.src_port = src_port = ntohs(src_port); 2510 args->f_id.dst_port = dst_port = ntohs(dst_port); 2511 } 2512 2513 IPFW_RLOCK(chain); 2514 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2515 if (args->rule) { 2516 /* 2517 * Packet has already been tagged. Look for the next rule 2518 * to restart processing. 2519 * 2520 * If fw_one_pass != 0 then just accept it. 2521 * XXX should not happen here, but optimized out in 2522 * the caller. 2523 */ 2524 if (V_fw_one_pass) { 2525 IPFW_RUNLOCK(chain); 2526 return (IP_FW_PASS); 2527 } 2528 2529 f = args->rule->next_rule; 2530 if (f == NULL) 2531 f = lookup_next_rule(args->rule, 0); 2532 } else { 2533 /* 2534 * Find the starting rule. It can be either the first 2535 * one, or the one after divert_rule if asked so. 2536 */ 2537 int skipto = mtag ? divert_cookie(mtag) : 0; 2538 2539 f = chain->rules; 2540 if (args->eh == NULL && skipto != 0) { 2541 if (skipto >= IPFW_DEFAULT_RULE) { 2542 IPFW_RUNLOCK(chain); 2543 return (IP_FW_DENY); /* invalid */ 2544 } 2545 while (f && f->rulenum <= skipto) 2546 f = f->next; 2547 if (f == NULL) { /* drop packet */ 2548 IPFW_RUNLOCK(chain); 2549 return (IP_FW_DENY); 2550 } 2551 } 2552 } 2553 /* reset divert rule to avoid confusion later */ 2554 if (mtag) { 2555 divinput_flags = divert_info(mtag) & 2556 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2557 m_tag_delete(m, mtag); 2558 } 2559 2560 /* 2561 * Now scan the rules, and parse microinstructions for each rule. 2562 */ 2563 for (; f; f = f->next) { 2564 ipfw_insn *cmd; 2565 uint32_t tablearg = 0; 2566 int l, cmdlen, skip_or; /* skip rest of OR block */ 2567 2568again: 2569 if (V_set_disable & (1 << f->set) ) 2570 continue; 2571 2572 skip_or = 0; 2573 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2574 l -= cmdlen, cmd += cmdlen) { 2575 int match; 2576 2577 /* 2578 * check_body is a jump target used when we find a 2579 * CHECK_STATE, and need to jump to the body of 2580 * the target rule. 2581 */ 2582 2583check_body: 2584 cmdlen = F_LEN(cmd); 2585 /* 2586 * An OR block (insn_1 || .. || insn_n) has the 2587 * F_OR bit set in all but the last instruction. 2588 * The first match will set "skip_or", and cause 2589 * the following instructions to be skipped until 2590 * past the one with the F_OR bit clear. 2591 */ 2592 if (skip_or) { /* skip this instruction */ 2593 if ((cmd->len & F_OR) == 0) 2594 skip_or = 0; /* next one is good */ 2595 continue; 2596 } 2597 match = 0; /* set to 1 if we succeed */ 2598 2599 switch (cmd->opcode) { 2600 /* 2601 * The first set of opcodes compares the packet's 2602 * fields with some pattern, setting 'match' if a 2603 * match is found. At the end of the loop there is 2604 * logic to deal with F_NOT and F_OR flags associated 2605 * with the opcode. 2606 */ 2607 case O_NOP: 2608 match = 1; 2609 break; 2610 2611 case O_FORWARD_MAC: 2612 printf("ipfw: opcode %d unimplemented\n", 2613 cmd->opcode); 2614 break; 2615 2616 case O_GID: 2617 case O_UID: 2618 case O_JAIL: 2619 /* 2620 * We only check offset == 0 && proto != 0, 2621 * as this ensures that we have a 2622 * packet with the ports info. 2623 */ 2624 if (offset!=0) 2625 break; 2626 if (is_ipv6) /* XXX to be fixed later */ 2627 break; 2628 if (proto == IPPROTO_TCP || 2629 proto == IPPROTO_UDP) 2630 match = check_uidgid( 2631 (ipfw_insn_u32 *)cmd, 2632 proto, oif, 2633 dst_ip, dst_port, 2634 src_ip, src_port, &fw_ugid_cache, 2635 &ugid_lookup, args->inp); 2636 break; 2637 2638 case O_RECV: 2639 match = iface_match(m->m_pkthdr.rcvif, 2640 (ipfw_insn_if *)cmd); 2641 break; 2642 2643 case O_XMIT: 2644 match = iface_match(oif, (ipfw_insn_if *)cmd); 2645 break; 2646 2647 case O_VIA: 2648 match = iface_match(oif ? oif : 2649 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2650 break; 2651 2652 case O_MACADDR2: 2653 if (args->eh != NULL) { /* have MAC header */ 2654 u_int32_t *want = (u_int32_t *) 2655 ((ipfw_insn_mac *)cmd)->addr; 2656 u_int32_t *mask = (u_int32_t *) 2657 ((ipfw_insn_mac *)cmd)->mask; 2658 u_int32_t *hdr = (u_int32_t *)args->eh; 2659 2660 match = 2661 ( want[0] == (hdr[0] & mask[0]) && 2662 want[1] == (hdr[1] & mask[1]) && 2663 want[2] == (hdr[2] & mask[2]) ); 2664 } 2665 break; 2666 2667 case O_MAC_TYPE: 2668 if (args->eh != NULL) { 2669 u_int16_t *p = 2670 ((ipfw_insn_u16 *)cmd)->ports; 2671 int i; 2672 2673 for (i = cmdlen - 1; !match && i>0; 2674 i--, p += 2) 2675 match = (etype >= p[0] && 2676 etype <= p[1]); 2677 } 2678 break; 2679 2680 case O_FRAG: 2681 match = (offset != 0); 2682 break; 2683 2684 case O_IN: /* "out" is "not in" */ 2685 match = (oif == NULL); 2686 break; 2687 2688 case O_LAYER2: 2689 match = (args->eh != NULL); 2690 break; 2691 2692 case O_DIVERTED: 2693 match = (cmd->arg1 & 1 && divinput_flags & 2694 IP_FW_DIVERT_LOOPBACK_FLAG) || 2695 (cmd->arg1 & 2 && divinput_flags & 2696 IP_FW_DIVERT_OUTPUT_FLAG); 2697 break; 2698 2699 case O_PROTO: 2700 /* 2701 * We do not allow an arg of 0 so the 2702 * check of "proto" only suffices. 2703 */ 2704 match = (proto == cmd->arg1); 2705 break; 2706 2707 case O_IP_SRC: 2708 match = is_ipv4 && 2709 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2710 src_ip.s_addr); 2711 break; 2712 2713 case O_IP_SRC_LOOKUP: 2714 case O_IP_DST_LOOKUP: 2715 if (is_ipv4) { 2716 uint32_t a = 2717 (cmd->opcode == O_IP_DST_LOOKUP) ? 2718 dst_ip.s_addr : src_ip.s_addr; 2719 uint32_t v = 0; 2720 2721 match = lookup_table(chain, cmd->arg1, a, 2722 &v); 2723 if (!match) 2724 break; 2725 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2726 match = 2727 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2728 else 2729 tablearg = v; 2730 } 2731 break; 2732 2733 case O_IP_SRC_MASK: 2734 case O_IP_DST_MASK: 2735 if (is_ipv4) { 2736 uint32_t a = 2737 (cmd->opcode == O_IP_DST_MASK) ? 2738 dst_ip.s_addr : src_ip.s_addr; 2739 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2740 int i = cmdlen-1; 2741 2742 for (; !match && i>0; i-= 2, p+= 2) 2743 match = (p[0] == (a & p[1])); 2744 } 2745 break; 2746 2747 case O_IP_SRC_ME: 2748 if (is_ipv4) { 2749 struct ifnet *tif; 2750 2751 INADDR_TO_IFP(src_ip, tif); 2752 match = (tif != NULL); 2753 } 2754 break; 2755 2756 case O_IP_DST_SET: 2757 case O_IP_SRC_SET: 2758 if (is_ipv4) { 2759 u_int32_t *d = (u_int32_t *)(cmd+1); 2760 u_int32_t addr = 2761 cmd->opcode == O_IP_DST_SET ? 2762 args->f_id.dst_ip : 2763 args->f_id.src_ip; 2764 2765 if (addr < d[0]) 2766 break; 2767 addr -= d[0]; /* subtract base */ 2768 match = (addr < cmd->arg1) && 2769 ( d[ 1 + (addr>>5)] & 2770 (1<<(addr & 0x1f)) ); 2771 } 2772 break; 2773 2774 case O_IP_DST: 2775 match = is_ipv4 && 2776 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2777 dst_ip.s_addr); 2778 break; 2779 2780 case O_IP_DST_ME: 2781 if (is_ipv4) { 2782 struct ifnet *tif; 2783 2784 INADDR_TO_IFP(dst_ip, tif); 2785 match = (tif != NULL); 2786 } 2787 break; 2788 2789 case O_IP_SRCPORT: 2790 case O_IP_DSTPORT: 2791 /* 2792 * offset == 0 && proto != 0 is enough 2793 * to guarantee that we have a 2794 * packet with port info. 2795 */ 2796 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2797 && offset == 0) { 2798 u_int16_t x = 2799 (cmd->opcode == O_IP_SRCPORT) ? 2800 src_port : dst_port ; 2801 u_int16_t *p = 2802 ((ipfw_insn_u16 *)cmd)->ports; 2803 int i; 2804 2805 for (i = cmdlen - 1; !match && i>0; 2806 i--, p += 2) 2807 match = (x>=p[0] && x<=p[1]); 2808 } 2809 break; 2810 2811 case O_ICMPTYPE: 2812 match = (offset == 0 && proto==IPPROTO_ICMP && 2813 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2814 break; 2815 2816#ifdef INET6 2817 case O_ICMP6TYPE: 2818 match = is_ipv6 && offset == 0 && 2819 proto==IPPROTO_ICMPV6 && 2820 icmp6type_match( 2821 ICMP6(ulp)->icmp6_type, 2822 (ipfw_insn_u32 *)cmd); 2823 break; 2824#endif /* INET6 */ 2825 2826 case O_IPOPT: 2827 match = (is_ipv4 && 2828 ipopts_match(ip, cmd) ); 2829 break; 2830 2831 case O_IPVER: 2832 match = (is_ipv4 && 2833 cmd->arg1 == ip->ip_v); 2834 break; 2835 2836 case O_IPID: 2837 case O_IPLEN: 2838 case O_IPTTL: 2839 if (is_ipv4) { /* only for IP packets */ 2840 uint16_t x; 2841 uint16_t *p; 2842 int i; 2843 2844 if (cmd->opcode == O_IPLEN) 2845 x = ip_len; 2846 else if (cmd->opcode == O_IPTTL) 2847 x = ip->ip_ttl; 2848 else /* must be IPID */ 2849 x = ntohs(ip->ip_id); 2850 if (cmdlen == 1) { 2851 match = (cmd->arg1 == x); 2852 break; 2853 } 2854 /* otherwise we have ranges */ 2855 p = ((ipfw_insn_u16 *)cmd)->ports; 2856 i = cmdlen - 1; 2857 for (; !match && i>0; i--, p += 2) 2858 match = (x >= p[0] && x <= p[1]); 2859 } 2860 break; 2861 2862 case O_IPPRECEDENCE: 2863 match = (is_ipv4 && 2864 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2865 break; 2866 2867 case O_IPTOS: 2868 match = (is_ipv4 && 2869 flags_match(cmd, ip->ip_tos)); 2870 break; 2871 2872 case O_TCPDATALEN: 2873 if (proto == IPPROTO_TCP && offset == 0) { 2874 struct tcphdr *tcp; 2875 uint16_t x; 2876 uint16_t *p; 2877 int i; 2878 2879 tcp = TCP(ulp); 2880 x = ip_len - 2881 ((ip->ip_hl + tcp->th_off) << 2); 2882 if (cmdlen == 1) { 2883 match = (cmd->arg1 == x); 2884 break; 2885 } 2886 /* otherwise we have ranges */ 2887 p = ((ipfw_insn_u16 *)cmd)->ports; 2888 i = cmdlen - 1; 2889 for (; !match && i>0; i--, p += 2) 2890 match = (x >= p[0] && x <= p[1]); 2891 } 2892 break; 2893 2894 case O_TCPFLAGS: 2895 match = (proto == IPPROTO_TCP && offset == 0 && 2896 flags_match(cmd, TCP(ulp)->th_flags)); 2897 break; 2898 2899 case O_TCPOPTS: 2900 match = (proto == IPPROTO_TCP && offset == 0 && 2901 tcpopts_match(TCP(ulp), cmd)); 2902 break; 2903 2904 case O_TCPSEQ: 2905 match = (proto == IPPROTO_TCP && offset == 0 && 2906 ((ipfw_insn_u32 *)cmd)->d[0] == 2907 TCP(ulp)->th_seq); 2908 break; 2909 2910 case O_TCPACK: 2911 match = (proto == IPPROTO_TCP && offset == 0 && 2912 ((ipfw_insn_u32 *)cmd)->d[0] == 2913 TCP(ulp)->th_ack); 2914 break; 2915 2916 case O_TCPWIN: 2917 match = (proto == IPPROTO_TCP && offset == 0 && 2918 cmd->arg1 == TCP(ulp)->th_win); 2919 break; 2920 2921 case O_ESTAB: 2922 /* reject packets which have SYN only */ 2923 /* XXX should i also check for TH_ACK ? */ 2924 match = (proto == IPPROTO_TCP && offset == 0 && 2925 (TCP(ulp)->th_flags & 2926 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2927 break; 2928 2929 case O_ALTQ: { 2930 struct pf_mtag *at; 2931 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2932 2933 match = 1; 2934 at = pf_find_mtag(m); 2935 if (at != NULL && at->qid != 0) 2936 break; 2937 at = pf_get_mtag(m); 2938 if (at == NULL) { 2939 /* 2940 * Let the packet fall back to the 2941 * default ALTQ. 2942 */ 2943 break; 2944 } 2945 at->qid = altq->qid; 2946 if (is_ipv4) 2947 at->af = AF_INET; 2948 else 2949 at->af = AF_LINK; 2950 at->hdr = ip; 2951 break; 2952 } 2953 2954 case O_LOG: 2955 if (V_fw_verbose) 2956 ipfw_log(f, hlen, args, m, 2957 oif, offset, tablearg, ip); 2958 match = 1; 2959 break; 2960 2961 case O_PROB: 2962 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2963 break; 2964 2965 case O_VERREVPATH: 2966 /* Outgoing packets automatically pass/match */ 2967 match = ((oif != NULL) || 2968 (m->m_pkthdr.rcvif == NULL) || 2969 ( 2970#ifdef INET6 2971 is_ipv6 ? 2972 verify_path6(&(args->f_id.src_ip6), 2973 m->m_pkthdr.rcvif) : 2974#endif 2975 verify_path(src_ip, m->m_pkthdr.rcvif, 2976 args->f_id.fib))); 2977 break; 2978 2979 case O_VERSRCREACH: 2980 /* Outgoing packets automatically pass/match */ 2981 match = (hlen > 0 && ((oif != NULL) || 2982#ifdef INET6 2983 is_ipv6 ? 2984 verify_path6(&(args->f_id.src_ip6), 2985 NULL) : 2986#endif 2987 verify_path(src_ip, NULL, args->f_id.fib))); 2988 break; 2989 2990 case O_ANTISPOOF: 2991 /* Outgoing packets automatically pass/match */ 2992 if (oif == NULL && hlen > 0 && 2993 ( (is_ipv4 && in_localaddr(src_ip)) 2994#ifdef INET6 2995 || (is_ipv6 && 2996 in6_localaddr(&(args->f_id.src_ip6))) 2997#endif 2998 )) 2999 match = 3000#ifdef INET6 3001 is_ipv6 ? verify_path6( 3002 &(args->f_id.src_ip6), 3003 m->m_pkthdr.rcvif) : 3004#endif 3005 verify_path(src_ip, 3006 m->m_pkthdr.rcvif, 3007 args->f_id.fib); 3008 else 3009 match = 1; 3010 break; 3011 3012 case O_IPSEC: 3013#ifdef IPSEC 3014 match = (m_tag_find(m, 3015 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 3016#endif 3017 /* otherwise no match */ 3018 break; 3019 3020#ifdef INET6 3021 case O_IP6_SRC: 3022 match = is_ipv6 && 3023 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 3024 &((ipfw_insn_ip6 *)cmd)->addr6); 3025 break; 3026 3027 case O_IP6_DST: 3028 match = is_ipv6 && 3029 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 3030 &((ipfw_insn_ip6 *)cmd)->addr6); 3031 break; 3032 case O_IP6_SRC_MASK: 3033 case O_IP6_DST_MASK: 3034 if (is_ipv6) { 3035 int i = cmdlen - 1; 3036 struct in6_addr p; 3037 struct in6_addr *d = 3038 &((ipfw_insn_ip6 *)cmd)->addr6; 3039 3040 for (; !match && i > 0; d += 2, 3041 i -= F_INSN_SIZE(struct in6_addr) 3042 * 2) { 3043 p = (cmd->opcode == 3044 O_IP6_SRC_MASK) ? 3045 args->f_id.src_ip6: 3046 args->f_id.dst_ip6; 3047 APPLY_MASK(&p, &d[1]); 3048 match = 3049 IN6_ARE_ADDR_EQUAL(&d[0], 3050 &p); 3051 } 3052 } 3053 break; 3054 3055 case O_IP6_SRC_ME: 3056 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3057 break; 3058 3059 case O_IP6_DST_ME: 3060 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3061 break; 3062 3063 case O_FLOW6ID: 3064 match = is_ipv6 && 3065 flow6id_match(args->f_id.flow_id6, 3066 (ipfw_insn_u32 *) cmd); 3067 break; 3068 3069 case O_EXT_HDR: 3070 match = is_ipv6 && 3071 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3072 break; 3073 3074 case O_IP6: 3075 match = is_ipv6; 3076 break; 3077#endif 3078 3079 case O_IP4: 3080 match = is_ipv4; 3081 break; 3082 3083 case O_TAG: { 3084 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3085 tablearg : cmd->arg1; 3086 3087 /* Packet is already tagged with this tag? */ 3088 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3089 3090 /* We have `untag' action when F_NOT flag is 3091 * present. And we must remove this mtag from 3092 * mbuf and reset `match' to zero (`match' will 3093 * be inversed later). 3094 * Otherwise we should allocate new mtag and 3095 * push it into mbuf. 3096 */ 3097 if (cmd->len & F_NOT) { /* `untag' action */ 3098 if (mtag != NULL) 3099 m_tag_delete(m, mtag); 3100 } else if (mtag == NULL) { 3101 if ((mtag = m_tag_alloc(MTAG_IPFW, 3102 tag, 0, M_NOWAIT)) != NULL) 3103 m_tag_prepend(m, mtag); 3104 } 3105 match = (cmd->len & F_NOT) ? 0: 1; 3106 break; 3107 } 3108 3109 case O_FIB: /* try match the specified fib */ 3110 if (args->f_id.fib == cmd->arg1) 3111 match = 1; 3112 break; 3113 3114 case O_TAGGED: { 3115 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3116 tablearg : cmd->arg1; 3117 3118 if (cmdlen == 1) { 3119 match = m_tag_locate(m, MTAG_IPFW, 3120 tag, NULL) != NULL; 3121 break; 3122 } 3123 3124 /* we have ranges */ 3125 for (mtag = m_tag_first(m); 3126 mtag != NULL && !match; 3127 mtag = m_tag_next(m, mtag)) { 3128 uint16_t *p; 3129 int i; 3130 3131 if (mtag->m_tag_cookie != MTAG_IPFW) 3132 continue; 3133 3134 p = ((ipfw_insn_u16 *)cmd)->ports; 3135 i = cmdlen - 1; 3136 for(; !match && i > 0; i--, p += 2) 3137 match = 3138 mtag->m_tag_id >= p[0] && 3139 mtag->m_tag_id <= p[1]; 3140 } 3141 break; 3142 } 3143 3144 /* 3145 * The second set of opcodes represents 'actions', 3146 * i.e. the terminal part of a rule once the packet 3147 * matches all previous patterns. 3148 * Typically there is only one action for each rule, 3149 * and the opcode is stored at the end of the rule 3150 * (but there are exceptions -- see below). 3151 * 3152 * In general, here we set retval and terminate the 3153 * outer loop (would be a 'break 3' in some language, 3154 * but we need to do a 'goto done'). 3155 * 3156 * Exceptions: 3157 * O_COUNT and O_SKIPTO actions: 3158 * instead of terminating, we jump to the next rule 3159 * ('goto next_rule', equivalent to a 'break 2'), 3160 * or to the SKIPTO target ('goto again' after 3161 * having set f, cmd and l), respectively. 3162 * 3163 * O_TAG, O_LOG and O_ALTQ action parameters: 3164 * perform some action and set match = 1; 3165 * 3166 * O_LIMIT and O_KEEP_STATE: these opcodes are 3167 * not real 'actions', and are stored right 3168 * before the 'action' part of the rule. 3169 * These opcodes try to install an entry in the 3170 * state tables; if successful, we continue with 3171 * the next opcode (match=1; break;), otherwise 3172 * the packet * must be dropped 3173 * ('goto done' after setting retval); 3174 * 3175 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3176 * cause a lookup of the state table, and a jump 3177 * to the 'action' part of the parent rule 3178 * ('goto check_body') if an entry is found, or 3179 * (CHECK_STATE only) a jump to the next rule if 3180 * the entry is not found ('goto next_rule'). 3181 * The result of the lookup is cached to make 3182 * further instances of these opcodes are 3183 * effectively NOPs. 3184 */ 3185 case O_LIMIT: 3186 case O_KEEP_STATE: 3187 if (install_state(f, 3188 (ipfw_insn_limit *)cmd, args, tablearg)) { 3189 retval = IP_FW_DENY; 3190 goto done; /* error/limit violation */ 3191 } 3192 match = 1; 3193 break; 3194 3195 case O_PROBE_STATE: 3196 case O_CHECK_STATE: 3197 /* 3198 * dynamic rules are checked at the first 3199 * keep-state or check-state occurrence, 3200 * with the result being stored in dyn_dir. 3201 * The compiler introduces a PROBE_STATE 3202 * instruction for us when we have a 3203 * KEEP_STATE (because PROBE_STATE needs 3204 * to be run first). 3205 */ 3206 if (dyn_dir == MATCH_UNKNOWN && 3207 (q = lookup_dyn_rule(&args->f_id, 3208 &dyn_dir, proto == IPPROTO_TCP ? 3209 TCP(ulp) : NULL)) 3210 != NULL) { 3211 /* 3212 * Found dynamic entry, update stats 3213 * and jump to the 'action' part of 3214 * the parent rule. 3215 */ 3216 q->pcnt++; 3217 q->bcnt += pktlen; 3218 f = q->rule; 3219 cmd = ACTION_PTR(f); 3220 l = f->cmd_len - f->act_ofs; 3221 IPFW_DYN_UNLOCK(); 3222 goto check_body; 3223 } 3224 /* 3225 * Dynamic entry not found. If CHECK_STATE, 3226 * skip to next rule, if PROBE_STATE just 3227 * ignore and continue with next opcode. 3228 */ 3229 if (cmd->opcode == O_CHECK_STATE) 3230 goto next_rule; 3231 match = 1; 3232 break; 3233 3234 case O_ACCEPT: 3235 retval = 0; /* accept */ 3236 goto done; 3237 3238 case O_PIPE: 3239 case O_QUEUE: 3240 args->rule = f; /* report matching rule */ 3241 if (cmd->arg1 == IP_FW_TABLEARG) 3242 args->cookie = tablearg; 3243 else 3244 args->cookie = cmd->arg1; 3245 retval = IP_FW_DUMMYNET; 3246 goto done; 3247 3248 case O_DIVERT: 3249 case O_TEE: { 3250 struct divert_tag *dt; 3251 3252 if (args->eh) /* not on layer 2 */ 3253 break; 3254 mtag = m_tag_get(PACKET_TAG_DIVERT, 3255 sizeof(struct divert_tag), 3256 M_NOWAIT); 3257 if (mtag == NULL) { 3258 /* XXX statistic */ 3259 /* drop packet */ 3260 IPFW_RUNLOCK(chain); 3261 return (IP_FW_DENY); 3262 } 3263 dt = (struct divert_tag *)(mtag+1); 3264 dt->cookie = f->rulenum; 3265 if (cmd->arg1 == IP_FW_TABLEARG) 3266 dt->info = tablearg; 3267 else 3268 dt->info = cmd->arg1; 3269 m_tag_prepend(m, mtag); 3270 retval = (cmd->opcode == O_DIVERT) ? 3271 IP_FW_DIVERT : IP_FW_TEE; 3272 goto done; 3273 } 3274 case O_COUNT: 3275 case O_SKIPTO: 3276 f->pcnt++; /* update stats */ 3277 f->bcnt += pktlen; 3278 f->timestamp = time_uptime; 3279 if (cmd->opcode == O_COUNT) 3280 goto next_rule; 3281 /* handle skipto */ 3282 if (cmd->arg1 == IP_FW_TABLEARG) { 3283 f = lookup_next_rule(f, tablearg); 3284 } else { 3285 if (f->next_rule == NULL) 3286 lookup_next_rule(f, 0); 3287 f = f->next_rule; 3288 } 3289 goto again; 3290 3291 case O_REJECT: 3292 /* 3293 * Drop the packet and send a reject notice 3294 * if the packet is not ICMP (or is an ICMP 3295 * query), and it is not multicast/broadcast. 3296 */ 3297 if (hlen > 0 && is_ipv4 && offset == 0 && 3298 (proto != IPPROTO_ICMP || 3299 is_icmp_query(ICMP(ulp))) && 3300 !(m->m_flags & (M_BCAST|M_MCAST)) && 3301 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3302 send_reject(args, cmd->arg1, ip_len, ip); 3303 m = args->m; 3304 } 3305 /* FALLTHROUGH */ 3306#ifdef INET6 3307 case O_UNREACH6: 3308 if (hlen > 0 && is_ipv6 && 3309 ((offset & IP6F_OFF_MASK) == 0) && 3310 (proto != IPPROTO_ICMPV6 || 3311 (is_icmp6_query(args->f_id.flags) == 1)) && 3312 !(m->m_flags & (M_BCAST|M_MCAST)) && 3313 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3314 send_reject6( 3315 args, cmd->arg1, hlen, 3316 (struct ip6_hdr *)ip); 3317 m = args->m; 3318 } 3319 /* FALLTHROUGH */ 3320#endif 3321 case O_DENY: 3322 retval = IP_FW_DENY; 3323 goto done; 3324 3325 case O_FORWARD_IP: { 3326 struct sockaddr_in *sa; 3327 sa = &(((ipfw_insn_sa *)cmd)->sa); 3328 if (args->eh) /* not valid on layer2 pkts */ 3329 break; 3330 if (!q || dyn_dir == MATCH_FORWARD) { 3331 if (sa->sin_addr.s_addr == INADDR_ANY) { 3332 bcopy(sa, &args->hopstore, 3333 sizeof(*sa)); 3334 args->hopstore.sin_addr.s_addr = 3335 htonl(tablearg); 3336 args->next_hop = 3337 &args->hopstore; 3338 } else { 3339 args->next_hop = sa; 3340 } 3341 } 3342 retval = IP_FW_PASS; 3343 } 3344 goto done; 3345 3346 case O_NETGRAPH: 3347 case O_NGTEE: 3348 args->rule = f; /* report matching rule */ 3349 if (cmd->arg1 == IP_FW_TABLEARG) 3350 args->cookie = tablearg; 3351 else 3352 args->cookie = cmd->arg1; 3353 retval = (cmd->opcode == O_NETGRAPH) ? 3354 IP_FW_NETGRAPH : IP_FW_NGTEE; 3355 goto done; 3356 3357 case O_SETFIB: 3358 f->pcnt++; /* update stats */ 3359 f->bcnt += pktlen; 3360 f->timestamp = time_uptime; 3361 M_SETFIB(m, cmd->arg1); 3362 args->f_id.fib = cmd->arg1; 3363 goto next_rule; 3364 3365 case O_NAT: { 3366 struct cfg_nat *t; 3367 int nat_id; 3368 3369 if (IPFW_NAT_LOADED) { 3370 args->rule = f; /* Report matching rule. */ 3371 t = ((ipfw_insn_nat *)cmd)->nat; 3372 if (t == NULL) { 3373 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 3374 tablearg : cmd->arg1; 3375 LOOKUP_NAT(V_layer3_chain, nat_id, t); 3376 if (t == NULL) { 3377 retval = IP_FW_DENY; 3378 goto done; 3379 } 3380 if (cmd->arg1 != IP_FW_TABLEARG) 3381 ((ipfw_insn_nat *)cmd)->nat = t; 3382 } 3383 retval = ipfw_nat_ptr(args, t, m); 3384 } else 3385 retval = IP_FW_DENY; 3386 goto done; 3387 } 3388 3389 case O_REASS: { 3390 int ip_off; 3391 3392 f->pcnt++; 3393 f->bcnt += pktlen; 3394 ip_off = (args->eh != NULL) ? ntohs(ip->ip_off) : ip->ip_off; 3395 if (ip_off & (IP_MF | IP_OFFMASK)) { 3396 /* 3397 * ip_reass() expects len & off in host 3398 * byte order: fix them in case we come 3399 * from layer2. 3400 */ 3401 if (args->eh != NULL) { 3402 ip->ip_len = ntohs(ip->ip_len); 3403 ip->ip_off = ntohs(ip->ip_off); 3404 } 3405 3406 m = ip_reass(m); 3407 args->m = m; 3408 3409 /* 3410 * IP header checksum fixup after 3411 * reassembly and leave header 3412 * in network byte order. 3413 */ 3414 if (m != NULL) { 3415 int hlen; 3416 3417 ip = mtod(m, struct ip *); 3418 hlen = ip->ip_hl << 2; 3419 /* revert len & off for layer2 pkts */ 3420 if (args->eh != NULL) 3421 ip->ip_len = htons(ip->ip_len); 3422 ip->ip_sum = 0; 3423 if (hlen == sizeof(struct ip)) 3424 ip->ip_sum = in_cksum_hdr(ip); 3425 else 3426 ip->ip_sum = in_cksum(m, hlen); 3427 retval = IP_FW_REASS; 3428 args->rule = f; 3429 goto done; 3430 } else { 3431 retval = IP_FW_DENY; 3432 goto done; 3433 } 3434 } 3435 goto next_rule; 3436 } 3437 3438 default: 3439 panic("-- unknown opcode %d\n", cmd->opcode); 3440 } /* end of switch() on opcodes */ 3441 3442 if (cmd->len & F_NOT) 3443 match = !match; 3444 3445 if (match) { 3446 if (cmd->len & F_OR) 3447 skip_or = 1; 3448 } else { 3449 if (!(cmd->len & F_OR)) /* not an OR block, */ 3450 break; /* try next rule */ 3451 } 3452 3453 } /* end of inner for, scan opcodes */ 3454 3455next_rule:; /* try next rule */ 3456 3457 } /* end of outer for, scan rules */ 3458 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3459 IPFW_RUNLOCK(chain); 3460 return (IP_FW_DENY); 3461 3462done: 3463 /* Update statistics */ 3464 f->pcnt++; 3465 f->bcnt += pktlen; 3466 f->timestamp = time_uptime; 3467 IPFW_RUNLOCK(chain); 3468 return (retval); 3469 3470pullup_failed: 3471 if (V_fw_verbose) 3472 printf("ipfw: pullup failed\n"); 3473 return (IP_FW_DENY); 3474} 3475 3476/* 3477 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3478 * These will be reconstructed on the fly as packets are matched. 3479 */ 3480static void 3481flush_rule_ptrs(struct ip_fw_chain *chain) 3482{ 3483 struct ip_fw *rule; 3484 3485 IPFW_WLOCK_ASSERT(chain); 3486 3487 for (rule = chain->rules; rule; rule = rule->next) 3488 rule->next_rule = NULL; 3489} 3490 3491/* 3492 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3493 * possibly create a rule number and add the rule to the list. 3494 * Update the rule_number in the input struct so the caller knows it as well. 3495 */ 3496static int 3497add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3498{ 3499 INIT_VNET_IPFW(curvnet); 3500 struct ip_fw *rule, *f, *prev; 3501 int l = RULESIZE(input_rule); 3502 3503 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3504 return (EINVAL); 3505 3506 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3507 if (rule == NULL) 3508 return (ENOSPC); 3509 3510 bcopy(input_rule, rule, l); 3511 3512 rule->next = NULL; 3513 rule->next_rule = NULL; 3514 3515 rule->pcnt = 0; 3516 rule->bcnt = 0; 3517 rule->timestamp = 0; 3518 3519 IPFW_WLOCK(chain); 3520 3521 if (chain->rules == NULL) { /* default rule */ 3522 chain->rules = rule; 3523 goto done; 3524 } 3525 3526 /* 3527 * If rulenum is 0, find highest numbered rule before the 3528 * default rule, and add autoinc_step 3529 */ 3530 if (V_autoinc_step < 1) 3531 V_autoinc_step = 1; 3532 else if (V_autoinc_step > 1000) 3533 V_autoinc_step = 1000; 3534 if (rule->rulenum == 0) { 3535 /* 3536 * locate the highest numbered rule before default 3537 */ 3538 for (f = chain->rules; f; f = f->next) { 3539 if (f->rulenum == IPFW_DEFAULT_RULE) 3540 break; 3541 rule->rulenum = f->rulenum; 3542 } 3543 if (rule->rulenum < IPFW_DEFAULT_RULE - V_autoinc_step) 3544 rule->rulenum += V_autoinc_step; 3545 input_rule->rulenum = rule->rulenum; 3546 } 3547 3548 /* 3549 * Now insert the new rule in the right place in the sorted list. 3550 */ 3551 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3552 if (f->rulenum > rule->rulenum) { /* found the location */ 3553 if (prev) { 3554 rule->next = f; 3555 prev->next = rule; 3556 } else { /* head insert */ 3557 rule->next = chain->rules; 3558 chain->rules = rule; 3559 } 3560 break; 3561 } 3562 } 3563 flush_rule_ptrs(chain); 3564done: 3565 V_static_count++; 3566 V_static_len += l; 3567 IPFW_WUNLOCK(chain); 3568 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3569 rule->rulenum, V_static_count);) 3570 return (0); 3571} 3572 3573/** 3574 * Remove a static rule (including derived * dynamic rules) 3575 * and place it on the ``reap list'' for later reclamation. 3576 * The caller is in charge of clearing rule pointers to avoid 3577 * dangling pointers. 3578 * @return a pointer to the next entry. 3579 * Arguments are not checked, so they better be correct. 3580 */ 3581static struct ip_fw * 3582remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 3583 struct ip_fw *prev) 3584{ 3585 INIT_VNET_IPFW(curvnet); 3586 struct ip_fw *n; 3587 int l = RULESIZE(rule); 3588 3589 IPFW_WLOCK_ASSERT(chain); 3590 3591 n = rule->next; 3592 IPFW_DYN_LOCK(); 3593 remove_dyn_rule(rule, NULL /* force removal */); 3594 IPFW_DYN_UNLOCK(); 3595 if (prev == NULL) 3596 chain->rules = n; 3597 else 3598 prev->next = n; 3599 V_static_count--; 3600 V_static_len -= l; 3601 3602 rule->next = chain->reap; 3603 chain->reap = rule; 3604 3605 return n; 3606} 3607 3608/** 3609 * Reclaim storage associated with a list of rules. This is 3610 * typically the list created using remove_rule. 3611 */ 3612static void 3613reap_rules(struct ip_fw *head) 3614{ 3615 struct ip_fw *rule; 3616 3617 while ((rule = head) != NULL) { 3618 head = head->next; 3619 if (DUMMYNET_LOADED) 3620 ip_dn_ruledel_ptr(rule); 3621 free(rule, M_IPFW); 3622 } 3623} 3624 3625/* 3626 * Remove all rules from a chain (except rules in set RESVD_SET 3627 * unless kill_default = 1). The caller is responsible for 3628 * reclaiming storage for the rules left in chain->reap. 3629 */ 3630static void 3631free_chain(struct ip_fw_chain *chain, int kill_default) 3632{ 3633 struct ip_fw *prev, *rule; 3634 3635 IPFW_WLOCK_ASSERT(chain); 3636 3637 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3638 for (prev = NULL, rule = chain->rules; rule ; ) 3639 if (kill_default || rule->set != RESVD_SET) 3640 rule = remove_rule(chain, rule, prev); 3641 else { 3642 prev = rule; 3643 rule = rule->next; 3644 } 3645} 3646 3647/** 3648 * Remove all rules with given number, and also do set manipulation. 3649 * Assumes chain != NULL && *chain != NULL. 3650 * 3651 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3652 * the next 8 bits are the new set, the top 8 bits are the command: 3653 * 3654 * 0 delete rules with given number 3655 * 1 delete rules with given set number 3656 * 2 move rules with given number to new set 3657 * 3 move rules with given set number to new set 3658 * 4 swap sets with given numbers 3659 * 5 delete rules with given number and with given set number 3660 */ 3661static int 3662del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3663{ 3664 struct ip_fw *prev = NULL, *rule; 3665 u_int16_t rulenum; /* rule or old_set */ 3666 u_int8_t cmd, new_set; 3667 3668 rulenum = arg & 0xffff; 3669 cmd = (arg >> 24) & 0xff; 3670 new_set = (arg >> 16) & 0xff; 3671 3672 if (cmd > 5 || new_set > RESVD_SET) 3673 return EINVAL; 3674 if (cmd == 0 || cmd == 2 || cmd == 5) { 3675 if (rulenum >= IPFW_DEFAULT_RULE) 3676 return EINVAL; 3677 } else { 3678 if (rulenum > RESVD_SET) /* old_set */ 3679 return EINVAL; 3680 } 3681 3682 IPFW_WLOCK(chain); 3683 rule = chain->rules; 3684 chain->reap = NULL; 3685 switch (cmd) { 3686 case 0: /* delete rules with given number */ 3687 /* 3688 * locate first rule to delete 3689 */ 3690 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3691 ; 3692 if (rule->rulenum != rulenum) { 3693 IPFW_WUNLOCK(chain); 3694 return EINVAL; 3695 } 3696 3697 /* 3698 * flush pointers outside the loop, then delete all matching 3699 * rules. prev remains the same throughout the cycle. 3700 */ 3701 flush_rule_ptrs(chain); 3702 while (rule->rulenum == rulenum) 3703 rule = remove_rule(chain, rule, prev); 3704 break; 3705 3706 case 1: /* delete all rules with given set number */ 3707 flush_rule_ptrs(chain); 3708 rule = chain->rules; 3709 while (rule->rulenum < IPFW_DEFAULT_RULE) 3710 if (rule->set == rulenum) 3711 rule = remove_rule(chain, rule, prev); 3712 else { 3713 prev = rule; 3714 rule = rule->next; 3715 } 3716 break; 3717 3718 case 2: /* move rules with given number to new set */ 3719 rule = chain->rules; 3720 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3721 if (rule->rulenum == rulenum) 3722 rule->set = new_set; 3723 break; 3724 3725 case 3: /* move rules with given set number to new set */ 3726 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3727 if (rule->set == rulenum) 3728 rule->set = new_set; 3729 break; 3730 3731 case 4: /* swap two sets */ 3732 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3733 if (rule->set == rulenum) 3734 rule->set = new_set; 3735 else if (rule->set == new_set) 3736 rule->set = rulenum; 3737 break; 3738 case 5: /* delete rules with given number and with given set number. 3739 * rulenum - given rule number; 3740 * new_set - given set number. 3741 */ 3742 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3743 ; 3744 if (rule->rulenum != rulenum) { 3745 IPFW_WUNLOCK(chain); 3746 return (EINVAL); 3747 } 3748 flush_rule_ptrs(chain); 3749 while (rule->rulenum == rulenum) { 3750 if (rule->set == new_set) 3751 rule = remove_rule(chain, rule, prev); 3752 else { 3753 prev = rule; 3754 rule = rule->next; 3755 } 3756 } 3757 } 3758 /* 3759 * Look for rules to reclaim. We grab the list before 3760 * releasing the lock then reclaim them w/o the lock to 3761 * avoid a LOR with dummynet. 3762 */ 3763 rule = chain->reap; 3764 chain->reap = NULL; 3765 IPFW_WUNLOCK(chain); 3766 if (rule) 3767 reap_rules(rule); 3768 return 0; 3769} 3770 3771/* 3772 * Clear counters for a specific rule. 3773 * The enclosing "table" is assumed locked. 3774 */ 3775static void 3776clear_counters(struct ip_fw *rule, int log_only) 3777{ 3778 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3779 3780 if (log_only == 0) { 3781 rule->bcnt = rule->pcnt = 0; 3782 rule->timestamp = 0; 3783 } 3784 if (l->o.opcode == O_LOG) 3785 l->log_left = l->max_log; 3786} 3787 3788/** 3789 * Reset some or all counters on firewall rules. 3790 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 3791 * the next 8 bits are the set number, the top 8 bits are the command: 3792 * 0 work with rules from all set's; 3793 * 1 work with rules only from specified set. 3794 * Specified rule number is zero if we want to clear all entries. 3795 * log_only is 1 if we only want to reset logs, zero otherwise. 3796 */ 3797static int 3798zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 3799{ 3800 INIT_VNET_IPFW(curvnet); 3801 struct ip_fw *rule; 3802 char *msg; 3803 3804 uint16_t rulenum = arg & 0xffff; 3805 uint8_t set = (arg >> 16) & 0xff; 3806 uint8_t cmd = (arg >> 24) & 0xff; 3807 3808 if (cmd > 1) 3809 return (EINVAL); 3810 if (cmd == 1 && set > RESVD_SET) 3811 return (EINVAL); 3812 3813 IPFW_WLOCK(chain); 3814 if (rulenum == 0) { 3815 V_norule_counter = 0; 3816 for (rule = chain->rules; rule; rule = rule->next) { 3817 /* Skip rules from another set. */ 3818 if (cmd == 1 && rule->set != set) 3819 continue; 3820 clear_counters(rule, log_only); 3821 } 3822 msg = log_only ? "All logging counts reset" : 3823 "Accounting cleared"; 3824 } else { 3825 int cleared = 0; 3826 /* 3827 * We can have multiple rules with the same number, so we 3828 * need to clear them all. 3829 */ 3830 for (rule = chain->rules; rule; rule = rule->next) 3831 if (rule->rulenum == rulenum) { 3832 while (rule && rule->rulenum == rulenum) { 3833 if (cmd == 0 || rule->set == set) 3834 clear_counters(rule, log_only); 3835 rule = rule->next; 3836 } 3837 cleared = 1; 3838 break; 3839 } 3840 if (!cleared) { /* we did not find any matching rules */ 3841 IPFW_WUNLOCK(chain); 3842 return (EINVAL); 3843 } 3844 msg = log_only ? "logging count reset" : "cleared"; 3845 } 3846 IPFW_WUNLOCK(chain); 3847 3848 if (V_fw_verbose) { 3849 int lev = LOG_SECURITY | LOG_NOTICE; 3850 3851 if (rulenum) 3852 log(lev, "ipfw: Entry %d %s.\n", rulenum, msg); 3853 else 3854 log(lev, "ipfw: %s.\n", msg); 3855 } 3856 return (0); 3857} 3858 3859/* 3860 * Check validity of the structure before insert. 3861 * Fortunately rules are simple, so this mostly need to check rule sizes. 3862 */ 3863static int 3864check_ipfw_struct(struct ip_fw *rule, int size) 3865{ 3866 int l, cmdlen = 0; 3867 int have_action=0; 3868 ipfw_insn *cmd; 3869 3870 if (size < sizeof(*rule)) { 3871 printf("ipfw: rule too short\n"); 3872 return (EINVAL); 3873 } 3874 /* first, check for valid size */ 3875 l = RULESIZE(rule); 3876 if (l != size) { 3877 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3878 return (EINVAL); 3879 } 3880 if (rule->act_ofs >= rule->cmd_len) { 3881 printf("ipfw: bogus action offset (%u > %u)\n", 3882 rule->act_ofs, rule->cmd_len - 1); 3883 return (EINVAL); 3884 } 3885 /* 3886 * Now go for the individual checks. Very simple ones, basically only 3887 * instruction sizes. 3888 */ 3889 for (l = rule->cmd_len, cmd = rule->cmd ; 3890 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3891 cmdlen = F_LEN(cmd); 3892 if (cmdlen > l) { 3893 printf("ipfw: opcode %d size truncated\n", 3894 cmd->opcode); 3895 return EINVAL; 3896 } 3897 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3898 switch (cmd->opcode) { 3899 case O_PROBE_STATE: 3900 case O_KEEP_STATE: 3901 case O_PROTO: 3902 case O_IP_SRC_ME: 3903 case O_IP_DST_ME: 3904 case O_LAYER2: 3905 case O_IN: 3906 case O_FRAG: 3907 case O_DIVERTED: 3908 case O_IPOPT: 3909 case O_IPTOS: 3910 case O_IPPRECEDENCE: 3911 case O_IPVER: 3912 case O_TCPWIN: 3913 case O_TCPFLAGS: 3914 case O_TCPOPTS: 3915 case O_ESTAB: 3916 case O_VERREVPATH: 3917 case O_VERSRCREACH: 3918 case O_ANTISPOOF: 3919 case O_IPSEC: 3920#ifdef INET6 3921 case O_IP6_SRC_ME: 3922 case O_IP6_DST_ME: 3923 case O_EXT_HDR: 3924 case O_IP6: 3925#endif 3926 case O_IP4: 3927 case O_TAG: 3928 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3929 goto bad_size; 3930 break; 3931 3932 case O_FIB: 3933 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3934 goto bad_size; 3935 if (cmd->arg1 >= rt_numfibs) { 3936 printf("ipfw: invalid fib number %d\n", 3937 cmd->arg1); 3938 return EINVAL; 3939 } 3940 break; 3941 3942 case O_SETFIB: 3943 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3944 goto bad_size; 3945 if (cmd->arg1 >= rt_numfibs) { 3946 printf("ipfw: invalid fib number %d\n", 3947 cmd->arg1); 3948 return EINVAL; 3949 } 3950 goto check_action; 3951 3952 case O_UID: 3953 case O_GID: 3954 case O_JAIL: 3955 case O_IP_SRC: 3956 case O_IP_DST: 3957 case O_TCPSEQ: 3958 case O_TCPACK: 3959 case O_PROB: 3960 case O_ICMPTYPE: 3961 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3962 goto bad_size; 3963 break; 3964 3965 case O_LIMIT: 3966 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3967 goto bad_size; 3968 break; 3969 3970 case O_LOG: 3971 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3972 goto bad_size; 3973 3974 ((ipfw_insn_log *)cmd)->log_left = 3975 ((ipfw_insn_log *)cmd)->max_log; 3976 3977 break; 3978 3979 case O_IP_SRC_MASK: 3980 case O_IP_DST_MASK: 3981 /* only odd command lengths */ 3982 if ( !(cmdlen & 1) || cmdlen > 31) 3983 goto bad_size; 3984 break; 3985 3986 case O_IP_SRC_SET: 3987 case O_IP_DST_SET: 3988 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3989 printf("ipfw: invalid set size %d\n", 3990 cmd->arg1); 3991 return EINVAL; 3992 } 3993 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3994 (cmd->arg1+31)/32 ) 3995 goto bad_size; 3996 break; 3997 3998 case O_IP_SRC_LOOKUP: 3999 case O_IP_DST_LOOKUP: 4000 if (cmd->arg1 >= IPFW_TABLES_MAX) { 4001 printf("ipfw: invalid table number %d\n", 4002 cmd->arg1); 4003 return (EINVAL); 4004 } 4005 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 4006 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 4007 goto bad_size; 4008 break; 4009 4010 case O_MACADDR2: 4011 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 4012 goto bad_size; 4013 break; 4014 4015 case O_NOP: 4016 case O_IPID: 4017 case O_IPTTL: 4018 case O_IPLEN: 4019 case O_TCPDATALEN: 4020 case O_TAGGED: 4021 if (cmdlen < 1 || cmdlen > 31) 4022 goto bad_size; 4023 break; 4024 4025 case O_MAC_TYPE: 4026 case O_IP_SRCPORT: 4027 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 4028 if (cmdlen < 2 || cmdlen > 31) 4029 goto bad_size; 4030 break; 4031 4032 case O_RECV: 4033 case O_XMIT: 4034 case O_VIA: 4035 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 4036 goto bad_size; 4037 break; 4038 4039 case O_ALTQ: 4040 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 4041 goto bad_size; 4042 break; 4043 4044 case O_PIPE: 4045 case O_QUEUE: 4046 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4047 goto bad_size; 4048 goto check_action; 4049 4050 case O_FORWARD_IP: 4051#ifdef IPFIREWALL_FORWARD 4052 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 4053 goto bad_size; 4054 goto check_action; 4055#else 4056 return EINVAL; 4057#endif 4058 4059 case O_DIVERT: 4060 case O_TEE: 4061 if (ip_divert_ptr == NULL) 4062 return EINVAL; 4063 else 4064 goto check_size; 4065 case O_NETGRAPH: 4066 case O_NGTEE: 4067 if (!NG_IPFW_LOADED) 4068 return EINVAL; 4069 else 4070 goto check_size; 4071 case O_NAT: 4072 if (!IPFW_NAT_LOADED) 4073 return EINVAL; 4074 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 4075 goto bad_size; 4076 goto check_action; 4077 case O_FORWARD_MAC: /* XXX not implemented yet */ 4078 case O_CHECK_STATE: 4079 case O_COUNT: 4080 case O_ACCEPT: 4081 case O_DENY: 4082 case O_REJECT: 4083#ifdef INET6 4084 case O_UNREACH6: 4085#endif 4086 case O_SKIPTO: 4087 case O_REASS: 4088check_size: 4089 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4090 goto bad_size; 4091check_action: 4092 if (have_action) { 4093 printf("ipfw: opcode %d, multiple actions" 4094 " not allowed\n", 4095 cmd->opcode); 4096 return EINVAL; 4097 } 4098 have_action = 1; 4099 if (l != cmdlen) { 4100 printf("ipfw: opcode %d, action must be" 4101 " last opcode\n", 4102 cmd->opcode); 4103 return EINVAL; 4104 } 4105 break; 4106#ifdef INET6 4107 case O_IP6_SRC: 4108 case O_IP6_DST: 4109 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 4110 F_INSN_SIZE(ipfw_insn)) 4111 goto bad_size; 4112 break; 4113 4114 case O_FLOW6ID: 4115 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4116 ((ipfw_insn_u32 *)cmd)->o.arg1) 4117 goto bad_size; 4118 break; 4119 4120 case O_IP6_SRC_MASK: 4121 case O_IP6_DST_MASK: 4122 if ( !(cmdlen & 1) || cmdlen > 127) 4123 goto bad_size; 4124 break; 4125 case O_ICMP6TYPE: 4126 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4127 goto bad_size; 4128 break; 4129#endif 4130 4131 default: 4132 switch (cmd->opcode) { 4133#ifndef INET6 4134 case O_IP6_SRC_ME: 4135 case O_IP6_DST_ME: 4136 case O_EXT_HDR: 4137 case O_IP6: 4138 case O_UNREACH6: 4139 case O_IP6_SRC: 4140 case O_IP6_DST: 4141 case O_FLOW6ID: 4142 case O_IP6_SRC_MASK: 4143 case O_IP6_DST_MASK: 4144 case O_ICMP6TYPE: 4145 printf("ipfw: no IPv6 support in kernel\n"); 4146 return EPROTONOSUPPORT; 4147#endif 4148 default: 4149 printf("ipfw: opcode %d, unknown opcode\n", 4150 cmd->opcode); 4151 return EINVAL; 4152 } 4153 } 4154 } 4155 if (have_action == 0) { 4156 printf("ipfw: missing action\n"); 4157 return EINVAL; 4158 } 4159 return 0; 4160 4161bad_size: 4162 printf("ipfw: opcode %d size %d wrong\n", 4163 cmd->opcode, cmdlen); 4164 return EINVAL; 4165} 4166 4167/* 4168 * Copy the static and dynamic rules to the supplied buffer 4169 * and return the amount of space actually used. 4170 */ 4171static size_t 4172ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4173{ 4174 INIT_VNET_IPFW(curvnet); 4175 char *bp = buf; 4176 char *ep = bp + space; 4177 struct ip_fw *rule; 4178 int i; 4179 time_t boot_seconds; 4180 4181 boot_seconds = boottime.tv_sec; 4182 /* XXX this can take a long time and locking will block packet flow */ 4183 IPFW_RLOCK(chain); 4184 for (rule = chain->rules; rule ; rule = rule->next) { 4185 /* 4186 * Verify the entry fits in the buffer in case the 4187 * rules changed between calculating buffer space and 4188 * now. This would be better done using a generation 4189 * number but should suffice for now. 4190 */ 4191 i = RULESIZE(rule); 4192 if (bp + i <= ep) { 4193 bcopy(rule, bp, i); 4194 /* 4195 * XXX HACK. Store the disable mask in the "next" 4196 * pointer in a wild attempt to keep the ABI the same. 4197 * Why do we do this on EVERY rule? 4198 */ 4199 bcopy(&V_set_disable, 4200 &(((struct ip_fw *)bp)->next_rule), 4201 sizeof(V_set_disable)); 4202 if (((struct ip_fw *)bp)->timestamp) 4203 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4204 bp += i; 4205 } 4206 } 4207 IPFW_RUNLOCK(chain); 4208 if (V_ipfw_dyn_v) { 4209 ipfw_dyn_rule *p, *last = NULL; 4210 4211 IPFW_DYN_LOCK(); 4212 for (i = 0 ; i < V_curr_dyn_buckets; i++) 4213 for (p = V_ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4214 if (bp + sizeof *p <= ep) { 4215 ipfw_dyn_rule *dst = 4216 (ipfw_dyn_rule *)bp; 4217 bcopy(p, dst, sizeof *p); 4218 bcopy(&(p->rule->rulenum), &(dst->rule), 4219 sizeof(p->rule->rulenum)); 4220 /* 4221 * store set number into high word of 4222 * dst->rule pointer. 4223 */ 4224 bcopy(&(p->rule->set), 4225 (char *)&dst->rule + 4226 sizeof(p->rule->rulenum), 4227 sizeof(p->rule->set)); 4228 /* 4229 * store a non-null value in "next". 4230 * The userland code will interpret a 4231 * NULL here as a marker 4232 * for the last dynamic rule. 4233 */ 4234 bcopy(&dst, &dst->next, sizeof(dst)); 4235 last = dst; 4236 dst->expire = 4237 TIME_LEQ(dst->expire, time_uptime) ? 4238 0 : dst->expire - time_uptime ; 4239 bp += sizeof(ipfw_dyn_rule); 4240 } 4241 } 4242 IPFW_DYN_UNLOCK(); 4243 if (last != NULL) /* mark last dynamic rule */ 4244 bzero(&last->next, sizeof(last)); 4245 } 4246 return (bp - (char *)buf); 4247} 4248 4249 4250/** 4251 * {set|get}sockopt parser. 4252 */ 4253static int 4254ipfw_ctl(struct sockopt *sopt) 4255{ 4256#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4257 INIT_VNET_IPFW(curvnet); 4258 int error; 4259 size_t size; 4260 struct ip_fw *buf, *rule; 4261 u_int32_t rulenum[2]; 4262 4263 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4264 if (error) 4265 return (error); 4266 4267 /* 4268 * Disallow modifications in really-really secure mode, but still allow 4269 * the logging counters to be reset. 4270 */ 4271 if (sopt->sopt_name == IP_FW_ADD || 4272 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4273 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4274 if (error) 4275 return (error); 4276 } 4277 4278 error = 0; 4279 4280 switch (sopt->sopt_name) { 4281 case IP_FW_GET: 4282 /* 4283 * pass up a copy of the current rules. Static rules 4284 * come first (the last of which has number IPFW_DEFAULT_RULE), 4285 * followed by a possibly empty list of dynamic rule. 4286 * The last dynamic rule has NULL in the "next" field. 4287 * 4288 * Note that the calculated size is used to bound the 4289 * amount of data returned to the user. The rule set may 4290 * change between calculating the size and returning the 4291 * data in which case we'll just return what fits. 4292 */ 4293 size = V_static_len; /* size of static rules */ 4294 if (V_ipfw_dyn_v) /* add size of dyn.rules */ 4295 size += (V_dyn_count * sizeof(ipfw_dyn_rule)); 4296 4297 /* 4298 * XXX todo: if the user passes a short length just to know 4299 * how much room is needed, do not bother filling up the 4300 * buffer, just jump to the sooptcopyout. 4301 */ 4302 buf = malloc(size, M_TEMP, M_WAITOK); 4303 error = sooptcopyout(sopt, buf, 4304 ipfw_getrules(&V_layer3_chain, buf, size)); 4305 free(buf, M_TEMP); 4306 break; 4307 4308 case IP_FW_FLUSH: 4309 /* 4310 * Normally we cannot release the lock on each iteration. 4311 * We could do it here only because we start from the head all 4312 * the times so there is no risk of missing some entries. 4313 * On the other hand, the risk is that we end up with 4314 * a very inconsistent ruleset, so better keep the lock 4315 * around the whole cycle. 4316 * 4317 * XXX this code can be improved by resetting the head of 4318 * the list to point to the default rule, and then freeing 4319 * the old list without the need for a lock. 4320 */ 4321 4322 IPFW_WLOCK(&V_layer3_chain); 4323 V_layer3_chain.reap = NULL; 4324 free_chain(&V_layer3_chain, 0 /* keep default rule */); 4325 rule = V_layer3_chain.reap; 4326 V_layer3_chain.reap = NULL; 4327 IPFW_WUNLOCK(&V_layer3_chain); 4328 if (rule != NULL) 4329 reap_rules(rule); 4330 break; 4331 4332 case IP_FW_ADD: 4333 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4334 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4335 sizeof(struct ip_fw) ); 4336 if (error == 0) 4337 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4338 if (error == 0) { 4339 error = add_rule(&V_layer3_chain, rule); 4340 size = RULESIZE(rule); 4341 if (!error && sopt->sopt_dir == SOPT_GET) 4342 error = sooptcopyout(sopt, rule, size); 4343 } 4344 free(rule, M_TEMP); 4345 break; 4346 4347 case IP_FW_DEL: 4348 /* 4349 * IP_FW_DEL is used for deleting single rules or sets, 4350 * and (ab)used to atomically manipulate sets. Argument size 4351 * is used to distinguish between the two: 4352 * sizeof(u_int32_t) 4353 * delete single rule or set of rules, 4354 * or reassign rules (or sets) to a different set. 4355 * 2*sizeof(u_int32_t) 4356 * atomic disable/enable sets. 4357 * first u_int32_t contains sets to be disabled, 4358 * second u_int32_t contains sets to be enabled. 4359 */ 4360 error = sooptcopyin(sopt, rulenum, 4361 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4362 if (error) 4363 break; 4364 size = sopt->sopt_valsize; 4365 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4366 error = del_entry(&V_layer3_chain, rulenum[0]); 4367 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4368 V_set_disable = 4369 (V_set_disable | rulenum[0]) & ~rulenum[1] & 4370 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4371 else 4372 error = EINVAL; 4373 break; 4374 4375 case IP_FW_ZERO: 4376 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 4377 rulenum[0] = 0; 4378 if (sopt->sopt_val != 0) { 4379 error = sooptcopyin(sopt, rulenum, 4380 sizeof(u_int32_t), sizeof(u_int32_t)); 4381 if (error) 4382 break; 4383 } 4384 error = zero_entry(&V_layer3_chain, rulenum[0], 4385 sopt->sopt_name == IP_FW_RESETLOG); 4386 break; 4387 4388 case IP_FW_TABLE_ADD: 4389 { 4390 ipfw_table_entry ent; 4391 4392 error = sooptcopyin(sopt, &ent, 4393 sizeof(ent), sizeof(ent)); 4394 if (error) 4395 break; 4396 error = add_table_entry(&V_layer3_chain, ent.tbl, 4397 ent.addr, ent.masklen, ent.value); 4398 } 4399 break; 4400 4401 case IP_FW_TABLE_DEL: 4402 { 4403 ipfw_table_entry ent; 4404 4405 error = sooptcopyin(sopt, &ent, 4406 sizeof(ent), sizeof(ent)); 4407 if (error) 4408 break; 4409 error = del_table_entry(&V_layer3_chain, ent.tbl, 4410 ent.addr, ent.masklen); 4411 } 4412 break; 4413 4414 case IP_FW_TABLE_FLUSH: 4415 { 4416 u_int16_t tbl; 4417 4418 error = sooptcopyin(sopt, &tbl, 4419 sizeof(tbl), sizeof(tbl)); 4420 if (error) 4421 break; 4422 IPFW_WLOCK(&V_layer3_chain); 4423 error = flush_table(&V_layer3_chain, tbl); 4424 IPFW_WUNLOCK(&V_layer3_chain); 4425 } 4426 break; 4427 4428 case IP_FW_TABLE_GETSIZE: 4429 { 4430 u_int32_t tbl, cnt; 4431 4432 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4433 sizeof(tbl)))) 4434 break; 4435 IPFW_RLOCK(&V_layer3_chain); 4436 error = count_table(&V_layer3_chain, tbl, &cnt); 4437 IPFW_RUNLOCK(&V_layer3_chain); 4438 if (error) 4439 break; 4440 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4441 } 4442 break; 4443 4444 case IP_FW_TABLE_LIST: 4445 { 4446 ipfw_table *tbl; 4447 4448 if (sopt->sopt_valsize < sizeof(*tbl)) { 4449 error = EINVAL; 4450 break; 4451 } 4452 size = sopt->sopt_valsize; 4453 tbl = malloc(size, M_TEMP, M_WAITOK); 4454 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4455 if (error) { 4456 free(tbl, M_TEMP); 4457 break; 4458 } 4459 tbl->size = (size - sizeof(*tbl)) / 4460 sizeof(ipfw_table_entry); 4461 IPFW_RLOCK(&V_layer3_chain); 4462 error = dump_table(&V_layer3_chain, tbl); 4463 IPFW_RUNLOCK(&V_layer3_chain); 4464 if (error) { 4465 free(tbl, M_TEMP); 4466 break; 4467 } 4468 error = sooptcopyout(sopt, tbl, size); 4469 free(tbl, M_TEMP); 4470 } 4471 break; 4472 4473 case IP_FW_NAT_CFG: 4474 if (IPFW_NAT_LOADED) 4475 error = ipfw_nat_cfg_ptr(sopt); 4476 else { 4477 printf("IP_FW_NAT_CFG: %s\n", 4478 "ipfw_nat not present, please load it"); 4479 error = EINVAL; 4480 } 4481 break; 4482 4483 case IP_FW_NAT_DEL: 4484 if (IPFW_NAT_LOADED) 4485 error = ipfw_nat_del_ptr(sopt); 4486 else { 4487 printf("IP_FW_NAT_DEL: %s\n", 4488 "ipfw_nat not present, please load it"); 4489 error = EINVAL; 4490 } 4491 break; 4492 4493 case IP_FW_NAT_GET_CONFIG: 4494 if (IPFW_NAT_LOADED) 4495 error = ipfw_nat_get_cfg_ptr(sopt); 4496 else { 4497 printf("IP_FW_NAT_GET_CFG: %s\n", 4498 "ipfw_nat not present, please load it"); 4499 error = EINVAL; 4500 } 4501 break; 4502 4503 case IP_FW_NAT_GET_LOG: 4504 if (IPFW_NAT_LOADED) 4505 error = ipfw_nat_get_log_ptr(sopt); 4506 else { 4507 printf("IP_FW_NAT_GET_LOG: %s\n", 4508 "ipfw_nat not present, please load it"); 4509 error = EINVAL; 4510 } 4511 break; 4512 4513 default: 4514 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4515 error = EINVAL; 4516 } 4517 4518 return (error); 4519#undef RULE_MAXSIZE 4520} 4521 4522/** 4523 * dummynet needs a reference to the default rule, because rules can be 4524 * deleted while packets hold a reference to them. When this happens, 4525 * dummynet changes the reference to the default rule (it could well be a 4526 * NULL pointer, but this way we do not need to check for the special 4527 * case, plus here he have info on the default behaviour). 4528 */ 4529struct ip_fw *ip_fw_default_rule; 4530 4531/* 4532 * This procedure is only used to handle keepalives. It is invoked 4533 * every dyn_keepalive_period 4534 */ 4535static void 4536ipfw_tick(void * __unused unused) 4537{ 4538 struct mbuf *m0, *m, *mnext, **mtailp; 4539 int i; 4540 ipfw_dyn_rule *q; 4541 4542 if (V_dyn_keepalive == 0 || V_ipfw_dyn_v == NULL || V_dyn_count == 0) 4543 goto done; 4544 4545 /* 4546 * We make a chain of packets to go out here -- not deferring 4547 * until after we drop the IPFW dynamic rule lock would result 4548 * in a lock order reversal with the normal packet input -> ipfw 4549 * call stack. 4550 */ 4551 m0 = NULL; 4552 mtailp = &m0; 4553 IPFW_DYN_LOCK(); 4554 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 4555 for (q = V_ipfw_dyn_v[i] ; q ; q = q->next ) { 4556 if (q->dyn_type == O_LIMIT_PARENT) 4557 continue; 4558 if (q->id.proto != IPPROTO_TCP) 4559 continue; 4560 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4561 continue; 4562 if (TIME_LEQ(time_uptime + V_dyn_keepalive_interval, 4563 q->expire)) 4564 continue; /* too early */ 4565 if (TIME_LEQ(q->expire, time_uptime)) 4566 continue; /* too late, rule expired */ 4567 4568 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4569 q->ack_fwd, TH_SYN); 4570 if (*mtailp != NULL) 4571 mtailp = &(*mtailp)->m_nextpkt; 4572 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4573 q->ack_rev, 0); 4574 if (*mtailp != NULL) 4575 mtailp = &(*mtailp)->m_nextpkt; 4576 } 4577 } 4578 IPFW_DYN_UNLOCK(); 4579 for (m = mnext = m0; m != NULL; m = mnext) { 4580 mnext = m->m_nextpkt; 4581 m->m_nextpkt = NULL; 4582 ip_output(m, NULL, NULL, 0, NULL, NULL); 4583 } 4584done: 4585 callout_reset(&V_ipfw_timeout, V_dyn_keepalive_period * hz, 4586 ipfw_tick, NULL); 4587} 4588 4589int 4590ipfw_init(void) 4591{ 4592 INIT_VNET_IPFW(curvnet); 4593 struct ip_fw default_rule; 4594 int error; 4595 4596 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 4597 4598 V_ipfw_dyn_v = NULL; 4599 V_dyn_buckets = 256; /* must be power of 2 */ 4600 V_curr_dyn_buckets = 256; /* must be power of 2 */ 4601 4602 V_dyn_ack_lifetime = 300; 4603 V_dyn_syn_lifetime = 20; 4604 V_dyn_fin_lifetime = 1; 4605 V_dyn_rst_lifetime = 1; 4606 V_dyn_udp_lifetime = 10; 4607 V_dyn_short_lifetime = 5; 4608 4609 V_dyn_keepalive_interval = 20; 4610 V_dyn_keepalive_period = 5; 4611 V_dyn_keepalive = 1; /* do send keepalives */ 4612 4613 V_dyn_max = 4096; /* max # of dynamic rules */ 4614 4615 V_fw_deny_unknown_exthdrs = 1; 4616 4617#ifdef INET6 4618 /* Setup IPv6 fw sysctl tree. */ 4619 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4620 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4621 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4622 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4623 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4624 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4625 &V_fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4626 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4627 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4628 &V_fw_deny_unknown_exthdrs, 0, 4629 "Deny packets with unknown IPv6 Extension Headers"); 4630#endif 4631 4632 V_layer3_chain.rules = NULL; 4633 IPFW_LOCK_INIT(&V_layer3_chain); 4634 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4635 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4636 UMA_ALIGN_PTR, 0); 4637 IPFW_DYN_LOCK_INIT(); 4638 callout_init(&V_ipfw_timeout, CALLOUT_MPSAFE); 4639 4640 bzero(&default_rule, sizeof default_rule); 4641 4642 default_rule.act_ofs = 0; 4643 default_rule.rulenum = IPFW_DEFAULT_RULE; 4644 default_rule.cmd_len = 1; 4645 default_rule.set = RESVD_SET; 4646 4647 default_rule.cmd[0].len = 1; 4648 default_rule.cmd[0].opcode = 4649#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4650 1 ? O_ACCEPT : 4651#endif 4652 O_DENY; 4653 4654 error = add_rule(&V_layer3_chain, &default_rule); 4655 if (error != 0) { 4656 printf("ipfw2: error %u initializing default rule " 4657 "(support disabled)\n", error); 4658 IPFW_DYN_LOCK_DESTROY(); 4659 IPFW_LOCK_DESTROY(&V_layer3_chain); 4660 uma_zdestroy(ipfw_dyn_rule_zone); 4661 return (error); 4662 } 4663 4664 ip_fw_default_rule = V_layer3_chain.rules; 4665 printf("ipfw2 " 4666#ifdef INET6 4667 "(+ipv6) " 4668#endif 4669 "initialized, divert %s, nat %s, " 4670 "rule-based forwarding " 4671#ifdef IPFIREWALL_FORWARD 4672 "enabled, " 4673#else 4674 "disabled, " 4675#endif 4676 "default to %s, logging ", 4677#ifdef IPDIVERT 4678 "enabled", 4679#else 4680 "loadable", 4681#endif 4682#ifdef IPFIREWALL_NAT 4683 "enabled", 4684#else 4685 "loadable", 4686#endif 4687 4688 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4689 4690#ifdef IPFIREWALL_VERBOSE 4691 V_fw_verbose = 1; 4692#endif 4693#ifdef IPFIREWALL_VERBOSE_LIMIT 4694 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4695#endif 4696 if (V_fw_verbose == 0) 4697 printf("disabled\n"); 4698 else if (V_verbose_limit == 0) 4699 printf("unlimited\n"); 4700 else 4701 printf("limited to %d packets/entry by default\n", 4702 V_verbose_limit); 4703 4704 error = init_tables(&V_layer3_chain); 4705 if (error) { 4706 IPFW_DYN_LOCK_DESTROY(); 4707 IPFW_LOCK_DESTROY(&V_layer3_chain); 4708 uma_zdestroy(ipfw_dyn_rule_zone); 4709 return (error); 4710 } 4711 ip_fw_ctl_ptr = ipfw_ctl; 4712 ip_fw_chk_ptr = ipfw_chk; 4713 callout_reset(&V_ipfw_timeout, hz, ipfw_tick, NULL); 4714 LIST_INIT(&V_layer3_chain.nat); 4715 return (0); 4716} 4717 4718void 4719ipfw_destroy(void) 4720{ 4721 struct ip_fw *reap; 4722 4723 ip_fw_chk_ptr = NULL; 4724 ip_fw_ctl_ptr = NULL; 4725 callout_drain(&V_ipfw_timeout); 4726 IPFW_WLOCK(&V_layer3_chain); 4727 flush_tables(&V_layer3_chain); 4728 V_layer3_chain.reap = NULL; 4729 free_chain(&V_layer3_chain, 1 /* kill default rule */); 4730 reap = V_layer3_chain.reap, V_layer3_chain.reap = NULL; 4731 IPFW_WUNLOCK(&V_layer3_chain); 4732 if (reap != NULL) 4733 reap_rules(reap); 4734 IPFW_DYN_LOCK_DESTROY(); 4735 uma_zdestroy(ipfw_dyn_rule_zone); 4736 if (V_ipfw_dyn_v != NULL) 4737 free(V_ipfw_dyn_v, M_IPFW); 4738 IPFW_LOCK_DESTROY(&V_layer3_chain); 4739 4740#ifdef INET6 4741 /* Free IPv6 fw sysctl tree. */ 4742 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4743#endif 4744 4745 printf("IP firewall unloaded\n"); 4746} 4747