ip_fw2.c revision 193744
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/ipfw/ip_fw2.c 193744 2009-06-08 19:57:35Z bz $"); 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 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/condvar.h> 51#include <sys/eventhandler.h> 52#include <sys/malloc.h> 53#include <sys/mbuf.h> 54#include <sys/kernel.h> 55#include <sys/lock.h> 56#include <sys/jail.h> 57#include <sys/module.h> 58#include <sys/priv.h> 59#include <sys/proc.h> 60#include <sys/rwlock.h> 61#include <sys/socket.h> 62#include <sys/socketvar.h> 63#include <sys/sysctl.h> 64#include <sys/syslog.h> 65#include <sys/ucred.h> 66#include <sys/vimage.h> 67#include <net/ethernet.h> /* for ETHERTYPE_IP */ 68#include <net/if.h> 69#include <net/radix.h> 70#include <net/route.h> 71#include <net/pf_mtag.h> 72#include <net/vnet.h> 73 74#define IPFW_INTERNAL /* Access to protected data structures in ip_fw.h. */ 75 76#include <netinet/in.h> 77#include <netinet/in_var.h> 78#include <netinet/in_pcb.h> 79#include <netinet/ip.h> 80#include <netinet/ip_var.h> 81#include <netinet/ip_icmp.h> 82#include <netinet/ip_fw.h> 83#include <netinet/ip_divert.h> 84#include <netinet/ip_dummynet.h> 85#include <netinet/ip_carp.h> 86#include <netinet/pim.h> 87#include <netinet/tcp_var.h> 88#include <netinet/udp.h> 89#include <netinet/udp_var.h> 90#include <netinet/sctp.h> 91#include <netinet/vinet.h> 92 93#include <netgraph/ng_ipfw.h> 94 95#include <netinet/ip6.h> 96#include <netinet/icmp6.h> 97#ifdef INET6 98#include <netinet6/scope6_var.h> 99#endif 100 101#include <machine/in_cksum.h> /* XXX for in_cksum */ 102 103#ifdef MAC 104#include <security/mac/mac_framework.h> 105#endif 106 107#ifndef VIMAGE 108#ifndef VIMAGE_GLOBALS 109struct vnet_ipfw vnet_ipfw_0; 110#endif 111#endif 112 113/* 114 * set_disable contains one bit per set value (0..31). 115 * If the bit is set, all rules with the corresponding set 116 * are disabled. Set RESVD_SET(31) is reserved for the default rule 117 * and rules that are not deleted by the flush command, 118 * and CANNOT be disabled. 119 * Rules in set RESVD_SET can only be deleted explicitly. 120 */ 121#ifdef VIMAGE_GLOBALS 122static u_int32_t set_disable; 123static int fw_verbose; 124static struct callout ipfw_timeout; 125static int verbose_limit; 126#endif 127 128#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 129static int default_to_accept = 1; 130#else 131static int default_to_accept; 132#endif 133static uma_zone_t ipfw_dyn_rule_zone; 134 135/* 136 * Data structure to cache our ucred related 137 * information. This structure only gets used if 138 * the user specified UID/GID based constraints in 139 * a firewall rule. 140 */ 141struct ip_fw_ugid { 142 gid_t fw_groups[NGROUPS]; 143 int fw_ngroups; 144 uid_t fw_uid; 145 int fw_prid; 146}; 147 148/* 149 * list of rules for layer 3 150 */ 151#ifdef VIMAGE_GLOBALS 152struct ip_fw_chain layer3_chain; 153#endif 154 155MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 156MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 157#define IPFW_NAT_LOADED (ipfw_nat_ptr != NULL) 158ipfw_nat_t *ipfw_nat_ptr = NULL; 159ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 160ipfw_nat_cfg_t *ipfw_nat_del_ptr; 161ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 162ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 163 164struct table_entry { 165 struct radix_node rn[2]; 166 struct sockaddr_in addr, mask; 167 u_int32_t value; 168}; 169 170#ifdef VIMAGE_GLOBALS 171static int autoinc_step; 172#endif 173 174extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 175 176#ifdef SYSCTL_NODE 177SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 178SYSCTL_V_PROC(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, enable, 179 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, fw_enable, 0, 180 ipfw_chg_hook, "I", "Enable ipfw"); 181SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, autoinc_step, 182 CTLFLAG_RW, autoinc_step, 0, "Rule number auto-increment step"); 183SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_fw, OID_AUTO, one_pass, 184 CTLFLAG_RW | CTLFLAG_SECURE3, fw_one_pass, 0, 185 "Only do a single pass through ipfw when using dummynet(4)"); 186SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, verbose, 187 CTLFLAG_RW | CTLFLAG_SECURE3, 188 fw_verbose, 0, "Log matches to ipfw rules"); 189SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, verbose_limit, 190 CTLFLAG_RW, verbose_limit, 0, 191 "Set upper limit of matches of ipfw rules logged"); 192SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 193 NULL, IPFW_DEFAULT_RULE, "The default/max possible rule number."); 194SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 195 NULL, IPFW_TABLES_MAX, "The maximum number of tables."); 196SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 197 &default_to_accept, 0, "Make the default rule accept all packets."); 198TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept); 199#endif /* SYSCTL_NODE */ 200 201/* 202 * Description of dynamic rules. 203 * 204 * Dynamic rules are stored in lists accessed through a hash table 205 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 206 * be modified through the sysctl variable dyn_buckets which is 207 * updated when the table becomes empty. 208 * 209 * XXX currently there is only one list, ipfw_dyn. 210 * 211 * When a packet is received, its address fields are first masked 212 * with the mask defined for the rule, then hashed, then matched 213 * against the entries in the corresponding list. 214 * Dynamic rules can be used for different purposes: 215 * + stateful rules; 216 * + enforcing limits on the number of sessions; 217 * + in-kernel NAT (not implemented yet) 218 * 219 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 220 * measured in seconds and depending on the flags. 221 * 222 * The total number of dynamic rules is stored in dyn_count. 223 * The max number of dynamic rules is dyn_max. When we reach 224 * the maximum number of rules we do not create anymore. This is 225 * done to avoid consuming too much memory, but also too much 226 * time when searching on each packet (ideally, we should try instead 227 * to put a limit on the length of the list on each bucket...). 228 * 229 * Each dynamic rule holds a pointer to the parent ipfw rule so 230 * we know what action to perform. Dynamic rules are removed when 231 * the parent rule is deleted. XXX we should make them survive. 232 * 233 * There are some limitations with dynamic rules -- we do not 234 * obey the 'randomized match', and we do not do multiple 235 * passes through the firewall. XXX check the latter!!! 236 */ 237#ifdef VIMAGE_GLOBALS 238static ipfw_dyn_rule **ipfw_dyn_v; 239static u_int32_t dyn_buckets; 240static u_int32_t curr_dyn_buckets; 241#endif 242 243static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 244#define IPFW_DYN_LOCK_INIT() \ 245 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 246#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 247#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 248#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 249#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 250 251/* 252 * Timeouts for various events in handing dynamic rules. 253 */ 254#ifdef VIMAGE_GLOBALS 255static u_int32_t dyn_ack_lifetime; 256static u_int32_t dyn_syn_lifetime; 257static u_int32_t dyn_fin_lifetime; 258static u_int32_t dyn_rst_lifetime; 259static u_int32_t dyn_udp_lifetime; 260static u_int32_t dyn_short_lifetime; 261 262/* 263 * Keepalives are sent if dyn_keepalive is set. They are sent every 264 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 265 * seconds of lifetime of a rule. 266 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 267 * than dyn_keepalive_period. 268 */ 269 270static u_int32_t dyn_keepalive_interval; 271static u_int32_t dyn_keepalive_period; 272static u_int32_t dyn_keepalive; 273 274static u_int32_t static_count; /* # of static rules */ 275static u_int32_t static_len; /* size in bytes of static rules */ 276static u_int32_t dyn_count; /* # of dynamic rules */ 277static u_int32_t dyn_max; /* max # of dynamic rules */ 278#endif /* VIMAGE_GLOBALS */ 279 280#ifdef SYSCTL_NODE 281SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_buckets, 282 CTLFLAG_RW, dyn_buckets, 0, "Number of dyn. buckets"); 283SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, 284 CTLFLAG_RD, curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 285SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_count, 286 CTLFLAG_RD, dyn_count, 0, "Number of dyn. rules"); 287SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_max, 288 CTLFLAG_RW, dyn_max, 0, "Max number of dyn. rules"); 289SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, static_count, 290 CTLFLAG_RD, static_count, 0, "Number of static rules"); 291SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, 292 CTLFLAG_RW, dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 293SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, 294 CTLFLAG_RW, dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 295SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, 296 CTLFLAG_RW, dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 297SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, 298 CTLFLAG_RW, dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 299SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, 300 CTLFLAG_RW, dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 301SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, 302 CTLFLAG_RW, dyn_short_lifetime, 0, 303 "Lifetime of dyn. rules for other situations"); 304SYSCTL_V_INT(V_NET, vnet_ipfw, _net_inet_ip_fw, OID_AUTO, dyn_keepalive, 305 CTLFLAG_RW, dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 306#endif /* SYSCTL_NODE */ 307 308#ifdef INET6 309/* 310 * IPv6 specific variables 311 */ 312#ifdef SYSCTL_NODE 313SYSCTL_DECL(_net_inet6_ip6); 314#endif /* SYSCTL_NODE */ 315 316static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 317static struct sysctl_oid *ip6_fw_sysctl_tree; 318#endif /* INET6 */ 319 320#ifdef VIMAGE_GLOBALS 321static int fw_deny_unknown_exthdrs; 322#endif 323 324/* 325 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 326 * Other macros just cast void * into the appropriate type 327 */ 328#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 329#define TCP(p) ((struct tcphdr *)(p)) 330#define SCTP(p) ((struct sctphdr *)(p)) 331#define UDP(p) ((struct udphdr *)(p)) 332#define ICMP(p) ((struct icmphdr *)(p)) 333#define ICMP6(p) ((struct icmp6_hdr *)(p)) 334 335static __inline int 336icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 337{ 338 int type = icmp->icmp_type; 339 340 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 341} 342 343#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 344 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 345 346static int 347is_icmp_query(struct icmphdr *icmp) 348{ 349 int type = icmp->icmp_type; 350 351 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 352} 353#undef TT 354 355/* 356 * The following checks use two arrays of 8 or 16 bits to store the 357 * bits that we want set or clear, respectively. They are in the 358 * low and high half of cmd->arg1 or cmd->d[0]. 359 * 360 * We scan options and store the bits we find set. We succeed if 361 * 362 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 363 * 364 * The code is sometimes optimized not to store additional variables. 365 */ 366 367static int 368flags_match(ipfw_insn *cmd, u_int8_t bits) 369{ 370 u_char want_clear; 371 bits = ~bits; 372 373 if ( ((cmd->arg1 & 0xff) & bits) != 0) 374 return 0; /* some bits we want set were clear */ 375 want_clear = (cmd->arg1 >> 8) & 0xff; 376 if ( (want_clear & bits) != want_clear) 377 return 0; /* some bits we want clear were set */ 378 return 1; 379} 380 381static int 382ipopts_match(struct ip *ip, ipfw_insn *cmd) 383{ 384 int optlen, bits = 0; 385 u_char *cp = (u_char *)(ip + 1); 386 int x = (ip->ip_hl << 2) - sizeof (struct ip); 387 388 for (; x > 0; x -= optlen, cp += optlen) { 389 int opt = cp[IPOPT_OPTVAL]; 390 391 if (opt == IPOPT_EOL) 392 break; 393 if (opt == IPOPT_NOP) 394 optlen = 1; 395 else { 396 optlen = cp[IPOPT_OLEN]; 397 if (optlen <= 0 || optlen > x) 398 return 0; /* invalid or truncated */ 399 } 400 switch (opt) { 401 402 default: 403 break; 404 405 case IPOPT_LSRR: 406 bits |= IP_FW_IPOPT_LSRR; 407 break; 408 409 case IPOPT_SSRR: 410 bits |= IP_FW_IPOPT_SSRR; 411 break; 412 413 case IPOPT_RR: 414 bits |= IP_FW_IPOPT_RR; 415 break; 416 417 case IPOPT_TS: 418 bits |= IP_FW_IPOPT_TS; 419 break; 420 } 421 } 422 return (flags_match(cmd, bits)); 423} 424 425static int 426tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 427{ 428 int optlen, bits = 0; 429 u_char *cp = (u_char *)(tcp + 1); 430 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 431 432 for (; x > 0; x -= optlen, cp += optlen) { 433 int opt = cp[0]; 434 if (opt == TCPOPT_EOL) 435 break; 436 if (opt == TCPOPT_NOP) 437 optlen = 1; 438 else { 439 optlen = cp[1]; 440 if (optlen <= 0) 441 break; 442 } 443 444 switch (opt) { 445 446 default: 447 break; 448 449 case TCPOPT_MAXSEG: 450 bits |= IP_FW_TCPOPT_MSS; 451 break; 452 453 case TCPOPT_WINDOW: 454 bits |= IP_FW_TCPOPT_WINDOW; 455 break; 456 457 case TCPOPT_SACK_PERMITTED: 458 case TCPOPT_SACK: 459 bits |= IP_FW_TCPOPT_SACK; 460 break; 461 462 case TCPOPT_TIMESTAMP: 463 bits |= IP_FW_TCPOPT_TS; 464 break; 465 466 } 467 } 468 return (flags_match(cmd, bits)); 469} 470 471static int 472iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 473{ 474 if (ifp == NULL) /* no iface with this packet, match fails */ 475 return 0; 476 /* Check by name or by IP address */ 477 if (cmd->name[0] != '\0') { /* match by name */ 478 /* Check name */ 479 if (cmd->p.glob) { 480 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 481 return(1); 482 } else { 483 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 484 return(1); 485 } 486 } else { 487 struct ifaddr *ia; 488 489 IF_ADDR_LOCK(ifp); 490 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 491 if (ia->ifa_addr->sa_family != AF_INET) 492 continue; 493 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 494 (ia->ifa_addr))->sin_addr.s_addr) { 495 IF_ADDR_UNLOCK(ifp); 496 return(1); /* match */ 497 } 498 } 499 IF_ADDR_UNLOCK(ifp); 500 } 501 return(0); /* no match, fail ... */ 502} 503 504/* 505 * The verify_path function checks if a route to the src exists and 506 * if it is reachable via ifp (when provided). 507 * 508 * The 'verrevpath' option checks that the interface that an IP packet 509 * arrives on is the same interface that traffic destined for the 510 * packet's source address would be routed out of. The 'versrcreach' 511 * option just checks that the source address is reachable via any route 512 * (except default) in the routing table. These two are a measure to block 513 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 514 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 515 * is purposely reminiscent of the Cisco IOS command, 516 * 517 * ip verify unicast reverse-path 518 * ip verify unicast source reachable-via any 519 * 520 * which implements the same functionality. But note that syntax is 521 * misleading. The check may be performed on all IP packets whether unicast, 522 * multicast, or broadcast. 523 */ 524static int 525verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 526{ 527 struct route ro; 528 struct sockaddr_in *dst; 529 530 bzero(&ro, sizeof(ro)); 531 532 dst = (struct sockaddr_in *)&(ro.ro_dst); 533 dst->sin_family = AF_INET; 534 dst->sin_len = sizeof(*dst); 535 dst->sin_addr = src; 536 in_rtalloc_ign(&ro, 0, fib); 537 538 if (ro.ro_rt == NULL) 539 return 0; 540 541 /* 542 * If ifp is provided, check for equality with rtentry. 543 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 544 * in order to pass packets injected back by if_simloop(): 545 * if useloopback == 1 routing entry (via lo0) for our own address 546 * may exist, so we need to handle routing assymetry. 547 */ 548 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 549 RTFREE(ro.ro_rt); 550 return 0; 551 } 552 553 /* if no ifp provided, check if rtentry is not default route */ 554 if (ifp == NULL && 555 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 556 RTFREE(ro.ro_rt); 557 return 0; 558 } 559 560 /* or if this is a blackhole/reject route */ 561 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 562 RTFREE(ro.ro_rt); 563 return 0; 564 } 565 566 /* found valid route */ 567 RTFREE(ro.ro_rt); 568 return 1; 569} 570 571#ifdef INET6 572/* 573 * ipv6 specific rules here... 574 */ 575static __inline int 576icmp6type_match (int type, ipfw_insn_u32 *cmd) 577{ 578 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 579} 580 581static int 582flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 583{ 584 int i; 585 for (i=0; i <= cmd->o.arg1; ++i ) 586 if (curr_flow == cmd->d[i] ) 587 return 1; 588 return 0; 589} 590 591/* support for IP6_*_ME opcodes */ 592static int 593search_ip6_addr_net (struct in6_addr * ip6_addr) 594{ 595 INIT_VNET_NET(curvnet); 596 struct ifnet *mdc; 597 struct ifaddr *mdc2; 598 struct in6_ifaddr *fdm; 599 struct in6_addr copia; 600 601 TAILQ_FOREACH(mdc, &V_ifnet, if_link) { 602 IF_ADDR_LOCK(mdc); 603 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) { 604 if (mdc2->ifa_addr->sa_family == AF_INET6) { 605 fdm = (struct in6_ifaddr *)mdc2; 606 copia = fdm->ia_addr.sin6_addr; 607 /* need for leaving scope_id in the sock_addr */ 608 in6_clearscope(&copia); 609 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { 610 IF_ADDR_UNLOCK(mdc); 611 return 1; 612 } 613 } 614 } 615 IF_ADDR_UNLOCK(mdc); 616 } 617 return 0; 618} 619 620static int 621verify_path6(struct in6_addr *src, struct ifnet *ifp) 622{ 623 struct route_in6 ro; 624 struct sockaddr_in6 *dst; 625 626 bzero(&ro, sizeof(ro)); 627 628 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 629 dst->sin6_family = AF_INET6; 630 dst->sin6_len = sizeof(*dst); 631 dst->sin6_addr = *src; 632 /* XXX MRT 0 for ipv6 at this time */ 633 rtalloc_ign((struct route *)&ro, 0); 634 635 if (ro.ro_rt == NULL) 636 return 0; 637 638 /* 639 * if ifp is provided, check for equality with rtentry 640 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 641 * to support the case of sending packets to an address of our own. 642 * (where the former interface is the first argument of if_simloop() 643 * (=ifp), the latter is lo0) 644 */ 645 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 646 RTFREE(ro.ro_rt); 647 return 0; 648 } 649 650 /* if no ifp provided, check if rtentry is not default route */ 651 if (ifp == NULL && 652 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 653 RTFREE(ro.ro_rt); 654 return 0; 655 } 656 657 /* or if this is a blackhole/reject route */ 658 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 659 RTFREE(ro.ro_rt); 660 return 0; 661 } 662 663 /* found valid route */ 664 RTFREE(ro.ro_rt); 665 return 1; 666 667} 668static __inline int 669hash_packet6(struct ipfw_flow_id *id) 670{ 671 u_int32_t i; 672 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 673 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 674 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 675 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 676 (id->dst_port) ^ (id->src_port); 677 return i; 678} 679 680static int 681is_icmp6_query(int icmp6_type) 682{ 683 if ((icmp6_type <= ICMP6_MAXTYPE) && 684 (icmp6_type == ICMP6_ECHO_REQUEST || 685 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 686 icmp6_type == ICMP6_WRUREQUEST || 687 icmp6_type == ICMP6_FQDN_QUERY || 688 icmp6_type == ICMP6_NI_QUERY)) 689 return (1); 690 691 return (0); 692} 693 694static void 695send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 696{ 697 struct mbuf *m; 698 699 m = args->m; 700 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 701 struct tcphdr *tcp; 702 tcp_seq ack, seq; 703 int flags; 704 struct { 705 struct ip6_hdr ip6; 706 struct tcphdr th; 707 } ti; 708 tcp = (struct tcphdr *)((char *)ip6 + hlen); 709 710 if ((tcp->th_flags & TH_RST) != 0) { 711 m_freem(m); 712 args->m = NULL; 713 return; 714 } 715 716 ti.ip6 = *ip6; 717 ti.th = *tcp; 718 ti.th.th_seq = ntohl(ti.th.th_seq); 719 ti.th.th_ack = ntohl(ti.th.th_ack); 720 ti.ip6.ip6_nxt = IPPROTO_TCP; 721 722 if (ti.th.th_flags & TH_ACK) { 723 ack = 0; 724 seq = ti.th.th_ack; 725 flags = TH_RST; 726 } else { 727 ack = ti.th.th_seq; 728 if ((m->m_flags & M_PKTHDR) != 0) { 729 /* 730 * total new data to ACK is: 731 * total packet length, 732 * minus the header length, 733 * minus the tcp header length. 734 */ 735 ack += m->m_pkthdr.len - hlen 736 - (ti.th.th_off << 2); 737 } else if (ip6->ip6_plen) { 738 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 739 hlen - (ti.th.th_off << 2); 740 } else { 741 m_freem(m); 742 return; 743 } 744 if (tcp->th_flags & TH_SYN) 745 ack++; 746 seq = 0; 747 flags = TH_RST|TH_ACK; 748 } 749 bcopy(&ti, ip6, sizeof(ti)); 750 /* 751 * m is only used to recycle the mbuf 752 * The data in it is never read so we don't need 753 * to correct the offsets or anything 754 */ 755 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 756 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 757#if 0 758 /* 759 * Unlike above, the mbufs need to line up with the ip6 hdr, 760 * as the contents are read. We need to m_adj() the 761 * needed amount. 762 * The mbuf will however be thrown away so we can adjust it. 763 * Remember we did an m_pullup on it already so we 764 * can make some assumptions about contiguousness. 765 */ 766 if (args->L3offset) 767 m_adj(m, args->L3offset); 768#endif 769 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 770 } else 771 m_freem(m); 772 773 args->m = NULL; 774} 775 776#endif /* INET6 */ 777 778#ifdef VIMAGE_GLOBALS 779static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 780#endif 781 782#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 783#define SNP(buf) buf, sizeof(buf) 784 785/* 786 * We enter here when we have a rule with O_LOG. 787 * XXX this function alone takes about 2Kbytes of code! 788 */ 789static void 790ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 791 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 792 struct ip *ip) 793{ 794 INIT_VNET_IPFW(curvnet); 795 struct ether_header *eh = args->eh; 796 char *action; 797 int limit_reached = 0; 798 char action2[40], proto[128], fragment[32]; 799 800 fragment[0] = '\0'; 801 proto[0] = '\0'; 802 803 if (f == NULL) { /* bogus pkt */ 804 if (V_verbose_limit != 0 && V_norule_counter >= V_verbose_limit) 805 return; 806 V_norule_counter++; 807 if (V_norule_counter == V_verbose_limit) 808 limit_reached = V_verbose_limit; 809 action = "Refuse"; 810 } else { /* O_LOG is the first action, find the real one */ 811 ipfw_insn *cmd = ACTION_PTR(f); 812 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 813 814 if (l->max_log != 0 && l->log_left == 0) 815 return; 816 l->log_left--; 817 if (l->log_left == 0) 818 limit_reached = l->max_log; 819 cmd += F_LEN(cmd); /* point to first action */ 820 if (cmd->opcode == O_ALTQ) { 821 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 822 823 snprintf(SNPARGS(action2, 0), "Altq %d", 824 altq->qid); 825 cmd += F_LEN(cmd); 826 } 827 if (cmd->opcode == O_PROB) 828 cmd += F_LEN(cmd); 829 830 if (cmd->opcode == O_TAG) 831 cmd += F_LEN(cmd); 832 833 action = action2; 834 switch (cmd->opcode) { 835 case O_DENY: 836 action = "Deny"; 837 break; 838 839 case O_REJECT: 840 if (cmd->arg1==ICMP_REJECT_RST) 841 action = "Reset"; 842 else if (cmd->arg1==ICMP_UNREACH_HOST) 843 action = "Reject"; 844 else 845 snprintf(SNPARGS(action2, 0), "Unreach %d", 846 cmd->arg1); 847 break; 848 849 case O_UNREACH6: 850 if (cmd->arg1==ICMP6_UNREACH_RST) 851 action = "Reset"; 852 else 853 snprintf(SNPARGS(action2, 0), "Unreach %d", 854 cmd->arg1); 855 break; 856 857 case O_ACCEPT: 858 action = "Accept"; 859 break; 860 case O_COUNT: 861 action = "Count"; 862 break; 863 case O_DIVERT: 864 snprintf(SNPARGS(action2, 0), "Divert %d", 865 cmd->arg1); 866 break; 867 case O_TEE: 868 snprintf(SNPARGS(action2, 0), "Tee %d", 869 cmd->arg1); 870 break; 871 case O_SETFIB: 872 snprintf(SNPARGS(action2, 0), "SetFib %d", 873 cmd->arg1); 874 break; 875 case O_SKIPTO: 876 snprintf(SNPARGS(action2, 0), "SkipTo %d", 877 cmd->arg1); 878 break; 879 case O_PIPE: 880 snprintf(SNPARGS(action2, 0), "Pipe %d", 881 cmd->arg1); 882 break; 883 case O_QUEUE: 884 snprintf(SNPARGS(action2, 0), "Queue %d", 885 cmd->arg1); 886 break; 887 case O_FORWARD_IP: { 888 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 889 int len; 890 struct in_addr dummyaddr; 891 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 892 dummyaddr.s_addr = htonl(tablearg); 893 else 894 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 895 896 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 897 inet_ntoa(dummyaddr)); 898 899 if (sa->sa.sin_port) 900 snprintf(SNPARGS(action2, len), ":%d", 901 sa->sa.sin_port); 902 } 903 break; 904 case O_NETGRAPH: 905 snprintf(SNPARGS(action2, 0), "Netgraph %d", 906 cmd->arg1); 907 break; 908 case O_NGTEE: 909 snprintf(SNPARGS(action2, 0), "Ngtee %d", 910 cmd->arg1); 911 break; 912 case O_NAT: 913 action = "Nat"; 914 break; 915 case O_REASS: 916 action = "Reass"; 917 break; 918 default: 919 action = "UNKNOWN"; 920 break; 921 } 922 } 923 924 if (hlen == 0) { /* non-ip */ 925 snprintf(SNPARGS(proto, 0), "MAC"); 926 927 } else { 928 int len; 929 char src[48], dst[48]; 930 struct icmphdr *icmp; 931 struct tcphdr *tcp; 932 struct udphdr *udp; 933#ifdef INET6 934 struct ip6_hdr *ip6 = NULL; 935 struct icmp6_hdr *icmp6; 936#endif 937 src[0] = '\0'; 938 dst[0] = '\0'; 939#ifdef INET6 940 if (IS_IP6_FLOW_ID(&(args->f_id))) { 941 char ip6buf[INET6_ADDRSTRLEN]; 942 snprintf(src, sizeof(src), "[%s]", 943 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 944 snprintf(dst, sizeof(dst), "[%s]", 945 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 946 947 ip6 = (struct ip6_hdr *)ip; 948 tcp = (struct tcphdr *)(((char *)ip) + hlen); 949 udp = (struct udphdr *)(((char *)ip) + hlen); 950 } else 951#endif 952 { 953 tcp = L3HDR(struct tcphdr, ip); 954 udp = L3HDR(struct udphdr, ip); 955 956 inet_ntoa_r(ip->ip_src, src); 957 inet_ntoa_r(ip->ip_dst, dst); 958 } 959 960 switch (args->f_id.proto) { 961 case IPPROTO_TCP: 962 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 963 if (offset == 0) 964 snprintf(SNPARGS(proto, len), ":%d %s:%d", 965 ntohs(tcp->th_sport), 966 dst, 967 ntohs(tcp->th_dport)); 968 else 969 snprintf(SNPARGS(proto, len), " %s", dst); 970 break; 971 972 case IPPROTO_UDP: 973 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 974 if (offset == 0) 975 snprintf(SNPARGS(proto, len), ":%d %s:%d", 976 ntohs(udp->uh_sport), 977 dst, 978 ntohs(udp->uh_dport)); 979 else 980 snprintf(SNPARGS(proto, len), " %s", dst); 981 break; 982 983 case IPPROTO_ICMP: 984 icmp = L3HDR(struct icmphdr, ip); 985 if (offset == 0) 986 len = snprintf(SNPARGS(proto, 0), 987 "ICMP:%u.%u ", 988 icmp->icmp_type, icmp->icmp_code); 989 else 990 len = snprintf(SNPARGS(proto, 0), "ICMP "); 991 len += snprintf(SNPARGS(proto, len), "%s", src); 992 snprintf(SNPARGS(proto, len), " %s", dst); 993 break; 994#ifdef INET6 995 case IPPROTO_ICMPV6: 996 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 997 if (offset == 0) 998 len = snprintf(SNPARGS(proto, 0), 999 "ICMPv6:%u.%u ", 1000 icmp6->icmp6_type, icmp6->icmp6_code); 1001 else 1002 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 1003 len += snprintf(SNPARGS(proto, len), "%s", src); 1004 snprintf(SNPARGS(proto, len), " %s", dst); 1005 break; 1006#endif 1007 default: 1008 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 1009 args->f_id.proto, src); 1010 snprintf(SNPARGS(proto, len), " %s", dst); 1011 break; 1012 } 1013 1014#ifdef INET6 1015 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1016 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 1017 snprintf(SNPARGS(fragment, 0), 1018 " (frag %08x:%d@%d%s)", 1019 args->f_id.frag_id6, 1020 ntohs(ip6->ip6_plen) - hlen, 1021 ntohs(offset & IP6F_OFF_MASK) << 3, 1022 (offset & IP6F_MORE_FRAG) ? "+" : ""); 1023 } else 1024#endif 1025 { 1026 int ip_off, ip_len; 1027 if (eh != NULL) { /* layer 2 packets are as on the wire */ 1028 ip_off = ntohs(ip->ip_off); 1029 ip_len = ntohs(ip->ip_len); 1030 } else { 1031 ip_off = ip->ip_off; 1032 ip_len = ip->ip_len; 1033 } 1034 if (ip_off & (IP_MF | IP_OFFMASK)) 1035 snprintf(SNPARGS(fragment, 0), 1036 " (frag %d:%d@%d%s)", 1037 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1038 offset << 3, 1039 (ip_off & IP_MF) ? "+" : ""); 1040 } 1041 } 1042 if (oif || m->m_pkthdr.rcvif) 1043 log(LOG_SECURITY | LOG_INFO, 1044 "ipfw: %d %s %s %s via %s%s\n", 1045 f ? f->rulenum : -1, 1046 action, proto, oif ? "out" : "in", 1047 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1048 fragment); 1049 else 1050 log(LOG_SECURITY | LOG_INFO, 1051 "ipfw: %d %s %s [no if info]%s\n", 1052 f ? f->rulenum : -1, 1053 action, proto, fragment); 1054 if (limit_reached) 1055 log(LOG_SECURITY | LOG_NOTICE, 1056 "ipfw: limit %d reached on entry %d\n", 1057 limit_reached, f ? f->rulenum : -1); 1058} 1059 1060/* 1061 * IMPORTANT: the hash function for dynamic rules must be commutative 1062 * in source and destination (ip,port), because rules are bidirectional 1063 * and we want to find both in the same bucket. 1064 */ 1065static __inline int 1066hash_packet(struct ipfw_flow_id *id) 1067{ 1068 INIT_VNET_IPFW(curvnet); 1069 u_int32_t i; 1070 1071#ifdef INET6 1072 if (IS_IP6_FLOW_ID(id)) 1073 i = hash_packet6(id); 1074 else 1075#endif /* INET6 */ 1076 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1077 i &= (V_curr_dyn_buckets - 1); 1078 return i; 1079} 1080 1081/** 1082 * unlink a dynamic rule from a chain. prev is a pointer to 1083 * the previous one, q is a pointer to the rule to delete, 1084 * head is a pointer to the head of the queue. 1085 * Modifies q and potentially also head. 1086 */ 1087#define UNLINK_DYN_RULE(prev, head, q) { \ 1088 ipfw_dyn_rule *old_q = q; \ 1089 \ 1090 /* remove a refcount to the parent */ \ 1091 if (q->dyn_type == O_LIMIT) \ 1092 q->parent->count--; \ 1093 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1094 (q->id.src_ip), (q->id.src_port), \ 1095 (q->id.dst_ip), (q->id.dst_port), V_dyn_count-1 ); ) \ 1096 if (prev != NULL) \ 1097 prev->next = q = q->next; \ 1098 else \ 1099 head = q = q->next; \ 1100 V_dyn_count--; \ 1101 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1102 1103#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1104 1105/** 1106 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1107 * 1108 * If keep_me == NULL, rules are deleted even if not expired, 1109 * otherwise only expired rules are removed. 1110 * 1111 * The value of the second parameter is also used to point to identify 1112 * a rule we absolutely do not want to remove (e.g. because we are 1113 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1114 * rules). The pointer is only used for comparison, so any non-null 1115 * value will do. 1116 */ 1117static void 1118remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1119{ 1120 INIT_VNET_IPFW(curvnet); 1121 static u_int32_t last_remove = 0; 1122 1123#define FORCE (keep_me == NULL) 1124 1125 ipfw_dyn_rule *prev, *q; 1126 int i, pass = 0, max_pass = 0; 1127 1128 IPFW_DYN_LOCK_ASSERT(); 1129 1130 if (V_ipfw_dyn_v == NULL || V_dyn_count == 0) 1131 return; 1132 /* do not expire more than once per second, it is useless */ 1133 if (!FORCE && last_remove == time_uptime) 1134 return; 1135 last_remove = time_uptime; 1136 1137 /* 1138 * because O_LIMIT refer to parent rules, during the first pass only 1139 * remove child and mark any pending LIMIT_PARENT, and remove 1140 * them in a second pass. 1141 */ 1142next_pass: 1143 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 1144 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q ; ) { 1145 /* 1146 * Logic can become complex here, so we split tests. 1147 */ 1148 if (q == keep_me) 1149 goto next; 1150 if (rule != NULL && rule != q->rule) 1151 goto next; /* not the one we are looking for */ 1152 if (q->dyn_type == O_LIMIT_PARENT) { 1153 /* 1154 * handle parent in the second pass, 1155 * record we need one. 1156 */ 1157 max_pass = 1; 1158 if (pass == 0) 1159 goto next; 1160 if (FORCE && q->count != 0 ) { 1161 /* XXX should not happen! */ 1162 printf("ipfw: OUCH! cannot remove rule," 1163 " count %d\n", q->count); 1164 } 1165 } else { 1166 if (!FORCE && 1167 !TIME_LEQ( q->expire, time_uptime )) 1168 goto next; 1169 } 1170 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1171 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1172 continue; 1173 } 1174next: 1175 prev=q; 1176 q=q->next; 1177 } 1178 } 1179 if (pass++ < max_pass) 1180 goto next_pass; 1181} 1182 1183 1184/** 1185 * lookup a dynamic rule. 1186 */ 1187static ipfw_dyn_rule * 1188lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1189 struct tcphdr *tcp) 1190{ 1191 INIT_VNET_IPFW(curvnet); 1192 /* 1193 * stateful ipfw extensions. 1194 * Lookup into dynamic session queue 1195 */ 1196#define MATCH_REVERSE 0 1197#define MATCH_FORWARD 1 1198#define MATCH_NONE 2 1199#define MATCH_UNKNOWN 3 1200 int i, dir = MATCH_NONE; 1201 ipfw_dyn_rule *prev, *q=NULL; 1202 1203 IPFW_DYN_LOCK_ASSERT(); 1204 1205 if (V_ipfw_dyn_v == NULL) 1206 goto done; /* not found */ 1207 i = hash_packet( pkt ); 1208 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q != NULL ; ) { 1209 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1210 goto next; 1211 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1212 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1213 continue; 1214 } 1215 if (pkt->proto == q->id.proto && 1216 q->dyn_type != O_LIMIT_PARENT) { 1217 if (IS_IP6_FLOW_ID(pkt)) { 1218 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1219 &(q->id.src_ip6)) && 1220 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1221 &(q->id.dst_ip6)) && 1222 pkt->src_port == q->id.src_port && 1223 pkt->dst_port == q->id.dst_port ) { 1224 dir = MATCH_FORWARD; 1225 break; 1226 } 1227 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1228 &(q->id.dst_ip6)) && 1229 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1230 &(q->id.src_ip6)) && 1231 pkt->src_port == q->id.dst_port && 1232 pkt->dst_port == q->id.src_port ) { 1233 dir = MATCH_REVERSE; 1234 break; 1235 } 1236 } else { 1237 if (pkt->src_ip == q->id.src_ip && 1238 pkt->dst_ip == q->id.dst_ip && 1239 pkt->src_port == q->id.src_port && 1240 pkt->dst_port == q->id.dst_port ) { 1241 dir = MATCH_FORWARD; 1242 break; 1243 } 1244 if (pkt->src_ip == q->id.dst_ip && 1245 pkt->dst_ip == q->id.src_ip && 1246 pkt->src_port == q->id.dst_port && 1247 pkt->dst_port == q->id.src_port ) { 1248 dir = MATCH_REVERSE; 1249 break; 1250 } 1251 } 1252 } 1253next: 1254 prev = q; 1255 q = q->next; 1256 } 1257 if (q == NULL) 1258 goto done; /* q = NULL, not found */ 1259 1260 if ( prev != NULL) { /* found and not in front */ 1261 prev->next = q->next; 1262 q->next = V_ipfw_dyn_v[i]; 1263 V_ipfw_dyn_v[i] = q; 1264 } 1265 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1266 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1267 1268#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1269#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1270 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1271 switch (q->state) { 1272 case TH_SYN: /* opening */ 1273 q->expire = time_uptime + V_dyn_syn_lifetime; 1274 break; 1275 1276 case BOTH_SYN: /* move to established */ 1277 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1278 case BOTH_SYN | (TH_FIN << 8) : 1279 if (tcp) { 1280#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1281 u_int32_t ack = ntohl(tcp->th_ack); 1282 if (dir == MATCH_FORWARD) { 1283 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1284 q->ack_fwd = ack; 1285 else { /* ignore out-of-sequence */ 1286 break; 1287 } 1288 } else { 1289 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1290 q->ack_rev = ack; 1291 else { /* ignore out-of-sequence */ 1292 break; 1293 } 1294 } 1295 } 1296 q->expire = time_uptime + V_dyn_ack_lifetime; 1297 break; 1298 1299 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1300 if (V_dyn_fin_lifetime >= V_dyn_keepalive_period) 1301 V_dyn_fin_lifetime = V_dyn_keepalive_period - 1; 1302 q->expire = time_uptime + V_dyn_fin_lifetime; 1303 break; 1304 1305 default: 1306#if 0 1307 /* 1308 * reset or some invalid combination, but can also 1309 * occur if we use keep-state the wrong way. 1310 */ 1311 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1312 printf("invalid state: 0x%x\n", q->state); 1313#endif 1314 if (V_dyn_rst_lifetime >= V_dyn_keepalive_period) 1315 V_dyn_rst_lifetime = V_dyn_keepalive_period - 1; 1316 q->expire = time_uptime + V_dyn_rst_lifetime; 1317 break; 1318 } 1319 } else if (pkt->proto == IPPROTO_UDP) { 1320 q->expire = time_uptime + V_dyn_udp_lifetime; 1321 } else { 1322 /* other protocols */ 1323 q->expire = time_uptime + V_dyn_short_lifetime; 1324 } 1325done: 1326 if (match_direction) 1327 *match_direction = dir; 1328 return q; 1329} 1330 1331static ipfw_dyn_rule * 1332lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1333 struct tcphdr *tcp) 1334{ 1335 ipfw_dyn_rule *q; 1336 1337 IPFW_DYN_LOCK(); 1338 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1339 if (q == NULL) 1340 IPFW_DYN_UNLOCK(); 1341 /* NB: return table locked when q is not NULL */ 1342 return q; 1343} 1344 1345static void 1346realloc_dynamic_table(void) 1347{ 1348 INIT_VNET_IPFW(curvnet); 1349 IPFW_DYN_LOCK_ASSERT(); 1350 1351 /* 1352 * Try reallocation, make sure we have a power of 2 and do 1353 * not allow more than 64k entries. In case of overflow, 1354 * default to 1024. 1355 */ 1356 1357 if (V_dyn_buckets > 65536) 1358 V_dyn_buckets = 1024; 1359 if ((V_dyn_buckets & (V_dyn_buckets-1)) != 0) { /* not a power of 2 */ 1360 V_dyn_buckets = V_curr_dyn_buckets; /* reset */ 1361 return; 1362 } 1363 V_curr_dyn_buckets = V_dyn_buckets; 1364 if (V_ipfw_dyn_v != NULL) 1365 free(V_ipfw_dyn_v, M_IPFW); 1366 for (;;) { 1367 V_ipfw_dyn_v = malloc(V_curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1368 M_IPFW, M_NOWAIT | M_ZERO); 1369 if (V_ipfw_dyn_v != NULL || V_curr_dyn_buckets <= 2) 1370 break; 1371 V_curr_dyn_buckets /= 2; 1372 } 1373} 1374 1375/** 1376 * Install state of type 'type' for a dynamic session. 1377 * The hash table contains two type of rules: 1378 * - regular rules (O_KEEP_STATE) 1379 * - rules for sessions with limited number of sess per user 1380 * (O_LIMIT). When they are created, the parent is 1381 * increased by 1, and decreased on delete. In this case, 1382 * the third parameter is the parent rule and not the chain. 1383 * - "parent" rules for the above (O_LIMIT_PARENT). 1384 */ 1385static ipfw_dyn_rule * 1386add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1387{ 1388 INIT_VNET_IPFW(curvnet); 1389 ipfw_dyn_rule *r; 1390 int i; 1391 1392 IPFW_DYN_LOCK_ASSERT(); 1393 1394 if (V_ipfw_dyn_v == NULL || 1395 (V_dyn_count == 0 && V_dyn_buckets != V_curr_dyn_buckets)) { 1396 realloc_dynamic_table(); 1397 if (V_ipfw_dyn_v == NULL) 1398 return NULL; /* failed ! */ 1399 } 1400 i = hash_packet(id); 1401 1402 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1403 if (r == NULL) { 1404 printf ("ipfw: sorry cannot allocate state\n"); 1405 return NULL; 1406 } 1407 1408 /* increase refcount on parent, and set pointer */ 1409 if (dyn_type == O_LIMIT) { 1410 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1411 if ( parent->dyn_type != O_LIMIT_PARENT) 1412 panic("invalid parent"); 1413 parent->count++; 1414 r->parent = parent; 1415 rule = parent->rule; 1416 } 1417 1418 r->id = *id; 1419 r->expire = time_uptime + V_dyn_syn_lifetime; 1420 r->rule = rule; 1421 r->dyn_type = dyn_type; 1422 r->pcnt = r->bcnt = 0; 1423 r->count = 0; 1424 1425 r->bucket = i; 1426 r->next = V_ipfw_dyn_v[i]; 1427 V_ipfw_dyn_v[i] = r; 1428 V_dyn_count++; 1429 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1430 dyn_type, 1431 (r->id.src_ip), (r->id.src_port), 1432 (r->id.dst_ip), (r->id.dst_port), 1433 V_dyn_count ); ) 1434 return r; 1435} 1436 1437/** 1438 * lookup dynamic parent rule using pkt and rule as search keys. 1439 * If the lookup fails, then install one. 1440 */ 1441static ipfw_dyn_rule * 1442lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1443{ 1444 INIT_VNET_IPFW(curvnet); 1445 ipfw_dyn_rule *q; 1446 int i; 1447 1448 IPFW_DYN_LOCK_ASSERT(); 1449 1450 if (V_ipfw_dyn_v) { 1451 int is_v6 = IS_IP6_FLOW_ID(pkt); 1452 i = hash_packet( pkt ); 1453 for (q = V_ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1454 if (q->dyn_type == O_LIMIT_PARENT && 1455 rule== q->rule && 1456 pkt->proto == q->id.proto && 1457 pkt->src_port == q->id.src_port && 1458 pkt->dst_port == q->id.dst_port && 1459 ( 1460 (is_v6 && 1461 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1462 &(q->id.src_ip6)) && 1463 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1464 &(q->id.dst_ip6))) || 1465 (!is_v6 && 1466 pkt->src_ip == q->id.src_ip && 1467 pkt->dst_ip == q->id.dst_ip) 1468 ) 1469 ) { 1470 q->expire = time_uptime + V_dyn_short_lifetime; 1471 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1472 return q; 1473 } 1474 } 1475 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1476} 1477 1478/** 1479 * Install dynamic state for rule type cmd->o.opcode 1480 * 1481 * Returns 1 (failure) if state is not installed because of errors or because 1482 * session limitations are enforced. 1483 */ 1484static int 1485install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1486 struct ip_fw_args *args, uint32_t tablearg) 1487{ 1488 INIT_VNET_IPFW(curvnet); 1489 static int last_log; 1490 ipfw_dyn_rule *q; 1491 struct in_addr da; 1492 char src[48], dst[48]; 1493 1494 src[0] = '\0'; 1495 dst[0] = '\0'; 1496 1497 DEB( 1498 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1499 __func__, cmd->o.opcode, 1500 (args->f_id.src_ip), (args->f_id.src_port), 1501 (args->f_id.dst_ip), (args->f_id.dst_port)); 1502 ) 1503 1504 IPFW_DYN_LOCK(); 1505 1506 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1507 1508 if (q != NULL) { /* should never occur */ 1509 if (last_log != time_uptime) { 1510 last_log = time_uptime; 1511 printf("ipfw: %s: entry already present, done\n", 1512 __func__); 1513 } 1514 IPFW_DYN_UNLOCK(); 1515 return (0); 1516 } 1517 1518 if (V_dyn_count >= V_dyn_max) 1519 /* Run out of slots, try to remove any expired rule. */ 1520 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1521 1522 if (V_dyn_count >= V_dyn_max) { 1523 if (last_log != time_uptime) { 1524 last_log = time_uptime; 1525 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1526 } 1527 IPFW_DYN_UNLOCK(); 1528 return (1); /* cannot install, notify caller */ 1529 } 1530 1531 switch (cmd->o.opcode) { 1532 case O_KEEP_STATE: /* bidir rule */ 1533 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1534 break; 1535 1536 case O_LIMIT: { /* limit number of sessions */ 1537 struct ipfw_flow_id id; 1538 ipfw_dyn_rule *parent; 1539 uint32_t conn_limit; 1540 uint16_t limit_mask = cmd->limit_mask; 1541 1542 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1543 tablearg : cmd->conn_limit; 1544 1545 DEB( 1546 if (cmd->conn_limit == IP_FW_TABLEARG) 1547 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1548 "(tablearg)\n", __func__, conn_limit); 1549 else 1550 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1551 __func__, conn_limit); 1552 ) 1553 1554 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1555 id.proto = args->f_id.proto; 1556 id.addr_type = args->f_id.addr_type; 1557 id.fib = M_GETFIB(args->m); 1558 1559 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1560 if (limit_mask & DYN_SRC_ADDR) 1561 id.src_ip6 = args->f_id.src_ip6; 1562 if (limit_mask & DYN_DST_ADDR) 1563 id.dst_ip6 = args->f_id.dst_ip6; 1564 } else { 1565 if (limit_mask & DYN_SRC_ADDR) 1566 id.src_ip = args->f_id.src_ip; 1567 if (limit_mask & DYN_DST_ADDR) 1568 id.dst_ip = args->f_id.dst_ip; 1569 } 1570 if (limit_mask & DYN_SRC_PORT) 1571 id.src_port = args->f_id.src_port; 1572 if (limit_mask & DYN_DST_PORT) 1573 id.dst_port = args->f_id.dst_port; 1574 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1575 printf("ipfw: %s: add parent failed\n", __func__); 1576 IPFW_DYN_UNLOCK(); 1577 return (1); 1578 } 1579 1580 if (parent->count >= conn_limit) { 1581 /* See if we can remove some expired rule. */ 1582 remove_dyn_rule(rule, parent); 1583 if (parent->count >= conn_limit) { 1584 if (V_fw_verbose && last_log != time_uptime) { 1585 last_log = time_uptime; 1586#ifdef INET6 1587 /* 1588 * XXX IPv6 flows are not 1589 * supported yet. 1590 */ 1591 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1592 char ip6buf[INET6_ADDRSTRLEN]; 1593 snprintf(src, sizeof(src), 1594 "[%s]", ip6_sprintf(ip6buf, 1595 &args->f_id.src_ip6)); 1596 snprintf(dst, sizeof(dst), 1597 "[%s]", ip6_sprintf(ip6buf, 1598 &args->f_id.dst_ip6)); 1599 } else 1600#endif 1601 { 1602 da.s_addr = 1603 htonl(args->f_id.src_ip); 1604 inet_ntoa_r(da, src); 1605 da.s_addr = 1606 htonl(args->f_id.dst_ip); 1607 inet_ntoa_r(da, dst); 1608 } 1609 log(LOG_SECURITY | LOG_DEBUG, 1610 "ipfw: %d %s %s:%u -> %s:%u, %s\n", 1611 parent->rule->rulenum, 1612 "drop session", 1613 src, (args->f_id.src_port), 1614 dst, (args->f_id.dst_port), 1615 "too many entries"); 1616 } 1617 IPFW_DYN_UNLOCK(); 1618 return (1); 1619 } 1620 } 1621 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1622 break; 1623 } 1624 default: 1625 printf("ipfw: %s: unknown dynamic rule type %u\n", 1626 __func__, cmd->o.opcode); 1627 IPFW_DYN_UNLOCK(); 1628 return (1); 1629 } 1630 1631 /* XXX just set lifetime */ 1632 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1633 1634 IPFW_DYN_UNLOCK(); 1635 return (0); 1636} 1637 1638/* 1639 * Generate a TCP packet, containing either a RST or a keepalive. 1640 * When flags & TH_RST, we are sending a RST packet, because of a 1641 * "reset" action matched the packet. 1642 * Otherwise we are sending a keepalive, and flags & TH_ 1643 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1644 * so that MAC can label the reply appropriately. 1645 */ 1646static struct mbuf * 1647send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1648 u_int32_t ack, int flags) 1649{ 1650 INIT_VNET_INET(curvnet); 1651 struct mbuf *m; 1652 struct ip *ip; 1653 struct tcphdr *tcp; 1654 1655 MGETHDR(m, M_DONTWAIT, MT_DATA); 1656 if (m == 0) 1657 return (NULL); 1658 m->m_pkthdr.rcvif = (struct ifnet *)0; 1659 1660 M_SETFIB(m, id->fib); 1661#ifdef MAC 1662 if (replyto != NULL) 1663 mac_netinet_firewall_reply(replyto, m); 1664 else 1665 mac_netinet_firewall_send(m); 1666#else 1667 (void)replyto; /* don't warn about unused arg */ 1668#endif 1669 1670 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1671 m->m_data += max_linkhdr; 1672 1673 ip = mtod(m, struct ip *); 1674 bzero(ip, m->m_len); 1675 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1676 ip->ip_p = IPPROTO_TCP; 1677 tcp->th_off = 5; 1678 /* 1679 * Assume we are sending a RST (or a keepalive in the reverse 1680 * direction), swap src and destination addresses and ports. 1681 */ 1682 ip->ip_src.s_addr = htonl(id->dst_ip); 1683 ip->ip_dst.s_addr = htonl(id->src_ip); 1684 tcp->th_sport = htons(id->dst_port); 1685 tcp->th_dport = htons(id->src_port); 1686 if (flags & TH_RST) { /* we are sending a RST */ 1687 if (flags & TH_ACK) { 1688 tcp->th_seq = htonl(ack); 1689 tcp->th_ack = htonl(0); 1690 tcp->th_flags = TH_RST; 1691 } else { 1692 if (flags & TH_SYN) 1693 seq++; 1694 tcp->th_seq = htonl(0); 1695 tcp->th_ack = htonl(seq); 1696 tcp->th_flags = TH_RST | TH_ACK; 1697 } 1698 } else { 1699 /* 1700 * We are sending a keepalive. flags & TH_SYN determines 1701 * the direction, forward if set, reverse if clear. 1702 * NOTE: seq and ack are always assumed to be correct 1703 * as set by the caller. This may be confusing... 1704 */ 1705 if (flags & TH_SYN) { 1706 /* 1707 * we have to rewrite the correct addresses! 1708 */ 1709 ip->ip_dst.s_addr = htonl(id->dst_ip); 1710 ip->ip_src.s_addr = htonl(id->src_ip); 1711 tcp->th_dport = htons(id->dst_port); 1712 tcp->th_sport = htons(id->src_port); 1713 } 1714 tcp->th_seq = htonl(seq); 1715 tcp->th_ack = htonl(ack); 1716 tcp->th_flags = TH_ACK; 1717 } 1718 /* 1719 * set ip_len to the payload size so we can compute 1720 * the tcp checksum on the pseudoheader 1721 * XXX check this, could save a couple of words ? 1722 */ 1723 ip->ip_len = htons(sizeof(struct tcphdr)); 1724 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1725 /* 1726 * now fill fields left out earlier 1727 */ 1728 ip->ip_ttl = V_ip_defttl; 1729 ip->ip_len = m->m_pkthdr.len; 1730 m->m_flags |= M_SKIP_FIREWALL; 1731 return (m); 1732} 1733 1734/* 1735 * sends a reject message, consuming the mbuf passed as an argument. 1736 */ 1737static void 1738send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1739{ 1740 1741#if 0 1742 /* XXX When ip is not guaranteed to be at mtod() we will 1743 * need to account for this */ 1744 * The mbuf will however be thrown away so we can adjust it. 1745 * Remember we did an m_pullup on it already so we 1746 * can make some assumptions about contiguousness. 1747 */ 1748 if (args->L3offset) 1749 m_adj(m, args->L3offset); 1750#endif 1751 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1752 /* We need the IP header in host order for icmp_error(). */ 1753 if (args->eh != NULL) { 1754 ip->ip_len = ntohs(ip->ip_len); 1755 ip->ip_off = ntohs(ip->ip_off); 1756 } 1757 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1758 } else if (args->f_id.proto == IPPROTO_TCP) { 1759 struct tcphdr *const tcp = 1760 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1761 if ( (tcp->th_flags & TH_RST) == 0) { 1762 struct mbuf *m; 1763 m = send_pkt(args->m, &(args->f_id), 1764 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1765 tcp->th_flags | TH_RST); 1766 if (m != NULL) 1767 ip_output(m, NULL, NULL, 0, NULL, NULL); 1768 } 1769 m_freem(args->m); 1770 } else 1771 m_freem(args->m); 1772 args->m = NULL; 1773} 1774 1775/** 1776 * 1777 * Given an ip_fw *, lookup_next_rule will return a pointer 1778 * to the next rule, which can be either the jump 1779 * target (for skipto instructions) or the next one in the list (in 1780 * all other cases including a missing jump target). 1781 * The result is also written in the "next_rule" field of the rule. 1782 * Backward jumps are not allowed, so start looking from the next 1783 * rule... 1784 * 1785 * This never returns NULL -- in case we do not have an exact match, 1786 * the next rule is returned. When the ruleset is changed, 1787 * pointers are flushed so we are always correct. 1788 */ 1789 1790static struct ip_fw * 1791lookup_next_rule(struct ip_fw *me, u_int32_t tablearg) 1792{ 1793 struct ip_fw *rule = NULL; 1794 ipfw_insn *cmd; 1795 u_int16_t rulenum; 1796 1797 /* look for action, in case it is a skipto */ 1798 cmd = ACTION_PTR(me); 1799 if (cmd->opcode == O_LOG) 1800 cmd += F_LEN(cmd); 1801 if (cmd->opcode == O_ALTQ) 1802 cmd += F_LEN(cmd); 1803 if (cmd->opcode == O_TAG) 1804 cmd += F_LEN(cmd); 1805 if (cmd->opcode == O_SKIPTO ) { 1806 if (tablearg != 0) { 1807 rulenum = (u_int16_t)tablearg; 1808 } else { 1809 rulenum = cmd->arg1; 1810 } 1811 for (rule = me->next; rule ; rule = rule->next) { 1812 if (rule->rulenum >= rulenum) { 1813 break; 1814 } 1815 } 1816 } 1817 if (rule == NULL) /* failure or not a skipto */ 1818 rule = me->next; 1819 me->next_rule = rule; 1820 return rule; 1821} 1822 1823static int 1824add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1825 uint8_t mlen, uint32_t value) 1826{ 1827 struct radix_node_head *rnh; 1828 struct table_entry *ent; 1829 struct radix_node *rn; 1830 1831 if (tbl >= IPFW_TABLES_MAX) 1832 return (EINVAL); 1833 rnh = ch->tables[tbl]; 1834 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1835 if (ent == NULL) 1836 return (ENOMEM); 1837 ent->value = value; 1838 ent->addr.sin_len = ent->mask.sin_len = 8; 1839 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1840 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1841 IPFW_WLOCK(ch); 1842 rn = rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent); 1843 if (rn == NULL) { 1844 IPFW_WUNLOCK(ch); 1845 free(ent, M_IPFW_TBL); 1846 return (EEXIST); 1847 } 1848 IPFW_WUNLOCK(ch); 1849 return (0); 1850} 1851 1852static int 1853del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1854 uint8_t mlen) 1855{ 1856 struct radix_node_head *rnh; 1857 struct table_entry *ent; 1858 struct sockaddr_in sa, mask; 1859 1860 if (tbl >= IPFW_TABLES_MAX) 1861 return (EINVAL); 1862 rnh = ch->tables[tbl]; 1863 sa.sin_len = mask.sin_len = 8; 1864 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1865 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1866 IPFW_WLOCK(ch); 1867 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1868 if (ent == NULL) { 1869 IPFW_WUNLOCK(ch); 1870 return (ESRCH); 1871 } 1872 IPFW_WUNLOCK(ch); 1873 free(ent, M_IPFW_TBL); 1874 return (0); 1875} 1876 1877static int 1878flush_table_entry(struct radix_node *rn, void *arg) 1879{ 1880 struct radix_node_head * const rnh = arg; 1881 struct table_entry *ent; 1882 1883 ent = (struct table_entry *) 1884 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1885 if (ent != NULL) 1886 free(ent, M_IPFW_TBL); 1887 return (0); 1888} 1889 1890static int 1891flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1892{ 1893 struct radix_node_head *rnh; 1894 1895 IPFW_WLOCK_ASSERT(ch); 1896 1897 if (tbl >= IPFW_TABLES_MAX) 1898 return (EINVAL); 1899 rnh = ch->tables[tbl]; 1900 KASSERT(rnh != NULL, ("NULL IPFW table")); 1901 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1902 return (0); 1903} 1904 1905static void 1906flush_tables(struct ip_fw_chain *ch) 1907{ 1908 uint16_t tbl; 1909 1910 IPFW_WLOCK_ASSERT(ch); 1911 1912 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1913 flush_table(ch, tbl); 1914} 1915 1916static int 1917init_tables(struct ip_fw_chain *ch) 1918{ 1919 int i; 1920 uint16_t j; 1921 1922 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1923 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1924 for (j = 0; j < i; j++) { 1925 (void) flush_table(ch, j); 1926 } 1927 return (ENOMEM); 1928 } 1929 } 1930 return (0); 1931} 1932 1933static int 1934lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1935 uint32_t *val) 1936{ 1937 struct radix_node_head *rnh; 1938 struct table_entry *ent; 1939 struct sockaddr_in sa; 1940 1941 if (tbl >= IPFW_TABLES_MAX) 1942 return (0); 1943 rnh = ch->tables[tbl]; 1944 sa.sin_len = 8; 1945 sa.sin_addr.s_addr = addr; 1946 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1947 if (ent != NULL) { 1948 *val = ent->value; 1949 return (1); 1950 } 1951 return (0); 1952} 1953 1954static int 1955count_table_entry(struct radix_node *rn, void *arg) 1956{ 1957 u_int32_t * const cnt = arg; 1958 1959 (*cnt)++; 1960 return (0); 1961} 1962 1963static int 1964count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1965{ 1966 struct radix_node_head *rnh; 1967 1968 if (tbl >= IPFW_TABLES_MAX) 1969 return (EINVAL); 1970 rnh = ch->tables[tbl]; 1971 *cnt = 0; 1972 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1973 return (0); 1974} 1975 1976static int 1977dump_table_entry(struct radix_node *rn, void *arg) 1978{ 1979 struct table_entry * const n = (struct table_entry *)rn; 1980 ipfw_table * const tbl = arg; 1981 ipfw_table_entry *ent; 1982 1983 if (tbl->cnt == tbl->size) 1984 return (1); 1985 ent = &tbl->ent[tbl->cnt]; 1986 ent->tbl = tbl->tbl; 1987 if (in_nullhost(n->mask.sin_addr)) 1988 ent->masklen = 0; 1989 else 1990 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1991 ent->addr = n->addr.sin_addr.s_addr; 1992 ent->value = n->value; 1993 tbl->cnt++; 1994 return (0); 1995} 1996 1997static int 1998dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1999{ 2000 struct radix_node_head *rnh; 2001 2002 if (tbl->tbl >= IPFW_TABLES_MAX) 2003 return (EINVAL); 2004 rnh = ch->tables[tbl->tbl]; 2005 tbl->cnt = 0; 2006 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 2007 return (0); 2008} 2009 2010static void 2011fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 2012{ 2013 struct ucred *cr; 2014 2015 cr = inp->inp_cred; 2016 ugp->fw_prid = jailed(cr) ? cr->cr_prison->pr_id : -1; 2017 ugp->fw_uid = cr->cr_uid; 2018 ugp->fw_ngroups = cr->cr_ngroups; 2019 bcopy(cr->cr_groups, ugp->fw_groups, sizeof(ugp->fw_groups)); 2020} 2021 2022static int 2023check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 2024 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 2025 u_int16_t src_port, struct ip_fw_ugid *ugp, int *ugid_lookupp, 2026 struct inpcb *inp) 2027{ 2028 INIT_VNET_INET(curvnet); 2029 struct inpcbinfo *pi; 2030 int wildcard; 2031 struct inpcb *pcb; 2032 int match; 2033 gid_t *gp; 2034 2035 /* 2036 * Check to see if the UDP or TCP stack supplied us with 2037 * the PCB. If so, rather then holding a lock and looking 2038 * up the PCB, we can use the one that was supplied. 2039 */ 2040 if (inp && *ugid_lookupp == 0) { 2041 INP_LOCK_ASSERT(inp); 2042 if (inp->inp_socket != NULL) { 2043 fill_ugid_cache(inp, ugp); 2044 *ugid_lookupp = 1; 2045 } else 2046 *ugid_lookupp = -1; 2047 } 2048 /* 2049 * If we have already been here and the packet has no 2050 * PCB entry associated with it, then we can safely 2051 * assume that this is a no match. 2052 */ 2053 if (*ugid_lookupp == -1) 2054 return (0); 2055 if (proto == IPPROTO_TCP) { 2056 wildcard = 0; 2057 pi = &V_tcbinfo; 2058 } else if (proto == IPPROTO_UDP) { 2059 wildcard = INPLOOKUP_WILDCARD; 2060 pi = &V_udbinfo; 2061 } else 2062 return 0; 2063 match = 0; 2064 if (*ugid_lookupp == 0) { 2065 INP_INFO_RLOCK(pi); 2066 pcb = (oif) ? 2067 in_pcblookup_hash(pi, 2068 dst_ip, htons(dst_port), 2069 src_ip, htons(src_port), 2070 wildcard, oif) : 2071 in_pcblookup_hash(pi, 2072 src_ip, htons(src_port), 2073 dst_ip, htons(dst_port), 2074 wildcard, NULL); 2075 if (pcb != NULL) { 2076 fill_ugid_cache(pcb, ugp); 2077 *ugid_lookupp = 1; 2078 } 2079 INP_INFO_RUNLOCK(pi); 2080 if (*ugid_lookupp == 0) { 2081 /* 2082 * If the lookup did not yield any results, there 2083 * is no sense in coming back and trying again. So 2084 * we can set lookup to -1 and ensure that we wont 2085 * bother the pcb system again. 2086 */ 2087 *ugid_lookupp = -1; 2088 return (0); 2089 } 2090 } 2091 if (insn->o.opcode == O_UID) 2092 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2093 else if (insn->o.opcode == O_GID) { 2094 for (gp = ugp->fw_groups; 2095 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2096 if (*gp == (gid_t)insn->d[0]) { 2097 match = 1; 2098 break; 2099 } 2100 } else if (insn->o.opcode == O_JAIL) 2101 match = (ugp->fw_prid == (int)insn->d[0]); 2102 return match; 2103} 2104 2105/* 2106 * The main check routine for the firewall. 2107 * 2108 * All arguments are in args so we can modify them and return them 2109 * back to the caller. 2110 * 2111 * Parameters: 2112 * 2113 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2114 * Starts with the IP header. 2115 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2116 * args->L3offset Number of bytes bypassed if we came from L2. 2117 * e.g. often sizeof(eh) ** NOTYET ** 2118 * args->oif Outgoing interface, or NULL if packet is incoming. 2119 * The incoming interface is in the mbuf. (in) 2120 * args->divert_rule (in/out) 2121 * Skip up to the first rule past this rule number; 2122 * upon return, non-zero port number for divert or tee. 2123 * 2124 * args->rule Pointer to the last matching rule (in/out) 2125 * args->next_hop Socket we are forwarding to (out). 2126 * args->f_id Addresses grabbed from the packet (out) 2127 * args->cookie a cookie depending on rule action 2128 * 2129 * Return value: 2130 * 2131 * IP_FW_PASS the packet must be accepted 2132 * IP_FW_DENY the packet must be dropped 2133 * IP_FW_DIVERT divert packet, port in m_tag 2134 * IP_FW_TEE tee packet, port in m_tag 2135 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2136 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2137 * 2138 */ 2139int 2140ipfw_chk(struct ip_fw_args *args) 2141{ 2142 INIT_VNET_INET(curvnet); 2143 INIT_VNET_IPFW(curvnet); 2144 2145 /* 2146 * Local variables holding state during the processing of a packet: 2147 * 2148 * IMPORTANT NOTE: to speed up the processing of rules, there 2149 * are some assumption on the values of the variables, which 2150 * are documented here. Should you change them, please check 2151 * the implementation of the various instructions to make sure 2152 * that they still work. 2153 * 2154 * args->eh The MAC header. It is non-null for a layer2 2155 * packet, it is NULL for a layer-3 packet. 2156 * **notyet** 2157 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2158 * 2159 * m | args->m Pointer to the mbuf, as received from the caller. 2160 * It may change if ipfw_chk() does an m_pullup, or if it 2161 * consumes the packet because it calls send_reject(). 2162 * XXX This has to change, so that ipfw_chk() never modifies 2163 * or consumes the buffer. 2164 * ip is the beginning of the ip(4 or 6) header. 2165 * Calculated by adding the L3offset to the start of data. 2166 * (Until we start using L3offset, the packet is 2167 * supposed to start with the ip header). 2168 */ 2169 struct mbuf *m = args->m; 2170 struct ip *ip = mtod(m, struct ip *); 2171 2172 /* 2173 * For rules which contain uid/gid or jail constraints, cache 2174 * a copy of the users credentials after the pcb lookup has been 2175 * executed. This will speed up the processing of rules with 2176 * these types of constraints, as well as decrease contention 2177 * on pcb related locks. 2178 */ 2179 struct ip_fw_ugid fw_ugid_cache; 2180 int ugid_lookup = 0; 2181 2182 /* 2183 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2184 * associated with a packet input on a divert socket. This 2185 * will allow to distinguish traffic and its direction when 2186 * it originates from a divert socket. 2187 */ 2188 u_int divinput_flags = 0; 2189 2190 /* 2191 * oif | args->oif If NULL, ipfw_chk has been called on the 2192 * inbound path (ether_input, ip_input). 2193 * If non-NULL, ipfw_chk has been called on the outbound path 2194 * (ether_output, ip_output). 2195 */ 2196 struct ifnet *oif = args->oif; 2197 2198 struct ip_fw *f = NULL; /* matching rule */ 2199 int retval = 0; 2200 2201 /* 2202 * hlen The length of the IP header. 2203 */ 2204 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2205 2206 /* 2207 * offset The offset of a fragment. offset != 0 means that 2208 * we have a fragment at this offset of an IPv4 packet. 2209 * offset == 0 means that (if this is an IPv4 packet) 2210 * this is the first or only fragment. 2211 * For IPv6 offset == 0 means there is no Fragment Header. 2212 * If offset != 0 for IPv6 always use correct mask to 2213 * get the correct offset because we add IP6F_MORE_FRAG 2214 * to be able to dectect the first fragment which would 2215 * otherwise have offset = 0. 2216 */ 2217 u_short offset = 0; 2218 2219 /* 2220 * Local copies of addresses. They are only valid if we have 2221 * an IP packet. 2222 * 2223 * proto The protocol. Set to 0 for non-ip packets, 2224 * or to the protocol read from the packet otherwise. 2225 * proto != 0 means that we have an IPv4 packet. 2226 * 2227 * src_port, dst_port port numbers, in HOST format. Only 2228 * valid for TCP and UDP packets. 2229 * 2230 * src_ip, dst_ip ip addresses, in NETWORK format. 2231 * Only valid for IPv4 packets. 2232 */ 2233 u_int8_t proto; 2234 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2235 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2236 u_int16_t ip_len=0; 2237 int pktlen; 2238 u_int16_t etype = 0; /* Host order stored ether type */ 2239 2240 /* 2241 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2242 * MATCH_NONE when checked and not matched (q = NULL), 2243 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2244 */ 2245 int dyn_dir = MATCH_UNKNOWN; 2246 ipfw_dyn_rule *q = NULL; 2247 struct ip_fw_chain *chain = &V_layer3_chain; 2248 struct m_tag *mtag; 2249 2250 /* 2251 * We store in ulp a pointer to the upper layer protocol header. 2252 * In the ipv4 case this is easy to determine from the header, 2253 * but for ipv6 we might have some additional headers in the middle. 2254 * ulp is NULL if not found. 2255 */ 2256 void *ulp = NULL; /* upper layer protocol pointer. */ 2257 /* XXX ipv6 variables */ 2258 int is_ipv6 = 0; 2259 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2260 /* end of ipv6 variables */ 2261 int is_ipv4 = 0; 2262 2263 if (m->m_flags & M_SKIP_FIREWALL) 2264 return (IP_FW_PASS); /* accept */ 2265 2266 dst_ip.s_addr = 0; /* make sure it is initialized */ 2267 pktlen = m->m_pkthdr.len; 2268 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 2269 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2270 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2271 2272/* 2273 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2274 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2275 * pointer might become stale after other pullups (but we never use it 2276 * this way). 2277 */ 2278#define PULLUP_TO(len, p, T) \ 2279do { \ 2280 int x = (len) + sizeof(T); \ 2281 if ((m)->m_len < x) { \ 2282 args->m = m = m_pullup(m, x); \ 2283 if (m == NULL) \ 2284 goto pullup_failed; \ 2285 } \ 2286 p = (mtod(m, char *) + (len)); \ 2287} while (0) 2288 2289 /* 2290 * if we have an ether header, 2291 */ 2292 if (args->eh) 2293 etype = ntohs(args->eh->ether_type); 2294 2295 /* Identify IP packets and fill up variables. */ 2296 if (pktlen >= sizeof(struct ip6_hdr) && 2297 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2298 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2299 is_ipv6 = 1; 2300 args->f_id.addr_type = 6; 2301 hlen = sizeof(struct ip6_hdr); 2302 proto = ip6->ip6_nxt; 2303 2304 /* Search extension headers to find upper layer protocols */ 2305 while (ulp == NULL) { 2306 switch (proto) { 2307 case IPPROTO_ICMPV6: 2308 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2309 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2310 break; 2311 2312 case IPPROTO_TCP: 2313 PULLUP_TO(hlen, ulp, struct tcphdr); 2314 dst_port = TCP(ulp)->th_dport; 2315 src_port = TCP(ulp)->th_sport; 2316 args->f_id.flags = TCP(ulp)->th_flags; 2317 break; 2318 2319 case IPPROTO_SCTP: 2320 PULLUP_TO(hlen, ulp, struct sctphdr); 2321 src_port = SCTP(ulp)->src_port; 2322 dst_port = SCTP(ulp)->dest_port; 2323 break; 2324 2325 case IPPROTO_UDP: 2326 PULLUP_TO(hlen, ulp, struct udphdr); 2327 dst_port = UDP(ulp)->uh_dport; 2328 src_port = UDP(ulp)->uh_sport; 2329 break; 2330 2331 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2332 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2333 ext_hd |= EXT_HOPOPTS; 2334 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2335 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2336 ulp = NULL; 2337 break; 2338 2339 case IPPROTO_ROUTING: /* RFC 2460 */ 2340 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2341 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2342 case 0: 2343 ext_hd |= EXT_RTHDR0; 2344 break; 2345 case 2: 2346 ext_hd |= EXT_RTHDR2; 2347 break; 2348 default: 2349 printf("IPFW2: IPV6 - Unknown Routing " 2350 "Header type(%d)\n", 2351 ((struct ip6_rthdr *)ulp)->ip6r_type); 2352 if (V_fw_deny_unknown_exthdrs) 2353 return (IP_FW_DENY); 2354 break; 2355 } 2356 ext_hd |= EXT_ROUTING; 2357 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2358 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2359 ulp = NULL; 2360 break; 2361 2362 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2363 PULLUP_TO(hlen, ulp, struct ip6_frag); 2364 ext_hd |= EXT_FRAGMENT; 2365 hlen += sizeof (struct ip6_frag); 2366 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2367 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2368 IP6F_OFF_MASK; 2369 /* Add IP6F_MORE_FRAG for offset of first 2370 * fragment to be != 0. */ 2371 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2372 IP6F_MORE_FRAG; 2373 if (offset == 0) { 2374 printf("IPFW2: IPV6 - Invalid Fragment " 2375 "Header\n"); 2376 if (V_fw_deny_unknown_exthdrs) 2377 return (IP_FW_DENY); 2378 break; 2379 } 2380 args->f_id.frag_id6 = 2381 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2382 ulp = NULL; 2383 break; 2384 2385 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2386 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2387 ext_hd |= EXT_DSTOPTS; 2388 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2389 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2390 ulp = NULL; 2391 break; 2392 2393 case IPPROTO_AH: /* RFC 2402 */ 2394 PULLUP_TO(hlen, ulp, struct ip6_ext); 2395 ext_hd |= EXT_AH; 2396 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2397 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2398 ulp = NULL; 2399 break; 2400 2401 case IPPROTO_ESP: /* RFC 2406 */ 2402 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2403 /* Anything past Seq# is variable length and 2404 * data past this ext. header is encrypted. */ 2405 ext_hd |= EXT_ESP; 2406 break; 2407 2408 case IPPROTO_NONE: /* RFC 2460 */ 2409 /* 2410 * Packet ends here, and IPv6 header has 2411 * already been pulled up. If ip6e_len!=0 2412 * then octets must be ignored. 2413 */ 2414 ulp = ip; /* non-NULL to get out of loop. */ 2415 break; 2416 2417 case IPPROTO_OSPFIGP: 2418 /* XXX OSPF header check? */ 2419 PULLUP_TO(hlen, ulp, struct ip6_ext); 2420 break; 2421 2422 case IPPROTO_PIM: 2423 /* XXX PIM header check? */ 2424 PULLUP_TO(hlen, ulp, struct pim); 2425 break; 2426 2427 case IPPROTO_CARP: 2428 PULLUP_TO(hlen, ulp, struct carp_header); 2429 if (((struct carp_header *)ulp)->carp_version != 2430 CARP_VERSION) 2431 return (IP_FW_DENY); 2432 if (((struct carp_header *)ulp)->carp_type != 2433 CARP_ADVERTISEMENT) 2434 return (IP_FW_DENY); 2435 break; 2436 2437 case IPPROTO_IPV6: /* RFC 2893 */ 2438 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2439 break; 2440 2441 case IPPROTO_IPV4: /* RFC 2893 */ 2442 PULLUP_TO(hlen, ulp, struct ip); 2443 break; 2444 2445 default: 2446 printf("IPFW2: IPV6 - Unknown Extension " 2447 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2448 if (V_fw_deny_unknown_exthdrs) 2449 return (IP_FW_DENY); 2450 PULLUP_TO(hlen, ulp, struct ip6_ext); 2451 break; 2452 } /*switch */ 2453 } 2454 ip = mtod(m, struct ip *); 2455 ip6 = (struct ip6_hdr *)ip; 2456 args->f_id.src_ip6 = ip6->ip6_src; 2457 args->f_id.dst_ip6 = ip6->ip6_dst; 2458 args->f_id.src_ip = 0; 2459 args->f_id.dst_ip = 0; 2460 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2461 } else if (pktlen >= sizeof(struct ip) && 2462 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2463 is_ipv4 = 1; 2464 hlen = ip->ip_hl << 2; 2465 args->f_id.addr_type = 4; 2466 2467 /* 2468 * Collect parameters into local variables for faster matching. 2469 */ 2470 proto = ip->ip_p; 2471 src_ip = ip->ip_src; 2472 dst_ip = ip->ip_dst; 2473 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2474 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2475 ip_len = ntohs(ip->ip_len); 2476 } else { 2477 offset = ip->ip_off & IP_OFFMASK; 2478 ip_len = ip->ip_len; 2479 } 2480 pktlen = ip_len < pktlen ? ip_len : pktlen; 2481 2482 if (offset == 0) { 2483 switch (proto) { 2484 case IPPROTO_TCP: 2485 PULLUP_TO(hlen, ulp, struct tcphdr); 2486 dst_port = TCP(ulp)->th_dport; 2487 src_port = TCP(ulp)->th_sport; 2488 args->f_id.flags = TCP(ulp)->th_flags; 2489 break; 2490 2491 case IPPROTO_UDP: 2492 PULLUP_TO(hlen, ulp, struct udphdr); 2493 dst_port = UDP(ulp)->uh_dport; 2494 src_port = UDP(ulp)->uh_sport; 2495 break; 2496 2497 case IPPROTO_ICMP: 2498 PULLUP_TO(hlen, ulp, struct icmphdr); 2499 args->f_id.flags = ICMP(ulp)->icmp_type; 2500 break; 2501 2502 default: 2503 break; 2504 } 2505 } 2506 2507 ip = mtod(m, struct ip *); 2508 args->f_id.src_ip = ntohl(src_ip.s_addr); 2509 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2510 } 2511#undef PULLUP_TO 2512 if (proto) { /* we may have port numbers, store them */ 2513 args->f_id.proto = proto; 2514 args->f_id.src_port = src_port = ntohs(src_port); 2515 args->f_id.dst_port = dst_port = ntohs(dst_port); 2516 } 2517 2518 IPFW_RLOCK(chain); 2519 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2520 if (args->rule) { 2521 /* 2522 * Packet has already been tagged. Look for the next rule 2523 * to restart processing. 2524 */ 2525 f = args->rule->next_rule; 2526 if (f == NULL) 2527 f = lookup_next_rule(args->rule, 0); 2528 } else { 2529 /* 2530 * Find the starting rule. It can be either the first 2531 * one, or the one after divert_rule if asked so. 2532 */ 2533 int skipto = mtag ? divert_cookie(mtag) : 0; 2534 2535 f = chain->rules; 2536 if (args->eh == NULL && skipto != 0) { 2537 if (skipto >= IPFW_DEFAULT_RULE) { 2538 IPFW_RUNLOCK(chain); 2539 return (IP_FW_DENY); /* invalid */ 2540 } 2541 while (f && f->rulenum <= skipto) 2542 f = f->next; 2543 if (f == NULL) { /* drop packet */ 2544 IPFW_RUNLOCK(chain); 2545 return (IP_FW_DENY); 2546 } 2547 } 2548 } 2549 /* reset divert rule to avoid confusion later */ 2550 if (mtag) { 2551 divinput_flags = divert_info(mtag) & 2552 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2553 m_tag_delete(m, mtag); 2554 } 2555 2556 /* 2557 * Now scan the rules, and parse microinstructions for each rule. 2558 */ 2559 for (; f; f = f->next) { 2560 ipfw_insn *cmd; 2561 uint32_t tablearg = 0; 2562 int l, cmdlen, skip_or; /* skip rest of OR block */ 2563 2564again: 2565 if (V_set_disable & (1 << f->set) ) 2566 continue; 2567 2568 skip_or = 0; 2569 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2570 l -= cmdlen, cmd += cmdlen) { 2571 int match; 2572 2573 /* 2574 * check_body is a jump target used when we find a 2575 * CHECK_STATE, and need to jump to the body of 2576 * the target rule. 2577 */ 2578 2579check_body: 2580 cmdlen = F_LEN(cmd); 2581 /* 2582 * An OR block (insn_1 || .. || insn_n) has the 2583 * F_OR bit set in all but the last instruction. 2584 * The first match will set "skip_or", and cause 2585 * the following instructions to be skipped until 2586 * past the one with the F_OR bit clear. 2587 */ 2588 if (skip_or) { /* skip this instruction */ 2589 if ((cmd->len & F_OR) == 0) 2590 skip_or = 0; /* next one is good */ 2591 continue; 2592 } 2593 match = 0; /* set to 1 if we succeed */ 2594 2595 switch (cmd->opcode) { 2596 /* 2597 * The first set of opcodes compares the packet's 2598 * fields with some pattern, setting 'match' if a 2599 * match is found. At the end of the loop there is 2600 * logic to deal with F_NOT and F_OR flags associated 2601 * with the opcode. 2602 */ 2603 case O_NOP: 2604 match = 1; 2605 break; 2606 2607 case O_FORWARD_MAC: 2608 printf("ipfw: opcode %d unimplemented\n", 2609 cmd->opcode); 2610 break; 2611 2612 case O_GID: 2613 case O_UID: 2614 case O_JAIL: 2615 /* 2616 * We only check offset == 0 && proto != 0, 2617 * as this ensures that we have a 2618 * packet with the ports info. 2619 */ 2620 if (offset!=0) 2621 break; 2622 if (is_ipv6) /* XXX to be fixed later */ 2623 break; 2624 if (proto == IPPROTO_TCP || 2625 proto == IPPROTO_UDP) 2626 match = check_uidgid( 2627 (ipfw_insn_u32 *)cmd, 2628 proto, oif, 2629 dst_ip, dst_port, 2630 src_ip, src_port, &fw_ugid_cache, 2631 &ugid_lookup, args->inp); 2632 break; 2633 2634 case O_RECV: 2635 match = iface_match(m->m_pkthdr.rcvif, 2636 (ipfw_insn_if *)cmd); 2637 break; 2638 2639 case O_XMIT: 2640 match = iface_match(oif, (ipfw_insn_if *)cmd); 2641 break; 2642 2643 case O_VIA: 2644 match = iface_match(oif ? oif : 2645 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2646 break; 2647 2648 case O_MACADDR2: 2649 if (args->eh != NULL) { /* have MAC header */ 2650 u_int32_t *want = (u_int32_t *) 2651 ((ipfw_insn_mac *)cmd)->addr; 2652 u_int32_t *mask = (u_int32_t *) 2653 ((ipfw_insn_mac *)cmd)->mask; 2654 u_int32_t *hdr = (u_int32_t *)args->eh; 2655 2656 match = 2657 ( want[0] == (hdr[0] & mask[0]) && 2658 want[1] == (hdr[1] & mask[1]) && 2659 want[2] == (hdr[2] & mask[2]) ); 2660 } 2661 break; 2662 2663 case O_MAC_TYPE: 2664 if (args->eh != NULL) { 2665 u_int16_t *p = 2666 ((ipfw_insn_u16 *)cmd)->ports; 2667 int i; 2668 2669 for (i = cmdlen - 1; !match && i>0; 2670 i--, p += 2) 2671 match = (etype >= p[0] && 2672 etype <= p[1]); 2673 } 2674 break; 2675 2676 case O_FRAG: 2677 match = (offset != 0); 2678 break; 2679 2680 case O_IN: /* "out" is "not in" */ 2681 match = (oif == NULL); 2682 break; 2683 2684 case O_LAYER2: 2685 match = (args->eh != NULL); 2686 break; 2687 2688 case O_DIVERTED: 2689 match = (cmd->arg1 & 1 && divinput_flags & 2690 IP_FW_DIVERT_LOOPBACK_FLAG) || 2691 (cmd->arg1 & 2 && divinput_flags & 2692 IP_FW_DIVERT_OUTPUT_FLAG); 2693 break; 2694 2695 case O_PROTO: 2696 /* 2697 * We do not allow an arg of 0 so the 2698 * check of "proto" only suffices. 2699 */ 2700 match = (proto == cmd->arg1); 2701 break; 2702 2703 case O_IP_SRC: 2704 match = is_ipv4 && 2705 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2706 src_ip.s_addr); 2707 break; 2708 2709 case O_IP_SRC_LOOKUP: 2710 case O_IP_DST_LOOKUP: 2711 if (is_ipv4) { 2712 uint32_t a = 2713 (cmd->opcode == O_IP_DST_LOOKUP) ? 2714 dst_ip.s_addr : src_ip.s_addr; 2715 uint32_t v = 0; 2716 2717 match = lookup_table(chain, cmd->arg1, a, 2718 &v); 2719 if (!match) 2720 break; 2721 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2722 match = 2723 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2724 else 2725 tablearg = v; 2726 } 2727 break; 2728 2729 case O_IP_SRC_MASK: 2730 case O_IP_DST_MASK: 2731 if (is_ipv4) { 2732 uint32_t a = 2733 (cmd->opcode == O_IP_DST_MASK) ? 2734 dst_ip.s_addr : src_ip.s_addr; 2735 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2736 int i = cmdlen-1; 2737 2738 for (; !match && i>0; i-= 2, p+= 2) 2739 match = (p[0] == (a & p[1])); 2740 } 2741 break; 2742 2743 case O_IP_SRC_ME: 2744 if (is_ipv4) { 2745 struct ifnet *tif; 2746 2747 INADDR_TO_IFP(src_ip, tif); 2748 match = (tif != NULL); 2749 } 2750 break; 2751 2752 case O_IP_DST_SET: 2753 case O_IP_SRC_SET: 2754 if (is_ipv4) { 2755 u_int32_t *d = (u_int32_t *)(cmd+1); 2756 u_int32_t addr = 2757 cmd->opcode == O_IP_DST_SET ? 2758 args->f_id.dst_ip : 2759 args->f_id.src_ip; 2760 2761 if (addr < d[0]) 2762 break; 2763 addr -= d[0]; /* subtract base */ 2764 match = (addr < cmd->arg1) && 2765 ( d[ 1 + (addr>>5)] & 2766 (1<<(addr & 0x1f)) ); 2767 } 2768 break; 2769 2770 case O_IP_DST: 2771 match = is_ipv4 && 2772 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2773 dst_ip.s_addr); 2774 break; 2775 2776 case O_IP_DST_ME: 2777 if (is_ipv4) { 2778 struct ifnet *tif; 2779 2780 INADDR_TO_IFP(dst_ip, tif); 2781 match = (tif != NULL); 2782 } 2783 break; 2784 2785 case O_IP_SRCPORT: 2786 case O_IP_DSTPORT: 2787 /* 2788 * offset == 0 && proto != 0 is enough 2789 * to guarantee that we have a 2790 * packet with port info. 2791 */ 2792 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2793 && offset == 0) { 2794 u_int16_t x = 2795 (cmd->opcode == O_IP_SRCPORT) ? 2796 src_port : dst_port ; 2797 u_int16_t *p = 2798 ((ipfw_insn_u16 *)cmd)->ports; 2799 int i; 2800 2801 for (i = cmdlen - 1; !match && i>0; 2802 i--, p += 2) 2803 match = (x>=p[0] && x<=p[1]); 2804 } 2805 break; 2806 2807 case O_ICMPTYPE: 2808 match = (offset == 0 && proto==IPPROTO_ICMP && 2809 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2810 break; 2811 2812#ifdef INET6 2813 case O_ICMP6TYPE: 2814 match = is_ipv6 && offset == 0 && 2815 proto==IPPROTO_ICMPV6 && 2816 icmp6type_match( 2817 ICMP6(ulp)->icmp6_type, 2818 (ipfw_insn_u32 *)cmd); 2819 break; 2820#endif /* INET6 */ 2821 2822 case O_IPOPT: 2823 match = (is_ipv4 && 2824 ipopts_match(ip, cmd) ); 2825 break; 2826 2827 case O_IPVER: 2828 match = (is_ipv4 && 2829 cmd->arg1 == ip->ip_v); 2830 break; 2831 2832 case O_IPID: 2833 case O_IPLEN: 2834 case O_IPTTL: 2835 if (is_ipv4) { /* only for IP packets */ 2836 uint16_t x; 2837 uint16_t *p; 2838 int i; 2839 2840 if (cmd->opcode == O_IPLEN) 2841 x = ip_len; 2842 else if (cmd->opcode == O_IPTTL) 2843 x = ip->ip_ttl; 2844 else /* must be IPID */ 2845 x = ntohs(ip->ip_id); 2846 if (cmdlen == 1) { 2847 match = (cmd->arg1 == x); 2848 break; 2849 } 2850 /* otherwise we have ranges */ 2851 p = ((ipfw_insn_u16 *)cmd)->ports; 2852 i = cmdlen - 1; 2853 for (; !match && i>0; i--, p += 2) 2854 match = (x >= p[0] && x <= p[1]); 2855 } 2856 break; 2857 2858 case O_IPPRECEDENCE: 2859 match = (is_ipv4 && 2860 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2861 break; 2862 2863 case O_IPTOS: 2864 match = (is_ipv4 && 2865 flags_match(cmd, ip->ip_tos)); 2866 break; 2867 2868 case O_TCPDATALEN: 2869 if (proto == IPPROTO_TCP && offset == 0) { 2870 struct tcphdr *tcp; 2871 uint16_t x; 2872 uint16_t *p; 2873 int i; 2874 2875 tcp = TCP(ulp); 2876 x = ip_len - 2877 ((ip->ip_hl + tcp->th_off) << 2); 2878 if (cmdlen == 1) { 2879 match = (cmd->arg1 == x); 2880 break; 2881 } 2882 /* otherwise we have ranges */ 2883 p = ((ipfw_insn_u16 *)cmd)->ports; 2884 i = cmdlen - 1; 2885 for (; !match && i>0; i--, p += 2) 2886 match = (x >= p[0] && x <= p[1]); 2887 } 2888 break; 2889 2890 case O_TCPFLAGS: 2891 match = (proto == IPPROTO_TCP && offset == 0 && 2892 flags_match(cmd, TCP(ulp)->th_flags)); 2893 break; 2894 2895 case O_TCPOPTS: 2896 match = (proto == IPPROTO_TCP && offset == 0 && 2897 tcpopts_match(TCP(ulp), cmd)); 2898 break; 2899 2900 case O_TCPSEQ: 2901 match = (proto == IPPROTO_TCP && offset == 0 && 2902 ((ipfw_insn_u32 *)cmd)->d[0] == 2903 TCP(ulp)->th_seq); 2904 break; 2905 2906 case O_TCPACK: 2907 match = (proto == IPPROTO_TCP && offset == 0 && 2908 ((ipfw_insn_u32 *)cmd)->d[0] == 2909 TCP(ulp)->th_ack); 2910 break; 2911 2912 case O_TCPWIN: 2913 match = (proto == IPPROTO_TCP && offset == 0 && 2914 cmd->arg1 == TCP(ulp)->th_win); 2915 break; 2916 2917 case O_ESTAB: 2918 /* reject packets which have SYN only */ 2919 /* XXX should i also check for TH_ACK ? */ 2920 match = (proto == IPPROTO_TCP && offset == 0 && 2921 (TCP(ulp)->th_flags & 2922 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2923 break; 2924 2925 case O_ALTQ: { 2926 struct pf_mtag *at; 2927 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2928 2929 match = 1; 2930 at = pf_find_mtag(m); 2931 if (at != NULL && at->qid != 0) 2932 break; 2933 at = pf_get_mtag(m); 2934 if (at == NULL) { 2935 /* 2936 * Let the packet fall back to the 2937 * default ALTQ. 2938 */ 2939 break; 2940 } 2941 at->qid = altq->qid; 2942 if (is_ipv4) 2943 at->af = AF_INET; 2944 else 2945 at->af = AF_LINK; 2946 at->hdr = ip; 2947 break; 2948 } 2949 2950 case O_LOG: 2951 if (V_fw_verbose) 2952 ipfw_log(f, hlen, args, m, 2953 oif, offset, tablearg, ip); 2954 match = 1; 2955 break; 2956 2957 case O_PROB: 2958 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2959 break; 2960 2961 case O_VERREVPATH: 2962 /* Outgoing packets automatically pass/match */ 2963 match = ((oif != NULL) || 2964 (m->m_pkthdr.rcvif == NULL) || 2965 ( 2966#ifdef INET6 2967 is_ipv6 ? 2968 verify_path6(&(args->f_id.src_ip6), 2969 m->m_pkthdr.rcvif) : 2970#endif 2971 verify_path(src_ip, m->m_pkthdr.rcvif, 2972 args->f_id.fib))); 2973 break; 2974 2975 case O_VERSRCREACH: 2976 /* Outgoing packets automatically pass/match */ 2977 match = (hlen > 0 && ((oif != NULL) || 2978#ifdef INET6 2979 is_ipv6 ? 2980 verify_path6(&(args->f_id.src_ip6), 2981 NULL) : 2982#endif 2983 verify_path(src_ip, NULL, args->f_id.fib))); 2984 break; 2985 2986 case O_ANTISPOOF: 2987 /* Outgoing packets automatically pass/match */ 2988 if (oif == NULL && hlen > 0 && 2989 ( (is_ipv4 && in_localaddr(src_ip)) 2990#ifdef INET6 2991 || (is_ipv6 && 2992 in6_localaddr(&(args->f_id.src_ip6))) 2993#endif 2994 )) 2995 match = 2996#ifdef INET6 2997 is_ipv6 ? verify_path6( 2998 &(args->f_id.src_ip6), 2999 m->m_pkthdr.rcvif) : 3000#endif 3001 verify_path(src_ip, 3002 m->m_pkthdr.rcvif, 3003 args->f_id.fib); 3004 else 3005 match = 1; 3006 break; 3007 3008 case O_IPSEC: 3009#ifdef IPSEC 3010 match = (m_tag_find(m, 3011 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 3012#endif 3013 /* otherwise no match */ 3014 break; 3015 3016#ifdef INET6 3017 case O_IP6_SRC: 3018 match = is_ipv6 && 3019 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 3020 &((ipfw_insn_ip6 *)cmd)->addr6); 3021 break; 3022 3023 case O_IP6_DST: 3024 match = is_ipv6 && 3025 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 3026 &((ipfw_insn_ip6 *)cmd)->addr6); 3027 break; 3028 case O_IP6_SRC_MASK: 3029 case O_IP6_DST_MASK: 3030 if (is_ipv6) { 3031 int i = cmdlen - 1; 3032 struct in6_addr p; 3033 struct in6_addr *d = 3034 &((ipfw_insn_ip6 *)cmd)->addr6; 3035 3036 for (; !match && i > 0; d += 2, 3037 i -= F_INSN_SIZE(struct in6_addr) 3038 * 2) { 3039 p = (cmd->opcode == 3040 O_IP6_SRC_MASK) ? 3041 args->f_id.src_ip6: 3042 args->f_id.dst_ip6; 3043 APPLY_MASK(&p, &d[1]); 3044 match = 3045 IN6_ARE_ADDR_EQUAL(&d[0], 3046 &p); 3047 } 3048 } 3049 break; 3050 3051 case O_IP6_SRC_ME: 3052 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3053 break; 3054 3055 case O_IP6_DST_ME: 3056 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3057 break; 3058 3059 case O_FLOW6ID: 3060 match = is_ipv6 && 3061 flow6id_match(args->f_id.flow_id6, 3062 (ipfw_insn_u32 *) cmd); 3063 break; 3064 3065 case O_EXT_HDR: 3066 match = is_ipv6 && 3067 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3068 break; 3069 3070 case O_IP6: 3071 match = is_ipv6; 3072 break; 3073#endif 3074 3075 case O_IP4: 3076 match = is_ipv4; 3077 break; 3078 3079 case O_TAG: { 3080 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3081 tablearg : cmd->arg1; 3082 3083 /* Packet is already tagged with this tag? */ 3084 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3085 3086 /* We have `untag' action when F_NOT flag is 3087 * present. And we must remove this mtag from 3088 * mbuf and reset `match' to zero (`match' will 3089 * be inversed later). 3090 * Otherwise we should allocate new mtag and 3091 * push it into mbuf. 3092 */ 3093 if (cmd->len & F_NOT) { /* `untag' action */ 3094 if (mtag != NULL) 3095 m_tag_delete(m, mtag); 3096 } else if (mtag == NULL) { 3097 if ((mtag = m_tag_alloc(MTAG_IPFW, 3098 tag, 0, M_NOWAIT)) != NULL) 3099 m_tag_prepend(m, mtag); 3100 } 3101 match = (cmd->len & F_NOT) ? 0: 1; 3102 break; 3103 } 3104 3105 case O_FIB: /* try match the specified fib */ 3106 if (args->f_id.fib == cmd->arg1) 3107 match = 1; 3108 break; 3109 3110 case O_TAGGED: { 3111 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3112 tablearg : cmd->arg1; 3113 3114 if (cmdlen == 1) { 3115 match = m_tag_locate(m, MTAG_IPFW, 3116 tag, NULL) != NULL; 3117 break; 3118 } 3119 3120 /* we have ranges */ 3121 for (mtag = m_tag_first(m); 3122 mtag != NULL && !match; 3123 mtag = m_tag_next(m, mtag)) { 3124 uint16_t *p; 3125 int i; 3126 3127 if (mtag->m_tag_cookie != MTAG_IPFW) 3128 continue; 3129 3130 p = ((ipfw_insn_u16 *)cmd)->ports; 3131 i = cmdlen - 1; 3132 for(; !match && i > 0; i--, p += 2) 3133 match = 3134 mtag->m_tag_id >= p[0] && 3135 mtag->m_tag_id <= p[1]; 3136 } 3137 break; 3138 } 3139 3140 /* 3141 * The second set of opcodes represents 'actions', 3142 * i.e. the terminal part of a rule once the packet 3143 * matches all previous patterns. 3144 * Typically there is only one action for each rule, 3145 * and the opcode is stored at the end of the rule 3146 * (but there are exceptions -- see below). 3147 * 3148 * In general, here we set retval and terminate the 3149 * outer loop (would be a 'break 3' in some language, 3150 * but we need to do a 'goto done'). 3151 * 3152 * Exceptions: 3153 * O_COUNT and O_SKIPTO actions: 3154 * instead of terminating, we jump to the next rule 3155 * ('goto next_rule', equivalent to a 'break 2'), 3156 * or to the SKIPTO target ('goto again' after 3157 * having set f, cmd and l), respectively. 3158 * 3159 * O_TAG, O_LOG and O_ALTQ action parameters: 3160 * perform some action and set match = 1; 3161 * 3162 * O_LIMIT and O_KEEP_STATE: these opcodes are 3163 * not real 'actions', and are stored right 3164 * before the 'action' part of the rule. 3165 * These opcodes try to install an entry in the 3166 * state tables; if successful, we continue with 3167 * the next opcode (match=1; break;), otherwise 3168 * the packet * must be dropped 3169 * ('goto done' after setting retval); 3170 * 3171 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3172 * cause a lookup of the state table, and a jump 3173 * to the 'action' part of the parent rule 3174 * ('goto check_body') if an entry is found, or 3175 * (CHECK_STATE only) a jump to the next rule if 3176 * the entry is not found ('goto next_rule'). 3177 * The result of the lookup is cached to make 3178 * further instances of these opcodes are 3179 * effectively NOPs. 3180 */ 3181 case O_LIMIT: 3182 case O_KEEP_STATE: 3183 if (install_state(f, 3184 (ipfw_insn_limit *)cmd, args, tablearg)) { 3185 retval = IP_FW_DENY; 3186 goto done; /* error/limit violation */ 3187 } 3188 match = 1; 3189 break; 3190 3191 case O_PROBE_STATE: 3192 case O_CHECK_STATE: 3193 /* 3194 * dynamic rules are checked at the first 3195 * keep-state or check-state occurrence, 3196 * with the result being stored in dyn_dir. 3197 * The compiler introduces a PROBE_STATE 3198 * instruction for us when we have a 3199 * KEEP_STATE (because PROBE_STATE needs 3200 * to be run first). 3201 */ 3202 if (dyn_dir == MATCH_UNKNOWN && 3203 (q = lookup_dyn_rule(&args->f_id, 3204 &dyn_dir, proto == IPPROTO_TCP ? 3205 TCP(ulp) : NULL)) 3206 != NULL) { 3207 /* 3208 * Found dynamic entry, update stats 3209 * and jump to the 'action' part of 3210 * the parent rule. 3211 */ 3212 q->pcnt++; 3213 q->bcnt += pktlen; 3214 f = q->rule; 3215 cmd = ACTION_PTR(f); 3216 l = f->cmd_len - f->act_ofs; 3217 IPFW_DYN_UNLOCK(); 3218 goto check_body; 3219 } 3220 /* 3221 * Dynamic entry not found. If CHECK_STATE, 3222 * skip to next rule, if PROBE_STATE just 3223 * ignore and continue with next opcode. 3224 */ 3225 if (cmd->opcode == O_CHECK_STATE) 3226 goto next_rule; 3227 match = 1; 3228 break; 3229 3230 case O_ACCEPT: 3231 retval = 0; /* accept */ 3232 goto done; 3233 3234 case O_PIPE: 3235 case O_QUEUE: 3236 args->rule = f; /* report matching rule */ 3237 if (cmd->arg1 == IP_FW_TABLEARG) 3238 args->cookie = tablearg; 3239 else 3240 args->cookie = cmd->arg1; 3241 retval = IP_FW_DUMMYNET; 3242 goto done; 3243 3244 case O_DIVERT: 3245 case O_TEE: { 3246 struct divert_tag *dt; 3247 3248 if (args->eh) /* not on layer 2 */ 3249 break; 3250 mtag = m_tag_get(PACKET_TAG_DIVERT, 3251 sizeof(struct divert_tag), 3252 M_NOWAIT); 3253 if (mtag == NULL) { 3254 /* XXX statistic */ 3255 /* drop packet */ 3256 IPFW_RUNLOCK(chain); 3257 return (IP_FW_DENY); 3258 } 3259 dt = (struct divert_tag *)(mtag+1); 3260 dt->cookie = f->rulenum; 3261 if (cmd->arg1 == IP_FW_TABLEARG) 3262 dt->info = tablearg; 3263 else 3264 dt->info = cmd->arg1; 3265 m_tag_prepend(m, mtag); 3266 retval = (cmd->opcode == O_DIVERT) ? 3267 IP_FW_DIVERT : IP_FW_TEE; 3268 goto done; 3269 } 3270 case O_COUNT: 3271 case O_SKIPTO: 3272 f->pcnt++; /* update stats */ 3273 f->bcnt += pktlen; 3274 f->timestamp = time_uptime; 3275 if (cmd->opcode == O_COUNT) 3276 goto next_rule; 3277 /* handle skipto */ 3278 if (cmd->arg1 == IP_FW_TABLEARG) { 3279 f = lookup_next_rule(f, tablearg); 3280 } else { 3281 if (f->next_rule == NULL) 3282 lookup_next_rule(f, 0); 3283 f = f->next_rule; 3284 } 3285 goto again; 3286 3287 case O_REJECT: 3288 /* 3289 * Drop the packet and send a reject notice 3290 * if the packet is not ICMP (or is an ICMP 3291 * query), and it is not multicast/broadcast. 3292 */ 3293 if (hlen > 0 && is_ipv4 && offset == 0 && 3294 (proto != IPPROTO_ICMP || 3295 is_icmp_query(ICMP(ulp))) && 3296 !(m->m_flags & (M_BCAST|M_MCAST)) && 3297 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3298 send_reject(args, cmd->arg1, ip_len, ip); 3299 m = args->m; 3300 } 3301 /* FALLTHROUGH */ 3302#ifdef INET6 3303 case O_UNREACH6: 3304 if (hlen > 0 && is_ipv6 && 3305 ((offset & IP6F_OFF_MASK) == 0) && 3306 (proto != IPPROTO_ICMPV6 || 3307 (is_icmp6_query(args->f_id.flags) == 1)) && 3308 !(m->m_flags & (M_BCAST|M_MCAST)) && 3309 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3310 send_reject6( 3311 args, cmd->arg1, hlen, 3312 (struct ip6_hdr *)ip); 3313 m = args->m; 3314 } 3315 /* FALLTHROUGH */ 3316#endif 3317 case O_DENY: 3318 retval = IP_FW_DENY; 3319 goto done; 3320 3321 case O_FORWARD_IP: { 3322 struct sockaddr_in *sa; 3323 sa = &(((ipfw_insn_sa *)cmd)->sa); 3324 if (args->eh) /* not valid on layer2 pkts */ 3325 break; 3326 if (!q || dyn_dir == MATCH_FORWARD) { 3327 if (sa->sin_addr.s_addr == INADDR_ANY) { 3328 bcopy(sa, &args->hopstore, 3329 sizeof(*sa)); 3330 args->hopstore.sin_addr.s_addr = 3331 htonl(tablearg); 3332 args->next_hop = 3333 &args->hopstore; 3334 } else { 3335 args->next_hop = sa; 3336 } 3337 } 3338 retval = IP_FW_PASS; 3339 } 3340 goto done; 3341 3342 case O_NETGRAPH: 3343 case O_NGTEE: 3344 args->rule = f; /* report matching rule */ 3345 if (cmd->arg1 == IP_FW_TABLEARG) 3346 args->cookie = tablearg; 3347 else 3348 args->cookie = cmd->arg1; 3349 retval = (cmd->opcode == O_NETGRAPH) ? 3350 IP_FW_NETGRAPH : IP_FW_NGTEE; 3351 goto done; 3352 3353 case O_SETFIB: 3354 f->pcnt++; /* update stats */ 3355 f->bcnt += pktlen; 3356 f->timestamp = time_uptime; 3357 M_SETFIB(m, cmd->arg1); 3358 args->f_id.fib = cmd->arg1; 3359 goto next_rule; 3360 3361 case O_NAT: { 3362 struct cfg_nat *t; 3363 int nat_id; 3364 3365 if (IPFW_NAT_LOADED) { 3366 args->rule = f; /* Report matching rule. */ 3367 t = ((ipfw_insn_nat *)cmd)->nat; 3368 if (t == NULL) { 3369 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 3370 tablearg : cmd->arg1; 3371 LOOKUP_NAT(V_layer3_chain, nat_id, t); 3372 if (t == NULL) { 3373 retval = IP_FW_DENY; 3374 goto done; 3375 } 3376 if (cmd->arg1 != IP_FW_TABLEARG) 3377 ((ipfw_insn_nat *)cmd)->nat = t; 3378 } 3379 retval = ipfw_nat_ptr(args, t, m); 3380 } else 3381 retval = IP_FW_DENY; 3382 goto done; 3383 } 3384 3385 case O_REASS: { 3386 int ip_off; 3387 3388 f->pcnt++; 3389 f->bcnt += pktlen; 3390 ip_off = (args->eh != NULL) ? ntohs(ip->ip_off) : ip->ip_off; 3391 if (ip_off & (IP_MF | IP_OFFMASK)) { 3392 /* 3393 * ip_reass() expects len & off in host 3394 * byte order: fix them in case we come 3395 * from layer2. 3396 */ 3397 if (args->eh != NULL) { 3398 ip->ip_len = ntohs(ip->ip_len); 3399 ip->ip_off = ntohs(ip->ip_off); 3400 } 3401 3402 m = ip_reass(m); 3403 args->m = m; 3404 3405 /* 3406 * IP header checksum fixup after 3407 * reassembly and leave header 3408 * in network byte order. 3409 */ 3410 if (m != NULL) { 3411 int hlen; 3412 3413 ip = mtod(m, struct ip *); 3414 hlen = ip->ip_hl << 2; 3415 /* revert len & off for layer2 pkts */ 3416 if (args->eh != NULL) 3417 ip->ip_len = htons(ip->ip_len); 3418 ip->ip_sum = 0; 3419 if (hlen == sizeof(struct ip)) 3420 ip->ip_sum = in_cksum_hdr(ip); 3421 else 3422 ip->ip_sum = in_cksum(m, hlen); 3423 retval = IP_FW_REASS; 3424 args->rule = f; 3425 goto done; 3426 } else { 3427 retval = IP_FW_DENY; 3428 goto done; 3429 } 3430 } 3431 goto next_rule; 3432 } 3433 3434 default: 3435 panic("-- unknown opcode %d\n", cmd->opcode); 3436 } /* end of switch() on opcodes */ 3437 3438 if (cmd->len & F_NOT) 3439 match = !match; 3440 3441 if (match) { 3442 if (cmd->len & F_OR) 3443 skip_or = 1; 3444 } else { 3445 if (!(cmd->len & F_OR)) /* not an OR block, */ 3446 break; /* try next rule */ 3447 } 3448 3449 } /* end of inner for, scan opcodes */ 3450 3451next_rule:; /* try next rule */ 3452 3453 } /* end of outer for, scan rules */ 3454 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3455 IPFW_RUNLOCK(chain); 3456 return (IP_FW_DENY); 3457 3458done: 3459 /* Update statistics */ 3460 f->pcnt++; 3461 f->bcnt += pktlen; 3462 f->timestamp = time_uptime; 3463 IPFW_RUNLOCK(chain); 3464 return (retval); 3465 3466pullup_failed: 3467 if (V_fw_verbose) 3468 printf("ipfw: pullup failed\n"); 3469 return (IP_FW_DENY); 3470} 3471 3472/* 3473 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3474 * These will be reconstructed on the fly as packets are matched. 3475 */ 3476static void 3477flush_rule_ptrs(struct ip_fw_chain *chain) 3478{ 3479 struct ip_fw *rule; 3480 3481 IPFW_WLOCK_ASSERT(chain); 3482 3483 for (rule = chain->rules; rule; rule = rule->next) 3484 rule->next_rule = NULL; 3485} 3486 3487/* 3488 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3489 * possibly create a rule number and add the rule to the list. 3490 * Update the rule_number in the input struct so the caller knows it as well. 3491 */ 3492static int 3493add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3494{ 3495 INIT_VNET_IPFW(curvnet); 3496 struct ip_fw *rule, *f, *prev; 3497 int l = RULESIZE(input_rule); 3498 3499 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3500 return (EINVAL); 3501 3502 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3503 if (rule == NULL) 3504 return (ENOSPC); 3505 3506 bcopy(input_rule, rule, l); 3507 3508 rule->next = NULL; 3509 rule->next_rule = NULL; 3510 3511 rule->pcnt = 0; 3512 rule->bcnt = 0; 3513 rule->timestamp = 0; 3514 3515 IPFW_WLOCK(chain); 3516 3517 if (chain->rules == NULL) { /* default rule */ 3518 chain->rules = rule; 3519 goto done; 3520 } 3521 3522 /* 3523 * If rulenum is 0, find highest numbered rule before the 3524 * default rule, and add autoinc_step 3525 */ 3526 if (V_autoinc_step < 1) 3527 V_autoinc_step = 1; 3528 else if (V_autoinc_step > 1000) 3529 V_autoinc_step = 1000; 3530 if (rule->rulenum == 0) { 3531 /* 3532 * locate the highest numbered rule before default 3533 */ 3534 for (f = chain->rules; f; f = f->next) { 3535 if (f->rulenum == IPFW_DEFAULT_RULE) 3536 break; 3537 rule->rulenum = f->rulenum; 3538 } 3539 if (rule->rulenum < IPFW_DEFAULT_RULE - V_autoinc_step) 3540 rule->rulenum += V_autoinc_step; 3541 input_rule->rulenum = rule->rulenum; 3542 } 3543 3544 /* 3545 * Now insert the new rule in the right place in the sorted list. 3546 */ 3547 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3548 if (f->rulenum > rule->rulenum) { /* found the location */ 3549 if (prev) { 3550 rule->next = f; 3551 prev->next = rule; 3552 } else { /* head insert */ 3553 rule->next = chain->rules; 3554 chain->rules = rule; 3555 } 3556 break; 3557 } 3558 } 3559 flush_rule_ptrs(chain); 3560done: 3561 V_static_count++; 3562 V_static_len += l; 3563 IPFW_WUNLOCK(chain); 3564 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3565 rule->rulenum, V_static_count);) 3566 return (0); 3567} 3568 3569/** 3570 * Remove a static rule (including derived * dynamic rules) 3571 * and place it on the ``reap list'' for later reclamation. 3572 * The caller is in charge of clearing rule pointers to avoid 3573 * dangling pointers. 3574 * @return a pointer to the next entry. 3575 * Arguments are not checked, so they better be correct. 3576 */ 3577static struct ip_fw * 3578remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 3579 struct ip_fw *prev) 3580{ 3581 INIT_VNET_IPFW(curvnet); 3582 struct ip_fw *n; 3583 int l = RULESIZE(rule); 3584 3585 IPFW_WLOCK_ASSERT(chain); 3586 3587 n = rule->next; 3588 IPFW_DYN_LOCK(); 3589 remove_dyn_rule(rule, NULL /* force removal */); 3590 IPFW_DYN_UNLOCK(); 3591 if (prev == NULL) 3592 chain->rules = n; 3593 else 3594 prev->next = n; 3595 V_static_count--; 3596 V_static_len -= l; 3597 3598 rule->next = chain->reap; 3599 chain->reap = rule; 3600 3601 return n; 3602} 3603 3604/* 3605 * Hook for cleaning up dummynet when an ipfw rule is deleted. 3606 * Set/cleared when dummynet module is loaded/unloaded. 3607 */ 3608void (*ip_dn_ruledel_ptr)(void *) = NULL; 3609 3610/** 3611 * Reclaim storage associated with a list of rules. This is 3612 * typically the list created using remove_rule. 3613 */ 3614static void 3615reap_rules(struct ip_fw *head) 3616{ 3617 struct ip_fw *rule; 3618 3619 while ((rule = head) != NULL) { 3620 head = head->next; 3621 if (ip_dn_ruledel_ptr) 3622 ip_dn_ruledel_ptr(rule); 3623 free(rule, M_IPFW); 3624 } 3625} 3626 3627/* 3628 * Remove all rules from a chain (except rules in set RESVD_SET 3629 * unless kill_default = 1). The caller is responsible for 3630 * reclaiming storage for the rules left in chain->reap. 3631 */ 3632static void 3633free_chain(struct ip_fw_chain *chain, int kill_default) 3634{ 3635 struct ip_fw *prev, *rule; 3636 3637 IPFW_WLOCK_ASSERT(chain); 3638 3639 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3640 for (prev = NULL, rule = chain->rules; rule ; ) 3641 if (kill_default || rule->set != RESVD_SET) 3642 rule = remove_rule(chain, rule, prev); 3643 else { 3644 prev = rule; 3645 rule = rule->next; 3646 } 3647} 3648 3649/** 3650 * Remove all rules with given number, and also do set manipulation. 3651 * Assumes chain != NULL && *chain != NULL. 3652 * 3653 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3654 * the next 8 bits are the new set, the top 8 bits are the command: 3655 * 3656 * 0 delete rules with given number 3657 * 1 delete rules with given set number 3658 * 2 move rules with given number to new set 3659 * 3 move rules with given set number to new set 3660 * 4 swap sets with given numbers 3661 * 5 delete rules with given number and with given set number 3662 */ 3663static int 3664del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3665{ 3666 struct ip_fw *prev = NULL, *rule; 3667 u_int16_t rulenum; /* rule or old_set */ 3668 u_int8_t cmd, new_set; 3669 3670 rulenum = arg & 0xffff; 3671 cmd = (arg >> 24) & 0xff; 3672 new_set = (arg >> 16) & 0xff; 3673 3674 if (cmd > 5 || new_set > RESVD_SET) 3675 return EINVAL; 3676 if (cmd == 0 || cmd == 2 || cmd == 5) { 3677 if (rulenum >= IPFW_DEFAULT_RULE) 3678 return EINVAL; 3679 } else { 3680 if (rulenum > RESVD_SET) /* old_set */ 3681 return EINVAL; 3682 } 3683 3684 IPFW_WLOCK(chain); 3685 rule = chain->rules; 3686 chain->reap = NULL; 3687 switch (cmd) { 3688 case 0: /* delete rules with given number */ 3689 /* 3690 * locate first rule to delete 3691 */ 3692 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3693 ; 3694 if (rule->rulenum != rulenum) { 3695 IPFW_WUNLOCK(chain); 3696 return EINVAL; 3697 } 3698 3699 /* 3700 * flush pointers outside the loop, then delete all matching 3701 * rules. prev remains the same throughout the cycle. 3702 */ 3703 flush_rule_ptrs(chain); 3704 while (rule->rulenum == rulenum) 3705 rule = remove_rule(chain, rule, prev); 3706 break; 3707 3708 case 1: /* delete all rules with given set number */ 3709 flush_rule_ptrs(chain); 3710 rule = chain->rules; 3711 while (rule->rulenum < IPFW_DEFAULT_RULE) 3712 if (rule->set == rulenum) 3713 rule = remove_rule(chain, rule, prev); 3714 else { 3715 prev = rule; 3716 rule = rule->next; 3717 } 3718 break; 3719 3720 case 2: /* move rules with given number to new set */ 3721 rule = chain->rules; 3722 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3723 if (rule->rulenum == rulenum) 3724 rule->set = new_set; 3725 break; 3726 3727 case 3: /* move rules with given set number to new set */ 3728 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3729 if (rule->set == rulenum) 3730 rule->set = new_set; 3731 break; 3732 3733 case 4: /* swap two sets */ 3734 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3735 if (rule->set == rulenum) 3736 rule->set = new_set; 3737 else if (rule->set == new_set) 3738 rule->set = rulenum; 3739 break; 3740 case 5: /* delete rules with given number and with given set number. 3741 * rulenum - given rule number; 3742 * new_set - given set number. 3743 */ 3744 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3745 ; 3746 if (rule->rulenum != rulenum) { 3747 IPFW_WUNLOCK(chain); 3748 return (EINVAL); 3749 } 3750 flush_rule_ptrs(chain); 3751 while (rule->rulenum == rulenum) { 3752 if (rule->set == new_set) 3753 rule = remove_rule(chain, rule, prev); 3754 else { 3755 prev = rule; 3756 rule = rule->next; 3757 } 3758 } 3759 } 3760 /* 3761 * Look for rules to reclaim. We grab the list before 3762 * releasing the lock then reclaim them w/o the lock to 3763 * avoid a LOR with dummynet. 3764 */ 3765 rule = chain->reap; 3766 chain->reap = NULL; 3767 IPFW_WUNLOCK(chain); 3768 if (rule) 3769 reap_rules(rule); 3770 return 0; 3771} 3772 3773/* 3774 * Clear counters for a specific rule. 3775 * The enclosing "table" is assumed locked. 3776 */ 3777static void 3778clear_counters(struct ip_fw *rule, int log_only) 3779{ 3780 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3781 3782 if (log_only == 0) { 3783 rule->bcnt = rule->pcnt = 0; 3784 rule->timestamp = 0; 3785 } 3786 if (l->o.opcode == O_LOG) 3787 l->log_left = l->max_log; 3788} 3789 3790/** 3791 * Reset some or all counters on firewall rules. 3792 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 3793 * the next 8 bits are the set number, the top 8 bits are the command: 3794 * 0 work with rules from all set's; 3795 * 1 work with rules only from specified set. 3796 * Specified rule number is zero if we want to clear all entries. 3797 * log_only is 1 if we only want to reset logs, zero otherwise. 3798 */ 3799static int 3800zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 3801{ 3802 INIT_VNET_IPFW(curvnet); 3803 struct ip_fw *rule; 3804 char *msg; 3805 3806 uint16_t rulenum = arg & 0xffff; 3807 uint8_t set = (arg >> 16) & 0xff; 3808 uint8_t cmd = (arg >> 24) & 0xff; 3809 3810 if (cmd > 1) 3811 return (EINVAL); 3812 if (cmd == 1 && set > RESVD_SET) 3813 return (EINVAL); 3814 3815 IPFW_WLOCK(chain); 3816 if (rulenum == 0) { 3817 V_norule_counter = 0; 3818 for (rule = chain->rules; rule; rule = rule->next) { 3819 /* Skip rules from another set. */ 3820 if (cmd == 1 && rule->set != set) 3821 continue; 3822 clear_counters(rule, log_only); 3823 } 3824 msg = log_only ? "All logging counts reset" : 3825 "Accounting cleared"; 3826 } else { 3827 int cleared = 0; 3828 /* 3829 * We can have multiple rules with the same number, so we 3830 * need to clear them all. 3831 */ 3832 for (rule = chain->rules; rule; rule = rule->next) 3833 if (rule->rulenum == rulenum) { 3834 while (rule && rule->rulenum == rulenum) { 3835 if (cmd == 0 || rule->set == set) 3836 clear_counters(rule, log_only); 3837 rule = rule->next; 3838 } 3839 cleared = 1; 3840 break; 3841 } 3842 if (!cleared) { /* we did not find any matching rules */ 3843 IPFW_WUNLOCK(chain); 3844 return (EINVAL); 3845 } 3846 msg = log_only ? "logging count reset" : "cleared"; 3847 } 3848 IPFW_WUNLOCK(chain); 3849 3850 if (V_fw_verbose) { 3851 int lev = LOG_SECURITY | LOG_NOTICE; 3852 3853 if (rulenum) 3854 log(lev, "ipfw: Entry %d %s.\n", rulenum, msg); 3855 else 3856 log(lev, "ipfw: %s.\n", msg); 3857 } 3858 return (0); 3859} 3860 3861/* 3862 * Check validity of the structure before insert. 3863 * Fortunately rules are simple, so this mostly need to check rule sizes. 3864 */ 3865static int 3866check_ipfw_struct(struct ip_fw *rule, int size) 3867{ 3868 int l, cmdlen = 0; 3869 int have_action=0; 3870 ipfw_insn *cmd; 3871 3872 if (size < sizeof(*rule)) { 3873 printf("ipfw: rule too short\n"); 3874 return (EINVAL); 3875 } 3876 /* first, check for valid size */ 3877 l = RULESIZE(rule); 3878 if (l != size) { 3879 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3880 return (EINVAL); 3881 } 3882 if (rule->act_ofs >= rule->cmd_len) { 3883 printf("ipfw: bogus action offset (%u > %u)\n", 3884 rule->act_ofs, rule->cmd_len - 1); 3885 return (EINVAL); 3886 } 3887 /* 3888 * Now go for the individual checks. Very simple ones, basically only 3889 * instruction sizes. 3890 */ 3891 for (l = rule->cmd_len, cmd = rule->cmd ; 3892 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3893 cmdlen = F_LEN(cmd); 3894 if (cmdlen > l) { 3895 printf("ipfw: opcode %d size truncated\n", 3896 cmd->opcode); 3897 return EINVAL; 3898 } 3899 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3900 switch (cmd->opcode) { 3901 case O_PROBE_STATE: 3902 case O_KEEP_STATE: 3903 case O_PROTO: 3904 case O_IP_SRC_ME: 3905 case O_IP_DST_ME: 3906 case O_LAYER2: 3907 case O_IN: 3908 case O_FRAG: 3909 case O_DIVERTED: 3910 case O_IPOPT: 3911 case O_IPTOS: 3912 case O_IPPRECEDENCE: 3913 case O_IPVER: 3914 case O_TCPWIN: 3915 case O_TCPFLAGS: 3916 case O_TCPOPTS: 3917 case O_ESTAB: 3918 case O_VERREVPATH: 3919 case O_VERSRCREACH: 3920 case O_ANTISPOOF: 3921 case O_IPSEC: 3922#ifdef INET6 3923 case O_IP6_SRC_ME: 3924 case O_IP6_DST_ME: 3925 case O_EXT_HDR: 3926 case O_IP6: 3927#endif 3928 case O_IP4: 3929 case O_TAG: 3930 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3931 goto bad_size; 3932 break; 3933 3934 case O_FIB: 3935 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3936 goto bad_size; 3937 if (cmd->arg1 >= rt_numfibs) { 3938 printf("ipfw: invalid fib number %d\n", 3939 cmd->arg1); 3940 return EINVAL; 3941 } 3942 break; 3943 3944 case O_SETFIB: 3945 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3946 goto bad_size; 3947 if (cmd->arg1 >= rt_numfibs) { 3948 printf("ipfw: invalid fib number %d\n", 3949 cmd->arg1); 3950 return EINVAL; 3951 } 3952 goto check_action; 3953 3954 case O_UID: 3955 case O_GID: 3956 case O_JAIL: 3957 case O_IP_SRC: 3958 case O_IP_DST: 3959 case O_TCPSEQ: 3960 case O_TCPACK: 3961 case O_PROB: 3962 case O_ICMPTYPE: 3963 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3964 goto bad_size; 3965 break; 3966 3967 case O_LIMIT: 3968 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3969 goto bad_size; 3970 break; 3971 3972 case O_LOG: 3973 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3974 goto bad_size; 3975 3976 ((ipfw_insn_log *)cmd)->log_left = 3977 ((ipfw_insn_log *)cmd)->max_log; 3978 3979 break; 3980 3981 case O_IP_SRC_MASK: 3982 case O_IP_DST_MASK: 3983 /* only odd command lengths */ 3984 if ( !(cmdlen & 1) || cmdlen > 31) 3985 goto bad_size; 3986 break; 3987 3988 case O_IP_SRC_SET: 3989 case O_IP_DST_SET: 3990 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3991 printf("ipfw: invalid set size %d\n", 3992 cmd->arg1); 3993 return EINVAL; 3994 } 3995 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3996 (cmd->arg1+31)/32 ) 3997 goto bad_size; 3998 break; 3999 4000 case O_IP_SRC_LOOKUP: 4001 case O_IP_DST_LOOKUP: 4002 if (cmd->arg1 >= IPFW_TABLES_MAX) { 4003 printf("ipfw: invalid table number %d\n", 4004 cmd->arg1); 4005 return (EINVAL); 4006 } 4007 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 4008 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 4009 goto bad_size; 4010 break; 4011 4012 case O_MACADDR2: 4013 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 4014 goto bad_size; 4015 break; 4016 4017 case O_NOP: 4018 case O_IPID: 4019 case O_IPTTL: 4020 case O_IPLEN: 4021 case O_TCPDATALEN: 4022 case O_TAGGED: 4023 if (cmdlen < 1 || cmdlen > 31) 4024 goto bad_size; 4025 break; 4026 4027 case O_MAC_TYPE: 4028 case O_IP_SRCPORT: 4029 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 4030 if (cmdlen < 2 || cmdlen > 31) 4031 goto bad_size; 4032 break; 4033 4034 case O_RECV: 4035 case O_XMIT: 4036 case O_VIA: 4037 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 4038 goto bad_size; 4039 break; 4040 4041 case O_ALTQ: 4042 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 4043 goto bad_size; 4044 break; 4045 4046 case O_PIPE: 4047 case O_QUEUE: 4048 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4049 goto bad_size; 4050 goto check_action; 4051 4052 case O_FORWARD_IP: 4053#ifdef IPFIREWALL_FORWARD 4054 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 4055 goto bad_size; 4056 goto check_action; 4057#else 4058 return EINVAL; 4059#endif 4060 4061 case O_DIVERT: 4062 case O_TEE: 4063 if (ip_divert_ptr == NULL) 4064 return EINVAL; 4065 else 4066 goto check_size; 4067 case O_NETGRAPH: 4068 case O_NGTEE: 4069 if (!NG_IPFW_LOADED) 4070 return EINVAL; 4071 else 4072 goto check_size; 4073 case O_NAT: 4074 if (!IPFW_NAT_LOADED) 4075 return EINVAL; 4076 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 4077 goto bad_size; 4078 goto check_action; 4079 case O_FORWARD_MAC: /* XXX not implemented yet */ 4080 case O_CHECK_STATE: 4081 case O_COUNT: 4082 case O_ACCEPT: 4083 case O_DENY: 4084 case O_REJECT: 4085#ifdef INET6 4086 case O_UNREACH6: 4087#endif 4088 case O_SKIPTO: 4089 case O_REASS: 4090check_size: 4091 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4092 goto bad_size; 4093check_action: 4094 if (have_action) { 4095 printf("ipfw: opcode %d, multiple actions" 4096 " not allowed\n", 4097 cmd->opcode); 4098 return EINVAL; 4099 } 4100 have_action = 1; 4101 if (l != cmdlen) { 4102 printf("ipfw: opcode %d, action must be" 4103 " last opcode\n", 4104 cmd->opcode); 4105 return EINVAL; 4106 } 4107 break; 4108#ifdef INET6 4109 case O_IP6_SRC: 4110 case O_IP6_DST: 4111 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 4112 F_INSN_SIZE(ipfw_insn)) 4113 goto bad_size; 4114 break; 4115 4116 case O_FLOW6ID: 4117 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4118 ((ipfw_insn_u32 *)cmd)->o.arg1) 4119 goto bad_size; 4120 break; 4121 4122 case O_IP6_SRC_MASK: 4123 case O_IP6_DST_MASK: 4124 if ( !(cmdlen & 1) || cmdlen > 127) 4125 goto bad_size; 4126 break; 4127 case O_ICMP6TYPE: 4128 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4129 goto bad_size; 4130 break; 4131#endif 4132 4133 default: 4134 switch (cmd->opcode) { 4135#ifndef INET6 4136 case O_IP6_SRC_ME: 4137 case O_IP6_DST_ME: 4138 case O_EXT_HDR: 4139 case O_IP6: 4140 case O_UNREACH6: 4141 case O_IP6_SRC: 4142 case O_IP6_DST: 4143 case O_FLOW6ID: 4144 case O_IP6_SRC_MASK: 4145 case O_IP6_DST_MASK: 4146 case O_ICMP6TYPE: 4147 printf("ipfw: no IPv6 support in kernel\n"); 4148 return EPROTONOSUPPORT; 4149#endif 4150 default: 4151 printf("ipfw: opcode %d, unknown opcode\n", 4152 cmd->opcode); 4153 return EINVAL; 4154 } 4155 } 4156 } 4157 if (have_action == 0) { 4158 printf("ipfw: missing action\n"); 4159 return EINVAL; 4160 } 4161 return 0; 4162 4163bad_size: 4164 printf("ipfw: opcode %d size %d wrong\n", 4165 cmd->opcode, cmdlen); 4166 return EINVAL; 4167} 4168 4169/* 4170 * Copy the static and dynamic rules to the supplied buffer 4171 * and return the amount of space actually used. 4172 */ 4173static size_t 4174ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4175{ 4176 INIT_VNET_IPFW(curvnet); 4177 char *bp = buf; 4178 char *ep = bp + space; 4179 struct ip_fw *rule; 4180 int i; 4181 time_t boot_seconds; 4182 4183 boot_seconds = boottime.tv_sec; 4184 /* XXX this can take a long time and locking will block packet flow */ 4185 IPFW_RLOCK(chain); 4186 for (rule = chain->rules; rule ; rule = rule->next) { 4187 /* 4188 * Verify the entry fits in the buffer in case the 4189 * rules changed between calculating buffer space and 4190 * now. This would be better done using a generation 4191 * number but should suffice for now. 4192 */ 4193 i = RULESIZE(rule); 4194 if (bp + i <= ep) { 4195 bcopy(rule, bp, i); 4196 /* 4197 * XXX HACK. Store the disable mask in the "next" 4198 * pointer in a wild attempt to keep the ABI the same. 4199 * Why do we do this on EVERY rule? 4200 */ 4201 bcopy(&V_set_disable, 4202 &(((struct ip_fw *)bp)->next_rule), 4203 sizeof(V_set_disable)); 4204 if (((struct ip_fw *)bp)->timestamp) 4205 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4206 bp += i; 4207 } 4208 } 4209 IPFW_RUNLOCK(chain); 4210 if (V_ipfw_dyn_v) { 4211 ipfw_dyn_rule *p, *last = NULL; 4212 4213 IPFW_DYN_LOCK(); 4214 for (i = 0 ; i < V_curr_dyn_buckets; i++) 4215 for (p = V_ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4216 if (bp + sizeof *p <= ep) { 4217 ipfw_dyn_rule *dst = 4218 (ipfw_dyn_rule *)bp; 4219 bcopy(p, dst, sizeof *p); 4220 bcopy(&(p->rule->rulenum), &(dst->rule), 4221 sizeof(p->rule->rulenum)); 4222 /* 4223 * store set number into high word of 4224 * dst->rule pointer. 4225 */ 4226 bcopy(&(p->rule->set), 4227 (char *)&dst->rule + 4228 sizeof(p->rule->rulenum), 4229 sizeof(p->rule->set)); 4230 /* 4231 * store a non-null value in "next". 4232 * The userland code will interpret a 4233 * NULL here as a marker 4234 * for the last dynamic rule. 4235 */ 4236 bcopy(&dst, &dst->next, sizeof(dst)); 4237 last = dst; 4238 dst->expire = 4239 TIME_LEQ(dst->expire, time_uptime) ? 4240 0 : dst->expire - time_uptime ; 4241 bp += sizeof(ipfw_dyn_rule); 4242 } 4243 } 4244 IPFW_DYN_UNLOCK(); 4245 if (last != NULL) /* mark last dynamic rule */ 4246 bzero(&last->next, sizeof(last)); 4247 } 4248 return (bp - (char *)buf); 4249} 4250 4251 4252/** 4253 * {set|get}sockopt parser. 4254 */ 4255static int 4256ipfw_ctl(struct sockopt *sopt) 4257{ 4258#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4259 INIT_VNET_IPFW(curvnet); 4260 int error; 4261 size_t size; 4262 struct ip_fw *buf, *rule; 4263 u_int32_t rulenum[2]; 4264 4265 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4266 if (error) 4267 return (error); 4268 4269 /* 4270 * Disallow modifications in really-really secure mode, but still allow 4271 * the logging counters to be reset. 4272 */ 4273 if (sopt->sopt_name == IP_FW_ADD || 4274 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4275 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4276 if (error) 4277 return (error); 4278 } 4279 4280 error = 0; 4281 4282 switch (sopt->sopt_name) { 4283 case IP_FW_GET: 4284 /* 4285 * pass up a copy of the current rules. Static rules 4286 * come first (the last of which has number IPFW_DEFAULT_RULE), 4287 * followed by a possibly empty list of dynamic rule. 4288 * The last dynamic rule has NULL in the "next" field. 4289 * 4290 * Note that the calculated size is used to bound the 4291 * amount of data returned to the user. The rule set may 4292 * change between calculating the size and returning the 4293 * data in which case we'll just return what fits. 4294 */ 4295 size = V_static_len; /* size of static rules */ 4296 if (V_ipfw_dyn_v) /* add size of dyn.rules */ 4297 size += (V_dyn_count * sizeof(ipfw_dyn_rule)); 4298 4299 /* 4300 * XXX todo: if the user passes a short length just to know 4301 * how much room is needed, do not bother filling up the 4302 * buffer, just jump to the sooptcopyout. 4303 */ 4304 buf = malloc(size, M_TEMP, M_WAITOK); 4305 error = sooptcopyout(sopt, buf, 4306 ipfw_getrules(&V_layer3_chain, buf, size)); 4307 free(buf, M_TEMP); 4308 break; 4309 4310 case IP_FW_FLUSH: 4311 /* 4312 * Normally we cannot release the lock on each iteration. 4313 * We could do it here only because we start from the head all 4314 * the times so there is no risk of missing some entries. 4315 * On the other hand, the risk is that we end up with 4316 * a very inconsistent ruleset, so better keep the lock 4317 * around the whole cycle. 4318 * 4319 * XXX this code can be improved by resetting the head of 4320 * the list to point to the default rule, and then freeing 4321 * the old list without the need for a lock. 4322 */ 4323 4324 IPFW_WLOCK(&V_layer3_chain); 4325 V_layer3_chain.reap = NULL; 4326 free_chain(&V_layer3_chain, 0 /* keep default rule */); 4327 rule = V_layer3_chain.reap; 4328 V_layer3_chain.reap = NULL; 4329 IPFW_WUNLOCK(&V_layer3_chain); 4330 if (rule != NULL) 4331 reap_rules(rule); 4332 break; 4333 4334 case IP_FW_ADD: 4335 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4336 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4337 sizeof(struct ip_fw) ); 4338 if (error == 0) 4339 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4340 if (error == 0) { 4341 error = add_rule(&V_layer3_chain, rule); 4342 size = RULESIZE(rule); 4343 if (!error && sopt->sopt_dir == SOPT_GET) 4344 error = sooptcopyout(sopt, rule, size); 4345 } 4346 free(rule, M_TEMP); 4347 break; 4348 4349 case IP_FW_DEL: 4350 /* 4351 * IP_FW_DEL is used for deleting single rules or sets, 4352 * and (ab)used to atomically manipulate sets. Argument size 4353 * is used to distinguish between the two: 4354 * sizeof(u_int32_t) 4355 * delete single rule or set of rules, 4356 * or reassign rules (or sets) to a different set. 4357 * 2*sizeof(u_int32_t) 4358 * atomic disable/enable sets. 4359 * first u_int32_t contains sets to be disabled, 4360 * second u_int32_t contains sets to be enabled. 4361 */ 4362 error = sooptcopyin(sopt, rulenum, 4363 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4364 if (error) 4365 break; 4366 size = sopt->sopt_valsize; 4367 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4368 error = del_entry(&V_layer3_chain, rulenum[0]); 4369 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4370 V_set_disable = 4371 (V_set_disable | rulenum[0]) & ~rulenum[1] & 4372 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4373 else 4374 error = EINVAL; 4375 break; 4376 4377 case IP_FW_ZERO: 4378 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 4379 rulenum[0] = 0; 4380 if (sopt->sopt_val != 0) { 4381 error = sooptcopyin(sopt, rulenum, 4382 sizeof(u_int32_t), sizeof(u_int32_t)); 4383 if (error) 4384 break; 4385 } 4386 error = zero_entry(&V_layer3_chain, rulenum[0], 4387 sopt->sopt_name == IP_FW_RESETLOG); 4388 break; 4389 4390 case IP_FW_TABLE_ADD: 4391 { 4392 ipfw_table_entry ent; 4393 4394 error = sooptcopyin(sopt, &ent, 4395 sizeof(ent), sizeof(ent)); 4396 if (error) 4397 break; 4398 error = add_table_entry(&V_layer3_chain, ent.tbl, 4399 ent.addr, ent.masklen, ent.value); 4400 } 4401 break; 4402 4403 case IP_FW_TABLE_DEL: 4404 { 4405 ipfw_table_entry ent; 4406 4407 error = sooptcopyin(sopt, &ent, 4408 sizeof(ent), sizeof(ent)); 4409 if (error) 4410 break; 4411 error = del_table_entry(&V_layer3_chain, ent.tbl, 4412 ent.addr, ent.masklen); 4413 } 4414 break; 4415 4416 case IP_FW_TABLE_FLUSH: 4417 { 4418 u_int16_t tbl; 4419 4420 error = sooptcopyin(sopt, &tbl, 4421 sizeof(tbl), sizeof(tbl)); 4422 if (error) 4423 break; 4424 IPFW_WLOCK(&V_layer3_chain); 4425 error = flush_table(&V_layer3_chain, tbl); 4426 IPFW_WUNLOCK(&V_layer3_chain); 4427 } 4428 break; 4429 4430 case IP_FW_TABLE_GETSIZE: 4431 { 4432 u_int32_t tbl, cnt; 4433 4434 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4435 sizeof(tbl)))) 4436 break; 4437 IPFW_RLOCK(&V_layer3_chain); 4438 error = count_table(&V_layer3_chain, tbl, &cnt); 4439 IPFW_RUNLOCK(&V_layer3_chain); 4440 if (error) 4441 break; 4442 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4443 } 4444 break; 4445 4446 case IP_FW_TABLE_LIST: 4447 { 4448 ipfw_table *tbl; 4449 4450 if (sopt->sopt_valsize < sizeof(*tbl)) { 4451 error = EINVAL; 4452 break; 4453 } 4454 size = sopt->sopt_valsize; 4455 tbl = malloc(size, M_TEMP, M_WAITOK); 4456 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4457 if (error) { 4458 free(tbl, M_TEMP); 4459 break; 4460 } 4461 tbl->size = (size - sizeof(*tbl)) / 4462 sizeof(ipfw_table_entry); 4463 IPFW_RLOCK(&V_layer3_chain); 4464 error = dump_table(&V_layer3_chain, tbl); 4465 IPFW_RUNLOCK(&V_layer3_chain); 4466 if (error) { 4467 free(tbl, M_TEMP); 4468 break; 4469 } 4470 error = sooptcopyout(sopt, tbl, size); 4471 free(tbl, M_TEMP); 4472 } 4473 break; 4474 4475 case IP_FW_NAT_CFG: 4476 if (IPFW_NAT_LOADED) 4477 error = ipfw_nat_cfg_ptr(sopt); 4478 else { 4479 printf("IP_FW_NAT_CFG: %s\n", 4480 "ipfw_nat not present, please load it"); 4481 error = EINVAL; 4482 } 4483 break; 4484 4485 case IP_FW_NAT_DEL: 4486 if (IPFW_NAT_LOADED) 4487 error = ipfw_nat_del_ptr(sopt); 4488 else { 4489 printf("IP_FW_NAT_DEL: %s\n", 4490 "ipfw_nat not present, please load it"); 4491 error = EINVAL; 4492 } 4493 break; 4494 4495 case IP_FW_NAT_GET_CONFIG: 4496 if (IPFW_NAT_LOADED) 4497 error = ipfw_nat_get_cfg_ptr(sopt); 4498 else { 4499 printf("IP_FW_NAT_GET_CFG: %s\n", 4500 "ipfw_nat not present, please load it"); 4501 error = EINVAL; 4502 } 4503 break; 4504 4505 case IP_FW_NAT_GET_LOG: 4506 if (IPFW_NAT_LOADED) 4507 error = ipfw_nat_get_log_ptr(sopt); 4508 else { 4509 printf("IP_FW_NAT_GET_LOG: %s\n", 4510 "ipfw_nat not present, please load it"); 4511 error = EINVAL; 4512 } 4513 break; 4514 4515 default: 4516 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4517 error = EINVAL; 4518 } 4519 4520 return (error); 4521#undef RULE_MAXSIZE 4522} 4523 4524/** 4525 * dummynet needs a reference to the default rule, because rules can be 4526 * deleted while packets hold a reference to them. When this happens, 4527 * dummynet changes the reference to the default rule (it could well be a 4528 * NULL pointer, but this way we do not need to check for the special 4529 * case, plus here he have info on the default behaviour). 4530 */ 4531struct ip_fw *ip_fw_default_rule; 4532 4533/* 4534 * This procedure is only used to handle keepalives. It is invoked 4535 * every dyn_keepalive_period 4536 */ 4537static void 4538ipfw_tick(void * __unused unused) 4539{ 4540 INIT_VNET_IPFW(curvnet); 4541 struct mbuf *m0, *m, *mnext, **mtailp; 4542 int i; 4543 ipfw_dyn_rule *q; 4544 4545 if (V_dyn_keepalive == 0 || V_ipfw_dyn_v == NULL || V_dyn_count == 0) 4546 goto done; 4547 4548 /* 4549 * We make a chain of packets to go out here -- not deferring 4550 * until after we drop the IPFW dynamic rule lock would result 4551 * in a lock order reversal with the normal packet input -> ipfw 4552 * call stack. 4553 */ 4554 m0 = NULL; 4555 mtailp = &m0; 4556 IPFW_DYN_LOCK(); 4557 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 4558 for (q = V_ipfw_dyn_v[i] ; q ; q = q->next ) { 4559 if (q->dyn_type == O_LIMIT_PARENT) 4560 continue; 4561 if (q->id.proto != IPPROTO_TCP) 4562 continue; 4563 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4564 continue; 4565 if (TIME_LEQ(time_uptime + V_dyn_keepalive_interval, 4566 q->expire)) 4567 continue; /* too early */ 4568 if (TIME_LEQ(q->expire, time_uptime)) 4569 continue; /* too late, rule expired */ 4570 4571 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4572 q->ack_fwd, TH_SYN); 4573 if (*mtailp != NULL) 4574 mtailp = &(*mtailp)->m_nextpkt; 4575 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4576 q->ack_rev, 0); 4577 if (*mtailp != NULL) 4578 mtailp = &(*mtailp)->m_nextpkt; 4579 } 4580 } 4581 IPFW_DYN_UNLOCK(); 4582 for (m = mnext = m0; m != NULL; m = mnext) { 4583 mnext = m->m_nextpkt; 4584 m->m_nextpkt = NULL; 4585 ip_output(m, NULL, NULL, 0, NULL, NULL); 4586 } 4587done: 4588 callout_reset(&V_ipfw_timeout, V_dyn_keepalive_period * hz, 4589 ipfw_tick, NULL); 4590} 4591 4592int 4593ipfw_init(void) 4594{ 4595 INIT_VNET_IPFW(curvnet); 4596 struct ip_fw default_rule; 4597 int error; 4598 4599 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 4600 4601 V_ipfw_dyn_v = NULL; 4602 V_dyn_buckets = 256; /* must be power of 2 */ 4603 V_curr_dyn_buckets = 256; /* must be power of 2 */ 4604 4605 V_dyn_ack_lifetime = 300; 4606 V_dyn_syn_lifetime = 20; 4607 V_dyn_fin_lifetime = 1; 4608 V_dyn_rst_lifetime = 1; 4609 V_dyn_udp_lifetime = 10; 4610 V_dyn_short_lifetime = 5; 4611 4612 V_dyn_keepalive_interval = 20; 4613 V_dyn_keepalive_period = 5; 4614 V_dyn_keepalive = 1; /* do send keepalives */ 4615 4616 V_dyn_max = 4096; /* max # of dynamic rules */ 4617 4618 V_fw_deny_unknown_exthdrs = 1; 4619 4620#ifdef INET6 4621 /* Setup IPv6 fw sysctl tree. */ 4622 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4623 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4624 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4625 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4626 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4627 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4628 &V_fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4629 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4630 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4631 &V_fw_deny_unknown_exthdrs, 0, 4632 "Deny packets with unknown IPv6 Extension Headers"); 4633#endif 4634 4635 V_layer3_chain.rules = NULL; 4636 IPFW_LOCK_INIT(&V_layer3_chain); 4637 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4638 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4639 UMA_ALIGN_PTR, 0); 4640 IPFW_DYN_LOCK_INIT(); 4641 callout_init(&V_ipfw_timeout, CALLOUT_MPSAFE); 4642 4643 bzero(&default_rule, sizeof default_rule); 4644 4645 default_rule.act_ofs = 0; 4646 default_rule.rulenum = IPFW_DEFAULT_RULE; 4647 default_rule.cmd_len = 1; 4648 default_rule.set = RESVD_SET; 4649 4650 default_rule.cmd[0].len = 1; 4651 default_rule.cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 4652 4653 error = add_rule(&V_layer3_chain, &default_rule); 4654 if (error != 0) { 4655 printf("ipfw2: error %u initializing default rule " 4656 "(support disabled)\n", error); 4657 IPFW_DYN_LOCK_DESTROY(); 4658 IPFW_LOCK_DESTROY(&V_layer3_chain); 4659 uma_zdestroy(ipfw_dyn_rule_zone); 4660 return (error); 4661 } 4662 4663 ip_fw_default_rule = V_layer3_chain.rules; 4664 printf("ipfw2 " 4665#ifdef INET6 4666 "(+ipv6) " 4667#endif 4668 "initialized, divert %s, nat %s, " 4669 "rule-based forwarding " 4670#ifdef IPFIREWALL_FORWARD 4671 "enabled, " 4672#else 4673 "disabled, " 4674#endif 4675 "default to %s, logging ", 4676#ifdef IPDIVERT 4677 "enabled", 4678#else 4679 "loadable", 4680#endif 4681#ifdef IPFIREWALL_NAT 4682 "enabled", 4683#else 4684 "loadable", 4685#endif 4686 4687 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4688 4689#ifdef IPFIREWALL_VERBOSE 4690 V_fw_verbose = 1; 4691#endif 4692#ifdef IPFIREWALL_VERBOSE_LIMIT 4693 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4694#endif 4695 if (V_fw_verbose == 0) 4696 printf("disabled\n"); 4697 else if (V_verbose_limit == 0) 4698 printf("unlimited\n"); 4699 else 4700 printf("limited to %d packets/entry by default\n", 4701 V_verbose_limit); 4702 4703 error = init_tables(&V_layer3_chain); 4704 if (error) { 4705 IPFW_DYN_LOCK_DESTROY(); 4706 IPFW_LOCK_DESTROY(&V_layer3_chain); 4707 uma_zdestroy(ipfw_dyn_rule_zone); 4708 return (error); 4709 } 4710 ip_fw_ctl_ptr = ipfw_ctl; 4711 ip_fw_chk_ptr = ipfw_chk; 4712 callout_reset(&V_ipfw_timeout, hz, ipfw_tick, NULL); 4713 LIST_INIT(&V_layer3_chain.nat); 4714 return (0); 4715} 4716 4717void 4718ipfw_destroy(void) 4719{ 4720 INIT_VNET_IPFW(curvnet); 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