ip_fw2.c revision 172387
1/*- 2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions 6 * are met: 7 * 1. Redistributions of source code must retain the above copyright 8 * notice, this list of conditions and the following disclaimer. 9 * 2. Redistributions in binary form must reproduce the above copyright 10 * notice, this list of conditions and the following disclaimer in the 11 * documentation and/or other materials provided with the distribution. 12 * 13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23 * SUCH DAMAGE. 24 * 25 * $FreeBSD: head/sys/netinet/ip_fw2.c 172387 2007-09-29 15:01:41Z maxim $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdivert.h" 38#include "opt_ipdn.h" 39#include "opt_inet.h" 40#ifndef INET 41#error IPFIREWALL requires INET. 42#endif /* INET */ 43#endif 44#include "opt_inet6.h" 45#include "opt_ipsec.h" 46#include "opt_mac.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 <net/if.h> 67#include <net/radix.h> 68#include <net/route.h> 69#include <net/pf_mtag.h> 70#include <netinet/in.h> 71#include <netinet/in_systm.h> 72#include <netinet/in_var.h> 73#include <netinet/in_pcb.h> 74#include <netinet/ip.h> 75#include <netinet/ip_var.h> 76#include <netinet/ip_icmp.h> 77#include <netinet/ip_fw.h> 78#include <netinet/ip_divert.h> 79#include <netinet/ip_dummynet.h> 80#include <netinet/ip_carp.h> 81#include <netinet/pim.h> 82#include <netinet/tcp.h> 83#include <netinet/tcp_timer.h> 84#include <netinet/tcp_var.h> 85#include <netinet/tcpip.h> 86#include <netinet/udp.h> 87#include <netinet/udp_var.h> 88#include <netinet/sctp.h> 89#ifdef IPFIREWALL_NAT 90#include <netinet/libalias/alias.h> 91#include <netinet/libalias/alias_local.h> 92#endif 93#include <netgraph/ng_ipfw.h> 94 95#include <altq/if_altq.h> 96 97#include <netinet/ip6.h> 98#include <netinet/icmp6.h> 99#ifdef INET6 100#include <netinet6/scope6_var.h> 101#endif 102 103#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 104 105#include <machine/in_cksum.h> /* XXX for in_cksum */ 106 107#include <security/mac/mac_framework.h> 108 109/* 110 * set_disable contains one bit per set value (0..31). 111 * If the bit is set, all rules with the corresponding set 112 * are disabled. Set RESVD_SET(31) is reserved for the default rule 113 * and rules that are not deleted by the flush command, 114 * and CANNOT be disabled. 115 * Rules in set RESVD_SET can only be deleted explicitly. 116 */ 117static u_int32_t set_disable; 118 119static int fw_verbose; 120static int verbose_limit; 121 122static struct callout ipfw_timeout; 123static uma_zone_t ipfw_dyn_rule_zone; 124#define IPFW_DEFAULT_RULE 65535 125 126/* 127 * Data structure to cache our ucred related 128 * information. This structure only gets used if 129 * the user specified UID/GID based constraints in 130 * a firewall rule. 131 */ 132struct ip_fw_ugid { 133 gid_t fw_groups[NGROUPS]; 134 int fw_ngroups; 135 uid_t fw_uid; 136 int fw_prid; 137}; 138 139#define IPFW_TABLES_MAX 128 140struct ip_fw_chain { 141 struct ip_fw *rules; /* list of rules */ 142 struct ip_fw *reap; /* list of rules to reap */ 143 LIST_HEAD(, cfg_nat) nat; /* list of nat entries */ 144 struct radix_node_head *tables[IPFW_TABLES_MAX]; 145 struct rwlock rwmtx; 146}; 147#define IPFW_LOCK_INIT(_chain) \ 148 rw_init(&(_chain)->rwmtx, "IPFW static rules") 149#define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx) 150#define IPFW_WLOCK_ASSERT(_chain) rw_assert(&(_chain)->rwmtx, RA_WLOCKED) 151 152#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 153#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 154#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 155#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 156 157/* 158 * list of rules for layer 3 159 */ 160static struct ip_fw_chain layer3_chain; 161 162MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 163MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 164 165struct table_entry { 166 struct radix_node rn[2]; 167 struct sockaddr_in addr, mask; 168 u_int32_t value; 169}; 170 171static int fw_debug = 1; 172static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 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_PROC(_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_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 182 &autoinc_step, 0, "Rule number autincrement step"); 183SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 184 CTLFLAG_RW | CTLFLAG_SECURE3, 185 &fw_one_pass, 0, 186 "Only do a single pass through ipfw when using dummynet(4)"); 187SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 188 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 189SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 190 CTLFLAG_RW | CTLFLAG_SECURE3, 191 &fw_verbose, 0, "Log matches to ipfw rules"); 192SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 193 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 194 195/* 196 * Description of dynamic rules. 197 * 198 * Dynamic rules are stored in lists accessed through a hash table 199 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 200 * be modified through the sysctl variable dyn_buckets which is 201 * updated when the table becomes empty. 202 * 203 * XXX currently there is only one list, ipfw_dyn. 204 * 205 * When a packet is received, its address fields are first masked 206 * with the mask defined for the rule, then hashed, then matched 207 * against the entries in the corresponding list. 208 * Dynamic rules can be used for different purposes: 209 * + stateful rules; 210 * + enforcing limits on the number of sessions; 211 * + in-kernel NAT (not implemented yet) 212 * 213 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 214 * measured in seconds and depending on the flags. 215 * 216 * The total number of dynamic rules is stored in dyn_count. 217 * The max number of dynamic rules is dyn_max. When we reach 218 * the maximum number of rules we do not create anymore. This is 219 * done to avoid consuming too much memory, but also too much 220 * time when searching on each packet (ideally, we should try instead 221 * to put a limit on the length of the list on each bucket...). 222 * 223 * Each dynamic rule holds a pointer to the parent ipfw rule so 224 * we know what action to perform. Dynamic rules are removed when 225 * the parent rule is deleted. XXX we should make them survive. 226 * 227 * There are some limitations with dynamic rules -- we do not 228 * obey the 'randomized match', and we do not do multiple 229 * passes through the firewall. XXX check the latter!!! 230 */ 231static ipfw_dyn_rule **ipfw_dyn_v = NULL; 232static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 233static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 234 235static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 236#define IPFW_DYN_LOCK_INIT() \ 237 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 238#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 239#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 240#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 241#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 242 243/* 244 * Timeouts for various events in handing dynamic rules. 245 */ 246static u_int32_t dyn_ack_lifetime = 300; 247static u_int32_t dyn_syn_lifetime = 20; 248static u_int32_t dyn_fin_lifetime = 1; 249static u_int32_t dyn_rst_lifetime = 1; 250static u_int32_t dyn_udp_lifetime = 10; 251static u_int32_t dyn_short_lifetime = 5; 252 253/* 254 * Keepalives are sent if dyn_keepalive is set. They are sent every 255 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 256 * seconds of lifetime of a rule. 257 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 258 * than dyn_keepalive_period. 259 */ 260 261static u_int32_t dyn_keepalive_interval = 20; 262static u_int32_t dyn_keepalive_period = 5; 263static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 264 265static u_int32_t static_count; /* # of static rules */ 266static u_int32_t static_len; /* size in bytes of static rules */ 267static u_int32_t dyn_count; /* # of dynamic rules */ 268static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 269 270SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 271 &dyn_buckets, 0, "Number of dyn. buckets"); 272SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 273 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 274SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 275 &dyn_count, 0, "Number of dyn. rules"); 276SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 277 &dyn_max, 0, "Max number of dyn. rules"); 278SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 279 &static_count, 0, "Number of static rules"); 280SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 281 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 282SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 283 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 284SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 285 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 286SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 287 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 288SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 289 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 290SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 291 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 292SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 293 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 294 295#ifdef INET6 296/* 297 * IPv6 specific variables 298 */ 299SYSCTL_DECL(_net_inet6_ip6); 300 301static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 302static struct sysctl_oid *ip6_fw_sysctl_tree; 303#endif /* INET6 */ 304#endif /* SYSCTL_NODE */ 305 306#ifdef IPFIREWALL_NAT 307MODULE_DEPEND(ipfw, libalias, 1, 1, 1); 308#endif 309static int fw_deny_unknown_exthdrs = 1; 310 311 312/* 313 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 314 * Other macros just cast void * into the appropriate type 315 */ 316#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 317#define TCP(p) ((struct tcphdr *)(p)) 318#define SCTP(p) ((struct sctphdr *)(p)) 319#define UDP(p) ((struct udphdr *)(p)) 320#define ICMP(p) ((struct icmphdr *)(p)) 321#define ICMP6(p) ((struct icmp6_hdr *)(p)) 322 323static __inline int 324icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 325{ 326 int type = icmp->icmp_type; 327 328 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 329} 330 331#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 332 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 333 334static int 335is_icmp_query(struct icmphdr *icmp) 336{ 337 int type = icmp->icmp_type; 338 339 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 340} 341#undef TT 342 343/* 344 * The following checks use two arrays of 8 or 16 bits to store the 345 * bits that we want set or clear, respectively. They are in the 346 * low and high half of cmd->arg1 or cmd->d[0]. 347 * 348 * We scan options and store the bits we find set. We succeed if 349 * 350 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 351 * 352 * The code is sometimes optimized not to store additional variables. 353 */ 354 355static int 356flags_match(ipfw_insn *cmd, u_int8_t bits) 357{ 358 u_char want_clear; 359 bits = ~bits; 360 361 if ( ((cmd->arg1 & 0xff) & bits) != 0) 362 return 0; /* some bits we want set were clear */ 363 want_clear = (cmd->arg1 >> 8) & 0xff; 364 if ( (want_clear & bits) != want_clear) 365 return 0; /* some bits we want clear were set */ 366 return 1; 367} 368 369static int 370ipopts_match(struct ip *ip, ipfw_insn *cmd) 371{ 372 int optlen, bits = 0; 373 u_char *cp = (u_char *)(ip + 1); 374 int x = (ip->ip_hl << 2) - sizeof (struct ip); 375 376 for (; x > 0; x -= optlen, cp += optlen) { 377 int opt = cp[IPOPT_OPTVAL]; 378 379 if (opt == IPOPT_EOL) 380 break; 381 if (opt == IPOPT_NOP) 382 optlen = 1; 383 else { 384 optlen = cp[IPOPT_OLEN]; 385 if (optlen <= 0 || optlen > x) 386 return 0; /* invalid or truncated */ 387 } 388 switch (opt) { 389 390 default: 391 break; 392 393 case IPOPT_LSRR: 394 bits |= IP_FW_IPOPT_LSRR; 395 break; 396 397 case IPOPT_SSRR: 398 bits |= IP_FW_IPOPT_SSRR; 399 break; 400 401 case IPOPT_RR: 402 bits |= IP_FW_IPOPT_RR; 403 break; 404 405 case IPOPT_TS: 406 bits |= IP_FW_IPOPT_TS; 407 break; 408 } 409 } 410 return (flags_match(cmd, bits)); 411} 412 413static int 414tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 415{ 416 int optlen, bits = 0; 417 u_char *cp = (u_char *)(tcp + 1); 418 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 419 420 for (; x > 0; x -= optlen, cp += optlen) { 421 int opt = cp[0]; 422 if (opt == TCPOPT_EOL) 423 break; 424 if (opt == TCPOPT_NOP) 425 optlen = 1; 426 else { 427 optlen = cp[1]; 428 if (optlen <= 0) 429 break; 430 } 431 432 switch (opt) { 433 434 default: 435 break; 436 437 case TCPOPT_MAXSEG: 438 bits |= IP_FW_TCPOPT_MSS; 439 break; 440 441 case TCPOPT_WINDOW: 442 bits |= IP_FW_TCPOPT_WINDOW; 443 break; 444 445 case TCPOPT_SACK_PERMITTED: 446 case TCPOPT_SACK: 447 bits |= IP_FW_TCPOPT_SACK; 448 break; 449 450 case TCPOPT_TIMESTAMP: 451 bits |= IP_FW_TCPOPT_TS; 452 break; 453 454 } 455 } 456 return (flags_match(cmd, bits)); 457} 458 459static int 460iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 461{ 462 if (ifp == NULL) /* no iface with this packet, match fails */ 463 return 0; 464 /* Check by name or by IP address */ 465 if (cmd->name[0] != '\0') { /* match by name */ 466 /* Check name */ 467 if (cmd->p.glob) { 468 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 469 return(1); 470 } else { 471 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 472 return(1); 473 } 474 } else { 475 struct ifaddr *ia; 476 477 /* XXX lock? */ 478 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 479 if (ia->ifa_addr->sa_family != AF_INET) 480 continue; 481 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 482 (ia->ifa_addr))->sin_addr.s_addr) 483 return(1); /* match */ 484 } 485 } 486 return(0); /* no match, fail ... */ 487} 488 489/* 490 * The verify_path function checks if a route to the src exists and 491 * if it is reachable via ifp (when provided). 492 * 493 * The 'verrevpath' option checks that the interface that an IP packet 494 * arrives on is the same interface that traffic destined for the 495 * packet's source address would be routed out of. The 'versrcreach' 496 * option just checks that the source address is reachable via any route 497 * (except default) in the routing table. These two are a measure to block 498 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 499 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 500 * is purposely reminiscent of the Cisco IOS command, 501 * 502 * ip verify unicast reverse-path 503 * ip verify unicast source reachable-via any 504 * 505 * which implements the same functionality. But note that syntax is 506 * misleading. The check may be performed on all IP packets whether unicast, 507 * multicast, or broadcast. 508 */ 509static int 510verify_path(struct in_addr src, struct ifnet *ifp) 511{ 512 struct route ro; 513 struct sockaddr_in *dst; 514 515 bzero(&ro, sizeof(ro)); 516 517 dst = (struct sockaddr_in *)&(ro.ro_dst); 518 dst->sin_family = AF_INET; 519 dst->sin_len = sizeof(*dst); 520 dst->sin_addr = src; 521 rtalloc_ign(&ro, RTF_CLONING); 522 523 if (ro.ro_rt == NULL) 524 return 0; 525 526 /* 527 * If ifp is provided, check for equality with rtentry. 528 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 529 * in order to pass packets injected back by if_simloop(): 530 * if useloopback == 1 routing entry (via lo0) for our own address 531 * may exist, so we need to handle routing assymetry. 532 */ 533 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 534 RTFREE(ro.ro_rt); 535 return 0; 536 } 537 538 /* if no ifp provided, check if rtentry is not default route */ 539 if (ifp == NULL && 540 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 541 RTFREE(ro.ro_rt); 542 return 0; 543 } 544 545 /* or if this is a blackhole/reject route */ 546 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 547 RTFREE(ro.ro_rt); 548 return 0; 549 } 550 551 /* found valid route */ 552 RTFREE(ro.ro_rt); 553 return 1; 554} 555 556#ifdef INET6 557/* 558 * ipv6 specific rules here... 559 */ 560static __inline int 561icmp6type_match (int type, ipfw_insn_u32 *cmd) 562{ 563 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 564} 565 566static int 567flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 568{ 569 int i; 570 for (i=0; i <= cmd->o.arg1; ++i ) 571 if (curr_flow == cmd->d[i] ) 572 return 1; 573 return 0; 574} 575 576/* support for IP6_*_ME opcodes */ 577static int 578search_ip6_addr_net (struct in6_addr * ip6_addr) 579{ 580 struct ifnet *mdc; 581 struct ifaddr *mdc2; 582 struct in6_ifaddr *fdm; 583 struct in6_addr copia; 584 585 TAILQ_FOREACH(mdc, &ifnet, if_link) 586 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 587 if (mdc2->ifa_addr->sa_family == AF_INET6) { 588 fdm = (struct in6_ifaddr *)mdc2; 589 copia = fdm->ia_addr.sin6_addr; 590 /* need for leaving scope_id in the sock_addr */ 591 in6_clearscope(&copia); 592 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 593 return 1; 594 } 595 } 596 return 0; 597} 598 599static int 600verify_path6(struct in6_addr *src, struct ifnet *ifp) 601{ 602 struct route_in6 ro; 603 struct sockaddr_in6 *dst; 604 605 bzero(&ro, sizeof(ro)); 606 607 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 608 dst->sin6_family = AF_INET6; 609 dst->sin6_len = sizeof(*dst); 610 dst->sin6_addr = *src; 611 rtalloc_ign((struct route *)&ro, RTF_CLONING); 612 613 if (ro.ro_rt == NULL) 614 return 0; 615 616 /* 617 * if ifp is provided, check for equality with rtentry 618 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 619 * to support the case of sending packets to an address of our own. 620 * (where the former interface is the first argument of if_simloop() 621 * (=ifp), the latter is lo0) 622 */ 623 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 624 RTFREE(ro.ro_rt); 625 return 0; 626 } 627 628 /* if no ifp provided, check if rtentry is not default route */ 629 if (ifp == NULL && 630 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 631 RTFREE(ro.ro_rt); 632 return 0; 633 } 634 635 /* or if this is a blackhole/reject route */ 636 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 637 RTFREE(ro.ro_rt); 638 return 0; 639 } 640 641 /* found valid route */ 642 RTFREE(ro.ro_rt); 643 return 1; 644 645} 646static __inline int 647hash_packet6(struct ipfw_flow_id *id) 648{ 649 u_int32_t i; 650 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 651 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 652 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 653 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 654 (id->dst_port) ^ (id->src_port); 655 return i; 656} 657 658static int 659is_icmp6_query(int icmp6_type) 660{ 661 if ((icmp6_type <= ICMP6_MAXTYPE) && 662 (icmp6_type == ICMP6_ECHO_REQUEST || 663 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 664 icmp6_type == ICMP6_WRUREQUEST || 665 icmp6_type == ICMP6_FQDN_QUERY || 666 icmp6_type == ICMP6_NI_QUERY)) 667 return (1); 668 669 return (0); 670} 671 672static void 673send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 674{ 675 struct mbuf *m; 676 677 m = args->m; 678 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 679 struct tcphdr *tcp; 680 tcp_seq ack, seq; 681 int flags; 682 struct { 683 struct ip6_hdr ip6; 684 struct tcphdr th; 685 } ti; 686 tcp = (struct tcphdr *)((char *)ip6 + hlen); 687 688 if ((tcp->th_flags & TH_RST) != 0) { 689 m_freem(m); 690 args->m = NULL; 691 return; 692 } 693 694 ti.ip6 = *ip6; 695 ti.th = *tcp; 696 ti.th.th_seq = ntohl(ti.th.th_seq); 697 ti.th.th_ack = ntohl(ti.th.th_ack); 698 ti.ip6.ip6_nxt = IPPROTO_TCP; 699 700 if (ti.th.th_flags & TH_ACK) { 701 ack = 0; 702 seq = ti.th.th_ack; 703 flags = TH_RST; 704 } else { 705 ack = ti.th.th_seq; 706 if ((m->m_flags & M_PKTHDR) != 0) { 707 /* 708 * total new data to ACK is: 709 * total packet length, 710 * minus the header length, 711 * minus the tcp header length. 712 */ 713 ack += m->m_pkthdr.len - hlen 714 - (ti.th.th_off << 2); 715 } else if (ip6->ip6_plen) { 716 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 717 hlen - (ti.th.th_off << 2); 718 } else { 719 m_freem(m); 720 return; 721 } 722 if (tcp->th_flags & TH_SYN) 723 ack++; 724 seq = 0; 725 flags = TH_RST|TH_ACK; 726 } 727 bcopy(&ti, ip6, sizeof(ti)); 728 /* 729 * m is only used to recycle the mbuf 730 * The data in it is never read so we don't need 731 * to correct the offsets or anything 732 */ 733 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 734 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 735#if 0 736 /* 737 * Unlike above, the mbufs need to line up with the ip6 hdr, 738 * as the contents are read. We need to m_adj() the 739 * needed amount. 740 * The mbuf will however be thrown away so we can adjust it. 741 * Remember we did an m_pullup on it already so we 742 * can make some assumptions about contiguousness. 743 */ 744 if (args->L3offset) 745 m_adj(m, args->L3offset); 746#endif 747 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 748 } else 749 m_freem(m); 750 751 args->m = NULL; 752} 753 754#endif /* INET6 */ 755 756static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 757 758#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 759#define SNP(buf) buf, sizeof(buf) 760 761/* 762 * We enter here when we have a rule with O_LOG. 763 * XXX this function alone takes about 2Kbytes of code! 764 */ 765static void 766ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 767 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 768 struct ip *ip) 769{ 770 struct ether_header *eh = args->eh; 771 char *action; 772 int limit_reached = 0; 773 char action2[40], proto[128], fragment[32]; 774 775 fragment[0] = '\0'; 776 proto[0] = '\0'; 777 778 if (f == NULL) { /* bogus pkt */ 779 if (verbose_limit != 0 && norule_counter >= verbose_limit) 780 return; 781 norule_counter++; 782 if (norule_counter == verbose_limit) 783 limit_reached = verbose_limit; 784 action = "Refuse"; 785 } else { /* O_LOG is the first action, find the real one */ 786 ipfw_insn *cmd = ACTION_PTR(f); 787 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 788 789 if (l->max_log != 0 && l->log_left == 0) 790 return; 791 l->log_left--; 792 if (l->log_left == 0) 793 limit_reached = l->max_log; 794 cmd += F_LEN(cmd); /* point to first action */ 795 if (cmd->opcode == O_ALTQ) { 796 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 797 798 snprintf(SNPARGS(action2, 0), "Altq %d", 799 altq->qid); 800 cmd += F_LEN(cmd); 801 } 802 if (cmd->opcode == O_PROB) 803 cmd += F_LEN(cmd); 804 805 if (cmd->opcode == O_TAG) 806 cmd += F_LEN(cmd); 807 808 action = action2; 809 switch (cmd->opcode) { 810 case O_DENY: 811 action = "Deny"; 812 break; 813 814 case O_REJECT: 815 if (cmd->arg1==ICMP_REJECT_RST) 816 action = "Reset"; 817 else if (cmd->arg1==ICMP_UNREACH_HOST) 818 action = "Reject"; 819 else 820 snprintf(SNPARGS(action2, 0), "Unreach %d", 821 cmd->arg1); 822 break; 823 824 case O_UNREACH6: 825 if (cmd->arg1==ICMP6_UNREACH_RST) 826 action = "Reset"; 827 else 828 snprintf(SNPARGS(action2, 0), "Unreach %d", 829 cmd->arg1); 830 break; 831 832 case O_ACCEPT: 833 action = "Accept"; 834 break; 835 case O_COUNT: 836 action = "Count"; 837 break; 838 case O_DIVERT: 839 snprintf(SNPARGS(action2, 0), "Divert %d", 840 cmd->arg1); 841 break; 842 case O_TEE: 843 snprintf(SNPARGS(action2, 0), "Tee %d", 844 cmd->arg1); 845 break; 846 case O_SKIPTO: 847 snprintf(SNPARGS(action2, 0), "SkipTo %d", 848 cmd->arg1); 849 break; 850 case O_PIPE: 851 snprintf(SNPARGS(action2, 0), "Pipe %d", 852 cmd->arg1); 853 break; 854 case O_QUEUE: 855 snprintf(SNPARGS(action2, 0), "Queue %d", 856 cmd->arg1); 857 break; 858 case O_FORWARD_IP: { 859 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 860 int len; 861 struct in_addr dummyaddr; 862 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 863 dummyaddr.s_addr = htonl(tablearg); 864 else 865 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 866 867 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 868 inet_ntoa(dummyaddr)); 869 870 if (sa->sa.sin_port) 871 snprintf(SNPARGS(action2, len), ":%d", 872 sa->sa.sin_port); 873 } 874 break; 875 case O_NETGRAPH: 876 snprintf(SNPARGS(action2, 0), "Netgraph %d", 877 cmd->arg1); 878 break; 879 case O_NGTEE: 880 snprintf(SNPARGS(action2, 0), "Ngtee %d", 881 cmd->arg1); 882 break; 883 case O_NAT: 884 action = "Nat"; 885 break; 886 default: 887 action = "UNKNOWN"; 888 break; 889 } 890 } 891 892 if (hlen == 0) { /* non-ip */ 893 snprintf(SNPARGS(proto, 0), "MAC"); 894 895 } else { 896 int len; 897 char src[48], dst[48]; 898 struct icmphdr *icmp; 899 struct tcphdr *tcp; 900 struct udphdr *udp; 901#ifdef INET6 902 struct ip6_hdr *ip6 = NULL; 903 struct icmp6_hdr *icmp6; 904#endif 905 src[0] = '\0'; 906 dst[0] = '\0'; 907#ifdef INET6 908 if (IS_IP6_FLOW_ID(&(args->f_id))) { 909 char ip6buf[INET6_ADDRSTRLEN]; 910 snprintf(src, sizeof(src), "[%s]", 911 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 912 snprintf(dst, sizeof(dst), "[%s]", 913 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 914 915 ip6 = (struct ip6_hdr *)ip; 916 tcp = (struct tcphdr *)(((char *)ip) + hlen); 917 udp = (struct udphdr *)(((char *)ip) + hlen); 918 } else 919#endif 920 { 921 tcp = L3HDR(struct tcphdr, ip); 922 udp = L3HDR(struct udphdr, ip); 923 924 inet_ntoa_r(ip->ip_src, src); 925 inet_ntoa_r(ip->ip_dst, dst); 926 } 927 928 switch (args->f_id.proto) { 929 case IPPROTO_TCP: 930 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 931 if (offset == 0) 932 snprintf(SNPARGS(proto, len), ":%d %s:%d", 933 ntohs(tcp->th_sport), 934 dst, 935 ntohs(tcp->th_dport)); 936 else 937 snprintf(SNPARGS(proto, len), " %s", dst); 938 break; 939 940 case IPPROTO_UDP: 941 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 942 if (offset == 0) 943 snprintf(SNPARGS(proto, len), ":%d %s:%d", 944 ntohs(udp->uh_sport), 945 dst, 946 ntohs(udp->uh_dport)); 947 else 948 snprintf(SNPARGS(proto, len), " %s", dst); 949 break; 950 951 case IPPROTO_ICMP: 952 icmp = L3HDR(struct icmphdr, ip); 953 if (offset == 0) 954 len = snprintf(SNPARGS(proto, 0), 955 "ICMP:%u.%u ", 956 icmp->icmp_type, icmp->icmp_code); 957 else 958 len = snprintf(SNPARGS(proto, 0), "ICMP "); 959 len += snprintf(SNPARGS(proto, len), "%s", src); 960 snprintf(SNPARGS(proto, len), " %s", dst); 961 break; 962#ifdef INET6 963 case IPPROTO_ICMPV6: 964 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 965 if (offset == 0) 966 len = snprintf(SNPARGS(proto, 0), 967 "ICMPv6:%u.%u ", 968 icmp6->icmp6_type, icmp6->icmp6_code); 969 else 970 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 971 len += snprintf(SNPARGS(proto, len), "%s", src); 972 snprintf(SNPARGS(proto, len), " %s", dst); 973 break; 974#endif 975 default: 976 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 977 args->f_id.proto, src); 978 snprintf(SNPARGS(proto, len), " %s", dst); 979 break; 980 } 981 982#ifdef INET6 983 if (IS_IP6_FLOW_ID(&(args->f_id))) { 984 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 985 snprintf(SNPARGS(fragment, 0), 986 " (frag %08x:%d@%d%s)", 987 args->f_id.frag_id6, 988 ntohs(ip6->ip6_plen) - hlen, 989 ntohs(offset & IP6F_OFF_MASK) << 3, 990 (offset & IP6F_MORE_FRAG) ? "+" : ""); 991 } else 992#endif 993 { 994 int ip_off, ip_len; 995 if (eh != NULL) { /* layer 2 packets are as on the wire */ 996 ip_off = ntohs(ip->ip_off); 997 ip_len = ntohs(ip->ip_len); 998 } else { 999 ip_off = ip->ip_off; 1000 ip_len = ip->ip_len; 1001 } 1002 if (ip_off & (IP_MF | IP_OFFMASK)) 1003 snprintf(SNPARGS(fragment, 0), 1004 " (frag %d:%d@%d%s)", 1005 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1006 offset << 3, 1007 (ip_off & IP_MF) ? "+" : ""); 1008 } 1009 } 1010 if (oif || m->m_pkthdr.rcvif) 1011 log(LOG_SECURITY | LOG_INFO, 1012 "ipfw: %d %s %s %s via %s%s\n", 1013 f ? f->rulenum : -1, 1014 action, proto, oif ? "out" : "in", 1015 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1016 fragment); 1017 else 1018 log(LOG_SECURITY | LOG_INFO, 1019 "ipfw: %d %s %s [no if info]%s\n", 1020 f ? f->rulenum : -1, 1021 action, proto, fragment); 1022 if (limit_reached) 1023 log(LOG_SECURITY | LOG_NOTICE, 1024 "ipfw: limit %d reached on entry %d\n", 1025 limit_reached, f ? f->rulenum : -1); 1026} 1027 1028/* 1029 * IMPORTANT: the hash function for dynamic rules must be commutative 1030 * in source and destination (ip,port), because rules are bidirectional 1031 * and we want to find both in the same bucket. 1032 */ 1033static __inline int 1034hash_packet(struct ipfw_flow_id *id) 1035{ 1036 u_int32_t i; 1037 1038#ifdef INET6 1039 if (IS_IP6_FLOW_ID(id)) 1040 i = hash_packet6(id); 1041 else 1042#endif /* INET6 */ 1043 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1044 i &= (curr_dyn_buckets - 1); 1045 return i; 1046} 1047 1048/** 1049 * unlink a dynamic rule from a chain. prev is a pointer to 1050 * the previous one, q is a pointer to the rule to delete, 1051 * head is a pointer to the head of the queue. 1052 * Modifies q and potentially also head. 1053 */ 1054#define UNLINK_DYN_RULE(prev, head, q) { \ 1055 ipfw_dyn_rule *old_q = q; \ 1056 \ 1057 /* remove a refcount to the parent */ \ 1058 if (q->dyn_type == O_LIMIT) \ 1059 q->parent->count--; \ 1060 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1061 (q->id.src_ip), (q->id.src_port), \ 1062 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1063 if (prev != NULL) \ 1064 prev->next = q = q->next; \ 1065 else \ 1066 head = q = q->next; \ 1067 dyn_count--; \ 1068 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1069 1070#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1071 1072/** 1073 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1074 * 1075 * If keep_me == NULL, rules are deleted even if not expired, 1076 * otherwise only expired rules are removed. 1077 * 1078 * The value of the second parameter is also used to point to identify 1079 * a rule we absolutely do not want to remove (e.g. because we are 1080 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1081 * rules). The pointer is only used for comparison, so any non-null 1082 * value will do. 1083 */ 1084static void 1085remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1086{ 1087 static u_int32_t last_remove = 0; 1088 1089#define FORCE (keep_me == NULL) 1090 1091 ipfw_dyn_rule *prev, *q; 1092 int i, pass = 0, max_pass = 0; 1093 1094 IPFW_DYN_LOCK_ASSERT(); 1095 1096 if (ipfw_dyn_v == NULL || dyn_count == 0) 1097 return; 1098 /* do not expire more than once per second, it is useless */ 1099 if (!FORCE && last_remove == time_uptime) 1100 return; 1101 last_remove = time_uptime; 1102 1103 /* 1104 * because O_LIMIT refer to parent rules, during the first pass only 1105 * remove child and mark any pending LIMIT_PARENT, and remove 1106 * them in a second pass. 1107 */ 1108next_pass: 1109 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1110 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1111 /* 1112 * Logic can become complex here, so we split tests. 1113 */ 1114 if (q == keep_me) 1115 goto next; 1116 if (rule != NULL && rule != q->rule) 1117 goto next; /* not the one we are looking for */ 1118 if (q->dyn_type == O_LIMIT_PARENT) { 1119 /* 1120 * handle parent in the second pass, 1121 * record we need one. 1122 */ 1123 max_pass = 1; 1124 if (pass == 0) 1125 goto next; 1126 if (FORCE && q->count != 0 ) { 1127 /* XXX should not happen! */ 1128 printf("ipfw: OUCH! cannot remove rule," 1129 " count %d\n", q->count); 1130 } 1131 } else { 1132 if (!FORCE && 1133 !TIME_LEQ( q->expire, time_uptime )) 1134 goto next; 1135 } 1136 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1137 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1138 continue; 1139 } 1140next: 1141 prev=q; 1142 q=q->next; 1143 } 1144 } 1145 if (pass++ < max_pass) 1146 goto next_pass; 1147} 1148 1149 1150/** 1151 * lookup a dynamic rule. 1152 */ 1153static ipfw_dyn_rule * 1154lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1155 struct tcphdr *tcp) 1156{ 1157 /* 1158 * stateful ipfw extensions. 1159 * Lookup into dynamic session queue 1160 */ 1161#define MATCH_REVERSE 0 1162#define MATCH_FORWARD 1 1163#define MATCH_NONE 2 1164#define MATCH_UNKNOWN 3 1165 int i, dir = MATCH_NONE; 1166 ipfw_dyn_rule *prev, *q=NULL; 1167 1168 IPFW_DYN_LOCK_ASSERT(); 1169 1170 if (ipfw_dyn_v == NULL) 1171 goto done; /* not found */ 1172 i = hash_packet( pkt ); 1173 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1174 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1175 goto next; 1176 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1177 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1178 continue; 1179 } 1180 if (pkt->proto == q->id.proto && 1181 q->dyn_type != O_LIMIT_PARENT) { 1182 if (IS_IP6_FLOW_ID(pkt)) { 1183 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1184 &(q->id.src_ip6)) && 1185 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1186 &(q->id.dst_ip6)) && 1187 pkt->src_port == q->id.src_port && 1188 pkt->dst_port == q->id.dst_port ) { 1189 dir = MATCH_FORWARD; 1190 break; 1191 } 1192 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1193 &(q->id.dst_ip6)) && 1194 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1195 &(q->id.src_ip6)) && 1196 pkt->src_port == q->id.dst_port && 1197 pkt->dst_port == q->id.src_port ) { 1198 dir = MATCH_REVERSE; 1199 break; 1200 } 1201 } else { 1202 if (pkt->src_ip == q->id.src_ip && 1203 pkt->dst_ip == q->id.dst_ip && 1204 pkt->src_port == q->id.src_port && 1205 pkt->dst_port == q->id.dst_port ) { 1206 dir = MATCH_FORWARD; 1207 break; 1208 } 1209 if (pkt->src_ip == q->id.dst_ip && 1210 pkt->dst_ip == q->id.src_ip && 1211 pkt->src_port == q->id.dst_port && 1212 pkt->dst_port == q->id.src_port ) { 1213 dir = MATCH_REVERSE; 1214 break; 1215 } 1216 } 1217 } 1218next: 1219 prev = q; 1220 q = q->next; 1221 } 1222 if (q == NULL) 1223 goto done; /* q = NULL, not found */ 1224 1225 if ( prev != NULL) { /* found and not in front */ 1226 prev->next = q->next; 1227 q->next = ipfw_dyn_v[i]; 1228 ipfw_dyn_v[i] = q; 1229 } 1230 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1231 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1232 1233#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1234#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1235 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1236 switch (q->state) { 1237 case TH_SYN: /* opening */ 1238 q->expire = time_uptime + dyn_syn_lifetime; 1239 break; 1240 1241 case BOTH_SYN: /* move to established */ 1242 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1243 case BOTH_SYN | (TH_FIN << 8) : 1244 if (tcp) { 1245#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1246 u_int32_t ack = ntohl(tcp->th_ack); 1247 if (dir == MATCH_FORWARD) { 1248 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1249 q->ack_fwd = ack; 1250 else { /* ignore out-of-sequence */ 1251 break; 1252 } 1253 } else { 1254 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1255 q->ack_rev = ack; 1256 else { /* ignore out-of-sequence */ 1257 break; 1258 } 1259 } 1260 } 1261 q->expire = time_uptime + dyn_ack_lifetime; 1262 break; 1263 1264 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1265 if (dyn_fin_lifetime >= dyn_keepalive_period) 1266 dyn_fin_lifetime = dyn_keepalive_period - 1; 1267 q->expire = time_uptime + dyn_fin_lifetime; 1268 break; 1269 1270 default: 1271#if 0 1272 /* 1273 * reset or some invalid combination, but can also 1274 * occur if we use keep-state the wrong way. 1275 */ 1276 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1277 printf("invalid state: 0x%x\n", q->state); 1278#endif 1279 if (dyn_rst_lifetime >= dyn_keepalive_period) 1280 dyn_rst_lifetime = dyn_keepalive_period - 1; 1281 q->expire = time_uptime + dyn_rst_lifetime; 1282 break; 1283 } 1284 } else if (pkt->proto == IPPROTO_UDP) { 1285 q->expire = time_uptime + dyn_udp_lifetime; 1286 } else { 1287 /* other protocols */ 1288 q->expire = time_uptime + dyn_short_lifetime; 1289 } 1290done: 1291 if (match_direction) 1292 *match_direction = dir; 1293 return q; 1294} 1295 1296static ipfw_dyn_rule * 1297lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1298 struct tcphdr *tcp) 1299{ 1300 ipfw_dyn_rule *q; 1301 1302 IPFW_DYN_LOCK(); 1303 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1304 if (q == NULL) 1305 IPFW_DYN_UNLOCK(); 1306 /* NB: return table locked when q is not NULL */ 1307 return q; 1308} 1309 1310static void 1311realloc_dynamic_table(void) 1312{ 1313 IPFW_DYN_LOCK_ASSERT(); 1314 1315 /* 1316 * Try reallocation, make sure we have a power of 2 and do 1317 * not allow more than 64k entries. In case of overflow, 1318 * default to 1024. 1319 */ 1320 1321 if (dyn_buckets > 65536) 1322 dyn_buckets = 1024; 1323 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1324 dyn_buckets = curr_dyn_buckets; /* reset */ 1325 return; 1326 } 1327 curr_dyn_buckets = dyn_buckets; 1328 if (ipfw_dyn_v != NULL) 1329 free(ipfw_dyn_v, M_IPFW); 1330 for (;;) { 1331 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1332 M_IPFW, M_NOWAIT | M_ZERO); 1333 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1334 break; 1335 curr_dyn_buckets /= 2; 1336 } 1337} 1338 1339/** 1340 * Install state of type 'type' for a dynamic session. 1341 * The hash table contains two type of rules: 1342 * - regular rules (O_KEEP_STATE) 1343 * - rules for sessions with limited number of sess per user 1344 * (O_LIMIT). When they are created, the parent is 1345 * increased by 1, and decreased on delete. In this case, 1346 * the third parameter is the parent rule and not the chain. 1347 * - "parent" rules for the above (O_LIMIT_PARENT). 1348 */ 1349static ipfw_dyn_rule * 1350add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1351{ 1352 ipfw_dyn_rule *r; 1353 int i; 1354 1355 IPFW_DYN_LOCK_ASSERT(); 1356 1357 if (ipfw_dyn_v == NULL || 1358 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1359 realloc_dynamic_table(); 1360 if (ipfw_dyn_v == NULL) 1361 return NULL; /* failed ! */ 1362 } 1363 i = hash_packet(id); 1364 1365 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1366 if (r == NULL) { 1367 printf ("ipfw: sorry cannot allocate state\n"); 1368 return NULL; 1369 } 1370 1371 /* increase refcount on parent, and set pointer */ 1372 if (dyn_type == O_LIMIT) { 1373 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1374 if ( parent->dyn_type != O_LIMIT_PARENT) 1375 panic("invalid parent"); 1376 parent->count++; 1377 r->parent = parent; 1378 rule = parent->rule; 1379 } 1380 1381 r->id = *id; 1382 r->expire = time_uptime + dyn_syn_lifetime; 1383 r->rule = rule; 1384 r->dyn_type = dyn_type; 1385 r->pcnt = r->bcnt = 0; 1386 r->count = 0; 1387 1388 r->bucket = i; 1389 r->next = ipfw_dyn_v[i]; 1390 ipfw_dyn_v[i] = r; 1391 dyn_count++; 1392 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1393 dyn_type, 1394 (r->id.src_ip), (r->id.src_port), 1395 (r->id.dst_ip), (r->id.dst_port), 1396 dyn_count ); ) 1397 return r; 1398} 1399 1400/** 1401 * lookup dynamic parent rule using pkt and rule as search keys. 1402 * If the lookup fails, then install one. 1403 */ 1404static ipfw_dyn_rule * 1405lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1406{ 1407 ipfw_dyn_rule *q; 1408 int i; 1409 1410 IPFW_DYN_LOCK_ASSERT(); 1411 1412 if (ipfw_dyn_v) { 1413 int is_v6 = IS_IP6_FLOW_ID(pkt); 1414 i = hash_packet( pkt ); 1415 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1416 if (q->dyn_type == O_LIMIT_PARENT && 1417 rule== q->rule && 1418 pkt->proto == q->id.proto && 1419 pkt->src_port == q->id.src_port && 1420 pkt->dst_port == q->id.dst_port && 1421 ( 1422 (is_v6 && 1423 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1424 &(q->id.src_ip6)) && 1425 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1426 &(q->id.dst_ip6))) || 1427 (!is_v6 && 1428 pkt->src_ip == q->id.src_ip && 1429 pkt->dst_ip == q->id.dst_ip) 1430 ) 1431 ) { 1432 q->expire = time_uptime + dyn_short_lifetime; 1433 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1434 return q; 1435 } 1436 } 1437 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1438} 1439 1440/** 1441 * Install dynamic state for rule type cmd->o.opcode 1442 * 1443 * Returns 1 (failure) if state is not installed because of errors or because 1444 * session limitations are enforced. 1445 */ 1446static int 1447install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1448 struct ip_fw_args *args, uint32_t tablearg) 1449{ 1450 static int last_log; 1451 ipfw_dyn_rule *q; 1452 struct in_addr da; 1453 char src[48], dst[48]; 1454 1455 src[0] = '\0'; 1456 dst[0] = '\0'; 1457 1458 DEB( 1459 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1460 __func__, cmd->o.opcode, 1461 (args->f_id.src_ip), (args->f_id.src_port), 1462 (args->f_id.dst_ip), (args->f_id.dst_port)); 1463 ) 1464 1465 IPFW_DYN_LOCK(); 1466 1467 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1468 1469 if (q != NULL) { /* should never occur */ 1470 if (last_log != time_uptime) { 1471 last_log = time_uptime; 1472 printf("ipfw: %s: entry already present, done\n", 1473 __func__); 1474 } 1475 IPFW_DYN_UNLOCK(); 1476 return (0); 1477 } 1478 1479 if (dyn_count >= dyn_max) 1480 /* Run out of slots, try to remove any expired rule. */ 1481 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1482 1483 if (dyn_count >= dyn_max) { 1484 if (last_log != time_uptime) { 1485 last_log = time_uptime; 1486 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1487 } 1488 IPFW_DYN_UNLOCK(); 1489 return (1); /* cannot install, notify caller */ 1490 } 1491 1492 switch (cmd->o.opcode) { 1493 case O_KEEP_STATE: /* bidir rule */ 1494 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1495 break; 1496 1497 case O_LIMIT: { /* limit number of sessions */ 1498 struct ipfw_flow_id id; 1499 ipfw_dyn_rule *parent; 1500 uint32_t conn_limit; 1501 uint16_t limit_mask = cmd->limit_mask; 1502 1503 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1504 tablearg : cmd->conn_limit; 1505 1506 DEB( 1507 if (cmd->conn_limit == IP_FW_TABLEARG) 1508 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1509 "(tablearg)\n", __func__, conn_limit); 1510 else 1511 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1512 __func__, conn_limit); 1513 ) 1514 1515 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1516 id.proto = args->f_id.proto; 1517 id.addr_type = args->f_id.addr_type; 1518 1519 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1520 if (limit_mask & DYN_SRC_ADDR) 1521 id.src_ip6 = args->f_id.src_ip6; 1522 if (limit_mask & DYN_DST_ADDR) 1523 id.dst_ip6 = args->f_id.dst_ip6; 1524 } else { 1525 if (limit_mask & DYN_SRC_ADDR) 1526 id.src_ip = args->f_id.src_ip; 1527 if (limit_mask & DYN_DST_ADDR) 1528 id.dst_ip = args->f_id.dst_ip; 1529 } 1530 if (limit_mask & DYN_SRC_PORT) 1531 id.src_port = args->f_id.src_port; 1532 if (limit_mask & DYN_DST_PORT) 1533 id.dst_port = args->f_id.dst_port; 1534 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1535 printf("ipfw: %s: add parent failed\n", __func__); 1536 IPFW_DYN_UNLOCK(); 1537 return (1); 1538 } 1539 1540 if (parent->count >= conn_limit) { 1541 /* See if we can remove some expired rule. */ 1542 remove_dyn_rule(rule, parent); 1543 if (parent->count >= conn_limit) { 1544 if (fw_verbose && last_log != time_uptime) { 1545 last_log = time_uptime; 1546#ifdef INET6 1547 /* 1548 * XXX IPv6 flows are not 1549 * supported yet. 1550 */ 1551 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1552 char ip6buf[INET6_ADDRSTRLEN]; 1553 snprintf(src, sizeof(src), 1554 "[%s]", ip6_sprintf(ip6buf, 1555 &args->f_id.src_ip6)); 1556 snprintf(dst, sizeof(dst), 1557 "[%s]", ip6_sprintf(ip6buf, 1558 &args->f_id.dst_ip6)); 1559 } else 1560#endif 1561 { 1562 da.s_addr = 1563 htonl(args->f_id.src_ip); 1564 inet_ntoa_r(da, src); 1565 da.s_addr = 1566 htonl(args->f_id.dst_ip); 1567 inet_ntoa_r(da, dst); 1568 } 1569 log(LOG_SECURITY | LOG_DEBUG, 1570 "ipfw: %d %s %s:%u -> %s:%u, %s\n", 1571 parent->rule->rulenum, 1572 "drop session", 1573 src, (args->f_id.src_port), 1574 dst, (args->f_id.dst_port), 1575 "too many entries"); 1576 } 1577 IPFW_DYN_UNLOCK(); 1578 return (1); 1579 } 1580 } 1581 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1582 break; 1583 } 1584 default: 1585 printf("ipfw: %s: unknown dynamic rule type %u\n", 1586 __func__, cmd->o.opcode); 1587 IPFW_DYN_UNLOCK(); 1588 return (1); 1589 } 1590 1591 /* XXX just set lifetime */ 1592 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1593 1594 IPFW_DYN_UNLOCK(); 1595 return (0); 1596} 1597 1598/* 1599 * Generate a TCP packet, containing either a RST or a keepalive. 1600 * When flags & TH_RST, we are sending a RST packet, because of a 1601 * "reset" action matched the packet. 1602 * Otherwise we are sending a keepalive, and flags & TH_ 1603 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1604 * so that MAC can label the reply appropriately. 1605 */ 1606static struct mbuf * 1607send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1608 u_int32_t ack, int flags) 1609{ 1610 struct mbuf *m; 1611 struct ip *ip; 1612 struct tcphdr *tcp; 1613 1614 MGETHDR(m, M_DONTWAIT, MT_DATA); 1615 if (m == 0) 1616 return (NULL); 1617 m->m_pkthdr.rcvif = (struct ifnet *)0; 1618 1619#ifdef MAC 1620 if (replyto != NULL) 1621 mac_create_mbuf_netlayer(replyto, m); 1622 else 1623 mac_create_mbuf_from_firewall(m); 1624#else 1625 (void)replyto; /* don't warn about unused arg */ 1626#endif 1627 1628 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1629 m->m_data += max_linkhdr; 1630 1631 ip = mtod(m, struct ip *); 1632 bzero(ip, m->m_len); 1633 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1634 ip->ip_p = IPPROTO_TCP; 1635 tcp->th_off = 5; 1636 /* 1637 * Assume we are sending a RST (or a keepalive in the reverse 1638 * direction), swap src and destination addresses and ports. 1639 */ 1640 ip->ip_src.s_addr = htonl(id->dst_ip); 1641 ip->ip_dst.s_addr = htonl(id->src_ip); 1642 tcp->th_sport = htons(id->dst_port); 1643 tcp->th_dport = htons(id->src_port); 1644 if (flags & TH_RST) { /* we are sending a RST */ 1645 if (flags & TH_ACK) { 1646 tcp->th_seq = htonl(ack); 1647 tcp->th_ack = htonl(0); 1648 tcp->th_flags = TH_RST; 1649 } else { 1650 if (flags & TH_SYN) 1651 seq++; 1652 tcp->th_seq = htonl(0); 1653 tcp->th_ack = htonl(seq); 1654 tcp->th_flags = TH_RST | TH_ACK; 1655 } 1656 } else { 1657 /* 1658 * We are sending a keepalive. flags & TH_SYN determines 1659 * the direction, forward if set, reverse if clear. 1660 * NOTE: seq and ack are always assumed to be correct 1661 * as set by the caller. This may be confusing... 1662 */ 1663 if (flags & TH_SYN) { 1664 /* 1665 * we have to rewrite the correct addresses! 1666 */ 1667 ip->ip_dst.s_addr = htonl(id->dst_ip); 1668 ip->ip_src.s_addr = htonl(id->src_ip); 1669 tcp->th_dport = htons(id->dst_port); 1670 tcp->th_sport = htons(id->src_port); 1671 } 1672 tcp->th_seq = htonl(seq); 1673 tcp->th_ack = htonl(ack); 1674 tcp->th_flags = TH_ACK; 1675 } 1676 /* 1677 * set ip_len to the payload size so we can compute 1678 * the tcp checksum on the pseudoheader 1679 * XXX check this, could save a couple of words ? 1680 */ 1681 ip->ip_len = htons(sizeof(struct tcphdr)); 1682 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1683 /* 1684 * now fill fields left out earlier 1685 */ 1686 ip->ip_ttl = ip_defttl; 1687 ip->ip_len = m->m_pkthdr.len; 1688 m->m_flags |= M_SKIP_FIREWALL; 1689 return (m); 1690} 1691 1692/* 1693 * sends a reject message, consuming the mbuf passed as an argument. 1694 */ 1695static void 1696send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1697{ 1698 1699#if 0 1700 /* XXX When ip is not guaranteed to be at mtod() we will 1701 * need to account for this */ 1702 * The mbuf will however be thrown away so we can adjust it. 1703 * Remember we did an m_pullup on it already so we 1704 * can make some assumptions about contiguousness. 1705 */ 1706 if (args->L3offset) 1707 m_adj(m, args->L3offset); 1708#endif 1709 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1710 /* We need the IP header in host order for icmp_error(). */ 1711 if (args->eh != NULL) { 1712 ip->ip_len = ntohs(ip->ip_len); 1713 ip->ip_off = ntohs(ip->ip_off); 1714 } 1715 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1716 } else if (args->f_id.proto == IPPROTO_TCP) { 1717 struct tcphdr *const tcp = 1718 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1719 if ( (tcp->th_flags & TH_RST) == 0) { 1720 struct mbuf *m; 1721 m = send_pkt(args->m, &(args->f_id), 1722 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1723 tcp->th_flags | TH_RST); 1724 if (m != NULL) 1725 ip_output(m, NULL, NULL, 0, NULL, NULL); 1726 } 1727 m_freem(args->m); 1728 } else 1729 m_freem(args->m); 1730 args->m = NULL; 1731} 1732 1733/** 1734 * 1735 * Given an ip_fw *, lookup_next_rule will return a pointer 1736 * to the next rule, which can be either the jump 1737 * target (for skipto instructions) or the next one in the list (in 1738 * all other cases including a missing jump target). 1739 * The result is also written in the "next_rule" field of the rule. 1740 * Backward jumps are not allowed, so start looking from the next 1741 * rule... 1742 * 1743 * This never returns NULL -- in case we do not have an exact match, 1744 * the next rule is returned. When the ruleset is changed, 1745 * pointers are flushed so we are always correct. 1746 */ 1747 1748static struct ip_fw * 1749lookup_next_rule(struct ip_fw *me) 1750{ 1751 struct ip_fw *rule = NULL; 1752 ipfw_insn *cmd; 1753 1754 /* look for action, in case it is a skipto */ 1755 cmd = ACTION_PTR(me); 1756 if (cmd->opcode == O_LOG) 1757 cmd += F_LEN(cmd); 1758 if (cmd->opcode == O_ALTQ) 1759 cmd += F_LEN(cmd); 1760 if (cmd->opcode == O_TAG) 1761 cmd += F_LEN(cmd); 1762 if ( cmd->opcode == O_SKIPTO ) 1763 for (rule = me->next; rule ; rule = rule->next) 1764 if (rule->rulenum >= cmd->arg1) 1765 break; 1766 if (rule == NULL) /* failure or not a skipto */ 1767 rule = me->next; 1768 me->next_rule = rule; 1769 return rule; 1770} 1771 1772static int 1773add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1774 uint8_t mlen, uint32_t value) 1775{ 1776 struct radix_node_head *rnh; 1777 struct table_entry *ent; 1778 1779 if (tbl >= IPFW_TABLES_MAX) 1780 return (EINVAL); 1781 rnh = ch->tables[tbl]; 1782 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1783 if (ent == NULL) 1784 return (ENOMEM); 1785 ent->value = value; 1786 ent->addr.sin_len = ent->mask.sin_len = 8; 1787 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1788 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1789 IPFW_WLOCK(&layer3_chain); 1790 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1791 NULL) { 1792 IPFW_WUNLOCK(&layer3_chain); 1793 free(ent, M_IPFW_TBL); 1794 return (EEXIST); 1795 } 1796 IPFW_WUNLOCK(&layer3_chain); 1797 return (0); 1798} 1799 1800static int 1801del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1802 uint8_t mlen) 1803{ 1804 struct radix_node_head *rnh; 1805 struct table_entry *ent; 1806 struct sockaddr_in sa, mask; 1807 1808 if (tbl >= IPFW_TABLES_MAX) 1809 return (EINVAL); 1810 rnh = ch->tables[tbl]; 1811 sa.sin_len = mask.sin_len = 8; 1812 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1813 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1814 IPFW_WLOCK(ch); 1815 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1816 if (ent == NULL) { 1817 IPFW_WUNLOCK(ch); 1818 return (ESRCH); 1819 } 1820 IPFW_WUNLOCK(ch); 1821 free(ent, M_IPFW_TBL); 1822 return (0); 1823} 1824 1825static int 1826flush_table_entry(struct radix_node *rn, void *arg) 1827{ 1828 struct radix_node_head * const rnh = arg; 1829 struct table_entry *ent; 1830 1831 ent = (struct table_entry *) 1832 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1833 if (ent != NULL) 1834 free(ent, M_IPFW_TBL); 1835 return (0); 1836} 1837 1838static int 1839flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1840{ 1841 struct radix_node_head *rnh; 1842 1843 IPFW_WLOCK_ASSERT(ch); 1844 1845 if (tbl >= IPFW_TABLES_MAX) 1846 return (EINVAL); 1847 rnh = ch->tables[tbl]; 1848 KASSERT(rnh != NULL, ("NULL IPFW table")); 1849 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1850 return (0); 1851} 1852 1853static void 1854flush_tables(struct ip_fw_chain *ch) 1855{ 1856 uint16_t tbl; 1857 1858 IPFW_WLOCK_ASSERT(ch); 1859 1860 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1861 flush_table(ch, tbl); 1862} 1863 1864static int 1865init_tables(struct ip_fw_chain *ch) 1866{ 1867 int i; 1868 uint16_t j; 1869 1870 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1871 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1872 for (j = 0; j < i; j++) { 1873 (void) flush_table(ch, j); 1874 } 1875 return (ENOMEM); 1876 } 1877 } 1878 return (0); 1879} 1880 1881static int 1882lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1883 uint32_t *val) 1884{ 1885 struct radix_node_head *rnh; 1886 struct table_entry *ent; 1887 struct sockaddr_in sa; 1888 1889 if (tbl >= IPFW_TABLES_MAX) 1890 return (0); 1891 rnh = ch->tables[tbl]; 1892 sa.sin_len = 8; 1893 sa.sin_addr.s_addr = addr; 1894 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1895 if (ent != NULL) { 1896 *val = ent->value; 1897 return (1); 1898 } 1899 return (0); 1900} 1901 1902static int 1903count_table_entry(struct radix_node *rn, void *arg) 1904{ 1905 u_int32_t * const cnt = arg; 1906 1907 (*cnt)++; 1908 return (0); 1909} 1910 1911static int 1912count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1913{ 1914 struct radix_node_head *rnh; 1915 1916 if (tbl >= IPFW_TABLES_MAX) 1917 return (EINVAL); 1918 rnh = ch->tables[tbl]; 1919 *cnt = 0; 1920 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1921 return (0); 1922} 1923 1924static int 1925dump_table_entry(struct radix_node *rn, void *arg) 1926{ 1927 struct table_entry * const n = (struct table_entry *)rn; 1928 ipfw_table * const tbl = arg; 1929 ipfw_table_entry *ent; 1930 1931 if (tbl->cnt == tbl->size) 1932 return (1); 1933 ent = &tbl->ent[tbl->cnt]; 1934 ent->tbl = tbl->tbl; 1935 if (in_nullhost(n->mask.sin_addr)) 1936 ent->masklen = 0; 1937 else 1938 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1939 ent->addr = n->addr.sin_addr.s_addr; 1940 ent->value = n->value; 1941 tbl->cnt++; 1942 return (0); 1943} 1944 1945static int 1946dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1947{ 1948 struct radix_node_head *rnh; 1949 1950 if (tbl->tbl >= IPFW_TABLES_MAX) 1951 return (EINVAL); 1952 rnh = ch->tables[tbl->tbl]; 1953 tbl->cnt = 0; 1954 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1955 return (0); 1956} 1957 1958static void 1959fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1960{ 1961 struct ucred *cr; 1962 1963 if (inp->inp_socket != NULL) { 1964 cr = inp->inp_socket->so_cred; 1965 ugp->fw_prid = jailed(cr) ? 1966 cr->cr_prison->pr_id : -1; 1967 ugp->fw_uid = cr->cr_uid; 1968 ugp->fw_ngroups = cr->cr_ngroups; 1969 bcopy(cr->cr_groups, ugp->fw_groups, 1970 sizeof(ugp->fw_groups)); 1971 } 1972} 1973 1974static int 1975check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 1976 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 1977 u_int16_t src_port, struct ip_fw_ugid *ugp, int *lookup, 1978 struct inpcb *inp) 1979{ 1980 struct inpcbinfo *pi; 1981 int wildcard; 1982 struct inpcb *pcb; 1983 int match; 1984 gid_t *gp; 1985 1986 /* 1987 * Check to see if the UDP or TCP stack supplied us with 1988 * the PCB. If so, rather then holding a lock and looking 1989 * up the PCB, we can use the one that was supplied. 1990 */ 1991 if (inp && *lookup == 0) { 1992 INP_LOCK_ASSERT(inp); 1993 if (inp->inp_socket != NULL) { 1994 fill_ugid_cache(inp, ugp); 1995 *lookup = 1; 1996 } 1997 } 1998 /* 1999 * If we have already been here and the packet has no 2000 * PCB entry associated with it, then we can safely 2001 * assume that this is a no match. 2002 */ 2003 if (*lookup == -1) 2004 return (0); 2005 if (proto == IPPROTO_TCP) { 2006 wildcard = 0; 2007 pi = &tcbinfo; 2008 } else if (proto == IPPROTO_UDP) { 2009 wildcard = INPLOOKUP_WILDCARD; 2010 pi = &udbinfo; 2011 } else 2012 return 0; 2013 match = 0; 2014 if (*lookup == 0) { 2015 INP_INFO_RLOCK(pi); 2016 pcb = (oif) ? 2017 in_pcblookup_hash(pi, 2018 dst_ip, htons(dst_port), 2019 src_ip, htons(src_port), 2020 wildcard, oif) : 2021 in_pcblookup_hash(pi, 2022 src_ip, htons(src_port), 2023 dst_ip, htons(dst_port), 2024 wildcard, NULL); 2025 if (pcb != NULL) { 2026 INP_LOCK(pcb); 2027 if (pcb->inp_socket != NULL) { 2028 fill_ugid_cache(pcb, ugp); 2029 *lookup = 1; 2030 } 2031 INP_UNLOCK(pcb); 2032 } 2033 INP_INFO_RUNLOCK(pi); 2034 if (*lookup == 0) { 2035 /* 2036 * If the lookup did not yield any results, there 2037 * is no sense in coming back and trying again. So 2038 * we can set lookup to -1 and ensure that we wont 2039 * bother the pcb system again. 2040 */ 2041 *lookup = -1; 2042 return (0); 2043 } 2044 } 2045 if (insn->o.opcode == O_UID) 2046 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2047 else if (insn->o.opcode == O_GID) { 2048 for (gp = ugp->fw_groups; 2049 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2050 if (*gp == (gid_t)insn->d[0]) { 2051 match = 1; 2052 break; 2053 } 2054 } else if (insn->o.opcode == O_JAIL) 2055 match = (ugp->fw_prid == (int)insn->d[0]); 2056 return match; 2057} 2058 2059#ifdef IPFIREWALL_NAT 2060static eventhandler_tag ifaddr_event_tag; 2061 2062static void 2063ifaddr_change(void *arg __unused, struct ifnet *ifp) 2064{ 2065 struct cfg_nat *ptr; 2066 struct ifaddr *ifa; 2067 2068 IPFW_WLOCK(&layer3_chain); 2069 /* Check every nat entry... */ 2070 LIST_FOREACH(ptr, &layer3_chain.nat, _next) { 2071 /* ...using nic 'ifp->if_xname' as dynamic alias address. */ 2072 if (strncmp(ptr->if_name, ifp->if_xname, IF_NAMESIZE) == 0) { 2073 mtx_lock(&ifp->if_addr_mtx); 2074 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) { 2075 if (ifa->ifa_addr == NULL) 2076 continue; 2077 if (ifa->ifa_addr->sa_family != AF_INET) 2078 continue; 2079 ptr->ip = ((struct sockaddr_in *) 2080 (ifa->ifa_addr))->sin_addr; 2081 LibAliasSetAddress(ptr->lib, ptr->ip); 2082 } 2083 mtx_unlock(&ifp->if_addr_mtx); 2084 } 2085 } 2086 IPFW_WUNLOCK(&layer3_chain); 2087} 2088 2089static void 2090flush_nat_ptrs(const int i) 2091{ 2092 struct ip_fw *rule; 2093 2094 IPFW_WLOCK_ASSERT(&layer3_chain); 2095 for (rule = layer3_chain.rules; rule; rule = rule->next) { 2096 ipfw_insn_nat *cmd = (ipfw_insn_nat *)ACTION_PTR(rule); 2097 if (cmd->o.opcode != O_NAT) 2098 continue; 2099 if (cmd->nat != NULL && cmd->nat->id == i) 2100 cmd->nat = NULL; 2101 } 2102} 2103 2104static struct cfg_nat * 2105lookup_nat(const int i) 2106{ 2107 struct cfg_nat *ptr; 2108 2109 LIST_FOREACH(ptr, &layer3_chain.nat, _next) 2110 if (ptr->id == i) 2111 return(ptr); 2112 return (NULL); 2113} 2114 2115#define HOOK_NAT(b, p) do { \ 2116 IPFW_WLOCK_ASSERT(&layer3_chain); \ 2117 LIST_INSERT_HEAD(b, p, _next); \ 2118} while (0) 2119 2120#define UNHOOK_NAT(p) do { \ 2121 IPFW_WLOCK_ASSERT(&layer3_chain); \ 2122 LIST_REMOVE(p, _next); \ 2123} while (0) 2124 2125#define HOOK_REDIR(b, p) do { \ 2126 LIST_INSERT_HEAD(b, p, _next); \ 2127} while (0) 2128 2129#define HOOK_SPOOL(b, p) do { \ 2130 LIST_INSERT_HEAD(b, p, _next); \ 2131} while (0) 2132 2133static void 2134del_redir_spool_cfg(struct cfg_nat *n, struct redir_chain *head) 2135{ 2136 struct cfg_redir *r, *tmp_r; 2137 struct cfg_spool *s, *tmp_s; 2138 int i, num; 2139 2140 LIST_FOREACH_SAFE(r, head, _next, tmp_r) { 2141 num = 1; /* Number of alias_link to delete. */ 2142 switch (r->mode) { 2143 case REDIR_PORT: 2144 num = r->pport_cnt; 2145 /* FALLTHROUGH */ 2146 case REDIR_ADDR: 2147 case REDIR_PROTO: 2148 /* Delete all libalias redirect entry. */ 2149 for (i = 0; i < num; i++) 2150 LibAliasRedirectDelete(n->lib, r->alink[i]); 2151 /* Del spool cfg if any. */ 2152 LIST_FOREACH_SAFE(s, &r->spool_chain, _next, tmp_s) { 2153 LIST_REMOVE(s, _next); 2154 free(s, M_IPFW); 2155 } 2156 free(r->alink, M_IPFW); 2157 LIST_REMOVE(r, _next); 2158 free(r, M_IPFW); 2159 break; 2160 default: 2161 printf("unknown redirect mode: %u\n", r->mode); 2162 /* XXX - panic?!?!? */ 2163 break; 2164 } 2165 } 2166} 2167 2168static int 2169add_redir_spool_cfg(char *buf, struct cfg_nat *ptr) 2170{ 2171 struct cfg_redir *r, *ser_r; 2172 struct cfg_spool *s, *ser_s; 2173 int cnt, off, i; 2174 char *panic_err; 2175 2176 for (cnt = 0, off = 0; cnt < ptr->redir_cnt; cnt++) { 2177 ser_r = (struct cfg_redir *)&buf[off]; 2178 r = malloc(SOF_REDIR, M_IPFW, M_WAITOK | M_ZERO); 2179 memcpy(r, ser_r, SOF_REDIR); 2180 LIST_INIT(&r->spool_chain); 2181 off += SOF_REDIR; 2182 r->alink = malloc(sizeof(struct alias_link *) * r->pport_cnt, 2183 M_IPFW, M_WAITOK | M_ZERO); 2184 switch (r->mode) { 2185 case REDIR_ADDR: 2186 r->alink[0] = LibAliasRedirectAddr(ptr->lib, r->laddr, 2187 r->paddr); 2188 break; 2189 case REDIR_PORT: 2190 for (i = 0 ; i < r->pport_cnt; i++) { 2191 /* If remotePort is all ports, set it to 0. */ 2192 u_short remotePortCopy = r->rport + i; 2193 if (r->rport_cnt == 1 && r->rport == 0) 2194 remotePortCopy = 0; 2195 r->alink[i] = LibAliasRedirectPort(ptr->lib, 2196 r->laddr, htons(r->lport + i), r->raddr, 2197 htons(remotePortCopy), r->paddr, 2198 htons(r->pport + i), r->proto); 2199 if (r->alink[i] == NULL) { 2200 r->alink[0] = NULL; 2201 break; 2202 } 2203 } 2204 break; 2205 case REDIR_PROTO: 2206 r->alink[0] = LibAliasRedirectProto(ptr->lib ,r->laddr, 2207 r->raddr, r->paddr, r->proto); 2208 break; 2209 default: 2210 printf("unknown redirect mode: %u\n", r->mode); 2211 break; 2212 } 2213 if (r->alink[0] == NULL) { 2214 panic_err = "LibAliasRedirect* returned NULL"; 2215 goto bad; 2216 } else /* LSNAT handling. */ 2217 for (i = 0; i < r->spool_cnt; i++) { 2218 ser_s = (struct cfg_spool *)&buf[off]; 2219 s = malloc(SOF_REDIR, M_IPFW, 2220 M_WAITOK | M_ZERO); 2221 memcpy(s, ser_s, SOF_SPOOL); 2222 LibAliasAddServer(ptr->lib, r->alink[0], 2223 s->addr, htons(s->port)); 2224 off += SOF_SPOOL; 2225 /* Hook spool entry. */ 2226 HOOK_SPOOL(&r->spool_chain, s); 2227 } 2228 /* And finally hook this redir entry. */ 2229 HOOK_REDIR(&ptr->redir_chain, r); 2230 } 2231 return (1); 2232bad: 2233 /* something really bad happened: panic! */ 2234 panic("%s\n", panic_err); 2235} 2236#endif 2237 2238/* 2239 * The main check routine for the firewall. 2240 * 2241 * All arguments are in args so we can modify them and return them 2242 * back to the caller. 2243 * 2244 * Parameters: 2245 * 2246 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2247 * Starts with the IP header. 2248 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2249 * args->L3offset Number of bytes bypassed if we came from L2. 2250 * e.g. often sizeof(eh) ** NOTYET ** 2251 * args->oif Outgoing interface, or NULL if packet is incoming. 2252 * The incoming interface is in the mbuf. (in) 2253 * args->divert_rule (in/out) 2254 * Skip up to the first rule past this rule number; 2255 * upon return, non-zero port number for divert or tee. 2256 * 2257 * args->rule Pointer to the last matching rule (in/out) 2258 * args->next_hop Socket we are forwarding to (out). 2259 * args->f_id Addresses grabbed from the packet (out) 2260 * args->cookie a cookie depending on rule action 2261 * 2262 * Return value: 2263 * 2264 * IP_FW_PASS the packet must be accepted 2265 * IP_FW_DENY the packet must be dropped 2266 * IP_FW_DIVERT divert packet, port in m_tag 2267 * IP_FW_TEE tee packet, port in m_tag 2268 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2269 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2270 * 2271 */ 2272int 2273ipfw_chk(struct ip_fw_args *args) 2274{ 2275 /* 2276 * Local variables holding state during the processing of a packet: 2277 * 2278 * IMPORTANT NOTE: to speed up the processing of rules, there 2279 * are some assumption on the values of the variables, which 2280 * are documented here. Should you change them, please check 2281 * the implementation of the various instructions to make sure 2282 * that they still work. 2283 * 2284 * args->eh The MAC header. It is non-null for a layer2 2285 * packet, it is NULL for a layer-3 packet. 2286 * **notyet** 2287 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2288 * 2289 * m | args->m Pointer to the mbuf, as received from the caller. 2290 * It may change if ipfw_chk() does an m_pullup, or if it 2291 * consumes the packet because it calls send_reject(). 2292 * XXX This has to change, so that ipfw_chk() never modifies 2293 * or consumes the buffer. 2294 * ip is the beginning of the ip(4 or 6) header. 2295 * Calculated by adding the L3offset to the start of data. 2296 * (Until we start using L3offset, the packet is 2297 * supposed to start with the ip header). 2298 */ 2299 struct mbuf *m = args->m; 2300 struct ip *ip = mtod(m, struct ip *); 2301 2302 /* 2303 * For rules which contain uid/gid or jail constraints, cache 2304 * a copy of the users credentials after the pcb lookup has been 2305 * executed. This will speed up the processing of rules with 2306 * these types of constraints, as well as decrease contention 2307 * on pcb related locks. 2308 */ 2309 struct ip_fw_ugid fw_ugid_cache; 2310 int ugid_lookup = 0; 2311 2312 /* 2313 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2314 * associated with a packet input on a divert socket. This 2315 * will allow to distinguish traffic and its direction when 2316 * it originates from a divert socket. 2317 */ 2318 u_int divinput_flags = 0; 2319 2320 /* 2321 * oif | args->oif If NULL, ipfw_chk has been called on the 2322 * inbound path (ether_input, ip_input). 2323 * If non-NULL, ipfw_chk has been called on the outbound path 2324 * (ether_output, ip_output). 2325 */ 2326 struct ifnet *oif = args->oif; 2327 2328 struct ip_fw *f = NULL; /* matching rule */ 2329 int retval = 0; 2330 2331 /* 2332 * hlen The length of the IP header. 2333 */ 2334 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2335 2336 /* 2337 * offset The offset of a fragment. offset != 0 means that 2338 * we have a fragment at this offset of an IPv4 packet. 2339 * offset == 0 means that (if this is an IPv4 packet) 2340 * this is the first or only fragment. 2341 * For IPv6 offset == 0 means there is no Fragment Header. 2342 * If offset != 0 for IPv6 always use correct mask to 2343 * get the correct offset because we add IP6F_MORE_FRAG 2344 * to be able to dectect the first fragment which would 2345 * otherwise have offset = 0. 2346 */ 2347 u_short offset = 0; 2348 2349 /* 2350 * Local copies of addresses. They are only valid if we have 2351 * an IP packet. 2352 * 2353 * proto The protocol. Set to 0 for non-ip packets, 2354 * or to the protocol read from the packet otherwise. 2355 * proto != 0 means that we have an IPv4 packet. 2356 * 2357 * src_port, dst_port port numbers, in HOST format. Only 2358 * valid for TCP and UDP packets. 2359 * 2360 * src_ip, dst_ip ip addresses, in NETWORK format. 2361 * Only valid for IPv4 packets. 2362 */ 2363 u_int8_t proto; 2364 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2365 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2366 u_int16_t ip_len=0; 2367 int pktlen; 2368 u_int16_t etype = 0; /* Host order stored ether type */ 2369 2370 /* 2371 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2372 * MATCH_NONE when checked and not matched (q = NULL), 2373 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2374 */ 2375 int dyn_dir = MATCH_UNKNOWN; 2376 ipfw_dyn_rule *q = NULL; 2377 struct ip_fw_chain *chain = &layer3_chain; 2378 struct m_tag *mtag; 2379 2380 /* 2381 * We store in ulp a pointer to the upper layer protocol header. 2382 * In the ipv4 case this is easy to determine from the header, 2383 * but for ipv6 we might have some additional headers in the middle. 2384 * ulp is NULL if not found. 2385 */ 2386 void *ulp = NULL; /* upper layer protocol pointer. */ 2387 /* XXX ipv6 variables */ 2388 int is_ipv6 = 0; 2389 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2390 /* end of ipv6 variables */ 2391 int is_ipv4 = 0; 2392 2393 if (m->m_flags & M_SKIP_FIREWALL) 2394 return (IP_FW_PASS); /* accept */ 2395 2396 pktlen = m->m_pkthdr.len; 2397 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2398 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2399 2400/* 2401 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2402 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2403 * pointer might become stale after other pullups (but we never use it 2404 * this way). 2405 */ 2406#define PULLUP_TO(len, p, T) \ 2407do { \ 2408 int x = (len) + sizeof(T); \ 2409 if ((m)->m_len < x) { \ 2410 args->m = m = m_pullup(m, x); \ 2411 if (m == NULL) \ 2412 goto pullup_failed; \ 2413 } \ 2414 p = (mtod(m, char *) + (len)); \ 2415} while (0) 2416 2417 /* 2418 * if we have an ether header, 2419 */ 2420 if (args->eh) 2421 etype = ntohs(args->eh->ether_type); 2422 2423 /* Identify IP packets and fill up variables. */ 2424 if (pktlen >= sizeof(struct ip6_hdr) && 2425 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2426 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2427 is_ipv6 = 1; 2428 args->f_id.addr_type = 6; 2429 hlen = sizeof(struct ip6_hdr); 2430 proto = ip6->ip6_nxt; 2431 2432 /* Search extension headers to find upper layer protocols */ 2433 while (ulp == NULL) { 2434 switch (proto) { 2435 case IPPROTO_ICMPV6: 2436 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2437 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2438 break; 2439 2440 case IPPROTO_TCP: 2441 PULLUP_TO(hlen, ulp, struct tcphdr); 2442 dst_port = TCP(ulp)->th_dport; 2443 src_port = TCP(ulp)->th_sport; 2444 args->f_id.flags = TCP(ulp)->th_flags; 2445 break; 2446 2447 case IPPROTO_SCTP: 2448 PULLUP_TO(hlen, ulp, struct sctphdr); 2449 src_port = SCTP(ulp)->src_port; 2450 dst_port = SCTP(ulp)->dest_port; 2451 break; 2452 2453 case IPPROTO_UDP: 2454 PULLUP_TO(hlen, ulp, struct udphdr); 2455 dst_port = UDP(ulp)->uh_dport; 2456 src_port = UDP(ulp)->uh_sport; 2457 break; 2458 2459 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2460 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2461 ext_hd |= EXT_HOPOPTS; 2462 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2463 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2464 ulp = NULL; 2465 break; 2466 2467 case IPPROTO_ROUTING: /* RFC 2460 */ 2468 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2469 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2470 case 0: 2471 ext_hd |= EXT_RTHDR0; 2472 break; 2473 case 2: 2474 ext_hd |= EXT_RTHDR2; 2475 break; 2476 default: 2477 printf("IPFW2: IPV6 - Unknown Routing " 2478 "Header type(%d)\n", 2479 ((struct ip6_rthdr *)ulp)->ip6r_type); 2480 if (fw_deny_unknown_exthdrs) 2481 return (IP_FW_DENY); 2482 break; 2483 } 2484 ext_hd |= EXT_ROUTING; 2485 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2486 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2487 ulp = NULL; 2488 break; 2489 2490 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2491 PULLUP_TO(hlen, ulp, struct ip6_frag); 2492 ext_hd |= EXT_FRAGMENT; 2493 hlen += sizeof (struct ip6_frag); 2494 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2495 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2496 IP6F_OFF_MASK; 2497 /* Add IP6F_MORE_FRAG for offset of first 2498 * fragment to be != 0. */ 2499 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2500 IP6F_MORE_FRAG; 2501 if (offset == 0) { 2502 printf("IPFW2: IPV6 - Invalid Fragment " 2503 "Header\n"); 2504 if (fw_deny_unknown_exthdrs) 2505 return (IP_FW_DENY); 2506 break; 2507 } 2508 args->f_id.frag_id6 = 2509 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2510 ulp = NULL; 2511 break; 2512 2513 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2514 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2515 ext_hd |= EXT_DSTOPTS; 2516 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2517 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2518 ulp = NULL; 2519 break; 2520 2521 case IPPROTO_AH: /* RFC 2402 */ 2522 PULLUP_TO(hlen, ulp, struct ip6_ext); 2523 ext_hd |= EXT_AH; 2524 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2525 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2526 ulp = NULL; 2527 break; 2528 2529 case IPPROTO_ESP: /* RFC 2406 */ 2530 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2531 /* Anything past Seq# is variable length and 2532 * data past this ext. header is encrypted. */ 2533 ext_hd |= EXT_ESP; 2534 break; 2535 2536 case IPPROTO_NONE: /* RFC 2460 */ 2537 PULLUP_TO(hlen, ulp, struct ip6_ext); 2538 /* Packet ends here. if ip6e_len!=0 octets 2539 * must be ignored. */ 2540 break; 2541 2542 case IPPROTO_OSPFIGP: 2543 /* XXX OSPF header check? */ 2544 PULLUP_TO(hlen, ulp, struct ip6_ext); 2545 break; 2546 2547 case IPPROTO_PIM: 2548 /* XXX PIM header check? */ 2549 PULLUP_TO(hlen, ulp, struct pim); 2550 break; 2551 2552 case IPPROTO_CARP: 2553 PULLUP_TO(hlen, ulp, struct carp_header); 2554 if (((struct carp_header *)ulp)->carp_version != 2555 CARP_VERSION) 2556 return (IP_FW_DENY); 2557 if (((struct carp_header *)ulp)->carp_type != 2558 CARP_ADVERTISEMENT) 2559 return (IP_FW_DENY); 2560 break; 2561 2562 case IPPROTO_IPV6: /* RFC 2893 */ 2563 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2564 break; 2565 2566 case IPPROTO_IPV4: /* RFC 2893 */ 2567 PULLUP_TO(hlen, ulp, struct ip); 2568 break; 2569 2570 default: 2571 printf("IPFW2: IPV6 - Unknown Extension " 2572 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2573 if (fw_deny_unknown_exthdrs) 2574 return (IP_FW_DENY); 2575 PULLUP_TO(hlen, ulp, struct ip6_ext); 2576 break; 2577 } /*switch */ 2578 } 2579 ip = mtod(m, struct ip *); 2580 ip6 = (struct ip6_hdr *)ip; 2581 args->f_id.src_ip6 = ip6->ip6_src; 2582 args->f_id.dst_ip6 = ip6->ip6_dst; 2583 args->f_id.src_ip = 0; 2584 args->f_id.dst_ip = 0; 2585 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2586 } else if (pktlen >= sizeof(struct ip) && 2587 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2588 is_ipv4 = 1; 2589 hlen = ip->ip_hl << 2; 2590 args->f_id.addr_type = 4; 2591 2592 /* 2593 * Collect parameters into local variables for faster matching. 2594 */ 2595 proto = ip->ip_p; 2596 src_ip = ip->ip_src; 2597 dst_ip = ip->ip_dst; 2598 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2599 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2600 ip_len = ntohs(ip->ip_len); 2601 } else { 2602 offset = ip->ip_off & IP_OFFMASK; 2603 ip_len = ip->ip_len; 2604 } 2605 pktlen = ip_len < pktlen ? ip_len : pktlen; 2606 2607 if (offset == 0) { 2608 switch (proto) { 2609 case IPPROTO_TCP: 2610 PULLUP_TO(hlen, ulp, struct tcphdr); 2611 dst_port = TCP(ulp)->th_dport; 2612 src_port = TCP(ulp)->th_sport; 2613 args->f_id.flags = TCP(ulp)->th_flags; 2614 break; 2615 2616 case IPPROTO_UDP: 2617 PULLUP_TO(hlen, ulp, struct udphdr); 2618 dst_port = UDP(ulp)->uh_dport; 2619 src_port = UDP(ulp)->uh_sport; 2620 break; 2621 2622 case IPPROTO_ICMP: 2623 PULLUP_TO(hlen, ulp, struct icmphdr); 2624 args->f_id.flags = ICMP(ulp)->icmp_type; 2625 break; 2626 2627 default: 2628 break; 2629 } 2630 } 2631 2632 ip = mtod(m, struct ip *); 2633 args->f_id.src_ip = ntohl(src_ip.s_addr); 2634 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2635 } 2636#undef PULLUP_TO 2637 if (proto) { /* we may have port numbers, store them */ 2638 args->f_id.proto = proto; 2639 args->f_id.src_port = src_port = ntohs(src_port); 2640 args->f_id.dst_port = dst_port = ntohs(dst_port); 2641 } 2642 2643 IPFW_RLOCK(chain); 2644 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2645 if (args->rule) { 2646 /* 2647 * Packet has already been tagged. Look for the next rule 2648 * to restart processing. 2649 * 2650 * If fw_one_pass != 0 then just accept it. 2651 * XXX should not happen here, but optimized out in 2652 * the caller. 2653 */ 2654 if (fw_one_pass) { 2655 IPFW_RUNLOCK(chain); 2656 return (IP_FW_PASS); 2657 } 2658 2659 f = args->rule->next_rule; 2660 if (f == NULL) 2661 f = lookup_next_rule(args->rule); 2662 } else { 2663 /* 2664 * Find the starting rule. It can be either the first 2665 * one, or the one after divert_rule if asked so. 2666 */ 2667 int skipto = mtag ? divert_cookie(mtag) : 0; 2668 2669 f = chain->rules; 2670 if (args->eh == NULL && skipto != 0) { 2671 if (skipto >= IPFW_DEFAULT_RULE) { 2672 IPFW_RUNLOCK(chain); 2673 return (IP_FW_DENY); /* invalid */ 2674 } 2675 while (f && f->rulenum <= skipto) 2676 f = f->next; 2677 if (f == NULL) { /* drop packet */ 2678 IPFW_RUNLOCK(chain); 2679 return (IP_FW_DENY); 2680 } 2681 } 2682 } 2683 /* reset divert rule to avoid confusion later */ 2684 if (mtag) { 2685 divinput_flags = divert_info(mtag) & 2686 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2687 m_tag_delete(m, mtag); 2688 } 2689 2690 /* 2691 * Now scan the rules, and parse microinstructions for each rule. 2692 */ 2693 for (; f; f = f->next) { 2694 ipfw_insn *cmd; 2695 uint32_t tablearg = 0; 2696 int l, cmdlen, skip_or; /* skip rest of OR block */ 2697 2698again: 2699 if (set_disable & (1 << f->set) ) 2700 continue; 2701 2702 skip_or = 0; 2703 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2704 l -= cmdlen, cmd += cmdlen) { 2705 int match; 2706 2707 /* 2708 * check_body is a jump target used when we find a 2709 * CHECK_STATE, and need to jump to the body of 2710 * the target rule. 2711 */ 2712 2713check_body: 2714 cmdlen = F_LEN(cmd); 2715 /* 2716 * An OR block (insn_1 || .. || insn_n) has the 2717 * F_OR bit set in all but the last instruction. 2718 * The first match will set "skip_or", and cause 2719 * the following instructions to be skipped until 2720 * past the one with the F_OR bit clear. 2721 */ 2722 if (skip_or) { /* skip this instruction */ 2723 if ((cmd->len & F_OR) == 0) 2724 skip_or = 0; /* next one is good */ 2725 continue; 2726 } 2727 match = 0; /* set to 1 if we succeed */ 2728 2729 switch (cmd->opcode) { 2730 /* 2731 * The first set of opcodes compares the packet's 2732 * fields with some pattern, setting 'match' if a 2733 * match is found. At the end of the loop there is 2734 * logic to deal with F_NOT and F_OR flags associated 2735 * with the opcode. 2736 */ 2737 case O_NOP: 2738 match = 1; 2739 break; 2740 2741 case O_FORWARD_MAC: 2742 printf("ipfw: opcode %d unimplemented\n", 2743 cmd->opcode); 2744 break; 2745 2746 case O_GID: 2747 case O_UID: 2748 case O_JAIL: 2749 /* 2750 * We only check offset == 0 && proto != 0, 2751 * as this ensures that we have a 2752 * packet with the ports info. 2753 */ 2754 if (offset!=0) 2755 break; 2756 if (is_ipv6) /* XXX to be fixed later */ 2757 break; 2758 if (proto == IPPROTO_TCP || 2759 proto == IPPROTO_UDP) 2760 match = check_uidgid( 2761 (ipfw_insn_u32 *)cmd, 2762 proto, oif, 2763 dst_ip, dst_port, 2764 src_ip, src_port, &fw_ugid_cache, 2765 &ugid_lookup, args->inp); 2766 break; 2767 2768 case O_RECV: 2769 match = iface_match(m->m_pkthdr.rcvif, 2770 (ipfw_insn_if *)cmd); 2771 break; 2772 2773 case O_XMIT: 2774 match = iface_match(oif, (ipfw_insn_if *)cmd); 2775 break; 2776 2777 case O_VIA: 2778 match = iface_match(oif ? oif : 2779 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2780 break; 2781 2782 case O_MACADDR2: 2783 if (args->eh != NULL) { /* have MAC header */ 2784 u_int32_t *want = (u_int32_t *) 2785 ((ipfw_insn_mac *)cmd)->addr; 2786 u_int32_t *mask = (u_int32_t *) 2787 ((ipfw_insn_mac *)cmd)->mask; 2788 u_int32_t *hdr = (u_int32_t *)args->eh; 2789 2790 match = 2791 ( want[0] == (hdr[0] & mask[0]) && 2792 want[1] == (hdr[1] & mask[1]) && 2793 want[2] == (hdr[2] & mask[2]) ); 2794 } 2795 break; 2796 2797 case O_MAC_TYPE: 2798 if (args->eh != NULL) { 2799 u_int16_t *p = 2800 ((ipfw_insn_u16 *)cmd)->ports; 2801 int i; 2802 2803 for (i = cmdlen - 1; !match && i>0; 2804 i--, p += 2) 2805 match = (etype >= p[0] && 2806 etype <= p[1]); 2807 } 2808 break; 2809 2810 case O_FRAG: 2811 match = (offset != 0); 2812 break; 2813 2814 case O_IN: /* "out" is "not in" */ 2815 match = (oif == NULL); 2816 break; 2817 2818 case O_LAYER2: 2819 match = (args->eh != NULL); 2820 break; 2821 2822 case O_DIVERTED: 2823 match = (cmd->arg1 & 1 && divinput_flags & 2824 IP_FW_DIVERT_LOOPBACK_FLAG) || 2825 (cmd->arg1 & 2 && divinput_flags & 2826 IP_FW_DIVERT_OUTPUT_FLAG); 2827 break; 2828 2829 case O_PROTO: 2830 /* 2831 * We do not allow an arg of 0 so the 2832 * check of "proto" only suffices. 2833 */ 2834 match = (proto == cmd->arg1); 2835 break; 2836 2837 case O_IP_SRC: 2838 match = is_ipv4 && 2839 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2840 src_ip.s_addr); 2841 break; 2842 2843 case O_IP_SRC_LOOKUP: 2844 case O_IP_DST_LOOKUP: 2845 if (is_ipv4) { 2846 uint32_t a = 2847 (cmd->opcode == O_IP_DST_LOOKUP) ? 2848 dst_ip.s_addr : src_ip.s_addr; 2849 uint32_t v; 2850 2851 match = lookup_table(chain, cmd->arg1, a, 2852 &v); 2853 if (!match) 2854 break; 2855 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2856 match = 2857 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2858 else 2859 tablearg = v; 2860 } 2861 break; 2862 2863 case O_IP_SRC_MASK: 2864 case O_IP_DST_MASK: 2865 if (is_ipv4) { 2866 uint32_t a = 2867 (cmd->opcode == O_IP_DST_MASK) ? 2868 dst_ip.s_addr : src_ip.s_addr; 2869 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2870 int i = cmdlen-1; 2871 2872 for (; !match && i>0; i-= 2, p+= 2) 2873 match = (p[0] == (a & p[1])); 2874 } 2875 break; 2876 2877 case O_IP_SRC_ME: 2878 if (is_ipv4) { 2879 struct ifnet *tif; 2880 2881 INADDR_TO_IFP(src_ip, tif); 2882 match = (tif != NULL); 2883 } 2884 break; 2885 2886 case O_IP_DST_SET: 2887 case O_IP_SRC_SET: 2888 if (is_ipv4) { 2889 u_int32_t *d = (u_int32_t *)(cmd+1); 2890 u_int32_t addr = 2891 cmd->opcode == O_IP_DST_SET ? 2892 args->f_id.dst_ip : 2893 args->f_id.src_ip; 2894 2895 if (addr < d[0]) 2896 break; 2897 addr -= d[0]; /* subtract base */ 2898 match = (addr < cmd->arg1) && 2899 ( d[ 1 + (addr>>5)] & 2900 (1<<(addr & 0x1f)) ); 2901 } 2902 break; 2903 2904 case O_IP_DST: 2905 match = is_ipv4 && 2906 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2907 dst_ip.s_addr); 2908 break; 2909 2910 case O_IP_DST_ME: 2911 if (is_ipv4) { 2912 struct ifnet *tif; 2913 2914 INADDR_TO_IFP(dst_ip, tif); 2915 match = (tif != NULL); 2916 } 2917 break; 2918 2919 case O_IP_SRCPORT: 2920 case O_IP_DSTPORT: 2921 /* 2922 * offset == 0 && proto != 0 is enough 2923 * to guarantee that we have a 2924 * packet with port info. 2925 */ 2926 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2927 && offset == 0) { 2928 u_int16_t x = 2929 (cmd->opcode == O_IP_SRCPORT) ? 2930 src_port : dst_port ; 2931 u_int16_t *p = 2932 ((ipfw_insn_u16 *)cmd)->ports; 2933 int i; 2934 2935 for (i = cmdlen - 1; !match && i>0; 2936 i--, p += 2) 2937 match = (x>=p[0] && x<=p[1]); 2938 } 2939 break; 2940 2941 case O_ICMPTYPE: 2942 match = (offset == 0 && proto==IPPROTO_ICMP && 2943 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2944 break; 2945 2946#ifdef INET6 2947 case O_ICMP6TYPE: 2948 match = is_ipv6 && offset == 0 && 2949 proto==IPPROTO_ICMPV6 && 2950 icmp6type_match( 2951 ICMP6(ulp)->icmp6_type, 2952 (ipfw_insn_u32 *)cmd); 2953 break; 2954#endif /* INET6 */ 2955 2956 case O_IPOPT: 2957 match = (is_ipv4 && 2958 ipopts_match(ip, cmd) ); 2959 break; 2960 2961 case O_IPVER: 2962 match = (is_ipv4 && 2963 cmd->arg1 == ip->ip_v); 2964 break; 2965 2966 case O_IPID: 2967 case O_IPLEN: 2968 case O_IPTTL: 2969 if (is_ipv4) { /* only for IP packets */ 2970 uint16_t x; 2971 uint16_t *p; 2972 int i; 2973 2974 if (cmd->opcode == O_IPLEN) 2975 x = ip_len; 2976 else if (cmd->opcode == O_IPTTL) 2977 x = ip->ip_ttl; 2978 else /* must be IPID */ 2979 x = ntohs(ip->ip_id); 2980 if (cmdlen == 1) { 2981 match = (cmd->arg1 == x); 2982 break; 2983 } 2984 /* otherwise we have ranges */ 2985 p = ((ipfw_insn_u16 *)cmd)->ports; 2986 i = cmdlen - 1; 2987 for (; !match && i>0; i--, p += 2) 2988 match = (x >= p[0] && x <= p[1]); 2989 } 2990 break; 2991 2992 case O_IPPRECEDENCE: 2993 match = (is_ipv4 && 2994 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2995 break; 2996 2997 case O_IPTOS: 2998 match = (is_ipv4 && 2999 flags_match(cmd, ip->ip_tos)); 3000 break; 3001 3002 case O_TCPDATALEN: 3003 if (proto == IPPROTO_TCP && offset == 0) { 3004 struct tcphdr *tcp; 3005 uint16_t x; 3006 uint16_t *p; 3007 int i; 3008 3009 tcp = TCP(ulp); 3010 x = ip_len - 3011 ((ip->ip_hl + tcp->th_off) << 2); 3012 if (cmdlen == 1) { 3013 match = (cmd->arg1 == x); 3014 break; 3015 } 3016 /* otherwise we have ranges */ 3017 p = ((ipfw_insn_u16 *)cmd)->ports; 3018 i = cmdlen - 1; 3019 for (; !match && i>0; i--, p += 2) 3020 match = (x >= p[0] && x <= p[1]); 3021 } 3022 break; 3023 3024 case O_TCPFLAGS: 3025 match = (proto == IPPROTO_TCP && offset == 0 && 3026 flags_match(cmd, TCP(ulp)->th_flags)); 3027 break; 3028 3029 case O_TCPOPTS: 3030 match = (proto == IPPROTO_TCP && offset == 0 && 3031 tcpopts_match(TCP(ulp), cmd)); 3032 break; 3033 3034 case O_TCPSEQ: 3035 match = (proto == IPPROTO_TCP && offset == 0 && 3036 ((ipfw_insn_u32 *)cmd)->d[0] == 3037 TCP(ulp)->th_seq); 3038 break; 3039 3040 case O_TCPACK: 3041 match = (proto == IPPROTO_TCP && offset == 0 && 3042 ((ipfw_insn_u32 *)cmd)->d[0] == 3043 TCP(ulp)->th_ack); 3044 break; 3045 3046 case O_TCPWIN: 3047 match = (proto == IPPROTO_TCP && offset == 0 && 3048 cmd->arg1 == TCP(ulp)->th_win); 3049 break; 3050 3051 case O_ESTAB: 3052 /* reject packets which have SYN only */ 3053 /* XXX should i also check for TH_ACK ? */ 3054 match = (proto == IPPROTO_TCP && offset == 0 && 3055 (TCP(ulp)->th_flags & 3056 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 3057 break; 3058 3059 case O_ALTQ: { 3060 struct pf_mtag *at; 3061 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 3062 3063 match = 1; 3064 at = pf_find_mtag(m); 3065 if (at != NULL && at->qid != 0) 3066 break; 3067 at = pf_get_mtag(m); 3068 if (at == NULL) { 3069 /* 3070 * Let the packet fall back to the 3071 * default ALTQ. 3072 */ 3073 break; 3074 } 3075 at->qid = altq->qid; 3076 if (is_ipv4) 3077 at->af = AF_INET; 3078 else 3079 at->af = AF_LINK; 3080 at->hdr = ip; 3081 break; 3082 } 3083 3084 case O_LOG: 3085 if (fw_verbose) 3086 ipfw_log(f, hlen, args, m, 3087 oif, offset, tablearg, ip); 3088 match = 1; 3089 break; 3090 3091 case O_PROB: 3092 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 3093 break; 3094 3095 case O_VERREVPATH: 3096 /* Outgoing packets automatically pass/match */ 3097 match = ((oif != NULL) || 3098 (m->m_pkthdr.rcvif == NULL) || 3099 ( 3100#ifdef INET6 3101 is_ipv6 ? 3102 verify_path6(&(args->f_id.src_ip6), 3103 m->m_pkthdr.rcvif) : 3104#endif 3105 verify_path(src_ip, m->m_pkthdr.rcvif))); 3106 break; 3107 3108 case O_VERSRCREACH: 3109 /* Outgoing packets automatically pass/match */ 3110 match = (hlen > 0 && ((oif != NULL) || 3111#ifdef INET6 3112 is_ipv6 ? 3113 verify_path6(&(args->f_id.src_ip6), 3114 NULL) : 3115#endif 3116 verify_path(src_ip, NULL))); 3117 break; 3118 3119 case O_ANTISPOOF: 3120 /* Outgoing packets automatically pass/match */ 3121 if (oif == NULL && hlen > 0 && 3122 ( (is_ipv4 && in_localaddr(src_ip)) 3123#ifdef INET6 3124 || (is_ipv6 && 3125 in6_localaddr(&(args->f_id.src_ip6))) 3126#endif 3127 )) 3128 match = 3129#ifdef INET6 3130 is_ipv6 ? verify_path6( 3131 &(args->f_id.src_ip6), 3132 m->m_pkthdr.rcvif) : 3133#endif 3134 verify_path(src_ip, 3135 m->m_pkthdr.rcvif); 3136 else 3137 match = 1; 3138 break; 3139 3140 case O_IPSEC: 3141#ifdef IPSEC 3142 match = (m_tag_find(m, 3143 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 3144#endif 3145 /* otherwise no match */ 3146 break; 3147 3148#ifdef INET6 3149 case O_IP6_SRC: 3150 match = is_ipv6 && 3151 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 3152 &((ipfw_insn_ip6 *)cmd)->addr6); 3153 break; 3154 3155 case O_IP6_DST: 3156 match = is_ipv6 && 3157 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 3158 &((ipfw_insn_ip6 *)cmd)->addr6); 3159 break; 3160 case O_IP6_SRC_MASK: 3161 case O_IP6_DST_MASK: 3162 if (is_ipv6) { 3163 int i = cmdlen - 1; 3164 struct in6_addr p; 3165 struct in6_addr *d = 3166 &((ipfw_insn_ip6 *)cmd)->addr6; 3167 3168 for (; !match && i > 0; d += 2, 3169 i -= F_INSN_SIZE(struct in6_addr) 3170 * 2) { 3171 p = (cmd->opcode == 3172 O_IP6_SRC_MASK) ? 3173 args->f_id.src_ip6: 3174 args->f_id.dst_ip6; 3175 APPLY_MASK(&p, &d[1]); 3176 match = 3177 IN6_ARE_ADDR_EQUAL(&d[0], 3178 &p); 3179 } 3180 } 3181 break; 3182 3183 case O_IP6_SRC_ME: 3184 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3185 break; 3186 3187 case O_IP6_DST_ME: 3188 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3189 break; 3190 3191 case O_FLOW6ID: 3192 match = is_ipv6 && 3193 flow6id_match(args->f_id.flow_id6, 3194 (ipfw_insn_u32 *) cmd); 3195 break; 3196 3197 case O_EXT_HDR: 3198 match = is_ipv6 && 3199 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3200 break; 3201 3202 case O_IP6: 3203 match = is_ipv6; 3204 break; 3205#endif 3206 3207 case O_IP4: 3208 match = is_ipv4; 3209 break; 3210 3211 case O_TAG: { 3212 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3213 tablearg : cmd->arg1; 3214 3215 /* Packet is already tagged with this tag? */ 3216 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3217 3218 /* We have `untag' action when F_NOT flag is 3219 * present. And we must remove this mtag from 3220 * mbuf and reset `match' to zero (`match' will 3221 * be inversed later). 3222 * Otherwise we should allocate new mtag and 3223 * push it into mbuf. 3224 */ 3225 if (cmd->len & F_NOT) { /* `untag' action */ 3226 if (mtag != NULL) 3227 m_tag_delete(m, mtag); 3228 } else if (mtag == NULL) { 3229 if ((mtag = m_tag_alloc(MTAG_IPFW, 3230 tag, 0, M_NOWAIT)) != NULL) 3231 m_tag_prepend(m, mtag); 3232 } 3233 match = (cmd->len & F_NOT) ? 0: 1; 3234 break; 3235 } 3236 3237 case O_TAGGED: { 3238 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3239 tablearg : cmd->arg1; 3240 3241 if (cmdlen == 1) { 3242 match = m_tag_locate(m, MTAG_IPFW, 3243 tag, NULL) != NULL; 3244 break; 3245 } 3246 3247 /* we have ranges */ 3248 for (mtag = m_tag_first(m); 3249 mtag != NULL && !match; 3250 mtag = m_tag_next(m, mtag)) { 3251 uint16_t *p; 3252 int i; 3253 3254 if (mtag->m_tag_cookie != MTAG_IPFW) 3255 continue; 3256 3257 p = ((ipfw_insn_u16 *)cmd)->ports; 3258 i = cmdlen - 1; 3259 for(; !match && i > 0; i--, p += 2) 3260 match = 3261 mtag->m_tag_id >= p[0] && 3262 mtag->m_tag_id <= p[1]; 3263 } 3264 break; 3265 } 3266 3267 /* 3268 * The second set of opcodes represents 'actions', 3269 * i.e. the terminal part of a rule once the packet 3270 * matches all previous patterns. 3271 * Typically there is only one action for each rule, 3272 * and the opcode is stored at the end of the rule 3273 * (but there are exceptions -- see below). 3274 * 3275 * In general, here we set retval and terminate the 3276 * outer loop (would be a 'break 3' in some language, 3277 * but we need to do a 'goto done'). 3278 * 3279 * Exceptions: 3280 * O_COUNT and O_SKIPTO actions: 3281 * instead of terminating, we jump to the next rule 3282 * ('goto next_rule', equivalent to a 'break 2'), 3283 * or to the SKIPTO target ('goto again' after 3284 * having set f, cmd and l), respectively. 3285 * 3286 * O_TAG, O_LOG and O_ALTQ action parameters: 3287 * perform some action and set match = 1; 3288 * 3289 * O_LIMIT and O_KEEP_STATE: these opcodes are 3290 * not real 'actions', and are stored right 3291 * before the 'action' part of the rule. 3292 * These opcodes try to install an entry in the 3293 * state tables; if successful, we continue with 3294 * the next opcode (match=1; break;), otherwise 3295 * the packet * must be dropped 3296 * ('goto done' after setting retval); 3297 * 3298 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3299 * cause a lookup of the state table, and a jump 3300 * to the 'action' part of the parent rule 3301 * ('goto check_body') if an entry is found, or 3302 * (CHECK_STATE only) a jump to the next rule if 3303 * the entry is not found ('goto next_rule'). 3304 * The result of the lookup is cached to make 3305 * further instances of these opcodes are 3306 * effectively NOPs. 3307 */ 3308 case O_LIMIT: 3309 case O_KEEP_STATE: 3310 if (install_state(f, 3311 (ipfw_insn_limit *)cmd, args, tablearg)) { 3312 retval = IP_FW_DENY; 3313 goto done; /* error/limit violation */ 3314 } 3315 match = 1; 3316 break; 3317 3318 case O_PROBE_STATE: 3319 case O_CHECK_STATE: 3320 /* 3321 * dynamic rules are checked at the first 3322 * keep-state or check-state occurrence, 3323 * with the result being stored in dyn_dir. 3324 * The compiler introduces a PROBE_STATE 3325 * instruction for us when we have a 3326 * KEEP_STATE (because PROBE_STATE needs 3327 * to be run first). 3328 */ 3329 if (dyn_dir == MATCH_UNKNOWN && 3330 (q = lookup_dyn_rule(&args->f_id, 3331 &dyn_dir, proto == IPPROTO_TCP ? 3332 TCP(ulp) : NULL)) 3333 != NULL) { 3334 /* 3335 * Found dynamic entry, update stats 3336 * and jump to the 'action' part of 3337 * the parent rule. 3338 */ 3339 q->pcnt++; 3340 q->bcnt += pktlen; 3341 f = q->rule; 3342 cmd = ACTION_PTR(f); 3343 l = f->cmd_len - f->act_ofs; 3344 IPFW_DYN_UNLOCK(); 3345 goto check_body; 3346 } 3347 /* 3348 * Dynamic entry not found. If CHECK_STATE, 3349 * skip to next rule, if PROBE_STATE just 3350 * ignore and continue with next opcode. 3351 */ 3352 if (cmd->opcode == O_CHECK_STATE) 3353 goto next_rule; 3354 match = 1; 3355 break; 3356 3357 case O_ACCEPT: 3358 retval = 0; /* accept */ 3359 goto done; 3360 3361 case O_PIPE: 3362 case O_QUEUE: 3363 args->rule = f; /* report matching rule */ 3364 if (cmd->arg1 == IP_FW_TABLEARG) 3365 args->cookie = tablearg; 3366 else 3367 args->cookie = cmd->arg1; 3368 retval = IP_FW_DUMMYNET; 3369 goto done; 3370 3371 case O_DIVERT: 3372 case O_TEE: { 3373 struct divert_tag *dt; 3374 3375 if (args->eh) /* not on layer 2 */ 3376 break; 3377 mtag = m_tag_get(PACKET_TAG_DIVERT, 3378 sizeof(struct divert_tag), 3379 M_NOWAIT); 3380 if (mtag == NULL) { 3381 /* XXX statistic */ 3382 /* drop packet */ 3383 IPFW_RUNLOCK(chain); 3384 return (IP_FW_DENY); 3385 } 3386 dt = (struct divert_tag *)(mtag+1); 3387 dt->cookie = f->rulenum; 3388 if (cmd->arg1 == IP_FW_TABLEARG) 3389 dt->info = tablearg; 3390 else 3391 dt->info = cmd->arg1; 3392 m_tag_prepend(m, mtag); 3393 retval = (cmd->opcode == O_DIVERT) ? 3394 IP_FW_DIVERT : IP_FW_TEE; 3395 goto done; 3396 } 3397 3398 case O_COUNT: 3399 case O_SKIPTO: 3400 f->pcnt++; /* update stats */ 3401 f->bcnt += pktlen; 3402 f->timestamp = time_uptime; 3403 if (cmd->opcode == O_COUNT) 3404 goto next_rule; 3405 /* handle skipto */ 3406 if (f->next_rule == NULL) 3407 lookup_next_rule(f); 3408 f = f->next_rule; 3409 goto again; 3410 3411 case O_REJECT: 3412 /* 3413 * Drop the packet and send a reject notice 3414 * if the packet is not ICMP (or is an ICMP 3415 * query), and it is not multicast/broadcast. 3416 */ 3417 if (hlen > 0 && is_ipv4 && offset == 0 && 3418 (proto != IPPROTO_ICMP || 3419 is_icmp_query(ICMP(ulp))) && 3420 !(m->m_flags & (M_BCAST|M_MCAST)) && 3421 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3422 send_reject(args, cmd->arg1, ip_len, ip); 3423 m = args->m; 3424 } 3425 /* FALLTHROUGH */ 3426#ifdef INET6 3427 case O_UNREACH6: 3428 if (hlen > 0 && is_ipv6 && 3429 ((offset & IP6F_OFF_MASK) == 0) && 3430 (proto != IPPROTO_ICMPV6 || 3431 (is_icmp6_query(args->f_id.flags) == 1)) && 3432 !(m->m_flags & (M_BCAST|M_MCAST)) && 3433 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3434 send_reject6( 3435 args, cmd->arg1, hlen, 3436 (struct ip6_hdr *)ip); 3437 m = args->m; 3438 } 3439 /* FALLTHROUGH */ 3440#endif 3441 case O_DENY: 3442 retval = IP_FW_DENY; 3443 goto done; 3444 3445 case O_FORWARD_IP: { 3446 struct sockaddr_in *sa; 3447 sa = &(((ipfw_insn_sa *)cmd)->sa); 3448 if (args->eh) /* not valid on layer2 pkts */ 3449 break; 3450 if (!q || dyn_dir == MATCH_FORWARD) { 3451 if (sa->sin_addr.s_addr == INADDR_ANY) { 3452 bcopy(sa, &args->hopstore, 3453 sizeof(*sa)); 3454 args->hopstore.sin_addr.s_addr = 3455 htonl(tablearg); 3456 args->next_hop = 3457 &args->hopstore; 3458 } else { 3459 args->next_hop = sa; 3460 } 3461 } 3462 retval = IP_FW_PASS; 3463 } 3464 goto done; 3465 3466 case O_NETGRAPH: 3467 case O_NGTEE: 3468 args->rule = f; /* report matching rule */ 3469 if (cmd->arg1 == IP_FW_TABLEARG) 3470 args->cookie = tablearg; 3471 else 3472 args->cookie = cmd->arg1; 3473 retval = (cmd->opcode == O_NETGRAPH) ? 3474 IP_FW_NETGRAPH : IP_FW_NGTEE; 3475 goto done; 3476 3477#ifdef IPFIREWALL_NAT 3478 case O_NAT: { 3479 struct cfg_nat *t; 3480 struct mbuf *mcl; 3481 /* XXX - libalias duct tape */ 3482 int ldt; 3483 char *c; 3484 3485 ldt = 0; 3486 args->rule = f; /* Report matching rule. */ 3487 retval = 0; 3488 t = ((ipfw_insn_nat *)cmd)->nat; 3489 if (t == NULL) { 3490 t = lookup_nat(cmd->arg1); 3491 if (t == NULL) { 3492 retval = IP_FW_DENY; 3493 goto done; 3494 } else 3495 ((ipfw_insn_nat *)cmd)->nat = 3496 t; 3497 } 3498 if ((mcl = m_megapullup(m, m->m_pkthdr.len)) == 3499 NULL) 3500 goto badnat; 3501 ip = mtod(mcl, struct ip *); 3502 if (args->eh == NULL) { 3503 ip->ip_len = htons(ip->ip_len); 3504 ip->ip_off = htons(ip->ip_off); 3505 } 3506 3507 /* 3508 * XXX - Libalias checksum offload 'duct tape': 3509 * 3510 * locally generated packets have only 3511 * pseudo-header checksum calculated 3512 * and libalias will screw it[1], so 3513 * mark them for later fix. Moreover 3514 * there are cases when libalias 3515 * modify tcp packet data[2], mark it 3516 * for later fix too. 3517 * 3518 * [1] libalias was never meant to run 3519 * in kernel, so it doesn't have any 3520 * knowledge about checksum 3521 * offloading, and it expects a packet 3522 * with a full internet 3523 * checksum. Unfortunately, packets 3524 * generated locally will have just the 3525 * pseudo header calculated, and when 3526 * libalias tries to adjust the 3527 * checksum it will actually screw it. 3528 * 3529 * [2] when libalias modify tcp's data 3530 * content, full TCP checksum has to 3531 * be recomputed: the problem is that 3532 * libalias doesn't have any idea 3533 * about checksum offloading To 3534 * workaround this, we do not do 3535 * checksumming in LibAlias, but only 3536 * mark the packets in th_x2 field. If 3537 * we receive a marked packet, we 3538 * calculate correct checksum for it 3539 * aware of offloading. Why such a 3540 * terrible hack instead of 3541 * recalculating checksum for each 3542 * packet? Because the previous 3543 * checksum was not checked! 3544 * Recalculating checksums for EVERY 3545 * packet will hide ALL transmission 3546 * errors. Yes, marked packets still 3547 * suffer from this problem. But, 3548 * sigh, natd(8) has this problem, 3549 * too. 3550 * 3551 * TODO: -make libalias mbuf aware (so 3552 * it can handle delayed checksum and tso) 3553 */ 3554 3555 if (mcl->m_pkthdr.rcvif == NULL && 3556 mcl->m_pkthdr.csum_flags & 3557 CSUM_DELAY_DATA) 3558 ldt = 1; 3559 3560 c = mtod(mcl, char *); 3561 if (oif == NULL) 3562 retval = LibAliasIn(t->lib, c, 3563 MCLBYTES); 3564 else 3565 retval = LibAliasOut(t->lib, c, 3566 MCLBYTES); 3567 if (retval != PKT_ALIAS_OK) { 3568 /* XXX - should i add some logging? */ 3569 m_free(mcl); 3570 badnat: 3571 args->m = NULL; 3572 retval = IP_FW_DENY; 3573 goto done; 3574 } 3575 mcl->m_pkthdr.len = mcl->m_len = 3576 ntohs(ip->ip_len); 3577 3578 /* 3579 * XXX - libalias checksum offload 3580 * 'duct tape' (see above) 3581 */ 3582 3583 if ((ip->ip_off & htons(IP_OFFMASK)) == 0 && 3584 ip->ip_p == IPPROTO_TCP) { 3585 struct tcphdr *th; 3586 3587 th = (struct tcphdr *)(ip + 1); 3588 if (th->th_x2) 3589 ldt = 1; 3590 } 3591 3592 if (ldt) { 3593 struct tcphdr *th; 3594 struct udphdr *uh; 3595 u_short cksum; 3596 3597 ip->ip_len = ntohs(ip->ip_len); 3598 cksum = in_pseudo( 3599 ip->ip_src.s_addr, 3600 ip->ip_dst.s_addr, 3601 htons(ip->ip_p + ip->ip_len - 3602 (ip->ip_hl << 2)) 3603 ); 3604 3605 switch (ip->ip_p) { 3606 case IPPROTO_TCP: 3607 th = (struct tcphdr *)(ip + 1); 3608 /* 3609 * Maybe it was set in 3610 * libalias... 3611 */ 3612 th->th_x2 = 0; 3613 th->th_sum = cksum; 3614 mcl->m_pkthdr.csum_data = 3615 offsetof(struct tcphdr, 3616 th_sum); 3617 break; 3618 case IPPROTO_UDP: 3619 uh = (struct udphdr *)(ip + 1); 3620 uh->uh_sum = cksum; 3621 mcl->m_pkthdr.csum_data = 3622 offsetof(struct udphdr, 3623 uh_sum); 3624 break; 3625 } 3626 /* 3627 * No hw checksum offloading: do it 3628 * by ourself. 3629 */ 3630 if ((mcl->m_pkthdr.csum_flags & 3631 CSUM_DELAY_DATA) == 0) { 3632 in_delayed_cksum(mcl); 3633 mcl->m_pkthdr.csum_flags &= 3634 ~CSUM_DELAY_DATA; 3635 } 3636 ip->ip_len = htons(ip->ip_len); 3637 } 3638 3639 if (args->eh == NULL) { 3640 ip->ip_len = ntohs(ip->ip_len); 3641 ip->ip_off = ntohs(ip->ip_off); 3642 } 3643 3644 args->m = mcl; 3645 retval = IP_FW_NAT; 3646 goto done; 3647 } 3648#endif 3649 3650 default: 3651 panic("-- unknown opcode %d\n", cmd->opcode); 3652 } /* end of switch() on opcodes */ 3653 3654 if (cmd->len & F_NOT) 3655 match = !match; 3656 3657 if (match) { 3658 if (cmd->len & F_OR) 3659 skip_or = 1; 3660 } else { 3661 if (!(cmd->len & F_OR)) /* not an OR block, */ 3662 break; /* try next rule */ 3663 } 3664 3665 } /* end of inner for, scan opcodes */ 3666 3667next_rule:; /* try next rule */ 3668 3669 } /* end of outer for, scan rules */ 3670 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3671 IPFW_RUNLOCK(chain); 3672 return (IP_FW_DENY); 3673 3674done: 3675 /* Update statistics */ 3676 f->pcnt++; 3677 f->bcnt += pktlen; 3678 f->timestamp = time_uptime; 3679 IPFW_RUNLOCK(chain); 3680 return (retval); 3681 3682pullup_failed: 3683 if (fw_verbose) 3684 printf("ipfw: pullup failed\n"); 3685 return (IP_FW_DENY); 3686} 3687 3688/* 3689 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3690 * These will be reconstructed on the fly as packets are matched. 3691 */ 3692static void 3693flush_rule_ptrs(struct ip_fw_chain *chain) 3694{ 3695 struct ip_fw *rule; 3696 3697 IPFW_WLOCK_ASSERT(chain); 3698 3699 for (rule = chain->rules; rule; rule = rule->next) 3700 rule->next_rule = NULL; 3701} 3702 3703/* 3704 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3705 * possibly create a rule number and add the rule to the list. 3706 * Update the rule_number in the input struct so the caller knows it as well. 3707 */ 3708static int 3709add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3710{ 3711 struct ip_fw *rule, *f, *prev; 3712 int l = RULESIZE(input_rule); 3713 3714 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3715 return (EINVAL); 3716 3717 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3718 if (rule == NULL) 3719 return (ENOSPC); 3720 3721 bcopy(input_rule, rule, l); 3722 3723 rule->next = NULL; 3724 rule->next_rule = NULL; 3725 3726 rule->pcnt = 0; 3727 rule->bcnt = 0; 3728 rule->timestamp = 0; 3729 3730 IPFW_WLOCK(chain); 3731 3732 if (chain->rules == NULL) { /* default rule */ 3733 chain->rules = rule; 3734 goto done; 3735 } 3736 3737 /* 3738 * If rulenum is 0, find highest numbered rule before the 3739 * default rule, and add autoinc_step 3740 */ 3741 if (autoinc_step < 1) 3742 autoinc_step = 1; 3743 else if (autoinc_step > 1000) 3744 autoinc_step = 1000; 3745 if (rule->rulenum == 0) { 3746 /* 3747 * locate the highest numbered rule before default 3748 */ 3749 for (f = chain->rules; f; f = f->next) { 3750 if (f->rulenum == IPFW_DEFAULT_RULE) 3751 break; 3752 rule->rulenum = f->rulenum; 3753 } 3754 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3755 rule->rulenum += autoinc_step; 3756 input_rule->rulenum = rule->rulenum; 3757 } 3758 3759 /* 3760 * Now insert the new rule in the right place in the sorted list. 3761 */ 3762 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3763 if (f->rulenum > rule->rulenum) { /* found the location */ 3764 if (prev) { 3765 rule->next = f; 3766 prev->next = rule; 3767 } else { /* head insert */ 3768 rule->next = chain->rules; 3769 chain->rules = rule; 3770 } 3771 break; 3772 } 3773 } 3774 flush_rule_ptrs(chain); 3775done: 3776 static_count++; 3777 static_len += l; 3778 IPFW_WUNLOCK(chain); 3779 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3780 rule->rulenum, static_count);) 3781 return (0); 3782} 3783 3784/** 3785 * Remove a static rule (including derived * dynamic rules) 3786 * and place it on the ``reap list'' for later reclamation. 3787 * The caller is in charge of clearing rule pointers to avoid 3788 * dangling pointers. 3789 * @return a pointer to the next entry. 3790 * Arguments are not checked, so they better be correct. 3791 */ 3792static struct ip_fw * 3793remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 3794 struct ip_fw *prev) 3795{ 3796 struct ip_fw *n; 3797 int l = RULESIZE(rule); 3798 3799 IPFW_WLOCK_ASSERT(chain); 3800 3801 n = rule->next; 3802 IPFW_DYN_LOCK(); 3803 remove_dyn_rule(rule, NULL /* force removal */); 3804 IPFW_DYN_UNLOCK(); 3805 if (prev == NULL) 3806 chain->rules = n; 3807 else 3808 prev->next = n; 3809 static_count--; 3810 static_len -= l; 3811 3812 rule->next = chain->reap; 3813 chain->reap = rule; 3814 3815 return n; 3816} 3817 3818/** 3819 * Reclaim storage associated with a list of rules. This is 3820 * typically the list created using remove_rule. 3821 */ 3822static void 3823reap_rules(struct ip_fw *head) 3824{ 3825 struct ip_fw *rule; 3826 3827 while ((rule = head) != NULL) { 3828 head = head->next; 3829 if (DUMMYNET_LOADED) 3830 ip_dn_ruledel_ptr(rule); 3831 free(rule, M_IPFW); 3832 } 3833} 3834 3835/* 3836 * Remove all rules from a chain (except rules in set RESVD_SET 3837 * unless kill_default = 1). The caller is responsible for 3838 * reclaiming storage for the rules left in chain->reap. 3839 */ 3840static void 3841free_chain(struct ip_fw_chain *chain, int kill_default) 3842{ 3843 struct ip_fw *prev, *rule; 3844 3845 IPFW_WLOCK_ASSERT(chain); 3846 3847 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3848 for (prev = NULL, rule = chain->rules; rule ; ) 3849 if (kill_default || rule->set != RESVD_SET) 3850 rule = remove_rule(chain, rule, prev); 3851 else { 3852 prev = rule; 3853 rule = rule->next; 3854 } 3855} 3856 3857/** 3858 * Remove all rules with given number, and also do set manipulation. 3859 * Assumes chain != NULL && *chain != NULL. 3860 * 3861 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3862 * the next 8 bits are the new set, the top 8 bits are the command: 3863 * 3864 * 0 delete rules with given number 3865 * 1 delete rules with given set number 3866 * 2 move rules with given number to new set 3867 * 3 move rules with given set number to new set 3868 * 4 swap sets with given numbers 3869 * 5 delete rules with given number and with given set number 3870 */ 3871static int 3872del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3873{ 3874 struct ip_fw *prev = NULL, *rule; 3875 u_int16_t rulenum; /* rule or old_set */ 3876 u_int8_t cmd, new_set; 3877 3878 rulenum = arg & 0xffff; 3879 cmd = (arg >> 24) & 0xff; 3880 new_set = (arg >> 16) & 0xff; 3881 3882 if (cmd > 5 || new_set > RESVD_SET) 3883 return EINVAL; 3884 if (cmd == 0 || cmd == 2 || cmd == 5) { 3885 if (rulenum >= IPFW_DEFAULT_RULE) 3886 return EINVAL; 3887 } else { 3888 if (rulenum > RESVD_SET) /* old_set */ 3889 return EINVAL; 3890 } 3891 3892 IPFW_WLOCK(chain); 3893 rule = chain->rules; 3894 chain->reap = NULL; 3895 switch (cmd) { 3896 case 0: /* delete rules with given number */ 3897 /* 3898 * locate first rule to delete 3899 */ 3900 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3901 ; 3902 if (rule->rulenum != rulenum) { 3903 IPFW_WUNLOCK(chain); 3904 return EINVAL; 3905 } 3906 3907 /* 3908 * flush pointers outside the loop, then delete all matching 3909 * rules. prev remains the same throughout the cycle. 3910 */ 3911 flush_rule_ptrs(chain); 3912 while (rule->rulenum == rulenum) 3913 rule = remove_rule(chain, rule, prev); 3914 break; 3915 3916 case 1: /* delete all rules with given set number */ 3917 flush_rule_ptrs(chain); 3918 rule = chain->rules; 3919 while (rule->rulenum < IPFW_DEFAULT_RULE) 3920 if (rule->set == rulenum) 3921 rule = remove_rule(chain, rule, prev); 3922 else { 3923 prev = rule; 3924 rule = rule->next; 3925 } 3926 break; 3927 3928 case 2: /* move rules with given number to new set */ 3929 rule = chain->rules; 3930 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3931 if (rule->rulenum == rulenum) 3932 rule->set = new_set; 3933 break; 3934 3935 case 3: /* move rules with given set number to new set */ 3936 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3937 if (rule->set == rulenum) 3938 rule->set = new_set; 3939 break; 3940 3941 case 4: /* swap two sets */ 3942 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3943 if (rule->set == rulenum) 3944 rule->set = new_set; 3945 else if (rule->set == new_set) 3946 rule->set = rulenum; 3947 break; 3948 case 5: /* delete rules with given number and with given set number. 3949 * rulenum - given rule number; 3950 * new_set - given set number. 3951 */ 3952 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3953 ; 3954 if (rule->rulenum != rulenum) { 3955 IPFW_WUNLOCK(chain); 3956 return (EINVAL); 3957 } 3958 flush_rule_ptrs(chain); 3959 while (rule->rulenum == rulenum) { 3960 if (rule->set == new_set) 3961 rule = remove_rule(chain, rule, prev); 3962 else { 3963 prev = rule; 3964 rule = rule->next; 3965 } 3966 } 3967 } 3968 /* 3969 * Look for rules to reclaim. We grab the list before 3970 * releasing the lock then reclaim them w/o the lock to 3971 * avoid a LOR with dummynet. 3972 */ 3973 rule = chain->reap; 3974 chain->reap = NULL; 3975 IPFW_WUNLOCK(chain); 3976 if (rule) 3977 reap_rules(rule); 3978 return 0; 3979} 3980 3981/* 3982 * Clear counters for a specific rule. 3983 * The enclosing "table" is assumed locked. 3984 */ 3985static void 3986clear_counters(struct ip_fw *rule, int log_only) 3987{ 3988 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3989 3990 if (log_only == 0) { 3991 rule->bcnt = rule->pcnt = 0; 3992 rule->timestamp = 0; 3993 } 3994 if (l->o.opcode == O_LOG) 3995 l->log_left = l->max_log; 3996} 3997 3998/** 3999 * Reset some or all counters on firewall rules. 4000 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 4001 * the next 8 bits are the set number, the top 8 bits are the command: 4002 * 0 work with rules from all set's; 4003 * 1 work with rules only from specified set. 4004 * Specified rule number is zero if we want to clear all entries. 4005 * log_only is 1 if we only want to reset logs, zero otherwise. 4006 */ 4007static int 4008zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 4009{ 4010 struct ip_fw *rule; 4011 char *msg; 4012 4013 uint16_t rulenum = arg & 0xffff; 4014 uint8_t set = (arg >> 16) & 0xff; 4015 uint8_t cmd = (arg >> 24) & 0xff; 4016 4017 if (cmd > 1) 4018 return (EINVAL); 4019 if (cmd == 1 && set > RESVD_SET) 4020 return (EINVAL); 4021 4022 IPFW_WLOCK(chain); 4023 if (rulenum == 0) { 4024 norule_counter = 0; 4025 for (rule = chain->rules; rule; rule = rule->next) { 4026 /* Skip rules from another set. */ 4027 if (cmd == 1 && rule->set != set) 4028 continue; 4029 clear_counters(rule, log_only); 4030 } 4031 msg = log_only ? "ipfw: All logging counts reset.\n" : 4032 "ipfw: Accounting cleared.\n"; 4033 } else { 4034 int cleared = 0; 4035 /* 4036 * We can have multiple rules with the same number, so we 4037 * need to clear them all. 4038 */ 4039 for (rule = chain->rules; rule; rule = rule->next) 4040 if (rule->rulenum == rulenum) { 4041 while (rule && rule->rulenum == rulenum) { 4042 if (cmd == 0 || rule->set == set) 4043 clear_counters(rule, log_only); 4044 rule = rule->next; 4045 } 4046 cleared = 1; 4047 break; 4048 } 4049 if (!cleared) { /* we did not find any matching rules */ 4050 IPFW_WUNLOCK(chain); 4051 return (EINVAL); 4052 } 4053 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 4054 "ipfw: Entry %d cleared.\n"; 4055 } 4056 IPFW_WUNLOCK(chain); 4057 4058 if (fw_verbose) 4059 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 4060 return (0); 4061} 4062 4063/* 4064 * Check validity of the structure before insert. 4065 * Fortunately rules are simple, so this mostly need to check rule sizes. 4066 */ 4067static int 4068check_ipfw_struct(struct ip_fw *rule, int size) 4069{ 4070 int l, cmdlen = 0; 4071 int have_action=0; 4072 ipfw_insn *cmd; 4073 4074 if (size < sizeof(*rule)) { 4075 printf("ipfw: rule too short\n"); 4076 return (EINVAL); 4077 } 4078 /* first, check for valid size */ 4079 l = RULESIZE(rule); 4080 if (l != size) { 4081 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 4082 return (EINVAL); 4083 } 4084 if (rule->act_ofs >= rule->cmd_len) { 4085 printf("ipfw: bogus action offset (%u > %u)\n", 4086 rule->act_ofs, rule->cmd_len - 1); 4087 return (EINVAL); 4088 } 4089 /* 4090 * Now go for the individual checks. Very simple ones, basically only 4091 * instruction sizes. 4092 */ 4093 for (l = rule->cmd_len, cmd = rule->cmd ; 4094 l > 0 ; l -= cmdlen, cmd += cmdlen) { 4095 cmdlen = F_LEN(cmd); 4096 if (cmdlen > l) { 4097 printf("ipfw: opcode %d size truncated\n", 4098 cmd->opcode); 4099 return EINVAL; 4100 } 4101 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 4102 switch (cmd->opcode) { 4103 case O_PROBE_STATE: 4104 case O_KEEP_STATE: 4105 case O_PROTO: 4106 case O_IP_SRC_ME: 4107 case O_IP_DST_ME: 4108 case O_LAYER2: 4109 case O_IN: 4110 case O_FRAG: 4111 case O_DIVERTED: 4112 case O_IPOPT: 4113 case O_IPTOS: 4114 case O_IPPRECEDENCE: 4115 case O_IPVER: 4116 case O_TCPWIN: 4117 case O_TCPFLAGS: 4118 case O_TCPOPTS: 4119 case O_ESTAB: 4120 case O_VERREVPATH: 4121 case O_VERSRCREACH: 4122 case O_ANTISPOOF: 4123 case O_IPSEC: 4124#ifdef INET6 4125 case O_IP6_SRC_ME: 4126 case O_IP6_DST_ME: 4127 case O_EXT_HDR: 4128 case O_IP6: 4129#endif 4130 case O_IP4: 4131 case O_TAG: 4132 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4133 goto bad_size; 4134 break; 4135 4136 case O_UID: 4137 case O_GID: 4138 case O_JAIL: 4139 case O_IP_SRC: 4140 case O_IP_DST: 4141 case O_TCPSEQ: 4142 case O_TCPACK: 4143 case O_PROB: 4144 case O_ICMPTYPE: 4145 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 4146 goto bad_size; 4147 break; 4148 4149 case O_LIMIT: 4150 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 4151 goto bad_size; 4152 break; 4153 4154 case O_LOG: 4155 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 4156 goto bad_size; 4157 4158 ((ipfw_insn_log *)cmd)->log_left = 4159 ((ipfw_insn_log *)cmd)->max_log; 4160 4161 break; 4162 4163 case O_IP_SRC_MASK: 4164 case O_IP_DST_MASK: 4165 /* only odd command lengths */ 4166 if ( !(cmdlen & 1) || cmdlen > 31) 4167 goto bad_size; 4168 break; 4169 4170 case O_IP_SRC_SET: 4171 case O_IP_DST_SET: 4172 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 4173 printf("ipfw: invalid set size %d\n", 4174 cmd->arg1); 4175 return EINVAL; 4176 } 4177 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4178 (cmd->arg1+31)/32 ) 4179 goto bad_size; 4180 break; 4181 4182 case O_IP_SRC_LOOKUP: 4183 case O_IP_DST_LOOKUP: 4184 if (cmd->arg1 >= IPFW_TABLES_MAX) { 4185 printf("ipfw: invalid table number %d\n", 4186 cmd->arg1); 4187 return (EINVAL); 4188 } 4189 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 4190 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 4191 goto bad_size; 4192 break; 4193 4194 case O_MACADDR2: 4195 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 4196 goto bad_size; 4197 break; 4198 4199 case O_NOP: 4200 case O_IPID: 4201 case O_IPTTL: 4202 case O_IPLEN: 4203 case O_TCPDATALEN: 4204 case O_TAGGED: 4205 if (cmdlen < 1 || cmdlen > 31) 4206 goto bad_size; 4207 break; 4208 4209 case O_MAC_TYPE: 4210 case O_IP_SRCPORT: 4211 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 4212 if (cmdlen < 2 || cmdlen > 31) 4213 goto bad_size; 4214 break; 4215 4216 case O_RECV: 4217 case O_XMIT: 4218 case O_VIA: 4219 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 4220 goto bad_size; 4221 break; 4222 4223 case O_ALTQ: 4224 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 4225 goto bad_size; 4226 break; 4227 4228 case O_PIPE: 4229 case O_QUEUE: 4230 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4231 goto bad_size; 4232 goto check_action; 4233 4234 case O_FORWARD_IP: 4235#ifdef IPFIREWALL_FORWARD 4236 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 4237 goto bad_size; 4238 goto check_action; 4239#else 4240 return EINVAL; 4241#endif 4242 4243 case O_DIVERT: 4244 case O_TEE: 4245 if (ip_divert_ptr == NULL) 4246 return EINVAL; 4247 else 4248 goto check_size; 4249 case O_NETGRAPH: 4250 case O_NGTEE: 4251 if (!NG_IPFW_LOADED) 4252 return EINVAL; 4253 else 4254 goto check_size; 4255 case O_NAT: 4256#ifdef IPFIREWALL_NAT 4257 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 4258 goto bad_size; 4259 goto check_action; 4260#else 4261 return EINVAL; 4262#endif 4263 case O_FORWARD_MAC: /* XXX not implemented yet */ 4264 case O_CHECK_STATE: 4265 case O_COUNT: 4266 case O_ACCEPT: 4267 case O_DENY: 4268 case O_REJECT: 4269#ifdef INET6 4270 case O_UNREACH6: 4271#endif 4272 case O_SKIPTO: 4273check_size: 4274 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4275 goto bad_size; 4276check_action: 4277 if (have_action) { 4278 printf("ipfw: opcode %d, multiple actions" 4279 " not allowed\n", 4280 cmd->opcode); 4281 return EINVAL; 4282 } 4283 have_action = 1; 4284 if (l != cmdlen) { 4285 printf("ipfw: opcode %d, action must be" 4286 " last opcode\n", 4287 cmd->opcode); 4288 return EINVAL; 4289 } 4290 break; 4291#ifdef INET6 4292 case O_IP6_SRC: 4293 case O_IP6_DST: 4294 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 4295 F_INSN_SIZE(ipfw_insn)) 4296 goto bad_size; 4297 break; 4298 4299 case O_FLOW6ID: 4300 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4301 ((ipfw_insn_u32 *)cmd)->o.arg1) 4302 goto bad_size; 4303 break; 4304 4305 case O_IP6_SRC_MASK: 4306 case O_IP6_DST_MASK: 4307 if ( !(cmdlen & 1) || cmdlen > 127) 4308 goto bad_size; 4309 break; 4310 case O_ICMP6TYPE: 4311 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4312 goto bad_size; 4313 break; 4314#endif 4315 4316 default: 4317 switch (cmd->opcode) { 4318#ifndef INET6 4319 case O_IP6_SRC_ME: 4320 case O_IP6_DST_ME: 4321 case O_EXT_HDR: 4322 case O_IP6: 4323 case O_UNREACH6: 4324 case O_IP6_SRC: 4325 case O_IP6_DST: 4326 case O_FLOW6ID: 4327 case O_IP6_SRC_MASK: 4328 case O_IP6_DST_MASK: 4329 case O_ICMP6TYPE: 4330 printf("ipfw: no IPv6 support in kernel\n"); 4331 return EPROTONOSUPPORT; 4332#endif 4333 default: 4334 printf("ipfw: opcode %d, unknown opcode\n", 4335 cmd->opcode); 4336 return EINVAL; 4337 } 4338 } 4339 } 4340 if (have_action == 0) { 4341 printf("ipfw: missing action\n"); 4342 return EINVAL; 4343 } 4344 return 0; 4345 4346bad_size: 4347 printf("ipfw: opcode %d size %d wrong\n", 4348 cmd->opcode, cmdlen); 4349 return EINVAL; 4350} 4351 4352/* 4353 * Copy the static and dynamic rules to the supplied buffer 4354 * and return the amount of space actually used. 4355 */ 4356static size_t 4357ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4358{ 4359 char *bp = buf; 4360 char *ep = bp + space; 4361 struct ip_fw *rule; 4362 int i; 4363 time_t boot_seconds; 4364 4365 boot_seconds = boottime.tv_sec; 4366 /* XXX this can take a long time and locking will block packet flow */ 4367 IPFW_RLOCK(chain); 4368 for (rule = chain->rules; rule ; rule = rule->next) { 4369 /* 4370 * Verify the entry fits in the buffer in case the 4371 * rules changed between calculating buffer space and 4372 * now. This would be better done using a generation 4373 * number but should suffice for now. 4374 */ 4375 i = RULESIZE(rule); 4376 if (bp + i <= ep) { 4377 bcopy(rule, bp, i); 4378 /* 4379 * XXX HACK. Store the disable mask in the "next" pointer 4380 * in a wild attempt to keep the ABI the same. 4381 * Why do we do this on EVERY rule? 4382 */ 4383 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 4384 sizeof(set_disable)); 4385 if (((struct ip_fw *)bp)->timestamp) 4386 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4387 bp += i; 4388 } 4389 } 4390 IPFW_RUNLOCK(chain); 4391 if (ipfw_dyn_v) { 4392 ipfw_dyn_rule *p, *last = NULL; 4393 4394 IPFW_DYN_LOCK(); 4395 for (i = 0 ; i < curr_dyn_buckets; i++) 4396 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4397 if (bp + sizeof *p <= ep) { 4398 ipfw_dyn_rule *dst = 4399 (ipfw_dyn_rule *)bp; 4400 bcopy(p, dst, sizeof *p); 4401 bcopy(&(p->rule->rulenum), &(dst->rule), 4402 sizeof(p->rule->rulenum)); 4403 /* 4404 * store set number into high word of 4405 * dst->rule pointer. 4406 */ 4407 bcopy(&(p->rule->set), 4408 (char *)&dst->rule + 4409 sizeof(p->rule->rulenum), 4410 sizeof(p->rule->set)); 4411 /* 4412 * store a non-null value in "next". 4413 * The userland code will interpret a 4414 * NULL here as a marker 4415 * for the last dynamic rule. 4416 */ 4417 bcopy(&dst, &dst->next, sizeof(dst)); 4418 last = dst; 4419 dst->expire = 4420 TIME_LEQ(dst->expire, time_uptime) ? 4421 0 : dst->expire - time_uptime ; 4422 bp += sizeof(ipfw_dyn_rule); 4423 } 4424 } 4425 IPFW_DYN_UNLOCK(); 4426 if (last != NULL) /* mark last dynamic rule */ 4427 bzero(&last->next, sizeof(last)); 4428 } 4429 return (bp - (char *)buf); 4430} 4431 4432 4433/** 4434 * {set|get}sockopt parser. 4435 */ 4436static int 4437ipfw_ctl(struct sockopt *sopt) 4438{ 4439#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4440 int error; 4441 size_t size; 4442 struct ip_fw *buf, *rule; 4443 u_int32_t rulenum[2]; 4444 4445 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4446 if (error) 4447 return (error); 4448 4449 /* 4450 * Disallow modifications in really-really secure mode, but still allow 4451 * the logging counters to be reset. 4452 */ 4453 if (sopt->sopt_name == IP_FW_ADD || 4454 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4455 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4456 if (error) 4457 return (error); 4458 } 4459 4460 error = 0; 4461 4462 switch (sopt->sopt_name) { 4463 case IP_FW_GET: 4464 /* 4465 * pass up a copy of the current rules. Static rules 4466 * come first (the last of which has number IPFW_DEFAULT_RULE), 4467 * followed by a possibly empty list of dynamic rule. 4468 * The last dynamic rule has NULL in the "next" field. 4469 * 4470 * Note that the calculated size is used to bound the 4471 * amount of data returned to the user. The rule set may 4472 * change between calculating the size and returning the 4473 * data in which case we'll just return what fits. 4474 */ 4475 size = static_len; /* size of static rules */ 4476 if (ipfw_dyn_v) /* add size of dyn.rules */ 4477 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4478 4479 /* 4480 * XXX todo: if the user passes a short length just to know 4481 * how much room is needed, do not bother filling up the 4482 * buffer, just jump to the sooptcopyout. 4483 */ 4484 buf = malloc(size, M_TEMP, M_WAITOK); 4485 error = sooptcopyout(sopt, buf, 4486 ipfw_getrules(&layer3_chain, buf, size)); 4487 free(buf, M_TEMP); 4488 break; 4489 4490 case IP_FW_FLUSH: 4491 /* 4492 * Normally we cannot release the lock on each iteration. 4493 * We could do it here only because we start from the head all 4494 * the times so there is no risk of missing some entries. 4495 * On the other hand, the risk is that we end up with 4496 * a very inconsistent ruleset, so better keep the lock 4497 * around the whole cycle. 4498 * 4499 * XXX this code can be improved by resetting the head of 4500 * the list to point to the default rule, and then freeing 4501 * the old list without the need for a lock. 4502 */ 4503 4504 IPFW_WLOCK(&layer3_chain); 4505 layer3_chain.reap = NULL; 4506 free_chain(&layer3_chain, 0 /* keep default rule */); 4507 rule = layer3_chain.reap; 4508 layer3_chain.reap = NULL; 4509 IPFW_WUNLOCK(&layer3_chain); 4510 if (rule != NULL) 4511 reap_rules(rule); 4512 break; 4513 4514 case IP_FW_ADD: 4515 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4516 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4517 sizeof(struct ip_fw) ); 4518 if (error == 0) 4519 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4520 if (error == 0) { 4521 error = add_rule(&layer3_chain, rule); 4522 size = RULESIZE(rule); 4523 if (!error && sopt->sopt_dir == SOPT_GET) 4524 error = sooptcopyout(sopt, rule, size); 4525 } 4526 free(rule, M_TEMP); 4527 break; 4528 4529 case IP_FW_DEL: 4530 /* 4531 * IP_FW_DEL is used for deleting single rules or sets, 4532 * and (ab)used to atomically manipulate sets. Argument size 4533 * is used to distinguish between the two: 4534 * sizeof(u_int32_t) 4535 * delete single rule or set of rules, 4536 * or reassign rules (or sets) to a different set. 4537 * 2*sizeof(u_int32_t) 4538 * atomic disable/enable sets. 4539 * first u_int32_t contains sets to be disabled, 4540 * second u_int32_t contains sets to be enabled. 4541 */ 4542 error = sooptcopyin(sopt, rulenum, 4543 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4544 if (error) 4545 break; 4546 size = sopt->sopt_valsize; 4547 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4548 error = del_entry(&layer3_chain, rulenum[0]); 4549 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4550 set_disable = 4551 (set_disable | rulenum[0]) & ~rulenum[1] & 4552 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4553 else 4554 error = EINVAL; 4555 break; 4556 4557 case IP_FW_ZERO: 4558 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 4559 rulenum[0] = 0; 4560 if (sopt->sopt_val != 0) { 4561 error = sooptcopyin(sopt, rulenum, 4562 sizeof(u_int32_t), sizeof(u_int32_t)); 4563 if (error) 4564 break; 4565 } 4566 error = zero_entry(&layer3_chain, rulenum[0], 4567 sopt->sopt_name == IP_FW_RESETLOG); 4568 break; 4569 4570 case IP_FW_TABLE_ADD: 4571 { 4572 ipfw_table_entry ent; 4573 4574 error = sooptcopyin(sopt, &ent, 4575 sizeof(ent), sizeof(ent)); 4576 if (error) 4577 break; 4578 error = add_table_entry(&layer3_chain, ent.tbl, 4579 ent.addr, ent.masklen, ent.value); 4580 } 4581 break; 4582 4583 case IP_FW_TABLE_DEL: 4584 { 4585 ipfw_table_entry ent; 4586 4587 error = sooptcopyin(sopt, &ent, 4588 sizeof(ent), sizeof(ent)); 4589 if (error) 4590 break; 4591 error = del_table_entry(&layer3_chain, ent.tbl, 4592 ent.addr, ent.masklen); 4593 } 4594 break; 4595 4596 case IP_FW_TABLE_FLUSH: 4597 { 4598 u_int16_t tbl; 4599 4600 error = sooptcopyin(sopt, &tbl, 4601 sizeof(tbl), sizeof(tbl)); 4602 if (error) 4603 break; 4604 IPFW_WLOCK(&layer3_chain); 4605 error = flush_table(&layer3_chain, tbl); 4606 IPFW_WUNLOCK(&layer3_chain); 4607 } 4608 break; 4609 4610 case IP_FW_TABLE_GETSIZE: 4611 { 4612 u_int32_t tbl, cnt; 4613 4614 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4615 sizeof(tbl)))) 4616 break; 4617 IPFW_RLOCK(&layer3_chain); 4618 error = count_table(&layer3_chain, tbl, &cnt); 4619 IPFW_RUNLOCK(&layer3_chain); 4620 if (error) 4621 break; 4622 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4623 } 4624 break; 4625 4626 case IP_FW_TABLE_LIST: 4627 { 4628 ipfw_table *tbl; 4629 4630 if (sopt->sopt_valsize < sizeof(*tbl)) { 4631 error = EINVAL; 4632 break; 4633 } 4634 size = sopt->sopt_valsize; 4635 tbl = malloc(size, M_TEMP, M_WAITOK); 4636 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4637 if (error) { 4638 free(tbl, M_TEMP); 4639 break; 4640 } 4641 tbl->size = (size - sizeof(*tbl)) / 4642 sizeof(ipfw_table_entry); 4643 IPFW_RLOCK(&layer3_chain); 4644 error = dump_table(&layer3_chain, tbl); 4645 IPFW_RUNLOCK(&layer3_chain); 4646 if (error) { 4647 free(tbl, M_TEMP); 4648 break; 4649 } 4650 error = sooptcopyout(sopt, tbl, size); 4651 free(tbl, M_TEMP); 4652 } 4653 break; 4654 4655#ifdef IPFIREWALL_NAT 4656 case IP_FW_NAT_CFG: 4657 { 4658 struct cfg_nat *ptr, *ser_n; 4659 char *buf; 4660 4661 buf = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO); 4662 error = sooptcopyin(sopt, buf, NAT_BUF_LEN, 4663 sizeof(struct cfg_nat)); 4664 ser_n = (struct cfg_nat *)buf; 4665 4666 /* 4667 * Find/create nat rule. 4668 */ 4669 IPFW_WLOCK(&layer3_chain); 4670 ptr = lookup_nat(ser_n->id); 4671 if (ptr == NULL) { 4672 /* New rule: allocate and init new instance. */ 4673 ptr = malloc(sizeof(struct cfg_nat), 4674 M_IPFW, M_NOWAIT | M_ZERO); 4675 if (ptr == NULL) { 4676 IPFW_WUNLOCK(&layer3_chain); 4677 free(buf, M_IPFW); 4678 return (ENOSPC); 4679 } 4680 ptr->lib = LibAliasInit(NULL); 4681 if (ptr->lib == NULL) { 4682 IPFW_WUNLOCK(&layer3_chain); 4683 free(ptr, M_IPFW); 4684 free(buf, M_IPFW); 4685 return (EINVAL); 4686 } 4687 LIST_INIT(&ptr->redir_chain); 4688 } else { 4689 /* Entry already present: temporarly unhook it. */ 4690 UNHOOK_NAT(ptr); 4691 flush_nat_ptrs(ser_n->id); 4692 } 4693 IPFW_WUNLOCK(&layer3_chain); 4694 4695 /* 4696 * Basic nat configuration. 4697 */ 4698 ptr->id = ser_n->id; 4699 /* 4700 * XXX - what if this rule doesn't nat any ip and just 4701 * redirect? 4702 * do we set aliasaddress to 0.0.0.0? 4703 */ 4704 ptr->ip = ser_n->ip; 4705 ptr->redir_cnt = ser_n->redir_cnt; 4706 ptr->mode = ser_n->mode; 4707 LibAliasSetMode(ptr->lib, ser_n->mode, ser_n->mode); 4708 LibAliasSetAddress(ptr->lib, ptr->ip); 4709 memcpy(ptr->if_name, ser_n->if_name, IF_NAMESIZE); 4710 4711 /* 4712 * Redir and LSNAT configuration. 4713 */ 4714 /* Delete old cfgs. */ 4715 del_redir_spool_cfg(ptr, &ptr->redir_chain); 4716 /* Add new entries. */ 4717 add_redir_spool_cfg(&buf[(sizeof(struct cfg_nat))], ptr); 4718 free(buf, M_IPFW); 4719 IPFW_WLOCK(&layer3_chain); 4720 HOOK_NAT(&layer3_chain.nat, ptr); 4721 IPFW_WUNLOCK(&layer3_chain); 4722 } 4723 break; 4724 4725 case IP_FW_NAT_DEL: 4726 { 4727 struct cfg_nat *ptr; 4728 int i; 4729 4730 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 4731 IPFW_WLOCK(&layer3_chain); 4732 ptr = lookup_nat(i); 4733 if (ptr == NULL) { 4734 error = EINVAL; 4735 IPFW_WUNLOCK(&layer3_chain); 4736 break; 4737 } 4738 UNHOOK_NAT(ptr); 4739 flush_nat_ptrs(i); 4740 IPFW_WUNLOCK(&layer3_chain); 4741 del_redir_spool_cfg(ptr, &ptr->redir_chain); 4742 LibAliasUninit(ptr->lib); 4743 free(ptr, M_IPFW); 4744 } 4745 break; 4746 4747 case IP_FW_NAT_GET_CONFIG: 4748 { 4749 uint8_t *data; 4750 struct cfg_nat *n; 4751 struct cfg_redir *r; 4752 struct cfg_spool *s; 4753 int nat_cnt, off; 4754 4755 nat_cnt = 0; 4756 off = sizeof(nat_cnt); 4757 4758 data = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO); 4759 IPFW_RLOCK(&layer3_chain); 4760 /* Serialize all the data. */ 4761 LIST_FOREACH(n, &layer3_chain.nat, _next) { 4762 nat_cnt++; 4763 if (off + SOF_NAT < NAT_BUF_LEN) { 4764 bcopy(n, &data[off], SOF_NAT); 4765 off += SOF_NAT; 4766 LIST_FOREACH(r, &n->redir_chain, _next) { 4767 if (off + SOF_REDIR < NAT_BUF_LEN) { 4768 bcopy(r, &data[off], 4769 SOF_REDIR); 4770 off += SOF_REDIR; 4771 LIST_FOREACH(s, &r->spool_chain, 4772 _next) { 4773 if (off + SOF_SPOOL < 4774 NAT_BUF_LEN) { 4775 bcopy(s, 4776 &data[off], 4777 SOF_SPOOL); 4778 off += 4779 SOF_SPOOL; 4780 } else 4781 goto nospace; 4782 } 4783 } else 4784 goto nospace; 4785 } 4786 } else 4787 goto nospace; 4788 } 4789 bcopy(&nat_cnt, data, sizeof(nat_cnt)); 4790 IPFW_RUNLOCK(&layer3_chain); 4791 error = sooptcopyout(sopt, data, NAT_BUF_LEN); 4792 free(data, M_IPFW); 4793 break; 4794 nospace: 4795 IPFW_RUNLOCK(&layer3_chain); 4796 printf("serialized data buffer not big enough:" 4797 "please increase NAT_BUF_LEN\n"); 4798 free(data, M_IPFW); 4799 } 4800 break; 4801 4802 case IP_FW_NAT_GET_LOG: 4803 { 4804 uint8_t *data; 4805 struct cfg_nat *ptr; 4806 int i, size, cnt, sof; 4807 4808 data = NULL; 4809 sof = LIBALIAS_BUF_SIZE; 4810 cnt = 0; 4811 4812 IPFW_RLOCK(&layer3_chain); 4813 size = i = 0; 4814 LIST_FOREACH(ptr, &layer3_chain.nat, _next) { 4815 if (ptr->lib->logDesc == NULL) 4816 continue; 4817 cnt++; 4818 size = cnt * (sof + sizeof(int)); 4819 data = realloc(data, size, M_IPFW, M_NOWAIT | M_ZERO); 4820 if (data == NULL) { 4821 IPFW_RUNLOCK(&layer3_chain); 4822 return (ENOSPC); 4823 } 4824 bcopy(&ptr->id, &data[i], sizeof(int)); 4825 i += sizeof(int); 4826 bcopy(ptr->lib->logDesc, &data[i], sof); 4827 i += sof; 4828 } 4829 IPFW_RUNLOCK(&layer3_chain); 4830 error = sooptcopyout(sopt, data, size); 4831 free(data, M_IPFW); 4832 } 4833 break; 4834#endif 4835 4836 default: 4837 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4838 error = EINVAL; 4839 } 4840 4841 return (error); 4842#undef RULE_MAXSIZE 4843} 4844 4845/** 4846 * dummynet needs a reference to the default rule, because rules can be 4847 * deleted while packets hold a reference to them. When this happens, 4848 * dummynet changes the reference to the default rule (it could well be a 4849 * NULL pointer, but this way we do not need to check for the special 4850 * case, plus here he have info on the default behaviour). 4851 */ 4852struct ip_fw *ip_fw_default_rule; 4853 4854/* 4855 * This procedure is only used to handle keepalives. It is invoked 4856 * every dyn_keepalive_period 4857 */ 4858static void 4859ipfw_tick(void * __unused unused) 4860{ 4861 struct mbuf *m0, *m, *mnext, **mtailp; 4862 int i; 4863 ipfw_dyn_rule *q; 4864 4865 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4866 goto done; 4867 4868 /* 4869 * We make a chain of packets to go out here -- not deferring 4870 * until after we drop the IPFW dynamic rule lock would result 4871 * in a lock order reversal with the normal packet input -> ipfw 4872 * call stack. 4873 */ 4874 m0 = NULL; 4875 mtailp = &m0; 4876 IPFW_DYN_LOCK(); 4877 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4878 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4879 if (q->dyn_type == O_LIMIT_PARENT) 4880 continue; 4881 if (q->id.proto != IPPROTO_TCP) 4882 continue; 4883 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4884 continue; 4885 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4886 q->expire)) 4887 continue; /* too early */ 4888 if (TIME_LEQ(q->expire, time_uptime)) 4889 continue; /* too late, rule expired */ 4890 4891 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4892 q->ack_fwd, TH_SYN); 4893 if (*mtailp != NULL) 4894 mtailp = &(*mtailp)->m_nextpkt; 4895 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4896 q->ack_rev, 0); 4897 if (*mtailp != NULL) 4898 mtailp = &(*mtailp)->m_nextpkt; 4899 } 4900 } 4901 IPFW_DYN_UNLOCK(); 4902 for (m = mnext = m0; m != NULL; m = mnext) { 4903 mnext = m->m_nextpkt; 4904 m->m_nextpkt = NULL; 4905 ip_output(m, NULL, NULL, 0, NULL, NULL); 4906 } 4907done: 4908 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4909} 4910 4911int 4912ipfw_init(void) 4913{ 4914 struct ip_fw default_rule; 4915 int error; 4916 4917#ifdef INET6 4918 /* Setup IPv6 fw sysctl tree. */ 4919 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4920 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4921 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4922 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4923 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4924 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4925 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4926 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4927 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4928 &fw_deny_unknown_exthdrs, 0, 4929 "Deny packets with unknown IPv6 Extension Headers"); 4930#endif 4931 4932 layer3_chain.rules = NULL; 4933 IPFW_LOCK_INIT(&layer3_chain); 4934 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4935 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4936 UMA_ALIGN_PTR, 0); 4937 IPFW_DYN_LOCK_INIT(); 4938 callout_init(&ipfw_timeout, CALLOUT_MPSAFE); 4939 4940 bzero(&default_rule, sizeof default_rule); 4941 4942 default_rule.act_ofs = 0; 4943 default_rule.rulenum = IPFW_DEFAULT_RULE; 4944 default_rule.cmd_len = 1; 4945 default_rule.set = RESVD_SET; 4946 4947 default_rule.cmd[0].len = 1; 4948 default_rule.cmd[0].opcode = 4949#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4950 1 ? O_ACCEPT : 4951#endif 4952 O_DENY; 4953 4954 error = add_rule(&layer3_chain, &default_rule); 4955 if (error != 0) { 4956 printf("ipfw2: error %u initializing default rule " 4957 "(support disabled)\n", error); 4958 IPFW_DYN_LOCK_DESTROY(); 4959 IPFW_LOCK_DESTROY(&layer3_chain); 4960 uma_zdestroy(ipfw_dyn_rule_zone); 4961 return (error); 4962 } 4963 4964 ip_fw_default_rule = layer3_chain.rules; 4965 printf("ipfw2 " 4966#ifdef INET6 4967 "(+ipv6) " 4968#endif 4969 "initialized, divert %s, " 4970 "rule-based forwarding " 4971#ifdef IPFIREWALL_FORWARD 4972 "enabled, " 4973#else 4974 "disabled, " 4975#endif 4976 "default to %s, logging ", 4977#ifdef IPDIVERT 4978 "enabled", 4979#else 4980 "loadable", 4981#endif 4982 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4983 4984#ifdef IPFIREWALL_VERBOSE 4985 fw_verbose = 1; 4986#endif 4987#ifdef IPFIREWALL_VERBOSE_LIMIT 4988 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4989#endif 4990 if (fw_verbose == 0) 4991 printf("disabled\n"); 4992 else if (verbose_limit == 0) 4993 printf("unlimited\n"); 4994 else 4995 printf("limited to %d packets/entry by default\n", 4996 verbose_limit); 4997 4998 error = init_tables(&layer3_chain); 4999 if (error) { 5000 IPFW_DYN_LOCK_DESTROY(); 5001 IPFW_LOCK_DESTROY(&layer3_chain); 5002 uma_zdestroy(ipfw_dyn_rule_zone); 5003 return (error); 5004 } 5005 ip_fw_ctl_ptr = ipfw_ctl; 5006 ip_fw_chk_ptr = ipfw_chk; 5007 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 5008#ifdef IPFIREWALL_NAT 5009 LIST_INIT(&layer3_chain.nat); 5010 ifaddr_event_tag = EVENTHANDLER_REGISTER(ifaddr_event, ifaddr_change, 5011 NULL, EVENTHANDLER_PRI_ANY); 5012#endif 5013 return (0); 5014} 5015 5016void 5017ipfw_destroy(void) 5018{ 5019 struct ip_fw *reap; 5020#ifdef IPFIREWALL_NAT 5021 struct cfg_nat *ptr, *ptr_temp; 5022#endif 5023 5024 ip_fw_chk_ptr = NULL; 5025 ip_fw_ctl_ptr = NULL; 5026 callout_drain(&ipfw_timeout); 5027 IPFW_WLOCK(&layer3_chain); 5028 flush_tables(&layer3_chain); 5029#ifdef IPFIREWALL_NAT 5030 LIST_FOREACH_SAFE(ptr, &layer3_chain.nat, _next, ptr_temp) { 5031 LIST_REMOVE(ptr, _next); 5032 del_redir_spool_cfg(ptr, &ptr->redir_chain); 5033 LibAliasUninit(ptr->lib); 5034 free(ptr, M_IPFW); 5035 } 5036 EVENTHANDLER_DEREGISTER(ifaddr_event, ifaddr_event_tag); 5037#endif 5038 layer3_chain.reap = NULL; 5039 free_chain(&layer3_chain, 1 /* kill default rule */); 5040 reap = layer3_chain.reap, layer3_chain.reap = NULL; 5041 IPFW_WUNLOCK(&layer3_chain); 5042 if (reap != NULL) 5043 reap_rules(reap); 5044 IPFW_DYN_LOCK_DESTROY(); 5045 uma_zdestroy(ipfw_dyn_rule_zone); 5046 if (ipfw_dyn_v != NULL) 5047 free(ipfw_dyn_v, M_IPFW); 5048 IPFW_LOCK_DESTROY(&layer3_chain); 5049 5050#ifdef INET6 5051 /* Free IPv6 fw sysctl tree. */ 5052 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 5053#endif 5054 5055 printf("IP firewall unloaded\n"); 5056} 5057